real estate investments, product market competition, and ... · this study approaches real estate...
TRANSCRIPT
Real Estate Investments, Product Market
Competition, and Stock Returns
Moussa DiopUniversity of Wisconsin-Madison
April 16, 2017
Abstract
By limiting operating flexibility, real estate investments are found to increase firm risk, thus
expected returns. This study introduces product market competition as a critical determinant
of the relation between real estate investments and stock returns. As part of capacity strate-
gies, these investments are generally associated with increased market power and lower cash
flow volatility in oligopolistic industries. I present a simple model of oligopolistic competition
showing a negative relation between real estate holdings and firm beta, and empirically con-
firm this prediction. Controlling for product market competition enhances identification of the
endogenous relation between real estate investments and stock returns.
JEL Classification: G11, G31, L11, L13
Keywords: Real Estate Investments, Stock Returns, Product Market Competition, Return Volatil-
ity, Oligopolistic Competition
Introduction
“The company generally owns the land and building or secures long-term leasesfor restaurant sites, which ensures long-term occupancy rights and helps controlrelated costs... The company identifies and develops sites that offer convenience tocustomers and long-term sales and profit potential.” McDonald’s Corporation
As this quote from McDonald Corporation’s 2011 annual report shows, securing long-term rights
to real estate may be of strategic importance for some firms, particularly those operating in real
estate intensive industries.1 The amount and type of real estate a firm owns reflect industry and
firm characteristics as well as location and general market conditions. This study explores the
impact of product market competition on the relation between real estate investments and stock
returns. More specifically, I show that pricing power derived from capital (real estate) investments
in oligopolistic industries reduces cash flow volatility, resulting in a negative relation between real
estate and stock returns.2 This study complements and extends Tuzel (2010), who does not consider
industry structure, by highlighting a potential benefit from real estate investments in oligopolistic
industries. This study assumes industry structure as exogenous. In contrast, Ambrose et al. (2016)
examine the endogenous interaction between capital investments and industry structure under
demand uncertainty, and incidentally note the possible effect of demand volatility on the volatility
of firm value as industry concentation increases.
Capital investments generally involve a sizable real estate component. By altering financial and
operating leverage, growth options, and a firm’s ability to benefit from positive economic shocks,
these investments certainly affect stock returns (Berk et al. (1999); Carlson et al. (2004); Cooper
(2006); Kogan and Papanikolaou (2013)). As shown by Tuzel (2010), frictions in capital adjustment
mechanisms, such as limited investment reversibility, may prevent firms from maintaining optimal
levels of capital stocks in the event of adverse market shocks, hence leading to a positive relation
between capital assets (real estate) and stock returns.3 Though silent about firm risk, the industrial
1Firm specificity of capital investments tied to the real estate is also a key determinant of the length and type oftenure that management decides to maintain on that real estate (Fisher (2004)).
2In this study, competitive industries refer to industries where firms are price takers. In contrast, firms operatingin oligopolistic industries enjoy some pricing power. The terms oligopolistic industries (oligopolies) and concentratedindustries have the sane meaning and are used interchangeably in this paper.
3The real estate literature reviewed later has also explored with various degrees of success the relation betweenreal estate ownership and stock returns (Deng and Gyourko (1999)), systematic risk (Brounen and Eichholtz (2005)),production uncertainty (Zhao and Sing (2015)), or demand uncertainty ( Ambrose et al. (2016)).
1
organization literature has extensively studied and identified benefits from capital investments for
oligopolies in terms of pricing power, market protection, and cash flow stability (e.g., Wenders
(1971); Spence (1977); Eaton and Lipsey (1980)). In summary, the net effect of capital investments,
including real estate, on firm risk likely depends on market structure. I argue that the relation
between real estate (capital) investments and stock returns is positive in competitive industries, but
non-positive in oligopolistic industries due to stronger pricing power and lower cash flow volatility
derived from these investments.
The real estate owned by non-real estate firms generally consists of real property assets used in
their production activities.4 Production and marketing strategies normally drive these investments,
which affect operational and financial risks. Despite its bulkiness, illiquidity, limited reversibility
in the short term, and large asymmetric adjustment costs, real estate represents roughly 28% of
corporate balance sheets in 2010 (Ambrose et al. (2016)), consisting largely of production facilities,
warehouses, office buildings, land, and retail outlets. Figure 1 shows that the real estate intensity
of corporations from 1970 to 2010. Given these undesirable attributes of real estate that increase
firm risk, why do firms still own a considerable amount of real estate?
This study approaches real estate as a factor of production, but a factor different from most
inputs since its short-term inflexibility makes it hard for firms to maintain an optimal level in
response to demand shocks.5 I argue that product market competition is a determining factor in
the type and amount of real estate that firms own. Capacity and output decisions are important
strategic tools available to firms operating in oligopolistic markets. Moreover, capacity expansions
generally involve a substantial real estate component. Dixit and Pindyck (1994) note that real
estate investments provide firms with options to grow production. But as noted by Grenadier
(2002) and Novy-Marx (2007), competition erodes and may even eliminate the value of growth
options. In competitive industries, capacity-increasing investments are likely to yield minimal
economic benefits since firms have little to no market power. Capacity investments will generally
increase cash flow, unless the additional capacity is left idle for strategic reasons. However, it is
4Non-real estate firms are firms whose primary business is not directly related to real estate development, invest-ment, management, or financing. The vast majority of these firms acquire or lease real estate to secure an inputresource (similar to labor or equipment) to meet future production objectives.
5It is recognized that some firms may approach real estate investments from a wealth maximization perspective.This dimension of corporate real estate investments is not addressed in this study.
2
unlikely that capacity investments lower cash flow volatility in competitive markets since all firms
are price takers. Consequently, firms undertaking real estate investments in competitive industries
will be burdened by the undesirable attributes of real estate and their inability to affect output
prices will make them more vulnerable to adverse market shocks. As a result, investors will view
these firms as riskier and will therefore require higher returns.6
In contrast, the effect of real estate investments on firm risk in concentrated (oligopolistic) indus-
tries is somewhat more favorable. In this operating environment, capacity investments aimed at
creating, consolidating, or protecting economic rents (i.e., profit margins) may stabilize cash flows.
Capacity investments may also lead to stronger product market presence and enhanced bargaining
power relative to suppliers. Peress (2010) shows that firms can use their market power to pass on
shocks to customers, thereby insulating profits. Thus, capacity investments may act as a hedge
against negative supply shocks by allowing firms to transfer these shocks to customers to the extent
permitted by the elasticity of demand or to negotiate more favorable pricing terms from suppliers
than smaller competitors can. Again, these strategies would only succeed in oligopolistic industries
since competition erodes market power. Consequently, it is possible that these economic benefits
from real estate investments in oligopolistic markets can temper or even overwhelm the associated
undesirable effect on firm risk discussed earlier.
To formally explore this theory and motivate the empirical analysis I propose a simple partial
equilibrium model of competition in an oligopolistic output market characterized by heterogeneous
firms. Firms use capital assets (real estate) and a fully depreciating input to produce a homoge-
neous final consumption good. They face asymmetric adjustment costs for capital investments and
use a Cobb-Douglas production technology. A detailed description of the model follows later. Since
the model cannot be solved algebraically, I carefully choose the parameters following the litera-
ture, simulate the firms’ investment and production decisions, and compute their average capital
investment characteristics and betas. These simulations reveal a negative relation between firm
beta and the ratio of capital assets (real estate) to the other input (Figure 2). Furthermore, the
graphs in Figure 2 show that the slope of the relation between real estate and firm risk flattens
as the number of firms increases, but remains non-positive. The fewer the number of competing
6Higher firm risk prompts investors to require higher returns, which ex post translate into a positive relationbetween real estate and observed returns.
3
firms (i.e., the higher industry concentration), the greater a firm’s pricing power and the steeper
the negative slope of the relation between firm beta and real estate investments.
Next, I empirically test this prediction by examining the return characteristics of a large sample of
public non-real estate corporations from 1973 to 2010 using a portfolio approach. I show that the
positive relation between real estate investments and stock returns documented in the literature is
specific to competitive industries. As predicted by the model, the relation between real estate and
stock returns turns negative in concentrated industries, as confirmed by large negative returns on
a synthetic long real estate position capturing the difference in returns between the high and low
real estate firms. In contrast, a similar synthetic long real estate position in competitive industries
produces positive returns, confirming the positive effect of real estate on firm risk documented by
Tuzel (2010) since most firms operate in competitive industries. Furthermore, I show that the
positive (negative) relation between real estate and stock returns in competitive (concentrated)
industries persists after controlling for conventional risk factors. These findings are also robust
to various measures of real estate and industry concentration. Since market efficiency normally
precludes consistent abnormal returns, these results imply the pricing of some risk factor(s) not
captured in the pricing model used to estimate abnormal returns. Finally, real estate appears to
alter exposure to the conventional risk factors included in the model in a similar manner in both
market structures.
Obviously, intense competition requires nimbleness at all levels of the firm. At the extreme, firms
facing cut-throat competition may be better off contracting out production in some cases. But this
cannot be true in all competitive industries, especially for oligopolies, since a firm’s competitive
advantages may be tied to firm-specific assets tied to real estate. These findings do not mean that
real estate investments are good or bad for shareholders in competitive or concentrated industries.
Real estate assets affect firm risk, thus ultimately altering the category of investors attracted to
these firms.
Possible endogeneity between real estate investments and returns makes identification of the true
relation difficult. Some business activities require more real estate than others (e.g., car manufac-
turing vs. computer production, or retail vs. business services). At the firm level, capital structure
may affect real estate decisions, which in turn feed back into capital structure and returns. Con-
4
trolling for product market competition permits a partial disentangling of the endogenous relation
between real estate investments and stock returns.7 I show that the findings persist in the firms’
industry-adjusted returns. In summary, corporate real estate strategies should consider indus-
try structure. Renting may be preferable from the perspective of firm risk for firms operating in
competitive markets. But the lower cash flow volatility associated with real estate ownership in
oligopolistic industries may yield significant benefits.
This paper proceeds as follows. Next, I review the relevant literature. The third section presents a
model of oligopolistic competitive equilibrium to motivate the study. The fourth section describes
the research methodology and data, followed by a discussion of the main empirical results and
various robustness checks. Finally, the last section concludes.
Related Literature
Several studies have examined the equilibrium relation between real estate investments and stock
returns. Deng and Gyourko (1999) document a weak negative relation between abnormal stock
returns and real estate investments for industrial firms. In contrast, Michael et al. (2001) and
Brounen et al. (2005) find no significant link between real estate and abnormal returns. But
both studies show that real estate has a significant negative effect on systematic risk, which is
not surprising since real estate generally has a low beta. The inconclusiveness of these studies
regarding the relation between abnormal stock returns and real estate may be due to endogeneity
issues. Also, the use of property, plant, and equipment (PPE) to total assets (TA) as a proxy for
real estate may be problematic. Deng and Gyourko (1999) find this ratio has a low cross-industry
variance. To avoid potential identification problems, several authors focus on retail firms, given the
importance of real estate in that industry, and generally document a positive relation between real
estate investments and stock returns (Brounen and Eichholtz (2005); Brounen et al. (2005); Yu and
Liow (2009)). Unfortunately, this finding is not necessarily generalizable to other industries.
7A simpler approach to dealing with this endogeneity issue is to restrict the analysis to one industry (Brounenand Eichholtz (2005); Yu and Liow (2009); Ling et al. (2012)). However, it is difficult to make general statementsabout the relation between real estate and stock returns from these studies.
5
Tuzel (2010) takes a broader approach in examining the relation between real estate investments
and stock returns. Using a portfolio formation methodology, which I adopt in this study, that
study finds that real estate increases firm risk resulting in a positive relation with returns. Tuzel
argues that the positive effect of real estate on firm risk stems from the operating inflexibility,
low depreciation rate, and significant asymmetric adjustment costs associated with real estate as
a factor of production. Somewhat similarly, Zhao and Sing (2015) examine corporate real estate
ownership decisions in the presence of productivity uncertainty and show that high-risk firms are
expected to hold less real estate assets in order to reduce potential operating losses in the event
of negative productivity shocks. This documented positive relation between real estate and stock
returns normally implies the pricing of some risk directly or indirectly associated with real estate.
