<%BANNER%>

Analysis of markets in the presence of network effects and standards competition

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 E20101123_AAAADX INGEST_TIME 2010-11-23T21:56:42Z PACKAGE UFE0011601_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 14611 DFID F20101123_AACMIH ORIGIN DEPOSITOR PATH wang_q_Page_25.QC.jpg GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
a452f58b69c3da1dbbadaf1aed7066ea
SHA-1
ed9c9fee577528f2f82513918278cc529db14a36
1762 F20101123_AACMDK wang_q_Page_28.txt
0ea5a5f458037f2a612ceb3acd5ac0f0
9306875b6e7e086b09d6488615ef953d2a3073f3
105715 F20101123_AACLWN wang_q_Page_75.jp2
32b54ec1c356054f4b7582fcad933cee
ec5ee60baf47d325a579d39c97472b620fbcf641
71269 F20101123_AACLRP wang_q_Page_18.jpg
ca71884c0b6917fc484bf4f996018eda
984055d086d74e724cc0f2a926f2657bf94dc3f4
4878 F20101123_AACMII wang_q_Page_25thm.jpg
61aa9243a005a5a92dfca2c5c590a5c3
ce04d35697f05ad34fd5cc5caebd8c2de052fbac
1751 F20101123_AACMDL wang_q_Page_29.txt
a3f26590bbc8d0d3cec638597df9e6d6
ea1a7b4afdc085de3d1ea6399814c03784dbbde3
26209 F20101123_AACLWO wang_q_Page_76.jp2
5ddbecb098d904ea5961da749e772b44
d943e7a66ad52e67d122d5ba4f58faa06fa31aab
70904 F20101123_AACLRQ wang_q_Page_19.jpg
d7713ba674e11f77ca26816d2961467d
2db52f9636ec55c8710d3c9cbf48e3c2c9b493dd
6029 F20101123_AACMIJ wang_q_Page_26thm.jpg
bf791bcce4e4fd034e2e6494b468f026
1fd511cf3dc9605b7452b15638abb644bdbbb347
1886 F20101123_AACMDM wang_q_Page_30.txt
c4a3297e56be678ee3628a5eb8d99f4a
910e7165616ecb6472812a3f4d69f5d68f0ac8a8
109506 F20101123_AACLWP wang_q_Page_77.jp2
f8b2d3d8e80683289e33ea23155fb41e
57afb4199a584134705680a2935cbfd33c33108b
80758 F20101123_AACLRR wang_q_Page_20.jpg
a57f387099cd10c1539b8fedb61e7a8c
99d24e6c1a1351e2add9e1d65c4f74397adebb54
18188 F20101123_AACMIK wang_q_Page_28.QC.jpg
ae0fd16b6e723b606da57fe0c7b8b673
26f715355f79d1f3bea65cf7b4519943259149af
1641 F20101123_AACMDN wang_q_Page_31.txt
f076143820e11a7eb27522e995400151
a1f839d76f729728dc23b5cae9a1d20757b30fec
128659 F20101123_AACLWQ wang_q_Page_78.jp2
ca1afcc3ccd20daa3effb0ca655d363f
53b78d7d75a66ae45089016526af9c0bc04385bc
71024 F20101123_AACLRS wang_q_Page_21.jpg
db7a9a39cafd806b5f5f89af16be64fe
6939e5ad46aca0d3bacc09a0acb5dc8aa034ce5a
6007 F20101123_AACMIL wang_q_Page_29thm.jpg
46122db72c65a3ebb64f9ca081030832
54d8d960b06d129814449077189908efcc3c757e
2070 F20101123_AACMDO wang_q_Page_32.txt
72e04dbc6da75284899fd2c3668f26a9
1b0bda467f0e9b70c3d412598398039343bad145
133052 F20101123_AACLWR wang_q_Page_79.jp2
9b8d05855af5cefa7ed01191f99585bc
050b3deffcccac3befe62b09bce12f03de27786d
69624 F20101123_AACLRT wang_q_Page_22.jpg
c88949a388bdf1c6a245a7437b04bc0f
646f555a186d5adad18bf1ea4b3e5bbd6f41e27a
21960 F20101123_AACMIM wang_q_Page_30.QC.jpg
1b4f084b9dd4ddd175c7568e383f8e59
501a8a6054e0666f01d5dbf75eeba4c0c8397564
2341 F20101123_AACMDP wang_q_Page_33.txt
514000f34b07f8fa0e94733674c8b9e5
bed82bfdda4349d7b46d766e5b709ef5eba8e8c0
71242 F20101123_AACLRU wang_q_Page_23.jpg
eec22b9197815917a198a348209b171e
2a4374b573e61e23695e008b8af069f07a5dd769
6070 F20101123_AACMIN wang_q_Page_31thm.jpg
23d142835bf44a7983fe37afb51e042c
eb36f5decb8518b2f2aea4ee4bbc80fa2e921e89
1799 F20101123_AACMDQ wang_q_Page_34.txt
e3c0b2dd7e8000e0bc0ce40c9a2df81e
c1da5dee18644b034db3bc7fa06095f61690f9ac
74575 F20101123_AACLWS wang_q_Page_80.jp2
24a6918efb809236016173cf7d2c468b
0861fed880856f313b15b6db439f855b5f0a165e
60002 F20101123_AACLRV wang_q_Page_24.jpg
cf95ce6434c1f9c56047222f1f997865
c78c531241c4116699e23a8c37f56b8f84dab31c
22027 F20101123_AACMIO wang_q_Page_32.QC.jpg
0ff012bddb3fc26bbe5d84dbb97d7c4f
0a26cc2983bdcb64cd529a36c127945f752d865a
2062 F20101123_AACMDR wang_q_Page_35.txt
990276795bac603bc94b33eb20876be8
ff748c0549f8d62ee9691b230be57ff4cf618c4a
46885 F20101123_AACLWT wang_q_Page_81.jp2
9b76a0801eb204679c2affed016e663a
a704c5eb1c9981c07fc585310271f398ddf242c1
63034 F20101123_AACLRW wang_q_Page_26.jpg
3337cc17c67dcb42e231c8cf537ed987
6eef8c337e1d878ef41c9c2b05b44732cf98fc67
6241 F20101123_AACMIP wang_q_Page_32thm.jpg
a8979a4707cc9d3238ae8830b6c6126a
e446c3e24aecd0e83b4458ac34a16597cefb1c94
2720 F20101123_AACMDS wang_q_Page_36.txt
c1d0754f443b34f7322ed4e064092441
c0ca177fc8167dd8bf718ada40769b7f7a0bb5c4
1053954 F20101123_AACLWU wang_q_Page_02.tif
469c6c0146c68988a3b758e654b8875a
e9567f1da48c8fe630eee65603c72b45b7233cd7
57355 F20101123_AACLRX wang_q_Page_27.jpg
1d94ce9303b0920aaaee235101146e94
5938a98712c77f0f96cd29d6d79589d9e110db91
23822 F20101123_AACMIQ wang_q_Page_33.QC.jpg
1739a322aa5b013e8391efcb89defaa9
aa3021e91f41e18b772621b0e4553c2880d7e025
1839 F20101123_AACMDT wang_q_Page_37.txt
0b8d41c8c0dc95eb811c6cc61ee3e53b
e92399534d59084b31037874e52184735e9ab9b1
F20101123_AACLWV wang_q_Page_03.tif
9ede2b5a2bcde07bc015cce38e558fc1
58e696575561fcb1b13b204f0f4de4e142cc189a
6492 F20101123_AACLPA wang_q_Page_21thm.jpg
db3ad18c424a124d8c8eb22371bd09a5
1c8940f79784291f5d452fcdd0a125f1bec6a512
51645 F20101123_AACLRY wang_q_Page_28.jpg
03707db8d8307ae95bbbeadc21f7a954
5ef99f799bd5eef6d699888fa201893fafbaba67
6818 F20101123_AACMIR wang_q_Page_33thm.jpg
3238d439cd326997f9b3de2c94320884
026a2054d7400fedee9261a1801d5b9231c84ab5
1887 F20101123_AACMDU wang_q_Page_38.txt
4c493189607931986e2e42f54ae8c15f
95a92211b3c04328f4463c03534f5c58e2b775d8
F20101123_AACLWW wang_q_Page_04.tif
177832f125e75f1101063a76fc27c52d
19baf6fa87220af3a2b648ae67160b69991e690a
20861 F20101123_AACLPB wang_q_Page_37.QC.jpg
975ac6a4432284b745d64404cf19eae7
c77951ed0c1d0717f521251b6836a3b3005ffd16
68445 F20101123_AACLRZ wang_q_Page_29.jpg
c88007de91d09b8795b10e9bb8e55b20
1b7a240dc20d91c9063b56b45507f9f6a713d47b
21277 F20101123_AACMIS wang_q_Page_34.QC.jpg
6f0a2d734a3afa4f9858546a91468b87
9140be7f712f71f6fc79fc7f0403a40e79296d41
1859 F20101123_AACMDV wang_q_Page_39.txt
fb4ad908a0e9a2cafc1baa7ad7fcbf4f
b2b1514fc44505a86db8dfc5d78f0cad8083f93f
25271604 F20101123_AACLWX wang_q_Page_06.tif
1e0b5ced44ab04158c5d4fea47470be1
66fdb5effbe4fa36cbd0afd1699f9d9672c6e2fe
468 F20101123_AACLPC wang_q_Page_76.txt
ce8b17bab4fda0c97494b6b39d4782c7
d2d3ffd6a99ca444b87d7a927454754516e5ea5d
5890 F20101123_AACMIT wang_q_Page_34thm.jpg
7e896b808878ac568a7fd1b5ea61ad62
d1c9c7b22ac5afeb23306f4f0141b2007a9f4d8f
1051929 F20101123_AACLUA wang_q_Page_06.jp2
b9ec6007d158f98e709bb2b738f1d8c1
91eb02297329e1e2925728e82867a16ea6197f3a
F20101123_AACLWY wang_q_Page_07.tif
c2d8407c7bbf4d54fdeb0dd6b817d87e
ccfea170f1613ec60e083f50d6dd662531ba9e94
1980 F20101123_AACLPD wang_q_Page_77.txt
5a5014eca4664e7b0d49fc420f41a111
55b0548660d3e780cd4574c6f1bdec24a38cd7aa
2346 F20101123_AACMDW wang_q_Page_40.txt
d1e5491898f180f15278ae79b0299d87
80214b86aebf9c04d2dccdfe75a7ffa019444c38
22380 F20101123_AACMIU wang_q_Page_35.QC.jpg
789870a9ea88dda7f6492b1a49ac1024
457afc0f56823c3b2518f9c0ef907a579f85a1bb
978637 F20101123_AACLUB wang_q_Page_07.jp2
ead7c471a266ffd8466b2094ae0ea8b1
44e1b0fdb79eeed10db77a0eba62be2fa9f1e01c
F20101123_AACLWZ wang_q_Page_08.tif
35387447f651cf3a9784814a2e5ff257
790c98db0bb447d6f3e214a273e52deb4f1f4b7f
10577 F20101123_AACLPE wang_q_Page_07.QC.jpg
00f8475aac7512b910ad31d427768786
9b0544b4cdeb8cdf5b3646d7ee758654dd84711b
1599 F20101123_AACMDX wang_q_Page_41.txt
42acbab06e6c6a2ff4b4bcdb14b3b600
f28988cbab6e14c904c77d2029143192159e8041
6340 F20101123_AACMIV wang_q_Page_35thm.jpg
6a6f2b3bf916bd14f31b8f3e5e7491cc
a44fae0bd4bfa8eb9b28f7a14a0dc2d6309fb364
6570 F20101123_AACLPF wang_q_Page_54thm.jpg
929faab4cb358f6cc9f6717c32051402
3391deaa94fd9481b6ae0626a3fff0106d7fd5ef
1745 F20101123_AACMDY wang_q_Page_42.txt
e8f60b5ded4c65800ed0e5f186bc33b3
e3c96ccf7edf5c859c7c86d66e5733fe2196d8fd
635196 F20101123_AACLUC wang_q_Page_09.jp2
83393d516c4ac80fcc7ee78d05304198
b144305fded180cc83cba6af0df8faa657777066
21509 F20101123_AACMIW wang_q_Page_36.QC.jpg
e5fc98dd5ffabdbe7bd1213abdfae1d2
ef32e9355b6935c9f4d1f90ebe345675b27606be
74763 F20101123_AACLPG wang_q_Page_44.jpg
0aaafcf4e270500ab475b0debccf4a9e
58095938e8754e89a4ed98491254268d35e9b895
F20101123_AACLZA wang_q_Page_64.tif
c812aeb11f3e4b7d54d5d16d5e8547fd
5b88bedb2b951c67c57a4b1ddb4869c00520a422
31125 F20101123_AACMBA wang_q_Page_42.pro
7d882bb78eaeb7fd1a5f0ffc368dc5f6
f5d6917f282c8fad75df8ba02c4c859ddef04030
1498 F20101123_AACMDZ wang_q_Page_43.txt
132ef6b4f59c507433328dc58cc84153
1e7eb7b3f0a2543438153e1c166d1207fa70d22d
86590 F20101123_AACLUD wang_q_Page_10.jp2
3382ce2740838f87dd52c1e4e1f61958
a95d972f95e932df0e7cf1de2e7a3a35f51ef071
20896 F20101123_AACMIX wang_q_Page_38.QC.jpg
f3aa1f0f2be70e5cda1bab93e6f4e60d
b574706a5e474a008183a0343f57b15f91d929fa
6323 F20101123_AACLPH wang_q_Page_45thm.jpg
5301d01ca86321004e6d1bb0700eeee7
b82be3f15b732b4449c53387bbfba899f600bf64
F20101123_AACLZB wang_q_Page_65.tif
678cd2bd40e62cde220a7aec331949a4
b57ffdcd073fb2bb238c00d6005d85ea6877276a
31294 F20101123_AACMBB wang_q_Page_43.pro
94a496b0274572653ae113349130bdde
18b057f194af8a0dee84afbbfa8c6c4e6b30985d
12549 F20101123_AACLUE wang_q_Page_11.jp2
67b1c7394d5d9528776e8fa441e17379
40673379eeb21150e9b2e9589bb2043e3c080b47
6119 F20101123_AACMIY wang_q_Page_38thm.jpg
afc1aa8fe625aa1bd042f34fb457985e
a10017b93ea1f09ec796c987cce8c640f3c54b92
21490 F20101123_AACMGA wang_q_Page_77.QC.jpg
90ae373bf85a766a9b0f13c03abcd015
6e55a712cab9f6268589e624a569df9f9bd60ed0
2365 F20101123_AACLPI wang_q_Page_78.txt
b01ced55e6adfd6de607733e035180a4
6834a05d87e937042a45e00d6ba1bc237e015fec
F20101123_AACLZC wang_q_Page_66.tif
6cd4b311050658a0509132db228f517b
05b33b0b4aed84d235e3fe0fbcd6eb38f66a65e7
60220 F20101123_AACMBC wang_q_Page_44.pro
55109e5f9d5b52f110a7850eb6a6b880
726969fc9b4350035f68a5ff1a2184a932a77cf4
98162 F20101123_AACLUF wang_q_Page_12.jp2
75c0f1591ceff520028b43070bb09f9f
e7704584f3ae399af9a059e27eb3d7568a05c580
21945 F20101123_AACMIZ wang_q_Page_39.QC.jpg
7b9d7375aabc1da1abc76c9de8b72d59
cc8bb3589ee55d2a96e3e7bae3773d6c7f026f46
F20101123_AACLPJ wang_q_Page_79.tif
db4feba7e9bb3d792bffe459f960b2fe
070fc7a9c8e4c738694d00323cc0b6061db13482
F20101123_AACLZD wang_q_Page_67.tif
05c6fc570ae59b837dad8d46b42e11f9
681bb1d2aaf81bd067538a46482d370cbe4f4313
46304 F20101123_AACMBD wang_q_Page_45.pro
86d25cde6ef3bab2e94efab489a1210c
8cebc738ae62c5d2d6a898fe998e82a185085ed0
106479 F20101123_AACLUG wang_q_Page_14.jp2
d3037af67bb210db07f603e5a34583aa
da03ed586d0ba12e0bcb37934a21c2b8fa8d6a83
19328 F20101123_AACMGB wang_q_Page_68.QC.jpg
63881ff16066e49a32e16431e903abe7
43322720f8f68fb896db5b650db42b4e5ad6e03d
49718 F20101123_AACLPK wang_q_Page_23.pro
10ac4b0d0cddfabac9a177ffd53ccf05
ed28c99e226c21f8fd590cfc5ccb71ba51deb65e
F20101123_AACLZE wang_q_Page_68.tif
49aeecd1a1610d4b9fba3b4fb0e297e9
a563f585ebb16f65ff48337f3be84ca5a027423a
67250 F20101123_AACMBE wang_q_Page_46.pro
41bec6aa14fc98d9350fa7bf29c8b08b
780fb6c8546c8600c4aedb6b95faca1441c9ed68
111386 F20101123_AACLUH wang_q_Page_15.jp2
ca09e1c9b6df6eaba4d1d9004d28fac8
3bd3595811f30979082c8369f7d10e9ad929cc7a
5308 F20101123_AACMGC wang_q_Page_50thm.jpg
4469d0059b551662e20e699162a3e6b3
203cb87c5f41bbde8e9132c12979c8a711e4fad4
F20101123_AACLZF wang_q_Page_69.tif
329c5cca5f7b64f899a248840eeb9802
b3ff9e4f491961e2d95a3092a62f39cbd9ae7eba
53987 F20101123_AACMBF wang_q_Page_47.pro
1dabfa062a1305db5767ac74f9533616
6cf4c3816d02fe58ae59e57ae1317e64fc2530f7
96374 F20101123_AACLUI wang_q_Page_16.jp2
7f59f53c069b5d09cb165f5c6471a47c
b756661998fa1161cae5c1c9ddadec39bbea3433
20770 F20101123_AACMGD wang_q_Page_12.QC.jpg
517618360559d061243f69aa49e642bb
96329dbe9ae464be2ff4d0ff69432f2e55bc8131
1699 F20101123_AACLPL wang_q_Page_62.txt
fb310d51e0bc8b9e36e0f078c83bf527
178c1451b686ef408923aef0ea4f9e8ea75b29c0
F20101123_AACLZG wang_q_Page_70.tif
444f54492d23b2b964221f74d2e34112
c4cecb8d5ead846b84b0cbe09171026097f76799
51052 F20101123_AACMBG wang_q_Page_48.pro
c04b7378d43c78abb81dc398769ccaa2
cbe693d0bbea98c726094d010a991720ea1a1548
978746 F20101123_AACLUJ wang_q_Page_17.jp2
7c10029a61951276c001ac0f5df792e7
9a464771e6f2b60a98c59c5d5b1896ea3b8ffcf2
21284 F20101123_AACMGE wang_q_Page_29.QC.jpg
050b351d9205f520ef684ba7766a5b54
9d62b0501b18e29f778a17ff20a40de71130369e
2863 F20101123_AACLPM wang_q_Page_09thm.jpg
b4181b18702c7717cdb52445651a9ac4
21721ed44947f46649055423f38d5dce1fc95e18
F20101123_AACLZH wang_q_Page_71.tif
44de5d4dcc6660662a43d02b069525cb
63792d8e769c5808fec71b99e0dae8c80766ac92
44416 F20101123_AACMBH wang_q_Page_49.pro
2dc87465fd126cea6c6867fd5593df3f
f2df88c1bd6d695f1232c95df8dd5d765dc68f7b
109351 F20101123_AACLUK wang_q_Page_18.jp2
b7d578ea1f7cbb546966f07513f3d51c
a06aca46c35a237548d6ae66f3739b9058d072aa
5889 F20101123_AACMGF wang_q_Page_37thm.jpg
3ea3d1a33c580c1e1455ff28c42370ef
89d035bf17a23e847038a00e94f9be47ba73c948
44203 F20101123_AACLPN wang_q_Page_37.pro
d74f7474d5ee96b15b2e32f5c267b9de
be54b5d21e11d6c1d3e9ab5ada47db8fbc635b95
F20101123_AACLZI wang_q_Page_72.tif
9e2ee4937506310ab8adbdd4e43a5b77
bcaea2ec9a0adcfff17cc113fe6f30e53b84b811
33476 F20101123_AACMBI wang_q_Page_50.pro
3a378df625549d40e9df3f29faf3d554
f92ae2020e3fd3d5c726e0a8c8597a18efeedfc6
109980 F20101123_AACLUL wang_q_Page_19.jp2
da78f9e9d2f2331bad6c0f61e95013e6
303138d8903f88ff168a5e5660839a538c22ddc8
5852 F20101123_AACMGG wang_q_Page_04thm.jpg
375571731e00fd975c8c0a2a4a74958f
41518467478f4e6671b5ca9a1c79851c401af79f
5045 F20101123_AACLPO wang_q_Page_63thm.jpg
e09eac3874cf85bb0645374459ed263b
224b2542eac72f3dcfb49fc507ee481a55c88a3a
F20101123_AACLZJ wang_q_Page_73.tif
282e6d4aa56da2e176515de468b60591
5b9eef2dad11b6f03a2dca72624346d0d205fe77
48104 F20101123_AACMBJ wang_q_Page_51.pro
e09588e8a7ac161daab7df28b0ecbfb5
7e6fb000bef295f3886ca6ceb7d80c0a411394b4
1051944 F20101123_AACLUM wang_q_Page_20.jp2
687db5a3d9eb9597ff5c17644016d800
4d2503766b984c843b97e8f1b346b367513f6f6e
6850 F20101123_AACMGH wang_q_Page_79thm.jpg
42f5a8d2ce40edcdec4ee90bd85e62d3
8a967fc5ad90e6b42c2a81640c8657a6fc546de0
10570 F20101123_AACLPP wang_q_Page_76.pro
ffe874c918901d3558256acfdd5e9129
77d5dd20b934878d4a8248bc33781c75f2ce966a
F20101123_AACLZK wang_q_Page_75.tif
aede73fa7b5d83d771df8460ebf08aab
f45c2a99ef5ac099273e2ebb0e117e916438cfb2
23070 F20101123_AACMBK wang_q_Page_52.pro
703f147805138ac238ee20fb11a5dc9c
5a228356b7b84d68d440756ffbce32d3c7b09c52
107071 F20101123_AACLUN wang_q_Page_21.jp2
54ed8a1d969ebf16ffcd8bf9ec60deec
9628d968f1a675adbbd52b40adedef445841358d
5618 F20101123_AACMGI wang_q_Page_41thm.jpg
1b16c8f33eb61ce794469c7241df559e
37bd2cdd1b0e77eadbb7b1f0a93d7c2375218cb4
68109 F20101123_AACLPQ wang_q_Page_74.jpg
fb98f96d4f43d7c83843a74c61e0143e
543ad42d3c77490bc71dcad52b76b122979c9ade
F20101123_AACLZL wang_q_Page_77.tif
9718f6ce1e409ee6b4cb3494f2c6e804
35fab6f01e0cb0957964503f7d627b2b0d859329
37976 F20101123_AACMBL wang_q_Page_53.pro
e450c442ed254eb8fc7860bc2c074a42
c2bcda7ab61b89ad41a4856229615681569e75aa
107694 F20101123_AACLUO wang_q_Page_23.jp2
c89ba81271e3fd66cdbca69c02fc7515
1c957011b541305ecf44c52ce4aa137913673803
15066 F20101123_AACMGJ wang_q_Page_60.QC.jpg
153695d353569fe73e4a15e85062c87f
d342154406e4beb93d54ed951e25580605fc14e9
1717 F20101123_AACLPR wang_q_Page_10.txt
0dd0002f2addd00907952e8e3791d93a
04457a5c210f821106aef9bd25828a689e9fd430
F20101123_AACLZM wang_q_Page_78.tif
e171c496391424d83c22c75ba87cf80c
662ae559da92ac72106be736dba65b6c79baae6b
50409 F20101123_AACMBM wang_q_Page_54.pro
b6a36e268c0a6326cfcf3d5ed39cbf4a
13a0d124a617d71fcfe10fa0a5ec053471367e2c
94338 F20101123_AACLUP wang_q_Page_24.jp2
919002d6ea0f61416dd9d8a0fdb3a244
f3610f953f6ef860ea7277597494e987859f80e2
5360 F20101123_AACMGK wang_q_Page_27thm.jpg
7fb5d1555a819ec25b0dd96ced9a6677
2d93f975657e12d80edc38792784f2c5df1470e6
52660 F20101123_AACLPS wang_q_Page_62.jpg
f312b6ad70f8e7d479c89de9947ab007
2be1f1ea67017b27908892e6278298f1536d0928
F20101123_AACLZN wang_q_Page_80.tif
ce52f10d5dc7aff43caa466d8c3b4c80
a596d4ba41a192e5c4a28ca5b78d157dc401b5c4
45798 F20101123_AACMBN wang_q_Page_55.pro
58213abbccac07bc8b21d2f703181478
99fd876c3aed13269da999110dd3ffe3020d78de
18690 F20101123_AACMGL wang_q_Page_46.QC.jpg
e67fb81e35ce450050b6acb95ac81ca4
6a8b1d7c129e000c9cfa5fa83a91c048e76cde13
F20101123_AACLPT wang_q_Page_76.tif
0a792974beed27848ecc3575c9582997
4adb57b07ce6a59fa5d94da41cfdff86899e53dd
F20101123_AACLZO wang_q_Page_81.tif
477fe1a01648b0b350e6f9959aa2b5e2
6038461118ba9ff816f1e067ccda0768544fba04
45190 F20101123_AACMBO wang_q_Page_56.pro
b99eca397c7d5eea2368d29cfb4480f2
723c0014096fe814ae8d0f07a3d79de5050e980b
67212 F20101123_AACLUQ wang_q_Page_25.jp2
08a2c55f772e2b2daa39f8da94dfb2be
87fe71df1a6ef2675be8fafd8b6f21de70c019c0
25409 F20101123_AACMGM wang_q_Page_79.QC.jpg
3bd53606d5a5e501eac4adfb893ccbb6
1da264d798f0459a9f3381f9aab0eea8b65c1d21
F20101123_AACLPU wang_q_Page_27.tif
37129173ff7646a99fa489fd7658bd3a
cc9cf71e0c30137bc8c21a9b952e4be2fa6673ba
7834 F20101123_AACLZP wang_q_Page_01.pro
a604283df72647a872639785890ca6db
2f3eeb85fbdba89ac448e5852ee72b605c7fa543
49568 F20101123_AACMBP wang_q_Page_57.pro
cd163f769a802734e64807e38b5b07e7
cfc1d4fc6001ef3ec557ce9b835b398daa4524aa
97191 F20101123_AACLUR wang_q_Page_26.jp2
4475b30b0e8b76bc0cbc13b6935b7b30
37ce2c2f9c659233b36dbfc0fdc58dbcd429bb11
25456 F20101123_AACMGN wang_q_Page_20.QC.jpg
a0908351f3cc433352762f53681a70fa
97d054ecf7519343d006124b0e7d69dd2357b8bd
6494 F20101123_AACLPV wang_q_Page_44thm.jpg
f17e61fc1eab65de453872375194f0a3
3f9e6161b083259d0fc20d948376c0fe2b08b15c
964 F20101123_AACLZQ wang_q_Page_02.pro
c6f9523afc2038db3ff7ed20b30ef6a2
6435968eab5b8cf7b5e0541f47316d7ca9519674
47916 F20101123_AACMBQ wang_q_Page_58.pro
e77e32808ae118c8549cd818286b771a
35fa845623f8b419542310ec2aabe68148f84105
85609 F20101123_AACLUS wang_q_Page_27.jp2
f585808decd10c8c59d582cb35dc4118
49f6a41ff5f581d9f0c76324c22a2d1ba34f29e8
14348 F20101123_AACMGO wang_q_Page_40.QC.jpg
d46a8943423fe95a527c16fe4fdf3137
613f18f0897ad5bb8d33a1f1e1807095c15ead34
1897 F20101123_AACLPW wang_q_Page_24.txt
f9a9ea084e2d75da91f2a747ad2b2132
f77b8d2dd881f3cf0f02ecc644e72ac4db4eab78
622 F20101123_AACLZR wang_q_Page_03.pro
3a062f7a1a0e7f5d91490f474570fee2
47d527632524adf29ed8340f4941e78c7077dfa6
36596 F20101123_AACMBR wang_q_Page_59.pro
805232a1cea881a30f552c878fbd1fd8
ac206540f7cf710c53c0de73f1402e5388d87687
77416 F20101123_AACLUT wang_q_Page_28.jp2
8797530db4f5216319ed773c8ea51943
b632b46469516c67ff1bff75d036c2b971deda21
20602 F20101123_AACMGP wang_q_Page_66.QC.jpg
4d14af87de8f78a9399f0336e31f3c91
4ae94257f1104868287dcec7161fb7ce98b333c8
43630 F20101123_AACLPX wang_q_Page_25.jpg
a45366c5bfb5522f42cb3acab8cee39c
4a53e5f84c29986fe1437fc66edb7593edb86d99
42308 F20101123_AACLZS wang_q_Page_04.pro
6950e667092b58253334f8c901c0968d
6e338093ed06208335b5b78c221e1ab4e166a371
32105 F20101123_AACMBS wang_q_Page_60.pro
5df03d1cc7314a973ff53092e1e8e153
a71b04e44f935637cf31c3bf567b58da90a8b872
933645 F20101123_AACLUU wang_q_Page_29.jp2
5de86fecd955af1681efef82c13631ff
dfac35e660bade0c45508a77f7b374665b5e18e6
18773 F20101123_AACMGQ wang_q_Page_27.QC.jpg
3eeab324b299c7963ad9554bc02ea932
b93a22d0b10542e5eb4b7895965143940ef5b326
23280 F20101123_AACLPY wang_q_Page_18.QC.jpg
bf7f82137a698b431ac0efac62717657
f64322abdef6aedc35f2ca5cde40255d49bd138e
6840 F20101123_AACLZT wang_q_Page_05.pro
1d42e8ba97edecc373b524ab1275608c
3cce7db10f459fffa3a6dc51b7b77840ae3c11af
34068 F20101123_AACMBT wang_q_Page_61.pro
8fa1578e49eb980b3801f22fee636997
8c43e478f410d8e8089dfd00f8b18366123a00a1
102194 F20101123_AACLUV wang_q_Page_30.jp2
2d3cd181bb8879bb9424ca40542af0c6
f3cb9c5a4d030f1077123ee7e8e5ea6c3314fa1a
5178 F20101123_AACMGR wang_q_Page_28thm.jpg
e73b55c0f7e2c4eba1e6e449e2ce5036
02eb10d5056329430562d13c695c1a4df1dda401
59355 F20101123_AACLPZ wang_q_Page_69.jp2
84147e7b4cc909160b7209ff51f11625
3db1a0a3317e6fe3d4e520884755516d6294008c
75118 F20101123_AACLZU wang_q_Page_06.pro
7858e237d0d27c2099e327c29acb18fe
a16a10c9a2fa63a083c50eb908380ef14dddf973
35499 F20101123_AACMBU wang_q_Page_63.pro
ed9ed629bd90f106d91018249743fde5
bbdd4ebf636ea8745a786c0432d40684c103810e
909674 F20101123_AACLUW wang_q_Page_31.jp2
32953862a9100d3f22142611b1f6fbca
92b42a2cd171940bc298dabb9dc6eafe7fb9ed96
22257 F20101123_AACMGS wang_q_Page_67.QC.jpg
c67ee3238d5b2634b4484eb3158c8c0c
6d10714dbff647dfc1e658b804125d71128b7078
43043 F20101123_AACMBV wang_q_Page_64.pro
cfbb1e5c597268065cc46fb949b46ab8
cf6cab11c1dc0864707860feac69741748f56600
104695 F20101123_AACLUX wang_q_Page_32.jp2
656a1e9c2a9a23a1696ef5acd2fcee40
ec766edec03e2a4814ed5018c2c028460b2d84c5
18028 F20101123_AACMGT wang_q_Page_72.QC.jpg
92758ed30e00ff7e302fa1ffb7030f27
a0bb0ae6987fcbc336f0af543080615abdf19407
37554 F20101123_AACLZV wang_q_Page_07.pro
e5dc300ad1caaa091f677772c3725925
94bfcc0108b0c826f3419d18aed1e4859594eed0
46088 F20101123_AACMBW wang_q_Page_65.pro
c3c39a88cc90a52f80f5d755f74301c3
411dc20b85fec27269d16df2dd3f9ba1c1a58f67
68296 F20101123_AACLSA wang_q_Page_30.jpg
93434c497507b66fbf37a8c09205e529
12c00ed930c23acf8cb03c7b8c7a063127480b92
1051970 F20101123_AACLUY wang_q_Page_33.jp2
bdddb77e0cbc0cdefce6f7dbf9799b17
65698433d43a7f7486d369027edf93bb498c3474
5874 F20101123_AACMGU wang_q_Page_16thm.jpg
6a53b174428d61ea3eca82851a4d4e15
47b98ca38da57ac54c4a3d4f6716b68a45f997dd
17506 F20101123_AACLZW wang_q_Page_09.pro
76a13b41275dc428eaec739e64d45686
ba792a424e5e954514e7ba6d9da3c6ba1d786c94
53661 F20101123_AACMBX wang_q_Page_66.pro
3071590f94f5121b73b93f37a7f90905
c13eec24bbeecba1015b3b81f750754f16eb59a3
66043 F20101123_AACLSB wang_q_Page_31.jpg
6297213d4ecb12a2f270f5c302fa26e6
812589c0bf76f6e9a76f5b627ed16752c955340e
98513 F20101123_AACLUZ wang_q_Page_34.jp2
85ef4c05e01b53832fdc441f0e99b79d
8bd223411b958ef3690173637c4141917dfcd83c
18501 F20101123_AACMGV wang_q_Page_71.QC.jpg
14f2ccb3ad04e2507d5950cb8468e742
a5e33fac370671739dd42c447c1da57cb173bf80
38716 F20101123_AACLZX wang_q_Page_10.pro
9d998d34cd775fdc177ac3485df81264
122b05fc9dccc2f673396977973a1f422b8acb3d
47774 F20101123_AACMBY wang_q_Page_67.pro
edbae253f96b9ea3309e7f22c989e8c6
4eca0634dd28c571b4cc05dd5bd6e5f8739dfccf
68083 F20101123_AACLSC wang_q_Page_32.jpg
cf7da004eb1ffe92bd3124e9a1f42e66
4ef3742669d32331da13573a33ca7078da510e89
5945 F20101123_AACMGW wang_q_Page_12thm.jpg
fb15138b6f87b6f46c33ebb89863fcfd
086b92a7f9f4868361472c9ac6713231fe96e7fc
4173 F20101123_AACLZY wang_q_Page_11.pro
1562bdaff992fed5088de956e3fae50a
0fc5e2abf9d9a6c24a4e0df5911ce8ef039a5937
51715 F20101123_AACMBZ wang_q_Page_68.pro
3657eb437e307d54827e5c873b70d40d
3742e7e6ebbb37131e78e2d9c4ac8f380974515d
75292 F20101123_AACLSD wang_q_Page_33.jpg
57b6184ec454f108c73758d047d603c9
c01254328dc14ef9ef6250258b0a1776d58162ce
F20101123_AACLXA wang_q_Page_09.tif
892b02f959ec4e97bd80bc13e976d992
0c73a3fb929301a59775ca861682c2c2a1fe31f3
21367 F20101123_AACMGX wang_q_Page_31.QC.jpg
5dc49831f17fd46e0a532ca1c2686574
1d5ebf543413923ebb0da6bb01a07b0faedfb6b8
44908 F20101123_AACLZZ wang_q_Page_12.pro
8cab42794e93d29053db459d2364a07f
eef54bd923eeb9475e963c87202cbfd54d12a665
65171 F20101123_AACLSE wang_q_Page_34.jpg
717c29b0054021630dd70cfbeb2fbc26
67154f6406446a6180147be674764045bd2df0d4
F20101123_AACLXB wang_q_Page_10.tif
50be36b3598c2aa64edc7b21701faa37
f1079160a6a4d2c6fcc971d7bea9223ae1227e05
6132 F20101123_AACMGY wang_q_Page_36thm.jpg
3361990e77e27e5395e117da053b9787
9522184dfc4685950569eca5faa369ea8d868891
69158 F20101123_AACLSF wang_q_Page_35.jpg
53c34035995fac973ddd238bf0cd6b72
bedc2d7c049aeb7c56124f84a104a3d98c4616c0
F20101123_AACLXC wang_q_Page_11.tif
b3ce5d3a603af7ac2bf9d63fff65d298
d7961cc6d351d1fc3b018ebd9eb8952578e7be8f
4963 F20101123_AACMGZ wang_q_Page_70thm.jpg
5d528fd9d060b7969a6bf48a404630b1
9973cf239d4a34fcc0bff0cff0e4d25990cf563c
69068 F20101123_AACLSG wang_q_Page_36.jpg
90d93b976c6ad7e4828440c0c9c093d6
9e051ddf76d28cf527f4dd6f5a332beab57a41a7
2329 F20101123_AACMEA wang_q_Page_44.txt
44423fc0bcf4fca303f5cc7ff077991d
098b6de74b1821dd74a92a454f18c8ba187001ca
F20101123_AACLXD wang_q_Page_12.tif
d9fec1c1c2d68b7e1755845c8ef40a1a
1de920839e092cd617ec79bfee7ad78568d6fc20
61428 F20101123_AACLSH wang_q_Page_37.jpg
1ad107be2e2562fdd50a4d95e3d73ce0
d98fb163aea3e2a3a64560acbecb4d49c5aa0ba2
1830 F20101123_AACMEB wang_q_Page_45.txt
b6a770190f2c2f2f157226f6846650e6
28dee2e489a86e7d54b276f89562301326ef8701
F20101123_AACLXE wang_q_Page_13.tif
dd82da9b50de1224dc920f7935cfc16b
a69908fd0a313874da940f127fbf4372885ef4d5
4442 F20101123_AACMJA wang_q_Page_40thm.jpg
95ed05fb639b7d777d82e193ba369553
581445dfc296e87b23df50f7a90e07c15da2d047
63363 F20101123_AACLSI wang_q_Page_38.jpg
f6f7946e3b82296fbd6460ae75fe5ec9
fb394ae9e5cc2d615dd935422bae6ea706dc4fb8
4355 F20101123_AACMEC wang_q_Page_46.txt
162fbd006d765284762ed65241c85322
593687d3f656d43080f1a0990c0b2ada1235d7eb
F20101123_AACLXF wang_q_Page_14.tif
bf008ae38e9151beac277bfe4c56dd3f
04a02152b4eecbd2e6c27004a35ee08074e78544
13623 F20101123_AACMJB wang_q_Page_42.QC.jpg
bd7679d4cc7acad9b3c545d4e7970372
ad8795d324f05c1ee2afdf4e8b79b03150bc633c
66248 F20101123_AACLSJ wang_q_Page_39.jpg
82408e53cd7a0eb5fe9b88dc2507f8b0
a7239f261df999f0ad2390c8c8e7f9f29c859e9e
2110 F20101123_AACMED wang_q_Page_47.txt
c3c6151cac317deb34e0174311895f79
1c7ad14824f2da2dfba083e6e535dce585756280
F20101123_AACLXG wang_q_Page_15.tif
9149242936d1ac71efe87317c4e9c6c3
13ea41c308997814021e58a7f5f6939ea8a51f26
18458 F20101123_AACMJC wang_q_Page_43.QC.jpg
07170ff462d1a8ddb5f7e3a2ff11bce0
25501e3b87a7e667e2c2658ed24512c75f6d334f
45692 F20101123_AACLSK wang_q_Page_40.jpg
36393e7768265575c8800071cfa4e84b
b2dc769b260a9e00b542137c6d2032ed1ddbcd7b
1764 F20101123_AACMEE wang_q_Page_49.txt
5d0bdccd5a0bde21ded472b49bb10a65
1e0b8ed894d96aaf65c9aed834edaf593d3a7055
F20101123_AACLXH wang_q_Page_16.tif
c06617da4f297721723426ee3c634e8d
3351728d7af0957fb5dcc175ab7d1a9102b59d98
F20101123_AACMJD wang_q_Page_44.QC.jpg
c1329ba29a714c6d61dcdb89aef100c3
8e070d0e6f812ab7e82c39c0cc806d1f67e9b39c
58338 F20101123_AACLSL wang_q_Page_41.jpg
053592d0316705c3bdb0845338c4fd88
e5c2ffba2869a84ce79ed26b60257f2533e4a7c4
1576 F20101123_AACMEF wang_q_Page_50.txt
0096290474a54a82c4e395f850da5f4c
7340de7ed6ce6981b942858f4d07297f1ed93933
F20101123_AACLXI wang_q_Page_17.tif
46ba28968ae61ff36f2d1b9fad5d2212
7dc193758f95051c74ff8111d64f1afe97c16480
44332 F20101123_AACLSM wang_q_Page_42.jpg
c9b23638b80dc02f97cdb55d0cf36ad8
acc3dfbb26d30ea3771436a200608f2824863a91
1899 F20101123_AACMEG wang_q_Page_51.txt
d92c749d63e2b1d6a7efdf09e594fc69
f7094752dac6326eb6f7fb40edf8694cc23e70cc
F20101123_AACLXJ wang_q_Page_18.tif
756fdf25cd1d13c3fb5b44bef0c35e71
fd1888957170f6beb511bbe77c2c7257722bcf0d
21422 F20101123_AACMJE wang_q_Page_45.QC.jpg
e98c0f4755a51711330b2211759ff1ce
858495a3aa7f48df45dded51547dfdbd1c40674a
55634 F20101123_AACLSN wang_q_Page_43.jpg
273fd784f89518aee62ec6cc21fdfea7
fdd6c288e85d00ad346f32955a1856c1a3edd143
932 F20101123_AACMEH wang_q_Page_52.txt
dac8c7e4c71edb5d40e8a7acee2b1f4e
303262570b650fe8170f03a3169b938ed42d39be
F20101123_AACLXK wang_q_Page_19.tif
67fdcb4f13397c9971c972d444cd0dc1
87647b6c5e3204bb13df27f261850a1db24a9b40
5590 F20101123_AACMJF wang_q_Page_46thm.jpg
abbd32a47122df6c8aff2cd733af4015
caf80689b8d75392513090646cece9b099a5cedd
1610 F20101123_AACMEI wang_q_Page_53.txt
efdfd38f2b7ab072ab2a5f777652592f
e7612e67026b86cbbba0fb15717fd556f7e43761
F20101123_AACLXL wang_q_Page_20.tif
7f267357a282e82ad1522fa3225083c3
020c4c34b207065a877005e637626a24201614b1
6758 F20101123_AACMJG wang_q_Page_47thm.jpg
da026d98bb73cf48b8f2bdc9f86bcd16
9a6d4e6639782dfc227a61ddd117e3c83467521f
1985 F20101123_AACMEJ wang_q_Page_54.txt
309c55bac62542999db4f5eefd34575f
f18d0173919cd2e9b6da2307b947f7415e7cfab2
66457 F20101123_AACLSO wang_q_Page_45.jpg
8c20fc076d2e5816c82905ba959b5e9e
6e1717c81e08ea3ef2a125b5a1dec9208c5a41f6
F20101123_AACLXM wang_q_Page_21.tif
533be572ea71b66ab46f555d1f45ee0a
5ac262db5a60098c0bb04e9ebbb74772fd6758b3
6548 F20101123_AACMJH wang_q_Page_48thm.jpg
ac3776eae6a70d60548f75fb55a33368
b68c66606e588ea19e081e50dd5ffba78d6a3e91
1811 F20101123_AACMEK wang_q_Page_55.txt
2fa67a267ce18f06ed2ba4dd187af49c
f1125d84007eaec272f8c8a4a02f939c563633b1
61941 F20101123_AACLSP wang_q_Page_46.jpg
0e0ffaf293847c1eabdd172c66ddcc18
8ae8290b80331ef9a71dd51c7cf56df641e4b10f
F20101123_AACLXN wang_q_Page_22.tif
c0c62c12b75588794e51c7c1e0c1ed69
28185685f0bddbbcabae3063a8478ed9e816cb97
21014 F20101123_AACMJI wang_q_Page_49.QC.jpg
93cd27fb677bbfe375b55b6f6a50cff9
60f67f7b127419c4f43fc0b3d035febe224413e5
1790 F20101123_AACMEL wang_q_Page_56.txt
b5e67a8253cd4b413342899fc6f447b6
68ce9002f729d6ba9fef7aaee6f36e728461e72a
76021 F20101123_AACLSQ wang_q_Page_47.jpg
6b2b8f86dc770270bb317ebfc1a14d5f
3fe6e05b378041191e5f553e9a056d2922f3b120
F20101123_AACLXO wang_q_Page_23.tif
b2c4ff1de7b13fa50f7c9fe292f9bf3c
9b1e62eca45c2a6bc1002ac0d4c26d23c3874f5f
5965 F20101123_AACMJJ wang_q_Page_49thm.jpg
597d90c297d9b38b536386631b541b29
9daeb244dc4299984197557f6e95339108907bd5
1964 F20101123_AACMEM wang_q_Page_57.txt
5645b1cb915d33b16019d209319a056f
eb575dc3c900dce9cf9f98a490bd76a3fdfcee28
65412 F20101123_AACLSR wang_q_Page_49.jpg
6430f3c99cd663007ec81020fb0c3a98
cb147bf681a91f16e491ed84a175aeed8559c0a8
F20101123_AACLXP wang_q_Page_24.tif
b961d29c8b48b4f97d3690928db13744
95cd37e69dfa7f06674513112a1776b970a46b06
18409 F20101123_AACMJK wang_q_Page_50.QC.jpg
4837d0fa8668d6f2f46755a7425cab89
10d7cea805999bd683cbad2104e49ada40c39e7e
1935 F20101123_AACMEN wang_q_Page_58.txt
571d83b69de26b895dc2d82f2ca9a2ca
8654a3d61cb11c62d4d8b001adc123f94e37b215
58926 F20101123_AACLSS wang_q_Page_50.jpg
828a602ec3c68fa312f79beb67d25d9d
2b4fba396f5fd1a13588d75f1239bf035ccabd0b
F20101123_AACLXQ wang_q_Page_25.tif
ce9bbe905dc3394b782d00a10bfd6853
bfebc0ac4179c1be0490394218b37b47f6d2e37d
22067 F20101123_AACMJL wang_q_Page_51.QC.jpg
2773ca6fa27ecabbf1fc0aa357c71bc6
c1f9d00a80993e7d6c6a991715e90ff4d83e00c9
1702 F20101123_AACMEO wang_q_Page_59.txt
e96cd8697e54a69d306ba4dbf72dc6e7
f9525e6965f1d343840bee4b69871aa9c5015a88
68585 F20101123_AACLST wang_q_Page_51.jpg
dadd9a671688fd6c0e1a01f09829b2c9
8fa7a22dffa3403f1d7eaf2898c17608cd79ee2a
F20101123_AACLXR wang_q_Page_26.tif
f2660552ff0e5915e693af398cc45265
75da9fa30531976b00a0186a6bfd05b3b7e1361d
6558 F20101123_AACMJM wang_q_Page_51thm.jpg
7e4aea1d3b929d7d6dd1052991b6455f
092d3915a1d69367df506422dca7957c27cd4b29
1516 F20101123_AACMEP wang_q_Page_60.txt
cd13a6c518503eef09d2e727422744d9
c46d8a229565bae06f7261a9d6b27648fe5ab680
37111 F20101123_AACLSU wang_q_Page_52.jpg
9fc366f5ac32f2be19abfc8d7da92053
0c1b25739c3ec25c9d0a954d40a3d0d4f6037a52
F20101123_AACLXS wang_q_Page_28.tif
e4329e3c8dea7286ca75468575394ce9
b897a43e45e443d9e75b795b751650e596c25632
12098 F20101123_AACMJN wang_q_Page_52.QC.jpg
f56af55a8f6f8f2a84665fa333b16f8c
fd6b1167d37b656c7661b80d9c37a81703c22a8e
1580 F20101123_AACMEQ wang_q_Page_61.txt
c6aa6894fb2abb17f256cca50dbb104f
48907b6e81dc2a1d35eed7f50d9716e150b1a25c
58162 F20101123_AACLSV wang_q_Page_53.jpg
45f79de5b71e4fb3c656be1bc530a1b2
e06435a1bbfe4a83a9b75e4df4fe4db9d90b89bd
3511 F20101123_AACMJO wang_q_Page_52thm.jpg
ffe61af248dd53f2e9792edc85993720
a0b56db80595a447d82c91d36da45857aafeadc9
1522 F20101123_AACMER wang_q_Page_63.txt
10982e59e59e6996e7a95bcda13d3f8d
5db212ae17be9a3f5586edcf440e4ea681b40332
72650 F20101123_AACLSW wang_q_Page_54.jpg
8df78aef25b56bea506d987e09e261bb
aa74979b51232cc1c33e13da8fad22d216a645a9
F20101123_AACLXT wang_q_Page_30.tif
c3495d3ddc426a03e8727a6b448a4686
ba2fa685290b380e5845af0ca5a86d4524ec4719
21254 F20101123_AACMJP wang_q_Page_55.QC.jpg
51b636c29a8f6bb1a73d53369320bd36
bba4fe0f98aab59fb07a2c3cc4506c8fcb0d122a
1793 F20101123_AACMES wang_q_Page_64.txt
c7a13c6fdd51f5f9a1fee97292b1ade5
a7af65713cf02dc76cb9be18d41843f8e42bb3f7
66957 F20101123_AACLSX wang_q_Page_55.jpg
7f9fdfd297de46084831596ec579dbcd
8951a3e09db55fc76c87672e5032bf54bb29d224
F20101123_AACLXU wang_q_Page_31.tif
92566bc9f9dc51af3851d24b45a71080
59311b20c9046e1352ce4ba28bfcf718c81a273d
6269 F20101123_AACMJQ wang_q_Page_55thm.jpg
9a61591413ae8d3f3c980186b09847b4
1fa461c7088693829f6d4223352232bf22c13b74
1852 F20101123_AACMET wang_q_Page_65.txt
1ff218456d1df83601b7a7312ee56e93
4eb5de92ca6807d2c04a919af29a21d1079a6d41
66335 F20101123_AACLSY wang_q_Page_56.jpg
d832e3b516bab710dde4de4b95a953a6
ddcfd1e4065db73cba5b8499a350bada7f39b559
F20101123_AACLXV wang_q_Page_32.tif
2dc755369bd7eac159a88907587c3064
3972b817b392b9451b45cfafcb73317f99cacb89
F20101123_AACLQA wang_q_Page_74.tif
8338423b17d94988ec5ec8d3e1b641c7
cb0178c47b82253cc5f98e7df7b6b5516298ae87
21360 F20101123_AACMJR wang_q_Page_56.QC.jpg
60346cc4dc8ef9ef803fee51a3924688
6b7c918b8982a884d9bfdeb2400fd0368a675235
2576 F20101123_AACMEU wang_q_Page_66.txt
f5f6c2511b1fffdacf1009bf610854e0
e5fc06b243f5f47fc03c48e533542e33d86a75d0
70793 F20101123_AACLSZ wang_q_Page_57.jpg
3da371ee75a5b35c726c5706ba7282cd
d85eca26a60275587e8c68dbf039f7e10ac112b2
F20101123_AACLXW wang_q_Page_33.tif
2fbbd14fcc2d5fb6f659b6a0dc291841
dcc2716c0346b2338cc140a8465ca4edcebf7663
898774 F20101123_AACLQB wang_q_Page_08.jp2
9cab4d5e07f1978d1fecd8c87dd0143f
193b84859657682987b5b24152246a206bf8a184
6090 F20101123_AACMJS wang_q_Page_56thm.jpg
3fe60931ff7c8881d386dca91fccbe5d
4f39ec0da01824c6764281ed5ed777df1e2e9351
1895 F20101123_AACMEV wang_q_Page_67.txt
1040cf92e16bc50092d37ac011e08ee8
d538167d3edc19f9897b6b971d4864b475a4cca0
F20101123_AACLXX wang_q_Page_35.tif
dd0f64409a600fb8c45d3f1306beed8a
3ca569da8751d1deec744a0fa9d2a8f1be757ebe
2323 F20101123_AACLQC wang_q_Page_68.txt
376a5518514c7b41efb4105a1964df9e
7fe7f70283318f50c09782e259196d1bb247b512
22900 F20101123_AACMJT wang_q_Page_57.QC.jpg
a5ca7b196667f42212f9c395c286b7f5
d3a94576ca6833c6c3df93ccea3519a607c01aac
