<%BANNER%>

Understanding education

University of Florida Institutional Repository
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20101124_AAAADT INGEST_TIME 2010-11-25T00:50:04Z PACKAGE UFE0011369_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 2085 DFID F20101124_AACUPI ORIGIN DEPOSITOR PATH luallen_j_Page_033.txt GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
cc9a9d51977e949791d0fc1251024611
SHA-1
babb8703c2abf186f638970e050daea208f08dd7
1945 F20101124_AACUOU luallen_j_Page_016.txt
77c19a1e6eafcc4a9dd2d34c5cd724b7
ddfbeb4fd94f5a1bd61954bcdcc563f25149dc6a
2018 F20101124_AACUPJ luallen_j_Page_034.txt
549b07237c84118598bb765dcaafcb3b
0f17e1d044802f3a23eee4e1f90ff5954f9bf33a
2342 F20101124_AACUOV luallen_j_Page_017.txt
1e4c94f9d530e391e443fd6210fd171b
a403adc8089e2347f6f0c945b7830254820c7fef
2602 F20101124_AACUPK luallen_j_Page_035.txt
2b1d1a6bbc7c3874943eac6b9ec4ed83
a54d01e98bd5ca60f707cdfdf4a96603d5015c38
2346 F20101124_AACUOW luallen_j_Page_018.txt
efb50102fdc94859ea0ae76f8b2a98d3
70dc0f94a0972fa7e7758e5b89720e01f9f2ab16
2097 F20101124_AACUPL luallen_j_Page_036.txt
18ad9d31ec3a84c18e6d5954a3d5c441
7eb1724f6bf7d7f2c2869fb12854e7b0ec3657e3
2102 F20101124_AACUOX luallen_j_Page_020.txt
bfbe48bd82b1787bc0d4ced91c17443e
b8af9d61c3f5761d7485dce5f1473f3a371652a6
1926 F20101124_AACUQA luallen_j_Page_052.txt
e462b4c99be6f33f5fd062b54aead7b3
1a0575fccf68ad44c827664ca8bc6eab043ce618
2019 F20101124_AACUPM luallen_j_Page_037.txt
0fd22e6fa2517a1a008c2d073a8b9f12
7985d8319d43232d667129cf7a08fa4b9a84e227
2255 F20101124_AACUOY luallen_j_Page_021.txt
73b36a42ae51b8402999a176197758ad
fe9a1c64ee8caee65ef53078b260f2e3991e1698
1710 F20101124_AACUQB luallen_j_Page_053.txt
51382a0cf715a7971c76d14abd50d760
7cfc887f6f07dac93ca1b0f08945cad71eb11924
1904 F20101124_AACUPN luallen_j_Page_038.txt
ad48226b5797146f861883c74d36725d
4a1c6868c327b3f4f4f2e21ff457772132fcc8e9
2144 F20101124_AACUOZ luallen_j_Page_022.txt
4eeff0224dea1cfe38376005db0ae898
81b59e149cbb90cc413ab2d6fa1a55efeb3929c1
2044 F20101124_AACUQC luallen_j_Page_054.txt
3a56523e15d3ebef5d27b80ad6acafa4
b8daa7df80e3f71beea605ba3eb0e7202c0a7009
2132 F20101124_AACUPO luallen_j_Page_039.txt
dca3a00119676b1cb83a046ee867ca54
8ec6167ab7759ce574f2fc6da1a5ef7fdc80bc0d
2106 F20101124_AACUQD luallen_j_Page_055.txt
633967df333cef887c7fb329b3cec23f
aa459846680c4c0cbf640aa7a77bf11a5265e562
1055 F20101124_AACUPP luallen_j_Page_040.txt
70973ae9b04ed5def4e140d6b55133d6
c545f86212d9c5fa95134c78927a4cfbc8b5b7aa
2532 F20101124_AACUPQ luallen_j_Page_041.txt
f5805a8729ed70d4df868d4bb8e52d67
657d76dd067b77f0c853df76fa50d80df4966666
2016 F20101124_AACUQE luallen_j_Page_056.txt
9873adcb5c1005bd2cdbf22e094c9587
f24599bf5b04d041447a357f4aec390ccf726989
2971 F20101124_AACUPR luallen_j_Page_042.txt
ba74079a9558455207dba22e7e189371
97db0143303b7e2893e812f960bb19ddef1fb302
1703 F20101124_AACUQF luallen_j_Page_057.txt
56c62f5b7da8be99d0cb410b201bf0e0
0fd1c1b61bf582ec1a0d095db7a02db076edc8bd
2727 F20101124_AACUPS luallen_j_Page_044.txt
c3db38c2574e45c1eca92f209db12c7a
77c82c93830ea2e512d6a230ae60ba8f9e7ea876
F20101124_AACUQG luallen_j_Page_058.txt
3c01cd5a44846705357113bd0fab57df
635facb4c6e69bbe5368a611a22c3889d95aacc4
2528 F20101124_AACUPT luallen_j_Page_045.txt
d5e65bf4c366f422f9f4b00d7656acf5
6e35d969430e6a25070aca89e9855de9fff95de5
1846 F20101124_AACUQH luallen_j_Page_059.txt
990bf71add6cdf8de2d41c60cff33716
a9694c61390ca19ffd6e144390dc9d54d18232d9
2437 F20101124_AACUPU luallen_j_Page_046.txt
5291d1e9c57cfa61a70f6651f1208a2d
174c5be6f49aaf8843f4869fa8cdddbf51371ef5
2261 F20101124_AACUQI luallen_j_Page_060.txt
e49fc45d0dd88b4ca8e7d815f589121d
bd3ba6be874cffa506cda2b66b28c7d00ed95430
1935 F20101124_AACUPV luallen_j_Page_047.txt
d98f3e0c698d68cc27d6dc77806c5f8d
03ff31fcec1458062f646db31834a7164c05bc1a
2001 F20101124_AACUQJ luallen_j_Page_061.txt
74c0b5c30366c4be258328aea6f19625
7bf6249a9cfaa0e03579a1680429eca773839565
1518 F20101124_AACUPW luallen_j_Page_048.txt
64d62a84bf330116e8bcf74fca196ca7
60ecca4b5092d2d526a6776d79e4276adcd52f11
2006 F20101124_AACUQK luallen_j_Page_062.txt
7078f4b4e41ab8aa4ce7f1c5d38c12fa
fb5df8376d460a55202a534a02c0add7f40cf4d4
1467 F20101124_AACUPX luallen_j_Page_049.txt
496559ebb71287635703e21df733258f
591e6444c3235af6f32911feb506c7bd99c17cc7
2051 F20101124_AACURA luallen_j_Page_080.txt
a16336652cce3b90d3596752df7da9d3
307b7f2d6b1062eee7a1bcae40ba5e0311aa0f89
1922 F20101124_AACUQL luallen_j_Page_063.txt
9eb8d4802a1f342d37d18f05f696ab7d
8bcfb81cbc5e2bdafeea2aa867505282bfcc3f31
701 F20101124_AACUPY luallen_j_Page_050.txt
b8bea1c2c457c64d764471c3c5615fd9
180febf5730ea34381664e419ecbde29bbd1e758
1990 F20101124_AACURB luallen_j_Page_081.txt
1a0e814c9c018ff7c7011d138c49fb92
dd0f4a1763548502ee994e98f94709abfabd0597
2003 F20101124_AACUQM luallen_j_Page_064.txt
b0b92507eb5120acdf8bc211e0275278
f6c3cad7353443f73091ad24fe0dfa854f84ad27
1998 F20101124_AACUPZ luallen_j_Page_051.txt
c2fb78e4f08481443d8beb312a3a0d10
168e1d91c2f12059b285e86d6ecd9bf88b654454
2323 F20101124_AACURC luallen_j_Page_082.txt
98d151bf5a8e9099dcc76f30996e7a53
cec2880f34113e63fa89f78d994d00a00e1ba8ea
2168 F20101124_AACUQN luallen_j_Page_065.txt
ca11c04268e7a30c543b27fa55af9167
bb2ab572c36df14d0ad04b2cf822b16b9e01a2b3
1947 F20101124_AACURD luallen_j_Page_083.txt
0f5745a09cb45319512687fa249a01a5
944bb6b0a0167cd8a578547dae1c31e543c35cf3
1989 F20101124_AACUQO luallen_j_Page_066.txt
52bfd36888086d6abc79d06bfde5f919
c01d9aa91f6afeead0a9e3b1294da59778dc5e5a
2004 F20101124_AACURE luallen_j_Page_084.txt
9d497968eb733a52ca1b7386f10a96d9
d80d87c876a95966a624b3bd69224aa8c019d717
1993 F20101124_AACUQP luallen_j_Page_067.txt
bdeb18f855d71abd7f6398733361254c
29cac61143d5901800b798432039d246215d9c3e
F20101124_AACUQQ luallen_j_Page_068.txt
c75c13262e1db5e75a11ae0113f05bcf
ffd11cc65f7f901dee9457fbbb492c550fb41da2
1863 F20101124_AACURF luallen_j_Page_085.txt
d64ecc8934d988fdfa9005733aca547d
8b3e41be68bf55bcb2017f09ba340c19f8c1a461
1997 F20101124_AACUQR luallen_j_Page_069.txt
b25cc7c38605f62bcfcdd77f1510784b
8ea6512adb0ca741222bbe78e1224292cf535523
2343 F20101124_AACURG luallen_j_Page_086.txt
f74e306d69fa831482bdec5851225af1
049a9bdb967a134856f22403db4b77a83206e59b
221 F20101124_AACUQS luallen_j_Page_070.txt
5a95166fc10a9e80204cd399653a382f
113a3d41082a1a07cbc1036f633d1f30409c06a3
1868 F20101124_AACURH luallen_j_Page_087.txt
282885397dd8042e39144190a8736e64
4c10ffb26f282535c47cb3ba52d90f889081d902
2104 F20101124_AACUQT luallen_j_Page_071.txt
85de26be7fbb6ae83ebbc62acb229e13
34cd68255425a038211c9821bbd4d05ec47c1245
1889 F20101124_AACURI luallen_j_Page_089.txt
ec808a298983dea07143e9c962cb181b
62c01aad17fea42aec2f4a60e1d65b355c8fd501
1894 F20101124_AACUQU luallen_j_Page_073.txt
bc32b728bd0ca75b2b94fa8b25b50149
91db52098d4a47717a1f135c5d0432cbaf37573c
F20101124_AACURJ luallen_j_Page_090.txt
ab12f7969117b44b79114fa5798af4ff
9949405ad543f39ff9fd52aa4d3b8e8a645c95a5
698 F20101124_AACUQV luallen_j_Page_074.txt
90f2f4c976d4f7aeecdc66506498cffe
1dd0b995553fb07de11e11cd5cde84eb6b42efa1
2041 F20101124_AACURK luallen_j_Page_091.txt
3bcce9457b2198ead350146a73683e3c
db8ccfa947ce53451198dc3a7ef0f7ae0d5fa499
2661 F20101124_AACUQW luallen_j_Page_075.txt
4c30bd1886b65b4515ac4441ffc95b25
3118145f7e67f0914621a1fc3fc0fc17d7eba12a
1485 F20101124_AACUSA luallen_j_Page_108.txt
1d026aae5b3cae275ed0366a242f7ccb
68a1a9ef63cf92055c92b192c535f546faebfccb
2068 F20101124_AACURL luallen_j_Page_092.txt
d9c4a92abc4a149c99b0f9c0cadf3794
af00f7f8e5d0be7f98517d4c601986b455a83450
F20101124_AACUQX luallen_j_Page_076.txt
2823e59e3dca65a507523b3caa25d197
c83778f22425466100837e14c9a4b03dc29b9b4d
97 F20101124_AACUSB luallen_j_Page_109.txt
5ec42c152152b25f76c6a8e41fd4b40c
85cb759fb13b5fefd6f70344c45816fd6982c98e
2000 F20101124_AACURM luallen_j_Page_093.txt
3ad31baae62e87ec564504cf08bf27f2
3c336b695c383e4804900f737cd3e12f6630e42d
1976 F20101124_AACUQY luallen_j_Page_077.txt
cad8d20a560a8ce2169f739d8994cf35
aa1a029be88ae0c1e7f71d69bf8137bbfaca389a
4028 F20101124_AACUSC luallen_j_Page_110.txt
976374a2083f5b0224b7fd1d7804ad47
19b3205f5e88ad1dc6ea89ca00bd83784ccafc31
1810 F20101124_AACURN luallen_j_Page_094.txt
55afb3c9b694abdf4b36c8e0e8268438
202c569afdca5ab4976ca9c982490079db74ecdf
1982 F20101124_AACUQZ luallen_j_Page_079.txt
561c7d6c3278ab859fb75868bdaf6874
015f0e65ba7b91b2cc52cee06301ab5c7e281c7b
4338 F20101124_AACUSD luallen_j_Page_111.txt
d6a3c495997024dae42f905fe4605fa1
1b8c2fc7871edf78195426782e266a771f2a6518
1215 F20101124_AACURO luallen_j_Page_095.txt
e0b0b4557dd2d1714adbef2c90c90b9d
120696d4e40196113703d8821c4a3ec3002ff676
4216 F20101124_AACUSE luallen_j_Page_112.txt
d2041368c17a7f73f8f2b6ec4ab69d5e
57a3353a1b338c312d7a17b3defd3dacea1b3320
1940 F20101124_AACURP luallen_j_Page_096.txt
4bf8509c12dd6dd8a9b7f112ec6d3157
159b2e4e5f797aefc4be59ac2e6abf82c74104e7
4355 F20101124_AACUSF luallen_j_Page_113.txt
dc2c61aee76f8d68d3de04bf0a50ce3f
da7f41d45f573a57ffadacc93e3fef8e5b2ca843
1382 F20101124_AACURQ luallen_j_Page_097.txt
f73f7d613b7efce20f8659309d46bcfb
ef10b6c2b90187d57a5e2743bc4d49b2990d1140
3704 F20101124_AACURR luallen_j_Page_098.txt
ca1526b097354f89ddd044512a1378e4
91dc348ff215047865c5204e07fd397ad3e85b8b
2203 F20101124_AACUSG luallen_j_Page_114.txt
fd24b204f32ac237e08a5512270fffcc
7a0546941efb7630810281b84dd999a515e0e668
965 F20101124_AACURS luallen_j_Page_099.txt
d0ee91742f2d4fbff6c7e0835d777f97
1a1972e6d06d100d0d5d45115276de5970e9c4a5
2523 F20101124_AACUSH luallen_j_Page_115.txt
c466055985fc7f62a880b764c63a0e21
8bb9af5b3491a81db6a07d0ae09b79b48f2f62e9
926 F20101124_AACURT luallen_j_Page_100.txt
d235fe7d926774287caafba8d37059df
8bd4eed95dcf5a5d615e7c623310fde950b2d1d3
2318 F20101124_AACUSI luallen_j_Page_117.txt
3005d645b7be003b1abc149f2ad34e13
70522d236b2100bd6e87730b768edac8e185723c
507 F20101124_AACURU luallen_j_Page_101.txt
b00ca5342580b97fad0d315ce69c33d1
d23b71456c6feb61697e0dc92df4c991364b771f
1239 F20101124_AACUSJ luallen_j_Page_118.txt
289eb3f4c89c210dd128604ca158bdc9
352f013ef27c69f40353e35933f8ddb05bad1315
736 F20101124_AACURV luallen_j_Page_102.txt
2a6f163bd8a5a301782fd7da5b347e43
ee476f6aa14feaaab80a04e64bf3a22606f2b194
974005 F20101124_AACUSK luallen_j.pdf
2cff7e123cb9f05d6b03b4b235c8f6d0
964dce4713654ed27466dafe7ec9f0598b86728a
3564 F20101124_AACURW luallen_j_Page_104.txt
e1a340eb8289ad20db93ef138dcbf955
c3c4dba5313330cc61a69110b3dd501a3d076b26
137280 F20101124_AACUSL UFE0011369_00001.mets FULL
2603ca6ddf17a607f8ab4aeadb063c55
e0c54b400987bb6015abdf7310487b6a9aa75218
3553 F20101124_AACURX luallen_j_Page_105.txt
4f3ebd9f46bb97880bc332d89c6f6019
868098dc3ad334373dd4cbad0fe3f589eec341fb
3830 F20101124_AACUTA luallen_j_Page_008thm.jpg
6000117df733264d42da084cadb962e9
7f456ab6eebc692de232a556057863ad610c6833
7673 F20101124_AACUSM luallen_j_Page_001.QC.jpg
b3dc384bbdc217bf92c69ddf709c6b6f
c4ac8716b4cdc50401760221d3748cd1714dfa4c
3420 F20101124_AACURY luallen_j_Page_106.txt
e1d711f06e846b668f79af5f181948e4
2bc5849883b2db58ccac62efc56ffb2ddb8742cc
5843 F20101124_AACUTB luallen_j_Page_009.QC.jpg
5387568a03e7b898b96908ee62a69b34
fafa5c8d64a0d7264329a1e1d7e9bc15d7654021
2762 F20101124_AACURZ luallen_j_Page_107.txt
d243478b0faf9fa6fc62d3fc99592c6b
6ada56b625bd3b383525bfa8fef0fcc47c9babb2
1983 F20101124_AACUTC luallen_j_Page_009thm.jpg
5b65d7e8972bc72a6993f87339bafb73
982a4ad1f228898c2f28413f04d599d0396db0ef
2505 F20101124_AACUSN luallen_j_Page_001thm.jpg
9bba38bd385ab6d3ea3aa24ef6d1d252
ac007546a5e381d4ecfce0659616077b008e4dc5
17165 F20101124_AACUTD luallen_j_Page_010.QC.jpg
d5fb9920bf60441bc2ee87a7d3005428
5e4e861652b6dcb6b86c3fea77795caa1bec7bc1
3399 F20101124_AACUSO luallen_j_Page_002.QC.jpg
c590d0977440a7edfc2ebcbe4ed902e9
c86c24d08fdbb9d33417c1c540ba0c9791767cc9
4902 F20101124_AACUTE luallen_j_Page_010thm.jpg
993dc50f963b52181e81d311713e4d9d
80e633ca8795c2cf68ce50f496e3d12407f81a77
1374 F20101124_AACUSP luallen_j_Page_002thm.jpg
8a7bab3926e8c46f7afda4175e7ba5a0
c730d0eaa66ec3976a667c64028963205ec3ac17
16801 F20101124_AACUTF luallen_j_Page_011.QC.jpg
febeb391d3107e1ee11e0598cc6ee9aa
59fb3042e2c330c7ea04310e507b9ecc8f578993
20333 F20101124_AACUSQ luallen_j_Page_003.QC.jpg
5b3f09af0b50e0222ef4d4449a46ba50
b0a5026da2840e12da3918e6ff6247e6330ab073
4792 F20101124_AACUTG luallen_j_Page_011thm.jpg
75702fbdc9f5be9d3395203bfa74d954
21ed3965c34c1637e0b574d5761ceb8ffde06d89
5625 F20101124_AACUSR luallen_j_Page_003thm.jpg
78403085cb2d562d3aabde8258b81c7c
02492b486515ee76eef209b506b06b5be365b343
13512 F20101124_AACUSS luallen_j_Page_004.QC.jpg
8b25cf5a7b727405f5e2267417e5f476
266e79a8f6ed6486306c2f6e77adaa373ac35c5e
20700 F20101124_AACUTH luallen_j_Page_012.QC.jpg
917b6ee51f7e6c30f85a31bdc0d45fc2
35d317ebb1a1b9cf00253a4eaee6ee291f2da2bb
3778 F20101124_AACUST luallen_j_Page_004thm.jpg
c0c458b9b6ac3b9d0a34c63c1d6ce8b3
6757de2ee97c3cf558fbec5fca4d55120d336676
23010 F20101124_AACUTI luallen_j_Page_013.QC.jpg
f458b2510449acee2f1d9ed116117846
5bab926cc455419a6a828b7adc302822fad1687e
18168 F20101124_AACUSU luallen_j_Page_005.QC.jpg
a044ab6b091b9aa2be0e18ef13479881
84b6db695c8c8f1f99ff1a571556d57619d83945
6423 F20101124_AACUTJ luallen_j_Page_013thm.jpg
c5f0d2cd51e41f25083c6a2d7806f84e
5a106a85efebe4b914adebf1cfdbb00b50cf945b
4948 F20101124_AACUSV luallen_j_Page_005thm.jpg
e1d2046625c5ad1e7e8ed980623be3e4
e2437fb5713d30b6f10654388c0c2dfdd393888f
22575 F20101124_AACUTK luallen_j_Page_014.QC.jpg
b534877cc2bcfe772e9b3f7361199263
45f1fb89b380d1a7c14e269ab94340da7dd9303c
4128 F20101124_AACUSW luallen_j_Page_006thm.jpg
f606b333e8a4d4c529e7967e5ff62395
3767d9a5aa751d1bb9067a1e4206ad6246114437
23418 F20101124_AACUUA luallen_j_Page_022.QC.jpg
3af755f0f1673aae0ace13a593086643
9d6b8519ed44b89d0f266ca433ace2f5a3090d20
6375 F20101124_AACUTL luallen_j_Page_014thm.jpg
826ae9ce8181eb1a0ff950d12a108df3
6b1e45d70a3ec73446530b8c17ee19a79ae2cf73
23540 F20101124_AACUSX luallen_j_Page_007.QC.jpg
e246a028a588a226053c3d39626fbae8
fe24f1b2cf556e113352469735d8cdd3d6cc0e7d
6387 F20101124_AACUUB luallen_j_Page_022thm.jpg
a985f104547a7669b15c56440e2eaeb2
4a10f185cc6c312164c725320f0b7d5402781956
25048 F20101124_AACUTM luallen_j_Page_015.QC.jpg
bcd1f0ebbf85ec5d1c65964a8878e4eb
e4083111f8d5b97a8ac4648eb03b5bce47866282
6340 F20101124_AACUSY luallen_j_Page_007thm.jpg
01ee222b86aee7702b91d8a642e2177a
f6880e7ba14faf1dd3d0d67c4d940c846e6344a3
22788 F20101124_AACUUC luallen_j_Page_023.QC.jpg
80106376d4ec54a1faf71d2b4507b413
647b69900bca50d6388eb46697838a5821ee0cf6
6858 F20101124_AACUTN luallen_j_Page_015thm.jpg
26e7ffaae6d0856b601b36d7cb63bdd4
49603157dcf7c56d5950d0e0241da237cb43d819
12777 F20101124_AACUSZ luallen_j_Page_008.QC.jpg
9a32df3b47956776554289062a7d4da8
95dfb505800805f3c55149eb69e34300fcc0ae37
6293 F20101124_AACUUD luallen_j_Page_023thm.jpg
deca6c0afeca99fd40fd157886b9eed4
8cfd74a8eb62178d487457a5ca9bb9a133f2be40
22279 F20101124_AACUTO luallen_j_Page_016.QC.jpg
cc0a1679233ccf5da3806315a38431df
da838b2d4ea9f3151cbd5068dc7371967f9bcfb3
6310 F20101124_AACUUE luallen_j_Page_024thm.jpg
5e3337630909642f7029eda65dd2d1c1
20b3455abb7ff945def8ed81efafa19cd586005e
6193 F20101124_AACUTP luallen_j_Page_016thm.jpg
09c5d083b9bae9bf0bb31cbf24a7a2f1
86befc49790ca4c7d1092a084d3e5198304e1de1
23091 F20101124_AACUUF luallen_j_Page_025.QC.jpg
181349dbd5d69d8b0f3e16636a7298bf
020f739555f1cabb8495c3c0139d0c39a8e10a5e
23897 F20101124_AACUTQ luallen_j_Page_017.QC.jpg
3cbfa8d0a30d18b734e5feb890753462
519e83b84ef77e4687f3146deba829124eefda46
28109 F20101124_AACVAA luallen_j_Page_105.QC.jpg
702e535227b473176948faa590afa6dc
f49b3614f284eef1635ffbf15995003507ab7909
6425 F20101124_AACUUG luallen_j_Page_025thm.jpg
1ec1c40e369b0f0a9fff8fe110bf89ed
394d24782d115e57359f73189aa599407ed3150f
6811 F20101124_AACUTR luallen_j_Page_017thm.jpg
e21bb4bc3713ea28aaa22e7e5b1747f8
f204e76c98145503c08bfb29daf3fb73b1188219
7117 F20101124_AACVAB luallen_j_Page_105thm.jpg
f9f2eecd3ae1d7305fee80e239e9d91f
4c6b34c151adfa88241662dc55a865afa14360c8
23991 F20101124_AACUUH luallen_j_Page_026.QC.jpg
fbedf800876fc0ee3c8e06b005ce57fa
84541e8d708fda33209b3efbd5dff17b55e80814
24143 F20101124_AACUTS luallen_j_Page_018.QC.jpg
322b78ee7e0deaef495042ac3dd62357
dac629a43660c789f696067d1327f9a7fdd73a0a
28157 F20101124_AACVAC luallen_j_Page_106.QC.jpg
f15de5aff63134fae3e213be4e1a0256
8aad8036b7cf4177733a501c9dfb8f495163c2c3
6671 F20101124_AACUTT luallen_j_Page_018thm.jpg
30f18b0ecc78b4b0b92dd3519296ff9a
be51c19922a175320f720ee4d414831e7ec61483
7198 F20101124_AACVAD luallen_j_Page_106thm.jpg
a0cf2f2e0304b1891ec3a8257be75688
106b4a83b7e7670103be8a81e7759217d83985d4
6555 F20101124_AACUUI luallen_j_Page_026thm.jpg
37a22c14a82f50229450c889a9271561
78325d324dd7a31cab69ca919751ae4178268da2
21825 F20101124_AACUTU luallen_j_Page_019.QC.jpg
7683c07383fb9a3ccdff5accc51bf2b8
67bfba97879c69d7190897af28286141a91a1f1c
13008 F20101124_AACVAE luallen_j_Page_107.QC.jpg
30a59ed9a090a8c851d23891341bf3f2
50d21427772df4bca217248fa2156fcde6cba3c9
22753 F20101124_AACUUJ luallen_j_Page_027.QC.jpg
e7276c574ac0be7892976a333a4318d5
33960318c024452411e5c807c940952a8135be8b
6199 F20101124_AACUTV luallen_j_Page_019thm.jpg
6131e055c48c85cfa4166e0dee5bcd25
8e625d463684d6eeb379a96f9a9931cfd58e239d
4144 F20101124_AACVAF luallen_j_Page_107thm.jpg
79b0b701d3c62ed279efdfef8718b017
d270551ab15183bd267da8532fd3230f52067471
6572 F20101124_AACUUK luallen_j_Page_027thm.jpg
e3dcea634dbd7a3fec91279b79acccd4
15ce73a5f3051dc8ae47f3c974f9a795f062f686
22426 F20101124_AACUTW luallen_j_Page_020.QC.jpg
673cc6f0d5f04f1d29e7bc60dd64b996
2a590274bb95c899a2306ae46b25ff2cd6b79632
14274 F20101124_AACVAG luallen_j_Page_108.QC.jpg
426c17ceee6a439f2cfc26730668c985
2d0d441ae5b5b6dbcb1502eb97f7e6e15b90d682
21111 F20101124_AACUUL luallen_j_Page_028.QC.jpg
64bf87818edaea35eccb722ae8bb3c22
86cef9b80a9e6d624fb206f03584a31a9925db00
6369 F20101124_AACUTX luallen_j_Page_020thm.jpg
cc0cf2b688e09161503f26021f38f61f
0a5add10ca839824449abe75f38ccf20ac9beaf2
23176 F20101124_AACUVA luallen_j_Page_036.QC.jpg
6e62f878685218884d436b4e13bc5dca
89499cdafe4269e7deeca4adc081e7db22327a15
4386 F20101124_AACVAH luallen_j_Page_108thm.jpg
9fdb0dfca6848d0144b72f2c4f193f91
5a6dcc540a92990d47e62fe26cb33dce415224e1
6200 F20101124_AACUUM luallen_j_Page_028thm.jpg
4374f672da2aa9059f65a66a99301076
7eb637641227713dafbc71a639ae4e69204dcea4
24273 F20101124_AACUTY luallen_j_Page_021.QC.jpg
44e0403570c63ba47b62adb02346f19a
db7a8b610f689c02753c5c36d9174d589a01f1a9
6619 F20101124_AACUVB luallen_j_Page_036thm.jpg
03b8b674af9ca3f975e5ae5c408da313
133e1907bd1dec550cc73906e13aba0f8ad22507
3737 F20101124_AACVAI luallen_j_Page_109.QC.jpg
ffde69a14aa0701bcfd0341b67011197
e4cfcda7f5649a1a42b283069063e0fc520f284a
24209 F20101124_AACUUN luallen_j_Page_029.QC.jpg
15823eab416982f13b19757f8af30a79
977fcc0b79372c3d2a6e028c954192cd9a08d923
6600 F20101124_AACUTZ luallen_j_Page_021thm.jpg
944bf7de004ccf21ed559bd6a248a6c6
e9215c587568581c9a47c19223003297383a1b94
22843 F20101124_AACUVC luallen_j_Page_037.QC.jpg
e8eadc12535cf47034ea3860399e09ff
4cfcd2d5ec07b37b148513e0ebf0db7d5c35ba11
1427 F20101124_AACVAJ luallen_j_Page_109thm.jpg
d954c49dc1899d758cff6fb6f6a62167
624af02557b14b412ab3cdcd6bee5dd0ee312943
F20101124_AACUUO luallen_j_Page_029thm.jpg
4a3ff4f130373b8937e291fc8742323d
214400d5567073fdd362ab1ee03b7aa47e082c30
F20101124_AACUVD luallen_j_Page_037thm.jpg
51e4b40d4d3c1ca3272490920d63416e
5b05c2745bcecfd17c2387db16545ae66d28cb87
29317 F20101124_AACVAK luallen_j_Page_110.QC.jpg
32f1f42dcac0b349f066de75ee11804d
2453539dd2e07eef3c180f56ffe3367f0afdb285
22483 F20101124_AACUUP luallen_j_Page_030.QC.jpg
ccca5e30c2bf8e7037a7ec30ac1fc8ef
7c5f02523520de11e613258fdbda05f59260fdd5
22013 F20101124_AACUVE luallen_j_Page_038.QC.jpg
c69992c651c45f3be358034ac2e37d88
e41455f0e495b37f16b8b77d900d60d6f25461ac
7395 F20101124_AACVAL luallen_j_Page_110thm.jpg
efd664c9038a6aa21e8b0704d22846ea
b1d55fb53af0a6e9e91d9a364a557a1b7267f6af
6497 F20101124_AACUUQ luallen_j_Page_030thm.jpg
b8312680b1d2b551ee464895176b97c9
4b381c7a0e0de0b64a520017887004f9a8e54382
6281 F20101124_AACUVF luallen_j_Page_038thm.jpg
3ccab43832c3e93852b29661b1de36fd
a3d22b58a87ab2e6b7c199531f3d83f080a5aef0
15684 F20101124_AACVBA luallen_j_Page_118.QC.jpg
0c6d7b38c036132e8c3f5ca7636aced1
f0d18e3e0e0c0b85b4ab4d77ab1bcef55357571e
25602 F20101124_AACVAM luallen_j_Page_111.QC.jpg
98a8fd59274160b18287d60169036898
e4c21099d8ce0c82b1a13549a8a47ba254682d3e
23011 F20101124_AACUUR luallen_j_Page_031.QC.jpg
aaeea51639837e0cbe9ab18dac8afe41
822cabeea1f289815f1f87be36834c28fdec34b1
23896 F20101124_AACUVG luallen_j_Page_039.QC.jpg
2594e866917743a4d1e4793a9f2c2f54
2edaf3a7c3b33630524ab0a2bafe4f24e82309d6
6550 F20101124_AACVAN luallen_j_Page_111thm.jpg
1da1fdec3bddb7896ad8ccf90b69700c
be7d96456662b68627d52dc2182cfbf4fad5e937
6295 F20101124_AACUUS luallen_j_Page_031thm.jpg
b9bf2b034dbc19c652a9aaaf861d69c8
f2632ca953170f5a29471381e651bc98afdf5ea9
6516 F20101124_AACUVH luallen_j_Page_039thm.jpg
272beb5557728d766176d82128138352
c70a0060c12f3127879fee9a1b4dc0797f2f6654
25864 F20101124_AACVAO luallen_j_Page_112.QC.jpg
1223a0c9ba6ee99d200f83a9ec23807c
d973a7cb0edf1a17bb460b4ffa8935a6fcef6f8a
21929 F20101124_AACUUT luallen_j_Page_032.QC.jpg
4175fc68965f218e62aa1a6a79697be3
e144c358b78bba55f4b5029d763c400dc4c3c67a
13872 F20101124_AACUVI luallen_j_Page_040.QC.jpg
35b766b9c6fd6a66bd7c09da60ddc08f
6b6b6af4f52cdec9cc07d8388eecb4da0db34fd8
6620 F20101124_AACVAP luallen_j_Page_112thm.jpg
c4ceb1fc1d22fafc0134905c538f2647
054534a436b29e392f583b06b54258f4e38d1048
6231 F20101124_AACUUU luallen_j_Page_032thm.jpg
f9fe349d0a072adcfbc9a57352abf0fd
e55153f407b2bbe157396b65fce58a981f2c0d99
25658 F20101124_AACVAQ luallen_j_Page_113.QC.jpg
3c7151a6442fe1491831cab9754b3a18
bb6ac57fcf3a994c6cc1fb6a16200ab8e9f315ca
23346 F20101124_AACUUV luallen_j_Page_033.QC.jpg
66ad36401f2d1887f9c4db914df3649c
bf684ac51812ca8337a7bcad3807f58fd48716ad
3897 F20101124_AACUVJ luallen_j_Page_040thm.jpg
a5d4f5955ea6e2d88eea96eb1ac277b4
09403d3707eca5a77517c635f824be1a7a94fa2a
6746 F20101124_AACVAR luallen_j_Page_113thm.jpg
8037671199ef90bac50f833cf765e27a
d43f14b2ceae03c00e73cb2584dea4576e550fe0
6563 F20101124_AACUUW luallen_j_Page_033thm.jpg
79df1089ba5ca0b6252d611e3614e1b2
c7d1b111024e6a4491ddbc1c03d4ddb00059a64d
20248 F20101124_AACUVK luallen_j_Page_041.QC.jpg
54418c6ffcfadcd1915a457f64744945
944861d8d832d780108d8e18a8c61bc1f93094bb
22726 F20101124_AACVAS luallen_j_Page_114.QC.jpg
30fd8b77585716e8d3a2f9ad89637cb1
2aa397585fed407143fec3fbee097e57f62fb24d
22777 F20101124_AACUUX luallen_j_Page_034.QC.jpg
a57b07fe9c7baf6de28c46eb37525821
c42b8c1de9251332c2ae2de66e75973b6635b71f
9921 F20101124_AACUWA luallen_j_Page_050.QC.jpg
28a6945f494fe342f23b0f9f2e6eabf1
6aecaef1266a9ebde50caa99198d91ad6579fe53
5655 F20101124_AACUVL luallen_j_Page_041thm.jpg
77ca3c2803930261800e936d388c914b
ffcf5eec2fbf489e50de2d5a96bbf1f264f26676
6329 F20101124_AACUUY luallen_j_Page_034thm.jpg
ed909ef4251a0818905e32681ba7da8c
d798ea2cfb9f1fdd00ef68c8b5ae3aa833cf615a
3152 F20101124_AACUWB luallen_j_Page_050thm.jpg
b0273c4b56f7c9118e48030253263a3e
33fafda9c05e6527c3e2184bd41b616760f62fd4
20765 F20101124_AACUVM luallen_j_Page_042.QC.jpg
3fa3cd27eb230ca9403b41975e89d5a5
307bc150a00733794959a3de8b1a937cf4152ed4
6172 F20101124_AACVAT luallen_j_Page_114thm.jpg
f994653b918d68dd5694830a842d9569
39bcf268f6466f5ff4d4aea00e075c575bf720d3
6740 F20101124_AACUUZ luallen_j_Page_035thm.jpg
4c7e427bc4e3514c3efb9e5423298061
318714c00c1758d2f0a72cb5a3a9a2bbd6bad329
16069 F20101124_AACUWC luallen_j_Page_051.QC.jpg
7d65da7d2eddee172e247ea4ade90168
fe51b2d10f65afecdba4c2ff6abb89bb7fde3a2e
5653 F20101124_AACUVN luallen_j_Page_042thm.jpg
a172acfd5f66ec67bb645ee9bf776f71
c07e4590a3f6d66542526bd7b04d2fb8a158fe26
26204 F20101124_AACVAU luallen_j_Page_115.QC.jpg
f32abc48bf6dbd54e7dbcdea88356635
89fc5d54fd16644eb2be8093e19d4e969ae65795
4627 F20101124_AACUWD luallen_j_Page_051thm.jpg
0374d0ee323cde032c2537e8cfe67852
251f15ff6058fa41cf24f9c44dbe46d9b3a140b0
5379 F20101124_AACUVO luallen_j_Page_043thm.jpg
ef9b00c71c07a93ffc3a85a2831161e2
5bee83b168228f1b214d96038157f32ffbdaf01e
7176 F20101124_AACVAV luallen_j_Page_115thm.jpg
eff8f85e683758cfda0b173343b32766
909d6abe6eeb063cee59f867ebb167c61fccaa82
14230 F20101124_AACUWE luallen_j_Page_052.QC.jpg
0948e8aa90d4e42e1b818b7b36cbd070
6a88ea96d4447965901286b88f1087fdc60dfac8
18620 F20101124_AACUVP luallen_j_Page_044.QC.jpg
1a9707c1ff95346d4d2d2af8eaafcdfd
4b79c96c7e1f074db2d53382564a546f8d623115
25039 F20101124_AACVAW luallen_j_Page_116.QC.jpg
4aba244c8082a5697bed7063b6f329d0
af56ebf925f528178be4a6a8567f164917f2235b
19027 F20101124_AACUWF luallen_j_Page_053.QC.jpg
61efe2ff25d3a788c1152264186adb8d
f1cdd467aff41a7adff197db98a539531c53e418
5169 F20101124_AACUVQ luallen_j_Page_044thm.jpg
9dc06b2e8cd52ef89fd60a29f6462fe8
c9fa32dd2e8e28e07f65a10a8c1bfa06790e8d91
6732 F20101124_AACVAX luallen_j_Page_116thm.jpg
9155d62144bfedded3e49a4538785d3f
d725d8cb07439bdc5288f8aa04a5eee71c27253d
23222 F20101124_AACUWG luallen_j_Page_054.QC.jpg
fe4219f6460fbad20730fd96913dcf96
b5eb450b8e62b2f3c35a36262a89fdb1af7a1bad
16462 F20101124_AACUVR luallen_j_Page_045.QC.jpg
2753e423eefc42ca3611619f4ccd5ba9
7908e5328f546638dc4f26ef4232bd4d4d74a4d4
23685 F20101124_AACVAY luallen_j_Page_117.QC.jpg
8ae267dc6966d9ea9872bd94ecd8d11d
5e1d2c52a2a927e78f69749b5bd86c6f7dc2ed15
F20101124_AACUWH luallen_j_Page_054thm.jpg
9b4fd44d9570cc8ef303ee405cacc949
94dffff5838e5469d71abf310ed8a4d225241d30
4540 F20101124_AACUVS luallen_j_Page_045thm.jpg
c7587eb716b57fed25a100a0de01cafa
03c321f82aed24fa0aa517f731a81bc32e853499
6681 F20101124_AACVAZ luallen_j_Page_117thm.jpg
ec014ec5b6fb0478087ba0234acdb120
89235d2a6ccd3026c75c82818df66257f0f2dd89
23493 F20101124_AACUWI luallen_j_Page_055.QC.jpg
05e952c8ab4e548ce9484f6b8771b814
4009519f71090bd165a7d935cd156cdad1fbe929
14485 F20101124_AACUVT luallen_j_Page_046.QC.jpg
64e4564bfbb33e24ec69012e236c2960
f643f50330db7fa9713e1e782a69d41943ce3927
6648 F20101124_AACUWJ luallen_j_Page_055thm.jpg
3e72c913e6fcc0ba70b06f9136480274
20efbef4b2a48140de32741174183d2ebc17fefd
4202 F20101124_AACUVU luallen_j_Page_046thm.jpg
469e754c5518e0eaf33ab5ac0b08c30d
15e8a4d34b626a5cffd2e5355c623044ac33f240
15634 F20101124_AACUVV luallen_j_Page_047.QC.jpg
e501c75685d73c8826d6b4e076a01cfe
5a2ce0fdf95b4a5794b9c2b6d0dfc3b595598dfc
23772 F20101124_AACUWK luallen_j_Page_056.QC.jpg
06ef02a10795786526d280c906b1bdf9
5b167acf4170d50b6cfceb76e79c76ae37ff3493
4716 F20101124_AACUVW luallen_j_Page_047thm.jpg
4e4841b5e34ed4b67e8ceb82523b9483
bc5937916258d4aebdd47c8569029d93907bdae1
23962 F20101124_AACUXA luallen_j_Page_065.QC.jpg
b0bb90eefc5280845a9f60107b3827c0
a4eaa835b27465d0ec86e2692a329ba011af6218
6402 F20101124_AACUWL luallen_j_Page_056thm.jpg
efe57cc69410a52d5b80e57655acd50c
55c8167f54537ad2f871185777af45ba6b70da2d
10818 F20101124_AACUVX luallen_j_Page_048.QC.jpg
447425bca9d210de65622e58a6fa3927
5622c5fd13c9e66c5a343fc8fe4a61b5aa3dda95
6739 F20101124_AACUXB luallen_j_Page_065thm.jpg
d23e926a9c803e5404ac2cf46776605c
a2279a129f4c683009d7609a3d5b734b0a9492b6
19350 F20101124_AACUWM luallen_j_Page_057.QC.jpg
de3309ef0c5cb6c5b88cf7458b2ed00b
23d6ef0532bfe5ad13b3d77ded537bb55e1e2784
3306 F20101124_AACUVY luallen_j_Page_048thm.jpg
6f5bef656d7f0df4aa0229232f300387
8b4ac8c88b28508527a7e6e91fd80b0027bc2729
23443 F20101124_AACUXC luallen_j_Page_066.QC.jpg
0bbeaa72bc57b8eee0f7c7aff856d390
f8d46e876413fc91fe869c148c7500fe27a71cd0
5698 F20101124_AACUWN luallen_j_Page_057thm.jpg
84f661df5bb71f5cd2ebd81163b479ba
1b6d040af334210f050a6d03b41854e00d5eddbf
13145 F20101124_AACUVZ luallen_j_Page_049.QC.jpg
54df852a3e9d164c85a2cd875c0b6aba
78d275217a7d469f04cf3ea7ab7311374a216672
6513 F20101124_AACUXD luallen_j_Page_066thm.jpg
371ee394d3249af1b6e850735510a5c5
1218c50bf9ebcea646c7d7108003a6fe45b8e19f
6331 F20101124_AACUWO luallen_j_Page_058thm.jpg
9ef908365ba7b6d8aebceece31be12f9
b38f1bf96935f5e6539d11b77479aa4befb67e77
23559 F20101124_AACUXE luallen_j_Page_067.QC.jpg
f8a1a1994186f8d6e1562c17c25170e8
1e45ec413249102e11b5261865b9e8b534ee56dd
21458 F20101124_AACUWP luallen_j_Page_059.QC.jpg
c021a72cd2e62d520ffb0941bbb28835
a055a6465891c17d8c4aecff01ba4a914da8d424
6639 F20101124_AACUXF luallen_j_Page_067thm.jpg
dd5de3e76e4ba90b281e142e70b980ed
dd056eab9a14de2bd01100ab3f74164389959c88
6123 F20101124_AACUWQ luallen_j_Page_059thm.jpg
2f28a48584afccec34697db17feff0a8
cecbdae5fe4a2954e5f1e77da77a901d53e6952d
6655 F20101124_AACUXG luallen_j_Page_068thm.jpg
ea0b7491b9eff117b9dd9ebe44db86e1
37b2dc4bf0ca764883beca6cab193e47ec4d19a1
24412 F20101124_AACUWR luallen_j_Page_060.QC.jpg
ff783a4ad3770499592d8a9143f810c1
27c56ca3ac4d41f7be3db9d710b0d2571da72230
23182 F20101124_AACUXH luallen_j_Page_069.QC.jpg
d8d414468aa9f51126e5c8d07b75a22c
3dd3404bd4a9cd7f8c23d0719bf263ae61b37ce7
6846 F20101124_AACUWS luallen_j_Page_060thm.jpg
fa2eea2ce2a7fa3b67dc02422579bb27
8c3864d2837f0267513a2628cba2d2cd85b4ad8d
6358 F20101124_AACUXI luallen_j_Page_069thm.jpg
dec797d6e0c82b75fef0fb704218feb7
7fb77998b5e655b4c80b2530f2a1bc5f8d082f4e
23468 F20101124_AACUWT luallen_j_Page_061.QC.jpg
217beaa53cb78da09ba84864322d7dba
71c10428784b9e2f305f2109eba8e2b8f86b028a
22348 F20101124_AACUAA luallen_j_Page_072.jpg
e009ba8ec30951eb205805ca9d00b895
f48364ee3d064a24a3d5e37ef6e816e85a75fb5f
4890 F20101124_AACUXJ luallen_j_Page_070.QC.jpg
57cc3d4c65956773146cf11765b01d6c
c3f93b7c1ede6f83e46fb877daa4539df50f651f
6419 F20101124_AACUWU luallen_j_Page_061thm.jpg
92f7f6289cd40b175e616de6187629ad
c189aa5cb474eaf93ad964bd6423323480f11cfe
51161 F20101124_AACUAB luallen_j_Page_073.jpg
ccf88f84462a31ae30063409f507a93a
b4f4e1221c93da953f50d21731d43da896bfa6cc
1762 F20101124_AACUXK luallen_j_Page_070thm.jpg
f24ab20d5a2db8546ba0024cc877734a
d8ce6a4f7827d448d4e6c7d314728af0b875bf2d
23606 F20101124_AACUWV luallen_j_Page_062.QC.jpg
dcaf41d2f95078fcea88e7d1a3e90c0d
84031029a7c5e414f532befb6eb88b935a60e398
23855 F20101124_AACUAC luallen_j_Page_074.jpg
680419329f00c84ee9d1a070c7ef3384
a0a9692b01738aad572970a0d21d9b5c22437189
22252 F20101124_AACUWW luallen_j_Page_063.QC.jpg
18749839c282cdcdf85cda0661c09f96
d5636fccd1d8f7ae6d449d978ca763af74aec909
63914 F20101124_AACUAD luallen_j_Page_075.jpg
3094a2f3e8f0cf9453ba43ed23d50141
2b48abafa76be597d512165efca093dda9539296
6418 F20101124_AACUWX luallen_j_Page_063thm.jpg
f832159d24caa6e0e1740ad718cc3db8
af3ac2779f5a920129cdbcd779b3ca997f2c08fa
65356 F20101124_AACUAE luallen_j_Page_076.jpg
e5fce71316f4dc9692a265399113f979
453825cb01d8b63a795cc0fdcec7a4894d8ec943
6521 F20101124_AACUYA luallen_j_Page_078thm.jpg
69e5c871353452c76e5061494a7c6295
c8cd8da347ded37c461a26a638ba062f395f2a42
17838 F20101124_AACUXL luallen_j_Page_071.QC.jpg
bb19eca846e0f0655b4f057b2ed09e4a
3e3f6f20314b57d21fe2f5dd849a438c76f8d6e3
23399 F20101124_AACUWY luallen_j_Page_064.QC.jpg
d3d0eec89ee063009ffa1681e4b736e4
665d1952e531d0474d8f56841b065d372a7725c1
70004 F20101124_AACUAF luallen_j_Page_077.jpg
cb3c3cf03b8cee7e06d2ef2d1d62874b
7a9a400fe528af270db35463a2962340cc099797
22905 F20101124_AACUYB luallen_j_Page_079.QC.jpg
11c91c725fe6d9d90cc1e34274939baf
aecf991a3f2ba444069dc778377ccb4540804676
5366 F20101124_AACUXM luallen_j_Page_071thm.jpg
015f4f09fc1beeb7a859fd8dc3d2af9f
b2cfebff627aa78a1dff180cd07bef52f8f018ec
6427 F20101124_AACUWZ luallen_j_Page_064thm.jpg
858f2e400e620f6907af0ed37908cd08
00109fa18c1e1d0c19b8450f6d056db83f90e6fa
71084 F20101124_AACUAG luallen_j_Page_078.jpg
7a91758841c7b083523d28c3cd9c9a80
5811fcbbdbbfbe607564afa3551cd446ce0556d7
6378 F20101124_AACUYC luallen_j_Page_079thm.jpg
21bc86ad6d8d32bca6f0f7440ac120d8
91068550b249152644689902c85bce726406e2db
6784 F20101124_AACUXN luallen_j_Page_072.QC.jpg
36a2165dca84eb0862d9906ef43d8342
12dce2fc92616184c72699cb27fc1d4487f5a50f
70901 F20101124_AACUAH luallen_j_Page_079.jpg
cb287d70cf504e8b96844032f6ccc798
111ac7e0fe2222b3f12bfbf72a5442bfd63853a7
22765 F20101124_AACUYD luallen_j_Page_080.QC.jpg
e5cc40c85f9f9d0bd462466997e702a2
6e1cc576fcf03b856d4f6ce872409bc48e881599
2309 F20101124_AACUXO luallen_j_Page_072thm.jpg
e27d70b174bce42fa8ce568ab91b3cbe
409583e6e43e441346846e3565b592a2213ca1b7
72610 F20101124_AACUAI luallen_j_Page_080.jpg
00849b2a2cd36b8802c906ad77a2384d
06791fbfa875afc9423714bd59d721bb65f0acee
50498 F20101124_AACTVC luallen_j_Page_064.pro
b8b64ed99af2d2de50d8524604448ff8
b87588fd8a32270a90326a77020e4182e9a5161a
6568 F20101124_AACUYE luallen_j_Page_080thm.jpg
c2d12c9790d64eab14854b47ff126e86
c5f6e97bf7ea8c694720ee5ce64338f821c18bc7
16584 F20101124_AACUXP luallen_j_Page_073.QC.jpg
fa72ea330533cd1c7dacda6c0b43ee67
a484d5d15319aee68f6e3ede17e01eaa794e61d1
69196 F20101124_AACUAJ luallen_j_Page_081.jpg
b734e776d8464bb3dbc135176b0a24bb
186be130811db0e0610e8ba957a39667e8ef1864
13734 F20101124_AACTVD luallen_j_Page_006.QC.jpg
2bc84a668189e4209e971ec3c52fb29c
3c788f08471665e24083d4b1d4335872de60a956
22667 F20101124_AACUYF luallen_j_Page_081.QC.jpg
21e2a9f147f1887de468afd3a734848a
6df9aa1f123e5202a7a2fd358d8d6bc7b00ca2ea
4793 F20101124_AACUXQ luallen_j_Page_073thm.jpg
206d139ebac74923e81a4dff624c1e4d
0c9418b955d436d709d51788d4d3f948bf4bccd6
74439 F20101124_AACUAK luallen_j_Page_082.jpg
af29f7811718ba8f049d594ba98cd86f
1d96eed478e285d6d61ddab345d7804ab755dfaa
2393 F20101124_AACTVE luallen_j_Page_116.txt
dd22ff18b6bf543a3155a1ce442bcb8b
7c5f891880b30b4605eef0a3cbab2ad398cd7ba0
6382 F20101124_AACUYG luallen_j_Page_081thm.jpg
bb899874675123018882606e523b5fc7
1a05ca5a3d0951230d0ec5207fcb64e0dfa34a27
7106 F20101124_AACUXR luallen_j_Page_074.QC.jpg
7a795aefd22008de6e2344b21566230d
5157fa8258dc8fe026731eb700d58646cfb83fbf
68433 F20101124_AACUAL luallen_j_Page_084.jpg
4b034b4860626b965c3205dc5cbe1f11
533c720f06da14e47382ba6f3e5de3fab4694215
137474 F20101124_AACTVF luallen_j_Page_098.jp2
aa0c4d1e0e42c5d2f7e6e2204809e74d
3430e12c0df104b6a9094e101f5fb6425c5ec148
23040 F20101124_AACUYH luallen_j_Page_082.QC.jpg
2c18ab5ec49247ee358c1c0557dcc4b4
0f8c88c1bbd74923270b5d2dce9399b35fa8ca09
2326 F20101124_AACUXS luallen_j_Page_074thm.jpg
5227b2e5e109e0c427cd94aca216c1cb
928646b935c9ee929d7a1a7dcae3531e67dc4743
65069 F20101124_AACUAM luallen_j_Page_085.jpg
3b9b213bbb576b0f66ba04cb3752efa6
d77e8e0b774f631dca5393b34251d77e962f0cbe
1053954 F20101124_AACTVG luallen_j_Page_063.tif
822c3d89c2477749483eaa61da628f7e
3f7c3e73e2cf16a844bc0d9dd0134e9ec20dc868
6297 F20101124_AACUYI luallen_j_Page_082thm.jpg
ea3ca32145b5e6d1253bb9329ce7a9f7
3ea4f5c5898669fa754d6f656e77d19f090077f9
20938 F20101124_AACUXT luallen_j_Page_075.QC.jpg
293c77f92cd476211cd405bb65348ddd
caf10bf323029aa21bf53ccd6653711b212bee64
35119 F20101124_AACUBA luallen_j_Page_100.jpg
6356062bdec4e28aa80a35bbf9ed0ec7
4ab5bd2f6b63f53803f07e9808d9e21da662540e
73500 F20101124_AACUAN luallen_j_Page_086.jpg
ead83e53d9c5bbadf6361955ba59d81f
951681785ef0fc2f2d155774da2882abee37240c
71772 F20101124_AACTVH luallen_j_Page_036.jpg
7f46995a2b80b395c844baf9b2cef93b
439f9c76c18e29e4428cfd72eb5997931e4537c1
23304 F20101124_AACUYJ luallen_j_Page_083.QC.jpg
a625fe212ad3f42c89e2212ec4683435
2d9a295c7808317ab758d561945c5774411c011c
5992 F20101124_AACUXU luallen_j_Page_075thm.jpg
4b3b3ab4ac540d92fa64f884f29f47ab
36ebfe601b38d471d44b39e4eadd975eee01ef29
32603 F20101124_AACUBB luallen_j_Page_101.jpg
6c193caa2d21f5fb43e2ded3a5723506
00d30eadb0e4fa65cef14b38a955e55c65d91884
70983 F20101124_AACUAO luallen_j_Page_088.jpg
5603ecb77a08c772f0152d947e5fdc7f
2e68895c811522856a8e1ab2d219aca09b7324ac
53607 F20101124_AACTVI luallen_j_Page_046.jpg
e7c9827c6157e5c9d6a8e8938848b7d1
57f34d308026b85e4b6916a366c62fbe67dcbe0a
6502 F20101124_AACUYK luallen_j_Page_083thm.jpg
a01ce6360ac984f2c5ef62687bdaeca4
69d9d57f20e5e8cb1ab9e34e8f8846152d07e03c
21360 F20101124_AACUXV luallen_j_Page_076.QC.jpg
f2a6b54e90081a16515197254ed00327
383e3954be659d7eecc8ddef42b4024412aaffbf
32061 F20101124_AACUBC luallen_j_Page_102.jpg
520f7424b46a5f40fe3e6a422fbf7f50
1cd10ded03bf768f49b2993da27f1fd101b37808
65405 F20101124_AACUAP luallen_j_Page_089.jpg
43aba21727999ac93680c12cbf85e180
c9f6db3bdff2d0783181296f2b4a2473b1340288
4407 F20101124_AACTVJ luallen_j_Page_118thm.jpg
52a3e14b494f64483a10349107593e61
81d963468e19ed76d40d9fb502c11c468d36ce13
21968 F20101124_AACUYL luallen_j_Page_084.QC.jpg
e0cbb1e432afb0c1402556211cb3dd67
6e20c535d6ebfa080e1fb93e901a7f9033b08971
6056 F20101124_AACUXW luallen_j_Page_076thm.jpg
f764d919337926adf58359d85b1ad1a6
4d5c1362ff2295aaf923d9171fb215c777c8d46e
19734 F20101124_AACUBD luallen_j_Page_103.jpg
78a57c9b84d62fde322e11c20af38696
61a66a45df7eede3c6b0a6c319517d41079c6db2
68981 F20101124_AACUAQ luallen_j_Page_090.jpg
8afbbb4fdf7e827ef01c6a646b8a7e59
3bd870c880de17a1a9b22c57cba81e8fe5435889
6661 F20101124_AACUZA luallen_j_Page_091thm.jpg
a2adc3fef0515eec056d792527e99b5c
cd22ce72081a9f81448cf508af399ad3f827accc
22705 F20101124_AACUXX luallen_j_Page_077.QC.jpg
8d15e0af68a6419b4d55db70f8e8d1dc
6ca42ab669be911695f3152ca870e2ea11efcd4b
91148 F20101124_AACUBE luallen_j_Page_104.jpg
74641d80dba2631876ebdbae01c47591
011904604001d05d35e39ed43dd6d21f3e945b15
71066 F20101124_AACUAR luallen_j_Page_091.jpg
7b2ca0a51eb09850d4025234ed37be4f
bf05828d0ecfa2bafbbf7dc9966cee11e820d67d
24275 F20101124_AACUZB luallen_j_Page_092.QC.jpg
86a03a637f5af5f3506b675a56fccc76
38f35ca3397d167e4633705b735f793d68b53bc5
F20101124_AACTVK luallen_j_Page_104.tif
fdfcfe6735d07ab21019b9a0487bf3cd
b61f59b72292692abb94a5cc4c88ac3327d85111
6283 F20101124_AACUYM luallen_j_Page_084thm.jpg
cb0867d6c66f75e441ec0d645531e768
bb7c428caf56e1614d28c79f411fd959cff1b12e
6456 F20101124_AACUXY luallen_j_Page_077thm.jpg
9d3234578ae0d53103623409657709f8
47f62a405166f9f1b58641310d01dc64bace1fa7
91400 F20101124_AACUBF luallen_j_Page_105.jpg
42d7b121ea3d757ec22952f260e3cbb8
620fa484a9310e6387f7c0daa9ea3ea877d1e66e
2714 F20101124_AACTWA luallen_j_Page_043.txt
2477c938029b147746a36137489288e4
e06a44b2fe6410c821331dea4915b104c62f45e6
73845 F20101124_AACUAS luallen_j_Page_092.jpg
40575ba57a03068fa4fcd122707e9a26
28490ed217e42c217feaa671b60ff40984f715e4
6546 F20101124_AACUZC luallen_j_Page_092thm.jpg
57f2ab025979d73060beaf64321f7fc5
d743abd7f398f5c656da56be9cbddfae7660d6fe
64805 F20101124_AACTVL luallen_j_Page_015.pro
80c1e53ce8fa4dff2444bf3e29481e09
dde3e4e2bc35c2460f76b70d7e3a5357a17450c0
20295 F20101124_AACUYN luallen_j_Page_085.QC.jpg
db081b14b66949c8cae685b98404ee77
dda3833dba5adda6b4048d377385364b23e2bfad
23607 F20101124_AACUXZ luallen_j_Page_078.QC.jpg
e6531ff061d574dedd8f9af9d9d832ef
23747678fc13d7c3b1cd9c0d152ec239fc0f88d8
92368 F20101124_AACUBG luallen_j_Page_106.jpg
0818c368102d95fa6840e5ef279c2314
372c3daa1b41fe4383a44dd39dcd497d8beda9c7
277 F20101124_AACTWB luallen_j_Page_103.txt
263ac4cafa8d38bd7b65f2e6067e9b30
cfcb5e6d5f8e3a2a8827bc18c49a9ee2751d4d2e
71807 F20101124_AACUAT luallen_j_Page_093.jpg
1ac1efc106ef5bcdcf941893486f3990
5f9c8389cd4d0b9044898d561fe7a52d6c04cefe
23439 F20101124_AACUZD luallen_j_Page_093.QC.jpg
bc05419549e87642782978009527cace
d8f26fb5ab128f36c7f960232df0a652c2a0379e
307977 F20101124_AACTVM luallen_j_Page_009.jp2
b47efefd2a39d66fd715714439dc8cac
078ee02b0617b99849df73946b18889cc45b655c
5747 F20101124_AACUYO luallen_j_Page_085thm.jpg
94be82270e05a413aade6cb5f5d247da
96019830eda8c857a5660c2694c9b39e8ddad47d
41974 F20101124_AACUBH luallen_j_Page_107.jpg
8e45f2c531309bf750cc4ea063db582e
23c537280befafbd16de64cf0d2106ceac580007
52879 F20101124_AACTWC luallen_j_Page_033.pro
cae31882dcdf46104cdca3658cda96d1
bded8610b70cbc17d769580549c993f4bc45cfd9
6412 F20101124_AACUZE luallen_j_Page_093thm.jpg
532dfa38289a76c0b32fda6c3abf35c0
f8892c5e15e3e40a79457c9fa342aa82ef0ca25b
2179 F20101124_AACTVN luallen_j_Page_088.txt
d4fee5586a6377b21542ef2277a5792d
2b513a222595fc9ebb39e6fe34aaba8868751f32
23719 F20101124_AACUYP luallen_j_Page_086.QC.jpg
b8b7f3c96d582ce2658a588829f60572
6d81524c1d3049a4833812868545ed4cd0ed8877
47864 F20101124_AACUBI luallen_j_Page_108.jpg
69cae2c79f69dc465255c249e23ca044
916a2edade774cafe72ef107b65ac4c7079e799d
71330 F20101124_AACTWD luallen_j_Page_056.jpg
5d283d805ed0d93896e975bda576331e
0fa365551217baf52acec27d00dcca0cbd5dd009
64517 F20101124_AACUAU luallen_j_Page_094.jpg
6405dbbc5f7bd47f72a7238147ea2891
7d2d9d877d13434451cad123df64a74d7cb95dc4
21313 F20101124_AACUZF luallen_j_Page_094.QC.jpg
e4a974056f03f08b384ef0a0fad648ce
8b00710641a0f1e14d1f64be2e4a6277bea9f961
62351 F20101124_AACTVO luallen_j_Page_087.jpg
ca9c6b72f64d283434a3ba9ac9fde77f
9f78628e573a09384d568c1827063d11ff5d510d
6560 F20101124_AACUYQ luallen_j_Page_086thm.jpg
32567aaa0b7df2607c37fbff150e0e0b
f5396f6fca2d431232ebd0aeb59a7999a5f7a0a3
11529 F20101124_AACUBJ luallen_j_Page_109.jpg
1d4731a0a35273a1098a51f3fd0c524d
f74d281037d6c6b5051e9286f627edf1001bcdf9
71600 F20101124_AACTWE luallen_j_Page_020.jpg
db0da44d62248a582079eb32106e30e2
5876fc8b3f4ea0b3d6f0a3efc7d00f4ec47905ff
33986 F20101124_AACUAV luallen_j_Page_095.jpg
53a1b374576ba8b141d4321d01e906ec
3da6cf8139646e842457a39ccc9cabd19e8e9455
6022 F20101124_AACUZG luallen_j_Page_094thm.jpg
09550934db38199499e8f9cf58ab9ba5
29a4f343893acfa71b12c7659cd3d6c88d3589b7
6494 F20101124_AACTVP luallen_j_Page_062thm.jpg
ebd60aec7ac73578cafede0ec5fd918e
803197f484222b19a8ba6e6450a0c415ac716867
19485 F20101124_AACUYR luallen_j_Page_087.QC.jpg
f058f2234759e7ce72bba1202d6ba75f
a14a0f93a9fc172e572532c121940e5ea0b5fd15
96105 F20101124_AACUBK luallen_j_Page_110.jpg
0e9696a2743bd7ac70c21178b87a79e4
f8025f3907140bde11860e7810f1777073bcd858
1958 F20101124_AACTWF luallen_j_Page_031.txt
fc74b34303222b016ec9855df06c916e
3ca71120b10b4736c69a3302a33cb37f7aa271f1
36309 F20101124_AACUAW luallen_j_Page_096.jpg
fda63ececac6616f830874a63142df86
b9cd833989537bc37d03a63cfbef897c349172df
10820 F20101124_AACUZH luallen_j_Page_095.QC.jpg
d520be8839588fe164d660a2c7e758b6
fbf1f1b8f562e884ab1c0dde735db657bbf9b58c
78324 F20101124_AACTVQ luallen_j_Page_105.pro
f4b1cfcfcead515adaa2df6cdd25725a
94889359267b09885a3ce077ce45d89c6b5170e7
5660 F20101124_AACUYS luallen_j_Page_087thm.jpg
41032381126caa26cd73915958ee2101
186dff1394ba8f8617abbdd6c6b5e431b91072db
84207 F20101124_AACUBL luallen_j_Page_111.jpg
d8cf9053cf8cf63b6ae8d69ca14b3def
533d11796991a49a628df2211d80969299d418ad
3509 F20101124_AACUZI luallen_j_Page_095thm.jpg
31a0cd6513231d17536fb0bb32da93e6
ca81453bb95a5b82f0aca813bfd2d4d49e1914a8
F20101124_AACTWG luallen_j_Page_089.tif
32635b4406b1c9ef09a76c0e0c96a28c
6bbdfb8f5533ab7792bed5f17d69b0c4cb27d893
38751 F20101124_AACUAX luallen_j_Page_097.jpg
6c48e0039a846cb705bc10c920e4be1f
79a021d1f2b2911a106c6ab8806ace537b713a82
5364 F20101124_AACTVR luallen_j_Page_053thm.jpg
261c22a6dbb1008cf18eb4e2a57e7c55
4ee6146d941e7ed3d4ed8c07b63c75948f888bb3
F20101124_AACUYT luallen_j_Page_088.QC.jpg
46fb5880ccfd4e223747c8d1e70f45f8
5e9a98be973a6679b3d4bdce1f1b40f3538fa40a
80585 F20101124_AACUCA luallen_j_Page_010.jp2
0844ebf8f6fbfceb87b54498f5a12cc2
b28e7b52f5e3ffbec6290bf51fab9867a7941a10
85506 F20101124_AACUBM luallen_j_Page_112.jpg
055ededf651cb2ab2643473ba70d0002
857058a43a40c4b1f702efdde2ba18a3aa8e47a5
12516 F20101124_AACUZJ luallen_j_Page_096.QC.jpg
488a6baa5749ef86e279469db8a71cca
8f772530d60d04725445b9fc4d2cc1b9f871e539
25271604 F20101124_AACTWH luallen_j_Page_099.tif
085e4a6820084881ae52aeadc07e4013
af0c544c5e5d67d367b07b6b238eb6d1155c7d5c
89158 F20101124_AACUAY luallen_j_Page_098.jpg
ed785ac0d4c874f1c045730cd250ffaa
66c2662e5f594cfa36933f7525fa54a4eb88346a
F20101124_AACTVS luallen_j_Page_025.tif
f54c7d3a8ab0dbb3632f78e284de05d0
72eeed27a4f65faa8ed5e31a70d98459efc02157
6367 F20101124_AACUYU luallen_j_Page_088thm.jpg
b181b27310b0352943451bd86a41b3c2
777b319bb0337898aaeab1044f652976d5599bf6
74696 F20101124_AACUCB luallen_j_Page_011.jp2
4ba2529c1d286b4f632736eae5e18ec7
7cf20fc02fbbd3dae021b61644cff6e07d42d423
84941 F20101124_AACUBN luallen_j_Page_113.jpg
fea682d64b71ecf8a20bc319b02300c2
31477998e73639bc8368ee6a3bb3c2fd5445aa8e
4173 F20101124_AACUZK luallen_j_Page_096thm.jpg
f7bfd9d85c9c6ee4ef4999601b3e7049
eb2e9d5e228d6e55b3072acf2c72faaed9018d0b
70444 F20101124_AACTWI luallen_j_Page_083.jpg
a3634f621eafea5aeed88911dd98ad22
7140af87afad26bc488ac2f1dac32c0831a3e1bc
36857 F20101124_AACUAZ luallen_j_Page_099.jpg
feb2d8a218ce87d24c70fc449045f228
f65de075896c1b61138e8ff1ca1935ee105f68cd
41850 F20101124_AACTVT luallen_j_Page_040.jpg
dad5bf956ced339b807540e000754781
6911cdfc7836ca99e3f8e754e462361041b85070
20858 F20101124_AACUYV luallen_j_Page_089.QC.jpg
1a6cc074007952783eed46b71f44feb4
8b037f79ced551e0b09c359bef488408ca34f6aa
92578 F20101124_AACUCC luallen_j_Page_012.jp2
97e38818c563ec800a388802f9bd5376
1a14ae524e6d630bff42e45a47f39079cfc061a0
90277 F20101124_AACUBO luallen_j_Page_115.jpg
6f2425d7d6fb239d60679dca8ab2eeb4
98815ac67a6dd229dd9585113c046e3e7d28e43b
12117 F20101124_AACUZL luallen_j_Page_097.QC.jpg
6ecc8972411f24103353866b40ed03ba
9c0628fdb4cfdc3857a430c7562fae467d5ecef9
290415 F20101124_AACTWJ luallen_j_Page_050.jp2
a27d31bcba1247288fb2339c5bc99155
daec8aeb8f92fbaf692f5eebc250a72cb2a34905
19263 F20101124_AACTVU luallen_j_Page_043.QC.jpg
07f1b348db49d41d50b77b5b987523d2
0405f86e388a3995194e480bdb5b3232abcba93b
6090 F20101124_AACUYW luallen_j_Page_089thm.jpg
79a00e88db922af3edeeec38e436f1a0
d822b1288cf92aaad3b71dced56c4a8de03739b6
106919 F20101124_AACUCD luallen_j_Page_013.jp2
f983c4cb4800cf4cc9f21d38f486d367
5783f8cc4d3d046e6c7922eef00e8f03e6ce7efd
85065 F20101124_AACUBP luallen_j_Page_116.jpg
12c60ffe7578f2d12e0cc4e1d335df32
31fabc5cfd082317123293d4b876f6bb0df775c8
3864 F20101124_AACUZM luallen_j_Page_097thm.jpg
e63f4e3ec0666f312371e05e743b3c5d
57ea406eccd00827dae58a1df1c078808e9a84c1
111109 F20101124_AACTWK luallen_j_Page_055.jp2
7ae5ff1edf12d2ca0829f69f6167af07
58f464631f5a071200f27e3188224ea75d48572f
F20101124_AACTVV luallen_j_Page_093.tif
10c5247f2d32e66ee00e8fa834ccb83c
e1c6070d336f8565eafbfbf5197946d7f15f17a2
22075 F20101124_AACUYX luallen_j_Page_090.QC.jpg
93802752a02ba8a6ed28cc29432126e7
d993e0b40c69e0e502e170484172faaa079e180b
111113 F20101124_AACUCE luallen_j_Page_014.jp2
f476aee9e580382b7dde7c800dcd28d2
223c3710b42fa9c6da493d879fc661675f8a20a4
84816 F20101124_AACUBQ luallen_j_Page_117.jpg
725dd62852367f16e09aa34d688953db
e3737fe6ae40d947a74182252be58068624cb745
1968 F20101124_AACTVW luallen_j_Page_025.txt
ced15a1d3035889806d841bce177b4db
2baa29d47ff95d4034ec09551bf7dcaa2ae25ebe
6189 F20101124_AACUYY luallen_j_Page_090thm.jpg
4c88641d8adbca0db8eea812177e46a4
9743394071a688e9845aa5578fe6994b3ac9b3e2
125640 F20101124_AACUCF luallen_j_Page_015.jp2
06ebc63e56269f648a24e85cf041d7c1
3a740cce11d08ca7b39a638c602b45991d463bb5
47748 F20101124_AACUBR luallen_j_Page_118.jpg
0ca48cc8c84493b228c9674d31d5bc1a
c742cf6b887d1af5d8b545855e16d2ef4bc557e4
7184 F20101124_AACUZN luallen_j_Page_098thm.jpg
12b834c97351e88656f9eeb29aa1471b
3791120306051ec27d056757ce3878aaf7c8ba79
48611 F20101124_AACTWL luallen_j_Page_063.pro
c54ee224712ff49bda6155e50a2fefb0
26dc57f21f5a8283a332e3eccdb11fff20c42d27
54226 F20101124_AACTVX luallen_j_Page_022.pro
3105c230e173d166b6bde220de9386b3
c40bec005425ad27bb1c7aa86123a3a4a6ac9234
23453 F20101124_AACUYZ luallen_j_Page_091.QC.jpg
64c52b9d0be59cebb5449dd60d0c9a7c
8fae52f5aa406ad9070bb7203b7048a41ec1bce3
103194 F20101124_AACUCG luallen_j_Page_016.jp2
6218255d2208adffb0b67cd917c1f29f
b0efd4e30746a22ced474742fc8cf79b032c0a33
F20101124_AACTXA luallen_j_Page_078.txt
33fd68ae7a4502366ddfc976d3c0b003
d543e7a22972062f981c2a29a795f4a9e0ad7ce6
25805 F20101124_AACUBS luallen_j_Page_001.jp2
04a947aa1a7cf9170a3df958a13d98d1
ebf9bc5eaf395b935328728133c128583012ba28
12573 F20101124_AACUZO luallen_j_Page_099.QC.jpg
c7418a443ef8485d1e4ca8d72b5863bd
2fbed3d2fe099285f736284d0cd473b3f5814b29
24155 F20101124_AACTWM luallen_j_Page_035.QC.jpg
4bc113809ef88a7e601ae104b8a56a41
c2e0fc700a16ef6c7eb23e024f6e482b63dd2378
F20101124_AACTVY luallen_j_Page_107.tif
f3cca836e364abbc27c959c483bcabba
89b3125bf9b0d67ed5da00dde7c009d593b2c2a8
117237 F20101124_AACUCH luallen_j_Page_017.jp2
fced9096eaf7de6e8ad1e4e3ef7e8f4a
035e482fb4129c6b18dda9d2f2e67a95047ca8eb
22263 F20101124_AACTXB luallen_j_Page_058.QC.jpg
cad05a4269ca820dd958dc482a180408
fe026a038bd0b6dda6394d44090abab56e6221e7
6029 F20101124_AACUBT luallen_j_Page_002.jp2
1dbeef6c57fb44e596b1702b746e4165
16b221fd21768ef47446f9fb177960d5980c6938
4308 F20101124_AACUZP luallen_j_Page_099thm.jpg
a7e4cd2ccf6b32cc2a3a339ec129b078
80bf42e329758c23fd73395ffbcb128879c3ebd9
1670 F20101124_AACTWN luallen_j_Page_003.txt
f7589c6bdbb77f4655905429032b3ce8
34a051986f022423d4df34fe429344c52090ca62
33828 F20101124_AACTVZ luallen_j_Page_052.pro
3e8033d8c04631069857cdaaa961471d
7b4c201cd3c6a3d0067a6803854b96af606d965c
117849 F20101124_AACUCI luallen_j_Page_018.jp2
91f0d24b0a0d6f899233f1c3c008c328
9dc051c153e3ac79741da606216acb6193b4d05f
4097 F20101124_AACTXC luallen_j_Page_049thm.jpg
5f001c616e60f38c8a525f630e794690
c916232cdfa38ed2b04e5d6011ce2861bd8f658d
90464 F20101124_AACUBU luallen_j_Page_003.jp2
e3ce46c0488acc8094f042518c70e3d1
61fb08a8f5802a0554bfc56c55d9fa886dc939db
12237 F20101124_AACUZQ luallen_j_Page_100.QC.jpg
580a42ebc3495fe0c088666909f615a7
b39d2126a39223f60e090a5c9c4f13e8391177d5
5869 F20101124_AACTWO luallen_j_Page_012thm.jpg
1c62460e8aa1be204223c05deed4dfaf
83aa1eef31a2139c9e6a80f54aac507c735abcb6
102262 F20101124_AACUCJ luallen_j_Page_019.jp2
f312347cd00982e42d60dcf8b84b8469
3dc74803478453e286d4aa71b2fdbb3cdc083ccd
F20101124_AACTXD luallen_j_Page_065.tif
9b29180f8a6260e7d8cad912b29b7d17
676e7cb9b7e785ca11c29bb1397db69f7acbedc1
4071 F20101124_AACUZR luallen_j_Page_100thm.jpg
8f1419a759f0aff4e24a1856764e5ffc
e90388757ec253d9a2ff48e9c62947493463c8af
111234 F20101124_AACTWP luallen_j_Page_092.jp2
db51762a1d2a2c30eaa72388bdb52ec4
8441f74230090efd2decf56d17b38c4dab62608f
108403 F20101124_AACUCK luallen_j_Page_020.jp2
235973268e9567ef54490561fa7a42ad
c9ad385899cb17990c753ada108c05fac2c03e8e
F20101124_AACTXE luallen_j_Page_030.tif
e9a83d02457086a9a183c8ee49408608
0d2de84313b883514e3e78f49d83dc6cb61f53d1
55609 F20101124_AACUBV luallen_j_Page_004.jp2
fc9078eec6f95a34b53a71005771cbdd
06d89f09ba7e87590280a8640b8232b05794c48a
11326 F20101124_AACUZS luallen_j_Page_101.QC.jpg
15ae24669cef5167bd029408925c8917
e734bd43233c9c05eb60e76cea65028297866b0c
23888 F20101124_AACTWQ luallen_j_Page_068.QC.jpg
89c3aeab17df11c878c2082ada727ab6
6eedcc8b52358c17ecb40a5e66e06ef58def9576
