Citation
The efficiency of the single-family housing market

Material Information

Title:
The efficiency of the single-family housing market an empirical study
Creator:
Gatzlaff, Dean H
Publication Date:
Language:
English
Physical Description:
x, 145 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Abscission ( jstor )
Avocados ( jstor )
Branches ( jstor )
Fruits ( jstor )
Inflorescences ( jstor )
Leaves ( jstor )
Phloem ( jstor )
Starches ( jstor )
Sugars ( jstor )
Testa ( jstor )
Dissertations, Academic -- Finance, Insurance, and Real Estate -- UF
Finance, Insurance, and Real Estate thesis Ph. D
Housing ( lcsh )
Housing, Single family ( lcsh )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1990.
Bibliography:
Includes bibliographical references (leaves 138-144).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Dean H. Gatzlaff.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
001640228 ( ALEPH )
24229030 ( OCLC )
AHR5313 ( NOTIS )

Downloads

This item has the following downloads:


Full Text
Figure 8-8. Total Sucrose Content of Avocado Leaves from Three Successive Growth Flushed on Girdled and Non-Girdled Branches. Leaf disks (0.618 cm ) were removed from each leaf of each flush at harvest. The samples from each flush were pooled, chopped and boiled in 85% EtOH. The ethanolic extracts were concentrated and the sugar content measured by the phenol-sulfuric acid method. The mean sucrose content for each entire flush was calculated.


155
the least massive (Fig. 8-4) yet it had the greatest values for the Hpm fixed per area and dry mass. As the plastochron leaf age increased, leaf mass increased and the relative dpm content decreased. The relative decrease in assimilation observed between flushes of different ages probably resulted from the decrease in soluble and fraction 1 proteins (82,83) combined with mass increases associated with an increase in starch and structural components. Starch and other structural carbohydrates add to the leaf mass but did not enhance photoassimilation.
In non-girdled flushes between 42-61% (Fig. 8-5) of the initially
1 4
fixed C remained in the leaves after 21 hours. The oldest flush had the slowest rate of translocation, translocating 25% of the fixed label in 21 hours, while the youngest flush had the greatest rate of translocation, translocating 50% of the fixed label during the same time period. The rate of translocation has been correlated to net photosynthesis (184). Rapid photosynthesis usually accompanies rapid translocation and rapid utilization. Girdling branches decreased
translocation from the flushes by almost 50%. At 70 hours the non-girdled branches had exported approximately 72% of the initially fixed CO^ while girdled branches had exported only about 29% of the initially assimilated label. Fondy and Geiger (67) described lower rates of translocation in girdled bean leaves than in control plants. They suggested that since the net carbon exchange rate did not change, that girdling alters the distribution of photoassimilates to various sink/pools.
The relative percentages of label in the neutral (sugars), basic (amino acids), acid 1 (organic acids and sugar monophosphates), acid 2 (phosphoglyceric acid and sugar diphosphates) fractions were not


186
flush the least (Chap. 8). Although the net carbon exchange (NCE)
rates of girdled and defruited plants do not change immediately after
sink manipulation (67,71) it is evident from our data and others (137)
that the absolute rate of photosynthesis decreased within 24 hours of sink manipulation.
The most distal flush (Fig. 9-3) had the least mass/unit area and
the most proximal flush had the greatest mass per unit area. Defruiting did not affect flush dry mass; girdling did increase dry
mass particularly in the oldest flush (flush 1) suggesting that the leaves on the girdled branches were acting as alternate sinks (39) for assimilates which would under non-girdled conditions be translocated to the trunk, roots or other sinks. Furthermore, the leaves of fruit-bearing branches were significantly more massive than flushes of non-fruit-bearing branches (Chap. 8)
The rate of translocation of labeled assimilates (Fig. 9-4) from the leaves on non-girdled fruit-bearing branches varied only slightly between flushes. At 24 hours post-labeling, between 39-48% of the initially fixed label remained in the leaves. Girdling decreased the rate of translocation with 48 to 62% of the label remaining in the leaves at 24 hours. Greater differences between the translocation rates of individual flushes on the girdled fruit-bearing branches were noted, for example at 19 hours flush 1 contained 62% of its initially fixed C and flush 3 contained 48% of its ini tially fixed "C. The translocation rates from defruited flushes were similar to those observed in girdled fruit-bearing branches. Translocation was inhibited between 10 to 15% by girdling. This suggests that in addition to the fruit, the roots and trunk were sinks for these leaves.


CD
n
(0
o
DC
n c
O
u
0)
DC
(D O
60
50
40
30
20
o
10
0
0
a
i
2
3
4
5
6
7
8
9
10
0 HOUR
VZZZ 1 HOUR
SSSSa 4 HOURS
11
+
12
a
13
14
MOBILITY ( cm from TLC plate edge)


187
Further extrapo]ation suggested that the leaves produced enough assimilate for both sinks, since the removal of one sink by girdling decreased photoassimilation. If photoassimilates vere limiting, girdling should not alter the rate of translocation. In non-fruit-bearing branches the removal of the major sinks by girdling inhibited translocation by 50% (Chap. 8)
The relative distribution of label in the neutral (sugars), basic (amino acids), acid 1 (organic acids and sugar monophosphates), acid 2 (phosphoglyceric acid and sugar diphosphates) fractions vas not significantly different betveen fruit-bearing and non-fruit-bearing branches or the flushes on these branches (Table 9-2); suggesting, that recent photoassimilates vere incorporated the into various soluble pools in the same proportions in both branch types. Even though relative pool sizes vere the same, the absolute pool size of neutral sugars vas larger in fruit-bearing branches. Similar observations have been made in citrus (72) and apple (84). The absolute quantity of ethanol soluble sugars (gm sucrose eq./gm fresh veight) (Fig. 9-5) vas higher in leaves of girdled branches of both non-fruit bearing and fruit-bearing branches. Hovever vithin a given treatment the fruit-bearing branches had the higher sucrose content. The most recently formed flush on the non-girdled branch, flush 3, had the highest sucrose content on a fresh veight basis (100 mg/gm fresh veight). Flushes 2 and 1 contained similar amounts of sucrose (approximately 50 mg sucrose/gm fresh tissue). The sucrose content for the three girdled fruit-bearing flushes ranged betveen 110-120 mg sucrose/gm fresh tissue. Defruited branches contained less than 50 mg sucrose/gm fresh tissue in the tvo most recent flushes.


Figure 2-1. Profile of Radiolabel Distribution in Leaf Extracts at 0, 1 and 4 Hours after Labeling. Leaves were fed CO and harvested immediately, or at 1 or 4 hours after labeling anh then separated into lamina and major vein fractions and dropped into boiling 85% EtOH. The leaves and veins were boiled, filtered and the residue was washed with 1 volume each of 50% EtOH and water. The extract was rotoevaporated and then spotted on pre-activated silica FG plates which were developed in a chloroform:acetic acid:water (3:3.5:0.5) solvent. The plates were air dried after the solvent front had moved at least 10 cm. One cm squares were then scraped into scintillation vials and counted. The peak of radioactivity which comigrates with sucrose is marked by an *. Location of mannoheptulose is indicated by a + .


Figure 2-2. Profile of radiolabel distribution from phloem exudates. Leaves were fed CO and harvested immediately, or at 4 or 24 hours after labeling. Phloem exudates were collected in capillary pipettes and applied in 0.25 ul quantities to pre-activated silica FG chromatography plates. Similar patterns were seen with all races of avocado. The peak of radioactivity which comigrates with sucrose is marked by an *. The location of mannoheptulose is indicated by a +.


221
74. Goldring A, S Gazit, C Degani (1987) Isozyme analysis of mature
avocado embryos to determine outcrossing rate in a 'Hass' plot. J Amer Soc Hort Sci 112:389-392
75. Goldschmidt EE, N Aschkenazi, Y Herzano, AA Schaffer, SP Monselise
(1985) A role for carbohydrate levels in the control of flovering in citrus. Scientia Hortic 26:159-166
76. Goldschmidt EE, A Golomb (1982) The carbohydrate balance of
alternate bearing citrus trees and the significance for flovering and fruiting. J Amer Soc Hort Sci 107:204-208
77. Grange RI (1987) Carbon partitioning in mature leaves of pepper:
Effects of transfer to high or lov irradiance. J Exp Bot 38:77-83
78. Grochovska MJ (1973) Comparative studies on physiological and
morphological features of bearing and non-bearing spurs of the apple tree. I. Changes in starch content during grovth. J Hort Sci 48:347-356
79. Grochovska MJ, A Karaszevska (1976) The production of grovth
promoting hormones and their active diffusion from immature, developing seeds of four apple cultivars. Fruit Sci Reports 3:5-14
80. Haas ARC (1936) Grovth and vater relations of the avocado fruit.
Plant Physiol 11:383-400
81. Halevy AH, CS Whitehead, AM Kofranek (1984) Does pollination
induce corolla abscission of Cyclamen flovers by promoting ethylene production? Plant Physiol 75:1090-1093
82. Hall AJ, CJ Brody (1977) Assimilate source-sink relationship in
Capsicum annum L. II. effects of fruiting and defloration on the photosynthetic capacity and senescence of the leaves. Aust J Plant Physiol 4:623-636
83. Hall AJ, FL Milthrope (1977) Assimilate source-sink relationships
in Capsicum annum L. III. The effects of fruit excision on photosynthesis and leaf and stem carbohydrates. Aust J Plant Physiol 4:771-783
84. Hall NT, JM Smoot, RJ Knight Jr, S Nagy (1980) Protein and amino
acid compositions of ten tropical fruits by gas-liquid chromatography. J Agric Food Chem 28:1217-1221
85. Hansen P (1967) 14C-Studies on apple trees. I. The effect of the
fruit on the translocation and distribution of photosynthates. Physiol Plant 20:382-391
1 4
86. Hansen P (1970) C-Studies on apple trees. V. Translocation of
labelled compounds from leaves to fruit and their conversion vithin the fruit. Physiol Plant 23:564-573


104
Translocation of assimilates was monitored by measuring the decrease in radioactivity in leaf punches taken from the labeled source leaf. Two punches, totaling .618 cm2, were sampled from the leaf immediately after labeling and again 48 hr later. Treatment branches, including all of the leaves, flowers, and fruitlets located distal to the girdle were harvested 48 hours after labeling. Two leaf punch samples were taken from each of the mature, proximal leaves (approximately 10 per branch). Distal tissues were sampled in their entirety. Cross-sectional slices (2-3 mm wide) of each branch (wood,bark) were sampled at locations both basipetal and acropetal to the labeled leaf. All samples were oven dried at 56C until reaching constant mass, weighed, digested with a commercial tissue solubiliser (Scintigest Fisher Scientific) and decolorized with 30% H202 as described by Burrel and Brunt (25) prior to liquid scintillation counting.
Forty-five separate labeling experiments were done over a period of two years. The results are the means of at least ten replicate samples taken at four different developmental stages described in Results.
Results
The results of the labeling experiments were grouped into four stages based on the average midvein length of the distal leaves; stage 1 the average midvein length of the distal leaves was <20 mm, stage 2 the average midvein length of the distal leaves was between 20 and 30 mm, stage 3 the average midvein length was between 30 and 50 mm and stage 4 the average midvein length was greater than 50 mm. During the


CHAPTER 8 ASSIMTLATTON OF 14CO BY NON-FRUIT RF.ARTNG FLUSHES
OF AVOCADO: PERSEA AMERICANA MILL. CV. 'PETERSEN'____150
Introduction.......................................150
Materials and Methods..............................151
Results and Discussion.............................152
CHAPTER 9 TRANSLOCATION OF RECENT PHOTOASSIMILATES IN
FRUIT-BEARING FLUSHES OF AVOCADO: PERSEA AMERICANA
MILL. CV. 'PETERSEN'.................................180
Introduction.......................................180
Materials and Methods..............................182
Results and Discussion.............................183
CHAPTER 10 SUMMARY..............................................210
CITATIONS........................................................216
BIOGRAPHICAL SKETCH..............................................231
iv


158
relative distribution of sucrose and starch within the leaves of the branch. Girdling the branch increased the sucrose content, sucrose specific activity and starch content of the oldest flush. Although mature leaves lose the ability to actively import sucrose (73), the present data suggested otherwise. In conclusion, the age of leaves on non-fruit-bearing branches did affect photoassimilation and the rate of translocation with the youngest leaves assimilating and translocating the majority of the recent radiolabled photoassimilate.


43
Table 3-1. Abscised Fruitlets on Singly-fruited and Multiple-fruited Inflorescences.
Inflorescences bearing a total of 106 fruit were monitored between February and August of 1986. There were 22 multiple-fruited inflorescences bearing a total of 74 fruit. Of these 22
inflorescences, 15 bore at least two-fruit each on a single peduncle. In the table below, "pairs" refers to two fruit borne on the same peduncle while "individual" refers to two or more fruit on the same inflorescence but borne on different peduncles. Pairs and individual fruit did exist on the same inflorescence. A total of 32 inflorescences bearing only one fruit were observed. These fruit are refered to as single fruit in the table below. The heading "% of abscised fruit" refers to the abscised fruit within each of the following categories; pairs, individual or single classifications of fruit. A total of 84 fruit abscised during the period of observation.
Inflorescence Type
Total # of Fruit Examined
% of Abscised
X of Total Abscised
% of Total Fruit
Multipie-frui ted Pairs 30
Individual
44
83
79.5
29.8 41.6
23.1 32.4
Single-frui ted Single
32
75
28.6
22.2
*
At a 95% confidence level these % are not significantly different.


86
during the female stage had very different dry mass (0.00/i5 vs. 0.0039). Because of the dicogamous flowering habit of avocado, the age of the post-anthesis flowers was different in the male and female
Flowers that were post-anthesis during the female stage had first opened 24 or 48 hours earlier. Flowers that were post-anthesis at the male stage could have opened 4 or 24 hours earlier. The lower
dry mass of the older post-anthesis flowers of the female stage
suggested that as the flower matured or senesced the dry mass
decreased.
Experiments were done to compare the relative 14C02 fixation by various components of a branch (Table 5-4). Results shown in Table 5-4 are typical of 5 replicate branch labelings done in this series.
The branch described here had a main axis and four subterminal
branches bearing new flushes of leaves with average midvein lengths
less than 20 mm. In addition to flower buds and flowers there were 21
fruits which ranged in diameter from 1 to 11 mm. The older, proximal
a 14
leaves fixed the majority (86.3%) of C02 recovered from the branch.
14
The new leaves fixed only about 10% of the total CO recovered from
J 2
the branch. However, they were 50% as efficient as the mature leaves in fixing 4C02 on a dry mass basis.
Fruits and flowers fixed very small amounts of 14C02 The total
14 1 -
C02 fixed (1.63%) and the specific activity (1.2 dpm mg ) of the flowers were greater than that of the fruits (.57%,.98 dpm mg"1, respectively). Fixation of 14C02 by the fruits increased with the diameter and paralleled the increase in dry mass (Fig. 5-2). The


Figure 2-4. The Relative Specific Activity of Avocado Leaf Sugars at Various Temperatures. Leaves were incubated at three temperatures and fed CO At the end of the incubation period the leaves were macerated and boiled in 85% EtOH. Leaf extracts were spotted on TLC plates which were developed in a chloroformracetic acid:water solution. The plates were removed from the developing solution and scraped for sugar extraction and scintillation spectroscopy. The silica gel scrapings were extracted with 500 ul of water. Sugar content was measured by the phenol-sulfuric acid method. Radioactivity was measured by scintillation spectroscopy. The relative specific activity (% total dpm/% total sugar) describes the increase in sugar radioactivity in relation to the amount of sugar present.


121
14 14
period between 0900 and 1100 hr. C0^ was generated from NaH CO^ (Amersham, specific activity of 54 mCi/mmol) using equimolar amounts of
5% lactic acid.
Development of the distal flush was examined by labeling individual leaves located proximal to the inflorescence (assimilate import studies). Branches were harvested at 24 or 48 hrs. and processed as described below.
Leaf and branch samples were dried at 56C for three days. The
samples were weighed, digested with a commercial tissue solubiliser
(Scintigest Fisher) and decolorised with 30% HO (25). When the samples were decolorized, 20 mL of scintillation fluid was added to the samples for counting.
The entire distal leaf mass (assimilate export studies, net photosynthetic capacity studies) or individual distal leaves
(assimilate export studies, phyllotaxy studies) were also labeled.
Entire branches were harvested at 0,24 or 48 hours after labeling and dissected immediately.
The leaves were removed from the branches after noting their relative orthostichy. The length of the midvein (MVL) was measured. Leaves with MVL less than 30 mm were processed in their entirety for scintillation counting. Leaf samples (.618 cm2 discs) were taken from
leaves with MVL>30 mm and processed for scintillation counting.
Positions of flowers and fruit were also noted before removal from
the branch. Reproductive tissues were used in their entirety and
processed for scintillation counting.
Net photosynthesis was calculated from leaf punch samples (.618 cm2) removed from the source leaf immediately after the one hour pulse
1 4
of The amount of translocated assimilate was calculated by


139


51
- 4 1 I I 1 1 I 1 I I I I I 1 1 I 1 1 I 1 I 1 1 I 1 I 1 1 1 I I I I 1 0 15 30 45 60 75 90 105 120 135 150 165
DAYS POST-ANTHESIS
+ FRUIT # 76
* FRUIT # 77 O FRUIT # 86 A FRUIT # 87 ? FRUIT # 88
# FRUIT # 89 $ FRUIT # 90


Figure 7-3. Distribution of Assimilates to Developing Distal Leaves vhen the Average Distal Midvein Length vas less than 20 mm. The black-dotted leaf represents the radiolabeled leaf. The distribution of total radioactivity to each orthostichy is shovn in this representative schmatic. These are the mean value from at least 10 replicates.


2 0mm
16.2%
2%


3
The flower-bud system is derived from terminal and subterminal buds on the previous season's growth. The buds are usually mixed and contain both vegetative and floral primordia (44,46,158,164,165). As

the shoot elongates the lateral branches become simply or highly branched structures which bear single flowers forming a cymose inflorescence (46). The terminal portion of the main axis remains vegetative and continues to grow. Occasionally shoot tips terminate in a flower bud; hence, no further elongation occurs (158,165). In Florida inflorescence development begins in November (44,46,165) and continues over a period of several months.
The flowers when fully open are approximately 12 mm wide (158). They are borne on cymose panicles and are hypogynous, trimerous and perfect in form. The calyx and corolla are called the perianth lobes. The calyx and corolla can be distinguished only by their position. The perianth is composed of two whorls (a three-parted calyx and a three-parted corolla) which are coherent and form a cup-like structure. Four whorls of stamens, each containing three members, are inserted into this cup structure. The fourth whorl is reduced to glandlike staminodes. The flower contains a single pistil which consists of one carpel and one ovule. The ovule is anatropus and has two integuments.
The vascular tissue (41,169) in the pedicel is arranged in a siphonostele. Cross sections indicate six areas of primary xylem. Toward the receptacle the six perianth traces depart from the stele and pass through the cortex. The gaps left by the shift of the perianth vascular traces disappear; however, the vascular tissue of the stele shrinks as it approaches the receptacle. The carpellary traces depart near the tip of the axis and enter the ovary just before the vascular


90
permeability, wilting and death (22,63,64,142,143) or abscission (81,178). The induction of enzyme systems involved in flower senescence may account for the increased sink intensity of the corolla; however, remobilization of metabolites within the flower, particularly from the nectaries (119) and the stamens (103), may have contributed to the accumulated label recovered in the corolla.
Avocado flowers are light green in color and exhibited some photosynthetic capabilities. Male flowers fixed 28% more 4C02 on a dry mass and a per flower basis than did female flowers (Table 5-3). Flowers labeled during the male or female stage exhibited similar mass
and radiolabel distributions. The pistils of male and female flowers fixed similar percentages of the total amount of 14C02 fixed and comprised similar percentages of the total dry mass (Table 5-2) of the
14 a _
flower. The dry mass and C02 content of the corolla (including nectaries) and the stamens varied slightly with the sexual stage of the flower. Male flowers had a greater mass and dpm percentage in their stamens than did open female flowers (Table 5-3).
Post-anthesis flowers were more massive during the male stage than the female stage although their photosynthetic capacity had decreased
(Table 5-3). The initiation of senescence events after anthesis may account for the progressive decrease in photosynthetic abilities of post-anthesis flowers. The decreased mass of post-anthesis flowers (female stage) may be attributed to respirational mass loss or to advanced senescence events since these flowers are 24 to 48 hours older than male-stage post-anthesis flowers. However, the cause for the lower mass of open flowers during the female stage remains unexplained.


120
vegetative sink strength of the apex by shoot tipping i.e., removing the developing leaves distal to the inflorescence, allowed the fruits to establish their mobilizing ability and increased yield (20).
The hypothesis that fruit abscission is induced by the limited availability of carbohydrates is based on two assumptions. First, that the leaves proximal to the inflorescence cannot provide adequate assimilate for the concurrent growth of fruits and new leaves. Second, that the distal leaves are immature and thus importers not exporters of assimilate during this period of development and fruit abscission. However, work done in this laboratory indicated that older leaves located proximal to the inflorescence (proximal leaves) supplied enough assimilate to support the development of both the new leaves and fruits (Chap 6). Furthermore, the growth of the vegetative bud is temporary; abscission occurs during flowering and early fruit set and continues well after the termination of distal vegetative growth.
The development of the new distal leaves and their contribution to fruit growth has not been examined. The purposes of this research were two-fold. First, to determine when the distal leaves become exporters of assimilate. Second, to characterize the movement of assimilates from mature distal source leaves to developing fruit.
Materials and Methods
Ten-year-old, field-grown trees of Persea americana Mill. cv.
'Petersen' were used in this study. Leaves to be labeled were enclosed
in a sealed plastic bag and labeled with 20 uCi of 14C02 for a one hour


87
majority (>90%) of label remained in the peel and mesocarp (Table 5-5) of the labeled fruit. Between 1 and 3% of the label was recovered in tissues other than the fruit.
1 4
The majority (>90%) of the C02 fixed by the fruit remained in the exocarp or peel (Table 5-5). A significant amount of label was recovered in the mesocarp. Little label was recovered in the seed coat, cotyledons or endosperm. No differences were observed in assimilate distribution in fruit harvested 24 or 48 hours after labeling. The limited translocation of fruit-fixed assimilates was not surprising, since avocado fruit undergo continuous cell division and expansion throughout development (167). Although the mesocarp is highly vascularised any available assimilate would be rapidly metabolized by the growing cells. A small amount of radioactivity was recovered from other tissues (leaves, apex, wood). Minor, but significant translocation from the fruit has been noted in apple (7,86) and strawberry (6).
Discussion
Source/sink relations in avocado during reproductive growth has been an area of active research in this laboratory, particularly with regard to source-limitation-induced abscission. The avocado is a good model for this type of study, since multiple sinks represented by concurrently developing flowers, fruit and leaves, compete for
I
hypothetically limiting amounts of photoassimilates. Work in this laboratory has shown that locally produced photoassimilates are not limiting reproductive growth and that source/sink relations in avocado


Figure 7-5. Distibution of Assimilates to Developing Distal Leaves vhen the Average Distal Midvein Length vas betveen 30 and 40 mm. The black-dotted leaf represents the radiolabeled leaf. The distribution of total radioactivity to each orthostichy is shovn in this representative schmatic. These are the mean values from at least 10 replicates.


Figure 3-6. Distribution of Abscised Fruitlets by Size. The majority of fruit which abscised had a diameter less than 1 cm. A total of 84 of the originally tagged 108 fruit abscised.


213
the developing vegetative and fruit mass than basipetally from the proximal leaves. In the final stages of distal leaf development the percentage of label going from the proximal leaves to distal reproductive and vegetative organs decreased while the percentage traveling basipetally increased. At all periods of development, the sink strength of flowers, fruits and leaves as determined by accumulated labeled assimilate was proportional to their dry mass. Thus, on a per unit mass basis developing fruitlets and distal leaves were equally effective in mobilizing recent photoassimilates. The availability of recent photoassimilates from local sources did not appear limiting since radiolabel was recovered acropetally and basipetally to fruit and does not explain the high abscission rate observed in avocado.
The leaves developing distal to the avocado inflorescence initially represent a potentially competing sink and eventually at maturation a source of assimilate for the developing fruits. The partitioning of recent photoassimilates during development of distal leaves was studied in relation to the leaf's sink/source transition. Avocado leaves reached their peak import rate at 22% final midvein length. Import decreased and was neglible at 37% final midvein length. Export of newly fixed assimilates from the developing leaves began at 20% final midvein length and peaked at 60% final midvein length. Fruits having the same orthostichy as the labeled source leaf received up to 95% of the radioactivity recovered from all of the fruit sampled from the same branch. Distribution of label to fruit with the same aligned orthostichy was a function of fruit mass and distance from the source leaf. The apparent precocious maturation of the leaves distal


124
distal leaves was between 30 and 40 mm (Fig. 7-5), 98% of the label was present in the leaves in direct alignment or adjacent to the source
leaf position. Leaves in direct alignment accumulated 50% of the label recovered in leaf tissue.
Distal Leaf Labeling
The partitioning of translocated assimilates from the mature distal leaves was examined by labeling individual distal leaves and then determining the distribution of radioactivity at each orthostichy.

Vhen there were no fruits present on the branch with the same orthostichy as the source leaf, the distribution of the translocated assimilate was skewed to favor a single adjacent orthostichy (Fig. 7-6). However, when fruits were present which had the same orthostichy as the labeled source leaf, these fruits received greater than 94% of the label recovered in fruit tissues (Fig. 7-7). The remainder of the radiobabeled assimilates was dispersed to the fruits at other orthostichies.
Effect of Distance and Fruit Mass on Assimilate Distribution
Vhen several fruits on separate peduncles had the same phyllotaxy as the labeled source leaf, then fruit size and parastichy determined assimilate distribution (Fig. 7-8). For example, fruits (5,8) which have the same phyllotaxy as the source leaf contained only 52% of the radiolabeled assimilate while fruits (3,7) at the adjacent orthostichy contained 45% of the radiolabel. In this particular case, the fruits at the adjacent position were nearer the source leaf and their mass was twice that of the other fruit.


*


15
Historically, research on the identity and quantity of substances translocated in the phloem has been retarded by difficulty collecting unadulterated phloem samples. Hartig (90) developed a bark incision technique which involved cutting through the phloem members and collecting the exuded sap in capillaries. Although this technique is commonly used, several factors may affect the integrity of the sample. These include loss of exudate to the xylem (206,209,210), alteration source loading (52), loss of phloem function (28,173), dilution and contamination of phloem exudate by the contents of damaged cells and the pressure release of non-phloem constituents (37,129,206). An alternate technique uses feeding aphids to penetrate individual sieve tube members (61,62,136,210). This technique is limited by the restricted feeding locations of aphids and by the limited host range of aphids. Avocado does not have any known aphid pest. Fortunately, under well-watered conditions, avocado petioles exude a sizable droplet of phloem contents when severed from the plant with little contamination from exposed cortical cells (personal observation). This novel characteristic allowed us to examine phloem exudates in individual leaves over time as related to sugar constituents.
Persea americana Mill, is comprised of three races or ecological groups: Mexican, Guatemalan and Vest Indian. These races vary in their cold tolerance (15) and fruit characteristics such as size, oil content, and skin texture (15,128,170). The possible discrepancy in the identity of the translocated sugars noted above may reflect inherent differences in races. This study was undertaken to identify the primary and


Figure 7-9. Translocation of Assimilates from Mature Leaves Distal to the Avocado Inflorescence. Within 48 hours a mean of 69% of the radiolabeled assimilates had been exported from the source leaf. Less than half of the exported assimilates were recovered in the fruit tissues. This is a representative schmatic of at least 10 replicate experiments having a standard error of 5% or less.


225
129. Lehman J (1973) Unterstichungen an Phio^mexsudat von Cucurbita pepo
L. 1. Enzymaktivitaten von Glykolyse, Garung and Citrat-Cyclus. Planta 114:41-50
130. Lewis LN, CV Coggins Jr, HZ Hield (1964) The effects of biennial
bearing and NAA on the carbohydrate and nitrogen composition of 'Vilking' mandarin leaves. Proc Amer Soc Hort Sci 84:147-151
131. Louie DS, JR Addicott, FT Addicott (1970) Applied auxin gradients
and abscission in explants. Plant Physiol 46:654-657
132. Malo SE (1971) Girdling increases avocado yields in South Florida.
Proc of the Tropical Region, Am Soc Hort Sci 15:19-22
133. Marsh R (1935) Rate of growth of Fuerte fruit. Yrbk Calif Avoc Soc
19:89-91
134. Mika A, R Antoszewski (1973) Photosynthesis and distribution in
apple shoots treated by pinching and bark ringing. Biol Plant 15:202-207
135. Milborrow BV, DR Robinson (1971) Factors affecting the
biosynthesis of abscisic acid. J Exp Bot 24:537-548
136. Mittler TE (1953) Amino acids in the phloem sap and their
excretion by aphids. Nature 172:207
137. Mondal MH, VA Brun, ML Brenner (1978) Effects of sink removal on
photosynthesis and senescence in leaves of soybean (Glycine max. L) plants. Plant Physiol 61:394-397
138. Monselise SP, R Goren, I Vallerstein (1972) Girdling effects on
orange fruit set and young fruit abscission. HortScience 7:514-515
139. Mor Y, AH Halevy (1980) Promotion of sink activity of developing
rose shoots by light. Plant Physiol 66:990-995
140. Morgan JM (1977) Changes in diffusive conductance and water
potential of wheat before and after anthesis. Aust J Plant Physiol 4:75-85
141. Mullins MG (1970) Hormone-directed transport of assimilates in
decapitated internodes of Phaseolus vulgaris L. Ann Bot 34:897-909
142. Nichols R (1966) Ethylene production during senescence of flowers.
J Hort Sci 41:279-290
143. Nichols R (1971) Induction of flower senescence and gynoecium
development in carnation (Dianthus caryophyllus) during senescence. Ann Bot 39:287-296


183
Scientific). The tissues vere decolorised vith 30* H 0 as described
22
by Burrel and Brunt (25). The sample vials vere then filled vith scintillation fluid and counted.
The metabolite pool vas examined by determining the proportion of radioactivity in the acid, neutral and basic fractions of the ethanolic leaf extract. Leaf punches vere taken at various times and packed in ice for transport to the laboratory. Flush samples vere pooled and boiled in 80% ethanol. Samples vere processed after the method of Housely et al. (102). The residue from this extraction vas used to determine the starch content by the phenol-sulfuric acid method (53).
Results and Discussion The importance of carbohydrate levels to flovering and fruiting has been revieved by others (67,71,162). Generally, research on the carbohydrate levels in perennial tree crops has focused on seasonal fluctuations in the starch levels of the roots, trunk and branches of the tree (8,9,12,76,88,89,162). Work by Scholefield et al. (162) and Cameron and Borst (26) has shovn that avocado exhibits trends similar to deciduous trees in starch content during the groving season, i.e., carbohydrate levels, specifically the starch levels decrease sharply during flovering and fruit development. A decrease in total carbohydrates has been associated vith alternate bearing in "off season" avocado trees (124,162).
The effect of fruit on carbohydrate partitioning has also been studied. Altering source sink relations has effectively altered leaf and stem carbohydrate content in such diverse crops as soybean (14,23,


123
similar to that of mature proximal leaves. The peak in assimilate import and the onset of assimilate export occurred during the same stage of leaf maturation. Whole Branch Labeling
The relationship between leaf age, midvein length and the leaf specific activity is shown in Fig. 7-2. The entire flush of distal leaves was labeled and the branch was harvested either immediately after labeling or 24 hours post-labeling. The specific activity of the leaves increased gradually with increasing midvein length until the midvein length approached 30 mm. The specific activity of leaves with midvein lengths between 30 and 40 mm increased 7 fold. However, translocation of assimilates from the distal leaves to the fruit had apparently not begun since radiolabel was not recovered in the inflorescence or proximal to the inflorescence. Proximal Leaf Labeling
Individual proximal leaves were labeled at various stages of distal leaf development. The amount of label going into each orthostichy of distal growth was determined. At the earliest stage of distal leaf development (average midvein length <20 mm), labeled assimilates from the proximal leaves were equally distributed among all orthostichies (Fig. 7-3). Vhen the proximal leaves were labeled at the next stage of development (average midvein length between 20-30 mm), the distribution of radiolabeled assimilates was skewed (Fig. 7-4), with the greatest amount of label (42%) going to the leaves in alignment with the source leaf. A total of 73% of the radiolabel in the new leaves was present in leaves in direct alignment with or adjacent to the source leaf position. Vhen the average midvein length of the


96
Table 5-4. Early Season Branch Labeling.
Entire branches were labeled for 1 hour and then immediately harvested. The branches were dissected and processed for scintillation
counting.
Tissue # Midvein dpm %14C %14C _Length (mm)_(% of total) (%dpm/mg)
Old Proximal Leaves 5 70. 9 + 14 2,737,446 86. 3 63
New Distal Leaves 31 18. 8 + 4 320,466 10. 1 33
Fruit 21 18,072 0. 57 0. 98
Fruit Receptacles 6,024 0. 19 0. 92
Fruit Peduncles 38,046 1. 2 1. 4
Miscellaneous
(flowers,buds, panicals)
51,680
1.63
1.2


152
mCi/mmol) using equimolar amounts of 5% lactic acid. Three branches vere chosen and individually labeled as described above for each experiment. Branches vere chosen from the upper, middle and lover canopy positions at each labeling. The experiment vas performed at least 3 times. The most distal flush vill be designated as flush 3, the flush proximal to flush 3 vill be designated as flush 2 and the most proximal flush vill be flush 1.
Translocation from the leaves vas measured by taking .309 cm2 leaf punches throughout the experimental period. Each leaf vas sampled tvice daily during the experiment. The entire branch vas harvested 70 hours after labeling. The leaf samples vere dried, veighed and digested vith a commercial tissue solubiliser (Scintigest Fisher Scientific). The tissues vere decolorised vith H 0 as described by
' 2 2 J
Burrel and Brunt (25). The sample vials vere then filled vith scintillation fluid and counted.
The metabolite pool vas studied by examining the proportion of radioactivity in the acid, neutral and basic fractions of the ethanol extract. Leaf punches vere taken at various times and packed in ice for transport to the laboratory. Flush samples vere pooled and boiled in 85% ethanol. Samples vere processed after the method of Housely et al. (102). The residue from this extraction vas used to determine the starch content by the phenol-sulfuric acid method (53).
Results and Discussion The importance of carbohydrate levels on flovering and fruiting has been revieved by others (45). Generally research on the carbohydrate levels in perennial tree crops has focused on seasonal


Figure 7-8. The Interaction of Mass and Linear Distance on the Distribution of Radiolabeled Assimilates. A single leaf distal to the fruit vas labeled and fruit vere processed as described in the Materials and Methods. Fruit number represents the fruit's linear proximity to the source leaf. Fruit 1 is closest to and fruit 8 is farthest from the source leaf. % C represents the amount of radiolabel recovered from the individual fruit as a proportion of the total radiolabel recovered from all of the fruit.


Figure 7-1. Relationship of Import and Export to Final Midvein Length. The relative import rate or capacity was defined as the mean percentage of the total translocated assimilate imported by the leaf in 24 hours. The relative export rate capacity was defined as the mean percentage of net photosynthate translocated from the source leaf in 24 hours. The experiment was performed 4 times.


72
Table 4-2. Relative Specific Activities of Individual Fruit Tissues during Avocado Fruit Development.
Source leaves on branches were fed CO for a one hour period. The fruit were harvested 24 or 48 hours later and dissected, dried and processed for scintillation counting. Tissue specific activity is defined as dpm mg~ dry mass. The data represent the percentage of the sample in which that tissue had the greatest specific activity; e.g. in fruit of the 0 to 0.5 gm range, the seed coat had the greatest specific activity in 5.3% of the fruit.
Fruit # Fruit Mesocarp Seed Coat Endosperm Cotyledon
Dry Mass Sampled
(gm)_
(percentage of samples in which specific activity was greastest)
0 0.5 76 23.7 5.3 51.3 19.7
.51 1.0 49 10.2 4.1 61.2 24.5
1.01 1.5 15 20.0 0.0 66.6 13.3
1.51 3.0 8 37.5 0.0 50.0 12.5


was evident within 24 hours of puncturing and assimilate movement into the fruit ceased within 72 hours. Fruit separation occurred 24-48 hours later. Avocado flowers were photosynthetic; hovever, they contributed little photoassimilate to fruit growth. At the earliest stage of distal leaf development, assimlates from proximal source leaves were translocated in equal proportions acropetally to the inflorescence and developing distal leaves and basipetally toward the roots. As the inflorescence began to bear fruit, 5-times more label moved acropetally toward the fruit and expanding distal leaves than basipetally from the source leaves. In the final stage of distal leaf development fewer radiolabeled assimilates moved acropetally toward the fruit and mature distal leaves than basipetally. Fruit and distal leaves were equally effective in mobilizing recent photoassimilates.
The import rate of developing leaves peaked at 22% final midvein length (FMVL) and was negligible at 37% FMVL. Export from the leaves began at 20% FMVL and peaked at 60% FMVL. The distribution of assimilates from source leaves depended on sink phyllotaxy; fruit having the same orthostichy as the labeled source leaf received up to 95% of the radioactivity recovered from all of the fruit. Flushes of leaves on fruit-bearing branches fixed more radiolabel and translocated radiolabeled assimilates more rapidly than leaves on non-fruit-bearing branches on the same tree. Girdling branches decreased translocation from leaves on non-fruit-bearing branches by 50% and from leaves on fruit-bearing branches by 15%.
I
These results suggest that there was no apparent competition for recent photoassimilates among or between fruit and developing leaves.
vi


CHAPTER 10 SUMMARY
Approximately 20% of flowers produced by avocado trees will form fruit but only 1 to 7% of the fruit will remain at harvest. Competition between developing reproductive structures and leaves for limiting carbohydrates may cause the abscission of avocado flowers and fruit because, they form in a pseudoterminal position and develop
concurrently with new apical leaves. This study was undertaken to determine if the availability of local photosynthates affected fruit abscission. Several lines of inquiry were made to study this hypothesis, including the identification of the phloem mobile sugars, measurement of fruit growth and yield, examination of assimilate movement into abscising and non-abscising fruit, and studies on the movement of radiolabel on fruit-bearing and non-fruit-bearing branches.
The major sugar components in alcoholic extracts and phloem exudates from representatives of the three races of avocado (Persea americana Mill.) were analyzed using thin layer chromatography and differential staining procedures. Sucrose and mannoheptulose were the predominant sugars found in leaf extracts and phloem exudates. Leaves fed 14C02 produced radiolabeled sucrose exclusively as the labeled translocated sugar for up to 24 hours in all three races of avocado during the summer months (May to August). During the summer months radiolabeled sucrose was observed to move from the lamina to the major veins within one hour after labeling. During the winter months (Nov.
210


Figure 7-4. Distribution of Assimilates to Developing Distal Leaves when the Average Distal Midvein Length was between 20 and 30 mm. The distribution of total radioactivity to each orthostichy is shown in this representative schmatic. These are the mean values fron at least 10 replicates. The black-dotted leaf represents the radiolabeled leaf.


39
described as a single sigmoidal curve. However, the relationship between fruit dry mass and fruit volume (Fig. 3-2) suggested that avocado exhibited a normal fruit growth pattern, i.e., initial growth during the first month post-anthesis was due primarily to cell division and later growth was due to cell expansion. Similar conclusions were drawn by Kushan et al. (122) from their examinations of DNA and polyamine content during fruit development of another common Florida cultivar, 'Simmonds'. They found that the DNA and polyamine content of the avocado fruit peaked at one month from full bloom which is the stage at which we observed the change in growth phases. This pattern of DNA and polyamine accumulation is similar to that observed in ovaries exhibiting a normal growth pattern .
Fruit growth rates were dependent on the stage of development. Fruit on different branches (Fig. 3-4 and 3-5) grew at the same rate at a given developmental stage. As expected the RGR values were greatest immediately post anthesis and the AGR values were greatest at mid-season. There was no difference; however, in the AGR or RGR values for abscising and retained fruit until 4-5 days prior to abscission. At this point growth rate of abscising fruit decreased. In some cases fruit volume increased marginally until fruit separation. In others, fruit volume decreased slightly before separation.
Fruit load on a branch did not affect fruit growth rates. Fruit on multiple-fruited branches (Fig. 3-4) grew at the same rate as fruit of
I
similar size on singly-fruited branches (Fig. 3-5). In fact, phyllotaxy which strongly governs assimilate movement in avocado (Chap 7), had no


60
determine if recent photoassimilates distribute differently within the fruits during abscission. Second, individual fruit will be punctured at time intervals before and after source leaf labeling in order to establish a time frame for the restriction of assimilate movement, senescence and abscission.
Materials and Methods
Fruit Labeling
The trees in this study were ten-year-old 'Peterson' orchard trees growing on Walden rootstock at the Tropical Research and Education Center in Homestead, FL.
Branches (approximately 1 cm in diameter) were girdled 24 hours before labeling. Individual leaves were enclosed in Plexiglas containers and fed with 20 uCi of 14C0^ generated from NaH 14co3 (Amersham with a specific activity of 54 m Ci/mmol) using an equimolar amount of 5% lactic acid. After 1 hour the containers were removed and two disks (.309 cm2) were taken from the source leaves using a hole punch. Source leaves were subsequently sampled twice daily for the remainder of the experiment. Fruit diameters were measured twice daily as described previously (Chap. 3). Growth rates were calculated after the method of Hunt (106) and as described in Chap. 3. Parti tioning Studies
Source leaves were labeled as described above. After 48 hours the branch (1-2 cm in diameter, 30-45 cm in length) bearing the fruit was removed. The fruit were dissected with a sharp knife and forceps and individual tissues dried at 56C. The dried tissue was weighed, digested, and decolorized by the method of Burrel and Brunt (25) in preparation for scintillation counting.


Figure 3-7. Abscission of Fruit over Time. The number of fruit remaining on the trees decreased exponentially with days post-anthesis.


85
stamens and nectaries all decreased post-anthesis which may reflect thp higher rate of flower abscission occurring at this point or the lack of growth in these mature tissues.
The fruit had the greatest sink strength but the lowest sink intensity and RSA. Fruit incorporated 35.2% of the total radioactivity recovered from the inflorescence yet the fruit comprised 71.5% of the total dry mass. Sink strength in this case was more a reflection of
mass than of sink intensity.
CO^ Fixation by Reproductive Structures
14
The relative distribution of dry mass and fixed CO^ within a flower was the same at both the male and female stages (Table 5-2). The corolla and nectaries comprised approximately 60% of the flowers dry mass. The stamens and pistil comprised approximately 26% and 13%, respectively, of the flower dry mass. Greater than 85% of the total flower radioactivity was recovered in the corolla.
Differences were observed in the mass and specific activity of flowers at different stages of development (Table 5-3). The mass of flowers open at the female stage (.00450 gm) was significantly less than that of flowers open during the male stage (.00509 gm). The specific activity of the open male flower was greater than the open female flower (data not shown). The flower buds fixed more C02 than the open or post-anthesis flowers.
The mass of post-anthesis flowers was also significantly greater at the male stage (.0052 gm) than the female stage (.0039 gm). During the male stage, open flower and post-anthesis flowers had similar mass (0.0050 gm vs. 0.0052 gm) while open flowers and post-anthesis flowers


Figure 6-1. Distribution within the Branch of C-Assimilates from the Proximal Source Leaf when the mean Midvein Length of the Distal Leaves was less than 20 mm. Source leaves were fed CO for a one hour period. The entire branch was harvested 48 hours later. The branch was dissected, dried and processed for scintillation counting. The schematic drawing indicates the approximate developmental stage of the reproductive flush from mid-March to early April. The black spots indicate the labeled sourse leaf. The percentages are based on the total radiolabel recovered after the ini tial 1 hr CO fixat ion period. The arrows indicate the direction of assimilate flow. The data represent the mean values of at least 10 replicates. Standard error of measurement was 5% or less.


188
The specific activity (Fig. 9-6) of sucrose extracted from the older flushes (1 and 2) was elevated in the girdled (3,000 and 5,000) and defruited (4,000 and 5,000) flushes compared to their non-girdled counterparts (2,500 and 4,000). The specific activity of the sucrose in the most distal flush vas greater in defruited (15,000) and girdled (10,000) branches than in non-girdled fruit-bearing branches (7,000). Although the distal flushes of girdled and non-girdled fruit-bearing and defruited flushes had similar dry mass, there vas an apparent difference in hov recent photoassimilates vere distributed to the carbohydrate pools.
Girdling and defruiting produced a highly labeled sucrose pool in the most recent flush of the branches. Girdling and defruiting decreased the sink demand on the flush. Hov can the higher specific activity of sucrose in the girdled fruit-bearing and defruited flushes be explained? One explanation involves the allocation of sucrose to a non-transport pool. Mature leaves exhibit compartmentation of sucrose vithin their cells (72,73). Tvo pools of sucrose have been suggested (72,73) one for export from the mesophyll into the apoplast for phloem loading and the other non-transport pool vhich is mobilized on demand. In this case, a portion of the labeled sucrose vas present and maintained in a non-transport pool vhile the remainder vas translocated or converted to starch. The size of this non-transport pool may correlate to sink demand (73). Hovever, over the short period of this experiment the pool sizes may not have compensated for the change in sink demand, therefore the specific activity of sucrose in flush 3 vas greater than expected.


Figure 9-3. Specific Leaf Veight of Avocado Leaves from Three Successive Growth Flushes on Girdled and Non-Girdled; Fruit-Bearing and Defruited Branches. Leaf disks (0.618 cm ) vere removed from each leaf of each flush on both girdled and non-girdled branches. The leaf disks vere dried and veighed as described in the Materials and Methods. The fruit vere removed from defruited branches one veek prior to labeling. Each flush bore at least 10 leaves and each leaf vas sampled at least A times during the experiment.


Figure 9-6. Specific Activity of Sucrose Extract from Avocado Leaves of Three Successive Flushes on Girdled and Non-Girdled Fruit-Bearing and Defruited Branches. At harvest, leaf disks (0.618 cm ) were removed from each leaf of each flush. The samples from each flush were pooled, chopped and boiled in 85% EtOH. The ethanolic extracts were concentrated. Sugar content was quantitated by the phenol-sulfuric acid method. Samples of known sucrose concentration were decolorized for scintillation counting. The fruit were removed from defruited branches one week prior to labeling.


62
variations observed in the specific activities of tissues between and within fruits corresponded to developmental and phyllotactic differences.
Piercing the fruit did not affect the translocation rates of assimilates from source leaves during the course of the experiments. An average of 33.3 +/- 18% of the initially fixed dpm were recovered in the fruit 96 hours after labeling. This equates to 48.7 +/- 24.3% of the total translocated dpm. Individual fruit averaged 1.9 x 106 +/- 1.6 x 106 dpm. The large standard deviation reflects the strong effect of phyllotaxy on the partitioning of recent photoassimilates (Chap. 7). For example if the source leaf and fruit had similar phyllotaxy then nearly 95% of the translocated label was recovered in that fruit. However, if the fruit and source leaf were not in direct alignment only a small percentage of label would move into the fruit (Chap. 7). Since source leaves were labeled without regard to phyllotaxy, a large discrepancy in the absolute imported dpm's was observed.
Relative assimilate import by the tissues of different sized fruits is shown in Table 4-2. The numbers given under each tissue type represented the percentage of the total fruit in which that tissue had the greatest specific activity. For example, the endosperm had the greatest tissue specific activity in 51.3% of the fruit in the 0-0.5 gms dry mass range. In approximately 5% of the fruits with dry masses less than 1 gram, the seed coat was the most actively importing tissue. Fruitlets in this size range also composed the greatest percentage of abscised fruit (Chap. 3). Fewer fruit in the 1 to 3 gram range abscised, and interestingly the seed coats of fruit in this sample were


94
Table 5-2. Fixation of 14C0 by Open Flowers.
Open flowers were labeled during both the male and female stages. The flowers were harvested, dissected and processed for scintillation counting immediately after labelling. The nectaries were included with the corolla.
Male Flowers Female Flowers
Tissue X Dry Mass X dpm X Dry Mass X dpm
Pistil 13.6 + 2.4 4.2 + 1.5 12.6 + 1.8 4.5 + 0.3 Stamens 29.1 + 0.2 9.5 + 0.3 23.0 + 2.5 5.9 + 0.8 Corolla 57.2 + 2.2 86.3 + 1.3 64.4 + 2.0 89.6 + 0.5


12
Assimilates moving into the reproductive structures are generated in the vegetative portions of the plant. Recently fixed photosynthetic assimilates are preferentially translocated to the sink; however, remobilization of storage compounds may also occur. In soybean (29) girdling did not affect the 14C-assimilate accumulation in pod walls or seeds; however, it did decrease the starch levels of the subtending leaf. Although sink demand remained constant, the source of readily available assimilate decreased when the plant was girdled. The plant compensated for this change by altering the partitioning of fixed carbon. The developing embryo may exert a direct effect on photosynthetic efficiency. The net C02 uptake in Capsicum annum was reduced by 30% when the fruits were removed. The stems of defruited plants exhibited pronounced changes in their soluble sugar and polysaccharide content. The decrease in photosynthate accumulation reflected the decrease in sink demand. The loss of photosynthetic capacity was due to the loss of both soluble fraction one proteins and RuBP carboxylase (82).
The ability of a plant to utilize stored reserves and the decrease in these reserves after a heavy fruit load have been used to explain alternate bearing (75,76,162). A heavy fruit load decreases the amount of stored reserves in the plant and decreases the plant's ability to produce fruit the next season. Olive is a notorious alternate bearing crop. Stutte and Martin (181) have shown that the carbohydrate reserves in olive have no affect on alternate bearing. They used C02 enrichment techniques to increase the starch content tenfold over the controls yet they saw no differences in the flowering pattern of the bearing and nonbearing trees.


QUANTITY OF FRUIT


103
noted during this period (44) has fostered the assumption that the developing floral and vegetative tissues compete for limiting amounts of assimilates (162). It follows from the commonly observed, high abscission rates of reproductive organs that fruitlets are less competitive than the developing distal leaves or other major sinks, i.e., branches and roots, in mobil izing assimilates. The purpose of this study was to examine the contribution of leaves, located proximal to the inflorescence, to the development of the distal tissues and to determine if the leaves distal to the inflorescence compete with the developing reproductive organs for recent photoassimilates.
Materials and Methods Ten-year-old trees of Persea americana Mill. cv. 'Peterson', growing in Rockdale soil at the Tropical Research and Education Center, Homestead, FL were used in this study. Branches (1-2 cm diameter) were girdled 2 to 3ft from the tip 24 hours prior to each labeling experiment. A single, intact leaf, located proximal to the inflorescence on each experimental branch, was enclosed in a 2-liter,
1 4
plastic bottle and fed 20 uCi of C02 during an hour period between
1 4
0900 and 1100 hr NaH C03 (Amersham with a specific activity of 54 mCi/mmol) was injected through a serum vial stopper into the bottle. An equimolar amount of 5% lactic acid was injected into the bottle to generate C02. Leaves were labeled on various dates during inflorescence, and fruitlet development, and distal bud expansion (January through June).


116


42
typical of avocado, the mass and volume data suggested it follows a normal pattern of growth and did not exhibit cell division and expansion
throughout development.