Ling et al. (2012) empirically examine the pricing of real estate risk for retail firms by examining
the sensitivity of firm market and real estate betas to real estate investments. They document
a positive relation between firm real estate betas and real estate investments, but unlike other
studies (e.g., Deng and Gyourko (1999)) they find that real estate has no effect on market beta.
In summary, this literature generally documents a positive relation between real estate and stock
returns, although real estate appears to reduce firm systematic risk.
The above literature does not consider the possible role of product market competition. Relying on
the industrial organization literature on barriers to entry, Ambrose et al. (2016) explores the effect
of demand uncertainty on the endogenous interaction of strategic capital investment, industry con-
centration, and incidentally volatility of firm value. As predicted by Spence (1977), Dixit (1980),
and Tirole (1988), Ambrose et al. (2016) show that capital investments affect industry concentra-
tion by deterring entry. Furthermore, they document that demand volatility decreases capacity
investments and increases the volatility of firm value as concentration increases. But capital invest-
ments also provide other benefits in oligopolistic industries in addition to entry deterrence. Gaspar
and Massa (2006) note that market power reduces information uncertainty for investors and thus
dampens return volatility. Peress (2010) shows that firms can use their monopoly power to pass
shocks on to customers, thereby protecting profits and reducing cash flow volatility. Aguerrevere
(2009) shows that the effect of competition on systematic risk is conditional on demand at the
industry level, but systematic risk generally increases with competition even if installed capacity is
6
not sufficient to accommodate current demand, for which Hou and Robinson (2006) present sup-
portive empirical evidence. Finally, Irvine and Pontiff (2009) identify product market competition
as a driver of the increase in firm-level volatility documented by Campbell et al. (2001).
Undoubtedly, product market competition affects strategic operating decisions, which in turn im-
pact the riskiness of operating cash flows. The complicated interaction between capacity decisions
(real estate investments), product market competition, and stock returns is the primary focus of
this study. In contrast to Ambrose et al. (2016), I assume that industry structure is exogenous,
do not model demand volatility, and focus on oligopoly pricing power and its effect on cash flow
volatility, rather than on entry deterrence. But similar to Tuzel (2010) and Ambrose et al. (2016)
I view real estate investment from a technological perspective.8
Model
The following model describes an oligopolistic output market in which firms elect whether to invest
in capital assets (real estate) every period and use a fully depreciating input resource (e.g., raw
materials, labor, or another form of fully depreciating capital) to produce a homogeneous final
consumption good. The model of production and asset pricing borrows somewhat from Tuzel
(2010) while the equilibrium concept known as extended oblivious equilibrium fully detailed in
the equilibrium concept and compution section that follows is taken from Weintraub et al. (2010).
Whereas the model in Tuzel (2010) is a general equilibrium model that assumes perfect competition
in the output market, this model is a partial market equilibrium model in which the consumer’s
decision problem is not explicitly modeled. But as a important differentiating characteristic of
this model, firms are allowed to have pricing power based on capacity installed in an oligopolistic
competition setting. I allow heterogeneity in production capacity, which leads to heterogeneity in
market power.
8Ambrose et al. (2016) classify real estate into strategic and general-purpose real estate. It must be stressed thatthe argument advanced in this paper applies more to strategic than general-purpose real estate. Figure 1 shows thereal estate owned by corporations is most likely firm-specific in nature. While there may be valid economic reasonsfor a firm to own strategic real estate assets, leasing may be the optimal solution for general-purpose real estate,particularly with the growth of the equity real estate investment trust (REIT) industry.
7
Demand
The focus of this paper is the real estate investment decisions and stock returns for firms in an
oligopolistic industry of N firms that produce a homogeneous final good Qi,t with i ∈ {1, 2, . . . N}
and t ∈ {1, 2, . . .∞}. The aggregate demand for the final good in period t, Qt =∑N
i=1Qi,t, is given
by
Qt(Pt) = γ0ZtP−γ1t ,
where Pt is the good’s equilibrium market price, Zt is a market-wide demand shock, γ1 is the
elasticity of demand, and γ0 is a positive demand shifter. The inverse aggregate demand function
is then given by
Pt(Qt) =
(γ0ZtQt
) 1γ1
.
I assume that Zt, the market-wide demand shock, follows the AR(1) process in logs of the form
lnZt = ρZ lnZt−1 + ut, ut ∼ N(0, σ2
Z
),
with ρZ representing the persistence of the shock and ut representing the stochastic component of
the shock that is distributed IID normal with variance σ2Z .
Supply
I assume that the N firms are equity-financed. Firm i produces output Qi,t at the beginning of
period t using real estate capital Ki,t and the fully depreciating input Li,t. I assume that firms
must use the entirety of installed capacity to produce the final good. The output is given by the
Cobb-Douglas production function
Qi,t = F (Ai,t,Ki,t, Li,t) = Ai,tKαi,tL
1−αi,t .
Ai,t is a Hicks-neutral idiosyncratic productivity shock which also follows the AR(1) process in
logs
lnAi,t = ρA lnAi,t−1 + vi,t, vi,t ∼ N(0, σ2
A
)
8
with ρA representing the persistence of the shock. vi,t is the stochastic component of the shock
that is distributed IID normal with variance σ2A. I assume vi,t and vj,t are uncorrelated for any pair
(i, j) with i 6= j and that ut and vi,t are independent.
Real estate capital Ki,t is purchased in a perfectly competitive market and depreciates over time
at rate δ. The law of motion for capital is therefore
Ki,t+1 = (1− δ)Ki,t + Ii,t
where Ii,t is firm i’s capital investment at time t, which will be available for production at the
beginning of period t+ 1. This capital investment is subject to asymmetric quadratic adjustment
costs given by
g (Ii,t,Ki,t) =1
2η
(Ii,t)2
Ki,t
where
η =
ηlow if Iit > 0
ηhigh otherwise, with ηhigh ≥ ηlow.
The asymmetric adjustment costs capture the fact that it is more costly to decrease real estate
holdings than to expand them. Firms incur no adjustment costs during periods when investment
is zero.
Firms seek to maximize their expected present discounted value of profits by making a sequence of
investment and input purchasing decisions. Formally, firm i’s sequence problem is
max{Li,t,Ii,t}∞t=0
E0
∞∑t=0
ψtΠ
Qi,t,∑j 6=i
Qj,t, Zt
where ψ = 1
1+r and r is the real interest rate, which is exogenously determined since the consumer
problem is not modeled. Assuming that input markets are perfectly competitive so that firms
take input prices as given, firm i’s profit function at time t, which also represents the amount of
9
dividends to shareholders since firms are equity financed, is
Π
Qi,t,∑j 6=i
Qj,t, Zt
= Pt(Qt) ·Qi,t − wLi,t − Ii,t − g (Ii,t,Ki,t) (1)
=
( γ0Zt
Ai,tKαi,tL
1−αi,t +
∑j 6=iQj,t
) 1γ1
Ai,tKαi,tL
1−αi,t − wLi,t
− (Ki,t+1 − (1− δ)Ki,t)−1
2η
(Ki,t+1 − (1− δ)Ki,t)2
Ki,t
where∑
j 6=iQj,t =∑
j 6=iAj,tKαj,tL
1−αj,t . Thus, firm i’s Bellman equation can be written as
V
Ki,t, Ai,t, Zt|∑j 6=i
Qj,t
= maxLi,t,Ki,t+1
ΠQi,t,∑
j 6=iQj,t, Zt
(2)
+ψE
VKi,t+1, Ai,t+1, Zt+1|
∑j 6=i
Qj,t+1
|Ai,t, Zt .
Asset Pricing
Firm i ’s consumption cost of capital (Tobin’s q) at time t from Equation 1 can be expresssed as
follows
qi,t = 1 + ηIi,tKi,t
.
Assuming the firm can liquidate its capital stock at this price, the value of firms equity at time t
(Ei,t) is then equal to the market value of its assets in place, namely
Ei,t = qi,tKi,t+1,
and the dividend firm i pays to shareholders is
Di,t = Π
Qi,t,∑j 6=i
Qj,t, Zt
.
10
The rate of return for a firm’s equity shareholders is therefore
rei,t+1 = ln
(Ei,t+1 +Di,t+1
Ei,t
).
In this paper, the main objects of interest are the firm betas and their relation with firm real estate
capital holdings Ki,t. Let rvw,t+1 be the value-weighted average firm return (i.e., the aggregate
stock market return). Then firm i’s beta at time t is
βi,t =Covt
(rvw,t+1, r
ei,t+1
)Vart (rvw,t+1)
where the variance and covariance are calculated using the previous 100 periods.
Equilibrium Concept and Computation
The firms’ investment and input decisions cannot be solved algebraically. In order to determine the
relation between market structure, capital holdings, and firm betas, an equilibrium concept must
be chosen that is both computationally feasible and allows independent variation of the number of
firms N . To do so, the study adopts a slightly modified version of extended oblivious equilibrium
from Weintraub et al. (2010) that disallows firm entry. Conditional on today’s value of the aggregate
industry shock, each firm tracks the long-run average industry state. That is, rather than tracking
every other firm’s capital holdings and productivity and responding appropriately, each firm follows
a strategy that is “oblivious” in the sense that it responds to a hypothetical average firm and is
oblivious to the actions of any particular competitor. This equilibrium concept is explained in
greater detail in the appendix section. The equilibrium is computed using an algorithm described
in the appendix. After an equilibrium is found, a panel dataset of variables from the model is
simulated using the equilibrium strategies.
Parameterization and Results
The parameters of the model are not calibrated to match any particular industry. They are chosen
to be plausible as a real data generating process and closely match the parameter values in previous
11
empirical studies (Kydland and Prescott (1982); Cooley and Prescott (1995); Tuzel (2010)). Table
1 presents the chosen parameter values for the model. The persistence and standard deviation of
systematic demand shocks (ρZ and σZ) are set to 0.8 and 0.025, respectively. I set the demand
elasticity (γ1) to 2 and the demand shifter (γ0) to 25 in order to generate variation in capital
holding across firms. The capital share (α), persistence of idiosyncratic productivity shocks (ρA),
and standard deviation of idiosyncratic productivity shocks (σA) are 0.36, 0.8, and 0.05, respectively.
I assume a reasonable capital depreciation rate (δ) of 0.02 and, and set the low and high capital
adjustment costs (ηLow and ηhigh) to 0.8 and 1.6, respectively – I set the value of the high capital
adjustment cost equal to twice that of the low adjustment cost.9 Finally, I use a real interest rate
of 0.04, which is adequate for the period covered by the study, and normalize the price of the fully
depreciating input to 1.
Figure 2 displays results from simulations of the model. In particular, it shows scatter plots of
average firm betas and average capital-to-other-input ratios for firms sorted on their capital-to-
other-input ratio in each period for various industry sizes. In contrast to the results of the perfectly
competitive model presented in Table 3 of Tuzel (2010), this model of an oligopolistic industry
produces a slightly negative relation between relative capital holdings and firm risk as measured by
firm betas. Furthermore, this relation appears to weaken as the industry becomes more competitive;
the trend is very flat when there are 100 firms in the industry compared to when there are just 8
or 10 firms. Since firms with market power may face incentives to hold relatively illiquid capital
assets such as real estate capital in order to maintain their competitive position, investments in
such firms may be less risky than investments in firms with less productive capacity.
Empirical Analysis
Methodology
The design of the empirical analysis is as follows. First, I assign firms to industries every year
based on their SIC codes10 and estimate the level of competition in the industries. I compute the
9The results are unchanged when the ratio of high to low adjustment costs is set even higher10I check the robustness of the main findings to alternative SIC classifications and the allocation of firms to
industries according to Fama and French (1992) industries.