115499 F20101124_AACUCL luallen_j_Page_021.jp2
29da99d9cdd8edc6a61038f534308a84
fb9f35cfd0179d7cb1da9fcb91be611b48473a65
134525 F20101124_AACTXF luallen_j_Page_115.jp2
60dcf72ac00ad0bed7740197511e9d52
7b45ea6e2fc4e3965a85d55021873fcb991a184c
1051982 F20101124_AACUBW luallen_j_Page_005.jp2
7f06f214e92087498ddeef9ae9d4e087
2428e870b71a43d0978e2dc153c13a33d0874f9c
4088 F20101124_AACUZT luallen_j_Page_101thm.jpg
6c48322fbec6b5083014bf8537394b3b
7b586c352ef40759a137e3b9446d02f7a1ee077c
27576 F20101124_AACTWR luallen_j_Page_098.QC.jpg
25a72d423a58afdc95025d38140e4753
564b92e7f810063073e56ff2cb0d4e53069708f2
109162 F20101124_AACUDA luallen_j_Page_036.jp2
40ae5fb23f64ca5ec31c5349cc2f145d
8c19022a691100216efdbd93a9dfd8d79113dfd5
110965 F20101124_AACUCM luallen_j_Page_022.jp2
7e71f39fbcfd3e2617507fadec21f1d6
0afc1947908dcd307ff644b040126affeb105415
763601 F20101124_AACTXG luallen_j_Page_051.jp2
d164d3ff4dd32b26dd0c56cdc2ae39a9
67f73c77a0dd1ede0221b84f3f0cacecfa0ecbdf
1051984 F20101124_AACUBX luallen_j_Page_006.jp2
3db7ec9a0f9ae93ca8af8f47ae2f4eca
a583341d0daa04fdf4ad906bc3ccb3494c674423
10857 F20101124_AACUZU luallen_j_Page_102.QC.jpg
c9a4b362042854e4ddbe1840ca5b7045
28a068bd943e7eaa6f51ab46214abd7d5b3877ac
76666 F20101124_AACTWS luallen_j_Page_114.jpg
9ab221b6beb5d5537ee30edd5d886993
288ecbf9829baed236e550cb42cc3e940442884b
105817 F20101124_AACUDB luallen_j_Page_037.jp2
73bcf38e9056e6add2eaf28aa886bf2f
6f0145a60bd804221988b1b00bddd8192e03b0b2
103972 F20101124_AACUCN luallen_j_Page_023.jp2
ed0427a2b339a1508ed0e00977da39e4
975251e9f394bb325ad05db210bff46b56323b48
26583 F20101124_AACTXH luallen_j_Page_040.pro
a0dc56d13ca40f0dc4611f00d81b80df
02bf7948a48c12fa88ea219e7891cca03e626b7f
1051983 F20101124_AACUBY luallen_j_Page_007.jp2
a86b31ae5d21a4fd6bb5db816b6c0d49
58856996b5c0a175d97cda2a2e0039dc423637f0
3610 F20101124_AACUZV luallen_j_Page_102thm.jpg
ec9ee592ddb348ce185c03c3eed1654f
831ecfb24c172c68ba21596497138d789a489681
612 F20101124_AACTWT luallen_j_Page_072.txt
fa0dda720ccff2bbffb7cdd9c40685aa
720ca3c9ea027d56dbf1eb594bbc09f9234cd4ad
101807 F20101124_AACUDC luallen_j_Page_038.jp2
73d48a68ddf5148ca62b930769ed1978
22a0a76afbf9db62e45603343af82371647e9cc2
103099 F20101124_AACUCO luallen_j_Page_024.jp2
6833fcb3a9c660951824674d7c43153a
9f11794a46b78991d6e36f9548f14ca4569ddd53
21902 F20101124_AACTXI luallen_j_Page_024.QC.jpg
1be4f9a4dccd161acd9f29f42c06089f
e5fc2adb337c12a6972915789c72a0b2cf0ac7fb
1051978 F20101124_AACUBZ luallen_j_Page_008.jp2
e435838a360abb16a6a8c58968cbd376
fb163847d1aed8ff7dc94c5404bf686a4788ce66
6690 F20101124_AACUZW luallen_j_Page_103.QC.jpg
f83259a82b80a8653097b67730c8ee49
16191be40f029d4e1f6585050b687ae17dbeae0c
50161 F20101124_AACTWU luallen_j_Page_077.pro
542872fb9716cd59d7a0eebe9803c81d
7da8ea139f9bff0472869ba9cee813f201b74398
F20101124_AACUDD luallen_j_Page_039.jp2
2c263451bfd56e4fd87d3fe13e88200e
33e7ad64dd7fa606cd385f10b99149a037245c77
106210 F20101124_AACUCP luallen_j_Page_025.jp2
516a5a61f1a3a3fb03a48150014044e2
68cc35b1a355ee66e7e1a3bcdec2a02b0bcd6ffd
177974 F20101124_AACTXJ UFE0011369_00001.xml
10444f1a9e003c6982a6bab343a9ffec
8564dd368bd120c932352f4a5757532e1d8425cd
2547 F20101124_AACUZX luallen_j_Page_103thm.jpg
8a80cc0a265c670b0ba9eb8e98ed2165
43df39971a4beaff5b794466ae88f9ba654fc9c9
50074 F20101124_AACTWV luallen_j_Page_096.jp2
84d1ded7f9df8f4d3bd7de176d918e85
ff7b3019e933e997437b387ea04f791553387f7c
58543 F20101124_AACUDE luallen_j_Page_040.jp2
bd4b55aad24c3eb4324111238b32d0e3
b9c87898ae850b1962952a4de19e5f5750feb572
116494 F20101124_AACUCQ luallen_j_Page_026.jp2
48bcaf58af2ce2bd35eb90badec1253a
59a77109b11075103ce577928847c09a79dfe54e
28002 F20101124_AACUZY luallen_j_Page_104.QC.jpg
0c210209641bfdf8f9adcc125707e58b
de93fab36bd1437357e58ddaf52db2343e857603
999 F20101124_AACTWW luallen_j_Page_004.txt
af16f41f3446b9a2aae649de21ef0ec8
63b0c0b09e3b3e0f64de210d0f02f67d5ae67869
980652 F20101124_AACUDF luallen_j_Page_041.jp2
bea03ee23f9d6299e814b7a408d45ab2
035aa04fa50c9c5b7aba4ac1249dbaf51c97afd7
111338 F20101124_AACUCR luallen_j_Page_027.jp2
cd667f7d00dc666ef1669198a934fb9a
842a81ed83bb2d2cce45ac4acfa58e817a713a03
7209 F20101124_AACUZZ luallen_j_Page_104thm.jpg
badff7cc039662a8c3e39946ab562944
0d5c92adbaaf47ce5ae8e898771511150a7b7476
1910 F20101124_AACTWX luallen_j_Page_019.txt
2dc9f6a3b4742e958ce3b7fc44237d73
09f8819367aa6a46ac88ee0093bf8603b23c916f
983263 F20101124_AACUDG luallen_j_Page_042.jp2
7d60f929ff9d86264eb48808e56bf7f0
f91f936db6422ed9a1afc077f9283c4a47f4ce87
81655 F20101124_AACTYA luallen_j_Page_015.jpg
f3d69a8ca45bd56f3c36d04b8fb499ed
86a7d076bb397e5c96da7cb38597c78d6211fcc1
98518 F20101124_AACUCS luallen_j_Page_028.jp2
0e3efa403bfee6a21afc7deaa9f4f560
7b863e13a0e66efd5d9a1861c641e19361304eed
874728 F20101124_AACUDH luallen_j_Page_043.jp2
785127ebb4c9cef8ced53da27c8fcf3d
126c21a7f88c55a2ed5360a72b4868979deda494
67262 F20101124_AACTYB luallen_j_Page_016.jpg
d14c08210efb345ebaa8b8c00801ccce
57f46f4aeaa7c98c32d3ab6768e29317db3941dd
116721 F20101124_AACUCT luallen_j_Page_029.jp2
99ab1bad5733d1973fe68f1c6a5cec21
9b4ea5486f299b8218d4ee512755d9751e7340d8
24222 F20101124_AACTXM luallen_j_Page_001.jpg
6f64840b5a584727448f47d0468360b1
a2a9a95ca6027e717e5c70680c706516fdaebee9
40976 F20101124_AACTWY luallen_j_Page_047.pro
3f19e72811476b37c4034c18bd5642be
6592a074b93310b96b9ff2ef7de513f8cdb2ad31
867602 F20101124_AACUDI luallen_j_Page_044.jp2
93f8138653af6577aec3f425da349bf9
020ddd15de7bc8639fe7508b1d33fe8f717128c9
75960 F20101124_AACTYC luallen_j_Page_017.jpg
9dee664d44cfce174950f6097ad59396
bdb5694a08b787fc6414124a5230d259b17d5f31
104575 F20101124_AACUCU luallen_j_Page_030.jp2
6fd2851919476f9a2ede7c7707afed7f
d9ee18865290c731822f950444f61e47b629b228
10562 F20101124_AACTXN luallen_j_Page_002.jpg
a2d2cd6f46a0ae1e8884d38965006f80
0c4f1171ff9e0d6c1fd0685f989a3bd586d8e6a0
4267 F20101124_AACTWZ luallen_j_Page_052thm.jpg
976ba802e62ae09be55ad55ba97d9344
7649cbe6a5811ead299c7c5078523405b1aad62d
837599 F20101124_AACUDJ luallen_j_Page_045.jp2
18075fede7baeb9e173fadd14b4e106a
4596652c6802f83263f2d52cb820e049202edb78
76131 F20101124_AACTYD luallen_j_Page_018.jpg
89b52fb8480c82de8b2d46e6b7ee180d
f051f3aaa26c56d3bc5e6ffa06405c46e0eafc15
108523 F20101124_AACUCV luallen_j_Page_031.jp2
d5cc6349812b4832f740842026a10e79
1f5006845f38af433a65ba8ce83e379edb83e3fb
61082 F20101124_AACTXO luallen_j_Page_003.jpg
0cb4b2f8bcf81d01624334bd91518387
959b6201f4584e4ddeac52ce955701cbecf80292
724382 F20101124_AACUDK luallen_j_Page_046.jp2
68814d081149124ef84a4fe936b0aed4
7867bc213ed9ffb33f229af5f8169aee921e176f
66894 F20101124_AACTYE luallen_j_Page_019.jpg
64b4e50ce3a98809cdc0a52f06e7f930
4f87e062809700831b262c7ae3a70860b01be65b
40838 F20101124_AACTXP luallen_j_Page_004.jpg
0f42bcee8aea3726dca97bf87d992e0b
88030e61ffaf12d1a777c630224758a7e8596f0e
740712 F20101124_AACUDL luallen_j_Page_047.jp2
21aff6c93d6325fd2672ff4ac67ab768
136f3f6a086d248b15475df06a0a670757f91e78
76026 F20101124_AACTYF luallen_j_Page_021.jpg
28b449b2cfd2fc6dd99554051210b7f0
137d1fc6d1c8c0eb30142cab3308b70a5a603d4a
100344 F20101124_AACUCW luallen_j_Page_032.jp2
238baf00c64c5cc840f7d8c8582b3344
6faca6f01c5f4648b2a6e6d6a611bc35c1ea9dc0
70257 F20101124_AACTXQ luallen_j_Page_005.jpg
6d2ff6e5abe0b9c991bae0224b0931a0
1a4e5146cc48cc486dfde80f12396d280bb37406
488620 F20101124_AACUDM luallen_j_Page_048.jp2
c6a2dafd3512aa5421b588ebc0932f21
896e624b01a4cd17e4c4e4a6d60014911ca31953
72491 F20101124_AACTYG luallen_j_Page_022.jpg
d6695763e0b6ab91257bfbd47906fb9f
46cc3ef4cf2655334b0820c532dfeb461ff07d67
109964 F20101124_AACUCX luallen_j_Page_033.jp2
3ecf57fc3c6321c71d24f6645edcac31
cebc26577a9bfb4a2de8cd33b4709d9860a321a7
52538 F20101124_AACTXR luallen_j_Page_006.jpg
8781f40fff1e08019c03b5df1de99256
7fd4b187e962fd067d3959735d97b7da2edd228c
112015 F20101124_AACUEA luallen_j_Page_065.jp2
a1db4ed72123de52f831af9a2824d449
050902edcb33a98c6352f901cc5dc517dfa032f5
557721 F20101124_AACUDN luallen_j_Page_049.jp2
7d414ce4b709a2b58ecd053c7866c0d8
060bbd96d5345e88c57d32cd90606cd42c1232f8
69342 F20101124_AACTYH luallen_j_Page_023.jpg
86774c566cf739e25f2bb064d1221c3c
e9eeab96c77de67a6360856d13722293fe19e2fb
106013 F20101124_AACUCY luallen_j_Page_034.jp2
e90b2332c169ac9da695f60c7e49791e
49876a6ab16109e8a5b0a449f1e482b0b370313d
82823 F20101124_AACTXS luallen_j_Page_007.jpg
0fa176e025e1735baf3fa4a4d5d4e6a2
a55aae2c130b3b59f3744eeb55b8ffc978b0addf
107515 F20101124_AACUEB luallen_j_Page_066.jp2
ee6cb6c62ff0d12b87ffb26adc0c09cc
efe72994d3b157c302c5c9e0bdde6478bc09d0c7
612397 F20101124_AACUDO luallen_j_Page_052.jp2
86baec43cb597fa9cf99d9c53a643cfd
8427e755e532657bcd6d256564b228fa4f4bc859
67847 F20101124_AACTYI luallen_j_Page_024.jpg
ceecf85f3f639e28b7cf6c70b1f293f9
b912da7c71dc73625d3c95b03caca55d2378711c
120941 F20101124_AACUCZ luallen_j_Page_035.jp2
4360d10623b9c102e323ed590b34c009
0b3c6624f16833dd5c06d31b057814bf76d1fd78
43550 F20101124_AACTXT luallen_j_Page_008.jpg
1f77caa6fe8e571340cb627cfbe1e537
6bf846a18e683920bde92e444c7c08eeb555bc84
107670 F20101124_AACUEC luallen_j_Page_067.jp2
faa653567e5c7235756a8730196da0b4
0d0fa476366b02ce078189818daed3fcc1f12c80
86534 F20101124_AACUDP luallen_j_Page_053.jp2
6fb3c65940be27915b7fced0d5975e5f
bee63f100a8ad86612f95751c331eda969ce081f
69738 F20101124_AACTYJ luallen_j_Page_025.jpg
5a314064752f962a4a3ae88fcdbead6a
a1f320c10d7386b82dcfa328b14c62e102d54b46
18034 F20101124_AACTXU luallen_j_Page_009.jpg
e6c91941401efc38b4dba9f2c61f8452
0de3be4dcd5065a71394609bc8d155b5acd2ddc5
110742 F20101124_AACUED luallen_j_Page_068.jp2
3bf5d4aa2c6fbd52faab20d40e6ba990
8e07436a7f66ab1608a8eb35047d78a0a4854658
109371 F20101124_AACUDQ luallen_j_Page_054.jp2
8f794b2406fbbf246df4c9f361993161
b9708490163ce09cbda3841c0ee8abd049bc152c
77084 F20101124_AACTYK luallen_j_Page_026.jpg
9bdf64eb2bde201e946e2e6fda20c101
0658972482ca897282bd67f90f28767a7f76be22
55447 F20101124_AACTXV luallen_j_Page_010.jpg
3794b080ac0297a35b203d7959e09e23
8fa441e1d6bf99dfa722e8004c51f9808ccf6776
107071 F20101124_AACUEE luallen_j_Page_069.jp2
7c951df87c1124be65034e48aeaade08
82f1a75cb44b355d2fd71a52a8cf60af02a8ae24
109507 F20101124_AACUDR luallen_j_Page_056.jp2
098df0e8f4fc779b3e8a9c71ea552395
033cf64219a0bd03211141226c7e714ca436c18f
73240 F20101124_AACTYL luallen_j_Page_027.jpg
273435133c5eb224985da6fe56adc643
4808d21ae09d97093b743b4313aaf4775305f3c9
50350 F20101124_AACTXW luallen_j_Page_011.jpg
3c6f60f87f412fd1e4dab102386f28dc
89a912954809c5172173ed1724867bad5fcc3d2f
12882 F20101124_AACUEF luallen_j_Page_070.jp2
25e040660b2d1c5cb333d589f2dc061c
97fb896cb4e226adf38f3467c0dbdd12335616ff
89091 F20101124_AACUDS luallen_j_Page_057.jp2
128613efc8abc81483ea9f5f3690e69d
a53659297190af7554671b29364fa9264d016a7c
65122 F20101124_AACTYM luallen_j_Page_028.jpg
28a5cbcf856b7517eea7460c77e744d4
23ff0f008c3c90a9d5a28d1a47614551fc41600e
62492 F20101124_AACTXX luallen_j_Page_012.jpg
0a9282fc6e77083e93b9884c7bc1c1b5
6f75456fb86ce402334ad9f3819a36bf2f72bb46
82308 F20101124_AACUEG luallen_j_Page_071.jp2
5f7aad09a77cba59cfc6d5c2b8f1c7cb
ec1659dab43654871fe3f0268353e32bcfd3ceb8
65776 F20101124_AACTZA luallen_j_Page_044.jpg
553ef836d0b5ec10ee7e9d79c47a97bc
b13d732511663b56922df279cc03084365b0a1ed
103334 F20101124_AACUDT luallen_j_Page_058.jp2
f4b72270d4cadea85afdcd9a7fb0297d
8c4b311a60588fd69ba19a50305744b99d82875f
70123 F20101124_AACTXY luallen_j_Page_013.jpg
b20356dbd4a9b2915c42d0758752cc77
c81526e905dabb4794c62b2b92eb671f0434c8be
25868 F20101124_AACUEH luallen_j_Page_072.jp2
0ac47d2c5be79d1587df8170dc347155
76c0f4e17f124d3b6e97a6b910d44eff7c581b36
60311 F20101124_AACTZB luallen_j_Page_045.jpg
1623804138252582a50004993a4f1584
007b7bb4b8e7c55e9c3ac2642261b49fc6cf5378
100008 F20101124_AACUDU luallen_j_Page_059.jp2
06d7dcef72aedbfc932c4c238b047793
34bf8eb2d1a4065c5cc901a1fb73461f2dd1d7a4
75668 F20101124_AACTYN luallen_j_Page_029.jpg
dc1176bf0b953be50fe57ced35ac841c
20a0890e15715eaffce3e1ac0bfd525721e01844
72717 F20101124_AACTXZ luallen_j_Page_014.jpg
5e527533398d54779af02ef029262382
301153f7fdb3d17ffa3b73e2abc6b9800fc3e3e8
78542 F20101124_AACUEI luallen_j_Page_073.jp2
585b3a9d5589acc87e8d10332543d933
c76d78562f322db3ab496e140f7388a838df472e
55930 F20101124_AACTZC luallen_j_Page_047.jpg
84b657ba77cdc0d672cde7a94a66ab71
3120ac6c6ea29bc7bd81f4a20ccc9a7af666cd9e
116672 F20101124_AACUDV luallen_j_Page_060.jp2
908aec12e9a7b1f4ab9e355b396c885d
d07fc4e72dec07668de2e74b29fcb9a38ccea28f
68908 F20101124_AACTYO luallen_j_Page_030.jpg
b2cf0dc98731fccf8180059946e4ccd2
312afb85bdeac6398f94c7837486a1a6bc24ad39
27760 F20101124_AACUEJ luallen_j_Page_074.jp2
ed11bc4d45143d6ee0a167bd27fbe21a
2ef671ed289c1da79cf61823a74edadfcc444c14
37630 F20101124_AACTZD luallen_j_Page_048.jpg
c318d03a82f953a7cc9419148dcc7261
9dca976bec4a04749bdc509ccf29ce561b52f3e6
70553 F20101124_AACTYP luallen_j_Page_031.jpg
db8fd8966da6a06bb7575d3bb25e4ad1
d5ae2c81823374b046ea77a88671a7cfaa1b086c
101235 F20101124_AACUEK luallen_j_Page_075.jp2
39d1b083fa78777b8a6402318c28156b
713426da7467a144cf909e4c5b17769dfb47ba1f
45717 F20101124_AACTZE luallen_j_Page_049.jpg
588491b654a2da15e73f411a31093641
fa7497a2597816993606f58a641da2a7968a714a
109747 F20101124_AACUDW luallen_j_Page_061.jp2
9c01eabeafebad674d6406e3ae018499
2bc9c174878511019ab7507bc7cb35c39c89b494
66508 F20101124_AACTYQ luallen_j_Page_032.jpg
76e6c13c99ffff9dad4d38331de5513e
49d3239d58c0e93a75cba096aca5b9ef50248bcd
96400 F20101124_AACUEL luallen_j_Page_076.jp2
521173182fe3848be097e434504e71bd
1c2b65e61df4131eb97ab5dd92cf5da63746e351
30401 F20101124_AACTZF luallen_j_Page_050.jpg
74f81193bdb53431ed3abc1bcbbb14e7
745167cce371285f1b3fb7c9bc4cb33d60e5e031
72592 F20101124_AACTYR luallen_j_Page_033.jpg
4dfa3e1fe993e48c0be971cf1958af7d
1a6b48e11779f4f583f92b6b20c8f3ebfdeaa538
109265 F20101124_AACUFA luallen_j_Page_091.jp2
8085a5f39df54c0453d6e0b0602dab76
a7e5fdb948facde9201d879087e719f3f32eeebb
106301 F20101124_AACUEM luallen_j_Page_077.jp2
c065c36ef578559e77068fce55cd212e
dde1df929aa4ba8c77e7d5d47d2d9351de346b74
57186 F20101124_AACTZG luallen_j_Page_051.jpg
3f0e5a790cbe6786d3682b38fa252b2e
14fa1d0cf8e1f511a2e95efd11fc48577dd1dcc7
108377 F20101124_AACUDX luallen_j_Page_062.jp2
31507a2dec118495f038d69c237449f5
2342221b61a59d5e8324c8ad0bd0568cf19c96ab
70690 F20101124_AACTYS luallen_j_Page_034.jpg
b4221d4559bbccdbb199cd411fec79a8
983edb54ad9ecf060f30b25d369971c0210ce7fa
106875 F20101124_AACUFB luallen_j_Page_093.jp2
71eb04669b6c9306e22efa01a7d36da4
5df1f2193cfe562371defdd0aa13bc966c838c29
107541 F20101124_AACUEN luallen_j_Page_078.jp2
5a5db85ca42650e0e3ff6f0522617148
2b6f7dce43aeda3345df818ca933fbb7f3359a6b
48570 F20101124_AACTZH luallen_j_Page_052.jpg
89e0f1f0a5461888b9b548ba3c337265
21a0402600a6cd40f52cbc03e87d88769a440f51
104897 F20101124_AACUDY luallen_j_Page_063.jp2
0ac35556500c43cca60f9ce86d9742e3
679a7447f349d96964e3f0ad54c933683b16e7c0
77972 F20101124_AACTYT luallen_j_Page_035.jpg
80e5eb17f00b0496ccdaf9c581fb5059
a3105f3c4c9dcdd46366021a3ab5c206a9c96d33
97880 F20101124_AACUFC luallen_j_Page_094.jp2
f68b7564738319c1f65d0593a94b03b4
3c9ce88a8ea2f4300b1fe2da227d35f941fc7f0c
107442 F20101124_AACUEO luallen_j_Page_079.jp2
c3642dac4eecb46fb8abcf893dcc8c04
5d7595e3a258de9f7eec61b8b3259aa435f58955
59677 F20101124_AACTZI luallen_j_Page_053.jpg
10a3776a09c9a3f35f90a9a055150fa6
17ea1b7d76258cc040a7733196fd7365d0f5da27
108742 F20101124_AACUDZ luallen_j_Page_064.jp2
11b3b93c62b64c72551eee6c0a24b760
af005d88a6e3f897ba7138760ae2350e4394c1d2
71891 F20101124_AACTYU luallen_j_Page_037.jpg
20b692a8c6b1646151fbb15670d5326c
318313696d59dedca3980135e85c02ff81ebce91
42393 F20101124_AACUFD luallen_j_Page_095.jp2
f408adf4997d30112a38f7e423a61675
1e6d5d5306c5780d3d73401207a54a3f0c87453c
111350 F20101124_AACUEP luallen_j_Page_080.jp2
7a64703f8ccdb1a3fa5b0912af867497
80b72bd7d857fc7977c0c00fdc49fc00882c0a5b
71923 F20101124_AACTZJ luallen_j_Page_054.jpg
8a12f8f4ea27a9637ed41e799b3a49bf
ae79516a6e55cd6c9866e629ea7f0f107dc7d534
67684 F20101124_AACTYV luallen_j_Page_038.jpg
82fb60f8b42279fa0e5785ca5a06afb6
85b93dbd4d430538731c9ab933f722de7cc6b3db
53800 F20101124_AACUFE luallen_j_Page_097.jp2
4ab7351015dc12aa2fa2cc78937512f1
d91a8e2a2e93c7e379523fc694f24657e16dd525
106313 F20101124_AACUEQ luallen_j_Page_081.jp2
1de21e8b7194b0b0de78fe693d9db281
18dde5164f76bd54ef9d905df0355344ce36d1da
73289 F20101124_AACTZK luallen_j_Page_055.jpg
c1efe502ebaefdff0698f622ed865866
176684bdd841950b1005d0c7962cfcc35071122b
72871 F20101124_AACTYW luallen_j_Page_039.jpg
fbcad6cd7912c17e522a810cdcbd2dd7
0eef447c4b333b82401547a146d41ad10788621a
515671 F20101124_AACUFF luallen_j_Page_099.jp2
4728e2ae8a59076a8d4643fe068484fd
4123a12100f0bec4e415211a4309c700340f7bf2
113802 F20101124_AACUER luallen_j_Page_082.jp2
81dd5adb95b9fb02a4fc4c68388a4655
443d0700bd617d1edd09f272dce285351b44e876
60189 F20101124_AACTZL luallen_j_Page_057.jpg
1495662f409d16ec44faa2d836798609
f54e52b0ded587d61338db69266378283a9c93fd
71461 F20101124_AACTYX luallen_j_Page_041.jpg
8f532a8f875845d43467c78a0133bce4
5cc45988aef9bbc0534c3352f3c4718c8f310352
478019 F20101124_AACUFG luallen_j_Page_100.jp2
5f675f1278b75242d891c0dd41aaeef1
5fa40f776278faa6c983a7501dcf1cc1c0bd0703
105953 F20101124_AACUES luallen_j_Page_083.jp2
e6b98d66a52fe832104eac903a61b1d8
f97859b819fa75dc58149ee752ffc21aa6bf459a
67946 F20101124_AACTZM luallen_j_Page_058.jpg
5bb2c2eea280e7583c1dcb5e7da9a9d7
c8ca6b4155e212abc5459936c451a8da508bd407
71999 F20101124_AACTYY luallen_j_Page_042.jpg
ba2753a986a39b4d8dd390c86a136580
2d0ee604f4f27eee0332fdff42cd98b62de490b2
466030 F20101124_AACUFH luallen_j_Page_101.jp2
09bb3bb1dbb641830242143c98497fb8
f77ecf3f86976f1b621a1d9954ebd173f8fbae2d
105128 F20101124_AACUET luallen_j_Page_084.jp2
458f7ba5c941509d810b984d1e00764d
ac4ffd91b45b297b591af8d0b70cd2eabcb3192f
65146 F20101124_AACTZN luallen_j_Page_059.jpg
2d48253e6664fc62c61c26f911f55d42
6bb5b6736c926a1c7e9423de06c9abd062b0d499
65660 F20101124_AACTYZ luallen_j_Page_043.jpg
ccaf77026b4ddd32def711907a179569
dc2b843d3ac5153c807b6f628dc158f0b9ead391
431793 F20101124_AACUFI luallen_j_Page_102.jp2
672c3889fc178fe3125dcca5113f3f6b
d456cbae9d6805b7a0b8c38e7ee8644dfc551c23
96951 F20101124_AACUEU luallen_j_Page_085.jp2
ec64560016aad6804b52fcd4582cc33e
a9c40460a0a1d25fdd4e73bd9324ff9dec9a8fd9
236519 F20101124_AACUFJ luallen_j_Page_103.jp2
fe1d222c50d938d0be9e10dedc6142fb
26798347c7d1a9271bda76cfbab77941d6b5c2a5
112537 F20101124_AACUEV luallen_j_Page_086.jp2
bea274a99ab4a740d1e3fbdaa314ac9c
caf638874d697fc4018d024f058b6d45f77b6993
76156 F20101124_AACTZO luallen_j_Page_060.jpg
1ff429bd4b01931430e24cf750824543
8cca38b30f8f54d43ceadb87e950051c1fbb2c5c
140363 F20101124_AACUFK luallen_j_Page_104.jp2
5642027a032870802bdf96d31293aa54
e99cde564f7d5c241487b690e782baa5c5c009d6
94192 F20101124_AACUEW luallen_j_Page_087.jp2
e865a1a5e75e45be59dad91ae5634132
a1c900d28ae486b0c7821c9fc90c12339e174f6e
71934 F20101124_AACTZP luallen_j_Page_061.jpg
ec416ce80d857676094cb7f18d8b2e11
c6a18bda1807b682fbdd71caaff1f7dc5ef0a44e
139649 F20101124_AACUFL luallen_j_Page_105.jp2
6b55248729faccd7fa2dc89e80f90753
c2b2bbd722d9ed14cdcbe4287eebdcfb10afeb1f
110388 F20101124_AACUEX luallen_j_Page_088.jp2
b5102587bbe6286563ca659788aeae15
5c237e4debcf4f70330a765d98c0074d03ac2994
72014 F20101124_AACTZQ luallen_j_Page_062.jpg
b993793b354564fd5d75edc96e7c2840
fd77acf17de2b64bed5b87aa0182f7169cf25005
139681 F20101124_AACUFM luallen_j_Page_106.jp2
683bf101a0e387b9b671e7f755ddf9af
caa311507a5ce19ad41d54ad87137360185d12e2
68211 F20101124_AACTZR luallen_j_Page_063.jpg
64e5b58a019ba8cbfa98e9ed257f3ac6
89fa2cf9e3348306538c75f443279f8615d42281
F20101124_AACUGA luallen_j_Page_003.tif
ee3b59c43f98f32dba87342a57e933da
072f8152038db0f4ada360cfa5e3879b7e30c042
62421 F20101124_AACUFN luallen_j_Page_107.jp2
c5f007670ebda4b27ce31cb17967f0b6
d2d0e26a8f8880ae4de2a9300d5e322964bd1edd
95006 F20101124_AACUEY luallen_j_Page_089.jp2
60632581190de1bf4305308686bd1e9e
bb26b45ada883a02c415bb6bf87a1f5e0a1b7a1b
70946 F20101124_AACTZS luallen_j_Page_064.jpg
bd837f2d52476b8ef6165da39596b45b
713a6b62c1b082df6eac95b1431edbff4c27ec10
F20101124_AACUGB luallen_j_Page_004.tif
4f5ce4d83794e437631b6e917fd20794
8720570d955034f419788658bcd31db5e2cab9ee
575009 F20101124_AACUFO luallen_j_Page_108.jp2
0e486b1fb1d4d8f08c338ff8b5920b2f
ba12bbbd7d5f9d82e4a271dd7bc652bbdcb2f7f1
104850 F20101124_AACUEZ luallen_j_Page_090.jp2
83557d1c64009f88d60b5b97c162777a
757f2546d5e9c999855cbc3aee33752e9279ceae
75016 F20101124_AACTZT luallen_j_Page_065.jpg
f13e69cfcfa02cae8c5680bc16064012
d60d17d04e16db164abf73c3959969366d95795a
F20101124_AACUGC luallen_j_Page_005.tif
289b1c6c8bca0b526095621e607c1cf4
0d316f31ee27a24b3ce096a261824c902862f53c
7331 F20101124_AACUFP luallen_j_Page_109.jp2
46498e3064a8555a5a8b2fb18ef3fcf8
447e446227c6cd288b75ed0708d7ece19ca09ff6
69920 F20101124_AACTZU luallen_j_Page_066.jpg
99fb49766779b6225046eaeb8cee8303
67b3906cd28122209c5640a28e25a6f113346933
F20101124_AACUGD luallen_j_Page_006.tif
2c5f4856955be68128a09ff9b7f79706
bad12c1665f05f4ae18a6fb7b625fa4f63900a61
149071 F20101124_AACUFQ luallen_j_Page_110.jp2
c7b8a31dd59f8f6c01e75d7676a8ae57
8a4a81eabbee1682d74a11165111950fb293ea21
71403 F20101124_AACTZV luallen_j_Page_067.jpg
d79502b587d343e2e040fd4697f6af31
de376ce4940dc8794413d29c90dd25890d570e8c
F20101124_AACUGE luallen_j_Page_007.tif
abd4884f1af3fe1a83bba173bccd6a11
5de5dd93a023d1a267966c17e3359b758177cb25
140503 F20101124_AACUFR luallen_j_Page_111.jp2
615da2ca5a36c3cfb47d9e6e92595e1c
afb0e38a838a7971f974b269eaab4a7969c46242
72761 F20101124_AACTZW luallen_j_Page_068.jpg
f00e21e105516fd900f56908f42745d0
06284d0ce0d8a8393eeaa63df74a804fedfac56d
F20101124_AACUGF luallen_j_Page_008.tif
9ad70d68317c624c3e727af0f6bd24c4
8a25b7815a0df41eb9ce9b064f47d79fc76bd5f7
142237 F20101124_AACUFS luallen_j_Page_112.jp2
61f45aca474c4bf3c2cf287b051c5bbd
3ec166887d4a45da0aa2edcef07defa167b26ecb
69953 F20101124_AACTZX luallen_j_Page_069.jpg
f5f03393840e614afd5821999cd5b324
0b4567577d3bc1a572247a9d7bc0d724f2ea802c
F20101124_AACUGG luallen_j_Page_009.tif
e636dba5faeb7d4b1090ced4ef10e148
9f6a30e0d8f0f34c55c05f948b83e1686ca6e90f
140047 F20101124_AACUFT luallen_j_Page_113.jp2
738feec2c8caa59a4f6a8b954b539020
23c527ea48246dc246d4558096e7f5ca38adf697
14016 F20101124_AACTZY luallen_j_Page_070.jpg
f3ca0cc43a481e7670792c9124e10ccc
d1799b3e2b6b70bc4ce74e6cebd1434c50cd5835
F20101124_AACUGH luallen_j_Page_010.tif
a334ddf1ab6aed0bc28ccb96dd7ecbc0
0b0d8f3c81710acdfdef91c5307146ecc703a7e4
116906 F20101124_AACUFU luallen_j_Page_114.jp2
31733c9e662829e87f1284fbf2f54845
836c9484c023d907f7b1dd961f671ea7a8ad0344
F20101124_AACTZZ luallen_j_Page_071.jpg
e5ff2365a7fc7e475de349d52822dd45
4ad6dec4dcfe332237c26ecce29e97910b28e009
F20101124_AACUGI luallen_j_Page_011.tif
a97338211a34d11591e80207af16dbb9
b885df10b548f2877e9a414dfc4045a5ff060577
128053 F20101124_AACUFV luallen_j_Page_116.jp2
cbb22e4c7b1baa5316a02d646108446d
cb01e0b1166279bfc9d36167e6d81ec96f3b4de2
F20101124_AACUGJ luallen_j_Page_012.tif
e35ebd23ba52bfbd5fa96955d7a5c9c2
916417a888e730caccfc17546f90b6f212de202d
123588 F20101124_AACUFW luallen_j_Page_117.jp2
40aef13aa452f0f64dc81c6e94aa7c9d
bc49fa64df2e4cb38e1aed9ca045d20ea1ed692c
F20101124_AACUGK luallen_j_Page_013.tif
56507cab2d2f33a47fb5fbbb91d711cb
05c2c1f3b0284f207e4dc8702c47fb4cb3c786db
67875 F20101124_AACUFX luallen_j_Page_118.jp2
8d045c872f6bab191b29f5a7a3923e8e
e81c444b78e1ae1b0ee76aa040f019a8089a5476
F20101124_AACUGL luallen_j_Page_014.tif
a6e3bec4a9619cfc6359b3c45dca14f4
aac4efdfa245d2c004e0e84ce60b9a98cec76e1b
F20101124_AACUFY luallen_j_Page_001.tif
4a51f8c77113857d61a601ca049d92e5
ce6b74fa0a3c6ad08b941bbe195c216ca9f1b6aa
F20101124_AACUHA luallen_j_Page_031.tif
f987d452ee969bf3993857eef55e1d0b
31fd08aa2ae4f9b6f92d6bb8ec9a4a40e5bed340
F20101124_AACUGM luallen_j_Page_015.tif
43e5afacc6b7206a616253035d2563bf
b7889ff5e451acadd851011f4060801ecc01ab7f
F20101124_AACUHB luallen_j_Page_032.tif
77ef02758644f73c33c96bafbf353271
c792de45a3a4394cf02595994eea15cfc8bff578
F20101124_AACUGN luallen_j_Page_016.tif
e9409125d5f2424b52f18b4954793b2a
efb7ab46b96f8982c473069fb59961f94fc72fe8
F20101124_AACUFZ luallen_j_Page_002.tif
a3766276255050e533d7c79bfcb26dc1
04c32531788c6eb225d4beac87b73d4129f5d76b
F20101124_AACUHC luallen_j_Page_033.tif
52a04102f61d17ad731b3f9ca2c4b965
9996f3b63da2ecfb6e0c615676f1bbd657a44d2e
F20101124_AACUGO luallen_j_Page_017.tif
2535b4900f629ed4792b095c92de6a25
64d99732164c75a66f243ce4f374e377dc516ff3
F20101124_AACUHD luallen_j_Page_034.tif
fb7b9c88e7c435f1faa5a89c7a55e062
43dfd3e110ef57a59ce5156b296b5bd6e0e69a75
F20101124_AACUGP luallen_j_Page_018.tif
696ceed3de4e007f06647fe15273eb9f
66913746144e7b9ae912fcb41e0ff56f4f96605c
F20101124_AACUHE luallen_j_Page_035.tif
ebd3ab817cf281b6a5af090fd865aa40
2aad59f70683e98c95b39ba809a27227c9f94810
F20101124_AACUGQ luallen_j_Page_019.tif
2b3dcd5bc79d9baff54471c460c5ebcf
6d584d2a48aab2bf88e50ab71881af07a9a68646
F20101124_AACUHF luallen_j_Page_036.tif
05a1203fc48a8ca65cec4c3a6ece45be
fa2b15a8b7aa654babdf60bd4b98e36264bcc5a5
F20101124_AACUGR luallen_j_Page_020.tif
9b5e0daad2a4479216e4ef66112d4b49
0353d29334748aad7507cee29534408e9d9b4a0d
F20101124_AACUHG luallen_j_Page_037.tif
3d3fe292301e9293eee5d858d45e5028
ca08ad23f27e8fbbdf454680c157d29720f2e79f
F20101124_AACUGS luallen_j_Page_021.tif
0f1ce66022650991e569f19a498a2c16
5b6a340e6875f5b4222346f2e1f16795838ee781
F20101124_AACUHH luallen_j_Page_038.tif
930a594ee04403f394ea1d3a9286c611
0ed297c0aeec0cc5dcc9c7a96f081eca1bdf3fe4
F20101124_AACUGT luallen_j_Page_022.tif
e669b2bbdfc22756849d32331a027391
f1b2ceabed356660239f4e66989b9edb4ed4e79d
F20101124_AACUHI luallen_j_Page_039.tif
ba0ce9c6dcb5f22eda5fc447b658a7ce
789efece494895675407cbaf151b1f966d1ae14c
F20101124_AACUGU luallen_j_Page_023.tif
deba6f85d28777c365d9b6de6ca50349
816c61f1b0dac6abfb6686cfea259d9de2f8f57f
F20101124_AACUHJ luallen_j_Page_040.tif
0c56049cf2a093891d63f4a39db138ad
cd125e26bdd398ac004f8ef6a5e3bdd3c810eba4
F20101124_AACUGV luallen_j_Page_024.tif
48af20d82e041f66ed1bfcd87bad4696
bc0106319483cff4f5e48dfe1baabed8105776f7
F20101124_AACUHK luallen_j_Page_041.tif
3492fe6aa95d18ddd2a26c090296307b
3ad83b2b4f3a1dc78f033de348eebee258322fa9
F20101124_AACUGW luallen_j_Page_026.tif
9e5e482801136658e7d3239917d3f8e6
ce1a81d80040bfef0a21892f1284ce5eb61b82a4
F20101124_AACUHL luallen_j_Page_042.tif
15780c25c008065797655856039a107d
43cce7458f85855df329445fd886bbd823a7be88
F20101124_AACUGX luallen_j_Page_027.tif
95701c2a032522eaaf4161b84a65ca66
23f6e33aa3d22f49f1dc41c12fa71001a334706a
F20101124_AACUIA luallen_j_Page_057.tif
b004ba618600c2e039f367441b738ca9
179329946fdc42ca0b240f33529eedf7beaa2d37
F20101124_AACUHM luallen_j_Page_043.tif
dc0d849831a35c54ee1414dd8d1d3d18
a5c996c4c1f42b96f0a3a401376cf576cd376cc0
F20101124_AACUGY luallen_j_Page_028.tif
7bb69cf1dfac7e659f135c1dea16473c
21f0d471448e3a0d9adf8c780795d689b4977bd5
F20101124_AACUIB luallen_j_Page_058.tif
3d904683b596189c6f6f2fc87299ce7a
92ca8448ab05879fcc7d5f7ea7174791304e9e82
F20101124_AACUHN luallen_j_Page_044.tif
dd53b571963d6a18b802a9d9e73d0f0a
3b12f8aa6ce7369635256cd67d68e1451c351258
F20101124_AACUGZ luallen_j_Page_029.tif
a9b3baa57db77c21522d21829a1c5978
9abc74dfa196acb8964b96938b3ccbb09bf33b99
F20101124_AACUIC luallen_j_Page_059.tif
dd2fd4a9134bc191845678e67be34144
71e12cd98a5aaed28eed098ea6a30c0282130f67
F20101124_AACUHO luallen_j_Page_045.tif
79f2400e5b7558b098e4fab7270d378f
0ee6487aaa78c8937b36c2cfb449e46f0f450557
F20101124_AACUID luallen_j_Page_060.tif
5cb2705d9c52e1dd601e46d487e5650f
3d97036825e0568eceef5c267aa09cbd7815f13b
F20101124_AACUHP luallen_j_Page_046.tif
059c0b246133253eee793d604afd3908
7b58e47816e53ef918a35b99df1f114cab5d855c
F20101124_AACUIE luallen_j_Page_061.tif
47c7754b4fc4d090ec1ab498a3dfcf95
2f9f98d14cfbf4f2c0b50a7e7e3b2e1447828794
F20101124_AACUHQ luallen_j_Page_047.tif
c6ce7b219b3b668e6fc2a2366e25632e
cfa5c510332673bbef3fa4ceba0027202f31deb2
F20101124_AACUIF luallen_j_Page_062.tif
9d1642cf81b35d1b6a7917cd982dbe44
a885cf94459d34a1957aad9ce8cec7c945b94e22
F20101124_AACUHR luallen_j_Page_048.tif
ac0f1905c31f568b7e752ddc022b68ab
e29a9b2fd912de4c8d5e82b459eada527802c8d8
F20101124_AACUIG luallen_j_Page_064.tif
6522d8f631be4548f9ae99c9f4fff7f9
4ea071f5ad122f5474f94f8f8067e2beef9db8fb
F20101124_AACUHS luallen_j_Page_049.tif
0e77f3dcd42c3c519c651c766624299f
6bbec4339f89615c722706dd82566295cafbc32f
F20101124_AACUIH luallen_j_Page_066.tif
5e2d90fb5e4302bde8b03647becd9d5e
4574f4e2a8a0e70ed33308930d1b5873208be888
F20101124_AACUHT luallen_j_Page_050.tif
35d6380286764be6dff9daefd10ff31a
5296e0430b52449f67f93cecd6cf6104ecaf8e10
F20101124_AACUII luallen_j_Page_067.tif
0fcf94cd513e00a4445d41884b6d9e25
bdaa3bd2a8bce00c8fccf166e6fcc5a26c2721a8
F20101124_AACUHU luallen_j_Page_051.tif
6e1bbd40b548ad3b56f020a237710b1e
7d56bede2fba730ce9632a1caee1ad500c58eeb8
F20101124_AACUIJ luallen_j_Page_068.tif
f15674e6ebe995f53432ba99ee2fe7ad
315b634469e57a853ae2711f784f2f3fdaa31a76
F20101124_AACUHV luallen_j_Page_052.tif
816b4f094f8023dbf15e580c751b25a9
54ca9f325e85ea2abfd691b76c988e17f6907b5e
F20101124_AACUIK luallen_j_Page_069.tif
b43b22df2e23c7f42cd62619515923d7
a15e993d93345770b60a8e45119ad6b7c4068b5a
F20101124_AACUHW luallen_j_Page_053.tif
c9008a6a53e7e7e0cd27d85ce77687a5
31557cb83f8544d93011ca2728dbbff8bef5c284
F20101124_AACUIL luallen_j_Page_070.tif
08e6252135ed8271acde83cddd670bdc
bafb1b7c2a20f992fa72cc81160ba4a4610a613e
F20101124_AACUHX luallen_j_Page_054.tif
f5a974ade5583096f7255ac21ddec17f
2c333f7c7b0b41e092402b51ec5b8ea417dfb9c0
F20101124_AACUIM luallen_j_Page_071.tif
187107c6330e5158460f6ce280bb13f9
14307270fde03def7b324ca2cb18506573a076f3
F20101124_AACUHY luallen_j_Page_055.tif
95d81bb070c401641dbfb0381c6b8806
34f66bb85d73f6b94ed3182cfb5df0bd7f8384c3
F20101124_AACUJA luallen_j_Page_085.tif
b3af23c2d2ff65c2d464169512ee8ddc
b53f1805896f031fd343b8ee24f6ee4a1c9c2756
F20101124_AACUIN luallen_j_Page_072.tif
e892ee73d37ab7e47380d52b4a4b2be7
5ae89a6831c1842260e9c305d894975390a382eb
F20101124_AACUHZ luallen_j_Page_056.tif
432548e0b35ac4011703c93f27a90de8
ed8198e04ee81151cea0bc4e35e3d42a0c86b34a
F20101124_AACUJB luallen_j_Page_086.tif
5a141d24f9f26ee309bd24dc43b96ade
cf5799fe38b17468db5e4d2b2ab36e169eeda667
F20101124_AACUIO luallen_j_Page_073.tif
fd0441c06019474bef9bf290c52147e7
799a5bd43a1fc916ebe8d12df4652ac025583e4a
F20101124_AACUJC luallen_j_Page_087.tif
832bf7691d62846558f78e9330ef810f
ae9d7e04a564d7264295fd0ca649b3ab226705c2
F20101124_AACUIP luallen_j_Page_074.tif
49792c78233f46d108501a34218b8c2d
0547df2c6dc57f843210220dc79191c7c0f992d2
F20101124_AACUJD luallen_j_Page_088.tif
254d24e7f90c4fad2ffa9be48de00857
c7c4ef1fdee860e97f4a59e50d5fc1f8f8f68f68
F20101124_AACUIQ luallen_j_Page_075.tif
0836118344c7b424c5a8d55b7fb2ee0b
9a89e74715802e0019f1e113a3d5e16ba9e9fdc2
F20101124_AACUJE luallen_j_Page_090.tif
e4940fcd7296d8089f26fbd1dc9c139a
87217accdf83d7c6bb8f2ee2db7d54fd8eb49082
F20101124_AACUIR luallen_j_Page_076.tif
58753e07d35ba1fe8c1148dda6fe1d8d
99509b19ccc503cb387d851f20ff7a81616e768e
F20101124_AACUJF luallen_j_Page_091.tif
f7ca2057a6d2584a819b53d937968ea7
4edc618912cb3670fd3240140b655139ab58dc20
F20101124_AACUIS luallen_j_Page_077.tif
1ac6c649a4330d0911f69faaf0f57e61
fa22eb08fce82ee66da829ad133e2deed7219f0f
F20101124_AACUJG luallen_j_Page_092.tif
5aad8fbe92f7faee8f0ce2c84993ae46
77c5f10e71c81bb5ededb4c8b9097e2c6953f876
F20101124_AACUJH luallen_j_Page_094.tif
8c1a02d010e4695f35e868f8105ea5d8
08ca157db6d0a739149b2449a0b40ce0ca1339f8
F20101124_AACUIT luallen_j_Page_078.tif
f80c584338ba0393ac3ed3a7e86dbae4
a10e013a9713eec66ff19f791934888d6ffe73a4
F20101124_AACUJI luallen_j_Page_095.tif
9997c733f5c36cff676349001fe5b22b
42f844df5f8f41ea27f1ffc229cf42944d66cffa
F20101124_AACUIU luallen_j_Page_079.tif
57f96df1b9e53df94ce5b6498b5a63c9
2487de512c89b3d19a052b158cd98f5887b925e6
F20101124_AACUJJ luallen_j_Page_096.tif
45cf3aded8cc8cdfd5ed93592a43471a
01b116f2463319008c7929ea6dc6c52aa195dfb5
F20101124_AACUIV luallen_j_Page_080.tif
05ac073bfbe191bb4ba7e88ffd612733
ba5d27766c35630ce61ed45090e38e8558c1c2c2
F20101124_AACUJK luallen_j_Page_097.tif
08523410b1c57cf2e84b4c5184db601e
25be7ecf16f9b1762e133fd1041a6ca58589a99c
F20101124_AACUIW luallen_j_Page_081.tif
c07093a8e36ed9db341649782d504406
a4373d2f97094de90798c7e087b67ec5d5216a1f
F20101124_AACUJL luallen_j_Page_098.tif
84da7addff399face2071db1630fec4a
b978707c12766fea8bb65c9c8824bf575316de20
F20101124_AACUIX luallen_j_Page_082.tif
8455c5a1148d4dcc15362988c3cc94f6
57639cfe46a0a5f14865835f0c429bba37598ca5
F20101124_AACUKA luallen_j_Page_116.tif
23b75f5504927816957aa0d1f39b9f13
f45a58246cbb81fab2f0e8412a84563167705b5e
F20101124_AACUJM luallen_j_Page_100.tif
22b0a346da548e5b7f4e9e848d9c947a
6ed870c54bc32e956db18ab3fd140c2db9968665
F20101124_AACUIY luallen_j_Page_083.tif
3de2195955d1cda09a1d8bcc07d9d318
6b473ccc5318ab7a1084df173a69d3c63c2ae3e5
F20101124_AACUKB luallen_j_Page_117.tif
7ca55abd24eeec40f34e11687963e4fd
4cc5625a344d55717a3ab0f41d520de374ea538f
F20101124_AACUJN luallen_j_Page_101.tif
ebccf122c7fc560d785157bcab9959c6
9495e8891848ce728b331bb991af716a3ea14642
F20101124_AACUIZ luallen_j_Page_084.tif
541d8d248c55ce937bf0f020bc41e6ff
d6b145326cfa748037d31714f09217ef413c9e56
F20101124_AACUKC luallen_j_Page_118.tif
d75e6fab940ff8612c79ecf2a4561310
2fe186d5aa274c8e9363a2d7281523d66b32c551
F20101124_AACUJO luallen_j_Page_102.tif
c30403dfd12058d9a739c88cddffece3
6535bf59af208f003c2cb97b9edb90b82ff068de
8680 F20101124_AACUKD luallen_j_Page_001.pro
7f7f31e3481689447c97b227c02746b4
1c803e2bb8eae1c4ee68d1adb1beb2f86aba58be
F20101124_AACUJP luallen_j_Page_103.tif
548fb3301176cd9a2d00c4977967faa4
0f0e46e629a865750043990269bd5fd500e20ec5
1357 F20101124_AACUKE luallen_j_Page_002.pro
fbacb03a105587d71d8d28b341fd3e67
79715131c130c05043bf26228d45348465d2efa3
F20101124_AACUJQ luallen_j_Page_105.tif
42160e07cdb5687e2f19b0c118d4437d
fe8c3bb8073a91c9bb41b31d15a0dc8ebc815ab4
41241 F20101124_AACUKF luallen_j_Page_003.pro
6e0fc21973a888bebd2eb7ea86259ae7
a27e4b4d23ec7abab434af19c2fcbde246133b1e
F20101124_AACUJR luallen_j_Page_106.tif
41e205c9b9c252a2ddc80bcdb5b0a07f
3933598e58a20487bd0fdb6509fcb2f89786adba
25034 F20101124_AACUKG luallen_j_Page_004.pro
bb03565fd1406adc6744f7e0c1f1dfbc
534058671144761ac984beab025a09fdf084aa3b
F20101124_AACUJS luallen_j_Page_108.tif
e9688a34317513f50b71daaf2f076814
d47e5d260073af50f56150641982de1364ab9726
73703 F20101124_AACUKH luallen_j_Page_005.pro
955b6f93cf25a139bd0a9b288f801acd
d99d913eeaef8e8728726559ca28bcd046b2494a
F20101124_AACUJT luallen_j_Page_109.tif
6fbd4287325fef18263c8c99b4210777
757c5b79daa9fbcf6ed3204e7ddadf4b2bad919f
47417 F20101124_AACUKI luallen_j_Page_006.pro
6839aac9611d816d0ddbfac521f730ed
c2ea02c3e2df52faad98709bcb16175927c407eb
F20101124_AACUJU luallen_j_Page_110.tif
22bab0db42220be4a6de04e106b34a14
400c3e63f4c371d8e35dd7e5af868b00b626f22f
59030 F20101124_AACUKJ luallen_j_Page_007.pro
44c8b10ffe364d5c25db3e81ed1e2838
dc1822aa113ea05687d83b71355407babf209ba5
F20101124_AACUJV luallen_j_Page_111.tif
5a1aae0e7487e92eb030a9db2b20343d
e3fc349dd47e00dee07321e77ea3f2ae78f70daa
35940 F20101124_AACUKK luallen_j_Page_008.pro
e448b5fb96a9e543b45a46d7b6fdada7
136ef6c12d9ea0d23cec0cc278b499c250a87e44
F20101124_AACUJW luallen_j_Page_112.tif
9521f7a82150e7941a39889fdde8ed33
d373d6995254a443c0329d015643fe1bcc05c0e8
8199 F20101124_AACUKL luallen_j_Page_009.pro
90c8123cb5a4f7a1ae7d33e2e83eedde
3f5115dce5f29c88fc0d09ff85354cd8eb3ba134
F20101124_AACUJX luallen_j_Page_113.tif
af704b2cd22e50c24b9129aaf9b7d82e
df938b5b52e6c4c1a05960352cbde9921cbcffce
36374 F20101124_AACUKM luallen_j_Page_010.pro
55917c1e5d80bb1726d6e46f972bafdc
a94c4f09b5455b87e64e20628a3619bb8a6a5889
F20101124_AACUJY luallen_j_Page_114.tif
eb8e04f77c20567bbc52b6ace6b3b8f0
1160022a16cde66cbd6fd9173b6f4ab830ecb8b2
61051 F20101124_AACULA luallen_j_Page_026.pro
e1b10754855ff739b554aa485f80387c
a99a5df114b5354f75f08c23b02006cee28ca7d1
33258 F20101124_AACUKN luallen_j_Page_011.pro
fab4f1e139772aafbbf4022645fe9688
b40517401e9d744273100316c41248eb578b217b
F20101124_AACUJZ luallen_j_Page_115.tif
033eca0a0cb028b53b7accc21f29f291
6fddea85489a9614e88256acbfabb5ba280029b1
55297 F20101124_AACULB luallen_j_Page_027.pro
57591a0f382b2c7d7d190f455812a762
114836ac980fbd02247421dce7a87b92545d2c33
43114 F20101124_AACUKO luallen_j_Page_012.pro
5cda90fe3ef7b947ebf5fc1895739bc7
d96aafb71ddd94fe268bb7cbb3f740c1f0396e87
46750 F20101124_AACULC luallen_j_Page_028.pro
d70d4b838c4db9507243a3942fa02673
2c03f14b3780c06383b9be94ce97fc140196a30e
50033 F20101124_AACUKP luallen_j_Page_013.pro
2260c3572599868d4a14aada27d9b082
a427ba895b2918689def029318da9dd15979ccf0
58020 F20101124_AACULD luallen_j_Page_029.pro
052e1b089db19d957e17979c3a97b216
a4d10555a5d32df229fb2434da2aff6b2d17e51b
57685 F20101124_AACUKQ luallen_j_Page_014.pro
4e32a12e9247d286eba50b2c49530ecc
b608f4e9d8ba2633923394ac7c27b287b302fcc4
50321 F20101124_AACULE luallen_j_Page_030.pro
19c85aba0728c44e6d338ca347c2e0a6
5f071984f2f3762102a4dd5ecdcdf7e2af0bcfb7
49220 F20101124_AACUKR luallen_j_Page_016.pro
08b8e80a8b76b70b701a5dd3fdb1f734
eab4a52d83508236b24ca2fa0f3836c2ba1f035f
49636 F20101124_AACULF luallen_j_Page_031.pro
5db7ad574e9e26102db2b29d4d6dbcde
ce61dc5659b76526b37f293957e6a08eac75397d
60456 F20101124_AACUKS luallen_j_Page_017.pro
f922f2a56218c43fb8aafe7f0c8b656d
083ce435eda38d63244b1a354f7444b86f7bfa3f
46799 F20101124_AACULG luallen_j_Page_032.pro
d7770a8d6b5e29f8dfa1124e5903a97c
446489f0102cc271a520b2d2b7fd66be8f62faf8
60487 F20101124_AACUKT luallen_j_Page_018.pro
6227c75bfd02ef72c33ff2381e34742d
11800376a8c966ab5b7af9369ff92e1df48ef507
51287 F20101124_AACULH luallen_j_Page_034.pro
794955567c9ce4fb5a393c84c510b066
d49fd950cac6a59ac0ef95a9cf79e1daf1465c14
48038 F20101124_AACUKU luallen_j_Page_019.pro
2425d9d103ceaa6630ff0dfb7372e2f2
b85872241b2d0290aa4a1605fa4a9e7dbbf08ff4
66656 F20101124_AACULI luallen_j_Page_035.pro
588367a4955aea4c5c0959bf33127205
1efd6d0e8ff67022abdc6d9cab8f64280049853a
50857 F20101124_AACUKV luallen_j_Page_020.pro
761c98d8b3dd11b31ccd2c955a18bbf4
acc3cf9c754d9334463ae078de6d2eaceb9840c6
52247 F20101124_AACULJ luallen_j_Page_036.pro
b7fda0d56dbd70dff070289524108b7e
afcecf490e375773b95efd43b712cba8950df371
58017 F20101124_AACUKW luallen_j_Page_021.pro
613846335b3122dd85c19af87f82b13d
721e1aee59fe6fe3faa2281d47750cb2e35d3c93
51478 F20101124_AACULK luallen_j_Page_037.pro
1e64ba238db0e741e86044661c5acead
8e2700e220620f57c4f3f73fadd60c36fd99918c
47896 F20101124_AACUKX luallen_j_Page_023.pro
29f8f1a62ea81a9f787359411e8b66fe
56951bcd101c42471806f2b6ce383df791e6cbbb
47411 F20101124_AACULL luallen_j_Page_038.pro
678914fb51f9f15352cbf16c7dbde6e9
bea56683d21c849c3757409ace6b26a394e82715
49366 F20101124_AACUKY luallen_j_Page_024.pro
7d754e684478d80a5ef4d4bddbeae376
fe31fa90b43be9ba9ec55359ff817abd252e1cf2
51265 F20101124_AACUMA luallen_j_Page_056.pro
01795adb72ca7dc8a09970a8299c781e
e1391f66bae673686375f045a2646e7c64901432
52046 F20101124_AACULM luallen_j_Page_039.pro
11163654582d5e899688e3c7c0394fc3
5119d76910edd70c4b358a5ffbdffc3ee72028ea
49972 F20101124_AACUKZ luallen_j_Page_025.pro
ce57f913fc9e8246a1654e2d9ad72e25
6371ceabb7f81669f2625874947cce6639f173d5
41633 F20101124_AACUMB luallen_j_Page_057.pro
bc21587aa9e8f5cdf58a2a29e12ae926
2ef71627c74175f06d7bcafc52462d6bd0b5b48b
55422 F20101124_AACULN luallen_j_Page_041.pro
2ae759a95f39a83a9b892a73021f6385
e83cab1d46ba1a4acfdc0f5cbd9c7cb3c0f70379
48667 F20101124_AACUMC luallen_j_Page_058.pro
3ca2ac78412202601318b066730a432b
fb8192231e82b42b200d94ba75c39a65b3df9aa3
59889 F20101124_AACULO luallen_j_Page_042.pro
239032cac107a24070f978a3b47352d8
daae51c409f03646a6211bac62ea47f49020a971
46475 F20101124_AACUMD luallen_j_Page_059.pro
aa8129a0a3f10bbab9da294befcdf8ea
0517e366a5fc269d856a7d75d59d31c7330d2ffc
52825 F20101124_AACULP luallen_j_Page_043.pro
956156d2ce4f40ca4aac4794ac6135b1
3ac8fe6dea8a7831c54e7f87373844cb713654d4
57765 F20101124_AACUME luallen_j_Page_060.pro
db46c1118e8d824a4960a0b874ef7f8c
f336ab1b1cf545130a0962cccd3cae94e9d27b25
52623 F20101124_AACULQ luallen_j_Page_044.pro
d67b93d5c323eebacdf25d1c5c8dc01d
45ed706ea73dd7fc3542fcae480603aa051362d3
50852 F20101124_AACUMF luallen_j_Page_061.pro
0697a1445c67544f73dc47dd891999cd
6c8e86cace3265e9e5c846e4c7712a8cb5c9a98d
50951 F20101124_AACULR luallen_j_Page_045.pro
1e7206f30cea238998a9143d5bab65ef
5f58d0629b5d2a2c7eef4868e216561536e07581
50971 F20101124_AACUMG luallen_j_Page_062.pro
3cc71f908200f77aced35ec90229aa05
9b79ffa498b3ec7f07a9d539e4f34069a34345ab
42822 F20101124_AACULS luallen_j_Page_046.pro
70f39539d535a5248b198d7cb31e4a66
0e450b740aa60f6a99663e84513c4a4a0776deca
55493 F20101124_AACUMH luallen_j_Page_065.pro
67c6d63172b272ad8132ff7765ea534f
7dc58c09b78c293bc89cddac78d8ff2a104f1069
27509 F20101124_AACULT luallen_j_Page_048.pro
16c66adb768984b4ef67d8d23e8a543d
ab9fbbb8559b8905a837f87e34cba7099216e503
49808 F20101124_AACUMI luallen_j_Page_066.pro
6379140388fbcd756d6d699751b31fac
36c7c86b0954d6cdd30364474407eba76255fb7d
31390 F20101124_AACULU luallen_j_Page_049.pro
3225a07dfc9430b99328ff8ec689fa7c
6b37c2adcdeeb056dde155101c1ab847679986c3
50355 F20101124_AACUMJ luallen_j_Page_067.pro
5ed2720978ebb82c009f77e47e6c5036
daccb306c96ec1e9c5d2067bf4cfcfc6f18b8898
10893 F20101124_AACULV luallen_j_Page_050.pro
228c739560305d7c76798b0745829179
44b925efbc8c9fa1f4a4cc132d14820b67e995d2
51899 F20101124_AACUMK luallen_j_Page_068.pro
31b376f39fa59c7763758cfc64776947
a40721b8202061172448b0620021ba76557accc1
43120 F20101124_AACULW luallen_j_Page_051.pro
5587806a9600efebd2c0f77a83d9a726
44eb72ea32ad1ffcd67c1047430d7e032418351f
45549 F20101124_AACUNA luallen_j_Page_085.pro
01fdbf3bd765f07d0ae6a4f8942939d8
854683d378fb87aab1c8a50dca4d77d1e151c736
49835 F20101124_AACUML luallen_j_Page_069.pro
b2bfdea7bc41e8fb27531b6e7dbc178f
557240e1230e4c44e7df85e123f84679b61af25c
40360 F20101124_AACULX luallen_j_Page_053.pro
152281e9ff663ea704dfe843aef6f3c0
38b7e78051d80c370681c77f1fcf5c65b5f6f0ca
4496 F20101124_AACUMM luallen_j_Page_070.pro
c163b1c9682e29c427f489c2d5229faf
a139fbc58105b5852002f15c5c7b8c6b815eefd9
51138 F20101124_AACULY luallen_j_Page_054.pro
5142f4535193f1f2e67a15e1583539e2
d2ede89964d217bb728d0ad61b823b7bb1e01c49
59341 F20101124_AACUNB luallen_j_Page_086.pro
fc5c690f30e6150a4ce6880b5cc9eff4
0ff25270beff75feff7704b03b22599c7c80ab76
42565 F20101124_AACUMN luallen_j_Page_071.pro
2f42722e5f4bc760f5bdc75e74cfba89
6aee1cb897fc8f40c182cdd0db74c03d0a0dba88
53645 F20101124_AACULZ luallen_j_Page_055.pro
9b7d643af0db52811b25bff19e4a3f54
e6e7a81d1293033054b8077f665795265489c718
46529 F20101124_AACUNC luallen_j_Page_087.pro
28a6ee67cb8ddea3900be7e9d8a5814a
3100ba4dc8d89b7b5cea2174b32744d989386652
11945 F20101124_AACUMO luallen_j_Page_072.pro
7ab5f32f60deb913a1c4b400a22d4a75
7ebd88b16efee323f1d00810180f7ede3640cb2e
54986 F20101124_AACUND luallen_j_Page_088.pro
d0b64f65bf894d2acd09d911da0c9d26
300fd0b9beedaa3793e34ba0be480375d5f7fe31
37297 F20101124_AACUMP luallen_j_Page_073.pro
2999eadb49b99b52ba9d77ee7fb03786
57812222ffc778bac92e702a9e8910e1a6bc7470
46422 F20101124_AACUNE luallen_j_Page_089.pro
3971a33a08bde93339b965ed10f797ff
0866a20618613ade2fc349e48775c8abc7710c36
13414 F20101124_AACUMQ luallen_j_Page_074.pro
16c9ccff72696b9bf50ae1f1803bf586
dfb1b81301e36e7321e36c047045c252279084d6
48989 F20101124_AACUNF luallen_j_Page_090.pro
9e111e6e025e53d3d3d836b41907d362
bda18ed72ea56fe3477793c0054aaba97a674870
53592 F20101124_AACUMR luallen_j_Page_075.pro
015e642ee7316798092a1e290cdd9800
1d568a20c4f5cc26368350bd036bdcda199c94c8
50954 F20101124_AACUNG luallen_j_Page_091.pro
f4337763a81b832ac408c7bd44bf0b4c
2469ef9bbfb46eeaf60c9e72d1e93bfa8acdbfe3
44266 F20101124_AACUMS luallen_j_Page_076.pro
b6c07c1442df14886adab8f304387c4e
b9b756d08a527281bc40c44ead6f2a38b5d2d62d
52590 F20101124_AACUNH luallen_j_Page_092.pro
541ff6a2cd367025186e6b47bc5246f9
c565b0e3a43c1a2cd2117f044aeeced153a7900c
49883 F20101124_AACUMT luallen_j_Page_078.pro
7b2bcf328424921bf4cd83daef96f0d6
c3e8a81aeb521a3343b394ee60da00862b1c461f
50818 F20101124_AACUNI luallen_j_Page_093.pro
597a01f4a3ce0d93b53dd133cf6c8c56
9f14d976c3929a9df50d69971ce025a3f37d0dbe
50248 F20101124_AACUMU luallen_j_Page_079.pro
75f47d486be087868bd7ff11b71c616e
bbec73c3865d354b451c821fcfe7d1e0ae028d55
44706 F20101124_AACUNJ luallen_j_Page_094.pro
35856aeead8f71312ccb81288b7692fe
74af5c896801d648026f7742f020d09c675d114f
F20101124_AACUMV luallen_j_Page_080.pro
28c154243e890a5458f1a795b3737922
e405ecf2b1fccab58e19791684ec0690f9eacff7
20857 F20101124_AACUNK luallen_j_Page_095.pro
ad51e59f01b1dff3085d50dafb269edf
1fdcb414357c7f8c88fb92765955526edc6acd87
50521 F20101124_AACUMW luallen_j_Page_081.pro
76e9173875c700f301832f81926aa43b
24e1a3fb5475b565b40bc77983572235254c20a9
28595 F20101124_AACUNL luallen_j_Page_096.pro
1db71a1dc2833a06e40b4075965e4ea4
b7bb7cf3902e04375492e2ca09996d05a2c3b32b
59921 F20101124_AACUMX luallen_j_Page_082.pro
078af1162150525a8d33a05a32bb5cb7
2c2efc0209a3da54afea43ecf35c5d2e18da837b
92891 F20101124_AACUOA luallen_j_Page_112.pro
e1e0fa3a1f68e1c51765265b6749b982
a27f5c3c5724c6cde97436719dbd3fc7461a1585
28389 F20101124_AACUNM luallen_j_Page_097.pro
3e950b79dc548ecc88870472e59e42fd
fc6e087c697f272b10620714d83c2ccbbbac557e
49280 F20101124_AACUMY luallen_j_Page_083.pro
b5d4e75758fcf39e03b291859740678a
09572a3cd0aaf2f7c7233829c42263f3577666ca
90608 F20101124_AACUOB luallen_j_Page_113.pro
283eb0234ad4934483c34a50889f5a26
9b77e7f23a18dba6d6f949a76a3d64ff3b4721dd
77987 F20101124_AACUNN luallen_j_Page_098.pro
a77b406d9ba2b874d9c4603984e50af5
d88e4978d52f0eee5a7e804767bc6410d195876c
50072 F20101124_AACUMZ luallen_j_Page_084.pro
43ae7680667f4c707c71afec36fa1511
58985acbe4baaac8df5e2d02673dad704cadaf8e
16283 F20101124_AACUNO luallen_j_Page_099.pro
cef52dd62311d199edb5e47b68ba1677
6b384ee03c81a3e754bbc28c42f7e1efad1a5977
54118 F20101124_AACUOC luallen_j_Page_114.pro
9faa709cf226b9640a6c92e48e01c092
9c05b64ce5f2aa42b79090337461eb022ea11b3e
11214 F20101124_AACUNP luallen_j_Page_100.pro
f2d8c484b9304d5e438d66691d11b7c8
96bf3bbcaa5cc94a57dd21cd1a81c32c868533cd
62327 F20101124_AACUOD luallen_j_Page_115.pro
045c6f0a7bc84d4349d49d9c351a3f46
fce945e2c5b358edc01425b6c11b8d3a3d5d6170
59077 F20101124_AACUOE luallen_j_Page_116.pro
7b9cd109cf896c9ea12a25af952e24da
278f5289b399303226561a79eedcd4b957cef742
10458 F20101124_AACUNQ luallen_j_Page_101.pro
76738b22196693b8983629c643fb6634
3a43fcee7bb68cbbd4e0842eb90bbd4a0887258a
57079 F20101124_AACUOF luallen_j_Page_117.pro
8a5e5f8d1e2351c160ca893472798a61
ed309b0e208687ddb26877cc70652dbffb8190bf
9502 F20101124_AACUNR luallen_j_Page_102.pro
c5d264536ee6b60eae88adeceb2fddec
74814ed8678fee7dd44f1bc6c44d0290989136b0
29852 F20101124_AACUOG luallen_j_Page_118.pro
231ca3c5fcfdd7029f4be0819c9cc1a6
d3a1a6857fde166b9b66dd271fed6b148bd0ab3b
5193 F20101124_AACUNS luallen_j_Page_103.pro
97115227bf23c263ee8518ef5c8c862f
efda32f5848de69ab3ac45f3d1638f3b3dfd176f
473 F20101124_AACUOH luallen_j_Page_001.txt
3ccedd9c010a8b5b710ee1121625333c
e4cf0c3e78d2780ab2a951352a1c6cadcfa3f03a
78663 F20101124_AACUNT luallen_j_Page_104.pro
a23f77516d70bc3c47853f23b553a1fb
75676ed437d64c03892f80ce39ff92b27759226b
124 F20101124_AACUOI luallen_j_Page_002.txt
cd7671ea94fea3b879281de1a2ab0e6c
d5b6fd7fd2b118bd5cb9bdf8562d75ffbd31e348
75731 F20101124_AACUNU luallen_j_Page_106.pro
b5d74fedc79811b8d7b73d8b8bd1ad62
88292628b9a7c1625d0d1976d672e2b5ff812b36
3131 F20101124_AACUOJ luallen_j_Page_005.txt
647b0a500f955173697e5eaf0570d6f2
fe7a58ffb9f785a5cc3b206c28dfde17168c5c8d
47555 F20101124_AACUNV luallen_j_Page_107.pro
30ea03f39e657fc36fb66dab917755d1
27594313a604311b040a636b66625f3f71e97796
1962 F20101124_AACUOK luallen_j_Page_006.txt
89b4c0e2f968323099d62135cec3270e
8aaf34353bb1077f09d255804f00192213dad756
30482 F20101124_AACUNW luallen_j_Page_108.pro
a08eb0e28ff3cfeb308ed7a9f767cec2
6be694fca3e53032a49da111136c258ba01a46a3
1896 F20101124_AACUPA luallen_j_Page_023.txt
5054fdbf0cf483280e7056ded542250e
e7648a7eb179724a85c6ab7fedd9a1919c5e470b
2388 F20101124_AACUOL luallen_j_Page_007.txt
201ddb1019e3bab38f00bde861d9b8a4
c9b015bccc79671ac8b18f2fafa9726b89b014b5
1654 F20101124_AACUNX luallen_j_Page_109.pro
cac0d7c3412f87a485baace0d42be18b
11e998097d2b7dbeb2030fd872739278c4fd0ef8
1986 F20101124_AACUPB luallen_j_Page_024.txt
715169a498fb60cb134bc625820ad456
628071bd79392b2bf0ebd7a97dc36de6b9120934
1436 F20101124_AACUOM luallen_j_Page_008.txt
8f124214436d54dc275b9c4b751db843
72525cb90ec786d875ca6151f59e12c272625922
85668 F20101124_AACUNY luallen_j_Page_110.pro
a14a55c5b3910b7df6e75baa7f54bc50
acdb4bf68b4bf5a130dbb13ebf9b88db5a918b69
2366 F20101124_AACUPC luallen_j_Page_026.txt
fdb2d6dd567c47ae9de0ae734b503e16
b43c855192ba5f2d7b9853dfefefa229362d747f
350 F20101124_AACUON luallen_j_Page_009.txt
0b7a3f218326e322ccb062a186fbd06b
91e0cc9f3354c20b4ac75cf90571230036969611
91840 F20101124_AACUNZ luallen_j_Page_111.pro
292f12f59ebb2b86cd566829c17970f9
566822295124d8b417f40891921e430fce02afbd
1621 F20101124_AACUOO luallen_j_Page_010.txt
587918ded006c75dbd275cb78b519b84
e7aec36618495ad87778dac71506089f2f9f4fd5
2165 F20101124_AACUPD luallen_j_Page_027.txt
ce552752d33002cf3c15257b40b7555b
4ebba5234092f4311e6e5e57bd7b21762e87efb4
1325 F20101124_AACUOP luallen_j_Page_011.txt
accf160b4b6d9ba1d5756cb4c0c6b926
4b3fedab70a164000d564b6ba59d2fc4fd6c08f2
1854 F20101124_AACUPE luallen_j_Page_028.txt
303419942658a2ff9bb0628d6a1438b7
f973dbbcc6f7bd4e71df9d07748c20ff22ecc138
1805 F20101124_AACUOQ luallen_j_Page_012.txt
e26f5ea2a42c82c8437f665a3ac78f69
58c76bf614bfe3ba64d662e49e229787aa6f4308
2260 F20101124_AACUPF luallen_j_Page_029.txt
2cdbb0b5c728e9c3596dde1495bdf0dc
6a65cb117b4498a54843e7a3a419b4f031c50347
1975 F20101124_AACUOR luallen_j_Page_013.txt
c85baa57f18cc45ba504cb42bb9d3a5d
c544097d3f81126bbbc7085e3450bce5b009eaee
1987 F20101124_AACUPG luallen_j_Page_030.txt
59a1f62a148d5263efadc5a29f2ec84b
e759827d4753a35d2856e1bca35634940dce7534
2274 F20101124_AACUOS luallen_j_Page_014.txt
6bce89003710c8e73e54dd62655bd6b3
1c862b1c133ee8ee61d8bd51ee5633909aed7366
F20101124_AACUPH luallen_j_Page_032.txt
4098da3f87b22b1c7b917d83017e51d1
cccd872ee2ccb3fbdeabf1e27d4118501218ee9d
2515 F20101124_AACUOT luallen_j_Page_015.txt
c132699785c84cfe815c32bbedc08529
465595beafc5600bd8d1a3fc148292c06f360645



