REAL EsrATE DIVESITIURES AND STOCK. PRICE PERFORMANCE:
USING .ALWRNATIVE EVENT STUDY VARIANCE EsTIMATORS

Thomas H. Strickland
Middle Tennessee State University

Murfreesboro, Tennessee

Edward L. Buhnys
Suffolk University

Boston, Massachusetts

Chen-Chin Chu
Memphis State University

Memphis, Tennessee

Donald H. Wort
California State University - Hayward

Hayward, California

Introduction

The question of pricing efficiency in securities markets is an important research topic
that generates considerable interest. However, while many researchers have examined the
general question of market efficiency, few have concentrated on the behavior of real estate
markets. Recent empirical work has been done by Shilling, Sirmans and Benjamin [11],
Guntermann and Smith [6], Gau ([4], [5]), and Skantz and Strickland [12]. Of particular
interest to this research is a study by Rite, Owers and Rogers (BOR) [7] who examined
the level and patterns of abnormal returns for firms separating real estate assets from their
remaining operations through spinoffs.

One of the questions addressed by HOR was related to the claim by many managers
of large publicly traded real estate :firms that certain accounting restrictions cause their
shares to be systemically undervalued in the stock market [8]. While HOR were unable
to determine if real estate assets were really undervalued, they did find that the asset
restructuring resulting from spinoff of real estate operations tended to increase the firms'
aggregate market values.

The objective of this study is to investigate the impact of a firm's decision to divest
real estate assets. The decision's impact is viewed from the perspective of the firm's
owner, i.e., change in shareholder value. The study includes a sample of forty-three
divestitures between 1971 and 1984. Event study methodology employed in this study
is widely used in accounting and finance research and is becoming more common in the
management literature [15].

The paper is outlined as follows. Section two discusses issues and motivations pertaining
to real estate divestitmes. Section three explain'i the event study methodology, describes the
data sample, and compares the two variance estimators used in the study. Section four de-
scnbes empirical results. A summary and conclusions are presented in the last section.



28 Journal ofBusiness Strategies Vol. S, No.2

Background and Motivations for Selling Real Estate Holdings

While this study can neither value individual assets held by public firms, nor ascertain
the exact selling prices for a majority of the transactions, the net gains/losses experienced
by firms divesting real estate assets are analyzed by using security market price data. This
methodology is explained in the next sectiQn. A brief discussion of why firms might
dispose of real estate assets follows.

A firm's stock price response to the sale of a large real estate asset may be closely
related to the reason for the sale. For example, if a firm is in financial difficulty its
bargaining position for selling assets of any type may be impaired.

The firm may wish to eam an accounting profit thus increasing reported earnings per
share. Some firms selling real estate for this reason may be be experiencing financial
difficulties and are attempting to avoid disastrous drops in reported earnings. Also, the
finn may be in need of cash for financing other projects when alternative financing sources
are limited. These reasons could explain "forced sales" at prices below market

Real estate sales may occur if previous plans for asset use have changed. Perhaps the
assets have lost strategic value to the firm. Dispositions may also be due to the mar-
ginal tax benefits between the sellers and acquirers.

likewise the sale may be an attempt by the firm to recognize the asset's "true value."
Several real estate firms, in particular, have given this reason. They felt their holdings
were not properly valued by analysts and investors in the firms' securities [8]. If this
perception is correct it might imply market inefficiency because investors and analysts
are ignoring factors that become clear when the real estate asset is isolated. This may
be analogous to a company take-over situation, where the break-up value may be greater
than that of the firm as a whole.

The actual reasons for selling real estate assets, however, are difficult to determine be-
cause the real motivations for asset disposition are often clouded by managements' vague
and general statements. This research, therefore, tests only for the aggregate divestiture-
related impact on shareholder wealth.

Methodology, Data and Experimental Design

The ideal approach to analyze real estate disposition decisions would be to compare
asset selling prices to pre-disposition ''fair market values." Unfortunately, this direct ap-
proach is precluded by one of the salient features of real estate markets - the dearth of
publicly available :financial data. With the exception of single-family residential homes,
sales prices are extremely difficult and expensive, if not impossible, to obtain. Since
neither sales prices nor ''fair market values" can be directly observed for these real es-
tate assets, another approach is required. This investigation compares the :firm's market
value before the sale with its post-divestiture value in order to measure shareholder gain
or loss on each transaction. The methodology described below does just this.