12
Herfindahl-Hirschman Index (HHI) of the industries as
HHIj =N∑i=1
s2ij (3)
where sij is the market share of firm i in industry j, which contains N firms. HHI is a measure
of market concentration. As is common in the literature, I estimate market share as the ratio of
firm sales to total industry sales. Therefore, a high HHI is indicative of a concentrated industry
dominated by few large firms, whereas a low HHI is indicative of a competitive industry. Since
changes in industry concentrations are likely to be gradual, following Hou and Robinson (2006)
I assign to each industry its three-year average HHI and classify the industries according to that
measure into concentration quintiles. The industries in the first quintile represent competitive
industries and those in the fifth quintile represent concentrated (oligopolistic) industries. Again,
I perform this classification annually. Next, I separately mix the firms in the competitive and
concentrated industry groups and sort each group of firms into decile portfolios according to the
amount of real estate on the firms’ balance sheets. I then track the performance of the portfolios
over the next 12 months before starting over the double sorting along industry concentration and
then real estate assets.
If product market competition has no effect on the relation between real estate investments and firm
risk, the behavior of portfolio returns relative to the firms’ real estate assets should be similar in
the competitive and concentrated industry groups. However, the model predicts that the relation
between returns and real estate will be non-positive in concentrated industries. First, I compare
the performance of the portfolios using average excess returns over the risk-free rate and average
industry-adjusted returns (i.e., average of the firms’ returns over their respective industry’s average
returns).11 If industry concentration and real estate are not determining return factors, and absent
any serious sampling problems, there should be no significant differences between the portfolio
returns in competitive and concentrated industries. Based on the findings of the existing literature,
I expect average portfolio returns to be positively correlated with real estate assets in competitive
industries. As argued, the story is more complicated in concentrated industries. Even though
11By controlling for average industry returns, the latter measure lead to a more accurate comparison of firms acrossdifferent industries.
13
real estate investments still reduce operating flexibility in those industries due to the associated
asymmetric adjustment costs, the resulting increase in pricing power will weaken or even cancel
out that negative effect as shown by the model’s predictions.
Even though the finding of significant differences in the portfolios’ excess and industry-adjusted
returns suffices to conclude that real estate affects stock returns, it cannot be automatically inter-
preted as evidence that the market prices some risk associated with real estate investments. To
explore this further, the next step of the analysis considers whether the portfolios generate signif-
icant differences in abnormal returns after controlling for conventional risk factors and the extent
to which these abnormal returns are related to the amount of real estate owned by the portfolios’
component firms. The portfolios’ abnormal returns are estimated from monthly return series using
the following three-factor market model:
ri,t − rf,t = αi + βi(rm,t − rf,t) + γismbt + δihmlt + εi,t, (4)
where the dependent variable is the monthly returns of portfolio i (ri,t) minus the risk-free rate
(rf,t) and the pricing factors are monthly excess market returns (rm,t − rf,t) and monthly returns
on the Fama and French (1992) benchmark size and book-to-market portfolios, smbt and hmlt,
respectively. The scalar αi represents portfolio i ’s abnormal return (alpha) during the period. The
coefficients βi, γi, and δi represent portfolio i ’s systematic risk and loadings on the Fama-French
size and book-to-market risk factors, respectively. Finally, the vector εi contains the portfolio’s
return error terms.
If equation (4) properly accounts for all risk factors priced by the market, the estimated portfolio
abnormal returns should not be statistically different from zero. Otherwise, I cannot categori-
cally reject that product market competition, real estate ownership, or both affect stock returns.
However, the finding of insignificant abnormal returns does not contradict the main argument of
this paper either. Therefore, it is difficult to make definitive predictions about the relation be-
tween abnormal portfolio returns and real estate investments. However, I expect this relation to
be non-negative in competitive industries and non-positive in concentrated industries.
14
Sample Selection
I use the Center for Research in Security Prices (CRSP) public company stock return data and the
corresponding Compustat accounting data for this empirical analysis. The initial sample consists
of U.S. firms with SIC numbers between 2000 and 5999 (i.e., industrial firms) with CRSP ordinary
share code 10 and 11 listed on the NYSE, AMEX, and Nasdaq between January 1970 and December
2010 that are contained at the intersection of the CRSP monthly stock return data series and the
merged Compustat annual industrial accounting database.12 Therefore, the sample excludes the
following industries: real estate investment trusts, construction, financial services, mining and oil,
agriculture, services, and healthcare.
I match the stock return and accounting data following Fama and French (1992) to ensure that
the accounting information used to create the portfolios is available prior to the stock return data
it is meant to explain and already impounded into stock prices. This is done by matching CRSP
monthly stock returns from July of year t to June of year t + 1 with accounting information for the
fiscal year t – 1, which results in stock returns lagging the release of the accounting data by at least
six months. Then, I assign firms to industries at the end of June every year according to the first
three digits of their Compustat SIC codes.13 Next, I compute the industries’ HHI using net sales as
explained earlier. I then drop industries with fewer than three firms, which eliminates monopolies,
and classify the remaining industries into concentration quintiles according to the industries’ three-
year average HHI. This three-year smoothing and the lagging of stock returns relative to accounting
data reduce the final sample to 37 years, spanning July 1973 through June 2010. Again, I perform
the industry classification at the end of June every year and keep the resulting industry quintiles
for the next 12 months from July through June.
Next, I separately classify the firms in the competitive and concentration industry quintiles into
ten portfolios each according to the amount of real estate assets in PPE at the end of June and
track the performance of the portfolios over the next 12 months.14
12I use 1970 as a starting date to reduce any potential bias toward large firms since Nasdaq firms were added toCRSP in 1973. Also, the Compustat PPE component accounts required for this analysis were sparsely populatedprior to 1970.
13Even though Compustat and CRSP SIC codes do not perfectly match (Kahle and Walkling, 1996), the outcomeof the analysis is unchanged when I use CRSP SIC codes for the industry classification. I also show that the resultsobtain when I use two-digit SIC codes or the Fama-French industries.
14Compustat breaks down PPE accounts into buildings, machinery and equipment, capitalized leases, land and
15
Descriptive Statistics
The final sample contains 7,736 firms classified into 171 industries according to their three-digit
SIC codes and consists of 71,885 firm-years, representing on average 1,943 firms per year. Table
2 summarizes the distributional characteristics of the industries’ annual and three-year moving
average HHI. I also report in Table 2 concentration figures based on total assets for comparison
purposes. Despite the exclusion of industries with fewer than three firms, the industries’ HHI vary
considerably, from 0.04 (indicative of a highly competitive industry) to 0.997 (reflective of a highly
concentrated industry). Table 2 also shows that the distributional characteristics of the four mea-
sures of industry concentration are very similar. Additionally, Figure 3 reveals that the distribution
of industry HHI is not constant over the sample period. The first moments of the distributions
vary considerably over time but remain interlocked, moving together over time. Hence, any of these
concentration measures should adequately capture the variation in industry concentration present
in the data and can be used for the classification of industries into quintile concentration groups.
As noted previously, I use three-year moving average sales HHI (HHI MA (sales)) in Table 2 and
throughout the rest of the paper as the main industry concentration measure.
The sample period witnessed two distinct waves of industry consolidation. The first and strongest
wave started in the mid-1970s and crested in the early 1990s. It resulted in average HHI increasing
by roughly 45% from 0.29 in 1976 to 0.42 in 1990. Also, the last decade witnessed a milder
surge in industry consolidation. The first wave coincided with a surge in hostile takeovers, partly
facilitated by an expansion of the junk bond market that collapsed as the economy went into
recession in the early 1990s (Carney (2009); Lipton (2006)). The latter wave was driven by a
number of factors, including globalization, a rise in commodity prices, low interest rates, shareholder
activism, hedge funds, and growth in private equity funds (Lipton (2006)). These exogenous shocks
to industry concentration should facilitate identification of the role of product market competition
in the relation between real estate investments and stock returns.
Industry Concentration Quintiles
improvements, construction in progress, natural resources, and other assets. Following Tuzel (2010), I use buildingsand capitalized leases, the largest components of PPE and the closest to production capacity, as the primary measureof real estate investments. Since Compustat only reports PPE accounts net of depreciations prior to 1985, I use netfigures for those years and gross figures after 1985 to compute real estate variables.
16
Table 3 summarizes the composition of the HHI-derived industry concentration quintiles, which
include 26 or 27 industries each. The average HHI of the concentrated industries group (the fifth
quintile) is roughly four times that of the competitive industries group (the first quintile): 0.66
versus 0.14.15 The difference between the competitive and concentrated industries is even more
pronounced at the firm level. The average HHI contribution16 of firms in the concentrated industries
is 30 times that of firms in the competitive industries. The difference in mean tests reported in the
last row of Table 3 unequivocally reject the null hypothesis of equal means between the two groups.
Figure 4 further highlights the difference in competitiveness between the two groups of industries.
The gap between the two curves tracing average HHI of the high and low concentration groups
never narrows over the study period. Furthermore, each series displays the two waves of industry
consolidation noted earlier, indicating that they were not limited to specific industries.17 Given the
pronounced structural differences between these two groups of industries, a comparison the effects
of real estate investments on stock returns in these industries should yield valuable insight into the
role of industry structure, if any.
Table 4 lists the average characteristics of firms in the competitive and concentrated industry
quintiles. As the number of firms drops with industry concentration, average firm size (measured
by sales, total assets, or market value of equity) predictably increases. Firms operating in con-
centrated industries are roughly twice the size of those in competitive industries. In addition,
book-to-market value increases with industry concentration; firms operating in competitive indus-
tries are more growth-oriented due to their constant need to undertake value-enhancing innovations
(Hou and Robinson (2006)). Brander and Lewis (1986, 1988) note that product market compe-
tition affects capital structure decisions. Both leverage and long-term debt ratios are positively
correlated with industry concentration. Since firms operating in concentrated industries are larger
and experience less cash flow volatility, they are able to sustain higher levels of debt (Jensen and
Meckling (1976); Myers (1984); Myers and Majluf (1984); Jensen (1986)). Table 4 clearly reveals
significant differences in size and capital structure between firms in competitive and concentrated
15For example, two representative industries of the concentrated and competitive industry groups are rubberproducts and restaurants, respectively.
16Firm i ’s contribution to industry j ’s HHI is s2ij in Equation (3).17Figure 4 shows that an alternative industry classification technique consisting of setting fixed HHI bands through-
out the sample period, rather than annually resetting these thresholds, would not materially alter the findings.
17
industries, as evidenced by the strong rejection of the null of equal mean values for all variables.
These differences certainly affect cash flows and should therefore be reflected in stock returns.
There is also a clear difference between the two groups of industries with respect to the amount
of real estate and other productive assets they own. The last four columns of Table 4 reveal an
interesting fact about the nature of the firms’ real estate investments. Columns 9 and 10 show that
firms operating in concentrated industries invest significantly more in buildings, land, and fixed
assets in general than those operating in competitive industries. Relatively, firms in competitive
industries are more likely to use leases (columns 7 and 8). Firms operating in concentrated industries
may be more reluctant to use leasing due to a higher proportion of firm-specific investments, stronger
cash flows, or both. Their higher capital intensity may also be a factor.
Real Estate Decile Portfolios
After classifying the sample into quintile industry concentration groups, I separately sort the firms in
the low and high quintile groups into decile portfolios according to the firms’ real estate investments.
Table 5 summarizes the characteristics of the firms in the first (low) and the tenth (high) real
estate decile portfolios of the competitive and concentrated quintile industries. As in Tuzel (2010),
I sort the firms into decile portfolios according to the ratio of buildings and capitalized leases
to PPE.18 Whether measured by sales, assets, or market value, the average size of the firms in
the tenth real estate decile is two to three times that of the firms in the first real estate decile
in competitive industries. The difference in average firm size between the first and tenth decile
portfolios is even more pronounced in concentrated industries where firms in the top decile portfolio
are on average more than 8 times larger in terms of sales. As noted previously, firms operating
in concentrated industries are generally larger, have higher book-to-market values, and are more
levered, as evidenced by the significant cross-industry difference in mean tests reported in Table
5.