PAGE 1

UNDERSTANDING EDUCATION: THREE ESSAYS ANALYZING UNINTENDED OUTCOMES OF SCHOOL POLICIES By JEREMY CLAYTON LUALLEN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Jeremy Clayton Luallen

PAGE 3

iii ACKNOWLEDGMENTS Throughout the course of my graduate and di ssertation work here at the University of Florida, there have been many friends, re latives and colleagues wh om I have relied on for much professional and emotional support. It is for this reason that I wish to acknowledge the contributions of all these important individu als, who were key to the success of this dissertation. In addition to the academic and financial support these individuals have provided, they have also made outstandi ng sacrifices to better my growth, not only as a professi onal, but also personally. First and foremost I would like to recogni ze my parents Dean and Nancy for their amazing support of my efforts. Their guidance through the years has bettered me in ways that I cannot even begin to measure. I w ould also like to thank my fiance Jaclyn for never-ending support of me and my work over the past several year s. I would like to specially acknowledge Dr. Lawrence Kenny for his wonderful friendship and guidance as a mentor over the years. I would also like to thank my sister Jessica who has inspired me in many ways. There are also many othe rs I would like to acknowledge for their assistance in making this disse rtation a success. I would lik e to thank Steven Slutsky, David Figlio, David Denslow, David He dge, Randall Reback, Richard Romano, Sarah Hamersma, David Brasington, David Sappingt on, Bill Bomberger, Chunrong Ai, Scott Hankins, Mark Hoekstra and Jim Dewey for th eir many comments, incisive criticism, and overall enthusiasm for my research.

PAGE 4

iv I would like to thank Anne Stahl at the National Juvenile Court Data Archive for her guidance and patience in help ing me to acquire data for th is project. Thanks also go to Kathy Budge, Dennis Small, and many othe rs at the Superintendent of Public Instruction Office in Washington State for pr oviding supplementary data that has been crucial to the success of my research. I woul d also like to thank Martha Haynes and the Staff of the Education Information and Acc ountability Services di vision of the Florida Department of Education, who supplied vari ous reports and data utilized in this dissertation. I would also like to thank Kerry Gruber and the entire staff at the National Center of Education Statistics that supp lies and distributes the Schools and Staffing Survey. I am also especially grateful to a ll of the teachers and administrators throughout the State of Florida who shared their time, their thoughts and their experiences with me.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES............................................................................................................vii LIST OF FIGURES...........................................................................................................ix ABSTRACT....................................................................................................................... ..x CHAPTER 1 SCHOOL’S OUT…FOREVER: A ST UDY OF JUVENILE CRIME, ATRISK YOUTHS AND TEACHER STRIKES.........................................................1 1.1 Introduction..................................................................................................1 1.2 The Identification Strategy..........................................................................3 1.2.1 Zip-Code Matching..........................................................................5 1.2.2 The Basic Model..............................................................................8 1.3 Data..............................................................................................................9 1.3.1 Methodology..................................................................................10 1.3.2 Defining the Parameters.................................................................11 1.4 Regression Analysis...................................................................................13 1.4.1 Community Differences.................................................................16 1.4.2 Differences in Offense Types........................................................18 1.4.3 Types of Offenders........................................................................19 2 THE ROLE OF TEACHER NETWOR KING IN TEACHER TRANSFER DECISIONS AND TE ACHER MOBILITY.........................................................42 2.1 Introduction................................................................................................42 2.2 Discussion of Networking..........................................................................43 2.2.1 How Teachers Network.................................................................43 2.2.2 How Teachers Use Their Networks...............................................45 2.3 The Basic Model........................................................................................46 2.4 Data and Empirical Study..........................................................................47 2.4.1 Defining the Parameters.................................................................48 2.4.2 Empirical Testing...........................................................................52 2.5 Robustness Checks.....................................................................................55 2.5.1 Re-examining the Data..................................................................55

PAGE 6

vi 2.5.2 Unobserved Correlation.................................................................57 2.6 Conclusion.................................................................................................58 3 WITH A LITTLE HELP FROM MY FRIENDS: EVIDENCE OF TEACHER NETWORKS USING MICRO DATA..................................................................65 3.1 Introduction................................................................................................65 3.2 The Identification Strategy........................................................................67 3.2.1 What Are Networks And How Are They Useful...........................67 3.2.2 How Do Teachers Build Networks?..............................................69 3.2.3 Measuring Teacher Networks........................................................70 3.2.4 The Basic Model............................................................................72 3.3 Data............................................................................................................76 3.4 Empirical Evidence....................................................................................77 3.5 Robustness Checks.....................................................................................80 3.6 Conclusion.................................................................................................83 APPENDIX A SUMMARY STATISTICS AND SUPPLEMENTARY REGRESSION ANALYSIS............................................................................................................96 B COMPLETE RESULTS OF SPE CIFIED REGRESSIONS.................................98 LIST OF REFERENCES.................................................................................................103 BIOGRAPHICAL SKETCH...........................................................................................107

PAGE 7

vii LIST OF TABLES Table page 1-1 Negative Binomial Regression onto Total Crime with Full Data Set....................30 1-2 Urban Subsample Negative Binomia l Regression onto Total Crime....................31 1-3 Suburban Subsample Negative Bino mial Regression onto Total Crime...............32 1-4 Rural Subsample Negative Binom ial Regression onto Total Crime.....................33 1-5 Effects of Strike Days by Crime Type and by Community Type..........................34 1-6 Effects of Strike Days by Crime Type and by Offender Type..............................35 1-7 Temporal Displacement of Crime by Crime Type and by Offender Type............36 1-8 Temporal Displacement of Crime through Lagged Strike Variable......................37 1-9 Changes in Total Juvenile Crime, by Da ys Elapsed since the Start of a Strike.....38 1-10 Effects of Strike Days with Sp ecific Years Dropped from the Sample.................40 1-11 Average Effect of Randomly Gene rated Strike Days on Total Crime...................41 1-12 Effects of Strike Days by Offender T ype with the Last 4 Years of the Sample Dropped..................................................................................................................41 2-1 Ordinary Least Squares Regression ont o teacher transfers using full sample with and without robust standard errors.................................................................60 2-2 Percent change teacher transfers pe r school for a one standard deviation increase in each significan t variable with full sample...........................................61 2-3 Ordinary Least Squares Regression ont o teacher transfers using sample with biasing outlier omitted...........................................................................................62 2-4 Outlier excluded percent change teacher transfers per school for a one standard deviation increase in e ach significant variable......................................................63 2-5 Ordinary Least Squares Regression of other types of teacher movements with outlier excluded......................................................................................................64

PAGE 8

viii 3-1 Professional Development Ac tivities Summary Statistics.....................................84 3-2 Summary Statistics fo r Various Teacher Groups...................................................85 3-3 Probit Regression of Intra-District Movement......................................................86 3-4 Probit Regression of Intra-District Movement......................................................87 3-5 Within State Probit Regression of Interdistrict Movement....................................93 3-6 Out Of State Probit Regressi on of Interdistrict Movement...................................94 3-7 Probit Regression of Intradistrict Mo vement with District Fixed Effects.............95 A-1 Strike Summary Statistics......................................................................................96 A-2 Regressions of Full Data Set excludin g Zip Codes with less than 500 students...97 B-1 Reported on Table 3-4...........................................................................................99 B-2 Reported on Table 3-5.........................................................................................100 B-3 Reported on Table 3-6.........................................................................................101 B-4 Reported on Table 3-7.........................................................................................102

PAGE 9

ix LIST OF FIGURES Figure page 1-1 Average Percent of Total Arrests as Function of Days between Offense and Capture ...................................................................................................................39 3-1 Comparisons of Teacher Characteris tics Along the Distribution of PDA...............88

PAGE 10

x Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy UNDERSTANDING EDUCATION: THREE ESSAYS ANALYZING UNINTENDED OUTCOMES OF SCHOOL POLICIES By Jeremy Clayton Luallen August 2005 Chair: Lawrence Kenny Major Department: Economics The goal of my study is to examine speci fic school policies to determine if unintentional consequences resu lt from these policies. Specifically, I focus on two main issues as they relate to student and teachers outcomes. I begin by looki ng at the effect of incapacitating juveniles in school as a force influencing juvenile crime. I exploit teacher strikes as a measure of unexpect ed student absence from school to measure the effect of school in preventing juvenile crime. My data set consists of information on every juvenile arrest made in Washington State ove r a 22-year period. I show that previous estimates of the effect of school incapacitation are systema tically underestimated, that criminal activity increa ses as students continue to remain out of school. I also show that these increases in crime reflect an increase in overall crime, not a di splacement. Lastly, I show that repeat juvenile offenders are more likely to have committed their first crime on a strike day, relative to a normal school day.

PAGE 11

xi Chapters 2 and 3 of the study focus on the role of teacher networks in influencing teacher mobility. Specifically, my study deve lops a model of teacher networks that describes how teachers assemble networks th rough professional development activities (PDAs) and how these networks provide an effective sorting mechanism for public school teachers. I empirically test the exis tence of teacher networks with 2 distinct datasets. The dataset in Chapter 2 is compri sed of various reports covering all 67 Florida school districts. Besides examining how professional development affects teacher movement, I am able to exploit the macro nature of the data to compare district characteristics (such as differences in compen sation levels and school district density) to examine how these factors also influence teach er mobility. The dataset in Chapter 3 uses survey data from the “Schools and Staffing Survey” and includes over 17,000 teachers. The high-powered nature of this dataset allows me to identify specific details, such as teacher salary incentives, individual networ k strength and union membership. Ultimately I conclude that teacher networks are an integr al part of a teacher’s transfer decision and have a sizable impact on intra-district teacher mobility.

PAGE 12

1 CHAPTER 1 SCHOOL’S OUT…FOREVER: A STUD Y OF JUVENILE CRIME, AT-RISK YOUTHS AND TEACHER STRIKES 1.1 Introduction Although we have taken important strides in understanding the economics of crime, there remains a great deal which we have yet to fully understand. The focus of this chapter is juvenile crime in pa rticular. Because juvenile crim e can be especially hard to study, due to general limitations on access to da ta, economists still have a lot to explore on this front. Over the past 20 years, economists and soci al scientists have attacked the problem of crime in four distinct ways. Specifically they have analyzed the potential effects of deterrence, retribution, rehabilitation and incapacitation as forces for reducing crime (Ehrlich 1981).1 Deterrence, a widely explored topic, stresses the importance of imposing penalties as an effective means of preventing crime, because the perceived costs to criminals of crim inal activity increases (Mocan and Rees 1999, Levitt 1998, Freeman 1996, Ehrlich and Gibbons 1977, Loc hner 2003). Retribu tion addresses the actual punishment of criminals. It suggests th at a criminal experiences an increased cost to his crime when a punishment is imposed on him and/or other criminals (Levitt 1998). Rehabilitation stresses the importance of reforming criminals, through treatment and rehabilitative programs, in preventing future crime (Cuellar, Markowitz and Libby 2003). Finally, incapacitation deals with the notion that the physical lock-up and detaining of 1 I am not arguing that these four ideas are mutually exclusive ways of preventing or reducing crime.

PAGE 13

2 criminals may be an effective means of pr eventing crime (Freeman 1996, Lochner 2004). This paper will closely examine the role of incapacitation in preventing juvenile crime. Specifically I deal with the effect of in capacitating juveniles in school on juvenile crime. Brian Jacob and Lars Lefgren (2003) we re the first to addre ss this topic in their paper, “Are Idle Hands the Devil’s Wo rkshop? Incapacitation, Concentration and Juvenile Crime”. They examine the effect of school attendance on juvenile crime by using teacher in-service days, days where teac hing professionals are required to attend work when children are not requ ired to be at school, as a source of variation in student attendance. While their use of in-service days is very clever, I utilize teache r strikes as a source of variation of studen t school attendance to gain additional insights into the relationship between juvenile crime and school incapacitation. One primary advantage to using teacher stri kes over in-service days as a source of variation in school attendance is that strikes often occur in blocks larger than one individual day. This allows me to observe the effect of prolonge d absence from school on juvenile crime. If juvenile criminal habits do change as absence from school increases, then a one day absence cannot reflect the averag e effect of school incapacitation. Another advantage of utilizing teacher strikes is that they are relatively unpredictable. Since in-service days are pl anned at the beginning of the regular school year, parents have adequate time to make arrangements for thei r children during these off-days. In contrast, school closures caused by teacher strikes are often reported in local newspapers only a couple of days before they are likely to take place, and even then, there is no guarantee that these reported strikes will actually occur. Additionally, a

PAGE 14

3 parent’s ability to plan for their children’s activities during strike days is complicated by the fact that details revealed in newspapers and other information sources may prove to be inaccurate or unjustified.2 The late-breaking and incomplete nature of information leaves parents little time to plan their children’s activities during the days of a teacher strike. Thus teacher strike days are likely to result in more unsupervised students, and more juvenile crime. The main drawback to this use of strike as variation is that t eacher strikes are often very sparse, and mostly occur in a only a few school districts. In order to overcome this difficulty, I rely on data from Washington State. Washington provides an ideal environment for this type of study because it has an extensive strike history as well as a detailed juvenile arrest data se t that dates back as far as 1980.3 1.2 The Identification Strategy As stated earlier, the objective of this anal ysis is to measure the impact of school incapacitation on juvenile crime rates by using t eacher strike days as variation in student absence from school on an ordinary school da y. Arguably, teacher stri kes are a source of variation that is exogenous to variations in j uvenile crime. Issues that lead teachers to strike include pay, class size and teacher planning time, none of which are likely to influence daily variations in juvenile crime.4 2 For instance, a school district may report that it is sure teachers will return to work when a court injunction is issued, and subsequently the teachers may defy the injunction. 3 Unlike most states in the U.S., teach er strike activity in Washington rema ins very much active to this day. The most recent strike event took place in 2003 in Marysville school distri ct. This strike lasted nearly two months (approximately 50 total days). A summary table describing teacher strikes in Washington can be seen in Table A-1. 4 In fact, areas where juvenile crim e may be an increasing problem are ar guably less-likely to see a teacher strike, because greater crime has been shown to be positively correlated with higher teacher pay (Grogger 1997).

PAGE 15

4 In order to draw direct comparisons between crime on strike vs. nonstrike days, I must be able to control for days in which schools are closed for other reasons. This would include days that coincide with sp ring, summer and winter vacations, weekends, teacher in-service and half-days, national holid ays, etc. However since the data cover a large span of time (22 years), and since each Washington school district determines their own school calendar, it is impossible to know where all of these nontypical school days occur.5 Further, a prolonged teacher strike sometimes results in a reworking of the existing school calendar (largely unobservable) so that a school di strict can meet the requirement set by the state that the school year last for 180 days.6 Since I cannot control for these school holidays with complete accu racy, I must include only those days which I am strongly confident students are regul arly scheduled to be in school. To minimize any possibility that school ho lidays could bias the results, I take extra measures when eliminating questionable da ys. I begin by eliminating weekends and national holidays.7 If a national holiday falls on a Saturday or Sunday, I eliminate the preceding Friday or subsequent Monday, respec tively. The first two Fridays of October are dropped because they are both traditiona l days for the state mandated teacher inservice day. The entire months of June and July were al so excluded, and only the last three school days in August are allowed in the sample. Finally I elim inate the first week 5Each school district may vary on the scheduling of breaks, half-days, inservice days, etc. To the best of my knowledge, there is no state or federal agency th at has archived these calendars over the 21 years. Since the advent of increased techno logical integration with school systems, some districts are beginning to electronically archive these calendars, however these cale ndars generally only go back a couple of years. The Washington State Superintendent of Public Instruction was able supply me with sufficient calendars for most districts, for each year from Aug. 2000 to June 2004. 6 The shortening of spring and winter breaks is comm on, as well as the lengthening of the school year into summer vacation. 7 Thanksgiving falls on Thursday, however both Thursd ay and Friday are dropped from the sample because students are given both of those days off from school.

PAGE 16

5 in January, the last two full weeks (at least) of March, the first tw o weeks of April (at least 8 school days), and the last two full w eeks (at least) of December. Based on these criteria, I can be reasonably certain that each observation in the analysis is a typical school day.8 Days which are included in the sample but are not ordinary school days will serve to downwardly bias the results, because more juvenile crime should be naturally occurring on those days. This is only true because this measurement of strikes as a treatment is limited to the days where students have missed an expected day of school.9 1.2.1 Zip-Code Matching My data set consists of individual-arrest da ta, where each juvenile is matched to his home address zip code. Since zip codes are set by the Addr ess Information division of the U.S. Post Office, they look very differe nt from other boundari es like county borders, congressional districts, and most importantly school districts. In order to match each juvenile (and his/her arrest) to his/her a ppropriate school district, I must address two problems. The first problem is that zip code s often spill over multiple school districts. Therefore it is often uncertain which school district a juveni le is in, given their home zip code. The second problem is that zip codes ar e redefined over time. First I will focus on how zip codes change, and how this problem is resolved so that identification strategy is preserved. As cities and towns develop both inside a nd outside city limits, the Post Office creates new zip codes and new zi p-code offices to handle increasing mail traffic. When 8 This is not a perfect matching process because half days and minor differences between districts are impossible to pinpoint. 9 The timing of strikes is very particular and seldom overlaps with nonschool days, except weekends.

PAGE 17

6 these new zip codes are created, they are draw n strictly as subsets of existing zip codes.10 The creation of these new zip codes makes zipcode fixed effects useless from one year to the next. In addition, it is also very difficult to pinpoint when new zip codes are created. However it is possible to see which zip codes were create d and what preexisting zip code they were originally a part of. By using zip-co de maps of Washington State from 1984 and 2000, I am able to map each subset zip code back into its original (1984) zip code.11 Essentially I treat zip codes in Washington as if they are never partitioned, beginning in 1984. This method leaves me with 510 zip codes in total and a stable zip code definition over a 22 year time frame. The second problem is matching these zip codes to their corresponding school district. Zip-code overlap with school dist ricts make this matching process difficult, however the approach I take is sound, and fair ly straight forward. I began by totaling the number of schools in each zip code noting which district they serve.12 I then divided the total number of schools in that zip code for a given district by the to tal number of schools in that zip code, regardless of their district Essentially what I am left with is a probability, that I will call p *, which reflects the probability that a child living in a particular zip code is also part of a particular school district. This p measure allows me to assign a proper strike tr eatment to zip codes. 10 So far I have not been able to find an instance in Washington State since 1984 where a zip code was created from two or more existing zip codes 11 Four maps of Washington State zip codes were provided by the Western Economics Research Co., Inc. courtesy of the Suzzallo Library at the Washington Library. They encompass they following years: 1984, 1989, 1992 and 1994. Zip code maps of Washington State for the year 2000 are made available by the Census. 12 I did this using a master list of schools (as of the 2003-04 school year) that includes each school’s address, grade level served, school code and district name. This list was provided by the Office of Superintendent of Public Instruction in Washington State. It is also publicly available.

PAGE 18

7 There are several features of zip codes th at simplify this matching process. One helpful factor is that many zip codes do not ove rlap school district bound aries at all. This leaves out any guess work. Another helpful poi nt is that zip codes and school districts often share a significant numb er of common boundaries. Th ese shared boundaries occur where divisions seem logical, such as along a county boundary, or a major river or highway. These convenient features lessen the confusion of the matching process. It is also worth noting that not all zip codes contain schools.13 In my sample there were four zip codes without any schools; two were in th e heart of Seattle, and two were located in rural areas. Conveniently, none of these four zip codes overla pped school district boundaries, so assigning them to the appr opriate school distri ct was simple. The last concern I am le ft with deals with how school districts change over time, if at all. If school district boundaries are changing frequent ly and dramatically over time, then the matching process breaks down. Howe ver, what I find is that school district boundaries are stable over time. In order for a district to change its boundaries, it must engage in costly and time-consuming legal pr ocesses, however that does not imply that such adjustments never occur.14 In fact, the two most freque nt changes to school district boundaries seem to be district cons olidation and district dissolution.15 Neither changes turn out not to be pr oblematic. These changes only occu r a total of four times over the 13 The zip codes I use in the sample are zip codes with residences. Many zip codes like business zips, P.O. Box zips, and other nonresidential zips, do not contain schools. These kinds of zips are not included in the data sample. 14 For an outline of regulations and requirements surrou nding district boundary changes, please refer to the publication, “Changing School District Boundaries: A Lay Person’s Guide” published by the Washington State Board of Education in conjunction with the Office of Superintendent of Public Instruction 15 Consolidation is when two or more districts form a new superdistrict. Dissolution is when a one or more districts are absorbed in to existing districts.

PAGE 19

8 course of the 22 years and involve districts wh ich do not experience any teacher strikes. I treat these integrated districts as if they had never been separate. The school district zip code matching pr ocess I invoke does not provide perfectly accurate matches in all cases, but I can be certain of how any mism atching will bias the results. If I say there was a st rike in an unaffected zip code, th e effect of the strike will be biased toward zero because juvenile crime s hould be unchanged. If I say that there was not a strike in an affected group, then the stri ke effect will fail to pick up any increase in juvenile crime rates. Any mismatching that arises from the imperfect matching process will downwardly bias the results. 1.2.2 The Basic Model In this basic model I am attempting to explain changes in juvenile crime for ordinary school days as a function of teacher strikes. I can start by expressing a simple regression model in the following form: Juvenile School Day Crime = + ( p *)(Teacher Strike) Again, all of the information from the crime data is reported at the zip-code level, however the teacher strikes occur at the schoo l-district level. To adjust I introduce p where p* represents the probability that the student population of a zip code is treated by the teacher strike. Rather than have fixed effects to take into account differences across zip codes, I want to initially include specifi c zip-code characteristics to make sure the data are well-behaved. I consider income levels, welfare status, parental education, juvenile work status, juvenile gender, single parent households and community characteristics in this model. Also I need time-fixed effects to control for temporal changes over 22 years. I ther efore include year, month a nd day fixed effects in the model.

PAGE 20

9 The regression model now takes the form: Juvenile School Day Crimemyd = + 1( p *)(Teacher Strike)myd +… +… 2(Median Income)myd + 3(Welfare)myd + 4(Parent Education)myd +… 5(Student Employment)myd + 6(Juvenile Gender)myd +… +… 7(Urban)myd + 8(Single Parent)myd 1(Year) + 2(Month) + 3(Day) Once I am satisfied that the data exhibit a ll of the normal properties one would expect from specifying a model of juvenile crime, I can then shift the analysis to include zip code fixed effects take into account differences which I cannot control for. When I move to this specification, the final regr ession model looks like the following: Juvenile School Day Crimemydz = + 1( p *)(Teacher Strike) mydz +… +… 1(Year) + 2(Month) + 3(Day) + 4(Zip) 1.3 Data My juvenile-arrest data comes from “Was hington Juvenile Court Case Records” made available by the National Juvenile Court Da ta Archive. It prov ides juvenile arrest data over 22 years, spanning from 1980 to 2001. This data set is both lengthy as well as highly detailed at the individual level. It provides the home address zip codes of the offenders, the nature of the crimes committed, the dates of the offenses (as well as the arrests), as well as many other important ch aracteristics surrounding th e reported arrests. In total when I include only Washington juve niles who were arrested for crimes that occurred during the ordinary school days preselected for my study, I have 401,864 arrest cases starting in 1980. In addition to the juvenile arrest data for Washington State, I use data provided by the “Census 2000 Summary File 3” and “1990 Su mmary Tape File 3” to derive annual zip-code characteristics. These Census files give zip-code level information on population and school enrollment numbers, as well adult educationa l attainment, the number of households who are welfare reci pients, type of comm unity, median household

PAGE 21

10 income and student employment. Because th ese characteristics are only available for these two periods (1990 and 2000) I trend them over the 22 years using an exponential function rather than a li near function to describe the path of the trend.16 Finally, teacher strike information came from the report “Public Employee Strikes in Washington” published in 2003 by the Publ ic Employees Relations Commission in Washington State. It covers all Wa shington public employee strikes since 196717; however since this publication lacks some im portant details about these strikes, it has been supplemented with specific informa tion from news articles provided by the Associated Press and the Seattle Times to streng then the accuracy of the treatment effect. These articles are useful in pinpointing specific details surro unding each strike such as whether the school remained opened with emer gency substitutes, and what information was distributed to parents in di stricts where strikes occurred. 1.3.1 Methodology Choosing a proper methodology for this analys is requires some extra care. Because the analysis is done at the daily level, a huge percentage of observations for any given day will be zero. The distribution of the de pendent variable is skewed towards zero, making an OLS regression analysis inappropr iate. This could be solved by using a poisson regression analysis, however the fit fo r a poisson regression for this dataset is poor.18 Also the poisson analysis does not accoun t for strangely behave d standard errors 16 I also trended these variables linearly, however when I do this I end up with some numbers that are not feasible do to the imprecise nature of this process. When I impose minimum and maximum constraints on these variables, I find that these results are insens itive to whether these variables are defined linearly or exponentially. 17 This report began with the enactment of Public Employees Collective Bargaining Act in 1967. 18 A preliminary test of the data shows that the dependent variable generally has a standard deviation roughly 3 times larger than the mean. In addition, I observe large values of the Chi squared terms in the

PAGE 22

11 and overdispersion. In this case, overdisper sion may take the form of a single juvenile committing several crimes on a single strike day, or a gang of juveniles committing what could be considered one crime. Ultimately the chosen method of regression analysis is a negative binomial regression analysis. This deals with problems of overdispersion that a poisson does not correct for.19 The dependent variable is aggregate crime per day (crime count) in a zip code, and the exposure is total student population in said zip code. 1.3.2 Defining the Parameters As described earlier, I begin by controlli ng for differences among zip codes directly rather than using zip-code fixed effects. E ach variable is derived from data taken from the 1990 and 2000 Census. Since I have only two observations for each variable, I interpolate annual values usi ng an exponential path function. I started by solving for the growth rate (Gi, zip) of each variable izip: Gi, zip = ((Censusi, zip 2000/ Censusi, zip 1990)(1/10)) 1 Then I interpolated each year’s obse rvation according to the equation: Variable izip in year t = Censusi, zip 1990*(1 + Gi, zip)(n), where n = t – 1990 Overall this analysis uses 7 variables to cap ture differences in communities. Income is measured using Median Income, which comes directly from the Census data. The Welfare variable is a percentage of households in a zip code that currently receive public assistance income. Of course I expect that local poverty is a serious factor in juvenile crime, so both of these variables seem to be relevant (Mocan and Rees 1999). Juveniles from poor neighborhoods are more likely to be less educated, and ther efore more likely to poisson “goodness-of-fit” tests. Both of these char acteristics suggest that the poisson model in not the correct model for these regressions. 19 The overdispersion parameter alpha is shown to be significantly different from zero. Thus I can conclude that the negative binomial regression significantly different from the poisson regression.

