Real Estate Investment and Portfolio Theory
Author(s): Harris C. Friedman
Source: The Journal of Financial and Quantitative Analysis, Vol. 6, No. 2 (Mar., 1971), pp.
861-874
Published by: Cambridge University Press on behalf of the University of Washington
School of Business Administration
Stable URL: https://www.jstor.org/stable/2329720
Accessed: 21-10-2018 06:12 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Cambridge University Press, University of Washington School of Business
Administration are collaborating with JSTOR to digitize, preserve and extend access to The
Journal of Financial and Quantitative Analysis
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
REAL ESTATE INWESTMENT AND PORTFOLIO THEORY
Harris C. Friedman*
,. Introduction
The purpose of this paper is to investigate the concept of mathematically
selected real estate portfolios and their relationships with mathematically se-
lected common stock portfolios. Markowitz [7, 8] demonstrated the desirability
of diversifying among common stock investments, but the theory has not been
extended to equity investment in real estate. In the past there have been suc-
cessful attempts made to use quantitative techniques to evaluate real estate in-
vestments on a per parcel basis, but there have been no real attempts to extend
these methods to a portfolio framework.
It is the intention of this study to show that, i.e., using reasonable
assumptions, the mathematical models used to select and evaluate common stock
portfolios can also be used for the selection and evaluation of real estate
portfolios. It will also be shown that these same models can also be used to
choose portfolios containing both real estate and common stocks.
The first section of the paper explains the methodology and models used in
this study. The second section identifies the data that was used and how it
was adjusted to fit into the framework of the specific models. The third section
deals with the empirical results, and in the final section implications for in-
vestors are discussed.
II. Methodology
Two similar models were used for the empirical tests performed in the study.
For identifying real estate or common stock efficient frontiers, the Sharpe [11]
diagonal model was used. To identify portfolios consisting of both assets, the
Cohen-Pogue [1] multi-index model was used.
When Markowitz [7] introduced his work, a primary hypothesis was that an
investor wanted to maximize return for a given level of risk or minimize risk for
a given level of return. This means that at each level of return or risk only
one set of stocks would satisfy the constraints. These portfolios are denoted
to be "efficient." At any level of return there is only one efficient portfolio
and at any level of risk there is only one efficient portfolio. It should be
noted that one efficient portfolio is not clearly better than any other efficient
portfolio. The investor's risk-return preference function determines which port-
folio is selected.
Assistant Director, Office of Economic Research, FederaZ Home Loan Bank
Board, and Zecturer, University of MaryZand. The author wishes to express his
appreciation to his dissertation committee with specific thanks to Keith V. Smith
and R. Bruce Ricks for their many heZpfuZ comments.
861
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
Markowitz developed an algorithm for identifying all efficient portfolios.
These various portfolios constitute what is called an efficient set. Each point
is obtainable and is not dominated by any other point. It is then up to the
investor to decide which one of the portfolios along the efficient set he prefers.
The curve representing all efficient portfolios is called the efficient frontier.
Each portfolio is identified by the percentage (x) of each stock in the portfolio.
A simplified method to obtain efficient sets was needed to increase the use
of the model. Sharpe [11, 12] expanded the Markowitz idea and created an algo-
rithm for performing the calculations for deriving efficient sets. The basic
portfolio relationships for N stocks are:
(1) R. = A. + B. I+C. i=l,.* ii N
where
(2) Ri = return on asset i
(2') I = random variable denoting the level of an index
= AN+l + CN+l and
(2") E(I) = AN+,
(2"') E(C N+1) = O
(2"") Var (I) = E(C N+) =Q
With assumptions concerning the covariance of the returns, the calculations
are simplified with the following results for portfolio return (P).
N+l
(3) E(P) = Z XiAi = A'X
i=l
N+l 2
(4) Var (P)= Z X Q X'O X
i=l ~ i N+l
where
(4') X,A,Q N+l have N+l elements and
(44") oN+l is a diagonal covariance matrix
where
Xi = proportion of wealth invested in security i
The full model is:
(5) Maximize XA'X-X' 0 N+X for all X.
862
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
Subject to:
(6) XN+l = Z X iB
i=l1
N
(6") Z X.= 1
i=l 111
(6"') X > 0, for all i, i=l . ,N
where
X is the "shadow price" of risk.
The Cohen-Pogue model used for the two asset case is a modification of the
Sharpe model, inwhich different indexes are used for different asset classes.1
The portfolio relationships are similar, but in the model the various indexes
are also related to a general economic index. With each of the models used, the
time horizon used for regressing past returns against the index as in equation
(1) was five years.