A close comparison of the first three columns of tables 4 and 5 reveals an interesting fact about
the relation between firm size and real estate investments. These columns show that real estate
is positively correlated with firm size in concentrated industries. For example, the average firm
18This study’s main findings remain unchanged when I use alternative real estate measures as discussed later.
18
sales of $2.90 billion in concentrated industries (column 1 in Table 4) lies between the first decile
portfolio’s average firm sales of $900 million and the tenth decile portfolio’s average firm sale $3.06
billion (column 1 of Table 5). In contrast, firms in competitive industries do not display the same
monotonic pattern: the average firm sales of $1.28 billion for the industry group in Table 4 is outside
those of the first and tenth decile portfolios ($331 and $703 million) in Table 5. This non-monotonic
relation between real estate investments and firm size may be due to greater heterogeneity within
that group of industries. Table 5 also shows a striking difference in the portfolios’ real estate and
fixed assets (columns 7 to 10). Depending on the measure of real estate used, firms in the high
real estate decile have on average 8 to 22 times more real estate than those in the low real estate
decile in the competitive and concentrated industry groups, despite a much smaller difference is
fixed assets.
In summary, Table 5 highlights two important points. First, average characteristics of firms in
competitive and concentrated industries are significantly different; competitive industries largely
include growth firms. Second, the low real estate decile firms in both industry groups are smaller,
more growth-oriented, and less levered. These significant differences in firm characteristics between
the low and high decile portfolios should normally affect the portfolios’ performance.
Main Results
In this section, I present the performance of the real estate sorted portfolios in competitive and
concentrated industries. Without controlling for industry structure, Tuzel (2010) documents a pos-
itive relation between real estate and returns from 1971 to 2005. I argue that this positive relation
applies to competitive industries only. As shown in the theory section, real estate investments
may lead to greater pricing power and lower firm risk in oligopolistic industries. Consequently, I
expect a non-positive relation between real estate and returns in those industries. These predicted
relations should materialize in the portfolios’ returns.
Average Portfolio Returns:
Table 6 presents the portfolios’ average excess and industry-adjusted returns during the 37-year
period from July 1973 through June 2010. The portfolios’ average excess returns in competitive
19
industries in Panel A are positive and statistically significant. As predicted, they increase with
the portfolios’ real estate rankings from 4.7% for the low real estate decile to 10.6% for the high
real estate decile, representing a 5.9% average annual return difference between the two. Also,
the portfolios’ average industry adjusted returns trend positively with their real estate rankings,
resulting in a 4.8% return premium for the top decile portfolio. These results shown in the top-
left corner of Figure 5 corroborate Tuzel’s findings. Managers of firms operating in competitive
industries should be aware that significant real estate ownership is likely to increase firm risk due
to the associated decrease in operating flexibility.19 Yet the argument that all firms should avoid
real estate ownership regardless of the competitive environment may be shortsighted. Panel B of
Table 6 repeats the analysis with the portfolios of firms operating in concentrated industries. In
contrast to the previous results, both portfolio return series decrease with the portfolios’ real estate
rankings. The portfolios’ average excess returns drop from 11.3% for the low decile to 4.8% for the
top decile (shown in the top-right quadrant of Figure 5). Basically, high real estate firms appear
to be less risky in concentrated industries compared to competitive industries. The above portfolio
returns are value-weighted averages, but the findings are unchanged with equally-weighted average
returns (Table .1 of the appendix).
As noted, real estate investments affect capital structure and may therefore also alter stock re-
turns through that channel. Even though it is unclear how leverage may differently affect stock
returns in competitive and concentrated industry, it is important that I adjust portfolio returns for
leverage since the model assumes that firms are equity financed.20 Table 7 reports value-weighted
average unlevered excess and industry-adjusted returns of the decile portfolios in competitive and
concentrated industries. As expected, average unlevered portfolio returns are lower than the cor-
responding levered returns presented in Table 6, reflecting the positive effect of leverage on equity
returns (in bottom-left quadrant of Figure 5). For example, the average unlevered excess returns
19The fact that real estate ownership is associated with higher firm risk is not necessarily bad. As firms undertakesuch investments, they may become more attractive to different investor clienteles.
20I compute unlevered stock returns (rU ) using the following formula that assumes no taxes:
rU = rLD + E
E− rD
D
E, (5)
where rL and rD are levered stock return (return from CRSP in this case) and average cost of debt, respectively. Dand E stand for market value of debt and equity. I assume an average cost of debt of 7% across the board – theresults are robust to changes in the average cost of debt.
20
of the low and high decile portfolios in competitive industries are 2.6% and 6.7%, compared to
levered returns of 4.7% and 10.6%, respectively. Furthermore, the relation between average unlev-
ered returns and real estate investments remains positive and generally monotonic in competitive
industries (Panel A of Table 7). Despite controlling for leverage, the difference between the high
and low decile portfolios’ returns remain large and statistically significant. For example, the differ-
ence between the high and low portfolios’ average unlevered excess returns is 4.1%, though lower
than the 5.9% difference in levered returns. More importantly, the behavior of unlevered portfolio
excess and industry-adjusted returns in concentrated markets (Panel B Table 7 and bottom-right
quadrant of Figure 5) is similar to that of levered returns discussed earlier. The portfolios’ average
unlevered excess returns decrease from 8.2% to 3.0%, giving a difference unlevered excess returns
of -5.2% between the high and low decile portfolios, compared to a difference in levered returns of
-6.5%.
The unlevered portfolio returns generally confirm the conclusions derived from levered returns. The
portfolios’ returns increase with their real estate rankings in competitive industries and decline in
concentrated industries. Again, the results are unchanged when equally weighted returns are used.
However, controlling for leverage may improve identification of the potential role of product market
competition on the relation between real estate investments and stock returns. But it is important
to control for industry fixed-effects as well since some industries are more real estate intensive than
others. Consequently, it is comforting that unlevered industry adjusted returns reported in Panel
B of Table 7 support this study’s predictions.
Abnormal Portfolio Returns:
So far, I have largely abstracted from risk considerations. In this section, I explore whether the
documented difference in portfolio performance in competitive and concentrated industries also
materializes in abnormal portfolio returns. The finding of significant abnormal portfolio returns
after controlling for conventional pricing factors may reveal the channel through which product
market competition alters the relation between real estate and stock returns.
Table 8 reports abnormal portfolio returns from the estimation of model (4) using the portfolios’
monthly value-weighted unlevered excess returns. The abnormal returns of competitive industry
21
portfolios are generally significant (both statistically and economically) and increase from -3.1% for
the low real estate decile portfolio to 1.9% for the high real estate decile portfolio, a difference in
abnormal return of 4.9%. Table 8 also shows estimated abnormal portfolio returns in concentrated
industries, which are negatively related to the portfolios’ real estate rankings. The estimated ab-
normal returns of the low and high decile portfolios are 2.5% and -2.5%, respectively, resulting in
a significant difference in abnormal return of -5.1%. This significant negative relation between real
estate and stock returns in concentrated industries is an important finding, for it goes beyond the
weak prediction of a non-positive relation set out initially. Untabulated estimations using equally
weighted unlevered portfolio returns produce similar results in competitive and concentrated in-
dustries. These findings also obtain when I use levered excess returns (Table .2 of the appendix).
Therefore, these documented relations between abnormal portfolio returns and real estate invest-
ments are not driven by either leverage or exposure to the conventional risk factors included in
model (4).21
Finding significant abnormal returns does not necessary imply that the market prices some real
estate risk. It is possible that real estate ownership is correlated with an omitted risk factor. The
challenge is that any search for abnormal return is conditional on the pricing model and the data.
Assuming that the model is correctly specified and that there are no significant data issues, the
portfolios should normally produce insignificant average abnormal returns. I follow the literature
in terms of model selection, even running the risk of sacrificing efficiency by adopting a multi-
factor market model rather than the more parsimonious capital asset pricing model. And even
though the estimated abnormal returns could be data-driven, the long sample period of 37 years
and the relatively low risk of selection bias, given the sample size, provide some reassurance. Also
the conditioning of the portfolios on real estate seems reasonable since it is unlikely that investors
would disregard such information when pricing stocks while generally reacting to corporate real
estate acquisitions and disposals as documented by event studies (Hite et al. (1984); Slovin et al.
(1990); Glascock et al. (1991); Myer et al. (1992); McIntosh et al. (1995); Rodriguez and Sirmans
(1996)).
In summary, the evidence presented shows that real estate investments and stock returns are
21The results remain materially unchanged when I use a four-factor model adding a market momentum factor tomodel (4).
22
positively related in competitive industries and negatively related in concentrated industries. Since
the positive return effect in competitive industries appears to be statistically stronger and most
industries are relatively competitive, it is likely that this effect dominates in studies that do not
control for industry concentration.
Real Estate and Portfolio Risk-Factor Loadings:
The effects of real estate on stock returns is not restricted to the evidence from abnormal returns
discussed in the previous section. Real estate investments also affect exposure to conventional risk
factors. To capture these effects I estimate equation (4) using monthly returns on the high real
estate decile portfolio minus monthly returns on the low real estate decile portfolio, which can be
interpreted as a synthetic long real estate position. The regression coefficient estimates from this
high-minus-low real estate position in Table 9 provides valuable insights. Regardless of industry
structure, real estate investments appear to lower market beta and sensitivity to the size risk factor,
and to increase exposure to the book-to-market risk factor. These findings are quite intuitive. It has
been documented that real estate investment lowers market beta (Deng and Gyourko (1999)). Also,
the positive correlation between real estate and size may explain the negative size effect and positive
book-to-market effect of real estate. The fact that the directions of these effects transcend industry
considerations means that the opposing effects of product market competition on the interaction
between real estate investments and stock returns are captured in the portfolios’ abnormal returns.
It is thus conceivable that factors beyond those included in the model might explain the observed
return reversal.
Furthermore, a comparison of the portfolios’ estimated factor loadings (unreported) confirms that
firms operating in competitive industries have on average higher market betas, as predicted by
Aguerrevere (2009), and are more exposed to the size risk factor, which should normally lead to
higher returns overall as documented by Hou and Robinson (2006).
Robustness Checks
The previous portfolio returns apply to the 37-year period from July 1973 to June 2010. I assigned
firms to industries according to the first three digit of their SIC codes and I classified industries into
23
HHI quintiles and sort firms in the low and high quintile groups into decile portfolios according the
ratio of buildings and capitalized leases in PPE. In this section, I check that the study’s findings
are not specific to the selected sample period and are robust to alternative industry classification
methods and real estate measures.
The U.S. economy is still recovering from the Great Recession brought about by the collapse of the
housing market and resulting financial crisis. Although it is not clear how such a tail event toward
the end of the sample period might have affected the results, it is important to ensure that the
findings are robust to the exclusion of that period of relatively high uncertainty. Table 10 reports
the abnormal portfolio returns from July 1973 to June 2005, which almost coincides with the period
covered by Tuzel (2010). To save space, I only report the returns on the High–Low decile portfolios
which capture the effects of real estate as discussed earlier. As Table 10 shows, the results are
unaffected by the exclusion of the Great Recession. The difference in abnormal returns between
the high and low real estate decile portfolios is positive and significant in competitive industries and
negative and significant in concentrated industries. In fact, these results are even slightly stronger
than the base-case results, indicating that the inclusion of the real estate driven recession might
have biased the main findings downward.
The evidence presented so far results from the performance of real estate decile portfolios. But if the
argument advanced in this paper is correct, quintile portfolios should also provide some supporting,
albeit weaker, evidence since the difference in real estate holdings between the high and low quintile
portfolios is smaller relative to that of corresponding decile portfolios. The outcome of such an
analysis presented in column 2 of Table 11 confirms this prediction – column 1 reports High-Low
returns from Table 7. The differences in average excess and industry-adjusted returns between
the high and low real estate quintile portfolios are lower in absolute terms than the corresponding
differences in decile returns reproduced in column 1. Also, a comparison of the average returns on
the low and high quintile portfolios (not shown in Table 11) and the corresponding decile portfolio
returns further corroborates this study’s predictions. In competitive industries, the average return
on the low quintile portfolio is higher than on the low decile portfolio, while the corresponding
return on the high quintile portfolio is lower than on the high decile portfolio. This interesting
finding further confirms the positive relation between real estate and stock returns in competitive
24
industries. In contrast, the average return on the low quintile portfolio in concentrated industries
is lower than on the low decile portfolio, as expected, but the corresponding average returns on the
high quintile and decile portfolios are roughly similar, hence confirming the predicted non-positive
relation between real estate and stock returns in concentrated industries.