PAGE 23

12 be criminally active overall (Freeman 1992). Since juveniles from poor households have a smaller opportunity cost of committing crime, I expect Median Income to be negatively correlated with juvenile crime, and naturally the converse should be true for Welfare. The Poor Parental Education variable is de fined as the percent of adults (25 years+) who have not completed the equivalent of a high school education. My prediction here is that as adult dropout rates incr ease, juvenile crime increases Because poor adult/parental education equates to less adu lt human capital, parents are likely to have lower demand for child quality (Becker and Tomes 1976, 1986) Therefore juveniles in these neighborhoods should also have a lower opportunity cost of crime. In addition, one can argue that poor adult/parental education may also imply that parents have fewer resources to provide monitoring for juveniles, less overa ll time to provide supervision of their own children (parents may be working long hour s or more than one job), or possibly uninformed/undesirable parenting techniques. Our Student Employment variable describe s the percent of 16-19 year olds who are both employed and also enrolled in school. The prediction of this variable is not clear. On one hand, students who are employed in af ter-school jobs may develop a sense of responsibility from working a j ob, or that they may fear losi ng future wages should they be caught committing a crime. This suggests that Student Employment should negatively influence juvenile crime. On the other ha nd, juveniles who work may have greater access to reliable transportation, or less restrictive parents, both of which could contribute to a greater propensity to commit crime. In th is case, Student Employ ment would positively influence juvenile crime.

PAGE 24

13 The Juvenile Male variable measures the pe rcent of the juvenile population in a zip code who are male. Since juvenile males trad itionally make up the majority of juvenile crime, juvenile crime for a zip code shoul d increase as a juvenile population there becomes proportionately more male.20 The Single parent variable is the percent of single parent house holds in that zip code. One expects that this variable should be positively co rrelated with juvenile crime so that more single parent households lead to more potentially unsupervised juveniles, and potentially greater juvenile crime. The Urban variable is the percent of th e total zip code popula tion who live in an urban community. This should pick up ba sic differences among community types, however I also interact this variable with the strike measure to see if strikes have differential impacts in different types of co mmunities. For several reasons, I expect crime in Urban areas to differ from that of Rural or Suburban areas. Urban areas feature many characteristics, such as increased crim inal opportunity, highe r pecuniary benefit, and increased criminal anonymity, which make s criminal activity in these areas more desirable (Glaeser and Sacerdote 1999). As su ch this Urban variable should be positively correlated with teacher strikes. Lastly, the Strike variable is a dummy variable, with a value of 1 describing a teacher stri ke event, and zero otherwise. 1.4 Regression Analysis The initial results of the negative binomi al regression analysis seem to confirm previous findings about the nature of sc hool incapacitation. Column 1 in Table 1-1 20 From 1993 to 2002, males made up about 71-75% of total juvenile arrests. For more information about juvenile crime by gender, please reference the Crime in the U.S. annual report published by the Federal Bureau of Investigation.

PAGE 25

14 shows that the presence of teacher strikes seems to have a positive and significant effect on juvenile criminal behavior. But before I begin to quantify the effect of school incapacitation on juvenile crime, I want to make sure that the data are behaving properly. The signs of the covariates in Column 1 of Table 1-1 seem to support the predictions made on the effects of income, we lfare, parent’s educat ion and urban status with respect to juvenile crime. First, we see that median income and welfare are negatively and positively significant, respectively. This implies that children in lower income households and welfare receiving house holds are more likely to engage in juvenile crime. These results support th e predictions associated with poorer living conditions. In addition, lower pa rental education leads to high er crime as well. Again, for several reasons this result makes sens e. The Urban variable is positive and significant, showing that urba n communities experience more overall juvenile crime. Our Student Employment variable is positive and statistically significant. This may imply after-school jobs provide students with greater resources which they may use to commit crime. Further, it may reflect the fact that as more student s get after-school jobs, they are spending less time increasing their human capit al (through study, after-school activities, etc.) One problem is that the effect of the Juvenile Male variab le is negative and significant, which contradicts the predicted effect. It says that across juvenile populations, those that are comprised of re latively more females experience more juvenile crime. One possible explanation for this result is that it may be reflecting problems in variable measurement for sma ll communities. For rural communities there are two basic problems. The first is that with a small populati on of juveniles, the

PAGE 26

15 variation of the Juvenile Male variable increases significantly.21 The second is that this unusual variation makes interpolating this variab le equally unreliable. Since a significant portion of zip codes have very small populati ons of juveniles, it is possible that the negative result I observe is being driven by ru ral zip codes that ha ve atypical juvenile gender characteristics.22 To check the sensitivity of the initial resu lts I begin by eliminating months in which strikes did not occur from the sample: January, March, and May (Specification II).23 I then went a step further and eliminated ev ery month other than September, October and April from the sample (Specification III). September and October are preserved because they are the most common months for strike s, and April was included because in 1991 there was a “state-wide” teacher strike that involved at l east 41 school districts. The results of Specifications II & III are repor ted in Columns 2 and 3 of Table 1-1 respectively. These results are not sign ificantly different from Specification I. Recalling the zip code matching process descri bed earlier in the paper, I am able to accurately match back to 1984 zip codes. Howe ver the crime data dates as far back as 1980. To deal with potential zip code mismat ches before 1984, I drop the first four years of the sample (Specification IV).24 This fourth specificati on shows the strike variable 21 For zip codes with juvenile populations less than 500, the standard deviation of the Juvenile Male variable is approximately 3 times larger than the standard deviation of those zip codes with 500 or more juveniles. 22 In fact when I look at only urban zip codes, I see th at this variable seems to correct itself, which gives credence to this argument. Further, Table A-2 shows that when I drop the smallest zip codes (population less than 500 juveniles) I see that the negative significance of the coefficient is destroyed. Thus it seems plausible that poor interpolation of this variable for sparsely populated areas has tainted the coefficients. 23 Over the 22 year period, there wa s one school district that orchestrated a strike in January, but it only lasted for one day. 24 In doing this I eliminate 9 strike events, 3 of which lasted 9 or more days.

PAGE 27

16 remains significant as the sample is modifi ed (Column 4 of Table 1-1). These results seem robust to many different repr esentations of the data set. It is also important to verify that thes e results are not being driven by zip code characteristics being specified inaccurately. The most immediate check of this is to use zip code fixed effects in lieu of explicit vari ables to capture the difference. Column 5 shows the addition of zip code fixed effects does not significantly alter the results. The coefficient of the strike variable Colu mn 5 of Table 1-1 shows that strikes have a positive effect on juvenile crime that is sta tistically significant. To quantify the impact of this increase, I take the partial derivative of the dependant variable with respect to the strike variable ( Totalcrime/ Strike). When I divide this marginal effect by the average of the dependent variable, I am left with a ch ange that reflects a pr oportion of the mean of the dependent variable. In the full data set, the marginal effect of the strike variable suggests that total juvenile crime increases by 56.71% on days when strikes occur. This is more than a modest change in total crime. 1.4.1 Community Differences Since the full data set includes all co mmunities, some of which may look and behave very differently from one another, I want to test whether juvenile crime in different communities is differentially affected on strike days. To capture community differences, I run separate regressions for each community type, where I restrict the sample to only those communities in which 51% or more of the population in a zip code live in one of three types of co mmunities: urban, suburban or rural.25 The results of these 25 The reader should be aware that this specification implies that zip codes which are considered one-third urban, one-third suburban, and one-third rural, will th erefore be excluded from this analysis. However because of the way in which the Census defines thes e communities, these kinds of zip codes make up a very small portion of the total sample.

PAGE 28

17 regressions can be viewed in the Columns of Tables 1-2, 1-3 and 1-4. These results show that unexpected school absences significantly influence juvenile crime only in urban communities. This result is not at all unreasonable, given there are many differences among these types of communities that no doubt influence cr iminal behavior. It seems understandable that rural communities experience no change in juvenile crime when a strike occurs. Rural communities may provide less opportunity for crime, as well as a small town atmosphere that makes anonymity difficult. The lack of a significant effect of strikes in suburban communities is less believable sin ce they arguably provide greater opportunity and anonymity, however there may yet be other, (possibly unobservabl e), characteristics about suburban characteristics th at naturally deter juvenile crime on these strike days. For instance, suburban communities may be pop ulated by more involved parents, may have more effective emergency resources, etc. The strike effects in urban communities however are positive and significant. In ur ban communities juvenile crime increases 19.68% on strike days.26 Again I run the same specification tests to test the robustness of these results. The signs of the covariates in ev ery urban subsample are the same as in Table 1-1, except for Juvenile Male. The Juvenile Male covariate does take on a positive and significant coefficient in the urban subsample, however in the suburban and rural subsamples it remains negative and significant. 26 The huge decrease in the magnitude of the strike coefficient from the full sample to urban subsample indicates that small populations in rural communities are exaggerating the effects of the strikes in the full sample.

PAGE 29

18 1.4.2 Differences in Offense Types Since school incapacitation may be differentially affecting the types of crimes that juveniles are committing, I partition the dependent variable measure of total crime into 5 specific crime types. These types of crim es include: drug and alcohol related crimes, mischievous crimes, property crimes, vi olent crimes, and finally weapons and endangerment crimes. I still consider all th ree community classifica tions when doing this analysis because I may yet find significant re sults in types of crimes for rural and/or suburban communities.27 Table 1-5 shows the results of this extended analysis. Rural and suburban communities continue to experience no change in juvenile crime resulting from strike days regardless of the type of crime that is being committed. Urban communities though do experience an a rray of changes in juvenile criminal behavior. Both mischievous crimes and propert y crimes seem to increase in the presence of a teacher strike. I estimate that misc hievous crime increases by 48% on average for days when strikes occur.28 Property crime increases by an average of 28.81%. This change in property crime is nearly double prev ious estimates in the literature. Violent crime decreases by approximately 31.53%.29 Drug and alcohol crimes, and weapons and endangerment crimes are unaffected by school incapacitation. These changes in criminal activity seem to make sense. It seems logical that mischievous crime would be most affected by school incapacitation. After all, these are 27 This could arise if increases in certain crimes were offsetting decreases in other types of crimes. 28 It is difficult to compare my measure of mischiev ous crime to Jacob and Lefgren’s measure of minor offenses. They define minor offenses as “NIBRS Group B offenses”, however my measure of mischievous crime is a select subset of these crimes. 29 Jacob and Lefgren estimate that school attendance decreases property crime by approximately 14%. They also estimate that violent crim e increases on school days by about 28%. Therefore, my estimate of 31.53% for violent crime is approximatel y 10% larger than previous estimates.

PAGE 30

19 the types of crimes that often result from boredom rather than calculated criminal thought. On the other hand, the decrease in violen t crime that occurs is harder to explain. I use, as Jacob and Lefgren do, social interacti on theory to explain th is result. Jacob and Lefgren argue that juvenile violence against other juveniles arises, at least in part, from disputes formed in and around the classroom. When these juveniles are not forced to be at school, there is a decreas e in the amount of overall vi olent crime that reflects a decrease in juvenile violence against other j uveniles. This explan ation seems credible.30 Given the limits of the data set, I cannot test this theory further.31 1.4.3 Types of Offenders In addition to the topics already spoken to, the detailed level of the data also allows me to also explore the nature of the crim inal, beyond superficial characteristics. By observing criminal record, I attempt to answer whether these additional crimes are caused by normal delinquents facing greater criminal opportunity, or by juveniles who do not normally engage in criminal activity. I agai n partition the dependent variable into two measures: 1) total crime by repeat offenders and 2) total crime by one-time offenders. I define repeat offenders not just as those juveniles who have committed previous crimes, but also those juveniles who will commit future crimes. Thus a juvenile who commits multiple crimes is considered to be a repeat offender juvenile even on their first crime. The results of these regressions can be seen in Table 1-6. Additional overall crime is induced for both offender types, however the average increase in crime for one-time offenders as a percent of the mean crime level for that 30 In 1998 and 1997, 62% of victims of juvenile violence were 17 or younger. 31 I have no information concerning the victims of the crimes in my data set.

PAGE 31

20 group is much higher than for the repeat o ffenders. On strike days there is a 33.45% increase in average total crime for one-time offenders, and a 13.89% increase in average total crime for repeat offenders. Essentiall y, one-time offenders as a group seem to be much more affected by the lack of school in capacitation then the repeat offenders. This evidence suggests that the incr eased crime on strike days is less about criminal motive, and is more a result of boredom. If these two groups are motivated differen tly to commit crimes, these effects may not be expressing differences ac ross groups in the same types of crimes. As a society we may be more concerned if the lack of school incapacitation were i nducing more property crime by one-time offenders rather than repe at offenders (where the crime may have eventually happened anyway). Therefore I am interested in examining whether these two groups differ in offense types on these strike days. To do this I create a dependent variable that equals total crime conditional on offender type and offense type. The results are expressed in Table 1-6. The decrease in violent crime comes mainly from the repeat offenders. Violent crimes committed by repe at offenders drop by 36.72% on strike days. In addition, both groups are cont ributing to the overall increas es in mischievous crimes and property crimes. Property crimes by one-time offenders and repeat offenders increase by 56.75% and 25.40% re spectively. Likewise mischi evous crimes increase by 83.11% for one-time offenders and by 30.82% for repeat offenders. 1.4.4 Strikes Days as Gateway Crimes Since the increase in crimes committed on strike days is substantially larger for first-time offenders, it may also be true that these juveniles may carry this behavior forward after their first arrest. To see if this is the case, I examine whether or not a

PAGE 32

21 current day repeat offender was significantl y more likely to haven gotten his/her start (commit his/her very first crime) on a strike day versus a normal school day. To do this, I generate a daily crime total that expresses the total of “first time” crimes by subsequent repeat offenders. The results of this regression can be also vi ewed in Table 1-6. A future repeat offender is found to be significantly more likely have committed his first crime on a strike day rather than an ordinary school day. This sugge sts that not only is school incapacitation preventing the creation of crimes, but it may also be preventing the creation of criminals. 1.4.5 Displacement of Crime Thus far it seems the evidence suggests juve nile crime increases when students are not incapacitated in school. However what is not clear is whether th ese changes describe an overall increase in the amount of juvenile crime, or a di splacement of crime from one day to another. Jacob and Lefgren try to sp eak to this question of temporal displacement in their paper, however their use of in-s ervice days again limits the power of their analysis. Jacob and Lefgren find that there is no temporal displacement of crime to inservice days, however because in-service da ys are completely known, this result may simply be showing that juvenile criminals will plan to commit more crime when facing more criminal opportunity. What we are really interested in is wh ether a juvenile will temporally displace a “planned” crime given an unexpected opportunity to do so, or whether an unexpected opportunity to commit cr ime results in more “unplanned” crime. Again teacher strikes provide a more ideal measure of unexpected criminal opportunity, so I revisit this question of temporal displacement.

PAGE 33

22 To test whether crime is being displace d, I begin by aggregating Total Crime to a weekly measurement rather than daily32. If juvenile crime is simply being displaced from weekend crimes to weekday crimes, then I s hould not observe an effect for the strike variable. To capture the week ly aggregated effect of stri kes, I proportion the fraction of the week where strikes occurred as a new meas ure of the strike treatment. Despite this aggregation, I find that strike s are still a significant factor in contributing to additional juvenile crime (Table 1-7). Table 1-7 shows a positive and significant st rike coefficient on Total Crime in the full data set. This is a strong indication th at overall juvenile crime is increasing, rather than being displaced across time. However, if displacement of crim e varies across crime type, or criminal type, we cannot conclude th at no crimes are being displaced. To see if the type of crime or the characteristics of the perpetrator influences whether displacement of crime occurs, I partition weekly total crim e by crime type and offender as before. The results of these regressions ar e also reported in Table 1-7. These strike coefficients on Table 1-7 show that every specification of cr ime type and criminal type results in an increase of overall juvenile crime, and not temporal displacement of crime. I confirm the lack of displacement in total crime by aggr egating further to a monthly crime total. Weekly and Monthly aggregations of juvenile crime provide a good test of displacement of crime. I perform another test by lagging the strike effe ct, in order to pick up acute changes in crime shortly after the strike ends. If fewer crimes are committed on days following the strike, then the coefficient of the lagge d strike variable should be 32 This weekly measurement includes weekends and ho lidays, since strike day crime could displaced from usual weekend crime. In addition, week fixed effects are used in place of day fixed effects.

PAGE 34

23 negative and significant.33 I run seven separate regre ssions, shown in Table 1-8, labeled “1 day” through “7 days.” For a strike that ends on day t the first treatment of the lagged strike variable under “1 day” begins on day ( t + 1), under “2 days” begins on day ( t + 2), and so on. Table 1-8 shows that there is no significant effect from the lagged strike treatment. These results provide further evid ence that juvenile crime occurring on strike days is additional crime, rather than displaced crime. 1.4.6 Duration Effects on Criminal Activity One last advantage I gain from using teacher strikes as an instrument, in conjunction with the strength of the daily arre st data, is that I can track how juvenile crime changes over time with prolonged absen ce. If it were true that an increase in juvenile crime is at least partially cause d by boredom and inactivity on the part of juveniles, then I should expect that the longer juveniles rema in “inactive”, the more crime is likely to result. Arguably there is just as much uncertainty over when teacher strikes will end as when they will begin. As such, tr ends in juvenile crimes up until the start of the strike and over the period of the strike should not be biased by student or parent foresight. In addition, the longe r a strike continues, the less able a parent or guardian is to provide adequate supervision to students. For example, a parent may be able to take one or two days away from work to supe rvise their child, but the cost of taking 10 consecutive days off work is substantially larger. To examine if juvenile crime is changi ng with the length of a teacher strike I partition the strike variable in to 4 groups of new independent variables. Three of these variables are clustered groups of 3 consecutive strike days, and the la st variable pertains 33 This sample period also includes w eekends. I also revert back to day fixed effects rather than week FE.

PAGE 35

24 to a strike in its 10th day or higher.34 These results can be view ed in Table 1-9. Table 1-9 shows that for the first 3 days, there is no si gnificant change in ove rall juvenile crime.35 After the 5th day or so, the strike coefficient becomes positive and significant, reflecting an increase in overall crime. Further, after the 5th or so strike day there is a gradual increase in the level of juvenile crime as strike length continues. In unreported regressions I find that as time passes, property and misc hievous crime trends upward, while the decrease in violent crime remains re latively constant across days. The positive coefficient shows that these increases in property and mischievous crime overpower the decreases in violent crime. Column 2 of Table 1-9 repeats the regression in Column 1, only with the strike variable clustered at the twoday level rather than at the three-day level. Clearly the evidence sugge sts that the longer juveniles remain absent from school, the more overall daily juvenile crime will be observed.36 1.5 Robustness Checks The results of the analysis up to this point seem to point to a very clear story about the relationship between juveni le crime and school incapaci tation. In this section I perform some robustness checks to ensure that these results are not spurious. I begin by 34 I do this because if I were to partition the strike variable for each separate day (1st day, 2nd day, etc.), then the number of treatments would be very small for the later strike days and the coefficients would be unreliable. 35 When I break up the dependent variable (total crim e) by crime type as before, violent, property and mischievous crime types are individually significant. However in the first 3 days, the opposite signs and magnitudes of the coefficients wash out aggregate effects in total crime. These results are consistent with Jacob and Lefgren’s findings (2003). 36 It is worth keeping in mind that these strike day counts are also limited to school days. That is to say that when a strike has reached its 10th day, a student has missed exactly 2 full weeks of school. It is also very likely that these criminal behaviors will change after a strike has gone on for an extensive amount of time. At that point, juvenile crime may level-off or even decrease substantially. Of course given the sparse and limited nature of extremely lengthy teach er strikes, an analysis of that na ture is inherently difficult not to mention highly unreliable.

PAGE 36

25 testing whether the strike variable is corre lated with some unobservable characteristic, and that perhaps the results are picking up some alternate relationship. For instance, if a strike event is in fact very foreseeable, th en one might speculate that the increase in juvenile crime predicted by the strike variab le is really picking up an increased police presence on days that strike occur. That is to say that the results are not reflecting a true increase in juvenile crime, but that there ar e simply more police on the streets capturing a larger portion of the same total amount of juvenile crime. If this were true, then one would expect th at the number of days between an offense and an arrest would be shorter on strike days. The graphs on Figure 1-1 show that this is not the case. When I consider only zip c odes where strikes did occur, I see that the percent of quick arrests37 is significantly higher on regular school days than on strike days. I explain this by sugges ting that on regular school days police have a better idea who the likely criminals are (repeat offenders), as opposed to strike days where offenders are more likely to be (previously unidentifie d) first time, or one-time offenders. 1.5.1 Importance of Singular Strikes Given the uniqueness of strikes, I must confirm that these results are not being solely driven by a small subsets of strikes. To do this I systematically eliminate each individual year from my sample starting with 1984. Table 1-10 shows that strikes are a significant determinant of juvenile crime in every year with the possible exception of 1985 (z = 1.70). When I drop 1985 from the sample I find a much weaker significance than when I eliminate any other year. In September of 1985, the Seattle school district engaged in a teacher strike that lasted 20 school days. Given that the evidence has 37 I define these as arrests where the associated offe nse occurred less than 48 hours before the arrest. 25

PAGE 37

26 suggested that juvenile crime in urban districts is most affect ed by teacher strikes, it is possible that these results are being solely driven by the circumstances surrounding this one district. To test this condition I restri ct the sample twice, once to look at the sample without the Seattle school district, and then again to look at only the Seattle school district. If Seattle is driving all of the results, then the strike variable should be insignificant when I exclude Seattle, and extremely significant when I exclude all other districts except for Seattle. Table 1-10 shows that strikes still have a positive and signi ficant effect when I drop Seattle from the analysis. In addition the strike variable in the Seattle subsample is marginally significant (z = 1.82). Therefore I can be confident that while this one event obviously contributes to the re sult, (as I would expect the second longest strike in the most urban school district in Washington State to do), it is not the sole force driving the results.38 1.5.2 Reproducibility of Results Another robustness check I perform is to make sure that the results cannot be easily reproduced with random generations of the strike variable. To check if this is the case, I use a uniformly generated random variable on the interval from 0 to 1 to re-specify a random dummy strike variable. Since teach ers are on strike 2,351 days out of the 1,703,910 day sample, strike days occur 0.13797 % of the time. I assign a strike value of 1 to the random variable, when the value of the random observation generated is less than or equal to 0.0013797. If the randomly gene rated observation is greater than 0.0013797, then it is reassigned a value of zero. The ne w random strike variable should approximate 38 The Seattle strike of 1985 is actually tied with the Mukilteo strike of 1990 (at 20 days) for the second longest strike in Washington State history from 1980-2001.

PAGE 38

27 the number of strikes in the original sample. The results in Table 1-11 show the mean of the random strike coefficient over 25 trial re gressions, as well as the average number of random strike days generated. Of the 25 trials only 2 random strike variable coefficients were significant. One of the statistically si gnificant coefficients was positive, while the other was negative. 1.5.3 Further Tests One final test of the data is to make sure that my finding of new criminal participation on strike days (when the samp le is broken up into one-time and repeat offenders) is not a byproduct of mislabeling offenders. Since the crime data ends in 2001, it could be that in the years leading up to 2001 many future career criminals are being mislabeled as one-time offenders due to the fact that future arrests past 2001 are unknown. To test this possibilit y, I drop the last 4 years of th e data sample and repeat the regressions using offender type as the dependent variable. Table 1-12 shows that this mislabeling does not affect the results. The effects of a strike on repeat offenders and one-time offenders are nearly iden tical for either specification. 1.6 Conclusion The results presented in this chapter provide a clearer pi cture concerning the relationship between schooling and urban juve nile crime. My main findings are the following: 1) The effects of a lack of school incapacitation are larger than those found in Jacob and Lefgren’s study. The evidence sugge sts that property crime rises by as much as 29%, almost twice as much as what is pr edicted in the existing literature. Likewise, violent juvenile crime decrea ses by as little as 31.53%, a nd as much as 36.72%. These estimates of increased violent crime represent a 10%-25% increase in previous estimates.

PAGE 39

28 2) I confirm that these change s in daily crime reflect change s in total crime, and not a displacement of crime from one day to the next, regardless of crime type and criminal type. 3) Different types of juveniles are affected differently by a lack school incapacitation. A significant propo rtion of the decrease in the level of violent crime can be attributed to repeat juvenile offenders, while one-time offenders contribute more to the increases in property crime and crimes of mischi ef. 4). A failure to incapacitate juveniles will result in significantly more crime by those juveniles who might not have otherwise engaged in criminal acts or who rarely enga ge in such acts. In fact, juveniles who become repeat offenders are more likely to have gotten their criminal start when incapacitation was expected but not implement ed, as opposed to or dinary school days. Incapacitation seems to have the greatest implications for new and seemingly preventable urban crime. I find that on strike days, new criminals are engaging in new criminal acts. Incapacitating these juveniles seems to be e ffective at preventing at-risk urban youths from engaging in new and ris ky behavior. I do not find that school incapacitation has any significant effect in suburban and rural communities. It may be true that the characteristics that differe ntiate these community types from urban communities, as well as each other, inherent ly reduce juvenile criminal behaviors, however such characteristics may be largely unobservable to us. If the evidence presented in this paper is ac curate, then what we are learning is that how we manage children’s time outside of sc hool is very important. If parents are strained to provide adequate supervision for their children, juveniles who we might consider to be at-risk may be the ones who are affected most, though at-risk juveniles are not the only ones affected. Families and bus inesses also bear the burden of increased 28 28

PAGE 40

29 juvenile delinquency. These results have profound implications for many urban school policies and programs. How school districts bud get their students’ days off is no longer a trivial matter. For example, school districts with long breaks (like in a traditional school calendar) may have very different juvenile crim inal behaviors than a school district with frequent but shorter breaks (like in a year -round school calendar). Differences in how school calendars overlap within and among dist ricts should have an effect on the nature of juvenile crime. The length of the school da y or school year in a di strict, school district policies regarding student atte ndance requirements, and even the nature of after school programs within a district are all policy issu es that determine more than just a child’s educational outcome. It is my goal to conti nue to explore this da ta set and scrutinize these kinds of school policies, so that I can better understand the impact that these kinds of policies may have.

PAGE 41

30 Table 1-1: Negative Bino mial Regression onto Total Crime with Full Data Set Variables ( ) ( ) ( ) ( V) (V) Strike Median Income Welfare Urban Poor Parental Educ. Juvenile Maleness Student Employment Single Parent House Alpha 0.195** 0.187** 0.210** 0.260** 0.220** (4.06)** (3.86)** (4.16)** (5.06)** (4.38)** -0.000058** -0.000059** -0.000059** -0.000055** — (102.47)** (83.82)** (60.76)** (52.80)** — 0.475** 0.452** 0.520** 0.821** — (7.71)** (5.96)** (4.97)** (6.28)** — 0.132** 0.112** 0.097** 0.055** — (26.96)** (18.45)** (11.54)** (6.14)** — 0.329** 0.346** 0.398** 0.482** — (10.40)** (8.88)** (7.45)** (6.67)** — -0.330** -0.341** -0.355** -0.396** — (28.06)** (26.24)** (22.32)** (22.72)** — 0.291** 0.297** 0.290** 0.296** — (17.49)** (14.43)** (10.23)** (9.16)** — 0.500** 0.495** 0.466** 0.985** — (9.52)** (7.64)** (5.20)** (9.67)** — 0.989** 1.009** 0.997** 0.956** 0.759** (156.56)** (126.04)** (91.12)** (87.74)** (79.06)** Number of obs. Time Fixed Effects Zip Fixed Effects R-Squared 1,703,910 1,128,630 583,440 479,400 479,400 Y Y Y Y Y N N N N Y 0.0367 0.0364 0.0367 0.0279 0.0559 In Columns – V, the sample is re-specified as follows (Type through Type V): (I) – Sample includes all ordinary school days from 1980-2001 (II) – January, March and May are excluded from the sample 1980-2001 (III) – Includes only April, September and October from 1980 2001 (IV) – Only April, September and October from 1984 – 2001 (V) – Only April, September and October from 1984 – 2001, with zip code fixed effects Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller. To interpret these regressions, I have to calculate the marginal effect of these variables. The marginal effect of th e strike variable is 0.149, which represents a 56.71% increase in average total juvenile crime on strike days.

PAGE 42

31 Table 1-2: Urban Subsample Negative Bi nomial Regression onto Total Crime Variables ( ) ( ) ( ) ( V) (V) Strike Median Income Welfare Poor Parental Educ. Juvenile Maleness Student Employment Single Parent House Alpha 0.151** 0.129* 0.178** 0.211** 0.214** (3.07)** (2.60)* (3.50)** (4.11)** (4.26)** -0.000061** -0.000062** -0.000062** -0.000057** — (83.07)** (67.39)** (48.42)** (42.17)** — 1.394** 1.369** 1.418** 1.703** — (16.40)** (13.01)** (9.79)** (9.46)** — 0.398** 0.403** 0.449** 0.462** — (9.49)** (7.79)** (6.31)** (4.65)** — 0.602** 0.577** 0.482** 0.397** — (11.13)** (8.60)** (5.16)** (3.62)** — 0.214** 0.224** 0.214** 0.157** — (9.10)** (7.67)** (5.34)** (3.47)** — 0.739** 0.790** 0.945** 1.276** — (8.28)** (7.12)** (6.24)** (7.55)** — 0.676** 0.685** 0.666** 0.638** 0.494** (114.06)** (91.00)** (65.18)** (62.65)** (54.87)** Number of obs. Time Fixed Effects Zip Fixed Effects R-Squared 454,953 301,312 155,601 129,593 129,593 Y Y Y Y Y N N N N Y 0.0484 0.0484 0.0487 0.0400 0.0635 Table 1-2 limits the sample to include only zip codes where 51% or more of the population lives in an area considered by the Census to be Urban In Columns – V, the sample is re-specified as follows (Type through Type V): (I) – Sample includes all ordinary school days from 1980-2001 (II) – January, March and May are excluded from the sample 1980-2001 (III) – Includes only April, September and October from 1980 2001 (IV) – Only April, September and October from 1984 – 2001 (V) – Only April, September and October from 1984 – 2001, with zip code fixed effects Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller. To inte rpret these regressions, I have to calculate the marginal effect of these variables. The marginal effect of th e strike variable is 0.121, which represents a 19.68% increase in average total juvenile crime on strike days.

PAGE 43

32 Table 1-3: Suburban Subsample Negative Binomial Regression onto Total Crime Variables ( ) ( ) ( ) ( V) (V) Strike Median Income Welfare Poor Parental Educ. Juvenile Maleness Student Employment Single Parent House Alpha -0.027 -0.064 -0.159 -0.162 -0.449 (0.12) (0.29) (0.61) (0.62) (1.74) -0.000015** -0.000016** -0.000013** -0.000012** — (8.01)** (6.54)** (4.74)** (5.66)** — -1.929** -2.146** -1.483** -2.005** — (9.21)** (8.31)** (4.20)** (4.43)** — 1.639** 1.856** 1.683** 2.393** — (13.33)** (12.29)** (8.09)** (8.55)** — -1.432** -1.973** -2.151** -2.899** — (5.33)** (5.94)** (4.74)** (5.66)** — -0.121* -0.198** -0.172 -0.064 — (2.23)* (2.96)** (1.87) (0.59) — 2.285** 2.369** 0.217** 3.062** — (14.13)** (11.97)** (7.83)** (8.82)** — 1.326** 1.303** 1.240** 1.203** 0.969** (59.08)** (54.84)** (39.01)** (37.78)** (34.42)** Number of obs. Time Fixed Effects Zip Fixed Effects R-Squared 134,100 88,841 45,794 37,992 37,992 Y Y Y Y Y N N N N Y 0.0294 0.0300 0.0302 0.0227 0.051 Table 1-3 limits the sample to include only zip codes where 51% or more of the population lives in an area considered by the Census to be Suburban In Columns – V, the sample is re-specified as follows (Type through Type V): (I) – Sample includes all ordinary school days from 1980-2001 (II) – January, March and May are excluded from the sample 1980-2001 (III) – Includes only April, September and October from 1980 2001 (IV) – Only April, September and October from 1984 – 2001 (V) – Only April, September and October from 1984 – 2001, with zip code fixed effects Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller

PAGE 44

33 Table 1-4: Rural Subsample Negative Binomial Regression onto Total Crime Table 1-4 limits the sample to include only zip codes where 51% or more of the population lives in an area considered by the Census to be Rural In Columns – V, the sample is re-specified as follows (Type through Type V): (I) – Sample includes all ordinary school days from 1980-2001 (II) – January, March and May are excluded from the sample 1980-2001 (III) – Includes only April, September and October from 1980 2001 (IV) – Only April, September and October from 1984 – 2001 (V) – Only April, September and October from 1984 – 2001, with zip code fixed effects Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller Variables ( ) ( ) ( ) ( V) (V) Strike Median Income Welfare Poor Parental Educ. Juvenile Maleness Student Employment Single Parent House Alpha 0.017 0.027 -0.343 -0.285 -0.392 (0.09) (0.14) (1.45) (1.20) (1.65) -0.000034** -0.000036** -0.000036** -0.000033** — (26.68)** (22.50)** (16.62)** (13.59)** — -0.077 -0.063 -0.108 0.117 — (0.68) (0.45) (0.56) (0.48) — -0.282** -0.332** -0.218 -0.200 — (3.81)** (3.63)** (1.77) (1.32) — -0.357** -0.362** -0.370** -0.421** — (37.69)** (32.83)** (25.81)** (25.64)** — 0.142** 0.156** 0.180** 0.267** — (4.34)** (3.86)** (3.25)** (4.15)** — 0.379** 0.401** 0.317* 0.864** — (4.41)** (3.81)** (2.15)* (5.21)** — 2.679** 2.767** 2.810** 2.712** 2.197** (80.71)** (66.36)** (49.08)** (46.97)** (44.48)** Number of obs. Time Fixed Effects Zip Fixed Effects R-Squared 1,101,359 729,469 377,244 304,491 304,491 Y Y Y Y Y N N N N Y 0.0313 0.0308 0.0316 0.0218 0.0551

PAGE 45

34 Table 1-5: Effects of Strike Days by Crime Type and by Community Type Dependant Variable Strike Coefficient for Urban SubSample Strike Coefficient for Suburban SubSample Strike Coefficient for Rural SubSample Drugs and Alcohol Mischievous Property Violent Weapon and Endangerment 0.200 -1.188 -0.113 (1.36) (1.48) (0.20) 0.536** -0.813 -0.241 (3.88)** (0.73) (0.34) 0.257** -0.128 -0.276 (3.79)** (0.34) (0.80) -0.364* -1.103 -0.266 (2.50)* (1.44) (0.52) -0.014 -53.096 — (0.05) (0.02) — Number of obs. Time Fixed Effects Zip Fixed Effects Ave. R-Squared 129,593 37,992 304,491 Y Y Y Y Y Y 0.0506 0.0423 0.0579 Each row represents a new regression with the listed crime type as the dependant variable. The reported number is the strike coefficient for the regression with the corresponding dependant variable. Each column is a different subsample group: urban, suburban and rural. All 14 regressions are defined as a Type V regression, including data from only April, September and October from 1984 – 2001, and zip code fixed effects. The Pseudo R-squared reported at the bottom of each column is the average R-squared over the 5 regressions. Z-statistics are given in the parenthese s for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller. The marginal effects of these variables show that violent crime decreases by 31.53% on strike days. Property crime and mischievous crime increase by 28.81% and 48% respectively. Again, these eff ects reflect changes in juvenile crim e as a proportion of the mean. The coefficients of the marginal effects are (–0.032), (0.072) and (0.025) for violent, property and mischievous crime respectively.

PAGE 46

35 Table 1-6: Effects of Strike Days by Crime Type and by Offender Type Dependant Variable Strike Coefficient for One-Time Offenders Strike Coefficient for Repeat Offenders Total Crime Drugs and Alcohol Mischievous Property Violent Weapon and Endangerment Gateway Crime 0.382** 0.156** (4.05)** (2.77)** 0.337 0.153 (1.18) (0.93) 1.024** 0.597** (3.66)** (3.99)** 0.534** 0.220** (4.44)** (2.81)** -0.567 -0.408** (1.48) (2.59)** -1.330 -0.009 (1.20) (0.03) — 0.286* — (2.00)* Number of obs. Time Fixed Effects Zip Fixed Effects 129,593 129,593 Y Y Y Y Table 1-7 includes only urban zip codes. All 13 regressions are defined as a Type V regression, including data from only April, September and October from 1984 – 2001, and zip code fixed effects. Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller. The ma rginal effects of these vari ables show that for repeat offenders, violent crime decreases by 36.72%, prope rty crime increases by 25.4%, and mischievous crime increases by 30.82% on strike days. The coefficients of these marginal effect s are (-0.032), (0.046) and (0.012). For one-time offenders pr operty crime increases by 56.75% an d mischievous crime increases by 83.11% on strike days. The coefficients of these ma rginal effects are (0.030) an d (0.008). Again, these effects reflect changes in juvenile cr ime as a proportion of the mean.

PAGE 47

36 Table 1-7: Temporal Displacement of Cr ime by Crime Type and by Offender Type Weekly Aggregated Dependant Variable Strike Coefficient for Full Data Set Strike Coefficient for Urban Subsample Strike Coefficient for One-Time Offenders Strike Coefficient for Repeat Offenders Total Crime Mischievous Property Violent Monthly Aggregated Dependant Variable Total Crime 0.235** 0.243** 0.418** 0.176** (4.50)** (4.67)** (4.51)** (3.11)** — 0.570** 0.820** 0.501** — (4.39)** (3.09)** (3.55)** — 0.312** 0.492** 0.225** — (4.73)** (4.27)** (3.06)** — -0.300* -0.223 -0.317* — (2.32)* (0.73) (2.27)* 0.534** 0.485** — — (3.86)** (3.73)** — — Number of obs. Time Fixed Effects Zip Fixed Effects 440,640 117,491 117,491 117,491 Y Y Y Y Y Y Y Y Each row represents a new regression with the listed crime type aggregated at the weekly level as the dependant variable. The reported number is the strike coefficient for the regressi on with the corresponding dependant variable. Each column is a different sample group. Column 1 is the full sample. The remaining three Columns are from the urban sample. The top 10 regressions are defined as a Type I regressions including zip code fixed effects. The bottom 2 regression coefficients represent the strike effect on Monthly Total Crime. Z-statistics are given in the parenthese s for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller

PAGE 48

37 Table 1-8: Temporal Displacement of Crime through Lagged Strike Variable Lagged Strike Variable Strike Coefficient For Urban SubSample 1 Day -0.033 (0.62) 2 Days -0.028 (0.54) 3 Days -0.030 (0.57) 4 Days -0.046 (0.87) 5 Days -0.041 (0.78) 6 Days -0.093 (1.75) 7 Days -0.070 (1.33) Number of obs. Time Fixed Effects Zip Fixed Effects Ave. R-Squared 293,112 Y Y 0.0597 Each row represents a new regression with strike variable lagged the by the corresponding row. Each regression is defined as a Type IV regression including zip code fixed effects of the urban subsample. The notable exception is that weekends are included to allow for the correct number of strike treatments. Therefore, days where lagged strike treatments overlap breaks, holidays and other nontypical school days are included in the sample. Z-statistics are given in th e parentheses for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller

PAGE 49

38 Table 1-9: Changes in Total Juvenile Crime, by Days Elapse d since the Start of a Strike Partitioned Strike Variable Strike Coefficient Partitioned Strike Variable Strike Coefficient Days 1 3 Days 4 6 Days 7 9 Day 10+ — — Alpha -0.005 Days 1 2 0.058 (0.06) (0.58) 0.185 Days 3 4 0.006 (1.91) (0.07) 0.261* Days 5 6 0.262* (2.54)* (2.37)* 0.313** Days 7 8 0.284** (2.93)** (2.62)** — Days 9 10 0.228 — (0.98) — Day 11+ 0.305** — (2.67)** 0.533** Alpha 0.533** (101.05)** (101.04)** Number of obs. Time Fixed Effects Zip Fixed Effects R-Squared 454,953 Number of obs. 454,953 Y Time Fixed Effects Y Y Zip Fixed Effects Y 0.0713 R-Squared 0.0713 Table 1-8 includes only urban zip codes. These two regression are defined as a Type I regressions, using data including all ordinary school days from 1980-2001, and zip code fixed effects. Both regressions use Total Crime as their dependant variable. Z-statistic s are given in the parentheses for every table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller

PAGE 50

39 0 .05 .1 .15 .2 PctArrests 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 DaysToCapture 0 .05 .1 .15 .2 PctArrests 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 DaysToCapture Figure 1-1: Average Percent of Total Arrests as Function of Days between Offense and Capture A) NonStrike Days for Zip C odes Where Strikes Occur, B) Strike Days for Zip Codes Where Strikes Occur A B

PAGE 51

40 Table 1-10: Effects of Strike Days with Specific Years Dropped from the Sample Year Dropped Strike Coefficient Year Dropped Strike Coefficient 1984 0.164** 1993 0.166** (3.39)** (3.41)** 1985 0.093 1994 0.163** (1.70) (3.28)** 1986 0.166** 1995 0.166** (3.43)** (3.40)** 1987 0.160** 1996 0.157** (3.16)** (3.23)** 1988 0.163** 1997 0.153** (3.37)** (3.16)** 1989 0.148** 1998 0.156** (3.02)** (3.23)** 1990 0.189** 1999 0.157** (3.77)** (3.26)** 1991 0.229** 2000 0.144** (3.18)** (2.99)** 1992 0.157** 2001 0.150** (3.24)** (3.11)** Without Seattle 0.133* Seattle Only 0.153 (2.04)* (1.82) Ave.Number of obs. Time Fixed Effects Zip Fixed Effects Ave. R-Squared 358274 Y Y 0.0603 358274 Y Y 0.0603 The top half of this table is a regression of Strike s on Total Crime with each year systematically dropped from the sample. Each regression includes only urban zip codes. All 18 regressions are defined as a Type I regressions, using data including all ordinary school days from 1984-2001, and zip code fixed effects. The bottom half of the table reports the effect of strike days, for the same Type I sample, with Seattle dropped from the sample, and with Seattle making up the entir e sample. The bottom of the table shows the average number of observations for the top 18 regressions, as well as the average R-squared term. Z-statistics are given in the parentheses for every table. Indicates sign ificance at the 5% level,** is significant at the 1% level or smaller.

PAGE 52

41 Table 1-11: Average Effect of Randomly Generated Strike Days on Total Crime Dependant Variable: Total Crime Random Generations of Strike Days Ave. Random Strike Coefficient Average Number of Strike Days Number of Trials 0.002 (0.05) 2,355 25 Number of obs. Time Fixed Effects Zip Fixed Effects Ave. R-Squared 1,703,910 Y N 0.0380 Table 1-12: Effects of Strike Days by Offender Type with the Last 4 Years of the Sample Dropped Dependant Variable Strike Coefficient for One-Time Offenders Strike Coefficient for Repeat Offenders Total Crime Total Crime without 1998-2001 0.382** 0.156** (4.05)** (2.77)** 0.383** 0.128* (4.08)** (2.31)* Number of obs. Time Fixed Effects Zip Fixed Effects 129,593 129,593 Y Y Y Y Table 1-11 is the average strike coefficient over 25 trials using a randomly generated strike variable. It includes the full data set. Table 1-12 is a Type V regression using just the urban subsample, and includes zip code fixed effects. Z-statistic s are given in the parentheses for ever y table. Indicates significance at the 5% level, ** is significant at the 1% level or smaller

PAGE 53

42 CHAPTER 2 THE ROLE OF TEACHER NETWORKING IN TEACHER TRANSFER DECISIONS AND TEACHER MOBILITY 2.1 Introduction In recent years, there has been growing research to support the notion that teacher movement is an important topic, and not one to be overlooked. Thus far, research on teacher sorting behavior has shown that st udent characteristics seem to influence a teacher’s transfer decision (Hanushek, Kai n, Rivkin 2001). More specifically, teachers sort away from relatively low-income, low-achieving schools. As a result, less experienced teachers are more typicall y found in poor, nonwhite, low-performing schools, particularly those in urban areas (Lankford, Loeb, Wyckoff 2002). This is most important because teacher experience seems to be directly linked to student achievement (Hanushek, Kain, Rivkin 1998, 1999)1. In addition to student char acteristics, other characteristics (such as salary differences) have been studied in an effort to better understand the factors that influence teacher movement. Research aimed at identif ying specific characteristics and conditions that promote teacher mobility have given mo re insight into the questions of when and why teachers move, however researchers stil l do not have the complete picture on how teachers move and why. 1 Hanuchek, Kain and Rivkin (1998) show that variatio ns in teacher quality accoun t for 7.5% of the total variation in student achievement. In addition, teache r quality differences tend to outweigh school quality differences considerably.

PAGE 54

43 The focus of this paper is to better unders tand the nature of teac her transfers, with special attention given to teacher intradis trict transfers from school to school. Specifically this paper will identify the us e of Master Inservice workshops aimed at professional development as a powerful means of building teacher networks. Ultimately I test the hypothesis that Inservice work shops give teachers greater opportunities to network with teachers from other schools in thei r district, and theref ore result in a higher rate of teacher transfer. The evidence will s how that a one standard deviation increase in workshop attendance contributes to a 12.5% increas e in teacher transfer rates per school. Thus teacher networks are a crucial element in influencing teacher transfer rates. Section 2.2 of this paper will provide a c ontext for the basic assumptions outlying the arguments set forth, Section 2.3 will specif y the model therein, Section 2.4 reports the preliminary results, Section 2.5 will check the validity of the results, and Section 2.6 will make the concluding remarks. 2.2 Discussion of Networking 2.2.1 How Teachers Network The idea that professionals develop networks is not a new concept in the social sciences. The term “networking” span s across many professions and takes many different forms. Like many professionals, t eachers also engage in networking for various reasons I will address later in this section. First I will address th e issue of how teachers build networks. Clearly some teacher networking takes plac e in private. Teachers may spend recreational time together getting to know one another, or they may talk over lunch. However these examples of the netw orking process, though valid, are not feasibly quantified or qualified. In order to quantify teacher networks, I rely on teacher Inservice activity to determine if teacher networks aid in mobility. Specifically I focus on the

PAGE 55

44 number of hours teachers spend in Inse rvice activity as my measure of network size/strength. I argue that Master Inse rvice workshops, a part of the teacher recertification process, provide an effective forum for networking. To understand why this is the case, it is necessary to first revi ew how the re-certifi cation process works. Every teacher in the state of Florida is required to have an up-to-date license that enables him or her to teach in Florida public schools. This license is earned by every teacher at the beginning of his or her career and must be re-c ertified every five years. Teachers earn their re-c ertification through any combinati on of three things: they may take state-issued Florida Subject Area Exam in appropriate fields to prove their level of competency, they may enroll in and complete required coursework (6 credits minimum) at an approved university or college, and/ or they may choose to complete up to 120 hours of Master Inservice courses/workshop s offered by the school district. This state mandated re-certi fication process provides an ideal environment for testing the effects of teacher networking. Teachers who take the Florida Subject Area Exam in their appropriate fields or who decide to take classes at a College or University have little to no interaction th eir colleagues in their school dist ricts. However, the Master Inservice option is quit e different from other re-certifi cation options. First, Inservice courses/workshops are almost exclusively made up of other teachers, all within the same district and from differing schools.2 Thus teachers interact with others who share a common profession, and often common interest s. Secondly, the environment of most workshops is designed in such a way that encourages interacti on among participants. 2 It is possible that a marginal num ber of administrators, county officials, etc. attend these teacher components for other reasons (professional or personal). They are not included in the Inservice data in this paper.

PAGE 56

45 Workshop assemblies are long in duration (often several hours per session) and can span weeks or months. In addition, individuals ar e often called upon to sh are experiences with all the participants, or present material re garding how the workshop has influenced their classroom activity. Thirdly, Inservice is an attendance based program, as opposed to the graded college coursework or graded test format. One could argue that the relatively stress-free nature of the Inservice allows teachers to spend more time getting to know each other, free from the distr action of grade requirements. 2.2.2 How Teachers Use Their Networks There are many reasons why any professi onal or businessperson would network with others in their field of work. Th ey yearn to share ideas, understand their competition, and exchange information that can benefit their personal interests, sometimes in a mutual way. Why teachers netw ork with each other is not so different. They exchange classroom experiences and teaching methods, information on workplace environments, as well as build personal bonds that enhance their feelings towards the profession, or even their persona l lives. As a result they could possibly use their networks to plan intradistrict school events, find work in the summer, or even to socialize. However, there are specific applications of strong teacher networks that would enhance mobility within a district. First, te achers use their networks to gain information about the various student and administrative ch aracteristics of schools in their district. Greater access to this information lowers the cost of job searching and enhances mobility. Secondly, teachers use their networ ks to reveal their teacher quality in a less costly way for the purpose of being hired at other schools within their own distri ct. If teachers do use their networks to sort into preferred school s in a more efficient, less costly way, then

PAGE 57

46 there are obvious implications for student out comes, since others have successfully linked student outcomes to teacher mobility. 2.3 The Basic Model A teacher’s desire to transfer to school n (new school) is based on the utility they receive at school n minus the utility they receive at school o (old school). The utility at the old school is given by: Vo(so, qo, ao, do, o) where o indexes the old school, so is the salary of the old school qo is a measure of student/classroom quality, ao (administration quality), do (distance from school to home), and o (opportunity cost of teaching). Similarly the utility at the new school is given by: Vn(sn, qn, an, dn, o) where n indexes the new school, sn is the salary of the new school qn is a measure of student/classroom quality, an (administration quality), dn (distance from school to home), and o (opportunity cost of teaching). Teachers in cur explicit costs to movement. These costs arise from the search and application pro cess. More specifically, teachers may have to submit paperwork, spend some time interv iewing, or generally gather information about the school they are transferring to. Thus I write this teacher decision as: Transfer when: Vn – Vo – C(in, rn, nj*) 0 Do not transfer when: Vn – Vo – C(in, rn, nj*) < 0 In this case C( ) represents the explicit cost borne by the teacher, where in is the amount of relevant school information a teacher desires to gather, rn represents district and school requirements for transfers (paper work, interviews, etc.), and nj* representing the network

PAGE 58

47 size of the jth individual. I further note the relationship of each parameter to this explicit cost function. It is reasonable to assume that the less a teacher knows about a prospective school, the more time he/she spends gathering relevant informati on. Therefore we expect an increase in necessary or relevant information (in) to translate to a direct increase in C( ), such that C/ in > 0. In addition an increase in dist rict and/or school requirements for transfer will increase time and resour ces spent towards transfer, such that C/ rn > 0. When looking at network size it becomes appare nt that a larger network size will lend itself to better flow of school information, pos sibly fewer or shorter interviews, and more inside information that helps one to eliminat e unnecessary measures. Therefore it seems logical that a larger network will decrease external cost, such that C/ nj* 0. Lastly we must qualify the network to external cost relatio nship as not being strict. This allows for the possibility that for school n one’s network is totally ineffective. That is to say that one’s network is not the releva nt network, i.e., a network in another state. From these equations it is clear why teachers may be more likely to transfer as their relevant network size increases 2.4 Data and Empirical Study My data for my study comes from a variet y of sources. I utilize individual school data taken from the “Florida School Indicators Repor t” (FSIR) and the “School Advisory Council Report” (SACR). Countywide data was also supplied in various forms from the Florida State Department of Education (F LDOE). These forms include the “FLDOE State Survey #2”, “Teacher Exit Interview In formation”, an “Inservice Hours Report”, “The Profiles of Florida Distri ct Schools (Students and Sta ff Data)”, and “Teacher Salary,

PAGE 59

48 Experience and Degree Level”. The samp le includes 67 observations, one for each Florida school district. Because several variables introduced in th e following section are highly collinear to district size, such variables have been defined on a per school basis.3 2.4.1 Defining the Parameters The dependent variable for my study ( Transfers) is the number of transfers within a district per school. In or der to approximate the closen ess of different school job options, or capture the density of school job choice in a di strict, I use the total area, expressed in sq. miles, divided by the number of schools in the district ( School Distance ). Since a denser school di strict gives a mobile teacher more viable, less costly job options, I expect this variable to be positively correlated with the total number of transfers per school. Student/classroom quality is measured by taking the standard deviation of the percentage of free lunch eligible students among all schools in a given district. This measure ( School Difference ) should capture the variation of the percentage of the student body that is of low socio-economic status (SES) across schools. In essence, this variable measures how schools in a district differ along family backgrounds. Since greater variation in work environments should give teachers a greater incentive to sort, this variable should be positively corr elated with transfers per school. Wage differences may also play an importa nt role in teacher transfer decisions. There are no wage differences among public schools in a given district, but wage differences among districts still exist and ar e important to account for. As such, Salary 3Transfers Network Dislike Supervisor Principal Moves School Changes

PAGE 60

49 Difference is defined as the average salary of the highest paying district directly adjacent to the current district minus the average sala ry of the district in which a teacher is currently employed. If a teacher considers sa lary difference in a move between districts, it is plausible that they may consider the firs t-best alternative salary difference above all else. If higher salaries provide incentives fo r teachers to leave thei r district rather than move within district, then there should be less movement between schools within district. Therefore this variable should be negativel y correlated with within-district movement. In addition to the interdistrict salary di fference, a wage concern for teachers should also be the wage forgone from remaining a teacher. To capture the opportunity wage forgone, I construct a variable ( Wage Ratio ) using a county-level wage index. This wage index reflects the average wages across a predetermined set of occupations for each county. By dividing average salary for a dist rict by its county wage index, I measure how well teachers in a county are paid relative to other occupations. The larger the Wage Ratio the less appealing other jobs are. As a result this variable should be positively correlated with teacher transfers. My measure of network size ( Network ) equals total hours spent in Inservice per school.4 Since more hours spent in Inservice sh ould lead to larger and more effective networks, this variable should be positively co rrelated with teacher tr ansfers per school. It is also true that the Network variable is positively corr elated with larger school size. This is a potential problem since sc hool size, measured in average number of teacher faculty per school, is positively correlated with district size itself. That is to say 4Component hours geared towards non-instructional staff (i.e., food service employees, transportation staff, administrative staff, substitute teacher s, etc.) are not included in this measurement. Since each district adheres to a statewide numbering standard for component identification, I can be sure that I only include components which were teacher-oriented.

PAGE 61

50 that as school districts grow, the size of th e schools as measured in total faculty also grows. Since teachers in larger districts (with larger schools) have more positions to choose from, they should have greater transfer opportunities. To prev ent this relationship from biasing any estimates of the Network variable, I create a vari able to control for the effect of larger school size. This variable, School Size, measures the average teaching staff per school in the a district. New public schools opening and old public sc hools closing also present teachers with new choices in making transfer decisions New schools that open in a district may provide up-to-date facilities and technology in the classroom making teaching there more desirable. School closures force teachers to reorganize into other schools or leave teaching entirely. The variable ( School Changes ) captures these el ements of school turnover. This variable is measured by assigning a value of positive one for every school gained and/or lost from the 2000-01 school year to the 2001-02 school year, and then summing the totals for each county and dividi ng by the number of schools. Since school districts where schools open a nd/or close should result in mo re teacher transfers, this variable should be positively co rrelated with transfers. There is another variable emphasized in th is model that is based on administrative movement, specifically movement between schools by existing principals. When principals transfer from school to school w ithin a district, positive teacher/principal relationships may induce principa ls to recruit teachers to wo rk at a new assignment. In essence principals may bring their favorite te achers with them to another school under the expectation/promise of a positive work environment. This variable ( Principal Moves ) is measured as the total number of principa ls who moved from one school to another

PAGE 62

51 school, divided by the number of schools. If principals take teachers with them when they move between schools, then principal transfers should be posit ively correlated with teacher transfers. Another variable that should affect teach er movement within a district is the number of relevant alternatives. It is ar guably difficult for teachers to move between high schools and elementary/middle schools. Obviously high school teachers who work in districts where there is only one high school cannot move to another high school in their district. As a result, one would exp ect that school district s with only one high school will have an inherently lower rate of teacher transfer compared to similar districts with multiple high schools. To account for this disparity I include a dummy variable ( One HS ) that receives a value of 1 when a school district ha s only one high school (zero otherwise). This variable should be nega tively correlated with teacher transfers. The last two variables I include in the re gression help to desc ribe the state of a district’s preferential treatment towards transferring existing teachers to fill new vacancies when they open, rather than hiring new teachers. Some districts require that open positions are filled with teachers al ready employed before a new teacher may be hired; while some other distri cts give no special consideratio n to existing teachers over new applicants. Since more relaxed or favor able policies towards teacher transfers may enhance teacher mobility, I introduce two dummy variables that describe the priority given to teacher transfers. The first variable ( Transfer PC ) describes a district that affords partial consideration to existing teach ers. This variable takes on a value of 1 when a district gives preferential treatment, but not exclusive privilege, to its teacher transfers (zero otherwise). The second dummy variable ( Transfer CC ) describes a

PAGE 63

52 district’s policy to necessarily fill a new position with a teacher transfer request before considering a new applicant. This variable takes on a value of 1 when a district offers complete consideration to existing teachers befo re all others (zero otherwise). In both cases, these variables should be positively correlated with tr ansfers per school to reflect the enhanced opportunity for mobility given by preference. 2.4.2 Empirical Testing The analysis begins with a simple OLS regression onto the dependent variable Transfers The results are expressed in Table 21, listed by coefficient with the relevant t-statistic listed in parenthesis. Columns 1 and 2 express the OLS regre ssion with our 10 major variables of interest. Some of these va riables that do not have the predicted sign, in this case School Distance Transfer PC and Transfer CC and are not significantly different from zero. Of the variables that are significant, ( Network School Difference School Changes and School Size ), each has the predicted sign. The positive and significant coefficient of the Network variable in Column 2 seems to suggest that networks exist, and th at in fact they do se rve to enhance teacher intradistrict mobility. As a teacher networks gr ows, teacher mobility rises. This provides evidence that teachers are using their netw orks to gain information and enhance the transfer process. To determine the magnit ude of the impact of networking on transfer rates, I look at the percent change in the av erage number of teacher transfers per school when the amount of networking increases by one standard deviation. I find that an increase in the number of inservice h ours per school by approximately 1000 hours increases the teacher transfer rate per sc hool by approximately 11.49%. These results can be viewed in Table 2-2.