2085 F20101123_AACMEW wang_q_Page_69.txt
95d3bf4628ce83e40bc78021fc2473f0
b7bc8ee8ae3f42aee64cc4761b45e24cf6dabcba
107007 F20101123_AACLVA wang_q_Page_35.jp2
9a5f781a6db5fc2bbc75f63140af7a41
820632b468499b072b3c75d4e0bb8cb54b8953bc
F20101123_AACLXY wang_q_Page_36.tif
808a1e6502beba8ae20d7dbc00e9e8de
333355273fa8c0d6c191d24cc97567e3e501c704
18007 F20101123_AACLQD wang_q_Page_62.QC.jpg
83860af35e3363ce9940bcfe7abf8a36
40d006745c2d84ea56b9dd2b6ab18b392311d410
6343 F20101123_AACMJU wang_q_Page_57thm.jpg
fd515cf4ba911caeaf568a6e598b3b70
0a5df45a8c5ea04dc96fd11039f90810f85fa6cb
1229 F20101123_AACMEX wang_q_Page_70.txt
757614b1f4f7a6c0644ee46c418952e7
a785acd7a4196aa06e682cd3428f92a9638e1038
103167 F20101123_AACLVB wang_q_Page_36.jp2
9c6b486f0f9187eac16dfae41f9e6151
3f922b3390cb242ed52db51371fbb8a8566f383b
F20101123_AACLXZ wang_q_Page_37.tif
b19be77b6b697cb9875dfdacb6b7fe23
4aec6fac54b65508928a72b3bf56dbe08e646f5d
6476 F20101123_AACLQE wang_q_Page_15thm.jpg
7e2ead45923426cffa85188be83e6dfc
b166f4f7cc46041c77f0d01a3f5a682b1ac709d4
23096 F20101123_AACMJV wang_q_Page_58.QC.jpg
3f4fbdeb4e175e673d3082d958a6ae27
8a6b85c4af09cad2bc5cc4017a4496c535e000ff
2187 F20101123_AACMEY wang_q_Page_71.txt
b491ef1c41d91aaa749a1cc5939bbc62
1759e2ae76be2ef707af229ae8ebf84e8b857681
93221 F20101123_AACLVC wang_q_Page_37.jp2
674b68e08f2c8fbc5a7de1f0b2ac2fd3
bf557e97a503fd1e51831060f584341f988dcdda
F20101123_AACLQF wang_q_Page_05.tif
d2689d218d0591847407beb3f4fa9bd6
0093890030a010d9843f6ba9dc26775ee091ac6b
16755 F20101123_AACMJW wang_q_Page_59.QC.jpg
02557c5f99bcdfcb5fc7d50e085bb84b
14321358e2cf2b3b1f2b8846b0082b5dd2e2e4dd
2349 F20101123_AACMEZ wang_q_Page_72.txt
c092b3988bbd0c9037f5b906ab95f8f6
767fc27f079d1f42f50d117c6388330632ef3cfb
35445 F20101123_AACMCA wang_q_Page_69.pro
7c2824c1a9d6f2df6a2df86b0eb48e26
24a87071a001d34ea2050d3235cffa34fa5f5cd0
96520 F20101123_AACLVD wang_q_Page_38.jp2
2ce2815d05e52131c11283ed4589c33a
955d398ad81cb916bda1a4edd0ed2cf29cc36a73
71563 F20101123_AACLQG wang_q_Page_48.jpg
f3fbb93ffde677f1734ffb316e075783
39b1f169c6d1f09a0ec7c90063f613252e156397
5380 F20101123_AACMJX wang_q_Page_59thm.jpg
329b80196f1be42f41de0fbbed1aa9b2
5c3f3f5ae6ba9a6115d1ce24db22f64de2c0863b
99364 F20101123_AACLVE wang_q_Page_39.jp2
7da2febc7b01194a879468c8c7706967
f801dbb95f87f82d0500cacfc0b303ae8efc0452
51178 F20101123_AACLQH wang_q_Page_20.pro
095ed4a6fa5318b6a5169bd844d8c388
b9a480d763044160a748cdec3c24250576466ea1
27714 F20101123_AACMCB wang_q_Page_70.pro
f6f7155ea3b5a1593b803375e22bbbea
736439b642a9b8e7c26ad54135e211cfac13fbf9
16093 F20101123_AACMJY wang_q_Page_61.QC.jpg
68274c3d8432d84a51b090a54715d23f
31eb300ca2b01bbfd6084746370d4b57bceca7f9
7013 F20101123_AACMHA wang_q_Page_20thm.jpg
82c319609481658658a8ba17bdfb751c
d56d28bd7d356f1a1a30733e75cc2827bd61ab45
69162 F20101123_AACLVF wang_q_Page_40.jp2
6190cc0977843e4311238a327119a8ce
b9288a7153eba0ab8e187307e0f5b1265f5c959f
2009 F20101123_AACLQI wang_q_Page_48.txt
d54106d9473938badeb3a587d472309e
1d1636e7eaa2b9a2a53dbab330051a0d2c6939a8
43440 F20101123_AACMCC wang_q_Page_71.pro
6b4be3e67b9547f04b7b4788f18af23b
49262582a679eb47425c0cec55213f6ac5af6271
5029 F20101123_AACMJZ wang_q_Page_61thm.jpg
18032e1374f88d869f438f5ee6fed5e1
af0627f49ad91fb6ebe8684ad6a34a2d26ce5cc8
121353 F20101123_AACMHB UFE0011601_00001.xml FULL
eb33110098fce58dd1833fc7f7c1569d
0a2ccc60c65e3a4abfa2b23bad37f32b4378f5d4
87614 F20101123_AACLVG wang_q_Page_41.jp2
fe77e93f5e2d7ef53a957f9753bef26c
0b3196543f9aed64f125ed7ed625191ac71d9300
20060 F20101123_AACLQJ wang_q_Page_41.QC.jpg
2454b8a1daf1dfd46e4072d498af92e1
44a54e876adf8fe7b1aa15d91a2f527beb00bb64
43597 F20101123_AACMCD wang_q_Page_72.pro
8c1053fa00c1ef9d454b319433326698
e5468edf34ad0cda975b39a522fe26523c388c60
63565 F20101123_AACLVH wang_q_Page_42.jp2
9d4343311b0c64515996acef448dd205
b5a6dacb334ce43fdb83e99084d2a59831c95325
F20101123_AACLQK wang_q_Page_29.tif
12bbcb61ec4e5f3a8efa63c4c6a26290
3c657b7d68463561410ef53d0a3f81f2a065b86e
40857 F20101123_AACMCE wang_q_Page_73.pro
a84f28e23cca026f9ba6598806e48f3c
338055f5d091f6bbc5087aa47a6e39db5c9699f6
7135 F20101123_AACMHC wang_q_Page_01.QC.jpg
491990772daf6ee32a14ef940b4547ca
64484b88e5247e89ee731a6502a9bd8f442e478a
766758 F20101123_AACLVI wang_q_Page_43.jp2
391aa8e038beab8d8e1e01257fcccd17
6cfd8c3f22d49dde3373558d9b2b71068a5990c7
6134 F20101123_AACLQL wang_q_Page_65thm.jpg
820765be0eb085b694ea12d847ebd1b9
42d4aeefd6f92480aa86611f85b287bd7b71ac4e
46994 F20101123_AACMCF wang_q_Page_74.pro
b557e998eb1998b879d68ad36382dd73
a7a4f10a92c7bbdca020784308d84ff3d91c28bc
2390 F20101123_AACMHD wang_q_Page_01thm.jpg
51bf963c15de00911d41839c291492f7
6bba8233a367818668dd2acf698dfc7fd3d937b2
116504 F20101123_AACLVJ wang_q_Page_44.jp2
a8a0e092a65d465bfbc4c4d9a6e09d3d
8ae7032a5692da1417a98b7d1c792b78bf90e7f7
49116 F20101123_AACMCG wang_q_Page_75.pro
25300f3e80fa01ffd04234304e17896f
d49d78bbc0b925ccae8fcf11872ddf6304aac417
3224 F20101123_AACMHE wang_q_Page_02.QC.jpg
b473068c5743f8736baf85b123fa83d3
28d5d100e5a2432bdbeb9c0dfa736e4fc0a2817c
101317 F20101123_AACLVK wang_q_Page_45.jp2
d03b8cb710b26e391db35dcf6ded6946
86d6a4b90b871ae7c6f8938c991a44b15ff4cad5
21193 F20101123_AACLQM wang_q_Page_73.QC.jpg
03df4e2af621d4e7017cdc043ce41b7f
ca8809382d300456bc5e88942712f97387146808
48707 F20101123_AACMCH wang_q_Page_77.pro
7ee1abb902d30b282d228eeb9a9cb614
65255db90208d73936afeceaebc2a140fad27ef7
1363 F20101123_AACMHF wang_q_Page_02thm.jpg
ee1db41df3f490f7c1abbf88bed80f49
fa1dbf8e719abf03ff3a6ec99d5898cf63b6e36b
96751 F20101123_AACLVL wang_q_Page_46.jp2
7ac8628376f0ccd4e803127c74cbc0cf
a8d450807bb1de15ef5f66a5f3eee725f4268baa
19751 F20101123_AACLQN wang_q_Page_53.QC.jpg
d968246ab15400f1ce8ee67a01b2c494
14a7df67df011fbe06c66dc54dade0babb357565
58479 F20101123_AACMCI wang_q_Page_78.pro
7c677ebfbb18637662c36eec4b673fab
7464e2f07b51b90d1fa131be9466c22533061a98
2950 F20101123_AACMHG wang_q_Page_03.QC.jpg
af5b8e12fab4c07beb95350acf012ac9
8662765354e19072108fd1bdd2993394b57f070e
115205 F20101123_AACLVM wang_q_Page_47.jp2
633ceb1e71486ee6293a6e85e804bbbe
e08188230b3ce052089ce4aafd6c14ee6d805218
105460 F20101123_AACLQO wang_q_Page_22.jp2
445721672a3ddc46328f02b6de6c6632
2cb6fa7f82814b74ef701714ed8caf3d02474e21
60522 F20101123_AACMCJ wang_q_Page_79.pro
a6636200a814de0d11158b8c49d52d31
43643f5346247719b28fe33fa8196b4c145aae19
1280 F20101123_AACMHH wang_q_Page_03thm.jpg
cc0149a34813e03d7a040fd9042bbdca
2b39c5d4983b934367687d455d6d625c40dec540
110040 F20101123_AACLVN wang_q_Page_48.jp2
fc83980a41b671fb157d9614297718c1
c531ef70c160389b52b91bee7aca478444a30cd6
38853 F20101123_AACLQP wang_q_Page_62.pro
fb3a71bd775680001cc126766ba642de
b06faefda9c223dd8c6ebe060357ed22334b77d9
33003 F20101123_AACMCK wang_q_Page_80.pro
9ab290f31b487f0bde11dcaa3c999bf0
0c1ca416fa307a8ef35a7d6e17e726983fdf51c1
20443 F20101123_AACMHI wang_q_Page_04.QC.jpg
f9e42aeb4093c13f1331bdf5a56651af
213e3f74c37944a2b6b138e2bc792104bacb9206
97901 F20101123_AACLVO wang_q_Page_49.jp2
b0852928bd17c0d0243b286740bfcd6a
f873c5a06ce4fb314e9f336bb70b3db05b91d10b
F20101123_AACLQQ wang_q_Page_34.tif
2c2275b44eac41e410d9be987ec597bd
a1bc3ab227078c3db6d703f569a9503b50a28036
19467 F20101123_AACMCL wang_q_Page_81.pro
65cea217403924d934b7eb82cac91709
66c8948d58c4a3f4da53e72f82d9f9e0d65bd6d8
5794 F20101123_AACMHJ wang_q_Page_05.QC.jpg
1aa6fbf61387f9bb569fe5400c5a303c
39bb1de20f7e1b21ea5ca6df71feace10590bd54
832977 F20101123_AACLVP wang_q_Page_50.jp2
4b89c254f5738edb299ac45517ad5bca
03651caf03c8e7837493f74f167416a794de4b04
6092 F20101123_AACLQR wang_q_Page_39thm.jpg
5cacc547b64b45dfde51a111a75648fc
81b08d59675cf5c146054459d0f8972044095728
457 F20101123_AACMCM wang_q_Page_01.txt
ae445ee06363df8cd2ec87337388e201
eea43430d2044533cb4ef0943179f83e8370221b
19438 F20101123_AACMHK wang_q_Page_06.QC.jpg
cea7e4337e14a3acb371cd67bc543364
9add57d0c1f6bcb97b1e95ff4bf21bf69feb7ba7
103878 F20101123_AACLVQ wang_q_Page_51.jp2
b61e2a10b57867701baaa8ced58a350a
471778a08026a356e831b3c863734f638748ed04
69159 F20101123_AACLQS wang_q_Page_67.jpg
65af63c0bd4a308234d3bcf0fd617352
3b2554acb43b359d3aa232a32847868926863e30
103 F20101123_AACMCN wang_q_Page_02.txt
0afa9a6f5e2cff5d06ccaeb9c43dcd3f
f2911efa68021d6e15361e3482c1c7f9d09a2a86
F20101123_AACMHL wang_q_Page_06thm.jpg
db325432ad84fda80f5c7b329828ca62
01643a74570a03ec14dee9156813a71a463c253b
F20101123_AACLQT wang_q_Page_01.tif
e26d15b14f050686a9115487d363a799
2c985bda78bb637467f46235176ddda3593afe4f
82 F20101123_AACMCO wang_q_Page_03.txt
ee0c4e59fdfb904300d6b8ec02b1ff63
53c6fcdb5c4f459738ed3a3591290a8585cb188b
2836 F20101123_AACMHM wang_q_Page_07thm.jpg
57468eb73ee54d2e9c17996b909dbffc
ba59db6cd6223bbc236f767f74217390c619ba21
52546 F20101123_AACLVR wang_q_Page_52.jp2
b17284a7f6f718f16b24b6a440b36b44
67f7f4724a91e11e524d683be7daaba429a3d085
734 F20101123_AACLQU wang_q_Page_09.txt
84497bcd6e9f6e5b6f26e4cc202faa6b
a05d5fb20e27662e73024fb2c929b8ed0b684b70
1712 F20101123_AACMCP wang_q_Page_04.txt
e7d97f2d96476fe81fa8618ed5bd736c
331d2695dd7adabc361290d9d402c82ea895cf0b
11750 F20101123_AACMHN wang_q_Page_08.QC.jpg
a24c186ff90f3220b5050d638052fad9
276ec436a3e494d0b6d04546912e49c9296e3131
87326 F20101123_AACLVS wang_q_Page_53.jp2
b72704a0fce80463f8e7b73067f93d0a
61cfd479602477a24abc4367734f9f231cf5ed25
93849 F20101123_AACLQV UFE0011601_00001.mets
dc981f1b4f2fcf5e8f319741b6905ecc
ee7df90f7e02e5aa676b1c9e32a3fe4b2d9f51bc
274 F20101123_AACMCQ wang_q_Page_05.txt
389d54a6430172f19e8fd4453c12f923
8be3174f41f054627c39a934ebf1a942a1eb5643
3429 F20101123_AACMHO wang_q_Page_08thm.jpg
c2e0bf0fa1634573734508ce981601fe
1a712752f81e5ee009a525bd9b60849dc0d21c57
110102 F20101123_AACLVT wang_q_Page_54.jp2
233507e88b5b91cb8135d8e888431a8d
5ea78b59e94de9bb0a8839da7ebc0ca9c08a7d8b
3106 F20101123_AACMCR wang_q_Page_06.txt
9a470e92c32b87eb8ea0076e276b4608
d0494618c8c0698ea7a04ea1bb6cbd7664d89cbd
9263 F20101123_AACMHP wang_q_Page_09.QC.jpg
3a9e6ac652c095bf830f88dfbc2d7b5b
935e4fbfd261106e6ef04c978ce6bd79f058294e
103005 F20101123_AACLVU wang_q_Page_55.jp2
fcd97c91c84d7a6ecc1e4bc57e6ac5c3
9caf4a8c8e54494c6d4486ddd052613a484393ae
1557 F20101123_AACMCS wang_q_Page_07.txt
0a69232aae1831b0494e1f75071c9a41
4d973d6b77f974c4df94cedf17832732e0d1aaf3
4703 F20101123_AACMHQ wang_q_Page_11.QC.jpg
6ca39f8a1cc3689ace43a62c3624d284
f536917c00a6f2a58e337386a3d73abc13dc8134
100496 F20101123_AACLVV wang_q_Page_56.jp2
196b11956c362c07e9a6522fa8e05dcf
189a9f5b2bbc2f3b4d7679409b78c6243ee23d2f
22519 F20101123_AACLQY wang_q_Page_01.jpg
0a850df56528ff9bc6debae9b3b742df
4f95032746061a208a8391d1b33fc48e4a74b689
1250 F20101123_AACMCT wang_q_Page_08.txt
256aa609993f4fbae14fa3f0cc685254
fdb90082368a89697bec6dfd71b9eb457cc5bafb
1697 F20101123_AACMHR wang_q_Page_11thm.jpg
3e053d104c2ac3fb336388a8da6c72c3
faf4ccdb77d0cd66ba3851530005d7954687a934
10107 F20101123_AACLQZ wang_q_Page_02.jpg
622cfaea9db08a1da25c5ae8eeee346d
4da57af881dce0809ee66749c16af83a60ee30aa
169 F20101123_AACMCU wang_q_Page_11.txt
390ef4ffe1b4e4b276e868f3bfe2c35b
6ed425fc9c4e8b621e50a6d8fc211d624649d701
107478 F20101123_AACLVW wang_q_Page_57.jp2
03bd193e00524afad07d9ba030842f38
1a56364ab0070cea70c4b488e09a3582d241e182
6249 F20101123_AACMHS wang_q_Page_13thm.jpg
0f6998b7dcc2f89f246d8cc47e0fd967
496ef42d856477f2afb928a8b9ea749ae1d4be08
1844 F20101123_AACMCV wang_q_Page_12.txt
a3eb5ce23b742df7c101aaf10f2d1506
2d9561cd2e9ece7f0422b40a1edd9c85ef6ce6a2
103707 F20101123_AACLVX wang_q_Page_58.jp2
1c12ecb5d6e10abab845f57e6bdd851f
2f2f244705b1ddcc988dfb2abb5ab97dd9239a4f
6359 F20101123_AACMHT wang_q_Page_14thm.jpg
12f831d2d9169ba59c8aeb8c427c52fe
4e553190a09f00729668edcaacd5f452f9a4edea
F20101123_AACMCW wang_q_Page_13.txt
f9cfee5b466ecf8a356a640ba4fe87c6
499a9823291fc21282f74e919c3320454202989c
69546 F20101123_AACLTA wang_q_Page_58.jpg
5de2d8445b04c89006b05168e5f712ea
439694653ed4102f6b8d52709941818ec0bc3cab
77725 F20101123_AACLVY wang_q_Page_59.jp2
90633eeb3de029a2bd292249a8f9ee56
e8a2a3912ebb0072872c7882d9111fb823e5a42f
23910 F20101123_AACMHU wang_q_Page_15.QC.jpg
4d8afb36c4896534f0e604b8b0c44e1f
600185c31f40857806360832377de017cd91ff1b
1928 F20101123_AACMCX wang_q_Page_14.txt
4183c477a3723cba03c12ee67808baf5
93bbc9820be363ed03b43fb0aba96fc70b840476
50300 F20101123_AACLTB wang_q_Page_59.jpg
792feae6f8fd71160cc49178950843ea
5f9bc2249bd8250e53adccc25a43c144659eb029
68140 F20101123_AACLVZ wang_q_Page_60.jp2
e8536fbfdf0cd18956401d8264325449
dc66833d95ccff668825d0ec46d0873cdeb165a9
22118 F20101123_AACMHV wang_q_Page_17.QC.jpg
6aa5882003ac7d09760debcd8ec5ad6d
605620af342125de17777e6a917a6cfeee931009
2033 F20101123_AACMCY wang_q_Page_15.txt
0a7fd635aa21667273e67c75996b3af6
e97128eb219070b31e2db5679e4133916129ca05
47147 F20101123_AACLTC wang_q_Page_60.jpg
a75e27bd5887dffab2834a2cbe36313f
fce78e4fb67f695ca9eb2024a71a15251b25083d
6193 F20101123_AACMHW wang_q_Page_17thm.jpg
2fa41933fb758811d488fbebf45bec02
f397715e4d6d853c7f351c8ff4f567349e659e91
F20101123_AACLYA wang_q_Page_38.tif
03e476fb98f1c2c430cae9e8cacb5916
71e5997842c260292eddf444224d455cb1e01955
50303 F20101123_AACMAA wang_q_Page_13.pro
ae8ccb66373f18938aff8125485c9c70
26ce6159f1884002ae03e131d9c533546d47271e
F20101123_AACMCZ wang_q_Page_16.txt
d39fc70d621ca75445469b0cc8064c66
c2056f4c410d298ef07187f73c67171bf562e11e
49865 F20101123_AACLTD wang_q_Page_61.jpg
7cd087e4c788387c339009ac4e251e74
ba6873a7b9b3a77df4f8468309990b10a52a2b90
6337 F20101123_AACMHX wang_q_Page_18thm.jpg
806ac6a27f18eaec35315f38a279ecd8
07413d9f096201b4514f018beb93ba85f49d0e47
F20101123_AACLYB wang_q_Page_39.tif
a5ae4352389ffb2d2504eb45ecd3ba14
5eea49ad09dcca0bcf86e8d6eaf443858acd2d1c
48789 F20101123_AACMAB wang_q_Page_14.pro
7575c0da52ad8572c0f8f6218bc58266
99271a9ce47c4b5f00a4d3f92bb944baf801c03c
48968 F20101123_AACLTE wang_q_Page_63.jpg
a7ec8c10320163bf5678529e4dc930f1
e273712d2d06d372431589773e128e8e68dac1c9
22972 F20101123_AACMHY wang_q_Page_19.QC.jpg
875dfc5fff181a92a471c6e2ac6db315
826ea8dc479e878d3c24a3de6e810eb5b21e251e
F20101123_AACLYC wang_q_Page_40.tif
3d97b3265ffe538841429a08d1cfc769
6752ad2f34439f49a6b615c538fc8d15819f7644
51470 F20101123_AACMAC wang_q_Page_15.pro
25493abf4241b86c175aa4f5af543542
d1b35d6d9320fb2dda6d78947d0b6ca7bcf645d5
63837 F20101123_AACLTF wang_q_Page_64.jpg
fee17597f98389a46570c46178f564fc
97bb95c3f5f5233fbe220b3339de0fd0a6d3c488
6373 F20101123_AACMHZ wang_q_Page_19thm.jpg
42c99e39e68d260e33163c81187b952b
e89678aa7f7e57962f261040ad032fe8859c0b66
1901 F20101123_AACMFA wang_q_Page_73.txt
885bd7d26c7fd69c986a3eac4629268f
b6d1dde56786ed405a1611ae93754badd58df1e5
F20101123_AACLYD wang_q_Page_41.tif
715d8d56895d5bc513b031314135bcdc
cd401c6f96e7bdcfa86217aaa9813a24ef81932a
43684 F20101123_AACMAD wang_q_Page_16.pro
5be9104f2cb8fc0a654cc72be559b5d0
2722aa9628374570c603d946ca03620e12b8bd74
68035 F20101123_AACLTG wang_q_Page_65.jpg
072fbcba8eec83885dba24b1ef949320
284b6b097043981cceb13e168e23d784c5f11cb5
1865 F20101123_AACMFB wang_q_Page_74.txt
d59d11270ce9ef23de4723d27d29675b
03da6c18e00ca1448addc0fc0b116216f12a6fb3
F20101123_AACLYE wang_q_Page_42.tif
e3214f48b1966e6614427f779f13d879
58f0116a418f2f94db28d802c498a45870832613
42281 F20101123_AACMAE wang_q_Page_17.pro
34169bed3f65f0445e8cacb81d296834
b3cb12a78443032459f3b7e87522e154f6c4c29f
64671 F20101123_AACLTH wang_q_Page_66.jpg
9236e377212b4eec3dd1f4b4dd79827b
43f594c7073a97a96cc914eec4c711a6bf6cff5d
5607 F20101123_AACMKA wang_q_Page_62thm.jpg
89f6059f4f1c6f3ec7b02d8064711089
968995ce151cdaa72c1d4cd2530e345c856dc24c
1948 F20101123_AACMFC wang_q_Page_75.txt
9a325efab0222c3e0405e5396edc7bca
37e855b06b0670b79bdacef13ab6411f6f5f331d
F20101123_AACLYF wang_q_Page_43.tif
abf2f24622850e656147d6d4329b5b14
c938307f7e4f6f0fb2832a01328770f1f7fcf2fd
49811 F20101123_AACMAF wang_q_Page_18.pro
edb3c2c0b8561c378ec274789e4e6bee
d5ccf116e3da46590b5b512478ec06b45cec4b77
62147 F20101123_AACLTI wang_q_Page_68.jpg
29ead716e2d17a1be4ab0b3bf1474ee0
48d2f4fb23a9b13eee0123f6a7d47cfb604b2bc2
16337 F20101123_AACMKB wang_q_Page_63.QC.jpg
293db09d77e2c0cac990cb893ae19da5
55706bfc2ec6188e5163e7da12a6c4b17afe928d
2451 F20101123_AACMFD wang_q_Page_79.txt
1fe6a0c4f05ba6d96f37df1687ae5e89
4e311333b67db9bf44d829cd171ec369cbfea702
F20101123_AACLYG wang_q_Page_44.tif
00514287095c8e52cd9dc21a385e8eb3
e2baa0940a7d785a9a9f5f799ba9d9ffaa2c3881
49262 F20101123_AACMAG wang_q_Page_19.pro
fb0302cd086ea48471a6bfae44fa4ab4
2b066e39ebc8c440abc5df04ea175a3d8d1d4c0e
40538 F20101123_AACLTJ wang_q_Page_69.jpg
8631fc44841c974c3e2308d45557ab0a
49992b7bcf086a9cba6a26861f0dadc04b7a58dd
20056 F20101123_AACMKC wang_q_Page_64.QC.jpg
d33260c726893ec6cf49c8858af673b3
3fa33188ef1389bf7bd81b8b16c864771fbcb478
1352 F20101123_AACMFE wang_q_Page_80.txt
22462ee17931cf4f084cdc3866aa705b
5058082a5efbebf903779fce5e95d1aecab7eabb
F20101123_AACLYH wang_q_Page_45.tif
a02270ab328ab5e395e91d5ccdad39d0
2fe4fce2c02566ae13701f02ad313c94d8bbfecb
49342 F20101123_AACMAH wang_q_Page_21.pro
95061b7e3317fca0e20ae2d45b8a4bd5
0bd334e52fec02b7ada683bed5ae46f395b1c070
51169 F20101123_AACLTK wang_q_Page_70.jpg
e1833671375c92c56a8d36d5a5be6c90
c8340661943a5dc60cd9f121a6153033f23826fb
5781 F20101123_AACMKD wang_q_Page_64thm.jpg
6445343d0b0f46ee2a082526118c899a
5ecc01b454f17e3618c46bf14e32dc2e58181dd7
831 F20101123_AACMFF wang_q_Page_81.txt
a020296b3ccf01ce2f092b2e72a5c786
a3fd426973345de5f573fe6f4deb2f7de59b4e4c
F20101123_AACLYI wang_q_Page_46.tif
dd5d300c0796fc1c64663e529c7f0100
fe201c334878a9b0803c2f0048d4eb980f405a5b
49210 F20101123_AACMAI wang_q_Page_22.pro
c5e6ef01806efa2df235f3bcf02d4fbf
48faeaded1b13d1ea0d1993941041d0594256a2b
62690 F20101123_AACLTL wang_q_Page_71.jpg
e717d4c2d833982569133565f39548cd
c50c1cde94d157f1990e57af5e6a4310ec2c7ae9
5691 F20101123_AACMKE wang_q_Page_66thm.jpg
4d88f2b5c08b41833861e3035702bb5d
9b87174585f55848be752aa7ec549246d424c0fa
616845 F20101123_AACMFG wang_q.pdf
290e69b4ad7dcc082a5559de86425a47
589ec635af6659e02843f2e7189eabe4950ff87b
F20101123_AACLYJ wang_q_Page_47.tif
50b1db91e499091bfe87293c70d37db0
7c067997883a9e1eb2633a9056bdecab7764c96c
46219 F20101123_AACMAJ wang_q_Page_24.pro
e496ceb72815ad8be63a86d1aab5047e
701779ee34958837611b9b6e152149a8524c129e
59713 F20101123_AACLTM wang_q_Page_72.jpg
842265caced98508a174c0fc1f8642fa
d7d4adf1964b274ea15e0ed7fde43aafeecd99a3
18569 F20101123_AACMFH wang_q_Page_10.QC.jpg
910bfa313449da5185126d9be5fc6a7c
dee42c7404fa513969b4916b4ff1745670908408
F20101123_AACLYK wang_q_Page_48.tif
27975eb80a9871651b65fa14d48ddfa9
f653849a6727a58a73adff83d9fe0405649031d0
33898 F20101123_AACMAK wang_q_Page_25.pro
8d0c52a4ddd0902031042b8858b87244
3bb7c8a8c6d7b68940d28dd184dfa971ecd834e5
67360 F20101123_AACLTN wang_q_Page_73.jpg
fabc2a02bd216bfbb254caf28e40ef78
b0be8cdfb8f33c1607b068a116546d1b90f2fb4d
6191 F20101123_AACMKF wang_q_Page_67thm.jpg
d35e2b57e51fecbb9721fa3a928e213d
358b768e479762cd64d40bf06864333a1fece90f
6047 F20101123_AACMFI wang_q_Page_30thm.jpg
f2db3632e49844ce833518c1ed031cad
91edde8eff6b3ab22c6ccc3a2a73379f9877425f
F20101123_AACLYL wang_q_Page_49.tif
3383c346d73df107dfa3ada7f62a4e14
99eb49a1680d2fb7d2185d1aed8b28434ab3cf65
45081 F20101123_AACMAL wang_q_Page_26.pro
296d61c224fd4edc8420505c2be5f0f6
41932a2bc7a3ff6010e703d901513a1507d925ea
69976 F20101123_AACLTO wang_q_Page_75.jpg
9ac010062b6626f1450815a7d7684ea8
b1ca166b11edebbb869be1c3cd9da0a961d7f469
5581 F20101123_AACMKG wang_q_Page_68thm.jpg
805f4af88eabdfb29a6651c973cd3e86
00510d5ed2d7f89b1561db58eec1ac2542aa51de
21019 F20101123_AACMFJ wang_q_Page_16.QC.jpg
e4bd166415e9aa37324e00d458534911
6890d14c8b07d2ea233b5398460e2947989be73c
F20101123_AACLYM wang_q_Page_50.tif
b9a63c3a6256e3be9439c4fcc4d1dfa6
eac9b14f67745d22e778b5f9bfd3a04d65eb15d8
42091 F20101123_AACMAM wang_q_Page_27.pro
d69de0cfc8683ff8f2a2d46e4b822b33
4ef7ffeadcb7ac81c52de146c60a2274fc38b06f
13223 F20101123_AACMKH wang_q_Page_69.QC.jpg
18195316a04eeaded58461bf8716e996
e0e99be9da1595b7714d53dc7f5097f5311a2c10
15140 F20101123_AACMFK wang_q_Page_80.QC.jpg
56c82edc2f8aa56da58238a2d02c9cad
76ecd0c5ab6b7a57e23a762617909f57f8174bda
F20101123_AACLYN wang_q_Page_51.tif
fe391406c911e6356a9e0e24cea9964e
e007be720db8e1ed49421e0bba264af85096a2e4
39413 F20101123_AACMAN wang_q_Page_28.pro
156f21d62786b64a312047fffc52160e
46826dff55e178748c09eedfa4b2cd5d056d0d07
21204 F20101123_AACLTP wang_q_Page_76.jpg
f18fe669683652c0b52d9946f780c7c8
dc790e5ce7b546fab21e1f792ee771ba39dc8e2c
4147 F20101123_AACMKI wang_q_Page_69thm.jpg
74c73b2a9cbdd27acfc8806089509399
409d371b66090c2966e84579ccdc9777849e265a
23612 F20101123_AACMFL wang_q_Page_54.QC.jpg
f26666bad2348aba188d6fb78269d691
237cd135df49fcd4b19e53c503de16981db66846
F20101123_AACLYO wang_q_Page_52.tif
14affe2e781fa855aa835f3dc6ed5651
ac5aebd0316ab69d5ba1a8729795698a9a0fed68
41873 F20101123_AACMAO wang_q_Page_29.pro
32ef1cfc916b9bc60d49b0b14b484005
96b1f3cb05d337f5112a0fc2988c7a0e9c50fee3
75540 F20101123_AACLTQ wang_q_Page_77.jpg
ec5fafc938cde9c86ccd16868041c530
a9dfd8858ac2fe041f05dc25f0ef3eff14dd0759
16882 F20101123_AACMKJ wang_q_Page_70.QC.jpg
45dbd326ed3d3c2af76cb3f99f65c2d4
baf2caacd7ec81eac974a3beb2c3c926a32cfcd8
23154 F20101123_AACMFM wang_q_Page_13.QC.jpg
cc3bde390050754a5660ef64c2d669e8
59e6ecf93266372614a302548674603aeb87ba77
F20101123_AACLYP wang_q_Page_53.tif
7f0e0e29704694c5ff06367d9eff65ce
8f17fcb7d51f513c6b57d1bf6a72aaa22d7b750f
47709 F20101123_AACMAP wang_q_Page_30.pro
fbeb82c2d7ddd988d88a5a36be454a1d
18572e70e5ad7668233d692f3ee7fd196f49a52a
85914 F20101123_AACLTR wang_q_Page_78.jpg
69d2bf4273db9c0b5d3a1a55de0e248e
574020f68d77fb48c82b914e604f6897b6db6108
5558 F20101123_AACMKK wang_q_Page_71thm.jpg
519d916aa7d6cb104e3356b2c28cc498
b1c49f356205d4c88b05f7c23d13aeffc1b3d5e3
6234 F20101123_AACMFN wang_q_Page_58thm.jpg
440f18390dbbfa8d3968ce039d833a06
372de6f5f78663d9ec7e1adf06809e2ffe860824
F20101123_AACLYQ wang_q_Page_54.tif
bd2f5bbb803725ee24766b38d4b864e9
c3bb51bbca3b70b238a2e3b84068c99e07a30cbc
40894 F20101123_AACMAQ wang_q_Page_31.pro
483d89af5114e2b0cab68d09759948e5
bd3f24fbaa88cfbe122d13fa2eaf2b6ef274e240
92347 F20101123_AACLTS wang_q_Page_79.jpg
e4730915ff62ebb9c1d7aa77339080ef
28902abda42f514f115f1869107f8832e8002eaa
5548 F20101123_AACMKL wang_q_Page_72thm.jpg
76b63eec042bb54cb885514110607d92
5c565a67c95c88b9030b81d5d9dd5ce79366ee7f
23478 F20101123_AACMFO wang_q_Page_48.QC.jpg
e884169dbe7c7d87aa63a4a9531597c2
ca7b7cdcb8ec5d53d0a5c6ae0ed1a3b541b2c30a
F20101123_AACLYR wang_q_Page_55.tif
7635b51f105be828b71e8109a992773d
5874159110a50b521a1a41f642eb287dc5b080d9
52515 F20101123_AACMAR wang_q_Page_32.pro
f0de912e79bd45b971f127429225cafe
5ea05288f845cf3cf43bf6bfd32c4a15e572027c
50190 F20101123_AACLTT wang_q_Page_80.jpg
57d594c1c1a500fd5ca9a83fb87b61b4
3385aea8e09c062f60c41f90254fffe2920f0ef4
6334 F20101123_AACMKM wang_q_Page_73thm.jpg
645b17557f114a5ef8972329e1848e26
2687dbc3c0fa93953c6d7c4c518510900d1e811c
F20101123_AACMFP wang_q_Page_05thm.jpg
793c025cffd251fef30bfe16a3e3c43b
28bcad8c509485c38b6c45552475c54cdb24a886
F20101123_AACLYS wang_q_Page_56.tif
f54f4149fa70ae15c53c5eb99c2039c6
1e42231a440dbc30772dd9522f9b78e7cfa21c52
53938 F20101123_AACMAS wang_q_Page_33.pro
5888155392b013f05fa7ccded69ea9f8
ed686ce231b3b174ed8ada4e84ee69fbf8e3f558
34806 F20101123_AACLTU wang_q_Page_81.jpg
3767bf2749eb95eca1977c5090688bb2
0138fd4086ff7ae7c0ac6dd630aa51c0da45157c
5446 F20101123_AACLOX wang_q_Page_53thm.jpg
5274656072bcabd3bd80cf3c6ca7f0eb
d1ef574de5367e9c0e92730ca35980029eb1d630
22725 F20101123_AACMKN wang_q_Page_75.QC.jpg
cfb5e502b7ec36bc9a44a1b24a6f034b
f3ebb54e6d6d62052dc753ab3fcbc3bb98a407c3
4655 F20101123_AACMFQ wang_q_Page_60thm.jpg
8a1292dcfa3239e61473c0768e5373b6
e862ddc7706aab4aff60f707e543bde5aeb8a422
45487 F20101123_AACMAT wang_q_Page_34.pro
2396648f5186cbda9811dfbf08f1bf07
0e574e99288db8b0a1920c99a14039dbef9a0aa5
23502 F20101123_AACLTV wang_q_Page_01.jp2
7fa44b0ef439b8a4430a9ccd3920fb59
bc581835552aee98269c34f938065de8570fabe2
107841 F20101123_AACLOY wang_q_Page_13.jp2
9231810c64bf8fa8e69cd2e6a507a666
8d30d020bae1e561c6516db258250704d45f966c
F20101123_AACLYT wang_q_Page_57.tif
543acea7198561d7a1a2b378f84b5aa6
f2f4249add74f4bb77b0d6cac95093472e69479a
6522 F20101123_AACMKO wang_q_Page_75thm.jpg
eb0d717b35f98ab168e8a5ee0eca8fd6
70f34401cff3837e1c0c7548e31f53707600e89c
5256 F20101123_AACMFR wang_q_Page_10thm.jpg
f60e4c957fe657f30451d2988330ac0c
331439b9f659e9de2bf3576be59975f4abaceec2
52312 F20101123_AACMAU wang_q_Page_35.pro
0cbd537f8206403983a6e351e7d6b323
755b003c31e964056d82ac5f4109c1fc5f44b858
5339 F20101123_AACLTW wang_q_Page_02.jp2
90f44de1488f1227f95f5aaf014e6138
6a6fa89101daef36c8f5808bb6ee761e4bcb282b
29726 F20101123_AACLOZ wang_q_Page_08.pro
d4274d8654fbe2196ff862468e1e761b
e751cc3396d969d309fb76a42ad389eb6ad71760
6889 F20101123_AACMKP wang_q_Page_76.QC.jpg
211dfe71764b4b7db5bda007e05d674e
ab59e3901288770e6d4c90a20bed135e9b7d4d15
20594 F20101123_AACMFS wang_q_Page_26.QC.jpg
6afe7a62c7016bf7ef85c8d298617bf9
86e5e5921058e616c19e41e2d935963704a0d808
56519 F20101123_AACMAV wang_q_Page_36.pro
25f1575aceca894b71983eee15140e51
6dd7f6b63ba027dbb745a24c584b669c6dfe831d
4140 F20101123_AACLTX wang_q_Page_03.jp2
20bf5491aa62876de3b29f76114531d0
d40ecbbc8b3acc45656c328a2189ddcce717c4df
F20101123_AACLYU wang_q_Page_58.tif
3d36396ae981e53266ae2b05bbda4c35
1c32bfa64467c974a6ff33ec4a67bd3cca3d49cd
2260 F20101123_AACMKQ wang_q_Page_76thm.jpg
e37162600429ace3668b05466215dfcc
94ada004493f80538cb719a3bd538be2580f6884
5266 F20101123_AACMFT wang_q_Page_43thm.jpg
566fa6aaffd6c2e93f103db4533350b7
2274d1b9cb9a38de16a4b262cad166c91a402557
46454 F20101123_AACMAW wang_q_Page_38.pro
5b23bc221f512d860c102736efad19bc
c677789cd9a8347dc318330191fa41cbf7233143
9473 F20101123_AACLRA wang_q_Page_03.jpg
15bdde51c5a74230ed0b08516a1c865b
0d62810124275a58a99a51d62a24b7d1f44f37a4
91833 F20101123_AACLTY wang_q_Page_04.jp2
140a685af8bdbab7bc2288f620e4af43
509f30ab0f7226d7ed818a8158285c6d116995e9
F20101123_AACLYV wang_q_Page_59.tif
373b9d697e37a9192822b80b8b587ec0
87788db3571abc6d26d24449e438b8b91a041237
5917 F20101123_AACMKR wang_q_Page_77thm.jpg
8b837014e9c92b3e5996a7f4a0923f26
e2a4d454f945f17451897fa1340fd01f57bcceab
22399 F20101123_AACMFU wang_q_Page_14.QC.jpg
24d67993a4adbc06f5845f4c9932b49f
b1fba7494807c2552a33bca39643ebeaa4ddf184
46190 F20101123_AACMAX wang_q_Page_39.pro
a728357a172752486accd76e7142a30f
8cc537f0ef520244f9162f6ecec5f0c15b962500
62360 F20101123_AACLRB wang_q_Page_04.jpg
ba4296cda2ba73c3dc0f0f13d3b37369
6a5a7f7e9a89029eba29869a6ec1af96aa4ba4a1
17682 F20101123_AACLTZ wang_q_Page_05.jp2
13c7c0793d9f118740fe6de76a109474
fc882f5a45cdfb3ce80369b401661f9836b1b415
F20101123_AACLYW wang_q_Page_60.tif
f5910a13d3499cf7e43e75997c28be91
7fc6a6bd36ad4ba383ba7c9d04b2f05735843a62
24383 F20101123_AACMKS wang_q_Page_78.QC.jpg
c709f74ff964514797913c9266595a37
01fcaf540a9fc226cd7a6e4e240e1b799a8b8ab9
6319 F20101123_AACMFV wang_q_Page_74thm.jpg
f16164deb302fc922de5d70ab3f86f3b
c4f39d09377e08261982bd6e34bab0839a9ff2ab
37558 F20101123_AACMAY wang_q_Page_40.pro
3ffb55dcfda280e3ab0a1e0b88752e02
e0f1df1575a0bc9eb5a9c7dc04138a8311e844a3
16623 F20101123_AACLRC wang_q_Page_05.jpg
1aca95758cfcd5438aeb569a3d68a5df
fa9885df2f461e31fd3123891ec9d6259a1a665e
F20101123_AACLYX wang_q_Page_61.tif
b5e27ca4ab8ea75bfa0ca6c7d7c682a4
dc1e99836cfab72fda4e523e1ffb3f5fe7dd4603
6635 F20101123_AACMKT wang_q_Page_78thm.jpg
db22b49c544f5c60ed40152cbec5ec01
7920c6eb89b4bb33c4263fd95fc836c868575f9f
24648 F20101123_AACMFW wang_q_Page_47.QC.jpg
5c7fd3c2a3ecb58e52027f204fcb9a22
5e77d2abc8daa5f280da2932c58097498b21ea24
39970 F20101123_AACMAZ wang_q_Page_41.pro
2c8475ab02cb3ec2fc3d936370b6f2d2
94a8c2cb8fa4f6bf0c9807b0405ec88b9d65088f
74758 F20101123_AACLRD wang_q_Page_06.jpg
83d8e0ce22c47bf712fc682039fa5c4c
41e172decbe8f4c6b65d82a4dbaa08e3abdfdaf1
75234 F20101123_AACLWA wang_q_Page_61.jp2
b4c813f8ad424f1737566c2c907887d1
4765e1ce22eb856a07ddc55b55fa4676e2a9e923
F20101123_AACLYY wang_q_Page_62.tif
c8b875a297001323641566481e736d48
7a6c86b355feda112bf776fafa6c3a9d00763b06
4377 F20101123_AACMKU wang_q_Page_80thm.jpg
47da216292d99085fce534c5cec357ba
8155f045def833d1343c892e3ba1cd2054aa87c5
4251 F20101123_AACMFX wang_q_Page_42thm.jpg
5de5376d0917a96276b549d3385bc4d5
f62f7d907f313cd93896cfc5b9b4bd4021bad688
40527 F20101123_AACLRE wang_q_Page_07.jpg
c12c84f53c3bf08a61201bcacf56a79d
16f4674c7b9757d983c18eb638bba77d9660f90d
80215 F20101123_AACLWB wang_q_Page_62.jp2
96d9c357f22ed665b34438d940702fd4
e8be8a482dac56cfb36838ff076b67d1fc8c9657
F20101123_AACLYZ wang_q_Page_63.tif
a362c20c133f582cab264250e1d97180
22f55829896e5d5efd18b858a0b67ddeb11dc579
11773 F20101123_AACMKV wang_q_Page_81.QC.jpg
89a56cd97a6720ca4ee3f39c74b846e9
3137b93c3118ec0df00657b1641dc9ec403c9088
21896 F20101123_AACMFY wang_q_Page_74.QC.jpg
506b93977d558618c59ae16155b29952
669139e70b0709aaa0e4a3214ed51d5c17324d50
37992 F20101123_AACLRF wang_q_Page_08.jpg
7f4de862a11b108507c5d530c525d1fc
34e4ebc1f241183dd9e10d6384a6371cc1bf6329
75501 F20101123_AACLWC wang_q_Page_63.jp2
34c20cf40074443856f4f669a9fd5f41
c2a61677c4241617bc2e8edf44fb15fed2890cee
3556 F20101123_AACMKW wang_q_Page_81thm.jpg
89cebe5beba565ea7c53ca303ff2efa0
bcac67e9ec01f6429ce3b6ccaeb23a5bc0729a6c
21958 F20101123_AACMFZ wang_q_Page_65.QC.jpg
5fb3ddb825bcbfe82c383fd846b6228f
d13cc717f52f81235d895f1ec7a8c9c106ded00e
1765 F20101123_AACMDA wang_q_Page_17.txt
a47f7b4397baf55abc4d6d44ee6c2598
034a7a25488d52803037bbda0a8a1267b0850aef
92526 F20101123_AACLWD wang_q_Page_64.jp2
9634b685393f3aabb3cbe653fe94464c
44e0450eebc28aa57dbb46c24341a203bad3238a
28882 F20101123_AACLRG wang_q_Page_09.jpg
75538dfaa9be66d5b43931a54d6bafea
159adea129b3fd55894753c749ca7bc38ad45ab8
1967 F20101123_AACMDB wang_q_Page_18.txt
3decbb3c9c1976373c8622eb41436869
2a46b29d30efbb0865dcbeda2cba9b1f65810ce2
103416 F20101123_AACLWE wang_q_Page_65.jp2
6cb80c734a31ec78c61df292c03bba24
5820fc4874efb60f64b194a9ba97902a91c22286
59795 F20101123_AACLRH wang_q_Page_10.jpg
4de5ea31f810b1ddd32b0d8e149dad4e
f0962cd736ee79ca1882e579dd10ee8a09cf4ccb
22798 F20101123_AACMIA wang_q_Page_21.QC.jpg
3b46412a84d6f0785a770ae92827a8ad
a98fbac5aa77f3e4bdb9bd381b46620124a4f1e6
1944 F20101123_AACMDC wang_q_Page_19.txt
ef691f9521c7fa79b651c89135b2e451
d1c942b7b78176dd8a412a1ad35edc827eb53916
95606 F20101123_AACLWF wang_q_Page_66.jp2
3681e15e78514081a9bcf035d6678c52
ab02c80b67db74c028712939b0491f01a2a5ed99
13756 F20101123_AACLRI wang_q_Page_11.jpg
8c88120f123a6fa891cdc40c109ae13c
772d8ffa1453372c4af904f363bfe20f2f074e52
22423 F20101123_AACMIB wang_q_Page_22.QC.jpg
1062bb7dcda269b8de59b734661da8e5
6153a962cb7de34c5f0df45814521664feb2f7d9
2014 F20101123_AACMDD wang_q_Page_20.txt
7eca0068ce0aec583c73cb25e0e30719
d1cbb639ba99c24ffac06ae3444ffd702e6d619f
104407 F20101123_AACLWG wang_q_Page_67.jp2
4023b7dfbd6b6c505657fa4db5639bb8
f08f6a510f8e9b1565384d1d3198f4e64cfd5236
64388 F20101123_AACLRJ wang_q_Page_12.jpg
fd8c93bf7a07c12a10b8ead6caa8d97b
02c7d005d8bb09b33581d93f691bdc70f85b6ff3
6250 F20101123_AACMIC wang_q_Page_22thm.jpg
845d6c65a4167303dab8825b2e89845b
7c5f1c152f6863caa68324223d6da2c677f9599d
1942 F20101123_AACMDE wang_q_Page_21.txt
21bfeb6633b0fae09c048b17357d5973
3f5997231dd4e9a1e808b6fab980baefeebf886d
91862 F20101123_AACLWH wang_q_Page_68.jp2
b1075172bbf110e0e57b290771245585
5ad8cf00c1b0ce965afd1db637a0fc69fd722d06
71181 F20101123_AACLRK wang_q_Page_13.jpg
6e3e812b68bb67b232cf4f0df82d15c5
db34a3031c711eeb1f8a89922222c0a4e2463737
F20101123_AACMDF wang_q_Page_22.txt
8f221797a9be80847bdf9f512e9c8028
e140567f0531351291d5d06824b3c32f5df7f6a8
689265 F20101123_AACLWI wang_q_Page_70.jp2
ab8b7981b5f8112674aeb9e464f80328
f2c2a24a1dbd2ec14a4e2b34ed64cac41aedfc57
70071 F20101123_AACLRL wang_q_Page_14.jpg
c663ed45ede2a7d687c136fa84197c39
12d06b3436ae414ece25e576e49b97088bfe4c36
22858 F20101123_AACMID wang_q_Page_23.QC.jpg
107b5e6807e632df4fd80ee2e6b73c08
f7ec1a25a61b046b56c8985213a32b5e2052c9d2
1962 F20101123_AACMDG wang_q_Page_23.txt
aae54e7f00f0f510088e1317f8ba256c
06c33d2788abd85e0ec4d4e4875b2a7a06c8460a
853335 F20101123_AACLWJ wang_q_Page_71.jp2
8bd33d6616f3eaec20cade839b92c86e
eadd105c318528e0fc0f107813ecfefcf9993341
73172 F20101123_AACLRM wang_q_Page_15.jpg
c33ba363e4139e38f5675e73cfd45b6a
ef4c15be07dd1ff08fee2c9016a95690ad4c27b4
6609 F20101123_AACMIE wang_q_Page_23thm.jpg
0d77db65aa65ed2cbb4f014887c53f42
d2934431e07346673421338fda7c4ccb30704ab3
1590 F20101123_AACMDH wang_q_Page_25.txt
daba132fcecef35ea24dd530aff6203a
b9e1792b24f7156f30f9c2d3c821dbf9160aeb55
830645 F20101123_AACLWK wang_q_Page_72.jp2
88e6cf9361e107968cdb53856a9ccca6
624fcd9fb16ab2a8c4dc1cf0d355006b8542b412
20821 F20101123_AACMIF wang_q_Page_24.QC.jpg
1b0949f77301ac7afc02ca2b72a35113
892b578fa8e4784fd74497ba5c488c385088ad1d
1834 F20101123_AACMDI wang_q_Page_26.txt
a0ef71cb8296650ad27d5c5a1211fdb2
c722bbe7b53132be51247154a51c7ba512776888
969922 F20101123_AACLWL wang_q_Page_73.jp2
d8de9be3d0daf88c7f37a6c2cc5769e7
f8d2acf0c58d02b750c0fc1e8ed84d2dfe68c4f2
64116 F20101123_AACLRN wang_q_Page_16.jpg
2827c69d7a4d30dc4c322fe4404aea21
6a3c8aa01e02f83c5dbedfd53b4e94e92260620c
6016 F20101123_AACMIG wang_q_Page_24thm.jpg
ca828a5d1e06036caac2ca79d01d2bb9
0ccb80239ddb9aed1add60d6f222a1e23f51775a
1766 F20101123_AACMDJ wang_q_Page_27.txt
b8178a7cbc8549cb768bf1f5808cd53e
6cab93c2b4a66d6ac07c7e8d4251a69468704ca6
102372 F20101123_AACLWM wang_q_Page_74.jp2
f854b04c67238f70d3476e0668a25765
9a9b98bd5a5037cfe433ca9aa7b6c980f3ddeec5
70043 F20101123_AACLRO wang_q_Page_17.jpg
f0fd07c47c7e50862c7943dfc60094e4
17ac266fc3631fe9f772988159fe7448aae425b2