Table 11 also reports the difference in excess and industry-adjusted returns between the high and
low decile portfolios when firms are classified into industries according to their two-digit SIC codes
(Column 3) and the 48 Fama-French industries (Column 4), a widely used industry classifica-
tion method in finance studies.22 Overall, these results confirm the previous findings, However,
the Fama-French industry classification method produces stronger results than the two-digit SIC
industry classification method since it involves more industries. Surprisingly, the Fama-French clas-
sification method performs as well as the three-digit SIC classification used in the base case, even
though it results in a much smaller number of industry groups (48 versus 171). It is possible that
the Fama-French classification technique results in more homogeneous industry groups, which may
help facilitate identification. The last column of Table 11 gives differences in average returns when
I use total assets in lieu of net sales to compute industry concentrations. Since total assets and net
sales are highly correlated and share similar distributional characteristics as Table 2 and Figure 3
show, the results in column 5 are almost identical to the base-case results in column 1.
Table 12 reports excess and industry-adjusted returns under alternative measures of real estate
ownership. In the base case (column 1), firms are classified into portfolios according to RER1, the
ratio of buildings and capitalized leases to PPE. Using RER2, a broader measure of real estate that
adds construction in progress and land to RER1, does not significantly alter the results (column 2).
RER3, the ratio of buildings and land to PPE, also leads to similar results (untabulated). PPER,
the ratio of PPE to TA, however, appears to be a poor proxy for real estate, as this measure
produces negative but insignificant results in concentrated industries (column 3). Intuitively, the
more broadly that real estate is defined, the more serious the identification problem becomes,
leading to insignificant results. This may explain why Deng and Gyourko (1999) and other authors
who use this measure find weak or insignificant results. Measuring a firm’s real estate intensity in
terms of deviation from its industry’s average using ARER, which is equal to RER1 minus average
22The two-digit SIC classification results in 36 industry groups. The 48 Fama-French industries are publishedonline at http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data library.html.
25
industry RER1, in column 4 leads to slightly stronger results in competitive industries and confirms
the non-positive relation between real estate and stock returns in concentrated industries.
As Table 3 shows, firms in competitive and concentrated industries are significantly different,
particularly with respect to size: firms operating in concentrated industries are considerably larger.
Could this size difference between firms in competitive and concentrated industries explain the
findings? This study is not comparing portfolio returns in competitive industries to portfolio returns
in concentrated industries. Rather, it compares portfolio returns within each industry concentration
group separately. Therefore, the difference in average firm size between the two groups should be
immaterial. The documented relations are more about differences in the portfolios’ returns relative
to real estate assets than overall portfolio returns, which normally capture any size effect. Finally,
the findings materialize not only in average excess and industry-adjusted portfolio returns, but also
in abnormal portfolio returns generated from a pricing model controlling for the size risk factor.
The fact that these relations evidence themselves in abnormal returns represents strong evidence
that firm size is probably not a determining factor in the importance of product market competition
in the relation between real estate and stock returns.
Conclusion
This study extends the literature by introducing product market competition as a determining
factor in the relation between real estate investments and firm risk. Approaching these capital in-
vestments from a technology perspective and borrowing from the industrial organization literature,
I propose a model of capital investment decisions in oligopolistic industries and empirically show
that the relation between real estate investments and firm risk is non-positive to negative in that
environment. Therefore, the positive relation between real estate investments and stock returns
documented in the literature applies to competitive industries where firms enjoys no pricing power,
resulting in more volatile cash flows. The findings persist after controlling for capital structure
and adjusting returns for exposure to conventional risk factors. Managers must certainly consider
the competitive environment and their firms’ strategic objectives when considering real estate in-
vestments, particularly for firms operating in highly competitive industries. Capital markets being
26
relatively efficient at pricing risk, whether from real estate or other sources, the amount of real
estate assets owned by a firm ultimately affects the type of investors attracted to that stock.
I would like to thank Brent W. Ambrose, Charles Cao, Ed. Coulson, Austin Jaffe, and Jiro Yoshidafor helpful comments. Research assistance was ably provided by Dennis McWeeny. All errors aremy own.
27
References
Aguerrevere, F. L. (2009). Real options, product market competition, and asset returns. Journalof Finance, 64(2):957–983.
Aiyagari, R. S. (1994). Uninsured idiosyncratic risk and aggregate saving. The Quarterly Journalof Economics, 109(3):659–684.
Ambrose, B. W., Diop, M., and Yoshida, J. (2016). Product market competition and corporatereal estate investment under demand uncertainty. Real Estate Economics.
Berk, J. B., Green, R. C., and Naik, V. (1999). Optimal investment, growth options, and securityreturns. Journal of Finance, 54(5):1553–1607.
Brander, J. A. and Lewis, T. R. (1986). Oligopoly and financial structure: The limited liabilityeffect. American Economic Review, 76(5):956–970.
Brander, J. A. and Lewis, T. R. (1988). Bankruptcy costs and the theory of oligopoly. CanadianJournal of Economics, 21(2):221–243.
Brounen, D., Colliander, G., and Eichholtz, P. M. (2005). Corporate real estate and stock perfor-mance in the international retail sector. Journal of Corporate Real Estate, 7(4):287–299.
Brounen, D. and Eichholtz, P. M. (2005). Corporate real estate ownership implications: Interna-tional performance evidence. Journal of Real Estate Finance and Economics, 30(4):429–445.
Campbell, J. Y., Lettau, M., Malkiel, B. G., and Xu, Y. (2001). Have individual stocks becomemore volatile? An empirical exploration of idiosyncratic risk. Journal of Finance, 56(1):1–43.
Carlson, M., Fisher, A., and Giammarino, R. (2004). Corporate investment and asset price dy-namics: Implications for the cross-section of returns. Journal of Finance, 59(6):2577–2603.
Carney, W. J. (2009). Mergers and Acquisitions. Aspen Publishers Online.
Cooley, T. F. and Prescott, E. C. (1995). Economic Growth and Business Cycles: Frontiers ofBusiness Cycle Research. Princeton, NJ: Princeton University Press.
Cooper, I. (2006). Asset pricing implications of nonconvex adjustment costs and irreversibility ofinvestment. Journal of Finance, 61(1):139–170.
Deng, Y. and Gyourko, J. (1999). Real estate ownership by non-real estate firms: An estimate ofthe impact on firm returns. Wharton Working Paper.
Dixit, A. (1980). The role of investment in entry-deterrence. Economic Journal, 90(357):95–106.
Dixit, A. K. and Pindyck, R. S. (1994). Investment under uncertainty. Princeton, NJ: PrincetonUniversity Press.
Eaton, B. C. and Lipsey, R. G. (1980). Exit barriers are entry barriers: The durability of capitalas a barrier to entry. Bell Journal of Economics, 11(2):721–729.
Ericson, R. and Pakes, A. (1995). Markov-perfect industry dynamics: A framework for empiricalwork. Review of Economic Studies, 62(1):53–82.
28
Fama, E. F. and French, K. R. (1992). The cross-section of expected stock returns. Journal ofFinance, 47(2):427–465.
Fisher, L. M. (2004). The wealth effects of sale and leasebacks: New evidence. Real Estate Eco-nomics, 32(4):619–643.
Gaspar, J.-M. and Massa, M. (2006). Idiosyncratic volatility and product market competition.Journal of Business, 79(6):3125–3152.
Glascock, J. L., Davidson, W. N., and Sirmans, C. (1991). The gains from corporate selloffs: Thecase of real estate assets. Real Estate Economics, 19(4):567–582.
Grenadier, S. R. (2002). Option exercise games: An application to the equilibrium investmentstrategies of firms. Review of financial studies, 15(3):691–721.
Hite, G. L., Owers, J. E., and Rogers, R. C. (1984). The separation of real estate operations byspin-off. Real Estate Economics, 12(3):318–332.
Hou, K. and Robinson, D. T. (2006). Industry concentration and average stock returns. Journalof Finance, 61(4):1927–1956.
Huggett, M. (1993). The risk-free rate in heterogeneous-agent incomplete-insurance economies.Journal of Economic Dynamics and Control, 17:953 – 969.
Irvine, P. J. and Pontiff, J. (2009). Idiosyncratic return volatility, cash flows, and product marketcompetition. Review of Financial Studies, 22(3):1149–1177.
Jensen, M. C. (1986). Agency costs of free cash flow, corporate finance, and takeovers. AmericanEconomic Review, 76(2):323–329.
Jensen, M. C. and Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costsand ownership structure. Journal of Financial Economics, 3(4):305–360.
Kahle, K. M. and Walkling, R. A. (1996). The impact of industry classifications on financialresearch. Journal of Financial and Quantitative Analysis, 31(03):309–335.
Kogan, L. and Papanikolaou, D. (2013). Firm characteristics and stock returns: The role ofinvestment-specific shocks. Review of Financial Studies, 26(11):2718–2759.
Kydland, F. E. and Prescott, E. C. (1982). Time to build and aggregate fluctuations. Econometrica:Journal of the Econometric Society, pages 1345–1370.
Ling, D. C., Naranjo, A., and Ryngaert, M. (2012). Real estate ownership, leasing intensity, andvalue: Do stock returns reflect a firm’s real estate holdings? Journal of Real Estate Finance andEconomics, 44(1-2):184–202.
Lipton, M. (2006). Merger waves in the 19th, 20th and 21st centuries. The Davies Lecture, OsgoodeHall Law School, York University.
McIntosh, W., Ott, S. H., and Liang, Y. (1995). The wealth effects of real estate transactions: Thecase of REITs. Journal of Real Estate Finance and Economics, 10(3):299–307.
Michael, S. J., Arjun, C., and James, W. R. (2001). Real asset ownership and the risk and returnto stockholders. Journal of Real Estate Research, 22(1):199–212.
29
Myer, F., He, L. T., and Webb, J. R. (1992). Sell-offs of U.S. real estate: The effect of domesticversus foreign buyers on shareholder wealth. Real Estate Economics, 20(3):487–500.
Myers, S. C. (1984). The capital structure puzzle. Journal of Finance, 39(3):574–592.
Myers, S. C. and Majluf, N. S. (1984). Corporate financing and investment decisions when firmshave information that investors do not have. Journal of Financial Economics, 13(2):187–221.
Novy-Marx, R. (2007). An equilibrium model of investment under uncertainty. Review of FinancialStudies, 20(5):1461–1502.
Peress, J. (2010). Product market competition, insider trading, and stock market efficiency. Journalof Finance, 65(1):1–43.
Rodriguez, M. and Sirmans, C. (1996). Managing corporate real estate: Evidence from the capitalmarkets. Journal of Real Estate Literature, 4(1):13–33.
Slovin, M. B., Sushka, M. E., and Polonchek, J. A. (1990). Corporate sale-and-leasebacks andshareholder wealth. Journal of Finance, 45(1):289–299.
Spence, A. M. (1977). Entry, capacity, investment and oligopolistic pricing. Bell Journal of Eco-nomics, 8(2):534–544.
Tauchen, G. (1986). Finite state markov chain approximations to univariate and vector autoregres-sions. Economics Letters, 20:177–181.
Tirole, J. (1988). The Theory of Industrial Organization. Cambridge, MA: MIT press.
Tuzel, S. (2010). Corporate real estate holdings and the cross-section of stock returns. Review ofFinancial Studies, 23(6):2268–2302.
Weintraub, G. Y., Benkard, C. L., and Roy, B. V. (2009). Computational methods for obliviousequilibrium. Working Paper.