PAGE 64

53 The results in Table 2-1 also suggest that student characteris tics, school size, and schools opening and closing all influence teacher movement. Teachers seem to sort more frequently in districts wher e variation in student quality across schools is high. The variable School Difference has a significant impact in t eacher movement and shows that teachers are more likely to move around, or so rt, when there are sizeable differences among student populations at different school s. According to the regression, a one standard deviation increase in the measure of school heterogene ity implies a nearly 15.25% increase in teacher transfers per school Teachers also seem to be more mobile in districts where schools are larger. The School Size variable is positive and significant, suggesting teachers face greater opportunity to sort when there are more total positions in a school. The coefficient of the School Size variable implies that a one standard deviation incr ease in the average size of a school leads to a roughly 17.85% increa se in transfer rates per school. Besides student and school characteristics, teachers also have greater mobility in districts where schools are opening and closing with greater frequency. The School Changes variable suggests that as schools open a nd close in a district, teacher mobility rises. One would expect that as schools ope n in a district, teachers seek new facilities, new classrooms, etc. Further, as schools clos e teachers are forced to shuffle and sort to other schools so that they may continue teach ing in their district. The evidence suggests that when 5 schools are added in a district with 100 schools, teacher transfers increase by nearly 30% for every school in the district. Column 2 runs the same regression usi ng robust standard errors. Using these robust standard errors helps to ease worries ab out heteroskedasticity in the regression. If

PAGE 65

54 it were true that smaller districts have sma ller measurement error than larger districts (which seems plausible), then r obust standard errors help to control for this problem. We see roughly the same results as the regression in Column 1. School Size and School Difference have the predicted sign and are significant at the 1% level. School Changes also has the predicted sign and is significant at the 0.05% level. Column 3 reports the same regression as in Column 1 except that one additional variable is included. This variable ( Dislike Supervisor ) is a rough measure of teacher dissatisfaction with administra tive oversight in one’s own sc hool. It is calculated for each district as the number of teachers pe r school who cited “dissatisfaction with supervisor” as a reason for voluntarily termin ating their employment on the Teacher Exit Interview from the 2000-2001 school year.5 If teacher discontent with administration leads to greater mobility, then this variable should be positively correlated with teacher transfers. In Column 3, the variable Dislike Supervisor has the predicted sign but is not significant. It suggests that job environments and teacher attitudes are not necessarily an important part of the movement decision. This result seems to defy intuition, however in the next section I will show later that this result is marginal, and that Dislike Supervisor is sensitive to model specification. Column 4 reports the same regression as in Column 3 using robust standard errors. It is also worth noting that the Dislike Supervisor variable is not an ideal measure of overall teacher a ttitudes towards administ ration because it does not include teachers who continued to teach in their own districts and only reflects a self5 This survey includes teachers who co ntinued to teach in other districts or other states, teachers who stayed employed in education departments in and outside of their district, teacher s who left the education profession altogether, those who retired, and teachers who left for private scho ols in and out of their district.

PAGE 66

55 selecting subgroup of teachers. It does however have some benefits. First, since there is no descriptive teacher evaluation of administ ration formally collected by the FLDOE or school districts, nor is there us eable county data of the same na ture, this variable offers at least some approximation of what that data might yield. Secondly, because the teachers in these interviews have already committed themselves to alternative employment, the responses given in these interviews are assu red to be candid and honest. It is unlikely that a teacher respondent fears having their identity becoming known to others in these interviews. Overall, these results seem at least superficially consistent with Hanushek, Kain and Rivkin findings (1999, 2001) th at student characteristics in fluence mobility, and that salary changes at best have only a modest im pact. However none of these variables seem to have a powerful an impact as the School Changes variable. There seems to be a large disparity between the predictive power of this variable, and other variables. To ensure that our results are valid, and to understand w hy each variable induces the kind of change it does, I check the robustness of the data. 2.5 Robustness Checks 2.5.1 Re-examining the Data With respect to the School Changes variable, one observa tion in the dataset experienced an abnormally high rate of school change between the years of 2000-2001 and 2001-2002. This county in particular closed two school facilities to open one, much larger school, that would incorporate all of the existing student population of the other schools, however the small size of this distri ct seems to exacerbate the measurement of the School Changes variable. It is possible that this one observation alone is driving the results of the School Changes variable. If this observation is more of an outlier, then it

PAGE 67

56 may be overstating the overall importance of the variable it influences. An OLS regression analysis without this outlier is reported in Table 2-3. Once the sample is modified to eliminate the possible outlier, the magnitude of the coefficient of the School Changes variable is reduced by almo st half, though the variable remains significant. This doe s not necessarily mean that th is variable is not important, however it does show that perhaps this outlier should be dropped to maintain the integrity of these results. These new results suggest that a one standard deviation increase in school openings/closings leads to a 14.55% increa se in teacher transfers per school. This change is considerably lower than the previous estimate. All other variables in the re gression have the predicted si gn and all of the formerly significant variables ar e still significant. Network and School Size are significant at the 1% level, while School Difference becomes marginally significant at the 10% level. Column 2 reports the same regression with r obust standard errors. Column 3 includes the Dislike Supervisor variable, which is significant at the 10% level. This result suggests that there may be some value to job envir onment and administrative quality that teachers consider in their movement decisions. Finally Column 4 presents the same regression as in Column 3, including robust standard errors. I again assess the impact of each variable independently, by determining the effect of an increase of one standard deviation of each variable. These results ca n be viewed in Table 2-4. With the exception of School Changes there are no substantial differences in th e magnitudes of the coefficients of the other variables. The key variable of interest, Network is positive and signif icant in each regression specification on Table 2-3, suggesting that netw orks are an important factor contributing

PAGE 68

57 to teacher movement. However, to confirm that the results are credible, I further test the robustness of these results. Specifically I address whether the variable itself is endogenous to the system. In the following se ction I test whether the measurement of networking is correlated with some unobserved teacher ch aracteristic(s). 2.5.2 Unobserved Correlation If my measure of teacher networks is correlated with some unobserved teacher characteristic(s), then it may be true that there is no real netw orking taking place. Instead, teachers who complete high levels of Inservice may simply be more likely to transfer for other reasons. If that were true then one should expect that teachers who complete Inservice would also be just as likely to move in ways other than just intradistrict transfers. To test whether the Network variable is reflec ting true networking or is just reflecting self-selection of highly mobile teachers into Inservice, I utilize new dependent variables that depict other types of teacher movement. Private Move equals the number of teachers who claim a move into private schools per school. The second variable, District Move measures the number of teachers moving into other Florida districts per school. The last variable is teacher movement to other states to teach. This variable, State Move equals the number of teachers moving to other states per school in the district. Regression results with these new dependent variables are reported in Table 2-5 using robust standard errors. Columns 1 and 2 report the regressions with Private Move as the new dependent variable, and using robust standard erro rs. Columns 3 and 4 do the same with District Move as the new dependent variab le, and Columns 5 and 6 have State Move as the new dependent variable. In each of these specifications, networ k size does not seem to be a significant factor influencing te acher movement. Since Inservi ce is not significant to any

PAGE 69

58 of these different kinds of movement, it seem s plausible that teachers who tend to move around more generally, are not se lf-select into attending a grea ter amount of Inservice. Thus I can be more confident that the positive effect of networking is not simply a reflection of the preferences of more mobile teachers. 2.6 Conclusion The evidence set forth in this paper s upports the argument that Master Inservice components provide an effective environment for teachers to network with each other. These networks seem to provide a basis for teachers to gather important information about schools and allow teachers access to tran sfer options that mi ght have previously been closed to them. Overall, after c ontrolling for determinants of mobility and eliminating biasing outlier effect s, I show that an increase of one standard deviation in Inservice hours causes roughly a 12.74% increase in teacher transfers per school. The results suggest that teachers exploit these networks to sort into schools that they find desirable. If teachers can more effectively so rt away from less desirable schools, such as low-income or failing schools, into more desirabl e schools, then it is easy to conceive that worse schools will be worse off as teacher networks improve. Of course the goal of the Inservice plans set forth by school dist ricts is to improve teacher quality. However, the unintended consequences of such staff development policies that encourage, or at least facilitate, teacher networking seems to exist as well. Further, the consequence of this particular policy seems to be quite substantial in promoting a potentially harmful action such as teacher transfer. This result may have considerable meaning for the future of public education policy. State officials must begin to weigh the possible negative effects of incr eased teacher networking with the potential

PAGE 70

59 (and realized) benefits of Inservice, so that such policy analysis can be conducted in the future with a complete understanding of social welfare consequence.

PAGE 71

60 Table 2-1: Ordinary Least Squares Regressi on onto teacher transfers using full sample with and without robust standard errors Variable Column 1 Column 2 Column 3 Column 4 Dependent Transfers Transf ers Transfers Transfers Network 0.0003261 (1.56) 0.0003261 (2.00)** 0.0003474 (1.67) 0.0003474 (2.19)** School Distance 0.0065532 (1.12) 0.0065532 (1.16) 0.0078015 (1.33) 0.0078015 (1.33) Principal Moves 0.3216886 (0.12) 0.3216886 (0.12) 0.9270341 (0.34) 0.9270341 (0.33) School Changes 17.86031 (4.88)** 17.86031 (3.42)** 17.55322 (4.84)** 17.55322 (3.20)** School Difference 0.0584747 (2.10)** 0.0584747 (2.34)** 0.0557304 (2.01)** 0.0557304 (2.16)** Salary Difference -0.0000891 (1.61) -0.0000891 (1.50) -0.0000766 (1.38) -0.0000766 (1.21) Wage Ratio -0.0018573 (0.27) -0.0018573 (0.22) 0.0001112 (0.02) 0.0001112 (0.02) School Size 0.0501213 (2.46)** 0.0501213 (2.80)** 0.0448708 (2.19)** 0.0448708 (2.96)** Dislike Supervisor — — — — 4.17856 (1.43) 4.17856 (1.17) One HS -0.7262709 (1.32) -0.7262709 (1.27) -0.7582897 (1.39) -0.7582897 (1.31) Transfer Policy PC -0.2608151 (0.65) -0.2608151 (0.64) -0.2026472 (0.51) -0.2026472 (0.48) Transfer Policy CC -0.2381146 (0.57) -0.2381146 (0.67) -0.1001645 (0.24) -0.1001645 (0.25) Constant 0.2509715 (0.09) 0.2509715 (0.08) -0.5541842 (0.20) -0.5541842 (0.19) R-Squared 0.6154 0.6154 0.6295 0.6295 Number of Obs. 67 67 67 67 Robust SE N Y N Y denotes significance at the 10% level **denotes significance at the 5% level

PAGE 72

61 Table 2-2: Percent change teacher transfer s per school for a one standard deviation increase in each significant variable with full sample Variable Mean Standard Deviation Range of % Change School Changes 0.027 0.05 30.06 to 30.58 % School Size 33.418 11.61 17.85 to 19.94 % School Difference 15.945 7.99 15.25 to 16.00 % Network 1758.317 1047.62 11.49 to 12.46 %

PAGE 73

62 Table 2-3: Ordinary Least Squares Regression onto teacher transfers using sample with biasing outlier omitted Variable Column 1 Column 2 Column 3 Column 4 Dependent Transfers Transf ers Transfers Transfers Network 0.0003563 (1.79)* 0.0003563 (2.22)** 0.0003822 (1.95)* 0.0003822 (2.46)** School Distance 0.0010389 (0.18) 0.0010389 (0.20) 0.0021892 (0.37) 0.0021892 (0.41) Principal Moves 0.3430332 (0.13) 0.3430332 (0.12) 1.147493 (0.46) 1.147493 (0.36) School Changes 9.11579 (1.90)* 9.11579 (1.91)* 8.322631 (1.76)* 8.322631 (1.73)* School Difference 0.0433008 (1.60) 0.0433008 (1.74)* 0.0393933 (1.48) 0.0393933 (1.55) Salary Difference -0.0000671 (1.26) -0.0000671 (1.06) -0.0000516 (0.97) -0.0000516 (0.79) Wage Ratio -0.0000751 (0.01) -0.0000751 (0.01) 0.0020135 (0.30) 0.0020135 (0.30) School Size 0.0502127 (2.60)** 0.0502127 (2.89)*** 0.0442077 (2.29)** 0.0442077 (2.95)*** Dislike Supervisor — — — — 4.78275 (1.74)* 4.78275 (1.52) One HS -0.4594238 (0.86) -0.4594238 (0.94) -0.4825945 (0.92) -0.4825945 (0.96) Transfer Policy PC -0.2247365 (0.59) -0.2247365 (0.55) -0.1563356 (0.41) -0.1563356 (0.37) Transfer Policy CC -0.3207089 (0.81) -0.3207089 (0.89) -0.1669839 (0.42) -0.1669839 (0.42) Constant 0.0560173 (0.02) 0.0560173 (0.02) -0.8754046 (0.33) -0.8754046 (0.33) R-Squared 0.6033 0.6033 0.6247 0.6247 Number of Obs. 66 66 66 66 Robust SE N Y N Y

PAGE 74

63 Table 2-4: Outlier excluded percent cha nge teacher transfers per school for a one standard deviation increase in each significant variable Variable Mean Standard Deviation Range of % Change School Size 33.696 11.48 17.40 to 19.73 % Network 1777.904 1043.22 12.73 to 13.65 % School Difference 15.881 8.03 10.84 to 11.91 % School Changes 0.023 0.04 0.37 to 11.36 % Dislike Supervisor 0.027 0.05 8.78 %

PAGE 75

64 Table 2-5: Ordinary Least Squares Regression of other types of teacher movements with outlier excluded Variable Column 1 Column 2 Column 3 Column 4 Column 5 Column 6 Dependent Private Move Private Move District Move District Move State Move State Move Network 0.0000101 (1.21) 0.0000104 (1.22) 0.0000342 (0.49) 0.0000305 (0.43) 0.0000014 (0.06) 0.0000034 (0.01) School Distance -0.000214 (1.09) -0.000196 (1.02) 0.0024853 (0.95) 0.0021902 (0.85) -0.000891 (1.36) -0.000813 (1.22) Principal Moves -0.001503 (0.02) 0.0073589 (0.07) -1.829443 (2.09)** -1.930217 (2.01)** -0.190106 (0.53) -0.142873 (0.39) School Changes 0.125987 (1.11) 0.1214916 (1.07) 4.121108 (1.44) 4.236288 (1.46) 0.6497598 (0.90) 0.5957743 (0.78) School Difference -0.001018 (0.73) -0.001058 (0.74) 0.0119132 (0.99) 0.0124806 (0.99) 0.0015131 (0.44) 0.0012472 (0.35) Salary Difference -0.000001 (0.43) -0.000001 (0.31) -0.000001 (0.06) -0.000003 (0.18) -0.000002 (0.37) -0.000001 (0.20) Wage Ratio 0.0002086 (0.96) 0.0002341 (1.02) -0.006608 (2.47)** -0.006912 (2.35)** -0.002273 (2.41)** -0.002131 (2.18)** School Size -0.000328 (0.42) -0.000405 (0.51) 0.0019503 (0.32) 0.0028223 (0.47) 0.0017899 (0.76) 0.0013812 (0.58) Dislike Supervisor — — 0.0611693 (0.66) — — -0.694535 (0.70) — — 0.3255332 (1.10) One HS -0.014332 (1.04) -0.014801 (1.08) 0.2299846 (0.87) 0.2333494 (0.88) 0.0774991 (0.95) 0.0759221 (0.91) Transfer Policy PC 0.0199595 (1.06) 0.0197389 (1.05) -0.111869 (0.73) -0.121802 (0.81) -0.007048 (0.17) -0.002392 (0.06) Transfer Policy CC 0.0022509 (0.12) 0.0042704 (0.21) -0.174603 (1.20) -0.196926 (1.29) -0.050714 (1.04) -0.040251 (0.78) Constant -0.041984 (0.50) -0.053770 (0.59) 2.621647 (2.52)** 2.756905 (2.38)** 1.011499 (2.84)** 0.9481022 (2.54)** R-Squared 0.1931 0.1977 0.3416 0.3461 0.2398 0.2511 Number of Obs. 66 66 66 66 66 66 Robust SE Y Y Y Y Y Y Outlier N N N N N N

PAGE 76

65 CHAPTER 3 WITH A LITTLE HELP FROM MY FRIENDS: EVIDENCE OF TEACHER NETWORKS USING MICRO DATA 3.1 Introduction In the last several years, the issue of teacher mobility has come to the foreground of discussions about the public education. Pa rents and families who value teacher quality have come to realize that teacher mobility is a topic that has pot entially far reaching consequences, not just for education labor ma rkets, but also in terms of levels of education and student achievement. Schools that struggle with teacher retention often find job vacancies difficult to fill, and may compensate by hiring less than fully qualified teachers, expanding class sizes, canceling cour se offerings and assi gning teachers from other subject areas (NCES 1997). Of course these kinds of actions may have adverse effects on student learning and achievement. As a result, economists and policy makers have done more to explore the issues of t eacher attrition and movement within school districts. The goals of such recent studies have been to shed light on the questions of which teachers are the most mobile, how do these teachers move, when do these movements take place, and what are the potential implications for students. The most recent studies focused on teacher transfer behavior and sorting within education have provided evidence that supports the notion that stude nt characteristics and school quality are important factors in a teach er’s transfer decision (Hanushek, Kain, and Rivkin 2001). This also seems consistent wi th teacher self-reported information. In the 2000-2001 Teacher Follow-up Survey, administered by the National Center of Education

PAGE 77

66 Statistics, 32% of teachers reported poor wor kplace conditions as a primary reason for the movement from one school to another. Rese arch shows that teachers sort away from low-income, low-achieving schools. Conseque ntly, less experienced teachers are more typically found in poor, nonwh ite, low-performing schools, pa rticularly those in urban areas (Lankford, Loeb, and Wyckoff 2002). For policymakers, this can be a troubling re sult. If more experienced teachers tend to sort away from poorly performing schools, then it is the students at these schools who stand to lose the most, because experien ced teachers have been found to be more effective than novice teachers in terms of higher student achievement (Hanushek, Kain, and Rivkin 1998, 1999).1 In addition, urban schools a nd highly urban school districts face high rates of teacher tur nover (Imazeki 2003). If teache rs are sorting away from these areas, then these urban districts may ha ve the most difficult time filling vacancies with replacement staff. Aside from student characteristics, econom ists have also studied the potential effects of salary changes on t eacher mobility. So far researchers have found that salary levels have only a modest impact in terms of student achievement and almost no impact on teacher mobility (Hanushek, Kain, and Rivk in 1999). However, others have also found that salary can influence a teacher’s m ovement choice in ways other than strict transfer, such as length of stay in te aching (Murnane and Olsen 1990). Although much has been learned about teacher mobility, resear chers still do not have the complete picture on how teachers move and why. 1 Hanushek, Kain and Rivkin (1998) s how that variations in teacher qu ality account for 7.5% of the total variation in student achievement. In addition, teache r quality differences tend to outweigh school quality differences considerably.

PAGE 78

67 The focus of this paper is to better unde rstand the nature of teacher transfers from school to school by introducing the idea of t eacher networks as an effective sorting mechanism for public school teachers. To st udy this issue I rely on information provided by the 1999-2000 Schools and Staffing Survey (SA SS) put out by the National Center for Education Statistics. The individually detailed and specific nature of the dataset not only allows me to consider how individual networ ks affect mobility but also allows me to control for a well-defined set of variables, gi ving my study a more complete and detailed look at networks. This paper will begin disc ussion on the notion of teacher networks, and will address how these networks influence teacher movement and to what extent. Specifically this paper descri bes how professional developm ent activities (PDAs) provide opportunity for teachers to netw ork with other teachers. PD As will be shown to have a sizable impact on intradistrict teacher mobility. 3.2 The Identification Strategy Before I describe how I identify teacher ne tworks, let me first discuss what teacher networks are, how they are useful, a nd how teachers build these networks. 3.2.1 What Are Networks And How Are They Useful The term network in this paper is define d as group or “family” of co-workers or colleagues within one’s field of work. Specifically, in this paper a network refers to a teacher’s associations with other teachers w ho work within their own district. Although the context of a network here is limited to the field of education, people from all professions form networks for various reasons. There are many reasons why any professi onal or businessperson would network with others in their field of work. Th ey yearn to share ideas, understand their competition, and exchange information that can benefit their personal interests,

PAGE 79

68 sometimes in a mutual way. Why teachers netw ork with each other is not so different. They exchange classroom experiences and teaching methods as well as information on workplace environments, as well as build pe rsonal bonds that enhance their feelings towards the profession, or even their personal lives. As a result, teachers may be able to use their networks to plan intradistrict school events more efficiently, find work in the summer time, or sometimes just to socialize w ith others in their field. However, there are two key uses of teacher networks that are the focus of this paper. First I argue that teachers use their networks to gain greater information about school characteristics. Because there are many characteristics about schools that may be difficult to observe (existing job vacanci es, quality of administrative support, departmental environments, etc.), networks may allow teachers to evaluate other schools in their district more efficiently. Greater information about schools should allow teachers to sort into their pref erred school more easily. Secondly, I argue that teachers use their networks to reve al their quality as a job candidate in a less costly wa y. Because it can be costly (or sometimes impossible) to reveal one’s aptitude for a job vacancy, t eachers may call on their networks to provide important or relevant information to potential employers. For example, a teacher may ask another teacher to put in a favorable word to a hiring principal. In this way, networks provide a valuable advantage to those seeking employment, while at the same time being a more credible or reliable source of info rmation for potential employers. If teachers do use their networks to foster more efficient, less costly sorting behavi or, then policies that encourage network building may have unint ended outcomes for teacher mobility and possibly student achievement.

PAGE 80

69 3.2.2 How Do Teachers Build Networks? Clearly some teacher networking takes pl ace in private forums. Teachers may spend recreational time together getting to know one another, or they may talk over lunch. However these examples of the networ king process are not f easibly quantified. To find a more measurable environment where networking takes place, I focus on professional development activities. Profe ssional Development Activities (PDAs) consist of an array of different activities geared towards maintaining and enhancing teacher competency and knowledge of various issues (including such issu es as the use of technology in the classroom, st udent assessment and state ed ucation standards, teaching methods, etc.). PDAs can include many different types of activities such as workshops, mentoring and/or peer observation, collabo rative research, University coursework, etc. Table 3-1 reports a summary of self-reported profe ssional development activity from the 1999-2000 Schools and Staffing Survey.2 Though each of these activities may be available to most teachers in the sample, attending workshops/c onferences/training clearly seems to be the most popular form of PDA in the sample, w ith nearly 94% of teachers reporting they have engaged in this form of PDA in the pa st year (prior to the 1999-2000 school year). PDAs provide an ideal environment for t eachers to network with one another. Teachers who participate in workshops or c onferences have greater opportunity to meet teachers from other schools within their distri ct and enhance their networks. In addition most PDA is geared to a specific employm ent group within the school system. This 2 The 1999-2000 Schools and Staffing Survey (SASS) is a survey of involving approximately 56,000 public school teachers from all over the U.S., and is produced by the National Center of Education Statistics. It is the main data source for this chapter and is described further in Section 3.3.

PAGE 81

70 means that administrative personnel do not usuall y engage in the same PDA as teachers. Likewise teachers do not usually engage in th e same PDA as other s upport staff (such as bus drivers, cafeteria workers, etc.). Thus I can be somewhat confident that if teachers are building networks through PDA, they are generally relevant networks. Besides simply meeting other teachers th rough PDA, these activities often provide an environment conducive to promoting teache r interaction. Workshops especially are designed in such a way that encourages interaction among participants. Workshop assemblies are usually long in duration (often several hours per session) and can span weeks or months. In addition, individuals ar e often called upon to sh are experiences with all the participants, or present material re garding how the workshop has influenced their classroom activity. 3.2.3 Measuring Teacher Networks In order to measure the size of a teacher’s network, I utilize national survey data that provides information on professional development activity. The 1999-2000 Schools and Staffing Survey (SASS), the survey from which the overall data set is derived, asks respondents to report the number of hours they spent in e ach of six major categories3 of PDA within the last school y ear (1998-99). Rather than re port the specific number of hours, respondents are asked to select a (pre determined) range of hours that includes the actual number of hours they have spent in each category of PDA. Using the midpoint of the reported range as my best estimate of repor ted hours, I create a co ntinuous variable of 3 These six categories are: PDAs w ith focus on 1) in-depth study of main assignment, 2) content and performance standards of main assignmen t, 3) methods of teaching, 4) uses of computers for instruction, 5) student assessment/methods of testing, and 6) student management in the classroom.

PAGE 82

71 PDA Hours ( Hours) by summing the reported hours acro ss all six PDA categories for each individual.4 In general the distribution of PDA Hours across my sample is skewed to the left with approximately 3.1 % of all teachers repo rting that they have not engaged in any form of PDA within the past year. Bo th the level of skewness of the sample (approximately1.30) and percent of teachers re porting no PDA within the past year are roughly equivalent across m ovement categories as well.5 In addition, the distribution of PDA Hours across these movement categories is also roughly identical. Of those teachers who did not move, 24.8 % (or 332) are in the upper quartile of the entire distribution. Likewise 25.2 % of teachers (or 88) who transfe rred within district were in the upper quartile of the enti re distribution of PDA Hours.6 It is also worth noting that in each individual PDA category, only a small fr action of teachers are top-coded (reporting the maximum number of hours allowed in the survey).7 Although for each PDA category the percent of teachers who ar e top-coded ranges from 16.4 % to 2.1 %, these differences do not vary significantly across movement types. Overall the data seems to indicate that 4 It is important to note that the highest range of hour s that teachers could report is “33 or more”. Where this happens, I use 60 hours as the maximum value in this range. 60 seems to be a reasonable upper limit of hours, however I have also tested maximum values at 48 and 64. These chan ges do not significantly impact the results. 5 The PDA distribution of t eachers who move intradistrict is slightly more skewed (1.37) than the same distribution for those who did not move (1.25). Additionally, the kurtosis of the PDA distribution for all movement categories is nearly equivalent (4.34 for non-movers, 4.64 for intradistrict movers, and 4.61 for those who leave thei r district). 6 Of those teachers who moved out of their district, 21.7 % (or 98) are in the upper quartile of the entire distribution. This includes both interdistrict and interstate movement. 7 There is only one teacher in my sample who is to p-coded for every category of PDA (and has thus taken the maximum amount of PDA measured in the survey)

PAGE 83

72 along the PDA measurement, teachers in va rious movement categories do not look different enough from one another to warrant an y concern these groups are incomparable. 3.2.4 The Basic Model In this basic model I am attempting to explain how a teacher’s decision to transfer to another school (in year t ) within their own district is affected by their network. My measure of a teacher network here is the num ber of hours a teacher spent in professional development activities in th e previous year (in year t – 1). I can start by expressing a simple regression model in the following form: Transfer(t) = + (PDA Hours(t – 1)) + X Here the dependent variable is a dummy th at expresses a teacher’s transfer within his own district. The dependent variable (Tra nsfer) is assigned a value of 1 for a within district move, and 0 otherwise. In additi on to the amount of networking a teacher does, the basic characteristics of th e most mobile teachers are somewhat different from those who are less mobile. To capture these diffe rences I include variable X, a vector of individual and school characteri stics that influence a transfer decision. Table 3-2 reports some descriptive statistics about the charac teristics of teachers who do and do not move around in various ways. The most notable di fference is that teach ers who move within district tend on average to be about 3-6 years less experi enced (and younger) then those who do not move. Since inexperienced teachers are generally not as far along the career path as more experienced teachers, this di fference seems intuitive. In addition, I must also consider marital status and job status in the model. Those teachers who are not married may find a move to a new school to be more costly since spousal relocation can complicate matters. Those teachers who are employed part-time by their district should

PAGE 84

73 have a greater overall benefit to searching fo r job vacancies since they stand to benefit from becoming a full-time employee. Aside from the natural differences that exist among teachers, a teacher’s transfer decision should also depend on the working co nditions he/she faced the previous year. Specifically, a teacher’s decision to tr ansfer may be influenced by student quality/characteristics, administrative quality/c haracteristics, or comp ensation levels. As a measure of student quality ( L ), I construct a measure: %Free Lunch Eligible = (%free-lunch eligibleschool,(t – 1) %free-lunch eligibledistrict,(t – 1)) This measure is the difference between the percent of the student population who are free-lunch eligible at a teacher’s school and the average percent of the student population who are free-lunc h eligible at the dist rict level. It is designed to capture the socio-economic status (SES) of the student popul ation at a teacher’s in itial school relative to the SES of the district as a whole. Give n that teachers desire to sort away from low SES schools, this variable should ha ve a positive impact on transfers. To measure administrative quality ( Adm ), I utilize the variable: Poor Administration = a dummy variable, receives a valu e of 1 is a teacher reports that administrative support is poor within the past year ( t – 1 ) This variable captures a teacher’s per ception of how supportive/encouraging the administration at their school is towards its st aff. The predicted sign of this variable should be positive, since in schools where admi nistrative quality is perceived to be poor, teacher mobility should be higher.8 Some additional measures of the working conditions in my regressions specification include: 8 I have also measured poor administrative quality us ing the average teacher response for each school. My results do not change with this different measure.

PAGE 85

74 Student Threat = a dummy variable, equals 1 if a teacher reports being threatened by a student within the past year ( t – 1 ) ( Thr ). Again because receiving a student threat in dicates poor working conditions, the sign of the coefficient should be positive to express teachers’ preferences to sort away from undesirable schools.9 Bonus Pay = a dummy variable, equals 1 if a teacher reports receiving bonus pay, separate from salary and extracurri cular pay within the past year ( BP ). Bonus pay should give teachers an incentive to stay at their existing school, and should therefore be negatively correlated with transfer rates. I also use the self-reporte d data to construct a meas ure of job satisfaction ( Job ) that reflects a teacher’s general attit ude toward their workplace: Job Dissatisfaction = a dummy variable, equals 1 if a teac her reports being strongly dissatisfied with being a teacher at their school in year ( t – 1 ) ( Job ) This variable is intended to capture not ju st the quality of the students, but also a teacher’s perception about the quality of thei r co-workers, their administrative superiors, school resource availability, and general i ssues surrounding their employment. Since greater general job dissatisfaction should lead to greater mobility, I expect the sign of this variable to be positive. While it is true that salary differences between schools seem to affect teacher transfer decisions, this analysis is limited to those teachers whose movement is within their district. Given that sala ries schedules are generally se t at the district level and do 9 I have also measured student threat using the aver age teacher response for each school. My results do not change with this different measure.

PAGE 86

75 not vary across schools in the same district, base salary differences between schools are negligible10. As a result, I do not include base salary as a variable in my analysis.11 I also control for whether a teacher belongs to a union in this model for two main reasons. The first is that union membersh ip may provide existing teachers with preferential treatment in filling a job vacancy. Another reason is that districts may find unionized teachers more difficult to fire. As a result, problem teachers may be shuffled around more often. These two explanations of how unions affect tran sfer rates are very different, however in both scenarios union membership should positively affect the probability of transferring. Finally I add state fixed-effects into th e regression model to control for basic institutional and cultural differences across states, and a dummy = 1 for missing data in the % Free-Lunch ( L) variable.12 It would be ideal to use sc hool district fixed-effects, however, doing this results in a dram atic reduction of my sample size.13 Once this is done, the regression takes the following form: Transfer(t) = + 1Hours(t – 1) + 2L(t – 1) + 3Thr(t – 1) + 4BP(t – 1) + 5PT(t – 1) + 6Exp(t – 1) + 7Adm(t – 1) + 8Mar(t – 1) + 9Job(t – 1) + 10Age + 11Union + 12LDummy + 13StateFE where PT is a dummy for job type (1 if part-time, 0 otherwise), Exp is the total years of full-time experience, Age is teacher age, Union is a dummy for union membership (1 is 10 In the SASS, over 98% of school districts reported having a set salary schedule for teachers in their district. This means that virtually every teacher in the sample who moved within district did not face salary differences between the schools in their district 11 Although other types of compensation such as incentive pay are an important component to one’s decision to transfer, those types of compensati on could be endogenously determined and therefore inappropriate to include. 12 The Free-Lunch variable is the only variable in the regression where observations were missing. Roughly 157 (or 9%) of observations were missing this data. 13 Never the less I show on Table 3-7 that the inclusion of district fixed-effects does not alter the results.

PAGE 87

76 unionized, 0 otherwise) and Mar is a dummy for marriage status (1 is married, 0 otherwise). 3.3 Data The data for this model comes from the 1999 – 2000 Schools and Staffing Survey (SASS) and the 2000 – 2001 Teacher Followup Survey (TFS) produced by the National Center for Education Statistics (NCES).14 The data are comprised of individual (teacher) level survey information that can be matched across surveys, as well as relevant survey information for the relevant schools, principals and school districts. In the original 19992000 SASS survey there were over 52,000 respo ndents, but the TFS consists of only a subset of those orig inal respondents. In aggregate the TFS includes 6,758 teachers, however not all of the teachers in the survey are included in this analysis.15 Since this analysis focuses on public school teachers who move within district relativ e to those who do not change schools, I eliminate those respondents who either leave th eir district, or leave teaching altogether. Later in the paper I will use other types of movement (int erdistrict and interstate movements) to show that PDA hours do not influence other mobility decisions, particularly those decisions wher e networks should not matter. 14 The datasets utilized in this paper contain iden tifiable and sensitive information. While there are versions of these data that are available to the general public, the datasets used in this paper are only available at the discretion of the NCES and with their proper approval. 15 Those respondents in the TFS who le ft the teaching profession after th e 1999-2000 school year (2,374 respondents) are given a version of the TFS that is different from the survey given to those who did not leave teaching.

PAGE 88

77 In addition, I eliminate those teachers who initially teach at private, Indian, or charter schools, teachers who reporte d their move was mostly involuntary16, and those teachers who were not considered full or part time regular teachers in 1999 (e.g., specialists, substitutes, student teachers and itinerant teachers). Once these criteria are imposed on the data set, 1,688 observations remain. Of the 1,688 observations in the dataset, 349 (or approximately 20.7%) of those were teachers who transferred within their district. 3.4 Empirical Evidence Because the dependent variable is a binary choice variable with a successful transfer equal to 1, probit estimation is appr opriate. Column 1 of Table 3-3 reports a preliminary regression of PDA Hours onto th e dependent variable, with the specified covariates also included.17 At first glance it seems th at networking does not have a significant impact on the likelihood of transfer. Networks th ough are a story of collegial kinship, and like any friendship, benefits may not be realized until the parties involved are familiar and comfortable enough with one another to begin a relationship. Thus networks may be difficult to establish initiall y. Further, networks may require a level of maintenance. If that is the case, then t eachers who engage in PDA frequently may have an increasing advantage over those who do not. This may imply that there are increasing returns to hours spent in PDA in terms of networking. A simple test of this assumption is to include a squared te rm of PDA Hours. 16 It is also worth noting that an y involuntary teacher movements who ma y be unavoidably included in the sample (thus creating noisy measures of intradistrict m ovement), are likely to bias any positive estimates of the effect of networking towards zero. 17 All regression estimates are clustered at the district level to control for possible standard error correlation at the district level. In addition, all regression estimates include robu st standard errors, to account for possible heteroskedasticity.

PAGE 89

78 In Column 2 of Table 3-3, the addition of a squared measure of PDA Hours in the regression seems to suggest that there may be some value to networking. Specifically, the coefficient on the squared term is posit ive and significant at the 0.025 level, which may indicate that while ha ving only a few hours networki ng may have no impact on transferring, positive returns to networking may exist for large values of PDA Hours. To further test whether this is th e case, I use a spline to allow the effect of an additional hour of PDA to be greater at high levels of PD A hours then at low levels of PDA hours. The spline estimates a continuous, piecewise-li near relationship between PDA Hours and transfer probability, and is defined as follows: PDA Hours (H) : from [0 – H*] hours: = H if H < H* = H* if H H* PDA Hours (H) : from [H* Hmax] hours: = 0 if H H* = (H – H*) if H > H* Table 3-4 reports the coeffici ents of several regressions using the spline. I test various cutoffs of H*, where H* is chosen at decile intervals al ong the distribution of PDA Hours. Table 3-4 illustrates that the 2nd segment of the spline is positive and significant for most decile cutoffs18, but the best fit for the regression is where the cutoff is at the 80th percentile. This table confirms that additional PDA Hours, particularly for those teachers in the top 20% of the overall distribution of PDA Hours, seem to have a positive and significant impact on a teacher’s tr ansfer decision. Interpretation of the coefficients suggests that a one standard deviation increase in the total number if PDA 18 For every cutoff chosen at or above the 20th percentile of the distribution, the 2nd segment of the spline estimation is positive and significant.

PAGE 90

79 Hours (approximately 46.6) can lead to a 0.020 to 0.078 increase in the probability of transferring within one’s own school district. Other covariates also seem to have a si gnificant impact of the transfer decision. The evidence seems to indicate that teacher s with less experience and those who work only part-time are significantly more likely to move between schools. Intuitively this seems valid since teachers with li ttle experience have had arguably less time to sort into a desirable school and should therefore be more likely to transfer than teachers who are further along the career pat h. In addition, part-time empl oyees stand to secure higher wages if they search for full-time employment, and have lower time-cost to job searching. Besides the natural sorting taking place for younger/less experienced teachers, teachers who work in low SES schools a nd who face undesirable working conditions clearly have a greater incen tive to sort into better sc hools, especially given that additional/incentive pay is rarely available for teachers to work in less desirable schools. The evidence in Tables 3-3 and 3-4 seem to support this claim. Teachers at schools where student SES as a whole is lower than the district averag e, schools where their overall job satisfaction is generally low, and teachers that feel administrative quality is generally poor, seem to have a higher propensity to sort. Lastly, Tables 3-3 and 3-4 show that ther e is also some advantage to a teacher being part of a teacher union in terms of increased mobility. This could be because teachers who are part of a union receive priority in filling a position vacancy. Another interpretation of this union effect could also be that unionized teachers are difficult to fire, and therefore get shuffled around more of ten. Overall the results shown in Tables 3-

PAGE 91

80 3 and 3-4 are consistent with previous findings in the literature, and not at all surprising given the economics going on. These regressi on specifications provi des some valuable insight into the teacher transfer decision, and also tells us that teacher networks are not an integral component to teacher mobility. 3.5 Robustness Checks The evidence in the previous section sugge sts that there is a significant relationship between teacher networks and transfer behavi or. However it is important to test the robustness of these results. If the time a te acher spends in PDA is simply correlated with some other (possibly unobservable) characte ristic(s), then the explanation provided by networking may be inaccurate. If it is true th at teachers who spend a great deal of time doing professional development activities are s imply more likely to transfer, then the results are spurious in a networking framework. Before I test the robustness of the results in the previous section, I look to see if those teachers who engage in the most pr ofessional development look substantially different from the average teacher. I compar e the characteristics of teachers in every decile of the distribution of PD A hours to the average characteristics of the entire sample of full and part-time teachers (2,280 observations ). The results of these comparisons are reported in Figure 3-1. E ach graph on Figure 3-1 shows the mean of the given characteristic for each interval (0% – 10% 10% 20%, etc) of the PDA distribution, centered about the mean for the entire sample. I use the standard errors at each interval to construct a 95% confidence interval on which the mean lies, and test whether the sample average falls in the range of th e confidence interval. There are only a few characteristics wher e teachers who engage in the most PDA look significantly different from the sample average. The graphs of Age and Full-Time

PAGE 92

81 Experience illustrate that teachers in the top end of the distribution of PDA Hours are significantly older and more experienced. However, previous ev idence suggests that more experienced teachers, who are also generally older, are also less likely to transfer schools. Therefore, if older and more expe rienced teachers are ta king the most hours of professional development, then we should expe ct PDA Hours to be negatively correlated with the probability of transferring, which does not seem to be the case. The graphs in Figure 3-1 may also suggest th at those teachers at the very bottom of the distribution of PDA Hours may be disproportionately white, male part-time teachers, with respect to the entire dist ribution of hours in PDA. Howeve r it is not clear that this presents a problem of selection bias in who d ecides to take PDA as it relates to transfer rates. Currently there is no evidence to sugge st that qualities of teacher race or gender are effective at predicting/explaining teacher movement. In addition, Table 3-4 showed that part-time teachers are more likely to m ove than full-time teachers. If part-time teachers are more likely to spend only a small amount of time in PDA, this should also downwardly bias the estimated effect of PD A Hours on transfers. These results are promising, however they are not conclusive. Teachers may still be sorting into the distribution of PDA Hours based on some characteristic that is not observable. To test whether teachers ar e sorting into high or lo w levels of professional development activity, I begin by looking at other types of teacher moveme nt. If it is true that teachers with high levels of PDA participation are simply more likely to move, then one might expect these same teachers are also mo re likely to move in ways other than just intradistrict transfer. However, since mo st PDAs are activities limited to one’s own school district, then networks should be la rgely ineffective for movement outside the

PAGE 93

82 scope of the immediate district. So if PDA hours are a strong predicto r of other types of movement, then that may be an indication th at the network relations hip is unjustified. Tables 3-5 and 3-6 report similar regressions similar to those reported in Table 3-4. Table 3-5 shows the effect of a change in PDA Hours on the probability of movement to another school dist rict within one’s state. Ta ble 3-6 shows the effect of a change in PDA Hours on the probability of movement to another school district in another state. In both cases, the effect of increased PDA Hours is not significantly different from zero. For Tables 3-5 and 3-6, neither spline segment is significant, which suggests that the amount of time a teacher sp ends in PDAs has no significant impact on how likely they are to move out-o f-district (within the state), or even out-of-state. This helps to quell some suspicion that teachers with high levels of PDA are simply more mobile. It may also be possible that the rela tionship between PDA and mobility is a function of district characteristic s or behavior. For example, if districts with “naturally” high levels of teacher mobility also demand an unusually high number of PDA hours for their faculty (arguably this may be the case with some urban school districts), then the positive effect of networking is simply a reflection of high levels of mobility in a subset of districts. To account for this possibility, I re peat the regressions in Table 3-4 with district fixed effects. Because the use of district-fixed effects limits the sample to districts where both movement and nonmoveme nt occur, the number of observations drops to only 398. Naturally this limits the power of the analysis, and is the main problem with using district fixed effects thr oughout. Table 3-7 reports the results of these regressions. According to the data, high levels of PDA have a positive effect on the

PAGE 94

83 transfer decision and low levels of PDA have a negative effect on th e transfer decision. The results in Table 3-7 confir m that district specific beha vior does not fully explain the relationship hours in PDA and mobility. 3.6 Conclusion The evidence presented in this paper clea rly seems to indicate that time spent in professional development activities leads to networks that enhance teacher mobility within their own district. These PDAs ar guably provide a forum in which teachers can build networks to gather im portant information about sc hool characteristics, work environments and job availability from which th ey can sort in a less costly, more efficient way. It also seems plausible that teachers can use their networks with favored colleagues to reveal their aptitude for a job vacancy, which may allow teachers access to transfer options that might have previous ly been closed to them. Results show that a one standard deviat ion increase in the amount of networking leads to a 0.020 to 0.078 increase in the proba bility of transferri ng within one’s own school district. If teachers use these networks to more e ffectively sort away from less desirable schools, such as low-income or failing schools, then networks arising from PDAs may contribute to the disparities in teach er quality across school types. It is also possible that if networks make it more difficult for low achieving schools to retain qualified teachers, then student achievement at these schools may also be diminished by networks. Oddly enough, this unintended c onsequence of staff development policies would be subversive to the goal of PDAs, wh ich is to improve education through greater teacher quality.

PAGE 95

84 Table 3-1: Professional Developm ent Activities Summary Statistics Number of Teachers Who Reported Yes No Percent Who Participated University Course in your Main 785 1495 34.4 % Teaching Field University Course NOT in your Main 614 1666 26.9 % Teaching Field Observational Visits to Other Schools 766 1514 33.6 % Individual or Collaborative Reseach on 1032 1248 45.3 % a Topic of Interest Regularly Scheduled Collaboration with 1561 719 68.5 % other Teachers on Issues of Instruction Mentoring or Peer Observation 1014 1266 44.4 % Participating in a Network of Teachers 559 1721 24.5 % (e.g. Internet Organization) Attending Workshops, Conferences 2141 139 93.9 % or Training Presenter at a Workshop, Conference 452 1828 19.8 % or Training

PAGE 96

85 Table 3-2: Summary Statistics for Various Teacher Groups No Move Move Intra-District InterDistrict Inter-State Number of Teachers 1400 (61.4 %) 370 (16.2 %) 484 (21.2 %) 96 (3.9 %) Mean Std. Dev. Mean Mean Mean PDA Hours 55.728 46.662 59.000 49.636 48.220 % Free-Lunch Eligible 0.002 0.225 0.045 -0.014 -0.021 Bonus Pay 0.137 0.344 0.113 0.101 0.146 Student Threat 0.095 0.294 0.127 0.130 0.083 Job Dissatisfaction 0.021 0.144 0.048 0.066 0.072 Full-Time Experience 10.915 9.856 7.959 5.134 4.583 Part-time employment 0.040 0.197 0.081 0.041 0.010 Poor Administration 0.069 0.254 0.102 0.101 0.104 Union Member 0.783 0.411 0.778 0.652 0.667 Age 40.484 11.340 37.956 33.634 32.030 Gender (Male) 0.282 0.450 0.235 0.299 0.292 Race (White) 0.902 0.297 0.862 0. 902 0.927 Marital Status 0.695 0.460 0.678 0.665 0.667

PAGE 97

86 Table 3-3: Probit Regression of Intra-District Movement Variables 1 2 0.0010626 0.0006944 PDA Hours (1.27) (0.62) — 0.00003 (PDA Hours)2 — (2.30)** 0.301017 0.3069657 % Free-Lunch Eligible (1.84)* (1.87)* 0.1669206 0.1629131 Bonus Pay (1.40) (1.37) 0.1009573 0.1002069 Student Threat (0.87) (0.86) 0.4224374 0.4353552 Job Dissatisfaction (1.85)* (1.92)* 0.0143332 0.0131213 Full-Time Experience (2.34)** (2.13)** 0.5452476 0.5394231 Part-time employment (3.43)** (3.41)** 0.3542717 0.3600298 Poor Administration (2.55)** (2.59)** 0.2372197 0.2386774 Union Member (2.36)** (2.37)** -0.005178 0.0055949 Age (1.01) (1.09) 0.0237581 0.0261624 Marital Status (0.29) (0.32) Dummy for Missing 0.1770793 0.176266 % Free Lunch Eligible (1.20) (1.20) 0.2702219 0.2997908 Constant (1.37) (1.50) Pseudo R-squared 0.0955 0.0984 Number of Obs. 1627 1627 Robust SE, clustered Y Y State FE Y Y denotes significance at the 0.05 level ** denotes significance at the 0.025 level

PAGE 98

87 Table 3-4: Probit Regression of Intra-District Movement Regression of PDA Hours with H* Chosen at Decile Intervals Along the Distribution of PDA Hours H* = 9 H* = 17 H* = 22.5 H* = 29.5 H* = 40.5 H* = 51.5 H* = 68 H* = 87.5 H* = 122.5 Variables (10%) (20%) (30%) (40%) (50%) (60%) (70%) (80%) (90%) PDA Hours 0 H* -0.0262 -0.0170 -0.0117 -0.0074 -0.0047 -0.0033 -0.0022 -0.0018 -0.0006 (1.12) (1.68) (1.63) (1.43) (1.34) (1.27) (1.12) (1.15) (0.55) PDA Hours > H* 0.0013 0.0017 0.0018 0.0020 0.0022 0.0026 0.0031 0.0043 0.0061 (1.52) (1.85)* (1.94)* (1.94)* (2.00)** (2.08)** (2.16)** (2.45)** (2.51)** % Free-Lunch Eligible 0.3041 0.3066 0.3077 0.3074 0.3067 0.3086 0.3117 0.3081 0.3010 (1.86)* (1.87)* (1.88)* (1.88)* (1.88)* (1.89)* (1.91)* (1.88)* (1.83)* Job Dissatisfaction 0.4214 0.4254 0.4282 0.4285 0.4306 0.4338 0.4377 0.4385 0.4288 (1.86)* (1.87)* (1.88)* (1.88)* (1.89)* (1.90)* (1.92)* (1.93)* (1.89)* Full-Time Experience -0.0141 -0.0136 -0.0135 -0.0134 -0.0132 -0.0130 -0.0130 -0.0130 -0.0134 (2.29)** (2.22)** (2.19)** (2.17)** (2.13)** (2.10)** (2.10)** (2.10)** (2.18)** Part-time employment 0.5383 0.5362 0.5351 0.5373 0.5417 0.5458 0.5479 0.5477 0.5381 (3.37)** (3.36)** (3.36)** (3.38)** (3.41)** (3.44)** (3.45)** (3.46)** (3.41)** Poor Administration 0.3575 0.3616 0.3636 0.3641 0.3638 0.3621 0.3595 0.3596 0.3593 (2.57)** (2.59)** (2.61)** (2.61)** (2.61)** (2.60)** (2.58)** (2.58)** (2.59)** Union Member 0.2344 0.2406 0.2419 0.2411 0.2415 0.2416 0.2399 0.2387 0.2373 (2.32)** (2.38)** (2.40)** (2.39)** (2.39)** (2.39)** (2.37)** (2.37)** (2.36)** Student Threat 0.1011 0.1016 0.1022 0.1025 0.1016 0.1000 0.0984 0.0987 0.0991 (0.87) (0.87) (0.87) (0.88) (0.87) (0.86) (0.85) (0.85) (0.85) Bonus Pay -0.1674 -0.1639 -0.1639 -0.1657 -0.1658 -0.1650 -0.1635 -0.1633 -0.1658 (1.41) (1.38) (1.38) (1.39) (1.40) (1.39) (1.38) (1.38) (1.40) Age -0.0053 -0.0054 -0.0054 -0.0055 -0.0055 -0.0056 -0.0057 -0.0058 -0.0056 (1.02) (1.05) (1.06) (1.07) (1.08) (1.09) (1.10) (1.12) (1.09) Dummy for Missing 0.1789 0.1779 0.1798 0.1796 0.1790 0.1777 0.1772 0.1807 0.1776 % Free Lunch Eligible (1.22) (1.21) (1.22) (1.22) (1.22) (1.21) (1.21) (1.23) (1.21) Marital Status 0.0232 0.0224 0.0232 0.0245 0.0256 0.0265 0.0266 0.0277 0.0264 (0.28) (0.27) (0.28) (0.30) (0.31) (0.32) (0.33) (0.34) (0.32) Constant -0.9070 0.5313 0.4935 0.4260 -0.9811 0.3674 0.3596 -0.9753 0.3402 (1.29) (2.15)** (2.09)** (1.93)* (1.45) (1.78)* (1.75)* (1.44) (1.69) Pseudo RSquared 0.0993 0.1004 0.1004 0.1002 0.1002 0.1003 0.1005 0.1012 0.1009 Number of Obs. 1627 1627 1627 1627 1627 1627 1627 1627 1627 State FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level ** denoted significance at the 0.025 level

PAGE 99

88 A) Percent of teachers receiving poor administrative support B) Percent of teachers receiving bonus pay C) Percent of teachers reporti ng overall job dissatisfaction Figure 3-1. Comparisons of Teacher Charact eristics Along the Distribution of PDA.

PAGE 100

89 D) Percent of teachers with a master's degree E) Percent of free-lunch eligible relative to the district average F) Percent of teachers who belong to a union Figure 3-1. Continued.

PAGE 101

90 G) Percent of teachers who are employed part-time H) Percent of teachers who are married I) Average teacher age Figure 3-1. Continued.

PAGE 102

91 J) Average years of full-time experience K) Percent of teachers who are white L) Percent of teachers who are male Figure 3-1. Continued.