Event study methodology, similar to that first used by Fama, Fisher, Jensen and Roll
[3], measures the impact of a specific management decision on security market returns.



Spring 1991 Strickland, Bubnys, Chen-Chin & Wort: Real Estate 29

The management decijion, Le., "event," in this study is the divestiture or spinoff of a
firm's real estate assets.

Suppose the research question is, for example, to measure the single, one-day effect of
an event. Three major steps are required for each firm in the sample. First, the actual
return for a firm's stock is observed on the event day. Second, by utilizing the familiar
security market "Beta" model, a prediction is made of what the stock's "normal" return
should have been had the event not occurred. Third, the impact of the event, defined as
abnormal return, is the difference between the stock's actual market return - presum-
ably affected by the event's occurence - and its "normal" return. This difference, if
any, measures the event's one-day impact on stockholder wealth. The Beta model and
abnormal return are formally defined below in Equations (1) and (2), respectively.

Daily security return data from the Center for Research in Security Prices (CRSP) were
obtained for the period 1963 through 1984. Value-weighted returns including dividends
for all NYSE and AMEX stocks were used as the market index proxy.

The final data set of 43 transactions was selected from an original list of approximately
110 firms obtained from several sources. These included the National Newspaper Index,
the Predicast Index, and the Wall Street Journal Index. The underlying assumption is
that an event is material for a firm if it is reported and indexed in a major financial
newspaper such as The Wall Street Journal, The New York Times, or The Los Angeles
Times. The event date is defined as the time when the transaction's information was
first made publicly available. This was determined in one of several ways. A newspa-
per index reference describing a firm's intent to enter into the transaction was considered
the event. Otherwise the related article was used to verify the event. When a reference
alluded to an earlier event, for example "...the firm had a gain of $ X million on sale of
land ...", we looked for an earlier reference in the index. Finally, when several related
references were discovered, a thorough search was conducted to determine whether the
transaction had been made or not. If an actual event date could not be verified the
potential event was not used. Table 1 lists information about each of the 43 events in-
cluded in the final sample. The finn's name is followed by its Compustat industry number
(approximately equivalent to 4-digit SIC categories) and calendar date of the event.

The return data for each divestiture was divided into three major periods based on event
date: a Beta estimation period of 120 days, a settling period of 90 days, and an analysis
period of 181 days. A total of 391 daily return observations are used for each event
with day t = 0 designated as the event date.

Data from the 120-day estimation period were used to estimate each firm's Beta model.
The model is,

A A A

Rt,1 = Uj + ()iRm,I' (1)

A

where values of t range from -300 to -181 days relative to the event date; Rt,1 is the
expe'%,ted re~rn for security i on day t; Rut,1 is the return on the market index on day t;
and Uj and ()i are OLS coefficients from the estimation period.



30 Journal ofBusiness Strategies Vol. 8, No.2

The next 90 days of returns (day -180 through day -91) are excluded from the study
to allow for a settling-down period between the Beta estimation and analysis periods. This
is consistent with Reints and Vandenberg [9] and Sanger and McConnell [10], who sug-
gest excluding time ~riods in which information concerning the event might influence
the estimation of the <ij and ~j coefficients.

The results are actually evaluated for the analysis period. For each finn i, and for
each day t, each day's impact due to the event is the difference between the actual re-
turn and the "normal" return. This difference is defined as the abnormal return,

A

A., = R t - R,~ '1, 1, 1, ' (2)

where t ranges 181 days from -90 to +90; ~ t is the security's actual daily return; and
Ri,t is its "normal" return. "Normal" return ~ predicted using the estimators aj and ~i
from the Beta model described in Eq. (1) in conjunction with the actual returns of the
market index, Rm,t, during the analysis period. For each day in the analysis period, ab-
normal returns are averaged over all firms in the sample to obtain event-related impact
results on a day by day basis. When the results for a cumulative effect of more than
one day are desired, daily abnormal returns are summed over the relevant time window
and designated Cumulative Abnormal Returns (CARs).