Weintraub, G. Y., Benkard, C. L., and Roy, B. V. (2010). Computational methods for obliviousequilibrium. Operations Research, 58(4):1247–1265.
Wenders, J. T. (1971). Excess capacity as a barrier to entry. Journal of Industrial Economics,20(1):14–19.
Yu, S. M. and Liow, K. H. (2009). Do retail firms benefit from real estate ownership? Journal ofProperty Research, 26(1):25–60.
Zhao, D. and Sing, T. F. (2015). Corporate real estate ownership and productivity uncertainty.Real Estate Economics, 44(2):521–547.
30
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
RE to TA(Market Value)
RE to TA (Book Value)
Non-residential RE to TA (Replacement Cost)
Figure 1: Ratios of real estate (RE) owned by non-farm non-financial corporations to total assets(TA). This real estate includes land and non-residential structures (production facilities, ware-houses, office buildings, land, and retail outlets (Source: Board of Governors of the Federal ReserveSystem)
31
8 firms 10 firmsSlope of line of best fit: -0.0026 Slope of line of best fit: -0.0050
.98
.99
11.
011.
02Av
erag
e be
ta
12 13 14 15 16 17Average capital-to-other-input ratio
.98
.99
11.
011.
02Av
erag
e be
ta
12 13 14 15 16 17Average capital-to-other-input ratio
15 firms 20 firmsSlope of line of best fit: -0.0009 Slope of line of best fit: -0.0015
.98
.99
11.
011.
02Av
erag
e be
ta
12 13 14 15 16 17Average capital-to-other-input ratio
.98
.99
11.
011.
02Av
erag
e be
ta
12 13 14 15 16 17Average capital-to-other-input ratio
30 firms 100 firmsSlope of line of best fit: -0.0019 Slope of line of best fit: -0.0002
.98
.99
11.
011.
02Av
erag
e be
ta
12 13 14 15 16 17Average capital-to-other-input ratio
.98
.99
11.
011.
02Av
erag
e be
ta
12 13 14 15 16 17Average capital-to-other-input ratio
Figure 2: Scatter plots of average firm betas against average capital-to-other-input ratios for firmssorted on their capital-to-other-input ratio in each period. Betas are computed using the previous100 periods and averages are taken over all 10,000 simulated periods.
32
0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.400
0.450
0.500
1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009
Median (Net Sales)
Mean (Net Sales)
Mean (Total Assets)
Figure 3: Mean and median of the sample’s three-year average HHI based on net sales and totalassets from 1973 to 2010
0.000
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
0.900
1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009
Low-concentration industries
High-concentration industries
Figure 4: Mean three-year average HHI (based on net sales) of the sample’s competitive andconcentrated industry groups from 1973 to 2010
33
————————————————————————–
34
0
1
2
3
4
5
6
7
8
9
10
1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
5
6
7
8
9
10
1 2 3 4 5 6 7 8 9 10
I I
II
(UL) (UH)
II
0
2
4
6
8
10
12
14
16
1 2 3 4 5 6 7 8 9 100
2
4
6
8
10
12
14
16
1 2 3 4 5 6 7 8 9 10
II
I I
II
(LL) (LH)
Decile-RE Portfolios Decile-RE Portfolios
Leve
red
Exce
ss R
etur
n (%
) U
nlev
ered
Exc
ess R
etur
n (%
)
Decile-RE Portfolios Decile-RE Portfolios
Leve
red
Exce
ss R
etur
n (%
) U
nlev
ered
Exc
ess R
etur
n (%
)
Figure 5: These are average decile portfolio returns. The two top quadrants (LL and LH) showaverage levered (L) excess portfolio returns in competitive (L: low concentration) industries andconcentrated (H: high concentration) industries, respectively. The bottom quadrants (UL and UH)show the average unlevered (U) excess portfolio returns in competitive (L) and concentrated (H )industries, respectively. The graphs labeled I (II ) show equally-weighted (value-weighted) averagereturns from 1973 to 2010.
35
Parameter Description Value
γ0 Demand shifter 25γ1 Demand elasticity 2ρZ Persistence of aggregate demand shock 0.8σZ Standard deviation of aggregate demand shock 0.025α Capital share of productivity 0.36ρA Persistence of idiosyncratic productivity shock 0.8σA Standard deviation of idiosyncratic productivity shock 0.05δ Capital depreciation rate 0.02
ηlow Low capital adjustment cost 0.8ηhigh High capital adjustment cost 1.6r Real interest rate 0.04w Price of fully depreciating input 1
Table 1: Model parameterization
36
Table 2: Sample’s Industry Concentrations
Mean Median SDV Min. Max. P25 P75
HHI (sales) 0.363 0.325 0.203 0.043 0.997 0.211 0.472HHI MA (sales) 0.361 0.327 0.194 0.043 0.996 0.216 0.469HHI (assets) 0.375 0.331 0.208 0.037 0.997 0.217 0.487HHI MA (assets) 0.373 0.334 0.199 0.039 0.997 0.222 0.487
These are descriptive statistics of the sample’s industry concentrations. The dataconsist of 7,736 industrial firms spanning the 37-year period from 1973 to 2010. Eachfirm is assigned to an industry according to its three-digit SIC codes, resulting in171 industries. HHI (sales) and HHI (assets) measure industry concentration usingnet sales and total assets, respectively, with HHI MA (sales) and HHI MA (assets)representing their respective 3-year moving averages. SDV, P25, and P75 stand forstandard deviation, 25th percentile, and 75th percentile, respectively.
Table 3: Industry Concentration Quintiles
Industry Level Firm Level(1) (2) (3) (1’) (2’) (3’)
Quintile HHI N. Ind. Ind. Years HHI Cont. N. Firms Firm Years
1 (Competitive) 0.142 26 988 0.004 869 33,0252 0.238 27 1,015 0.016 392 14,8843 0.329 27 1,012 0.031 287 10,9024 0.434 27 1,015 0.058 200 7,5835 (Concentrated) 0.663 26 995 0.121 145 5,491t-statistic 5–1 115.32 33.77
This table reports the characteristics of industry concentration quintiles. Each firm is assigned to anindustry according to its three-digit SIC codes, resulting in 171 industries. At the end of June eachyear, the industries are then grouped into quintiles according to their three-year average salses HHI. Thestatistics are the 37-year averages at the industry level in columns 1 to 3 and at the firm level in columns1’ to 3’. The variable HHI Cont. (HHI Contribution) is the average of the firms’ squared ratios of netsales to aggregate industry net sales.
37
Table 4: Characteristics of Firms in Competitive and Concentrated Industry Quintiles
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)Sales TA MV BM LEV LTDR RER1 RER2 RER3 PPER
Quintile 1: Competitive 1,275 1,062 1,654 0.783 0.332 0.177 0.300 0.365 0.246 0.283
Quintile 5: Concentrated 2,889 2,460 2,425 1.043 0.435 0.251 0.288 0.359 0.274 0.304
t-statistic 5–1 8.27 9.03 4.03 13.18 28.30 22.39 -4.50 -2.05 8.71 7.71
The study sample consists of 7,736 industrial firms spanning the 37-year period from 1973 to 2010. Firms are assigned toindustries according to their three-digit SIC codes, when are then divided into HHI quintiles. The above statistics are the37-year averages. Sales, TA, and MV, stand respectively for net sales, total assets, and market value of equity, in 2010U.S. dollars (millions). Book to market (BM ) is the ratio of book value of equity to market value of equity. LEV is theratio of book value of total liabilities to total market value of firm. LTDR is the ratio of book value of long-term debt toMV plus book value of long-term debt. RER1, RER2, and RER3 are, respectively, the ratio of buildings and capitalizedleases to property, plant, and equipment (PPE); the ratio of buildings, capitalized leases, construction in progress, andland to PPE; and the ratio of buildings and land to PPE. PPER is the ratio PPE to TA.
Table 5: Characteristics of Real Estate Decile Portfolios
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)TA Sales MV BM LTDR LEV RER1 RER2 RER3 PPER
Competitive IndustriesDecile 1: Low RE Portf. (1a) 331 373 301 0.647 0.126 0.267 0.031 0.076 0.031 0.206Decile 10: High RE Portf. (2a) 703 905 931 0.818 0.213 0.358 0.651 0.716 0.401 0.387t-statistic 10–1 7.60 9.90 8.32 5.67 16.15 14.31 321.3 206.4 64.47 32.73
Concentrated IndustriesDecile 1: Low RE Portf. (1b) 900 701 823 0.807 0.211 0.366 0.028 0.098 0.045 0.322Decile 10: High RE Portf. (2b) 3,059 6,011 4,580 1.201 0.252 0.438 0.678 0.754 0.531 0.324t-statistic 10–1 3.27 3.43 3.03 4.00 2.83 4.49 93.98 70.19 32.29 0.21
Cross Tests t-statistics1b vs. 1a 5.16 2.83 3.95 2.45 7.71 8.33 -2.27 3.12 2.84 10.082b vs. 2a 3.61 3.31 2.95 4.81 3.52 6.42 3.89 5.27 8.45 -7.0)
The study sample consists of 7,736 industrial firms spanning the 37-year period from 1973 to 2010. Firms are assigned to industriesaccording to their three-digit SIC codes. At the end of June each year, the industries are then grouped into quintiles according totheir three-year average salesHHI. Firms in the competitive and concentrated industry groups are then ranked into decile portfoliosaccording to their real estate ownership proxied by REI1 defined below. The above statistics are the portfolios’ 37-year averages.Sales, TA, and MV, stand respectively for net sales, total assets, and market value of equity, in 2010 U.S. dollars (millions). Bookto market (BM ) is the ratio of book value of equity (TA minus total liabilities plus balance sheet deferred taxes and investmenttax credit minus book value of preferred stocks) to market value of equity. LEV is the ratio of book value of total liabilities tototal market value of firm. LTDR is the ratio of book value of long-term debt to MV plus book value of long-term debt. RER1,RER2, and RER3 are, respectively, the ratio of buildings and capitalized leases to property, plant, and equipment (PPE); theratio of buildings, capitalized leases, construction in progress, and land to PPE; and the ratio of buildings and land to PPE.PPER is the ratio PPE to TA.
38
Tab
le6:
Aver
age
Lev
ered
Por
tfol
ioR
etu
rns
Portfolios
Low
23
45
67
89
High
High-L
ow
Panel
A:Competitive
Industries
Exce
ssR
etu
rn4.7
08.0
69.3
510.3
49.3
99.0
99.4
610.1
210.1
710.5
95.8
9(1.02)
(2.00)
(2.43)
(2.96)
(2.74)
(2.83)
(3.01)
(2.98)
(3.05)
(2.82)
(2.72)
Ind
.A
dj.
Ret
urn
-3.6
5-0
.51
-0.1
60.5
10.0
40.4
2-0
.37
0.6
21.2
21.1
64.8
1(-2.48)
(-0.53)
(-0.20)
(0.65)
(0.06)
(0.51)
(-0.40)
(0.78)
(1.67)
(1.20)
(3.06)
Panel
B:ConcentratedIn
dustries
Exce
ssR
etu
rn11.2
65.1
37.5
14.9
67.5
45.0
18.5
510.5
78.1
24.7
9-6
.46
(2.12)
(1.02)
(2.01)
(1.36)
(2.16)
(1.44)
(2.55)
(2.92)
(2.24)
(1.28)
(-1.45)
Ind
.A
dj.