Since, except for purchasing shares in real estate investment trusts or
somehow adjusting his equity position, an investor cannot buy 62 percent or any
other percent of a real estate property. Real estate is an all-or-none decision
at the going market price. He can buy one, two or one thousand shares of common
stock, but he cannot buy real estate in the same manner, i.e., it is indivisible.
Since we are dealing with risk and return, this presents a quadratic programming
problem. To add the integer dimension greatly complicates the computations.
There may be an integer solution to quadratic programming problems,- even though
to date no algorithm has been written.
To condense this problem and make it more manageable, we have used a method
to simulate the "true" integer solution until such time as an efficient integer
code can be written. The modification procedure is based on the premise that the
model's most useful purpose is to identify the securities (assets) that should
be in the portfolio. It was assumed that in each corner portfolio, any property
(j) having a nonzero X value was purchased. New X values were calculated based
on the properties' values (Pj) and real estate's total share in the portfolio.
The new X value for property j is:
P. *he i I all real estate
i Pi. 1
i assets
alI
where Xa > L|
Wben we are discussing only real estate portfolios, EX. will always equal 1.0.
Figure 1 represents a short hypothetical example to portray the solution. It
might be argued that this modification procedure will actually distort the true
picture and lead to unsatisfactory portfolios. This conclusion is not supported
by the results to be presented. It is true that other methods may lead to dif-
ferent and possibly "better" portfolios, but this still must be proven.
1
The following indexes were used: for real estate, an average of the Boeckh
construction cost indexes for residences, apartments, hotels, and for commercial
construction and factories, and the American Appraisal Association index; for
common stocks-, Standard and Poor's Composite 500.
863
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
FIGURE I
HYPOTHETICAL PORTFOLIO MODIFI CATION
Asset X Price X (adjusted)
Property 1 .15 $100-- .25
Property 2 .30 $100-- .25
Property 3 .15 $ 50-- .12
Property 4 .25 $ 90-- .23
Property 5 .15 $ 60-- .15
1) Z P. = $100 + $100 + $50 + $90 + $60 = $400
all i
2) Z X = 1.00
all i
3) X (lo?)X 1.0 = .25
4) X2= 400)x 1.O = .25
5) X 3= 400
50)x 1.0 = .12
6) X4 ~_4OOI 1.0 = .23
7) X5 60 x 1.0 = .15
III. Sample
The real estate sample used in this study consisted of fifty properties and
came from two sources. The first source was the real estate investment portfolio
of the Pacific Mutual Life Insurance Company. In February 1969 the portfolio
consisted of forty properties, but due to continuity and the time horizons in-
volved, only thirty-three of the properties were used in this study. The remain-
ing seventeen properties were owned by savings and loan associations. The data
on individual properties was collected with the assistance of the Federal Home
Loan Bank Board. The real estate data consisted of property type, market values,
and yearly operating performance for the years 1963-1968 inclusive. Even though
the majority of the sample came from one portfolio, the geographic and property
type dispersion was very great. This fact can somewhat alleviate the problem
of nonrandomness.
The common stock sample consisted of a random sample of fifty stocks listed
on the NYSE and picked from Standard & Poor's Compustat Tape. The data obtained
consisted of the closing price (adjusted for stock splits and dividends) and the
annual dividend for the years 1963-1968 inclusive. The average returns on the
two samples are presented graphically in Figures II and III.
864
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
FIGURE II
DISTRIBUTION OF COMMON STOCK RETURNS
Number
-5 -0 -5 -1 1 2 2 3 3 4 4 5
O 5 O 5 O 5 O 5 O
Return (%)
FIGURE III
DISTRIBUTION OF REAL ESTATE RETURNS
Number
-5 -O -5 -1 _1 2 _2 _3 _3
O 5 O 5 O 5
Return (%)
865
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
The methods of evaluating a given property and the measurement of yield will
not be discussed at this time. These issues have been contested elsewhere, and
in this paper it will be assumed that the correct methods are used. A good
appraiser would be counted on to perform the same input estimation that a security
analyst now must perform for common stocks. The fine points of real estate in-
vestment analysis are not relevant to the specific subject, even though they are
important when estimating the input parameters. In this paper we use a method
closely related to the internal rate of return in order to measure yields. The
method is that of yearly relatives. These reduce all returns to a common
denominator. The computation for a common stock in year t is:
P D
~t-1
where
y
t = Return relative in year t
Pt = Stock price at end of year t
Dt = Dividends paid in year t
The expected return or relative is the average of the past five years' yield.