PAGE 1

ANALYSIS OF MARKETS IN THE PRESENCE OF NETWORK EFFECTS AND STANDARDS COMPETITION By QI WANG 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 QI WANG

PAGE 3

To My Parents

PAGE 4

ACKNOWLEDGMENTS This dissertation is indebted to many people. First, I would like to thank my advisor, Jinhong Xie, for guiding me through my entire Ph.D. study and the entire dissertation process. She is not only the chair of my dissertation committee, but also a mentor and a lifelong friend to me. Without her constant support, indefatigable encouragement, steady mentoring and patient guidance, I could not have developed strong interests in research, accomplished this dissertation and gone through the difficulties in my graduate study and my personal life. There is no single word in the world that could express my great appreciation to her. Second, I would also like to thank my committee members and other faculty members for their extensive comments and feedback for this dissertation. In particular, I benefited immensely from Barton Weitz, Steven Shugan, Chunrong Ai, Mike Lewis and Debi Mitra. I am also indebted to my friends for their dependable help, especially Yubo Chen, Qiong Wang, Qian Tang, Younghan Bae, Xu Jun, Lucy Liu, Shuman Li and Song Xue. The friendship with them makes my graduate study enjoyable and fun; the help from them made it possible for me to complete my Ph.D. study while bringing up my daughter. Last, but certainly not least, I am extremely grateful from the bottom of my heart to my family--my husband, my daughter, my parents, my brother and sister. I feel extremely lucky to grow up in a family where my parents, my brother and sister are always selflessly with me for every decision I make, even though they are physically far from iv

PAGE 5

me. I am thankful to my husband for his understanding and to my daughter for all of the joy that she brought to me in my life, especially in the five-year-long study. It is their unconditional love and support that make all my accomplishment possible. v

PAGE 6

TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT.........................................................................................................................x CHAPTER 1 INTRODUCTION........................................................................................................1 2 A STRUCTURAL ANALYSIS OF INTRA-STANDARD PRICE COMPETITION IN MARKETS WITH NETWORK EFFECTS................................6 2.1 Introduction.............................................................................................................6 2.2 The Baseline Model..............................................................................................13 2.2.1 Consumer Behavior....................................................................................13 2.2.2 Firm Behavior.............................................................................................14 2.3 Empirical Analysis of the Baseline Model...........................................................18 2.3.1 The 3.5-inch Floppy Disk Drive Market....................................................18 2.3.2 Data.............................................................................................................21 2.3.3 Identification...............................................................................................23 2.3.4 Empirical Results of Baseline Model.........................................................23 2.4 Modified Model and Empirical Results................................................................26 2.4.1 Modified Model..........................................................................................26 2.4.2 Empirical Results: Time-Varying CV Approach.......................................27 2.4.3 Validation: Stage-based Menu Approach...................................................33 2.4.4 Discussion of Key Findings........................................................................35 2.5 Conclusion............................................................................................................39 3 DYNAMIC ANALYSIS OF NEW TECHNOLOGY DIFFUSION IN MAREKTS WITH COMPETING STANDARDS.........................................................................42 3.1 Introduction...........................................................................................................42 3.2 The Baseline Model..............................................................................................47 3.2.1 Consumer Utility Equation.........................................................................47 3.2.2 Market Share Equations.............................................................................48 vi

PAGE 7

3.2.3 Diffusion Equation at Three Levels...........................................................49 3.3 Dynamic Model with Consumer Expectation.......................................................50 3.3.1 Consumer Expectation of User-Base and Firm-Base.................................51 3.3.2 Consumer Dynamic Decision.....................................................................51 3.3.3 Choice Probability......................................................................................53 3.4 Empirical Analysis................................................................................................54 3.4.1 Data.............................................................................................................54 3.4.2 Nested Logit Model Estimation..................................................................56 3.4.3 Diffusion Equation Estimation...................................................................57 Impacts of Network Effects and Standard Competition on Market Potential..................................................................................................57 3.5 Conclusion and Future Research..........................................................................62 LIST OF REFERENCES...................................................................................................66 BIOGRAPHICAL SKETCH.............................................................................................70 vii

PAGE 8

viii LIST OF TABLES Table page 2-1 Descriptive Statistics................................................................................................22 2-2 Baseline Model: Estimated Competitive Structure..................................................25 2-3 Revised Model 1, Time-Varying CV Approach......................................................29 2-4 Revised Model 2, Time-Varying CV Approach......................................................31 2-5 Revised Model 3, Stage-Based Menu Approach.....................................................35 3-1 Descriptive Statistics................................................................................................55 3-2 Nested Logit Model Estimation...............................................................................57 3-3 Market Potential Equation Estimation.....................................................................58 3-4 The Diffusion Equations Estimatio n: Category and Standard Level.......................60 3-5 The Diffusion Equations Estimation: Brand Level..................................................61

PAGE 9

LIST OF FIGURES Figure page 2-1 The Changes of Conjectural Variation Parameters with Time.................................32 2-2 Sales: 3.5 inch Floppy Disk Drive vs Competing Products....................................39 3-1 Impacts of Installed-User-Base on Market Potential...............................................59 3-2 Impacts of Supporting-Firm-Base on Market Potential...........................................60 3-3 Impacts of Word of Mouth on Category Diffusion..................................................61 3-4 Impact of Betamax User-Base on Category and Standard Diffusion.......................62 ix

PAGE 10

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 ANALYSIS OF MARKETS IN THE PRESENCE OF NETWORK EFFECTS AND STANDARDS COMPETITION By Qi Wang August 2005 Chair: Jinhong Xie Major Department: Marketing The rapid development of information technology and the digital revolution in recent years have significantly increased the importance of network effects in the success of many new products. When multiple technology standards compete in markets with network effects, consumers willingness to pay for any given standard depends much more strongly on market size and market share than in markets without network effects. This creates new challenges for marketing strategy. There is a pressing need to understand the marketing implications of technology standards in markets with network effects, for both firms pushing different standards (inter-standards competition) and firms adopting the same standard (intra-standards competition). This dissertation applies theoretical and empirical analyses to address this need. The dissertation consists of two essays. Chapter 2 provides a structural analysis of intra-standard price competition between an innovating firm and imitating firms. Chapter 3 provides a dynamic analysis of new technology diffusion at three levels: product category, platform and brand levels in x

PAGE 11

the presence of inter-standard competition. Data collected from different industries with standards competition are used to test the proposed models. xi

PAGE 12

CHAPTER 1 INTRODUCTION The rapid development of information technology and the digital revolution in recent years have significantly increased the importance of network effects in the success of many new products. When multiple technology standards compete in markets with network effects, consumers willingness to pay for any given standard depends much more strongly on market size and market share than in markets without network effects. This creates new challenges for marketing strategy. There is a pressing need to understand the marketing implications of technology standards in markets with network effects, for both firms pushing different standards (inter-standards competition) and firms adopting the same standard (intra-standards competition). This dissertation applies theoretical and empirical analyses to address this need. The dissertation consists of two chapters. Chapter 2 provides a structural analysis of intra-standard price competition between an innovating firm and imitating firms. Chapter 3 provides a dynamic analysis of new product diffusion at both product category and platform levels in the presence of inter-standard competition. In Chapter 2, I analyze the intra-standards price competition in markets with network effects by using a structural analysis approach. In the literature on network effects, marketing scholars have highlighted one characteristic, installed-user-base and demonstrated its implications to marketing strategies, but under-explored the other characteristic, supporting-firm-base. Although the supporting-firm-base is important to the success of firms in face of standard competition and impacts marketing strategies 1

PAGE 13

2 differently from the installed-user-base, there are NO empirical works that have demonstrated its existence and analyzed its important implications to marketing strategies. This paper empirical examines the existence of supporting-firm-base and investigates the impacts of supporting-firm-base and installed-user-base on the competitive interactions between an innovating firm and its imitators in markets influenced by network effects. The competitive interaction among firms supporting the same technological standard (i.e., intra-standard competition) is also an important but underexplored issue. Theorists differ widely in the assumptions that they have made about the structure of competition among firms supporting the same technological standard (i.e., intra-standard competition). Although they have assumed a fixed competitive structure (e.g., bertrand, leader-follower, or cooperation) over product life cycle, we argue that the competitive structure of intra-standard competition is likely to be stage-dependent. In this study, I apply the new empirical industrial organization framework (NEIO) to examine the supporting-firm-base effects and intra-standard competition in 3.5-inch floppy disk drives during the period 1983-1998. My time-varying structural model estimations yield several interesting results. First, I find empirical evidence that consumers valuations are not only positively related to the installed base, but also positively related to the number of supporting firms. Second, my results show an opposite relationship of supporting-firm-base and installed-user-base with the competitive interactive behavior --a positive relationship between the competitive interaction and the supporting-firm-base and a negative relationship between the competitive interaction and the installed-user-base. The positive relationship suggests that the supporting-firm

PAGE 14

3 base effect discourages competitive behavior. In contrast, the negative relationship suggests that installed-user-base motivates competitive behavior. Third, my results show that the competitive structure of the floppy drive market did change over the period studied. Firms behaved collectively in the early stage, competitively in the middle stage, and collectively again in the late stage of the product life cycle. This study fills a gap in the network effects literature by providing empirical evidence on the pattern of intra-standards competition and demonstrating the implications of such intra-standards competition for innovating firms marketing strategy. The empirical findings of this study raise some important new theoretical and empirical issues for future research on markets with network effects and standards competition. In Chapter 3, I dynamically analyze the diffusion of new technology in markets with inter-standards competition by incorporating consumer expectation on network effects. Understanding the impact of inter-standards competition on new technology diffusion is extremely important to the success of new products in the presence of network effects and competing technological standards. In this study, I empirically examine how inter-standard competition affects new technology diffusion at three levels--the category-level, the standard-level and the brand-level. Specifically, I examine the impact of competition between incompatible standards on new technology diffusion by incorporating (1) market share differences, which measure the relative strength between competing standards, and (2) two important network effect factors: installed-user-base and supporting-firm-base of competing standards into our model. The network effect factors measure the absolute strength of each competing standard.