Ret
urn
3.0
9-1
.21
-0.1
4-1
.70
1.4
2-0
.19
1.0
61.2
01.5
6-1
.31
-4.4
1(1.46)
(-0.52)
(-0.09)
(-1.19)
(1.14)
(-0.15)
(0.86)
(0.88)
(0.96)
(-0.81)
(-1.58)
Th
eta
ble
rep
ort
san
nu
alize
dva
lue-weigh
ted
aver
age
month
lyle
ver
edex
cess
retu
rns
(in
per
cent)
earn
edby
the
reales
tate
ran
ked
port
folios
inco
mp
etit
ive
an
dco
nce
ntr
ate
din
du
stry
gro
up
sfo
rth
e444-m
onth
per
iod
Ju
ly1973–Ju
ne
2010
an
dth
et-
stati
stic
s(i
talici
zed
figu
res
inp
are
nth
eses
).C
olu
mn
s1
(lab
eled
Low
)th
rou
gh
10
(lab
eled
High
)re
pre
sent
stock
port
folios
ran
ked
inin
crea
sin
gord
erof
real
esta
teow
ner
ship
,p
roxie
dby
the
rati
oof
bu
ild
ings
an
dca
pit
alize
dle
ase
sto
pro
per
ty,
pla
nt,
an
deq
uip
men
t(R
ER1
).ExcessReturn
isth
eex
cess
retu
rnover
the
risk
-fre
era
te,
pro
xie
dby
the
on
e-m
onth
Tre
asu
ryb
ill
rate
sfr
om
Ibb
ots
on
an
dA
ssoci
ate
s.In
d.Adj.
Return
stan
ds
for
ind
ust
ry-a
dju
sted
retu
rn,
the
exce
ssre
turn
over
the
aver
age
retu
rnof
an
ind
ust
ryp
ort
folio
incl
ud
ing
firm
ssh
ari
ng
the
sam
eth
ree-
dig
itS
ICco
des
.T
he
colu
mn
lab
eled
High-L
ow
rep
ort
sd
iffer
ence
sin
retu
rns
bet
wee
nth
eh
igh
an
dlo
wre
al
esta
ted
ecile
port
folios.
39
Tab
le7:
Aver
age
Un
leve
red
Por
tfol
ioR
etu
rns
Portfolios
Low
23
45
67
89
High
High-L
ow
Panel
A:Competitive
Industries
Exce
ssR
etu
rn2.5
95.0
06.2
37.2
36.4
36.3
86.2
56.5
76.6
16.7
34.1
4(0.25)
(0.98)
(1.45)
(2.02)
(1.79)
(1.92)
(1.92)
(1.91)
(1.91)
(1.64)
(2.23)
Ind
.A
dj.
Ret
urn
-3.0
4-0
.68
-0.0
70.5
60.2
00.5
2-0
.31
0.2
31.0
70.6
93.7
3(-2.22)
(-0.81)
(-0.25)
(0.90)
(0.36)
(0.78)
(-0.36)
(0.38)
(1.98)
(0.83)
(2.85)
Panel
B:ConcentratedIn
dustries
Exce
ssR
etu
rn8.1
73.8
54.2
92.3
24.5
61.6
45.6
46.3
54.9
02.9
9-5
.18
(1.61)
(0.44)
(0.71)
(-0.06)
(0.93)
(-0.35)
(1.56)
(1.65)
(1.17)
(0.31)
(-1.65)
Ind
.A
dj.
Ret
urn
3.5
2-0
.24
-0.5
5-1
.69
0.8
4-0
.87
0.6
30.7
51.1
4-0
.30
-3.8
3(2.27)
(-0.12)
(-0.67)
(-1.69)
(0.84)
(-1.05)
(0.72)
(0.79)
(1.19)
(-0.10)
(-1.88)
Th
eta
ble
rep
ort
san
nu
alize
dvalu
e-w
eighte
daver
age
month
lyu
nle
ver
edex
cess
retu
rns
(in
per
cent)
earn
edby
the
reales
tate
ran
ked
port
folios
inco
mp
etit
ive
an
dco
nce
ntr
ate
din
du
stry
gro
up
sfo
rth
e444-m
onth
per
iod
Ju
ly1973–Ju
ne
2010
an
dth
et-
stati
stic
s(i
talici
zed
figu
res
inp
are
nth
eses
).U
nle
ver
edre
turn
calc
ula
tion
sass
um
en
ota
ke
taxes
an
da
cost
of
the
deb
tof
7%
acr
oss
the
board
.C
olu
mn
s1
(lab
eled
Low
)th
rou
gh
10
(lab
eled
High
)re
pre
sent
the
stock
port
folios
ran
ked
inin
crea
sin
gord
erof
real
esta
teow
ner
ship
,p
roxie
dby
the
rati
oof
bu
ild
ings
an
dca
pit
alize
dle
ase
sto
pro
per
ty,
pla
nt,
an
deq
uip
men
t(R
ER1
).ExcessReturn
isth
eex
cess
retu
rnover
the
risk
-fre
era
te,
pro
xie
dby
the
on
e-m
onth
Tre
asu
ryb
ill
rate
sfr
om
Ibb
ots
on
an
dA
ssoci
ate
s.In
d.Adj.
Return
stan
ds
for
ind
ust
ry-a
dju
sted
retu
rn,
the
exce
ssre
turn
over
the
aver
age
retu
rnof
an
ind
ust
ryp
ort
folio
incl
ud
ing
firm
ssh
ari
ng
the
sam
eth
ree-
dig
itS
ICco
des
.T
he
colu
mn
lab
eled
High-L
ow
rep
ort
sd
iffer
ence
sin
retu
rns
bet
wee
nth
eh
igh
an
dlo
wre
al
esta
ted
ecile
port
folios.
40
Tab
le8:
Ab
nor
mal
Por
tfol
ioR
etu
rns
Portfolios
Low
23
45
67
89
High
High-L
ow
Com
pet
itiv
eIn
du
stri
es(1
)-3
.08
-0.0
21.3
23.0
11.7
21.5
42.1
02.4
11.9
41.7
64.8
5(-1.76)
(-0.01)
(1.12)
(2.85)
(1.86)
(1.74)
(2.23)
(2.51)
(1.81)
(1.31)
(2.80)
Con
centr
ate
dIn
du
stri
es(2
)2.5
2-1
.63
-1.0
8-2
.71
-0.8
7-4
.21
0.9
4-0
.17
-1.0
8-2
.53
-5.0
5(0.96)
(-0.73)
(-0.72)
(-1.69)
(-0.61)
(-3.20)
(0.77)
(-0.10)
(-0.72)
(-1.39)
(-1.65)
Observa
tions
444
444
444
444
444
444
444
444
444
444
444
Adj.
R2Estim
ation(1)
0.787
0.813
0.838
0.843
0.872
0.864
0.844
0.859
0.817
0.791
0.171
Adj.
R2Estim
ation(2)
0.505
0.564
0.609
0.570
0.633
0.672
0.642
0.537
0.618
0.513
0.083
Th
ista
ble
rep
ort
ses
tim
ate
dab
norm
al
port
folio
retu
rns
(in
per
cent
an
dan
nu
alize
d)
from
valu
e-w
eighte
dm
onth
lyu
nle
ver
edex
cess
port
folio
retu
rns
inco
mp
etit
ive
an
dco
nce
ntr
ate
din
du
stri
es.
Th
eit
alici
zed
figu
res
inp
are
nth
eses
are
the
rob
ust
t-st
ati
stic
sof
the
coeffi
cien
tes
tim
ate
s.U
nle
ver
edre
turn
calc
ula
tion
sass
um
en
ota
xes
an
da
cost
of
deb
tof
7%
acr
oss
the
board
.T
hes
ees
tim
ate
sare
base
don
the
thre
e-fa
ctor
mod
el4
con
sist
ing
of
the
exce
ssm
ark
etre
turn
(rm−r f
)an
dth
etw
oF
am
aan
dF
ren
chst
ock
risk
fact
ors
,sm
ban
dhml.
Colu
mn
s1
(lab
eled
Low
)th
rou
gh
10
(lab
eled
High
)re
pre
sent
the
stock
port
folios
ran
ked
inin
crea
sin
gord
erof
real
esta
teow
ner
ship
,p
roxie
dby
the
rati
oof
bu
ild
ings
an
dca
pit
alize
dle
ase
sto
pro
per
ty,
pla
nt,
an
deq
uip
men
t(R
ER1
).r m−r f
rep
rese
nts
the
valu
e-w
eighte
dm
onth
lyre
turn
son
CR
SP
-lis
ted
stock
sm
inu
sth
eon
e-m
onth
Tre
asu
rybil
lra
tes
from
Ibb
ots
on
an
dA
ssoci
ate
s.sm
ban
dhml
are
the
aver
age
retu
rns
on
small-s
tock
port
folios
min
us
the
aver
age
retu
rnon
big
-sto
ckp
ort
folios
an
dth
eaver
age
retu
rnon
valu
e-st
ock
port
folios
min
us
the
aver
age
retu
rnon
gro
wth
-sto
ckp
ort
folios,
resp
ecti
vel
y.M
onth
lyst
ock
retu
rnd
ata
are
from
CR
SP
,w
ithr m
,r f
,sm
b,an
dhml
from
Ken
net
hF
ren
ch’s
web
site
.T
he
colu
mn
lab
eled
High-L
ow
rep
ort
sab
norm
al
retu
rns
from
diff
eren
ces
inex
cess
retu
rns
bet
wee
nth
eh
igh
an
dlo
wre
al
esta
tedec
ile
port
folios.
41
Table 9: High-Low Portfolio Factor Loadings
Competitive ConcentratedIndustries Industries
alpha 4.85 -5.05(2.80) (1.65)
rm − rf -0.147 -0.165(4.56) (2.87)
smb -0.215 -0.188(4.66) (2.29)
hml 0.164 0.282(3.39) (3.26)
Observations 444 444Adj. R-squared 0.171 0.083
This table presents the regression results of differences in value-weightedunlevered excess returns between the high and low real estate decile port-folios. Average unexplained returns (alpha) are in percent and annualized.rm,t−rf,t represents the value-weighted monthly returns on CRSP-listedstocks minus the one-month Treasury bill rates from Ibbotson and As-sociates. smb and hml are the average returns on small-stock portfoliosminus the average return on big-stock portfolios and the average returnon value-stock portfolios minus the average return on growth-stock port-folios, respectively. The italicized figures in parentheses are the robustt-statistics of the coefficient estimates.
Table 10: Abnormal Returns on High–Low Portfolio from July 1973 to June 2005
Competitive Industries Concentrated Industries(1) (2) (1´) (2´)
Levered Returns Unlevered Returns Levered Returns Unlevered Returns
alpha 6.73 4.64 -10.63 -6.84(3.08) (2.53) (-2.31) (-2.10)
Risk Factors Yes Yes Yes Yes
Observations 390 390 390 390Adj. R2 0.161 0.186 0.127 0.095
This table presents abnormal returns from regression of differences value-weighted levered and unlevered excessreturns between the high and low real estate decile portfolios from 1973 to 2005. Columns (1 ) and (2 ) applyto competitive industries; columns (1´) and (2´) are for concentrated industries. These estimates are based onthe three-factor model 4 consisting of the excess market return (rm,t − rf,t) and the two Fama and French stockrisk factors, smb and hml. Average unexplained excess returns are annualized values. rm,t − rf,t represents thevalue-weighted monthly returns on CRSP-listed stocks minus the one-month Treasury bill rates from Ibbotson andAssociates. smb and hml are the average returns on small-stock portfolios minus the average return on big-stockportfolios and the average return on value-stock portfolios minus the average return on growth-stock portfolios,respectively.
42
Table 11: Quintile Portfolios and Alternative Industry Classifications
(1) (2) (3) (4) (5)Base Case Quintile Portf. 2-digit SIC FF Industries HHI (asset)
Competitive IndustriesExcess Return 4.14 2.47 3.44 3.45 4.17
(2.23) (1.60) (1.79) (2.22) (2.25)Ind. Adj. Return 3.73 2.42 3.62 3.02 4.07
(2.85) (2.92) (2.39) (2.58) (2.95)
Concentrated IndustriesExcess Return -5.18 -3.21 -4.46 -12.15 -4.06
(-1.65) (-1.43) (-1.46) (-2.30) (-1.45)Ind. Adj. Return -3.83 -1.85 -1.67 -9.31 -1.85
(-1.88) (-1.42) (-0.68) (-1.91) (-0.93)
The table reports differences in value-weighted average unlevered excess and industry-adjusted returns (in per-cent) between the high and low real estate decile portfolios during the 444-month period from July 1973 to June2010 and their t-statistics (italicized figures in parentheses). Column 1 represents the Base Case (High-Lowreturns from Table 7). Column 2 lists returns on quintile, rather than decile, portfolios based on RER1. Incolumn 3, firms are assigned to industries according to their two-digit SIC codes and in column 4 according tothe 48 Fama and industry groups. In column 5, industry concentrations are based on total assets, rather thannet sales.