PAGE 103

92 M) Percent of teachers who reporte d being threatened by a student Figure 3-1. Continued.

PAGE 104

93 Table 3-5: Within State Probit Regr ession of Interdistrict Movement Regression of PDA Hours with H* Chosen at Decile Intervals Along the Distribution of PDA Hours H* = 9 H* = 17 H* = 22.5 H* = 29.5 H* = 40.5 H* = 51.5 H* = 68 H* = 87.5 H* = 122.5 Variables (10%) (20%) (30%) (40%) (50%) (60%) (70%) (80%) (90%) PDA Hours 0 H* -0.0035 -0.0053 -0.0059 -0.0063 -0.0045 -0.0033 -0.0026 -0.0020 -0.0012 (0.16) (0.54) (0.86) (1.29) (1.34) (1.31) (1.36) (1.27) (1.03) PDA Hours > H* -0.0002 -0.0000 0.00016 0.00051 0.00079 0.00100 0.00151 0.00211 0.00348 (0.24) (0.05) (0.16) (0.51) (0.70) (0.81) (1.03) (1.17) (1.25) % Free-Lunch Eligible -0.1968 -0.1979 -0.1992 -0.2014 -0.2036 -0.2032 -0.2021 -0.2028 -0.2019 (1.11) (1.12) (1.12) (1.14) (1.15) (1.15) (1.14) (1.14) (1.13) Job Dissatifaction 0.76287 0.76015 0.76046 0.76279 0.76613 0.76885 0.77090 0.77125 0.76829 (3.67)** (3.65)** (3.65)** (3.65)** (3.65)** (3.66)** (3.67)** (3.67)** (3.68)** Full-Time Experience -0.0331 -0.0331 -0.0330 -0.0328 -0.0327 -0.0326 -0.0326 -0.0326 -0.0326 (4.51)** (4.51)** (4.50)** (4.48)** (4.44)** (4.42)** (4.42)** (4.40)** (4.41)** Part-time employment 0.00475 0.00152 -0.0011 -0.0045 -0.0010 0.00166 0.00252 0.00442 0.00656 (0.02) (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) (0.02) (0.03) Poor Administration 0.03717 0.03778 0.03775 0.03936 0.03753 0.03605 0.03455 0.03386 0.03687 (0.24) (0.24) (0.24) (0.25) (0.24) (0.23) (0.22) (0.21) (0.23) Union Member -0.0888 -0.0900 -0.0909 -0.0910 -0.0904 -0.0887 -0.0877 -0.0874 -0.0863 (0.91) (0.92) (0.93) (0.93) (0.92) (0.91) (0.90) (0.89) (0.88) Student Threat 0.12178 0.12207 0.12313 0.12641 0.12717 0.12681 0.12620 0.12655 0.12332 (0.96) (0.97) (0.97) (1.00) (1.00) (1.00) (0.99) (1.00) (0.97) Bonus Pay -0.2414 -0.2381 -0.2352 -0.2327 -0.2327 -0.2332 -0.2324 -0.2331 -0.2390 (1.79)* (1.77)* (1.74)* (1.72)* (1.72)* (1.73)* (1.72)* (1.73)* (1.78)* Age -0.0157 -0.0156 -0.0156 -0.0155 -0.0156 -0.0157 -0.0157 -0.0158 -0.0159 (3.03)** (3.02)** (3.01)** (3.00)** (3.02)** (3.03)** (3.03)** (3.04)** (3.06)** Dummy for Missing -0.3381 -0.3357 -0.3342 -0.3336 -0.3331 -0.3369 -0.3376 -0.3342 -0.3324 % Free Lunch Eligible (1.80)* (1.79)* (1.78)* (1.78)* (1.78)* (1.79)* (1.79)* (1.78)* (1.77)* Marital Status -0.0054 -0.0049 -0.0043 -0.0019 -0.0013 -0.0014 -0.0006 0.00009 -0.0000 (0.06) (0.06) (0.05) (0.02) (0.02) (0.02) (0.01) (0.00) (0.00) Constant -0.0104 0.03170 0.32786 0.35249 0.33986 0.32819 0.32307 0.31912 0.28890 (0.02) (0.05) (0.69) (0.75) (0.72) (0.70) (0.68) (0.67) (0.6) Pseudo RSquared 0.1686 0.1689 0.1694 0.1695 0.1695 0.1696 0.1695 0.1694 0.1689 Number of Obs. 1635 1635 1635 1635 1635 1635 1635 1635 1635 State FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level ** denoted significance at the 0.025 level

PAGE 105

94 Table 3-6: Out Of State Probit Regr ession of Interdistrict Movement Regression of PDA Hours with H* Chosen at Decile Intervals Along the Distribution of PDA Hours H* = 9 H* = 17 H* = 22.5 H* = 29.5 H* = 40.5 H* = 51.5 H* = 68 H* = 87.5 H* = 122.5 Variables (10%) (20%) (30%) (40%) (50%) (60%) (70%) (80%) (90%) PDA Hours 0 H* 0.02337 0.01003 -0.0003 -0.0014 0.00059 0.00215 0.00216 0.00292 0.00281 (0.65) (0.62) (0.03) (0.18) (0.10) (0.48) (0.63) (1.03) (1.35) PDA Hours > H* 0.00101 0.00089 0.00142 0.00168 0.00148 0.00088 0.00053 -0.0014 -0.0079 (0.65) (0.54) (0.85) (0.97) (0.78) (0.42) (0.21) (0.44) (1.27) % Free-Lunch Eligible -0.6608 -0.6655 -0.6616 -0.6636 -0.6633 -0.6593 -0.6593 -0.6549 -0.6604 (2.51)** (2.54)** (2.52)** (2.52)** (2.51)** (2.50)** (2.51)** (2.50)** (2.53)** Job Dissatifaction 0.41903 0.41354 0.41948 0.42437 0.41961 0.41483 0.41369 0.40598 0.40404 (1.05) (1.03) (1.05) (1.06) (1.05) (1.03) (1.03) (1.01) (1.00) Full-Time Experience -0.0440 -0.0440 -0.0434 -0.0433 -0.0435 -0.0437 -0.0438 -0.0440 -0.0443 (2.79)** (2.77)** (2.75)** (2.74)** (2.75)** (2.76)** (2.77)** (2.81)** (2.82)** Part-time employment -0.8839 -0.8876 -0.8871 -0.8903 -0.8874 -0.8842 -0.8842 -0.8852 -0.8912 (1.63) (1.64) (1.63) (1.64) (1.64) (1.63) (1.63) (1.62) (1.63) Poor Administration 0.27761 0.27740 0.26798 0.26595 0.26815 0.27250 0.27307 0.27725 0.27726 (1.05) (1.05) (1.01) (1.00) (1.01) (1.03) (1.03) (1.05) (1.05) Union Member -0.0159 -0.0188 -0.0172 -0.0164 -0.0170 -0.0186 -0.0191 -0.0213 -0.0249 (0.11) (0.13) (0.12) (0.11) (0.11) (0.12) (0.13) (0.14) (0.16) Student Threat 0.05111 0.05212 0.05034 0.05093 0.05066 0.05101 0.05180 0.05261 0.06140 (0.23) (0.24) (0.23) (0.23) (0.23) (0.23) (0.24) (0.24) (0.28) Bonus Pay 0.20266 0.20067 0.20190 0.20185 0.20131 0.20238 0.20180 0.20068 0.21043 (1.09) (1.08) (1.09) (1.09) (1.09) (1.10) (1.09) (1.08) (1.14) Age -0.0394 -0.0396 -0.0394 -0.0394 -0.0394 -0.0394 -0.0394 -0.0392 -0.0388 (3.67)** (3.69)** (3.67)** (3.67)** (3.67)** (3.66)** (3.65)** (3.64)** (3.59)** Dummy for Missing 0.03832 0.04128 0.05240 0.05589 0.05176 0.04787 0.04829 0.04654 0.03925 % Free Lunch Eligible (0.11) (0.12) (0.15) (0.16) (0.15) (0.14) (0.14) (0.14) (0.11) Marital Status 0.06546 0.06408 0.06081 0.06151 0.06151 0.06037 0.05910 0.05554 0.04640 (0.46) (0.45) (0.43) (0.44) (0.44) (0.43) (0.42) (0.39) (0.33) Constant 0.76726 0.07371 0.96390 0.98918 0.95216 0.91634 0.34887 0.87351 0.86222 (1.02) (0.10) (1.36) (1.42) (1.39) (1.33) (0.41) (1.28) (1.27) Pseudo RSquared 0.2489 0.2488 0.2484 0.2485 0.2484 0.2484 0.2485 0.2492 0.2508 Number of Obs. 1108 1108 1108 1108 1108 1108 1108 1108 1108 State FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level ** denoted significance at the 0.025 level

PAGE 106

95 Table 3-7: Probit Regression of Intradistric t Movement with District Fixed Effects Regression of PDA Hours with H* Chosen at Decile Intervals Along the Distribution of PDA Hours H* = 9 H* = 17 H* = 22.5 H* = 29.5 H* = 40.5 H* = 51.5 H* = 68 H* = 87.5 H* = 122.5 Variables (10%) (20%) (30%) (40%) (50%) (60%) (70%) (80%) (90%) PDA Hours 0 H* -0.2037 -0.0707 -0.0443 -0.0282 -0.0147 -0.0099 -0.0065 -0.0047 -0.0035 (2.12)** (2.20)** (2.03)** (1.82)* (1.45) (1.31) (1.22) (1.25) (1.26) PDA Hours > H* 0.00242 0.00303 0.00330 0.00357 0.00371 0.00416 0.00509 0.00678 0.01231 (1.13) (1.38) (1.49) (1.52) (1.44) (1.48) (1.59) (1.79)* (2.34)** % Free-Lunch Eligible 0.57160 0.50149 0.46890 0.45768 0.44608 0.45569 0.45798 0.44090 0.45394 (1.10) (0.95) (0.89) (0.87) (0.85) (0.87) (0.88) (0.85) (0.87) Job Dissatifaction 1.24178 1.22087 1.22145 1.21289 1.21081 1.20811 1.20478 1.17851 1.13832 (2.09)** (2.02)** (1.98)** (1.96)** (1.96)** (1.96)** (1.96)** (1.93)* (1.88)* Full-Time Experience -0.0119 -0.0098 -0.0099 -0.0094 -0.0086 -0.0089 -0.0093 -0.0099 -0.0098 (0.68) (0.55) (0.56) (0.53) (0.48) (0.51) (0.53) (0.56) (0.56) Part-time employment 0.51540 0.69751 0.72376 0.73810 0.73554 0.73359 0.72474 0.71323 0.71233 (1.09) (1.51) (1.59) (1.65) (1.67) (1.67) (1.65) (1.62) (1.62) Poor Administration 0.77191 0.73451 0.73095 0.72802 0.73031 0.72283 0.71226 0.70846 0.67968 (1.79)* (1.68) (1.66) (1.65) (1.66) (1.64) (1.61) (1.60) (1.55) Union Member 0.01757 0.08987 0.09814 0.09679 0.09406 0.09087 0.08021 0.08181 0.09168 (0.06) (0.31) (0.33) (0.33) (0.32) (0.31) (0.27) (0.28) (0.32) Student Threat 0.41600 0.42832 0.42130 0.41446 0.41373 0.40975 0.40086 0.39627 0.39301 (1.30) (1.32) (1.29) (1.29) (1.30) (1.30) (1.27) (1.26) (1.26) Bonus Pay -0.4831 -0.4094 -0.3959 -0.3940 -0.3839 -0.3700 -0.3651 -0.3550 -0.3483 (1.26) (1.08) (1.03) (1.02) (0.99) (0.96) (0.95) (0.93) (0.90) Age -0.0154 -0.0161 -0.0156 -0.0154 -0.0155 -0.0154 -0.0153 -0.0152 -0.0156 (0.96) (1.01) (0.98) (0.97) (0.98) (0.98) (0.98) (0.98) (1.00) Dummy for Missing -0.0759 -0.0769 -0.0674 -0.0824 -0.1056 -0.1187 -0.1165 -0.0904 -0.0660 % Free Lunch Eligible (0.20) (0.19) (0.17) (0.21) (0.27) (0.30) (0.30) (0.23) (0.17) Marital Status -0.0479 -0.0328 -0.0319 -0.0328 -0.0396 -0.0419 -0.0485 -0.0521 -0.0487 (0.21) (0.15) (0.14) (0.15) (0.18) (0.19) (0.22) (0.23) (0.22) Constant 2.09461 1.55653 1.29478 1.16611 1.02917 0.98271 0.96313 0.99048 1.00050 (2.62)** (2.25)** (2.00)** (1.82)* (1.62) (1.57) (1.54) (1.57) (1.62) Pseudo RSquared 0.1760 0.1710 0.1693 0.1675 0.1640 0.1632 0.1631 0.1646 0.1683 Number of Obs. 398 398 398 398 398 398 398 398 398 State FE Y Y Y Y Y Y Y Y Y District FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level ** denoted significance at the 0.025 level

PAGE 107

96 APPENDIX A SUMMARY STATISTICS AND SUPPLEMENTARY REGRESSION ANALYSIS Table A-1: Strike Summary Statistics Total Number of Strike and Strike Treatments in the 22 Year Sample: Total Strikes: 102 (64)** Total Strike Treatments: 2,351 Total Strikes and Strike Treatments in Each Community: Urban: Total Strikes: 74 (43)** Total Strike Treatments: 1,858 (1616)** Suburban Total Strikes: 31 (17)** Total Strike Treatments: 97 (67)** Rural Total Strikes: 56 (32)** Total Strike Treatments: 398 (275)** Longest Strike Length: 25 Days Mean Strike Length: 6.7 Days (Median is 5 Days) Distribution of Strike Days: Total Strike Treatments by Strike Year: 1 Day 321 1980 106 2 Days 211 1981 24 3 Days 202 1982 0 4 Days 197 1983 74 5 Days 188 1984 2 6 Days 177 1985 567 7 Days 179 1986 27 8 Days 170 1987 87 9 Days 58 1988 0 10 Days 49 1989 0 11 Days 45 1990 222 12 Days 44 1991 1016 13 Days 44 1992 9 14 Days 39 1993 4 15 Days 39 1994 90 16 Days 39 1995 75 17 Days 38 1996 0 18 Days 38 1997 0 19 Days 34 1998 18 20 Days 34 1999 1 21 Days 30 2000 0 22 Days 3 2001 11 23 Days 3 24 Days 3 25 Days 3 **The numbers in parenthesis represent the value in Type IV specifications

PAGE 108

97 Table A-2: Regressions of Full Data Set excluding Zip C odes with less than 500 students Variables ( ) ( ) ( ) ( V) (V) Strike Median Income Welfare Urban Poor Parental Educ. Juvenile Maleness Student Employment Alpha 0.187** 0.177** 0.200** 0.246** 0.212** (3.88)** (3.64)** (3.94)** (4.77)** (4.27)** -0.00003** -0.00003** -0.00003** -0.00003** (103.76)** (85.14)** (61.69)** (54.36)** 1.120** 1.149** 1.250** 1.572** (18.28)** (14.20)** (11.22)** (11.09)** 0.110** 0.089** 0.071** 0.046** (22.30)** (14.42)** (8.38)** (5.23)** 0.503** 0.523** 0.570** 0.802** (14.02)** (11.83)** (9.34)** (9.39)** 0.266** 0.195** 0.0829 0.265** (5.66)** (3.33)** (1.01) (2.66)** 0.224** 0.228** 0.225** 0.192** (11.67)** (9.61)** (6.88)** (5.07)** 0.922** 0.941** 0.927** 0.889** 0.705** (153.67)** (117.63)** (92.70)** (85.48)** (76.58)** Number of obs. Time Fixed Effects Zip Fixed Effects Pseudo R-Squared 879,441 582,543 301,056 249,745 249,745 Y Y Y Y Y N N N N Y 0.0397 0.0396 0.040 0.0308 0.0556

PAGE 109

APPENDIX B COMPLETE RESULTS OF SPECIFIED REGRESSIONS

PAGE 110

99 Table B-1: Reported on Table 3-4 Regression of PDA Hours with H* chosen at a cutoff on the Distri bution of PDA Hours Variables 10% 20% 30% 40% 50% 60% 70% 80% 90% PDA Hours 0 H* -0.0262 -0.0170 -0.0117 -0.0074 -0.0047 -0.0033 -0.0022 -0.0018 -0.0006 (1.12) (1.68) (1.63) (1.43) (1.34) (1.27) (1.12) (1.15) (0.55) PDA Hours > H* 0.0013 0.0017 0.0018 0.0020 0.0022 0.0026 0.0031 0.0043 0.0061 (1.52) (1.85)* (1.94)* (1.94)* (2.00)** (2.08)** (2.16)** (2.45)** (2.51)** % Free-Lunch Eligible 0.3041 0.3066 0.3077 0.3074 0.3067 0.3086 0.3117 0.3081 0.3010 (1.86)* (1.87)* (1.88)* (1.88)* (1.88)* (1.89)* (1.91)* (1.88)* (1.83)* Job Dissatifaction 0.4214 0.4254 0.4282 0.4285 0.4306 0.4338 0.4377 0.4385 0.4288 (1.86)* (1.87)* (1.88)* (1.88)* (1.89)* (1.90)* (1.92)* (1.93)* (1.89)* Full-Time Experience -0.0141 -0.0136 -0.0135 -0.0134 -0.0132 -0.0130 -0.0130 -0.0130 -0.0134 (2.29)** (2.22)** (2.19)** (2.17)** (2.13)** (2.10)** (2.10)** (2.10)** (2.18)** Part-time employment 0.5383 0.5362 0.5351 0.5373 0.5417 0.5458 0.5479 0.5477 0.5381 (3.37)** (3.36)** (3.36)** (3.38)** (3.41)** (3.44)** (3.45)** (3.46)** (3.41)** Poor Administration 0.3575 0.3616 0.3636 0.3641 0.3638 0.3621 0.3595 0.3596 0.3593 (2.57)** (2.59)** (2.61)** (2.61)** (2.61)** (2.60)** (2.58)** (2.58)** (2.59)** Union Member 0.2344 0.2406 0.2419 0.2411 0.2415 0.2416 0.2399 0.2387 0.2373 (2.32)** (2.38)** (2.40)** (2.39)** (2.39)** (2.39)** (2.37)** (2.37)** (2.36)** Race (White) -0.0906 -0.0924 -0.0931 -0.0950 -0. 0970 -0.0987 -0.10 00 -0.1015 -0.0953 (0.76) (0.77) (0.78) (0.80) (0.81) (0.83) (0.84) (0.85) (0.80) Marital Status 0.0232 0.0224 0.0232 0.0245 0.0256 0.0265 0.0266 0.0277 0.0264 (0.28) (0.27) (0.28) (0.30) (0.31) (0.32) (0.33) (0.34) (0.32) Dummy for Missing 0.1789 0.1779 0.1798 0.1796 0.1790 0.1777 0.1772 0.1807 0.1776 % Free Lunch Eligible (1.22) (1.21) (1.22) (1.22) (1.22) (1.21) (1.21) (1.23) (1.21) Bonus Pay -0.1674 -0.1639 -0.1639 -0.1657 -0.1658 -0.1650 -0.1635 -0.1633 -0.1658 (1.41) (1.38) (1.38) (1.39) (1.40) (1.39) (1.38) (1.38) (1.40) Student Threat 0.1011 0.1016 0.1022 0.1025 0.1016 0.1000 0.0984 0.0987 0.0991 (0.87) (0.87) (0.87) (0.88) (0.87) (0.86) (0.85) (0.85) (0.85) Age -0.0053 -0.0054 -0.0054 -0.0055 -0.0055 -0.0056 -0.0057 -0.0058 -0.0056 (1.02) (1.05) (1.06) (1.07) (1.08) (1.09) (1.10) (1.12) (1.09) Gender (Male) -0.1157 -0.1180 -0.1186 -0.1188 -0.1188 -0.1188 -0.1188 -0.1176 -0.1132 (1.35) (1.37) (1.38) (1.38) (1.38) (1.38) (1.38) (1.37) (1.32) Master's Degree 0.1211 0.1211 0.1212 0.1204 0.1202 0.1210 0.1215 0.1210 0.1180 (1.43) (1.43) (1.43) (1.42) (1.42) (1.43) (1.43) (1.43) (1.39) Constant -0.9070 0.5313 0.4935 0.4260 -0.9811 0.3674 0.3596 -0.9753 0.3402 (1.29) (2.15)** (2.09)** (1.93)* (1.45) (1.78)* (1.75)* (1.44) (1.69) R-Squared 0.0993 0.1004 0.1004 0.1002 0.1002 0.1003 0.1005 0.1012 0.1009 Number of Obs. 1627 1627 1627 1627 1627 1627 1627 1627 1627 State FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level, ** denoted significance at the 0.025 level

PAGE 111

100 Table B-2: Reported on Table 3-5 Regression of PDA Hours with H* chosen at a cutoff on the Distri bution of PDA Hours Variables 10% 20% 30% 40% 50% 60% 70% 80% 90% PDA Hours 0 H* -0.0035 -0.0053 -0.0059 -0.0063 -0. 0045 -0.0033 -0.0026 -0.0020 -0.0012 (0.16) (0.54) (0.86) (1.29) (1. 34) (1.31) (1.36) (1.27) (1.03) PDA Hours > H* -0.0002 -0.0000 0.00016 0.00051 0.00079 0.00100 0.00151 0.00211 0.00348 (0.24) (0.05) (0.16) (0.51) (0. 70) (0.81) (1.03) (1.17) (1.25) % Free-Lunch Eligible -0.1968 -0.1979 -0.1992 -0.2014 -0. 2036 -0.2032 -0.2021 -0.2028 -0.2019 (1.11) (1.12) (1.12) (1.14) (1. 15) (1.15) (1.14) (1.14) (1.13) Job Dissatifaction 0.76287 0.76015 0.76046 0.76279 0. 76613 0.76885 0.77090 0.77125 0.76829 (3.67)** (3.65)** (3.65)** (3.65)** (3. 65)** (3.66)** (3.67)** (3.67)** (3.68)** Full-Time Experience -0.0331 -0.0331 -0.0330 -0.0328 -0. 0327 -0.0326 -0.0326 -0.0326 -0.0326 (4.51)** (4.51)** (4.50)** (4.48)** (4. 44)** (4.42)** (4.42)** (4.40)** (4.41)** Part-time employment 0.00475 0.00152 -0.0011 -0.0045 -0 .0010 0.00166 0.00252 0.00442 0.00656 (0.02) (0.01) (0.01) (0.02) (0. 01) (0.01) (0.01) (0.02) (0.03) Poor Administration 0.03717 0.03778 0.03775 0.03936 0. 03753 0.03605 0.03455 0.03386 0.03687 (0.24) (0.24) (0.24) (0.25) (0. 24) (0.23) (0.22) (0.21) (0.23) Union Member -0.0888 -0.0900 -0.0909 -0 .0910 -0.0904 -0.0887 -0.0877 -0.0874 -0.0863 (0.91) (0.92) (0.93) (0.93) (0. 92) (0.91) (0.90) (0.89) (0.88) Race (White) -0.1287 -0.1324 -0.1355 -0 .1382 -0.1378 -0.1367 -0.1374 -0.1366 -0.1335 (1.00) (1.03) (1.05) (1.07) (1. 07) (1.06) (1.06) (1.06) (1.03) Marital Status -0.0054 -0.0049 -0.0043 -0.0019 -0.0013 -0.0014 -0.0006 0.00009 -0.0000 (0.06) (0.06) (0.05) (0.02) (0. 02) (0.02) (0.01) (0.00) (0.00) Dummy for Missing -0.3381 -0.3357 -0.3342 -0.3336 -0. 3331 -0.3369 -0.3376 -0.3342 -0.3324 % Free Lunch Eligible (1.80)* (1.79)* (1.78)* (1.78)* (1. 78)* (1.79)* (1.79)* (1.78)* (1.77)* Bonus Pay -0.2414 -0.2381 -0.2352 -0. 2327 -0.2327 -0.2332 -0.2324 -0.2331 -0.2390 (1.79)* (1.77)* (1.74)* (1.72)* (1. 72)* (1.73)* (1.72)* (1.73)* (1.78)* Student Threat 0.12178 0.12207 0. 12313 0.12641 0.12717 0.12681 0.12620 0.12655 0.12332 (0.96) (0.97) (0.97) (1.00) (1. 00) (1.00) (0.99) (1.00) (0.97) Age -0.0157 -0.0156 -0.0156 -0.0155 -0 .0156 -0.0157 -0.0157 -0.0158 -0.0159 (3.03)** (3.02)** (3.01)** (3.00)** (3. 02)** (3.03)** (3.03)** (3.04)** (3.06)** Gender (Male) 0.0217 0.0199 0.0181 0.0167 0.0176 0.0194 0.0212 0.0227 0.0245 (0.26) (0.23) (0.21) (0.20) (0. 21) (0.23) (0.25) (0.27) (0.29) Master's Degree -0.0153 -0.0162 -0.0173 -0 .0156 -0.0156 -0.0129 -0.0115 -0.0126 -0.0139 (0.17) (0.18) (0.19) (0.17) (0. 17) (0.14) (0.13) (0.14) (0.15) Constant -0.0104 0.03170 0.32786 0.35249 0.33986 0.32819 0.32307 0.31912 0.28890 (0.02) (0.05) (0.69) (0.75) (0 .72) (0.70) (0.68) (0.67) (0.6) R-squared 0.1686 0.1689 0.1694 0.1695 0.1695 0.1696 0.1695 0.1694 0.1689 Number of Obs. 1635 1635 1635 1635 1635 1635 1635 1635 1635 State FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level ** denoted significance at the 0.025 level

PAGE 112

101 Table B-3: Reported on Table 3-6 Regression of PDA Hours with H* chosen at a cutoff on the Distri bution of PDA Hours Variables 10% 20% 30% 40% 50% 60% 70% 80% 90% PDA Hours 0 H* 0.02337 0.01003 -0.0003 -0.0014 0.00059 0.00215 0.00216 0.00292 0.00281 (0.65) (0.62) (0.03) (0.18) (0. 10) (0.48) (0.63) (1.03) (1.35) PDA Hours > H* 0.00101 0.00089 0.00142 0.00168 0.00148 0.00088 0.00053 -0.0014 -0.0079 (0.65) (0.54) (0.85) (0.97) (0. 78) (0.42) (0.21) (0.44) (1.27) % Free-Lunch Eligible -0.6608 -0.6655 -0.6616 -0.6636 -0. 6633 -0.6593 -0.6593 -0.6549 -0.6604 (2.51)** (2.54)** (2.52)** (2.52)** (2. 51)** (2.50)** (2.51)** (2.50)** (2.53)** Job Dissatifaction 0.41903 0.41354 0.41948 0.42437 0. 41961 0.41483 0.41369 0.40598 0.40404 (1.05) (1.03) (1.05) (1.06) (1. 05) (1.03) (1.03) (1.01) (1.00) Full-Time Experience -0.0440 -0.0440 -0.0434 -0.0433 -0. 0435 -0.0437 -0.0438 -0.0440 -0.0443 (2.79)** (2.77)** (2.75)** (2.74)** (2. 75)** (2.76)** (2.77)** (2.81)** (2.82)** Part-time employment -0.8839 -0.8876 -0.8871 -0.8903 -0. 8874 -0.8842 -0.8842 -0.8852 -0.8912 (1.63) (1.64) (1.63) (1.64) (1. 64) (1.63) (1.63) (1.62) (1.63) Poor Administration 0.27761 0.27740 0.26798 0.26595 0. 26815 0.27250 0.27307 0.27725 0.27726 (1.05) (1.05) (1.01) (1.00) (1. 01) (1.03) (1.03) (1.05) (1.05) Union Member -0.0159 -0.0188 -0.0172 -0 .0164 -0.0170 -0.0186 -0.0191 -0.0213 -0.0249 (0.11) (0.13) (0.12) (0.11) (0. 11) (0.12) (0.13) (0.14) (0.16) Race (White) 0.23557 0.2448 0.24363 0.24274 0.24431 0.24747 0.24979 0.25548 0.24977 (0.89) (0.92) (0.92) (0.91) (0. 92) (0.93) (0.94) (0.95) (0.93) Marital Status 0.06546 0.06408 0. 06081 0.06151 0.06151 0.06037 0.05910 0.05554 0.04640 (0.46) (0.45) (0.43) (0.44) (0. 44) (0.43) (0.42) (0.39) (0.33) Dummy for Missing 0.03832 0.04128 0.05240 0.05589 0. 05176 0.04787 0.04829 0.04654 0.03925 % Free Lunch Eligible (0.11) (0.12) (0.15) (0.16) (0. 15) (0.14) (0.14) (0.14) (0.11) Bonus Pay 0.20266 0.20067 0.20190 0. 20185 0.20131 0.20238 0.20180 0.20068 0.21043 (1.09) (1.08) (1.09) (1.09) (1. 09) (1.10) (1.09) (1.08) (1.14) Student Threat 0.05111 0.05212 0. 05034 0.05093 0.05066 0.05101 0.05180 0.05261 0.06140 (0.23) (0.24) (0.23) (0.23) (0. 23) (0.23) (0.24) (0.24) (0.28) Age -0.0394 -0.0396 -0.0394 -0.0394 -0 .0394 -0.0394 -0.0394 -0.0392 -0.0388 (3.67)** (3.69)** (3.67)** (3.67)** (3. 67)** (3.66)** (3.65)** (3.64)** (3.59)** Gender (Male) -0.1214 -0.1187 -0.1257 -0 .1286 -0.1267 -0.1219 -0.1209 -0.1176 -0.1264 (0.78) (0.76) (0.80) (0.82) (0. 81) (0.78) (0.77) (0.75) (0.81) Master's Degree 0.53769 0.53978 0. 53363 0.53165 0.5316 0.53155 0.53111 0.53137 0.54125 (3.21) (3.23) (3.19) (3.19) (3. 19) (3.19) (3.18) (3.18) (3.23) Constant 0.76726 0.07371 0.96390 0. 98918 0.95216 0.91634 0.34887 0.87351 0.86222 (1.02) (0.10) (1.36) (1.42) (1. 39) (1.33) (0.41) (1.28) (1.27) R-squared 0.2489 0.2488 0.2484 0.2485 0.2484 0.2484 0.2485 0.2492 0.2508 Number of Obs. 1108 1108 1108 1108 1108 1108 1108 1108 1108 State FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level, ** denoted significance at the 0.025 level

PAGE 113

102 Table B-4: Reported on Table 3-7 Regression of PDA Hours with H* chosen at a cutoff on the Distri bution of PDA Hours Variables 10% 20% 30% 40% 50% 60% 70% 80% 90% PDA Hours 0 H* -0.2037 -0.0707 -0.0443 -0.0282 -0. 0147 -0.0099 -0.0065 -0.0047 -0.0035 (2.12)** (2.20)** (2.03)** (1.82)* (1. 45) (1.31) (1.22) (1.25) (1.26) PDA Hours > H* 0.00242 0.00303 0.00330 0.00357 0. 00371 0.00416 0.00509 0.00678 0.01231 (1.13) (1.38) (1.49) (1.52) (1. 44) (1.48) (1.59) (1.79)* (2.34)** % Free-Lunch Eligible 0.57160 0.50149 0.46890 0.45768 0. 44608 0.45569 0.45798 0.44090 0.45394 (1.10) (0.95) (0.89) (0.87) (0. 85) (0.87) (0.88) (0.85) (0.87) Job Dissatifaction 1.24178 1.22087 1.22145 1.21289 1. 21081 1.20811 1.20478 1.17851 1.13832 (2.09)** (2.02)** (1.98)** (1.96)** (1. 96)** (1.96)** (1.96)** (1.93)* (1.88)* Full-Time Experience -0.0119 -0.0098 -0.0099 -0.0094 -0. 0086 -0.0089 -0.0093 -0.0099 -0.0098 (0.68) (0.55) (0.56) (0.53) (0. 48) (0.51) (0.53) (0.56) (0.56) Part-time employment 0.51540 0.69751 0.72376 0.73810 0. 73554 0.73359 0.72474 0.71323 0.71233 (1.09) (1.51) (1.59) (1.65) (1. 67) (1.67) (1.65) (1.62) (1.62) Poor Administration 0.77191 0.73451 0.73095 0.72802 0. 73031 0.72283 0.71226 0.70846 0.67968 (1.79)* (1.68) (1.66) (1.65) (1. 66) (1.64) (1.61) (1.60) (1.55) Union Member 0.01757 0.08987 0.09814 0.09679 0. 09406 0.09087 0.08021 0.08181 0.09168 (0.06) (0.31) (0.33) (0.33) (0. 32) (0.31) (0.27) (0.28) (0.32) Race (White) -0.0808 -0.083 -0.0838 -0 .0862 -0.0885 -0.0908 -0.0924 -0.0937 -0.0864 (0.65) (0.67) (0.68) (0.69) (0. 71) (0.73) (0.74) (0.75) (0.69) Marital Status -0.0479 -0.0328 -0.0319 -0 .0328 -0.0396 -0.0419 -0.0485 -0.0521 -0.0487 (0.21) (0.15) (0.14) (0.15) (0. 18) (0.19) (0.22) (0.23) (0.22) Dummy for Missing -0.0759 -0.0769 -0.0674 -0.0824 -0. 1056 -0.1187 -0.1165 -0.0904 -0.0660 % Free Lunch Eligible (0.20) (0.19) (0.17) (0.21) (0. 27) (0.30) (0.30) (0.23) (0.17) Bonus Pay -0.4831 -0.4094 -0.3959 -0. 3940 -0.3839 -0.3700 -0.3651 -0.3550 -0.3483 (1.26) (1.08) (1.03) (1.02) (0. 99) (0.96) (0.95) (0.93) (0.90) Student Threat 0.41600 0.42832 0. 42130 0.41446 0.41373 0.40975 0.40086 0.39627 0.39301 (1.30) (1.32) (1.29) (1.29) (1. 30) (1.30) (1.27) (1.26) (1.26) Age -0.0154 -0.0161 -0.0156 -0.0154 -0 .0155 -0.0154 -0.0153 -0.0152 -0.0156 (0.96) (1.01) (0.98) (0.97) (0. 98) (0.98) (0.98) (0.98) (1.00) Gender (Male) -0.107 -0.1103 -0.1111 -0 .1113 -0.1114 -0.1117 -0.1118 -0.1099 -0.1043 (1.30) (1.33) (1.34) (1.34) (1. 35) (1.35) (1.36) (1.33) (1.26) Master's Degree 0.1495 0.1497 0.15 0.149 0. 1485 0.1495 0.1502 0.1497 0.1463 (1.76)* (1.75)* (1.75)* (1.74)* (1. 73)* (1.74)* (1.75)* (1.74)* (1.71)* Constant 2.09461 1.55653 1.29478 1. 16611 1.02917 0.98271 0.96313 0.99048 1.00050 (2.62)** (2.25)** (2.00)** (1.82)* (1. 62) (1.57) (1.54) (1.57) (1.62) R-squared 0.1760 0.1710 0.1693 0.1675 0.1640 0.1632 0.1631 0.1646 0.1683 Number of Obs. 398 398 398 398 398 398 398 398 398 State FE Y Y Y Y Y Y Y Y Y Robust SE, clustered Y Y Y Y Y Y Y Y Y denoted significance at the 0.05 level, ** denoted significance at the 0.025 level

PAGE 114

103 LIST OF REFERENCES Allan, Emilie A., and Steffensmeier, Darrell J., “Youth, Underemployment, and Property Crime: Differential Effects of Job Availa bility and Job Quality on Juvenile and Young Adult Arrest Rates” American Sociological Review Vol. 54, No.1 (February 1989), pp.107-123 Angrist, Joshua D. and Guryan, Jonathan “D oes Teacher Testing Raise Teacher Quality? Evidence from State Certification Requirements” NBER Working Paper No 9545 March 2003. Ballou, Dale and Podgursky, Michael “Teach ers’ Attitudes Towards Merit Pay: Examining Conventional Wisdom” Industrial and Labor Relations Review Vol. 47, No. 1, (October 1993) p. 50-61 Becker, Gary S. and Tomes, Nigel, “Human Capital and the Rise and Fall of Families” Journal of Labor Economics Vol. 4, No. 3, Part 2: The Family and the Distribution of Economic Rewards. (July 1986), pp. S1-S39. Becker, Gary S. and Tomes, Nigel, “Child Endowments and the Quantity and Quality of Children” (in Part II: Labor Supply and the Family) The Journal of Political Economy Vol. 84, No. 4, Part 2: Essays in Labor Economics in Honor of H. Gregg Lewis. (August 1976), pp. S143-S162. Brewer, Dominic J. “Career Paths and Quits Decisions: Evidence From Teaching” Journal of Labor Economics Vol. 14, No. 2 (April 1996) p. 313-339 Cameron, Colin A., and Trivedi, Pravin K., “Econometric Models Based on Count Data: Comparisons and Applications of Some Estimators and Tests” Journal of Applied Econometrics Vol. 1, No. 1. (January 1986), pp. 29-53. Carson, R.T. and Grogger, Jefferey T. “Models for Truncated Counts” Journal of Applied Econometrics Vol. 6, No. 3. (July September 1991), pp. 225238. Corman, Hope, and Mocan, Naci H., “Carro ts, Sticks and Broken Windows”, National Bureau of Economic Research: Working Paper 9061 (July 2002). Cuellar, Alison Evans, Markowitz, Sara and Libby, Anne M., "The Relationships Between Mental Health and Substance Abuse Treatment and Juvenile Crime", National Bureau of Economic Research Working Paper No. W9952 (September 2003).

PAGE 115

104 Eberts, Randall, Hollenbeck, Kevin and Stone Joe “Teacher Performance Incentives and Student Outcomes” The Journal of Human Resources, Vol. 37, No. 4 (Autumn 2002), p. 913-927 Ehrlich, Isaac, “On the Useful ness of Controlling Individual s: An Economic Analysis of Rehabilitation, Incapacitation and Deterrence” The American Economic Review Vol. 71, No. 3. (June 1981), pp. 307-322. Ehrlich, Isaac and Gibbons, Joel C., “On the Measurement of the Deterrent Effect of Capital Punishment and the Theory of Deterrence” The Journal of Legal Studies Vol. 6, No. 1. (January 1977), pp. 35-50. Feld, Barry C., “Juvenile and Criminal Just ice Systems' Responses to Youth Violence” Crime and Justice Vol. 24, Youth Violence. (1998), pp. 189-261. Fleisher, Belton M., “The Effect of Unemployment on Juvenile Delinquency” The Journal of Political Economy Vol. 71, No. 6 (December 1963), pp.543-555 Freeman, Richard B., “Why Do So Many Y oung American Men Commit Crimes and What Might I Do About It?” Journal of Economic Perspectives Vol 10:1; pp.25-42 (Winter 1996). Freeman, Richard B., “Crime and The Empl oyment of Disadvantaged Youths” in Urban Labor Markets and Job Opportunity by George Peterson and WayneVroman (Washington, D.C.: Urban Press Institute, 1992). Glaeser, Edward L. and Sacerdote, Bruce, “Why is There More Crime in Cities?” The Journal of Political Economy Vol. 107, No. 6, Part 2: Symposium on the Economic Analysis of Social Behavior in Honor of Gary S. Becker. (December 1999), pp. S225-S258. Glaeser, Edward L. and Sacerdote, Bruce, “Crime and Social Interactions” The Quarterly Journal of Economics Vol. 111, No. 2. (May 1996), pp. 507-548. Greenberg, David and McCall, John “Teacher Mobility and Allocation” The Journal of Human Resources, Vol. 9, No. 4 (Autumn 1974), p. 480-502. Griliches, Zvi, Hall, Bronwyn H. and Haus man, Jerry, “Econometric Models for Count Data with an Application to the Patents-R & D Relationship” Econometrica Vol. 52, No. 4. (July 1984), pp. 909-938. Gritz, Mark R. and Theobald, Neil D. “The E ffects of School District Spending Priorities On Length of Stay in Teaching” The Journal of Human Resources, Vol. 31, No. 3 (Summer 1996), p. 477-512 Grogger, Jeffrey T., "Local Violence, E ducational Attainment, and Teacher Pay" National Bureau of Economic Research Working Paper No. 6003 (April 1997).

PAGE 116

105 Hanushek, Eric A., Kain, John F. and Rivkin, St even G. “Do Higher Salaries Buy Better Teachers?” NBER Working Paper No 7082, April 2002, JEL No. I2, J4 Hanushek, Eric A., Kain, John F. and Rivkin, Steven G. “Why Public Schools Lose Teachers” NBER Working Paper No 8599, November 2001, JEL No. I20, J45 Hanushek, Eric A., Kain, John F. and Rivkin, Steven G. “Teachers, Schools, and Academic Achievement” NBER Working Paper No 6691, August 1998, JEL No. I2, H4 Haveman, Robert and Wolfe, Barbara, “The Determinants of Children's Attainments: A Review of Methods and Findings” Journal of Economic Literature Vol. 33, No. 4. (December 1995), pp. 1829-1878. Hill, Anne M. and O’Neill, June, “Family Endowments and the Achievement of Young Children with Special Reference to the Underclass” The Journal of Human Resources Vol. 29, No. 4, Special Issue: Th e Family and Intergenerational Relations. (Autumn 1994), pp. 1064-1100. Hoxby, Caroline “Would School Choice Change The Teaching Profession?” Journal of Human Resources, Vol. 37, No. 4 (Autumn 2002), p. 846-891 Imazeki, Jennifer “Teacher Salaries and Teach er Attrition: How Much is Enough?” (May 2003) Imazeki, Jennifer “Teacher Attrition and Mobili ty in Urban Districts: Evidence From Wisconsin” In Fiscal Issues in Urban Schools; Research in Education: Fiscal Policy and Practice Volume 1, Jennifer King Rice and Christopher Roelke, eds. Jacob, Brian and Lefgren, Lars, “Are Idle Hands the Devil’s Workshop? Incapacitation, Concentration and Juvenile Crime” The American Economic Review Vol. 93, No. 5. (December 2003), pp.1560-1577. Lankford, Hamilton, Loeb, Susanna and Wyckoff, James “Teacher Sorting and the Plight of Urban Schools: A Descriptive Analysis” Educational Evaluation and Policy Analysis Vol. 24, No. 1 (Spring 2002) p. 37-62 Levitt, Steven, “Juvenile Crime and Punishment” The Journal of Political Economy Vol. 106, No. 6. (December 1998), pp. 1156-1185. Lochner, Lance, "Education, Work, and Cr ime: A Human Capital Approach" National Bureau of Economic Research Working Paper No. 10478 (May 2004). Lochner, Lance, "Individual Perceptions of the Criminal Justice System", National Bureau of Economic Research Working Paper No. 9474 (February 2003).

PAGE 117

106 Lochner, Lance and Moretti, Enrico, "The Effect of Education on Criminal Activity: Evidence from Prison Inmates, Arrests and Self-Reports" American Economic Review Vol. 94, No. 1. (March 2004). McDowall, David and Singer, Simon I., “C riminalizing Delinquency: The Deterrent Effects of the New York Juvenile Offender Law” Law & Society Review Vol. 22, No. 3. (1988), pp. 521-536. Mocan, Naci H., Rees, Daniel I., “Economic Conditions, Deterrence and Juvenile Crime: Evidence From Micro Data”, National Bureau of Economic Research: Working Paper No. 7405 (October 1999). Mocan, Naci H., Scafidi, Benjamin, and Tekin, Erdal, “Catholic Schools and Bad Behavior”, National Bureau of Ec onomic Research: Working Paper 9172 (September 2002). Murnane, Richard J. “Selection and Su rvival in the Teacher Labor Market” The Review of Economics and Statistics, Vol. 66, No. 3 (August 1984), p. 513-518 Murnane, Richard J. “Teacher Mobility Revisited” The Journal of Human Resources, Vol. 16, No. 1 (Winter 1981), p. 3-19 Murnane, Richard J. and Olsen, Randall J. “T he Effects of Salaries and Opportunity Costs on Length of Stay in Teaching: Evidence from North Carolina” The Journal of Human Resources, Vol. 25, No. 1 (Winter 1990), p. 106-124 Pfeiffer, Christian, “Juvenile Crime and Violence in Europe” Crime and Justice Vol. 23. (1998), pp. 255-328. Rees, Daniel “Grievance Procedur e Strength and Teacher Quits” Industrial and Labor Relations Review Vol. 45, No. 1, (October 1991), p. 31-43 Tatum, Becky L., “An Analysis of Factors Contributing to the Delinquency of the Black Youth” Journal of Black Studies Vol. 26, No. 3 (January 1996), pp. 356-368 U.S. Department of Education, National Center for Education Statistics (1997), America’s Teachers: Profile of a Profession, 1993-94 Washington, D.C.: U.S. Government Printing Office, NCES 97-460. Wilson, O.W., “How to Measure the Extent of Juvenile Delinquency” Journal of Criminal Law and Criminology (1931-1951) Vol. 41, No. 4. (November December 1950), pp. 435-438. Zabalza, A. “Internal Labour Mobi lity and the Teaching Profession” The Economic Journal Vol. 88, No. 350 (June 1978), p. 314-330

PAGE 118

107 BIOGRAPHICAL SKETCH Before beginning his doctoral studies, Jere my Luallen attended the University of Florida as an undergraduate where he receiv ed two bachelor degrees (one in economics and one in political science) in three years. As a graduate student Jeremy has been recognized numerous times for scholarly achie vement in his work. In 2004 he was the recipient of Edward Zabel Award, an award gi ven for excellence in dissertation research and publication potential. In addition he was also recognized for excellence in completed research, as well as potential for future res earch, in the field of Public Policy as a recipient of the Walter-Lanzillotti Award. Jeremy is a member of the American Education Finance Association, has been invited to present his research at several coll egial conferences, such as the annual meeting of the Southern Economic Association, a nd at the American Education Finance Association annual conference. In addition to his schooling Jeremy has worked as a consultant for the Naples Children and E ducation Foundation, and will begin working as a Senior Analyst for Abt A ssociates after graduation.


Permanent Link: http://ufdc.ufl.edu/UFE0011369/00001

Material Information

Title: Understanding education : three essays analyzing unintended outcomes of school policies
Physical Description: xi, 107 p.
Language: English
Creator: Luallen, Jeremy Clayton ( Dissertant )
Kenny, Lawrence W. ( Thesis advisor )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2005
Copyright Date: 2005

Subjects

Subjects / Keywords: Economics thesis, Ph. D.   ( local )
Dissertations, Academic -- UF -- Economics   ( local )

Notes

Abstract: The goal of my study is to examine specific school policies to determine if unintentional consequences result from these policies. Specifically, I focus on two main issues as they relate to student and teachers outcomes. I begin by looking at the effect of incapacitating juveniles in school as a force influencing juvenile crime. I exploit teacher strikes as a measure of unexpected student absence from school to measure the effect of school in preventing juvenile crime. My data set consists of information on every juvenile arrest made in Washington State over a 22-year period. I show that previous estimates of the effect of school incapacitation are systematically underestimated, that criminal activity increases as students continue to remain out of school. I also show that these increases in crime reflect an increase in overall crime, not a displacement. Lastly, I show that repeat juvenile offenders are more likely to have committed their first crime on a strike day, relative to a normal school day. Chapters 2 and 3 of the study focus on the role of teacher networks in influencing teacher mobility. Specifically, my study develops a model of teacher networks that describes how teachers assemble networks through professional development activities (PDAs) and how these networks provide an effective sorting mechanism for public school teachers. I empirically test the existence of teacher networks with 2 distinct datasets. The dataset in Chapter 2 is comprised of various reports covering all 67 Florida school districts. Besides examining how professional development affects teacher movement, I am able to exploit the macro nature of the data to compare district characteristics (such as differences in compensation levels and school district density) to examine how these factors also influence teacher mobility. The dataset in Chapter 3 uses survey data from the ''Schools and Staffing Survey'' and includes over 17,000 teachers. The high-powered nature of this dataset allows me to identify specific details, such as teacher salary incentives, individual network strength and union membership. Ultimately I conclude that teacher networks are an integral part of a teacher's transfer decision and have a sizable impact on intra-district teacher mobility.
Subject: binomial, crime, development, economics, education, incapacitation, inservice, intradistrict, juvenile, Luallen, mobility, negative, networks, professional, property, schools, strikes, transfer, violent, Washington
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 118 pages.
General Note: Includes vita.
Thesis: Thesis (Ph. D.)--University of Florida, 2005.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0011369:00001

Permanent Link: http://ufdc.ufl.edu/UFE0011369/00001

Material Information

Title: Understanding education : three essays analyzing unintended outcomes of school policies
Physical Description: xi, 107 p.
Language: English
Creator: Luallen, Jeremy Clayton ( Dissertant )
Kenny, Lawrence W. ( Thesis advisor )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2005
Copyright Date: 2005

Subjects

Subjects / Keywords: Economics thesis, Ph. D.   ( local )
Dissertations, Academic -- UF -- Economics   ( local )

Notes

Abstract: The goal of my study is to examine specific school policies to determine if unintentional consequences result from these policies. Specifically, I focus on two main issues as they relate to student and teachers outcomes. I begin by looking at the effect of incapacitating juveniles in school as a force influencing juvenile crime. I exploit teacher strikes as a measure of unexpected student absence from school to measure the effect of school in preventing juvenile crime. My data set consists of information on every juvenile arrest made in Washington State over a 22-year period. I show that previous estimates of the effect of school incapacitation are systematically underestimated, that criminal activity increases as students continue to remain out of school. I also show that these increases in crime reflect an increase in overall crime, not a displacement. Lastly, I show that repeat juvenile offenders are more likely to have committed their first crime on a strike day, relative to a normal school day. Chapters 2 and 3 of the study focus on the role of teacher networks in influencing teacher mobility. Specifically, my study develops a model of teacher networks that describes how teachers assemble networks through professional development activities (PDAs) and how these networks provide an effective sorting mechanism for public school teachers. I empirically test the existence of teacher networks with 2 distinct datasets. The dataset in Chapter 2 is comprised of various reports covering all 67 Florida school districts. Besides examining how professional development affects teacher movement, I am able to exploit the macro nature of the data to compare district characteristics (such as differences in compensation levels and school district density) to examine how these factors also influence teacher mobility. The dataset in Chapter 3 uses survey data from the ''Schools and Staffing Survey'' and includes over 17,000 teachers. The high-powered nature of this dataset allows me to identify specific details, such as teacher salary incentives, individual network strength and union membership. Ultimately I conclude that teacher networks are an integral part of a teacher's transfer decision and have a sizable impact on intra-district teacher mobility.
Subject: binomial, crime, development, economics, education, incapacitation, inservice, intradistrict, juvenile, Luallen, mobility, negative, networks, professional, property, schools, strikes, transfer, violent, Washington
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 118 pages.
General Note: Includes vita.
Thesis: Thesis (Ph. D.)--University of Florida, 2005.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0011369:00001


This item has the following downloads:


Full Text












UNDERSTANDING EDUCATION: THREE ESSAYS ANALYZING UNINTENDED
OUTCOMES OF SCHOOL POLICIES
















By

JEREMY CLAYTON LUALLEN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Jeremy Clayton Luallen















ACKNOWLEDGMENTS

Throughout the course of my graduate and dissertation work here at the University

of Florida, there have been many friends, relatives and colleagues whom I have relied on

for much professional and emotional support. It is for this reason that I wish to

acknowledge the contributions of all these important individuals, who were key to the

success of this dissertation. In addition to the academic and financial support these

individuals have provided, they have also made outstanding sacrifices to better my

growth, not only as a professional, but also personally.

First and foremost I would like to recognize my parents Dean and Nancy for their

amazing support of my efforts. Their guidance through the years has bettered me in ways

that I cannot even begin to measure. I would also like to thank my fiancee Jaclyn for

never-ending support of me and my work over the past several years. I would like to

specially acknowledge Dr. Lawrence Kenny for his wonderful friendship and guidance as

a mentor over the years. I would also like to thank my sister Jessica who has inspired me

in many ways. There are also many others I would like to acknowledge for their

assistance in making this dissertation a success. I would like to thank Steven Slutsky,

David Figlio, David Denslow, David Hedge, Randall Reback, Richard Romano, Sarah

Hamersma, David Brasington, David Sappington, Bill Bomberger, Chunrong Ai, Scott

Hankins, Mark Hoekstra and Jim Dewey for their many comments, incisive criticism, and

overall enthusiasm for my research.









I would like to thank Anne Stahl at the National Juvenile Court Data Archive for

her guidance and patience in helping me to acquire data for this project. Thanks also go

to Kathy Budge, Dennis Small, and many others at the Superintendent of Public

Instruction Office in Washington State for providing supplementary data that has been

crucial to the success of my research. I would also like to thank Martha Haynes and the

Staff of the Education Information and Accountability Services division of the Florida

Department of Education, who supplied various reports and data utilized in this

dissertation. I would also like to thank Kerry Gruber and the entire staff at the National

Center of Education Statistics that supplies and distributes the Schools and Staffing

Survey. I am also especially grateful to all of the teachers and administrators throughout

the State of Florida who shared their time, their thoughts and their experiences with me.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ......... .................................................................................... iii

LIST OF TABLES ............................... .............. ................. vii

LIST OF FIGURES ......... ......................... ...... ........ ............ ix

A B STR A C T ................................................. ..................................... .. x

CHAPTER

1 SCHOOL'S OUT...FOREVER: A STUDY OF JUVENILE CRIME, AT-
RISK YOUTHS AND TEACHER STRIKES.............. .... ................... 1

1.1 Introdu action ................... ......... .......... ................. ............... 1
1.2 The Identification Strategy ........................................ ....... ............... 3
1.2.1 Zip-Code M atching................. ......... ...... .................5
1.2.2 The Basic M odel ........................ .. ........................ .8
1 .3 D a ta ................................................... ................ ................ 9
1.3.1 M ethodology .. ..................... .. .......... ...... .............. .... ............ ..10
1.3.2 Defining the Param eters............... ......................... .................11
1.4 R egression A nalysis........................................................ ............... 13
1.4.1 Community Differences...................................... .................. 16
1.4.2 Differences in Offense Types ................................................18
1.4.3 Types of Offenders .............. .... ............................ ........... 19

2 THE ROLE OF TEACHER NETWORKING IN TEACHER TRANSFER
DECISIONS AND TEACHER MOBILITY ....................................................... 42

2 .1 Introdu action ................................................................................ 42
2.2 Discussion of Networking...................... .... .......................... 43
2.2.1 H ow Teachers N etw ork ........................................ .....................43
2.2.2 How Teachers Use Their Networks.............................................45
2.3 The Basic Model .......................................... ...................46
2.4 D ata and Em pirical Study ........................................ ...................... 47
2.4.1 D efining the Param eters............................................................... 48
2.4.2 E m pirical T testing ....................................... ......... ............... 52
2.5 R obustness C hecks.......................................................... ............... 55
2.5.1 R e-exam ining the D ata ...................................... ............... 55









2.5.2 U observed C orrelation ........... ...................... ............ ........57
2 .6 C o n c lu sio n ........................................................................................... 5 8

3 WITH A LITTLE HELP FROM MY FRIENDS: EVIDENCE OF TEACHER
NETWORKS USING MICRO DATA......................................................65

3.1 Introduction .................. .................................................. 65
3.2 The Identification Strategy .............. ...................... ............... ... 67
3.2.1 What Are Networks And How Are They Useful...........................67
3.2.2 How Do Teachers Build Networks? ...........................................69
3.2.3 M easuring Teacher Networks ...................................................70
3.2.4 The B asic M odel ........................................ ........................ 72
3 .3 D ata ................................................................................. 7 6
3 .4 E m pirical E v iden ce ................................................... ..............................77
3.5 R obustness C hecks.......................................................... ............... 80
3 .6 C o n c lu sio n ........................................................................................... 8 3

APPENDIX

A SUMMARY STATISTICS AND SUPPLEMENTARY REGRESSION
A N A L Y S IS ................................................... ................ 9 6

B COMPLETE RESULTS OF SPECIFIED REGRESSIONS ..............................98

L IST O F R E F E R E N C E S ...................................................................... ..................... 103

BIOGRAPHICAL SKETCH .............................................................. ............... 107
















LIST OF TABLES


Table pge

1-1 Negative Binomial Regression onto Total Crime with Full Data Set....................30

1-2 Urban Subsample Negative Binomial Regression onto Total Crime ..................31

1-3 Suburban Subsample Negative Binomial Regression onto Total Crime ..............32

1-4 Rural Subsample Negative Binomial Regression onto Total Crime ...................33

1-5 Effects of Strike Days by Crime Type and by Community Type..........................34

1-6 Effects of Strike Days by Crime Type and by Offender Type ...........................35

1-7 Temporal Displacement of Crime by Crime Type and by Offender Type............36

1-8 Temporal Displacement of Crime through Lagged Strike Variable....................37

1-9 Changes in Total Juvenile Crime, by Days Elapsed since the Start of a Strike.....38

1-10 Effects of Strike Days with Specific Years Dropped from the Sample.................40

1-11 Average Effect of Randomly Generated Strike Days on Total Crime...................41

1-12 Effects of Strike Days by Offender Type with the Last 4 Years of the Sample
D ro p p e d ................................. ....................................................... ............... 4 1

2-1 Ordinary Least Squares Regression onto teacher transfers using full sample
with and without robust standard errors...................................... ............... 60

2-2 Percent change teacher transfers per school for a one standard deviation
increase in each significant variable with full sample ....................................... 61

2-3 Ordinary Least Squares Regression onto teacher transfers using sample with
biasing outlier om itted ....................... .. ...................... .... ........ ...... ............62

2-4 Outlier excluded percent change teacher transfers per school for a one standard
deviation increase in each significant variable .............. ........................ ......... 63

2-5 Ordinary Least Squares Regression of other types of teacher movements with
outlier ex clu ded ............................................................................ ............... 64









3-1 Professional Development Activities Summary Statistics................................84

3-2 Summary Statistics for Various Teacher Groups........................................85

3-3 Probit Regression of Intra-District Movement ............................................... 86

3-4 Probit Regression of Intra-District Movement ............................................... 87

3-5 Within State Probit Regression of Interdistrict Movement............................... 93

3-6 Out Of State Probit Regression of Interdistrict Movement .............................. 94

3-7 Probit Regression of Intradistrict Movement with District Fixed Effects.............95

A -1 Strike Sum m ary Statistics ......................................................... .............. 96

A-2 Regressions of Full Data Set excluding Zip Codes with less than 500 students ...97

B R ep orted on T ab le 3-4 ................................................................ .....................99

B -2 R reported on Table 3-5 ................................................ ............................. 100

B -3 R reported on T able 3-6 ........................................... ........................................ 10 1

B -4 R reported on T able 3-7 ........................................... ........................................ 102
















LIST OF FIGURES


Figure


1-1 Average Percent of Total Arrests as Function of Days between Offense and
C ap tu re ...................................................................................3 9

3-1 Comparisons of Teacher Characteristics Along the Distribution of PDA ..............88


page















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

UNDERSTANDING EDUCATION: THREE ESSAYS ANALYZING UNINTENDED
OUTCOMES OF SCHOOL POLICIES

By

Jeremy Clayton Luallen

August 2005

Chair: Lawrence Kenny
Major Department: Economics

The goal of my study is to examine specific school policies to determine if

unintentional consequences result from these policies. Specifically, I focus on two main

issues as they relate to student and teachers outcomes. I begin by looking at the effect of

incapacitating juveniles in school as a force influencing juvenile crime. I exploit teacher

strikes as a measure of unexpected student absence from school to measure the effect of

school in preventing juvenile crime. My data set consists of information on every

juvenile arrest made in Washington State over a 22-year period. I show that previous

estimates of the effect of school incapacitation are systematically underestimated, that

criminal activity increases as students continue to remain out of school. I also show that

these increases in crime reflect an increase in overall crime, not a displacement. Lastly, I

show that repeat juvenile offenders are more likely to have committed their first crime on

a strike day, relative to a normal school day.









Chapters 2 and 3 of the study focus on the role of teacher networks in influencing

teacher mobility. Specifically, my study develops a model of teacher networks that

describes how teachers assemble networks through professional development activities

(PDAs) and how these networks provide an effective sorting mechanism for public

school teachers. I empirically test the existence of teacher networks with 2 distinct

datasets. The dataset in Chapter 2 is comprised of various reports covering all 67 Florida

school districts. Besides examining how professional development affects teacher

movement, I am able to exploit the macro nature of the data to compare district

characteristics (such as differences in compensation levels and school district density) to

examine how these factors also influence teacher mobility. The dataset in Chapter 3 uses

survey data from the "Schools and Staffing Survey" and includes over 17,000 teachers.

The high-powered nature of this dataset allows me to identify specific details, such as

teacher salary incentives, individual network strength and union membership. Ultimately

I conclude that teacher networks are an integral part of a teacher's transfer decision and

have a sizable impact on intra-district teacher mobility.














CHAPTER 1
SCHOOL'S OUT...FOREVER: A STUDY OF JUVENILE CRIME, AT-RISK
YOUTHS AND TEACHER STRIKES

1.1 Introduction

Although we have taken important strides in understanding the economics of crime,

there remains a great deal which we have yet to fully understand. The focus of this

chapter is juvenile crime in particular. Because juvenile crime can be especially hard to

study, due to general limitations on access to data, economists still have a lot to explore

on this front.

Over the past 20 years, economists and social scientists have attacked the problem

of crime in four distinct ways. Specifically they have analyzed the potential effects of

deterrence, retribution, rehabilitation and incapacitation as forces for reducing crime

(Ehrlich 1981).1 Deterrence, a widely explored topic, stresses the importance of

imposing penalties as an effective means of preventing crime, because the perceived

costs to criminals of criminal activity increases (Mocan and Rees 1999, Levitt 1998,

Freeman 1996, Ehrlich and Gibbons 1977, Lochner 2003). Retribution addresses the

actual punishment of criminals. It suggests that a criminal experiences an increased cost

to his crime when a punishment is imposed on him and/or other criminals (Levitt 1998).

Rehabilitation stresses the importance of reforming criminals, through treatment and

rehabilitative programs, in preventing future crime (Cuellar, Markowitz and Libby 2003).

Finally, incapacitation deals with the notion that the physical lock-up and detaining of


SI am not arguing that these four ideas are mutually exclusive ways of preventing or reducing crime.









criminals may be an effective means of preventing crime (Freeman 1996, Lochner 2004).

This paper will closely examine the role of incapacitation in preventing juvenile crime.

Specifically I deal with the effect of incapacitating juveniles in school on juvenile

crime. Brian Jacob and Lars Lefgren (2003) were the first to address this topic in their

paper, "Are Idle Hands the Devil's Workshop? Incapacitation, Concentration and

Juvenile Crime". They examine the effect of school attendance on juvenile crime by

using teacher in-service days, days where teaching professionals are required to attend

work when children are not required to be at school, as a source of variation in student

attendance. While their use of in-service days is very clever, I utilize teacher strikes as a

source of variation of student school attendance to gain additional insights into the

relationship between juvenile crime and school incapacitation.

One primary advantage to using teacher strikes over in-service days as a source of

variation in school attendance is that strikes often occur in blocks larger than one

individual day. This allows me to observe the effect of prolonged absence from school

on juvenile crime. If juvenile criminal habits do change as absence from school

increases, then a one day absence cannot reflect the average effect of school

incapacitation.

Another advantage of utilizing teacher strikes is that they are relatively

unpredictable. Since in-service days are planned at the beginning of the regular school

year, parents have adequate time to make arrangements for their children during these

off-days. In contrast, school closures caused by teacher strikes are often reported in local

newspapers only a couple of days before they are likely to take place, and even then,

there is no guarantee that these reported strikes will actually occur. Additionally, a









parent's ability to plan for their children's activities during strike days is complicated by

the fact that details revealed in newspapers and other information sources may prove to

be inaccurate or unjustified.2 The late-breaking and incomplete nature of information

leaves parents little time to plan their children's activities during the days of a teacher

strike. Thus teacher strike days are likely to result in more unsupervised students, and

more juvenile crime.

The main drawback to this use of strike as variation is that teacher strikes are often

very sparse, and mostly occur in a only a few school districts. In order to overcome this

difficulty, I rely on data from Washington State. Washington provides an ideal

environment for this type of study because it has an extensive strike history as well as a

detailed juvenile arrest data set that dates back as far as 1980.3

1.2 The Identification Strategy

As stated earlier, the objective of this analysis is to measure the impact of school

incapacitation on juvenile crime rates by using teacher strike days as variation in student

absence from school on an ordinary school day. Arguably, teacher strikes are a source of

variation that is exogenous to variations in juvenile crime. Issues that lead teachers to

strike include pay, class size and teacher planning time, none of which are likely to

influence daily variations in juvenile crime.4


2 For instance, a school district may report that it is sure teachers will return to work when a court
injunction is issued, and subsequently the teachers may defy the injunction.

3 Unlike most states in the U.S., teacher strike activity in Washington remains very much active to this day.
The most recent strike event took place in 2003 in Marysville school district. This strike lasted nearly two
months (approximately 50 total days). A summary table describing teacher strikes in Washington can be
seen in Table A-1.

4 In fact, areas where juvenile crime may be an increasing problem are arguably less-likely to see a teacher
strike, because greater crime has been shown to be positively correlated with higher teacher pay (Grogger
1997).









In order to draw direct comparisons between crime on strike vs. nonstrike days, I

must be able to control for days in which schools are closed for other reasons. This

would include days that coincide with spring, summer and winter vacations, weekends,

teacher in-service and half-days, national holidays, etc. However since the data cover a

large span of time (22 years), and since each Washington school district determines their

own school calendar, it is impossible to know where all of these nontypical school days

occur.5 Further, a prolonged teacher strike sometimes results in a reworking of the

existing school calendar (largely unobservable) so that a school district can meet the

requirement set by the state that the school year last for 180 days.6 Since I cannot control

for these school holidays with complete accuracy, I must include only those days which I

am strongly confident students are regularly scheduled to be in school.

To minimize any possibility that school holidays could bias the results, I take extra

measures when eliminating questionable days. I begin by eliminating weekends and

national holidays. If a national holiday falls on a Saturday or Sunday, I eliminate the

preceding Friday or subsequent Monday, respectively. The first two Fridays of October

are dropped because they are both traditional days for the state mandated teacher in-

service day. The entire months of June and July were also excluded, and only the last

three school days in August are allowed in the sample. Finally I eliminate the first week

5Each school district may vary on the scheduling of breaks, half-days, inservice days, etc. To the best of
my knowledge, there is no state or federal agency that has archived these calendars over the 21 years.
Since the advent of increased technological integration with school systems, some districts are beginning to
electronically archive these calendars, however these calendars generally only go back a couple of years.
The Washington State Superintendent of Public Instruction was able supply me with sufficient calendars
for most districts, for each year from Aug. 2000 to June 2004.

6 The shortening of spring and winter breaks is common, as well as the lengthening of the school year into
summer vacation.

7 Thanksgiving falls on Thursday, however both Thursday and Friday are dropped from the sample because
students are given both of those days off from school.









in January, the last two full weeks (at least) of March, the first two weeks of April (at

least 8 school days), and the last two full weeks (at least) of December. Based on these

criteria, I can be reasonably certain that each observation in the analysis is a typical

school day.8 Days which are included in the sample but are not ordinary school days will

serve to downwardly bias the results, because more juvenile crime should be naturally

occurring on those days. This is only true because this measurement of strikes as a

treatment is limited to the days where students have missed an expected day of school.9

1.2.1 Zip-Code Matching

My data set consists of individual-arrest data, where each juvenile is matched to his

home address zip code. Since zip codes are set by the Address Information division of

the U.S. Post Office, they look very different from other boundaries like county borders,

congressional districts, and most importantly school districts. In order to match each

juvenile (and his/her arrest) to his/her appropriate school district, I must address two

problems. The first problem is that zip codes often spill over multiple school districts.