For real estate, dividends are replaced by net income.
It is argued here that risk in the real estate market can be tentatively
quantified just as risk in the stock market is. The measure used will be the
variance of return. It is not claimed that this is the best or even a good
measure, but for purposes of empirical testing of stock portfolios it has proven
adequate in the past. Risk in real estate may be quite different, but again for
the present this question will be relegated to the field of real estate invest-
ment analysis, which logically must precede portfolio selection. The variance
is measured as the variance of past yields.
Since yearly prices are necessary for the calculations and only two real
estate prices were known, a method was used to interpolate all intermediate real
estate prices based on the property values at the beginning of 1963 and at the
ending of 1968. The following methodology was used to obtain these intermediate
values:
(a) 6 P63 (l+g)
where P = price and g = growth rate
5
(b) (1+g) P
P63
(c) g = 6 1/58 1
631
(d) 63+t P63 (l+g)t
The compound growth rate was chosen because it was assumed that assets grow in
this fashion rather than on a straight line basis.
866
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
Another simplification was involved in generating the efficient portfolios
after taxes. In order to keep the two assets fairly comparable, the effects of
financing were disregarded. Because of this the only tax benefits accruing to
the real estate investments were in the form of depreciation. 2 The concept of
choosing real estate portfolios is unaltered by financing decisions, and the only
variation would have led to increased after-tax returns.
IV. Empirical Results
Single Asset Portfolios
For purposes of extending and implementing the theory in this paper, we only
identify efficient sets of portfolios. No attempt is made to choose the optimal
portfolio from this set, as this would require a discussion of utility theory
and the creation of utility curves.
The efficient curves in Figure IV represent real-estate efficient frontiers
before and after taxes and common stock frontiers before and after taxes. A tax
rate of 60 percent was chosen as representative of a high tax bracket investor.
The tax effects seen in the figure would not be as extreme for a lower tax
bracket investor.
As can be seen from the position of the various curves, on both a before-
tax basis and an after-tax basis, real estate portfolios dominate the stock
portfolios. These results seem to be unrealistic when the returns presented in
Figure II and III are compared. The rationale for these results will be dis-
cussed below. The stock portfolios are only more efficient in the high return
range of the curve because there were a few common stocks having higher returns
than any real estate assets. This implies that the general investor should hold
real estate portfolios rather than stock portfolios. Unfortunately, the sample
used in this study is not representative of the universe of real estate assets.
Also, the average investor does not have the capital needed to invest in most
large scale real-estate investments.
A major reason for the low risk associated with real estate portfolios is
the negative covariance among the assets. Stocks are traded in a national market
and tend to move together. Real estate is sold in a local market, and economic
conditions vary from region to region.
It should be noted that the two real-estate curves are much closer together
than are the common stock curves. This shows that taxes have more impact on com-
mon stock returns than they have on real estate returns. This fact is borne out
by the existence of many investors who use real estate as a tax shelter. Wendt
and Wong [15] show that the tax shelter for real estate represents 27.3 percent
of the return on total capital and 99.1 percent of the return on equity. These
high percentages show why the real estate curves are very close to each other.
Tax shelters such as depreciation are not available for common stock.
Ricks [10] has shown that depreciation and interest deductibility each ac-
count for an equal amount of the tax advantage.
867
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
One could argue that the only reason real estate dominates stocks is the
method used to interpolate real estate prices. This method tends to understate
variance. The same method was tried for common stocks, and the resulting effi-
cient curve shifted only slightly to the right of the one shown in Figure IV.
Also, part of the risk involved in holding common stocks arises because of daily
market fluctuations. Real estate is a longer term investment and, therefore, is
not subject to daily price changes. Tests were also performed by decreasing the
sale value of real estate by 15 percent to allow for thin markets. The composi-
tion and position of the curve shifts only slightly when this adjustment is made.
Given the available data, this brief exercise has shown that real estate
portfolios can theoretically be chosen using the same methods currently used for
common stocks. The practical problem involved is that of obtaining adequate
yield projections for real estate. This problem should be no more or less dif-
ficult than the corresponding problem for common stocks, even though the in-
formation is less readily available in the real estate field.