Al
partitioned normally into fruit which abscised 7 or more days after labeling. Radiolabeled assimilates were detected in fruit abscising between 4 and 7 days post-labeling although the distribution of label to the fruit tissues differed from control fruit. No label was recovered from the tissues of fruit which abscised within 4 days of labeling. This suggests that the initiation of senescence altered fruit assimilate import during a 4 to 7 day period prior to abscission; however, growth rates did not decrease until 4 to 5 days before fruit separation. The restriction of assimilate movement to fruit at 4 days before fruit separation agrees with our fruit growth data and supports the contention of Davenport and Manners that the cessation of fruit growth is due to the initiation of senescence and its accompanying phenomena.
This study was part of a larger effort to understand carbohydrate availability and abscission in avocado. Abscising fruit were indistinguisable from retained fruit until 4 to 5 days before fruit separation. Ve suggest that fruit abscission resulted from an initial senescence event perceived by the fruit 5-7 days before actual separation. This event initiates nucellar senescence and ethylene evolution. Ethylene induced cell division in the abscission zone restricted the movement of assimilates and water into the fruit causing a decrease in the rate of growth. Cessation of fruit growth therefore was the result and not the cause of abscission. Local source limitation also did not appear to be a likely factor in fruit abscission since the probability of abscission was the same for all fruit whether they were present on a branch singly, in pairs, or other multiples. Finally, although cv. 'Petersen' exhibited a single sigmoidal growth curve


88
change with leaf flush age (Chap.7,8). The leaves located proximal to the inflorescence supply assimilate to the developing leaves and fruit and to the roots. As the leaves developing distal to the inflorescence mature, assimilates from the leaves proximal to the inflorescence no longer move acropetally but move primarily towards the roots. When matured, leaves immediately distal to the inflorescence supply assimilate to the fruit, another developing flush of leaves and to the roots. Vhen the final flush of leaves has matured all of the translocated assimilates move basipetally toward the fruit and roots.
The study of source/sink relations would not be complete without examining how assimilates were distributed within the reproductive organs and if the reproductive organs could provide enough assimilates for their own growth. The photosynthetic capabilities of the reproductive organs of plants vary drastically (148,183). Although most of these studies have been done in agronomic crops, tree crops appear to exhibit the same diversity. Citrus flowers contribute very little photosynthate to their growth (13,186,196), while apple flowers (88) contribute significant photoassimilate to their growth. Although the flowers' role as a source of photoassimilate varies with the plant, its status as a weak sink is well established (199).
The RSA (%dpm Xdry mass"1), sink intensity (Xdpm mg dry mass"1 and sink strength (%dpm) values for flower buds all peaked when the buds were of intermediate size. The sink intensity (8.1 %dpm mg dry mass 1 and RSA (2.2 %dpm X dry mass"1) of small buds was also relatively high compared to the more developed elongated buds (4.5 Xdpm mg dry mass"1,1.23 %dpm X dry mass"1). The most actively assimilating tissues were the younger less developed flower buds. As the floral structures


89
developed (elongated hud stage) the RSA, sink intensity and sink strength decreased. Similar data were reported in apple (88). Apple flower buds were most active in fixing C02 and in accumulating translocated label in floral stage c (flowers are open in stage e). High sink activities have also been noted prior to anthesis in soybean (23). In soybean this high sink intensity is associated with the maturation of the embryo sac.
The tissues of the open flower were supplied assimilates by the leaves proximal to the inflorescence (Table 5-1). Partitioning of assimilate to the nectaries was accentuated, as indicated by the high RSA (3.17 %dpm %dry mass-1), during the 24 hour period before the flower initially opened. The increase in label in the nectaries probably resulted from the accumulation of sugars in the nectar. Tissues of flowers that were post-anthesis at harvest were weaker sinks than open flowers i.e., flowers post-anthesis accumulated less label than open flowers. Flower abscission at this time was very high. The one exception was the corolla of the post-anthesis flowers which had the greatest increase in the assimilation of radiolabel (RSA = 3.6 %dpm %dry mass"1) of any tissue examined. The increased sink intensity of the corolla post-anthesis may have reflected the initiation of senescence events. During the 24 hour period post-anthesis the petals and sepals of the corolla elongated and changed in color from a light
green to a pale yellow. Davenport (45) has shown that ethylene
i
production by avocado flowers increased nearly 20-fold within 24 hours post-anthesis (male stage). Similar climacteric-like rises in ethylene have been observed in rose petals (63) and carnation (142,143). The increase in ethylene production is followed by increased membrane


21
The profile of the radiolabel distribution at zero hour for the
laminate extract is shown in Fig. 2-1. The majority of disintegrations
per min (dpm) were found at the sucrose spot. Nonstaining metabolic
intermediates account for the remainder of the extracted label. No
radioactivity was present in the vein extract at this time point. At 1 hour, however, all of the ethanol extractable radioactivity was associated with the sucrose spot and radiolabel was detected in the vein extracts for the first time (Table 2-2). The radiolabel in the vein
extracts was associated with the sucrose spot. During the 4 hour course
of this experiment the proportion of sucrose in the laminate and vein
extracts remained the same as did the proportion of radiolabel after 1 hour, indicating that the system was not disrupted by the labeling
process and that the sample leaves were physiologically comparable.
During the winter months (Nov. to Jan.) recently fixed C02 was incorporated into both sucrose and mannoheptulose. The incorporation of
label into either the sucrose or mannoheptulose pool was dependent on temperature (Fig. 2-3). At low temperatures (22C) more label was incorporated into sucrose than into mannoheptulose (Fig. 2-3); more
sucrose than mannoheptulose was also produced under these conditions
(Fig. 2-4). These results were similar to those reported by Bean et al.
(12) at a similar temperature. At the ambient temperature for that time
o 14
of year (27 C), similar amounts of radiolabeled C02 were incorporated into sucrose and mannoheptulose, and they were being produced at similar
rates. At a higher temperature (37C), the amount of label recovered in
sucrose decreased although sucrose was still preferentially labeled over
mannoheptulose (Fig. 2-4).


This dissertation was submitted to the Graduate Faculty of the College of Agriculture and to the Graduate School and was accepted partial fulfillment of the requirements for the degree of Doctor of Philosophy- /)
U / >
May 1990
Dean, College of Agricultvfji
Dean, Graduate School
I


Figure 2-3. Distribution of C in Sugars of Avocado Leaves at Various Temperatures. Leaves were incubated at three temperatures and fed
CO At the end of the incubation period the leaves were macerated and foiled in 85% EtOH. Leaf extracts were spotted on TLC plates. The plates were developed in a chloroformracetic acidrwater solution. The plates were removed from the developing solution and scraped for sugar extraction and scintillation spectroscopy. The silica gel scrapings were extracted with 500 ul of water. Sugar content was measured by the phenol-sulfuric acid method. Radioactivity was measured by scintillation spectroscopy. The histogram bars indicate the relative percentage of total leaf dpm recovered from the the silica gel corresponding to the location of the sugar.


217
13. Bean RC, GV Todd (1960) Photosynthesis and respiration in
developing fruits. I. CO uptake by young oranges in light and in dark. Plant Physiol 35:^25-429
14. Bennett AB, BL Sveger, RG Spansvick (1984) Sink to source
translocation in soybean. Plant Physiol 74:434-436
15. Biale JB, RE Young (1971) The avocado pear. In: the biochemistry
of fruits and their products. Food Science and Technology, ed. A.C. Hulme. Vol. 2: 1-63. Academic Press, New York
16. Biddulph 0, R Cory (1965) Translocation of 14C-metabolites in the
phloem of bean plant. Plant Physiol 40:119-129
17. Blumenfeld A, S Gazit (1969) An endogenous inhibitor of auxins and
kinetin. Israel J Bot 18:217-219
18. Blumenfeld A, S Gazit (1970) The role of the seed coat in avocado
fruit (grovth and maturation). Yrbk Calif Avoc Soc 54:100-104
19. Blumenfeld A, S Gazit (1974) Development of seeded and seedless
avocado fruits. J Amer Soc Hort Sci 99:442-448
20. Blumenfeld A, S Gazit and E Argaman (1983) Factors involved in
avocado productivity. Spec Publ No 222:84-85 Volcani Center, Israel
21. Bogen E, A Wallace (1966) CO Fixation in preparations from
Tunisian sweet lemon and Eureka lemon fruits. Proc Am Soc Hortic Sci 88:298-307
22. Borochov A, AH Halevy, M Shinitzky (1976) Increase in
microviscosity vith aging in protoplast plasmallemna of rose petals. Nature 263:158-159
23. Brun WA, KJ Betts (1984) Source/sink relations of abscising and
non-abscising soybean flowers. Plant Physiol 75:187-191
24. Bukovac MJ (1971) The nature and chemistry of promotion of
abscission in maturing cherry fruit. HortSci 6:385-388
25. Burrel MM, P Brunt (1981) Preparation of green plant material for
liquid scintillation counting. Ann Bot 48:395-97
26. Cameron SH, G Borst (1938) Starch in the avocado tree. Proc Am Soc
Hort Sci 36:255-258
27. Cameron SH, RT Mueller, A Wallace (1952) Nutrient composition and
seasonal losses of avocado trees. Yrbk Calif Avoc Soc 36:201-209
28. Canny MJ (1973) Phloem translocation. Cambridge University Press,
Cambridge, England


19
Temperature Studi es
The most recent flushes of leaves evident during Nov., Dec. and Jan. were cut from the tree and transported into the laboratory. Three leaves of approximately the same age were cut from the branch and each suspended in a 2 liter beaker. The beakers were half filled with ice, hot water (37C) or water at ambient temperature (27C). A vial containing 3 mL of 5% lactic acid was taped to the inside of the beaker. The beakers were covered with Saran wrap and allowed to equilibrate for 30 min, 60cm below a 400 watt lamp. Leaf temperatures were measured before and after the labeling period. After the equilibration period,
14
10 uCi of NaH CO was added to the lactic acid. One hour later, the
3 9
leaf temperature was measured again and the Saran wrap was removed. The leaves were minced and boiled in 85% ET0H as described earlier. The alcoholic extracts were used to spot TLC plates. The experiment was repeated three times. Sugar Separation
Preparative TLC plates were spotted with control sugars and streaked with 100 ul of leaf extract and then developed in a chloroformracetic acidrwater solvent. The control strips were sprayed with aniline diphenylamine in acidified acetone. The areas of the unstained plate corresponding to the Rf of sucrose and mannoheptulose were scraped. The silica gel scrapings of each sugar were each extracted with 500 ul of water. The suspension was then centrifuged to sediment the silica gel. The sugar content of the supernatant was quantitated using the phenol-sulfuric acid method of Dubois et al. (53). Radioactivity in an aliquot of the supernatant was measured by liquid scintillation counting.


163
GIRDLED NON-GIRDLED
BRANCH TYPE
FLUSH 3
FLUSH 2
FLUSH 1


2 0mm
6.6%


5
develop, but the embryo remains small. Cell division during this period is very rapid and is marked by a homogeneous cell population size. During the second stage of growth the embryo develops, but the fruit overall does not enlarge. The third phase begins when the embryo stops growing and the fruit continues to grow. When the fruit has reached approximately 1/2 of its final size, the cells have also reached their final dimensions (average cell size of pericarp cells is 50-60 microns). Unlike most other fruits whose cells simply enlarge during phase three, the avocado cells continue to divide until harvest.
The gross development of the avocado is represented by a sigmoidal curve (167,169). However, when examined more closely, the growth of the avocado has a diurnal cycle (171) in which the fruit contracts during the day and swells at night getting progressively larger each day.
The avocado fruit is a berry with one carpel and contains a single seed (15). The pericarp consists of three layers: the exocarp (skin and rind), fleshy mesocarp (edible portion of the fruit), the endocarp (thin inner layer next to the outer seed coat). The exocarp is the outer layer of the fruit and is covered by a wax-like cuticle. The exocarp is composed of one layer of epidermis, 1-3 layers of hypodermal cells, several layers of parenchyma cells and a layer of scelerenchyma cells which line the inner surface of the peel. The mesocarp is made up of uniform isodiametric parenchyma cells. Oil cells or idioblasts are spread throughout the mesocarp and can be identified by their large size and the lignified walls. The mesocarp is well vascularized in an


18
Thin Layer Chromatography
Silica FG plates were pre-activated by heating at 110C for 1 hour. The cooled plates were spotted with aliquots of control sugars (glucose, fructose, sucrose, mannoheptulose, perseitol, galactose, sorbose, ribulose, sorbitol), laminate and vein extracts, invertase treated sugars and extracts and phloem exudates. The plates were air-dried and enclosed in a tank with a developing solution consisting of either chloroformracetic acid:water (3:3.5:0.5) or acetone:water (9:1). The plates were removed when the solvent front had moved at least 10 cm. They were air dried for 30 min, then sprayed with aniline diphenylamine in acidified acetone (Sigma) and heated for 30 minutes at 80C to enhance color development. Aniline diphenylamine is a differential spray reagent for the identification of non-reducing carbohydrates. Pentoses appear as red spots. Methylated aldopentoses appear cherry red. Aldohexoses appear brown. Hexauronic acid and methylated aldohexoses appear brownish red. Fully methylated aldohexoses appear maroon. Additionally, some plates were sprayed with concentrated H S0^ and heated at 80C for 30 min to detect any nonstaining organic components.
One-half of each plate was sprayed with the color developer. The unstained half of the plate with identical samples was scraped for scintillation counting. The Rf of radioactivity was determined by scraping 1 cm squares extending from below the origin to the solvent front of the TLC plate. The Rf and color of radioactive and stained spots of the extracts were compared to the stained standards for identification of the putative sugars in the extract.


% FRUIT REMAINING ON TREES
(% of initially monitored fruit)


2
brlov -?.8oC and killed by temperatures of -4.4C (15). The fruit is larger than that of the Mexican race but smaller than avocados of the Guatemalan race. The peel is smooth, leathery and sometimes glossy.
The Guatemalan cultivars are intermediate between the other races with respect to their climatic adaptation. This race is prevalent in the Central American highlands (1500-1800 m) (15). Here it grows under cool temperatures but is likely to be killed by temperatures less than -6.1C. The fruit weighs between 340 and 600 gr and it has a thick and brittle skin. The fruit matures later than either of the other two races.
The tree architecture of the genus Persea according to Rauh's model is determined by a monopodial trunk which grows rhythmically and so develops tiers of branches (195). The branches are morphogenetically identical to the trunk. The meristems are not equivalent, which leads to the differentiation between the trunk and the branches. However, branch development is closely correlated to the rhythmic growth of the axes. Branching is by syllepsis with the lateral axis developing during the phase of extension of the terminal bud on the current shoot. The branch tier then becomes more diffuse. The tree has a shallow root system and the leaves are arranged in spirals and come out in flushes.
The flowers are lateral and do not affect the growth of the shoot system. The inflorescence is consistently lateral or pseudoterminal. The flowers develop on the first part of the renewal shoot and their position as lateral appendages is soon evident as the distal leafy shoot elongates. The inflorescences appear by the thousands with each inflorescence bearing hundreds of flowers.


CHAPTER 9
TRANSLOCATION OF RECENT PHOTOASSIMILATES IN FRUIT-BEARING FLUSHES IN AVOCADO: PERSEA AMERICANA MILL. CV. 'PETERSEN'
Introduction
Flowering and fruiting have frequently been examined in relation to the carbohydrate balance of the plant (76,83,118,162). The changing seasonal levels of carbohydrates in plants has often been used to explain phenonena such as alternate bearing, flower initiation, and
flower and fruit abortion. In trees, the seasonal fluctuations in starch content exhibit a strong inverse correlation to the fruit load on the tree (75,76,162). It still has not been determined whether or not these seasonal starch fluctuations are the cause or coincidental with premature fruit abscission and related phenomena.
Many studies employ sink manipulation via girdling, shading or fruit removal to alter source sink relations (29,34,36,39,43,68,82,83, 92,107,124,132,134,137,145,159,177,180,181,190,198,200,201). Girdling has been shown to increase floral initiation and yields (34,98,99,107, 162) in several tree crops. This increase in productivity has been attributed to the accumulation of carbohydrates (134) and/or growth regulators (38,68) above the girdle. The role of fruit as a moderator of carbohydrate content has also been examined. Carbohydrates accumulate in the chloroplast when fruit are removed (93,94,99,161). Accumulation of starch may physically distort the chloroplast causing a decrease in C02 assimilation (140). This decrease in net C02 has been noted by many workers (39,43,92,93,118,140,145). Several authors
180


195
CO
D J
PL, O
a.
CO
w
<
CO
CO
<
CJ
I
a
>
o
< o
Eh
O
45.0
40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
\
V X
A-
' / / /
/
/
/
/
/
/
/
/
/
/
4_
girdled +fruit
non-girdled +fruit
defruited
branch type
flush 3
flush 2
flush 1


179
160
140
120
100
80
60
40
20
0
\\\
l X X X
X X x X
///.
/
/
J
/
/
/
/
/
/
girdled
non-girdled
branch type
flush 3
flush 2
flush 1


7 7c
28 7c
^^^^^^^^^^
2 9 %


106
dry mass distribution. At this stage the inflorescence had enlarged, accounting for most of the dry mass increase in the reproductive tissues.
During the period when midvein lengths were 30 to 50 mm (Fig. 6-3), changes in assimilate movement from the proximal source leaves suggested the sink to source transition (8,29) of the distal leaves was occurring. Total export of radiolabel from the source leaf decreased from 59% to 48%. The proportion of label traveling to the distal tissues also decreased by 27%. Reproductive tissue dry mass represented 75% of the expanding biomass.
When the average distal midvein length was greater than 50 mm the distal leaves were no longer sinks for assimilate from the proximal leaves. Very little radiolabel (.145% of the translocated label) was recovered in the distal tissues; however, the proportion of label in fruitlet and leaf tissues again paralleled the dry mass ratios (Fig. 6-4). Fruitlets at this stage represented 51.6% of the total mass (Table 6-1) and contained 47.4% of the label translocated acropetally.
This sink-to-source conversion was paralleled by a decrease in export from the older proximal leaves (Fig. 6-4). In the early stages of flush development (Figs. 6-1 and 6-2), the proximal leaves exported greater than 55% of their recently fixed photoassimilates in 48 hours. When the average midvein length was between 30-50 mm the export of assimilates from the proximal source leaves decreased to below 50%. The amount of label going to the distal tissues also decreased. In the final stage of sink to source transition, when the average midvein


LN VOLUME (cc~3)


Figure 3-2. Volume vs Dry Mass of Avocado Fruit. Avocado fruits were harvested periodically during the season for dry mass measurements. The increase in mass with diameter paralleled the increase in volume predicted by a spherical fruit model. Each point represents a single measuremnt.


109
apple (17,109) and orange (121) where mature leaves failed to import radiolabeled assimilates. The distal leaves do in fact become sources of assimilate for the fruitlets when their midvein length is between 35-45 mm (Chap. 7). Upon maturation of the leaves distal to the avocado inflorescence, even more assimilate was available; however, fruitlet abscission continued long after maturation of the distal leaves as has been described elsewhere (Chap. 2).
Fruitlet and developing leaves did not compete for recent photoassimilates. The amount of radiolabel assimilated by a tissue was directly proportional to the dry mass of the tissue i.e., the
mobilizing ability of floral and vegetative tissue was the same at the level of the whole fruitlet and flower (Table 1). Floral and
vegetative sinks in avocado have also been shown to have similar sink strengths for nitrogen (208). Floral tissue dry mass and leaf dry mass were about the same during the first stage (midvein length <20 mm). During the next two stages fruit mass was 2-3 times greater than leaf mass. Although the developing floral tissues were growing rapidly, no difference was seen in the specific activities of the two different sinks. This suggests that neither tissue was limited for assimilates and that both tissues were accumulating assimilates at their maximum rate.
In conclusion, the accumulation of recent photoassimilates by the developing floral and vegetative tissues was mass dependent. During the early stages of reproductive flush development carbohydrate availability appeared to be sufficient to support the growth of leaves and fruitlets and did not limit fruitlet growth or stimulate fruitlet abscission.


ASSIMILATE PARTITIONING IN AVOCADO, PERSEA AMERICANA MILL.
By
SUSAN F. FINAZZO
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
1990


CHAPTER 2
THE TRANSLOCATED SUGARS IN PERSEA AMERICANA MILL.
Introduction
Phloem exudates of some members of the Lauraceae family have been shown to contain several sugars. These include sucrose, raffinose,
stachyose, verbascose and myoinositol (211). The major component of these transported sugars is sucrose. Zimmerman and Ziegler (211) identified sucrose as the only sugar translocated in the phloem of avocado, a member of the Lauraceae. Recent unpublished results, however, have led Blumenfeld (BARD project #1-414-81 Final Report) to suggest that the C7 sugars, perseitol and manoheptulose are the major translocated sugars.
The occurrence of mannoheptulose in leaves and fruit is rare although not unique to avocado (30,203,205,210,211). Mannoheptulose was first identified in avocado fruit by LaForge (123). During avocado leaf photosynthesis labeling of sucrose is rapid and linear (11,12) while mannoheptulose is labeled at a linear rate one-third that of sucrose.
Mannoheptulose appears to be a normal product of photosynthesis (11,12,146) generated from mannoheptulose phosphate by a phosphatase. Nordal and Benson (146) isolated mannoheptulose and mannoheptulose monophosphate from the leaves of avocado and suggested that a slow enzymatic exchange occurred between free and phosphorylated mannoheptulose.
14


TABLE OF CONTENTS
ABSTRACT...........................................................v
CHAPTER 1 GENERAL BACKGROUND AND HISTORY.........................1
Research in Avocado..................................6
Abscission and Partitioning..........................9
CHAPTER 2 THE TRANSOCATED SUGARS IN PERSEA AMERICANA MILL.......14
Introduction........................................14
Materials and Methods...............................16
Results and Discussion..............................20
CHAPTER 3 GROWTH OF ABSCISING AND PERSISTING AVOCADO FRUIT......33
Introduction........................................33
Materials and Methods...............................35
Results.............................................36
Discussion..........................................38
CHAPTER 4 ASSIMILATE PARTITIONING BETWEEN ABSCISING AND
PERSISTING AVOCADO FRUITS.............................58
Introduction........................................58
Materials and Methods...............................60
Results.............................................61
Discussion..........................................64
CHAPTER 5 SINK/SOURCE RELATIONS IN THE AVOCADO INFLORESCENCE----79
Introduction........................................79
Materials and Methods...............................80
Results.............................................83
Discussion..........................................87
CHAPTER 6 DISTRIBUTION OF RECENT PHOTOASSIMILATES IN AVOCADO:
THE ROLE OF PROXIMAL LEAVES..........................102
Introduction.......................................102
Materials and Methods..............................103
Results............................................104
Discussion.........................................107
CHAPTER 7 SINK TO SOURCE TRANSITION OF DEVELOPING LEAVES
LOCATED DISTAL TO AVOCADO INFLORESCENCES.............119
Introduction.......................................119
Materials and Methods..............................120
Results............................................122
Discussion.........................................125


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 EHULBE2SX_N2C92D INGEST_TIME 2017-07-11T21:49:19Z PACKAGE AA00002111_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


228
175. Sharon-Paber 0, Kahn V (1983) Avocado mesocarp, browning
potential, carotenoid content polyphenol oxidase, catalase and peroxidase activites: comparison between six avocado cultivars. J Food Sci 48:1874-1875
176. Shiroya M, GR Lister, CD Nelson, G Krotkov (1961) Translocation of
C in tobacco at different stages of development following assimilation of CO^ by a single leaf. Can J Bot 39:855-64
177. Silvious JE, NJ Chatterton, DF Kremer (1979) Photosynthate
partitioning in soybean leaves at two irradiance levels. Plant Physiol 64: 872-875
178. Stead AD, KG Moore (1979) Studies on flower longevity in
digitalis: pollination-induced corolla abscission in digitalis flowers. Planta 146:409-414
179. Stout AB (1927) The flower behavior of avocados. J New York Bot
Gard 7:145-203.
180. Streeter JG, DL Jeffers (1979) Distribution of total
non-structural carbohydrates in soybean plants having increased reproductive load. Crop Sci 19:727-734
181. Stutte G, GC Martin (1984) Effect of carbohydrate reserves on
flower formation in Plea europaea L. Plant Physiol 75:136
182. Tamas IA, DH Wallace, PM Ludford, JL Ozbun (1979) Effect of older
fruits on abortion and abscisic acid concentration of younger fruits in Phaseolus vulgaris L. Plant Physiol 64:620-622
183. Thorne GN (1965) Photosynthesis of ears and flag leaves of wheat
and barley. Ann Bot 29:317-329
184. Thorne JH, HR Koller (1974) Influence of assimilate demand on
photosynthesis, diffusive resistances, translocation, and carbohydrate levels of soybean leaves. Plant Physiol 54:201-207
185. Thrower L (1962) Translocation of labelled assimilates in soybean
II. The pattern of translocation in intact and defoliated plants. Aust J Biol Sci 15:629-50
186. Todd GW, RC Bean, B Propst (1961) Photosynthesis and respiration
in developing fruits. II. Comparative rates at various stages of development. Plant Physiol 36:69-73
187. Tomer E, M Gottreich (1978) Abnormalities in avocado (Persea
americana Mill) ovule development. Bot Gaz 1:81-86
188. Tomer E, M Gottreich, S Gazit (1976) Defective ovules in avocado
cultivars. J Amer Soc Hort Sci 101:620-623


Figure 8-10. Specific Activity of Starch Extracted from Avocado Leaves of Three Successive Flushes on Girdled and Non-Girdled Branches. Leaf disks (0.618 cm2) from each flush vere pooled, chopped and boiled in 85% EtOH. The residues of the ethanolic extracts vere digested by the method of Housely et al. (102) and the starch quantitated by the phenol-sulfuric acid method. Samples of knovn starch content vere used to quantify radiolabel.


UNIVERSITY OF FLORIDA
3 1262 08394 222 6
A31262083942226A


193
Table 9-2. Time Course of Changes in the Relative Pool Sizes of Soluble Radiolabeled Compounds From Avocado Leaves.
Samples were taken from leaves at various time points and processed by the method of Housely et al. (102). The neutral fraction represents the sugars; the basic fraction consists of amino acids; organic acids and sugar monophosphates are included in the acid 1 fraction; phosphoglyceric acid and sugar diphosphates are included in the acid 2 fraction. These results are presented as percentages of the total radioactivity recovered from the columns. Time is given in hours post-labeling.
Time Neutral Basic Acid 1 Acid 2
(hr post-labeling) (% of total radiolabel recovered from column)
0 90.4 + 2.5 5.3 + 3.8 3.6 + 3.0 0.5 + 0.3
24 88.8 + 5.0 7.1 + 4.9 2.8 + 1.7 1.2 + 0.9
52 72.5 + 7.5 5.2 + 2.7 6.8 + 5.7 5.2 + 4.7
78 66.7 + 0.7 19.4 + 5.5 8.2 + 3.9 5.8 + 2.2


81
Entire inflorescences or fruit were radiolabeled at several times during reproductive development. A schematic representation of an early stage of reproductive development is shown in Fig. 5-1. In this figure, leaves located distal to the inflorescences are at the center. Four individual, cymose inflorescences are shown inserted into the main axis of the stem. The older (proximal) leaves are located basipetal to the inflorescences. Individual inflorescences may include small (round <2mm), intermediate (round 2-3.5mm), elongated flower buds (elongate >3.5 mm), as well as open flowers, post-anthesis flowers (opened 4 or 24 hours earlier), and fruits. A total of 434 small flower buds, 350 medium flower buds, 270 elongated flower buds, 101 open flowers, 113 post-anthesis flowers and 25 fruitlets were evaluated in this study. Flower Sink Strength
Single proximal leaves were labeled as previously described for the studies on reproductive structure sink strength. The branches composed of the distal leaves, inflorescences, and proximal leaves were removed and dissected into their component parts e.g., leaf, flower (corolla, stamens, pistil, nectaries and staminodes) 24 hours after labeling. The tissues were dried, weighed and digested with a commercially available tissue solubiliser (Fisher; Scintigest) as described by Burrel and Brunt (25). CO^ Fixation by Floral Structures
After removing leaves developing distal to the inflorescence, the flowering panicles were enclosed in plastic bags and fed with 20 uCi of
14
C02 for a one hour period. Branches bearing inflorescences and


7
ripening depends upon both the stage of development of the fruit and the cultivar. If the fruit is not picked, it may be 'stored' on the tree from anywhere between three months and one year.
The avocado is also unique in that it undergoes continuous drop throughout its development. In other fruits such as apple and orange a period of abscission follows anthesis (June drop). Fruit remaining on the tree through this period continues on to maturity. In avocado, depending on the cultivar, approximately 85% of the fruit set will abscise during the season (148). However, they do not abscise during a
limited period but may abscise anytime from immediately post anthesis until harvest. The fruitlets that abscise are characterized by a discoloration of the seed coat and mesocarp (1,10,18,47), by an increase in ethylene production (1), an increase in respiration (14), and an increase in peroxidase isoenzymes and peroxidase activity (8,9,114,115,116,174,175).
As avocado fruitlets mature their respiration decreases (4). Ethylene production also decreases. Ethylene can no longer be detected in the fruitlet once it reaches approximately 10% of its final mass. The initial burst of ethylene is thought to be stimulated by pollination (4) or result from the ongoing abscission process (46).
An increase in respiration and ethylene production occurs when the fruit is removed from the tree (1,46). Fruitlets at all developmental stages show a similar climacteric pattern when harvested (205). However, the postharvest preclimacteric period shortens as the fruit develops. Seed respiration is high in young fruits and decreases with
fruit development (205).


CHAPTER 3
GROWTH OF ABSCISING AND PERSISTING AVOCADO FRUIT
Introduction
Although the growth of avocado fruit has been well characterized (18,80,133,167,169,194,207), little work has been done on the growth kinetics of abscising fruits. Avocado varieties exhibit one of two possible fruit set patterns (44). Cultivars which exhibit the Type II
fruit set habit initially set few fruit which are retained until harvest. Type I fruit set pattern involves an initial heavy fruit set followed by an extended period of abscission. Typically up to 20% of the flowers will form fruitlets but only 1 to 7% of the fruit remain at harvest (44,148). This latter pattern is typical of the 'Petersen' cultivar used in this study and of the 'Hass' cultivar. Both are utilized in commercial production.
The causes of fruit abscission are unknown. Sedgely (172) has shown that 90% of the fruitlets which abscise within one week of flowering were unfertilized. Fruitlets which abscised after this period were fertile and anatomically normal (166,187,188,194). Recent studies (48,74) have suggested that genetic selection and outcrossing may play an important role in determining which fruitlets abscise.
Abscission of avocado fruitlets is characterized by the evolution of ethylene and the deterioration and browning of the seed coat (18,47) in an otherwise anatomically normal fruitlet. Davenport and Manners (47) have shown that the burst of ethylene produced by the seedcoat
33


BIOGRAPHICAL SKETCH Susan F. Finazzo, daughter of Carlo and Odessie Finazzo, was born in Orlando, FL on May 13,1957. She attended Caesar Rodney High School in Camden-Wyoming, DE. After graduation, she began studies at the University of Delaware. She graduated in 1979 with a Bachelor of Arts degree in biology and geology. In the fall of that year she began graduate studies at the Pennsylvania State University. She graduated from the Pennsylvania State University in 1982 with Master of Science degree in microbiology.
231


199
< <
<
J o
<
e
6
0.0150
0.0135
0.0120
0.0105
0.0090
0.0075
0.0060
0.0045
0.0030
0.0015
0.0000
\
\
\
/
/
////
////
\\
\
\
\
\
/
?
/A
'///*
/////
#56
girdled +fruit
non-girdled +fruit
defruited
branch type
flush 3
flush 2
flush 1


177
X
CO
D
x: s x:
W CO
O ^ U
EC
cj en
CO
45
40
35-
30
25
20
15
10 -
5
0
\ X
X
X
X
X
X
\
X
\
V
XX\
X V
X V
X
/
/
/
/
/
GIRDLED
NON-GIRDLED
BRANCH TYPE
FLUSH 3
FLUSH 2
FLUSH 1


107
length was 50 mm or greater, the proximal leaf exported only 31.4% of its recently fixed photoassimilates. Almost all of this label traveled basipetally.
Discussion
The concept of competition between individual fruitlets and leaves as an explanation for premature fruit abscission is well entrenched in the literature (35,36,76,138,162,191,197). Carbohydrates play an integral role in this explanation since fluctuation in plant starch content at times appears to parallel fruit yield. In fact treatments which alter starch content, such as girdling (35,36,107,124,132,138, 191, 197) often increase yields. These types of data are frequently used to support the notion that carbohydrate limitation induces abscission.
Scholefield et al. (162) noted seasonal fluctuations in starch content of avocado. They suggested that low yields were the result of competition between the developing vegetative terminus and the
fruitlets. Their data indicate, however, that the peak in starch content coincides with the peak in fruit and flower abscission. In this case starch, availability did not affect abscission. Conflicting data have been presented on the importance of the starch content in biennial bearing citrus trees. Some workers (76) have described significant correlations between the starch content and the bearing status of trees, while others (112,130) found no clear correlation between carbohydrate levels and tree productivity. The evidence in support of stored carbohydrates as the tangible link to fruit abscission is strictly correlative.


84
The data from proximal leaf labeling experiments are given in Table 5-1. The RSA expresses the relationship between the partitioning of current photosynthate and the matter previously allocated to that particular organ. Any deviation of the RSA from unity indicates a change in the growth pattern of the tissue. The competitive ability of a tissue can also be expressed by the sink intensity. Small and intermediate buds incorporated almost twice as much of the translocated labeled assimilate (dpm mg"1) as did the elongated buds (Table 5-1). Similarly, the sink intensity and RSA of small (8.1 %dpm mg"1, 2.20 %dpm %mg 1) and intermediate buds (9.4 %dpm mg"1, 2.6 %dpm %mg-1) were very similar to each other but were almost 2-times that of the elongated buds (4.5 %dpm mg"1, 1.23 %dpm %mg_1). In fact, intermediate buds were growing more rapidly as measured by the RSA (2.6 %dpm Xmg x) than the majority of the tissues examined. Flower bud growth as measured by sink intensity and RSA increased until the elongated bud stage when both sink intensity and RSA decreased dramatically consistent with impending flower opening.
Partitioning of the translocated labeled assimilates occurred within the flower (Table 5-1). Although all of the floral tissues actively incorporated radiolabeled assimilates, the nectaries of the open flowers exhibited the greatest sink intensity (11.6 %dpm mg 1) and RSA (3.17 Xdpm Xmg"1). The partitioning of labeled assimilates in the post-anthesis flower differed from that observed in open flowers. The corolla exhibited the greatest sink intensity and RSA in post-anthesis flowers. The sink strength, sink intensity and RSA of the pistil,


95
Table 5-3. Relative Specific Activity of Inflorescence Components.
Labeled inflorescences were harvested immediately after labeling and divided into the groups for processing. At the 95% confidence level the dry mass of male and female flowers are significantly different.
Male Flower Stage
Tissue Mean Dry Mass X dpm/mg X dpm/Flower
Whole Open Flower .0051 + .00037 17.7 26.7
Mature Flower .0052 + .00059 5.7 8.5
Small Bud <3mm .0022 + .00016 40.5 27.2
Medium Bud >3mm .0035 + .00029 36.1 37.5
Female Flower Stage
Tissue Mean Dry Mass X dpm/mg X dpm/Flower
Whole Open Flower .0042 + .00006 12.6 19.2
Mature Flower .0039 + .00001 10.8 13.7
Small Bud <3mm .0019 + .00001 45.0 32.3
Medium Bud >3mm .0031 + .00013 31.6 34.8
r


Figure 4-1. Distribution of C in Pierced Single Fruits. Source leaves on branches bearing one fruit were fed CO for a one hour period. The seed coat of the fruit was pierced by a surface sterilized stainless steel wire before (48,24 hrs), after (24, 48 hrs) or at the time of labeling (0 hr). The fruit were harvested 72 hrs after source leaf labeling, dissected, dried and processed for scintillation counting. Six branches were used per experiment, each of which was repeated 4 times. Histogram bars represent the percentage of total radioactivity in the plant tissue. Data was transformed by arcsine transformation for statistical analysis. Mean separation within columns by Tukey's studentized range test (P<0.05).


11
1 4
dropped dramatically. Sugars comprised 70~R6% of the C-lnbel of the
source leaf's soluble assimilates. The highest values occurred during late flowering and early pod filling stages.
Variation in the levels of assimilate accumulation had also been
reported (23) between abscising and non-abscising soybean flowers and
Mandarin orange (69). The sink intensity was very high in both
abscising and non-abscising flowers before anthesis. During the first
three days post-anthesis, the sink intensity decreased and remained low. In normally setting flowers the sink strength then began to
increase while abscising flowers failed to recover their sink strength.
Reproductive abscission in soybean is determined at a very early stage Photosynthate accumulation correlates to reproductive abscission.
However, whether variations in assimilate accumulation are a cause or
an effect of abscission has yet to be determined.
In tomato (94), assimilate accumulation in the fruit was dependent
on the developmental stage of the fruit and was independent of the rate
of the C02 fixation in the source leaf. Translocation of photosynthate was regulated by sink demand. However, soybeans given supplemental light exhibited both reduced abscission and increased seed weight/node
when compared to natural light controls. Shading of flowers and pods increased abscission and reduced seed weight. Shading of the flowers and pods resulted in a decrease in 14C-assimilates from the source
leaf. As the nodes were shaded the abscission percentage increased,
even when supplemental lighting was provided. This suggested that
flowers and pods play a role in regulating both their own assimilate
accumulation and their own abscission via a perceived light stimulus


222
87. Hansen P (1970) C-studies on apple trees. VI. The influence of
the fruit on the photosynthesis of the leaves, and the relative photosynthetic yields of fruit and leaves. Physiol Plant 25:805-815
88. Hansen P (1971) 14C-Studies on apple trees. VII. The early
seasonal growth in leaves, flovers and shoots as dependent upon photosynthates and existing reserves. Physiol Plant 25:469-473
89. Hansen P, J Grauslund (1973) 14C-studies on apple trees. VIII. The
seasonal variation and nature of reserves. Physiol Plant 28:24-32
90. Hartig T (1860) Beitrage zur physiologischen Forstbotanik. Allg
Forst Jagdztg 36:257-300
91. Hatton TT, PL Harding, VF Reeder, JN Yeastman, VH Krome (1963)
Fruit veights and corresponding diameters for Florida avocado. USDA, AMS 515:11
92. Heindl JC, VA Brun (1983) Light and shade effects on abscission
and C-photoassimilate partitioning among reproductive structures in soybean. Plant Physiol 73:434-439
93. Henson IE, V Mahalakshmi, G Alagarsvamy, FR Bidinger.(1983) An
association betveen flovering and reduced stomatal sensitivity to vater stress in pearl millet (Pennisetum americanum L. Leeke). Ann Bot 52:641-648
94. Ho LC (1979) Regulation of assimilate translocation betveen leaves
and fruits in the tomato. Ann Bot 43:437-448
95. Ho LC, RG Hurd, LJ Ludvig, AF Shav, JHM Thornley, AC Withers
(1984) Changes in photosynthesis, carbon budget and mineral content during the grovth of the first leaf of cucumber. Ann Bot 54:167-181
96. Ho LC, AJ Peel (1969) Transport of 14C-labelled assimilates and
P-labelled phosphate in Salix viminalis in relation to phyllotaxis and leaf age. Ann Bot 33:743-751
1 4
97. Ho LC, AF Shav (1977) Carbon economy and translocation of C in
leaflets of the 7th leaf of tomato during leaf expansion. Ann Bot 41:833-848
98. Hoffmann E, F Lenz (1974) Die photosyntheseraten und
kohlenhydragehalte der blatter bei fruchttragenden und nicht-fruchttragenden Auberginen und Erdbeepflanzen. Gartenbau 39:539-547
99. Hoffmann E, G Mix, F Lenz (1975) Der staregehalt der chloroplasten
bei fruchttragenden und nicht fruchttragenden Auberginen und Erdbeerpflanzen. Angev Bot 49:115-1213


135
5000
4500
4000
3500
to
CO
3000
<
J
a;
Q
CP B
2500
6
a
2000
1500
1000
500
0
0
3
MIDVEIN LENGTH
dpm/mg
1
2
3
4
5
6
7
8
9
LEAF POSITION


173
GIRDLED NON-GIRDLED
BRANCH TYPE
FLUSH 3 FLUSH 2 FLUSH 1


165
50000
45000
40000
35000
30000
25000
20000
15000
10000
5000
0
\
\

\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
v *
///
GIRDLED
NON-GIRDLED
BRANCH TYPE
FLUSH 3
FLUSH 2
FLUSH 1


CHAPTER 6
DISTRIBUTION OF RECENT PHOTOASSIMILATES IN AVOCADO:
THE ROLE OF PROXIMAL LEAVES
Introduction
Carbohydrate reserves have been shown to play an integral role in early season growth in deciduous trees (88,155) where the carbon supply comes solely from those reserves. In evergreen trees, however, both current photosynthates and stored reserves contribute metabolic resources to developing sink tissues (121). The cyclical pattern of starch accumulation in the wood of avocado is well documented (26,162) and has led to the hypothesis that low levels of starch reserves induce fruitlet abscission and alternate bearing syndrome. Existing mature leaves in evergreen trees also serve as sources of both reserve carbohydrates and current photoassimilates. Their role in early season growth of flowers and fruitlets has not been evaluated.
Depending on the cultivar, avocado flower initiation begins in late November to January in Florida (44,158). By mid-January the terminal buds destined to become inflorescences have enlarged and are morphologically distinct from vegetative and dormant axillary buds. The development of the inflorescence has been described elsewhere (44,46,158,164).
The development of the inflorescence is concurrent with resumption of vegetative growth by the terminal bud (158,164,197). Concurrent growth and development of the inflorescence, fruitlets, and vegetative termini coupled with the high flower and fruitlet abscission rates
102


49
0.20
2.5
0.15
Q
m
a 0
u
m
<
10
0.05
0.00
Q
\
m <
u
DAYS POST-ANTHESIS
+ RGR cnT3/cnT3/Day ? AGR cnT3/Day


223
100. Floffman, NE, SF Yang (1980) Changes of 1-aminocyclopropane-
1-carboxyleic acid content in ripening fruits in relation to their ethylene production rates. J Amer Soc Hort Sci 105:492-495
101. Hopkinson JM (1964) Studies on the expansion of the leaf surface.
IV. The carbon and phosphorus economy of a leaf. J Exp Bot 15:125-137
102. Housley TL, DM Peterson, LE Shraeder (1977) Long distance
translocation of sucrose, serine, leucine, lysine and C02 assimilates. I. Soybeans. Plant Physiol 59:217-220
103. Hsieh YC, J Sacalis (1986) Levels of ACC in various floral
portions during aging of cut carnations. J Amer Soc Hort Sci 111: 942-944
104. Huber SC, DV Israel (1982) Biochemical basis for partitioning of
photosynthetically fixed carbon betveen starch and sucrose in soybean (Glycine max Merr.) leaves. Plant Physiol 69:691-696
105. Huberman M, R Goren (1979) Exo- and endo-cellular cellulase and
polygalacturonase in abscission zones of developing orange fruits. Physiol Plant 45:189-196
106. Hunt R (1982) Plant grovth curves: the functional approach to
plant grovth analysis. Univ Park Press, Baltimore, Md
107. Ibrahim IM, S El-din Bahlool (1979) The effect of girdling on
flovering, fruiting and vegetative grovth of avocado trees. Agric Res Rev 57:55-65
108. Isebrands JG, PR Larson (1973) Anatomical changes during leaf
ontogeny in Populus deltoides. Amer J Bot 60:199-208
109. Jankievicz LS, R Antoszevski, E Klimovicz (1967) Distribution of
labelled assimilates vithin a young apple tree after supplying 14C0 to a leaf or shoot. Biol Plant 9:116-121
2
110. Jennings VM, RM Shibbles (1968) Genotypic differences in
photosynthetic contributions of plant parts to grain yield in oats. Crop Sci 8:173-175
14
111. Jones H, JE Eagles (1962) Translocation of Carbon vithin and
betveen leaves. Ann Bot 26:505-511
112. Jones WW, TV Embleton, EI Barnhart, CB Cree (1974) Effect of time
and amount of fruit thinning on leaf carbohydrates and fruit set in 'Valentia' oranges. Hilgardia 42:441-449
1 4
113. Joy KV (1964) Translocation in sugar beet I. Assimilation of C02
and distribution of materials from leaves. J Exp Bot 15:485-494


RELATIVE INPORT CAPACITY
(% of photosynthate imported)
i i* i* >*ro
oro tk.oiQDorui^aiODO
h-1-1-1-1-1-1-1-^-1-1
RELATIVE EXPORT CAPACITY
(% of net photosynthate exported)
+ ?
m m
s x
o o
D D
H H


207
///
'////
X
\
X
\
xxsx
\
V
\
-_1
X X
\ \ \ \
\
?
/ /
////
/
girdled +fruit
non-girdled +fruit
defruited
branch type
flush 3
flush 2
flush 1


chapter 8
assimilation of 14co by non-fruit bearing flushes of avocado:
persea Americana mill. cv. 'Petersen'
Introduction
In avocado, leaf flush development and fruit development coincide for much of the season. Persea americana cv. Mill, may be either deciduous, semi-deciduous or evergreen. 'Petersen', the cultivar used in this study, is an evergreen variety. Two vegetative flushes occur between January and August in Florida. Individual leaves survive between 2 and 3 years. Reproductive flushes are first evident in late January. The inflorescence emerges in a pseudoterminal position. Leaves develop at the apex as the inflorescence expands. These developing tissues are supplied assimilates by the older leaves proximal to the inflorescence (Chap. 6). The leaves distal to the inflorescence mature and become the primary supplier of photoassimilates for the developing fruit and for another apical flush (Chap. 7). This final flush begins to develop in early June. Similar periods of flush development occur on non-fruit-bearing branches.
Research on starch reserves has focused primarily on stores in the roots and in the trunk (162). Although the leaves do not represent as large a reservoir for starch as these other tissues, they may play a significant role in plant development, particularly in evergreen trees. How the starch reserves in flushes of varying ages contribute to plant
150


67
The distribution of radiolaheled assimilates to the various tissues of punctured single (Fig. 4-6) and double fruits (Fig. 4-7) also differed. The anatomy of the vascular system may explain these departures from control values and provide a time frame for abscission. The mesocarp (pericarp) and the seed coat are directly supplied by vascular strands. The endosperm and the cotyledons have no direct connection to the vascular system. The relative percentage radioactivity in the mesocarp of single pierced fruitlets was significantly lower if pierced 24 or 48 hours before labeling. The process of phloem restriction therefore began within 24 hours. The higher proportion of radioactivity in the seed coat suggests that the vascular strands supplying the seed coat were not affected or not restricted to the same degree as the vascular strands supplying the mesocarp. This may result from the asymmetric distribution of the vascular strands (41) and the mass of the tissues. The vascular strands supplying the seed coat pass through the pericarp in the thicker side of the fruit. If the abscission "signal" traveled cell to cell the mass of the pericarp would hinder the movement of the abscission signal to the phloem. The vascular strands supplying the mesocrap originate from the thinner side of the fruit. These strands may perceive an abscission signal sooner than the vascular strands supplying the seed coat because fewer cells lie between the vascular strands and the putative origin of the abscission signal. The comparative mass of the tissues may also contribute to the delay observed in the restriction of the vascular strands. The less massive seed coat imports less assimilate than the more massive mesocarp. mass flow of assimilate may be easier to curtail to the seed coat than to the mesocarp.


% of TOTAL FRUIT RADIOLABEL


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
ASSIMILATE PARTITIONING IN AVOCADO, PERSEA AMERICANA MILL.
By
Susan F. Finazzo May 1990
Chairperson: Dr. T. L. Davenport
Major Department: Horticultural Science (Fruit Crops)
Competition between developing reproductive structures and leaves for limiting carbohydrates may cause abscission of avocado fruitlets, because they form in a pseudoterminal position and develop concurrently with new, apical leaves. The assimilation, partitioning and translocation of recent radiolabeled photoassimilates were measured throughout fruit development to test this hypothesis. Sucrose and mannoheptulose were the major sugars in phloem exudates. During the summer months, only sucrose was radiolabeled. During the winter months, radiolabel was recovered in both sucrose and mannoheptulose.
Abscising and persisting fruit exhibited similar growth kinetics until 4-5 days before fruit separation when the growth rate of the abscising fruit decreased. Abscising fruit (puncture-induced) accumulated more radiolabel in the seed coat than did non-abscising controls. Restriction of radiolabel movement into the punctured fruit
v


Figure 7-2. Early Season Labeling of Distal Leaves. All of the leaves distal to the inflorescence were labeled as described in the Materials and Methods. The branch was harvested immediately after labeling or 24 hours after labeling. The branch was dissected and processed for scintillation counting. The midvein length (mm) is the length of the midvein in the blade of the leaf. Leaf position describes the leaf's age and location. Leaf 1 is the most distal or acropetal; leaf 8 is the most proximal or most basipetal. All of these leaves are distal to the inflorescence. The experiment was repeated four times.


68
Like the mesocarp, however, the percentage of radiolabel in the seed coat does not vary from control values if the seed coat is punctured at 0 hour or after labeling. The lower values found in the cotyledons and endosperm after labeling could be an injury response such as an increase in respiration or redistribution of the assimilates. These lower values could also be caused by restricted release of assimilates from the seed coat. Since no direct vascular connection exists to these tissues they are dependent on the release of assimilates by the seed coat into the apoplast. In any case, it was evident that puncturing the seed coat immediately set in motion processes that altered assimilate distribution to the endosperm and cotyledons independent of vascular restriction.
Similar trends were seen in the distribution of assimilates to pierced fruit on doublet inflorescences. The effects of vascular restriction seemed to occur more quickly since tlie puncture effects were noted even in fruit pierced 24-48 hours after labeling. Although initiation of abscission occurred quickly the completion of the process appeared delayed in multiple-fruited inflorescences.
Single fruit pierced 72 hours before labeling abscised 24 hours after labeling and contained no radiolabel. Phloem restriction was therefore complete within 72 hours; however, separation of the fruit required 24 more hours. In fact, the receptacles of fruit pierced 24 and 48 hours before labeling contained 50% as much radiolabel as their subtending fruit indicating either vascular restriction greatly increased metabolic activity in the abscission zone. Fruit pierced at 0 or 24 hours post labeling exhibited slightly


Figure 8-1. Relative Distribution of C-Labeled Assimilates in Three Successive Avocado Leaf Flushes. Leaf samples (0.618 cm2) were taken from each leaf of each flush on both girdled and non-girdled branches immediately after the 1 hr labeling period. These samples were dried, weighed, digested and the radiolabel quantified as described in the Material and Methods. At the end of the experiment, branches were harvested and leaf area was measured. Leaf area and label
14
incorporation after 1 hr. exposure to CO were used to calculate the proportion of label incorporated by each oi three consecutive, girdled or non-girdled flushes. Branches were approximately 70-80 cm in total length and 1 to 1.5 cm in diameter. Branches were girdled proximal to the oldest flush, adjacent to the main scaffolding branch. Three branches were chosen, 1 each from the lower, middle and upper canopy positions. The experiment was repeated four times.
i


184
29,39,102,103,118,137,177,180,184), citrus (76), pepper (77,82,83), tomato (5,77,94), Easter lily (198), nectarine (38,43), and apple (87,88,89,134). In avocado fruit removal increased leaf dry mass and increased the number of starch grains in the most recent flush of leaves (161). The presence of multiple leaf flushes concurrently on a branch during fruit development provides an interesting model for carbohydrate partitioning with respect to leaf age, leaf position and fruit load.
Fruit-bearing and defruited branches had a mean of 8 leaves in the third flush, 10 in the second and 7 in the first flush. All of the leaves on the branches had approximately the same lamina length (Table 9-1). The second flush had a slightly greater area because of the greater number of leaves. In fact, the flushes of fruit-bearing and non-fruit bearing (Chap. 8), branches were similar except for a slight increase in leaf number and size of the leaves in flush 3.
Each flush of leaves on the fruit-bearing branches fixed similar proportions or about 33% of the total label fixed (Fig. 9-1). Leaves in the second flush, immediately distal to the fruit fixed slightly (37%) more label than either flush 3 (33%) or flush 1 (30%). This slight increase is probably due to the slightly larger total leaf area associated with this flush. The assimilation of label by flushes on fruit-bearing branches was much different than the age-dependent decrease in photoassimilation noted in flushes of non-fruit bearing branches (Chap. 8). Fruit sinks are known to stimulate photosynthsis (82,137,184) and delay leaf senescence (39,82,137) probably by decreasing the age-related loss of RuBPcase activity (82).