PAGE 15

4 My model involves three components: (1) a consumer utility function that captures the influences of price, product attribute, and network effectsmeasured as installed-userbase and supporting-firm-base of two competing standards, (2) equations that derive the market share of two incompatible platforms based on the estimated nested-logit model, and (3) equations that derive the dynamic diffusion of the whole industry, two competing standards and each individual brand. I first estimate a baseline model which captures the impacts of the current network effects--current user-base and firm-base from two competing standards on consumers utility and consequently on new technology diffusion. I then extend the baseline model by incorporating the consumer expectations for the future user-base and firm-base of competing technological standards into the consumer utility function. A dynamic programming method and the maximum likelihood estimation method are thus used to solve and estimate the extended model. The model is estimated using data from the VCR industry. The estimated parameters reveal several interesting results. First, my results show that while the installed-user-base from the compatible standard increases the consumers evaluation to a new product, which is consistent with the literature on network effects, the installed-user-base from competing technology standard decreases the consumers evaluation to a new product, which is new to literature on network effects, and demonstrates the evidence for one of the important impacts of standard competition on new technology diffusion. Second, I find the empirical evidence that standard competition impacts new technology diffusion complimentarily. This can be seen from two results: (1) the supporting-firm-base from both of two competing standards--VHS and Betamax positively influence the consumers evaluation on each brand, and (2) the installed-user

PAGE 16

5 base from two competing standards generates positive word of mouth effects on the category diffusion as well as the standard and brand diffusion at the early stage of standard competition when none of the two competing standards dominates market. Third, I also find results that the network effects impact the market potential dynamically. While the installed-user-base from two standards both positively impact the new technology diffusion at the early stage of standard competition, the installed-user-base of Betamax turns to impact the diffusion of new technology at three levels negatively after the technology standard VHS dominates the market. Finally, my results suggest that network effects and standard competition not only impact the diffusion on individual consumers evaluation to each brand, but only on the market potential complementarily and dynamically. Similarly with the word of mouth impact of installed-user-base on new technology diffusion, the supporting-firm-base from both of two standards impact market potentials seems positively at the early stage of standard competition, but the supporting-firm-base of Betamax has no impact on the market potential later on when the VCR market is dominated by VHS format. This study extends research on network effects and provides important implications to markets with network effects by explicitly modeling and estimating the impacts of inter-standards competition on new technology diffusion. It also advances understanding of the formation of consumer expectation for the future installed based of competing technological standardsa very important but under-explored research area.

PAGE 17

CHAPTER 2 A STRUCTURAL ANALYSIS OF INTRA-STANDARD PRICE COMPETITION IN MARKETS WITH NETWORK EFFECTS 2.1 Introduction The rapid development of information technology and the digital revolution in recent years have significantly increased the importance of network effects and standard competition in the market place (Shapiro and Varian 1999). Marketing scholars are giving increasing attention to the implications of network effects and standards competition for marketing strategies. Several new studies, for example, address such important issues as the evolution of markets with indirect network externalities (Sachin Gupta et al., 1999, Nair, Chintaguta, and Dube 2004), asymmetry network effects (Shankar and Bayus 2003), pioneer survival in network markets (Srinivasan, Lilien, and Rangaswamy 2004), and product line and technology licensing in the presence of network effect (Sun, Xie, and Cao 2004). One important but underexplored issue in the research of network effects is competitive interactions among firms supporting the same technological standard. Understanding competitive interactions among innovating and imitating firms in the presence of network effect is important to practitioners because the strategic behavior of firms is crucial to the success of the technological standard as well as the fate of the individual firms in markets with network effects. Understanding firms competitive behavior in intra-standard competition is also important from a research perspective: competition among firms adopting the same technology standard in the presence of 6

PAGE 18

7 network effects exhibits certain unique characteristics that do not apply to traditional markets, a phenomenon that raises new, interesting theoretical and empirical issues. Different from traditional markets where competition generally has a negative effect on profit, intra-standard competition in markets with network effect (i.e., competition among firms adopting the same technological standard) can be beneficial. Analyses based on game theory have shown that in markets with network effects, innovating firms can earn a higher profit by having compatible competitors than being a monopolist (Katz and Shapiro 1985, Corner 1995, Xie and Sirbu 1995, Sun, Xie, and Cao 2004) or even benefit from subsidizing compatible entrants (Economides 1996). While theoretical research has provided motivations for innovating firms to encourage compatible entry, the literature has not addressed the issue of how network effects may affect the competition structure in such markets. In the literature on network effects, theorists differ widely in the assumptions that they have made about the structure of competition. For example, some assume that innovator and imitator behave as symmetric Nash game players (e.g., Xie and Sirbu 1995, Economides 1996) while others assume that firms play Stackelberg (leader-follower) games (e.g., Corner 1995). Still other theorists assume a cooperative market structure in which both innovator and imitator behave collectively (Gallini 1984). However, there is little empirical evidence on HOW innovating and imitating firms adopting the same standard actually compete in the market place. In this study, we empirically examine the existence of supporting-firm-base effect and the intra-standard price competition between innovating and imitating firms. We argue that, although theoretical research of network effects has assumed a single fixed

PAGE 19

8 competitive structure (e.g., bertrand, leader-follower, or cooperation), we expect that the competitive structure of intra-standard competition is not fixed but varies across different stages of the product life cycle: network externalities create interdependence between consumers and firms and among compatible rivals; as such interdependence changes with the development of markets and firms may change their competitive behavior accordingly. First, network effects create an installed-user-base effect that motivates competitive behavior. Theoretical research has suggested that consumer willingness to pay for a given product increases with the size of the user base of its technology standard (Farrell and Saloner 1986, Katz and Shapiro 1985). This positive relationship between consumer willingness to pay and the installed-user-base has been supported by recent empirical studies based on data from various industries affected by network effects, including software (Gandal 1995, Brynjolfsson and Kemerer 1996,), VCRs (Hiroshi Ohashi 2003), video games (Shankar and Bayus 2003), CD players (Basu, Mazumdar, and Raj 2003), and PDAs (Nair, Chintaguta, and Dube 2004). As suggested by the literature of network effect (e.g., Katz and Shapiro 1985, Corner 1995, Economides 1996), innovating firms can benefit by encouraging competition because competition leads to a lower price and a larger total user base, thus a higher product value. This positive link between competition and profit suggests that the installed-user-base effect motivates competitive behavior in intra-standard competition. Second, network effects may create a supporting-firm-base effect that motivates cooperative behavior. Based on the assumption of fulfilled expectation, several theoretical studies suggest that consumer willingness to pay is positively affected by the

PAGE 20

9 expected total market size, which is larger for a competitive market than for a monopoly market (e.g., Katz and Shapiro 1985, Sun, Xie, and Cao 2004) and increases with the number of firms supporting the same technological standards (e.g., Economides 1996). While this theory implies a positive relationship between consumer willingness to pay and the number of supporting firms, such a relationship has not been empirically studied. Testing this possible supporting-firm-base effect will provide the empirical evidence needed to validate the existing theory; in addition, the existence of a supporting-firm-base effect would suggest a negative link between competition and consumer willingness to paycompetition discourages market entry and leads to a small supporting-firm-base, thus a low consumer willingness to pay. This negative link between competition and profit implies that the supporting-firm-base effect discourages competitive behavior but motivates cooperative behavior in intra-standard competition. Since both user base and firm base change with time, the relative strength of the installed-user-base effect and the supporting-firm-base effect may vary across different stages of the product life cycle. As a result, the competitive structure of intra-standard competition may not be fixed but stage-specific. In this paper, we empirically examine the existence of supporting-firm-base and investigate the possible stage-dependent pattern of intra-standard price competition between innovating and imitating firms. Specifically, we apply the new empirical industrial organization framework (NEIO) to examine competitive interactions in the 3.5-inch floppy disk drive (FDD) market using 16 years of data from all manufactures of floppy disk drives during the period 1983-1998. Marketing scholars have recently applied the NEIO framework to examine firms competitive pricing behavior in several markets

PAGE 21

10 without network effects. For example, Sudhir investigates the competitive pricing behavior in the U.S. author market (2001a); he also examines manufacturers pricing interactions in the presence of strategic retailers, using data from yogurt and peanut butter markets (2001b). Kadiyali (1996) examines the incumbents strategic behavior (deterrence or accommodation) in face of a new entrant, using data from the photographic film industry. These previous studies based on the NEIO framework have illustrated the importance of understanding competitive interactions to firms marketing strategies and provided us with useful methodological guidance. We chose to examine intra-standard competition in the 3.5-inch FDD market for several reasons. First, the FDD industry is subject to network effects. Different types of drives in this market are not interchangeable and the value of a given type of disk drive depends on the number of people who have adopted the same drive (i.e., the larger the installed base of a given disk drive, the more people one can share files with). Second, the innovator, Sony, faced competition from other manufactures adopting its technological standard. As Disk Trend Report recorded, Sony was willing to work with U.S. manufacturers to build its technology as the industry standard. As a result, as many as 27 other manufacturers adopted Sonys technology and produced 3.5-inch FDDs in the same market. These imitating firms were both rivals of and allies to the innovator, Sony. Third, the 3.5-inch FDD encountered intensive standards competition when it was first introduced by Sony in 1981. Before its introduction, the 5.25-inch FDD had been widely adopted in the microcomputer market. Soon after its introduction, three other new incompatible disk drives (i.e., 3-inch, 3.25-inch, and 4-inch) were also introduced in the

PAGE 22

11 FDD market. Hence, the 3.5-inch FDD market has the main characteristics we discussed above. To empirically examine the proposed supporting-firm-base effect and the stage-dependent competitive interactions, in this paper, we (1) model consumer utility as a function of both the installed-user-base and the supporting-firm base; (2) incorporate an equation of the number of supporting firms in our structural equations; and (3) model stage-dependent competitive interactions by introducing a time-varying structure to the estimation of conduct parameters. Three different approaches (menu approach, weighted profit approach, and conjectural variation) are used complementarily in testing and validating the stage-dependent competitive structure. Our empirical investigation of the floppy disc drive market reveals several interesting findings. First, we find that consumer valuations are not only positively related to the installed base, as suggested by previous empirical studies, but are also positively related to the number of manufactures that adopt the same standard. In other words, consumers care about both the installed-user-base and the supporting-firm-base. As the value of a product increases with both, this result highlights the importance of building up the base of firms adopting the same technological standards, thus the importance of competitive interaction in intra-standard competition. Second, by modeling the interrelationship of the number of supporting firms and the prices of innovating and imitating firms at previous period, we are able to identify factors that affect the supporting-firm-base. Our results show that the number of supporting firms is positively influenced by the mean price level of imitating firms in a previous period, This finding is important because it implies that competitive behavior

PAGE 23

12 discourages, but cooperative behavior encourages, market entry and suggests that the supporting-firm-base effect has an opposite impact compared with the installed-user-base that encourages competitive behavior. Third, our results show that installed-user-base motivates competitive behavior and supporting-firm-base motivates cooperative behavior, the competitive structure of the floppy disk drive market changed over the period studied. Firms behaved collectively in the early stage, competitively in the middle stage, and collectively again in the late stage of the product life cycle. Such a stage-dependent competition pattern has never been suggested on either theoretical or empirical grounds, but is certainly plausible for markets where firms supporting the same standard are both rivals and allies. As our results show, consumers willingness to pay in markets with network effects depends on BOTH the size of the installed base and the number of manufactures adopting the same standard. The latter can motivate the innovating firm to collaborate rather than aggressively compete (on price) with imitating firms at the early and later stages, though for different reasons. The early collaborative behavior helps attract more manufacturers to adopt a new standard when the new standard is still struggling for market acceptance, and the late collaborative behavior discourages manufacturers from abandoning the existing standard when the next generation of technology appears. As a result, firms are more likely to cooperate in the early and late stages, and to compete on price in the middle stage. The rest of this study is organized as follows. We introduce the baseline model in section 2.2. We discuss our data and present the empirical results based on the baseline model in section 2.3. We then introduce the modified model in section 2.4 and report the

PAGE 24

13 empirical results and validation results in section 2.5. Finally, we discuss conclusions and future research in Section 2.6. 2.2 The Baseline Model This section defines our baseline model. The baseline model has two characteristics. First, it adopts the specification for consumer behaviors as suggested in the literature of network effects (e.g. Nair, Chintaguta, and Dube 2004), which considers the impact of installed-user base but not the impact of supporting-firm base on consumer utility. Second, as many studies of competitive interactions (e.g. Sudir 2001a) it assumes a constant competitive structure. We modify the baseline model in Section 4 by (1) incorporating supporting-firm base into functions of consumer utility and conjectural parameters and (2) assuming a stage-dependent competition pattern. 2.2.1 Consumer Behavior Following the literature of network effect, we assume that consumers make their purchase decisions by maximizing their utilities and consumer utility is affected by installed-user base and product attributes. Let j denote firms, where denotes the innovator and denotes imitator.1 The utility of consumer i at time 1j 2j t for firms product is defined as j 011ijtjktkjttjtijtjtijtkuxpNu (2.1) where j kt x and j t p are the attribute and price for product at time is the cumulative sales at time (i.e., a measurement of consumer installed-base). thk j t 1tN 1t jt is the unobserved characteristic for product at time t, j jtu is the average utility across 1 In real markets, there may be more than one imitator. Following the literature (Suslow (1986) and Putsis and Dhar (1999)), firm 2 in the duopoly model (1) can be interpreted as a representative imitator.

PAGE 25

14 consumers for product at time t, and j ijt is the random term across product and consumer at time t. Following the literature (e.g. Sudir 2001b), we assume that j i ijt has double exponential distribution and is independent, identical across products and consumers. We allow consumers to have an option of purchasing outside product and normalize the utility of the outside products to zero across times. Then the market share for product at time is given by 0j j t exp1ejtjtjus xpjtu (2.2) Similarly, the market share for outside product 0j at time t is given by 011exptjtjsu (2.3) where j s and 0t s denote the market share of product and outside product j 0j From the market share equations (i.e., (2.2) and (2.3)), we can derive 0ln(/). j ttjt s su (2.4) Since the demand for product at time can be expressed by where jtD j t jtjttDsM t M is the market potential at time the demand equations can hence be written as t 0ln()ln()jttjtDD u 0. j (2.5) 2.2.2 Firm Behavior Firms make price decisions by maximizing their profits. The firm s profit at time is given by j t (), jtjtjtjtpcDj 1,2. (2.6)

PAGE 26

15 where jt p are firm s price and marginal cost at time respectively. jtc j t Marginal Cost. Firms marginal costs are defined by jtgjgtjtgcz (2.7) where j gt z are the factors that influence firms marginal cost. Note that, j gt z includes product attributes jkt x and other influential factors such as time trend. j t is the unobservable cost factors for product j Strategic Interactions. Two approaches, the conjectural variation approach (CV), and the menu approach, have been used frequently in the empirical studies of firm competitive interactions (e.g., Kadiyali 1996, Vilcassim, Kadiyali and Chintagunta 1999). The CV approach examines the strategic interactions by estimating a set of parameters and inferring the competitive structure from the signs of the set of parameters, while the menu approach does so by separately estimating alternative models and inferring the competitive structure from the selected best-fitted model. It is common for empirical researchers to use only one approach to estimate the competitive interactions because different approaches often lead to consistent results given appropriate model specification (e.g. Bresnahan 1989). Since one of our objectives is to identify more appropriate model specification (e.g., examine if supporting-firm-base should be incorporated into the consumer utility function and if state-dependent competitive structure is more appropriate than a fixed competitive structure), in this paper, we use both approaches complementarily. The CV approach (pioneered by Iwata 1974) captures the competition structure simply by a constant conjectural variation parameter. In the CV approach, a firm is

PAGE 27

16 assumed to have conjectures about how its competitors will react to its price changes and incorporate the conjectures into its price decisions (Kadiyali et al. 2001). In the baseline model, the first-order condition for firm is j 1,2() 0p jtjtjtjtjtjtijtjtltlljDDDpcpp1,2j (2.8) where /ijt it p p 1,2i (, i j ) are called conjectural variation parameters (i.e., 12/tpp 1t and 21/tpp 2t ). The conjectural variation parameters thus tell us how a firm reacts to its competitors price changes and therefore indicate the competition structure. For example, 120 indicates Bertrand competition. When 10 but 20 we infer that the innovator is leader and the imitator is follower. Symmetrically, when 20 but all 10 we infer that the imitator is leader and the innovator is follower. 12, 0 implies a cooperative market structure (Kadiyali et al. 2001). Different from CV approach, in menu approach, the competitive structure of an underlying market is inferred by estimating the alternative models (i.e., models with different competitive structure) separately and selecting the best-fitted model. We consider four possible competitive structures: Bertrand, Innovator-Led leader-follower, Imitator-Led leader-follower, and Cooperation. Under the Bertrand competition assumption, each firm chooses its price assuming that its competitor does not react to its price changes, so the first order condition for firm is j ()0, 1jtjtjtjtjtjtjtDDpcjpp ,2. (2.9)

PAGE 28

17 Under the innovator-led leader-follower assumption, innovator anticipates imitators reaction to its price and incorporates it in its price decision. Therefore, the first order condition for innovator is 111111112()tttttttttDDpDpcppp 210.ttp (2.10) In contrast, the imitator makes its price decision assuming that innovator does not react to its price. That is, the first order condition for imitators is 222222() 20tttttttDDpcpp1t (2.11) Taking derivative of imitators first order condition (2.11) with respect to, we derive the expression of and substitute it in the first order condition of innovator (2.10). Symmetrically, we can derive the first order conditions for innovator and imitators under imitator-led leader-follower assumption as 1tp 2/tpp 111111()ttttttDDpcpp 10.t (2.12) 222222221()tttttttttDDpDpcppp 120.ttp2t (2.13) Similarly, we can derive the expression of 1/tpp in first order condition (2.13) by taking derivative of first order condition (2.12) with respect to and substituting into (2.13). 2tp Finally, under the cooperation assumption, both firms act as if they were monopoly selling the differentiated products. In other words, they maximize a total profit of it own and its competitor when making their price decisions. The total profit is ()tjtjtj jt p cD The first order condition for firm is j

PAGE 29

18 1,2()0, 1tltjtltltjtjtlljDDpcjpp ,2. (2.14) In the next section, we empirically estimate the baseline model using both CV and menu approach. We show that two approaches lead to different inferred competitive structures. In Section 4, we modify the baseline model by incorporating a new variable, supporting-firm-base, into the baseline model. We also allow conjectural variation parameters to be timing-varying. We then use the time-varying CV approach to examine the modified model. Our results show a significant effect of supporting-firm base and suggest that the stage-dependent competitive structure provides a better fit than the constant competitive structure. Finally, to validate our key results, we apply menu approach to the modified model and perform stage-based estimation. The results show that the two approaches, CV and menu approach, lead to consistent results when the modified model and stage-dependent competitive structure are used. 2.3 Empirical Analysis of the Baseline Model We estimate the baseline model using data of the 3.5-inch floppy disk drive market from 1983 to 1998. 2.3.1 The 3.5-inch Floppy Disk Drive Market A floppy disk is a data storage device that comprises a circular piece of thin, flexible magnetic medium encased in a square or rectangular plastic wallet.2 Floppy disks were ubiquitously used to distribute software, transfer data between computers, and create small backups in the 1980s and 1990s. Sony introduced the 3.5-inch floppy disk 2 From http://en.wikipedia.org/wiki/Floppy_disk

PAGE 30

19 drive in 1981. Before its introduction, there were two main floppy disk drives, 8-inch and 5.25-inch disk drives. The 8-inch floppy disk drives, the first floppy disk drive introduced in 1967 by IBM, were mainly used in mainframe computers and minicomputers. The 5.25-inch disk drives, introduced by Shugart in 1976, were mainly used in minicomputers and microcomputers (PCs). Floppy disks are read and written by a floppy disk drive (FDD). A floppy disk drive basically has four principal components: read-write heads, which are located on the both sides of a disk and move together on the same assembly; a drive motor, which is a very small spindle motor driving the rotation of the disks; a stepper motor that makes a precise number of stepped revolutions to move the read-write head assembly to the proper disk track position; and electronic circuitry, which handle the data read from or written to the disk and control the stepper-motor control circuits used to mover the read-write heads to each track, as well as the movement of the read-write heads toward the disk surface. With the increasing popularity of PCs in 1980s, small size, easy to carry and display were becoming the major concerns of computer manufacturers and hence disk drive manufacturers. The 5.25-inch floppy disk drives were clearly becoming a limitation to the improvements of microcomputer size. At this time, Sony introduced its 3.5-inch disk drive. Sony uses a single-crystal ferrite head in its 3.5-inch floppy disk drive instead of the poly-crystal ferrite head used by 8-inch and 5.25-inch disk drive. The single-crystal ferrite head is lighter, more stable and more reliable than the poly-crystal ferrite head. With this stable and reliable read-write head, Sony was able to put the same number of 80

PAGE 31

20 tracks as 5.25-inch disks on much smaller 3.5-inch disks and keep its 3.5-inch disk drive with equal or even better reliability than 5.25-inch and 8-inch disk drive.3 As a result, the disk density was increased from 48 tracks per inch (TPI) to 135 TPI, and disk size was therefore largely decreased. Along with the introduction of Sonys 3.5-inch disk drive, there were two other microfloppy disk drives contending for market standard: the 3-inch drive, introduced by Matsushita Electric Industrial, Hitachi and Hitachi/Maxell; and the 3.25-inch disk drive, introduced by Tabor and Seagate Technology. Besides these two disk drives, IBM, as an industry dominant manufacturer, was also expected to introduce a 3.9-inch disk drive on an unknown schedule. These floppy disk drives with different sizes were incompatible because disks could only be loaded in the correct size of disk drive. As Disk Trend Report recorded, Sony was willing to work with U.S. manufacturers to build its technology as the industry standard. As a result, as many as 27 other manufacturers adopted Sonys technology and produced 3.5-inch FDD in the same market. Sonys 3.5-inch disk drive eventually became the industry standard. To sum up, the 3.5 inch FDD market has the main characteristics in which we are interested. First, it is subject to network effects. Second, the innovator, Sony, faced competition from imitating firms adopting its technological standard. Third, the 3.5-inch FDD encountered intensive standards competition. 3 See Mini Micro Systems, April, 1981.

PAGE 32

21 2.3.2 Data Our annual data on product attributes, sales and market share in the 3.5-inch floppy disk drive market are collected from Disk Trend Report,4 a leading annual market research publication in the disk drive industry. It has been considered as a highly reliable and complete source of disk drive industry data (Josh Lerner 1997, Christensen et al 1998). We compose all imitators into one imitator, and use a composite imitator as representative imitator in estimation as did in Suslow (1986) and Putsis and Dhar (1999). We denote the innovator and imitator hereafter as and respectively. Data on the imitators product attributes are accordingly computed by averaging all imitators product attributes at each year. Imitators sales and market share are then derived by subtracting innovators sales and market share from the sales and market share of 3.5-inch floppy disk drive market. Data on the annual average price of innovator and imitators products at each year are computed by dividing their annual dollar sales by annual sales volume. 1j 2j Disk Trend Report records the attributes for each disk drive that is introduced by each manufacturer in each year. They are listed according to three categories: (i) disk capacity, which includes total capacity, density per track, data surface per spindle, and tracks per surface; (ii) drive performance, which includes the time spent from track to track, settling time, average rotational delay time and data transfer rate; and (iii) disk drive features, which includes height, width and depth of each disk drive. By analyzing the attributes in each category, we found that the attributes in first and second categories are highly correlated. To reduce the problems of multicollinearity, we conducted step4 We would like to thank James Porter for his generous offer of the series books of Disk Trend Report.

PAGE 33

22 wise regression to choose the most influential attribute variables. Our regression analyses identify five attributes that contribute to the most of variances of price and sales: disk density (Den), track to track time (TT), height (H), width (W) and depth (D). These attributes are used in our utility function (2.1). Following literature on network effects (e.g. Brynjolfsson and kemerer 1996), the consumer installed-base in demand equation that captures the network effects is measured as the cumulative sales of whole 3.5-inch floppy disk drive market at previous period. We also measure the supporting firm base using the total number of supporting firms at previous period. The potential market size ( t M ) in the demand equation is measured as the total of the floppy disk drive market sales at each time from 1983-1998. The total of the floppy disk drive market sales is measured by adding together all sales from different sizes of floppy disk drives that were introduced in FDD market. For example, it can be sales from 8-inch, 5.25-inch, 3.5-inch, 4-inch, 3.25-inch, 3-inch and high density FDD. We use all of the above five product attributes and one time trend variable in marginal cost equation. Table 2-1 provides the descriptive statistics of our data. From Table 2-1, we can see that the difference between innovator and imitators product mainly lies in disk density. And innovator charges higher average price than imitator. Table 2-1. Descriptive Statistics Innovator Imitator Mean Std. Mean Std. t-statistic of Mean Differences Sales (Million) 6.22 4.17 33.94 33.01 Average Price 52.82 34.87 48.07 27.68 Density 135 0 131.73 5.87 2.23** Track to Track 4.23 2.73 3.74 0.58 0.70 Height 1.06 0.34 1.01 0.23 0.49 Width 3.96 0.048 4.05 0.24 1.54 Depth 5.64 0.25 5.69 0.45 -0.36 Installed-base (Million) 142 170.20 Note: **: p<0.05.

PAGE 34

23 2.3.3 Identification We assume that the product attributes jkt x are exogenous and orthogonal to the error terms ( jt and jt ); price and sales are endogenous and correlated with the error terms ( jt and jt ). This identification assumption is reasonable and commonly used in the simultaneous equation estimation (e.g. Sudir 2001b). We use product attributes, their squares and the products of innovators attribute and imitators attributes as instruments. Since innovators density remained the same over 16 years, we exclude it as one instrument variable. So we have a total of 23 instruments in the demand equation: ten exogenous variables (a constant, four innovators attributes, and five imitators attributes), nine exogenous variables square, and four products of innovators and imitators attributes. We also apply these instruments to the two competitive interaction equations. In the baseline model, we need to estimate 9 parameters in the demand equation (constant, coefficients of five product attributes, price, installed-base and supporting firm-base), 7 parameters in the marginal cost equation (constant, coefficients of five product attributes and time trend), and 2 conjectural variation parameters in the time varying first order conditions if using conjectural variation approach. Therefore, our baseline model is identifiable. 2.3.4 Empirical Results of Baseline Model In the conjectural variation approach, we simultaneously estimate the demand equation (2.5), the marginal cost equation (2.7) and the competitive interaction equation (2.8). The competitive structure is inferred by examining the signs of the two conjectural variation parameters, 1 and 2 as discussed in Section 2.

PAGE 35

24 In the menu approach, we estimate the structural equations under four alternative assumptions of competitive structures separately, i.e., for each assumed competitive structure, we simultaneously estimate the demand equation (2.5), the marginal cost equation (2.7), and one of corresponding competitive interaction equations given in (2.9)-(2.14). We then infer the competitive structure by selecting the best-fitted model based on the results of the Vuong test.5 The three-stage least square (3SLS) method is used in the estimation of both approaches. Table 2-2 provides the results of the two approaches. As shown in Table 2-2, the results of CV approach show that both conjectural variation parameters, 1 and 2 are not significantly different from zero, indicating a Bertrand competition structure. The results of menu approach show that the cooperation model has a maximum log-likelihood value (-15.09), while the log-likelihood values for the other three alternative competitive structures (Bertrand, Innovator-Led leader-follower, and Imitator-Led leader-follower) are .00, -27.97, and .75, respectively. The results of the Vuong test statistics (see Table 2-2) indicate that the cooperation model is significantly better than the other three alternatives (p<0.01 for all three comparisons). Hence, our empirical analysis of the baseline model under the assumption of a constant competitive structure leads to different conclusions about the nature of competitive interaction in the studied marketBertrand based on conjectural variation 5 The Vuong test statistic is defined as 1ln()fVpgT q where and are the log-likelihood values of the two non-nested models, and and are the number of parameters estimated in each model, respectively. V is distributed as If and then the model corresponding to is rejected in favor of the model corresponding to ln()f )ln(g p q )1,0(N 0V critical valueVV f g and vice-versa (Kadiyali et al. 2000).

PAGE 36

25 approach but Cooperation based on menu approach. This indicates that the competitive structures derived from the two approaches are inconsistent when the baseline model is used. Table 2-2. Baseline Model: Estimated Competitive Structure Conjectural Variation Approach Conjectural Variation Parameter 1 2 The log-likelihood Estimation 0.34NS (1.24) 0.23NS (0.59) -33.22 Estimated Competitive Structure Bertrand Menu Approach Alternative Models Bertrand Innovator-Led Leader-Follower Imitator-Led Leader-Follower Cooperation Log-likelihood (Vuong Statistics) -24.00*** (2.23) -27.97*** (3.22) -28.75*** (3.42) -15.09 (--) Estimated Competitive Structure Cooperation Note: ***: p<0.01; NS: p>0.1 For conjectural variation approach, the numbers in parentheses are t-statistics. For menu approach, the numbers in parentheses are Vuong statistics. The critical value of Vuong statistics at p=0.05 is 1.64. The Vuong statistics are significant at p<0.01 in all cases (See footnote 7 for discussion of Vuong test). The inconsistency found in Table 2-2 can be the results of model misspecification. As we argued, network effects create two types of interdependences: (1) the interdependence among consumers (i.e., the value of a product to consumers is affected by the number of consumers bought the same product), and (2) the interdependence between consumers and firms (i.e., the value of a product to consumers is affected by the number of the firms supporting the same technological standard). However, as the previous empirical studies of network effects, our baseline model only considers the installed-use-base effect, but not the supporting-firm-base effect. In the next section, we modify our baseline model by incorporating the supporting-firm-base effect into consumer utility function, and modeling the competitive structure to be a function of (1) installed-user-base and supporting-firm-base, and (2) time trend.

PAGE 37

26 2.4 Modified Model and Empirical Results 2.4.1 Modified Model Demand. We modify the baseline model by incorporating a new variable, supporting-firm-base, into consumer behavior. Let t H denote the number of firms supporting the innovators technological standard at time t (i.e., a measurement of supporting-firm-base). We formally incorporate the supporting-firm-base into the consumers utility model: 011112121ijtjktkjtttttjtijtjtijtkuxpNHNTHTu (2.15) Equation (2.15) assumes that consumer utility is affected not only by the installed-user-base, as assumed in the existing literature and in our baseline model, but also by the supporting-firm-base. Furthermore, since both the installed-user-base effect and the supporting-firm-base effect may change with time, we include two interaction terms, and 1(tNT )) 1(t H T in Equation (2.15) to capture the possible interaction effect, where T donate time trend. To examine how the number of supporting firms may be affected by price competition, we also models the number of supporting firms as a function of imitators price and the number of supporting firms at the previous period: 2112121ttttHHHp t (2.16) The quadratic term in Equation (2.16) is intended to capture the saturation effect. Strategic Interactions. To test the possible stage-dependent competitive interactions, we relax the assumption of a fixed competitive structure by allowing the conjectural variation parameters in equation (2.8) to be timing-varying. Specifically, we define the conjectural parameters, 1 and 2 to be two types of functions. We first define

PAGE 38

27 the conjectural variation parameters to be a function of installed-user-base and supporting-firm-base, as shown in Equation (2.17): 0112 1 jtjjtjtNH 1,2j , (2.17) and second to be a function of time trend T, as shown in Equation (2.18): 2012,jtjjjTT 1,2j (2.18) Equation (2.17) allows us to examine how the installed-user-base and supporting-firm-base directly influences the competitive structure directly, and whether the competitive structure is stage-dependent, while equation (2.18) provides us an overall pattern of the stage-dependent competitive structure over time. 2.4.2 Empirical Results: Time-Varying CV Approach We first estimate the modified model using the time-varying conjectural variation approach by modeling the competitive structure to be function of installed-user-base and supporting-firm-base as well as function of time trend, and then validate the key findings using stage-based menu approach. The structural equations for CV approach include the demand equation (2.5), the marginal cost equation (2.7), the supporting-firm-base equation (2.16), and the strategic interaction equations (2.8). Note that the jtu in (2.5) is now derived from (2.15) rather than from (2.1) as in Section 2. Also, jt in (2.8) is now a function of installed-user-base and supporting-firm-base as defined in (2.17) or function of time trend as defined in (2.18) rather than a constant as in Section 2. To derive the best fitted model, we incorporate four time dummy variables in equation (2.17). The four time dummy variables are defined as 11d if the time period is at years 1 to 4, and otherwise 0; if the time period is at years 5 to 8 and otherwise 0; if the time 21d 31d

PAGE 39

28 period is at years 9 to 12 and otherwise 0; 41d if the time period is at years 13 to 16 and otherwise 0. The estimated function of conjectural variation parameters is then written as 012345678(*1)(*2)(*3)(*4) (*1)(*2)(*3)(*4)jjjjjjjjjjNdNdNdNdHdHdHdHd (2.17) where j =1 and 2. The results of the modified model using time-varying CV approach are given in Tables 2.3-2.4. Impact of Supporting-Firm-Base on Consumers. As shown in Table 2-3 (see the results of Demand Equation on the top of Table 2-3), the installed base is positively related with consumer valuation (p<0.01), suggesting a positive installed-user-base effect. This finding is consistent with previous studies (e.g. Brynjolfsson and kemerer 1996). More importantly, our result shows that the number of the supporting firms is also positively related to consumer valuation (p<0.01), suggesting a positive supporting-firm-base effect. Table 2-3 also shows that the parameters of both interaction terms in the demand equation are significant, suggesting that both the positive installed-user-base effect and the positive supporting-firm-base effect vary across time. Impact of Price on Supporting-Firm-Base. Table 2-3 shows that the number of supporting firms is positively related to the price of imitators (see the results of Supporting-Firm-Base Equation in the middle of Table 2-3). This result combined with the positive effect of supporting-firm-base on consumer valuation discussed above suggests that the supporting-firm-base effect reduce firms incentive to aggressively compete on price, because a high price encourage more firms to adopt the same technological standard and in turn increase the value of the product to consumers.

PAGE 40

29 Table 2-3. Revised Model 1, Time-Varying CV Approach Demand and Marginal Cost Equations Demand Marginal Cost Intercept -61.78*** (-2.54)a 3877.86*** (3.89) Den 14.14*** (2.91) -278.40 (-1.56) TT -0.36 (-0.77) -46.21*** (-2.42) Height 2.31 (1.36) 368.85** (3.36) Width 28.31*** (5.36) -1919.24** (-2.30) Depth -25.28*** (-5.91) 88.62 (0.73) Time Trend 6.22** (1.91) Price -0.075*** (-5.80) Installed-base 0.62E-07*** (5.03) Installed-base* Time -0.38E-08*** (-5.15) Firm-base 0.17** (4.28) Firm-base Time -0.032*** (-5.53) Supporting-Firm-Base Equation Firm-Base at Previous Period 1.23*** (5.27) Square of Firm-Base at Previous Period -0.018** (-1.70) Imitating Firms Price at Previous Period 0.055*** (3.51) Conjectural Variation Parameters Innovator Imitator 0i (constant) 0.57* (1.52) 0.22 (0.28) User-Base Firm-Base User-Base Firm-Base 1i (at first period) -0.13E-05*** (-2.00) 0.24*** (2.02) -0.33E-04 (-0.33) 5.51 (0.33) 2i (at second period) -0.46E-07*** (-3.94) 0.13*** (4.67) -0.71E-07*** (-2.67) 0.33*** (4.10) 3i (at third period) -0.35E-08*** (-3.00) 0.04** (1.67) -0.94E-08*** (-3.91) 0.13*** (2.84) 4i ( at fourth period) -0.14E-08*** (-1.91) 0.03** (1.62) -0.29E-08** (-1.69) 0.07* (1.38) Note: The number in parentheses is t statistic. ***: p<0.01. **: p<0.05***.

PAGE 41

30 Impacts of User-Base and Firm-Base on Competitive Structure. The estimation of the impacts of user-base and firm-base on competitive structure are given in the bottom of Table 2-3 (see Time-Varying CV Parameter Estimation). As shown in Table 2-3, for both innovating and imitating firms, the coefficient of user-base across four periods are all significantly negative, while the coefficient of firm-base across four periods are all significantly positive. The different signs of the impacts of user-base and firm-base imply that installed-user-base motivates competitive behavior, while the supporting-firm-base motivates cooperative behavior. Stage-Dependent Competitive Interaction. The estimations of time-varying CV parameters where the conjectural variation parameter is defined to be a function of time trend are given in Table 2-4. As show in Table 2-4, for both innovating and imitating firms, the constant, 0j is positive (p<0.01), the linear term, 1j is negative (p<0.01), and the quadratic term, 2j is positive (p<0.01). These results suggest that (1) firms behave cooperatively at the beginning of the new product introduction, and (2) the competitive structure may vary across different stages of product life cycle. To graphically illustrate the pattern of competitive behavior in the market examined, Figure 1 shows how the conjecture variation parameters, jt (j=1,2), given in (2.18) change with time when the estimated time-varying conjecture variation parameters given in Table 2-4 are used.

PAGE 42

31 Table 2-4. Revised Model 2, Time-Varying CV Approach Demand and Marginal Cost Equations Demand Marginal Cost Intercept -54.16*** (-5.23)a -20413.3 (-1.20) Den 12.80*** (5.81) 4097.15 (1.19) TT 1.75*** (4.56) -5.35 (-0.14) Height 3.21*** (4.64) -198.08** (-2.39) Width 27.57*** (6.85) -1073.70** (-2.18) Depth -26.90*** (-7.37) 1031.99** (2.56) Time Trend -3.25** (-1.91) Price -0.051*** (-4.21) Installed-base 0.44E-07*** (4.86) Installed-base* Time -0.28E-08*** (-5.22) Firm-base 0.044** (2.34) Firm-base Time -0.016*** (-3.45) Supporting-Firm-Base Equation Firm-Base at Previous Period 1.017*** (11.99) Square of Firm-Base at Previous Period -0.020*** (-5.35) Imitating Firms Price at Previous Period 0.14*** (14.60) Conjectural Variation Parameters Innovator Imitator 0i 4.56*** (7.55) 11.18*** (5.36) 1i -0.62*** (-5.36) -1.80*** (-4.90) 2i 0.023*** (3.96) 0.075*** (4.67) Comparison: Modified vs. Baseline Model Modified model Baseline model Log-likelihood ratio Log-likelihood value -13.88 -33.22 38.68*** Note: The number in parentheses is t statistic. ***: p<0.01. **: p<0.05***. The loglikelihood ratio has 2 distribution, the critical valueat p<0.01 with d.f.=10. 218.30

PAGE 43

32 As shown in Figure 2-1, both conjecture variation parameters, 1t and 2t follow a quadratic curve. Both parameters are positive in the early years, indicating a cooperative market structure in the early stage of standard development. Then both 1t and 2t decrease to near zero at around year 12, indicating a competitive market structure in the middle stage. In the late years, both 1and t 2t increase to be positive again, indicating that the market moved toward a cooperative structure again in the late stage. Hence, Figure 1 shows a very interesting three-stage pattern of competitive structure in floppy disk drive market: cooperation in the early stage, competition in the middle stage, and cooperation again in the later stage. 02468101212345678910111213141516TimeCV parameters lamda1 lamda2 Figure 2-1. The Changes of Conjectural Variation Parameters with Time To evaluate the superiority of the stage-dependent competitive structure model over the constant competitive structure model, we test if the former fits data better than the latter by comparing the log-likelihood values of the conjectural variation models given in

PAGE 44

33 Table 2 (-33.22) with that given in Table 3 (-13.88). The log-likelihood ratio test6 suggests a significant superiority of the stage-dependent competitive model (, p<0.01, d.f. =10). 218.30 To sum up, the empirical analysis of the modified model using time-varying conjecture variation approach leads to three key findings: (1) a positive effect of supporting-firm-base on consumer valuation, (2) a positive effect of price on the number of supporting firms, and (3) a three-stage competitive interaction pattern: cooperation-competition-cooperation. To further validate these key findings, we apply a stage-based menu approach to the modified model. 2.4.3 Validation: Stage-based Menu Approach Different from CV approach that captures the stage-dependent competitive interaction via a set of time-varying conjecture variation parameters, when stage-based menu approach is used to estimate stage-dependent competitive structure, one needs to first decide how stages are divided and then apply menu approach to data in each stages. Based on results from CV approach (see Figure 2-1), we divided the data set into three stages: an early stage (year 1-5), a middle stage (year 5-12), and a later stage (year 10-16).7 For each stage, menu approach is used to estimate all four alternative models (i.e., Bertrand, Innovator-Led leader-follower, Imitator-Led leader-follower, and Cooperation), 6 We use the log-likelihood ratio test to compare the goodness of fit between these two nested models. The log-likelihood ratio is 38.68, which is larger than the critical value of at p<0.01 with freedom degree=10, showing that our time-varying conjectural variation model is significantly better than the conjectural variation model with single competitive structure. 218.30 7 Duo to the limited data available, these stages partially overlap with each other so that there are sufficient data points within each stage to allow models to be identifiable and to use Vuong test statistics to select the best-fitted models. While different possible cutoff times were considered, our final definition of stages is consistent with the curves observed in Figure 1 and Vuong test statistics are significant in selecting the best-fitted models.