Mixed Portfolios
Once the mechanism was developed for selecting real estate portfolios, the
next logical step was to combine the two assets into somewhat complete invest-
ment portfolios. This meant chosing a set of portfolios from among a group of
common stocks and real estate assets. To perform the calculations, a multi-index
model was used.
The results are presented in Figure V and Tables 1 and 2. The most interest-
ing phenomenon is that, while the highest return portfolios are made up of com-
mon stocks, once real estate assets are chosen they quickly become the dominant
asset in the portfolios. On an after-tax basis this occurs even sooner than it
does before taxes. The reasons for these results are synonomous with the ones
previously mentioned that explain both the tax effects of real estate and the
reasons why the real estate curve dominated the common stock curve.
It is interesting to note in Figure V that the spread between the curves
before and after taxes is much greater at the higher end of the curve than at
the lower portion. This again can be explained by the entrance and dominance of
real estate in the lower return portfolios.
It should also be noted that the standard deviation of these mixed port-
folios quickly falls to a level approaching zero. This shows that the negative
covariance between real estate and common stocks greatly reduces the total port-
folio risk.
A problem arises with the selection method in that, near the lowest return
portfolios, almost all of the assets are selected with very small percentages.
This presents a problem in that no one would want to buy three shares of stock to
reduce his portfolio risk. This implies the necessity of minimum value con-
straints or the selection of portfolios nearer the high return segment of the
curve.
V. Implications
Financial Intermediaries
The empirical results, combined with the current status of investment port-
folios, suggest that financial institutions are underrating the value of investing
868
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
FIGURE IV
SINGLE ASSET EFFICIENT FRONTIERS
20
Stock
(after tax)
15
Stock
(before tax)
Standard 10
Deviation (%)
>% SrReal Estate
5 / (before tax)
Real Estate
(after tax)
0 10 20 30 40
Returns (%)
869
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
FIGURE V
EFFICIENT FRONTIERS
20
15
Standard 10 _ 60% Tax Rate
Deviation (%)
Before Taxes
l-. I I I I
0 10 20 30 40
Returns (%)
870
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
TABLE 1
CHARACTERISTICS AND COMPOSITION OF
MIXED-ASSET PORTFOLIOS
No Taxes
Selected Expected Standard
Corner Portfolio Deviation
Portfolios Return of Return Common Real
Stock Estate
1 48.2 36.6 100.0 (1)
10 41.0 22.6 100.0 (1)
20 36.4 15.1 88.2 (19) 11.8 (1)
30 30.8 11.0 60.1 (28) 39.9 (2)
40 26.6 8.6 42.3 (35) 57.7 (5)
50 15.9 2.2 11.8 (39) 88.4 (10)
60 14.2 1.8 10.8 (41) 90.0 (18)
70 13.7 1.8 8.6 (45) 91.4 (23)
80 12.0 1.4 7.6 (46) 92.4 (31)
90 11.1 1.2 7.2 (47) 92.8 (36)
100 10.6 1.1 5.9 (47) 94.1 (38)
120 10.1 1.0 5.1 (49) 94.9 (38)
140 8.9 0.9 3.0 (49) 97.0 (42)
159 5.9 0.8 3.0 (50) 97.0 (43)
indicates number of assets selected.
871
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
TABLE 2
CHARACTERISTICS AND COMPOSITION OF
MIXED-ASSET PORTFOLIOS
60% Tax Rate
Selected Expected Standard
Corner Portfolio Deviation
Portfolios Return Of Return Common Real
Stock Estate
1 34.7 29.7 100.0 (1)
10 29.1 20.9 100.0 (10)
20 26.1 17.8 93.2 (19) 6.8 (1)
30 22.0 11.3 64.9 (28) 35.1 (2)
40 19.3 8.5 49.2 (37) 50.8 (4)
50 12.2 4.3 25.8 (38) 74.2 (12)
60 8.3 1.4 7.4 (42) 92.6 (18)
70 7.4 0.8 4.2 (45) 95.8 (25)
80 7.0 0.7 2.6 (46) 97.4 (32)
90 6.6 0.6 2.8 (47) 97.2 (36)
100 6.4 0.6 6.9 (48) 93.1 (38)
120 5.8 0.5 9.3 (49) 90.7 (39)
140 4.7 0.5 9.3 (50) 90.7 (40)
151 2.5 0.5 6.2 (50) 93.8 (35)
indicates number of assets selected.