The t value statistics described by Brown and Warner (BW) [1], (equations 5,6,7,8 and
All) are calculated in two different ways. One uses the traditional variance estimator
and one uses an alternative method. Following the BW equation 7, the standard error
measure is the square root of the sum of squared excess return deviations over the 120-
day estimation period. This provides the traditional variance estimator.

The alternative variance estimator is Tiku's [14] MMLE, which censors the smallest
and largest values from the excess return deviations. In effect, the estimation period's
series of 120 deviations (BWs equations 7 and 8) are arranged in order and the 10%
lowest and highest deviations are dropped. The remaining 96 deviations are used in
calculating the robust MMLE estimate of variance and the corresponding MMLE t val-
ues.

The MMLE standard deviation a is defined as:

a = {B + (B2 - 4AC)5} / 2(A(A-1)}5
where

A = n - 2, B = L(~-1 - X2)
and

n-1
C = L X2i + (X22 + X2n-1) - 8 Jl2 ,

i = 2

(3)

(4)

where Xi is the deviation of mean excess returns for period i, (X2 and -"0-1 being the
new lowest and highest deviations, respectively, after MMLE censoring), and !!,L and e
are given in Tiku, Tan and Balakrishnan [14 pp. 74-75]. Tiku [13] investigated the ef-



Spring 1991 striclcland, Bubnys, Chen-Chin & Wort: Real Estate 31

ficiencies of MMLE's mean and variance for location and scale parameters of symmetric
non-normal distributions. He found that the MMIE parameters are more efficient than
other more prominent robust estimators. Also, the MMLE meao/variance statistia; are
explicit and simple functions of sample observations and are easy to compute.

am, Bubnys, and lee [2] used simulation analysis to investigate the robustness of the
MMI.E for samples of daily stock returns. Their results show that the MMLE and tta-
ditional estimators yield similar estimates of stock market risk premium. However, the
MMLE method produces a smaller variation over time because the censoring method
reduces the sensitivity of the MMLE to extreme daily market returns. Results using these
two different t value calculation methods are described in the next section.

Empirical Results

Selected results for the total sample of forty-three events are presented in Tables 2 and
3. Table 2 shows daily abnormal returns and corresponding t values for individual days
-to through +10 using the traditional and MMLE methods while Table 3 presents the
same information for different cumulative time windows.

Table 2 shows that the daily abnormal return for day -1 is +2.01%. Traditional and
MMLE t values are, respectively, 45940 and 4.3685, both significant at the .001 level
Test statistic values are also significant for both estimators in day 2 after the event.
Interestingly the abnormal returns are not significant in day 0 or +1 after the event, nor
are they of consequence during any other single day shown. The higher MMLE Vari-
ance estimate is reflected in MMLE t values, which are about 5 percent lower than the
traditional t values.

QmmIative abmrmal return results presented in Table 3 indicate that CARs are not signifi-
cantly different from uro during the period from day -90 to two days before the event (t
values are 1.3061 and 1.2417). The event apparently was not anticipated by the market.

There is a positive price response, however, for cumulative time windows starting at
day -1. For the time window from -1 to +90 days, the event-related CARs are +17.64%,
significant at the 0.001 level. Sample firms did well for approximately 90 trading days
after the event. Other time windows starting at day -1 show positive CARs and signifi-
cance levels at least at the 0.05 level. For the 4-day window, from -1 to +2, share prices
increased by 35%, significant at the 0.001 level. These results suggest that the impact
of real estate divestitures on stock prices was positive.

The effect of using the robust MMLE standard deviations was to reduce the test sta-
tistic values compared to the traditional approach. However, the t values were virtually
the same. The MMLE standard deviation for the estimation period was 0.004605 (from
Eq. (3», about 5% larger than the traditional estimate of 0.004378. Nevertheless, use of
the MMLE approach may be a correct response to one of Brown and Warner's [1] points
about variance increases during the event period.