205
o
u
D
to
CP
6
a
T3
14000
13000
12000
11000
10000
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
\
X
/ / / /
_
SN^s>
X
X
s
x x
X
\
X
s
WW
X x x X
-
Vf/A
////
girdled +fruit
non-girdled +fruit
\s
^ X
X
X
_\\
\ \ \
.\\>x
V
X X X X \
x
X
X
-1
f /
/
/
' / /A
defruited
branch type
-*-*-*-*
flush 3
flush 2
flush 1


131
the translocated assimilates. However, the movement of labeled assimilates to the fruit and basipetally to the roots suggests that assimilates were ample and did not limit fruit growth.
In summary, leaves distal to the avocado inflorescence reached their peak of assimilate import when they were approximately 22% final midvein length. Export from these leaves commenced at this stage and reached a plateau when the leaves were approximately 60% final midvein
length. At maturity, the distal leaves translocated 69% of their recently photoassimilated carbon in 48 hours. Radiolabel recovered in the fruit accounted for 29% of the assimilate, while the remainder of the label traveled basipetally. Assimilate movement from source leaves was strongly guided by phyllotaxy, although fruit size did influence label accumulation.
The apparent precocious maturation of the developing distal leaves suggested that the leaves were autonomous and hence, the period of competition between fruits and developing leaves for translocated assimilates was shorter than previously expected.


211
14
to Tan.) C02 incorporated during photosynthesis vas recovered in both sucrose and mannoheptulose. Although mannoheptulose vas present in avocado leaf extracts and leaf phloem exudates in quantities similar to sucrose and vas labeled during the vinter months, its role is presently unclear.
The diameters of abscising and persisting avocado fruits vere measured every other day in order to compare grovth and abscission kinetics. Abscising and persisting avocado fruit exhibited similar grovth kinetics until 4 to 5 days before fruit separation, at vhich time the rate of fruit expansion decreased in abscising fruit. The greatest relative grovth rate (0.18921 cm3 cm"3 d"1) occurred immediately post-anthesis. Maximum absolute grovth rate values (2.2897
cm3 d"1) occurred at approximately 110 days post-anthesis. Eighty percent of the measured fruit abscised during the experiment. The number of fruit on a branch did not affect the fruit's grovth rate or probability of abscission.
Fruit grovth and the distribution of radiolabeled assimilates vere studied in abscising and persisting fruit. Abscising fruit accumulated significantly greater proportions of radiolabel in the seed coat vhen compared to persisting fruit. Persisting fruit increased the longevity of punctured fruit on the same inflorescence and extended the period in vhich assimilate distribution vas altered. Puncturing the seed coat caused a transient increase in its sink strength. Restriction of assimilate movement in punctured fruit vas evident vithin 24 hours and complete vithin 72 hours. Fruit separation occurred 24 to 48 hours later.


97
Table 5-5. Radiolabel Distribution after Photosynthetic 14C0 Assimilation by Fruit.
Fruit were labeled for 1 hour in the light with CO and then harvested 24 or 48 hrs later. Fruit were peeled, dissected and tissues were processed as described in the Materials and Methods. A total of 4 and 9 fruit were harvested after 24 and 48 hr chase periods, respectively, with a mean radiolabel of 278,616 + 32,168 and 205,546 + 13446 dpm, respectively.
Tissue dpm Dry Mass dpm Dry Mass
X of total (X of total) X of total (X of total) _recovered_recovered_
Exocarp 52.2 + 10.3 21.4 + 1.6 56.8 + 7.2 22.3 + 4.3
Mesocarp 44.9 + 7.5 65.5 + 2.3 41.9 + 7.2 66.6 + 3.9
Seed Coat 0.7 + 0.2 9.2 + 0.4 0.7 + 0.2 7.1 + 0.4
Endosperm 0.12+ 0.03 1.1 + 0.1 0.15+ 0.03 1.1 + 0.1
Cotyledon 0.1 + 0.05 2.9 + 1.1 0.15+ 0.04 3.0 + 0.9



Figure 6-4. Distribution within the Branch of C-Assimilates from the Proximal Source Leaf when the mean Midvein Length of the Distal Leaves was greater than 50 mm. Source leaves were fed CO for a one hour period. The entire branch was harvested 48 hours later. The branch was dissected, dried and processed for scintillation counting. The schematic drawing indicates the approximate developmental stage of the reproductive flush from early to mid-May. The black spots indicate the Labeled source leaf. The percentages are based on the total radiolabel recovered after the ini tial lhr CO fixat ion period. The arrows indicate the direction of assimilate flow. The data represent the mean values of at least 10 replicates. Standard error of measurement was 5%
or less.


105
early stages of flush development (distal midvein length <20 mm, Fig. 6-1) the inflorescence was still expanding. The active sink tissues were the developing flower buds and distal leaves. Labeled assimilates from the source leaf moved in both the acropetal and basipetal directions. No label was recovered from the mature non-source leaves proximal or distal to the source leaf. Forty-three percent of the recently fixed carbon remained in the source leaf 48 hours after labeling (Fig. 6-1). Of the 29% of label recovered in the tissues distal to source leaf, 15.39% was recovered in the leaves and 13.61% was recovered in the inflorescence and developing fruitlet tissues.
1 A
The amount of C-assimlates recovered in each tissue was proportional to the dry mass of the tissue (Table 6-1). For example, during the earliest stage (midvein length <20 mm), reproductive mass represented 52.9% of the total harvested dry mass and accounted for 53.06% of the total C assimilated by the harvested tissues.
The average midvein length of the distal leaves was 20 to 30 mm at the next stage of development (Fig. 6-2). All stages of flower bud development were present on the inflorescence. Translocation of photoassimilates from the source leaf was occurring at the same rate as in the previous stage i.e., 57-59% of the labeled carbon was translocated from the source leaf in 48 hours; however, the proportion of label traveling basipetally was much lower. Only 10.7% of the radiolabel was traveling basipetally while 48.2% of the label was recovered in the distal tissues. The distribution of label between the reproductive and vegetative tissues again was directly related to the


126
midvein length. The increasing rate of assimilate import reflected the increase in sink leaf size. The subsequent decline of assimilate import reflected the decrease in areas of the leaf still importing assimilate. The increase and subsequent decrease of assimilate import noted in developing leaves of avocado has also been observed in the leaves of cucumber (101), squash (203), beet (65) and soybean (185) during their sink to source transitions.
The export of label from avocado leaves was first observed when the leaves were 20% final midvein length. Avocado leaves began export of assimilates at a younger developmental stage (20% final midvein length) than most other plants (65,172,187,203), with the exception of tomato leaves (97) which initiated export at 10% final lamina length. Cucumber (95) and beet (72) leaves initiate export of assimilates at 30% and 25% final lamina length, respectively. Early studies linked the initiation of export to the cessation of cell expansion (42); however, others (65) have recently suggested that initiation of export depends on the osmotic pressure or the soluble sugar content of the sieve elements and companion cells.
The increase in the export rate of developing avocado leaves was very gradual. The gradual increase in export capacity however may have been a reflection of the sampling technique combined with the leaf maturation and not a true measurement of export from the leaf. Samples were taken from the lateral vein sections between the midvein and the
growing edge. Since leaves mature basipetally (127) the central midvein sections may export assimilate to other less mature areas within the leaf. If so, this shift of label within the leaf would


82
leaves proximal to the inflorescence were removed immediately after the labeling period. The branch tissues were then dissected into leaf, branch (distal and proximal to the inflorescence), and reproductive structures (panicals, receptacles, fruit, corolla, pistil, staminodes and nectaries) and prepared as described above for scintillation counting. Some branches were labeled in the morning when the open flowers were functionally female and other branches were labeled in the afternoon when the open flowers were functionally male.
Relative CO Fixation
-2-
Branches composed of fruit- and flower-bearing inflorescences with leaves distal and proximal to the inflorescence were enclosed in plastic bags and fed with 20 uCi of for a one hour period. The
branches were harvested immediately after labeling and then dissected into reproductive structures (fruit, flowers, buds, receptacles and pedicels), new leaves (developing distal to the fruit), and old leaves (present on the proximal or basipetal side of the fruiting structures). These tissues were processed as described above for scintillation counting. In this paper flowers which had opened previously were referred to as post-anthesis flowers.
The average lamina length of the leaves distal to the inflorescence was less than 20 mm. The leaves distal to the inflorescence at this stage of development were sinks and not sources of assimilate (Chap. 7). Carbon Fixation by Fruit
Individual fruit (diameter 4 cm) were enclosed in sealed plastic containers and then labeled with 100 uCi of 14C0 as previously


17
determined by scintillation spectroscopy of 1 ml samples. Some samples were spiked with sucrose for treatment with invertase and for comparison to unspiked samples.
Invertase (Sigma) was added to some extracts and control sugar samples, after acidification to pH 6 with IN HC1, to reach a final concentration of 1 mg/ml. The samples were incubated for 30 minutes at 4C. The pH of the samples was then adjusted to 7.5, by the addition of 30% NaOH, to stabilize the fructose content. Thirty microliters were spotted on thin layer plates and developed as described below. Invertase was used to more firmly identify the spot which comigrated with sucrose. Phloem Exudates
The trees used in this study were mature, orchard trees growing in Rockdale soil at the Tropical Research and Education Center in Homestead, FL. The most recent growth flushes, each containing approximately 10 leaves, were enclosed in 3 gallon plastic bags and labeled with 20 uCi of 14C0 for 1 hour. Two branches of each cultivar
2
were labeled on three different occasions during the late summer. Representative cultivars of each race and racial hybrids were chosen, they included 'Peterson' and 'Fuchs' (Vest Indian), 'Winter Mexican' and 'Mexicola' (Mexican), 'Itzamana' (Guatemalan), 'Booth 8' (Guatemalan X West Indian), and 'Fuerte' (Mexican X Guatemalan). Flushes were harvested at 4 and 24 hours after labeling. Each branch was cut from the tree and the cut end placed in a beaker of water before immediate transfer to the laboratory. Leaves were excised at the base of the petioles and the exudate collected with 5 ul capillary pipettes. The exudate (3 to 5 ul) was spotted directly onto TLC plates.


70
Vascular restriction, as indicated by label accumulation in the abscission zone, began within 24 hours of seed coat puncturing. At 48 hours vascular restriction continued and seed coat metabolic activity, as measured by the specific activity began to decline. At 72 hours vascular restriction was complete. Fruit separation, however, required another 24 to 48 hours to complete.


197
CO CO
<
?J
Q
a
5500
5000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
\ \\\
\
V
V \
////
////
/
/
\
*"
\
\
\
\
\
\
girdled +fruit
non-girdled +fruit
defruited
branch type I I flush 3
flush 2
flush 1


CHAPTER 4
ASSIMILATE PARTITIONING BETWEEN ABSCISING AND PERSISTING AVOCADO FRUITS
Introduction
The causes of avocado fruit abscission remain unclear even though fruit loss is extensive and economically damaging. Avocado exhibits two fruit set patterns. In the Type II pattern very few fruit are initially set; however, these fruit are retained until harvest. In the Type I fruit set pattern a heavy fruit set is followed by a prolonged period of extensive fruit abscission. This pattern is typical of many commercial varieties and the 'Petersen' cultivar used in this study. Sedgely (172) has shown that 90% of the fruitlets which abscise within one week of flowering were unfertilized. Fruitlets which abscised after this period were fertile and anatomically normal (172,187,188). Recent studies (48,74) have shown that genetic selection and and outcrossing may play an important role in determining which fruitlets abscise. Abscission of fruitlets can also be induced by injuring or puncturing the seed coat (19).
The suggested time frame in the literature for the senescence and abscission of fruitlets ranges from 4 days to 3 weeks. Abscission of avocado fruitlets is characterized by deterioration and browning of the seed coat (18) in an otherwise anatomically normal fruitlet. Adato and Gazit (1) suggested that the seed coat degeneration was related to increased ethylene production. This in turn correlated with fruit abscission. They suggested a 3 week timespan between peak ethylene
58


53
D
O >
D PS
O
-3.5
0
15
30
45
60
75
90
105 120 135 150
I 165
DAYS POST-ANTHESIS
rj
FRUIT # 16
FRUIT # 17
FRUIT # 21
FRUIT # 27


229
1R9. Trip Pf PR Gorham (1968) Bidirectional translocation of sugars in
sieve tubes of squash plants. Plant Physiol 43:8/7-888
190. Trochoulias T (1973) Avocado cincturing. Ag Gazette of New South
Wales 84:127
191. T'se A, A Ramina, VP Hackett, RM Sachs (1974) Enhanced
inflorescence development in bougainvillea 'San Diego Red' by removal of young leaves and cytokinen treatments. Plant Physiol 54:404-407
192. Turgeon R (1984) Termination of nutrient import and development of
vein loading capacity in albino tobacco leaves. Plant Physiol 76:45-48
193. Turgeon R, JA Vebb (1975) Leaf development and phloem transport in
Cucurbita pepo:plant economy. Planta 123:53-62
194. Valmayor RV (1967) Cellular development of the avocado from
blossom to maturity. Phillipine Agr 40:907-976
195. Venning F0, FB Lincoln (1958) Developmental morphology of the
vegetative axis of avocado (Persea americana L (sic)) and its significance to spacing, pruning practices and yields of the grove. Proc Fla State Hort Soc 71:350-356
196. Vu JCV, G Yelenosky, MG Bausher (1985) Photosynthetic activity in
the flover buds of 'Valencia' orange (Ci trus sinensis [L.J Osbeck). Plant Physiol 78:420-423
197. Vallerstein I, R Goren, SP Monselise (1974) The effect of girdling
on starch accumulation in sour orange seedlings. Can J Bot 52:930-937
198. Vang YT, PJ Breen (1986) Partitioning of 14C-assimilatte in Easter
lily as affected by grovth stage and flover removal. Scien Hort 29:273-281
199. Vardlav IF (1968) The control and pattern of movement of
carbohydrates in plants. Biol Rev 34:79-105
200. Vardlav, IF (1975) Assimilate movement in lollium and sorghum
leaves. I. Irradiance effects on photosynthesis, export and the distribution of assimilates. Aust J Plant Physiol 2:377-387
201. Vardlav IF, C Marshall (1976) Assimilate movement in lollium and
sorghum leaves. II. Irradiance effects on the products of photosynthesis. Aust J Plant Physiol 3:389-400
202. Veaver RJ, V Shindy, VM Kliever (1968) Grovth regulator induced
movement of photosynthetic products into fruits of black Corinth grapes. Plant Physiol 48:183-188


Figure 8-4. Specific Leaf Weight of Avocado Leaves from Three Successive Growth Flushes on Girdled and Non-Girdled Branches. Leaf disks (0.618 an ) were removed from each leaf of each flush on both girdled and non-girdled branches. The leaf disks were dried and weighed as described in the Materials and Methods. Each flush bore at least 10 leaves and each leaf was sampled at least 4 times during the experiment.


Figure 4-2. Distribution of C in the Pierced Fruit of the Double-fruited Inflorescence. Source leaves on branches bearing two fruit were fed CO for a one hour period. The seed coat of one fruit was pierced by a surface sterilized stainless steel wire before (48,24 hrs), after (24, 48 hrs) or at the time of labeling (0 hr). Both fruit were harvested 72 hrs after source leaf labeling. The fruit were dissected, dried and processed for scintillation counting. Six branches were used per experiment, each of which was repeated 4 times. Histogram bars represent the percentages of total radioactivity in the plant tissues. Data was transformed by arcsine transformation for statistical analysis. Mean separation within columns by Tukey's studentized range test (P<0.05).


CITATIONS
1. Adato I, S Gazit (1977) Role of ethylene in avocado fruit development and ripening: I. Fruit drop. J Expt Bot 28:636-643
2. Adato I, S Gazit, A Blumenfeld (1976) Relationship between changes in abscisic acid and ethylene production during ripening of avocado fruits. Aus J Plant Physiol 3:555-558
3. Addicott, FT (1970) Plant hormones in the control of abscission. Biol Rev 45:485-524
4. Akamine EK, T Goo (1973) Respiration and ethylene production during ontogeny of fruit. J Amer Soc Hort Sci 98:381-383
5. Ammerlaan AVS, MMHAJ Joosten, RI Grange (1986) The starch content of tomato leaves grown under glass. Scientia Hort 28:1-9
6. Antoszewski R, E Lis (1968) Translocation of some radioactive compounds from the strawberry receptacle to the mother plant. Bull Acad Polon Sci CI 16:443-446
7. Antoszewski R, A Mika (1971) Translocation of some assimilates from the sink to the donor in apple tree. Biologia Plant 13:43-49
8. Awad M, RE Young (1979) Postharvest variation in cellulase, polygalacturonase, and pectinmethylesterase in a avocado (Persea americana Mill, cv. Fuerte) fruits in relation to respiration and ethylene production. Plant Physiol 64:306-308
9. Awad M, RE Young (1980) Avocado pectinmethylesterase activity in relation to temperature, ethylene, and ripening. J Amer Soc Hort Sci 105:638-641
10. Barnell E (1939) Studies in tropical fruits V. Some anatomical
aspects of fruit-fall in two tropical arboreal plants. Ann Bot 3:77-79
11. Bean RC, BK Barr, HV Welch, GG Porter (1962) Carbohydrate
metabolism of the avocado. I. Relations between sugars in the leaves during photosynthesis and subsequent dark periods. Arch Biochem Biophys 96:524-520
12. Bean RC, BK Barr, HV Welch, GG Porter (1962) Carbohydrate
metabolism of the avocado. II. Formation of sugars during short periods of photosynthesis. Plant Physiol 37:280-284
216


FRUIT
NO.
I
2
3 4 5 6 7
8
DRY MASS
0170 0275
14
%C
.021 2.1 00
0492 17.700 0401
0333 12.400
.0733
310
1289 27.800 0438 39.300
LABEL
4


22
Persea americana Mill, consists of three ecologically distinct races: Mexican, Guatemalan and Vest Indian. Ve did not observe any differences in the sugar compliment of the phloem of avocado varieties representative of the three races. In fact, all of the varieties tested had significant amounts of mannoheptulose and sucrose in their phloem and leaves. Race did not play an important role in the determination of
the translocated sugar.
Seasonal changes did influence the partitioning of recently fixed label into the sugar pools. Although the reasons for this are not clear, several avenues of investigation are available. The winter months in Florida are cool and dry, suggesting mannoheptulose may act as an osmoticant. Although the experiment was conducted under water saturating conditions, the leaves used in the experiments were removed from field grown trees which had acclimated to the current winter conditions. However, at the experimentally induced lover temperature, sucrose, not mannoheptulose, appeared to be preferentially formed suggesting that mannoheptulose was not formed as a protection from frost. Flower initiation also occurs during the winter months. The production of mannoheptulose during the period of flower initiation is
suggestive and deserves further examination.
In conclusion, avocado exhibited seasonal differences in the partitioning of recently photoassimilated carbon into its sugar pools. During the summer months, recently fixed carbon is recovered only in sucrose although'mannoheptulose was present. During the winter months sucrose and mannoheptulose were both labeled in a temperature dependent manner.


Figure 9-8. Specific Activity of Starch Extracted from Avocado Leaves of Three Successive Flushes on Girdled and Non-Girdled Fruit-Bearing and Defruited Branches. Leaf disks (0.618 cm ) from each flush were pooled, chopped and boiled in 85% EtOH. The residues of the ethanolic extracts were digested by the method of Housely et al. (102) and the starch quantitated by the phenol sulfuric acid method. Samples of known starch content were used to quantify radiolabel. The fruit were removed from defruited branches one week prior to labeling.


182
Materials and Methods Ten-year-old, field-grown trees of Persea americana Mill. cv. 'Peterson' located at the Tropical Research and Education Center, Homestead, FL were used in this study. The most distal 3 leaf flushes on a branch were enclosed in a sealed plastic bag and labeled with 100 uCi of 14C0 for one a hour period between 9:00 and 11:00 a.m. 14C0
2 2
was generated from NaH14C03 (Amersham with a specific activity of 54 mCi/mmol) using equimolar amounts of 5% lactic acid. Three branches were chosen and individually labeled as described above for each experiment. Branches were chosen from the upper, middle and lower canopy positions at each labeling. Reproductive branches each had two fruit approximately 6 cm in diameter. On defruited branches, fruit were removed either one week or two days prior to labeling. The experiment was performed 3 times.
The two distal flushes (3 and 2) represent current season's growth. The most proximal flush (flush 1) was produced during the previous season. Fruit were positioned between flush 1 and flush 2. This study was conducted late in the season when flush 3 was mature and the fruit were at their peak absolute growth rate (Chap. 2). Fruit photosynthesis was not included in this study since the magnitude of translocation of assimilate from the fruit is not significant (Chap.
5).
Translocation from the leaves was measured by taking leaf punches (.309 cm2) throughout the experimental period. The entire branch was harvested 70 hours after labeling. The leaf samples were weighed and digested with a commercial tissue solubiliser (Scintigest Fisher


224
114. Knhn V (1975) Polyphenol oxidase activity and browning of three
avocado varieties. J Sci Fd Agric 26:1319-1324
115. Kahn V (1976) Polyphenol oxidase isoenzymes in avocado. Phytochem
15:267-272
116. Kahn V (1983) Multiple effects of hydrogen peroxide on the
activity of avocado polyphenol oxidase. Phytochem 22:2155-2159
117. Koch KE (1984) Translocation of photosynthetic products from
source leaves to aligned juice segments in citrus fruit. Hort Science 19:260-261
118. Koch KE, LE Schrader. 1984. 14C-Photosynthate partitioning and
translocation in soybeans during reproductive development. Plant Physiol 75:1040-1043
119. Kratashova N, S Tsylenok (1968) On the biologic role of nectaries
and nectar. II. Studies on the resorption of nectar by flower parts. Akad Nauk SSSR Ser Biol Med Nauk 1:134-137
120. Kriedemann P (1966) The photosynthetic activity of the wheat ear.
Ann Bot 30:349-363
121. Kriedemann P (1969) 14C translocation in orange plants. Aust J Agr
Res 20:291-300
122. Kushan MM, DG Richardson, AJ Ferro (1983) Intermediates in the
recycling of 5-methylthioribose to methionine in fruits. Plant Physiol 73:257-261
123. LaForge EB (1916) D-Manno-ketoheptose, a new sugar from the
avocado. J Biol Chem 28:511-22
124. Lahav E, B Gefen, D Zamet (1971) The effect of girdling on the
productivity of the avocado. J Am Soc Hort Sci 96:3986-398
125. Lance C, GE Hobson, RE Young, JR Biale (1965) Metabolic processes
in cytoplasmic particles of the avocado fruit VII. Oxidative and phosphorylative activities throughout the climacteric cycle. Plant Physiol 54:1116-1123
126. Larson PR, RE Dickson (1973) Distribution of 14C in developing
leaves of eastern cottonwood according to phyllotaxy. Planta 111:95-112
127. Larson PR, JC Gordon (1960) Leaf development, photosynthesis and
C distribution in Populus deltoides seedlings. Amer J Bot 56:1058-1066
128. Lee, SK (1981) Methods for percentage oil analysis of avocado
fruits. Yrbk Calif Avoc Soc 65:133-1415


99


Figure 8-7. Specific Activity of Sucrose Extract from Avocado Leaves of Three Successive Flushes on Girdled and Non-Girdled Branches. At harvest, leaf disks (0.618 cm ) vere removed from each leaf of each flush. The samples from each flush vere pooled, chopped and boiled in 85% EtOH. The ethanolic extracts vere concentrated. Sugar content vas quantitated by the phenol-sulfuric acid method. Samples of knovn sucrose concentration vere decolorized for scintillation counting.


130
In cases where there were no fruit in alignment with the labeled source leaf, label distribution reflected the tangential movement of assimilates and was moderated by the fruit mass and linear distance from the source leaf.
Phyllotaxy alone does not determine assimilate distribution in avocado. When several fruits on separate pedicels had the same phyllotaxy as the labeled source leaf, then size and parastichy determined assimilate distribution (Fig. 7-8). If the fruits were of similar size, then the fruits nearest the source accumulated slightly more labeled assimilate. If the fruit nearest the source was significantly larger, then it was the major sink for labeled assimilates; however, if the fruit farthest from the sink was larger then a more equal distribution of radiolabel was observed. The effect of size on assimilate distribution was related to the absolute growth potential of the fruit at that particular stage (94).
The supply of recent photoassimilates did not appear to limit fruit or leaf growth, or to partition to one fruit at the expense of another. Studies of assimilate movement indicated that 69% of the label was exported out of the source leaves in 48 hours (Fig. 7-9). Only 29% of this label was recovered in fruit tissues. No label was recovered in the mature leaves proximal to the fruit or in non-labeled leaves distal to the fruit. Label was recovered in the stem sections proximal to the fruit. Although the fruit are strong sinks (Chap. 5) less than half of the available translocated assimilates from the mature leaves distal to the inflorescence went to the fruit. Being the nearest sink to the source of assimilate, the fruit would also exert the greatest "pull" on


32
///// S/\
22
27
37
TEMPERATURE (centigrade)
SUCROSE
MANNOHEPTULOSE


167
<
< j
M-l O
<
S
0.0120
0.0110
0.0100
0 0090
0.0080
0.0070
0.0060
0.0050
0.0040
0 0030
0 0020
0.0010
0 0000
AW
s
V X
WW
/
/
/
/
/
/
/
/
GIRDLED
NON-GIRDLED
BRANCH TYPE
FLUSH 3
FLUSH 2
__
FLUSH 1


189
Increased translocation rates have been associated vith increased sucrose phosphate synthase (SPS) activity (153) vhich in turn correlates to increased sink demand. Leaves on fruit-bearing branches may exhibit increased SPS activity resulting in an increase in sucrose labeling and a decrease in starch accumulation (101,171). In avocado, sink demand apparently did increase sucrose labeling. Leaves on fruit-bearing flushes maintained a highly-labeled, soluble sucrose pool.
The leaves of the most distal flush on the non-girdled branch contained the highest proportion of starch (54%) (Fig. 9-7). Generally, the flushes of girdled branches had higher percentages of starch than their non-girdled counterparts (28% girdled vs. 17% non-girdled). Defruiting depressed the starch content of the distal flush (54% vs. 34%, non-girdled vs defruited) yet caused an increase in the starch content of the oldest most acropetal flush (34 vs 44%, non-girled vs defruited). The flushes vhich flank the fruit (flush 1 and 2) had the lovest starch content. Flush 2 contained the lovest percentage of starch on both non-girdled and defruited branches. Other
vork done in this laboratory has shovn that the majority of assimilate translocated from the proximal flush travels basipetally at this stage of leaf development (Chap. 6). Flush 2, vhich vas distal to the fruit provided assimilate to both the fruit and to the developing apex. The depression in the starch content probably reflected the mobilization of recent reserves to the developing sinks (flush 3 and the fruit).
The specific activities (Fig. 9-8) of starch extracted from girdled fruit-bearing branches or defruited branches vere similar. The highest specific activity vas found in starch extracted from the


128
n^t gains in leaves as small as 22% final midvein length (51). Net carbon fixation in avocado was constant on an area basis until the leaves reached 23% final midvein length vhen a dramatic increase in photosynthesis occurred. At this stage leaf import of assimilates was at its peak and export of assimilates had just begun.
Net photosynthesis has been reported to decline between 50-75% final midvein length (65,101) in beet and cucumber. Hopkinson (101) suggested that this may result from self-shading or other age related phenomena such as increased photorespiration. Net photosynthesis did not decrease in mature leaves during this study perhaps because unlike cucumber and beet, avocado is a perennial evergreen. The age of a 'Petersen' avocado leaf may be measured in years rather than weeks. Furthermore, canopy shading may not have a significant effect on leaf photosynthesis since avocado leaves saturate at low light intensities (163).
The pattern of radiolabel distribution in developing leaves depended upon sink leaf age and sink and source leaf phyllotaxis. Avocado exhibits a 2/5 phyllotaxis. As the leaves distal to the inflorescence matured, the effect of phyllotaxy on label distribution
became pronounced. During the early stages of distal leaf development (MVL <20 mm) equivalent amounts of label were recovered from leaves at each orthostichy (Fig. 7-3). Vhen the average distal lamina length was between 20 and 30 mm (Fig. 7-4), two adjacent orthostichies, one of which was in alignment with the source leaf contained the majority of radiolabel. The final stage, when the average distal lamina length was


36
The following equations were used to describe growth, where V represents volume:
Relative growth rate (RGR) = 1/V dV/dT = P (T)
V
Absolute growth rate (AGR) = dV/dT = (T) exp[fv(T)] The RGR was determined in reference to a measurable attribute V (volume) which is a function of time, V= f(T). The RGR therefor may be expressed as; RGR = 1/V2 [(V2 VI) / (T2 Tl)] = P (T) or d ln/dT V (T).
The RGR describes the rate of deposition of matter in relation to the
bulk of existing tissue or the rate of growth in size divided by the
size of the fruit. The absolute growth rate is also a function of a measurable parameter and time V = f(t). The first derivative of this
function dV/dT is the AGR which describes the absolute increase in mass
per unit time. AGR is the first derivative or slope of the function describing the function of time and volume and the RGR is this quantity divided by the final V.
Results
'Petersen' exhibited a single sigmoidal growth curve (Fig. 3-1). Fruit volume increased from .015 cm3 at anthesis to greater than 400 cm3 at harvest 175 days post-anthesis. Diurnal fluctuations in fruit volume due to water stress and transpiration have been observed (171); however, these effects were minimized in the present experiments by using drip irrigation and by measuring the fruit early in the morning.
Fruit volume and the dry mass (Fig. 3-2) increased rapidly and then leveled off as the season progressed. The relationship between fruit dry mass and volume (Fig. 3-2) indicated that fruit underwent substantial relative dry mass accumulation during the initial


154
incorporated 30% and flush 1, 20% of the total label incorporated by the branch. The youngest, most recently formed flush therefore accounted for the majority of the carbon fixed by the branch. The age dependent decrease in photoassimilation has been linked to a decrease in the soluble and fraction 1 proteins and to a decrease in RuBPcase activity (82,137). These observations have been made primarily in annual agronomic crops (82,137) where age is determined by canopy position. However, in evergreen perennials such as avocado, flush development probably provides the best correlation for leaf age and physiological activities.
Vhen the branches were girdled the differences in carbon fixation between the flushes were accentuated with Flush 3 assimilating approximately 65% of the label while Flush 1 assimilated less than 15% of the label. Although girdling caused an apparent increase in the
14 _
relative fixation of C02 by leaves in Flush 3, it decreased absolute fixation by all three flushes (Fig. 8-2). Flushes on non-girdled branches had greater specific activities, whether compared on a dry weight (Fig. 8-2) basis or an area (Fig. 8-3) basis. On girdled branches the specific activities of the leaf flushes decreased particularly for the older flushes. Girdling within 24 hours of labeling adversely affected the assimilation of 14C0^ and skewed the relative efficiency of fixation in favor of the youngest flush. Decreasing the sink to source ratio has been shown to increase leaf and intercellular resistances and decrease carbon fixation (83,118).
An inverse relationship, with the oldest flush having lowest values and the youngest flush having the highest values, existed between assimilation and leaf mass or leaf area. The distal flush was


185
Girdling slightly decreased the ratio of assimilates fixed by flush 2. Flushes of leaves on the non-girdled branches, like their girdled counterparts fixed very similar proportions of 14C02 suggesting that girdling, which removed the trunk and roots as sinks, had very little effect on carbohydrate metabolism of fruit-bearing branches during the short period of this experiment. Branches defruited one week prior labeling differed markedly from the non-girdled branches in the relative fixation of label. The pattern of relative fixation in these branches was very similar to that observed in non-fruit-bearing branches (Chap. 8); the youngest flush fixed the most label (44%) and the older flushes proportionately less (flush 2-32%, flush 1-24%). Similar observations were made in experiments using soybean plants in which depodding was accompanied by decreased photosynthesis (137) and altered the partitioning of assimilates associated with senescence (39).
The flushes of non-girdled fruit-bearing branches (third, second, first) fixed 4800, 5400, 4200 dpm/mg, respectively on a dry weight basis (Fig. 9-2). The specific activity of the second flush which fixed approximately 5400 dpm/mg dry mass was only 20% higher than the value for flush one which had the lowest specific activity of the fruit-bearing flushes. Girdling and defruiting decreased the absolute specific activity of the flushes between 20 to 50%. The relative activity of flush 2 was affected the most; girdling decreased the specific activity by 65% while defruiting decreased the leaf specific activity by 55%. The trends observed for defruited and girdled fruit-bearing flushes mimiced those observed for non-fruit-bearing flushes where the youngest flush fixed the most DPM and the oldest


151
growth is not known. Non-fruit-bearing branches of avocaHo represent a significant portion of the tree's foliage and their contribution to the tree's carbon economy may be significant.
This paper examines non-fruit-bearing branches, while a companion paper (Chap. 9) examines fruit-bearing branches of avocado. Photoassimilation by both branch types was examined to provide a direct comparison between fruit-bearing and non-fruit-bearing branches; since, the ultimate goal of this research was to determine if carbohydrate partitioning directly controls fruit abscission. The immediate goals
of this research were three-fold. First, the assimilation and translocation of recently fixed photoassimilates by leaf flushes of different ages were measured to determine which flush was most efficient in fixing and translocating recently fixed photoassimilates. Second, the starch and sucrose content of the leaves was measured, so that pool sizes and pool turnover within the leaves of different flushes could be compared. Third, similar measurements were made in branches girdled 24 hours before labeling to determine what effect sink removal would have on carbohydrate metabolism.
Materials and Methods Ten-year-old, field-grown trees of Persea americana Mill., cv. 'Peterson', located at the Tropical Research and Education Center, Homestead, FL were used in this study. The most distal 3 leaf flushes on a branch were enclosed in a sealed plastic bag and labeled with 100 uCi of 14C0 for a one-hour period between 0900 and 1100 h 14C0 was
2 v 2
1 4
generated from NaH CO (Amersham with a specific activity of 54


8
The role of ethylene in the ripening of fruit has not been determined. Ethylene does not induce its own synthesis in immature fruitlets but it does induce its own synthesis in mature fruit (55,56). In fruit harvested during the early season, ethylene causes a temporary increase in respiration rates but does not trigger the climacteric pattern (55,56). Preclimacteric fruits contain less than .1 nmol/g of 1-amino-cyclopropane-l-carboxylic acid (ACC) (100). This immediate precursor of ethylene increases to 45 nmol/gr of tissue during the climacteric peak which in turn causes an increase in ethylene production. As ethylene is evolved the ACC content again drops. As the fruit becomes overripe ACC accumulates; however, ethylene is no longer generated. Ethylene has no effect on attached fruit (70). Freshly harvested fruit do not respond to ethylene for a least 24 hours suggesting that attached fruit are affected by some endogenous factor which inhibits ethylene stimulated ripening (17).
Blumenfeld and Gazit (17) described an endogenous growth regulator found in avocado fruit tissue. This agent inhibits indolacetic acid (IAA) and cytokinin-mediated activities. Avocado mesocarp also synthesizes a large quantity of ABA during development (135). The greatest synthesis occurs after the seed coat has broken its vascular connection with the mesocarp. The large increase in free ABA occurs two days after the peak of ethylene evolution. Although ABA may be involved in the ripening process, it is not responsible for ethylene evolution (17).
Various biochemical processes have been examined during the climacteric cycle. Lance et al. (125) reported that the increase in the respiratory activity of the preclimacteric fruit was probably due


Figure 5-1. Sink Strength of Floral Components. The schematic diagram represents the avocado inflorescence in relation to the new distal leaves and the older proximal leaves. The insert is a detailed representation of a pedicel and its component parts. Sink strength is defined as the relative content of radiolabel on a Xdpm mg dry mass"1 basis.


122
subtracting the amount of radiolabel remaining in the source leaf at 24 hr. post-labeling from the amount of label in the source leaf at 0 hr. The relative rate of export was defined as the percentage of net radioactive photosynthate translocated from the source leaf. The relative rate of import was defined as the percentage of the total radioactive translocate imported by the sink leaf.
The distribution of radiolabel in the leaves distal to the inflorescence was indicated as the percentage of the total radioactivity recovered from the distal leaves. Similarly, the distribution of label in fruit at a particular orthostichy was
calculated as a percentage of the radioactivity recovered from all of the fruit.
Results
Import/Export
The relative rate of import or the percentage of the translocated assimilate imported by distal leaves as they developed is shown in Fig. 7-1. Small leaves imported little of the total translocated assimilate. As the leaf enlarged to 20% of the final midvein length (FMVL) the relative import ratio peaked. As the leaf continued to enlarge the import of assimilates by the developing distal leaf decreased and was negligible when the leaf reached 37% FMVL.
The relative rate of export (Fig. 7-1) was measured by the difference in label present at 0 and 24 hours post-labeling in labeled source leaves located distal to the inflorescence. Export began at 20% final midvein length and reached a plateau at 60% final midvein length. The amount of radiolabeled assimilates translocated in 24 hours was


Figure 3-5. Growth of Individual Fruit on a Single-fruited Inflorescence. Fruit 16 and 17 abscised approximately 55 days after tagging, while fruits 21 and 27 were retained until harvest.


Figure 8-3. Radioactivity per Unit Avocado Leaf Area. Leaf samples (0.618 cm ) were taken from each leaf of each flush on both girdled and non-girdled branches immediately after the 1 hr labeling period. These samples were dried, weighed, digested and the radiolabel quantified as described in the Material and Methods. At the end of the experiment, branches were harvested and leaf area was measured. Leaf area and
14
label incorporation after 1 hr. exposure to CO were used to calculate the dpm/cm of label incorporated by each of three consecutive, girdled or non-girdled flushes. Branches were approximately 70-80 cm in total length and 1 to 1.5 cm in diameter. Branches were girdled proximal to the oldest flush, adjacent to the main scaffolding branch. Three branches were chosen, 1 each from the lower, middle and upper canopy positions. The experiment was repeated four times.


Figure 3-3. Absolute and Relative Growth Rates of Avocado. The AGR and RGR were calculated after Hunt (106). The maximum AGR was 2.2897 cm3 d"1 at 110 days post anthesis. The RGR was greatest (.18921 cm3 cm"3 d"1) early in the season.


13
Plant growth substances have been credited as the driving force of sink demand. Exogenous application of individual or multiple growth regulators to sinks increased the demand for assimilates (141). Growth regulators, therefore, directly affect translocation by stimulating the sink's synthetic activities which require assimilates. Gibberellic acid (GA) and its synthetic analogs decreased stylar abscission, decreased the early drop of fruit and increased the early size of fruit (3). Gibberellic acid is thought to act by increasing auxin production. Auxin prevented abscission when applied to the sink distal to the abscission zone (3,31,131,154,182).
The role of the fruit as a source of hormones (3,131,202) and nutrients has complicated the early models of source-sink relationships. Unlike source to sink translocation which occurs primarily in the phloem, export from the sink to the source is via the xylem (14). Since mass flow in the xylem is nonselective, any exported compound may be translocated to all the vegetative parts of the plant. Removing soybean pods from the plant decreases dry matter accumulation and reduces the nitrogen content of the whole plant (37). Depodded plants also exhibited delayed leaf senescence, suggesting that the pods alter the partitioning of plant constituents and affect the initiation of senescence. In Phaseolus vulgaris (182) the older fruits at the base of the raceme aborted less frequently than the younger fruits. When the older fruits were removed, the abortion rate of the younger fruits decreased as did the ABA content of the fruits. This suggested that the older fruits have a direct regulatory effect on the ABA content and abscission of younger fruit.


Copyright 1990
by
Susan F. Finazzo


Figure 6-3. Distribution within the Branch of C-Assimilates from the Proximal Source Leaf when the mean Midvein Length of the Distal Leaves was between 30-50 mm. Source leaves were fed CO for a one hour period. The entire branch was harvested 48 hours later. The branch was dissected, dried and processed for scintillation counting. The schematic drawing indicates the approximate developmental stage of the reproductive flush from mid-April to early May. The black spots indicate the labeled source leaf. The percentages are based on the total radiolabel recovered after the initial 1 hr CO fixation period. The arrows indicate the direction of assimilate flow. The data represent the mean values of at least 10 replicates. Standard error of measurement was 5% or less.


38
five fruit (86-89) shown in Fig. 3-4 indicated that growth of individual fruits on a branch was identical even though these fruit had different phyllotaxy's and different arrangements on the inflorescences (three separate inflorescences; two inflorescences bearing two fruit and one inflorescences bearing one fruit).
The growth rates (AGR) of the five fruit mentioned above were nearly identical to the growth rates observed for other fruit present singly on different branches (Fig. 3-5) or on multiple-fruited branches (fruit 86-89, fruit 76-77). The growth of abscising fruit (Fig. 3-5)
was similar to retained fruit until 4.54 +/- .35 days before abscission. Four to 5 days before fruit separation the AGR of the fruit destined to abscise decreased (data not shown). Generally fruit continued to increase slightly in volume until separation, however occasionally fruit shriveled (fruit 17) before separation.
Nearly 80% of the tagged fruit abscised during the experimental period (Fig. 3-6). The majority of these fruit (51%) were less than 1 cm in diameter. The number of fruit retained (Fig. 3-7) showed an exponential decrease over time. A large number of fruit abscised early in the season. Progressively fewer fruit abscised with time.
Discussion
Early workers in avocado growth and development determined that avocado exhibited single sigmoidal growth kinetics. These worker also showed that somewhat uniquely avocado fruit growth resulted from continuous cell division and expansion throughout development (167,169,194). The growth of the 'Petersen' cultivar can also be


92
losses exceeded photosynthetic gains. Tn citrus, the organic acids produced by PEPCase were used directly as respiratory substrates in the TCA cycle (21). Photosynthesis in avocado fruits may have a similar role.
In conclusion, all of the reproductive structures of the avocado
14
inflorescence were capable of fixing C02 However, assimilates translocated from other organs, such as the proximal leaves, provided the majority of necessary photosynthate for growth. The avocado flower contributed very little photosynthate to its own development. However, the flowers and flower buds were also relatively weak sinks so the significance of flower carbon fixation to growth and respiration has not been determined. Avocado flowers exhibited a stage-dependent mass difference; flowers in the male stage were significantly more massive than flowers in the female stage. Fixation of carbon by the fruit correlated to fruit dry mass. Very little translocation of fruit-fixed assimilates was noted. Although all of the structures of the avocado inflorescence fixed C02, their contribution to the carbon budget appeared to be minor.


218
29. Carlson OR, VA Brim (19R4) Alteration of 14C-assimi 1 ate
partitioning in leaves of soybeans having increased reproductive loads at one node. Plant Physiol 75:887-890
30. Charlson AJ, NK Richtmeyer (1960) The isolation of an octulose and
an octitol from natural sources: D-glycerol-D-manno-octulose and D-erythro-D-galacto-ocititol from avocado and D-glycerol-D-manno-octulose from Sedum species. J Am Chem Soc
31. Chatterjee SK, AC Leopold (1964) Kinetin and gibberillin actions
on abscission processes. Plant Physiol 39:334-337
and translocation of C-sucrose in mango. Scientia Hort 22:121-28
33. Clark RB, A Wallace, RT Mueller (1961) Dark CO fixation in
avocado roots, leaves and fruit. Yrbk Calif Avoc Soc 45:96-101
34. Cohen A (1981) Recent developments in girdling of citrus trees.
Proc Int Soc Citriculture 1981:196-199
35. Cohen A (1983) Recent developments in girdling of citrus trees.
Proc Int Soc Citriculture 1983:83-87
36. Cohen A, S Roisman (1971) Ringing avocado to promote fertility.
Hassadeh 51:47-48
37. Crafts AS, CE Crisp (1971) Phloem transport in plants. Freeman,
San Francisco, U.S.A 38. Crafts-Brandner SJ, FE Below, JE Harper, RH Hageman (1984) Effects of pod removal on metabolism and senescence of nodulating and nonnodulating soybean isolines. Plant Physiol 75:311-317
38. Crews CE, SL Williams, HM Vines (1975) Characteristics of
photosynthesis in peach leaves. Planta 126:97-104
39. Crisp CE (1973) Phoem transport. J Amer Soc Hort Sci 98:250-254
40. Crookston RK, J 0'Toole, JL Osbun (1974) Characterization of the
bean pod as a photosynthetic organ. Crop Sci 14:708-712
41. Cummings K, CA Shroeder (1942) Anatomy of the avocado. Yrbk Calif
Avocado Soc 26:56-64
42. Dale JE (1976) Cell division in leaves. In: Cell division in
higher plants, p76-94, ed. MM Yeoman. Academic Press, London
43. Dann IR, RA Wildes, DJ Chalmers (1984) Effects of limb girdling on
growth and development of competing fruit and vegetative tissues of peach trees. Aust J Plant Physiol 11:49-58
82:3428-34
32. Chauhan PS
by leaves and fruits,
44. Davenport TL (1982) Avocado growth and development. Proc Fla State
Hort Soc 95:92-96


Figure 5-2. Early Season Labeling of Avocado Fruit. Entire branches were labeled. The branches were dissected and processed for scintillation counting immediately after the labeling period. Dry mass and average dpm of radiolabel fixed per fruit is plotted against fruit diameter (mm).


66
assjmi 1ntes by fruit treated in this way mimiced the ratios found in non-punctured abscised fruitlets observed in preliminary studies (data not shown) suggesting the puncture treatment was a reasonable means of testing "natural" abscission events.
Piercing the seed coat drastically decreased the growth rate of all of the fruit on the inflorescence (Table 4-4). Cessation of growth by retained orange fruits has also been reported by Zucconi et al. (212). They found that persisting orange fruit exhibited stunted growth shortly before the peak drop of abscising fruit.
Fruitlets abscised 4 to 5 days after puncturing. The observation of diminished fruit growth by persisting fruit on inflorescences where abscission is occurring was noted in this study but not in the earlier fruit growth study from this laboratory (Chap. 3). This discrepancy is probably due to the frequency of fruit measurements. The twice daily measurements clearly showed the transient decreases in growth of persisting fruits which were not evident in the earlier study when fruit were measured every 2 days.
Although the growth rate of a non-punctured fruit on a multiple-fruited inflorescence decreased when another fruit was punctured, the presence of the non-punctured fruitlet appeared to increase the longevity of the pierced fruit. This was supported by the observation that pierced single fruitlets abscised sooner than their doublet counterparts (data not shown). Single fruitlets tended to abscise 4 to 5 days after puncturing, while their double-fruited counterparts required 5 to 6 days for separation.


Figure 7-7. The Distribution of Assimilates to Fruits vhen the Source Leaf and a Fruit have the Same Phyllotaxy. A single leaf (black-dotted) distal to the fruits vas labeled. The fruit vere harvested and processed for scintillation counting 24 hours after labeling. The percentages indicate the proportion of the total fruit radioactivity associated vith the fruit at each stelar location. The standard error for these measurements vas 10% or less. This is a representative schematic vhich assumes fruit are present at each stelar location. Fifteen branches vere used in this study.


2 0%
2 0mm


Figure 8-6. Sucrose Content per Unit Area of Avocado Leaves from Three Successive Growth Flushes on Girdled and Non-Girdled Branches. Leaf disks (0.618 cm') were removed from each leaf of each flush at harvest. The samples from each flush were pooled, chopped and boiled in 85% EtOH. The ethanolic extracts were concentrated and the sugar content was measured by the phenol-sulfuric acid method.


CHAPTER 7
SINK TO SOURCE TRANSITION OF DEVELOPING LEAVES LOCATED DISTAL
TO AVOCADO INFLORESCENCES
Introduction
Developing leaves are known to be net importers (sinks) and mature leaves net exporters (sources) of photoassimilated carbon. Recent photoassimilates from mature leaves of tomato were reported to translocate to the developing leaves at rates according to the age of the developing leaves and the leaves' phyllotactic relationship to the source leaf (96). The conversion of leaves from sinks to sources involves both physiological and structural changes. These include an increase in net photosynthesis (51,73,95,101), a decrease in assimilate import (101,192,193,203), the onset of phloem loading (65,192) and alterations in leaf anatomy (108).
The development of leaf autonomy in avocado is of particular interest, since immature leaves developing distal to the inflorescence represent a potentially 'competing' sink with developing flowers and fruits for available assimilate. The inflorescences of avocado emerge subterminally (44). New leaves form at the apex, distal to the inflorescence, shortly after the onset of flowering and develop concurrently with flowers and early setting fruit. The concurrent growth of the reproductive organs and new leaves coupled with the high rate of fruitlet abscission (44) observed during this period has led
some workers (20) to suggest that these developing sinks compete for a limited amount of available assimilate. Furthermore, reducing the
119


191
The values for the relative specific activity of starch extracted from different flushes on non-fruit-bearing branches were similar suggesting that recent photoassimilates were similarly allocated to this pool in all flushes; however, fruit-bearing flushes allocated recent photoassimilates to the starch and soluble sugar pools differently; the second flush directed more of its recent
photoassimilates into starch biosynthesis and less into soluble sugar pools while the most distal flush directed more assimilate to its soluble sugar pools than to starch synthesis. The translocation and fixation rates for flushes on the fruit-bearing flushes were nearly the
same.
Altering the source to sink ratio by girdling had its most striking effect on translocation rates. In non-fruit-bearing branches girdling decreased the movement of recent photoassimilates by approximately 50%. Girdling in this case removed the primary sink. However, in fruit-bearing branches girdling decreased the rate of translocation by only 15%. This indicated that the fruit was the major sink, but it also suggested that enough assimilate was produced to supply both the fruit and an alternate sink, the trunk and roots. If this were not the case, and only enough assimilate were produced to supply the fruit then no decrease in the translocation rate would be expected.


108
Furthermore increasing starch levels is not always effective in increasing yields. Girdling increases the starch content of the leaves and wood above the girdle yet in some varieties of avocado girdling does not increase yields (36). Stutte and Martin (181) increased the starch content of olive trees 5-fold yet they were unable to change the alternate bearing pattern. Interestingly, killing the olive seed embryo overcame the alternate bearing problem without affecting fruit dry mass accumulation. Similar results were obtained in apple (78,79) and orange (130) which further negates the suggestion that starch levels directly affect fruitlet dry mass accumulation or fruitlet
abscission.
Although carbohydrates are absolutely essential for the growth of fruitlets and new leaves, work in this laboratory has shown that carbohydrate supply from nearby leaves is not a factor in fruitlet abscission. Mature leaves proximal to the inflorescence were more than capable of providing assimilate for new growth. In fact, mature leaves proximal to the inflorescence translocated assimilate both acropetally towards the developing apex and basipetally towards the trunk. The basipetal sink for assimilates from the proximal leaves in avocado differed from the basipetal sink for assimilates observed in mango (32). Mature mango leaves proximal to the labeled source leaf accumulate translocated labeled assimilates; however, mature leaves of avocado did not act as sinks for translocated assimilates.
The paucity of label translocated to the distal leaves of stage 4 inflorescences (Fig. 6-4) from the proximal source leaf reflected the maturation and increased autonomy of the distal leaves which supply the developing fruits. Similar results were obtained in strawberry (160),


Figure 9-1. Relative Distribution of C-Labeled Assimilates in Three Successive Avocado Leaf Flushes on Fruit-Bearing and Defruited Branches. Leaf samples (0.618 cm ) were taken from each leaf of each flush on both girdled and non-girdled, fruit-bearing and defruited branches immediately after the 1 hr labeling period. These samples were dried, weighed, digested and the radiolabel quantified as described in the Material and Methods. At the end of the experiment, branches were harvested and leaf area was measured. Leaf area and
1 4
label incorporation after 1 hr exposure to C02 were used to calculate the proportion of label incorporated by each of three consecutive, girdled or non-girdled flushes. The fruit were removed from defruited branches one week prior to labeling. Branches were approximately 70-80 cm in total length and 1 to 1.5 cm in diameter. Branches were girdled proximal to the oldest flush, adjacent to the main scaffolding branch. Three branches were chosen, 1 each from the lower, middle and upper canopy positions. The experiment was repeated four times.