PAGE 45

34 and the best-fitted model in each stage is selected using Vuong test statistics. The results of model selecting test for each stages are given in the top part of Table 2-5. As shown in Table 2-5, according to the Vuong statistical test, the best-fitted model is Cooperation in the early stage, Innovator-Led Leader-follower in the middle stage, and Cooperation again in the later stage. All other alternative models are rejected. Hence, menu approach leads to the same three-stage pattern of competitive interaction as CV approach (i.e., cooperation-competition-cooperation). In other words, when we use the baseline model and assume a constant competitive structure, the two different approaches lead to inconsistent conclusions (see Table 2-2). However, when we use the modified model and allow the competitive structure to be stage-dependent, the two approaches lead to consistent conclusions. Table 2-5 also shows the results of parameter estimations of the best-fitted models in each stage. Note that the stage-based menu approach allows not only competitive structure to vary across stages, but also parameters in all structure equations vary across stages. As a result, the estimation results using menu approach reveal more details of the market dynamics, which will be discussed in more details later. Nevertheless, as in the estimation of CV approach, the results of demand model by menu approach show a positive effect of the number of supporting firm on consumer valuation (in the early and later stage), and the results of supporting firm model by menu approach show a positive relationship between the number of supporting firms and price. Hence, our three key findings derived in CV approach are also supported by the menu approach.

PAGE 46

35 Table 2-5. Revised Model 3, Stage-Based Menu Approach Model Selection Test Stage Bertrand Innovator-Led Leader-Follower Imitator-Led Leader-Follower Cooperation Best-fitted model Early (Year 1-5) 32.52*** (3.37) 32.66*** (3.31) 32.57*** (3.35) 40.07 (---) Cooperation Middle (Year 5-12) 21.58*** (8.29) 45.03 (---) 18.72*** (9.30) 35.84*** (3.25) Innovator-Led Leader-Follower Late (Year 10-16) 3.12*** (17.71) 19.56*** (11.50) 3.75*** (17.47) 50.04 (---) Cooperation Parameter Estimation: Demand Model (The Best-fitted Model) Parameter Early Stage (Cooperation) Middle Stage (Innovator-Led Leader-Follower) Late Stage (Coorperation) Intercept -26.18 (-0.99) -258.92** (-1.94) 527.20** (1.72) Den 2.05 (0.34) 52.49** (1.98) -98.82 (-1.57) TT 2.71 (1.45) 1.06** (2.39) 0.98 (0.58) Height -2.50 (-0.50) -1.67** (-1.87) 17.64*** (7.32) Width 15.72 (0.58) 38.45*** (9.81) 24.76*** (5.08) Depth -7.19 (-0.35) -29.85*** (-10.22) -43.01*** (-5.77) Price -0.024 (-1.40) -0.0098** (-1.65) -0.081*** (-2.40) Installed-base -0.92E-07 (-1.04) -0.29E-08 (-1.54) 0.35E-08** (1.82) Firm-base 0.19*** (2.88) -0.032 (-0.92) 0.071*** (2.42) Parameter Estimation: Supporting-Firm-Base Model Parameter Early Stage Middle Stage Late Stage Firm-Base at Previous Period 0.74 (1.28) 1.18*** (10.18) 0.83*** (7.19) Square of Firm-Base at Previous Period -0.0081 (-0.42) -0.018*** (-4.91) -0.038*** (-3.85) Imitating Firms Price at Previous Period 0.14*** (4.03) 0.061** (1.87) 0.39*** (4.48) Note: ***: p<0.01, **: p<0.05. For model selection test, the numbers in parentheses are Vuong statistics. The critical value of Vuong statistics at p=0.05 is 1.64. For parameter estimations, the numbers in parentheses are t-statistics. 2.4.4 Discussion of Key Findings First, previous empirical studies have shown a positive relationship between installed-user-base and consumer valuation (e.g. Nair, Chintaguta, and Dube 2004). Such

PAGE 47

36 a positive relationship has been explained by a positive network effect: the larger the user base, the higher the value of the product. By simultaneously examining the installed-user-base and supporting-firm-base, we find that both can have a positive effect on consumer adoption decision. More interestingly, our stage-based analyses show that the relative importance of installed-user-base effect and supporting-firm-base effect varies across stages of product life cycle. For example, in the early stage of 3.5-inch floppy disk drive, the consumer valuation is positively affected by the supporting-firm-base but not by the installed-use-base (see Table 2-5). While this result is not intuitive, it may result from the high uncertainty faced by consumers in the presence of standards competition. In the early stage of product introduction, 3.5-inch disk drive faced intense competition from multiple incompatible technological standards including the old 8-inch and 5.25-inch floppy disk drives as well as the new 3-inch, 3.25-inch and 4-inch floppy disk drives. Since different types of drives were not interchangeable, the adoption utility of a given type of disk drive depends on the outcome of the standards competition. While it is difficult for early buyers to predict if the new 3.5-inch floppy disk technology would be widely adopted by the market, the number of firms supporting the Sonys new innovation in the early stage could be useful information to consumers in evaluating the odds of the new technology because the larger the number of supporting firms, the more likely that the new technology will become industry standards. The fact that consume valuation is affected by the supporting-firm-base but not by the installed-user-base in the early stage suggests that at the beginning of 3.5-inch disk drive introduction, consumers were more concerned about the outcome of ongoing standards battle but less about the current user base of the drive. While conventional wisdom has emphasized the importance of

PAGE 48

37 building consumer base in markets with network effects, our finding highlights the importance of simultaneously considering both the installed-user-base and supporting-firm-base. As we will discuss next, the installed-user-base effect and supporting-firm-base effect can have opposite implications on price competition. Second, our empirically analyses show a positive relationship between the number of supporting firms and price. This result implies that the positive supporting-firm-base effect discussed above reduces firms incentive to cut price because a high price encourages more firms to adopt the innovators technological standards, which in turn increases consumer valuation. Note that it is commonly agreed that the positive installed-user-base effect motivates firms compete more aggressively because a low price encourages more consumers to adopt the new product, which in turn increases consumer valuation. Take the two effects together, our findings suggest that while both supporting-firm-base and installed-user-base increase consumer valuation, they have opposite impacts on firms competitive behavior: the former motivates cooperative behavior but the latter motivates competitive behavior. Hence, the nature of firms strategic interaction is expected to be jointly determined by the two effects and may change across different stages of its product life cycle. Third, our most interesting finding is the three-stage competitive interaction pattern found in 3.5-inch floppy disk drive market: firms cooperate in the early stage, compete in the middle stage, and then cooperate again in the late stage. This pattern can be explained by the relative importance of the two effects (i.e., installed-user-base effect and supporting-firm-base effect) to consumers in different stages, and the opposite impacts of the two effects on firms competitive behavior. In the early stage of 3.5-inch

PAGE 49

38 disk drive introduction, 3.5-inch disk drive faced intense competition from several competing standards (see Figure 2-2), which drawn consumers attention toward the supporting-firm-base rather than the current installed-use-base (see Table 4). Firms adopted cooperative behavior because cooperation helps but competition obstructs building up supporting-firm-base. In 1990 (i.e., 9 years after the introduction of 3.5 floppy disk drive), the sales of 3.5 floppy disk drive have surpassed the sales of all alternative technologies (see Figure 2), indicating that 3.5 floppy disk drive had established its technology standard position. In the next few years (the middle stage), neither installed-user-base nor supporting-firm-base have a significant effect on consumer valuation (see Table 4), implying that in this stage, the product is evaluated largely based on its performance rather than by network effects. As a result, there would be little incentive for firms to attract more imitators; rather, they had strong incentive to compete aggressively to achieve as much profit as possible. In 1995 (12 years after the introduction of 3.5-inch disk drive), new generation technology, high-density floppy disk drive, enters the market. In this late stage, consumers were concerned about both the installed-user-base and supporting-firm-base, perhaps because both indicate how likely the existing technology would be replaced by the new generation technology. Since the supporting-firm-base was more important to consumers than the installed-user-base (see Table 2-5), firms adopted cooperative behavior again to discourage manufactures from abandoning the existing standard.

PAGE 50

39 0.020000.040000.060000.080000.0100000.0120000.019831985198719891991199319951997TimeSales (Thousands) 3.5 inch 8+5.25 inch High Capacity Figure 2-2. Sales: 3.5 inch Floppy Disk Drive vs. Competing Products 2.5 Conclusion In this paper, we use the new empirical industrial organization framework to examine price competition between innovating and imitating firms. Different from the existing research of network effects, we consider not only an installed-user-base effect but also a supporting-firm-base effect on consumer utility. To examine the impact of the supporting-firm-base on intra-standard competition, we model the interrelationship of the number of imitating firms with the prices of innovating and imitating firms. To allow our models to capture the stage-dependent price competition structure, we introduce time-varying parameters to the conjectural variation approach and validate our estimation results using menu approach. Our results provide empirical evidence that markets with network effects generate both a positive consumer-installed-base effect and a positive supporting-firm-base effect (on consumer utilities). However, these two effects have opposite impacts on firms

PAGE 51

40 competitive behaviors. To increase consumer installed-base, firms need to keep price low, which motivates firms to compete. In contrast, to encourage other firms to adopt the innovators technology, firms need to keep price high, which motivates firms to cooperate. Furthermore, these two opposite effects impact the price competition between firms differently at different stage. Our results using both approaches suggest that the innovator and imitators have more incentive to encourage more manufactures to support a new standard in its early stage and discourage existing manufacturers from abandoning an established standard in its later stage when facing the challenge of next generation technology. As a result, firms are more likely to cooperate in the early and later stage, and compete on price in the middle stage. This paper makes several contributions to the literature of network effects. First, we examine an effect overlooked in the literature, the supporting-firm-base effect. Our results not only provide empirical evidence supporting a positive relationship between the number of firms adopting the same technological standard and consumers valuation, but also suggest that the supporting-firm-base effect works in an opposite direction compared with the well-known installed-user-base effect in terms of their impacts on intra-standard competition. Second, our study provides empirical evidence of a three-stage pattern of competitive interaction in intra-standard competition. Our results suggest such a pattern in the floppy drive market over the period studied: firms behaved collectively in the early stage, competitively in the middle stage, and collectively again in the late stage. Such a stage-dependent competition pattern is new to the literature on network effects and raises

PAGE 52

41 interesting theoretical and empirical questions for future research on the strategic implications of stage-dependent intra-standard competition. Finally, this paper illustrates the use of complementary approaches to evaluate different structural models and estimate stage-dependent competitive structures. This approach can be applied to other situations where a constant competitive structure is questionable. Overall, this study fills a gap in the network effects literature by providing empirical evidence on patterns of intra-standard competition. Our findings raise some important new theoretical and empirical issues for future research on markets with network effects and standards competition. However, since our study only uses data from the floppy disk drive industry, the generality of our findings needs to be tested in other industries with network effects.

PAGE 53

CHAPTER 3 DYNAMIC ANALYSIS OF NEW TECHNOLOGY DIFFUSION IN MAREKTS WITH COMPETING STANDARDS 3.1 Introduction The diffusion of new technology has been extensively studied since the traditional diffusion model was published (e.g., Bass 1969, Mahajan and Muller 1979, Norton and Bass 1987). Recently, the rapid development of information technology and the digital revolution attract researchers attention again to the diffusion of new technology, especially in markets with network effects. While studies on the adoption of the new technology such as the ATM machine (Saloner and Shepard 1995), Compact Disc (Gandal, Kende and Rob 2000), DVD player (Inceogle and Park 2003, Dranove and Gandal 2003), and ACH electronic payments system (Gowrisankaran and Stavins 2004) demonstrate the importance of the network effects (i.e. installed-user-base effect) on the new technology diffusion, they all focus on the new technology diffusion at category-level. There is NO empirical research examines the standard diffusion and how the competition between incompatible technologies impacts the new technology diffusion at three levels--category-level, standard-level and brand-level. Understanding the new technology diffusion at standard-level and the impacts of standards competition on the new technology diffusion at three levels is extremely important because the markets with network effects and standards competition exhibit 42

PAGE 54

43 certain unique characteristics that do not apply to the traditional markets and have not been addressed in the literature of new technology diffusion and network effects. First, in markets with network effects and standards competition, the new technology diffusion at three-levels is interdependent: the standard diffusion impacts the market potential for whole category diffusion and the consumers evaluation for an individual brand, while the category diffusion impacts also the market potentials for each standards diffusion and the brand diffusion comprises of the standards diffusion. Studying the interdependence of the new technology diffusion at three-levels requires the marketing scholars to integrate the category diffusion, standard diffusion and brand diffusion into a systematic framework. However, most of the traditional diffusion models (e.g. Bass 1969, Bass, Krishnan and Jain 1994, Horsky 1990) focus only on the category diffusion without considering the impacts of standards and brands diffusion on each other. Though there is little diffusion research in recent years that has proposed models to incorporate the competition impact on diffusion (e.g. Norton and Bass 1992, Mahajan and Muller 1996), there is NO research addressing the standard diffusion in a systematic framework. Second, standards competition impacts the new technology diffusion differently from the competitions among competing brands in the traditional markets. The differences exist in two aspects: (1) unlike the traditional markets where the competitions among competing brands may expend the whole category diffusion positively, the standards competition in markets with network effects may impact the market potential negatively. This is because the high uncertainty that consumers have on which standard will dominant market deters consumers from making choice immediately when there is

PAGE 55

44 no single technological standard dominating the market. (2) Standards competition generates different word of mouth effects on the new technology diffusion at the stage when the incompatible technologies are competing for market dominance from the stage when the market is dominated by one of the competing technologies. Recent studies about the impacts of network effects on new technology diffusion has revealed the positive impacts of network effects on a particular technology standard (e.g. Saloner and Shepard 1995, Gandal, Kende and Rob 2000, Park 2003 and Ohashi 2003) or the negative impacts of network effects on competing standard (e.g. Dranove and Gandal 2003), but there is NO empirical works demonstrating how the standard competition impacts the new technology diffusion on the market potential and consumers word of mouth effects. Third, literature has suggested (e.g. Katz and Shapiro 1985 and Shapiro and Varian 1999) that consumer expectation is a major factor in consumer decisions about whether or not to purchase a new technology, and the management of consumer expectation on network effects is a critical factor in standard competition and hence the technology diffusion, but no empirical works demonstrate to managers on how consumers form their expectations on network effects, especially on the future installed-user-base and supporting-firm-base. This study sets up a systematic framework by integrating the new technology diffusion at three levels. Specifically, this study builds a choice-base diffusion model by incorporating (1) the standard competition---measured as the market share differences between the competing standards, and (2) two important network effect factors: installed

PAGE 56

45 user-base and supporting-firm-base of competing standards into a baseline model and a dynamic model. Our choice-based diffusion model involves three components: (1) a consumer utility model that captures the influences of price, product attribute, and network effectsmeasured as installed-userbase and supporting-firm-base of two competing standards, (2) equations that derive the market share of two incompatible platforms based on the estimated nested-logit model, and (3) equations that derive the dynamic diffusion at category-level, standard-level and brand-level. Different from the baseline model that only captures the current network effects---current user-base and firm-base of two competing standards, our dynamic model captures the consumer expectations for the future user-base and firm-base of competing technological standards, and thus involves using a dynamic programming method to estimate the market share equations for each competing standard and brand. Our model is estimated using data from the VCR industry. The competition in the home VCR market is one of the best-known examples of technology standard competition between two incompatible format Betamax and VHS (Park 2003). Though the Betamax format was first introduced by Sony in February 1976, VHS format, which is introduced one and half year later by JVC in 1977 and incompatible with the Betamax format, was able to surpass the Betamax in the early years of 1980s and eventually dominated the U.S. VCR market in the late years of 1980s. Examining this market would provide valuable insights on how the inter-standard competition impacts the new product diffusion at category-level, standard-level and brand-level.

PAGE 57

46 The estimated parameters reveal several interesting results. First, we find that while the installed-user-base from the compatible standard increases the consumers evaluation to a new product, which is consistent to the literature on network effects, the installed-user-base from competing technology standard decreases the consumers evaluation to a new product, which is new to literature on network effects, and demonstrates the evidence for one of the important impacts of standard competition on new technology diffusion. Second, we find the empirical evidence that standard competition impacts new technology diffusion complimentarily. This can be seen from two results: (1) the supporting-firm-base from both of two competing standards---VHS and Betamax positively influence the consumers evaluation on each brand, and (2) the installed-user-base from both of two competing standards generates positive word of mouth effects on the category diffusion as well as the standard and brand diffusion at the early stage of standard competition when none of the two competing standards dominates market. Third, we also find results that the network effects impact the market potential dynamically. While the installed-user-base from two standards both positively impact the new technology diffusion at the early stage of standard competition, the installed-user-base of Betamax turns to impact the diffusion of new technology at three levels negatively after the technology standard VHS dominates the market. Finally, our results suggest that network effects and standard competition not only impact the diffusion on individual consumers evaluation to each brand, but only on the market potential complementarily and dynamically. Similarly with the word of mouth impact of installed-user-base on new technology diffusion, the supporting-firm-base from

PAGE 58

47 both of two standards impact market potentials seems positively at the early stage of standard competition, but the supporting-firm-base of Betamax has no impact on the market potential later on when the VCR market is dominated by VHS format. This study extends research on network effects and provides important implications to markets with network effects by explicitly modeling and estimating the impacts of inter-standards competition on new technology diffusion. It also advances understanding of the formation of consumer expectation for the future installed based of competing technological standardsa very important but under-explored research area. Our paper is organized as follows. We first introduce our baseline model and dynamic model in the section 3.2 and 3.3. Then we present our preliminary results in section 3.4. We discuss our results and future research directions in section 3.5. 3.2 The Baseline Model We start with analyzing the new product diffusion at three levels: category, platform and brand levels by incorporating the impact of network effects and standard competition into (1) the equation of consumers evaluation of each brand, (2) the market potential equation and (3) the diffusion equations at three levels. The network effect variables that are incorporated in the baseline model are the current installed-user-base and the current supporting-firm-base. The dynamic model by considering consumer expectation about future network effects will be introduced in the next section. 3.2.1 Consumer Utility Equation We use the nested logit model to estimate the market shares. Let i denote consumer and denote brand ( j 0,1,2,, j J ), where outside option is denoted as 0 j Different technology standard groups are denoted as g ( G g,,2,1,0 ). The outside

PAGE 59

48 option is the only product in group 0j 0 g Further, denote the set of brands in each group as () and denote the value of brand as g g Jgg j 01,1,1'',1'',1''1 jtkjktjtkgjgjgtjgjgtjgjgtjgjgtlljtggglxpNFNFMSD (3.1) where and denote the product attribute and price for brand ; jktx jtp k j l M SD denotes the market share difference between the compatible technology standard of brand j and other competing technology standards g We incorporate four network effect variables:, and in the utility function to measure the installed-user-base and the supporting-firm-base for each brands compatible technology standard 1,tjgN 1,'tjgN 1,tjgF 1,'tjgF g and competing technology standard respectively. 'g Then the utility for consumer i is accordingly modeled as (1).ijtjtigtijtU (3.2) where ijt is an identically and independently distributed extreme value; and ])1([ijtigt is also an extreme value random variable (i.e., double exponential distribution)). 3.2.2 Market Share Equations If brand is compatible to technology standard then the formula for the market share of brand as a fraction of the total standard share (see Berry 1994) is j g j ggjDesjt/][)1/(/ (3.3) where Similarly, the market share for the technology standard is gjtjgeD)1/( g

PAGE 60

49 ggggDDs)1()1( (3.4) Thus, the market share for brand as a fraction of all brands regardless its technology standard can be derive as j gggjDDesjt)1()1/( (3.5) With the market share of outside options ggDs)1(01 and )1/()]ln()[ln()ln(0 ssDgg we can then derive a simple analytic equation for mean utility levels as )ln()ln()ln(/0gjjtjsss (3.6) and the parameters in (3.6) can then be obtained by instrumental estimation methods. 3.2.3 Diffusion Equation at Three Levels To derive the diffusion equation at three levels, we need model the market potential first. Instead of modeling the market potential as TV households as in literature (e.g., Park 2003 and Ohashi 2003), we model it as a function of the average price of all products in market, product quality and network effects at different standard competition stages. Define a parameter as gtggtggtggkjkjkMSDFMSDNNap1,01,11,00))(max()( where represents the average price of all products in market; is the maximum level of each product attribute among different brands. The ap )(maxjkjx MSDNtg 1, and MSDFtg 1,

PAGE 61

50 denote the interactive terms of network effect variables---installed-user-base and supporting-firm-base with the market share difference. These interactive terms allow us to capture the dynamic influences of network effects across different stages of standard competition. The market potential is then modeled as a function of TV households (TVHH) and the defined parameter : )exp( TVHHMt (3.7) Accordingly, we model the diffusion equations for the whole category as 1100,11,1()(ttttggtggtg ) X XMXNNMSD (3.8) where the industry potential t M is defined in (3.7). Similarly, each standards and each brands diffusion is given by ,,1,00,11,1'0',1'1',1'((gtgtgtgggtggtggtggtgg )) X XMNNMSDNNMSD (3.9) And ,,1,0110,11,1'0',1'1',1'((jtjtjtjtgjgtgjgtgjgtgjgtjgg )) X XMXNNMSDNNMSD (3.10) where the market potential level for each technology standard and each brand are measured as 1()jtjtttMsMX and ,,(gtgtttMsMX 1) given the industry potential measure t M and market shares (3.4)-(3.6). 3.3 Dynamic Model with Consumer Expectation We now incorporating consumer expectation about network effects into the logit model and estimate it by using dynamic programming and maximum loglikelihood estimation method.

PAGE 62

51 3.3.1 Consumer Expectation of User-Base and Firm-Base The consumer expectations of the future installed-user-base g tN and supporting-firm-base g t F are modeled as functions of current installed-user-base ,1 g tN and supporting-firm-base ,1 g tF of competing standards. The function is written as ,,1,11(),1GgtglgtglgtgutlNNFg ,2,,.G1,2,,. (3.11) ,,1,11(),Ggtgfgtgfgtgftl F NFg G (3.12) where g ut and g ft are assumed to follow independent and identical multinomial distribution. 3.3.2 Consumer Dynamic Decision Different from the myopic assumption that a consumer makes purchase decision by only maximizing its single period utility, we assume that a dynamic oriented consumer looks forward periods ahead and maximize its sum of discounted expected future over T periods, given its expectations about the future network effects. The objective function of a dynamic oriented consumer sum of discounted expected future is then defined as T tU 1max()ijtjJTJttijtijtdtjEUd tS (3.13) where is the single-period utility at time t of a consumer i for a product defined as (3.2); is an indicator that equals to 1 if a consumer i chooses a product at time and equals to zero otherwise; and ijtU j ijtd j t is a discount factor. The is the mathematical expectation operator and the is the state space at time The state space ()E tS t

PAGE 63

52 tS includes installed-user-base g tN and supporting-firm-base g t F for different standard g A dynamic oriented consumers decision involves making a sequence of decisions on in each period to maximize its expected discounted utility. Denote the maximum value of discounted expected utility over the horizon T as ijtd ()max()ijtjJTtitttijtijttdtjVSEUdS (3.14) The value function can be rewritten as ()ittVS (3.15) (())max(()itijttjJVStVS where the alternative-specific expected value function, is the expected value of consumer selecting alternative at time According to Bellman (1957), we can derive the alternative specific value function at time t recursively by the following equation ()ijttVS i j t 11()()(,,1,)ijttijttitttijtijtVSEUSEVSSd (3.16) where the alternative specific value at last period is T ()()ijTTijTTVSEUS As seen in (3.16), the alternative-specific expected utility function assumes that decisions are made by maximizing the current period utility plus the expected total future value, given the current choice. The discounted expectation in (3.16) is taken over the distribution of conditional on the current state decision and current error term (1)St ()St iktd ijt with the transition density denoted by ((1)(),,)ijtijtfStStd

PAGE 64

53 3.3.3 Choice Probability Similarly as in standard choice models that the choice probability of a consumer for product is made by comparing the values of different alternatives, the choice probability of a dynamic oriented consumer i for product is also made by comparing the values of different alternatives. The major difference with the dynamic model is that the values of different alternatives include both the current period utility and the expected discounted utilities at time t over the remaining periods. Assuming that (1) the unobservables i j j ijt is conditional independent on observable state variables and (2) the unobservables ()St ijt is distributed as i.i.d. multivariate extreme value. Then we have the choice probability for consumer i to choose alternative at time as (Rust 1994): j t ''0exp()Pr(1).exp()jttikttJjttjVSdSVS (3.17) where () j ttVS represents the deterministic portion of the alternative specific value function. That is, 11()() j ttjttttVSUEVSS Given the choice probability (3.17) for an individual consumer i to brand and the observed choice quantities for each brand at each period, we can derive the loglikelihood function by assuming homogenous consumers or a particular heterogeneous function of different consumers. This loglikelihood function hence can be estimated by using maximum loglikelihood estimation methods. Consequently, the market shares and diffusion equations for each technology standard and brands can be finally estimated by using the estimated parameters from the maximum loglikelihood estimation methods. j

PAGE 65

54 3.4 Empirical Analysis We estimate the baseline model by using data from the U.S. home VCR market from 1978 to 1996. This time period ranges from the early stage of market emerge, the middle stage of the intensive standard competition and to the late market stable stage. The competition in the home VCR market is one of the best-known examples of technology standard competition between two incompatible format Betamax and VHS (Park 2003). As we know, though the Betamax format was first introduced by Sony in February 1976, VHS format, which is introduced one and half year later by JVC in 1977 and incompatible with the Betamax format, was able to surpass the Betamax in the early years of 1980s and eventually dominated the U.S. VCR market in the late years of 1980s. Examining this market would provide valuable insights on how the inter-standard competition impacts the new product diffusion at the market level, standard level and brand level. 3.4.1 Data Our data set includes annual sales, prices and product attributes on top 20 brands. The annual sales data for the whole VCR market are collected from Consumer Electronics in Review, published by Consumer Electronics Association. The market share data for top 20 brands in this market from 1981 to 1996 are collected from issues of Television Digest with Consumer Electronics, published by Warren Publishing Inc.. The market share data for each brand from 1978 to 1980 are adopted from Ohashi (2003) and Park (2003). The top 20 brands in our original sample from 1978 to 1996 count up to 90% of the total market share. Data on prices and product attributes are collected from New York Times, Orion Blue Book Video and Television (1993, 1997) and Consumer Reports.

PAGE 66

55 To obtain a preliminary result about the diffusion at each technology standard and brand, we include the top six brands---RCA, Panasonic, Sony, Zenith, JVC and Magnavox in our estimation. These brands are the major earliest brands in the VCR market, which count for 50% of the total market share on average. RCA, Panasonic and Magnavox adopted VHS format from the beginning when they entered this market. In contrast, Sony and Zenith first adopted format Betamax but switched to format VHS in the late 1980s. Each VCR producer usually introduced multiple models in each year, however since the sales data on model-level are not available, we select a basic model to represent a brand in each year (Park 2003). The product attributes that are included in our preliminary estimation are heads, programming events and programming days. The descriptive statistics for these variables are reported in table 3-1. Table 3-1. Descriptive Statistics Product Attribute Brand Heads Programming Events Programming Days Price Sales (Million) Market Share Format RCA 3.42 (0.96) 6.21 (2.37) 197 (182) 754.1 (305) 1.04 (0.56) 0.16 (0.0076) VHS Panasonic 3.16 (1.01) 5 (3.11) 18.63 (12.6) 710.3 (253.7) 0.79 (0.45) 0.10 (0.035) VHS Magnavox 3.16 (1.01) 5.68 (2.67) 125.3 (167.5) 734.5 (341.04) 0.64 (0.56) 0.065 (0.029) VHS JVC 3.26 (0.99) 5.32 (3.09) 124.2 (168.2) 862.1 (399.4) 0.37 (0.26) 0.036 (0.018) VHS Sony 3.05 (1.18) 5.32 (2.94) 16.42 (12.7) 819.7 (259.4) 0.42 (0.25) 0.098 (0.105) Beta to VHS Zenith 3.37 (1.16) 4.84 (3.48) 102.9 (161.1) 699 (285.8) 0.52 (0.29) 0.034 (0.017) Beta to VHS Network Effect Standard Installed-User-Base Supporting-Firm-Base VHS 19.34 (18.68) 31.21 (17.02) Betamax 2.57 (1.48) 4.21 (3.61)

PAGE 67

56 3.4.2 Nested Logit Model Estimation As introduced previously, to estimate diffusion equations at three levels, we need first estimate the market share for each brand and technology standard. We use GMM estimation method to estimate the nested logit model (3.6), where the product attributes in our preliminary estimation are Head, Programming Days and Programming Events. The estimation results of the nested logit model are presented in table 3-2. From table 3-2, we find that the installed-user-base from compatible technology standard positively influences consumers evaluation about a new product, which is consistent to literature on network effects (Gandal 1995, Brynjolfsson and Kemerer 1996). In contrast, we also find that the installed-user-base from competing technology influences consumers evaluation negatively. This is new to network effects literature and demonstrates the impacts of standard competition on consumers evaluation and in turn the market share and diffusion of each brand. Similarly as the impacts of user-base from compatible technology standard, we find that the supporting-firm-base from the compatible standard also positively influences consumers evaluation. This reinforces the results in Chapter 2 that supporting-firm-base, as one of important source of network effects which has been ignored in the literature of network effect, increases consumers evaluation about a new product. Different from the negative impact of user-base from competing standard, our results show that the supporting-firm-base from competing standard positively impacts the consumers evaluation. This interesting result suggests that the consumers evaluation of a brand increases with more entrants in the market no matter which competing standards that an entrant chooses.

PAGE 68

57 Table 3-2. Nested Logit Model Estimation Coefficient Estimate t-statistics Constant -4.67*** -12.33 Head 0.11 1.25 Programming Days -0.00025 -0.58 Programming Events -0.73E-03 -0.29 Price -0.12*** -3.54 User-base for compatible standard 0.043*** 5.63 User-base for competing standard -0.053** -2.32 Firm-base for compatible standard 0.019*** 4.03 Firm-base for competing standard 0.034*** 2.16 )/ln(gSj Cumulative Sales at t-1 Market Share Difference 0.87*** 0.014 8.48 10.18 0.63 0.32 ***: p<0.01, **: p<0.05, and *: p<0.1. 3.4.3 Diffusion Equation Estimation With the estimation of the nested logit model, we then derive the estimated market share for each technology standard and brand. Using the estimated market share, the diffusion equations at three levels---category-level (3.8), standard-level (3.9) and brand-level (3.10) are accordingly estimated by using General Linear Estimation method. Impacts of Network Effects and Standard Competition on Market Potential The estimation results of market potential equation (3.7) are reported in table 3-3. From table 3-3, we can see that the market potential is significantly influenced by the market average price and product quality. Specifically, we found that average market price significant decreases the market potential and one of the product quality variable---Programming Events significantly increases the market potential. In terms of the impact of network effects and standard competition, thought we only find significant results on firm-base from Betamax standard, the signs of user-base and firm-base from both of two competing standards---VHS and Betamax provide some interesting insights.

PAGE 69

58 Table 3-3. Market Potential Equation Estimation Coefficient Estimation Average Price -0.50E-03*** (-4.38) Head -0.062*** (-4.69) Programming Days -0.91E-03*** (-4.42) Programming Events 0.14** (1.94) VHS User-base 0.0040 (0.14) Betamax User-base -0.29 (-2.73) VHS Firm-base -0.26E-02 (-0.30) Betamax Firm-base 0.19*** (6.63) VHS User-base *Market Share Difference 0.012 (0.058) Betamax User-base *Market Share Difference -0.65 (-0.90) VHS Firm-base *Market Share Difference 0.078 (0.62) Betamax Firm-base *Market Share Difference -2.98*** (-6.83) ***: p<0.01, **: p<0.05, and *: p<0.1. First, we can see from figure 1 that the user-base of technology VHS influences the market potential positively and this positive impacts becomes stronger as it dominates its competing standards (i.e., the market share difference between VHS and Betamax increases). In contrast, the user-base of technology Betamax influences the market potential negatively, and this negative impact becomes stronger as it is dominated by its competing standards.

PAGE 70

59 Figure 1. Impacts of Installed-User-Base on Market Potential-2-1.5-1-0.500.512345678910111213141516171819YearImpact User-Base of VHS on Market Potential User-Base of Beta on Market Potential Figure 3-1. Impacts of Installed-User-Base on Market Potential Second, the impacts of firm-base on market potentials are different at two different stages: at the early stage, since the firm-base of technology Betamax is larger than that of technology VHS due to the first introduction of technology Betamax, the market potential increases as more firms support the technology Betamax and decreases with more firms supporting technology VHS. However, as more firms switch to technology VHS while the firm-base of technology Betamax turns to zero at the late years, the impacts of firm-base from two technology standards on market potentials reverse: the firm-base of VHS impacts market potentials positively while the firm-base of Betamax impacts market potentials negatively (see table 3-3 and figure 3-2).

PAGE 71

60 Figure 2. Impacts of Supporting-Firm-Base on Market Potential-0.200.20.40.60.811.21.412345678910111213141516171819YearImpact Impacts of VHS Firm-Base on Market Potential Impacts of Beta Firm-Base on Market Potential Figure 3-2. Impacts of Supporting-Firm-Base on Market Potential Impacts of Word of Mouth and Standard Competition on Diffusion As discussed previously, different from the traditional diffusion models that only study the word of mouth impacts from the whole markets cumulative sales, in our models, we not only incorporate the word of mouth impacts on diffusion from two competing standards separately, but also incorporate the interactive impacts of word of mouth with the stages of standards competition, where the stages of standards competition are measured as the extent of market share difference. The results of the diffusion equation estimations at three levels are reported in table 3-4 and 3-5. Table 3-4. The Diffusion Equations Estimation: Category and Standard Level Coefficient Category-Level Standard-Level Constant -2.42*** (-3.37) 0.068 (1.26) User-base of VHS 0.014* (1.38) 0.014 (1.44) User-base of Beta 0.53*** (2.64) 0.14*** (2.54) User-base of VHS*Market Share Difference 0.27*** (2.20) 0.026** (1.69) User-base of Beta*Market Share Difference -6.29** (-2.47) -0.60*** (-2.74) ***: p<0.01, **: p<0.05, and *: p<0.1.

PAGE 72

61 Table 3-5. The Diffusion Equations Estimation: Brand Level Coefficient Brand-Level Constant -0.29*** (-19.58) User-base of compatible standard 0.0015*** (5.37) User-base of competing standard 0.0013 (1.48) User-base of compatible standard*Time Trend 0.0036*** (13.68) User-base of Competing standard*Time Trend -0.058*** (16.47) Brand own user-base 0.0023*** (4.29) ***: p<0.01, **: p<0.05, and *: p<0.1. Consistent with the results in the literature of new product diffusion, we find that the word of mouth impacts from two competing standards---VHS and Betamax both positively influence the diffusion of new product at three levels. But differently, we also find that (1) the word of mouth impact of Betamax is larger than that of VHS at the early years, but (2) at the late years, the word of mouth impact of VHS increases while that of Betamax becomes negatively (see figure 3-38). Figure 3. Impacts of Word of Mouth from VHS and Betamax on Category Diffusion-5-4-3-2-101234512345678910111213141516171819YearImpact Impacts of VHS on Market Diffusion Impact of Betamax on Market Diffusion Figure 3-3. Impacts of Word of Mouth on Category Diffusion 8 We only present the impact of word of mouth from two competing standards on the whole market diffusion here. The word of mouth impacts from two competing standards on the standards and brands diffusion are similar as figure 3-3.

PAGE 73

62 More interestingly, we found that as the standard competition appears clear on which standard dominates market (i.e., the market share difference becomes large), the dominated standard (which is the Betamax technology in our data) will no longer generate positive word of mouth, but instead generate negative word of mouth on new product diffusion at three levels (see figure 3-4). This result further suggests that before the standard competition has settled to one of the competing standards, both of the two competing standards contribute to the diffusion at three levels positively and complementary. Once the market settled to one technology standard, the other technology standards would become barriers to the expansion of market, standards and brands. Figure 4. Impact of Betamax User-Base On Category and Standard Diffusion-5-4-3-2-10112345678910111213141516171819YearImpact Impact of Betamax User-Base on Market Diffusion Impact of Betamax on Standard Diffusion Figure 3-4. Impact of Betamax User-Base on Category and Standard Diffusion 3.5 Conclusion and Future Research This paper examines how the inter-standard competition affects new technology diffusion at three levels---the category-level, the standard-level and the brand-level. Specifically, we examine the impact of competition between incompatible standards on new technology diffusion by incorporating (1) the standard competition---measured as the market share differences between the competing standards, and (2) two important

PAGE 74

63 network effect factors: installed-user-base and supporting-firm-base of competing standards into (1) the consumer utility function, (2) market potential equation, and (3) the diffusion equations at three levels. The estimation of the diffusion equations at three levels reveals several interesting results. First, our results show that while the installed-user-base from the compatible standard increases the consumers evaluation to a new product, which is consistent to the literature on network effects, the installed-user-base from competing technology standard decreases the consumers evaluation to a new product, which is new to literature on network effects, and demonstrates the evidence for one of the important impacts of standard competition on new technology diffusion. Second, we find the empirical evidence that standard competition impacts new technology diffusion complimentarily. This can be seen from two results: (1) the supporting-firm-base from both of two competing standards---VHS and Betamax positively influence the consumers evaluation on each brand, and (2) the installed-user-base from both of two competing standards generates positive word of mouth effects on the category diffusion as well as the standard and brand diffusion at the early stage of standard competition when none of the two competing standards dominates market. Third, we also find results that the network effects impact the market potential dynamically. While the installed-user-base from two standards both positively impact the new technology diffusion at the early stage of standard competition, the installed-user-base of Betamax turns to impact the diffusion of new technology at three levels negatively after the technology standard VHS dominates the market.