Therefore it is often uncertain which school district a juvenile is in, given their home zip

code. The second problem is that zip codes are redefined over time. First I will focus on

how zip codes change, and how this problem is resolved so that identification strategy is

preserved.

As cities and towns develop both inside and outside city limits, the Post Office

creates new zip codes and new zip-code offices to handle increasing mail traffic. When




8 This is not a perfect matching process because half days and minor differences between districts are
impossible to pinpoint.

9 The timing of strikes is very particular and seldom overlaps with nonschool days, except weekends.









these new zip codes are created, they are drawn strictly as subsets of existing zip codes.10

The creation of these new zip codes makes zip-code fixed effects useless from one year

to the next. In addition, it is also very difficult to pinpoint when new zip codes are

created. However it is possible to see which zip codes were created and what preexisting

zip code they were originally a part of. By using zip-code maps of Washington State

from 1984 and 2000, I am able to map each subset zip code back into its original (1984)

zip code.1 Essentially I treat zip codes in Washington as if they are never partitioned,

beginning in 1984. This method leaves me with 510 zip codes in total and a stable zip

code definition over a 22 year time frame.

The second problem is matching these zip codes to their corresponding school

district. Zip-code overlap with school districts make this matching process difficult,

however the approach I take is sound, and fairly straight forward. I began by totaling the

number of schools in each zip code, noting which district they serve.12 I then divided the

total number of schools in that zip code for a given district by the total number of schools

in that zip code, regardless of their district. Essentially what I am left with is a

probability, that I will call p*, which reflects the probability that a child living in a

particular zip code is also part of a particular school district. Thisp* measure allows me

to assign a proper strike treatment to zip codes.


10 So far I have not been able to find an instance in Washington State since 1984 where a zip code was
created from two or more existing zip codes

1 Four maps of Washington State zip codes were provided by the Western Economics Research Co., Inc.
courtesy of the Suzzallo Library at the Washington Library. They encompass they following years: 1984,
1989, 1992 and 1994. Zip code maps of Washington State for the year 2000 are made available by the
Census.

12 I did this using a master list of schools (as of the 2003-04 school year) that includes each school's
address, grade level served, school code and district name. This list was provided by the Office of
Superintendent of Public Instruction in Washington State. It is also publicly available.









There are several features of zip codes that simplify this matching process. One

helpful factor is that many zip codes do not overlap school district boundaries at all. This

leaves out any guess work. Another helpful point is that zip codes and school districts

often share a significant number of common boundaries. These shared boundaries occur

where divisions seem logical, such as along a county boundary, or a major river or

highway. These convenient features lessen the confusion of the matching process. It is

also worth noting that not all zip codes contain schools.13 In my sample there were four

zip codes without any schools; two were in the heart of Seattle, and two were located in

rural areas. Conveniently, none of these four zip codes overlapped school district

boundaries, so assigning them to the appropriate school district was simple.

The last concern I am left with deals with how school districts change over time, if

at all. If school district boundaries are changing frequently and dramatically over time,

then the matching process breaks down. However, what I find is that school district

boundaries are stable over time. In order for a district to change its boundaries, it must

engage in costly and time-consuming legal processes, however that does not imply that

such adjustments never occur.14 In fact, the two most frequent changes to school district

boundaries seem to be district consolidation and district dissolution.15 Neither changes

turn out not to be problematic. These changes only occur a total of four times over the




13 The zip codes I use in the sample are zip codes with residences. Many zip codes like business zips, P.O.
Box zips, and other nonresidential zips, do not contain schools. These kinds of zips are not included in the
data sample.

14 For an outline of regulations and requirements surrounding district boundary changes, please refer to the
publication, "Changing School District Boundaries: A Lay Person's Guide" published by the Washington
State Board of Education in conjunction with the Office of Superintendent of Public Instruction
15 Consolidation is when two or more districts form a new superdistrict. Dissolution is when a one or more
districts are absorbed into existing districts.









course of the 22 years and involve districts which do not experience any teacher strikes. I

treat these integrated districts as if they had never been separate.

The school district zip code matching process I invoke does not provide perfectly

accurate matches in all cases, but I can be certain of how any mismatching will bias the

results. If I say there was a strike in an unaffected zip code, the effect of the strike will be

biased toward zero because juvenile crime should be unchanged. If I say that there was

not a strike in an affected group, then the strike effect will fail to pick up any increase in

juvenile crime rates. Any mismatching that arises from the imperfect matching process

will downwardly bias the results.

1.2.2 The Basic Model

In this basic model I am attempting to explain changes in juvenile crime for

ordinary school days as a function of teacher strikes. I can start by expressing a simple

regression model in the following form:

Juvenile School Day Crime = a + P(p*)(Teacher Strike)

Again, all of the information from the crime data is reported at the zip-code level,

however the teacher strikes occur at the school-district level. To adjust I introduce p*

where* represents the probability that the student population of a zip code is treated by

the teacher strike. Rather than have fixed effects to take into account differences across

zip codes, I want to initially include specific zip-code characteristics to make sure the

data are well-behaved. I consider income levels, welfare status, parental education,

juvenile work status, juvenile gender, single parent households and community

characteristics in this model. Also I need time-fixed effects to control for temporal

changes over 22 years. I therefore include year, month and day fixed effects in the

model.









The regression model now takes the form:

Juvenile School Day Crimemyd = a + 11(p*)(Teacher Strike)myd +...
+...2(Median Income)myd + P3(Welfare)myd + P4(Parent Education)myd
+... 5(Student Employment)myd + 36(Juvenile Gender)myd +...
+.. 07(Urban)myd + 8(Single Parent)myd 61(Year) + 82(Month) + 63(Day)

Once I am satisfied that the data exhibit all of the normal properties one would expect

from specifying a model of juvenile crime, I can then shift the analysis to include zip

code fixed effects take into account differences which I cannot control for. When I move

to this specification, the final regression model looks like the following:

Juvenile School Day Crimemydz = a + 11(p*)(Teacher Strike) mydz ...
+... 6i(Year) + 62(Month) + 63(Day) + 64(Zip)

1.3 Data

My juvenile-arrest data comes from "Washington Juvenile Court Case Records"

made available by the National Juvenile Court Data Archive. It provides juvenile arrest

data over 22 years, spanning from 1980 to 2001. This data set is both lengthy as well as

highly detailed at the individual level. It provides the home address zip codes of the

offenders, the nature of the crimes committed, the dates of the offenses (as well as the

arrests), as well as many other important characteristics surrounding the reported arrests.

In total when I include only Washington juveniles who were arrested for crimes that

occurred during the ordinary school days preselected for my study, I have 401,864 arrest

cases starting in 1980.

In addition to the juvenile arrest data for Washington State, I use data provided by

the "Census 2000 Summary File 3" and "1990 Summary Tape File 3" to derive annual

zip-code characteristics. These Census files give zip-code level information on

population and school enrollment numbers, as well adult educational attainment, the

number of households who are welfare recipients, type of community, median household









income and student employment. Because these characteristics are only available for

these two periods (1990 and 2000) I trend them over the 22 years using an exponential

function rather than a linear function to describe the path of the trend.16

Finally, teacher strike information came from the report "Public Employee Strikes

in Washington" published in 2003 by the Public Employees Relations Commission in

Washington State. It covers all Washington public employee strikes since 196717;

however since this publication lacks some important details about these strikes, it has

been supplemented with specific information from news articles provided by the

Associated Press and the Seattle Times to strengthen the accuracy of the treatment effect.

These articles are useful in pinpointing specific details surrounding each strike such as

whether the school remained opened with emergency substitutes, and what information

was distributed to parents in districts where strikes occurred.

1.3.1 Methodology

Choosing a proper methodology for this analysis requires some extra care. Because

the analysis is done at the daily level, a huge percentage of observations for any given

day will be zero. The distribution of the dependent variable is skewed towards zero,

making an OLS regression analysis inappropriate. This could be solved by using a

poisson regression analysis, however the fit for a poisson regression for this dataset is

poor.18 Also the poisson analysis does not account for strangely behaved standard errors


16 I also trended these variables linearly, however when I do this I end up with some numbers that are not
feasible do to the imprecise nature of this process. When I impose minimum and maximum constraints on
these variables, I find that these results are insensitive to whether these variables are defined linearly or
exponentially.

17 This report began with the enactment of Public Employees Collective Bargaining Act in 1967.

18 A preliminary test of the data shows that the dependent variable generally has a standard deviation
roughly 3 times larger than the mean. In addition, I observe large values of the Chi squared terms in the









and overdispersion. In this case, overdispersion may take the form of a single juvenile

committing several crimes on a single strike day, or a gang of juveniles committing what

could be considered one crime. Ultimately the chosen method of regression analysis is a

negative binomial regression analysis. This deals with problems of overdispersion that a

poisson does not correct for.19 The dependent variable is aggregate crime per day (crime

count) in a zip code, and the exposure is total student population in said zip code.

1.3.2 Defining the Parameters

As described earlier, I begin by controlling for differences among zip codes directly

rather than using zip-code fixed effects. Each variable is derived from data taken from

the 1990 and 2000 Census. Since I have only two observations for each variable, I

interpolate annual values using an exponential path function. I started by solving for the

growth rate (Gi, zip) of each variable izip:

Gi, zip = ((Censusi, zip 2000/ Censusi, zip 1990)(110)) 1

Then I interpolated each year's observation according to the equation:

Variable izip in year t = Censusi, zip 1990*(1 + Gi, zip)(n), where n = t 1990

Overall this analysis uses 7 variables to capture differences in communities. Income is

measured using Median Income, which comes directly from the Census data. The

Welfare variable is a percentage of households in a zip code that currently receive public

assistance income. Of course I expect that local poverty is a serious factor in juvenile

crime, so both of these variables seem to be relevant (Mocan and Rees 1999). Juveniles

from poor neighborhoods are more likely to be less educated, and therefore more likely to

poisson "goodness-of-fit" tests. Both of these characteristics suggest that the poisson model in not the
correct model for these regressions.

19 The overdispersion parameter alpha is shown to be significantly different from zero. Thus I can conclude
that the negative binomial regression significantly different from the poisson regression.









be criminally active overall (Freeman 1992). Since juveniles from poor households have

a smaller opportunity cost of committing crime, I expect Median Income to be negatively

correlated with juvenile crime, and naturally the converse should be true for Welfare.

The Poor Parental Education variable is defined as the percent of adults (25 years+)

who have not completed the equivalent of a high school education. My prediction here is

that as adult dropout rates increase, juvenile crime increases. Because poor adult/parental

education equates to less adult human capital, parents are likely to have lower demand for

child quality (Becker and Tomes 1976, 1986). Therefore juveniles in these

neighborhoods should also have a lower opportunity cost of crime. In addition, one can

argue that poor adult/parental education may also imply that parents have fewer resources

to provide monitoring for juveniles, less overall time to provide supervision of their own

children (parents may be working long hours or more than one job), or possibly

uninformed/undesirable parenting techniques.

Our Student Employment variable describes the percent of 16-19 year olds who are

both employed and also enrolled in school. The prediction of this variable is not clear.

On one hand, students who are employed in after-school jobs may develop a sense of

responsibility from working a job, or that they may fear losing future wages should they

be caught committing a crime. This suggests that Student Employment should negatively

influence juvenile crime. On the other hand, juveniles who work may have greater access

to reliable transportation, or less restrictive parents, both of which could contribute to a

greater propensity to commit crime. In this case, Student Employment would positively

influence juvenile crime.









The Juvenile Male variable measures the percent of the juvenile population in a zip

code who are male. Since juvenile males traditionally make up the majority of juvenile

crime, juvenile crime for a zip code should increase as a juvenile population there

becomes proportionately more male.20

The Single parent variable is the percent of single parent households in that zip

code. One expects that this variable should be positively correlated with juvenile crime

so that more single parent households lead to more potentially unsupervised juveniles,

and potentially greater juvenile crime.

The Urban variable is the percent of the total zip code population who live in an

urban community. This should pick up basic differences among community types,

however I also interact this variable with the strike measure to see if strikes have

differential impacts in different types of communities. For several reasons, I expect

crime in Urban areas to differ from that of Rural or Suburban areas. Urban areas feature

many characteristics, such as increased criminal opportunity, higher pecuniary benefit,

and increased criminal anonymity, which makes criminal activity in these areas more

desirable (Glaeser and Sacerdote 1999). As such this Urban variable should be positively

correlated with teacher strikes. Lastly, the Strike variable is a dummy variable, with a

value of 1 describing a teacher strike event, and zero otherwise.

1.4 Regression Analysis

The initial results of the negative binomial regression analysis seem to confirm

previous findings about the nature of school incapacitation. Column 1 in Table 1-1



20 From 1993 to 2002, males made up about 71-75% of total juvenile arrests. For more information about
juvenile crime by gender, please reference the Crime in the U.S. annual report published by the Federal
Bureau of Investigation.









shows that the presence of teacher strikes seems to have a positive and significant effect

on juvenile criminal behavior. But before I begin to quantify the effect of school

incapacitation on juvenile crime, I want to make sure that the data are behaving properly.

The signs of the covariates in Column 1 of Table 1-1 seem to support the

predictions made on the effects of income, welfare, parent's education and urban status

with respect to juvenile crime. First, we see that median income and welfare are

negatively and positively significant, respectively. This implies that children in lower

income households and welfare receiving households are more likely to engage in

juvenile crime. These results support the predictions associated with poorer living

conditions. In addition, lower parental education leads to higher crime as well. Again,

for several reasons this result makes sense. The Urban variable is positive and

significant, showing that urban communities experience more overall juvenile crime.

Our Student Employment variable is positive and statistically significant. This may

imply after-school jobs provide students with greater resources which they may use to

commit crime. Further, it may reflect the fact that as more students get after-school jobs,

they are spending less time increasing their human capital (through study, after-school

activities, etc.)

One problem is that the effect of the Juvenile Male variable is negative and

significant, which contradicts the predicted effect. It says that across juvenile

populations, those that are comprised of relatively more females experience more

juvenile crime. One possible explanation for this result is that it may be reflecting

problems in variable measurement for small communities. For rural communities there

are two basic problems. The first is that with a small population of juveniles, the









variation of the Juvenile Male variable increases significantly.21 The second is that this

unusual variation makes interpolating this variable equally unreliable. Since a significant

portion of zip codes have very small populations of juveniles, it is possible that the

negative result I observe is being driven by rural zip codes that have atypical juvenile

gender characteristics.22

To check the sensitivity of the initial results I begin by eliminating months in which

strikes did not occur from the sample: January, March, and May (Specification II).23 I

then went a step further and eliminated every month other than September, October and

April from the sample (Specification III). September and October are preserved because

they are the most common months for strikes, and April was included because in 1991

there was a "state-wide" teacher strike that involved at least 41 school districts. The

results of Specifications II & III are reported in Columns 2 and 3 of Table 1-1

respectively. These results are not significantly different from Specification I.

Recalling the zip code matching process described earlier in the paper, I am able to

accurately match back to 1984 zip codes. However the crime data dates as far back as

1980. To deal with potential zip code mismatches before 1984, I drop the first four years

of the sample (Specification IV).24 This fourth specification shows the strike variable


21 For zip codes with juvenile populations less than 500, the standard deviation of the Juvenile Male
variable is approximately 3 times larger than the standard deviation of those zip codes with 500 or more
juveniles.
22 In fact when I look at only urban zip codes, I see that this variable seems to correct itself, which gives
credence to this argument. Further, Table A-2 shows that when I drop the smallest zip codes (population
less than 500 juveniles) I see that the negative significance of the coefficient is destroyed. Thus it seems
plausible that poor interpolation of this variable for sparsely populated areas has tainted the coefficients.

23 Over the 22 year period, there was one school district that orchestrated a strike in January, but it only
lasted for one day.
24 In doing this I eliminate 9 strike events, 3 of which lasted 9 or more days.









remains significant as the sample is modified (Column 4 of Table 1-1). These results

seem robust to many different representations of the data set.

It is also important to verify that these results are not being driven by zip code

characteristics being specified inaccurately. The most immediate check of this is to use

zip code fixed effects in lieu of explicit variables to capture the difference. Column 5

shows the addition of zip code fixed effects does not significantly alter the results.

The coefficient of the strike variable Column 5 of Table 1-1 shows that strikes have

a positive effect on juvenile crime that is statistically significant. To quantify the impact

of this increase, I take the partial derivative of the dependant variable with respect to the

strike variable (aTotalcrime/aStrike). When I divide this marginal effect by the average

of the dependent variable, I am left with a change that reflects a proportion of the mean of

the dependent variable. In the full data set, the marginal effect of the strike variable

suggests that total juvenile crime increases by 56.71% on days when strikes occur. This

is more than a modest change in total crime.

1.4.1 Community Differences

Since the full data set includes all communities, some of which may look and

behave very differently from one another, I want to test whether juvenile crime in

different communities is differentially affected on strike days. To capture community

differences, I run separate regressions for each community type, where I restrict the

sample to only those communities in which 51% or more of the population in a zip code

live in one of three types of communities: urban, suburban or rural.25 The results of these


25 The reader should be aware that this specification implies that zip codes which are considered one-third
urban, one-third suburban, and one-third rural, will therefore be excluded from this analysis. However
because of the way in which the Census defines these communities, these kinds of zip codes make up a
very small portion of the total sample.









regressions can be viewed in the Columns of Tables 1-2, 1-3 and 1-4. These results show

that unexpected school absences significantly influence juvenile crime only in urban

communities.

This result is not at all unreasonable, given there are many differences among these

types of communities that no doubt influence criminal behavior. It seems understandable

that rural communities experience no change in juvenile crime when a strike occurs.

Rural communities may provide less opportunity for crime, as well as a small town

atmosphere that makes anonymity difficult. The lack of a significant effect of strikes in

suburban communities is less believable since they arguably provide greater opportunity

and anonymity, however there may yet be other, (possibly unobservable), characteristics

about suburban characteristics that naturally deter juvenile crime on these strike days.

For instance, suburban communities may be populated by more involved parents, may

have more effective emergency resources, etc. The strike effects in urban communities

however are positive and significant. In urban communities juvenile crime increases

19.68% on strike days.26

Again I run the same specification tests to test the robustness of these results. The

signs of the covariates in every urban subsample are the same as in Table 1-1, except for

Juvenile Male. The Juvenile Male covariate does take on a positive and significant

coefficient in the urban subsample, however in the suburban and rural subsamples it

remains negative and significant.





26 The huge decrease in the magnitude of the strike coefficient from the full sample to urban subsample
indicates that small populations in rural communities are exaggerating the effects of the strikes in the full
sample.









1.4.2 Differences in Offense Types

Since school incapacitation may be differentially affecting the types of crimes that

juveniles are committing, I partition the dependent variable measure of total crime into 5

specific crime types. These types of crimes include: drug and alcohol related crimes,

mischievous crimes, property crimes, violent crimes, and finally weapons and

endangerment crimes. I still consider all three community classifications when doing this

analysis because I may yet find significant results in types of crimes for rural and/or

suburban communities.27 Table 1-5 shows the results of this extended analysis.

Rural and suburban communities continue to experience no change in juvenile

crime resulting from strike days regardless of the type of crime that is being committed.

Urban communities though do experience an array of changes in juvenile criminal

behavior. Both mischievous crimes and property crimes seem to increase in the presence

of a teacher strike. I estimate that mischievous crime increases by 48% on average for

days when strikes occur.28 Property crime increases by an average of 28.81%. This

change in property crime is nearly double previous estimates in the literature. Violent

crime decreases by approximately 31.53%.29 Drug and alcohol crimes, and weapons and

endangerment crimes are unaffected by school incapacitation.

These changes in criminal activity seem to make sense. It seems logical that

mischievous crime would be most affected by school incapacitation. After all, these are

27 This could arise if increases in certain crimes were offsetting decreases in other types of crimes.

28 It is difficult to compare my measure of mischievous crime to Jacob and Lefgren's measure of minor
offenses. They define minor offenses as "NIBRS Group B offenses", however my measure of mischievous
crime is a select subset of these crimes.

29 Jacob and Lefgren estimate that school attendance decreases property crime by approximately 14%.
They also estimate that violent crime increases on school days by about 28%. Therefore, my estimate of
31.53% for violent crime is approximately 10% larger than previous estimates.









the types of crimes that often result from boredom rather than calculated criminal

thought. On the other hand, the decrease in violent crime that occurs is harder to explain.

I use, as Jacob and Lefgren do, social interaction theory to explain this result. Jacob and

Lefgren argue that juvenile violence against other juveniles arises, at least in part, from

disputes formed in and around the classroom. When these juveniles are not forced to be

at school, there is a decrease in the amount of overall violent crime that reflects a

decrease in juvenile violence against other juveniles. This explanation seems credible.30

Given the limits of the data set, I cannot test this theory further.31

1.4.3 Types of Offenders

In addition to the topics already spoken to, the detailed level of the data also allows

me to also explore the nature of the criminal, beyond superficial characteristics. By

observing criminal record, I attempt to answer whether these additional crimes are caused

by normal delinquents facing greater criminal opportunity, or by juveniles who do not

normally engage in criminal activity. I again partition the dependent variable into two

measures: 1) total crime by repeat offenders and 2) total crime by one-time offenders. I

define repeat offenders not just as those juveniles who have committed previous crimes,

but also those juveniles who will commit future crimes. Thus a juvenile who commits

multiple crimes is considered to be a repeat offender juvenile even on their first crime.

The results of these regressions can be seen in Table 1-6.

Additional overall crime is induced for both offender types, however the average

increase in crime for one-time offenders as a percent of the mean crime level for that



30 In 1998 and 1997, 62% of victims of juvenile violence were 17 or younger.

31 I have no information concerning the victims of the crimes in my data set.









group is much higher than for the repeat offenders. On strike days there is a 33.45%

increase in average total crime for one-time offenders, and a 13.89% increase in average

total crime for repeat offenders. Essentially, one-time offenders as a group seem to be

much more affected by the lack of school incapacitation then the repeat offenders. This

evidence suggests that the increased crime on strike days is less about criminal motive,

and is more a result of boredom.

If these two groups are motivated differently to commit crimes, these effects may

not be expressing differences across groups in the same types of crimes. As a society we

may be more concerned if the lack of school incapacitation were inducing more property

crime by one-time offenders rather than repeat offenders (where the crime may have

eventually happened anyway). Therefore I am interested in examining whether these two

groups differ in offense types on these strike days. To do this I create a dependent

variable that equals total crime conditional on offender type and offense type. The results

are expressed in Table 1-6. The decrease in violent crime comes mainly from the repeat

offenders. Violent crimes committed by repeat offenders drop by 36.72% on strike days.

In addition, both groups are contributing to the overall increases in mischievous crimes

and property crimes. Property crimes by one-time offenders and repeat offenders

increase by 56.75% and 25.40% respectively. Likewise mischievous crimes increase by

83.11% for one-time offenders and by 30.82% for repeat offenders.

1.4.4 Strikes Days as Gateway Crimes

Since the increase in crimes committed on strike days is substantially larger for

first-time offenders, it may also be true that these juveniles may carry this behavior

forward after their first arrest. To see if this is the case, I examine whether or not a









current day repeat offender was significantly more likely to haven gotten his/her start

(commit his/her very first crime) on a strike day versus a normal school day. To do this, I

generate a daily crime total that expresses the total of"first time" crimes by subsequent

repeat offenders. The results of this regression can be also viewed in Table 1-6. A future

repeat offender is found to be significantly more likely have committed his first crime on

a strike day rather than an ordinary school day. This suggests that not only is school

incapacitation preventing the creation of crimes, but it may also be preventing the

creation of criminals.

1.4.5 Displacement of Crime

Thus far it seems the evidence suggests juvenile crime increases when students are

not incapacitated in school. However what is not clear is whether these changes describe

an overall increase in the amount of juvenile crime, or a displacement of crime from one

day to another. Jacob and Lefgren try to speak to this question of temporal displacement

in their paper, however their use of in-service days again limits the power of their

analysis. Jacob and Lefgren find that there is no temporal displacement of crime to in-

service days, however because in-service days are completely known, this result may

simply be showing that juvenile criminals will plan to commit more crime when facing

more criminal opportunity. What we are really interested in is whether a juvenile will

temporally displace a "planned" crime given an unexpected opportunity to do so, or

whether an unexpected opportunity to commit crime results in more "unplanned" crime.

Again teacher strikes provide a more ideal measure of unexpected criminal opportunity,

so I revisit this question of temporal displacement.









To test whether crime is being displaced, I begin by aggregating Total Crime to a

weekly measurement rather than daily32. If juvenile crime is simply being displaced from

weekend crimes to weekday crimes, then I should not observe an effect for the strike

variable. To capture the weekly aggregated effect of strikes, I proportion the fraction of

the week where strikes occurred as a new measure of the strike treatment. Despite this

aggregation, I find that strikes are still a significant factor in contributing to additional

juvenile crime (Table 1-7).

Table 1-7 shows a positive and significant strike coefficient on Total Crime in the

full data set. This is a strong indication that overall juvenile crime is increasing, rather

than being displaced across time. However, if displacement of crime varies across crime

type, or criminal type, we cannot conclude that no crimes are being displaced. To see if

the type of crime or the characteristics of the perpetrator influences whether displacement

of crime occurs, I partition weekly total crime by crime type and offender as before. The

results of these regressions are also reported in Table 1-7. These strike coefficients on

Table 1-7 show that every specification of crime type and criminal type results in an

increase of overall juvenile crime, and not temporal displacement of crime. I confirm the

lack of displacement in total crime by aggregating further to a monthly crime total.

Weekly and Monthly aggregations of juvenile crime provide a good test of

displacement of crime. I perform another test by lagging the strike effect, in order to pick

up acute changes in crime shortly after the strike ends. If fewer crimes are committed on

days following the strike, then the coefficient of the lagged strike variable should be



32 This weekly measurement includes weekends and holidays, since strike day crime could displaced from
usual weekend crime. In addition, week fixed effects are used in place of day fixed effects.









negative and significant.33 I run seven separate regressions, shown in Table 1-8, labeled

"1 day" through "7 days." For a strike that ends on day t, the first treatment of the lagged

strike variable under "1 day" begins on day (t + 1), under "2 days" begins on day (t + 2),

and so on. Table 1-8 shows that there is no significant effect from the lagged strike

treatment. These results provide further evidence that juvenile crime occurring on strike

days is additional crime, rather than displaced crime.

1.4.6 Duration Effects on Criminal Activity

One last advantage I gain from using teacher strikes as an instrument, in

conjunction with the strength of the daily arrest data, is that I can track how juvenile

crime changes over time with prolonged absence. If it were true that an increase in

juvenile crime is at least partially caused by boredom and inactivity on the part of

juveniles, then I should expect that the longer juveniles remain "inactive", the more crime

is likely to result. Arguably there is just as much uncertainty over when teacher strikes

will end as when they will begin. As such, trends in juvenile crimes up until the start of

the strike and over the period of the strike should not be biased by student or parent

foresight. In addition, the longer a strike continues, the less able a parent or guardian is

to provide adequate supervision to students. For example, a parent may be able to take

one or two days away from work to supervise their child, but the cost of taking 10

consecutive days off work is substantially larger.

To examine if juvenile crime is changing with the length of a teacher strike I

partition the strike variable into 4 groups of new independent variables. Three of these

variables are clustered groups of 3 consecutive strike days, and the last variable pertains


33 This sample period also includes weekends. I also revert back to day fixed effects rather than week FE.










to a strike in its 10th day or higher.34 These results can be viewed in Table 1-9. Table 1-9

shows that for the first 3 days, there is no significant change in overall juvenile crime.35

After the 5th day or so, the strike coefficient becomes positive and significant, reflecting

an increase in overall crime. Further, after the 5th or so strike day there is a gradual

increase in the level of juvenile crime as strike length continues. In unreported

regressions I find that as time passes, property and mischievous crime trends upward,

while the decrease in violent crime remains relatively constant across days. The positive

coefficient shows that these increases in property and mischievous crime overpower the

decreases in violent crime. Column 2 of Table 1-9 repeats the regression in Column 1,

only with the strike variable clustered at the two-day level rather than at the three-day

level. Clearly the evidence suggests that the longer juveniles remain absent from school,

the more overall daily juvenile crime will be observed.36

1.5 Robustness Checks

The results of the analysis up to this point seem to point to a very clear story about

the relationship between juvenile crime and school incapacitation. In this section I

perform some robustness checks to ensure that these results are not spurious. I begin by



34 do this because if I were to partition the strike variable for each separate day (1st day, 2nd day, etc.), then
the number of treatments would be very small for the later strike days and the coefficients would be
unreliable.

35 When I break up the dependent variable (total crime) by crime type as before, violent, property and
mischievous crime types are individually significant. However in the first 3 days, the opposite signs and
magnitudes of the coefficients wash out aggregate effects in total crime. These results are consistent with
Jacob and Lefgren's findings (2003).
36 It is worth keeping in mind that these strike day counts are also limited to school days. That is to say that
when a strike has reached its 10th day, a student has missed exactly 2 full weeks of school. It is also very
likely that these criminal behaviors will change after a strike has gone on for an extensive amount of time.
At that point, juvenile crime may level-off or even decrease substantially. Of course given the sparse and
limited nature of extremely lengthy teacher strikes, an analysis of that nature is inherently difficult not to
mention highly unreliable.






25


testing whether the strike variable is correlated with some unobservable characteristic,

and that perhaps the results are picking up some alternate relationship. For instance, if a

strike event is in fact very foreseeable, then one might speculate that the increase in

juvenile crime predicted by the strike variable is really picking up an increased police

presence on days that strike occur. That is to say that the results are not reflecting a true

increase in juvenile crime, but that there are simply more police on the streets capturing a

larger portion of the same total amount of juvenile crime.

If this were true, then one would expect that the number of days between an offense

and an arrest would be shorter on strike days. The graphs on Figure 1-1 show that this is

not the case. When I consider only zip codes where strikes did occur, I see that the

percent of quick arrests37 is significantly higher on regular school days than on strike

days. I explain this by suggesting that on regular school days, police have a better idea

who the likely criminals are (repeat offenders), as opposed to strike days where offenders

are more likely to be (previously unidentified) first time, or one-time offenders.

1.5.1 Importance of Singular Strikes

Given the uniqueness of strikes, I must confirm that these results are not being

solely driven by a small subsets of strikes. To do this I systematically eliminate each

individual year from my sample starting with 1984. Table 1-10 shows that strikes are a

significant determinant of juvenile crime in every year with the possible exception of

1985 (z = 1.70). When I drop 1985 from the sample I find a much weaker significance

than when I eliminate any other year. In September of 1985, the Seattle school district

engaged in a teacher strike that lasted 20 school days. Given that the evidence has


37 I define these as arrests where the associated offense occurred less than 48 hours before the arrest.









suggested that juvenile crime in urban districts is most affected by teacher strikes, it is

possible that these results are being solely driven by the circumstances surrounding this

one district.

To test this condition I restrict the sample twice, once to look at the sample without

the Seattle school district, and then again to look at only the Seattle school district. If

Seattle is driving all of the results, then the strike variable should be insignificant when I

exclude Seattle, and extremely significant when I exclude all other districts except for

Seattle. Table 1-10 shows that strikes still have a positive and significant effect when I

drop Seattle from the analysis. In addition the strike variable in the Seattle subsample is

marginally significant (z = 1.82). Therefore I can be confident that while this one event

obviously contributes to the result, (as I would expect the second longest strike in the

most urban school district in Washington State to do), it is not the sole force driving the

results.38

1.5.2 Reproducibility of Results

Another robustness check I perform is to make sure that the results cannot be easily

reproduced with random generations of the strike variable. To check if this is the case, I

use a uniformly generated random variable on the interval from 0 to 1 to re-specify a

random dummy strike variable. Since teachers are on strike 2,351 days out of the

1,703,910 day sample, strike days occur 0.13797 % of the time. I assign a strike value of

1 to the random variable, when the value of the random observation generated is less than

or equal to 0.0013797. If the randomly generated observation is greater than 0.0013797,

then it is reassigned a value of zero. The new random strike variable should approximate

38 The Seattle strike of 1985 is actually tied with the Mukilteo strike of 1990 (at 20 days) for the second
longest strike in Washington State history from 1980-2001.









the number of strikes in the original sample. The results in Table 1-11 show the mean of

the random strike coefficient over 25 trial regressions, as well as the average number of

random strike days generated. Of the 25 trials, only 2 random strike variable coefficients

were significant. One of the statistically significant coefficients was positive, while the

other was negative.

1.5.3 Further Tests

One final test of the data is to make sure that my finding of new criminal

participation on strike days (when the sample is broken up into one-time and repeat

offenders) is not a byproduct of mislabeling offenders. Since the crime data ends in

2001, it could be that in the years leading up to 2001 many future career criminals are

being mislabeled as one-time offenders due to the fact that future arrests past 2001 are

unknown. To test this possibility, I drop the last 4 years of the data sample and repeat the

regressions using offender type as the dependent variable. Table 1-12 shows that this

mislabeling does not affect the results. The effects of a strike on repeat offenders and

one-time offenders are nearly identical for either specification.

1.6 Conclusion

The results presented in this chapter provide a clearer picture concerning the

relationship between schooling and urban juvenile crime. My main findings are the

following: 1) The effects of a lack of school incapacitation are larger than those found in

Jacob and Lefgren's study. The evidence suggests that property crime rises by as much

as 29%, almost twice as much as what is predicted in the existing literature. Likewise,

violent juvenile crime decreases by as little as 31.53%, and as much as 36.72%. These

estimates of increased violent crime represent a 10%-25% increase in previous estimates.







28

2) I confirm that these changes in daily crime reflect changes in total crime, and not a

displacement of crime from one day to the next, regardless of crime type and criminal

type. 3) Different types of juveniles are affected differently by a lack school

incapacitation. A significant proportion of the decrease in the level of violent crime can

be attributed to repeat juvenile offenders, while one-time offenders contribute more to the 28

increases in property crime and crimes of mischief. 4). A failure to incapacitate juveniles

will result in significantly more crime by those juveniles who might not have otherwise

engaged in criminal acts or who rarely engage in such acts. In fact, juveniles who

become repeat offenders are more likely to have gotten their criminal start when

incapacitation was expected but not implemented, as opposed to ordinary school days.

Incapacitation seems to have the greatest implications for new and seemingly

preventable urban crime. I find that on strike days, new criminals are engaging in new

criminal acts. Incapacitating these juveniles seems to be effective at preventing at-risk

urban youths from engaging in new and risky behavior. I do not find that school

incapacitation has any significant effect in suburban and rural communities. It may be

true that the characteristics that differentiate these community types from urban

communities, as well as each other, inherently reduce juvenile criminal behaviors,

however such characteristics may be largely unobservable to us.

If the evidence presented in this paper is accurate, then what we are learning is that

how we manage children's time outside of school is very important. If parents are

strained to provide adequate supervision for their children, juveniles who we might

consider to be at-risk may be the ones who are affected most, though at-risk juveniles are

not the only ones affected. Families and businesses also bear the burden of increased









juvenile delinquency. These results have profound implications for many urban school

policies and programs. How school districts budget their students' days off is no longer a

trivial matter. For example, school districts with long breaks (like in a traditional school

calendar) may have very different juvenile criminal behaviors than a school district with

frequent but shorter breaks (like in a year-round school calendar). Differences in how

school calendars overlap within and among districts should have an effect on the nature

of juvenile crime. The length of the school day or school year in a district, school district

policies regarding student attendance requirements, and even the nature of after school

programs within a district are all policy issues that determine more than just a child's

educational outcome. It is my goal to continue to explore this data set and scrutinize

these kinds of school policies, so that I can better understand the impact that these kinds

of policies may have.











Table 1-1: Negative Binomial Regression onto Total Crime with Full Data Set
Variables (I) (II) (III) (IV) (V)
Strike 0.195** 0.187** 0.210** 0.260** 0.220**


(4.06)**


(3.86)**


(4.16)**


(5.06)**


(4.38)**


Median Income



Welfare


Urban



Poor Parental Educ.



Juvenile Maleness



Student Employment


Single Parent House


Alpha


-0.000058**
(102.47)**

0.475**
(7.71)**

0.132**
(26.96)**

0.329**
(10.40)**

-0.330**
(28.06)**

0.291**
(17.49)**

0.500**
(9.52)**

0.989**
(156.56)**


-0.000059**
(83.82)**

0.452**
(5.96)**


-0.000059**
(60.76)**

0.520**
(4.97)**


0.112** 0.097**
(18.45)** (11.54)**


0.346**
(8.88)**

-0.341**
(26.24)**


0.398**
(7.45)**

-0.355**
(22.32)**


0.297** 0.290**
(14.43)** (10.23)**


0.495**
(7.64)**

1.009**
(126.04)**


0.466**
(5.20)**

0.997**
(91.12)**


Number of obs. 1,703,910 1,128,630 583,440 479,400 479,400
Time Fixed Effects Y Y Y Y Y
Zip Fixed Effects N N N N Y
R-Squared 0.0367 0.0364 0.0367 0.0279 0.0559
In Columns I V, the sample is re-specified as follows (Type I through Type V):
(I) Sample includes all ordinary school days from 1980-2001
(II) January, March and May are excluded from the sample 1980-2001
(III) Includes only April, September and October from 1980 2001
(IV) Only April, September and October from 1984 2001
(V) Only April, September and October from 1984 2001, with zip code fixed effects
Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level,
** is significant at the 1% level or smaller. To interpret these regressions, I have to calculate the marginal
effect of these variables. The marginal effect of the strike variable is 0.149, which represents a 56.71%
increase in average total juvenile crime on strike days.


-0.000055**
(52.80)**

0.821**
(6.28)**


0.055**
(6.14)**

0.482**
(6.67)**

-0.396**
(22.72)**

0.296**
(9.16)**

0.985**
(9.67)**

0.956**
(87.74)**


0.759**
(79.06)**










Table 1-2: Urban Subsample Negative Binomial Regression onto Total Crime
Variables (I) (II) (III) (IV) (V)
Strike 0.151** 0.129* 0.178** 0.211** 0.214**
(3.07)** (2.60)* (3.50)** (4.11)** (4.26)**

Median Income -0.000061** -0.000062** -0.000062** -0.000057** -
(83.07)** (67.39)** (48.42)** (42.17)** -

Welfare 1.394** 1.369** 1.418** 1.703** -
(16.40)** (13.01)** (9.79)** (9.46)** -

Poor Parental Educ. 0.398** 0.403** 0.449** 0.462** -
(9.49)** (7.79)** (6.31)** (4.65)** -

Juvenile Maleness 0.602** 0.577** 0.482** 0.397** -
(11.13)** (8.60)** (5.16)** (3.62)** -

Student Employment 0.214** 0.224** 0.214** 0.157** -
(9.10)** (7.67)** (5.34)** (3.47)** -

Single Parent House 0.739** 0.790** 0.945** 1.276** -
(8.28)** (7.12)** (6.24)** (7.55)** -

Alpha 0.676** 0.685** 0.666** 0.638** 0.494**
(114.06)** (91.00)** (65.18)** (62.65)** (54.87)**
Number of obs. 454,953 301,312 155,601 129,593 129,593
Time Fixed Effects Y Y Y Y Y
Zip Fixed Effects N N N N Y
R-Squared 0.0484 0.0484 0.0487 0.0400 0.0635
Table 1-2 limits the sample to include only zip codes where 51% or more of the population lives in an area
considered by the Census to be Urban
In Columns I V, the sample is re-specified as follows (Type I through Type V):
(I) Sample includes all ordinary school days from 1980-2001
(II) January, March and May are excluded from the sample 1980-2001
(III) Includes only April, September and October from 1980 2001
(IV) Only April, September and October from 1984 2001
(V) Only April, September and October from 1984 2001, with zip code fixed effects
Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level,
** is significant at the 1% level or smaller. To interpret these regressions, I have to calculate the marginal
effect of these variables. The marginal effect of the strike variable is 0.121, which represents a 19.68%
increase in average total juvenile crime on strike days.










Table 1-3: Suburban Subsample Negative Binomial Regression onto Total Crime
Variables (I) (II) (III) (IV) (V)
Strike -0.027 -0.064 -0.159 -0.162 -0.449
(0.12) (0.29) (0.61) (0.62) (1.74)

Median Income -0.000015** -0.000016** -0.000013** -0.000012** -
(8.01)** (6.54)** (4.74)** (5.66)** -

Welfare -1.929** -2.146** -1.483** -2.005** -
(9.21)** (8.31)** (4.20)** (4.43)** -

Poor Parental Educ. 1.639** 1.856** 1.683** 2.393** -
(13.33)** (12.29)** (8.09)** (8.55)** -

Juvenile Maleness -1.432** -1.973** -2.151** -2.899** -
(5.33)** (5.94)** (4.74)** (5.66)** -

Student Employment -0.121* -0.198** -0.172 -0.064 -
(2.23)* (2.96)** (1.87) (0.59) -

Single Parent House 2.285** 2.369** 0.217** 3.062** -
(14.13)** (11.97)** (7.83)** (8.82)** -

Alpha 1.326** 1.303** 1.240** 1.203** 0.969**
(59.08)** (54.84)** (39.01)** (37.78)** (34.42)**
Number of obs. 134,100 88,841 45,794 37,992 37,992
Time Fixed Effects Y Y Y Y Y
Zip Fixed Effects N N N N Y
R-Squared 0.0294 0.0300 0.0302 0.0227 0.051
Table 1-3 limits the sample to include only zip codes where 51% or more of the population lives in an area
considered by the Census to be Suburban
In Columns I V, the sample is re-specified as follows (Type I through Type V):
(I) Sample includes all ordinary school days from 1980-2001
(II) January, March and May are excluded from the sample 1980-2001
(III) Includes only April, September and October from 1980 2001
(IV) Only April, September and October from 1984 2001
(V) Only April, September and October from 1984 2001, with zip code fixed effects
Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level,
** is significant at the 1% level or smaller










Table 1-4: Rural Subsample Negative Binomial Regression onto Total Crime

Variables (I) (II) (III) (IV) (V)
Strike 0.017 0.027 -0.343 -0.285 -0.392
(0.09) (0.14) (1.45) (1.20) (1.65)

Median Income -0.000034** -0.000036** -0.000036** -0.000033** -
(26.68)** (22.50)** (16.62)** (13.59)** -


Welfare -0.077 -0.063 -0.108 0.117 -
(0.68) (0.45) (0.56) (0.48) -


Poor Parental Educ. -0.282** -0.332** -0.218 -0.200
(3.81)** (3.63)** (1.77) (1.32) -


Juvenile Maleness -0.357** -0.362** -0.370** -0.421** -
(37.69)** (32.83)** (25.81)** (25.64)** -


Student Employment 0.142** 0.156** 0.180** 0.267** -
(4.34)** (3.86)** (3.25)** (4.15)** -

Single Parent House 0.379** 0.401** 0.317* 0.864** -
(4.41)** (3.81)** (2.15)* (5.21)** -

Alpha 2.679** 2.767** 2.810** 2.712** 2.197**
(80.71)** (66.36)** (49.08)** (46.97)** (44.48)**
Number of obs. 1,101,359 729,469 377,244 304,491 304,491
Time Fixed Effects Y Y Y Y Y
Zip Fixed Effects N N N N Y
R-Squared 0.0313 0.0308 0.0316 0.0218 0.0551
Table 1-4 limits the sample to include only zip codes where 51% or more of the population lives in an area
considered by the Census to be Rural
In Columns I V, the sample is re-specified as follows (Type I through Type V):
(I) Sample includes all ordinary school days from 1980-2001
(II) January, March and May are excluded from the sample 1980-2001
(III) Includes only April, September and October from 1980 2001
(IV) Only April, September and October from 1984 2001
(V) Only April, September and October from 1984 2001, with zip code fixed effects
Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level,
** is significant at the 1% level or smaller










Table 1-5: Effects of Strike Days by Crime Type and by Community Type
Strike Strike Strike
Dependant Variable Coefficient for Coefficient for Coefficient for
Urban Sub- Suburban Sub- Rural Sub-
Sample Sample Sample
Drugs and Alcohol 0.200 -1.188 -0.113
(1.36) (1.48) (0.20)

Mischievous 0.536** -0.813 -0.241
(3.88)** (0.73) (0.34)

Property 0.257** -0.128 -0.276
(3.79)** (0.34) (0.80)

Violent -0.364* -1.103 -0.266
(2.50)* (1.44) (0.52)

Weapon and -0.014 -53.096
Endangerment (0.05) (0.02) -
Number of obs. 129,593 37,992 304,491
Time Fixed Effects Y Y Y
Zip Fixed Effects y y Y
Ave. R-Squared 0.0506 0.0423 0.0579
Each row represents a new regression with the listed crime type as the dependant variable. The reported
number is the strike coefficient for the regression with the corresponding dependant variable. Each column
is a different subsample group: urban, suburban and rural. All 14 regressions are defined as a Type V
regression, including data from only April, September and October from 1984 2001, and zip code fixed
effects. The Pseudo R-squared reported at the bottom of each column is the average R-squared over the 5
regressions. Z-statistics are given in the parentheses for every table. Indicates significance at the 5%
level, ** is significant at the 1% level or smaller. The marginal effects of these variables show that violent
crime decreases by 31.53% on strike days. Property crime and mischievous crime increase by 28.81% and
48% respectively. Again, these effects reflect changes in juvenile crime as a proportion of the mean. The
coefficients of the marginal effects are (-0.032), (0.072) and (0.025) for violent, property and mischievous
crime respectively.










Table 1-6: Effects of Strike Days by Crime Type and by Offender Type
Strike
Strike Coefficient Stri
Coefficient for
Dependant Variable for One-Time
Offenders Repeat
Offenders
Total Crime 0.382** 0.156**
(4.05)** (2.77)**

Drugs and Alcohol 0.337 0.153
(1.18) (0.93)

Mischievous 1.024** 0.597**
(3.66)** (3.99)**

Property 0.534** 0.220**
(4.44)** (2.81)**

Violent -0.567 -0.408**
(1.48) (2.59)**

Weapon and -1.330 -0.009
Endangerment (1.20) (0.03)

Gateway Crime -0.286*
-- (2.00)*
Number of obs. 129,593 129,593
Time Fixed Effects Y Y
Zip Fixed Effects Y Y

Table 1-7 includes only urban zip codes. All 13 regressions are defined as a Type V regression, including
data from only April, September and October from 1984 2001, and zip code fixed effects.
Z-statistics are given in the parentheses for every table. Indicates significance at the 5% level,
** is significant at the 1% level or smaller. The marginal effects of these variables show that for repeat
offenders, violent crime decreases by 36.72%, property crime increases by 25.4%, and mischievous crime
increases by 30.82% on strike days. The coefficients of these marginal effects are (-0.032), (0.046) and
(0.012). For one-time offenders property crime increases by 56.75% and mischievous crime increases by
83.11% on strike days. The coefficients of these marginal effects are (0.030) and (0.008). Again, these
effects reflect changes in juvenile crime as a proportion of the mean.











Table 1-7: Temporal Displacement of Crime by Crime Type and by Offender Type
Strike Strike Strike Strike
Weekly Aggregated Dependant Coefficient Coefficient Coefficient for Coefficient
Variable for Full for Urban One-Time for Repeat
Data Set Subsample Offenders Offenders


Total Crime



Mischievous


0.235**
(4.50)**


Property


Violent


0.243**
(4.67)**

0.570**
(4.39)**

0.312**
(4.73)**

-0.300*
(2.32)*


0.418**
(4.51)**

0.820**
(3.09)**

0.492**
(4.27)**

-0.223
(0.73)


0.176**
(3.11)**

0.501**
(3.55)**

0.225**
(3.06)**

-0.317*
(2.27)*


Monthly Aggregated Dependant
Variable


Total Crime


0.534** 0.485**
(3.86)** (3.73)**


Number of obs. 440,640 117,491 117,491 117,491
Time Fixed Effects Y Y Y Y
Y Y Y Y
Zip Fixed Effects

Each row represents a new regression with the listed crime type aggregated at the weekly level as the
dependant variable. The reported number is the strike coefficient for the regression with the corresponding
dependant variable. Each column is a different sample group. Column 1 is the full sample. The remaining
three Columns are from the urban sample. The top 10 regressions are defined as a Type I regressions
including zip code fixed effects. The bottom 2 regression coefficients represent the strike effect on Monthly
Total Crime. Z-statistics are given in the parentheses for every table. Indicates significance at the 5%
level, ** is significant at the 1% level or smaller










Table 1-8: Temporal Displacement of Crime through Lagged Strike Variable
Lagged Strike Variable Strike Coefficient
For Urban Sub-
Sample
1 Day -0.033
(0.62)

2 Days -0.028
(0.54)

3 Days -0.030
(0.57)

4 Days -0.046
(0.87)

5 Days -0.041
(0.78)

6 Days -0.093
(1.75)

7 Days -0.070
(1.33)
Number of obs. 293,112
Time Fixed Effects Y
Zip Fixed Effects Y
Ave. R-Squared 0.0597
Each row represents a new regression with strike variable lagged the by the corresponding row. Each
regression is defined as a Type IV regression including zip code fixed effects of the urban subsample. The
notable exception is that weekends are included to allow for the correct number of strike treatments.
Therefore, days where lagged strike treatments overlap breaks, holidays and other nontypical school days
are included in the sample. Z-statistics are given in the parentheses for every table. Indicates significance
at the 5% level, ** is significant at the 1% level or smaller











Table 1-9: Changes in Total Juvenile Crime, by Days Elapsed since the Start of a Strike
Partitioned Strike Partitioned Strike
Variable Strike Coefficient Variable Strike Coefficient


Days 1 3



Days 4 6



Days 7 9



Day 10+


Alpha


-0.005
(0.06)

0.185
(1.91)

0.261*
(2.54)*

0.313**
(2.93)**


Days 1 2



Days 3 4



Days 5 6



Days 7 8



Days 9 10



Day 11+


0.533**
(101.05)**


Alpha


0.058
(0.58)

0.006
(0.07)

0.262*
(2.37)*

0.284**
(2.62)**

0.228
(0.98)

0.305**
(2.67)**


0.533**
(101.04)**


Number of obs. 454,953 Number of obs. 454,953
Time Fixed Effects Y Time Fixed Effects Y
Zip Fixed Effects Y Zip Fixed Effects Y
R-Squared 0.0713 R-Squared 0.0713

Table 1-8 includes only urban zip codes. These two regression are defined as a Type I regressions, using
data including all ordinary school days from 1980-2001, and zip code fixed effects. Both regressions use
Total Crime as their dependant variable. Z-statistics are given in the parentheses for every table. *
Indicates significance at the 5% level, ** is significant at the 1% level or smaller





























C-
0)




LO











LO











q-




0-


0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
DaysToCapture


Figure 1-1: Average Percent of Total Arrests as Function of Days between Offense and
Capture A) NonStrike Days for Zip Codes Where Strikes Occur, B) Strike
Days for Zip Codes Where Strikes Occur


I I I I I I I I I I I I I I I
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
DaysToCapture











B











Table 1-10: Effects of Strike Days with Specific Years Dropped from the Sample
Year Dropped Strike Coefficient Year Dropped Strike Coefficient


1984


1985


1986


1987


1988


1989


1990


1991


1992


0.164**
(3.39)**

0.093
(1.70)

0.166**
(3.43)**

0.160**
(3.16)**

0.163**
(3.37)**

0.148**
(3.02)**

0.189**
(3.77)**

0.229**
(3.18)**

0.157**
(3.24)**


1993


1994


1995


1996


1997


1998


1999


2000


2001


0.166**
(3.41)**

0.163**
(3.28)**

0.166**
(3.40)**

0.157**
(3.23)**

0.153**
(3.16)**

0.156**
(3.23)**

0.157**
(3.26)**

0.144**
(2.99)**

0.150**
(3.11)**


Without Seattle 0.133* Seattle Only 0.153
(2.04)* (1.82)


Ave.Number of obs. 358274 358274
Time Fixed Effects Y Y
Zip Fixed Effects Y Y
Ave. R-Squared 0.0603 0.0603
The top half of this table is a regression of Strikes on Total Crime with each year systematically dropped
from the sample. Each regression includes only urban zip codes. All 18 regressions are defined as a Type I
regressions, using data including all ordinary school days from 1984-2001, and zip code fixed effects. The
bottom half of the table reports the effect of strike days, for the same Type I sample, with Seattle dropped
from the sample, and with Seattle making up the entire sample. The bottom of the table shows the average
number of observations for the top 18 regressions, as well as the average R-squared term. Z-statistics are
given in the parentheses for every table. Indicates significance at the 5% level,** is significant at the 1%
level or smaller.










Table 1-11: Average Effect of Randomly Generated Strike Days on Total Crime
Dependant Variable: Total Random Generations of
Crime Strike Days

Ave. Random Strike 0.002
Coefficient (0.05)

Average Number 2,355
of Strike Days

Number of Trials 25

Number of obs. 1,703,910
Time Fixed Effects Y
Zip Fixed Effects N
Ave. R-Squared 0.0380

Table 1-12: Effects of Strike Days by Offender Type with the Last 4 Years of the Sample
Dropped
Strike
Strike Coefficient Stri
Coefficient for
Dependant Variable for One-Time
Repeat
Offenders Repeat
Offenders

Total Crime 0.382** 0.156**
(4.05)** (2.77)**

Total Crime without 0.383** 0.128*
1998-2001 (4.08)** (2.31)*

Number of obs. 129,593 129,593
Time Fixed Effects Y Y
Zip Fixed Effects Y Y

Table 1-11 is the average strike coefficient over 25 trials using a randomly generated strike variable. It
includes the full data set. Table 1-12 is a Type V regression using just the urban subsample, and includes
zip code fixed effects. Z-statistics are given in the parentheses for every table. Indicates significance at
the 5% level, ** is significant at the 1% level or smaller















CHAPTER 2
THE ROLE OF TEACHER NETWORKING IN TEACHER TRANSFER DECISIONS
AND TEACHER MOBILITY

2.1 Introduction

In recent years, there has been growing research to support the notion that teacher

movement is an important topic, and not one to be overlooked. Thus far, research on

teacher sorting behavior has shown that student characteristics seem to influence a

teacher's transfer decision (Hanushek, Kain, Rivkin 2001). More specifically, teachers

sort away from relatively low-income, low-achieving schools. As a result, less

experienced teachers are more typically found in poor, nonwhite, low-performing

schools, particularly those in urban areas (Lankford, Loeb, Wyckoff 2002). This is most

important because teacher experience seems to be directly linked to student achievement

(Hanushek, Kain, Rivkin 1998, 1999)1.

In addition to student characteristics, other characteristics (such as salary

differences) have been studied in an effort to better understand the factors that influence

teacher movement. Research aimed at identifying specific characteristics and conditions

that promote teacher mobility have given more insight into the questions of when and

why teachers move, however researchers still do not have the complete picture on how

teachers move and why.




1 Hanuchek, Kain and Rivkin (1998) show that variations in teacher quality account for 7.5% of the total
variation in student achievement. In addition, teacher quality differences tend to outweigh school quality
differences considerably.









The focus of this paper is to better understand the nature of teacher transfers, with

special attention given to teacher intradistrict transfers from school to school.

Specifically this paper will identify the use of Master Inservice workshops aimed at

professional development as a powerful means of building teacher networks. Ultimately

I test the hypothesis that Inservice workshops give teachers greater opportunities to

network with teachers from other schools in their district, and therefore result in a higher

rate of teacher transfer. The evidence will show that a one standard deviation increase in

workshop attendance contributes to a 12.5% increase in teacher transfer rates per school.

Thus teacher networks are a crucial element in influencing teacher transfer rates.

Section 2.2 of this paper will provide a context for the basic assumptions outlying

the arguments set forth, Section 2.3 will specify the model therein, Section 2.4 reports the

preliminary results, Section 2.5 will check the validity of the results, and Section 2.6 will

make the concluding remarks.

2.2 Discussion of Networking

2.2.1 How Teachers Network

The idea that professionals develop networks is not a new concept in the social

sciences. The term "networking" spans across many professions and takes many

different forms. Like many professionals, teachers also engage in networking for various

reasons I will address later in this section. First I will address the issue of how teachers

build networks. Clearly some teacher networking takes place in private. Teachers may

spend recreational time together getting to know one another, or they may talk over

lunch. However these examples of the networking process, though valid, are not feasibly

quantified or qualified. In order to quantify teacher networks, I rely on teacher Inservice

activity to determine if teacher networks aid in mobility. Specifically I focus on the









number of hours teachers spend in Inservice activity as my measure of network

size/strength. I argue that Master Inservice workshops, a part of the teacher re-

certification process, provide an effective forum for networking. To understand why this

is the case, it is necessary to first review how the re-certification process works.

Every teacher in the state of Florida is required to have an up-to-date license that

enables him or her to teach in Florida public schools. This license is earned by every

teacher at the beginning of his or her career and must be re-certified every five years.

Teachers earn their re-certification through any combination of three things: they may

take state-issued Florida Subject Area Exam in appropriate fields to prove their level of

competency, they may enroll in and complete required coursework (6 credits minimum)

at an approved university or college, and/or they may choose to complete up to 120 hours

of Master Inservice courses/workshops offered by the school district.

This state mandated re-certification process provides an ideal environment for

testing the effects of teacher networking. Teachers who take the Florida Subject Area

Exam in their appropriate fields or who decide to take classes at a College or University

have little to no interaction their colleagues in their school districts. However, the Master

Inservice option is quite different from other re-certification options. First, Inservice

courses/workshops are almost exclusively made up of other teachers, all within the same

district and from differing schools.2 Thus teachers interact with others who share a

common profession, and often common interests. Secondly, the environment of most

workshops is designed in such a way that encourages interaction among participants.


2 It is possible that a marginal number of administrators, county officials, etc. attend these teacher
components for other reasons (professional or personal). They are not included in the Inservice data in this
paper.









Workshop assemblies are long in duration (often several hours per session) and can span

weeks or months. In addition, individuals are often called upon to share experiences with

all the participants, or present material regarding how the workshop has influenced their

classroom activity. Thirdly, Inservice is an attendance based program, as opposed to the

graded college coursework or graded test format. One could argue that the relatively

stress-free nature of the Inservice allows teachers to spend more time getting to know

each other, free from the distraction of grade requirements.

2.2.2 How Teachers Use Their Networks

There are many reasons why any professional or businessperson would network

with others in their field of work. They yearn to share ideas, understand their

competition, and exchange information that can benefit their personal interests,

sometimes in a mutual way. Why teachers network with each other is not so different.

They exchange classroom experiences and teaching methods, information on workplace

environments, as well as build personal bonds that enhance their feelings towards the

profession, or even their personal lives. As a result they could possibly use their networks

to plan intradistrict school events, find work in the summer, or even to socialize.

However, there are specific applications of strong teacher networks that would

enhance mobility within a district. First, teachers use their networks to gain information

about the various student and administrative characteristics of schools in their district.

Greater access to this information lowers the cost of job searching and enhances mobility.

Secondly, teachers use their networks to reveal their teacher quality in a less costly way

for the purpose of being hired at other schools within their own district. If teachers do

use their networks to sort into preferred schools in a more efficient, less costly way, then









there are obvious implications for student outcomes, since others have successfully linked

student outcomes to teacher mobility.

2.3 The Basic Model

A teacher's desire to transfer to school n (new school) is based on the utility they

receive at school n minus the utility they receive at school o (old school). The utility at

the old school is given by:

Vo(so, qo, ao, do, o)

where o indexes the old school, So is the salary of the old school, qo is a measure of

student/classroom quality, ao (administration quality), do (distance from school to home),

and o (opportunity cost of teaching). Similarly the utility at the new school is given by:

Vn(sn, qn, an, dn, o)

where n indexes the new school, sn is the salary of the new school, qn is a measure of

student/classroom quality, an (administration quality), dn (distance from school to home),

and o (opportunity cost of teaching). Teachers incur explicit costs to movement. These

costs arise from the search and application process. More specifically, teachers may have

to submit paperwork, spend some time interviewing, or generally gather information

about the school they are transferring to. Thus I write this teacher decision as:

Transfer when: Vn Vo C(in, rn, nj*) > 0

Do not transfer when: Vn Vo C(i, rn, nj*) < 0

In this case C(.) represents the explicit cost borne by the teacher, where in is the amount

of relevant school information a teacher desires to gather, rn represents district and school

requirements for transfers (paperwork, interviews, etc.), and nj* representing the network









size of the jth individual. I further note the relationship of each parameter to this explicit

cost function.