Figure 6-2. Distribution within the Branch of C-Assimilates from the Proximal Source Leaf when the mean Midvein Length of the Distal Leaves was between 20-30 mm. Source leaves were fed CO for a one hour period. The entire branch was harvested 48 hours later. The branch was dissected, dried and processed for scintillation counting. The schematic drawing indicates the approximate developmental stage of the reproductive flush from early to mid-April. The black spots indicate the labeled source leaf. The percentages are based on the total radiolabel recovered after the ini tial 1 hr 'CO fixat ion period. The arrows indicate the direction of assimilate flow. The data represent the mean values of at least 10 replicates. Standard error of measurement was 5% or less.


212
The ability of avocado reproductive structure?: to assimilate photosynthetic products either directly or via transport from source leaves vas evaluated. The accumulation of labeled assimilates in the reproductive organs and tissues vas estimated after flovers, fruits and source leaves vere fed 14C02- The mean sink strength (total dpm) of flovers and flover buds vas less than half that of fruits. The sink intensity (dpm mg"1 dry mass) of the flovers and flover buds vas 20 times that of the fruit. Developing fruit fixed 14C02 in proportion to their dry mass. Avocado leaves fixed the majority (96%) of the available labeled C02 vhile the fruit and flovers fixed only 4% of the label. Although some label fixed by the fruits vas translocated to other tissues, the majority (97-99%) of the radiolabel remained in the fruit. C02 fixation by the flovers and fruits of avocado did not contribute substantial photoassimilate to fruit grovth.
Inflorescences of avocado form in a pseudoterminal position and develop concurrently vith nev, apical (distal) leaves. The potential 'competition' betveen developing leaves, flovers and fruit for limiting
amounts of available assimilate as a cause of flover and fruitlet abscission vas studied by examining the movement of 14C-labeled
photoassimilates from mature leaves located proximal to the inflorescence at all stages of inflorescence and distal leaf development. Assimilates moving from the proximal leaves vere equally divided betveen acropetal and basipetal sinks at the earliest stage of distal leaf development (midvein length <20 mm). As the inflorescences began to bear developing fruitlets (<1 cm diameter) and distal leaves continued to develop, 5-times more label moved acropetally to supply


6
asymmetrical pattern. The endocarp consists of a few rows of parenchyma cells. These cells are smaller than those of the mesocarp and they adhere to the seed coat.
Research in Avocado Avocado research in the last 15 years has focused primarily on three different areas. These are the fruit's nutritional value for human consumption, post harvest and climacteric behavior and morphology during development and abscission.
The fruit has been a major food source for the people of Central America for several thousand years, and with the advent of cold storage the fruit's popularity has spread worldwide. The less tropical avocados contain about 124 calories per 100 grams of fresh weight (84). The fruit contains between 4 and 10 percent protein (84). Avocado fruit provides approximately 666 mg/100 gr fresh weight of the essential amino acids (84). This is six times the amount provided by apples and three times the amount provided by either peaches or oranges (84). The avocado is exceptionally rich in minerals, oils and a number of vitamins (84).
The avocado is unique in several ways. In many fruits the final stages of growth result from cell expansion; however, in avocado, cells continue to undergo mitosis throughout the fruits sojourn on the tree (169). Furthermore, the avocado fruit does not ripen while it is still attached to the tree (2). Once removed from the tree, the fruit if
|
sufficiently developed will undergo a climacteric rise in respiration and will ripen or soften. The length of time between picking and


215
proportional to photoassimi 1 ation. Girdling decreased the assimilation of label betveen 20 to 50% and the translocation of radiolabeled assimilates from the leaves by approximately 50%. Sucrose and starch content vas greater in the leaves of girdled than non-girdled branches. Flush age did affect photoassimilation and translocation in non-fruit-bearing branches of avocado. The partitioning of recent photoassimilates betveen soluble and insoluble carbohydrate pools in avocado varied betveen flushes and vas dependent on fruit-bearing status.
Carbohydrate limitation did not appear to influence avocado fruit abscission. Approximately 80% of the measured fruit in this study abscised regardless of fruit load on the branch. Furthermore, during the peak period of abscission, assimilates from the proximal source leaves vere supplying assimilates to both the acropetal sinks (fruit and developing leaves) and the basipetal sinks (roots). Avocado leaves exported assimilate at an early stage of development suggesting that the period of competition betveen fruit and developing leaves vas shorter than anticipated. Finally, fruit-bearing branches vere more efficient at fixing and translocating recent photoassimilates. Vhen these branches vere girdled, translocation decreased suggesting that the leaves provide more than enough assimilate for the developing fruit.


91
The fruits exhibited the lowest RSA (.49 Xdpm %mg dry mass"1) and sink intensity (1.8 %dpm dry mass"1) of any reproductive tissue (Table 5-1), yet because of their mass they had the greatest sink strength (35.2 %dpm). Brun (23) attributed the low sink intensity of post-anthesis fertilized soybean flowers to the availability of cytoplasmic starch grains. The reason for the low sink intensity in very small fruits of avocado is unknown; however, small avocado fruit have a very high respiratory rate (4,186) and C02 derived from translocated assimilates may be lost in respiration.
Respiratory losses in small avocado fruit (4,186) offset photosynthetic gains (186). Respiration and carbon fixation (Fig.5-2) correlated to the mass of the avocado fruit (186). In citrus, mass and respiration correlated only in young fruit (186). Todd et al. (186) suggested that the bimodal respiration curves of citrus corresponded to the transition from cell division to cell expansion. Since avocado fruit undergo continuous cell division (168) throughout development this bimodality would not be expected.
Avocado fruit fixed less than 4% of the radiolabel available to a branch (Table 5-4). Carbon fixation was correlated to fruit mass and diameter (Fig. 5-2); other studies using detached fruit (186) found that photosynthesis per unit area remained constant in growing avocados. Although avocado fruit photosynthesis clearly does not play a major role in fruit growth, it may be involved in some regulatory processes. Avocado fruit preparations contain RuBisCo and PEPCase (33). PEPCase activity has been noted in several plants (54,196) during early fruit development, particularly in fruit whose respiratory


23
Table 2-1. Thin layer chromatography of control sugars and leaf/vein extracts and phloem exudates.
Chloroform: Acetone:Water
Acetic acid : Water
Rf color* Rf
Leaf .252 brown .492
.333 orange .79
Vein .249 brown .492
.333 orange .79
Phloem exudate .249 brown
.333 orange
Invertase treatments
Sucrose .361 blue-grey .516
.385 orange .543
Leaf .333 orange .79
.361 blue-grey .516
.385 orange .543
Vein .333 orange .79
.361 blue-grey .516
.385 orange .543
Sucrose .249 brown .492
Mannoheptulose .333 orange .79
Glucose .361 blue-grey .516
Fructose .385 orange .543
Invertase ----
Phloem exudates were obtained with 5 ul capillary pipettes from cross-sectional cuts through the leaf petiole. Leaves were separated into laminate and major vein portions and dropped into boiling 85% EtOH. The leaves and veins were boiled for 15 and then filtered. The residue was washed with 1 volume of 50% EtOH followed by 1 volume of water. The extract was rotoevaporated and then spotted on pre-activated TLC plates. Samples were separated into vials for control and invertase treatments. Acidified extracts were treated with invertase (final concentration 1 mg/ml) and incubated at 4C for one hour. The pH of the sample was then increased to 7.5 and then spotted on pre-activated TLC plates. The plates were developed in either a chloroform:acetic acid:water (3:3.5:0.5) or an acetone:water (9:1) solvent system. When the solvent front had moved at least 10 cm the plates were removed and air dried for 30 min. The plates were sprayed with analine diphenylamine in acidified acetone and heated at 80C for 30 min. Analine diphenylamine in acidified acetone is a differential spray for the detection of non-reducing sugars.


101
0 080
0.070
0.060
0.050
6
0
PS
Q
040
0.030
0.020
0.010
0.000
3000
2500
2000
1500
1000
500
M
D
\
DIAMETER (mm)
?
dpm/FRUIT DRY MASS


181
(49,50,161) contend that fruit influence CO^ assimilation via leaf conductance and not by feedback mechanisms related to starch accumulation.
In avocado, leaf flush development and fruit development coincide for much of the season. Reproductive flushes are first evident in late January. The inflorescence emerges in a pseudoterminal position. Leaves develop at the apex as the inflorescence expands. These developing tissues are supplied assimilates by the older leaves proximal to the inflorescence (Chap. 6). The leaves distal to the inflorescence mature and become the primary supplier of photoassimilates for the developing fruit and for another apical flush (Chap. 7). This final flush begins to develop in early June. For the purpose of simplicity, the most distal flush vill be designated as the third flush, the flush immediately distal to the inflorescence vill be the second flush and the flush proximal to the inflorescence vill be the first flush.
This and a companion chapter on the flushes of non-fruit-bearing branches (Chap. 8) are part of a series of reports which address carbohydrate involvement in the growth and abscission of avocado fruit (Chap. 2 and 3). There were several immediate goals for this research. First the assimilation, translocation, and carbohydrate content of fruit-bearing and defruited branches was measured to determine which flush was the most active in photoassimilation and translocation. Second, the starch and sucrose content of the leaves in each flush was measured so that carbohydrate pool sizes could be compared in leaves of different ages. Third, sink ratios were altered by girdling and defruiting to determine how carbohydrate metabolism was
affected.


Figure 9-2. Specific Activity of Avocado Leaf Tissue from Three Successive Leaf Flushes on Fruit-Bearing and Defruited Branches. Leaf samples vere taken from each leaf of each flush on both girdled and non-girdled; fruit-bearing and defruited branches immediately after the 1 hr labeling period. These samples vere dried, veighed, digested and the radiolabel quantitated as described in the Materials and Methods. Leaf disk (0.618 cm') dry mass and label incorporation after 1 hr exposure to CO of all the leaves in the flush vere used to calculate the mean incorporation of label by each of three consecutive, girdled or non-girdled flushes. The fruit vere removed from defruited branches one veek prior to labeling. Branches vere approximately 70-80 cm in total length and 1 to 1.5 cm in diameter. Branches vere girdled proximal to the oldest flush, adjacent to the main scaffolding branch. Three branches vere chosen, 1 each from the lover, middle and upper canopy positions. The experiment vas repeated four times.


Figure 8-9. Relative Starch Content of Avocado Leaves from Three Successive Growth Flushes on Girdled vs. Non-Girdled Branches. Leaf disks (0.618 cm') were removed from each leaf of each flush at harvest. The samples from each flush were pooled, chopped and boiled in 85% EtOH. The residues of the ethanolic extracts were digested by the method of Housely et al. (102) and the starch quantitated by the phenol-sulfuric acid method. The mean starch content for each entire flush was calculated.


80
Avocado flowers have light, yellowish green petals. Their opening behavior has recently been reviewed by Davenport (46). The flowers are perfect, possessing both stamens and a single pistil. In the case of 'Petersen', a Type A cultivar (46), the flowers initially open in the morning as functionally female flowers. These flowers close at midday and then reopen the next day in the afternoon as functionally male flowers. Mature avocado trees produce up to 1.6 million flowers per tree, but only 0 to 0.66 percent of these flowers yield mature fruit (27). This large quantity of flowers may represent a substantial carbon expenditure; thus, the potential for assimilate competition among flowers and between flowers and fruits exists (20,156). A photosynthetic flower could benefit with regard to sink competition for stored or limited photosynthates. In companion reports, we describe the distribution of recent photoassimialates from source leaves into reproductive organs and other sinks of avocado. This report describes the distribution of those recent photoassimilates among and within buds, flowers and fruitlets.
Materials and Methods Ten year old trees of Persea americana Mill. cv. 'Petersen' growing on 'Walden' rootstock at the Tropical Research and Education Center, Homestead, FL were used in this study. Tissues to be labeled were enclosed in a sealed plastic bag and fed with 20 uCi of C02 for a one hour period between 0900 and 1100 a.m. 14C02 was generated from NaH14C03 (Amersham, specific activity 54 mCi mmol"1) using equimolar amounts of 5% lactic acid.




159
Table 8-1. Relative Pool sizes of Soluble Radiolabeled Compounds in Avocado Leaves.
Samples were taken from leaves at various times points and processed by the method of Housely et al. (102). The neutral fraction represents the sugars; the basic fraction consists of amino acids; organic acids and sugar monophosphates are included in the acid 1 fraction; phosphoglyceric acid and sugar diphosphates are included in the acid 2 fraction. These results are presented as percentages of the total radioactivity recovered from the columns. Time is given in hours post-labeling.
Time Neutral Basic Acid 1 Acid 2
24 52 78
0
88.8 +/- 4.1
86.6 +/- 6.6 83.3 +/- 6.3
76.7 +/- 7.5
8.56 +/- 4.2 6.5 +/- 1.7
9.5 +/- 0.2
1.53 +/- 1.6 4.37 +/- 2.3 4.82 +/- 4.0 4.95 +/- 4.0
0.12 +/- 0.02
3.47 +/- 2.5
2.18 +/- 1.8 5.32 +/- 3.6
11.27 +/- 1.7


74
Table 4-4. Fruit Growth
Fruit were measured twice daily at 0800 and 1600h with a fruit caliper. Fruit volume (cm ) was calculated using a sperical model as described in Chap. 3.
Experiment 1: Average increase in volume before piercing. Experiment 2: Average increase in volume after piercing.
SINGLE
DOUBLE (PIERCED)
DOUBLE
(NOT PIERCED)
Experiment 1
2
13.50 + 6.79 1.11 + 5.26
5.69 + 4.84 2.85 + 4.13
9.11 + 8.40 3.32 + 4.67


59
production and abscission. Sedgely (172) suggested that seed degeneration occurs during the 7 to 8 day period between the cessation of growth and the separation of fruit from the tree. Davenport and Manners (47) showed that senescence of the nucellus and the seed coat are relatively rapid events. Senescence of the nucellus and seed coat was first visually noted within 33 hours after excision of the fruit-bearing branches from trees, and abscission of the fruit occurred 2 days after the onset of senescence. Accordingly, they concluded that seed coat senescence occurs early in the scheme of events and plays an integral role in abscission. Ethylene production which is a product of the senescence process is responsible for the induction of flower and frui tiet abscission.
Work in this laboratory indicated that fruit which abscised within
1 4
4 days of source leaf labeling ( C02) contained no radiolabeled assimilates. Fruits which abscised between 4 and 8 days after source leaf labeling contained radiolabel; however, a greater percentage of the total label was recovered in the seedcoats of abscised versus non-abscised fruit. The distribution of label in fruits which abscised after 8 days was identical to the non-abscising controls. In this study the abscised fruit exhibited seed coat discoloration but no other damage or gross anatomic abnormalities were evident.
This research was part of a larger effort to delineate the mechanisms of the extensive and premature abscission of avocado fruit. The specific objectives of this research were two-fold. First, the accumulation of labeled recent photoassimilates by the tissues of control and puncture-induced-abscising fruit will be compared to


220
60. Evans LT, HM Ravson (1970) Photosynthesis and respiration by the
flag leaf and components of the ear during grain development in wheat. Aust J Biol Sci 23:245-254
61. Evert RF, WE Eschrich, SE Eichhorn, ST Limbach (1973) Observations
on penetration of barley leaves by the aphid Rhopalosiphum maidis Fitch. Protoplasma 77:95-110
62. Evert RF, V Eschrich, JT Medler, FJ Alfieri (1968) Observations on
penetration of linden branches by stylets of the aphid Longistigma caryae. Amer J Bot 55:860-874
63. Faragher JD, S Mayak (1984) Physiological responses of cut roses
flowers to exposure to low temperature: changes in membrane permeability and ethylene production. J Exp Bot 35:965-974
64. Faragher JD, E Vachtel, S Mayak (1987) Changes in the physical
state of membrane lipids during senscence of rose petals. Plant Physiol 83:1037-1042
65. Fellows RJ, DR Geiger (1974) Structural and physiological changes
in sugar beet leaves during sink to source conversion. Plant Physiol 54:877-885
66. Flinn AM (1974) Regulation of leaflet photosynthesis by developing
fruit in the pea. Physiol Planta 31:275-278
67. Fondy BR, DR Geiger (1981) Regulation of export by integration of
sink and source activity. What's New in Plant Physiology 12:33-36
68. Forney CF, PJ Breen (1985) Dry matter partitioning and
assimilation in fruiting and deblossomed strawberry. J Amer Soc Hort Sci 110:181-185
69. Garcia-Martinez JL, MA Garcia-Papi (1979) Influence of gibberellic
acid on early fruit development, diffusible growth substances and content of macronutrients in seedless Clementine Mandarin. Sci Hort 11:337-347
70. Gazit S, A Blumenfeld (1970) Response of mature avocado fruits to
ethylene treatments before and after harvest. J Amer Hort Sci 95: 229-231
71. Geiger DR (1979) Control of partitioning and export of carbon in
leaves of higher plants. Bot Gaz 140:241-248
72. Geiger DR, R Giaquinta, SA Sovonick, RJ Fellows (1973) Solute
distribution in sugar beet in relation to phloem loading and
translocation. Plant Physiol 52:585-589
73. Giaquinta R (1978) Source and sink leaf metabolism in relation to
phloem translocation. Carbon partitioning and enzymology. Plant Physiol 61:380-385


40
effect on the growth of fruit. Fruits on multiple-fruited branches all grew at the same rates even with different phyllotactic relationships to the source leaves.
Competition between fruits for available assimilates appeared to have very little influence on abscission in avocado. Hypothetically, a single fruit on an inflorescence should have a greater likelihood of reaching maturity since it is the sole recipient of assimilates. Conversely, competition should exist on multiple-fruited inflorescences. However the probability (Table 3-1) of abscission for single fruit (75%) is just as great as for fruit present in pairs (83%) or for individual fruit (79.5%) on a multiple-fruited inflorescence. Greater than 50% of the abscised fruit were less than 1 cm in diameter. The abscission of many of the fruitlets less than .5 cm could possibly be attributed to infertile or abnormal ovules (172). The causes of larger fruit abscission are still unknown, although it is apparently a "whole tree" and not a "local" phenomenon.
Abscising avocado fruit were indistinguishable from retained fruit until 4 to 5 days before separation, unlike orange and peach fruits (212) where growth ceased 20-30 days before separation. Furthermore, according to the literature available on avocado fruit abscission, fruit growth ceases anywhere between 4 days to 3 weeks prior to separation. This paper supports a shorter period between growth cessation and fruit separation and agrees with the abscission kinetics observed by Davenport
V
and Manners (47). They suggested that the cessation of growth results from the ongoing process of abscission and not vice versa. Other work from this laboratory (Chap. 4) on carbohydrate availability and restriction during abscission indicated that recent photoassimilates


4 8 #/c
1.5/
.6 9%
50/c
.31 %
2 0mm


CHAPTER 1 GENERAL BACKGROUND AND HISTORY
The avocado belongs to the Lauraceae family and is one of the few commercially important members of the genus Persea. Three races or ecological groups of avocado have been defined (15): Mexican, Guatemalan and Vest Indian. The Mexican race originated in the mountains of Mexico and Central America. The Guatemalan race is native to the highlands of Central America and the Vest Indian race is a native of the lowlands of Central America and northern South America. However, many of the commercial cultivars are hybrids of these races and the mixed genotypes have permitted a wider geographic distribution than the original races.
Avocado cultivars of the Mexican race are adapted to the cool temperatures of the tropical and subtropical areas with a Mediterranean-type climate. They are the most cold tolerant of the three races and can survive temperatures below -7.8 C (15). Some cultivars survive temperatures as low as -10C (15). The leaves characteristically have an anise odor. The fruit has a thin smooth skin and is relatively small ranging in size from 75-300 g. The fruit matures earlier than or simultaneously with the Vest Indian cultivars.
The Vest Indian race cultivars are well adapted to lowland tropical conditions, i.e., high rainfall and high temperatures. However, this race is highly cold sensitive, being severely injured
1


209
cj
<
CO
[14
O I*
>
Cm
CO
800
600
400
200
0
x X
XXX'.
girdled +fruit
nxxx:
'XXX/
X X XX X XX X /*
//
/
A AT AT
/
X
X
X A
////
x
x x
f X X X
#2%
non-girdled +fruit
defruited
branch type
flush 3
flush 2
-.
flush 1


214
to the inflorescence, as indicated by assimilate export, suggested that the developing leaves were autonomous earlier than previously thought. Subsequently, the period of possible leaf/fruit competition would be diminished.
Tropical evergreen trees frequently undergo several growth flushes each year. Persea americana cv Mill, growing in Florida commonly undergo 2-3 growth flushes a year; one prior to floral initiation, a second immediately post-anthesis, and a third approximately 8 weeks post-anthesis. Fruit-bearing branches of avocado consisting of 3
1 4
flushes were fed CO and then characterized for translocation rates
2
and carbohydrate content. The leaves on non-girdled reproductive and defruited branches translocated radiolabeled assimilate at similar rates. Girdling decreased translocation of radiolabeled assimilates from flushes by 15-20%. The leaf specific activity was 2-times greater in the most recent flush than in the oldest flush. Leaf dry mass increased in parallel with increased photoassimilation by leaves on fruit-bearing branches. Sucrose content was higher in girdled than non-girdled branches. Photoassimilation and assimilate movement within mature non-fruit-bearing flushes was also examined. Non-fruit bearing branches of avocado consisting of 3 recent flushes were fed C02 and then characterized for translocation rates and carbohydrate content. Leaf specific activity was 3-times greater in the most recent flush than in the leaves of the oldest flush. The youngest flush fixed greater than 50% of the total available label, while the oldest flush fixed less than 20% of the available label. Relative pool sizes were identical in all three flushes although the absolute amount of soluble sugars was greatest in the youngest flush. Leaf dry mass was inversely


10
Inhibition and senescence in relation to fruit abscission are actually two separate phenomena. Inhibition of abscission results from the interaction of hormones (3,24). The secretion of IAA, in particular, by leaves, fruits and other storage organs is proposed as one mechanism whereby abscission is inhibited. Abortion and senescence resulting in abscission occur when the organ in question physiologically degenerates. Degeneration is the result of improper development or a physiologically weak embryo (3).
The hypothesis that organ competition results in abscission stems from the concept that plants initiate and partially develop more leaves, buds, flowers and fruits than their stem and root systems can support. This results in competition between sinks for light and nutrients. This competition extends to vegetative and reproductive growth. Whether a fruitlet reaches maturity depends on its ability to remain a strong sink. If the fruitlet can not successfully compete, its reserves are remobilized (fruitlet now becomes a source) and eventually the fruitlet abscises.
Assimilate partitioning and translocation have been studied in several crops. Koch and Schrader (118) examined the allocation of 14C-photosynthates in soybean from anthesis to harvest. They found that the amount of 14C-labelled amino acids translocated from the source leaf to the pod varied with the developmental stage of the pod and these amino acids comprised 12-48% of the seeds' nitrogen. One-third of the 14C-label present in the leaf blade was in the form of starch. During the late pod filling stage; however, this proportion


192
Table 9-1. Summary of Avocado Branch Characteristics.
Each branch was dissected at harvest and the leaves were counted and measured. Lamina length equals that of the central vein in the blade portion of the leaf (petiole excluded). Area was determined using a leaf area meter.
Flush # # of Leaves
Lamina Length(mm)
Area (cm2)
# of Branches
Frui t-bearing
3 2 1
7.0 + 4.1 10.4 + 4.0 7.0 + 2.0
77.0 + 12 79.7 + 8.0 78.2 + 9.1
157 + 25.3 264 + 48.4 194 + 64.2
8
Defrui ted
3 2 1
9.0 + 3.3 10.0 + 4.4 7.0 + 3.1
78.4 + 6.2 79.4 + 3.5 81.7 + 5.8
175 + 303 214 + 73 206 + 34


156
significantly different between flushes at specific time points (Table 8-2). This suggested that recent photoassimilates were incorporated into various soluble pools in the same proportions.
Similar distributions of sucrose were noted in both girdled and non-girdled branches when based on area measurements (Fig. 8-6). The sucrose specific activity (Fig. 8-7) values for the distal flushes were identical on both the girdled and non-girdled branches suggesting that girdling within 24 hours did not negatively affect labeling kinetics or pool sizes in the youngest leaves. The second flush contained the most sucrose by area and the first flush the least sucrose. Girdling appeared to have little effect on sucrose specific activity (Fig. 8-7), except for slightly elevated values found in the leaves of the older flushes (1 and 2) of the girdled branches. The decreased translocation rate and the smaller sucrose pool size in these older flushes (Figs. 8-6, 8-8), coupled with the incorporation of label, even at the reduced rate noted in girdled branches, would elevate the pool specific activity. Alternatively, the older flushes on the girdled branches could act as an alternate sink or storage area (29, 39) for assimilates translocated from the younger flushes since girdling the branch removed the branches primary sink. This second alternative is supported by increased sucrose content observed in the older flushes of girdled branches.
In longer term, batch studies the leaves on girdled, non-fruit-bearing branches accumulated 50% more starch than similar non-girdled non-fruit-bearing branches (data not shown). Starch content of the leaves on the girdled branch peaked 6 days after girdling. The specific activity of starch extracted from the leaves of


127
appear as a gradual increase in export out of the leaf. Similar results were obtained in beet where phloem loading as determined by autoradiography (65) began at 25% final midvein length; however, export from the leaf was not observed until 35 percentage final midvein length.
Export from avocado leaves reached a plateau at a slightly later developmental age (60% final midvein length) than other plants (65,72,117,203). In these other studies export commenced between one third to one half leaf expansion and reached a plateau at approximately 50% final lamina length. Vhen avocado leaves were 65% final midvein length the rate of export reached a plateau of 56 +/- 12% of the daily fixed carbon. This value remained constant during continued leaf expansion. The rate of export or the daily export of carbon varies with the plant. Mature cucumber leaves export approximately 80% of their daily fixed carbon (95) and tomato leaves export around 70% (94),
while beet leaves translocate only 25% (65) of their daily fixed carbon.
Between 20 and 40% final midvein length, the leaf was both exporting and importing assimilates (Fig. 7-1). The bi-directional movement of assimilates has been noted elsewhere (16,111,189) and generally occurs in leaves and petioles that are approximately half-grown (65,189). The period of export/import overlap is generally very brief. During this time sink, source and transition regions exist in a single leaf.
Net carbon fixation was noted in avocado leaves as small as 6.6% final midvein length (Fig. 7-2). Some workers have found net losses of carbon in very small leaves in the light (42), while others have found


227
160. Schaffer B, JA Barden, JM Williams (1985) Partitioning of
( C)-photosynthate in fruiting and deblossomed day-neutral strawberry plants. HortScience 20:911-13
161. Schaffer B, L Ramos, SP Lara (1987) Effect of fruit removal on net
gas exchange of avocado leaves. HortScience 22:925-927
164. Scholefield, PB, M Sedgley, D Alexander (1985) Carbohydrate
cycling in relation to shoot growth, floral initiation and development and yield in the avocado. Scientia Hort 25:99-110
163. Scholefield PB, JJ Valcott, PE Kriedemann, A Ramadasan (1980) Some
environmental effects on photosynthesis and water relations of avocado leaves. Yrbk Calif Avoc Soc 64:38-43
164. Schroeder CA (1945) The avocado inflorescence. Yrbk Calif Avoc Soc
28:39-40
165. Schroeder CA (1951) Flower bud development in the avocado. Yrbk
Calif Avoc Soc 35:159-163
166. Schroeder CA (1952) Floral development, sporogenesis, and
embryology in the avocado, Persea americana. Bot Gaz 113:270-278
167. Schroeder CA (1952) Growth and development of the Fuerte avocado
fruit. Yrbk Calif Avoc Soc 36:103-109
168. Schroeder CA (1953) The growth and development of the Fuerte
avocado fruit. Proc Amer Soc Hort Sci 61:96-99
169. Schroeder CA (1957) Growth and development of the avocado fruit.
Yrbk Calif Avoc Soc 41:114-118
170. Shroeder CA (1985) Physiological gradient in avocado fruit. Yrbk
Calif Avoc Soc 69:137-143
171. Schroeder CA, PA Vielannd (1956) Diurnal fluctuations in size in
various parts of the avocado tree and fruit. Yrbk Calif Avoc Soc Yrbk 40:253-258
172. Sedgley M (1980) Anatomical investigation of abscissed avocado
flowers and fruitlets. Ann Bot 46:771-777
173. Sharkey PJ, JS Pate (1975) Selectivity in xylem to phloem transfer
of amino acids in fruiting shoots of white Lupin (Lupinus albus L.). Planta 127:251-262
174. Sharon 0, Kahn V (1979) Browning potential, PP0, catalase and acid
phosphatase activities during ripening of non-chilled and chilled avocado. J Sci Food Agric 31:634-638


93
Table 5-1. Distribution of Radioactivity Supplied by Proximal Source
Leaves to the Avocado Inflorescence.
The following data represent a composite of at least 20 inflorescences. Sink intensity is defined as the relative distribution of total radioactivity on a Xdpm mg dry mass basis. Dry mass is given in grams. Percentage C or the sink strength is the percentage of the total C recovered from the reproductive tissues. The relative specific activity (RSA) is defined as the % dpm % dry mass" .
Tissue Dpm Dry Mass dpm/mg %14c Sink Intensi ty RSA
Pedicel 3025840 0.4068 7438 18.90 10.3 2.9
Buds
Very Small 949771 0.1636 5805 5.90 8.1 2.2
Small 2155190 0.3185 6767 13.50 9.4 2.6
Medium 945325 0.2927 3230 5.90 4.5 0.8
Bud Scales 130580 0.0514 2540 0.8 3.5 1.0
Flowers
Open
Pistil 109078 0.0282 3868 0.7 5.5 1.4
Stamens 576404 0.0822 7012 3.60 9.7 2.8
Nectary 612686 0.0730 8393 3.80 11.6 3.2
Corolla 914801 0.1778 5145 5.70 7.1 1.9
Mature
Pistil 40799 0.0172 2372 0.3 3.3 0.8
Stamen 226814 0.0417 5439 1.40 7.5 2.0
Nectary 116555 0.0298 3911 0.7 5.4 1.5
Corolla 571739 0.0660 8863 3.60 12.3 3.6
Fruit 5630112 4.3966 1281 35.20 1.8 0.5


201

C
CO H
i-H
tn <1)
CO X)
D
r-H


l-H cH

o H
c


u


(G
CO
W c
H
< I/)
-J
u
1__1 a
r1 CO
CO
< 4J
O
CJ 4-)
1
M-l
rH O

hours after labeling
-d flush -Lr ~ flush o flush
-a flush
is
3 girdled
2 girdled 1 girdled
3 non-girdled
flush 2 non-girdled flush 1 non-girdled


203
as
o
M to
&
CO
u
33
CO
D J
O CP
to
CO
O
CJ D
CO
6
120
\
\
\
\
\
\
X
/
;\\\
/////
?
' ///
*
/ / /
/ ///}
s / / /
7/
y / / /
- / /
'/A
\\\
girdled +fruit
non-girdled + fruit
defruited
branch type
flush 3
flush 2
flush 1


4
tissue fades away. The vascular supply to the ovary consists of a dorsal trace, two or more lateral traces and two ventral traces which are united as a single vascular bundle.
The ventral trace enters the funiculus of the ovule and runs down the opposite side of the ovule near the dorsal part (41). This bundle branches profusely providing the integument with an ample vascular system. These branches are evident in the seed coats of mature fruits. The two dorsal and two lateral carpellary traces extend up through the style. The dorsal trace extends into the stigma. The lateral traces disappear in the style.
The flower although perfect is dicogomous. The pistil matures before the pollen is shed. Each flower opens twice at a distinctive time (179). At the first opening the pistil is receptive but the pollen is not shed. At the second opening the pollen is shed but the pistil is no longer receptive. The first and second opening under normal conditions are separated by one night. The distinctive time of opening has led to the formation of two classes, i.e., class A type and class B type. Class A cultivars open for the first time in the morning and for the second time during the afternoon of the next day. Class B cultivars open first in the afternoon and then reopen the next morning. In both cases during the first opening the flower is functionally pistillate. Even with the presence of active pollinating agents only .1% of the flowers will produce mature fruit (148).
The avocado has three distinct growing periods (44,133,168,169,171). The first period begins at the time of full bloom. The fruit enlarges rapidly and the integuments of the seed


Figure 9-7. Relative Starch Content of Avocado Leaves from Three Successive Growth Flushes on Girdled vs. Non-Girdled Fruit-Bearing and Defruited Branches. Leaf disks (0.618 cm ) were removed from each leaf of each flush at harvest. The samples from each flush were pooled, chopped and boiled in 85% EtOH. The residues of the ethanolic extracts were digested by the method of Housely et al. (102) and the starch quantitated by the phenol-sulfuric acid method. The mean starch content for each entire flush was calculated. The fruit were removed from defruited branches one week prior to labeling.


I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Thomas L. davenpdrt, Chai
Associate Pro
of Horticultural Science
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Karen Koch
Associate Professor of Horticultural Science
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Thnmac Hnmnhrpvc* VJ
Thomas Humphreys Professor of Horticultural Science
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully
Idi
adequate, in scope and qualit-Doctor of Philosophy.
tion for the degree of
Donald Huber
Associate Professor of Horticultural Science
I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy.
-5.
Kenneth Boote Professor of Agronomy


73
Table 4-3. Distribution of Dry Mass and Translocated C-Assimilates in Fruit Tissues.
Source leaves on branches were fed 1HC0 for a one hour period. The fruit were harvested 48 hours later, dissected, dried and processed for scintillation counting. The relative specific activity (RSA) is defined as X C/ X dry mass.
Dry mass #Fruit Tissue X of Total X of Total RSA
(gm) Sampled Fruit Fruit 1 Dry
Recovered Mass
0 0.5 28 Mesocarp 87.9 + 5.6 87.9 + 2.8 1.00
Seed coat 8.7 + 3.9 8.8 + 1.0 0.99
Endosperm 1.9 + 1.0 1.6 + 0.7 1.18
Cotyledon 1.9 + 1.6 1.4 + 1.0 1.35
.501 1.0 20 Mesocarp 87.4 + 5.2 89.4 + 0.9 0.98
Seed coat 8.6 + 3.5 7.4 + 0.7 1.16
Endosperm 1.3 + 0.7 0.9 + 0.3 1.44
Cotyledon 2.7 + 1.4 2.3 + 0.7 1.17
1.01 1.5 4 Mesocarp 83.5 + 4.2 88.3 + 1.6 0.95
Seed coat 9.7 + 3.6 7.9 + 0.8 1.23
Endosperm 1.9 + 0.4 1.1 + 0.3 1.72
Cotyledon 4.6 + 1.3 2.9 + 0.4 1.58


Figure 3-1. Growth Curve for Persea americana Mill. cv. 'Petersen'. This curve is a composite of sequential non-destructive data from at least 10 non-abscising fruit. Volume was calculated from the measured diameter using a spherical model. Volumes were transformed to their natural log values. The slope of this curve d In V/dT is the RGR.


2.5
1 .5
0 5
0 5
1.5
2.5
3.5
4 5
5 5
0.0
5.0
10 0
15 0
20 .0
25 0
30 .0
35.0
40 0
45
FRUIT VOLUME cm*3 (DAY POST-ANTHESIS)


Figure 9-5. Sucrose Content per Unit Fresh Veight of Avocado Leaves from Three Successive Growth Flushes on Girdled and Non-Girdled Fruit-Bearing and Defruited Branches. Leaf disks (0.618 cm2) were removed from each leaf of each flush at harvest. The samples from each flush were pooled, chopped and boiled in 85% EtOH. The ethanolic extracts were concentrated and the sugar content was measured by the phenol-sulfuric acid method. The fruit were removed from defruited branches one week prior to labeling.


61
Puncture Studies
The term, 'single fruit' refers to an inflorescence bearing only one fruit. Similarly, doublet fruit refers to 2 fruit sharing the same inflorescence. Six single fruit and 6 doublet fruit were used in each experiment, which was repeated 4 times.
Fruit were pierced deeply enough to injure the seed coat using a stainless steel wire (gauge #20) either before (24,48,72 and 96 hours), after (24,48,72 and 96 hours) or at the time of source leaf labeling (0 hour). This technique allowed us to arbitrarily induce abscission of individual fruitlets. Both the fruit and the wire were surface sterilized with dilute chlorox solution before and after puncturing the seed coat. Only one fruit on the inflorescence was punctured when two were present. Fruit Growth
Fruit were measured twice daily at their widest point with at fruit caliper.
Results
The vascular system of avocado is asymmetric (19,41) and gradients of dry mass and oil content in the fruit have been noted (170). These gradients are oriented radially around the seed and longitudinally from the stem to blossom end. Recent photoassimilates were distributed symmetrically throughout the radial and longitudinal quadrants of the fruit (Table 4-1). The data were obtained from fruits with different amounts of dry mass and different source leaf-to-fruit positions. The


226
144. Nirody PS (!<>??) Investigations in avocado breeding. Yrbk Calif
Avoc Soc 6:65-78
145. Noel ARA (1970) The girdled tree. Bot Rev 36:162-95
146. Nordal A, AA Benson (1954) Isolation of mannoheptulose and
identification of its phosphate in avocado leaves. J Am Chem Soc 76:5054-5055
147. Ong CK, KE Colvill, C Marshall (1978) Assimilation of 14C0 by the
inflorescence of Poa annua L. and Lolium perenne L. Ann Boi 42:855-862
148. Papademetriou MK (1975) Percentage fruit-set in avocado (Persea
americana Mill). Yrbk Calif Avoc Soc 59:135-142
149. Patrick JV (1972) Vascular system of the stem of the wheat plant
II. Development. Aust J Bot 20:65-78
150. Patrick JV (1972) Distribution of assimilate during stem
elongation in wheat. Aust J Biol Sci 25:455-67
1 4
151. Peel AJ (1964) Tangential movement of C-labeled assimilates in
stems of willow. J Exp Bot 15:104-114
152. Priestley CA (1973) Bases for the expression of the results of
chemical analyses of plant tissue. Ann Bot 37:943-953
153. Quebedeaux B, R Chollet (1975) Growth and development of soybean
(Glycine max L. Merr.) pods. CO exchange and enzyme studies. Plant Physiol 55:745-748 2
154. Quinby JR, PV Morgan (1987) Genetic regulation of development in
Sorghum bicolor IV. GA hastens floral differentiation but not floral development under nonfavorable photoperiods. Plant Physiol 85:615-620
1 4
155. Quinlan JD (1969) Mobilization of C in the spring following
autumn assimilation of CO by an apple rootstock. J Hort Sci 44:107-110 2
156. Quinlan JD, AP Preston (1068) Effects of thinning blossoms and
fruitlets on growth and cropping of sunset apples. J Hort Sci 44:107-110
157. Quinn PJ (1981) The fluidity of the cell membrane and its
regulation. Progr Biophys Mol Biol 38:1-104
158. Reece PC (1942) Differentiation of avocado blossom buds in
Florida. Bot Gaz 104:323-328
159. Schaffer AA, K Liw, EE Goldschmidt, CD Boyer, R Goren (1986)
Citrus leaf chlorosis induced by sink removal:starch, nitrogen, and chloroplast ultrastructure. J Plant Physiol 124:111-121


20
Results and Discussion The goal of the present research was to identify the translocated sugar containing recently photoassimilated C02 and to address the discrepancy between previously reported results. Earlier workers reported either sucrose as the only translocated sugar (211) or suggested both sucrose and mannoheptulose (11,12,146) are formed during photosynthesis and both play a role in translocation. We observed only 2 major sugars, sucrose and mannoheptulose, present in ethanol extracts (Table 2-1), and only sucrose contained the radiolabel during the summer months (Fig. 2-1).
The phloem exudates from the petiole incisions of the tested cultivars contained both mannoheptulose and sucrose. These sugars were present in the phloem in similar amounts as indicated by their almost identical spot sizes on thin layer chromatography plates and as quantitated by the phenol sulfuric acid method after separation on preparative TLC plates. However, sucrose was the only labeled compound in the phloem exudates at 4 and 24 hours (Fig. 2-2). This was true for all races of avocado. These results indicate that sucrose was the sugar responsible for the movement from the leaves of recently photoassimilated carbon and remobilized leaf starch reserves.
Chromatograms of both laminate and vein extracts produced two spots (Table 2-1). One spot had the same mobility as sucrose and comigrated with sucrose in spiked samples. When treated with invertase this spot
|
disappeared and two other spots appeared having the same mobility as fructose and glucose. The second spot was not affected by invertase (Table 2-1). It had the same mobility and staining characteristics as mannoheptulose. This sugar was never associated with the radiolabel during the summer months.


30
u
CP
(T3
O
4J
O
CO
60.0
55.0
50.0
45.0
40.0
\
22
27
\
^^^^
37
\
\
\
V
TEMPERATURE (C)
SUCROSE
MANNOHEPTULOSE


Figure 8-5. Export of Labeled Assimilates form Leaves of Three Successive Growth Flushes on Girdled and Non-Girdled Avocado Branches. Leaf disks taken throughout the experimental period were dried, weighed and digested as described in the Materials and Methods. Translocation was defined as the difference between the amount of radiolabel present immediately after a 1 hr pulse of CO and the given time point during the 75 hr chase.


34
occurs with the onset of nucellar senescence approximately 30 hours after excision from the tree. Exposure of excised fruit to aminoethoxyvinyl glycine did not affect the progression of senescence, but did delay fruit separation. They suggested that ethylene generated by the seed coat is essential for the activation of the cells in the abscission zone, but does not initiate senescence. An as yet undetermined event initiates seed coat senescence which in turn stimulates ethylene evolution and fruit separation.
Abscising and retained fruit may differ significantly in their development. Zucconi et al. (212) have shown that 'Shamouti' oranges stop growing 1-2 weeks before fruit separation. Similarly, the growth rate of peaches also decreased 10-20 days before abscission. The time span between growth termination and the abscission of avocado fruitlets suggested in the literature ranges from 4 days to 3 weeks (1,47,74). Adato and Gazit (1) suggested a 3 week timespan between peak ethylene production and fruit abscission. Sedgely (172) has suggested that seed coat degeneration occurs during the 7 to 8 day period between the cessation of growth and the separation of fruit from the tree. Davenport and Manners (47) however have shown that senescence of the nucellus and integuments is first visually noted within 27 hours after excision from the tree and abscission of the fruit occurs 2 days later. The overall goal of this research was to clarify the mechanism underlying the extensive, premature fruit drop in avocado. The growth of abscising and retained fruit was examined, to determine if abscising and retained fruit have inherently different growth patterns. Secondly, fruit growth in relation to abscission was used to establish a time frame for fruit abscission.


83
described. Fruit were harvested at 24 and 48 hrs after labeling. The fruit were separated into the peel (exocarp), mesocarp, seed coat, cotyledons and endosperm and processed for scintillation counting as described above.
The relative specific activity (RSA) as defined by Mor and Halevy (81,139) reflected the changes in the growth pattern of sinks. The RSA (%dpm Xdry mass"1) was calculated as the distribution of label or current photoassimilate content relative to the distribution of dry mass. Deviations from unity reflected changes in the allocation of assimilate and hence changes in growth patterns. Values greater than
unity indicate that proportionately more assimilate is being utilized by the organ in question in relation to the existing tissue. Sink intensity is defined as the relative distribution of radiolabel on a Xdpm mg~ basis or alternatively as the dpm mg" (139). Sink strength is defined as the the percentage of the total label recovered from the tissues.
Results
Proximal Leaf Labeling
A schematic representation of an early stage of reproductive
development is shown in Fig. 5-1. Typically at this stage of reproductive development, each panicle would be composed of several inflorescences bearing flower buds, flower buds and flowers or flower buds, flowers and fruit. The insert of Fig. 5-1 is an example of the last panicle type since it bears flower buds, open flowers, post-anthesis flowers, and fruit.


219
45. Davenport TL (1983) Ethylene production in avocado flovers and
fruit: its role in senescence and abscission. Proc 10th Ann Meeting Plant Grovth Reg Soc Am:213-215
46. Davenport TL (1986) Avocado flovering. Hort Rev 8:257-289
47. Davenport TL, MM Manners (1982) Nucellar senescence and ethylene
production as they relate to avocado fruitlet abscission. J Exp Bot 33:815-826
48. Degani C, A Goldring, S Gazit, U Lavi (1986) Genetic selection
during the abscission of avocado fruitlets. HortSci 21:1187-1188
49. DeJong TM. (1985) Fruit effects on photosynthesis in Prunus
persica. Physiol Plant 66:149-153
50. DeJong TM (1986) Effects of reproductive and vegetative sink
activity on leaf conductance and vater potential in Prunus persica L. Batsch. Scientia Horticul 29:131-137
51. Dickmann DI (1971) Photosynthesis and respiration by developing
leaves of cottonvood (Populus deltoides Barts.). Bot Gaz 132:253-259
52. Die J van, PML Tammes (1966) Studies on phloem exudation from
Yucca flaccida Hav. III. Prolonged bleeding from isolated parts of the young inflorescences. Koninkl Ned Adad Veterschap Proc 69: 648-654
53. Dubois M, KA Gilles, JK Hamilton, PA Rebers, F Smith
(1956)Colorimetric Method for determination of sugars and related substances. Anal Chem 28:350-356
54. Duffus CM, R Rosie (1973) Some enzyme activities associated vith
the chlorophyll containing layers of the immature barley pericarp. Planta 114:219-226
55. Eaks IL (1966) The effect of ethylene upon ripening and
respiration rate of avocado fruit. Yrbk Calif Avoc Soc 50:128-133
56. Eaks IL (1980) Respiratory rate, ethylene production, and ripening
response of avocado fruit to ethylene or propylene folloving harvest at different maturities. J Amer Soc Hort Sci 105:744-747
57. Elmore CD (1973) Contribution of the capsule vail and bracts to
the developing cotton fruit. Crop Sci 13:751-752
58. Engi BAC (1962) The contribution of different organs to grain
veight in upland svamp rice. Ann Bot 26:529-531
59. Esau K (1950) Development and structure of the phloem tissue. Bot
Rev 5:373-398


157
girdled branches declined vith time. The decline in the specific activity of starch extracted from the leaves paralleled the increase in starch content. The increase in starch during the course of the experiment diluted the radiolabeled pool. These leaves contained betveen 27 and 52 mg starch/gm tissue at the peak and lov point of their diurnal cycle.
Vhen this preliminary vork vas expanded to examine individual flushes, it vas found that the relative starch content vas greatest in the youngest flush (Fig. 8-9). Vhen branches vere girdled the relative distribution of starch shifted in favor of flush 3 and 2. This vas not surprising since these flushes initially assimilated 90% of the available label. The specific activities of starch (Fig. 8-10) extracted from the leaves vas greater for the flushes on the non-girdled branches than for their girdled counterparts. Likevise, the specific activity values for starch extracted from flush 3 vere greater than for flush 1 or 2.
Avocado provides an interesting model for assimilate partitioning for many reasons, not the least of vhich is the presence of multiple leaf flushes on a branch during fruit development. Assimilation and translocation in non-fruit-bearing branches exhibited an age dependent correlation. The youngest, mature flush fixed the greatest proportion of available label and translocated the recently fixed assimilate the fastest. Although the relative pool sizes (Table 8-1) vere the same for each flush the absolute sucrose content vas greatest in the youngest flush. The relative starch content of the youngest flush vas also greater than the other tvo flushes. Girdling decreased the absolute assimilation of label by all of the flushes and decreased the


169
100
CP
c
H
CO r1
w
<
rI
*J

n
4->
O H
c
M

M


03
CO
W C
H (D
< Ifl
(1)
M u

CO
CO
< 4J
O
u
1

O

HOURS AFTER LABELING
- -A- FLUSH 2 GIRDLED
FLUSH 1 GIRDLED
? FLUSH 3 NON-GIRDLED
A FLUSH 2 NON-GIRDLED
-1 )- FLUSH 1 NON-GIRDLED
Q FLUSH 3 GIRDLED
NEXT


37
phase of development which lasted approximately one month at which time the fruit volume was less than 4 cm (Fig. 3-1). The final phase of development, which was characterized by large increases in volume and smaller increases in dry mass, began approximately 1 month post-anthesis and lasted until harvest.
Relative growth rates were calculated using natural log transformations of the data after the manner of Hunt (106). The RGR (Fig.3-3) was highest during the early season having a maximum value of .18921 cm3 cm"3 d"1 and lowest late in the season with a value of .015034 cm3 cm"3 d"1. The RGR describes the accumulation of volume in relation to the bulk of the tissue. Defined in terms of RGR, the fruit grew more rapidly immediately after anthesis than later in the season;
however, an opposite statement can be made in terms of AGR.
The AGR increased from 0 to approximately .2 cm3 d~ during the first month post-anthesis. Between 30 and 110 days post-anthesis the AGR increased and reached a maximum value of 2.2897 crn d The AGR decreased between 110 days post-anthesis and harvest. The curve of the AGR suggested that fruit growth occurred in 3 phases; the initial phase (0-30 days), the second phase (30-110 days) which was characterized by a linear increase in growth rate and the final phase in which a linear decrease in the growth rate was observed. The initial phase ended when the fruit were about 2 cm in diameter. The fruit increased in diameter from 2 to 6 cm during the second phase and from 6 to 11 cm in the final phase.
The arrangement and number of fruit on a branch did not affect the growth rate or fruit abscission. Frequently, fruit of similar size were retained on the inflorescences (Fig. 3-4). The growth curves of the


Figure 7-6. The Distribution of Assimilates to Fruits when there were no Fruit with the Same Phyllotaxy as the Source Leaf. A single leaf (black-dotted) distal to the fruits was labeled. The fruit were harvested and processed for scintillation counting 24 hours after labeling. The percentages indicate the mean proportion of the total fruit radioactivity associated with the fruit at each stelar location. The standard error for these measurements was 10% or less. This is a representative schematic which assumes fruit are present at 4 of the 5 stelar locations. Fifteen branches were used in this study.