PAGE 75

64 Fourth, our results suggest that network effects and standard competition not only impact the diffusion on individual consumers evaluation to each brand, but only on the market potential complementarily and dynamically. Similarly with the word of mouth impact of installed-user-base on new technology diffusion, the supporting-firm-base from both of two standards impact market potentials seems positively at the early stage of standard competition, but the supporting-firm-base of Betamax has no impact on the market potential later on when the VCR market is dominated by VHS format. Since this is only a preliminary study, there are several directions on which we can work in the future. First, in our preliminary estimation, we only included top 6 brands rather than 20 brands that we have in our original datasets. Future work will include all of the 20 brands in our examination. Second, it would provide additional insights to us on how the competitive behavior among firms impacts the new technology diffusion at three levels. Specifically, we are interested in how the different competitive interactions (i.e., Bertrand, Leader-follower) among firms within a same technological standard (i.e., intra-standard firms) and firms across different technology standards (i.e., inter-standard firms) impact the new technology diffusion differently. Third, we are still working on our dynamic model by taking consumer expectation into consideration. Results from our dynamic model not only provide deep insights on how individual consumer forms its expectation on network effects, particularly on installed-user-base and firm-base for every competing standard, but also enable us to compare the results on new technology diffusion at three levels with our baseline model. The comparison between our baseline model and dynamic model provides

PAGE 76

65 methodological implications on whether considering the consumer dynamics improves the prediction accuracy on new technology diffusion at three levels. Finally, the generalization of our study is limited. To see if our results can be generalized across different markets with network effects and standard competition, additional studies are needed on other markets with standard competition.

PAGE 77

LIST OF REFERENCES Bass, Frank M. (1969), A New Product Growth Model for Consumer Durables, Management Science, 15 (January), 215-227. -----, Trichy V. Krishman and Dipak C. Jain (1994), Why the Bass Model Fits Without Decision Variables, Marketing Science, 13, 203-223. Basu, Amiya, Tridib Mazumdar and S P Raj (2003), Indirect Network Externality Effects on Product Attributes, Marketing Science, 22 (2), 209-221. Bellman, Richard (1957), Dynamic Programming, Princeton, NJ: Princeton University Press. Berry, S. (1994), Estimating Discrete-Choice Models of Product Differentiation, Rand Journal of Economics, Vol. 25 (2), 242-262. Besanko, D., Gupta S., and Jain D. (1998), Logit Demand Estimation under Competitive Pricing Behavior: An Equilibrium Framework, Management Science, 44 (11), 1533-1547. Bresnahan, T. (1989), Industries and Market Power, In: Schmalensee, R., Willig, R. (Eds), Handbook of Industrial Organization, North Holland, Amsterdam. Brynjolfsson, Erik and Chris F. Kemerer (1996), Network Externalities in Microcomputer Software: an Econometric Analysis of the Spreadsheet Market, Management Science, 42 (12), 1627-1647. Christensen, Clayton M., Fernando F. Suarez, and James M. Utterback (1998), Strategies for Survival in Fast-Changing Industries, Management Science, 44 (2), Part 2 of 2, S207-S220. Corner, Kathleen R. (1995), Obtaining Strategic Advantage from Being Imitated: When Can Encouraging Clones Pay? Management Science, 41 (2), 209-225. Dranove, D. and Gandel N. (2003), The DVD vs. Divx Standard War: Empirical Evidence of Network Effects and Preannouncemnt Effects, Journal of Economics and Management Science, 12 (3), 363-386. Economides, Nicholas (1996), Network Externalities, Complementarities, and Invitation to Enters, European Journal of Political Economy, 12, 211-233. 66

PAGE 78

67 Farrell, Joseph and Garth Saloner (1986), Installed base and Compatibility: Innovation, Product Preannouncement, and Predation, The American Economic Review, 76 (5), 940-955. Gallini, Nancy T. (1984), Detterence By Market Sharing: A Strategic Incentive for Licensing, The American Economic Review, 74 (5), 931-941. Gandal, N. (1995), Competing Compatibility Standards and Network Externalities in the PC software Market, Review of Economics and Statistics, 77 (4), 599-608. --------, M. Kende and R. Rob (2000), The Dynamics of Technological Adoption in Hardware/Software Systems: The Case of Compact Disc Players, Rand Journal of Economics, 31, 43-61. Garud, Raghu and Arun Kumaraswamy (1993), Changing Competitive Dynamics in Network Industries: An Exploration of Sun Microsystems Open Systems Strategy, Strategic Management Journal, 14, 351-369. Gasmi, F., Laffont J.J., and Vuong Q. (1992), Econometric Analysis of Collusive Behavior in a Soft Drink Market, Journal of Economics and Management Strategy, 1 (2), 277-311. Gowrisankaran, G. and J. Stavins (2004), Network Externalities and Technology Adoption: Lessons from Electronic Payments, Rand Journal of Economics, 35 (2), 260-276. Gupta, Sachin, Dipak C. Jain, and Mohanbir S. Sawhney (1999), Modeling the Evolution of Markets with Indirect Network Externalities; An Application to Digital Television, Marketing Science, 18 (3), 396-416. Horsky, Dan (1990), A Diffusion Model Incorporating Product Benefits, Price, Income and Information, Marketing Science, 9 (4), 324-385. Inceoglu, Firat and Minsoo Park (2003), Diffusion of a New Product under Network Effects: The Case of U.S. DVD Player Market, Boston University, working paper. Iwata, G. (1974), Measurement of Conjectural Variations in Oligopoly, Econometrica, 42 (5), 947-966. Kadiyali, Vrinda (1996), Entry, Its Deterrence, and Its Accommodation: A Study of the U.S. Photographic Film Industry, Rand Journal of Economics, 27 (3), 452-478. ------, Vrinda, Pradeep Chintagunta, and Naufel Vilcassim (2000), Manufacturer-Retailer Channel Interactions and Implications for Market Power: an Empirical Investigation of Pricing in a Local Market, Marketing Science, 19 (2), 127-148.

PAGE 79

68 -----, K. Sudhir, and Vithala R. Rao (2001), Structural Analysis of Competitive Behavior: New Empirical Industrial Organization Methods in Marketing, International Journal of Research in Marketing, 18, 161-186. Katz, Michael and Carl Shapiro (1985), Network Externalities, Competition and Compatibility, The American Economic Review, 75 (3), 424-440. Lerner, Josh (1997), an Empirical Exploration of a Technology Race, Rand Journal of Economics, 28 (2), 228-247. Mahajan, Vijay and Eitan Muller (1979), Innovation Diffusion and New Product Growth Models in Market, Journal of Marketing, 43, 55-68. ----------------and ---------------(1996), Timing, Diffusion, and Substitution of Successive Generations of Technological Innovation: The IBM Framework Case, Technological Forecasting and Social Change, 51, 109-132. Nair, Harikesh, Pradeep Chintagunta, and Jean-Pierre Dube (2004), Empirical Analysis of Indirect Network Effects in the Market for Personal Digital Assistants, Quantitative Marketing and Economics, 2, 23-58. Norton, John A. and Frank M. Bass (1987), A Diffusion Model of Adoption and Substitution for Successive Generations of High Technology Products, Management Science, 33 (September), 1069-1086. --------and --------(1992), Evolution of Technological Generations: The Law of Capture, Sloan Management Review, 33 (Winter), 66-77. Ohashi, H (2003), The Role of Network Effects in the U.S. VCR Market, 1978-1986, Journal of Economics and Management Strategy, Vol. 12 (4), 447-494. Park, S. (2003), Quantitative Analysis of Network Externalities in Competing Technologies: The VCR Case, The Review of Economics and Statistics, 86 (4), 937-945. Rust, J. (1987), Optimal Replacement of GMC Bus Engines: An Empirical Model of Harold Zurcher, Econometrica, 55 (5), 999-1033. Saloner, Garth and Andrea Shepard (1995), Adoption of Technologies with Network Effects: An Empirical Examination of the Adoption of Automated Teller Machines, Rand Journal of Economics, 26 (3), 479-501. Shankar, Venkatesh and Berry L. Bayus (2003), Network Effects and Competition: An Empirical Analysis of the Home Video Game Industry, Strategic Management Journal, 24, 375-384. Shapiro, C. and R. Varian (1999), The Art of Standard Wars, California Management Review, 41 (2), 8-32.

PAGE 80

69 -----and -----(1999), Information Rules: A Strategic Guide to the Network Economy, Harvard Business School Press, Boston, Massachusetts. Srinivasan, Raji, Gary L. Lilien, and Arvind Rangaswamy (2004), First in, First out? The Effects of Network Externalities on Pioneer Survival, Journal of Marketing, 68 (January), 41-58. Sudhir, K. (2001a), Competitive Pricing Behavior in the Auto Market: A Structural Analysis, Marketing Science, 20 (1), 42-60. ----(2001b), Structural Analysis of Manufacturer Pricing in the Presence of a Strategic Retailer, Marketing Science, 20 (3), 244-264. Sun, Baohong, Jinhong Xie, and H. Henry Cao (2004), Product Strategy for Innovators in Markets with Network Effects, Marketing Science, 23 (2), 243-254. Vilcassim Naufel J., Vrinda Kadiyali and Pradeep V. Chintagunta (1999), Investigating Dynamic Multifirm Market Interactions in Price and Advertising, Management Science, 45 (4), 499-518. Xie, Jinhong and Marvin Sirbu (1995), Price Competition and Compatibility in the Presence of Positive Demand Externalities, Management Science, 41, 909-926. Wang, Q. and Jinhong Xie (2005), A Structural Analysis of Intra-Standard Price Competition in Markets with Network Effects, University of Florida, working paper.

PAGE 81

BIOGRAPHICAL SKETCH Qi Wang was born in Henan, China, in 1969. She earned a Bachelor of Applied Mathematics at Zhengzhou University with highest honors in 1991, and subsequently a Master of Econometrics at Zhongshan University with highest honors in 1994. Before joining the Ph.D. program in the Marketing Department at the University of Florida, she had been an assistant professor at Zhongshan University for two years and an editor at Zhongshan University Press for five years. During her five years of study at the University of Florida, she taught an undergraduate course in Sales Management in 2003 and 2005. She has now accepted a position as an Assistant Professor of Marketing at State University of New York at Binghamton. 70


Permanent Link: http://ufdc.ufl.edu/UFE0011601/00001

Material Information

Title: Analysis of markets in the presence of network effects and standards competition
Physical Description: Mixed Material
Language: English
Creator: Wang, Qi ( Dissertant )
Xie, Jinhong ( Thesis advisor )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2005
Copyright Date: 2005

Subjects

Subjects / Keywords: Marketing thesis, Ph. D.
Dissertations, Academic -- UF -- Marketing
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
theses   ( marcgt )

Notes

Abstract: The rapid development of information technology and the digital revolution in recent years have significantly increased the importance of network effects in the success of many new products. When multiple technology standards compete in markets with network effects, consumers's willingness to pay for any given standard depends much more strongly on market size and market share than in markets without network effects. This creates new challenges for marketing strategy. There is a pressing need to understand the marketing implications of technology standards in markets with network effects, for both firms pushing different standards (inter-standards competition) and firms adopting the same standard (intra-standards competition). This dissertation applies theoretical and empirical analyses to address this need. The dissertation consists of two essays. Chapter 2 provides a structural analysis of intra-standard price competition between an innovating firm and imitating firms. Chapter 3 provides a dynamic analysis of new technology diffusion at three levels: product category, platform and brand levels in the presence of inter-standard competition. Data collected from different industries with standards competition are used to test the proposed models.
Subject: competition, effects, empirical, inter, intra, network, standards, structural
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 81 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.
Resource Identifier: aleph - 003398418
System ID: UFE0011601:00001

Permanent Link: http://ufdc.ufl.edu/UFE0011601/00001

Material Information

Title: Analysis of markets in the presence of network effects and standards competition
Physical Description: Mixed Material
Language: English
Creator: Wang, Qi ( Dissertant )
Xie, Jinhong ( Thesis advisor )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2005
Copyright Date: 2005

Subjects

Subjects / Keywords: Marketing thesis, Ph. D.
Dissertations, Academic -- UF -- Marketing
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
theses   ( marcgt )

Notes

Abstract: The rapid development of information technology and the digital revolution in recent years have significantly increased the importance of network effects in the success of many new products. When multiple technology standards compete in markets with network effects, consumers's willingness to pay for any given standard depends much more strongly on market size and market share than in markets without network effects. This creates new challenges for marketing strategy. There is a pressing need to understand the marketing implications of technology standards in markets with network effects, for both firms pushing different standards (inter-standards competition) and firms adopting the same standard (intra-standards competition). This dissertation applies theoretical and empirical analyses to address this need. The dissertation consists of two essays. Chapter 2 provides a structural analysis of intra-standard price competition between an innovating firm and imitating firms. Chapter 3 provides a dynamic analysis of new technology diffusion at three levels: product category, platform and brand levels in the presence of inter-standard competition. Data collected from different industries with standards competition are used to test the proposed models.
Subject: competition, effects, empirical, inter, intra, network, standards, structural
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 81 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.
Resource Identifier: aleph - 003398418
System ID: UFE0011601:00001


This item has the following downloads:


Full Text












ANALYSIS OF MARKETS IN THE PRESENCE OF NETWORK EFFECTS AND
STANDARDS COMPETITION















By

QI WANG


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

QI WANG



































To My Parents















ACKNOWLEDGMENTS

This dissertation is indebted to many people. First, I would like to thank my

advisor, Jinhong Xie, for guiding me through my entire Ph.D. study and the entire

dissertation process. She is not only the chair of my dissertation committee, but also a

mentor and a lifelong friend to me. Without her constant support, indefatigable

encouragement, steady mentoring and patient guidance, I could not have developed

strong interests in research, accomplished this dissertation and gone through the

difficulties in my graduate study and my personal life. There is no single word in the

world that could express my great appreciation to her.

Second, I would also like to thank my committee members and other faculty

members for their extensive comments and feedback for this dissertation. In particular, I

benefited immensely from Barton Weitz, Steven Shugan, Chunrong Ai, Mike Lewis and

Debi Mitra. I am also indebted to my friends for their dependable help, especially Yubo

Chen, Qiong Wang, Qian Tang, Younghan Bae, Xu Jun, Lucy Liu, Shuman Li and Song

Xue. The friendship with them makes my graduate study enjoyable and fun; the help

from them made it possible for me to complete my Ph.D. study while bringing up my

daughter.

Last, but certainly not least, I am extremely grateful from the bottom of my heart to

my family--my husband, my daughter, my parents, my brother and sister. I feel extremely

lucky to grow up in a family where my parents, my brother and sister are always

selflessly with me for every decision I make, even though they are physically far from









me. I am thankful to my husband for his understanding and to my daughter for all of the

joy that she brought to me in my life, especially in the five-year-long study. It is their

unconditional love and support that make all my accomplishment possible.
















TABLE OF CONTENTS

page

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

LIST OF TABLES .................................................... ....... .. .............. viii

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

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

CHAPTER

1 IN T R O D U C T IO N ............................................................................. .............. ...

2 A STRUCTURAL ANALYSIS OF INTRA-STANDARD PRICE
COMPETITION IN MARKETS WITH NETWORK EFFECTS.............................6

2.1 Introduction..................................................... ................... .. ....... ...... 6
2.2 The B baseline M odel .................. ............................ .... .... .. ........ .... 13
2.2.1 C onsum er B behavior ........................................................ ............. 13
2.2.2 Firm B ehavior................. ............................................ .. ...... .. .......... 14
2.3 Empirical Analysis of the Baseline M odel ............. .............................................18
2.3.1 The 3.5-inch Floppy Disk Drive Market..............................................18
2.3.2 D ata.................................... ........................ ..... ... ......... 21
2.3.3 Identification ................ .................... .............. ............. .. .......... 23
2.3.4 Empirical Results of Baseline M odel ............... ................ ............... 23
2.4 Modified Model and Empirical Results.....................................................26
2.4.1 M odified M odel ........................ ............. .... ........................... 26
2.4.2 Empirical Results: Time-Varying CV Approach ....................................27
2.4.3 Validation: Stage-based Menu Approach.......................................33
2.4.4 D discussion of K ey Findings............................................ .. .............. .35
2.5 C conclusion ..................................... ................................ .......... 39

3 DYNAMIC ANALYSIS OF NEW TECHNOLOGY DIFFUSION IN MAREKTS
W ITH COM PETING STANDARDS..................................... ........................ 42

3 .1 Introdu action ......... .... ....... .. .......................................................... 42
3.2 T he B baseline M odel .............................................. .. ...................... .... 47
3.2.1 Consum er U utility Equation..................................... ......... ............... 47
3.2.2 M market Share Equations ........................................ ......... ............... 48









3.2.3 Diffusion Equation at Three Levels ................................. ............... 49
3.3 Dynamic Model with Consumer Expectation.....................................................50
3.3.1 Consumer Expectation of User-Base and Firm-Base...............................51
3.3.2 Consumer Dynamic Decision..................................................................51
3.3.3 Choice Probability ......................................... .... .. .... .. .. ........ .... 53
3 .4 E m pirical A naly sis.......... .......................................................... ... .... .. .... .. 54
3 .4 .1 D ata ....................................................................................................54
3.4.2 Nested Logit Model Estimation....................................... ............... 56
3.4.3 D iffusion Equation Estim ation................... ............................................ 57
Impacts of Network Effects and Standard Competition on Market
P potential ................................................. ................ 57
3.5 Conclusion and Future Research ........................................ ....... ............... 62

L IST O F R EFE R E N C E S ............................................................................. ............. 66

B IO G R A PH IC A L SK E TCH ..................................................................... ..................70
















LIST OF TABLES

Table pge

2 -1 D escriptiv e Statistics ......................................... .. ................................ ............... 22

2-2 Baseline Model: Estimated Competitive Structure ..............................................25

2-3 Revised Model 1, Time-Varying CV Approach ............................................... 29

2-4 Revised Model 2, Time-Varying CV Approach ............................................... 31

2-5 Revised Model 3, Stage-Based Menu Approach .............................................. 35

3-1 D escriptiv e Statistics .......................................................... .......... ..................... 55

3-2 N ested Logit M odel Estim ation ........................................ ......................... 57

3-3 M market Potential Equation Estim action ........................................ ............... 58

3-4 The Diffusion Equations Estimation: Category and Standard Level .....................60

3-5 The Diffusion Equations Estimation: Brand Level ............................................61
















LIST OF FIGURES


Figure pge

2-1 The Changes of Conjectural Variation Parameters with Time..............................32

2-2 Sales: 3.5 inch Floppy Disk Drive vs. Competing Products ..................................39

3-1 Impacts of Installed-User-Base on Market Potential ............................................59

3-2 Impacts of Supporting-Firm-Base on Market Potential .......................................60

3-3 Impacts of Word of Mouth on Category Diffusion............................. ...............61

3-4 Impact of Betamax User-Base on Category and Standard Diffusion.......................62















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

ANALYSIS OF MARKETS IN THE PRESENCE OF NETWORK EFFECTS AND
STANDARDS COMPETITION


By

Qi Wang

August 2005

Chair: Jinhong Xie
Major Department: Marketing

The rapid development of information technology and the digital revolution

in recent years have significantly increased the importance of network effects in the

success of many new products. When multiple technology standards compete in markets

with network effects, consumers' willingness to pay for any given standard depends

much more strongly on market size and market share than in markets without network

effects. This creates new challenges for marketing strategy. There is a pressing need to

understand the marketing implications of technology standards in markets with network

effects, for both firms pushing different standards (inter-standards competition) and firms

adopting the same standard (intra-standards competition). This dissertation applies

theoretical and empirical analyses to address this need. The dissertation consists of two

essays. Chapter 2 provides a structural analysis of intra-standardprice competition

between an innovating firm and imitating firms. Chapter 3 provides a dynamic analysis of

new technology diffusion at three levels: product category, platform and brand levels in









the presence of inter-standard competition. Data collected from different industries with

standards competition are used to test the proposed models.














CHAPTER 1
INTRODUCTION

The rapid development of information technology and the digital revolution in

recent years have significantly increased the importance of network effects in the success

of many new products. When multiple technology standards compete in markets with

network effects, consumers' willingness to pay for any given standard depends much

more strongly on market size and market share than in markets without network effects.

This creates new challenges for marketing strategy. There is a pressing need to

understand the marketing implications of technology standards in markets with network

effects, for both firms pushing different standards (inter-standards competition) and firms

adopting the same standard (intra-standards competition). This dissertation applies

theoretical and empirical analyses to address this need. The dissertation consists of two

chapters. Chapter 2 provides a structural analysis of intra-standardprice competition

between an innovating firm and imitating firms. Chapter 3 provides a dynamic analysis of

new product diffusion at both product category and platform levels in the presence of

inter-standard competition.

In Chapter 2, I analyze the intra-standards price competition in markets with

network effects by using a structural analysis approach. In the literature on network

effects, marketing scholars have highlighted one characteristic, installed-user-base and

demonstrated its implications to marketing strategies, but under-explored the other

characteristic, supporting-firm-base. Although the supporting-firm-base is important to

the success of firms in face of standard competition and impacts marketing strategies









differently from the installed-user-base, there are NO empirical works that have

demonstrated its existence and analyzed its important implications to marketing

strategies. This paper empirical examines the existence of supporting-firm-base and

investigates the impacts of supporting-firm-base and installed-user-base on the

competitive interactions between an innovating firm and its imitators in markets

influenced by network effects.

The competitive interaction among firms supporting the same technological

standard (i.e., intra-standard competition) is also an important but underexplored issue.

Theorists differ widely in the assumptions that they have made about the structure of

competition among firms supporting the same technological standard (i.e., intra-standard

competition). Although they have assumed a fixed competitive structure (e.g., bertrand,

leader-follower, or cooperation) over product life cycle, we argue that the competitive

structure of intra-standard competition is likely to be stage-dependent.

In this study, I apply the new empirical industrial organization framework (NEIO)

to examine the supporting-firm-base effects and intra-standard competition in 3.5-inch

floppy disk drives during the period 1983-1998. My time-varying structural model

estimations yield several interesting results. First, I find empirical evidence that

consumers' valuations are not only positively related to the installed base, but also

positively related to the number of supporting firms. Second, my results show an opposite

relationship of supporting-firm-base and installed-user-base with the competitive

interactive behavior --- a positive relationship between the competitive interaction and

the supporting-firm-base and a negative relationship between the competitive interaction

and the installed-user-base. The positive relationship suggests that the supporting-firm-









base effect discourages competitive behavior. In contrast, the negative relationship

suggests that installed-user-base motivates competitive behavior. Third, my results show

that the competitive structure of the floppy drive market did change over the period

studied. Firms behaved collectively in the early stage, competitively in the middle stage,

and collectively again in the late stage of the product life cycle.

This study fills a gap in the network effects literature by providing empirical

evidence on the pattern of intra-standards competition and demonstrating the implications

of such intra-standards competition for innovating firms' marketing strategy. The

empirical findings of this study raise some important new theoretical and empirical issues

for future research on markets with network effects and standards competition.

In Chapter 3, I dynamically analyze the diffusion of new technology in markets

with inter-standards competition by incorporating consumer expectation on network

effects. Understanding the impact of inter-standards competition on new technology

diffusion is extremely important to the success of new products in the presence of

network effects and competing technological standards. In this study, I empirically

examine how inter-standard competition affects new technology diffusion at three levels-

-the category-level, the standard-level and the brand-level. Specifically, I examine the

impact of competition between incompatible standards on new technology diffusion by

incorporating (1) market share differences, which measure the relative strength between

competing standards, and (2) two important network effect factors: installed-user-base

and supporting-firm-base of competing standards into our model. The network effect

factors measure the absolute strength of each competing standard.









My model involves three components: (1) a consumer utility function that captures

the influences of price, product attribute, and network effects-measured as installed-

user- base and supporting-firm-base of two competing standards, (2) equations that derive

the market share of two incompatible platforms based on the estimated nested-logit

model, and (3) equations that derive the dynamic diffusion of the whole industry, two

competing standards and each individual brand. I first estimate a baseline model which

captures the impacts of the current network effects--current user-base and firm-base from

two competing standards on consumer's utility and consequently on new technology

diffusion. I then extend the baseline model by incorporating the consumer expectations

for the future user-base and firm-base of competing technological standards into the

consumer utility function. A dynamic programming method and the maximum likelihood

estimation method are thus used to solve and estimate the extended model.

The model is estimated using data from the VCR industry. The estimated

parameters reveal several interesting results. First, my results show that while the

installed-user-base from the compatible standard increases the consumer's evaluation to a

new product, which is consistent with the literature on network effects, the installed-user-

base from competing technology standard decreases the consumer's evaluation to a new

product, which is new to literature on network effects, and demonstrates the evidence for

one of the important impacts of standard competition on new technology diffusion.

Second, I find the empirical evidence that standard competition impacts new

technology diffusion complimentarily. This can be seen from two results: (1) the

supporting-firm-base from both of two competing standards--VHS and Betamax

positively influence the consumer's evaluation on each brand, and (2) the installed-user-









base from two competing standards generates positive word of mouth effects on the

category diffusion as well as the standard and brand diffusion at the early stage of

standard competition when none of the two competing standards dominates market.

Third, I also find results that the network effects impact the market potential

dynamically. While the installed-user-base from two standards both positively impact the

new technology diffusion at the early stage of standard competition, the installed-user-

base of Betamax turns to impact the diffusion of new technology at three levels

negatively after the technology standard VHS dominates the market.

Finally, my results suggest that network effects and standard competition not only

impact the diffusion on individual consumer's evaluation to each brand, but only on the

market potential complementarily and dynamically. Similarly with the word of mouth

impact of installed-user-base on new technology diffusion, the supporting-firm-base from

both of two standards impact market potentials seems positively at the early stage of

standard competition, but the supporting-firm-base of Betamax has no impact on the

market potential later on when the VCR market is dominated by VHS format.

This study extends research on network effects and provides important implications

to markets with network effects by explicitly modeling and estimating the impacts of

inter-standards competition on new technology diffusion. It also advances understanding

of the formation of consumer expectation for the future installed based of competing

technological standards-a very important but under-explored research area.














CHAPTER 2
A STRUCTURAL ANALYSIS OF INTRA-STANDARD PRICE COMPETITION IN
MARKETS WITH NETWORK EFFECTS

2.1 Introduction

The rapid development of information technology and the digital revolution in

recent years have significantly increased the importance of network effects and standard

competition in the market place (Shapiro and Varian 1999). Marketing scholars are

giving increasing attention to the implications of network effects and standards

competition for marketing strategies. Several new studies, for example, address such

important issues as the evolution of markets with indirect network externalities (Sachin

Gupta et al., 1999, Nair, Chintaguta, and Dube 2004), asymmetry network effects

(Shankar and Bayus 2003), pioneer survival in network markets (Srinivasan, Lilien, and

Rangaswamy 2004), and product line and technology licensing in the presence of

network effect (Sun, Xie, and Cao 2004).

One important but underexplored issue in the research of network effects is

competitive interactions among firms supporting the same technological standard.

Understanding competitive interactions among innovating and imitating firms in the

presence of network effect is important to practitioners because the strategic behavior of

firms is crucial to the success of the technological standard as well as the fate of the

individual firms in markets with network effects. Understanding firms' competitive

behavior in intra-standard competition is also important from a research perspective:

competition among firms adopting the same technology standard in the presence of









network effects exhibits certain unique characteristics that do not apply to traditional

markets, a phenomenon that raises new, interesting theoretical and empirical issues.

Different from traditional markets where competition generally has a negative

effect on profit, intra-standard competition in markets with network effect (i.e.,

competition among firms adopting the same technological standard) can be beneficial.

Analyses based on game theory have shown that in markets with network effects,

innovating firms can earn a higher profit by having compatible competitors than being a

monopolist (Katz and Shapiro 1985, Corer 1995, Xie and Sirbu 1995, Sun, Xie, and Cao

2004) or even benefit from subsidizing compatible entrants (Economides 1996). While

theoretical research has provided motivations for innovating firms to encourage

compatible entry, the literature has not addressed the issue of how network effects may

affect the competition structure in such markets. In the literature on network effects,

theorists differ widely in the assumptions that they have made about the structure of

competition. For example, some assume that innovator and imitator behave as symmetric

Nash game players (e.g., Xie and Sirbu 1995, Economides 1996) while others assume

that firms play Stackelberg (leader-follower) games (e.g., Corer 1995). Still other

theorists assume a cooperative market structure in which both innovator and imitator

behave collectively (Gallini 1984). However, there is little empirical evidence on HOW

innovating and imitating firms adopting the same standard actually compete in the market

place.

In this study, we empirically examine the existence of supporting-firm-base effect

and the intra-standard price competition between innovating and imitating firms. We

argue that, although theoretical research of network effects has assumed a single fixed









competitive structure (e.g., bertrand, leader-follower, or cooperation), we expect that the

competitive structure of intra-standard competition is not fixed but varies across different

stages of the product life cycle: network externalities create interdependence between

consumers and firms and among compatible rivals; as such interdependence changes with

the development of markets and firms may change their competitive behavior

accordingly.

First, network effects create an installed-user-base effect that motivates competitive

behavior. Theoretical research has suggested that consumer willingness to pay for a

given product increases with the size of the user base of its technology standard (Farrell

and Saloner 1986, Katz and Shapiro 1985). This positive relationship between consumer

willingness to pay and the installed-user-base has been supported by recent empirical

studies based on data from various industries affected by network effects, including

software (Gandal 1995, Brynjolfsson and Kemerer 1996,), VCRs (Hiroshi Ohashi 2003),

video games (Shankar and Bayus 2003), CD players (Basu, Mazumdar, and Raj 2003),

and PDAs (Nair, Chintaguta, and Dube 2004). As suggested by the literature of network

effect (e.g., Katz and Shapiro 1985, Corner 1995, Economides 1996), innovating firms

can benefit by encouraging competition because competition leads to a lower price and a

larger total user base, thus a higher product value. This positive link between

competition and profit suggests that the installed-user-base effect motivates competitive

behavior in intra-standard competition.

Second, network effects may create a supporting-firm-base effect that motivates

cooperative behavior. Based on the assumption of fulfilled expectation, several

theoretical studies suggest that consumer willingness to pay is positively affected by the









expected total market size, which is larger for a competitive market than for a monopoly

market (e.g., Katz and Shapiro 1985, Sun, Xie, and Cao 2004) and increases with the

number of firms supporting the same technological standards (e.g., Economides 1996).

While this theory implies a positive relationship between consumer willingness to pay

and the number of supporting firms, such a relationship has not been empirically studied.

Testing this possible supporting-firm-base effect will provide the empirical evidence

needed to validate the existing theory; in addition, the existence of a supporting-firm-base

effect would suggest a negative link between competition and consumer willingness to

pay-competition discourages market entry and leads to a small supporting-firm-base,

thus a low consumer willingness to pay. This negative link between competition and

profit implies that the supporting-firm-base effect discourages competitive behavior but

motivates cooperative behavior in intra-standard competition.

Since both user base and firm base change with time, the relative strength of the

installed-user-base effect and the supporting-firm-base effect may vary across different

stages of the product life cycle. As a result, the competitive structure of intra-standard

competition may not be fixed but stage-specific.

In this paper, we empirically examine the existence of supporting-firm-base and

investigate the possible stage-dependent pattern of intra-standard price competition

between innovating and imitating firms. Specifically, we apply the new empirical

industrial organization framework (NEIO) to examine competitive interactions in the 3.5-

inch floppy disk drive (FDD) market using 16 years of data from all manufactures of

floppy disk drives during the period 1983-1998. Marketing scholars have recently applied

the NEIO framework to examine firms' competitive pricing behavior in several markets









without network effects. For example, Sudhir investigates the competitive pricing

behavior in the U.S. author market (2001a); he also examines manufacturers' pricing

interactions in the presence of strategic retailers, using data from yogurt and peanut butter

markets (2001b). Kadiyali (1996) examines the incumbent's strategic behavior

(deterrence or accommodation) in face of a new entrant, using data from the photographic

film industry. These previous studies based on the NEIO framework have illustrated the

importance of understanding competitive interactions to firms' marketing strategies and

provided us with useful methodological guidance.

We chose to examine intra-standard competition in the 3.5-inch FDD market for

several reasons. First, the FDD industry is subject to network effects. Different types of

drives in this market are not interchangeable and the value of a given type of disk drive

depends on the number of people who have adopted the same drive (i.e., the larger the

installed base of a given disk drive, the more people one can share files with). Second,

the innovator, Sony, faced competition from other manufactures adopting its

technological standard. As Disk Trend Report recorded, Sony was willing to work with

U.S. manufacturers to build its technology as the industry standard. As a result, as many

as 27 other manufacturers adopted Sony's technology and produced 3.5-inch FDDs in the

same market. These imitating firms were both rivals of and allies to the innovator, Sony.

Third, the 3.5-inch FDD encountered intensive standards competition when it was first

introduced by Sony in 1981. Before its introduction, the 5.25-inch FDD had been widely

adopted in the microcomputer market. Soon after its introduction, three other new

incompatible disk drives (i.e., 3-inch, 3.25-inch, and 4-inch) were also introduced in the









FDD market. Hence, the 3.5-inch FDD market has the main characteristics we discussed

above.

To empirically examine the proposed supporting-firm-base effect and the stage-

dependent competitive interactions, in this paper, we (1) model consumer utility as a

function of both the installed-user-base and the supporting-firm base; (2) incorporate an

equation of the number of supporting firms in our structural equations; and (3) model

stage-dependent competitive interactions by introducing a time-varying structure to the

estimation of conduct parameters. Three different approaches ("menu approach,"

"weighted profit approach," and "conjectural variation") are used complementarily in

testing and validating the stage-dependent competitive structure.

Our empirical investigation of the floppy disc drive market reveals several

interesting findings. First, we find that consumer valuations are not only positively

related to the installed base, as suggested by previous empirical studies, but are also

positively related to the number of manufactures that adopt the same standard. In other

words, consumers care about both the installed-user-base and the supporting-firm-base.

As the value of a product increases with both, this result highlights the importance of

building up the base of firms adopting the same technological standards, thus the

importance of competitive interaction in intra-standard competition.

Second, by modeling the interrelationship of the number of supporting firms and

the prices of innovating and imitating firms at previous period, we are able to identify

factors that affect the supporting-firm-base. Our results show that the number of

supporting firms is positively influenced by the mean price level of imitating firms in a

previous period, This finding is important because it implies that competitive behavior









discourages, but cooperative behavior encourages, market entry and suggests that the

supporting-firm-base effect has an opposite impact compared with the installed-user-base

that encourages competitive behavior.

Third, our results show that installed-user-base motivates competitive behavior and

supporting-firm-base motivates cooperative behavior, the competitive structure of the

floppy disk drive market changed over the period studied. Firms behaved collectively in

the early stage, competitively in the middle stage, and collectively again in the late stage

of the product life cycle. Such a stage-dependent competition pattern has never been

suggested on either theoretical or empirical grounds, but is certainly plausible for markets

where firms supporting the same standard are both rivals and allies. As our results show,

consumers' willingness to pay in markets with network effects depends on BOTH the

size of the installed base and the number of manufactures adopting the same standard.

The latter can motivate the innovating firm to collaborate rather than aggressively

compete (on price) with imitating firms at the early and later stages, though for different

reasons. The early collaborative behavior helps attract more manufacturers to adopt a

new standard when the new standard is still struggling for market acceptance, and the late

collaborative behavior discourages manufacturers from abandoning the existing standard

when the next generation of technology appears. As a result, firms are more likely to

cooperate in the early and late stages, and to compete on price in the middle stage.

The rest of this study is organized as follows. We introduce the baseline model in

section 2.2. We discuss our data and present the empirical results based on the baseline

model in section 2.3. We then introduce the modified model in section 2.4 and report the









empirical results and validation results in section 2.5. Finally, we discuss conclusions and

future research in Section 2.6.

2.2 The Baseline Model

This section defines our baseline model. The baseline model has two

characteristics. First, it adopts the specification for consumer behaviors as suggested in

the literature of network effects (e.g. Nair, Chintaguta, and Dube 2004), which considers

the impact of installed-user base but not the impact of supporting-firm base on consumer

utility. Second, as many studies of competitive interactions (e.g. Sudir 2001a) it assumes

a constant competitive structure. We modify the baseline model in Section 4 by (1)

incorporating supporting-firm base into functions of consumer utility and conjectural

parameters and (2) assuming a stage-dependent competition pattern.

2.2.1 Consumer Behavior

Following the literature of network effect, we assume that consumers make their

purchase decisions by maximizing their utilities and consumer utility is affected by

installed-user base and product attributes. Letj denote firms, where j= 1 denotes the

innovator and j= 2 denotes imitator.1 The utility of consumer i at time t for firm j's

product is defined as

"ut = 8o + )XJktk -a + Pt + 71-1 + t +,jt = "jt +,t (2.1)
k

where xJ, and p, are the kth attribute and price for product j at time t. N, 1 is the

cumulative sales at time t -1 (i.e., a measurement of consumer installed-base). J,, is the

unobserved characteristic for product j at time t, uj, is the average utility across


1 In real markets, there may be more than one imitator. Following the literature (Suslow (1986) and Putsis and Dhar (1999)), firm 2 in
the duopoly model (1) can be interpreted as a representative imitator.






14


consumers for product j at time t, and y is the random term across product j and

consumer i at time t. Following the literature (e.g. Sudir 2001b), we assume that s, has

double exponential distribution and is independent, identical across products and

consumers.

We allow consumers to have an option of purchasing outside product j= 0, and

normalize the utility of the outside products to zero across times. Then the market share

for product j at time t is given by

exp(udt)
,t = exp (t) (2.2)
1+ exp( u)


Similarly, the market share for outside product j = 0 at time t is given by


Sot = +exp (2.3)
I+ + exp (u.)


where si and so, denote the market share of product j and outside product = 0.

From the market share equations (i.e., (2.2) and (2.3)), we can derive

In(sj, / s,) = jt,,. (2.4)

Since the demand D, for product j at time t can be expressed by Dj = s jMt, where M,

is the market potential at time t, the demand equations can hence be written as

ln(Djt)= In(Dot) + ui, j 0. (2.5)

2.2.2 Firm Behavior

Firms make price decisions by maximizing their profits. The firm j 's profit at time

t is given by

7rt = (Pt ct )D1t, j = 1,2. (2.6)









where pit, c,, are firm j's price and marginal cost at time t, respectively.

Marginal Cost. Firms' marginal costs are defined by

cJt = zgzjgt+ jt (2.7)
g

where ZJgt are the factors that influence firm's marginal cost. Note that, Zjgt includes

product attributes xjkt and other influential factors such as time trend. -, is the

unobservable cost factors for product j.