It is reasonable to assume that the less a teacher knows about a prospective school,

the more time he/she spends gathering relevant information. Therefore we expect an

increase in necessary or relevant information (in) to translate to a direct increase in C(.),

such that 8C/8i, > 0. In addition an increase in district and/or school requirements for

transfer will increase time and resources spent towards transfer, such that CC/Or, > 0.

When looking at network size it becomes apparent that a larger network size will lend

itself to better flow of school information, possibly fewer or shorter interviews, and more

inside information that helps one to eliminate unnecessary measures. Therefore it seems

logical that a larger network will decrease external cost, such that 8C/cnj* < 0. Lastly we

must qualify the network to external cost relationship as not being strict. This allows for

the possibility that for school n, one's network is totally ineffective. That is to say that

one's network is not the relevant network, i.e., a network in another state. From these

equations it is clear why teachers may be more likely to transfer as their relevant network

size increases

2.4 Data and Empirical Study

My data for my study comes from a variety of sources. I utilize individual school

data taken from the "Florida School Indicators Report" (FSIR) and the "School Advisory

Council Report" (SACR). Countywide data was also supplied in various forms from the

Florida State Department of Education (FLDOE). These forms include the "FLDOE

State Survey #2", "Teacher Exit Interview Information", an "Inservice Hours Report",

"The Profiles of Florida District Schools (Students and Staff Data)", and "Teacher Salary,









Experience and Degree Level". The sample includes 67 observations, one for each

Florida school district.

Because several variables introduced in the following section are highly collinear to

district size, such variables have been defined on a per school basis.3

2.4.1 Defining the Parameters

The dependent variable for my study (Transfers) is the number of transfers within

a district per school. In order to approximate the closeness of different school job

options, or capture the density of school job choice in a district, I use the total area,

expressed in sq. miles, divided by the number of schools in the district (School

Distance). Since a denser school district gives a mobile teacher more viable, less costly

job options, I expect this variable to be positively correlated with the total number of

transfers per school.

Student/classroom quality is measured by taking the standard deviation of the

percentage of free lunch eligible students among all schools in a given district. This

measure (School Difference) should capture the variation of the percentage of the student

body that is of low socio-economic status (SES) across schools. In essence, this variable

measures how schools in a district differ along family backgrounds. Since greater

variation in work environments should give teachers a greater incentive to sort, this

variable should be positively correlated with transfers per school.

Wage differences may also play an important role in teacher transfer decisions.

There are no wage differences among public schools in a given district, but wage

differences among districts still exist and are important to account for. As such, Salary


3Transfers, Network, Dislike Supervisor, Principal Moves, School Changes.









Difference is defined as the average salary of the highest paying district directly adjacent

to the current district minus the average salary of the district in which a teacher is

currently employed. If a teacher considers salary difference in a move between districts,

it is plausible that they may consider the first-best alternative salary difference above all

else. If higher salaries provide incentives for teachers to leave their district rather than

move within district, then there should be less movement between schools within district.

Therefore this variable should be negatively correlated with within-district movement.

In addition to the interdistrict salary difference, a wage concern for teachers should

also be the wage forgone from remaining a teacher. To capture the opportunity wage

forgone, I construct a variable (Wage Ratio) using a county-level wage index. This

wage index reflects the average wages across a predetermined set of occupations for each

county. By dividing average salary for a district by its county wage index, I measure how

well teachers in a county are paid relative to other occupations. The larger the Wage

Ratio, the less appealing otherjobs are. As a result this variable should be positively

correlated with teacher transfers.

My measure of network size (Network) equals total hours spent in Inservice per

school.4 Since more hours spent in Inservice should lead to larger and more effective

networks, this variable should be positively correlated with teacher transfers per school.

It is also true that the Network variable is positively correlated with larger school

size. This is a potential problem since school size, measured in average number of

teacher faculty per school, is positively correlated with district size itself. That is to say



4Component hours geared towards non-instructional staff (i.e., food service employees, transportation staff,
administrative staff, substitute teachers, etc.) are not included in this measurement. Since each district
adheres to a statewide numbering standard for component identification, I can be sure that I only include
components which were teacher-oriented.









that as school districts grow, the size of the schools as measured in total faculty also

grows. Since teachers in larger districts (with larger schools) have more positions to

choose from, they should have greater transfer opportunities. To prevent this relationship

from biasing any estimates of the Network variable, I create a variable to control for the

effect of larger school size. This variable, School Size, measures the average teaching

staff per school in the a district.

New public schools opening and old public schools closing also present teachers

with new choices in making transfer decisions. New schools that open in a district may

provide up-to-date facilities and technology in the classroom, making teaching there more

desirable. School closures force teachers to reorganize into other schools or leave

teaching entirely. The variable (School Changes) captures these elements of school

turnover. This variable is measured by assigning a value of positive one for every school

gained and/or lost from the 2000-01 school year to the 2001-02 school year, and then

summing the totals for each county and dividing by the number of schools. Since school

districts where schools open and/or close should result in more teacher transfers, this

variable should be positively correlated with transfers.

There is another variable emphasized in this model that is based on administrative

movement, specifically movement between schools by existing principals. When

principals transfer from school to school within a district, positive teacher/principal

relationships may induce principals to recruit teachers to work at a new assignment. In

essence principals may bring their favorite teachers with them to another school under the

expectation/promise of a positive work environment. This variable (Principal Moves) is

measured as the total number of principals who moved from one school to another









school, divided by the number of schools. If principals take teachers with them when

they move between schools, then principal transfers should be positively correlated with

teacher transfers.

Another variable that should affect teacher movement within a district is the

number of relevant alternatives. It is arguably difficult for teachers to move between

high schools and elementary/middle schools. Obviously high school teachers who work

in districts where there is only one high school cannot move to another high school in

their district. As a result, one would expect that school districts with only one high

school will have an inherently lower rate of teacher transfer compared to similar districts

with multiple high schools. To account for this disparity I include a dummy variable

(One HS) that receives a value of 1 when a school district has only one high school (zero

otherwise). This variable should be negatively correlated with teacher transfers.

The last two variables I include in the regression help to describe the state of a

district's preferential treatment towards transferring existing teachers to fill new

vacancies when they open, rather than hiring new teachers. Some districts require that

open positions are filled with teachers already employed before a new teacher may be

hired; while some other districts give no special consideration to existing teachers over

new applicants. Since more relaxed or favorable policies towards teacher transfers may

enhance teacher mobility, I introduce two dummy variables that describe the priority

given to teacher transfers. The first variable (Transfer PC) describes a district that

affords partial consideration to existing teachers. This variable takes on a value of 1

when a district gives preferential treatment, but not exclusive privilege, to its teacher

transfers (zero otherwise). The second dummy variable (Transfer CC) describes a









district's policy to necessarily fill a new position with a teacher transfer request before

considering a new applicant. This variable takes on a value of 1 when a district offers

complete consideration to existing teachers before all others (zero otherwise). In both

cases, these variables should be positively correlated with transfers per school to reflect

the enhanced opportunity for mobility given by preference.

2.4.2 Empirical Testing

The analysis begins with a simple OLS regression onto the dependent variable

Transfers. The results are expressed in Table 2-1, listed by coefficient with the relevant

t-statistic listed in parenthesis.

Columns 1 and 2 express the OLS regression with our 10 major variables of

interest. Some of these variables that do not have the predicted sign, in this case School

Distance, Transfer PC and Transfer CC, and are not significantly different from zero.

Of the variables that are significant, (Network, School Difference, School Changes and

School Size), each has the predicted sign.

The positive and significant coefficient of the Network variable in Column 2

seems to suggest that networks exist, and that in fact they do serve to enhance teacher

intradistrict mobility. As a teacher networks grows, teacher mobility rises. This provides

evidence that teachers are using their networks to gain information and enhance the

transfer process. To determine the magnitude of the impact of networking on transfer

rates, I look at the percent change in the average number of teacher transfers per school

when the amount of networking increases by one standard deviation. I find that an

increase in the number of inservice hours per school by approximately 1000 hours

increases the teacher transfer rate per school by approximately 11.49%. These results can

be viewed in Table 2-2.









The results in Table 2-1 also suggest that student characteristics, school size, and

schools opening and closing all influence teacher movement. Teachers seem to sort more

frequently in districts where variation in student quality across schools is high. The

variable School Difference has a significant impact in teacher movement and shows that

teachers are more likely to move around, or sort, when there are sizeable differences

among student populations at different schools. According to the regression, a one

standard deviation increase in the measure of school heterogeneity implies a nearly

15.25% increase in teacher transfers per school

Teachers also seem to be more mobile in districts where schools are larger. The

School Size variable is positive and significant, suggesting teachers face greater

opportunity to sort when there are more total positions in a school. The coefficient of the

School Size variable implies that a one standard deviation increase in the average size of

a school leads to a roughly 17.85% increase in transfer rates per school.

Besides student and school characteristics, teachers also have greater mobility in

districts where schools are opening and closing with greater frequency. The School

Changes variable suggests that as schools open and close in a district, teacher mobility

rises. One would expect that as schools open in a district, teachers seek new facilities,

new classrooms, etc. Further, as schools close teachers are forced to shuffle and sort to

other schools so that they may continue teaching in their district. The evidence suggests

that when 5 schools are added in a district with 100 schools, teacher transfers increase by

nearly 30% for every school in the district.

Column 2 runs the same regression using robust standard errors. Using these

robust standard errors helps to ease worries about heteroskedasticity in the regression. If









it were true that smaller districts have smaller measurement error than larger districts

(which seems plausible), then robust standard errors help to control for this problem. We

see roughly the same results as the regression in Column 1. School Size and School

Difference have the predicted sign and are significant at the 1% level. School Changes

also has the predicted sign and is significant at the 0.05% level.

Column 3 reports the same regression as in Column 1 except that one additional

variable is included. This variable (Dislike Supervisor) is a rough measure of teacher

dissatisfaction with administrative oversight in one's own school. It is calculated for

each district as the number of teachers per school who cited "dissatisfaction with

supervisor" as a reason for voluntarily terminating their employment on the Teacher Exit

Interview from the 2000-2001 school year.5 If teacher discontent with administration

leads to greater mobility, then this variable should be positively correlated with teacher

transfers.

In Column 3, the variable Dislike Supervisor has the predicted sign but is not

significant. It suggests that job environments and teacher attitudes are not necessarily an

important part of the movement decision. This result seems to defy intuition, however in

the next section I will show later that this result is marginal, and that Dislike Supervisor

is sensitive to model specification. Column 4 reports the same regression as in Column 3

using robust standard errors. It is also worth noting that the Dislike Supervisor variable

is not an ideal measure of overall teacher attitudes towards administration because it does

not include teachers who continued to teach in their own districts and only reflects a self-


5 This survey includes teachers who continued to teach in other districts or other states, teachers who stayed
employed in education departments in and outside of their district, teachers who left the education
profession altogether, those who retired, and teachers who left for private schools in and out of their
district.









selecting subgroup of teachers. It does however have some benefits. First, since there is

no descriptive teacher evaluation of administration formally collected by the FLDOE or

school districts, nor is there useable county data of the same nature, this variable offers at

least some approximation of what that data might yield. Secondly, because the teachers

in these interviews have already committed themselves to alternative employment, the

responses given in these interviews are assured to be candid and honest. It is unlikely

that a teacher respondent fears having their identity becoming known to others in these

interviews.

Overall, these results seem at least superficially consistent with Hanushek, Kain

and Rivkin findings (1999, 2001) that student characteristics influence mobility, and that

salary changes at best have only a modest impact. However none of these variables seem

to have a powerful an impact as the School Changes variable. There seems to be a large

disparity between the predictive power of this variable, and other variables. To ensure

that our results are valid, and to understand why each variable induces the kind of change

it does, I check the robustness of the data.

2.5 Robustness Checks

2.5.1 Re-examining the Data

With respect to the School Changes variable, one observation in the dataset

experienced an abnormally high rate of school change between the years of 2000-2001

and 2001-2002. This county in particular closed two school facilities to open one, much

larger school, that would incorporate all of the existing student population of the other

schools, however the small size of this district seems to exacerbate the measurement of

the School Changes variable. It is possible that this one observation alone is driving the

results of the School Changes variable. If this observation is more of an outlier, then it









may be overstating the overall importance of the variable it influences. An OLS

regression analysis without this outlier is reported in Table 2-3.

Once the sample is modified to eliminate the possible outlier, the magnitude of the

coefficient of the School Changes variable is reduced by almost half, though the variable

remains significant. This does not necessarily mean that this variable is not important,

however it does show that perhaps this outlier should be dropped to maintain the integrity

of these results. These new results suggest that a one standard deviation increase in

school openings/closings leads to a 14.55% increase in teacher transfers per school. This

change is considerably lower than the previous estimate.

All other variables in the regression have the predicted sign and all of the formerly

significant variables are still significant. Network and School Size are significant at the

1% level, while School Difference becomes marginally significant at the 10% level.

Column 2 reports the same regression with robust standard errors. Column 3 includes the

Dislike Supervisor variable, which is significant at the 10% level. This result suggests

that there may be some value to job environment and administrative quality that teachers

consider in their movement decisions. Finally Column 4 presents the same regression as

in Column 3, including robust standard errors. I again assess the impact of each variable

independently, by determining the effect of an increase of one standard deviation of each

variable. These results can be viewed in Table 2-4. With the exception of School

Changes, there are no substantial differences in the magnitudes of the coefficients of the

other variables.

The key variable of interest, Network, is positive and significant in each regression

specification on Table 2-3, suggesting that networks are an important factor contributing









to teacher movement. However, to confirm that the results are credible, I further test the

robustness of these results. Specifically I address whether the variable itself is

endogenous to the system. In the following section I test whether the measurement of

networking is correlated with some unobserved teacher characteristicss.

2.5.2 Unobserved Correlation

If my measure of teacher networks is correlated with some unobserved teacher

characteristicss, then it may be true that there is no real networking taking place.

Instead, teachers who complete high levels of Inservice may simply be more likely to

transfer for other reasons. If that were true then one should expect that teachers who

complete Inservice would also be just as likely to move in ways other than just

intradistrict transfers. To test whether the Network variable is reflecting true networking

or is just reflecting self-selection of highly mobile teachers into Inservice, I utilize new

dependent variables that depict other types of teacher movement.

Private Move equals the number of teachers who claim a move into private

schools per school. The second variable, District Move, measures the number of

teachers moving into other Florida districts per school. The last variable is teacher

movement to other states to teach. This variable, State Move, equals the number of

teachers moving to other states per school in the district. Regression results with these

new dependent variables are reported in Table 2-5 using robust standard errors.

Columns 1 and 2 report the regressions with Private Move as the new dependent

variable, and using robust standard errors. Columns 3 and 4 do the same with District

Move as the new dependent variable, and Columns 5 and 6 have State Move as the new

dependent variable. In each of these specifications, network size does not seem to be a

significant factor influencing teacher movement. Since Inservice is not significant to any









of these different kinds of movement, it seems plausible that teachers who tend to move

around more generally, are not self-select into attending a greater amount of Inservice.

Thus I can be more confident that the positive effect of networking is not simply a

reflection of the preferences of more mobile teachers.

2.6 Conclusion

The evidence set forth in this paper supports the argument that Master Inservice

components provide an effective environment for teachers to network with each other.

These networks seem to provide a basis for teachers to gather important information

about schools and allow teachers access to transfer options that might have previously

been closed to them. Overall, after controlling for determinants of mobility and

eliminating biasing outlier effects, I show that an increase of one standard deviation in

Inservice hours causes roughly a 12.74% increase in teacher transfers per school. The

results suggest that teachers exploit these networks to sort into schools that they find

desirable. If teachers can more effectively sort away from less desirable schools, such as

low-income or failing schools, into more desirable schools, then it is easy to conceive that

worse schools will be worse off as teacher networks improve.

Of course the goal of the Inservice plans set forth by school districts is to improve

teacher quality. However, the unintended consequences of such staff development

policies that encourage, or at least facilitate, teacher networking seems to exist as well.

Further, the consequence of this particular policy seems to be quite substantial in

promoting a potentially harmful action such as teacher transfer. This result may have

considerable meaning for the future of public education policy. State officials must begin

to weigh the possible negative effects of increased teacher networking with the potential






59


(and realized) benefits of Inservice, so that such policy analysis can be conducted in the

future with a complete understanding of social welfare consequence.












Table 2-1: Ordinary Least Squares Regression onto teacher transfers using full sample
with and without robust standard errors


Variable


Dependent

Network


School Distance


Principal Moves


School Changes


School Difference


Salary Difference


Wage Ratio


School Size


Dislike Supervisor


One HS


Transfer Policy
PC

Transfer Policy
CC


Constant


Column 1

Transfers

0.0003261
(1.56)

0.0065532
(1.12)

0.3216886
(0.12)

17.86031
(4.88)**

0.0584747
(2.10)**

-0.0000891
(1.61)

-0.0018573
(0.27)

0.0501213
(2.46)**


-0.7262709
(1.32)

-0.2608151
(0.65)

-0.2381146
(0.57)

0.2509715
(0.09)


Column 2

Transfers

0.0003261
(2.00)**

0.0065532
(1.16)

0.3216886
(0.12)

17.86031
(3.42)**

0.0584747
(2.34)**

-0.0000891
(1.50)

-0.0018573
(0.22)

0.0501213
(2.80)**


-0.7262709
(1.27)

-0.2608151
(0.64)

-0.2381146
(0.67)

0.2509715
(0.08)


Column 3

Transfers

0.0003474
(1.67)

0.0078015
(1.33)

0.9270341
(0.34)

17.55322
(4.84)**

0.0557304
(2.01)**

-0.0000766
(1.38)

0.0001112
(0.02)

0.0448708
(2.19)**

4.17856
(1.43)

-0.7582897
(1.39)

-0.2026472
(0.51)

-0.1001645
(0.24)

-0.5541842
(0.20)


Column 4

Transfers

0.0003474
(2.19)**

0.0078015
(1.33)

0.9270341
(0.33)

17.55322
(3.20)**

0.0557304
(2.16)**

-0.0000766
(1.21)

0.0001112
(0.02)

0.0448708
(2.96)**

4.17856
(1.17)

-0.7582897
(1.31)

-0.2026472
(0.48)

-0.1001645
(0.25)

-0.5541842
(0.19)


R-Squared 0.6154 0.6154 0.6295 0.6295
Number of Obs. 67 67 67 67
Robust SE N Y N Y
denotes significance at the 10% level
**denotes significance at the 5% level






61


Table 2-2: Percent change teacher transfers per school for a one standard deviation
increase in each significant variable with full sample
Variable Mean Standard Deviation Range of % Change
School Changes 0.027 0.05 30.06 to 30.58 %
School Size 33.418 11.61 17.85 to 19.94 %
School Difference 15.945 7.99 15.25 to 16.00 %
Network 1758.317 1047.62 11.49 to 12.46 %













Table 2-3: Ordinary Least Squares Regression onto teacher transfers using sample with


biasin outlier omitted
Variable Column 1


Dependent

Network


School Distance


Principal Moves


School Changes


School Difference


Salary Difference


Wage Ratio


School Size


Dislike Supervisor


One HS


Transfer Policy PC


Transfer Policy CC


Constant

R-Squared
Number of Obs.
Robust SE


Transfers
0.0003563
(1.79)*

0.0010389
(0.18)

0.3430332
(0.13)

9.11579
(1.90)*

0.0433008
(1.60)

-0.0000671
(1.26)

-0.0000751
(0.01)

0.0502127
(2.60)**




-0.4594238
(0.86)

-0.2247365
(0.59)

-0.3207089
(0.81)

0.0560173
(0.02)

0.6033
66
N


Column 2
Transfers
0.0003563
(2.22)**

0.0010389
(0.20)

0.3430332
(0.12)

9.11579
(1.91)*

0.0433008
(1.74)*

-0.0000671
(1.06)

-0.0000751
(0.01)

0.0502127
(2.89)***




-0.4594238
(0.94)

-0.2247365
(0.55)

-0.3207089
(0.89)

0.0560173
(0.02)

0.6033
66
Y


Column 3
Transfers
0.0003822
(1.95)*

0.0021892
(0.37)

1.147493
(0.46)

8.322631
(1.76)*

0.0393933
(1.48)

-0.0000516
(0.97)

0.0020135
(0.30)

0.0442077
(2.29)**

4.78275
(1.74)*

-0.4825945
(0.92)

-0.1563356
(0.41)

-0.1669839
(0.42)

-0.8754046
(0.33)

0.6247
66
N


Column 4
Transfers
0.0003822
(2.46)**

0.0021892
(0.41)

1.147493
(0.36)

8.322631
(1.73)*

0.0393933
(1.55)

-0.0000516
(0.79)

0.0020135
(0.30)

0.0442077
(2.95)***

4.78275
(1.52)

-0.4825945
(0.96)

-0.1563356
(0.37)

-0.1669839
(0.42)

-0.8754046
(0.33)

0.6247
66
Y






63


Table 2-4: Outlier excluded percent change teacher transfers per school for a one
standard deviation increase in each significant variable
Variable Mean Standard Deviation Range of % Change
School Size 33.696 11.48 17.40 to 19.73 %
Network 1777.904 1043.22 12.73 to 13.65 %
School Difference 15.881 8.03 10.84 to 11.91 %
School Changes 0.023 0.04 0.37 to 11.36 %
Dislike Supervisor 0.027 0.05 8.78 %







64


Table 2-5: Ordinary Least Squares Regression of other types of teacher movements with
outlier excluded


Variable


Dependent


Network


School Distance


Principal Moves


School Changes


School
Difference


Salary Difference


Wage Ratio


School Size


Dislike
Supervisor


One HS


Transfer Policy
PC

Transfer Policy
CC


Constant

R-Squared

Number of Obs.

Robust SE


Column 1 Column 2 Column 3 Column 4


Private
Move


Private District District
Move Move Move


Column 5 Column 6

State Move State Move


0.0000101 0.0000104 0.0000342 0.0000305 0.0000014 0.0000034
(1.21) (1.22) (0.49) (0.43) (0.06) (0.01)

-0.000214 -0.000196 0.0024853 0.0021902 -0.000891 -0.000813
(1.09) (1.02) (0.95) (0.85) (1.36) (1.22)

-0.001503 0.0073589 -1.829443 -1.930217 -0.190106 -0.142873
(0.02) (0.07) (2.09)** (2.01)** (0.53) (0.39)

0.125987 0.1214916 4.121108 4.236288 0.6497598 0.5957743
(1.11) (1.07) (1.44) (1.46) (0.90) (0.78)

-0.001018 -0.001058 0.0119132 0.0124806 0.0015131 0.0012472
(0.73) (0.74) (0.99) (0.99) (0.44) (0.35)

-0.000001 -0.000001 -0.000001 -0.000003 -0.000002 -0.000001
(0.43) (0.31) (0.06) (0.18) (0.37) (0.20)

0.0002086 0.0002341 -0.006608 -0.006912 -0.002273 -0.002131
(0.96) (1.02) (2.47)** (2.35)** (2.41)** (2.18)**

-0.000328 -0.000405 0.0019503 0.0028223 0.0017899 0.0013812
(0.42) (0.51) (0.32) (0.47) (0.76) (0.58)


0.0611693
(0.66)


-0.694535
(0.70)


0.3255332
(1.10)


-0.014332 -0.014801 0.2299846 0.2333494 0.0774991 0.0759221
(1.04) (1.08) (0.87) (0.88) (0.95) (0.91)

0.0199595 0.0197389 -0.111869 -0.121802 -0.007048 -0.002392
(1.06) (1.05) (0.73) (0.81) (0.17) (0.06)

0.0022509 0.0042704 -0.174603 -0.196926 -0.050714 -0.040251
(0.12) (0.21) (1.20) (1.29) (1.04) (0.78)


-0.041984
(0.50)

0.1931
66


-0.053770
(0.59)

0.1977
66


2.621647
(2.52)**

0.3416
66

Y
N


2.756905
(2.38)**

0.3461
66


1.011499
(2.84)**

0.2398
66


0.9481022
(2.54)**

0.2511
66


N N


Outlier














CHAPTER 3
WITH A LITTLE HELP FROM MY FRIENDS: EVIDENCE OF TEACHER
NETWORKS USING MICRO DATA

3.1 Introduction

In the last several years, the issue of teacher mobility has come to the foreground of

discussions about the public education. Parents and families who value teacher quality

have come to realize that teacher mobility is a topic that has potentially far reaching

consequences, not just for education labor markets, but also in terms of levels of

education and student achievement. Schools that struggle with teacher retention often

find job vacancies difficult to fill, and may compensate by hiring less than fully qualified

teachers, expanding class sizes, canceling course offerings and assigning teachers from

other subject areas (NCES 1997). Of course these kinds of actions may have adverse

effects on student learning and achievement. As a result, economists and policy makers

have done more to explore the issues of teacher attrition and movement within school

districts. The goals of such recent studies have been to shed light on the questions of

which teachers are the most mobile, how do these teachers move, when do these

movements take place, and what are the potential implications for students.

The most recent studies focused on teacher transfer behavior and sorting within

education have provided evidence that supports the notion that student characteristics and

school quality are important factors in a teacher's transfer decision (Hanushek, Kain, and

Rivkin 2001). This also seems consistent with teacher self-reported information. In the

2000-2001 Teacher Follow-up Survey, administered by the National Center of Education









Statistics, 32% of teachers reported poor workplace conditions as a primary reason for the

movement from one school to another. Research shows that teachers sort away from

low-income, low-achieving schools. Consequently, less experienced teachers are more

typically found in poor, nonwhite, low-performing schools, particularly those in urban

areas (Lankford, Loeb, and Wyckoff 2002).

For policymakers, this can be a troubling result. If more experienced teachers tend

to sort away from poorly performing schools, then it is the students at these schools who

stand to lose the most, because experienced teachers have been found to be more

effective than novice teachers in terms of higher student achievement (Hanushek, Kain,

and Rivkin 1998, 1999).1 In addition, urban schools and highly urban school districts

face high rates of teacher turnover (Imazeki 2003). If teachers are sorting away from

these areas, then these urban districts may have the most difficult time filling vacancies

with replacement staff.

Aside from student characteristics, economists have also studied the potential

effects of salary changes on teacher mobility. So far researchers have found that salary

levels have only a modest impact in terms of student achievement and almost no impact

on teacher mobility (Hanushek, Kain, and Rivkin 1999). However, others have also

found that salary can influence a teacher's movement choice in ways other than strict

transfer, such as length of stay in teaching (Murnane and Olsen 1990). Although much

has been learned about teacher mobility, researchers still do not have the complete picture

on how teachers move and why.



1 Hanushek, Kain and Rivkin (1998) show that variations in teacher quality account for 7.5% of the total
variation in student achievement. In addition, teacher quality differences tend to outweigh school quality
differences considerably.









The focus of this paper is to better understand the nature of teacher transfers from

school to school by introducing the idea of teacher networks as an effective sorting

mechanism for public school teachers. To study this issue I rely on information provided

by the 1999-2000 Schools and Staffing Survey (SASS) put out by the National Center for

Education Statistics. The individually detailed and specific nature of the dataset not only

allows me to consider how individual networks affect mobility but also allows me to

control for a well-defined set of variables, giving my study a more complete and detailed

look at networks. This paper will begin discussion on the notion of teacher networks, and

will address how these networks influence teacher movement and to what extent.

Specifically this paper describes how professional development activities (PDAs) provide

opportunity for teachers to network with other teachers. PDAs will be shown to have a

sizable impact on intradistrict teacher mobility.

3.2 The Identification Strategy

Before I describe how I identify teacher networks, let me first discuss what teacher

networks are, how they are useful, and how teachers build these networks.

3.2.1 What Are Networks And How Are They Useful

The term network in this paper is defined as group or "family" of co-workers or

colleagues within one's field of work. Specifically, in this paper a network refers to a

teacher's associations with other teachers who work within their own district. Although

the context of a network here is limited to the field of education, people from all

professions form networks for various reasons.

There are many reasons why any professional or businessperson would network

with others in their field of work. They yearn to share ideas, understand their

competition, and exchange information that can benefit their personal interests,









sometimes in a mutual way. Why teachers network with each other is not so different.

They exchange classroom experiences and teaching methods as well as information on

workplace environments, as well as build personal bonds that enhance their feelings

towards the profession, or even their personal lives. As a result, teachers may be able to

use their networks to plan intradistrict school events more efficiently, find work in the

summer time, or sometimes just to socialize with others in their field.

However, there are two key uses of teacher networks that are the focus of this

paper. First I argue that teachers use their networks to gain greater information about

school characteristics. Because there are many characteristics about schools that may be

difficult to observe (existing job vacancies, quality of administrative support,

departmental environments, etc.), networks may allow teachers to evaluate other schools

in their district more efficiently. Greater information about schools should allow teachers

to sort into their preferred school more easily.

Secondly, I argue that teachers use their networks to reveal their quality as a job

candidate in a less costly way. Because it can be costly (or sometimes impossible) to

reveal one's aptitude for a job vacancy, teachers may call on their networks to provide

important or relevant information to potential employers. For example, a teacher may

ask another teacher to put in a favorable word to a hiring principal. In this way, networks

provide a valuable advantage to those seeking employment, while at the same time being

a more credible or reliable source of information for potential employers. If teachers do

use their networks to foster more efficient, less costly sorting behavior, then policies that

encourage network building may have unintended outcomes for teacher mobility and

possibly student achievement.









3.2.2 How Do Teachers Build Networks?

Clearly some teacher networking takes place in private forums. Teachers may

spend recreational time together getting to know one another, or they may talk over

lunch. However these examples of the networking process are not feasibly quantified.

To find a more measurable environment where networking takes place, I focus on

professional development activities. Professional Development Activities (PDAs) consist

of an array of different activities geared towards maintaining and enhancing teacher

competency and knowledge of various issues (including such issues as the use of

technology in the classroom, student assessment and state education standards, teaching

methods, etc.).

PDAs can include many different types of activities such as workshops, mentoring

and/or peer observation, collaborative research, University coursework, etc. Table 3-1

reports a summary of self-reported professional development activity from the 1999-2000

Schools and Staffing Survey.2 Though each of these activities may be available to most

teachers in the sample, attending workshops/conferences/training clearly seems to be the

most popular form of PDA in the sample, with nearly 94% of teachers reporting they

have engaged in this form of PDA in the past year (prior to the 1999-2000 school year).

PDAs provide an ideal environment for teachers to network with one another.

Teachers who participate in workshops or conferences have greater opportunity to meet

teachers from other schools within their district and enhance their networks. In addition

most PDA is geared to a specific employment group within the school system. This



2 The 1999-2000 Schools and Staffing Survey (SASS) is a survey of involving approximately 56,000 public
school teachers from all over the U.S., and is produced by the National Center of Education Statistics. It is
the main data source for this chapter and is described further in Section 3.3.









means that administrative personnel do not usually engage in the same PDA as teachers.

Likewise teachers do not usually engage in the same PDA as other support staff (such as

bus drivers, cafeteria workers, etc.). Thus I can be somewhat confident that if teachers

are building networks through PDA, they are generally relevant networks.

Besides simply meeting other teachers through PDA, these activities often provide

an environment conducive to promoting teacher interaction. Workshops especially are

designed in such a way that encourages interaction among participants. Workshop

assemblies are usually long in duration (often several hours per session) and can span

weeks or months. In addition, individuals are often called upon to share experiences with

all the participants, or present material regarding how the workshop has influenced their

classroom activity.

3.2.3 Measuring Teacher Networks

In order to measure the size of a teacher's network, I utilize national survey data

that provides information on professional development activity. The 1999-2000 Schools

and Staffing Survey (SASS), the survey from which the overall dataset is derived, asks

respondents to report the number of hours they spent in each of six major categories3 of

PDA within the last school year (1998-99). Rather than report the specific number of

hours, respondents are asked to select a (predetermined) range of hours that includes the

actual number of hours they have spent in each category of PDA. Using the midpoint of

the reported range as my best estimate of reported hours, I create a continuous variable of





3 These six categories are: PDAs with focus on 1) in-depth study of main assignment, 2) content and
performance standards of main assignment, 3) methods of teaching, 4) uses of computers for instruction,
5) student assessment/methods of testing, and 6) student management in the classroom.









PDA Hours (Hours), by summing the reported hours across all six PDA categories for

each individual.4

In general the distribution of PDA Hours across my sample is skewed to the left

with approximately 3.1 % of all teachers reporting that they have not engaged in any

form of PDA within the past year. Both the level of skewness of the sample

(approximately 1.30) and percent of teachers reporting no PDA within the past year are

roughly equivalent across movement categories as well.5 In addition, the distribution of

PDA Hours across these movement categories is also roughly identical. Of those

teachers who did not move, 24.8 % (or 332) are in the upper quartile of the entire

distribution. Likewise 25.2 % of teachers (or 88) who transferred within district were in

the upper quartile of the entire distribution of PDA Hours.6 It is also worth noting that in

each individual PDA category, only a small fraction of teachers are top-coded (reporting

the maximum number of hours allowed in the survey).7 Although for each PDA category

the percent of teachers who are top-coded ranges from 16.4 % to 2.1 %, these differences

do not vary significantly across movement types. Overall the data seems to indicate that





4 It is important to note that the highest range of hours that teachers could report is "33 or more". Where
this happens, I use 60 hours as the maximum value in this range. 60 seems to be a reasonable upper limit
of hours, however I have also tested maximum values at 48 and 64. These changes do not significantly
impact the results.

5 The PDA distribution of teachers who move intradistrict is slightly more skewed (1.37) than the same
distribution for those who did not move (1.25). Additionally, the kurtosis of the PDA distribution for all
movement categories is nearly equivalent (4.34 for non-movers, 4.64 for intradistrict movers, and 4.61 for
those who leave their district).

6 Of those teachers who moved out of their district, 21.7 % (or 98) are in the upper quartile of the entire
distribution. This includes both interdistrict and interstate movement.

7 There is only one teacher in my sample who is top-coded for every category of PDA (and has thus taken
the maximum amount of PDA measured in the survey)









along the PDA measurement, teachers in various movement categories do not look

different enough from one another to warrant any concern these groups are incomparable.

3.2.4 The Basic Model

In this basic model I am attempting to explain how a teacher's decision to transfer

to another school (in year t) within their own district is affected by their network. My

measure of a teacher network here is the number of hours a teacher spent in professional

development activities in the previous year (in year t -1). I can start by expressing a

simple regression model in the following form:

Transfer) = a + P(PDA Hours(t 1)) + 6X

Here the dependent variable is a dummy that expresses a teacher's transfer within

his own district. The dependent variable (Transfer) is assigned a value of 1 for a within

district move, and 0 otherwise. In addition to the amount of networking a teacher does,

the basic characteristics of the most mobile teachers are somewhat different from those

who are less mobile. To capture these differences I include variable X, a vector of

individual and school characteristics that influence a transfer decision. Table 3-2 reports

some descriptive statistics about the characteristics of teachers who do and do not move

around in various ways. The most notable difference is that teachers who move within

district tend on average to be about 3-6 years less experienced (and younger) then those

who do not move. Since inexperienced teachers are generally not as far along the career

path as more experienced teachers, this difference seems intuitive. In addition, I must

also consider marital status and job status in the model. Those teachers who are not

married may find a move to a new school to be more costly since spousal relocation can

complicate matters. Those teachers who are employed part-time by their district should









have a greater overall benefit to searching for job vacancies since they stand to benefit

from becoming a full-time employee.

Aside from the natural differences that exist among teachers, a teacher's transfer

decision should also depend on the working conditions he/she faced the previous year.

Specifically, a teacher's decision to transfer may be influenced by student

quality/characteristics, administrative quality/characteristics, or compensation levels. As

a measure of student quality (L), I construct a measure:

%Free Lunch Eligible = (%free-lunch eligibleshool,(t -1) %free-lunch eligibledistrict,(t 1))

This measure is the difference between the percent of the student population who

are free-lunch eligible at a teacher's school and the average percent of the student

population who are free-lunch eligible at the district level. It is designed to capture the

socio-economic status (SES) of the student population at a teacher's initial school relative

to the SES of the district as a whole. Given that teachers desire to sort away from low

SES schools, this variable should have a positive impact on transfers.

To measure administrative quality (Adm), I utilize the variable:

Poor Administration = a dummy variable, receives a value of 1 is a teacher reports
that administrative support is poor within the past year (t 1)

This variable captures a teacher's perception of how supportive/encouraging the

administration at their school is towards its staff. The predicted sign of this variable

should be positive, since in schools where administrative quality is perceived to be poor,

teacher mobility should be higher.8 Some additional measures of the working conditions

in my regressions specification include:


8 I have also measured poor administrative quality using the average teacher response for each school. My
results do not change with this different measure.









Student Threat = a dummy variable, equals 1 if a teacher reports being threatened
by a student within the past year (t 1) (Thr).

Again because receiving a student threat indicates poor working conditions, the sign of

the coefficient should be positive to express teachers' preferences to sort away from

undesirable schools.9

Bonus Pay = a dummy variable, equals 1 if a teacher reports receiving bonus pay,
separate from salary and extracurricular pay within the past year (BP).

Bonus pay should give teachers an incentive to stay at their existing school, and should

therefore be negatively correlated with transfer rates.

I also use the self-reported data to construct a measure of job satisfaction (Job) that

reflects a teacher's general attitude toward their workplace:

Job Dissatisfaction = a dummy variable, equals 1 if a teacher reports being strongly
dissatisfied with being a teacher at their school in year (t 1) (Job)

This variable is intended to capture not just the quality of the students, but also a

teacher's perception about the quality of their co-workers, their administrative superiors,

school resource availability, and general issues surrounding their employment. Since

greater general job dissatisfaction should lead to greater mobility, I expect the sign of this

variable to be positive.

While it is true that salary differences between schools seem to affect teacher

transfer decisions, this analysis is limited to those teachers whose movement is within

their district. Given that salaries schedules are generally set at the district level and do





9 I have also measured student threat using the average teacher response for each school. My results do not
change with this different measure.









not vary across schools in the same district, base salary differences between schools are

negligible10. As a result, I do not include base salary as a variable in my analysis.1

I also control for whether a teacher belongs to a union in this model for two main

reasons. The first is that union membership may provide existing teachers with

preferential treatment in filling ajob vacancy. Another reason is that districts may find

unionized teachers more difficult to fire. As a result, problem teachers may be shuffled

around more often. These two explanations of how unions affect transfer rates are very

different, however in both scenarios union membership should positively affect the

probability of transferring.

Finally I add state fixed-effects into the regression model to control for basic

institutional and cultural differences across states, and a dummy = 1 for missing data in

the % Free-Lunch (L) variable.12 It would be ideal to use school district fixed-effects,

however, doing this results in a dramatic reduction of my sample size.13 Once this is

done, the regression takes the following form:

Transfer(,) = a + PiHours(t i) + p2L(t 1-) + 33Thr(t 1) + 04BP(t 1) + P5PT(t 1) + 36Exp(t 1)
+ 37Adm(t 1) + PsMar(t 1) + 390Job(t 1)+ 3ioAge + PiiUnion + p12LDummy
+ P13StateFE

where PTis a dummy for job type (1 if part-time, 0 otherwise), Exp is the total years of

full-time experience, Age is teacher age, Union is a dummy for union membership (1 is

10 In the SASS, over 98% of school districts reported having a set salary schedule for teachers in their
district. This means that virtually every teacher in the sample who moved within district did not face salary
differences between the schools in their district

1 Although other types of compensation such as incentive pay are an important component to one's
decision to transfer, those types of compensation could be endogenously determined and therefore
inappropriate to include.
12 The Free-Lunch variable is the only variable in the regression where observations were missing.
Roughly 157 (or 9%) of observations were missing this data.

13 Never the less I show on Table 3-7 that the inclusion of district fixed-effects does not alter the results.









unionized, 0 otherwise) and Mar is a dummy for marriage status (1 is married, 0

otherwise).

3.3 Data

The data for this model comes from the 1999 2000 Schools and Staffing Survey

(SASS) and the 2000 2001 Teacher Follow-up Survey (TFS) produced by the National

Center for Education Statistics (NCES).14 The data are comprised of individual (teacher)

level survey information that can be matched across surveys, as well as relevant survey

information for the relevant schools, principals and school districts. In the original 1999-

2000 SASS survey there were over 52,000 respondents, but the TFS consists of only a

subset of those original respondents.

In aggregate the TFS includes 6,758 teachers, however not all of the teachers in the

survey are included in this analysis.15 Since this analysis focuses on public school

teachers who move within district relative to those who do not change schools, I

eliminate those respondents who either leave their district, or leave teaching altogether.

Later in the paper I will use other types of movement (interdistrict and interstate

movements) to show that PDA hours do not influence other mobility decisions,

particularly those decisions where networks should not matter.








14 The datasets utilized in this paper contain identifiable and sensitive information. While there are
versions of these data that are available to the general public, the datasets used in this paper are only
available at the discretion of the NCES, and with their proper approval.

15 Those respondents in the TFS who left the teaching profession after the 1999-2000 school year (2,374
respondents) are given a version of the TFS that is different from the survey given to those who did not
leave teaching.









In addition, I eliminate those teachers who initially teach at private, Indian, or

charter schools, teachers who reported their move was mostly involuntary16, and those

teachers who were not considered full or part time regular teachers in 1999 (e.g.,

specialists, substitutes, student teachers and itinerant teachers). Once these criteria are

imposed on the data set, 1,688 observations remain. Of the 1,688 observations in the

dataset, 349 (or approximately 20.7%) of those were teachers who transferred within their

district.

3.4 Empirical Evidence

Because the dependent variable is a binary choice variable with a successful

transfer equal to 1, probit estimation is appropriate. Column 1 of Table 3-3 reports a

preliminary regression of PDA Hours onto the dependent variable, with the specified

covariates also included.17 At first glance it seems that networking does not have a

significant impact on the likelihood of transfer. Networks though are a story of collegial

kinship, and like any friendship, benefits may not be realized until the parties involved

are familiar and comfortable enough with one another to begin a relationship. Thus

networks may be difficult to establish initially. Further, networks may require a level of

maintenance. If that is the case, then teachers who engage in PDA frequently may have

an increasing advantage over those who do not. This may imply that there are increasing

returns to hours spent in PDA in terms of networking. A simple test of this assumption is

to include a squared term of PDA Hours.


16 It is also worth noting that any involuntary teacher movements who may be unavoidably included in the
sample (thus creating noisy measures of intradistrict movement), are likely to bias any positive estimates of
the effect of networking towards zero.

1 All regression estimates are clustered at the district level to control for possible standard error correlation
at the district level. In addition, all regression estimates include robust standard errors, to account for
possible heteroskedasticity.









In Column 2 of Table 3-3, the addition of a squared measure of PDA Hours in the

regression seems to suggest that there may be some value to networking. Specifically,

the coefficient on the squared term is positive and significant at the 0.025 level, which

may indicate that while having only a few hours networking may have no impact on

transferring, positive returns to networking may exist for large values of PDA Hours. To

further test whether this is the case, I use a spline to allow the effect of an additional hour

of PDA to be greater at high levels of PDA hours then at low levels of PDA hours.

The spline estimates a continuous, piecewise-linear relationship between PDA Hours and

transfer probability, and is defined as follows:

PDA Hours (H) : from [0 H*] hours:
=H ifH < H*
= H* ifH > H*

PDA Hours (H) : from [H* Hmax] hours:
= 0 ifH < H*
=(H-H*) ifH>H*

Table 3-4 reports the coefficients of several regressions using the spline. I test

various cutoffs of H*, where H* is chosen at decile intervals along the distribution of

PDA Hours. Table 3-4 illustrates that the 2nd segment of the spline is positive and

significant for most decile cutoffs18, but the best fit for the regression is where the cutoff

is at the 80th percentile. This table confirms that additional PDA Hours, particularly for

those teachers in the top 20% of the overall distribution of PDA Hours, seem to have a

positive and significant impact on a teacher's transfer decision. Interpretation of the

coefficients suggests that a one standard deviation increase in the total number if PDA



18 For every cutoff chosen at or above the 20th percentile of the distribution, the 2nd segment of the spline
estimation is positive and significant.









Hours (approximately 46.6) can lead to a 0.020 to 0.078 increase in the probability of

transferring within one's own school district.

Other covariates also seem to have a significant impact of the transfer decision.

The evidence seems to indicate that teachers with less experience and those who work

only part-time are significantly more likely to move between schools. Intuitively this

seems valid since teachers with little experience have had arguably less time to sort into a

desirable school and should therefore be more likely to transfer than teachers who are

further along the career path. In addition, part-time employees stand to secure higher

wages if they search for full-time employment, and have lower time-cost to job

searching.

Besides the natural sorting taking place for younger/less experienced teachers,

teachers who work in low SES schools and who face undesirable working conditions

clearly have a greater incentive to sort into better schools, especially given that

additional/incentive pay is rarely available for teachers to work in less desirable schools.

The evidence in Tables 3-3 and 3-4 seem to support this claim. Teachers at schools

where student SES as a whole is lower than the district average, schools where their

overall job satisfaction is generally low, and teachers that feel administrative quality is

generally poor, seem to have a higher propensity to sort.

Lastly, Tables 3-3 and 3-4 show that there is also some advantage to a teacher

being part of a teacher union in terms of increased mobility. This could be because

teachers who are part of a union receive priority in filling a position vacancy. Another

interpretation of this union effect could also be that unionized teachers are difficult to

fire, and therefore get shuffled around more often. Overall the results shown in Tables 3-









3 and 3-4 are consistent with previous findings in the literature, and not at all surprising

given the economics going on. These regression specifications provides some valuable

insight into the teacher transfer decision, and also tells us that teacher networks are not an

integral component to teacher mobility.

3.5 Robustness Checks

The evidence in the previous section suggests that there is a significant relationship

between teacher networks and transfer behavior. However it is important to test the

robustness of these results. If the time a teacher spends in PDA is simply correlated with

some other (possibly unobservable) characteristicss, then the explanation provided by

networking may be inaccurate. If it is true that teachers who spend a great deal of time

doing professional development activities are simply more likely to transfer, then the

results are spurious in a networking framework.

Before I test the robustness of the results in the previous section, I look to see if

those teachers who engage in the most professional development look substantially

different from the average teacher. I compare the characteristics of teachers in every

decile of the distribution of PDA hours to the average characteristics of the entire sample

of full and part-time teachers (2,280 observations). The results of these comparisons are

reported in Figure 3-1. Each graph on Figure 3-1 shows the mean of the given

characteristic for each interval (0% 10%, 10% 20%, etc) of the PDA distribution,

centered about the mean for the entire sample. I use the standard errors at each interval to

construct a 95% confidence interval on which the mean lies, and test whether the sample

average falls in the range of the confidence interval.

There are only a few characteristics where teachers who engage in the most PDA

look significantly different from the sample average. The graphs of Age and Full-Time









Experience illustrate that teachers in the top end of the distribution of PDA Hours are

significantly older and more experienced. However, previous evidence suggests that

more experienced teachers, who are also generally older, are also less likely to transfer

schools. Therefore, if older and more experienced teachers are taking the most hours of

professional development, then we should expect PDA Hours to be negatively correlated

with the probability of transferring, which does not seem to be the case.

The graphs in Figure 3-1 may also suggest that those teachers at the very bottom of

the distribution of PDA Hours may be disproportionately white, male part-time teachers,

with respect to the entire distribution of hours in PDA. However it is not clear that this

presents a problem of selection bias in who decides to take PDA as it relates to transfer

rates. Currently there is no evidence to suggest that qualities of teacher race or gender

are effective at predicting/explaining teacher movement. In addition, Table 3-4 showed

that part-time teachers are more likely to move than full-time teachers. If part-time

teachers are more likely to spend only a small amount of time in PDA, this should also

downwardly bias the estimated effect of PDA Hours on transfers. These results are

promising, however they are not conclusive. Teachers may still be sorting into the

distribution of PDA Hours based on some characteristic that is not observable.

To test whether teachers are sorting into high or low levels of professional

development activity, I begin by looking at other types of teacher movement. If it is true

that teachers with high levels of PDA participation are simply more likely to move, then

one might expect these same teachers are also more likely to move in ways other than just

intradistrict transfer. However, since most PDAs are activities limited to one's own

school district, then networks should be largely ineffective for movement outside the









scope of the immediate district. So if PDA hours are a strong predictor of other types of

movement, then that may be an indication that the network relationship is unjustified.

Tables 3-5 and 3-6 report similar regressions similar to those reported in Table 3-4.

Table 3-5 shows the effect of a change in PDA Hours on the probability of

movement to another school district within one's state. Table 3-6 shows the effect of a

change in PDA Hours on the probability of movement to another school district in

another state. In both cases, the effect of increased PDA Hours is not significantly

different from zero. For Tables 3-5 and 3-6, neither spline segment is significant, which

suggests that the amount of time a teacher spends in PDAs has no significant impact on

how likely they are to move out-of-district (within the state), or even out-of-state. This

helps to quell some suspicion that teachers with high levels of PDA are simply more

mobile.

It may also be possible that the relationship between PDA and mobility is a

function of district characteristics or behavior. For example, if districts with "naturally"

high levels of teacher mobility also demand an unusually high number of PDA hours for

their faculty (arguably this may be the case with some urban school districts), then the

positive effect of networking is simply a reflection of high levels of mobility in a subset

of districts. To account for this possibility, I repeat the regressions in Table 3-4 with

district fixed effects. Because the use of district-fixed effects limits the sample to

districts where both movement and nonmovement occur, the number of observations

drops to only 398. Naturally this limits the power of the analysis, and is the main

problem with using district fixed effects throughout. Table 3-7 reports the results of these

regressions. According to the data, high levels of PDA have a positive effect on the









transfer decision and low levels of PDA have a negative effect on the transfer decision.

The results in Table 3-7 confirm that district specific behavior does not fully explain the

relationship hours in PDA and mobility.

3.6 Conclusion

The evidence presented in this paper clearly seems to indicate that time spent in

professional development activities leads to networks that enhance teacher mobility

within their own district. These PDAs arguably provide a forum in which teachers can

build networks to gather important information about school characteristics, work

environments and job availability from which they can sort in a less costly, more efficient

way. It also seems plausible that teachers can use their networks with favored colleagues

to reveal their aptitude for a job vacancy, which may allow teachers access to transfer

options that might have previously been closed to them.

Results show that a one standard deviation increase in the amount of networking

leads to a 0.020 to 0.078 increase in the probability of transferring within one's own

school district. If teachers use these networks to more effectively sort away from less

desirable schools, such as low-income or failing schools, then networks arising from

PDAs may contribute to the disparities in teacher quality across school types. It is also

possible that if networks make it more difficult for low achieving schools to retain

qualified teachers, then student achievement at these schools may also be diminished by

networks. Oddly enough, this unintended consequence of staff development policies

would be subversive to the goal of PDAs, which is to improve education through greater

teacher quality.











Table 3-1: Professional Development Activities Summary Statistics
Number of Teachers Who Reported

Percent Who
Yes Nocip
Participated


785 1495 34.4 %


University Course in your Main
Teaching Field

University Course NOT in your Main
Teaching Field

Observational Visits to Other Schools

Individual or Collaborative Reseach on
a Topic of Interest

Regularly Scheduled Collaboration with
other Teachers on Issues of Instruction

Mentoring or Peer Observation

Participating in a Network of Teachers
(e.g. Internet Organization)

Attending Workshops, Conferences
or Training

Presenter at a Workshop, Conference
or Training


26.9 %


33.6 %

45.3 %


68.5 %


44.4 %

24.5 %


93.9 %


19.8 %


614 1666


766 1514

1032 1248



1561 719


1014 1266

559 1721


2141 139


452 1828










Table 3-2: Summary Statistics for Various Teacher Groups
No Move Move

Inter-
Intra-District Inter-State
District

484 (21.2
Number of Teachers 1400 (61.4%) 370(16.2%) 484(22 96(3.9%)


Mean Std. Dev. Mean Mean Mean

PDA Hours 55.728 46.662 59.000 49.636 48.220

% Free-Lunch Eligible 0.002 0.225 0.045 -0.014 -0.021

Bonus Pay 0.137 0.344 0.113 0.101 0.146

Student Threat 0.095 0.294 0.127 0.130 0.083

Job Dissatisfaction 0.021 0.144 0.048 0.066 0.072

Full-Time Experience 10.915 9.856 7.959 5.134 4.583

Part-time employment 0.040 0.197 0.081 0.041 0.010

Poor Administration 0.069 0.254 0.102 0.101 0.104

Union Member 0.783 0.411 0.778 0.652 0.667

Age 40.484 11.340 37.956 33.634 32.030

Gender (Male) 0.282 0.450 0.235 0.299 0.292

Race (White) 0.902 0.297 0.862 0.902 0.927

Marital Status 0.695 0.460 0.678 0.665 0.667










Table 3-3: Probit Regression of Intra-District Movement
Variables 1 2


PDA Hours


(PDA Hours)2


% Free-Lunch Eligible


Bonus Pay


Student Threat


Job Dissatisfaction


Full-Time Experience


Part-time employment


Poor Administration


Union Member


Age


Marital Status

Dummy for Missing
% Free Lunch Eligible

Constant


0.0010626 0.0006944
(1.27) (0.62)

0.00003
(2.30)**

0.301017 0.3069657
(1.84)* (1.87)*


0.1669206
(1.40)


0.1629131
(1.37)


0.1009573 0.1002069
(0.87) (0.86)

0.4224374 0.4353552
(1.85)* (1.92)*


0.0143332
(2.34)**


0.0131213
(2.13)**


0.5452476 0.5394231
(3.43)** (3.41)**

0.3542717 0.3600298
(2.55)** (2.59)**

0.2372197 0.2386774
(2.36)** (2.37)**


-0.005178 0.0055949
(1.01) (1.09)

0.0237581 0.0261624
(0.29) (0.32)

0.1770793 0.176266
(1.20) (1.20)


0.2702219
(1.37)


0.2997908
(1.50)


Pseudo R-squared 0.0955 0.0984
Number of Obs. 1627 1627
Robust SE, clustered Y Y
State FE Y Y
* denotes significance at the 0.05 level
** denotes significance at the 0.025 level







87


Table 3-4: Probit Regression of Intra-District Movement
Regression of PDA Hours with H* Chosen at Decile Intervals Along the Distribution of PDA Hours
H* = H* = H* = H H= H* = H* =
H* = 9 H* = 17 H* = 68
22.5 29.5 40.5 51.5 87.5 122.5
Variables (10%) (20%) (30%) (40%) (50%) (60%) (70%) (80%) (90%)


PDA Hours
0 H*


PDA Hours
>H*

% Free-Lunch
Eligible

Job
Dissatisfaction

Full-Time
Experience

Part-time
employment

Poor
Administration

Union
Member

Student Threat

Bonus Pay

Age

Dummy for
Missing
% Free Lunch
Eligible
Marital Status

Constant


Pseudo R-
Squared
Number of
Obs.
State FE
Robust SE,
clustered


-0.0262 -0.0170 -0.0117 -0.0074 -0.0047 -0.0033 -0.0022 -0.0018 -0.0006
(1.12) (1.68) (1.63) (1.43) (1.34) (1.27) (1.12) (1.15) (0.55)
0.0013 0.0017 0.0018 0.0020 0.0022 0.0026 0.0031 0.0043 0.0061

(1.52) (1.85)* (1.94)* (1.94)* (2.00)** (2.08)** (2.16)** (2.45)** (2.51)**
0.3041 0.3066 0.3077 0.3074 0.3067 0.3086 0.3117 0.3081 0.3010

(1.86)* (1.87)* (1.88)* (1.88)* (1.88)* (1.89)* (1.91)* (1.88)* (1.83)*
0.4214 0.4254 0.4282 0.4285 0.4306 0.4338 0.4377 0.4385 0.4288

(1.86)* (1.87)* (1.88)* (1.88)* (1.89)* (1.90)* (1.92)* (1.93)* (1.89)*
-0.0141 -0.0136 -0.0135 -0.0134 -0.0132 -0.0130 -0.0130 -0.0130 -0.0134

(2.29)** (2.22)** (2.19)** (2.17)** (2.13)** (2.10)** (2.10)** (2.10)** (2.18)**
0.5383 0.5362 0.5351 0.5373 0.5417 0.5458 0.5479 0.5477 0.5381

(3.37)** (3.36)** (3.36)** (3.38)** (3.41)** (3.44)** (3.45)** (3.46)** (3.41)**
0.3575 0.3616 0.3636 0.3641 0.3638 0.3621 0.3595 0.3596 0.3593

(2.57)** (2.59)** (2.61)** (2.61)** (2.61)** (2.60)** (2.58)** (2.58)** (2.59)**
0.2344 0.2406 0.2419 0.2411 0.2415 0.2416 0.2399 0.2387 0.2373

(2.32)** (2.38)** (2.40)** (2.39)** (2.39)** (2.39)** (2.37)** (2.37)** (2.36)**
0.1011 0.1016 0.1022 0.1025 0.1016 0.1000 0.0984 0.0987 0.0991
(0.87) (0.87) (0.87) (0.88) (0.87) (0.86) (0.85) (0.85) (0.85)
-0.1674 -0.1639 -0.1639 -0.1657 -0.1658 -0.1650 -0.1635 -0.1633 -0.1658
(1.41) (1.38) (1.38) (1.39) (1.40) (1.39) (1.38) (1.38) (1.40)
-0.0053 -0.0054 -0.0054 -0.0055 -0.0055 -0.0056 -0.0057 -0.0058 -0.0056
(1.02) (1.05) (1.06) (1.07) (1.08) (1.09) (1.10) (1.12) (1.09)
0.1789 0.1779 0.1798 0.1796 0.1790 0.1777 0.1772 0.1807 0.1776

(1.22) (1.21) (1.22) (1.22) (1.22) (1.21) (1.21) (1.23) (1.21)
0.0232 0.0224 0.0232 0.0245 0.0256 0.0265 0.0266 0.0277 0.0264
(0.28) (0.27) (0.28) (0.30) (0.31) (0.32) (0.33) (0.34) (0.32)
-0.9070 0.5313 0.4935 0.4260 -0.9811 0.3674 0.3596 -0.9753 0.3402
(1.29) (2.15)** (2.09)** (1.93)* (1.45) (1.78)* (1.75)* (1.44) (1.69)
0.0993 0.1004 0.1004 0.1002 0.1002 0.1003 0.1005 0.1012 0.1009


1627
Y
Y


1627
Y
Y


1627
Y
Y


1627
Y
Y


1627
Y
Y


1627
Y
Y


1627
Y
Y


1627
Y
Y


1627
Y
Y


* denoted significance at the 0.05 level
** denoted significance at the 0.025 level







88




W A











0-
0 20 40 60 80 100
PDA Distribution



B








8-0


0 20 40 60 80 100
FDA Distribution


8C




. A I I T I T T I I


i I I 1 I 1 4
O-1

0 20 40 60 80 100
PDA distribution

A) Percent of teachers receiving poor administrative support
B) Percent of teachers receiving bonus pay
C) Percent of teachers reporting overall job dissatisfaction

Figure 3-1. Comparisons of Teacher Characteristics Along the Distribution of PDA.





























0 20 40 60 80 100
PDA Distribution


E






ai)




o t
0-


L..


0 20 40 60 80 100
FDA Distribution

F



t-











0 20 40 60 80 100
PDA Distribution
D) Percent of teachers with a master's degree
E) Percent of free-lunch eligible relative to the district average
F) Percent of teachers who belong to a union


Figure 3-1. Continued.