230
203. Webb JA, PR Gorham (1964) Translocation of photosynthetically
assimilated C in straight-necked squash. Plant Physiol 39:663-667
204. Young TV, RCJ Koo (1977) Effects of age, position, and fruiting
status on mineral composition of 'Tonnage' avocado leaves. J Amer Soc Hort Sci 102:311-313
205. Zauberman G, M Schiffmann-Nadel (1972) Respiration of whole fruit
and seed of avocado at various stages of development. J Amer Soc Hort Sci 97:313-315
206. Ziegler H (1974) Biochemical aspects of phloem transport. Symp.
Soc. Exp. Biol. XXVIII. Transport at the Cellular Level. Cambridge University Press, Cambridge, MA
207. Zilkah S, I Klein (1987) Growth kinetics and determination of
shape and size of small and large avocado fruits cultivar 'Hass' on the tree. Sci Hortic 32:195-202
208. Zilkah S, I Klein, S Fei^enbaum, SA Veinbaum (1987) Translocation
of foliar-applied urea N to reproductive and vegetative sinks of avocado and its effect on initial fruit set. J Amer Soc Hort Sci 112:1061-1065
209. Zimmerman MH (1960) Transport in the phloem. Ann Rev Plant Physiol
11:167-169
210. Zimmerman MH (1969) Translocation of nutrients. In:The physiology
of plant growth and development. p34-47, ed. MB Vilkins Mcgraw Hill, Maidenhead, England.
211. Zimmerman MH, H Ziegler Appendix III: List of sugars and sugar
alcohols in sieve-tube exudates. In: Encyclopedia of plant physiology Vol 1 Transport in plants, ed. MH Zimmerman, JA Milburn. Springer-Verlag, New York.
212. Zucconi F, SP Monselise, R Goren (1978) Growth-abscission in
developing orange fruit. Scientia Hort 9:1376


64
than in the control fruit. Fruit pierced at the same time or 24 to 48 h after labeling had a greater percentage of label in the mesocarp than did control fruit. Data for fruit punctured 72 and 96 hours before labeling are not shown. These fruit abscised soon after labeling and contained no radiolabel. Data for fruit punctured 72 and 96 hours after labeling are also not shown. The relative percentage radioactivity in these fruit did not vary from control values.
The distribution of radiolabel in the pierced fruitlet on the double-fruited inflorescence is shown in Fig. 4-2. Data for the unpierced fruitlet were identical to the control.
In single fruit the percentage of radiolabel was significantly lower in the mesocarp and higher in the seed coat than in the control when the fruit was punctured 48 or 24 hours before labeling. The percentage of radiolabel was higher in the cotyledons and endosperm than in the control tissues if punctured 48 and 24 hours before labeling and lower than the control when the fruit was pierced at 0,24 or 48 hours post-labeling. Puncturing the seed coat of a single fruit on an inflorescence decreased its growth rate by 92%. The growth rates of both fruit on the doubly-fruited inflorescence were affected when only one was punctured. The growth rate of the punctured fruitlet decreased 50%, while the growth rate of the non-punctured fruitlet decreased 36%.
Discussion
The anatomy of the developing avocado fruit has been described by Cummings and Schroeder (41). The avocado seed develops off center producing a thin and a thick side to the fruit. The six major strands of vascular tissue are asymmetrically distributed (19,41). The vascular


Figure 9-4. Export of Labeled Assimilates form Leaves of Three Successive Growth Flushes on Girdled and Non-Girdled Fruit-Bearing and Defruited Avocado Branches. Leaf disks taken throughout the experimental period were dried, weighed and digested as described in the Materials and Methods. The fruit were removed from defruited branches one week prior to labeling. Translocation was defined as the difference between the amount of radiolabel present immediately after a 1 hr pulse of CO and the given time point during the 75 hr chase.


35
Materials and Methods Avocado (Persea americana Mill, cv 'Petersen') trees used in these experiments were 10-year-old grafted trees growing in Rockdale soil at the Tropical Research and Education Center, Homestead, FL. Individual fruit were tagged during February and March of 1986. Tagged fruit were less than 1 cm in diameter. Fruit diameters were measured at their widest point with a fruit caliper accurate to 0.1 mm. They were measured every other day between 0830 h and 1030 h. Fruit were gently pulled before they were measured to determine if the fruit removal force had decreased. Fruit which easily separated from the tree when pulled were considered abscised. Fruit comparable in size to the tagged fruit were harvested twice during the season to measure diameter and dry mass
accumulation.
The number and arrangement of fruit on each inflorescence was noted. A total of of 54 inflorescences (Table 3-1) were observed during the growing season. There were 32 singly-fruited inflorescences. There were 22 multiple-fruited inflorescences bearing a total of 74 fruit.
The growth of avocado fruits has been measured directly by displaced volume (18,133), dry mass (11) and diameter (18,91,171,207). Combinations of these parameters are commonly used since the avocado has an irregular shape which cannot be easily described by a common geometric form. Because 'Petersen' is nearly spherical in shape, volume was calculated from the measured diameter using a spherical model. The actual volumes of the fruit remaining at harvest were measured by water displacement. Actual measurements did not vary by more than 6% from those calculated using the measured diameter and a spherical model.


Table 6-1. Radiolabel Distribution to the Distal Tissues from the Labeled Source Leaves Located Proximal to the Avocado Inflorescence.
1 4
Source leaves, proximal to the inflorescence, were fed CO^ for a one hour period. The entire branch was harvested 48 hours later. The branch was dissected, dried and processed for scintillation counting. The reproductive tissues include the bud scales, flower buds, open flowers, senescent flowers, receptacles and peduncles. Mass is given in grams.
Dry Mass (g) 1 14C-Assimilate Localization (dpm)
Mean Reproductive Leaf | Reproductive Leaf Total Total export to
Midvein Structures Tissues | Structures Tissues Export Distal Tissue
Lenght
<20 0.2441 0.114 | | 3,126,317 2,761,286 11,590,578 5,887,036
<30 0.4069 0.1801 5,068,949 2,553,531 15,814,274 7,622,480
<50 1.0539 0.5546 11,251,902 4,753,648 36,369,420 16,005,551
>50 1.5547 1.4564 11,026 123,206 52,534,912 234,232
(X of total dry mass) j of total dpm recovered)
<20 52.9 47.1 | 53.1 46.9 56.5 28.7
<30 66.4 33.6 | 66.5 33.5 58.9 48.2
<50 67.3 32.7 | | 70.3 29.7 48.4 21.3
>50 51.8 48.4 47.4 52.6 31.4 0.15


125
Mature Distal Leaves
Individual mature leaves distal to the inflorescence were labeled and the rate of translocation and the sites of label distribution were observed. Within 48 hrs, 69% of the initially fixed carbon was translocated from the labeled source leaf. The fruit and reproductive structures accumulated 29% of the initially fixed carbon. No radiolabel was recovered from the leaves proximal to the inflorescence although the wood of the branch proximal to the inflorescence contained radioactivity.
Discussion
Other work from this laboratory has examined the source/sink relations of the flowers, fruit and leaves proximal to the inflorescence (Chap. 6). From these studies,it was learned that the leaves proximal to the inflorescence were the major sources of recent photoassimilate for the developing vegetative and reproductive tissues. However, the amount of assimilate moving acropetally to these sinks modulated with the developmental stage of the leaves distal to the inflorescence. The decrease in acropetal assimilate movement suggested that the sink to source conversion of the leaves occurred when the average midvein length of the leaves in the flush was between 30 and 40 mm. An in-depth study of the sink to source conversion of the distal leaves was not done in these earlier works.
Import of assimilate by developing avocado leaves increased dramatically between 10 and 20% final midvein length. The rate of import then declined until it was almost negligible at 37% final


129
between 30 and 40 mm (Fig. 7-5), showed an even more pronounced skewing of label assimilation with two adjacent orthostichies containing 98% of the label recovered from the leaves.
The transition from an equal to phyllotactically determined distribution of radiolabel reflected the maturation of the vascular system. The continuous, acropetal, helical differentiation of the procambium and phloem is well established (59,149,150) as is the helical movement of assimilates through the stem via the phloem (109). Leaves forming at the apex receive assimilates from the undifferentiated procambium (150). The differentiation of the vascular system with increasing plastochron age was evident in the final two developmental classifications where the leaves in direct alignment and at the next orthostichic position contained the majority of the radiolabel. Leaves with lamina lengths 30 mm or less contained the most label. Leaves larger than this were in transition from being sinks to sources or had become exporters of assimilate.
The path of assimilates exported from the mature leaves of avocado (Figs. 7-6, 7-7), as in other plants (85,94,113,117,121,126,176), was strongly guided by phyllotaxy. If avocado fruit were in phyllotactic alignment with the source leaf then these fruit contained 94.5% of the label recovered in the fruit (Fig. 7-6). If there were no fruit with the same orthostichy as the source leaf then the majority of the label (62.6%) recovered in the fruit was found in fruit located at the orthostichy 72 from the source leaf position (Fig. 7-7). Tangential movement of assimilates between orthostichies has also been shown to occur in beet (113), willow (151), bean (185), citrus (121) and tobacco (176).


% of TOTAL FRUIT RADIOLABEL
rn
a o
T)
m
33
O
o
m a o
m m a
o o >
rn in o o >
XI
9L


175
EC
CO
D J
Cu 4J
x:
Eh d)
u
CO CP
o e
D
CO
90
80
70
60
50
Cn 40
30
20
10
0
t> X N \ > V X
\v\\\\\
; xxxx x xx
K\\\\\ v
_
/
/
/
4
/ /
/ /
/
GIRDLED
NON-GIRDLED
BRANCH TYPE
FLUSH 3
FLUSH 2
FLUSH 1


171
\V\
/
/
GIRDLED
NON-GIRDLED
BRANCH TYPE
FLUSH 3
FLUSH 2
FLUSH 1


153
fluctuations in the starch levels of the roots, trunk and branches of the tree (5,11,16,23,42). Work by Scholefield et al. (162) and Cameron and Borst (26) has shown that in avocado carbohydrate levels, specifically the starch levels in the branches, trunk and roots, decrease sharply during flowering and fruit development. A decrease in total carbohydrates has been associated with alternate bearing in "off season" avocado trees. Fruit removal (161) increased leaf dry mass and increased the number of starch grains in the most recent flush of leaves.
The presence of multiple leaf flushes concurrently on a branch during fruit development provides an interesting model for carbohydrate partitioning with respect to leaf age, leaf position and fruit load. Even though fruit load is not specifically addressed in this paper the contribution of the non-fruit-bearing branches to the tree's carbon economy may be significant and warrants investigation. Eight branches, each bearing three mature flushes of leaves were chosen for these experiments. Flush 3, the most recently matured flush, consisted of 11 + 2.9 leaves, averaging 86 + 4 mm in length (lamina), and 304 + 45 cm2 in area. The values for Flush 2, the early spring flush and Flush 1, the flush formed during the previous season were very similar to those of Flush 3. The total leaf area of flushes on non-fruit-bearing branches was generally greater than on fruit-bearing branches (Chap. 9). Defruited and non-fruit-bearing plants have greater total leaf areas than the corresponding fruit-bearing plants (5,160).
The most recent flush (Flush 3) assimilated the greatest proportion of the available radiolabel (Fig. 8-1). Flush 3 incorporated 50% of the total label fixed by the branch. Flush 2


CHAPTER 5
SINK/SOURCE RELATIONS IN THE AVOCADO INFLORESCENCE
Introduction
The photosynthetic contribution of reproductive organs to their own growth varies among species. The reproductive structures of barley (183), wheat (60,120,183), rice (58), oats (110) and other grasses (147) contribute between 50 and 75% of the needed photosynthate to the developing grain. In contrast, the reproductive organs of bean (40), soybean (153), cotton (57) and peas (66) are supplied assimilate primarily by the leaves. Although C02 fixation occurs in the reproductive organs of the latter group respiratory losses offset any net photosynthetic gain (153).
Significant differences in the fixation of carbon by the flowers and fruit of perennial tree crops have also been noted. Hansen (88) noted that substantial amounts of 14C02 were fixed by apple flowers prior to flower opening. Fixation decreased with petal drop and remained depressed until the fruit enlarged. Moreover, developing lemon (21) and orange fruit (13), and Valencia orange flowers (196) have been shown to contribute quantitatively little photosynthate to the developing fruit. Young avocado and mango fruit also fixed C02 (186), but the large respiratory loss in young fruit offset the net gain (4,186). Similar investigations on avocado flowers have not, to our knowledge, been done.
79


24
Table 2-2. Pulse-Chase of Ethanol-Extracted Radiolabeled Sugars in Avocado
Tissue Time Extracted Sugars (mg) dpm/mg X Leaf Sugars X dpm
Lamina 0 20.64 4.25 69.60 41.00
Vein 9.45 0.00 31.40 0.00
Lamina 1 19.37 10.65 69.00 79.31
Vein 8.68 2.78 30.90 20.69
Lamina 4 28.50 7.18 68.20 78.04
Vein 13.26 2.02 31.80 21.96
Leaves were labeled, harvested and prepared as discribed in the Materials and Methods. Sugars were quantited by the phenol-sulfuric acid method of Dubois et al.(53). Sugars specific activity is recorded as dpm/mg sugar. The distribution of sugars between the laminate and vein portions of the blade is indicated by the percent leaf sugars. The percent dpm refers to the amount of radiolabel associated with the sucrose spot on TLC plates for each leaf blade (laminate vein).
*
^Associated with sugars
Associated with the sucrose spot


161
GIRDLED
NON-GIRDLED
BRANCH TYPE
FLUSH 3
FLUSH 2
FLUSH 1


143


Figure 3-4. Growth of Fruit on Multiple-fruited Inflorescence. Fruit 86-89 were present on the same inflorescence and none of these fruit abscised during the course of the experiment. Fruit 76 and 77 were also present on the same inflorscence, however fruit 76 abscised after 50 days of study.


Figure 8-2. Specific Activity of Avocado Leaf Tissue from Three Successive Leaf Flushes. Leaf samples were taken from each leaf of each flush on both girdled and non-girdled branches immediately after the 1 hr labeling period. These samples were dried, weighed, digested and the radiolabel quantitated as described in the Materials and Methods. Leaf disk (0.618 cm ) dry mass and label incorporation after 1 hr. exposure to CO of all the leaves in the flush were used to calculate the mean incorporation of label by each of three consecutive, girdled or non-girdled flushes. Branches were approximately 70-80 cm in total length and 1 to 1.5 cm in diameter. Branches were girdled proximal to the oldest flush, adjacent to the main scaffolding branch. Three branches were chosen, 1 each from the lower, middle and upper canopy positions. The experiment was repeated four times.
i


28
100
90
80
70
CO
D O CJ
< Eh
O
60
50
40
30
20
10
0
0 12 3 4
5 6 7
8 9 10 11 12 13 14
MOBILITY (cm)
4 HOURS
24 HOURS


65
strands on the thicker side enter and anastomose through the entire seed coat. The vascular strands on the thinner side anastomose through the pericarp. The asymmetry in the vascular system may explain the gradients in oil content (170) and dry weight (170) observed in the fruit; however, we found the specific activity (Table 4-1) of radiolabeled assimilates in each tissue was consistently similar throughout the tissue. This discrepancy between oil and dry mass content and radiolabel distribution results from the dependency of specific activity measurements on dry mass. As can be seen in Table 4-3 the distribution of "C assimilates paralleled dry mass content.
The import of assimilates as measured by the relative specific activity (Table 4-2) suggested that assimilation of recent photoassimilates by specific tissues depended upon fruit size. Generally, the endosperm and the mesocarp exhibited the greatest relative specific activity. The greater activity in the endosperm may correlate to this tissue's nutritive role during embryo development. The relative specific activity of the seed coat was appreciably greater than other fruit tissues only in fruits with dry masses' less than 1 gram. A large proportion of fruitlets in this size range were shown to abscise in a previous study (Chap. 3). The possible significance of this increased sink strength will be discussed later.
Inducing synchronous abscission of avocado fruitlets has been accomplished using excised fruit-bearing branches in the laboratory (47). However, in order to preserve the integrity of the source/sink pathway, an alternate method was employed in the field. Puncturing the seed coat of the fruitlet as described by Blumenfeld and Gazit (18) was effective in inducing abscission. The import of radiolabeled


63
not the most actively importing tissue. Generally, the endosperm was the most active importer of assimilates, but since it represented such a small percentage of the dry mass its sink activity was overshadowed by the more massive tissues.
The total amount of radiolabel in a tissue appeared to be a function of mass (Table 4-3). For example, the mesocarp comprised between 87 and 89% of the fruit dry mass for fruit with a total dry mass less than 1.5 grams. The mesocarp also imported 84 to 88% of the total radiolabel recovered from the fruit. Similarly, the seed coat comprised 7.4 to 9% of the fruit dry mass and imported 8.6 to 9.7% of the total radiolabel recovered from the fruit. Although all of the fruit tissues are importing radiolabel, the endosperm and cotyledons have a high 14C/dry mass ratio indicating these tissues are growing more rapidly than the mesocarp and seedcoat.
Fruit were pierced to initate seed coat senescence and abscission (19). The distribution of radiolabel in the tissues of pierced, single fruits is shown in Fig. 4-1. In the non-pierced control the label distribution was as described in Table 4-3. The mesocarp contained approximately 88% of the total label recovered from the fruit. In fruit pierced 48 h before source leaf labeling only about 52% of the total label recovered from the fruit was found in the mesocarp. The percentage of label found in the seed coat increased and was greater than 5-times the percentage found in control fruit. In fruit punctured 24 hours before source leaf labeling, the percentages of label recovered in the endosperm, cotyledons and seedcoat were significantly greater and the percentage of label recovered in the mesocarp was significantly less


190
second flushes of defruited and girdled fruit-bearing branches. So although the fruit-bearing flushes fixed similar amounts of 14C02, the second fruit-bearing flush directed more recently fixed carbon into gluconeogenesis when sink strength decreased. The first and third flushes had similar values which were about 60% that of flush 2. The starch extracted from leaves on non-girdled fruit-bearing flushes had the highest specific activity. Unlike the pattern observed in the girdled and defruited branches, the specific activity of the starch extracted from leaves on non-girdled fruit-bearing flushes increased with the age of the flush.
This paper and its companion paper (Chap. 8) have shown that fruit-bearing and non-fruit-bearing branches were very different in their abilities to assimilate and translocate recent photoasssimilates. Leaves of similar age on fruit-bearing branches were more photosynthetically competent than their non-fruit-bearing counterparts. All of the flushes on fruit-bearing branches fixed similar proportions of 14C02 and translocated their recently fixed photoassimilates more rapidly than non-fruit-bearing branches. Unlike the pattern observed in fruit-bearing branches, the fixation and translocation of radiolabeled assimilates by non-fruit-bearing branches was age dependent, with the youngest flushes being the most competent (Chap. 8). Non-fruit-bearing and fruit-bearing branches also differed in dry mass, soluble sugar content, and starch content. The pool sizes of the neutral fraction (sugars) was much larger for fruit-bearing flushes, although relative pool sizes (neutral, basic, acid 1, acid 2) were the same for all flushes.


71
Table 4-1. Symmetrical Partitioning of Assimilates within Tissues of Avocado Fruit.
1 4
Source leaves on branches were fed CO for a one hour period. The fruit were harvested either 24 or 48 hours later and dissected, dried and processed for scintillation counting. Fruit represent a variety of fruit sizes (.2 1.2 gms.).
Fruit # 1 2 3 4 5 6
(dpm mg" dry weight)
Dry Mass(gm) 0.21 0.36 0.67 0.93 0.72 1.20
TISSUE
Endosperm 2889 2889 8415 302 104 987
Endosperm 2996 2978 8422 278 100 914
Cotyledon 2300 3200 3776 894 42 916 Cotyledon 2100 3263 3985 919 44 902
Seed coat 1875 2313 2914 625 57 1631
Seed coat 1914 2266 2850 650 55 1560
Mesocarp 3818 1954 4923 1522 98 1909
Mesocarp 3209 2048 5550 1491 100 1694
Mesocarp 1464 1739
Mesocarp 1849
The data from six different fruit are presented. Replicate samples from longitudinal, radial or cross-sectional sections were tested from each tissue.


16
secondary translocated sugars in representatives of all three races of avocado as part of a larger effort to understand the partitioning of recent photoassimilates in avocado.
Materials and Methods
Leaf Extracts
The trees used in this study were three-year-old, containerized trees of Persea americana Mill. cv. 'Peterson' grafted onto 'Walden' root stock as well as 10-year-old 'Peterson' and 'Booth 8' orchard trees growing on 'Walden' root stock. Individual leaves on the most recent flush were sealed in 2-liter plastic containers and exposed to 2.25 uCi
14 14
of C02 generated from NaH C03 (Amersham, specific activity of 54 mCi/mmol) using equimolar amounts of 5% lactic acid. The leaves were labeled for 15 minutes between 0900 and 1100 hr on cloudless days. The experiment was replicated for each of two leaves on six occasions during May.
Leaves were harvested immediately following the labeling period (0 hr), at 1 and 4 hours. They were excised, quickly separated into laminate and major vein portions, cut into small pieces and dropped into separate flasks containing approximately 100 ml boiling 85% ET0H according to Housely et al. (102). The leaves/veins were boiled for 15 minutes and filtered. The residue was further extracted with similar volumes of 50% ET0H and water. The sample extracts were pooled, rotoevaporated to dryness and reconstituted with 20 ml H20. Sugar content of the extracts was quantitated by the phenol-sulfuric acid method of Dubois et al. (53) to calculate the relative sucrose content and specific activity of the radiolabeled products. Radioactivity was


69
enhanced values of radiolabel in the receptacles. Fruit pierced at 48 hours post labeling shoved no increase in radiolabel in their receptacles over control values.
Further insight into the physiological activity of the tissues may be gained by examining the tissue specific activity in abscised fruits (data not shown). The relative sink, strength of tissues in pierced single fruitlets was similar to non-pierced controls. Similarly, doublet fruit pierced before labeling and at 0 hour had similar relative specific activities. However, in doublet fruit pierced 24 or 48 hours after labeling, the seed coat and not the endosperm/mesocarp had the greatest sink strength. The increased sink strength of the seed coat may reflect the increased activity in the seed coat which preceded abscission. This increased sink strength decreased with time after puncturing.
In conclusion, the partitioning of assimilates and the growth of abscising fruit differed from, but was influenced by persisting fruit. Punctured fruit on multiple-fruited inflorescences were retained longer than their single-fruited counterparts. Although similar trends were observed in partitioning patterns in both single- and multiple-fruited inflorescences, the differences in the latter occurred over an extended period of time. It is unknown whether or not these patterns result strictly from vascular restrictions or from other inherent metabolic processes associated with abscission. It is clear that persisting fruit do in some way influence the abscission behavior of the abscising fruit.
Puncturing the seed coat caused an immediate increase in its sink strength. This increase may reflect metabolic activity associated with the initiation of the abscission processes or an injury response.


9
to the increased availability of cofactors and not due to an increase in the number of mitochondria or an increase in ATPase activity. The level of the enzyme pectinmethylesterase remains constant through the preclimacteric but the enzyme's activity decreases with initiation of the respiratory climacteric (9). Conversely, the activity of the enzyme which depolymerises pectin, polygalacturinase, increases during the climacteric and post-climacteric periods (8). Cellulase activity in avocado also increases during the preclimacteric stage (8). Exo-and endo-cellulase activities are stimulated by ethylene (105,157).
Climacteric and post-climacteric avocado are characterized by browning tissues. The browning is caused by oxidases such as polyphenol oxidase (PPO), peroxidase (P0), and catalase (114,115,116 respectively). Polyphenol oxidase is actually a family of isozymes. In avocado there are at least six forms of the enzyme ranging in molecular weight from 14,000 to 400,000 daltons. These isozymes are good developmental markers since they appear or disappear at specific times during the climacteric cycle. Peroxidases also have multiple forms which appear at specific times in the climacteric cycle.
Abscission and Partitioning The process of abscission is a widespread phenomenon in crop plants. There have been several hypotheses put forth to explain the abscission of plant organs. These hypotheses involve correlative plant behavior, inhibition and senescence, organ competition and plant homeostasis (3). Today, the relevant hypotheses concerning fruit abscission are inhibition and senescence, and organ competition (3).



PAGE 1

THE EFFICIENCY OF THE SINGLE-FAMILY HOUSING MARKET: AN EMPIRICAL STUDY By DEAN H. GATZLAFF 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 1990

PAGE 2

ACKNOWLEDGEMENTS This project has benefitted from the help and support of many individuals and organizations. I owe a great deal of gratitude to both the Real Estate Research Center at the University of Florida and the Urban Land Institute in Washington D.C. for providing research facilities and financial assistance. In addition, I want to acknowledge those who provided invaluable assistance and suggestions. Early discussions with Jerry Milliman and Marc Smith helped develop much of the motivation for selecting the research topic, and, along with Wayne Archer, Roy Crum, G.S. Maddala, and David Denslow, they provided insight into critical research issues through their seminars or comments. I am especially grateful to J. Sa-Aadu and Michael Ryngaert for their critical evaluation and direction throughout the entire course of this project and to Halbert Smith and David Ling for their thoughtful guidance and encouragement as the project progressed. Finally, I am, and will always be, deeply indebted to my parents for their unending support and to Kathy, my wife, who has always given immeasurable understanding, care, and love to our daughters and to me. ii

PAGE 3

TABLE OF CONTENTS ACKNOWLEDGEMENTS LIST OF TABLES LIST OF FIGURES. ABSTRACT CHAPTERS ii V . viii ix 1 2 3 4 5 INTRODUCTION 1 Purpose and Contribution of the Study. 1 Significance of Housing Market Efficiency. 3 THE EFFICIENT MARKET HYPOTHESIS .. Types and Forms of Market Efficiency ... Conventional Tests of Weak-Form Efficiency Random-Walk Theory and Autocorrelated Returns . . . . . . Trading Strategies .... Joint Hypotheses . . . . . Rational and Adaptive Expectations Price Extrapolation ......... Arbitrageurs ......... Real Estate Market Efficiency ....... Historical View ...... The Single-Family Housing Market .... Speculative "Bubbles" ... LITERATURE REVIEW. 7 7 11 11 16 18 19 19 21 23 23 26 27 29 Principal Studies . . . . . 29 Linneman (1986) . . . . . . 30 Rayburn, Devaney, and Evans (1987) . 31 Case and Shiller ( 1989) . . . 32 Guntermann and Smith (1987) . . . 34 Additional Research. . . . . . 37 Summary of the Literature. . . . 41 GENERAL HYPOTHESES TESTABLE HYPOTHESES. iii 43 48

PAGE 4

6 7 8 9 10 APPENDICES A B TEST METHODOLOGY, BASE MODEL, AND DATA Overview of the Test Methodology Description of the Base Model Real Returns to Housing After-Tax Excess Returns to Housing Description of the Data ... Base Model EMH Test Results. Base Model Assumptions .......... MARGINAL TAX RATES 51 51 54 54 55 56 58 70 75 Implied Bondholder Marginal Tax Rates. 75 Home Buyer Marginal Tax Rates. . . . 78 EMH Test Results: Excess Returns Adjusted for Time-Varying Marginal Tax Rates 80 INFLATION EXPECTATIONS Expected and Unexpected Inflation .... Past Inflation Models and Tests of Rationality . . . . . . . Inflation Expectations Model ... Unexpected Inflation and Housing Returns EMH Test Results: Residual Returns Adjusted for Unexpected Inflation ... TIME-VARYING RISK PREMIA 87 87 90 97 101 102 109 Evidence of Time-Varying Volatility. 109 Autoregressive Conditional Heteroskedasticity ........... 111 Modeling Risk-Premia . . . ... 113 ARCH-M Test Results ........... 114 Trading Strategies . . . . .. 116 SUMMARY AND CONCLUSIONS WEIGHTED-REPEAT-SALES (WRS) PRICE INDICES SUPPLEMENTARY TABLES AND STATISTICS. REFERENCE LIST BIOGRAPHICAL SKETCH. 120 122 122 126 138 145 iv

PAGE 5

LIST OF TABLES 2-01 U.S. Capital Market Total Annual Returns. 25 3-01 Tabular Survey of the Real Estate EMH Empirical Literature . . . . 38 6-01 Summary of the Data Used to Estimate the Base Model . . . . 58 6-02 Summary Statistics of Capital Market Real Quarterly Returns .......... 60 6-03 Regression of Annual Real Returns on Annual Real Returns Lagged (Full Sample Period) . . 64 6-04 Replication of Case and Shiller (1989) Statistics Regression of After-Tax Annual Excess Returns on After-Tax Annual Excess Returns Lagged. . 65 6-05 Regression of Quarterly Real Returns on Quarterly Real Returns Lagged (Full Sample Period) . . 68 6-06 Regression of After-Tax Quarterly Excess Returns on After-Tax Quarterly Excess Returns Lagged. . 69 6-07 Autocorrelation Function of Real Quarterly Returns . . . . . . . . 6-08 Autocorrelation Function of After-Tax Excess Quarterly Returns .......... 71 72 7-01 First Order Autocorrelation of Margi'nal Tax Rates 77 7-02 Sources of Data to Determine Home Buyer Marginal Tax Rates . . . . . . . . 79 7-03 Correlation Matrices of Quarterly Home Owner Excess Returns, Constructed Home Owner Marginal Tax Rates and Implied Bondholder Marginal Tax Rates . . . . . . . . . . 84 7-04 Regression of After-Tax 4-Quarter Excess Returns on After-Tax 4-Quarter Excess Returns Lagged V

PAGE 6

Adjusted Nonconstant Home Owner Marginal Tax Rates . . . . . . . 7-05 Regression of After-Tax Quarterly Excess Returns on After-Tax Quarterly Excess Returns Lagged Adjusted for Nonconstant Home Owner Marginal Tax 85 Rates . . . . . . . . . 86 8-01 Regression of Annual Excess Returns on Annual Excess Returns Lagged (Prior and Subsequent to Monetary Policy Shift) . . . . 96 8-02 Comparative Statistics: Livingston Survey vs ARIMA-1 Quarter Forward Model of Unexpected Inflation 99 8-03 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation: Atlanta ........ 104 8-04 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): Atlanta .... 104 8-05 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation: Chicago 105 8-06 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): Chicago ... 105 8-07 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation: Dallas ....... 106 8-08 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): Dallas ... 106 8-09 Regression of After-Tax Quarterly Excess Returns Adjusted for Home owner Marginal Tax Rates on Unexpected Inflation: San Francisco 107 8-10 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): San Francisco .. 107 9-01 Correlation of City Expected Quarterly After-Tax Excess Returns and Expected Volatility .... 111 vi

PAGE 7

9-02 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation and Volatility 115 9-03 Average Quarterly Returns Using Alternative Trading Strategies ......... B-01 Regression of After-Tax 4-Quarter Excess Returns 118 on After-Tax 4-Quarter Excess Returns Lagged 127 B-02 Regression of After-Tax Quarterly Excess Returns on After-Tax Quarterly Excess Returns Lagged Using Nonconstant Tenure Choice Marginal Tax Rates ... 128 B-03 Regression of After-Tax 4-Quarter Excess Returns on After-Tax 4-Quarter Excess Returns Lagged Using Nonconstant Tenure Choice Tax Rates ... 129 B-04 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Alternative Unexpected Inflation Models: Atlanta. 130 B-05 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged: Atlanta (Using Alternative Models) ..... 131 B-06 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Alternative Unexpected Inflation Models: Chicago. 132 B-07 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged: Chicago (Using Alterative Models) ..... 133 B-08 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Alternative Unexpected Inflation Models: Dallas 134 B-09 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged: Dallas (Using Alternative Models) ...... 135 B-10 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Alternative Unexpected Inflation Models: San Francisco . . . . . . 136 B-11 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged: San Francisco (Using Alternative Models) .... 137 vii

PAGE 8

LIST OF FIGURES 1-01 Composition of the u. s. Capital Market (December 31, 1986) . . . . . . . 4 6-01 Real and Excess Annual Returns to Housing: Atlanta, Chicago, Dallas, and San Francisco . 61 7-01 Implied Bondholder Marginal Tax Rates (1970:11986:2) . . . . . . 76 7-02 Homeowner Marginal Tax Rates and Quarterly Excess Returns: Atlanta . . . . . . 82 7-03 Homeowner Marginal Tax Rates and Quarterly Excess Returns: Chicago . . . . . . 82 7-04 Homeowner Marginal Tax Rates and Quarterly Excess Returns: Dallas . . . . . . . 83 7-05 Homeowner Marginal Tax Rates and Quarterly Excess Returns: San Francisco . . . . . . 83 8-01 Unexpected Annual Inflation (Livingston, ASA-NBER, ARIMA) . . . . . . . . . 93 8-02 Real 1-Year Treasury Bill Yields and Unexpected Annual Inflation . . . . . . . 94 8-03 Unexpected Annual Inflation, ARIMA Model: Quarterly, 4-Quarter, Annual . . . . . 100 viii

PAGE 9

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 THE EFFICIENCY OF THE SINGLE-FAMILY HOUSING MARKET: AN EMPIRICAL STUDY By Dean H. Gatzlaff December 1990 Chairperson: Dr. Halbert c. Smith Major Department: Finance, Insurance and Real Estate This study examines the efficiency of the single-family housing market. It is hypothesized that ex post returns to housing are weak-form efficient; thus, investors cannot construct ex ante investment rules based on historical information that consistently yield above-normal risk adjusted returns. Previous work is extended by examining the effects of time-varying marginal tax rates, unexpected inflation, and return volatility in the analysis of after tax excess returns. After-tax excess housing returns are found to be negatively correlated with marginal tax rates and positively correlated with unanticipated inflation. Furthermore, the data suggest that inflationary shocks were autocorrelated during the late 1970s and early 1980s, and not anticipated by either the bond or housing markets. When returns are ix

PAGE 10

adjusted for both time-varying marginal tax rates and unanticipated inflation, after-tax excess returns are no longer significantly autocorrelated in three of the four markets studied. While persistent return "momentum" exists, when all transaction costs are considered, no investment rule can be constructed to consistently exploit any remaining serial dependence present in the markets. Hence, the hypothesis of a weak-form efficient market cannot be rejected. Finally, limited preliminary evidence is found to suggest that expected after-tax excess returns to housing are weakly correlated (positively) with expected volatility. This provides provisional support that return volatility (risk) is priced by housing market participants. However, modeling expected volatility using an ARCH-M model reveals no evidence of time-varying risk premia in housing returns. X

PAGE 11

CHAPTER 1 INTRODUCTION Purpose and Contribution of the study This study evaluates the efficiency of the single family housing market using conventional methodologies reported in the finance literature. Although there exists a plethora of literature that models and tests the efficient market hypothesis (EMH), it is largely confined to corporate security market analysis. Research in the area of real estate market efficiency is quite limited. Few, if any, definitive answers to the efficiency question have emerged. Previous studies by Rayburn, Devaney ~nd Evans (1987) and Guntermann and Smith (1987) indicate that real returns to housing are not autocorrelated, hence, the hypothesis of a weak-form efficient market could not be rejected on that basis. However, Case and Shiller (1989) find that after-tax excess returns to housing (annual returns to housing in excess of the after-tax one-year treasury bill yield) are autocorrelated. 1 They indicate that home buyers, indifferent to time of purchase, can construct an investment 1 case and Shiller find only limited evidence of autocorrelated real price changes in housing and indicate that single-family housing prices are not forecastable. 1

PAGE 12

2 rule that yields above-normal returns. They suggest that real treasury bill yield information is not incorporated in the decision to invest in housing and on that basis reject the notion of a weak-form efficient market. It is well known that any test of market efficiency is a joint test of some market equilibrium model. Fama (1976) demonstrates that even a serially correlated return series is consistent with market efficiency if equilibrium returns are serially correlated as well. 2 Adopting the weighted repeat-sales housing price indices constructed by Case and Shiller (1987), this study examines whether their results are sensitive to alternative specifications of what a "fair" or equilibrium expected excess return is for single-family housing. This study explores the possibility that findings of serial correlation in the Case and Shiller study may be the result of serial 2 serial correlation of stock returns is generally taken as evidence of market inefficiency because it seems reasonable to assert that the variance of price changes due to new information arrival is large relative to the variance in expected equilibrium returns over time. Hence, if new information arrives randomly, autocorrelation in returns is expected to be low. Serial correlation in treasury bill returns, however, is not judged to be a violation of weak form efficiency because the treasury bill returns are expected to follow the level of inflation, which is a highly autocorrelated series. A discussion of the conditions necessary for market efficiency appears in Chapter 2.

PAGE 13

correlation in home buyer marginal tax rates, ex post inflationary shocks, and return volatility. 3 Significance of Housing Market Efficiency The single-family housing market is a prime component of the U.S. economy. A substantial portion of aggregate U.S. wealth, housing is the most widely held of all investment assets. As shown in Figure 1-01, the estimated value of single-family housing approximates $5 trillion. In addition, nearly one-third of all income property, valued at $2.5 trillion (debt and equity), consists of multi-family residential property. At $3 trillion, the estimated value of single-family home equity is approximately equal to that of corporate equity. The remaining $2 trillion value of single-family housing mortgages approaches the size of the U.S. Treasury debt. 4 Furthermore, the largest portion of many individuals' investment portfolios is their leveraged equity in their home. Efficient capital allocation within this market segment could strongly affect economic conditions. 3 3 case and Shiller assumed these factors do not change (are constant) or randomly change during the sample period. These assumptions are detailed later in Chapter 6. 4 statistics quoted are 1987 estimates of the Board of Governors of the Federal Reserve and Salomon Brothers, Inc.

PAGE 14

Commercial Single-Family Real Estate Real Estate Equity Debt Equity Debt Corporate U.S. Tax Equity Debt Treasury Exempt Debt Bonds FIGURE 1-0lCompostion of the U.S. Capital Market (December 31, 1986) Source: Hartzell, Lepcio, Fernald, and Jordan (1987) In addition to issues of general economic welfare, the efficiency of the market is important to the development of real estate economic theory. It is fundamental to fields such as valuation, investment, and urban economics. Valuation theory assumes that all buyers and sellers are well informed. Grossman and Stiglitz (1976), in reference to capital markets, have suggested that in an efficient market, it is the activity of informed individuals that serves to set prices for the entire market. The market prices aggregate information so that all participants (both informed and uninformed) become informed by observing prices.

PAGE 15

5 If single-family homes within a market are viewed as substitutes, then "the result of this process, which is the competitive process, is an arraying of all the sites {properties} in a hierarchy of land {property} values" (Ratcliff 1949, p.366). 5 Individual property value movements are not purely idiosyncratic. Under this condition, it is plausible that the activity of informed individuals in the real estate market highly influences, if not sets, the most probable selling price of homes in each market. While it is recognized that real estate market participants are not price-takers, findings of market efficiency strengthen the notion that informed individuals substantially influence market prices. This strengthens the concept of using the direct sales comparison approach in real estate valuation and may lead to alternative valuation methods. Institutional investment in single-family housing occurs almost solely as whole mortgages or mortgage-backed securities (MBS). Savings and loans, commercial banks, government agencies, pension funds, and insurance companies, along with others who manage large portfolios of assets, collateralized by housing, have an interest in the stability and predictability of the housing market. Understanding the degree of housing market efficiency lends valuable support 5 Parenthetical insertions are by this author.

PAGE 16

6 to this effort. Support of the EMH indicates that unanticipated returns to housing are random and normally distributed, an assumption central to much investment theory. Finally, the concept of real estate market efficiency is central to evaluating the economic effects of local government policies such as zoning, growth controls, and property taxation. For example, the assertion that real estate markets are inefficient supports proponents of suburban growth restrictions. Real estate inefficiencies, proponents argue, result in abnormally high risk-adjusted returns to developers of suburbs. This overstimulates investment in suburban real estate, resulting in excessive suburbanization--a misallocation of resources. The existence of efficient real estate markets, especially the single-family housing market, gives provisional strength to the argument that historical suburbanization trends may be allocationally efficient. It suggests that development patterns may arise primarily from individual locational preferences responding to demographic and technical change. However, it is important to note that informationally efficient markets do not necessarily imply allocational efficiency.

PAGE 17

CHAPTER 2 THE EFFICIENT MARKET HYPOTHESIS Types and Forms of Market Efficiency The concept of efficiency is central to understanding both the positive and normative economics of a market. Economists commonly describe efficiency as the allocation of resources such that no reallocation can occur that increases the utility of some without decreasing the utility of others. A new allocation of goods in an economy is said to be "Pareto-preferred" to another if all economic agents are at least as well off or better off than under the prior allocation. 1 Thus, efficiency is a relative, not absolute, concept. The financial economics literature distinguishes between three distinct, but interrelated, types of efficiency in capital markets: informational, allocational, and operational. 2 An "informationally efficient" market, as defined by Fama (1970), implies that all market prices fully and instantaneously reflect all relevant information. 1 This concept of efficiency is attributed to Vilfredo Pareto, a nineteenth-century Italian economist and sociologist. 2 This study focuses on informational efficiency. Unless otherwise noted, the use of the term efficiency refers to informational efficiency. 7

PAGE 18

8 In such a market, prices serve as an accurate signal for the allocation of scarce capital. 3 Funds flow to the most productive investments, those yielding the highest marginal risk-adjusted return, maximizing net social benefits. Funds continue to flow to productive investments until the marginal risk-adjusted return on the least profitable investment equals the opportunity cost of borrowing. Subsequently, when the risk-adjusted marginal rate of return is equated for all borrowers and lenders, the market is said to be "allocationally efficient.'' Fama states that in general terms, the ideal is a market in which prices provide accurate signals for resource allocation: that is, a market in which firms can make productive investment decisions, and investors can choose among the security prices that represent ownership of a firm's activities under the assumption that security prices at any time "fully reflect" all available information. A market in which prices always "fully reflect" available information is called "{informationally} efficient" (Fama (1970), p.383). Consistent with the traditional concept of Pareto efficiency, all economic agents (borrowers and lenders) are better off in an efficient capital market. 3 It is important to differentiate between the allocational efficiency defined in the finance literature and that (Pareto-efficiency) defined in the mainline economics literature. Whereas the finance literature describes allocational efficiency on an after-tax (corporate) basis, Pareto-efficiency of investment requires that risk-adjusted marginal returns be equated before-tax. It maintains that the substitution effect resulting from taxes creates allocational inefficiencies. Thus, full Pareto-efficiency requires an efficient tax structure, an optimal level of public goods and services, efficient private goods and factors, no externalities, and labor resources that are efficiently distributed across regions.

PAGE 19

Finally, "operational efficiency" considers the degree of transaction costs associated with the flow of funds. Perfect operational efficiency would require a market with no transaction costs. Informational efficiency is a necessary condition for both allocational and operational efficiency. In the single-family housing market, informational efficiency implies that the distribution of market prices accurately reflects the expected risk-adjusted marginal return associated with the bundle of housing services at each location. This suggests that the market prices both private and public goods and services embedded in the purchase of real estate assets. As in the financial markets, this distribution of prices is useful in directing investment and consumption decisions to maximize welfare. If the housing market is informationally efficient, firms and households will choose to invest in and consume an optimal amount of housing services, given the current state of nature. 4 9 Fama defines three "forms" of informational efficiency: strong-form, semistrong-form, and weak-form. Each form is based on a different notion of exactly what type of information is relevant in the phrase "all market prices 4 This condition assumes that housing capital is fluid and mobile, that negative externalities and congestible housing services are priced by the market, and that future conditions are accurately anticipated In addition, it assumes investors have identical preference sets.

PAGE 20

10 fully and instantaneously reflect all available relevant information." Strong-form efficiency states that investors cannot consistently earn above-normal risk-adjusted returns using any or all information, whether private or public. This, of course, is highly restrictive, suggesting that even insider information is reflected in current prices. Grossman and Stiglitz (1976) have shown that such efficiency is impossible since costless information is both a sufficient and necessary condition in order for prices to fully reflect all available information. When information is costly, as in both financial and real estate markets, prices cannot reflect all information since those who incur costs gathering it would not be fully compensated. However, weaker forms of efficiency are not precluded by the existence of costly information. Investors acquiring costly information in such environments could be fairly compensated with returns commensurate with the cost of the information acquired. 5 The information subset relevant for semistrong-form efficiency consists of all available current public 5 The cost of acquiring new information should equal the added return benefits such that the net present value, including entrepreneurial profit, will equal zero. Cornell and Roll (1981) argue that the average investor who uses costly security analysis data will outperform investors who use less information; but, their net returns will be identical. Some empirical evidence has been found to support this position.

PAGE 21

11 information. This would include newly released information such as government economic statistics, investment advisory data, or information in newspapers or news services. For weak-form efficiency the information set is further partitioned to include only publically available historical information, such as past asset prices or returns. Hence, by definition, weak-form efficient markets indicate that investors cannot construct investment strategies based on past returns alone that consistently yield above-normal risk-adjusted returns. Weak-form efficiency is a necessary condition for both strong and semistrong-form efficiency. Conventional Tests of Weak-Form Efficiency Random-Walk Theory and Autocorrelated Returns Conventional tests of efficiency reported in the finance literature are generally based on the assumption that market equilibrium conditions can be expressed in terms of expected returns. The earliest references to this concept have been attributed to Bachelier (1900) and are now referred to as expected return theory or the "fair-game" model. Bachelier hypothesized that speculation should be a "fair-game"--in the long-run (aggregate) excess profits to speculation should be zero. Formally, this is defined where Lt= rj,t E(rj,tltt_,), rj,t = (Pj,t-Pj,t-1) / pj,t-1' (2-01) (2-02)

PAGE 22

12 and Pj,t is the actual price of asset j at the end of period t, r j, t is the one-period return of asset j int; E(rj,tltt_ 1 ) is the expected one-period return of asset j int given the available relevant information set, t, at end of period t-1; Lt is the residual (abnormal) return int; and Lt and rj,t are randomly distributed variables at t. Taking (mathematical) expectations of (2-01) yields, ( 2-03) The fair-game model is based on the behavior of average residuals, not the entire distribution 6 In other words, on average, across a large number of observations, the expected return is equal to the average actual return. Therefore, by definition, the expected residual is equal to zero -a fair-game, (2-04) Modifications of the fair game model include the martingale, the submartingale, and the random-walk models. The martingale and the submartingale represent only minor modifications of the fair-game model. The random-walk model, 6 Fama (1970) indicates that the model properties are implied by the assumptions that (1) the conditions of market equilibrium can be stated in terms of expected returns, and (2) the relevant information set, tt, is fully utilized by the market in forming equilibrium expected returns.

PAGE 23

13 however, substantially strengthens the conditions necessary for efficiency. 7 Samuelson (1965) established that randomly fluctuating returns represent a sufficient condition for weak-form efficiency. 8 Others reasoned that, if price changes can be assumed to be attributable to new relevant information priced by the market, and there exists no reason to believe that this information enters the market other than randomly, then it is reasonable to assume that in an efficient market one-period returns are statistically independent of one another (such as Alexander, 1961). This indicates that not only should the expected residuals be zero, but also residual returns should follow a random-walk. 9 Although somewhat imprecise, this describes an enhanced return generating process from that of the fair-game model. Random-walk theory, with regard to returns, explicitly requires that a series of successive one-period returns is 7 The martingale is a fair game model where the expected price next period is the same as the price this period, E (Pj t+ 1 I tt) =Pjt" The submartingale simply says that the expected price next period is greater than the price this period, E (Pj t+, 1 t ) >P t. Tests of these models typically compare the 'models' Jabnormal returns with those of the buy and-hold strategy. 8 In analyzing commodity futures contracts, Samuelson indicates that the expected price change of such contracts from period to period will be zero and is deemed a "fair game." 9 Prices may be hypothesized to follow a random-walk if price changes are independent and identically distributed. If price changes are expected to be non-zero prices are said to follow a random-walk with "drift".

PAGE 24

both independent and identically distributed. Unlike the fair-game model, it is not based solely on average returns or average residuals. The probability density function of returns, f( ), must be the same for all t. Hence, the random-walk model can be described as 14 (2-05) where rj,t is the one-period return on asset j in time t. This says, that each one-period return can be characterized as an independent variable, where the conditional probability distribution (conditioned on the information set) is identical to the marginal probability distribution. Thus, the information set is independent of the sequence of returns. 10 Indicating, (2-06) since all past relevant information is priced by the market, consistent with weak-form efficiency. In addition, all of the moments of the return distribution (i.e. the mean, variance, skewness, and kurtosis) must be the same with or without the relevant information set for all t. Under these conditions, if we assume that the number of underlying transactions is sufficiently large in generating each one-period return, then by the Central Limit Theorem 10 This condition is described later to be consistent with the rational expectations hypothesis.

PAGE 25

15 and {2-03) we can expect returns to be normally and independently distributed, {Lt> N { 0, a 2 ) Previous empirical work indicates that stock returns do not meet all the distributional properties of the random walk. 11 However, it appears unlikely that investors can exploit any of the distributional anomalies. In addition, it is highly improbable that risk is constant over time; the variance of the return distribution is likely to be time dependent. 12 Therefore, applying strict distributional random-walk model criteria for tests of efficiency is widely viewed as quite restrictive. Since returns are required to be independent, the random-walk model requires no autocorrelation of returns for all lags, including one-period lags. However, significant one-period autocorrelated returns are not inconsistent with the fair-game model. 13 This then, for purposes of testing efficiency, is viewed as the primary statistical difference between the random-walk model and the fair-game model. 11 Both Fama {1970) and Copeland and Weston (1988) provide a good review of this evidence. 12 French, Schwert, and Stambaugh {1987) indicate that the volatility of stock returns is time-varying and autocorrelated. While random-walk theory assumes that the volatility of asset returns is constant, random changes in volatility should have no adverse affect on tests of efficiency. 13 copeland and Weston {1988) show that even though the residual, t' is a fair-game variable, the conditional expectation of returns fort can depend on the return observed for t-1. Thus, the autocorrelation of one-period returns need not be zero in a fair-game model.

PAGE 26

16 Together, the random-walk of independent returns and Equation (2-06), have important implications which form the basis of most efficient market tests. Conventional tests of weak-form efficiency take the form of: (1) tests of autocorrelation, (2) trading strategies, or (3) runs tests, to evaluate the stochastic characteristics of the return time series. All represent testable derivatives of the expected returns theory. Consistent with the random-walk literature, studies involving tests of autocorrelation, most commonly tests of first-order autocorrelation, are often applied. Findings of autocorrelated returns indicate that future returns may be predicted from historical returns. Such findings reject the random-walk model hypothesis. However, to rejection the efficient market hypothesis (EMH) additionally requires that trading strategies, which account for transaction costs, be identified that exploit any serial dependence. Trading Strategies Like most theoretical models, the fair-game and random walk models are limited in their ability to describe fully the necessary conditions of the efficient market hypothesis. The formal notation of this concept takes on a general notion that is not necessarily consistent with market efficiency. While the return generating process described by a random-walk implies independent and identical return distributions, variations of distributions are nearly

PAGE 27

17 impossible to exploit; hence, they do not necessarily imply a rejection of the EMH. In order for market efficiency hypotheses to be rejected, random-walk inconsistencies must be exploitable by market participants. The existence of predictable patterns (serial correlation) of residual returns potentially represents relevant information. If this information can be used by investors to construct trading strategies that yield excessive profits, then the market is considered inefficient. 14 Only if the strategies consistently outperform a buy-and-hold option is the EMH definitively rejected. If predictable return patterns exist but transaction costs are too high for the information to benefit investors they are rational not to use it. Consequently, under the latter scenario, the hypothesis of a weak-form efficient market cannot be rejected. 14 with respect to security returns, Alexander (1961) examines a number of trading rules (strategies) to determine whether they could outperform a buy-and-hold strategy. Of the strategies put forth they% filter rule was examined rigorously. It was used to exploit the presence of positively correlated returns. This rule suggests that if the security price rose y%, the investor should buy-and-hold the security until its price moved down at least y% from the next high, then the investor should sell and go short. The short position is held until the price again rises y% above the low, at which time the investor covers the short position and buys long. Alexander concludes that such filters cannot beat the buy-and-hold position supporting the EMH.

PAGE 28

18 Interestingly, the very use of a known pattern will serve to eliminate it. Weak-form efficiency does not rule out the fact that patterns in returns may exist over very short periods of time. For instance, if research by investors indicates that on Friday prices rise and on Monday prices fall, informed investors will sell at the end of Friday, forcing prices down, and buy at the end of Monday, raising prices. This, of course, will mitigate any long term "weekend effect". 15 Joint Hypotheses It is important to note that all tests of market efficiency are tests of joint hypotheses. While conducting primary tests of the EMH, constructed actual returns and expected equilibrium return models are typically applied. As with hedonic modeling, if a component either of the return or expected return series is misspecified or ignored, the primary test may yield spurious results. Thus, the equilibrium model is implicitly tested, as well as the explicit testable hypothesis. For example, assume a dividend stream is a significant component of a return series of a particular asset and the dividend series is autocorrelated. Neglecting to include it in the construction of a return series could produce 15 French (1980) found significant evidence of a weekend effect in security returns using data from 1953-1977. However, it is not exploitable when transactions costs are considered.