872
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
in real estate. It is impossible to say whether this stems from a lack of know-
ledge about real estate or a psychological prejudice against this type of in-
vestment. If the institution has a long-term portion of its portfolio invested
in bonds and mortgages, there is no theoretical reason why real estate cannot be
included. It is reasonable to assume that a property will be marketable some
time within a thirty-year period.
Intermediaries such as mutual funds and investment trusts do not hold a
variety of asset groups. The managers of these institutions should note that
they may not be receiving the maximum benefits of diversification and the
covariability of asset return. Of course there are legal constraints to invest-
ing in other assets, but generally a manager does have some discretionary power.
Individuals
Individual investment portfolios present the biggest challenge to research-
ers. There is much less known about these portfolios than is known about in-
stitutional portfolios. It may be argued that the costs involved in using a
model of the type presented here would be prohibitive to the individual. This
would not be the case if investment service bureaus prepared the input data for
everyone. An individual could then select his universe, decide on constraints,
and finally use the model for a very small cost. Once the efficient set was
generated, it would be up to the investor to posit his utility curve and choose
the correct portfolio.
Since, with a small portfolio, the transactions costs may become prohibi-
tive, the individual investor may want to place lower-bound constraints on the
assets selected. This would enable him to limit the number of assets in his
portfolio. Of course, there is the single valued problem of real estate, but
the individual can circumvent this by buying shares in a real estate investment
trust. He can examine the portfolios of many investment trusts to obtain the
approximate return-risk level for real estate that is desired.
VI. Summary
This paper has shown that the models developed to select common stock port-
folios can be adapted to the selection of real estate portfolios and mixed asset
portfolios. The concepts are all identical, and as long as return and risk can
be quantified, the problems are soluble.
The portfolios identified using a small sample indicate that real estate
portfolios can have more return and less risk than do common stock portfolios.
When the two assets are combined, the real estate assets dominate the resultant
portfolios. On an after-tax basis these results are more apparent. The local
aspect of real estate versus the national aspect of common stocks is primarily
responsible for these results.
873
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
REFERENCES
[1] Cohen, K. J., and J. A. Pogue, "An Empirical Evaluation of Alternative
Portfolio-Selection Models," JournaZ of Business (April 1967), pp. 166-193.
[2] Friedman, H., "Portfolio Approach to Real Estate Investment," unpublished
manuscript (Los Angeles: University of California, 1968).
[3] Friedman, H., "Mixed Asset Portfolio Selection," unpublished Ph.D. dis-
sertation (Los Angeles: University of California, 1970).
[4] Hastie, K. L., "The Determination of Optimal Investment Policy,"
Management Science (August 1967), pp. 757-774.
[5] Hayes, S. L.,and L. M. Harlan, "Real Estate as a Corporate Investment,"
Harvard Business Review (July-August 1967), pp. 144-160.
[6] Hoagland, H. W.,and L. D. Stone, Real Estate Finance, (Homewood, Illinois:
Richard D. Irwin, 1965).
[7] Markowitz, H. M., "Portfolio Selection," Journal of Finance, (March 1952),
pp. 77-91.
[8] Markowitz, H. M., "Portfolio Selection, Efficient Diversification of
Investments,"Cowles Foundation Monograph No. 16 (New York: Wiley, 1959).
[9] Ricks, R. B., Recent Trends in InstitutionaZ ReaZ Estate Investment,
Research Report 23, Center for Real Estate and Urban Economics (Berkeley,
California; University of California, 1964).
[10] Ricks, R. B., "Inputed Returns on Real Estate Financed with Life Insurance
Company Loans," JournaZ of Finance (December 1969), pp. 921-937.
[11] Sharpe, W. F., "A Simplified Model for Portfolio Analysis," Management
Science (January 1963), pp. 277-293.
[12] Sharpe, W. F., "Addendum to 'A Simplified Model for Portfolio Analysis',"
Management Science (April 1963).
[13] Sharpe, W. F., "Capital Asset Prices: A Theory of Market Equilibrium Under
Conditions of Risk," JournaZ of Finance (September 1964), pp. 425-442.
[14] Smith, K. V., "Stock Price and Economic Indexes for Generating Efficient
Portfolios," JournaZ of Business (July 1969), pp. 326-336.
[15] Wendt, P. F.5and S. N. Wong,"Investment Performance: Common Stocks Versus
Apartment Houses," JournaZ of Finance (December 1965), pp. 633-646.
874
This content downloaded from 198.91.37.2 on Sun, 21 Oct 2018 06:12:39 UTC
All use subject to https://about.jstor.org/terms