32 Journal ofBusiness Strategies Vol. 8, No.2

Table 1

Sample Firm Names, Industry Number, and Events

Industry Event Date
Finn Name Number MmlOl UBI :Im

u.s. Steel 2911 02 17 83
First Union Republic 6798 02 27 81
Oxford First 6150 05 28 80
San Diego Gas & Eectric 4931 07 01 80
American Standard Inc. 3580 06 23 71
Newhall 6552 01 06 72
Buttes Gas & Oil 1311 01 03 79
Delmarva Power Company 4931 08 05 83
FaiJChild Industries 3721 12 27 83
AT&T 4811 05 18 83
Arlen Corp. 6512 08 22 79
Citicorp 6025 07 05 83
Fluor 1600 06 29 83
Texas Pacific Light 6798 06 21 83
American Century 6199 05 17 83
Martin Marietta 3760 05 04 83
Koger Properties Inc. 6552 04 18 83
First City Properties 6552 04 05 83
Sears 5311 12 20 82
AMOCO 2911 04 20 82
Playboy 2721 03 25 82
RF. Ahmanson 6120 11 05 81
Pearce Urstadt 6199 03 19 81
U.S. Home Corp. 6552 01 09 81
Del E. Webb 1540 05 31 83
Western Union 4890 09 25 81
City Investing 6199 10 23 80
Loews 6199 03 14 83
Oxford First 6150 08 29 79
Bay Financial 6513 05 16 83
First Union Republic 6798 02 10 83
Aetna Life Insurance 6312 12 31 82
Federal Realty Investment 6798 12 09 82
Aluminum Co. of America 3330 01 04 71
McGtaw-Hill 2731 10 12 70
Holiday Inn 7011 01 06 71
Otrysler 3711 08 23 79
General Growth 6798 08 22 79
General Motors 3711 01 05 82
Horizon Corp. 6532 04 15 83
Arlen Corp. 1700 05 25 84
Dunlop 3000 11 26 84
Aetna Ufe Insurance 6312 11 13 84



Spring 1991 Strickland, Bubnys, Chen-Chin & Wort: Real Estate

Table 2

Abnormal Daily Returns and T Values for
Traditional and MMLE Variance Estimators

33

Abnonnal
Return

Traditional MMLE
Signtt Signtt

T Value l&n:l T Yalue Level

-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
o
1
2
3
4
5
6
7
8
9

10

0.0077
-0.0012
-0.0039
0.0026

-0.0044
0.0012

-0.0052
-0.0032
-0.0043
0.0201
0.0025

-0.0007
0.0131

-0.0070
0.0003

-0.0031
0.0061
0.0009
0.0020

-0.0007
-0.0015

1.7616
-0.2656
-0.8898
0.5931

-0.9944
0.2642

-1.1847
-0.7324
-0.9860
4.5940 ***
0.5778

-0.1649
2.9958 ***

-1.5996
0.0605

-0.7178
1.3822
0.2113
0.4488

-0.1590
-0.3410

1.6751
-0.2526
-0.8461
0.5640

-0.9456
0.2513

-1.1265
-0.6965
-0.9375

4.3685 ***
0.5495

-0.1568
2.8487 ***

-1.5211
0.0576

-0.6826
1.3143
0.2009
0.4267

-0.1512
-0.3242

lSignificance levels:
.. = 0.05; ** = 0.01; *.. = .001

Conclusions

The main question addressed in this study is whether or not the divestiture of substan-
tial real estate assets by publicly traded firms has a significant impact on shareholder value.
Forty-three divestitures over a fourteen year period are analyzed using event study meth-
odology.

The results show that shareholder wealth increases by economically and statistically sig-
nificant amounts. In a four-day period surrounding the event date, for example, the event-
related shareholder wealth increase was 3.5%, significant at the 0.001 level. Although
no definitive conclusions regarding real estate market efficiency or variance estimator
improvement can be reached, the results do provide some insight into effects ofreal es-
tate divestitures.