PAGE 29

19 findings that the market is inefficient (efficient) when, in fact, it is efficient (inefficient). Rational and Adaptive Expectations Price Extrapolation The rational expectations hypothesis (REH), posits that prices are formed on the basis of the expected future returns, including components of appreciation and dividends adjusted for taxes. As introduced by Muth (1961), the REH requires both expected prediction residuals (errors) to equal zero and that prediction residuals are uncorrelated with the information set that is available to the forecaster at the time of the forecast. If the residuals are correlated with the information set it implies that the forecaster has not used all available information. Maddala (1988) formalizes this as y\ = E(yt,~t 1)' (2-07) where Yt is the predicted value of Yt in period t. Maddala indicates that the left-hand side of (2-07) should be interpreted as the "subjective" expectation and the right hand side the "objective" expectation conditional on the data available when the expectation is formed. Both the descriptive and the formal definitions are consistent with random-walk theory (see Eq. 2-06); hence, the EMH implies

PAGE 30

20 rational rather than adaptive expectations on the part of investors. Adaptive expectations hypotheses indicate that economic agents revise their expectations upward or downward, based on the most recent errors. Formally, this may be modeled (2.08) with o < r < 1. This indicates that market participants adapt gradually to new information. 16 This is in contrast to the EMH where "all relevant information is fully and instantaneously reflected in the market". The popularity of technical analysis in security markets is evidence that many investors think the market is adaptive, that it overreacts (underreacts) to new information leading to predictable corrections (resistance targets). Similarly, the real estate market is often assumed to follow an adaptive pricing process. Models of expected housing returns found in the literature often extrapolate recent experience. In these models, high appreciation rates in period t raise the likelihood of high appreciation rates 16 other models of adaptive expectations such as distributive and polynomial lagged processes are well known and often applied.

PAGE 31

in period t+l, thus, implying returns are positively correlated. 17 Furthermore, at times reporters allege that some regional markets are highly overpriced, forecasting an impending long-term "bear" real estate market in these regions. 18 Here, market prices are believed to react slowly to excess supply and demand conditions. If true, investors may be able to take advantage of the predictable cyclical nature of these real estate markets. Arbitrageurs Arbitrageurs play an important role in maintaining market efficiency in both national and global security markets. The trading activity of both arbitrageurs and informed traders (those with better estimates of future states) serves to bring securities prices quickly into equilibrium. While it is impossible for investors to consistently earn abnormal risk-adjusted returns in an efficient market, short-term excessive returns are not 21 17 Since inflation, especially long-term inflation rates, are positively autocorrelated and returns to real estate are generally positively correlated with inflation, it is likely that such models largely capture the component of housing return due to general inflation. 18 In March of 1990, The Wall Street Journal indicated that both the real estate markets in the northeast and west were highly overvalued and long-run "bear" markets were widely predicted. In the short-run this information may be self-fulfilling; however, if this information is used by market participants it should serve to mitigate long-run forecasts.

PAGE 32

22 inconsistent with weak-form efficiency. In fact, the possibility of such returns is a necessity of market efficiency, since it serves as an incentive for price correcting arbitrage. Arbitrage trading serves to inform uninformed investors through the pricing mechanism. The absence of any formal arbitrage activity in real estate markets may justify arguing that the market is inefficient. Real estate markets tend to be local or regional rather than national, futures contracts in real estate equities do not exist, and short selling is not available. All this curtails or perhaps eliminates arbitrage activity. Nevertheless, informed corporate, institutional, and individual investors actively participate in both debt and equity investment in real estate. Investors, including realty groups, systematically analyze regional real estate markets for mispriced properties and hold portfolios of rental housing on the basis of expected (abnormal) returns. If informed buyers and sellers participate and if agents disburse relevant information, prices may adjust quickly to new information in active markets. The lack of formal arbitrage activity does not preclude the notion of market efficiency.

PAGE 33

23 Real Estate Market Efficiency Historical View Market imperfections. Historically, the real estate market has been widely perceived and casually asserted to be inefficient. 19 This is not surprising, since the market suffers high transaction and information costs. These costs result from a market distinguished by highly heterogeneous products cleared through a negotiated market rather than an auction market. Therefore, it is difficult for investors to identify and take advantage of market mispricing opportunities which would serve to correct the market. Nonetheless, the existence of transaction and information costs are not sufficient to preclude market efficiency. 20 In addition, other real estate market imperfections such a barriers to entry, indivisible assets, and limited liquidity have been suggested as potential sources of market inefficiency. 21 Liu, Grissom, and Hartzell (1988) theorize 19 Jaffe and Sirmans (1984), in discussing real estate investment decisions, offer evidence that the real estate market is perceived to be inefficient. However, they indicate that scholarly work on the issue is limited. 20 Analyzing security markets, Fama (1970) indicates that all available relevant information can be reflected in the market price despite the existence of high transaction costs. In addition, Copeland and Weston (1988) show that the existence of information costs is not inconsistent with weak-form market efficiency. 21 Roulac (1978) provides a discussion of these issues. It is important, though, to distinguish between perfect markets and efficient markets. Of the two concepts, market efficiency is much less restrictive. For example, it does

PAGE 34

24 that market imperfections are priced. Gau (1985) and others have argued that alternative forms of ownership and financing serve to mitigate market limitations to some extent. Unfortunately, the influence of such market innovations on market efficiency is not clear. Abnormal Returns. Empirical work reporting excessive returns to real estate is often cited as evidence that real estate markets are inefficient. 22 Wendt and Wong(1965), Coyne, Goulet and Piconni (1980), Kaplan (1985), as well as others, indicate that returns on nonsecuritized real estate investments have consistently outperformed alternative asset groups, such as stocks and bonds, on both a risk-adjusted or unadjusted basis. Bruggeman, Chen and Thibodeau (1984), Miles and Mccue (1984), and Hartzell, Hekman, and Miles (1987) find that returns on risk-adjusted commingled real estate funds (CREFs) dominate stock and bond returns, in addition to providing a superior inflation hedge. Finally, Burns and Epley (1982) find similar results for the returns on real estate investment trusts (REITs). not require markets to be frictionless or assets to be infinitely divisible. It simply requires market imperfections to be fully and instantaneously reflected in the market price. Hence, the EMH does not require conditions consistent with a perfect market. 22 While the majority of the work cited in this section analyzes income properties (securitized and unsecuritized), housing returns are viewed as substantially correlated to other real estate assets. Unfortunately, research which relates housing market returns to other capital asset groups is quite limited.

PAGE 35

25 In contrast, Smith and Shulman (1976) find no abnormal returns for REITs while Kapplan and Schwartz (1988) find stock and bond returns to substantially outperform those of real estate limited partnerships (RELPs). Finally, Ibbotson and Siegel (1984) indicate returns, unadjusted for risk, on direct real estate investments fall between those of stocks and bonds, as shown in Table 2-01 below. TABLE 2-01 U.S. Capital Market Total Annual Returns (1947 1982) Arithmetic Geometric Mean Mean Asset Group Return Return Common Stocks: NYSE 12.25% 10.90% AMEX (1960-1982) 9.33 6.66 OTC 15.17 13.05 Corporate Bonds: Long-Term 3.40 2.99 Intermediate-Term 4.84 5.03 Real Estate: Farms 11.32 11.09 Residential Housing 7.50 7.43 Commerical (1960-1982) 8.10 8.03 U.S. Government Securities: U.S. Treasury Notes: 4.38 4.49 u. s. Treasury Bonds: 2.99 3.31 Municipal Bonds: Long-Term: 2.21 1.65 Source: Ibbotson and Siegel (1984) Standard Deviation 17.16% 23.30 21.41 9.71 6.41 7.21 3.89 3.78 4.85 8.57 11.12 Liu, Grissom, and Hartzell (1988) have relaxed the standard assumptions of the capital asset pricing model (CAPM) to explicitly recognize limited marketability, high

PAGE 36

26 transactions costs, and a mildly segmented market. They suggest factors unique to the market imperfections of real estate, such as non-market risk and cost components, are priced by the market. These results are consistent with investment theory, past empirical findings, and the EMH. They argue that excessive risk-adjusted returns to real estate are an artifact of misspecified modeling. Unfortunately, infrequent transactions pose a difficultly to modeling and conducting extensive tests of directly held real estate investments. The Single-Family Housing Market Tests of efficiency conducted on returns to single family housing are marked by an additional difficulty. Unlike most real estate assets, single-family housing is viewed as a consumption good as well as an investment. In this case consumption characteristics of housing assets may dominate the purchase decision. If consumers have heterogeneous housing preferences, tests of efficiency using individual house returns are likely to be judged invalid. Assuming preferences are not time dependent, aggregating constant-quality housing returns across a large number of transactions will reduce this limitation to some extent. It is widely recognized, however, that many home buyers consider housing to be a major investment and price it with respect to its investment characteristics. In doing so, the home buyer considers both its implicit rental stream

PAGE 37

27 (dividends to housing) and the expected future selling price. While the dynamics of this process are poorly understood, these buyers form "subjective" estimates of the expected future value. 23 In this case, the price is formed as the "objective" present value of the expected after-tax "cash flows". Applications of investment motivated housing models are widely applied in the literature. 24 In contrast to directly held real estate investment markets, single-family housing transactions occur frequently, and comparative price data are readily available. In addition, economic factors impacting future housing prices, such as mortgage rates or employment rates, are widely published and often appear in local newspapers. Speculative "Bubbles" Speculative "bubbles" have been alleged in some housing markets. At times, persistent price rises caused by speculative buying in regional markets appear to suggest such conditions. In the 1980s, markets in the northeast and California experienced substantial price increases while markets in Texas declined. Case (1986) examines large reoccurring price increases in the Boston market and Gyouko and Voith (1990) indicate evidence of a "California effect". 23 This terminology is consistent with Maddala (1988) in discussing rational expectations (see Eq. 2-06). 24 Among other places, examples of this appear in the tenure choice literature such as Hendershott and Shilling (1982).

PAGE 38

28 Speculative price "bubbles" suggest upward (downward) price inertia, implying positively autocorrelated returns and market inefficiency. 25 While some supporting evidence exists, rigorous study of the rationale for such conditions is viewed as preliminary. 25 Hypotheses of rational speculative "bubbles" do exist in security markets; however, such theory is quite preliminary at this time.

PAGE 39

CHAPTER 3 LITERATURE REVIEW Principal studies Research on the efficiency of the housing market is an emerging topic in at least two areas of the real estate literature, investment and housing economics. 1 Each area has developed largely independent of the other. To date, no comprehensive effort has been made to review the full body of the literature. Single market studies by Linneman (1986) and Rayburn, Devaney and Evans (1987), and Case and Shiller (1989) and inter-market studies by Guntermann and Smith (1987) and Hamilton and Schwab (1985) are the principal attempts to analyze the efficiency of the single-family housing market. These studies are described in detail in the following sections. In addition, a complete listing of the real estate market efficiency literature appears in Table 3-01 following a review of the principal work. 1 Works by Roulac (1978), and Jaffe and Sirmans (1984), among other provide a basis for relating the EMH to real estate. In addition, empirical work on the efficiency of income properties by Gau (1984, 1985) provided substantial impetus for testing the efficiency of the single-family housing market. 29

PAGE 40

30 Linneman (1986} In what has been described as "a first step in testing the efficiency of the single-family housing market," Linneman conducted a semistrong-form efficiency test of the Philadelphia market. Using a sample of 3,776 owner-occupied residential properties, the owners' contemporaneous assessments of value in 1975 were regressed on nearly fifty detailed structural and neighborhood characteristics considered to be "current public information". In essence, the study described a trading rule that examined whether undervalued houses (those with negative residuals) in 1975 represented arbitrage opportunities in 1978, ex post. The findings indicate abnormal increases to the subsequent values of the undervalued properties. 2 However, when transaction costs were considered, assumed to be twelve percent, an insignificant number of properties were judged to be profitable arbitrage candidates. Thus, the hypothesis of a semistrong-form efficient market could not be rejected. 2 This result may be spurious. Since property values are owner-assessments, they can be assumed to be normally distributed across the actual "market" value. If it is assumed that the total variance remains constant between periods, it is likely that the residuals would have mean reverting tendencies. That is, properties with negative (positive) residuals would subsequently redistribute themselves toward the mean when reappraised, while those properties with zero residuals would randomly change to nullify any "abnormal returns". Thus, consistent with Linneman's findings, "undervalued" properties would appear to yield excessive returns and "overvalued" properties would yield less than normal returns.

PAGE 41

31 The Linneman study is quite limited because it examines one market over a single period. In addition, is assumed implicit rental income to owner-occupied housing to be invariant and proportionate to home value. Consequently, the distribution of implicit rents and risk are assumed equal across properties of equal value. Finally, returns were estimated using each owner's assessment of value rather than actual transaction prices, potentially biasing the results. 3 Rayburn. Devaney. and Evans (1987) Rayburn, Devaney, and Evans (ROE) test a weak-form efficiency hypothesis from a sample of 140,000 residential transactions in Memphis, Tennessee, occurring from 1970 to 1984. Ten submarkets were delineated based on the Memphis City Planning Commission planning districts. From actual transaction prices a monthly series of mean prices was constructed on a square foot basis from a minimum of ten transactions for each of the ten submarkets. Using log transformations of the time series and taking the first differences, subjective tests of autocorrelation 3 The use of subjective value estimates as a proxy for market value is likely to be biased. Since the housing market is heterogeneous, the implicit assumption that all homeowners have identical utility functions, equally valuing different housing characteristics, is highly restrictive. In addition, the estimates are severely contaminated with "private" information. Linneman examines the bias, arguing that it is insignificant. Yet, he concedes that the use of actual transaction price data would strengthen the study.

PAGE 42

32 at various lags were performed. A time pattern of returns was revealed in seven of the ten submarkets. 4 However, applying Alexander's filter rule, a buy-and-hold strategy could not be outperformed for the period sampled, assuming transactions costs as low as two percent. Therefore, the weak-form efficiency hypothesis for each submarket could not be rejected. It was, however, rejected for some markets during the period 1970 to 1975. Reasons for this anomaly could not be explained. Like Linneman, RDE's single market efficiency test was based solely on the change in prices; no explicit examination of the dividend stream to housing was conducted. It is particularly interesting to note also that no explanation was offered as to why submarkets appeared to yield different degrees of efficiency. Case and Shiller (1989) Case and Shiller examine four major metropolitan markets. Using actual transaction price data, repeated sales were identified and a citywide weighted repeat-sales (WRS) index was constructed. Case and Shiller argue that the use of repeat-sales price data for the construction of 4 As indicated later, Case and Shiller (1989] suggest the use of first differences of logged indices may induce spurious correlations.

PAGE 43

33 an index is an improvement over other methods, controlling for quality changes in the sample. 5 Weak-form efficiency tests were performed on the citywide index and on returns which included city-specific rental indices from the U.S. Bureau of Labor Statistics to proxy implicit rent. Consistent with previous work, they find that real changes in house prices are not predictable. However, Case and Shiller find that after-tax excess annual returns are positively autocorrelated and conclude that the single-family housing market is inefficient. Assuming round-trip transaction costs to be zero, they indicate it is possible to construct a trading rule that yields abnormal returns to home buyers who are unconstrained as to time of purchase. 6 They suggest the most likely reason for the apparent inefficiencies is that "predictable" real interest rates are not priced by the market. Nevertheless, their study is subject to rather strong assumptions. They assume that individual marginal income tax rates and property tax rates are constant during the study period. Additionally, they implicitly assume that treasury bill yields accurately predicted the volatile 5 A description of the WRS price index methodology appears in Appendix A, including a discussion of its strengths and limitations. 6They consider only first-time home buyers and home owners who want to trade up and therefore argue that transaction costs can be ignored since the purchase will be made anyway. In addition, they assume all costs associated with waiting to purchase are zero.

PAGE 44

inflation rates of the late 1970s and early 1980s and that the risk premia to single-family housing are constant throughout this time period. Guntermann and Smith (1987) 34 Guntermann and Smith, in a study of 57 metropolitan areas, test the housing market for weak-form efficiency using FHA-based housing price data from 1968 to 1982. They test for efficiency across major metropolitan markets rather than within each market, making their study distinctly different from the previously cited studies. In a second test of efficiency, they proxy implicit rental streams by including a city-specific rental variable computed from the rental component of the CPI for 27 of the metropolitan areas. To their advantage, the use of FHA-based price data imposes a degree of homogeneity on their study. Annually, FHA financing has accounted for ten to forty percent of the value of all new loans on oneto four-family houses in the United States. These homes are largely confined to the middle segment of the U.S. housing market due to statutory maximum mortgage limits. Furthermore, FHA financing insures that the property is in good condition by meeting FHA specifications and complying with standard building code regulations. Using a portfolio approach, Guntermann and Smith compare the correlation of aggregate unanticipated returns

PAGE 45

35 to the aggregate unanticipated returns of subsequent years. 7 Findings indicate no relationship of returns for lags of one to three years and a weak negative relationship in returns for lags of four to ten years. However, a trading rule based on ranked decile groupings of annual returns for the areas could not be constructed to outperform the buy-and-hold strategy after transactions costs were considered. Thus, the hypothesis of an efficient housing market could not be rejected. While Guntermann and Smith's approach is elegant and their findings interesting, the applicability is quite limited. Even if findings had indicated that the housing market was inefficient, it would be difficult for investors or homeowners to exploit the market failure, since investors would need to be able to hold city indexed portfolios of housing subject to homestead property taxes across multiple cities. Alternatively, highly mobile homeowners could earn abnormal returns only by identifying and trading mean return properties that were substitutable across cities. It is unreasonable to assume that investors, development firms, or individual homeowners would be able to 7 In using this approach they note that significant autocorrelation of actual returns is not inconsistent with market efficiency. Ex post actual returns may exhibit significant serial correlation over some time intervals even though expected returns would not. They point out that serial correlation of raw returns over long time intervals would be expected even in an efficient market due to such factors as changes in inflation rates and changes in tax rates.

PAGE 46

36 exploit such market inefficiencies. A more pertinent question is whether investors (development firms, landholding companies, or homeowners) can earn abnormal risk-adjusted returns by sequentially investing in housing on the basis of publically available information within a particular metropolitan area. Additionally, the FHA-based price data have their deficiencies. At least three important problems with the data can be identified. First, the maximum mortgage limit truncates the data base. Thus, there may be systematic moves in the index brought about by periodic statutory changes to the limit. With respect to the overall housing price index, Greenlees (1982) indicates that truncation causes the FHA price index to be biased downward prior to mortgage ceiling adjustments, and biased upward following ceiling increases. Second, homes within the sample that experience abnormal returns may subsequently move out of the sample again because of the truncated data. Finally, findings of efficiency in this market segment may not generalize to housing market segments using other types of financing. Segmented by value, location, and quality, FHA financed homes may appreciate in value at rates different from other market segments.

PAGE 47

37 Additional Research In other studies of housing market efficiency, Hamilton and Schwab {1985), using FHA-based price data, indicate that households failed to incorporate past appreciation in their housing return expectations during the period from 1976 to 1979. However, by incorporating more time periods Guntermann and Smith show that these findings cannot be generalized to include other periods. Finally, Essayyad, Marx, and Green {1987) also used FHA data to conduct conventional random-walk tests following Fama's {1970) methodology. They report findings of weak-form efficiency consistent with those of Guntermann and Smith. Unfortunately, their findings are severely limited by the length of the data series. Table 3-01 provides a listing of the real estate EMH empirical literature.

PAGE 48

38 TABLE 3-01 Tabular Survey of the Real Estate EMH Empirical Literature PANEL A Single-Family Residential Market Studies, (City-Specific) Case and Shiller (1989) Test Markets: Atlanta, GA; Chicago, IL; Dallas, TX; San Francisco, CA. Test Period: 1970:1-1986:2; 62 Quarters. Index: Quarterly index constructed using weighted-repeat sales methodology. Test Form: Weak-form efficiency. Data: Repeat sales were identified for each market from a data base of individual sales of single-family homes maintaine by the SREA Market Data Center in Atlanta, GA and the California Market Data Cooperative in Glendale, CA. Major Results: Hypothesis of weak-form efficient market is rejected since excess returns are found to be autocorrelated. Findings indicate buyers indifferent to time of purchase could earn abnormal returns. Suggest trends in real interest rates do not appear to be incorporated in prices. Rayburn, Devaney and Evans (1987) Test Market: Memphis, TN; subdivided into 10 single-family housing submarkets. Test Period: 1970:1-1984:12; 180 months. Index: Monthly mean price per square foot in each submarket, based on a minimum of Test Form: Data: ten transactions. Weak-form efficiency. Submarket transactions were identified from a data base of 140,000 sales maintained by the Chandler Company in Memphis, TN. Major Results: Hypothesis of a weak-form efficient market could not be rejected. Evidence of autocorrelated returns were found in seven of the ten submarkets. However, an investment rule could not be constructed to outperform the buy and-hold strategy. No seasonal patterns were found to the rate of change in the index. Skantz and Strickland (1987) Test Market: Houston, TX. Test Period: 1977 and 1981. Index: Test Form: Data: Major Results: Linneman (1986) Cross-sectional and time-series analysis using using ordinary least squares regression model Semistrong-form efficiency. Actual transaction data on 183 houses located in two Houston-area subdivisions. Each transaction included price, square footage, lot size, financing, month of sale, age of home, location data. Findings are consistent with a semistrong-form efficient single-family housing market. The market reacted quickly and rationally to new publicly available information. Test Market: Philadelphia, PA. Test Period: 1975 and 1978. Index: Cross-sectional analysis using hedonic model and a one-period modified Test Form: Data: Major Results: PANEL B Gunterman and Test Market: Test Period: Index: Test Form: Data: Major Results: capital asset pricing model. Semistrong-form efficiency. Annual Housing Survey, using owners assessment of value as well as structural and neighborhood variables. Hypothesis of a semistrong-form efficient market cannot be rejected when transaction costs are included. Single-Family Residential Market Studies, (U.S. Market) Smith (1988) U.S. single-family housing market using 57 metropolitan areas. 1967:1-1982:1; 15 years. Annual mean price per square foot for each metropolitan area for existing homes using FHA financing. Weak-form efficiency. FHA 203(b) data available in publications of the U.S. government. Hypothesis of a weak-form efficient market could not be rejected. Evidence of autocorrelated returns found, however, trading rule cannot be constructed to outperform buy-and-hold strategy.

PAGE 49

39 TABLE 3-01 -continued Essayyad, Marx Test Markets: Test Period: Index: and Green (1987) 73 U.S. metropolitan areas. 1966-1985; 19 years Annual mean price per square foot for each metropolitan area for existing homes using FHA financing. Test Form: Weak-form efficiency. Data: FHA 203(b) data available in publications of the U.S. government. Major Results: Hypothesis of a weak-form efficient market could not be rejected. Hamilton and Schwab (1985) Test Markets: U.S. single family housing market, using 49 metropolitan areas. Test Period: 1974-1976. Index: Cross-sectional analysis using hedonic model. Test Form: Rational expectations. Data: Annual Housing Survey. Estimates of appreciation are based on changes in each cities average sales price of existing single family homes insured under the FHA 203(b) program. Major Results: Results do not support the rational expectations hypothesis. Households fail to accurately incorporate past appreciation in housing into their future expectations of housing appreciation. PANEL C Gau (1985) Test Market: Test Period : Index: Test Form: Data: Major Results : Gau (1984) Test Market : Test Period: Index: Test Form: Data: Major Results: PANEL D Multi-Family Residential Market Studies, (City-Specific) Vancouver, B C.; apartment market. 1971-1980; 120 months. Price per square foot using a single, but different, property transaction each month within a defined submarket. Semistrong-form efficiency; event study. Apartment transaction data (See Gau (1984)) Could not reject semistrong-form efficiency hypothesis. Investors are unable to earn abnormal risk-adjusted returns on the basis of new government policy information or interest rate movements. Vancouver, B C ; apartment and commercial markets 1971-1980; 120 months. Price per square foot using a single, but different, property transaction each month within a defined submarket. Weak-form efficiency. Transaction data including price, date, location, use, square footage, and t he number of units. c ould not reject the hypothesis of a weak-form efficient market. study reports a general absence of significant autocorrelation and support for the fair game model as a representation of price behavior in real estate markets. Multi-Family Residential Market Studies, (U.S. Market) Isakson and Mcinish (1988) Test Market: 18 U.S. metropolitan areas. Test Period: 1978-1983; 5 years. Index: Examined the income component of multi-family housing returns for Test Form: Data: autocorrelation. Did not include the capital appreciation component. Weak-form efficiency. Average income and operating expenses of apartments reported by the Institute of Real Estate Management (IREM). Major Results: Hypothesis of a weak-form efficient market could not be rejected for lags of one year; however, the hypothesis was rejected for lags of two and three years.

PAGE 50

TABLE 3-01 -continued PANEL E Davies (1977) Test Market: Test Period: Index: Test Form: Data: Major Results: Urban Land Market Studies, (City-Specific) London, Canada 1966-1973; 93 months. Monthly median sales price of vacant lots. Semistrong-form efficiency. Land transactions reported in Teela Digest. Evidence indicates that builders form their expectations about future prices and costs primarily on the basis of the rate of change in the profitability of building and selling houses four months prior to the current period. However, evidence suggests that the adoption of restrictive land use policies triggered an immediate and sustained speculative increase in prices due to increased costs or uncertainty. Adams, Milgram, Green, and Mansfield (1968) Test Market: Northeast Philadelphia. Test Period: 1945-1962, 14 years. Index: Cross-section hedonic model with time trend dummy. Test Form: Residual analysis with macroeconomic variables. 40 Data: 1111 land transactions and associated land characteristics, such as access, condition, and zoning. Major Results: Interest rates were inversely related to regression residuals and positively related to construction activity and general economic conditions. Land development trends were anticipated and priced by the market PANEL F McIntosh and Test Market: Test Period: Indices: Test Form: Data: Commercial Market Studies, (City-Specific) Henderson (1989) Dallas-Fort Worth, Texas. 1979-1985, 73 months. Three indices were constructed: (1) price per square foot of net rentable area, (2) gross income multiplier, and (3) net income multiplier, each using a single, but different, property transaction each month. Weak-form efficiency 127 office transactions and associated characteristics, such as address, vacancy rates, financing, land area, age, condition, and parking. Major Results: Hypothesis of a weak-form efficient market could not be rejected. PANEL G Locke (1985) Test Market: Test Period: Indices: Test Form: Data: Major Results: Real Estate Securities Market Studies, (Foreign Market) British and Australian income producing real estate markets. 1980s, Specific period not by study. Indices were constructed using the ratio of the change in monthly prices of the property investment trusts. Weak-form efficiency Companies listed on each respective country's stock exchange which invested primarily in real estate were identified and aggregated to form an index of real estate company returns. Hypothesis of a weak-form efficient market could not be rejected.

PAGE 51

41 Summary of the Literature In summary, research in the area of real estate market efficiency is quite limited and findings have been mixed, largely due to insufficient data. Linneman (1986) indicates that the hypothesis of a semistrong-form efficient housing market cannot be rejected. Rayburn, Devaney and Evans (1987), and Guntermann and Smith (1988) report that the housing market is consistent with weak-form efficiency. In addition, these three studies indicate that trading rules cannot be constructed that consistently yield abnormal risk adjusted marginal returns to housing net of transaction costs. In contrast, Case and Shiller (1989) reject the hypothesis of a weak-form efficient housing market. Although, they find real house prices to be unpredictable, after-tax excess returns are found to be positively autocorrelated. Case and Shiller indicate that buyers indifferent to time of purchase can construct a trading rule that yields abnormal returns. However, their study is subject to some rather strong assumptions. They assume home buyer marginal tax rates and risk-premia are constant and although evidence exists to the contrary, inflation expectation errors by investors are assumed to be random during the unprecedented inflation of the 1970s. While important, these findings are partial and preliminary Data limitations are such that there exists

PAGE 52

little hope of definitively answering the efficiency question by conducting comprehensive studies on the total U.S. housing market. Until such time that adequate data become available, it appears that a piecewise research agenda is required to resolve definable market specific efficiency issues. 42

PAGE 53

CHAPTER 4 GENERAL HYPOTHESES This study hypothesizes that the single-family housing market is weak-form efficient. This implies that investors (home buyers) cannot construct ex ante investment rules, on the basis of historical information alone, that consistently yield above-normal risk-adjusted returns. The study, in examining both real returns and after-tax excess returns, relaxes the strong assumptions of Case and Shiller (1989) and tests the robustness of their results. More specifically, the assumptions that home buyer marginal tax rates are constant, inflationary shocks are stochastic, and the risk-premia to housing do not vary are relaxed. It is speculated that the findings of autocorrelated after-tax excess returns, reported by Case and Shiller, are largely induced by the autocorrelated attributes of these factors. Therefore, when ex post returns are constructed using time-varying marginal tax rates and are adjusted for unexpected inflation and time varying return volatility, abnormal returns (deviations from expected returns) are hypothesized to follow a random-walk. Thus, consistent with the EMH, there exists no information 43

PAGE 54

44 in the historical housing return series that can be used to consistently earn excessive profits. Expressed formally, the model defined earlier (see Eq. 2-01), is restated in terms of housing returns such that, E(fn t+1) = E[Rn t+1 E(~ t+1 I 1 t)] = 0, , ( 4-01) and (4-02) where Rn, t+ 1 is the actual nominal return in period t+l to housing, E(Rnt+ 1 1tt) is the expected nominal return in period t+l to housing given the information set, t, in period t, and n, t+1 is the abnormal return, which is assumed to be normally distributed with mean, zero, and variance, a 2 As described previously, if the market is efficient differences between the actual ex post nominal return series and the expected nominal return series can be characterized as a random-walk. Findings of autocorrelation in the predicted errors suggest that not all available historical information has been used in setting prices, and provisionally rejects the EMH. Expressing Equation 4-01 in terms of after-tax excess returns yields where RFt+ 1 is the risk-free rate in period t+l, T is the home buyer's marginal tax rate, and the difference between the actual nominal return and the after-tax risk free rate

PAGE 55

45 is defined as the after-tax excess return. 8 From this expression it can be seen that if income tax rates are time varying, they must be included in the construction of actual returns, since they are imbedded in the formation of home buyers return expectations. Thus, assuming marginal tax rates to be constant may induce spurious autocorrelation in the prediction errors. Additionally, tests of efficiency assume that market inflation expectations are accurately imbedded in both the risk-free rate and housing return expectations. If inflation prediction errors are random, they will have no impact on any subsequent tests of efficiency. However, there is evidence that during the volatile 1970s and early 1980s, markets in general systematically underestimated price level changes, resulting in autocorrelated inflationary shocks. 9 Since this condition appears to be an enigma confined to the time period studied, tests of efficiency over this sample period may be biased and judged 8 Income taxes on returns to housing are assumed to be zero. Since implicit rent is not taxed and capital gains due to appreciation can be avoided by continuously rolling over gains on principal residences, this assumption is reasonable. Additionally, consistent with this and other work, RR= MAX [MB, (1-t)TB] + RP. This states that the required (expected) return on housing, RR, is equal to the maximum rate between municipal bills and the after-tax treasury bill, plus some risk premium. As shown later, assuming the marginal tax rate of the median priced home buyer, the after-tax rate on the treasury bill is the dominant security in determining the required rate of return on housing. 9 A discussion of this evidence is found in Chapter 8.

PAGE 56

46 myopic. It is therefore argued that the after-tax excess returns to housing should be adjusted for unexpected inflation and the Case and Shiller (1989) results examined to determine their sensitivity to such conditions. Using Equation (4-03), the general condition can be expressed to explicitly consider the impact of inflationary shocks such that, E ( t+1) = E [ (Rn, t+1( 1-r) RF t+1-Rui, t+1) -E ( (Rn, t+1( 1-r) RF t+1) I t t' t", t) ] = o, ( 4-04) where Rui t+i is the after-tax excess return attributable to I unexpected inflation, and t 7 t denotes the information I available int to form inflationary expectations in t+l. Still, (4-04) assumes that the risk-premium associated with housing returns is constant. Recent research provided some evidence of time-varying risk-premia in security markets; however, such tests of risk-premia have yet to be conducted in real estate markets. 10 If changes in risk are random or constant, the removal of the return component due to risk is unnecessary and tests of efficiency are not biased. Omitting conditions of autocorrelated risk-premia however, will induce spurious conditions of autocorrelated returns and bias the subsequent tests of efficiency. 10 A discussion of time-varying risk-premia is found Chapter 9.

PAGE 57

47 Expressed to include volatility expectations, Equation (4-04) can be explicitly modeled by simply stating E ( t+1) = E[ (Rn t+1-(1-'T)RFt+1-Rui t+1)-E( (~ t+1-(1--r)RFt) I tt,t,.. t'ta, t)] I I I I I = o, ( 4-05) where ta, t denotes all historical information on risk in the housing market available int. This model formally states this study's general hypothesis; when after-tax excess returns to housing are constructed considering time-varying marginal tax rates, unexpected inflation, and return volatility, abnormal returns are hypothesized to follow a random-walk with mean equal to zero, (E(t)=0) 11 11 Alternatively, this condition can be formalized as follows: Assume R = MAX [MB,(1-t)TB] +RP+ e, where R is the actual nominal return to housing, MB is the yield on municipal bills, (1-t)TB is the after-tax yield on treasury bills, RP is the risk premium to housing and e is a random error term. Separating yields: r + r + r. = (1-t) (tb+ei) +RP+ e e1 Ul , where r is the expected real return to housing, rei is the expected return to housing due to inflation, rui is the unexpected return to housing from inflation, tb is the real treasury bill yield, and ei is the expected inflation rate imbedded in the treasury bill. Under this condition expected returns equal actual nominal returns less returns due to unexpected inflation, E(R) = R rui" Rearranging, we get: E(Re) = R (1-t) (TB) rui = RP + e, where E(Re) is the expected excess return, and e is a random variable. Therefore, E(e) = E[ (R-(1-t) (TB)-r i)-E(Re)] = 0, where E(e) is a random variable with mean zero and variance a 2 assuming the risk-premium is random.

PAGE 58

CHAPTER 5 TESTABLE HYPOTHESES To test the hypothesis that the housing market is weak form efficient, conventional methodologies reported in the finance literature are employed. Historical single-family housing returns are assumed to constitute the relevant information set and are hypothesized to be instantaneously captured in the future return expectations of home buyers. Therefore, in a weak-form efficient single-family housing market, home buyers cannot predict future abnormal returns to housing on the basis of past price or return information and consequently, cannot consistently "time the market". If home buyers expect a positive return on their investment, future price movements should follow a random walk return generating process with "drift". Consistent with conventional tests of the random-walk model, ex post returns to housing are constructed, tested for autocorrelation and the findings compared to previous findings. In addition, the test methodology in Chapter 8 examines the random-walk characteristics of any abnormal returns (return prediction errors) to housing. Hypotheses of the EMH are primarily tested by examining the first-order autocorrelation characteristics of each 48

PAGE 59

49 relevant return series. Regression coefficients of the alternative return series are estimated and evaluated using the general linear model, (5-01) where Rct> is the relevant ex post return in period t and L denotes the number of periods lagged. The expected return that is unrelated to the previous return is captured in the intercept term, B 0 Since home buyers expect a positive return in housing assets, the intercept is postulated to be positive when testing real housing returns and zero if testing abnormal returns to housing. The slope of the regression, B 1 measures the correlation between the return in period t-L and that in period t, given B 0 Under the hypothesis that the market is weak-form efficient, B 1 = o, should not be rejected. A significant positive or negative slope coefficient suggests that all historical return information is not used by investors. Under this scenario it may be possible, given transaction costs, to devise an investment rule based on this information to earn excessive profits. The error term, u, is assumed to be independent and normally distributed with mean zero and variance, a 2 In addition to testing annual returns to housing, the random-walk characteristics of the quarterly returns are briefly examined. Autocorrelation functions of the

PAGE 60

50 quarterly returns are estimated and examined at various lags. If there is no significant autocorrelation at important lags (first or fourth) then it is concluded that there is no autoregressive or moving average structure in the time series. In addition, the results of a chi-square test, the Q-statistic, for no time pattern in the series are reported for each quarterly return series. 1 If the statistic is close to zero, we accept the null hypothesis that there exists no time pattern in the return series. This result is consistent with the a weak-form efficient market. 1 If the autocorrelation function, ACF, is defined as, ACFL = covariance (R'R) / variance (R), the Q-statistic is composed of the first L residual autocorrelations ACF 1 ACFL, such that, Q = T :E ACF 2 (L), for L=l to L. Hence, the Q-statistic is the sum of squared independent normal random variables, with mean 0 and variance 1/T. Box and Pierce (1970) have shown that it is distributed as chi square.

PAGE 61

CHAPTER 6 TEST METHODOLOGY, BASE MODEL, AND DATA This chapter describes the methodology used to test the housing market EMH. An overview of the entire test methodology is first outlined. Following this, the initial model (base model) and data are described in detail. Initial tests of the EMH are then conducted using the base model and the results reported. The chapter concludes with a discussion of the base model assumptions and their potential impact on the initial test findings. overview of the Test Methodology The test methodology employed is comprised of four principal steps. In each step, an assumption of the preceding model is relaxed and a time-series of housing returns constructed. Conventional tests of the EMH outlined in Chapter 5 are applied to each of the constructed return series and the results evaluated. In the first step, real returns and after-tax excess returns to single-family housing are estimated for four U.S. metropolitan markets (Atlanta, Chicago, Dallas, and San Francisco) using methodologies similar to Case and Shiller 51

PAGE 62

52 (1989). 1 To insure the accuracy of the constructed data, tests of autocorrelation, replicating Case and Shiller, are conducted on each return series. These initial tests provide baseline statistics used to compare and evaluate the subsequent tests of weak-form efficiency conducted in the following phases under alternative model assumptions. The results of these initial tests are reported later in this chapter. In the second step, using data from Ling (1990) and other secondary sources, a series of time-varying home buyer marginal tax rates are estimated for each city. After-tax excess returns to housing are then reconstructed using the city-specific home buyer time-varying marginal tax rates rather than a constant marginal tax rate, as assumed previously by Case and Shiller. Tests of autocorrelation are conducted on both the annual and quarterly return data. These procedures and the data used are discussed in Chapter 7. After-tax excess returns to housing are adjusted for unexpected inflation in the third step. In this phase a step-ahead autoregressive integrated moving average (ARIMA) model is used to model annualized inflation expectations on a quarterly basis. The values forecasted by the ARIMA model are compared to survey data available every other quarter to 1 Base model after-tax excess returns to housing use the same procedure and are identical to those estimated by Case and Shiller.

PAGE 63

53 determine their validity. The after-tax excess returns are then regressed on the unexpected inflation series and the estimated regression coefficients evaluated. The estimated intercept and slope coefficient of the regression are interpreted as the expected return to housing and the after tax excess return component to housing associated with unexpected inflation, respectively. Additionally, the residual is defined as the abnormal after-tax excess return to housing. Abnormal after-tax excess return series are constructed for each city and their random-walk characteristics statistically examined. This phase of the EMH test is described in Chapter 8. Step four, discussed in Chapter 9, examines the impact of time-varying return volatility on expected returns. An autoregressive conditional heteroskedastic (ARCH) model is used to estimate expected after-tax excess returns adjusted for return volatility. If the risk-premium to housing is time-varying we would anticipate expected volatility (risk) to be positively correlated with expected after-tax excess returns and unexpected volatility to be negatively correlated. In this model, if volatility is shown to be time-varying and predictable, the random-walk properties of the model residuals, defined as abnormal returns, are examined and subsequently used to test the hypothesis that the single-family housing market is weak-form efficient.

PAGE 64

54 Description of the Base Model Real Returns to Housing Housing returns are constructed using both the house appreciation component and the dividend (implicit rent) stream. 2 Using the quarterly WRS housing price indices for each market reported by Case and Shiller (1987), ex post annual real returns for sample j in quarter t, Rr,j_Ct>, are defined as Rr,j_Ct> = exp(WRSj_Ct+ 4 > WRSj_Ct>J + CJ [ Oct> +Dct+1 > +Dct+2> +Dct+3>] exp [WRSLCt>] [CPI-UCt+ 4 /CPI-UCt>]. (6-01) WRSj_Ct> is the nominal WRS price index ( in logs) constructed from sample j in quarter t; c 1 is a constant used to calibrate the model; Dct> is the quarterly city-specific residential rent component of the consumer price index CPI U; and CPI-Uct> is the quarterly city-specific consumer price index, net of shelter. In Equation (6-01), the annual housing return due to the house price appreciation component is captured by the 2 Real changes in house prices were analyzed by Case and Shiller (1989). In contrast, this study includes the return component due to implicit rent and analyzes the predictability of total real returns to housing. If it is assumed that implicit rent is not taxed and capital gains can be avoided, these returns can be interpreted as real after-tax returns.

PAGE 65

55 first term, exp[WRSj_Ct+ 4 > WRSj_ct>l, while returns due to the dividend stream to housing are captured by the second, ( Cj [D +D +D +D]) over exp [WRSL] The nominal return to housing is then adjusted for changes in the general price level, [CPI-Ul, to derive real returns to housing. Cj is initially set so that the average dividend to price ratio is five percent during the study sample period, hence, = 0.05 (6-02) for each city. 3 After-Tax Excess Returns to Housing Excess returns, are defined as the realized return (actual) on a risky asset in excess of the contemporaneous risk-free interest rate. Thus, Re= Ra Rf, where Re denotes the excess return, Ra is the realized return on a risky asset, and R 1 is the risk-free rate of return. After3 rf c. is set so that the assumed dividend to housing ratio differs from the actual ratio and the actual ratio is not constant, subsequent tests of the random-walk properties of the time series will be biased. Since sizable returns due to price appreciation are commonly anticipated during the sample period, the five percent dividend to price ratio, assumed by Case and Shiller, appears quite high. However, the impact of this assumption on tests of autocorrelation is found to be minor.

PAGE 66

56 tax excess returns are defined simply as excess returns constructed on an after-tax basis. For housing, the aftertax excess return for sample j in period t, Re,j_Ct>, is formally defined, = exp [WRS j_Ct+ 4 > WRS j_Ct>] + exp [WRSj_Ct> J 1 [ ( 1'f < t> ) TB ct,/ 10 0 ] (6-03) where -r Ct> is the home buyer marginal personal income tax rate and TB is the one-year treasury bill rate. Here it is assumed that neither capital gains nor dividends to housing are subject to income taxes; however, interest on risk-free securities is considered taxable. 4 Description of the Data To estimate the models, quarterly WRS price indices, available for the metropolitan areas of Atlanta, Chicago, Dallas, and San Francisco, were obtained directly from Karl Case. Constructed using a weighted repeat-sales methodology 4 Adjusting after-tax housing returns to account for property tax changes and services received from the local government would be appropriate to model, however, adequate data do not appear to be available at this time. Therefore, any subsequent tests of autocorrelation assume that year-to year changes in these variables are not serially dependent. In addition, base model marginal tax rates, -r, are assumed to be constant at thirty percent, consistent with Case and Shiller.

PAGE 67

57 from a data base consisting of nearly one million recorded transactions (221,876 Atlanta, 397,183 Chicago, 211,638 Dallas, 121,909 San Francisco), each city index extends from the first quarter of 1970 and through the second quarter of 1986. 5 The underlying transactions data for Atlanta, Chicago and Dallas, as well as the transactions data from before 1979 for San Francisco, were originally obtained from the Society of Real Estate Appraisers' Market Data Center in Atlanta, Georgia. The transaction data for San Francisco from 1979 through 1986 were obtained from the California Market Data Cooperative in Glendale, California. To estimate the dividend to housing, city-specific residential rent indices are obtained from the U.S. Bureau of Labor Statistic (BLS) Consumer Price Index Division in Washington D.C. General inflation for each city is estimated using the city-specific CPI-U, net-of-shelter, price level index from the BLS. One-year and three-month treasury bill yields as reported in the Federal Reserve Bulletin are used as a proxy for the annual and quarterly risk-free rate. A description of the data used to construct home buyer marginal tax rates and inflationary expectations in the second and third steps of the analysis is deferred until 5 sales transaction data for San Francisco were available for the third quarter of 1986. The empirical results for this market include this additional period.

PAGE 68

Chapters 7 and 8. Table 6-01 below, summarizes the data used in testing the base model. TABLE 6-01 Summary of the Data Used to Estimate the Base Model Series Dates Used Freq Source Weighted Repeat Sales 70:1-86:2 Housing Return Indices (Samples A,B, and T). Q Yield on one-year 70: 01-86: 06 M treasury securities (Nominal). Yield on three-month 70:01-86:06 M treasury securities (Nominal). CPI-U, Net of Shelter 70:1-86:2 (City-Specific). CPI-U, Residential 70:1-86:2 Rent (City-Specific). Q Q Karl Case & Robert Shiller. Federal Reserve Bulletin. Federal Reserve Bulletin. Bureau of Labor Statistics (BLS). Bureau of Labor Statistics (BLS). 58 Note: Mand Q indicate that the data series used were available monthly or quarterly, respectively. Samples A and B refer to half-sample indices for each city, while sample T refers to the city index estimated from the complete sample of sales transactions. Base Model EMH Test Results Annual real returns and annual after-tax excess returns to single-family housing are first estimated on a quarterly basis from 1970:1 to 1986:2 for Atlanta, Chicago, and Dallas and from 1970:1 to 1986:3 for San Francisco. Later, quarterly returns are estimated for each series over the

PAGE 69

59 same sample period on a quarterly basis. Statistics of real quarterly returns for each city are reported in Table 6-02. Consistent with previous work, real housing returns are estimated to be similar to that of common stock during the study period, while the variances of housing returns are less than that of stock. In addition, the regional nature of the real estate market is evidenced by low cross correlation statistics, although real returns in each of the cities studied are positively correlated. However, estimated annual real returns and after-tax excess returns, shown in Figure 6-01, yield some common characteristics. Real and after-tax excess returns appear highly correlated with the period of rising inflation in the 1970s and the subsequent decline in inflation in the early 1980s. In addition, after-tax excess returns, in each market, are less than real returns for the periods after mid-1980. To insure the accuracy of the initially constructed return series and to provide baseline test statistics, tests of first-order autocorrelation are conducted that replicate those reported by Case and Shiller (1989). Commonly, tests of first-order autocorrelation involve regressing a time series of returns on the same time series lagged one period. However, Case and Shiller have shown that when using the housing return model above, spurious serial correlation may result from such tests. Alternatively, they indicate that

PAGE 70

TABLE 6-02 Summary Statistics of Capital Market Real Quarterly Returns (1970:1 1986:1) L.Term L.Term I.Term Real Change Real Real Real Real Common Small Corp. Gov. Gov. T-Bill in ReturnsReturnsReturnsReturns statistic stocks Stocks Bonds Bonds Bonds Rates CPI-U (ATL) (CHI) (DAL) (OAK) Mean .013 .031 .009 .008 .007 .003 .016 .013 .013 .018 .023 Std. Dev. .091 .143 .071 .072 .043 .009 .010 .028 .023 .027 .026 Skewness -.389 .204 .424 .629 .534 .365 .392 .502 -.859 .038 -.084 Kurtosis 3.618 3.344 3.934 4.147 3.822 2.626 2.734 4.864 3.416 2.914 5.044 CrossCorrelations common Stocksl.000 Small Stocks .800 1.000 LT Cor. Bonds .607 .380 1.000 LT Gov. Bonds .545 .331 .963 1.000 IT Gov. Bonds .487 .300 .955 .954 1.000 Real T-Bill .305 .157 .441 .470 .529 1.000 CPI-U Infl. -.413 -.238 -.517 -.511 -.536 -.731 1.000 Real Housing Returns (ATL) .212 .238 .216 .145 .142 .087 -.159 1.000 Real Housing Returns (CHI) .143 .017 .321 .283 .235 -.026 -.3 3 8 .066 1.000 Real Housing Returns (DAL)-.084 027 -.065 -.041 -.023 .010 -.172 .013 .142 1.000 Real Housing Returns (OAK) .292 .325 .267 .248 .190 -.216 -.148 .153 .174 .029 1.000 Stock, bond, and series as reported in SBBI 1989 Yearbook, Ibbotson Associates. 0\ 0

PAGE 71

OJ ~---------------------, 0.2 ~-----------------------, Ill h,, "'---"i'...,..----,::;;;r------------:--,---------->.;-----t -G.I ..___... _.__........_~.___~__._-~__._ _.__........_~~~~-~~ -~ m m m m m m m m m !Ill m __. R~ Reun --oElem Rehm Atlanta OJ r---------------------~ 0.2 r----------~,=......,,,'-++-------~ Ill ' a .. t, -di ~-'---~--'--..__-'----'---'--......L--'---'----'---'-~---'--'-' m m m m m ll7I m m 879 !Ill !!I m -~~ Dallas .I! .,._ : OJ 0.2 Ill I '\ I \ \ / ~ L--: / ~J BlU 872 m 1914 ffl ll7I BT! 1911 879 !Ill !!I m __. R~Rehm --oElem Rehm Chicago OJ 0.2 OJ ~J ..___... _.__ __._ _.__ ____._ _.______._ _.__ __.__...____._ _,_______.__...____.__.,_, m m m m m ll7I m m 579 !Ill !!I m San Francisco FIGURE 6-01 Real and Excess Annual Returns to Housing O'I ....

PAGE 72

62 such correlation is avoided by constructing two independent indices for each city and regressing the changes in one index on the lagged changes of the second index. Consequently, they randomly divide the underlying transaction data for each city into two half-samples and estimate WRS price indices for each sample. Each of the resultant indices (two half-sample indices for each city) were obtained from Karl Case and are used to estimate returns to housing (real and after-tax excess returns) using Equations 6-01 and 6-03. Annual real returns and annual after-tax excess returns are tested for first-order autocorrelation by adjusting Equation (5-01) to consider returns generated from each of the half-sample indices. Modifying (5-01) yields, Rr,j Ct> = Bo + B, Rr,k Ct-L> + uCt)' (6-04) and Re,j_Ct> = Bo + B, Re,k_Ct-L> + uCt>' (6-05) where j and k indicate that the respective return is constructed from each half-sample index and Lis the number of periods lagged. To increase the sample size, annual returns are calculated on a quarterly basis and L set equal to four quarters. This eliminates the possible influence of seasonality in this test. 6 Equations (6-04) and (6-05) are 6 case and Shiller (1987) report that while second quarter price changes tend to be high and third quarter changes low, the difference is small, and only in Chicago is seasonality significant at the five percent level.

PAGE 73

63 estimated using ordinary least squares (OLS) regression. To correct for the overlapping characteristics of the resulting data series, the estimated coefficient standard errors are adjusted using the methodology suggested by Hansen and Hodrick (1980). The results of these regressions are reported in Tables 6-03 and 6-04. For each city, the contemporaneous regression of one return series (constructed from half sample j) on the other return series (constructed from half sample k) is first reported. If the indices were measured without error, the estimated B 1 coefficient and the coefficient of determination, R 2 would both be equal to 1.0. However, the errors-in-variables characteristics should result in the estimated coefficients being somewhat less than 1.0. In this set of regressions the estimates of B 1 for each of the cities, are found to be significantly positive. In addition, the estimated B 1 coefficient for the cities of Atlanta, Chicago, and Dallas is close to the anticipated level of 1.0. However, both the estimated B 1 coefficient and the R 2 statistic reported for San Francisco are relatively small, suggesting that the series may contain substantial measurement error. 7 This result is slightly different from that reported by Case and Shiller. It is suspected that the difference is due to differences in the 7 This is later evaluated using quarterly return data.