Strategic Interactions. Two approaches, the conjectural variation approach (CV),

and the menu approach, have been used frequently in the empirical studies of firm

competitive interactions (e.g., Kadiyali 1996, Vilcassim, Kadiyali and Chintagunta 1999).

The CV approach examines the strategic interactions by estimating a set of parameters

and inferring the competitive structure from the signs of the set of parameters, while the

menu approach does so by separately estimating alternative models and inferring the

competitive structure from the selected best-fitted model. It is common for empirical

researchers to use only one approach to estimate the competitive interactions because

different approaches often lead to consistent results given appropriate model specification

(e.g. Bresnahan 1989). Since one of our objectives is to identify more appropriate model

specification (e.g., examine if supporting-firm-base should be incorporated into the

consumer utility function and if state-dependent competitive structure is more appropriate

than a fixed competitive structure), in this paper, we use both approaches

complementarily.

The CV approach (pioneered by Iwata 1974) captures the competition structure

simply by a constant conjectural variation parameter. In the CV approach, a firm is









assumed to have conjectures about how its competitors will react to its price changes and

incorporate the conjectures into its price decisions (Kadiyali et al. 2001). In the baseline

model, the first-order condition for firm j is


alrjt aD t D t
=D+ +(Pt-ct+ ) )-- =0, j=1,2 (2.8)
SDpt pjt 1=12 pclt


where 2 = pJt /Pp,, (i= 1,2, i j) are called conjectural variation parameters (i.e.,

a1, = 9p2t /9p1 and 2 = p9l /0p2t ). The conjectural variation parameters thus tell us how a

firm reacts to its competitor's price changes and therefore indicate the competition

structure. For example, = 22 = 0 indicates Bertrand competition. When # 0, but

12 = 0, we infer that the innovator is leader and the imitator is follower. Symmetrically,

when k2 # 0, but all 0 = 0, we infer that the imitator is leader and the innovator is

follower. >,22 >0 implies a cooperative market structure (Kadiyali et al. 2001).

Different from CV approach, in menu approach, the competitive structure of an

underlying market is inferred by estimating the alternative models (i.e., models with

different competitive structure) separately and selecting the best-fitted model. We

consider four possible competitive structures: Bertrand, Innovator-Led leader-follower,

Imitator-Led leader-follower, and Cooperation. Under the Bertrand competition

assumption, each firm chooses its price assuming that its competitor does not react to its

price changes, so the first order condition for firm j is

S=Dt -c, =0, j=,2. (2.9)
ajt = + ( c) jt = j = 1,









Under the innovator-led leader-follower assumption, innovator anticipates

imitator's reaction to its price and incorporates it in its price decision. Therefore, the first

order condition for innovator is

D1 = DI + (p t) D, l DIt P2t 0. (2.10)
pilt Lpt 8a p 2t pi1t 1

In contrast, the imitator makes its price decision assuming that innovator does not

react to its price. That is, the first order condition for imitators is

D2, +(2t -c2) =0. (2.11)


Taking derivative of imitator's first order condition (2.11) with respect to pit, we

derive the expression of p2t /ait and substitute it in the first order condition of innovator

(2.10). Symmetrically, we can derive the first order conditions for innovator and imitators

under imitator-led leader-follower assumption as

t(Plt t), =o. (2.12)
'-, ', '/, p


-- D2t +(P2t -C2t) D2t + D2t Opt 0. (2.13)
OP2t L P2t P pIt OP2t

Similarly, we can derive the expression of pit /SP2t in first order condition (2.13)

by taking derivative of first order condition (2.12) with respect to p2t and substituting

into (2.13).

Finally, under the cooperation assumption, both firms act as if they were monopoly

selling the differentiated products. In other words, they maximize a total profit of it own

and its competitor when making their price decisions. The total profit is

n, = (pi, -ct)D, The first order condition for firm j is
J










=, Dt + (Pt -clt) Dt = 0, j= 1,2. (2.14)
apt 1=1,2 apt
I#i

In the next section, we empirically estimate the baseline model using both CV and

menu approach. We show that two approaches lead to different inferred competitive

structures.

In Section 4, we modify the baseline model by incorporating a new variable,

supporting-firm-base, into the baseline model. We also allow conjectural variation

parameters to be timing-varying. We then use the time-varying CV approach to examine

the modified model. Our results show a significant effect of supporting-firm base and

suggest that the stage-dependent competitive structure provides a better fit than the

constant competitive structure. Finally, to validate our key results, we apply menu

approach to the modified model and perform stage-based estimation. The results show

that the two approaches, CV and menu approach, lead to consistent results when the

modified model and stage-dependent competitive structure are used.

2.3 Empirical Analysis of the Baseline Model

We estimate the baseline model using data of the 3.5-inch floppy disk drive market

from 1983 to 1998.

2.3.1 The 3.5-inch Floppy Disk Drive Market

A floppy disk is a data storage device that comprises a circular piece of thin,

flexible magnetic medium encased in a square or rectangular plastic wallet.2 Floppy disks

were ubiquitously used to distribute software, transfer data between computers, and

create small backups in the 1980s and 1990s. Sony introduced the 3.5-inch floppy disk


2 From http://en.wikipedia.org/wiki/Floppy disk.









drive in 1981. Before its introduction, there were two main floppy disk drives, 8-inch and

5.25-inch disk drives. The 8-inch floppy disk drives, the first floppy disk drive

introduced in 1967 by IBM, were mainly used in mainframe computers and

minicomputers. The 5.25-inch disk drives, introduced by Shugart in 1976, were mainly

used in minicomputers and microcomputers (PCs).

Floppy disks are read and written by a floppy disk drive (FDD). A floppy disk

drive basically has four principal components: read-write heads, which are located on the

both sides of a disk and move together on the same assembly; a drive motor, which is a

very small spindle motor driving the rotation of the disks; a stepper motor that makes a

precise number of stepped revolutions to move the read-write head assembly to the

proper disk track position; and electronic circuitry, which handle the data read from or

written to the disk and control the stepper-motor control circuits used to mover the read-

write heads to each track, as well as the movement of the read-write heads toward the

disk surface.

With the increasing popularity of PCs in 1980s, small size, easy to carry and

display were becoming the major concerns of computer manufacturers and hence disk

drive manufacturers. The 5.25-inch floppy disk drives were clearly becoming a limitation

to the improvements of microcomputer size. At this time, Sony introduced its 3.5-inch

disk drive.

Sony uses a single-crystal ferrite head in its 3.5-inch floppy disk drive instead of

the poly-crystal ferrite head used by 8-inch and 5.25-inch disk drive. The single-crystal

ferrite head is lighter, more stable and more reliable than the poly-crystal ferrite head.

With this stable and reliable read-write head, Sony was able to put the same number of 80









tracks as 5.25-inch disks on much smaller 3.5-inch disks and keep its 3.5-inch disk drive

with equal or even better reliability than 5.25-inch and 8-inch disk drive.3 As a result, the

disk density was increased from 48 tracks per inch (TPI) to 135 TPI, and disk size was

therefore largely decreased.

Along with the introduction of Sony's 3.5-inch disk drive, there were two other

microfloppy disk drives contending for market standard: the 3-inch drive, introduced by

Matsushita Electric Industrial, Hitachi and Hitachi/Maxell; and the 3.25-inch disk drive,

introduced by Tabor and Seagate Technology. Besides these two disk drives, IBM, as an

industry dominant manufacturer, was also expected to introduce a 3.9-inch disk drive on

an unknown schedule. These floppy disk drives with different sizes were incompatible

because disks could only be loaded in the correct size of disk drive.

As Disk Trend Report recorded, Sony was willing to work with U.S. manufacturers

to build its technology as the industry standard. As a result, as many as 27 other

manufacturers adopted Sony's technology and produced 3.5-inch FDD in the same

market. Sony's 3.5-inch disk drive eventually became the industry standard.

To sum up, the 3.5 inch FDD market has the main characteristics in which we are

interested. First, it is subject to network effects. Second, the innovator, Sony, faced

competition from imitating firms adopting its technological standard. Third, the 3.5-inch

FDD encountered intensive standards competition.


See Mini Micro Systems, April, 1981.









2.3.2 Data

Our annual data on product attributes, sales and market share in the 3.5-inch floppy

disk drive market are collected from Disk TrendReport,4 a leading annual market

research publication in the disk drive industry. It has been considered as a highly reliable

and complete source of disk drive industry data (Josh Lerner 1997, Christensen et al

1998).

We compose all imitators into one imitator, and use a composite imitator as

representative imitator in estimation as did in Suslow (1986) and Putsis and Dhar (1999).

We denote the innovator and imitator hereafter as j, and j2, respectively. Data on the

imitator's product attributes are accordingly computed by averaging all imitators' product

attributes at each year. Imitator's sales and market share are then derived by subtracting

innovator's sales and market share from the sales and market share of 3.5-inch floppy

disk drive market. Data on the annual average price of innovator and imitator's products

at each year are computed by dividing their annual dollar sales by annual sales volume.

Disk TrendReport records the attributes for each disk drive that is introduced by

each manufacturer in each year. They are listed according to three categories: (i) disk

capacity, which includes total capacity, density per track, data surface per spindle, and

tracks per surface; (ii) drive performance, which includes the time spent from track to

track, settling time, average rotational delay time and data transfer rate; and (iii) disk

drive features, which includes height, width and depth of each disk drive. By analyzing

the attributes in each category, we found that the attributes in first and second categories

are highly correlated. To reduce the problems of multicollinearity, we conducted step-


We would like to thank James Porter for his generous offer ofthe series books of Disk Trend Report.
We would like to thank James Porter for his generous offer of the series books of Disk Trend Report.










wise regression to choose the most influential attribute variables. Our regression analyses

identify five attributes that contribute to the most of variances of price and sales: disk

density (Den), track to track time (TT), height (H), width (W) and depth (D). These

attributes are used in our utility function (2.1).

Following literature on network effects (e.g. Brynjolfsson and kemerer 1996), the

consumer installed-base in demand equation that captures the network effects is

measured as the cumulative sales of whole 3.5-inch floppy disk drive market at previous

period. We also measure the supporting firm base using the total number of supporting

firms at previous period. The potential market size (AM,) in the demand equation is

measured as the total of the floppy disk drive market sales at each time from 1983-1998.

The total of the floppy disk drive market sales is measured by adding together all sales

from different sizes of floppy disk drives that were introduced in FDD market. For

example, it can be sales from 8-inch, 5.25-inch, 3.5-inch, 4-inch, 3.25-inch, 3-inch and

high density FDD. We use all of the above five product attributes and one time trend

variable in marginal cost equation. Table 2-1 provides the descriptive statistics of our

data. From Table 2-1, we can see that the difference between innovator and imitator's

product mainly lies in disk density. And innovator charges higher average price than

imitator.

Table 2-1. Descriptive Statistics
Innovator Imitator t-statistic of Mean
Mean Std. Mean Std. Differences
Sales (Million) 6.22 4.17 33.94 33.01
Average Price 52.82 34.87 48.07 27.68
Density 135 0 131.73 5.87 2.23**
Track to Track 4.23 2.73 3.74 0.58 0.70
Height 1.06 0.34 1.01 0.23 0.49
Width 3.96 0.048 4.05 0.24 1.54
Depth 5.64 0.25 5.69 0.45 -0.36
Installed-base (Million) 142 170.20
Note: **: p<0.05.










2.3.3 Identification

We assume that the product attributes xjk are exogenous and orthogonal to the

error terms (4, and ,, ); price and sales are endogenous and correlated with the error

terms (J, and ,,J). This identification assumption is reasonable and commonly used in

the simultaneous equation estimation (e.g. Sudir 200 b). We use product attributes, their

squares and the products of innovator's attribute and imitator's attributes as instruments.

Since innovator's density remained the same over 16 years, we exclude it as one

instrument variable. So we have a total of 23 instruments in the demand equation: ten

exogenous variables (a constant, four innovator's attributes, and five imitator's

attributes), nine exogenous variables' square, and four products of innovator's and

imitator's attributes. We also apply these instruments to the two competitive interaction

equations. In the baseline model, we need to estimate 9 parameters in the demand

equation (constant, coefficients of five product attributes, price, installed-base and

supporting firm-base), 7 parameters in the marginal cost equation (constant, coefficients

of five product attributes and time trend), and 2 conjectural variation parameters in the

time varying first order conditions if using conjectural variation approach. Therefore, our

baseline model is identifiable.

2.3.4 Empirical Results of Baseline Model

In the conjectural variation approach, we simultaneously estimate the demand

equation (2.5), the marginal cost equation (2.7) and the competitive interaction equation

(2.8). The competitive structure is inferred by examining the signs of the two conjectural

variation parameters, ) and ), as discussed in Section 2.









In the menu approach, we estimate the structural equations under four alternative

assumptions of competitive structures separately, i.e., for each assumed competitive

structure, we simultaneously estimate the demand equation (2.5), the marginal cost

equation (2.7), and one of corresponding competitive interaction equations given in (2.9)-

(2.14). We then infer the competitive structure by selecting the best-fitted model based

on the results of the Vuong test.5 The three-stage least square (3SLS) method is used in

the estimation of both approaches. Table 2-2 provides the results of the two approaches.

As shown in Table 2-2, the results of CV approach show that both conjectural

variation parameters, \ and ., are not significantly different from zero, indicating a

Bertrand competition structure. The results of menu approach show that the cooperation

model has a maximum log-likelihood value (-15.09), while the log-likelihood values for

the other three alternative competitive structures (Bertrand, Innovator-Led leader-

follower, and Imitator-Led leader-follower) are -24.00, -27.97, and -28.75, respectively.

The results of the Vuong test statistics (see Table 2-2) indicate that the cooperation model

is significantly better than the other three alternatives (p<0.01 for all three comparisons).

Hence, our empirical analysis of the baseline model under the assumption of a

constant competitive structure leads to different conclusions about the nature of

competitive interaction in the studied market-Bertrand based on conjectural variation



5 The Vuong test statistic is defined as V In (p q) where In(f) and In(g) are the log-

likelihood values of the two non-nested models, and p and q are the number of parameters estimated in
each model, respectively. V is distributed as N(0,1) If V > 0 and V > ritica allue then the model
corresponding to f is rejected in favor of the model corresponding to g, and vice-versa (Kadiyali et al.
2000).










approach but Cooperation based on menu approach. This indicates that the competitive

structures derived from the two approaches are inconsistent when the baseline model is

used.

Table 2-2. Baseline Model: Estimated Competitive Structure
Conjectural Variation Approach
Conjectural Variation Parameter
S22 The log-likelihood
n NS 0 2NS


Estimation .-33.22
(1.24) (0.59)
Estimated Bertrand
Bertrand
Competitive Structure
Menu Approach
Innovator-Led Imitator-Led
Alternative Models Bertrand Invar-Led Itator-Led Cooperation
Leader-Follower Leader-Follower
Log-likelihood -24.00* -27.97* -28.75* -15.09
(Vuong Statistics) (2.23) (3.22) (3.42) (-)
Estimated
Estimated Cooperation
Competitive Structure Coope
Note:
: p<0.01; N: p>0.1
For conjectural variation approach, the numbers in parentheses are t-statistics. For menu approach, the
numbers in parentheses are Vuong statistics. The critical value of Vuong statistics at p=0.05 is 1.64. The
Vuong statistics are significant atp<0.01 in all cases (See footnote 7 for discussion of Vuong test).

The inconsistency found in Table 2-2 can be the results of model misspecification.

As we argued, network effects create two types of interdependences: (1) the

interdependence among consumers (i.e., the value of a product to consumers is affected

by the number of consumers bought the same product), and (2) the interdependence

between consumers and firms (i.e., the value of a product to consumers is affected by the

number of the firms supporting the same technological standard). However, as the

previous empirical studies of network effects, our baseline model only considers the

installed-use-base effect, but not the supporting-firm-base effect.

In the next section, we modify our baseline model by incorporating the supporting-

firm-base effect into consumer utility function, and modeling the competitive structure to

be a function of (1) installed-user-base and supporting-firm-base, and (2) time trend.









2.4 Modified Model and Empirical Results

2.4.1 Modified Model

Demand. We modify the baseline model by incorporating a new variable,

supporting-firm-base, into consumer behavior. Let H, denote the number of firms

supporting the innovator's technological standard at time t (i.e., a measurement of

supporting-firm-base). We formally incorporate the supporting-firm-base into the

consumers' utility model:

u"t = o + xjktfk --Pjt +yNt 1- + Nt 71-1 1-H,- + -tT + 2Ht T+ ,+t = u jt + t (2.15)
k

Equation (2.15) assumes that consumer utility is affected not only by the installed-

user-base, as assumed in the existing literature and in our baseline model, but also by the

supporting-firm-base. Furthermore, since both the installed-user-base effect and the

supporting-firm-base effect may change with time, we include two interaction terms,

(N, 1T) and (H, T), in Equation (2.15) to capture the possible interaction effect, where T

donate time trend.

To examine how the number of supporting firms may be affected by price

competition, we also models the number of supporting firms as a function of imitator's

price and the number of supporting firms at the previous period:

Ht = (0Ht- + q2Ht-21 P2t-1 +1 (2.16)

The quadratic term in Equation (2.16) is intended to capture the saturation effect.

Strategic Interactions. To test the possible stage-dependent competitive

interactions, we relax the assumption of a fixed competitive structure by allowing the

conjectural variation parameters in equation (2.8) to be timing-varying. Specifically, we

define the conjectural parameters, A1 and 2 to be two types of functions. We first define









the conjectural variation parameters to be a function of installed-user-base and

supporting-firm-base, as shown in Equation (2.17):

jt = 2J +^NIt- +2 2Ht 1, j = 1,2, (2.17)

and second to be a function of time trend T, as shown in Equation (2.18):

Jt = 2J + ,T + j22, j = 1,2 (2.18)

Equation (2.17) allows us to examine how the installed-user-base and supporting-firm-

base directly influences the competitive structure directly, and whether the competitive

structure is stage-dependent, while equation (2.18) provides us an overall pattern of the

stage-dependent competitive structure over time.

2.4.2 Empirical Results: Time-Varying CV Approach

We first estimate the modified model using the time-varying conjectural variation

approach by modeling the competitive structure to be function of installed-user-base and

supporting-firm-base as well as function of time trend, and then validate the key findings

using stage-based menu approach. The structural equations for CV approach include the

demand equation (2.5), the marginal cost equation (2.7), the supporting-firm-base

equation (2.16), and the strategic interaction equations (2.8). Note that the uJ, in (2.5) is

now derived from (2.15) rather than from (2.1) as in Section 2. Also, 2,, in (2.8) is now a

function of installed-user-base and supporting-firm-base as defined in (2.17) or function

of time trend as defined in (2.18) rather than a constant as in Section 2. To derive the

best fitted model, we incorporate four time dummy variables in equation (2.17). The four

time dummy variables are defined as dl = 1 if the time period is at years 1 to 4, and

otherwise 0; d2 = 1 if the time period is at years 5 to 8 and otherwise 0; d3 = if the time









period is at years 9 to 12 and otherwise 0; d4 = 1 if the time period is at years 13 to 16 and

otherwise 0. The estimated function of conjectural variation parameters is then written as

~, =J 20 +2 J(N*dl)+-j2(N*d2)+2j3(N*d3)+2j4(N*d4) (2.17')
+ j5 (H *dl) + 6(H*d2) +j7 (H *d3) +2 (H *d4)

where j=1 and 2. The results of the modified model using time-varying CV approach are

given in Tables 2.3-2.4.

Impact of Supporting-Firm-Base on Consumers. As shown in Table 2-3 (see

the results of Demand Equation on the top of Table 2-3), the installed base is positively

related with consumer valuation (p<0.01), suggesting a positive installed-user-base effect.

This finding is consistent with previous studies (e.g. Brynjolfsson and kemerer 1996).

More importantly, our result shows that the number of the supporting firms is also

positively related to consumer valuation (p<0.01), suggesting a positive supporting-firm-

base effect. Table 2-3 also shows that the parameters of both interaction terms in the

demand equation are significant, suggesting that both the positive installed-user-base

effect and the positive supporting-firm-base effect vary across time.

Impact of Price on Supporting-Firm-Base. Table 2-3 shows that the number of

supporting firms is positively related to the price of imitators (see the results of

Supporting-Firm-Base Equation in the middle of Table 2-3). This result combined with

the positive effect of supporting-firm-base on consumer valuation discussed above

suggests that the supporting-firm-base effect reduce firms incentive to aggressively

compete on price, because a high price encourage more firms to adopt the same

technological standard and in turn increase the value of the product to consumers.










Table 2-3. Revised Model 1, Time-Varying CV Approach
Demand and Marginal Cost Equations
Demand
-61.78**
Intercept (2.54)a

Den 14.14**
Den
(2.91)
T -0.36
(-0.77)
2.31
Height .3
(1.36)
Width 28.31**
Width
(5.36)
Deh -25.28***
Depth
(-5.91)
Time Trend


Marginal Cost
3877.86***
(3.89)
-278.40
(-1.56)
-46.21**
(-2.42)
368.85*
(3.36)
-1919.24*
(-2.30)
88.62
(0.73)
6.22*
(1.91)


Price

Installed-base

Installed-base* Time

Firm-base

Firm-base Time


Firm-Base at Previous Period

Square of Firm-Base at
Previous Period
Imitating Firm's Price at
Previous Period



2,0 (constant)


A1 (at first period)

K22 (at second period)

2,3 (at third period)

2.4 ( at fourth period)


-0.075.
(-5.80)
0.62E-07**
(5.03)
-0.38E-08**
(-5.15)
0.17*
(4.28)
-0.032***
(-5.53)
Supporting-Firm-Base Equation
1.23***
(5.27)
-0.018*
(-1.70)
0.055***
(3.51)
Conjectural Variation Parameters
Innovator
0.57
(1.52)


User-Base
-0.13E-05
(-2.00)
-0.46E-07
(-3.94)
-0.35E-08
(-3.00)
-0.14E-08
(-1.91)


Firm-Base
0.24
(2.02)
0.13
(4.67)
0.04
(1.67)
0.03
(1.62)


Note: The number in parentheses is t statistic. :p<0.01. :p<0.05.


Imitator
0.22
(0.28)


User-Base
-0.33E-04
(-0.33)
-0.71E-07
(-2.67)
-0.94E-08
(-3.91)
-0.29E-08
(-1.69)


Firm-Base
5.51
(0.33)
0.33
(4.10)
0.13
(2.84)
0.07
(1.38)









Impacts of User-Base and Firm-Base on Competitive Structure. The estimation

of the impacts of user-base and firm-base on competitive structure are given in the

bottom of Table 2-3 (see Time-Varying CV Parameter Estimation). As shown in Table 2-

3, for both innovating and imitating firms, the coefficient of user-base across four periods

are all significantly negative, while the coefficient of firm-base across four periods are all

significantly positive. The different signs of the impacts of user-base and firm-base imply

that installed-user-base motivates competitive behavior, while the supporting-firm-base

motivates cooperative behavior.

Stage-Dependent Competitive Interaction. The estimations of time-varying CV

parameters where the conjectural variation parameter is defined to be a function of time

trend are given in Table 2-4. As show in Table 2-4, for both innovating and imitating

firms, the constant, 2j0, is positive (p<0.01), the linear term, 2J,, is negative (p<0.01),

and the quadratic term, 2J, is positive (p<0.01). These results suggest that (1) firms

behave cooperatively at the beginning of the new product introduction, and (2) the

competitive structure may vary across different stages of product life cycle. To

graphically illustrate the pattern of competitive behavior in the market examined, Figure

1 shows how the conjecture variation parameters, 2J, (/=1,2), given in (2.18) change with

time when the estimated time-varying conjecture variation parameters given in Table 2-4

are used.










Table 2-4. Revised Model 2, Time-Varying CV Approach
Demand and Marginal Cost Equations
Demand


-54.16
(-5.23)a
12.80
(5.81)
1.75
(4.56)
3.21
(4.64)
27.57***
(6.85)
-26.90
(-7.37)


Intercept

Den

TT

Height

Width

Depth

Time Trend


Price

Installed-base

Installed-base* Time

Firm-base

Firm-base Time


Firm-Base at Previous Period
Square of Firm-Base at
Previous Period
Imitating Firm's Price at
Previous Period


Log-likelihood value


-0.051
(-4.21)
0.44E-07
(4.86)
-0.28E-08
(-5.22)
0.044*
(2.34)
-0.016
(-3.45)
Supporting-Firm-Base Equation
1.017
(11.99)
-0.020
(-5.35)
0.14
(14.60)
Conjectural Variation Parameters
Innovator


4.56
(7.55)
-0.62
(-5.36)
0.023
(3.96)
Comparison: Modified vs. Baseline Model
Modified model Baseline model
-13.88 -33.22


Marginal Cost
-20413.3
(-1.20)
4097.15
(1.19)
-5.35
(-0.14)
-198.08
(-2.39)
-1073.70**
(-2.18)
1031.99
(2.56)
-3.25
(-1.91)


Imitator
11.18
(5.36)
-1.80
(-4.90)
0.075
(4.67)

Log-likelihood ratio
38.68


Note:
The number in parentheses is t statistic. : p<0.01. : p<0.05
The loglikelihood ratio has %2 distribution, the critical value %2 =18.30 at p<0.01 with d.f.=10.










As shown in Figure 2-1, both conjecture variation parameters, 4~ and ,2, follow a

quadratic curve. Both parameters are positive in the early years, indicating a cooperative

market structure in the early stage of standard development. Then both 4 and


2, decrease to near zero at around year 12, indicating a competitive market structure in

the middle stage. In the late years, both ), and )2 increase to be positive again, indicating

that the market moved toward a cooperative structure again in the late stage. Hence,

Figure 1 shows a very interesting three-stage pattern of competitive structure in floppy

disk drive market: cooperation in the early stage, competition in the middle stage, and

cooperation again in the later stage.





12
10 -
10

d 8
E
S6


2


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Time
-*-lamdal ---lamda2


Figure 2-1. The Changes of Conjectural Variation Parameters with Time

To evaluate the superiority of the stage-dependent competitive structure model over

the constant competitive structure model, we test if the former fits data better than the

latter by comparing the log-likelihood values of the conjectural variation models given in









Table 2 (-33.22) with that given in Table 3 (-13.88). The log-likelihood ratio test6

suggests a significant superiority of the stage-dependent competitive model (%2 = 18.30,

p<0.01, d.f. =10).

To sum up, the empirical analysis of the modified model using time-varying

conjecture variation approach leads to three key findings: (1) a positive effect of

supporting-firm-base on consumer valuation, (2) a positive effect of price on the number

of supporting firms, and (3) a three-stage competitive interaction pattern: cooperation-

competition-cooperation. To further validate these key findings, we apply a stage-based

menu approach to the modified model.

2.4.3 Validation: Stage-based Menu Approach

Different from CV approach that captures the stage-dependent competitive

interaction via a set of time-varying conjecture variation parameters, when stage-based

menu approach is used to estimate stage-dependent competitive structure, one needs to

first decide how stages are divided and then apply menu approach to data in each stages.

Based on results from CV approach (see Figure 2-1), we divided the data set into three

stages: an early stage (year 1-5), a middle stage (year 5-12), and a later stage (year 10-

16).7 For each stage, menu approach is used to estimate all four alternative models (i.e.,

Bertrand, Innovator-Led leader-follower, Imitator-Led leader-follower, and Cooperation),


6 We use the log-likelihood ratio test to compare the goodness of fit between these two nested models. The
log-likelihood ratio is 38.68, which is larger than the critical value of X2 = 18.30 at p<0.01 with freedom
degree= 10, showing that our time-varying conjectural variation model is significantly better than the
conjectural variation model with single competitive structure.

7 Duo to the limited data available, these stages partially overlap with each other so that there are sufficient
data points within each stage to allow models to be identifiable and to use Vuong test statistics to select the
best-fitted models. While different possible cutoff times were considered, our final definition of stages is
consistent with the curves observed in Figure 1 and Vuong test statistics are significant in selecting the
best-fitted models.









and the best-fitted model in each stage is selected using Vuong test statistics. The results

of model selecting test for each stages are given in the top part of Table 2-5. As shown in

Table 2-5, according to the Vuong statistical test, the best-fitted model is Cooperation in

the early stage, Innovator-Led Leader-follower in the middle stage, and Cooperation

again in the later stage. All other alternative models are rejected. Hence, menu approach

leads to the same three-stage pattern of competitive interaction as CV approach (i.e.,

cooperation-competition-cooperation). In other words, when we use the baseline model

and assume a constant competitive structure, the two different approaches lead to

inconsistent conclusions (see Table 2-2). However, when we use the modified model and

allow the competitive structure to be stage-dependent, the two approaches lead to

consistent conclusions.

Table 2-5 also shows the results of parameter estimations of the best-fitted models

in each stage. Note that the stage-based menu approach allows not only competitive

structure to vary across stages, but also parameters in all structure equations vary across

stages. As a result, the estimation results using menu approach reveal more details of the

market dynamics, which will be discussed in more details later. Nevertheless, as in the

estimation of CV approach, the results of demand model by menu approach show a

positive effect of the number of supporting firm on consumer valuation (in the early and

later stage), and the results of supporting firm model by menu approach show a positive

relationship between the number of supporting firms and price. Hence, our three key

findings derived in CV approach are also supported by the menu approach.










Table 2-5. Revised Model 3, Stage-Based Menu Approach
Model Selection Test
Innovator-Led Imitator-Led Best
Best-fitted
Stage Bertrand Leader- Leader- Cooperation model
model
Follower Follower
Early 32.52* 32.66* 32.57* 40.07
(Year 1-5) (3.37) (3.31) (3.35) (---)
Innovator-Led
Middle 21.58* 45.03 18.72* 35.84* n r-
Leader-
(Year 5-12) (8.29) (---) (9.30) (3.25) ollower
Follower
Late 3.12* 19.56** 3.75* 50.04
(Year 10-16) (17.71) (11.50) (17.47) (---)
Parameter Estimation: Demand Model (The Best-fitted Model)
Parameter Early Stage Middle Stage Late Stage
(Cooperation) (Innovator-Led (Coorperation)
Leader-Follower)
-26.18 -258.92 527.20
Intercept (-0.99) (-1.94) (1.72)
2.05 52.49 -98.82
(0.34) (1.98) (-1.57)
2.71 1.06** 0.98
(1.45) (2.39) (0.58)
-2.50 -1.67** 17.64**
g (-0.50) (-1.87) (7.32)
15.72 38.45 24.76
(0.58) (9.81) (5.08)
-7.19 -29.85 -43.01
h (-0.35) (-10.22) (-5.77)
-0.024 -0.0098 -0.081
Pnrce
(-1.40) (-1.65) (-2.40)
-0.92E-07 -0.29E-08 0.35E-08
Installed-base
(-1.04) (-1.54) (1.82)
0.19 -0.032 0.071*
Firm-base
(2.88) (-0.92) (2.42)
Parameter Estimation: Supporting-Firm-Base Model
Parameter Early Stage Middle Stage Late Stage
Firm-Base at Previous 0.74 1.18 0.83
Period (1.28) (10.18) (7.19)
Square of Firm-Base at -0.0081 -0.018" -0.038*
Previous Period (-0.42) (-4.91) (-3.85)
Imitating Firm's Price 0.14 0.061 0.39
at Previous Period (4.03) (1.87) (4.48)
Note:
Sp<0.01, p<0.05.
For model selection test, the numbers in parentheses are Vuong statistics. The critical value of Vuong
statistics atp=0.05 is 1.64.
For parameter estimations, the numbers in parentheses are t-statistics.



2.4.4 Discussion of Key Findings

First, previous empirical studies have shown a positive relationship between

installed-user-base and consumer valuation (e.g. Nair, Chintaguta, and Dube 2004). Such









a positive relationship has been explained by a positive network effect: the larger the user

base, the higher the value of the product. By simultaneously examining the installed-

user-base and supporting-firm-base, we find that both can have a positive effect on

consumer adoption decision. More interestingly, our stage-based analyses show that the

relative importance of installed-user-base effect and supporting-firm-base effect varies

across stages of product life cycle. For example, in the early stage of 3.5-inch floppy disk

drive, the consumer valuation is positively affected by the supporting-firm-base but not

by the installed-use-base (see Table 2-5). While this result is not intuitive, it may result

from the high uncertainty faced by consumers in the presence of standards competition.

In the early stage of product introduction, 3.5-inch disk drive faced intense competition

from multiple incompatible technological standards including the old 8-inch and 5.25-

inch floppy disk drives as well as the new 3-inch, 3.25-inch and 4-inch floppy disk

drives. Since different types of drives were not interchangeable, the adoption utility of a

given type of disk drive depends on the outcome of the standards competition. While it is

difficult for early buyers to predict if the new 3.5-inch floppy disk technology would be

widely adopted by the market, the number of firms supporting the Sony's new innovation

in the early stage could be useful information to consumers in evaluating the odds of the

new technology because the larger the number of supporting firms, the more likely that

the new technology will become industry standards. The fact that consume valuation is

affected by the supporting-firm-base but not by the installed-user-base in the early stage

suggests that at the beginning of 3.5-inch disk drive introduction, consumers were more

concerned about the outcome of ongoing standards battle but less about the current user

base of the drive. While conventional wisdom has emphasized the importance of









building consumer base in markets with network effects, our finding highlights the

importance of simultaneously considering both the installed-user-base and supporting-

firm-base. As we will discuss next, the installed-user-base effect and supporting-firm-

base effect can have opposite implications on price competition.

Second, our empirically analyses show a positive relationship between the number

of supporting firms and price. This result implies that the positive supporting-firm-base

effect discussed above reduces firms' incentive to cut price because a high price

encourages more firms to adopt the innovator's technological standards, which in turn

increases consumer valuation. Note that it is commonly agreed that the positive installed-

user-base effect motivates firms compete more aggressively because a low price

encourages more consumers to adopt the new product, which in turn increases consumer

valuation. Take the two effects together, our findings suggest that while both supporting-

firm-base and installed-user-base increase consumer valuation, they have opposite

impacts on firms' competitive behavior: the former motivates cooperative behavior but

the latter motivates competitive behavior. Hence, the nature of firms' strategic

interaction is expected to be jointly determined by the two effects and may change across

different stages of its product life cycle.

Third, our most interesting finding is the three-stage competitive interaction

pattern found in 3.5-inch floppy disk drive market: firms cooperate in the early stage,

compete in the middle stage, and then cooperate again in the late stage. This pattern can

be explained by the relative importance of the two effects (i.e., installed-user-base effect

and supporting-firm-base effect) to consumers in different stages, and the opposite

impacts of the two effects on firms' competitive behavior. In the early stage of 3.5-inch









disk drive introduction, 3.5-inch disk drive faced intense competition from several

competing standards (see Figure 2-2), which drawn consumers' attention toward the

supporting-firm-base rather than the current installed-use-base (see Table 4). Firms

adopted cooperative behavior because cooperation helps but competition obstructs

building up supporting-firm-base.

In 1990 (i.e., 9 years after the introduction of 3.5 floppy disk drive), the sales of

3.5 floppy disk drive have surpassed the sales of all alternative technologies (see Figure

2), indicating that 3.5 floppy disk drive had established its technology standard position.

In the next few years (the middle stage), neither installed-user-base nor supporting-firm-

base have a significant effect on consumer valuation (see Table 4), implying that in this

stage, the product is evaluated largely based on its performance rather than by network

effects. As a result, there would be little incentive for firms to attract more imitators;

rather, they had strong incentive to compete aggressively to achieve as much profit as

possible.

In 1995 (12 years after the introduction of 3.5-inch disk drive), new generation

technology, high-density floppy disk drive, enters the market. In this late stage,

consumers were concerned about both the installed-user-base and supporting-firm-base,

perhaps because both indicate how likely the existing technology would be replaced by

the new generation technology. Since the supporting-firm-base was more important to

consumers than the installed-user-base (see Table 2-5), firms adopted cooperative

behavior again to discourage manufactures from abandoning the existing standard.











120000.0

100000.0

0 80000.0

o 60000.0

40000.0

20000.0

0.0 ..--" , -
1983 1985 1987 1989 1991 1993 1995 1997
Time
-- 3.5 inch 8+5.25 inch High Capacity


Figure 2-2. Sales: 3.5 inch Floppy Disk Drive vs. Competing Products




2.5 Conclusion

In this paper, we use the new empirical industrial organization framework to

examine price competition between innovating and imitating firms. Different from the

existing research of network effects, we consider not only an installed-user-base effect

but also a supporting-firm-base effect on consumer utility. To examine the impact of the

supporting-firm-base on intra-standard competition, we model the interrelationship of the

number of imitating firms with the prices of innovating and imitating firms. To allow our

models to capture the stage-dependent price competition structure, we introduce time-

varying parameters to the conjectural variation approach and validate our estimation

results using menu approach.

Our results provide empirical evidence that markets with network effects generate

both a positive consumer-installed-base effect and a positive supporting-firm-base effect

(on consumer utilities). However, these two effects have opposite impacts on firms'









competitive behaviors. To increase consumer installed-base, firms need to keep price

low, which motivates firms to compete. In contrast, to encourage other firms to adopt the

innovator's technology, firms need to keep price high, which motivates firms to

cooperate. Furthermore, these two opposite effects impact the price competition between

firms differently at different stage. Our results using both approaches suggest that the

innovator and imitators have more incentive to encourage more manufactures to support a

new standard in its early stage and discourage existing manufacturers from abandoning

an established standard in its later stage when facing the challenge of next generation

technology. As a result, firms are more likely to cooperate in the early and later stage,

and compete on price in the middle stage.

This paper makes several contributions to the literature of network effects. First, we

examine an effect overlooked in the literature, the supporting-firm-base effect. Our

results not only provide empirical evidence supporting a positive relationship between the

number of firms adopting the same technological standard and consumer's valuation, but

also suggest that the supporting-firm-base effect works in an opposite direction compared

with the well-known installed-user-base effect in terms of their impacts on intra-standard

competition.

Second, our study provides empirical evidence of a three-stage pattern of

competitive interaction in intra-standard competition. Our results suggest such a pattern

in the floppy drive market over the period studied: firms behaved collectively in the early

stage, competitively in the middle stage, and collectively again in the late stage. Such a

stage-dependent competition pattern is new to the literature on network effects and raises









interesting theoretical and empirical questions for future research on the strategic

implications of stage-dependent intra-standard competition.

Finally, this paper illustrates the use of complementary approaches to evaluate

different structural models and estimate stage-dependent competitive structures. This

approach can be applied to other situations where a constant competitive structure is

questionable.

Overall, this study fills a gap in the network effects literature by providing

empirical evidence on patterns of intra-standard competition. Our findings raise some

important new theoretical and empirical issues for future research on markets with

network effects and standards competition. However, since our study only uses data from

the floppy disk drive industry, the generality of our findings needs to be tested in other

industries with network effects.