34 Journal ofBusiness Strategies

Table 3

Cumulative Abnormal Returns and T Values for
Traditional and MMLE Variance Estimators

Vol. 8, No.2

Cumulative Traditional MMLE
Time- Cumulative

Window Abnonnal Signfi Sigut"-
In Days Returns T Value Level T Value Level

-90 to -2 0.0539 1.3061 1.2417

-1 to +90 0.1764 4.1998 *** 3.9927 ***

-1 to 0 0.0226 3.6575 *** 3.4772 ***

-1 to +1 0.0219 2.8911 ** 2.7486 **

-1 to +2 0.0350 4.0018 *** 3.8046 ***

-1 to +3 0.0280 2.8639 *** 2.7227 ***

-1 to +8 0.0341 2.4631 * 2.3417 *

tSignificance levels:
• = 0.05; ** = 0.01; *** = .001

While results in this study strictly hold for the sample used and the time period ana-
lyzed, management's ability to dispose of property at relatively favorable prices increased
total wealth of shareholders. The question of presumably higher real estate project val-
ues to the buyer than to the seller, while interesting, is beyond the scope of this research.
It can be the focus of further research.

References

1. Brown, S.J. and Warner, J.B. "Using Daily Stock Returns: The Case of Event Studies."
Journal of Financial Economics. Vol. 14 (1985), pp. 3-31.

2. Chu, c.c., Bubnys, E.L. and Lee, C.F. "Estimates of the Expected Market Risk Premium:
Empirical Analysis." Working Paper presented at the Joint National Meeting of ORSA/TIMS,
(October 1987).

3. Farna, E.F., Fisher, L., Jensen, M. and Roll, R. "The Adjustment of Stock Prices to New
Information." International Economic Review, Vol. 10 (February 1969), pp. 1-21.



Spring 1991 Strickland, Bubnys, Chen-Chin & Wort: Real Estate 35

4. Gau, G.W. "Public Information and Abnormal Returns in Real Estate Investment." AREUEA
Journal Vol. 13 (1985), pp. 15-31.

5. Gao, G.W. ''Weak Form Tests of the Efficiency of Real Estate Investment Markets." The
Financial Review. Vol. 19 (1984), pp. 301-320.

6. Guntermann, K.L and Smith, R.L. «Efficjency of the Market for Residential Real Estate."
Land Economics, Vol. 63, No.1 (February, 1987), pp. 34-45.

7. Hite, G., Owers, J. and Rogers, R. «The Separation of Real Estate Operations by Spin-Off."
AREUEA Journal Vol. 12 (1984), pp. 318-332.

8. Palmon, D. and Sied1e, LJ. «Current Value Reporting of Real Estate Companies and a Pos-
sible Example of Market Inefficiency." The Accounting Review, Vol. 53 (1978), pp' 776-790.

9. Reints, W.W. and Vandenberg, PA "The Impact of Changes in Trading Location of a
Security's Systematic Risk." Journal of Financial and Quantitative Analysis. Vol. 10 (1975),
pp. 881.-889.

10. Sanger, G.C. and McConnell, JJ. «Stock Exchange Ustings, Firm Value, and Security Mar-
ket Efficiency: The Impact of NASDAQ." Journal of Financial and Ouantitative Analysis,
Vol. 21 (1986), pp. 1-25.

11. Shilling, J.D., Sirmans, C.F. and Benjamin, J.D. "Flood Insurance, Wealth Redistribution and
Urban Property Value." Working Paper No. 401, Louisiana State UnivelSity, (August 1984).

12. Skantz, T.R. and Strickland, T.H. "House Prices and a Flood Event: An Empirical Investi-
gation of Market Efficiency." Journal of Real Estate Researc~ Vol. 2 (1987), pp. 75-83.

13. Ttku, M.L. "Robustness of MMLE Estimators Based on Censored Samples and Robust Test
Statistics." Journal of Statistical P1annini and Inference, Vol. 4 (1980), pp. 123-143.

14. Tiku, M.L, Tan, W.Y. and Balakrishnan, N. Robust Inference. New York Uty: Marcel
Dekker, Inc. (1986).

15. Worrell, D.L., Davidson, W.M., Chandy, P.R. and Garrison, S.L "Management Turnover
Through Deaths of Key Executives: Effects on Investor Wealth." Academy of Management
Journal Vol. 29 (1986), pp. 674-694.


	Real Estate Divestitures and Stock Price Performance Using Alternative Event Study Variance Estimators