PAGE 74

TABLE 6-03 Regression of Annual Real Returns on Annual Real Returns Lagged (Full Sample Period) City Parameters Atlanta j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Chicago j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Dallas j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 San Francisco j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Rr_j(t) = Bo + B, Rr_kCt-L> + UCt> Period 1970:1-1985:2 1970:1-1985:2 1970:1-1985:2 1970:1-1985:3 Obs. Bo s, R 2 Root MSE(t-stat) (t-stat)adjR 2 62 0.028 58 0.045 58 0.049 62 0 027 58 0.058 58 0.057 62 0.048 58 0.059 58 0.053 63 0.069 59 0.073 59 0.090 0.012 ( 1. 180) 0.039 (2.716) 0.038 (2.501) 0.003 (0.291) 0.024 ( 1. 184) 0.028 ( 1. 428) 0.028 ( 1. 723) 0.055 (2.493) 0.044 (2.334) 0.049 ( 1. 878) 0.050 (2.090) 0.070 (2.234) 0.768 (6.175) 0.185 (1.090) 0.211 (1. 089) 0.948 (8.291) 0.540 (2.334) 0.452 (2.045) 0.724 (4.477) 0.328 (1. 558) 0.296 (1.759) 0.488 (2.803) 0.451 (2.800) 0.300 (1.320) .658 .652 .045 .028 .045 .028 .832 .829 .274 .261 .222 .208 .439 .429 .099 .083 .117 .101 .293 .282 .253 .240 .074 .058 64 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MA3 process as described by Hansen and Hodrick (1980).

PAGE 75

TABLE 6-04 Replication of Case and Shiller (1989) Statistics Regression of After-Tax Annual Excess Returns on After-Tax Annual Excess Returns Lagged City Parameters Atlanta j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Chicago j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Dallas j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 San Francisco j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Re_j(t) = Bo + B, Re_kCtL> + uCt) Period 1970:1-1985:2 1970:1-1985:2 1970:1-1985:2 1970:1-1985:3 Obs. B 0 B 1 R 2 Root MSE(t-stat) (t-stat)adjR 2 62 0.029 58 0.048 58 0.049 62 0.027 58 0.051 58 0.052 62 0.048 58 0.063 58 0.061 63 0.071 59 0.100 59 0.079 0.011 (0.960) 0.039 (2.163) 0.040 (2.199) 0.004 (0.427) 0.017 (0.879) 0.020 (1.040) 0.026 (1.589) 0.045 (1. 778) 0.032 (1.381) 0.042 (1. 585) 0.047 (1.722) 0.056 ( 1. 551) 0.825 (6.229) 0.350 (1.780) 0.328 ( 1. 610) 0.933 (8.593) 0.707 (3.531) 0.664 (3.391) 0.794 (5.714) 0.551 (2.654) 0.529 (2.965) 0.604 (3.758) 0.552 (3.512) 0.509 (2.200) .688 .683 .136 .120 .113 .097 .847 .844 .481 .472 .452 .442 .589 .582 .285 .272 .291 .279 .459 .450 .381 .370 .205 .191 65 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MA3 process as described by Hansen and Hodrick (1980).

PAGE 76

calibrating of the model and the inclusion of the entire return series. 66 Statistics are reported for the regression of annual real returns to housing constructed with sample A on the one-year lagged real returns constructed from sample Bin Table 6-03. The estimated B 1 coefficients from these regressions are consistently positive, suggesting that there is evidence of inertia in real returns. This is consistent with the findings of Case and Shiller with respect to changes in real house prices. Still, B 1 coefficients significant at the five percent level are found only in the Chicago market, the largest of the four markets. While some inertia does exist in each of the real return series, from these results it appears unlikely that excessive profits are consistently predictable or available. Comparing these results with those reported by Case and Shiller indicate that including the implicit rental stream has little impact on the random-walk properties of the real return series in all the markets studied, with the exception Chicago. In Chicago, the first-order autocorrelation properties of the real return series are stronger than that of the changes in real housing price series. The reason for this condition is not clear. Table 6-03 indicates that, in the four markets studied, real returns become more predictable as the size of the market increases. It is speculated that in larger markets, such as Chicago,

PAGE 77

67 competing real estate assets are more difficult to identify. City-wide information, while available, is cumbersome and difficult to assimilate; hence, transaction costs increase, resulting in decreased efficiency. In contrast, Table 6-03 shows that after-tax excess returns to housing are found to exhibit substantial serial dependency in all of the markets studied. The hypothesis that 8 1 = 0, is rejected at a 95 percent confidence level for the markets of Chicago, Dallas, and San Francisco. Although the hypothesis can not be rejected for Atlanta substantial return inertia exists. These findings replicate those reported by Case and Shiller (1989), confirming the underlying constructed after-tax excess return series. In addition to tests using annual returns to housing, quarterly real returns and quarterly after-tax excess returns in each city are examined for first-order autocorrelation. The results of these tests are reported in Tables 6-05 and 6-06. Regression of real quarterly returns on itself indicates that the two half-sample price indices are best matched in the Atlanta and Chicago markets. The data are most homogenous in Atlanta. The Dallas indices are less similar and remarkably the two Oakland indices are statistically independent. This suggests that quarterly returns in that market are measured with significant error or that the returns are highly unpredictable.

PAGE 78

TABLE 6-05 Regression of Quarterly Real Returns on Quarterly Real Returns Lagged (Full Sample Period) City Parameters Atlanta j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l Chicago j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l Dallas j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l San Francisco j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l Rr_j(t> = Bo + B, Rr_k,CtL> + u Period 1970:1-1986:1 1970:1-1986:1 1970:1-1986:1 1970:1-1986:2 Obs. B 0 B 1 R 2 Root MSE(t-stat) (t-stat)adjR 2 65 0.022 64 0.027 64 0.034 65 0.023 64 0.025 64 0.023 65 0.034 64 0.034 64 0.029 66 0.038 65 0.038 65 0.038 0.006 ( 1. 998) 0.016 (4.292) 0.017 (3.752) 0.006 (1.672) 0.007 (1.990) 0.012 (3.715) 0.014 (2.925) 0.016 (3.354) 0.017 (4.053) 0.026 (4.995) 0.019 (3.623) 0.005 ( 1. 033) 0.522 (6.601) -0.233 (-2.338) -0.314 (-2.088) .409 .400 .081 .066 .066 .051 0. 662 307 (5.288) .296 0.503 .181 (3.701) .168 0.030 .001 (0.273)-.015 0.220 .047 (1.770) .032 0.161 .027 (1.309) .011 -0.032 .002 (-0.305)-.015 -0.162 (-1.612) 0.158 (1.566) 0.767 (6.287) .039 .024 .038 .022 .386 .376 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. 68

PAGE 79

TABLE 6-06 Regression of After-Tax Quarterly Excess Returns on After-Tax Quarterly Excess Returns Lagged Re_j(t) = 8 0 + 8 1 Re_kCt L) + UCt> City Parameters Period Obs. B 0 B 1 R 2 Root MSE(t-stat) (t-stat)adjR 2 Atlanta j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l Chicago j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l Dallas j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l San Francisco j=A, k=B, L=0 j=A, k=B, L=l j=B, k=A, L=l 1970:1-1986:1 1970:1-1986:1 1970:1-1986:1 1970:1-1986:2 65 0.022 64 0.027 64 0.032 65 0.022 64 0.021 64 0.023 65 0.034 64 0.033 64 0.032 66 0.041 65 0.040 65 0.040 0.007 (2.323) 0.017 (4.522) 0.019 (4.176) 0.008 (2.625) 0.007 (2.376) 0.011 (3.073) 0.015 (3.198) 0.017 (3.508) 0.017 (3.603) 0.026 (4.656) 0.022 (3.877) 0.004 (0.734) 0.517 (6.085) -0.136 (-1.276) -0.219 (-1.494) 0.526 (4.721) 0.555 (5.121) 0.250 (2.053) 0.252 (2.182) 0.224 (1. 944) 0.088 (0.762) -0.033 (-0.341) 0.165 ( 1. 722) 0.844 (6.812) Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. .370 .360 .025 .010 .035 .019 .261 .250 .297 .286 .064 .049 .070 .055 .057 .042 .009 .000 .002 .000 .045 .030 .424 .415 69

PAGE 80

70 Regressing real quarterly returns using the full sample period leads to widely differing results in each of the cities. In Atlanta, real returns are negatively correlated to the previous quarter's return, while in Dallas we see reversed signs depending on the regression specification. Looking at quarterly after-tax excess returns, correlation patterns are much more consistent across the cities. While B 1 is not significantly different from zero in Atlanta, Dallas and Oakland, Chicago demonstrates significant patterns of a first-order autoregressive process. In Tables 6-07 and 6-08, autocorrelation functions of the real and after-tax excess quarterly return series are estimated at lags up to eight quarters and a chi-square test for no time pattern in returns is reported. These tables indicate that when the full sample is used, real quarterly returns for Chicago and quarterly after-tax excess returns for both Chicago and San Francisco exhibit some pattern in their return structure. However, real quarterly returns in Atlanta, Dallas, and San Francisco and quarterly after-tax excess returns in Atlanta and Dallas are found consistent with random "white-noise". Base Model Assumptions The return methodology described previously depends largely on four primary assumptions. It is assumed that (1)

PAGE 81

TABLE 6-07 Autocorrelation Function of Real Quarterly Returns CHI-SQ Series P1 P2 P3 P4 Ps p6 P7 Pa (12) PROB. PANEL A (1970:1-1986:1) Real Returns Atlanta -0.394 0.207 0.088 -0.045 0.086 0.041 -0.026 -0.114 17.01 0.149 Chicago 0.228 0.201 0.460 0.177 0.194 0.128 -0.100 0.153 33.70 0.001 Dallas -0.113 -0.047 0.165 0.095 0.067 0.188 -0.157 -0.112 11.92 0.452 San Fran. 0.216 0.171 0.198 0.139 0.153 0.018 0.040 0.038 12.09 0.438 PANEL B (1970:1-1979:3) Real Returns Atlanta -0.484 0.302 -0.003 0.045 0.142 -0.023 0.024 -0.096 16.50 0.169 Chicago 0.208 0.048 0.153 0.143 -0.074 -0.300 -0.248 0.072 17.32 0.138 Dallas -0.140 -0.069 0.151 0.059 0.031 0.212 -0.163 -0.165 8.40 0.753 San Fran. 0.046 -0.007 0.190 0 000 0.119 -0.071 -0 149 -0.053 6.17 0.907 PANEL C (1979:4-1986:1) Real Returns Atlanta 0.007 0.010 0.547 -0.114 0.083 0.286 -0.020 0.059 14.76 0.255 Chicago 0.003 -0.019 0.462 -0.067 0.038 0.270 -0.131 0.002 13.87 0.309 Dallas -0.112 0.003 -0.117 0.195 -0.190 -0.137 -0.061 0.109 7.54 0.820 San Fran. 0.204 0.105 -0.073 0.080 -0.169 -0.213 -0.060 -0.055 6.50 0.889 -.J ....

PAGE 82

TABLE 6-08 Autocorrelation Function of After-Tax Excess Quarterly Returns Series P1 P2 P3 P4 Ps p6 P1 Pa PANEL A (1970:1-1986:1) Exe. Returns Atlanta -0.284 0.156 0.104 0.069 0.148 -0.073 -0.011 -0.050 Chicago 0.452 0.367 0.389 0.527 0.305 0.187 0.212 0.297 Dallas 0.048 0.048 0.267 0.200 0.160 0.244 0.185 -0.028 San Fran. 0.308 0.252 0.343 0.309 0.313 0.072 0.117 0.164 PANEL B (1970:1-1979:3) Exe. Returns Atlanta -0.382 0.139 0.059 0.020 0.137 -0.077 0.015 -0.101 Chicago 0.111 0.032 0.015 0.323 -0.116 -0.320 -0.201 0.186 Dallas -0.075 -0.124 0.204 0.067 0.078 0.203 -0.110 -0.088 San Fran. 0.030 0.023 0.120 0.039 0.122 -0.110 -0.105 0.054 PANEL C (1979:4-1986:1) Exe. Returns Atlanta -0.064 -0.145 0.134 0.128 0.126 -0.273 -0.178 0.192 Chicago 0.150 -0.127 0.064 0.235 -0.020 -0.066 0.112 0.040 Dallas 0.079 0.186 -0.025 0.126 -0.228 -0.102 -0.263 0.020 San Fran. 0.208 -0.045 0.084 0.300 0.053 -0.323 -0.225 -0.005 CHI-SQ (12) PROB. 12.95 0.373 74.53 0.000 16.14 0.185 43.65 0.000 10.72 0.553 23.46 0.024 8.59 0.738 3.56 0.990 10.50 0.572 6.86 0.866 10.45 0.576 15.91 0.196 '--l t,,J

PAGE 83

73 the WRS price index accurately estimates movements in housing prices. The return model assumes, (2) that marginal income tax rates and property tax rates are constant, (3) inflation expectations were consistent with rational expectations during this period, and finally, (4) that the risk-premium associated with housing returns is not serially correlated. While the WRS index methodology attempts to control for constant-quality housing, it has its shortcomings. When used over short time spans, the index may restrict observations to a disproportionate number of high turnover properties, potentially reflecting forced sales or renovation projects. 8 This biases the index and if the proportion of high turnover properties is autocorrelated, tests of autocorrelation in returns may be biased. Despite these limitations, other price indices present other weaknesses. One-directional tests of efficiency can be conducted using the WRS index. If, for instance, autocorrelation is present in after-tax excess returns, it is not clear whether it is due to the model of housing returns or the actual nature of excess returns. However, if autocorrelation is rejected, the hypothesis of an efficient market cannot be rejected. 8 case and Shiller indicate that transactions were selected such that properties which changed condition or quality between the time of the first sale and the second (repeat sale) were excluded from the data set.

PAGE 84

74 The assumptions that marginal tax rates are constant, inflation accurately forecasted, and housing risk-premium constant are explicitly modeled in the following chapters and the robustness of the initial test findings, as reported in this chapter, examined.

PAGE 85

CHAPTER 7 MARGINAL TAX RATES Implied Bondholder Marginal Tax Rates In this chapter the base model assumption that marginal tax rates (MTRs) are constant is removed. It is hypothesized that home buyer MTRs are serially dependent and negatively correlated with the after-tax excess return series previously estimated in Chapter 6. This suggests that any autocorrelation found in these after-tax excess returns may result from assuming a constant marginal tax rate. Using yields on newly issued three-month prime-grade municipal securities and three-month U.S. Treasury securities, Figure 7-01 plots monthly estimates of the marginal bondholder's tax rate for the first quarter of 1970 through the second quarter of 1986. 1 This indicates that imputed bondholder marginal tax rates during this period 1 The marginal bondholder's tax rate, Tb' is estimated as rb=l-(M/TB), where M denotes the prime-grade municipal yield, and TB the treasury yield. At this rate, after-tax treasury yields equal tax-exempt yields. It is recognized that exist complications with assuming this simple relationship. For instance, the risk premia on municipal bonds is assumed constant, while it is quite evident that during recessions (1974-75 and 1981-82) the risk premia on municipal bonds increases reducing the spread. 75

PAGE 86

6 C,.) =ti 0... D.55 0.5 0.45 0 4 were not constant, but were in fact quite volatile. In addition, Panel A of Table 7-01 suggests that bondholder marginal tax rates exhibit significant first-order autoregressive, ARl, characteristics. Hence, they are not random, and would influence the patterns of any generated after-tax return series which neglected to take them into account. 2 76 Q35 0 3 0.25 01 B70 .1 B721 B74.I B76 .1 B78 .1 BOO.I B821 198 4.1 --4lm~e~ Marginal Tax Rate FJGURE 7-01 Implied Bondholder Marginal Tax Rates (1970:1 to 1986:2) \/ \ V h I 1 i I 1900 .1 2 Marginal tax rates estimated from national tax data as the percentage tax on each additional dollar of income are published for selected years in the U.S. Statistical Abstract. These data suggest marginal tax rate trends similar to that indicated by the imputed bondholder's marginal tax rate. Marginal income tax rates generally increased during the 1970s and decreased during the early 1980s.

PAGE 87

77 Unfortunately, implied bondholder marginal tax rates are not necessarily representative of the typical home buyer's marginal tax rate. They represent the marginal tax rate of the last (or marginal) bond investor for whom it is advantageous to invest in municipal securities rather than treasury bills. TABLE 7-01 PANEL A MTR Series USA_MT~ L=l PANEL B MTR Series ATL_MTRit L=4 CHI_MTRit L=4 DAL_MTRit L=4 SAN_MTRit L=4 First-Order Autocorrelation of Marginal Tax Rates Obs. Sample Period MSE 1970:1-1986:2 Sample Period 1970:1-1985:2 65 0.044 Obs. MSE 61 0.015 61 0.018 61 0.018 61 0.023 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.145 (3.685) 0.641 .411 (6.635) .402 B 0 B 1 R2 (t-stat) (t-stat)adjR2 0.106 0.537 .292 (2.338) (2.716) .280 0.076 0.661 .486 (2.130) (4.105) .477 0.064 0.738 .557 (1.806) (4.736) .549 0.054 0.812 .757 (1.907) (7.454) .753 Note: MTRit and MT~ denote the home owner marginal tax rates for each city and the implied U.S. bondholder marginal tax rate, respectively. L denotes number of lagged quarters. The coefficient standard errors in PANEL Bare adjusted as indicated by Hansen and Hodrick (1980).

PAGE 88

78 Home Buyer Marginal Tax Rates The marginal tax rate of the home buyer is dependent not only on statutory changes in the tax structure but in changes in the out-of-pocket' cost of housing. For instance, if the current real price of housing or real mortgage rates increase, some home buyers are priced out of the market, unable to afford the out-of-pocket' housing costs. When this occurs, a tax bracket 'shift' takes place. The average buyer of the median-priced home has a different marginal tax rate than the average buyer of the median priced home prior to price or mortgage increases. Average new buyers must have higher incomes in order to afford the increased out-of-pocket costs. consequently, the MTRs of home buyers are not captured by simply examining statutory changes in the tax structure. While it is impractical to model the entire distribution of MTRs for all home buyers, it is possible to estimate the average income of the buyer of the median priced home in each city. Using marginal tax rate data, which are available by income bracket, the MTR of this home buyer can be estimated for each quarter. Using the WRS housing price index, the median price of constant-quality housing is first estimated. Each city's respective WRS index is set to the median price of existing homes sold, as reported for that city in 1980 by the Office of Thrift Supervision (OTS, formerly the Federal Home Loan

PAGE 89

79 Bank Board, FHLBB). This yields a series prices for constant-quality 1980 median priced housing. Survey data reporting the national average loan-to value ratios are available from both the OTS and the Chicago Title Insurance Company. Using the average contract mortgage rate in each quarter, a series of average mortgage payments is constructed for each city. The average income TABLE 7-02 Sources of Data to Determine Home Buyer Marginal Tax Rates Series Dates Used Freq Source Yield on newly-issued 70: 01-86: 09 M three-month primegrade municipals Contract rate on 3070:01-86:09 M year fixed-rate conventional mortgages. Median priced single1980 family housing unit (1980). Marginal Federal 70:1-86:3 Income Tax (Married). Mortgage payment-to75:1-86:1 Income Ratio; Loan-to-Value Ratio. Loan-to-Value Ratio. 70:1-74:1 A Q A A Salomon Brothers, An Analytical Record of Yields and Spreads. Federal Reserve Bulletin. FHLBB, Savings and Home Financing Source Book. Ling, The Price of owner-Occupied Housing Services. Chicago Title Insurance Co., Chicago Guaranty. FHLBB, Savings and Home Financing Source Book. Note: M, Q, and A denote monthly, quarterly, and annual data series, respectively.

PAGE 90

80 of a representative home buyer of this representative housing is then determined from data indicating the average monthly payment as a percentage of income. 3 Finally, a series of median home buyer MTRs is estimated for each city using MTR data from Ling (1990). EMH Test Results: Excess Returns Adjusted for Time-Varying Marginal Tax Rates A series of home buyer marginal tax rates is estimated for each city using the methodology described in the previous section. These series are plotted in Figures 7-02 and 7-03, along with the previously estimated quarterly after-tax excess returns. These figures indicate that home buyer marginal tax rates were generally higher in the 1980s when after-tax excess returns, as previously estimated, declined. However, the inverted U shape, similar to that shown in Figure 7-01 remains prevalent. Panel B of Table 701 indicates that these series follow a first-order autoregressive (ARl) pattern, and quarterly after-tax excess returns to housing are shown to be negatively correlated with the constructed home buyer marginal tax rate series for each city in Panel C of Table 7-03. However, the correlation factors are relatively low. 3 Mortgage payment-to-income ratios were unavailable for the period 1970:1 to 1974:4. Interest rates during this period were relatively stable; hence an assumption that payment-to-income ratios were constant during this period was used.

PAGE 91

Regressing after-tax excess returns constructed using the constructed series of time-varying marginal tax rates shows little difference from previous tests of autocorrelation reported in Chapter 6. 81 Tables 7-04 and 7-05 indicate that the earlier assumption, that marginal tax rates are constant, does not materially affect the results of autocorrelation tests. While this methodology lacks precision, it is preferable to assuming a strictly constant home buyer marginal tax rate and is applied in further tests of efficiency conducted in Chapters 8 and 9. In Table 7-04, after-tax annual excess returns are estimated as the product of four quarterly after-tax excess returns. This methodology is identical to simply using annual returns. 4 In addition, Ling (1990) reports the tenure choice tax rate, using these data, results are reported in Tables B-02 and B-03 of Appendix B. While there is evidence that applying the tenure choice tax rate to the model reduces autocorrelation tendencies in the return series, Ling (1990) indicates that for the quantity of housing demanded decision (how much housing to invest in) the marginal tax rate is the applicable rate. Therefore, it is the rate applied to the models subsequently tested in Chapters 8 and 9. 4 Regression results applying this methodology to the base model appear in Appendix B, Table B-01. The results reported in Table B-01 are identical to those reported earlier in Table 6-04, indicating the test methodologies are equivalent.

PAGE 92

t) C) "' c t) (.) Ii a. 0 4 ,-----------------------------, ( Median Priced Home Buyer: Atlanta ) 0 3 ---r--------------------------------~ 0 1 -t-,.-,-.-,-.-,-r,r--,-r--,--,-,-,--,-,-,-,-,--,-,-..---r-T--r-T-,-,---.-,--.-,-.--,-.--,-,-.-,-.-,-r--,-r--,-r--,-,--,,--,-,-,--,-,--,-,--,-,-~ 1970 1 1972.1 1974.1 MTR ( owner ) 1976 1 1978.1 Quarter fac Return ( A I 1980.1 1982 1 1984 1 0 Exc.Return ( B ) FIGURE 7-02 Homeowner Marginal Tax Rates and Quarterly Excess Returns: Atlanta 1986 1 0 4 -r-----------------------------~ ( Median Priced Home Buyer: Chicago ) 0.3 -+------------------------------------< 0 1 -+----------------------------------, 0.1 --h-,-~r--,-r--,-r--,-,--,-,-,-,-,--,-,--,-,-~~~~--.-,--.-,-.--,-.--,-,-~.-,-r--,-r--,-,--,-,--,-,-,--,-,--,-,---r-T--r-T-,-,---.-,-1 1970 1 1972.1 1974 1 MTR ( owner l + 1976.1 1978 1 Quarter Exc Return ( A I 1980 1 1982 1 1984 1 0 Exc.Return ( B I FIGURE 7-03 Homeowner Marginal Tax Rates and Quarterly Excess Returns: Chicago 1986.1 82

PAGE 93

I) 0) 111 c I) u ... I) Q. 0 4 -,----------------------------------( Median Priced Home Buyer: Dallas ) 0 3 +--------------------------------_J -0 1 1970 1 1972.1 1974 1 1976.1 1978.1 1980 1 1982 1 1984 1 1986 1 Quarter MTR l owner ) + Eitc Return l A ) Eitc Return ( B ) FIGURE 7-04 Homeowner Marginal Tax Rates and Quarterly Excess Returns: Dallas 0 4 ( Median Priced Home Buyer: Oakland ) 0.3 o.2 1aEmeooai~~~-----l7Ft-------~ -0.1 -+-,~~~~~~~~----~---~~~,...,...S;L.-~~~~~--------1 1970 1 1972.1 1974 1 MTR ( owner ) 1976 1 1978 1 Quarter Eitc Return ( A ) 1980.1 1982 1 1984 1 Exc Return ( B ) FIGURE 7-05 Homeowner Marginal Tax Rates and Quarterly Excess Returns: San Francisco 1986 1 83

PAGE 94

84 TABLE 7-03 Correlation Matrices of Quarterly Home Owner Excess Returns, Constructed Home Owner Marginal Tax Rates and Implied Bondholder Marginal Tax Rates PANEL A ATL_Re CHI_Re DAL_Re SAN_Re ATL_Re 1.0000 CHI_Re 0.0675 1.0000 DAL_Re 0.1480 0.2939 1.0000 SAN_Re 0.3313 0.3225 0.1868 1.000 PANEL B ATL_MT~ CHI_MT~ DAL_MT~ SAN_MT~ ATL_MT~ 1.0000 CHI_MT~ 0.7277 1. 0000 DAL_MT~ 0.1388 0.5612 1. 0000 SAN_MT~ 0.3573 0.7691 0.8700 1.0000 USA_MT~ 0.5033 0.4702 0.0233 0.1796 PANEL C ATL R e CHI_Re DAL_Re SAN_Re ATL_MT~ -0.0728 CHI_MT~ -0.1538 DAL_MT~ -0.1767 SAN_MT~ -0.1914 USA_MT~ 0.1464 0.1824 0.2866 0.2022 Note: Excess returns, denoted Re, were computed using a constant 30 percent marginal tax rate for each city. Marginal home owner tax rates are denoted MT for each city and the implied bondholder marginal tax rate is USA_MT~.

PAGE 95

TABLE 7-04 Regression of After-Tax 4-Quarter Excess Returns on After-Tax 4-Quarter Excess Returns Lagged Adjusted for Nonconstant Home Owner Marginal Tax Rates Re,m_j(t) = Bo + B, Re,m_k(t-L) + u City Parameters Atlanta j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Chicago j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Dallas j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 San Francisco j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Sample Period Obs. 1970:1-1985:2 1970:1-1985:2 1970:1-1985:2 1970:1-1985:3 MSE 62 0.047 58 0.049 58 0.047 62 0.026 58 0.053 58 0.054 62 0.048 58 0.064 58 0.061 63 0.069 59 0.077 59 0.098 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.010 (0.906) 0.033 (1. 895) 0.034 (1.929) 0.004 (0.406) 0.015 (0.776) 0.017 (0.885) 0.023 ( 1. 510) 0.042 (1. 735) 0.029 (1. 315) 0.040 (1.543) 0.044 (1. 642) 0.053 (1.507) 0.830 (6.401) 0.383 (1.866) 0.348 (1.676) 0.938 (9.843) 0.706 (3.503) 0.671 (3.344) 0.814 (5.908) 0.540 (2. 540) 0.522 (2.956) 0.614 (3.873) 0.559 (3.541) 0.515 (2.208) .705 .700 .167 .152 .126 .111 .867 .865 .487 .477 .457 .447 .604 .597 .272 .259 .294 .282 .475 .466 .395 .384 .209 .195 85 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MA3 process as described by Hansen and Hodrick (1980).

PAGE 96

TABLE 7-05 Regression of After-Tax Quarterly Excess Returns on After-Tax Quarterly Excess Returns Lagged Adjusted for Nonconstant Home Owner Marginal Tax Rates Re,m_j(t) = Bo + B, Re,m_k(t L) + u City Sample Period Obs. B 0 B 1 R2 Parameters MSE (t-stat) (t-stat)adjR 2 Atlanta 1970:1-1986:1 j=A, k=B, L=0 65 0.006 0.521 .373 0.022 (2.112) (6.126) .363 j=A, k=B, L=l 64 0.016 -0.129 .023 0.028 (4.122) (-1.207) .007 j=B, k=A, L=l 64 0.018 -0.210 .032 0.033 (3.848) (-1.434) .017 Chicago 1970:1-1986:1 j=A, k=B, L=0 65 0.008 0.535 .269 0.021 (2.423) (4.820) .258 j=A, k=B, L=l 64 0.006 0.564 .306 0.021 (2.176) (5.231) .295 j=B, k=A, L=l 64 0.009 0.267 .073 0.024 (2.757) (2.207) .058 Dallas 1970:1-1986:1 j=A, k=B, L=0 65 0.014 0.253 .071 0.034 (3.034) (2.191) .056 j=A, k=B, L=l 64 0.016 0.224 .058 0.033 (3.353) (1. 947) .042 j=B, k=A, L=l 64 0.016 0.091 .010 0.032 (3.396) (0.793)-.006 San Francisco 1970:1-1986:2 j=A, k=B, L=0 66 0.026 -0.037 .002 0.041 (4.601) (-0.380)-.013 j=A, k=B, L=l 65 0.021 0.163 .044 0.040 (3.836) ( 1. 695) .028 j=B, k=A, L=l 65 0.004 0.842 .422 0.040 (0.734) (6.787) .413 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. 86

PAGE 97

CHAPTER 8 INFLATION EXPECTATIONS Expected and Unexpected Inflation Tests of efficiency using returns generated from the models described in Chapters 6 and 7 assume that unexpected inflation (differences between actual and expected inflation) is stochastic with mean zero. Under these conditions, market participants are said to hold "rational" expectations with respect to inflation. In defining after-tax excess returns as the nominal return to housing in excess of the after-tax risk-free interest rate, Re= Rn (1-r)TB, inflation expectations (embedded in the interest rate structure) are subtracted from the nominal return to housing. 1 If inflation expectations are rational during the sample period and returns to housing are consistently correlated with unexpected inflation, then returns due to unexpected inflation should provide be random and provide no new information in explaining ex post after-tax excess returns. Under this condition, the assumption, 1 consistent with the Fisher hypothesis, nominal risk free interest rates are set equivalent to the required real risk-free interest rate plus the expected rate of inflation. 87

PAGE 98

E(Ruit) = 0, 88 ( 8-01) holds and Equation (4-04), the formal market efficiency model rewritten below for convenience, is unaffected, (8-02) However, if unexpected inflation is systematically biased, any subsequent tests of the EMH will be affected. While the assumption that Equation (8-01) holds is consistent with the rational expectations hypothesis and tests of market efficiency, the position that market expectations of inflation are rational is replete with controversy. When considering the volatile inflationary period of the 1970s and early 1980s, substantial evidence exists suggesting that inflation expectations during this period are were not rational. This evidence suggests that both the rapid increases and decreases in the rates of inflation were systematically underestimated. This has led some to reject the entire notion of the rational expectations hypothesis. However, this position appears extreme and myopic. It is quite reasonable to expect that rational markets may not meet full rational expectations conditions (in a statistical sense) when examined over relatively short periods of time. 2 The variables are defined earlier in Equation (4-04).

PAGE 99

89 In addition, single factors that highly influence a market, such as inflation expectations relative to housing returns or bond returns, may not appear rational for short time intervals. This, however, should not necessarily lead to a full rejection of the EMH for these markets. Studies need to include sufficient time periods to fully reject the efficient markets hypothesis. It is possible, for example, to identify short periods, (days and weeks) where ex post price changes in the S&P 500 and ex post returns to bonds are serially dependent. However, these dependencies are not consistent when longer time spans are considered and should not lead to a rejection of the entire EMH hypothesis. Unfortunately, adequate housing market data for long periods do not currently exist. However, it is important to look at the sensitivity of the previous test results to the assumption that inflation expectations over this period were rational. This chapter first presents evidence suggesting that inflation surprises over the period studied were autocorrelated. Using a model to estimate unexpected inflation, after-tax excess returns to housing are adjusted for any returns due to unexpected inflation. Finally, the random-walk characteristics of the abnormal returns to housing, residuals from this model, are examined and the EMH test results reported.

PAGE 100

90 Past Inflation Expectation Models and Tests of Rationality Since inflation expectations are not observable, alternative methods have been used to estimate both expected and unexpected inflation. Although there has been much work done evaluating these methodologies, there is no general consensus as to the 'best' or 'dominant' method. In estimating expected inflation, one of three general methodologies is typically used. The first extracts inflation expectations from the market risk-free interest rate. For example, an interest rate model developed by Fama (1975) is used to predict the rate of inflation one-month ahead using the CPI. This model requires the assumption that real risk-free returns are constant. Later, Fama and Gibbons (1982, 1984) modify the interest rate model to account for nonconstant real interest rates. While they suggest this model dominates other methodologies, this finding is not consistent with work by others, such as Hafer and Hein (1990). Univariate time-series forecasting models represent the second method. A number of alternative model specifications have been suggested. Adaptations of autoregressive distributed lag (DL) models or autoregressive integrated moving average (ARIMA) models are most commonly used. All these models use some form of past inflation information, available to market participants, to estimate future inflation. The shortcoming of the DL model is that it will

PAGE 101

91 lead to systematic forecast errors during periods of rising or falling inflation. However, the ARIMA model yields residuals which are consistent with rational expectations, making that class of models highly attractive and widely used. Consensus survey data of market participants or experts, represents the third method. In this case survey data on expected inflation is used as a proxy for market inflation expectations. While this method is also appealing, widely used, and considered by many as the most viable method, the data are often found to reject the rational expectations hypothesis. This has led researchers to question both the survey procedures and whether survey respondents typify the sentiment of the aggregate market. A number of studies have examined observable inflation expectations using the Livingston survey data. 3 Work by Pesando (1975) found that the survey data reject the rational expectations hypothesis. 4 Carlson (1977), 3 These data come from a survey of professional economists conducted by Joseph Livingston. Taken semi annually (June and December) since 1947, the survey asks economists to estimate the consumer price index (CPI) that will exist six and twelve months from the survey date, as well as forecasts of other key economic variables. The forecasted levels are then typically converted to sixand twelve-month inflation forecasts by using the actual CPI level at the time the survey was taken. In addition, these data have been widely used to empirically test the effect of anticipated inflation on interest rates. 4 Pesando could not reject a weaker hypothesis of rationality using the six-month-ahead survey data and indicates that they are more consistent with the notion of

PAGE 102

92 examining the same data, indicates that respondents are likely to know only the April CPI figures when surveyed in June and the October CPI when surveyed in December. He suggests that the data reflect eightand fourteen-month forecasts, rather than sixand twelve-month. Using these modified assumptions, Carlson rejects the rational expectations hypothesis. Mullineax (1978), later suggests that, in fact, Carlson's revised data are consistent with rationality under alternative tests. While the above tests suggest that the data from 1959 to 1969 may be consistent with rationality, none of the tests includes data from the period of the 1970s, characterized by rapid inflation. Using data from 1959 to 1975, Pearce (1979) finds that the Livingston survey data are not rational. In addition, Hafer and Hein (1985) examine a survey conducted jointly by the American Statistical Association and the National Bureau of Economic Research (ASA-NBER). 5 They find that using data from the first quarter of 1970 through the first quarter of 1984 the ASA-NBER data display systematic bias in their forecast errors. The Livingston data, shown in Figure 8-01, appear to follow very similar trends during this period. rationality than the twelve-month-ahead data. 5 The ASA-NBER surveys economists quarterly, asking for expected levels of the GNP oneto sixquarters in the future. From this data an expected implicit GNP deflator is then derived which some studies have employed as a proxy for expected inflation.

PAGE 103

0.05 == :, I:: I:: 0 ::, I> ::; I> < 5 --ODS 93 These studies suggest that survey respondents were unable to forecast the historically high rates of inflation experienced during the 1970s and early 1980s. An exception to this result appears in a study by Pearce (1987). He examined a monthly survey of money market participants conducted by Money Market Services International and found that these data, including data from the 1970s and 1980s, are consistent with rational expectations. However, this survey typically takes place within two weeks of the actual CPI announcement date. This, in addition to the findings of Pesando (1977), suggests that the accuracy of survey respondents inflation forecasts deteriorate as the term of the forecast lengthens. -eLiv. S11vey FIGURE 8-01 --+ASA~BER Strvey --6ARMA lqtr.fwd Unexpected Annual Inflation (Livingston, ASA-NBER, ARIMA)

PAGE 104

94 In addition, changes in money and capital market expectations appear to lag movements of inflation rates during this period. Interest rates lag both increases and decreases in expected inflation rates. 6 This results in a autocorrelated series of real interest rates, typified by periods of prolonged negative real rates as in Figure 8-02. 0.1 ,------------------------e-Real !Yr. T-&l -erUnexpected Inflation FIGURE 8-02 sm S76.1 S78.I 9Kl.l !1821 Real 1-Year Treasury Bill Yields and Unexpected Annual Inflation !184.1 MU 6Typically, real interest rates are low when inflation expectations are high; and, conversely, real interest rates are relatively high when expected inflation rates are declining. See Taylor (1982) and Dokko and Edelstein (1987). Mundell (1963) and Tobin (1965) hypothesize that this results from people shifting to interest-bearing assets during times of high interest and expected inflation rates, reducing the equilibrium real rate of return on such assets.

PAGE 105

95 Finally, the study period is marked by changes in monetary policy, which have been characterized as a 'regime shift'. Blanchard (1984) indicates that the impact (or event) of changes in the monetary policy in October 1979 were not immediately believed. "Direct, informal evidence suggests that it took more than a year to fully change the beliefs of financial markets, perhaps more than that to change those of labor markets." (Blanchard, p212.) Regressing after-tax excess returns prior to and subsequent to the October 1979 policy shift indicates that the hypothesis of an efficient market can no longer be rejected using the base model data. Results of this regression appear in Table 8-01. While there exists little consensus to the 'dominant' methodology for estimating expected inflation, both the survey data and the interest rate structure suggest that inflation shocks during the 1970s and early 1980s were not fully anticipated. Systematic forecast errors appear evident during this period. Throop (1988) finds that adaptive expectation models, not rational expectation, best explain both investment and labor market inflationary expectations during this period. Given this possibility, inflation expectations are modeled in the following sections, and after-tax returns to housing are adjusted for autocorrelated inflation shocks.

PAGE 106

TABLE 8-01 Regression of Annual Excess Returns on Annual Excess Returns Lagged (Prior and Subsequent to Monetary Policy Shift) Re_jCt> = Bo + B, Re_kCtL> + u City Obs. B 0 8 1 R2 Obs. B 0 B 1 R2 Parameters MSE (t-stat) (t-stat)adjR 2 MSE (t-stat) (t-stat)adjR 2 Atlanta PANEL1970:1 -1979:3 PANEL 1979:4 -1985:2 j=A, k=B, L=0 [A] 39 0.020 0.772 .658 [ B] 23 0.007 0.763 .356 0.032 (1.187) (4.638) .649 0.024 (0.426) (1.872) .325 j=A, k=B, L=4 35 0.058 0.296 .091 19 0.023 0.266 .042 0.052 (2.124) (1.190) .063 0.029 (1.086) (0.486)-.012 j=B, k=A, L=4 35 0.057 0.302 .082 19 0.033 0.002 .ooo 0.056 (2.017) (1.112) .055 0.025 (2.226) (0.007)-.056 Chicago j=A, k=B, L=0 39 -0.007 1.049 .800 23 0.003 0.610 .360 0.024 (-0.446) (6.697) .795 0.029 (0.310) ( 1. 889) .330 j=A, k=B, L=4 35 0.067 0.338 .051 19 0.007 0.245 .057 0.022 ( 1. 078) (0.648) .022 0.034 (0.504) (0.576) .005 j=B, k=A, L=4 35 0.097 0.096 .008 19 -0.001 0.001 .000 0.047 (1.927) (0.231)-.022 0.037 (-0.045) (0.001)-.056 Dallas j=A, k=B, L=0 39 0.041 0.712 .472 23 0.018 0.763 .366 0.055 (1.332) (3.161) .458 0.032 (0.969) (1.913) .336 j=A, k=B, L=4 35 0.084 0.397 .172 19 0.035 -0.000 .000 0.067 (2.207) (1.545) .147 0.028 (1.962) (-0.000)-.056 j=B, k=A, L=4 35 0.070 0.390 .180 19 0.024 0.041 .005 0.067 (2.102) (1.796) .155 0.024 (1.448) (0.164)-.050 Oakland j=A, k=B, L=0 39 0.117 0.273 .109 24 0.012 0.604 .438 0.073 (2.626) (1.173) .085 0.045 (0.647) (2.279) .413 j=A, k=B, L=4 35 0.101 0.410 .149 20 0.014 0.255 .282 0.073 ( 1. 899) (1.417) .123 0.051 (0.702) ( 1. 615) .249 j=B, k=A, L=4 35 0.156 0.127 .010 20 0.016 0.172 .072 0.094 (2.371) (0.335)-.021 0.064 (0.591) (0.591) .030 A and B denote half sample indices for each city and L denotes number of lagged quarters. Coefficient standard errors adjusted for MA3 process (Hansen and Hodrick, 1980) \0 O'I

PAGE 107

97 Inflation Expectations Model This study models inflation expectations using one step-ahead forecasts from an ARIMA (0,2,1) model and employing a similar methodology to that used by Pearce (1979). 7 Following Schwert (1987) the autocorrelation function of the second differences of the natural log of quarterly CPI-U, net-of-shelter, index is estimated for the period 1960:1 to 1989:1. The first twelve autocorrelations of the second difference of the variable, which represent changes in the inflation rate, are close to zero with the exception of the first. This is typical of a first-order moving average process, ( 1L) xt = ( 1-a 1 L) at (8-01) were Xt is the inflation rate; at, the independent errors; 8 the moving average parameter; and L the lag operator. 8 The ARIMA model specified by Equation (8-01), using the time series techniques of Box and Jenkins (1970), is estimated for the quarterly CPI-U series from, 1962:3 to 1972:2. 9 Using the coefficient estimates of the period selected, the 7 The autoregressive integrated moving average model is usually denoted by the notation ARIMA(p,d,q), where pis the degree of the autoregressive process, dis the degree of differencing, and q is the degree of the moving-average process. 8 Box and Jenkins (1970) describe this process in detail. 9 The use of future inflation information to estimate the model coefficients is consistent with the notion that market participants anticipate future events.

PAGE 108

98 CPI-U price level in quarter one of 1970 is forecasted using a 'forecast back' procedure of the SAS step-wise model. The model is reestimated for the period 1962:4 to 1972:3 and then used to forecast the CPI-U for 1970:2. This procedure is repeated until the last forecast date, 1986:3, is reached. 10 The estimates of the CPI-U, net-of-shelter, index are used to derive estimates of the expected quarterly inflation. Unexpected quarterly inflation (ARIMA lqtr.fwd.) is then defined as the difference between actual inflation and the estimated expected inflation. Two methods were considered to convert the unexpected quarterly inflation rate to unexpected annual inflation rates on a quarterly basis. The first method used the following formula, (8-02) where Uia is the unexpected annual inf lat ion (ARIMA 4qtrs. cum.) in period t and UiqCt> is the unexpected quarterly inflation in period t from the ARIMA model. Thus, the annual inflation is simply the product of the unexpected inflation in each subsequent quarter. However, this should underestimate annual inflation, since inflation expectations 10 The model residuals are evaluated by examining the statistics for each quarter forecast. Residuals for all periods were determined to be "white-noise".

PAGE 109

99 are generally more accurate for shorter terms. Considering a second methodology, the quarterly inflation series already includes information on future inflation in the estimation of the coefficients, thus quarterly unexpected inflation rates (ARIMA annualized) were annualized by multiplying each value by four. The unexpected inflation values are plotted in Figure 8-03. Comparing the annualized estimate with the Livingston survey data on a semiannual basis, Table 8-02 indicates that the ARIMA (0,2,1) model yields a lower value of mean absolute unexpected inflation. While the data do not follow a random-walk, (see Figure 8-01) they have characteristics nearly identical to the Livingson survey data (the correlation coefficient is equal to .81). In addition, quarterly inflation expectations using the model, plotted in Figure 8-01, appear consistent with rational expectations. TABLE 8-02 Comparative Statistics: Livingston Survey vs ARIMA-1 Quarter Forward Model of Unexpected Inflation Iu,LIV Ct> = Bo + B, Iu,n Ct> + uCt> PANEL A Period 1970:2 1985:4 Obs. BO Bl R2 Root MSE(t-stat) (t-stat)adjR2 32 0.016 -0.003 0.913 -1. 222 7. 221 .635 .622 Absolute Unexpected Inflation PANEL B Livingston Survey ARIMA-1 Qtr.Frwd. Mean Std.Dev. 0.0216 0.0189 0.0160 0.0132

PAGE 110

100 '370.1 1972.1 '374 .1 '376.1 '378.1 '380 .1 '382 1 1984 .1 -aARMA lqtr.f wd '370 .1 !!72.1 'S74.1 'S76.1 '378.1 '380 .1 '382 1 '384 .1 --4ARMA 1-qtr.fwd ( 4 qhs. cllll~ative ) '370 .1 1972.1 '374.1 '376 .1 '378.1 '380 .1 '382 1 1984 .1 -trARMA ~qtr.fwd ( annuar1Zed l FIGURE 8-03 Unexpected Inflation ARIMA Model: Quarterly, 4-Quarter, Annual

PAGE 111

101 Unexpected Inflation and Housing Returns Studies by Fama and Schwert (1977) and Hartzell, Hekman, and Miles (1987) indicate that both expected and unexpected inflation are highly correlated with housing returns (estimated coefficients are close to 1.0). If unexpected inflation is autocorrelated, it is necessary to adjust ex post housing returns due to unexpected inflation when evaluating ex ante after-tax excess returns. To remove the component of housing returns associated with unexpected inflation, after-tax excess returns, Rem' are regressed on I unexpected inflation using the linear model Rem Ct> = Bo + B, UI + e. I (8-03) The intercept, B 0 is interpreted as the expected after-tax excess returns uncorrelated with unexpected inflation, while the coefficient, B 1 reflects unexpected after-tax excess returns associated with unexpected inflation. The model residual, e' is then interpreted as the abnormal return. It is anticipated that, consistent with past work, unexpected inflation is positively correlated with returns to housing. If inflation is systematically underestimated or overestimated by the market, treasury bill yields, as well as nominal housing returns, are expected to be affected. If inflation rates were one percent higher than expected this would result in, given previous findings, ex post nominal returns to housing being overstated by one

PAGE 112

102 percent. In addition, after-tax treasury bill yields may be understated by about three-quarters percent, assuming a marginal tax rate of about twenty-five percent. Hence, the estimated B 1 coefficient in Equation (8-01) is anticipated to be approximately 1.75. The results of regressing after-tax excess returns for each city on unexpected inflation are listed in Tables 8-03, 8-05, 8-07, and 8-09. 11 These results suggest that after tax excess returns, as modeled, are significantly correlated with unexpected inflation, which is serially dependent. In addition, the estimated B 0 and B 1 coefficients are close to their anticipated values. EMH Test Results: Residual Returns Adjusted for Unexpected Inflation Further tests of first-order autocorrelation are then conducted on the abnormal return series (the series of model residuals for Eq. 8-03) using a linear model such that, Resn i = B 0 + B 1 Resn k + e (8-04) where Resn_Ht> is the abnormal return in period t constructed from sample j, Resn_k is the abnormal return in period t 11 The results of a strong positive relationship between unexpected inflation and excess return are quite robust. Unexpected inflation was modeled in a number of different ways on both an annual and quarterly basis. The results of these regressions are found in Tables B-04, B-06, B-08, and B-10 in Appendix B. In every case, results indicate that housing returns provide a significant hedge to unexpected inflation.

PAGE 113

103 constructed from sample k lagged L periods; and e an error term. This model represents a minor modification of the previous first-order tests of autocorrelation. The results of the regressions for each city are indicated in Tables 8-04, 8-06, 8-08, and 8-10. These results indicate that after the effect of unexpected inflation is removed from the after-tax excess return series, the hypothesis B 1 = 0 is rejected at the 5 percent confidence level consistently in only one city, Chicago. Both the B 0 and the B 1 coefficients are close to zero for Atlanta, suggesting little possibility for investors to identify trading rules which yield abnormal returns. In Dallas and in San Francisco findings suggest that abnormal returns exhibit some inertia. However, these series are not significantly different from zero at a five percent confidence level, and little confidence is placed in the San Francisco series because of the previously demonstrated measurement error. 12 While there exist autocorrelated returns in Chicago, they are substantially less significant than those previously reported by Case and Shiller (1989). 12 In addition, Proposition 13 was enacted in California during the sample period. This resulted in a substantial decrease in property taxes and a two percent ceiling was placed on annual property reassessments.

PAGE 114

104 TABLE 8-03 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation: Atlanta City Parameters Atlanta j=A, n=l j=B, n=l j=T, n=l j=T residuals: Sample Period Obs. MSE 1970:1-1986:1 65 0.026 65 0.031 65 0.025 DW=2.925 Q(24)=27.12 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.015 1.822 .142 (4.685) (3.226) .128 0.016 1.839 .105 (4.067) (2.717) .091 0.015 1.827 .153 (4.943) (3.371) .139 Note: A and B denote half sample indices and T denotes a full sample index while L indicates the number of lagged quarters. ARIMA 1-qtr. forward (annualized) model of inflation indicated by n=l. TABLE 8-04 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): Atlanta Rres,n_j,(t) = Bo + B, Rres,n k,(t L) + U(t) City Parameters Sample Period Obs. MSE Atlanta 1970:1-1985:2 j=A;k=B;n=l;L=0 j=A;k=B;n=l;L=4 j=B;k=A;n=l;L=4 62 0.025 58 0.028 58 0.034 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.001 (0.153) -0.005 (-0.820) -0.007 (-1.033) 0.612 .461 (4.526) .452 0.122 .026 (1.010) .008 0.000 .000 (0.002)-.018 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MAJ process as described by Hansen and Hodrick (1980).

PAGE 115

105 TABLE 8-05 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation: Chicago City Parameters Chicago j=A, n=l j=B, n=l j=T, n=l Sample Period Obs. 1970:1-1986:1 MSE 65 0.024 65 0.023 65 0.020 j=T residuals: DW=l.309, Q(24)=51.65 Bo B1 R2 (t-stat) (t-stat)adjR 2 0.016 (5.277) 0.015 (5.180) 0.015 (5.884) 1.484 .007 (2.838)-.009 1.647 103 (3.309) .089 1.403 .144 (3.283) .131 Note: A and B denote half sample indices and T denotes a full sample index while L indicates the number of lagged quarters. ARIMA 1-qtr forward (annualized) model of inflation indicated by n=l. TABLE 8-06 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): Chicago R =B+BR +u res,n_j, Ct) 0 1 res,n_k,Ct L) Ct) City Parameters Sample Period Obs. Chicago 1970:1-1985:2 j=A;k=B;n=4;L=0 j=A;k=B;n=4;L=4 j=B;k=A;n=4;L=4 MSE 62 0.025 58 0.047 58 0.049 Bo B1 R2 (t-stat) (t-stat)adjR 2 -0.004 (-0.697) -0.006 (-0.469) 0.001 (0.066) 0.897 (8.062) 0.593 (2.911) 0.549 (2.567) .843 .841 .359 .348 .304 .292 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MAJ process as described by Hansen and Hodrick (1980).

PAGE 116

106 TABLE 8-07 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation: Dallas City Parameters Dallas j=A, n=l j=B, n=l j=T, n=l Sample Period Obs. 1970:1-1986:1 MSE 65 0.033 65 0.035 65 0.025 j=T residuals: DW=2.878 Q(24)=25.63 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.021 (5.094) 0.021 (4.738) 0.017 (5.341) 2.097 (2.937) 1.754 (2.282) 1.622 (3. 052) .120 .106 .076 .062 .127 .113 Note: A and B denote half sample indices and T denotes a full sample index while L indicates the number of lagged quarters. ARIMA 1-qtr. forward (annualized) model of inflation indicated by n=l. TABLE 8-08 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): Dallas Rres,n_j,(t) = Bo + B, Rres,n_k,(t L) + U(t) City Parameters Sample Period Obs. MSE Dallas 1970:1-1985:2 j=A;k=B;n=l;L=0 j=A;k=B;n=l;L=4 j=B;k=A;n=l;L=4 62 0.045 58 0.054 58 0.053 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.007 (0.757) 0.005 (0.376) -0.008 (-0.605) 0.730 (4.522) 0.410 (1.949) 0.260 (1.409) .458 .449 .159 .144 .085 .068 Note: Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MAJ process as described by Hansen and Hodrick (1980). See notes above.