CHAPTER 3
DYNAMIC ANALYSIS OF NEW TECHNOLOGY DIFFUSION IN MAREKTS WITH
COMPETING STANDARDS

3.1 Introduction

The diffusion of new technology has been extensively studied since the traditional

diffusion model was published (e.g., Bass 1969, Mahajan and Muller 1979, Norton and

Bass 1987). Recently, the rapid development of information technology and the digital

revolution attract researchers' attention again to the diffusion of new technology,

especially in markets with network effects.

While studies on the adoption of the new technology such as the ATM machine

(Saloner and Shepard 1995), Compact Disc (Gandal, Kende and Rob 2000), DVD player

(Inceogle and Park 2003, Dranove and Gandal 2003), and ACH electronic payments

system (Gowrisankaran and Stavins 2004) demonstrate the importance of the network

effects (i.e. installed-user-base effect) on the new technology diffusion, they all focus on

the new technology diffusion at category-level. There is NO empirical research examines

the standard diffusion and how the competition between incompatible technologies

impacts the new technology diffusion at three levels--category-level, standard-level and

brand-level.

Understanding the new technology diffusion at standard-level and the impacts of

standards competition on the new technology diffusion at three levels is extremely

important because the markets with network effects and standards competition exhibit









certain unique characteristics that do not apply to the traditional markets and have not

been addressed in the literature of new technology diffusion and network effects.

First, in markets with network effects and standards competition, the new

technology diffusion at three-levels is interdependent: the standard diffusion impacts the

market potential for whole category diffusion and the consumer's evaluation for an

individual brand, while the category diffusion impacts also the market potentials for each

standard's diffusion and the brand diffusion comprises of the standards' diffusion.

Studying the interdependence of the new technology diffusion at three-levels requires the

marketing scholars to integrate the category diffusion, standard diffusion and brand

diffusion into a systematic framework. However, most of the traditional diffusion models

(e.g. Bass 1969, Bass, Krishnan and Jain 1994, Horsky 1990) focus only on the category

diffusion without considering the impacts of standards' and brands' diffusion on each

other. Though there is little diffusion research in recent years that has proposed models to

incorporate the competition impact on diffusion (e.g. Norton and Bass 1992, Mahajan

and Muller 1996), there is NO research addressing the standard diffusion in a systematic

framework.

Second, standards competition impacts the new technology diffusion differently

from the competitions among competing brands in the traditional markets. The

differences exist in two aspects: (1) unlike the traditional markets where the competitions

among competing brands may expend the whole category diffusion positively, the

standards competition in markets with network effects may impact the market potential

negatively. This is because the high uncertainty that consumers have on which standard

will dominant market deters consumers from making choice immediately when there is









no single technological standard dominating the market. (2) Standards competition

generates different word of mouth effects on the new technology diffusion at the stage

when the incompatible technologies are competing for market dominance from the stage

when the market is dominated by one of the competing technologies. Recent studies

about the impacts of network effects on new technology diffusion has revealed the

positive impacts of network effects on a particular technology standard (e.g. Saloner and

Shepard 1995, Gandal, Kende and Rob 2000, Park 2003 and Ohashi 2003) or the

negative impacts of network effects on competing standard (e.g. Dranove and Gandal

2003), but there is NO empirical works demonstrating how the standard competition

impacts the new technology diffusion on the market potential and consumers' word of

mouth effects.

Third, literature has suggested (e.g. Katz and Shapiro 1985 and Shapiro and Varian

1999) that consumer expectation is a major factor in consumer decisions about whether

or not to purchase a new technology, and the management of consumer expectation on

network effects is a critical factor in standard competition and hence the technology

diffusion, but no empirical works demonstrate to managers on how consumers form their

expectations on network effects, especially on the future installed-user-base and

supporting-firm-base.

This study sets up a systematic framework by integrating the new technology

diffusion at three levels. Specifically, this study builds a choice-base diffusion model by

incorporating (1) the standard competition---measured as the market share differences

between the competing standards, and (2) two important network effect factors: installed-









user-base and supporting-firm-base of competing standards into a baseline model and a

dynamic model.

Our choice-based diffusion model involves three components: (1) a consumer

utility model that captures the influences of price, product attribute, and network

effects-measured as installed-user- base and supporting-firm-base of two competing

standards, (2) equations that derive the market share of two incompatible platforms based

on the estimated nested-logit model, and (3) equations that derive the dynamic diffusion

at category-level, standard-level and brand-level. Different from the baseline model that

only captures the current network effects---current user-base and firm-base of two

competing standards, our dynamic model captures the consumer expectations for the

future user-base and firm-base of competing technological standards, and thus involves

using a dynamic programming method to estimate the market share equations for each

competing standard and brand.

Our model is estimated using data from the VCR industry. The competition in the

home VCR market is one of the best-known examples of technology standard

competition between two incompatible format Betamax and VHS (Park 2003). Though

the Betamax format was first introduced by Sony in February 1976, VHS format, which

is introduced one and half year later by JVC in 1977 and incompatible with the Betamax

format, was able to surpass the Betamax in the early years of 1980s and eventually

dominated the U.S. VCR market in the late years of 1980s. Examining this market would

provide valuable insights on how the inter-standard competition impacts the new product

diffusion at category-level, standard-level and brand-level.









The estimated parameters reveal several interesting results. First, we find that while

the installed-user-base from the compatible standard increases the consumer's evaluation

to a new product, which is consistent to the literature on network effects, the installed-

user-base from competing technology standard decreases the consumer's evaluation to a

new product, which is new to literature on network effects, and demonstrates the

evidence for one of the important impacts of standard competition on new technology

diffusion.

Second, we find the empirical evidence that standard competition impacts new

technology diffusion complimentarily. This can be seen from two results: (1) the

supporting-firm-base from both of two competing standards---VHS and Betamax

positively influence the consumer's evaluation on each brand, and (2) the installed-user-

base from both of two competing standards generates positive word of mouth effects on

the category diffusion as well as the standard and brand diffusion at the early stage of

standard competition when none of the two competing standards dominates market.

Third, we also find results that the network effects impact the market potential

dynamically. While the installed-user-base from two standards both positively impact the

new technology diffusion at the early stage of standard competition, the installed-user-

base of Betamax turns to impact the diffusion of new technology at three levels

negatively after the technology standard VHS dominates the market.

Finally, our results suggest that network effects and standard competition not only

impact the diffusion on individual consumer's evaluation to each brand, but only on the

market potential complementarily and dynamically. Similarly with the word of mouth

impact of installed-user-base on new technology diffusion, the supporting-firm-base from









both of two standards impact market potentials seems positively at the early stage of

standard competition, but the supporting-firm-base of Betamax has no impact on the

market potential later on when the VCR market is dominated by VHS format.

This study extends research on network effects and provides important implications

to markets with network effects by explicitly modeling and estimating the impacts of

inter-standards competition on new technology diffusion. It also advances understanding

of the formation of consumer expectation for the future installed based of competing

technological standards-a very important but under-explored research area.

Our paper is organized as follows. We first introduce our baseline model and

dynamic model in the section 3.2 and 3.3. Then we present our preliminary results in

section 3.4. We discuss our results and future research directions in section 3.5.

3.2 The Baseline Model

We start with analyzing the new product diffusion at three levels: category,

platform and brand levels by incorporating the impact of network effects and standard

competition into (1) the equation of consumer's evaluation of each brand, (2) the market

potential equation and (3) the diffusion equations at three levels. The network effect

variables that are incorporated in the baseline model are the current installed-user-base

and the current supporting-firm-base. The dynamic model by considering consumer

expectation aboutfuture network effects will be introduced in the next section.

3.2.1 Consumer Utility Equation

We use the nested logit model to estimate the market shares. Let i denote

consumer and j denote brand (j = 0,1, 2,.-.,J ), where outside option is denoted as j = 0.

Different technology standard groups are denoted as g (g = 0,1,2, G). The outside









option j = 0 is the only product in group g = 0. Further, denote the set of brands in each

group g as 3, ( 3, = J) and denote the value of brand j as
g

5t, = A +i AkXkt aP]t
k
g-1 (3.1)
+7 AgNjgt + jgFgt-1 + Igrg t-1 + gF)g', g +7t ; ZMSD + ,t
g' g' g =1

where xJ, and p,, denote the product attribute k and price for brand j; MSD1 denotes

the market share difference between the compatible technology standard of brand j and

other competing technology standards g'. We incorporate four network effect

variables: Njg, N jg't-, Fjgt-i and Fjgt-i in the utility function to measure the installed-

user-base and the supporting-firm-base for each brand's compatible technology standard

g and competing technology standard g', respectively.

Then the utility for consumer i is accordingly modeled as

U,t = 6jt + ,;g, + (1 U)E,. (3.2)

where s,, is an identically and independently distributed extreme value; and


[,, + (1 -o-)s, ] is also an extreme value random variable (i.e., double exponential

distribution)).

3.2.2 Market Share Equations

If brand j is compatible to technology standard g, then the formula for the market

share of brand j as a fraction of the total standard share (see Berry 1994) is

s ,ji = [e /, 5(I-)]/Dg (3.3)


where Dg = e Similarly, the market share for the technology standard g is
J3,










D( -)
s = (3.4)
g

Thus, the market share for brand j as a fraction of all brands regardless its

technology standard can be derive as

e 5j1/(1-(T)
s = ~ D(3.5)
1 D, D 9)
g


With the market share of outside options so -- and

g

ln(D ) = [ln(s,) ln(so)] /(1- a), we can then derive a simple analytic equation for mean


utility levels as

ln(s) ln(s)= ,, + o ln(s,, ) (3.6)

and the parameters in (3.6) can then be obtained by instrumental estimation methods.

3.2.3 Diffusion Equation at Three Levels

To derive the diffusion equation at three levels, we need model the market potential

first. Instead of modeling the market potential as TV households as in literature (e.g.,

Park 2003 and Ohashi 2003), we model it as a function of the average price of all

products in market, product quality and network effects at different standard competition

stages. Define a parameter 0 as


0 = (o (ap)+ k (max(%zk ))+ I o ig,t 1 + ,g1 g,tl -MSD + goFgt- MSD
k g

where ap represents the average price of all products in market; max(xjk) is the maximum


level of each product attribute among different brands. The Ng,,_i -MSD and Fg,t,- -MSD










denote the interactive terms of network effect variables---installed-user-base and

supporting-firm-base with the market share difference. These interactive terms allow us

to capture the dynamic influences of network effects across different stages of standard

competition. The market potential is then modeled as a function of TV households

(TVHH) and the defined parameter 0:

Mt =TVHH-exp() (3.7)

Accordingly, we model the diffusion equations for the whole category as

-Xt = (M,-X, )( + gONg, -i ^glNg,,t -MSD) (3.8)
g

where the industry potential M, is defined in (3.7).

Similarly, each standard's and each brand's diffusion is given by

Xg,, -,,,1 =Mgt(,,A o +go^,Ng, + ^,,l .MSD+ (gg,,ti +,Ng,,-I .MSD)) (3.9)
g' g

And

X,, -X,,t =J,,t(0 +IX, l + oNgJ g + Ng,t .MSD+ (goNjg, 1 + t,,l .-MSD))(3.10)
jg', g

where the market potential level for each technology standard and each brand are

measured as MJ = s, (Mt Xt-) and M g, = sg,t(M -Xt 1), given the industry potential


measure M, and market shares (3.4)-(3.6).

3.3 Dynamic Model with Consumer Expectation

We now incorporating consumer expectation about network effects into the logit

model and estimate it by using dynamic programming and maximum loglikelihood

estimation method.









3.3.1 Consumer Expectation of User-Base and Firm-Base

The consumer expectations of the future installed-user-base NV, and supporting-

firm-base Fg,t are modeled as functions of current installed-user-base Nt, and

supporting-firm-base Fg,t of competing standards. The function is written as

G
Ng, = (OgiNgt, +OgFg,t -)+ =gut,g 1,2,..-,G. (3.11)
/=1

G
Fgt =Z(OsgNg-t 1 +IfFg,t-1)+ g,g = 1,2,.-, G. (3.12)
/=1

where ug,, and g,, are assumed to follow independent and identical multinomial

distribution.

3.3.2 Consumer Dynamic Decision

Different from the myopic assumption that a consumer makes purchase decision by

only maximizing its single period utility, we assume that a dynamic oriented consumer

looks forward T periods ahead and maximize its sum of discounted expected future U,

over T periods, given its expectations about the future network effects. The objective

function of a dynamic oriented consumer sum of discounted expected future is then

defined as

T J
max Et--' t (UJdJit S,) (3.13)
d I rI t [ t ( 3 1 3 )


where u,, is the single-period utility at time t of a consumer i for a product j

defined as (3.2); d,, is an indicator that equals to 1 if a consumer i chooses a product j

at time t and equals to zero otherwise; and 5 is a discount factor. The E(.) is the

mathematical expectation operator and the S, is the state space at time t. The state space









S, includes installed-user-base Ng and supporting-firm-base Fgt for different standard

g.

A dynamic oriented consumer's decision involves making a sequence of decisions

on d,, in each period to maximize its expected discounted utility. Denote the maximum

value of discounted expected utility over the horizon T as

Tr T n
t(St)= max Et (tZ(UjtdjSt) (3.14)


The value function V,(St) can be rewritten as

V,(S(t)) = max [(V, (S)] (3.15)

where V,t(S,), the alternative-specific expected value function, is the expected value of

consumer i selecting alternative j at time t. According to Bellman (1957), we can

derive the alternative specific value function at time t recursively by the following

equation

t (S) = {E(UJ, St,)+ SE[V+, (St+1 St, t, dt = 1,)] (3.16)

where the alternative specific value at last period T is V, (S,) = E(Ur IS,).

As seen in (3.16), the alternative-specific expected utility function assumes that

decisions are made by maximizing the current period utility plus the expected total future

value, given the current choice. The discounted expectation in (3.16) is taken over the

distribution of S(t+1) conditional on the current state S(t), decision d,kt, and current error


term Ey with the transition density denoted by f(S(t+l) S(t),djt,,t) .









3.3.3 Choice Probability

Similarly as in standard choice models that the choice probability of a consumer i

for product j is made by comparing the values of different alternatives, the choice

probability of a dynamic oriented consumer i for product j is also made by comparing

the values of different alternatives. The major difference with the dynamic model is that

the values of different alternatives include both the current period utility and the expected

discounted utilities at time t over the remaining periods. Assuming that (1) the

unobservables c, is conditional independent on observable state variables S(t), and (2)

the unobservables s, is distributed as i.i.d. multivariate extreme value. Then we have the

choice probability for consumer i to choose alternative j at time t as (Rust 1994):

exp f17,(Sol
Pr(d,-kt = t) exp t (St) (3.17)
exp f,-t (So)
J'=0

where V,,(S,) represents the deterministic portion of the alternative specific value

function. That is, V,(S,)= UJ, +s8EVI, (S, +s,) .

Given the choice probability (3.17) for an individual consumer i to brand j and

the observed choice quantities for each brand at each period, we can derive the

loglikelihood function by assuming homogenous consumers or a particular heterogeneous

function of different consumers. This loglikelihood function hence can be estimated by

using maximum loglikelihood estimation methods. Consequently, the market shares and

diffusion equations for each technology standard and brands can be finally estimated by

using the estimated parameters from the maximum loglikelihood estimation methods.









3.4 Empirical Analysis

We estimate the baseline model by using data from the U.S. home VCR market

from 1978 to 1996. This time period ranges from the early stage of market emerge, the

middle stage of the intensive standard competition and to the late market stable stage.

The competition in the home VCR market is one of the best-known examples of

technology standard competition between two incompatible format Betamax and VHS

(Park 2003). As we know, though the Betamax format was first introduced by Sony in

February 1976, VHS format, which is introduced one and half year later by JVC in 1977

and incompatible with the Betamax format, was able to surpass the Betamax in the early

years of 1980s and eventually dominated the U.S. VCR market in the late years of 1980s.

Examining this market would provide valuable insights on how the inter-standard

competition impacts the new product diffusion at the market level, standard level and

brand level.

3.4.1 Data

Our data set includes annual sales, prices and product attributes on top 20 brands.

The annual sales data for the whole VCR market are collected from Consumer

Electronics in Review, published by Consumer Electronics Association. The market share

data for top 20 brands in this market from 1981 to 1996 are collected from issues of

Television Digest / i/h Consumer Electronics, published by Warren Publishing Inc.. The

market share data for each brand from 1978 to 1980 are adopted from Ohashi (2003) and

Park (2003). The top 20 brands in our original sample from 1978 to 1996 count up to

90% of the total market share. Data on prices and product attributes are collected from

New York Times, Orion Blue Book Video and Television (1993, 1997) and Consumer

Reports.










To obtain a preliminary result about the diffusion at each technology standard and

brand, we include the top six brands---RCA, Panasonic, Sony, Zenith, JVC and

Magnavox in our estimation. These brands are the major earliest brands in the VCR

market, which count for 50% of the total market share on average. RCA, Panasonic and

Magnavox adopted VHS format from the beginning when they entered this market. In

contrast, Sony and Zenith first adopted format Betamax but switched to format VHS in

the late 1980s. Each VCR producer usually introduced multiple models in each year,

however since the sales data on model-level are not available, we select a basic model to

represent a brand in each year (Park 2003). The product attributes that are included in our

preliminary estimation are heads, programming events and programming days. The

descriptive statistics for these variables are reported in table 3-1.

Table 3-1. Descriptive Statistics
Product Attribute
Brand Heads Programming Programming Price Sales Market Format
Events Days (Million) Share
RCA 3.42 6.21 197 754.1 1.04 0.16 VHS
(0.96) (2.37) (182) (305) (0.56) (0.0076)
Panasonic 3.16 5 18.63 710.3 0.79 0.10 VHS
(1.01) (3.11) (12.6) (253.7) (0.45) (0.035)
Magnavox 3.16 5.68 125.3 734.5 0.64 0.065 VHS
(1.01) (2.67) (167.5) (341.04) (0.56) (0.029)
JVC 3.26 5.32 124.2 862.1 0.37 0.036 VHS
(0.99) (3.09) (168.2) (399.4) (0.26) (0.018)
Sony 3.05 5.32 16.42 819.7 0.42 0.098 Beta to
(1.18) (2.94) (12.7) (259.4) (0.25) (0.105) VHS
Zenith 3.37 4.84 102.9 699 0.52 0.034 Beta to
(1.16) (3.48) (161.1) (285.8) (0.29) (0.017) VHS
Network Effect
Standard Installed-User-Base Supporting-Firm-Base
VHS 19.34 31.21
(18.68) (17.02)
Betamax 2.57 4.21
(1.48) (3.61)









3.4.2 Nested Logit Model Estimation

As introduced previously, to estimate diffusion equations at three levels, we need

first estimate the market share for each brand and technology standard. We use GMM

estimation method to estimate the nested logit model (3.6), where the product attributes

in our preliminary estimation are Head, Programming Days and Programming Events.

The estimation results of the nested logit model are presented in table 3-2. From

table 3-2, we find that the installed-user-base from compatible technology standard

positively influences consumer's evaluation about a new product, which is consistent to

literature on network effects (Gandal 1995, Brynjolfsson and Kemerer 1996). In contrast,

we also find that the installed-user-base from competing technology influences

consumer's evaluation negatively. This is new to network effects literature and

demonstrates the impacts of standard competition on consumer's evaluation and in turn

the market share and diffusion of each brand.

Similarly as the impacts of user-base from compatible technology standard, we find

that the supporting-firm-base from the compatible standard also positively influences

consumer's evaluation. This reinforces the results in Chapter 2 that supporting-firm-base,

as one of important source of network effects which has been ignored in the literature of

network effect, increases consumers' evaluation about a new product.

Different from the negative impact of user-base from competing standard, our results

show that the supporting-firm-base from competing standard positively impacts the

consumer's evaluation. This interesting result suggests that the consumer's evaluation of

a brand increases with more entrants in the market no matter which competing standards

that an entrant chooses.










Table 3-2. Nested Logit Model Estimation
Coefficient Estimate t-statistics
Constant -4.67 -12.33
Head 0.11 1.25
Programming Days -0.00025 -0.58
Programming Events -0.73E-03 -0.29
Price -0.12* -3.54
User-base for compatible standard 0.043* 5.63
User-base for competing standard -0.053** -2.32
Firm-base for compatible standard 0.019* 4.03
Firm-base for competing standard 0.034 2.16
In(S, /g) 0.87*** 10.18
Cumulative Sales at t-1 0.014 0.63
Market Share Difference 8.48 0.32
: p<0.01, : p<0.05, and *: p<0.1.


3.4.3 Diffusion Equation Estimation

With the estimation of the nested logit model, we then derive the estimated market

share for each technology standard and brand. Using the estimated market share, the

diffusion equations at three levels---category-level (3.8), standard-level (3.9) and brand-

level (3.10) are accordingly estimated by using General Linear Estimation method.

Impacts of Network Effects and Standard Competition on Market Potential

The estimation results of market potential equation (3.7) are reported in table 3-3.

From table 3-3, we can see that the market potential is significantly influenced by the

market average price and product quality. Specifically, we found that average market

price significant decreases the market potential and one of the product quality variable---

Programming Events significantly increases the market potential.

In terms of the impact of network effects and standard competition, thought we

only find significant results on firm-base from Betamax standard, the signs of user-base

and firm-base from both of two competing standards---VHS and Betamax provide some

interesting insights.










Table 3-3. Market Potential Equation Estimation
Coefficient Estimation
Average Price -0.50E-03*
(-4.38)
Head -0.062
(-4.69)
Programming Days -0.91E-03*
(-4.42)
Programming Events 0.14
(1.94)
VHS User-base 0.0040
(0.14)
Betamax User-base -0.29
(-2.73)
VHS Firm-base -0.26E-02
(-0.30)
Betamax Firm-base 0.19**
(6.63)
VHS User-base *Market 0.012
Share Difference (0.058)
Betamax User-base -0.65
*Market Share Difference (-0.90)
VHS Firm-base *Market 0.078
Share Difference (0.62)
Betamax Firm-base -2.98
*Market Share Difference (-6.83)
: p<0.01, : p<0.05, and : p<0.1.



First, we can see from figure 1 that the user-base of technology VHS influences the

market potential positively and this positive impacts becomes stronger as it dominates its

competing standards (i.e., the market share difference between VHS and Betamax

increases). In contrast, the user-base of technology Betamax influences the market

potential negatively, and this negative impact becomes stronger as it is dominated by its

competing standards.


































Figure 3-1. Impacts of Installed-User-Base on Market Potential

Second, the impacts of firm-base on market potentials are different at two different

stages: at the early stage, since the firm-base of technology Betamax is larger than that of

technology VHS due to the first introduction of technology Betamax, the market potential

increases as more firms support the technology Betamax and decreases with more firms

supporting technology VHS. However, as more firms switch to technology VHS while

the firm-base of technology Betamax turns to zero at the late years, the impacts of firm-

base from two technology standards on market potentials reverse: the firm-base of VHS

impacts market potentials positively while the firm-base of Betamax impacts market

potentials negatively (see table 3-3 and figure 3-2).


Figure 1. Impacts of Installed-User-Base on Market Potential

0.5

0-
1 ~2 -T~~ 6 E 8 9 10 11 12 13 14 15 16 1. 18 1I
-0.5


- -1

-1.5

-2
Year

-- User-Base of VHS on Market Potential U--User-Base of Beta on Market Potential






























Figure 3-2. Impacts of Supporting-Firm-Base on Market Potential

Impacts of Word of Mouth and Standard Competition on Diffusion

As discussed previously, different from the traditional diffusion models that only

study the word of mouth impacts from the whole market's cumulative sales, in our

models, we not only incorporate the word of mouth impacts on diffusion from two

competing standards separately, but also incorporate the interactive impacts of word of

mouth with the stages of standards competition, where the stages of standards

competition are measured as the extent of market share difference. The results of the

diffusion equation estimations at three levels are reported in table 3-4 and 3-5.

Table 3-4. The Diffusion Equations Estimation: Category and Standard Level
Coefficient Category-Level Standard-Level
Constant -2.42 0.068
(-3.37) (1.26)
User-base of VHS 0.014 0.014
(1.38) (1.44)
User-base of Beta 0.53 0.14
(2.64) (2.54)
User-base of VHS*Market 0.27* 0.026*
Share Difference (2.20) (1.69)
User-base of Beta*Market -6.29** -0.60**
Share Difference (-2.47) (-2.74)
: p<0.01, : p<0.05, and : p<0.1.


Figure 2. Impacts of Supporting-Firm-Base on Market Potential

1.4
1.2

t; 0.8


0.2
0
-0.2 1 2 3 4 :.. F ': 10 11 12 13 1-4 1 i 1i. 17 10 1'
Year

--Impacts of VHS Firm-Base on Market Potential
--Impacts of Beta Firm-Base on Market Potential











Table 3-5. The Diffusion Equations Estimation: Brand Level
Coefficient Brand-Level
Constant -0.29*
(-19.58)
User-base of compatible 0.0015*
standard (5.37)
User-base of competing 0.0013
standard (1.48)
User-base of compatible 0.0036*
standard*Time Trend (13.68)
User-base of Competing -0.058
standard*Time Trend (16.47)
Brand own user-base 0.0023
(4.29)
***: p<0.01, **: p<0.05, and : p<0.1.

Consistent with the results in the literature of new product diffusion, we find that

the word of mouth impacts from two competing standards---VHS and Betamax both

positively influence the diffusion of new product at three levels. But differently, we also

find that (1) the word of mouth impact of Betamax is larger than that of VHS at the early

years, but (2) at the late years, the word of mouth impact of VHS increases while that of

Betamax becomes negatively (see figure 3-38).


Figure 3. Impacts of Word of Mouth from VHS and Betamax on
Category Diffusion

5
4-
3
2

o. 0
S-1 1 2 3 4 5 6 8 10 11 1 17 18 19
-2
-3
-4
-5
Year

-*- Impacts of VHS on Market Diffusion --- Impact of Betamax on Market Dffusion


Figure 3-3. Impacts of Word of Mouth on Category Diffusion


8 We only present the impact of word of mouth from two competing standards on the whole market
diffusion here. The word of mouth impacts from two competing standards on the standards' and brands'
diffusion are similar as figure 3-3.










More interestingly, we found that as the standard competition appears clear on

which standard dominates market (i.e., the market share difference becomes large), the

dominated standard (which is the Betamax technology in our data) will no longer

generate positive word of mouth, but instead generate negative word of mouth on new

product diffusion at three levels (see figure 3-4). This result further suggests that before

the standard competition has settled to one of the competing standards, both of the two

competing standards contribute to the diffusion at three levels positively and

complementary. Once the market settled to one technology standard, the other technology

standards would become barriers to the expansion of market, standards and brands.


Figure 4. Impact of Betamax User-Base On Category and Standard
Diffusion

1
0-
-1 1 2 3 5 9 10 11 12 15

-2
-3
-4
-5
Year
-Impact of Betamax User-Base on Market Diffusion --- Impact of Betamax on Standard Diffusion



Figure 3-4. Impact of Betamax User-Base on Category and Standard Diffusion

3.5 Conclusion and Future Research

This paper examines how the inter-standard competition affects new technology

diffusion at three levels---the category-level, the standard-level and the brand-level.

Specifically, we examine the impact of competition between incompatible standards on

new technology diffusion by incorporating (1) the standard competition---measured as

the market share differences between the competing standards, and (2) two important









network effect factors: installed-user-base and supporting-firm-base of competing

standards into (1) the consumer utility function, (2) market potential equation, and (3) the

diffusion equations at three levels.

The estimation of the diffusion equations at three levels reveals several interesting

results. First, our results show that while the installed-user-base from the compatible

standard increases the consumer's evaluation to a new product, which is consistent to the

literature on network effects, the installed-user-base from competing technology standard

decreases the consumer's evaluation to a new product, which is new to literature on

network effects, and demonstrates the evidence for one of the important impacts of

standard competition on new technology diffusion.

Second, we find the empirical evidence that standard competition impacts new

technology diffusion complimentarily. This can be seen from two results: (1) the

supporting-firm-base from both of two competing standards---VHS and Betamax

positively influence the consumer's evaluation on each brand, and (2) the installed-user-

base from both of two competing standards generates positive word of mouth effects on

the category diffusion as well as the standard and brand diffusion at the early stage of

standard competition when none of the two competing standards dominates market.

Third, we also find results that the network effects impact the market potential

dynamically. While the installed-user-base from two standards both positively impact the

new technology diffusion at the early stage of standard competition, the installed-user-

base of Betamax turns to impact the diffusion of new technology at three levels

negatively after the technology standard VHS dominates the market.









Fourth, our results suggest that network effects and standard competition not only

impact the diffusion on individual consumer's evaluation to each brand, but only on the

market potential complementarily and dynamically. Similarly with the word of mouth

impact of installed-user-base on new technology diffusion, the supporting-firm-base from

both of two standards impact market potentials seems positively at the early stage of

standard competition, but the supporting-firm-base of Betamax has no impact on the

market potential later on when the VCR market is dominated by VHS format.

Since this is only a preliminary study, there are several directions on which we can

work in the future. First, in our preliminary estimation, we only included top 6 brands

rather than 20 brands that we have in our original datasets. Future work will include all of

the 20 brands in our examination.

Second, it would provide additional insights to us on how the competitive behavior

among firms impacts the new technology diffusion at three levels. Specifically, we are

interested in how the different competitive interactions (i.e., Bertrand, Leader-follower)

among firms within a same technological standard (i.e., intra-standard firms) and firms

across different technology standards (i.e., inter-standard firms) impact the new

technology diffusion differently.

Third, we are still working on our dynamic model by taking consumer expectation

into consideration. Results from our dynamic model not only provide deep insights on

how individual consumer forms its expectation on network effects, particularly on

installed-user-base and firm-base for every competing standard, but also enable us to

compare the results on new technology diffusion at three levels with our baseline model.

The comparison between our baseline model and dynamic model provides






65


methodological implications on whether considering the consumer dynamics improves

the prediction accuracy on new technology diffusion at three levels.

Finally, the generalization of our study is limited. To see if our results can be

generalized across different markets with network effects and standard competition,

additional studies are needed on other markets with standard competition.















LIST OF REFERENCES


Bass, Frank M. (1969), A New Product Growth Model for Consumer Durables,
Management Science, 15 (January), 215-227.

-----, Trichy V. Krishman and Dipak C. Jain (1994), Why the Bass Model Fits Without
Decision Variables, Marketing Science, 13, 203-223.

Basu, Amiya, Tridib Mazumdar and S. P. Raj (2003), Indirect Network Externality
Effects on Product Attributes, Marketing Science, 22 (2), 209-221.

Bellman, Richard (1957), Dynamic Programming, Princeton, NJ: Princeton University
Press.

Berry, S. (1994), Estimating Discrete-Choice Models of Product Differentiation, Rand
Journal ofEconomics, Vol. 25 (2), 242-262.

Besanko, D., Gupta S., and Jain D. (1998), Logit Demand Estimation under Competitive
Pricing Behavior: An Equilibrium Framework, Management Science, 44 (11),
1533-1547.

Bresnahan, T. (1989), Industries and Market Power, In: Schmalensee, R., Willig, R.
(Eds), Handbook ofIndustrial Organization, North Holland, Amsterdam.

Brynjolfsson, Erik and Chris F. Kemerer (1996), Network Externalities in
Microcomputer Software: an Econometric Analysis of the Spreadsheet Market,
Management Science, 42 (12), 1627-1647.

Christensen, Clayton M., Fernando F. Suarez, and James M. Utterback (1998), Strategies
for Survival in Fast-Changing Industries, Management Science, 44 (2), Part 2 of 2,
S207-S220.

Corner, Kathleen R. (1995), Obtaining Strategic Advantage from Being Imitated: When
Can Encouraging "Clones" Pay? Management Science, 41 (2), 209-225.

Dranove, D. and Gandel N. (2003), The DVD vs. Divx Standard War: Empirical
Evidence of Network Effects and Preannouncemnt Effects, Journal of Economics
and Management Science, 12 (3), 363-386.

Economides, Nicholas (1996), Network Externalities, Complementarities, and Invitation
to Enters, European Journal of Political Economy, 12, 211-233.









Farrell, Joseph and Garth Saloner (1986), Installed base and Compatibility: Innovation,
Product Preannouncement, and Predation, The American Economic Review, 76 (5),
940-955.

Gallini, Nancy T. (1984), Detterence By Market Sharing: A Strategic Incentive for
Licensing, The American Economic Review, 74 (5), 931-941.

Gandal, N. (1995), Competing Compatibility Standards and Network Externalities in the
PC software Market, Review of Economics and Statistics, 77 (4), 599-608.

--------, M. Kende and R. Rob (2000), The Dynamics of Technological Adoption in
Hardware/Software Systems: The Case of Compact Disc Players, Rand Journal of
Economics, 31, 43-61.

Garud, Raghu and Arun Kumaraswamy (1993), Changing Competitive Dynamics in
Network Industries: An Exploration of Sun Microsystems' Open Systems Strategy,
Strategic Management Journal, 14, 351-369.

Gasmi, F., Laffont J.J., and Vuong Q. (1992), Econometric Analysis of Collusive
Behavior in a Soft Drink Market, Journal of Economics and Management Strategy,
1 (2), 277-311.

Gowrisankaran, G. and J. Stavins (2004), Network Externalities and Technology
Adoption: Lessons from Electronic Payments, Rand Journal ofEconomics, 35 (2),
260-276.

Gupta, Sachin, Dipak C. Jain, and Mohanbir S. Sawhney (1999), Modeling the Evolution
of Markets with Indirect Network Externalities; An Application to Digital
Television, Marketing Science, 18 (3), 396-416.

Horsky, Dan (1990), A Diffusion Model Incorporating Product Benefits, Price, Income
and Information, Marketing Science, 9 (4), 324-385.

Inceoglu, Firat and Minsoo Park (2003), Diffusion of a New Product under Network
Effects: The Case of U.S. DVD Player Market, Boston University, working paper.

Iwata, G. (1974), Measurement of Conjectural Variations in Oligopoly, Econometrica, 42
(5), 947-966.

Kadiyali, Vrinda (1996), Entry, Its Deterrence, and Its Accommodation: A Study of the
U.S. Photographic Film Industry, Rand Journal ofEconomics, 27 (3), 452-478.

------, Vrinda, Pradeep Chintagunta, and Naufel Vilcassim (2000), Manufacturer-Retailer
Channel Interactions and Implications for Market Power: an Empirical
Investigation of Pricing in a Local Market, Marketing Science, 19 (2), 127-148.









-----, K. Sudhir, and Vithala R. Rao (2001), Structural Analysis of Competitive Behavior:
New Empirical Industrial Organization Methods in Marketing, International
Journal ofResearch in Marketing, 18, 161-186.

Katz, Michael and Carl Shapiro (1985), Network Externalities, Competition and
Compatibility, The American Economic Review, 75 (3), 424-440.

Lerner, Josh (1997), an Empirical Exploration of a Technology Race, Rand Journal of
Economics, 28 (2), 228-247.

Mahaj an, Vijay and Eitan Muller (1979), Innovation Diffusion and New Product Growth
Models in Market, Journal ofMarketing, 43, 55-68.

----------------- and ---------- (1996), Timing, Diffusion, and Substitution of
Successive Generations of Technological Innovation: The IBM Framework Case,
Technological Forecasting and Social Change, 51, 109-132.

Nair, Harikesh, Pradeep Chintagunta, and Jean-Pierre Dube (2004), Empirical Analysis
of Indirect Network Effects in the Market for Personal Digital Assistants,
Quantitative Marketing and Economics, 2, 23-58.

Norton, John A. and Frank M. Bass (1987), A Diffusion Model of Adoption and
Substitution for Successive Generations of High Technology Products,
Management Science, 33 (September), 1069-1086.

--------- and --------- (1992), Evolution of Technological Generations: The Law of
Capture, Sloan Management Review, 33 (Winter), 66-77.

Ohashi, H (2003), The Role of Network Effects in the U.S. VCR Market, 1978-1986,
Journal of Economics andManagement Strategy, Vol. 12 (4), 447-494.

Park, S. (2003), Quantitative Analysis of Network Externalities in Competing
Technologies: The VCR Case, The Review of Economics and Statistics, 86 (4),
937-945.

Rust, J. (1987), Optimal Replacement of GMC Bus Engines: An Empirical Model of
Harold Zurcher, Econometrica, 55 (5), 999-1033.

Saloner, Garth and Andrea Shepard (1995), Adoption of Technologies with Network
Effects: An Empirical Examination of the Adoption of Automated Teller Machines,
Rand Journal of Economics, 26 (3), 479-501.

Shankar, Venkatesh and Berry L. Bayus (2003), Network Effects and Competition: An
Empirical Analysis of the Home Video Game Industry, Strategic Management
Journal, 24, 375-384.

Shapiro, C. and R. Varian (1999), The Art of Standard Wars, California Management
Review, 41 (2), 8-32.









------ and ------ (1999), Information Rules: A Strategic Guide to the Network Economy,
Harvard Business School Press, Boston, Massachusetts.

Srinivasan, Raji, Gary L. Lilien, and Arvind Rangaswamy (2004), "First in, First out?
The Effects of Network Externalities on Pioneer Survival," Journal ofMarketing,
68 (January), 41-58.

Sudhir, K. (2001a), Competitive Pricing Behavior in the Auto Market: A Structural
Analysis, Marketing Science, 20 (1), 42-60.

----- (200 b), Structural Analysis of Manufacturer Pricing in the Presence of a Strategic
Retailer, Marketing Science, 20 (3), 244-264.

Sun, Baohong, Jinhong Xie, and H. Henry Cao (2004), Product Strategy for Innovators in
Markets with Network Effects, Marketing Science, 23 (2), 243-254.

Vilcassim Naufel J., Vrinda Kadiyali and Pradeep V. Chintagunta (1999), Investigating
Dynamic Multifirm Market Interactions in Price and Advertising, Management
Science, 45 (4), 499-518.

Xie, Jinhong and Marvin Sirbu (1995), Price Competition and Compatibility in the
Presence of Positive Demand Externalities, Management Science, 41, 909-926.

Wang, Q. and Jinhong Xie (2005), A Structural Analysis of Intra-Standard Price
Competition in Markets with Network Effects, University of Florida, working
paper.















BIOGRAPHICAL SKETCH

Qi Wang was born in Henan, China, in 1969. She earned a Bachelor of Applied

Mathematics at Zhengzhou University with highest honors in 1991, and subsequently a

Master of Econometrics at Zhongshan University with highest honors in 1994. Before

joining the Ph.D. program in the Marketing Department at the University of Florida, she

had been an assistant professor at Zhongshan University for two years and an editor at

Zhongshan University Press for five years.

During her five years of study at the University of Florida, she taught an

undergraduate course in Sales Management in 2003 and 2005. She has now accepted a

position as an Assistant Professor of Marketing at State University of New York at

Binghamton.