PAGE 117

107 TABLE 8-09 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Unexpected Inflation: San Francisco City Parameters San Francisco j=A, n=l j=B, n=l Sample Period Obs. 1970:1-1986:2 MSE 66 0.040 66 0.050 j=T, n=l 66 0.030 j=T residuals: DW=l.500 Q(24)=46.32 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.025 (5.059) 0.027 (4.306) 0.025 (6.841) 0.825 .015 (0.969)-.001 2.061 .055 (1.933) .040 1.290 .060 (2.035) .046 Note: A and B denote half sample indices and T denotes a full sample index while L indicates the number of lagged quarters. ARIMA 1-qtr. forward model of inflation indicated by n=l. TABLE 8-10 Regression of After-Tax 4-Quarter Residual Returns After-Tax 4-Quarter Residual Returns Lagged (Adjusted for Inflation Shocks): San Francisco Rres,n_j,(t) = Bo + B, Rres,n_k,(t) + U(t) City Parameters Sample Period Obs. San Francisco 1970:1-1985:3 j=A;k=B;n=l;L=0 j=A;k=B;n=l;L=4 j=B;k=A;n=l;L=4 MSE 63 0.065 59 0.073 59 0.086 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.002 (0.145) 0.001 (0.072) 0.001 (0.037) 0.598 (3.510) 0.513 (2.882) 0.422 (1.831) .410 .401 .301 .290 .155 .141 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MAJ process as described by Hansen and Hodrick (1980).

PAGE 118

108 In addition, these results suggest that the degree of efficiency in the single-family housing market is related to city size. It is suspected that search costs are positively correlated to city size, limiting the ability for market participants to correct pricing errors. Finally, the Chicago market was less active over the period studied. Atlanta, Dallas, and San Francisco all experienced substantial growth during this time period. These findings preliminarily suggest that active markets are relatively more efficient than inactive markets. This suggests abnormal returns may be easier to identify in slower markets periods; however, this issue requires more research.

PAGE 119

CHAPTER 9 TIME-VARYING RISK PREMIA Evidence of Time-Varying Volatility Given the "momentum" in the after-tax excess returns adjusted for unexpected inflation, this chapter explores the possible intertemporal relationship between expected risk and expected housing returns. The chapter concludes by examining a trading strategy to determine if the resultant after-tax excess return, the study's adjusted return structure, is exploitable. Returns to real estate, including returns to housing, have often been described as dominating other asset classes both before and after adjusting returns for risk. However, these returns are often generated from appraisal based indices, which are argued to smooth the volatility of the return series. 1 This data set provides possibly a unique opportunity to analyze the relationship between expected risk, as proxied by expected return volatility, and expected returns to housing. In particular, expected after-tax excess returns to housing (the expected risk premia) are examined and 1 see Geltner (1989) and Giliberto (1988). 109

PAGE 120

110 hypothesized to be positively correlated to expected housing return volatility. 2 In addition, if housing expected return volatility is serially dependent and correlated with the expected risk premium to housing, it will bias subsequent tests of autocorrelation of the after-tax excess return series. This condition is commonly referred to as autoregressive conditional heteroskedasticity (ARCH). To date, no real estate studies have been published which address this issue. Casual observation of the ex post quarterly after-tax excess returns (Figures 7-02, 7-03, 7-04, and 7-05) suggest that volatility may be time-varying and correlated with returns. It appears that ex post return volatility in each market is greater in the periods prior to 1979. This would relate to the periods marked by high ex post after-tax excess annual returns shown previously in Figure 6-01. Using the model described by Equation 8-03, a series of ex ante after-tax excess returns can be constructed, excluding the return associated with unexpected inflation. A moving average variance using subsequent four periods is calculated from the previously constructed ex ante after-tax excess return series. This series of variances is defined 2 French, Schwert, and Stambaugh (1987) report that the expected stock market risk premium is positively correlated to expected return volatility. In addition, they find that unexpected return volatility is negatively correlated with the expected market risk premium, providing further indirect evidence.

PAGE 121

111 as expected volatility. Using 16 data observations, every fourth period, Table 9-01 reports the correlation coefficients of expected after-tax excess housing returns in each city with expected volatility. While the correlations are found to be positive, as hypothesized, the number of observations is so small that there is little reliability in any drawing conclusions from these data. TABLE 9-01 Correlation of Expected Quarterly After-Tax Excess Returns and Expected Volatility City Atlanta Chicago Dallas San Fran Parameters E[Re] E[Re] E[Re] E[Re] E[Atl_vol] 0.65 E[Chi_vol] 0.01 E[Dal_vol] 0.27 E[San_vol] 0.13 Autoregressive Conditional Heteroskedasticity The autoregressive conditional heteroskedastic (ARCH) model, initially developed by Engle (1982), allows the full data set of 62 quarterly observations to be used. In examining the a simple autoregressive model, (9-01)

PAGE 122

112 where et is assumed independent and normally distributed with mean, o, and variance, a 2 Maddala (1988) indicates, that the conditional mean, E(YtlYt ,> = r yt_ 1 is a function of t while the conditional variance, var(yt1Yt_ 1 ) is constant. While the simple autoregressive model assumes that the variance is constant, the ARCH model allows the variance to be a function of the past. Thus, the variance is autocorrelated. Engle (1982), generalizing Equation (901) specifies the conditional mean as a linear function such that, Yt = r Yt-1 + et, with et -IN(0,ht), where + .. + (9-02) Under this condition the trend in the variance is conveyed into the mean, resulting in a trend in the mean. Therefore, while diagnostic statistics may suggest that an autocorrelated error structure of the OLS regression exists, it is simply the result of serially dependent variance. The ARCH model has been recently used effectively by Engle, Lilien, and Robins (1985), and French, Schwert, and Stambaugh (1987), and others, to model speculative security prices. They indicate that the variance structure of the series analyzed is predictable and this is translated to the characteristics of the expected mean.

PAGE 123

113 Modeling Risk-Premia This study uses a modification of the ARCH model developed by Engle, Lilien, and Robins (1982) to adjust expected after-tax excess returns for systematic variance in the return series. It is commonly referred to as the ARCH-M model (ARCH in mean). The ARCH-M model allows the conditional variance to be a determinant in modeling the expected mean. Hence, expected after-tax excess returns to housing are modeled as, Re,m_jCt> = B 0 + B 1 Uin_Ct> + B 1 act> + e where, 2 2 2 a = ao + a, e Ct-1> + a2 e Ct-2> (9-03) This suggests that the variance of the return series follows an autoregressive process. The expected variance is modeled as a function of the squared residuals of the previous two quarters. Equation 9-03 is estimated using the quarterly returns generated from the after-tax excess returns constructed in Chapter 7, for each city. The nonlinear model is estimated using a maximium likelihood function. Employing the Regression Analysis of Time Series (RATS) software available from VAR Econometrics, the ARCH model coefficients specified by Eq. 9-03 are estimated.

PAGE 124

114 ARCH-M Test Results The test results indicate a wide dispersion of model estimates across the four cities. Two model specifications were estimated, the first does not control for unexpected inflation, while the second model simultaneously estimates a B 1 coefficient. The first model suggests that with the exception of Dallas, which is negatively related, there is little significance in any of the estimated a 1 or a 2 coefficients. This indicates that there exists little predictability in the variance structure of quarterly after-tax excess returns to housing. San Francisco indicates signs of positive momentum in the variance structure, however the a 1 is not significant at the 5 percent level. Estimating the second model yields results similar to those of the first. Again, in this model the variance structure does not appear to be consistently predictable. Only in San Francisco do the coefficient estimates follow the hypothesized results. In the other cities, contrary to prior expectations, the expected variance appears to be negatively correlated with the expected risk-premium to housing. This condition, while not robust, is speculated to be due to the measurement error in the return series itself and the limited number of observations available. Despite using quarterly return data as opposed to the annual data used previously, the B 1 coefficient remains quite

PAGE 125

115 significant and the coefficient estimate near 1.75 as anticipated. Tests of the model residuals are determined to be invalid due to the unpredictability of the expected variance term. Thus, returns to housing are not consistent with the ARCH-M model as specified. While the estimated model is quite sensitive to the specification of the expected variance function, other model specifications were examined with little change in the results. TABLE 9-02 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates City Obs. Atlanta 61 Chicago 61 Dallas 61 San Fran. 61 on Unexpected Inflation and Volatility Re,m_jCt> = Bo + B, Uin_Ct> + Bz a Ct> B 0 B 1 B 2 a 0 a 1 a 2 (t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat) 0.023 -0.385 0.001 0.019 -0.428 (0.352) (-0.151) (4.532) (0.060) (-0.818) 0.041 1.897 -1.344 0.001 -0.181 -0.549 (5.885) (5.095) (-2.990) (5.277) (-2.840) (-4.049) 0.021 -0.308 0.000 0.255 0.131 (0.691) (-0.211) (2.739) (0.623) (0.213) -0.283 1.642 15.656 0.000 0.006 0.017 (-0.028) (3.281) (0.029) (2.898) (0.030) (0.028) 0.033 -0.780 0.001 -0.151 -0.345 (10.714) (-3.919) (6.114) (-2.334) (-0.816) 0.042 1.900 -1.242 0.001 -0.169 -0.503 (7.307) (5.063) (-3.419) (4.337) (-1.195) (-0.271) 0.020 0.746 0.001 0.836 -0.457 (2.235) (0.235) (3.037) (1.693) (-0.702) 0.017 1. 504 0.252 0.000 1.010 -0.533 (1.346) (4.147) (0.527) ( 2. 64 7) (2.237) (-3.090) Note: Unexpected inflation model with ARIMA 1-qtr. forward model.

PAGE 126

116 Trading Strategies As described previously in Chapter 2, formal testing of the random-walk model is limited in it's ability to describe fully the necessary conditions of the EMH. In order to reject the EMH, random-walk inconsistencies must be exploitable. Autocorrelation in the return structure alone, does not imply a rejection of the EMH. The results reported in Chapter 8 indicate that after adjusting for unexpected inflation, abnormal returns are significantly correlated in Chicago. In addition, although not statistically significant, Dallas and San Francisco exhibit substantial inertia in their abnormal return structure. However, the returns to housing in Atlanta appear consistent with that of the random-walk and unexploitable. Exploiting any known serial dependency in the housing market appears quite difficult. Their currently exist no futures contracts or short selling opportunities for the housing asset. Thus, any expected decrease in the price of housing is very difficult to exploit. In addition, transaction costs, which are assumed by Linneman (1986) to be as high as 12 percent, make it largely prohibitive for

PAGE 127

the typical home buyer to exploit any expected price increases. 3 1 1 7 Case and Shiller (1989) indicate, home buyers who already intend to purchase a home, but are indifferent to the time of purchase, may be able to exploit the market. In this case, all transaction costs are irrelevant, since they will be incurred anyway. The following trading rule is suggested by Case and Shiller: buy now if the after-tax excess return modeled by the regression in Table 6-04 is greater than the after-tax excess full sample mean, otherwise delay the purchase for one year. Using this methodology they indicate the average trading profits for the years invested earn range about 1 to 3 percent of the value of the house. 4 Looking closer at this strategy, it implies that there are no costs to waiting a year. However, under such a strategy any no excess housing returns can be earned in the year waited. Hence, it is important to include the "opportunity cost" of waiting a year. Under this scenerio, which includes the "cost" of delaying the purchase, one buys if the expected after-tax 3 Transaction costs often include such items as brokers fees, moving expenses, and search costs. In addition, other types of costs may be incurred which are difficult to quantify, such as new schools or neighborhood amenities. 4 case and Shiller define the average trading profit as the difference between the buy signal average returns and the total sample average returns times the proportion of quarters indicating a buy signal.

PAGE 128

118 excess return is greater than the sample mean, otherwise the home buyer delays purchase for one year and invests in one year treasury bills. The average return from this strategy is listed below in column B of Table 9-03. Column C indicates a strategy where the buyer buys if the expected after-tax excess return is greater than the treasury bill, otherwise dely the purchase and invest in the security. Only in San Francisco does this trading rule earn excessive returns; however, the rule is not consistent across cities. TABLE 9-03 City Regression Atlanta AB BA Chicago AB BA Dallas AB BA San Fran. AB BA Average Quarterly Returns Using Alternative Trading Strategies A Ave. Sample Re .0615 .0622 .0646 .0636 .0949 .0846 .1105 .1141 B Mean Strat. E[Re] .0376 .0398 .0605 .0550 .0551 .0477 .0923 .1018 C TBill Strat. E[Re] .0615 .0622 .0644 .0646 .0949 .0827 .1128 .1152 D Ave. Sample Re,m .0536 .0541 .0569 .0559 .0870 .0766 .1059 .1094 E Mean strat. E [Re, m] .0213 .0290 .0549 .0519 .0477 .0423 .0685 .0801 F TBill Strat. E[Re, m] .0009 .0000 .0528 .0509 .0553 .0414 .0685 .0842 Note: Column A, B, and C denotes the average quarterly return for each sample using unadjusted after-tax excess return data. Columns D, denotes the sample average quarterly return using return data adjusted for home buyer marginal tax rates. Columns E and F applies trading rules on the basis of the expected after-tax excess returns adjusted for unexpected inflation.

PAGE 129

119 Column D reports the average annual after-tax excess returns using time-varying home buyer marginal tax rates. Using expected after-tax returns calculated from the regression in Tables 8-03, 8-05, 8-07, 8-09 in Chapter 8 to signal the buy or delay purchase signal. The actual returns that would have resulted from such a signal are listed in columns E and F of Table 9-01. Column E gave a buy signal for values above the mean of the expected after-tax returns adjusted for unexpected inflation, while column F gave a buy signal for expected returns greater than the after-tax one year treasury bill. These strategies do not consistently outperform the buy-and-hold strategy, hence the EMH cannot be rejected.

PAGE 130

CHAPTER 10 SUMMARY AND CONCLUSIONS Previous results examining the efficiency of the single-family housing market have been mixed. This study extends past work by incorporating time-varying home buyer marginal tax rates, unexpected inflation and volatility into the analysis of after-tax excess returns to housing. Findings indicate that excess housing returns are negatively correlated with home buyer marginal tax rates, which are determined to be autocorrelated. However, when time-varying marginal tax rates are introduced into the housing return model and the resulting return series tested for autocorrelation, there is not significant change from previous estimates. This suggests that findings of autocorrelated after-tax excess returns to housing are robust to alternative marginal tax rate specifications. More importantly, excess housing returns are found to be positively correlated with unexpected inflation. Evidence is presented that suggests inflation shocks were autocorrelated during the late 1970s and early 1980s, and not fully anticipated by either the bond or housing markets. When after-tax excess returns are adjusted for both time varying marginal tax rates and autocorrelated inflation 120

PAGE 131

121 shocks, findings indicate that there remains persistent return inertia in three of the four markets studied, however, trading rules cannot exploit this autocorrelation. Hence the hypothesis of an efficient market cannot be rejected. In addition, preliminary evidence suggests that autocorrelated return patterns are correlated with city size. It is speculated that inactive markets may induce inefficient results and that substantial search costs in larger cities may inhibit competing properties to be identified allowing greater price inertia to occur. This would appear to explain the conditions of predictable returns consistently found in Chicago. Finally, the volatility of the return structure is explored. While expected return volatility is found to be positively correlated to expected after-tax quarterly excess returns, employing an ARCH-M model using quarterly returns yields little predictability in the volatility of city return series. Hence, there appears to be no ARCH impact on the expected after-tax excess returns. Although the data do not confirm the ARCH model, it seems likely that home buyers would require greater expected returns during times of volatile inflation (and volatile housing returns), however, this remains an area to be examined in the future.

PAGE 132

APPENDIX A WEIGHTED-REPEAT-SALES (WRS) PRICE INDICES The median sales price of single-family housing for different markets is widely reported in the popular literature. Such price statistics have been highly criticized as measures of housing appreciation primarily because price changes are contaminated by changes in the age, characteristics and size of the units sold over the period reported. Several methods have been suggested in the literature that attempt to identify the changes in price of "equivalent quality" housing. 1 Case and Shiller (1987) propose a weighted-repeat-sales (WRS) methodology for constructing existing home price indices. They argue that the WRS methodology is superior to other return indices because it most accurately controls for size and quality changes in house transaction data. The WRS method is a modification of the work of Bailey, Muth, and Nourse (1963) (BMN), controlling for heteroskedasticity in the regression error term. The BMN method can be described by first assuming: (A-01) 1 Nourse (1982) and Greenlees (1984) provide useful critics of alternative housing price indices. 122

PAGE 133

123 where Pi is the initial sales price; P 8 is the second sales price; rt is the rate of appreciation in period t; Dt is the exponent equal to -1 if period tis the initial sale, +1 if it is the repeat sale, and O for all other t. Taking the natural log of (6.01) yields, (A-02) thus, (A-03) where ( ) 'denote the log function. Using data which identify a large number of repeat sales, (B.03) can be estimated by linear regression. The independent variables, Dt, represent a series of dummy variables for each period tin the sample, except for the initial period o. The dummy variables take the value -1 for the year in which the initial sales tales place, 1 for the year in which the second sale takes place, otherwise O. The estimated coefficients represent the natural logs of the constructed price index. Case and Shiller indicate that the variance of the regression error term from (A-03) is likely to be related to the time interval between repeat sales. Further, they argue that there is to be a random drift component due to the differences in the amount of house depreciation and upkeep among houses. A modification is made to the BMN method by

PAGE 134

assuming that the log price of the jth house at time tis given by: 124 (A-04) where ct is the log of the citywide level of housing prices at time t; Hjt is a random walk drift component of zero mean and variance a\; and Ujt is a house specific random error term having zero mean and variance a 2 n. The WRS method consists of three stages. First, the BMN procedure is followed by estimating (B.03), and a vector of regression residuals is calculated. A weighted regression of the squared residuals in the first stage is then regression on the time interval between the sales using a constant term. The constant term of the second-stage regression is an estimate of the house specific random error (2a 2 n), and the slope term is the estimate of drift (a\). Finally, a generalized least squares regression (weighted) is run by first dividing each observation in the first-stage regression by the square root of the fitted value in the stage-two regression and then running the stage-one regression again. Sample bias can enter the WRS methodology in applying the data to the regressions. When used to estimate indices for repeat-sales over short time periods, the index restricts observations to a disproportionate number of high turnover properties, potentially reflecting the pricing of

PAGE 135

125 forced sales or renovation properties. This effect will influence the index, particularly if the proportion of high turnover properties are not constant over time. In this case any subsequent tests of autocorrelation will be biased. Despite these limitations, it is unlikely that spurious correlation of this type, introduced by the methodology, could result in the degree of autocorrelation reported by Case and Shiller. It appears reasonable to assume that one directional tests of efficiency can be conducted using the methodology. While the WRS methodology may induce some limited time pattern to the return series, real estate transactions costs are generally sufficient to suppress the likelihood of establishing any profitable investment rules under such conditions. Therefore, if the return series are found to follow random-walk properties, it is reasonable not to reject the hypothesis of an efficient market.

PAGE 136

APPENDIX B SUPPLEMENTARY TABLES AND STATISTICS

PAGE 137

TABLE B-01 Regression of After-Tax 4-Quarter Excess Returns on After-Tax 4-Quarter Excess Returns Lagged City Parameters Atlanta j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Chicago j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Dallas j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 San Francisco j=A, k=B, L=0 j=A, k=B, L=4 j=B, k=A, L=4 Sample Period Obs. 1970:1-1985:2 1970:1-1985:2 1970:1-1985:2 1970:1-1985:3 MSE 62 0.029 58 0.049 58 0.046 62 0.026 58 0.053 58 0.053 62 0.048 58 0.064 58 0.061 63 0.069 59 0.099 59 0.077 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.011 (1.106) 0.038 (2.085) 0.037 (2.147) 0.004 (0.504) 0.019 (1.132) 0.017 (1.008) 0.024 (1.440) 0.045 (1.845) 0.031 ( 1. 292) 0.040 (1.736) 0.053 (1.498) 0.044 (1.676) 0.820 (6.428) 0.331 (1.478) 0.363 (1.741) 0.935 (10.616) 0.663 (3.697) 0.698 (3.921) 0.810 (5.213) 0.540 (2.508) 0.520 (2.629) 0.619 ( 4. 150) 0.523 (2.177) 0.562 (3.374) .695 .690 .114 .098 .152 .137 .861 .859 .447 .437 .476 .466 .600 .593 .271 .258 .290 .277 .483 .474 .216 .202 .398 .387 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MAJ process as described by Hansen and Hodrick (1980). 127

PAGE 138

128 TABLE B-02 Regression of After-Tax Quarterly Excess Returns on After-Tax Quarterly Excess Returns Lagged Using Nonconstant Tenure Choice Marginal Tax Rates Re_Ht> = B 0 + 8 1 Re_k(tL) + u City Sample Period Obs. B 0 8 1 R2 Parameters MSE (t-stat) (t-stat)adjR 2 Atlanta 1970:1-1986:1 j=A, k=B, L=0 65 0.007 0.516 .368 0.022 (2.212) (6.058) .358 j=A, k=B, L=l 64 0.016 -0.139 .027 0.028 (4.326) (-1.302) .011 j=B, k=A, L=l 64 0.018 -0.225 .037 0.032 (4.026) (-1.534) .021 Chicago 1970:1-1986:1 j=A, k=B, L=0 65 0.008 0.534 .268 0.022 (2.449) (4.804) .266 j=A, k=B, L=l 64 0.006 0.562 .304 0.021 (2.205) (5.200) .292 j=B, k=A, L=l I 64 0.010 0.263 .071 0.024 (2.800) (2.172) .056 Dallas 1970:1-1986:1 j=A, k=B, L=0 65 0.015 0.244 .066 0.034 (3.166) (2.103) .051 j=A, k=B, L=l 64 0.017 0.215 .053 0.033 (3.493) (1.863) .038 j=B, k=A, L=l 64 0.017 0.078 .007 0.032 (3.583) (0.673)-.009 San Francisco 1970:1-1986:2 j=A, k=B, L=0 66 0.026 -0.050 .004 0.040 (4.833) (-0.508)-.012 j=A, k=B, L=l 65 0.022 0.155 .040 0.039 (4.006) (1.613) .024 j=B, k=A, L=l 65 0.005 0.837 .416 0.040 (0.775) (6.701) .407 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters.

PAGE 139

129 TABLE B-03 Regression of After-Tax 4-Quarter Excess Returns on After-Tax 4-Quarter Excess Returns Lagged Using Nonconstant Tenure Choice Tax Rates Re_Ht> = Bo + B, Re_kCt-L) + u Obs. R2 City Sample Period B 0 B 1 Parameters MSE (t-stat) (t-stat)adjR 2 Atlanta 1970:1-1985:2 j=A, k=B, L=0 62 0.011 0.818 .690 0.029 (1.092) (6.353) .685 j=A, k=B, L=4 58 0.036 0.363 .115 0.048 (2.031) (1.485) .099 j=B, k=A, L=4 58 0.035 0.333 .151 0.046 (2.096) (1.737) .136 Chicago 1970:1-1985:2 j=A, k=B, L=0 62 0.004 0.937 .866 0.026 (0.468) (10.822) .864 j=A, k=B, L=4 58 0.017 0.669 .455 0.054 (1.040) (3.758) .445 j=B, k=A, L=4 58 0.015 0.705 .486 0.053 (0.921) (4.166) 477 Dallas 1970:1-1985:2 j=A, k=B, L=0 62 0.024 0.801 .583 0.048 (1.464) (5.042) .577 j=A, k=B, L=4 58 0.045 0.528 .260 0.062 (1.903) (2.440) .247 j=B, k=A, L=4 58 0.032 0.507 .278 0.060 (1.341) (2.558) .266 San Francisco 1970:1-1985:3 j=A, k=B, L=0 63 0.043 0.597 .452 0.069 (1.832) (3.903) .443 j=A, k=B, L=4 59 0.058 0.496 .193 0.097 (1.606) (2.030) .179 j=B, k=A, L=4 59 0.047 0.550 .383 0.076 (1.774) (3.269) .371 Note: A and B denote half sample indices for each city respectively and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MA3 process as described by Hansen and Hodrick (1980).

PAGE 140

130 TABLE B-04 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Alternative Unexpected Inflation Models: Atlanta Re,m_j(t) = Bo + B, Uin_Ct> + u City Parameters Atlanta j=A, n=2 j=B, n=2 j=A, n=3 j=B, n=3 j=A, n=4 j=B, n=4 j=A, n=5 j=B, n=5 j=A, n=6 j=B, n=6 Sample Period Obs. 1970:1-1986:1 MSE 65 0.027 65 0.032 65 0.026 65 0.032 65 0.028 65 0.033 65 0.027 65 0.032 65 0.027 65 0.032 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.012 (3.597) 0.013 (3.216) 0.015 (4.526) 0.015 (3.840) 0.013 (3.563) 0.015 (3.436) 0.012 (2.743) 0.011 (3.101) 0.014 (4.195) 0.015 (1.882) 1.242 .095 (2.577) .081 1. 000 045 (1.719) .030 1.487 .123 (2.970) .109 1.839 .060 (2.717) .045 0.886 .016 (1.015) .001 -0.103 .000 (-0.100)-.016 1. 983 062 (2.041) .047 1. 946 082 (2.377) .068 0.899 .053 (1.882) .038 0.911 .040 (1.613) .024 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. n denotes the model used for inflation expectations n=2: ARIMA model (in sample) n=3: ASA-NBER n=4: Polynomial Distributed Lag (6,3) n=5: Harris POL n=6: ARIMA model (1-step forward, using city specific inflation data.

PAGE 141

131 TABLE B-05 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged: Atlanta (Using Alternative Models) Rres,n_j,(t) = Bo + s, Rres,n_k,(t-L) + u City Sample Period Obs. B 0 B 1 R2 Parameters MSE (t-stat) (t-stat)adjR 2 Atlanta 1970:1-1985:2 j=A;k=B;n=2;L=0 62 0.002 0.770 .647 0.027 (0.472) (5.767) .641 j=A;k=B;n=2;L=4 58 -0.003 0.380 .183 0.039 (-0.632) (1.948) .168 J=B;k=A;n=2;L=4 58 -0.006 0.346 .120 0.043 (-1.113) (1.522) .105 j=A;k=B;n=3;L=0 62 0.001 0.689 .561 0.026 (0.346) (4.816) .554 j=A;k=B;n=3;L=4 58 -0.004 0.355 .180 0.033 (-0.813) (1.925) .165 j=B;k=B;n=3;L=4 58 -0.006 0.317 .094 0.039 (-1.270) (1.326) .078 j=A;k=B;n=4;L=0 62 0.001 0.766 .690 0.026 (0.245) (6.357) .685 j=A;k=B;n=4;L=4 58 -0.002 0.321 .133 0.044 (-0.207) (1.610) .117 j=B;k=A;n=4;L=4 58 -0.005 0.355 .156 0.048 (-0.463) (1.488) .100 j=A;k=B;n=5;L=0 62 0.002 0.699 .581 0.026 (0.276) (5.017) .574 j=A;k=B;n=5;L=4 58 -0.001 0.233 .068 0.038 (-0.076) (1.115) .051 j=B;k=B;n=5;L=4 58 -0.003 0.205 .038 0.042 (-0.343) (0.813) .020 j=A;k=B;n=6;L=0 62 0.002 0.738 .563 0.026 (0.262) (4.837) .556 j=A;k=B;n=6;L=4 58 -0.004 0.243 .079 0.035 (-0.513) (1.205) .063 j=B;k=A;n=6;L=4 58 -0.007 0.140 .023 0.038 (-0.806) (0.627) .005 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MA3 process as described by Hansen and Hodrick (1980).

PAGE 142

TABLE B-06 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates 132 on Alternative Unexpected Inflation Models: Chicago Re m Ht> = Bo + B, Uin + u Obs. R2 City Sample Period B 0 B 1 Parameters MSE (t-stat) (t-stat)adjR 2 Chicago 1970:1-1986:1 j=A, n=2 65 0.015 0.048 .000 0.026 (4.561) (0.106)-.016 j=B, n=2 65 0.012 0.832 .055 0.024 (4.084) (1.908) .040 j=A, n=3 65 0.015 0.320 .007 0.025 (4.739) (0.659)-.009 j=B, n=3 65 0.014 1.205 0.023 (4.914) (2.693) j=A, n=4 65 0.013 1.187 0.025 (4.108) (1.505) j=B, n=4 65 0.012 1.208 0.024 (3.865) (1.582) j=A, n=5 65 0.011 2.263 0.024 (3.634) (3.119) j=B, n=5 65 0.010 2.432 0.023 (3.340) (3.517) j=A, n=7 65 0.016 1.066 0.024 (5.320) (2.698) j=B, n=7 65 0.015 1.117 0.023 (5.160) (2.941) Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. n denotes the model used for inflation expectations n=2: ARIMA model (in sample) n=3: ASA-NBER n=4: Polynomial Distributed Lag (6,3) n=5: Harris PDL n=7: ARIMA model (1-step forward, using city specific inflation data. .103 .089 .035 .019 .038 .023 .134 .120 .164 .151 .104 .089 .121 .107

PAGE 143

133 TABLE B-07 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged: Chicago (Using Alternative Models) Rres,n j,(t) = B 0 + s, Rres,n k,Ct-L) + u City Sample Period Obs. B 0 B 1 R2 Parameters MSE (t-stat) (t-stat)adjR 2 Chicago 1970:1-1985:2 j=A;k=B;n=2;L=0 62 -0.003 0.955 .862 0.025 (-0.675) (10.628) .859 j=A;k=B;n=2;L=4 58 -0.005 0.744 .521 0.049 (-0.461) (4.289) .512 j=B;k=A;n=2;L=4 58 -0.001 0.607 .394 0.053 (-0.046) (3.317) .383 j=A;k=B;n=3;L=0 62 -0.004 0.961 .838 0.027 (-0.616) (9. 681) .835 j=A;k=B;n=3;L=4 58 -0.005 0.760 .522 0.048 (-0.474) (4.303) .514 j=B;k=B;n=3;L=4 58 0.000 0.548 .334 0.054 (0.021) (2.916) .322 j=A;k=B;n=4;L=0 62 -0.005 0.932 .690 0.025 (-0.790) (9.886) .685 j=A;k=B;n=4;L=4 58 -0.005 0.641 .133 0.050 (-0.374) (3.372) .117 j=B;k=A;n=4;L=4 58 -0.003 0.592 .156 0.052 (-0.223) (3.079) .100 j=A;k=B;n=5;L=0 62 -0.005 0.905 .792 0.025 (-0.794) (8.303) .788 j=A;k=B;n=5;L=4 58 -0.002 0.514 .259 0.048 (-0.210) (2.435) .246 j=B;k=B;n=5;L=4 58 -0.004 0.436 .195 0.050 (-0.332) (2.025) .181 j=A;k=B;n=7;L=0 62 -0.005 0.873 .790 0.025 (-0.820) (8.270) .787 j=A;k=B;n=7;L=0 58 -0.006 0.489 .250 0.048 (-0.555) (2.375) .236 j=A;k=B;n=7;L=0 58 -0.002 0.469 .218 0.049 (-0.021) (2.173) .204 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MAJ process as described by Hansen and Hodrick (1980).

PAGE 144

134 TABLE B-08 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Alternative Unexpected Inflation Models: Dallas City Parameters Dallas j=A, n=2 j=B, n=2 j=A, n=3 j=B, n=3 j=A, n=4 j=B, n=4 j=A, n=5 j=B, n=5 j=A, n=8 j=B, n=8 Sample Period Obs. 1970:1-1986:1 MSE 65 0.033 65 0.036 65 0.033 65 0.035 65 0.034 65 0.037 65 0.034 65 0.036 65 0.033 65 0.036 B 0 B 1 R 2 (t-stat) (t-stat)adjR 2 0.017 (4.075) 0.018 (3.939) 0.020 (4.914) 0.020 (4.771) 0.017 (3.848) 0.019 (4.008) 0.017 (3.661) 0.017 (3.503) 0.021 (5.152) 0.021 (4.598) 1.613 .103 (2.689) .089 1.388 .069 (2.162) .054 1.629 .009 (2.566) .008 1.765 .101 (2.653) .086 1. 975 051 (1.845) .036 0.493 .003 (0.428)-.013 1.965 .054 (1.892) .039 2.093 .055 (1.919) .040 1.578 .130 (3. 061) .116 0.958 .043 (1.687) .028 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. n denotes the model used for inflation expectations n=2: ARIMA model (in sample) n=3: ASA-NBER n=4: Polynomial Distributed Lag (6,3) n=5: Harris POL n=8: ARIMA model (1-step forward, using city specific inflation data.

PAGE 145

135 TABLE B-09 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Excess Returns Lagged: Dallas (Using Alternative Models) Rres,n_j,Ct) = Bo + B, Rres,n_k,Ct-L) + uct> City Sample Period Obs. B 0 B 1 R2 Parameters MSE (t-stat) (t-stat)adjR 2 Dallas 1970:1-1985:2 j=A;k=B;n=2;L=0 62 0.008 0.771 .535 0.046 (0.773) (4.569) .527 j=A;k=B;n=2;L=4 58 0.005 0.514 .252 0.045 (0.377) (2.391) .239 J=B;k=A;n=2;L=4 58 -0.006 0.500 .290 0.056 (-0.837) (2.627) .277 j=A;k=B;n=3;L=0 62 0.008 0.777 .511 0.045 (0. 791) (4.357) .503 j=A;k=B;n=3;L=4 58 0.005 0.530 .251 0.054 (0.407) (2.382) .238 j=B;k=B;n=3;L=4 58 -0.010 0.443 .241 0.050 (-0.852) (2.320) .228 j=A;k=B;n=4;L=0 62 0.008 0.754 .561 0.025 (0.746) (4.814) .554 j=A;k=B;n=4;L=4 58 0.007 0.530 .284 0.050 (0.520) (2.595) .272 j=B;k=A;n=4;L=4 58 -0.009 0.523 .284 0.052 (-0.637) (2.594) .272 j=A;k=B;n=5;L=0 62 0.008 0.786 .537 0.025 (0.831) (4.584) .529 j=A;k=B;n=5;L=4 58 0.009 0.514 .254 0.048 (0.672) (2.404) .241 j=B;k=B;n=5;L=4 58 -0.007 0.431 .224 0.050 (-0.519) (2.212) .210 j=A;k=B;n=8;L=0 62 0.007 0.672 .418 0.046 (1.148) (6.561) .408 j=A;k=B;n=8;L=0 58 0.002 0.401 .161 0.053 (0.232) (1.802) .146 j=A;k=B;n=8;L=0 58 -0.010 0.289 .098 0.053 (-1.415) (1.357) .082 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MA3 process as described by Hansen and Hodrick (1980).

PAGE 146

136 TABLE B-10 Regression of After-Tax Quarterly Excess Returns Adjusted for Home Owner Marginal Tax Rates on Alternative Unexpected Inflation Models: San Francisco Re,m j(t) = Bo + B, Uin_Ct> + uct> Obs. R2 City Sample Period B 0 B 1 Parameters MSE (t-stat) (t-stat)adjR 2 San Francisco 1970:1-1986:2 j=A, n=2 66 0.025 -0.059 .ooo 0.041 (4.840) (2.689)-.016 j=B, n=2 66 0.022 2.138 .085 0.050 (3.554) (2.436) .071 j=A, n=3 66 0.025 0.193 .001 0.040 (4.964) (0.252)-.015 j=B, n=3 66 0.027 2.450 .098 0.049 (4.420) (2.639) .084 j=A, n=4 66 0.024 0.788 .006 0.040 (4.525) (0.631)-.009 j=B, n=4 66 0.024 1.129 .008 0.052 (3.625) (0.708)-.008 j=A, n=5 66 0.021 2.507 0.039 (4.057) (2.097) j=B, n=5 66 0.022 2.336 0.051 (3.312) (1.504) j=A, n=9 66 0.026 0.797 0.039 (2.884) (0.854) j=B, n=9 66 0.027 1.594 0.050 (2.389) (1.361) Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters. n denotes the model used for inflation expectations n=2: ARIMA model (in sample) n=3: ASA-NBER n=4: Polynomial Distributed Lag (6,3) n=5: Harris POL n=9: ARIMA model (1-step forward, using city specific inflation data. .064 .050 .034 .019 .037 .022 .089 .074

PAGE 147

137 TABLE B-11 Regression of After-Tax 4-Quarter Residual Returns on After-Tax 4-Quarter Residual Returns Lagged: San Francisco (Using Alternative Models) Rres,n j, (t) = Bo + B, Rres,n lc,(t) + u City Obs. B 0 B 1 R 2 Parameters Period MSE (t-stat) (t-stat)adjR 2 San Francisco 1970:1-1985:3 j=A;k=B;n=2;L=0 63 0.002 0.643 .458 0.065 (0.153) (3.952) .450 j=A;k=B;n=2;L=4 59 0.002 0.596 .393 0.071 (0.089) (3.340) .382 J=B;k=A;n=2;L=4 59 -0.001 0 457 .187 0.087 (-0.058) (1.993) .173 j=A;k=B;n=3;L=0 63 0.003 0.628 .419 0.067 (0.171) (3.650) .410 j=A;k=B;n=3;L=4 59 0.103 0.603 .385 0.071 (-0.441) (3.283) .374 j=B;k=B;n=3;L=4 59 -0.001 0.451 .191 0.083 (-0.073) (2.015) .176 j=A;k=B;n=4;L=0 63 0.002 0.591 .442 0.064 (0.155) (3.826) .433 j=A;k=B;n=4;L=4 59 0.002 0.530 .356 0.071 (0.114) (3.086) .345 j=B;k=A;n=4;L=4 59 0.001 0.462 .168 0.091 (0.028) (1.866) .153 j=A;k=B;n=5;L=0 63 0.002 0.526 .367 0.061 (0.166) (3.272) .357 j=A;k=B;n=5;L=4 59 0.004 0.431 .259 0.068 (0.256) (2 454) .246 j=B;k=B;n=5;L=4 59 0.003 0.368 .099 0.087 (0.127) (1.380) .084 j=A;k=B;n=9;L=0 63 0.001 0.598 .358 0.065 (0.053) (3.209) .348 j=A;k=B;n=9;L=4 59 -0.001 0.579 .338 0.064 (-0.037) (2.969) 327 j=B;k=A;n=9;L=4 59 -0.004 0.366 .135 0.067 (-0.211) (1.638) .120 Note: A and B denote half sample indices for each city respectively, and L denotes number of lagged quarters Coefficient standard errors have been adjusted to yield estimates of the corrected standard errors for a MAJ process as described by Hansen and Hodrick (1980).

PAGE 148

REFERENCE LIST Adams, F.G., G. Milgram, E.W. Green, and C. Mansfield, "Undeveloped Land Prices During Urbanization: A Micro Empirical study Over Time," The Review of Economics and Statistics, 50:1968, 248-258. Alexander, S.S., "Price Movements in Speculative Markets: Trends or Random Walks," Industrial Management Review, 2:1961, 7-12. Bachelier, L., "Th~orie de la Speculation" (Paris: Gauthier Villars, 1900), and reprinted in P. Cootner, ed. The Random Character of Stock Market Prices, M.I.T. Press, Cambridge, 1964, 17-78. Bailey, M.J., R.F. Muth, and H.O. Nourse, "A Regression Method for Real Estate Price Index Construction," American Statistical Association Journal, 58:1963, 933942. Blanchard, O.J., Deflation," Proceedings 211-215. "The Lucas Critique and the Volcker American Economic Association Papers and in The American Economic Review, 74:1984, Box, G.E., and G.M. Jenkins, Time Series Analysis Forecasting and Control, Holden-Day, San Francisco, 1970. Box, G.E., and D.A. Pierce, Distribution of Residual Autocorrelations in Autoregressive-Integrated Moving Average Time Series Models," Journal of the American Statistical Association, 65:1970, 1509-1526. Bruggeman, W.B., A.H. Chen, and T.G. Thibodeau, "Real Estate Invesment Funds: Performance and Portfolio Considerations," Journal of the American Real Estate and Urban Economics Association, 12:1984, 333-354. Burns, w., and D. Epley, "The Performance of Portfolios of REITs and Stocks," Journal of Portfolio Management, 8:1982, 37-42. 138

PAGE 149

Carlson, J.A., "A Study of Price Forecasts," Annals of Economic and Social Measurement, 6:1977, 27-56. 139 Case, K.E., "The Market for Single-Family Homes in Boston," New England Economic Review, May/June, 1986, 38-48. Case, K.E., and R.J. Shiller, "Prices of Single-Family Homes since 1970: New Indexes for Four Cities," The New England Economic Review, September, 1987, 45-56. Case, K.E., and R.J. Shiller, "The Efficiency of the Market for Single-Family Homes," The American Economic Review, 79:1989, 125-137. Copeland, T., and J. Weston, Financial Theory and Corporate Policy, Third Edition, Addison-Wesley, Reading, 1988. Cornell, B., and R. Roll, "Strategies for Pairwise Competitions in Markets and Organizations," Bell Journal of Economics, 12:1981, 201-213. Coyne, T., w. Goulet, and M. Picconi, "Residential Real Estate Versus Financial Assets," Journal of Portfolio Management, 7:1980, 20-24. Davies, G.W., "An Examination of the Market for Urban Single-Family Detached Lots," Journal of Regional Science, 17:1977, 243-253. Dokko, Y., and R.H. Edelstein, "The Empirical Interrelationship Among the Mundell and Darby Hypotheses and Expected Stock Market Returns," Review of Economics and Statistics, 69:1987, 161-166. Dokko, Y., R.H. Edelstein, M. Pomer, and E.S. Urdang, "Determinants of the Rate of Return for Commercial Real Estate: Inflation Expectations and Market Adjustment Lags," Unpublished Manuscript, Center for Real Estate and Urban Economics, University of California, Berkeley, 1990. Engle, R.F., "Autoregressive Conditional Heteroscedasiticity with Estimates of the Variance of United Kingdom Inflation," Econometrica, 50:1982, 987-1007. Engle, R.F., D.M. Lilien, and R.P. Robins, "Estimating Time Varying Risk Premia in the Term Structure: The ARCH-M Model," Econometrica, 55:1987, 391-407. Essayyad, M., D. Marx, and H. Green, "An Empirical Investigation of Efficiency in the Real Estate Market,"

PAGE 150

140 Unpublished Manuscript, University of Alaska, Anchorage, 1987. Fama, E.F., "Efficient Capital Markets: A Review of Theory and Empirical Work," Journal of Finance, 25:1970, 383417. Fama, E.F., "Short-Term Interest Rates as Predictors of Inflation," The American Economic Review, 65:1975, 269282. Fama, E.F., Foundations in Finance, Basic Books, New York, 1976. Fama, E.F., and M.R. Gibbons, "Inflation, Real Returns, and Capital Investment," Journal of Monetary Economics, 9:1982, 297-323. Fama, E.F., and M.R. Gibbons, "A Comparison of Inflation Forecasts," Journal of Monetary Economics, 13:1984, 327-348. Fama, E.F., and G.W. Schwert, "Asset Returns and Inflation," Journal of Financial Economics, 5:1977, 115-146. French, K.R., "Stock Returns and the Weekend Effect," Journal of Financial Economics, 8:1980, 55-69. French, K.R., G.W. Schwert, and R.F. Stambaugh, "Expected Stock Returns and Volatility," Journal of Financial Economics, 19:1987, 3-19. Gau, G., "Weak-Form Tests of Efficiency of Real Estate Investment Markets," Financial Review, 19:1984, 301320. Gau, G., "Public Information and Abnormal Returns in Real Estate Investment," Journal of the American Real Estate and Urban Economics Association, 13:1985, 15-31. Geltner, D., "Bias in Appraisal Based Returns." Journal of the American Real Estate and Urban Economics Association, 17:1989, 338-352. Giliberto, S.M "A Note on the Use of Appraisal Data in Indexes of Performance Measurement." Journal of the American Real Estate and Urban Economics Association, 16:1988, 77-83. Greenlees, J., "Sample Truncation in FHA Data: Implications for Home Purchase Indexes," Southern Economic Journal, 48:1982, 917-931.

PAGE 151

Grossman, S.J., "The Impossibility of Informationally Efficient Markets," The American Economic Review, 70:1980, 393-408. Grossman, S.J., and J. stiglitz, "Information and Competitive Price Systems," The American Economic Review, 66:1976, 246-253. 141 Guntermann, K.L., and R.L. Smith, "Efficiency of the Market for Residential Real Estate," Land Economics, 63:1987, 34-45. Gyourko, J., and P. Linneman, "Owner-Occupied Homes, Income Producing Properties, and REITs as Inflation Hedges: Empirical Findings," Journal of Real Estate Finance and Economics, 1:1988, 347-372. Gyourko, J., and R. Voith, "Local Market and National Components in House Price Appreciation," Unpublished Manuscript, University of Pennsylvania, Philadelphia, 1990. Hafer R.W., and S.E. Hein, "On the Accuracy of Time-Series, Interest Rate, and Survey Forecasts of Inflation," Journal of Business, 58:1985, 377-398. Hafer R.W., and S.E. Hein, "Forecasting Inflation Using Interest-Rate and Time-Series Models: Some International Evidence," Journal of Business, 63:1990, 1-17. Hamilton, B.W., and R.M. Schwab, "Expected Appreciation in Urban Housing Markets," Journal of Urban Economics, 18:1985, 103-118. Hansen, L.P., and R.J. Hodrick, "Forward Exchange Rates as Optimal Predictors of Future Spot Rates: An Econometric Analysis," Journal of Political Economy, 88:1980, 829853. Hartzell, D.J., J.S. Hekman, and M.E. Miles, "Real Estate Returns and Inflation," Journal of the American Real Estate and Urban Economics Association, 15:1, 1987, 617-637. Hartzell, D.J., A. Lepcio, J.D. Fernald, and S. Jordan, "Commercial Mortgage-Backed Securities: An Investor's Primer," Housing Finance Review, 6:1987, 169-191. Hendershott, P.H., and J.D. Shilling, "The Economics of Tenure Choice, 1955-1979," Research in Real Estate, 1:1982, 105-133.

PAGE 152

142 Ibbotson, R.G., and L.B. Siegel, "Real Estate Returns: A Comparison with Other Investments," Journal of the American Real Estate and Urban Economics Association, 12:1984, 219-242. Isakson, H.R., and T.H. Mcinish, "Efficiency of the Market for Multi-Family Housing," Unpublished Manuscript, Department of Finance and Real Estate, The University of Texas at Arlington, 1988. Jaffe, A.J., and C.F. Sirmans, "The Theory and Evidence on Real Estate Financial Decisions: A Review of the Issues," Journal of the American Real Estate and Urban Economics Association, 12:1984, 378-400. Kaplan, H., "Farmland as a Portfolio Investment," Journal of Portfolio Management, 11:1985, 73-78. Kapplin, S.D., and A.L. Schwartz, "Public Real Estate Limited Partnership Returns: A Preliminary Comparison with Other Investments," Journal of the American Real Estate and Urban Economics Association, 16:1988, 63-76. Ling, D.C., "The Price of Owner-Occupied Housing Services: 1973-1989," Unpublished Manuscript, University of Florida, Gainesville, 1990. Linneman, P., "An Empirical Test of the Efficiency of the Housing Market," Journal of Urban Economics, 20:1986, 140-154. Liu, C.H., T.V. Grissom, and D.J. Hartzell, "Market Imperfections, Omitted Asset Markets, and Abnormal Real Estate Returns: A Theoretical Investigation," Unpublished Manuscript, New York University, New York, 1988. Locke, s., "Real Estate Market Efficiency," Land Development Studies, 3:1986, 171-178. Maddala, G.S., Introduction to Econometrics, McMillan Publishing Co., New York, 1988. McIntosh, w., and G.V. Henderson, "Efficiency of the Office Properties Market," Journal of Real Estate Finance and Economics, 2:1989, 61-70. Miles, M., and T. Mccue. "Commercial Real Estate Returns," Journal of the Real Estate and Urban Economics Association, 12:1984, 355-377.

PAGE 153

143 Mullineaux, D., "On testing for Rationality: Another look at the Livingston Price Expectations Data," Journal of Political Economy, 86:1978, 329-336. Mundell, R., "Inflation and Real Interest," Journal of Political Economy," 71:1965, 280-283. Muth, J.F., "Rational Expectations and the Theory of Price Movements," Econometrica, 29:1961, 315-335. Pearce, D.K., "Comparing Survey and Rational Measures of Expected Inflation," Journal of Money, Credit and Banking, 11:1979, 447-456. Pearce, D.K., "Short-term Inflation Expectations: Evidence from a Monthly Survey," Journal of Money, Credit, and Banking, 19:1987, 388-395. Pesando, J.E., "A Note on the Rationality of the Livingston Price Expectations." Journal of Political Economy, 83:1975, 849-858. Ratcliff, R., Urban Land Economics, McGraw-Hill Book Company, Inc., New York, 1949. Rayburn, w., M. Devaney and R. Evans, "A Test of Weak-Form Efficiency in Real Estate Residential Returns," Journal of the American Real Estate and Urban Economics Association, 15:1987, 220-223. Roulac, S.E., "Influence of Capital Market Theory on Real Estate Returns and the Value of Economic Analysis," Real Estate Appraiser and Analyst, 44:1978, 62-71. Samuelson, P.A., "Proof that Properly Anticipated Prices Fluctuate Randomly," Industrial Management Review, 6:1965, 41-49. Schwert, G.W., "Effects of Model Specification on Tests for Unit Roots in Macroeconomic Data," Journal of Monetary Economics, 20:1987, 73-103. Skantz, T.R., and T.H. Strickland, "House Prices and a Flood Event: An Empirical Investigation of Market Efficiency," Journal of Real Estate Research, 2:1987, 75-83. Smith, K., and D. Shulman, "The Performance of Equity Real Estate Investment Trusts," Financial Analysts Journal, 32:1976, 61-66.

PAGE 154

Taylor, H., "Interest Rates: How Much Does Expected Inflation Matter?," Business Review Federal Reserve Bank of Philadelphia, July/August, 1982, 3-12. 144 Throop, A.W., "An Evalution of Alternative Measures of Expected Inflation," Economic Review, Summer, 1988, 2742. Tobin, J. "Money and Economic Growth," Econometrica, 33:1965, 671-784. Wendt, P., ands. Wong, "Investment Performance: Common Stocks vs Apartment Houses," Journal of Finance, 20:1965, 633-646.

PAGE 155

BIOGRAPHICAL SKETCH Dean Gatzlaff received Bachelor of Architecture and Bachelor of Environmental Design degrees from the University of Minnesota in 1979. After receiving bachelor degrees, he practiced architecture and became professionally registered in 1982. As a registered architect, he was responsible for the design and construction of over $40,000,000 worth of real estate development. In 1986 he received a Master in Business Administration degree from the University of Wisconsin Lacrosse, and entered the doctoral program in the Department of Finance, Real Estate, and Insurance at the University of Florida. His doctoral thesis was partially funded through competitive scholarships from the Urban Land Institute. He currently holds the position of Assistant Professor in the School of Business at the Florida State University, and is an active member of the American Real Estate and Urban Economics Association, the American Real Estate Society, the Urban Land Institute, and the American Institute of Architects. 145

PAGE 156

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor o~ Philosophy. ? / ~ v -!J. / ~.1,)vvf C. /v / l Halbert c. Smith, Chairman Professor of Finance, Insurance and Real Estate I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. /) ~ ,~ / ~ David c. Lin / Associate Profe or of Finance, Insurance and Real Estate I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. David A. Denslow Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Michael. Ryngae Assistant Professor of Finance, Insurance and Real Estate I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Cl C----Jf w. --<-' J Marc T. Smith Assistant Professor of Finance, Insurance and Real Estate

PAGE 157

This dissertation was submitted to the Graduate Faculty of the Department of Finance, Insurance and Real Estate in the College of Business Administration and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. December, 1990 Dean, Graduate School

PAGE 158

UNIVERSITY OF FLORIDA lllllllllllllllllllllllllllllllllllllll llllll 1111111111111111111 3 1262 08556 7245