Table 12: Alternative Real Estate Measures
(1) (2) (3) (4)Base Case RER2 PPER ARER
Competitive IndustriesExcess Return 4.14 4.70 4.33 4.01
(2.23) (2.18) (1.56) (2.69)Ind. Adj. Return 3.73 4.28 4.86 4.53
(2.85) (2.84) (2.92) (3.26)
Concentrated IndustriesExcess Return -5.18 -5.01 -3.02 -1.03
(-1.65) (-1.52) (-0.99) (-0.41)Ind. Adj. Return -3.83 -2.78 -0.36 -1.11
(-1.88) (-1.33) (-0.17) (-0.47)
The table reports differencesf in value-weighted average unlevered excess and industry-adjustedreturns (in percent) between the high and low real estate decile portfolios during the 444-monthperiod from July 1973 to June 2010 and their t-statistics (italicized figures in parentheses). Incolumn 1, the Base Case (High-Low returns from Table 7), firms are sorted into decile portfoliosaccording to RER1, the ratio of buildings and capitalized leases to property, plant, and equipment(PPE). The figures in columns 2 to 4 are based on different measures of real estate ownership. RER2and PPER are respectively the ratio of buildings, capitalized leases, construction in progress, andland to PPE, and the ratio PPE to total assets (TA). ARER is firm RER1 minus the averageindustry RER1 (i.e., firms belonging to the same three-digit SIC codes).
43
Tab
le.1
:E
quall
y-W
eigh
ted
Ave
rage
Por
tfol
ioR
etu
rns
Portfolios
Low
23
45
67
89
High
High-L
ow
Panel
A:Competitive
Industries
Exce
ssR
etu
rn9.7
612.6
611.9
214.0
613.9
211.7
914.8
914.1
112.6
114.9
25.1
6(1.95)
(2.85)
(2.73)
(3.65)
(3.51)
(3.21)
(3.91)
(3.72)
(3.29)
(3.57)
(2.25)
Ind
.A
dj.
Ret
urn
-2.7
10.2
0-1
.35
0.7
10.6
2-0
.68
0.7
40.5
0-0
.04
2.1
64.8
6(-1.86)
(0.18)
(-1.36)
(0.76)
(0.59)
(-0.67)
(0.73)
(0.51)
(-0.05)
(2.01)
(2.80)
Panel
B:ConcentratedIn
dustries
Exce
ssR
etu
rn17.1
512.1
110.5
77.1
811.0
09.4
111.8
912.0
27.7
48.8
9-8
.26
(2.82)
(1.95)
(2.38)
(1.75)
(2.79)
(2.40)
(2.78)
(2.45)
(1.93)
(1.99)
(-1.65)
Ind
.A
dj.
Ret
urn
4.2
41.2
9-0
.03
-2.3
30.6
5-0
.33
2.1
5-0
.24
-1.6
5-2
.18
-6.4
2(1.47)
(0.44)
(-0.02)
(-1.08)
(0.34)
(-0.18)
(1.03)
(-0.09)
(-0.69)
(-0.93)
(-1.72)
Th
eta
ble
rep
ort
san
nu
alize
deq
ually-w
eighte
daver
age
month
lyle
ver
edre
turn
s(i
np
erce
nts
)ea
rned
by
the
real
esta
tera
nked
port
folios
inth
eco
mp
etit
ive
an
dco
nce
ntr
ate
din
du
stry
gro
up
s)fo
rth
e444-m
onth
per
iod
from
Ju
ly1973
toJu
ne
2010
an
dth
eirt-
stati
stic
s(i
talici
zed
figu
res
inp
are
nth
eses
).C
olu
mn
s1
(lab
eled
Low
)th
rou
gh
10
(lab
eled
High
)re
pre
sent
the
stock
port
folios
ran
ked
inin
crea
sin
gord
erof
real
esta
teow
ner
ship
,p
roxie
dby
the
rati
oof
bu
ild
ings
an
dca
pit
alize
dle
ase
sto
pro
per
ty,
pla
nt,
an
deq
uip
men
t(R
ER1
).ExcessReturn
isth
eex
cess
retu
rnover
the
risk
-fre
era
te,
pro
xie
dby
the
on
e-m
onth
Tre
asu
ryb
ill
rate
sfr
om
Ibb
ots
on
an
dA
ssoci
ate
s.In
d.
Adj.
Return
stan
ds
for
ind
ust
ry-a
dju
sted
retu
rnan
dis
the
exce
ssre
turn
over
the
aver
age
retu
rnof
an
indu
stry
port
folio
regro
up
ing
firm
ssh
ari
ng
the
sam
eth
ree-
dig
itS
ICco
des
.T
he
colu
mn
lab
eled
High-L
ow
rep
ort
sd
iffer
ence
sin
retu
rns
bet
wee
nth
eh
igh
an
dlo
wre
al
esta
ted
ecile
port
folios.
44
Tab
le.2
:A
bn
orm
alP
ortf
olio
Ret
urn
sfr
omL
ever
edE
xce
ssR
eturn
s
Portfolio
Low
23
45
67
89
High
High-L
ow
Com
pet
itiv
eIn
du
stri
es(1
)-3
.54
0.2
51.5
73.5
42.2
91.9
13.2
03.8
23.1
73.0
26.5
6(-1.90)
(0.17)
(1.23)
(2.90)
(2.08)
(1.72)
(2.80)
(3.17)
(2.35)
(2.01)
(3.20)
Con
centr
ate
dIn
du
stri
es(2
)3.7
9-3
.68
-2.0
6-3
.47
-1.5
4-4
.76
0.5
00.2
8-0
.56
-3.7
6-7
.56
(1.07)
(-1.15)
(-0.91)
(-1.51)
(-0.76)
(-2.44)
(0.25)
(0.12)
(-0.24)
(-1.51)
(-1.75)
Observa
tions
444
444
444
444
444
444
444
444
444
444
444
Adj.
R2Estim
ation(1)
0.844
0.866
0.895
0.884
0.902
0.887
0.874
0.881
0.844
0.847
0.143
Adj.
R2Estim
ation(2)
0.569
0.616
0.647
0.618
0.676
0.701
0.666
0.600
0.607
0.576
0.109
Th
ista
ble
rep
ort
ses
tim
ate
dab
norm
al
port
folio
retu
rns
(in
per
cent
an
dan
nu
alize
d)
from
valu
e-w
eighte
dm
onth
lyle
ver
edex
cess
port
folio
retu
rns
inco
mp
etit
ive
ind
ust
ries
.T
he
italici
zed
figu
res
inp
are
nth
eses
are
the
rob
ust
t-st
ati
stic
sof
the
coeffi
cien
tes
tim
ate
s.T
hes
ees
tim
ate
sare
base
don
the
thre
e-fa
ctor
mod
el4
con
sist
ing
of
the
exce
ssm
ark
etre
turn
(rm,t−r f,t
)an
dth
etw
oF
am
aan
dF
ren
chst
ock
risk
fact
ors
,sm
ban
dhml.
Colu
mn
s1
(lab
eled
Low
)th
rou
gh
10
(lab
eled
High
)re
pre
sent
the
stock
port
folios
ran
ked
inin
crea
sing
ord
erof
real
esta
teow
ner
ship
,p
roxie
dby
the
rati
oof
bu
ild
ings
an
dca
pit
alize
dle
ase
sto
pro
per
ty,
pla
nt,
an
deq
uip
men
t(R
ER1
).r m
,t−r f,t
rep
rese
nts
the
valu
e-w
eighte
dm
onth
lyre
turn
son
CR
SP
-lis
ted
stock
sm
inu
sth
eon
e-m
onth
Tre
asu
ryb
ill
rate
sfr
om
Ibb
ots
on
an
dA
ssoci
ate
s.sm
ban
dhml
are
the
aver
age
retu
rns
on
small-s
tock
port
folios
min
us
the
aver
age
retu
rnon
big
-sto
ckp
ort
folios
an
dth
eaver
age
retu
rnon
valu
e-st
ock
port
folios
min
us
the
aver
age
retu
rnon
gro
wth
-sto
ckp
ort
folios,
resp
ecti
vel
y.T
he
colu
mn
lab
eled
High-L
ow
rep
ort
sab
norm
al
retu
rns
from
diff
eren
ces
inre
turn
sb
etw
een
the
hig
han
dlo
wre
al
esta
ted
ecile
port
folios.
45
Appendix: Model
Equilibrium
This appendix explains the equilibrium concept. This concept, extended oblivious equilibrium, istaken from Weintraub et al. (2010). In an extended oblivious equilibrium, rather than tracking thestate of every other firm as a state variable (as would be the case in a Markov perfect equilibrium ofEricson and Pakes (1995)), each firm instead tracks the long-run average industry state, conditionalon today’s value of the aggregate industry shock. That is, rather than tracking every other firm’scapital holdings and productivity and responding appropriately, each firm follows a strategy thatis “oblivious” in the sense that it responds to a hypothetical average firm and is oblivious to theactions of any particular competitor. Furthermore, the equilibrium definition in Weintraub et al.(2010) is simplified by disallowing firm entry and exit and by allowing the firms to condition theirbehavior on only the current value of the aggregate shock, rather than a long history of shocks.This greatly reduces the computational burden of solving for an equilibrium and also allows thenumber of firms in the industry to be varied easily.
In the model, an extended oblivious equilibrium is a strategy µ∗ = µ∗ (Ki,t, Ai,t, Zi,t) such that
supµV (Ki,t, Ai,t, Zt|µ, µ∗) = V (Ki,t, Ai,t, Zt|µ∗, µ∗) .
for every Ki,t, Ai,t, and Zi,t and for every i where V (Ki,t, Ai,t, Zt|µ′, µ) is the value function solvingthe Bellman equation (2) that results when firm i plays strategy µ′ and all N − 1 of his opponentsplay strategy µ. This equilibrium concept has the advantage that it converges to a perfectlycompetitive equilibrium as the number of firms increases.
Computational Approach
This appendix describes the algorithm used to numerically solve the model for an extended obliviousequilibrium with N firms and all other parameters given. This algorithm is presented in a moregeneral form in Weintraub et al. (2009).
The steps of the algorithm are as follows:
1. Use the method of Tauchen (1986) to discretize the AR(1)’s for Zt and Ai,t, obtaining fivenodes and the corresponding transition matrices for each.
2. Discretize the Ki,t and Li,t spaces on a fine grid.
3. Conjecture an extended oblivious equilibrium strategy µGuess (Ki,t, Ai,t, Zi,t) (i.e., the policyfunctions for Li,t and Ki,t+1).
4. Using µGuess (Ki,t, Ai,t, Zt), compute the average firm state conditional on Zt; call it sµ(Zt).This is done in a way that is similar to finding the stationary distribution of assets in aHuggett (1993) or Aiyagari (1994) economy for each value of the aggregate shock. Note thatsµ(Zt) is simply a scalar for each Zt.
5. Solve firm i’s Bellman equation (2) using value function iteration with∑
j 6=iQj,t = (N − 1) ·sµ(Zt). The implied policy functions/extended oblivious equilibrium strategy µNew (Ki,t, Ai,t, Zi,t)is obtained in the process.
46
6. Check if sup ‖ µGuess (Ki,t, Ai,t, Zi,t) − µNew (Ki,t, Ai,t, Zi,t) ‖< ε for some small toleranceε is true. If so, µGuess (Ki,t, Ai,t, Zi,t) is an extended oblivious equilibrium, µ∗. If not, setµGuess (Ki,t, Ai,t, Zi,t) = µNew (Ki,t, Ai,t, Zi,t) and return to step 4.
After an extended oblivious equilibrium has been found, the behavior of the N firms is simulatedfor 11,000 time periods (discarding the first 1,000 to eliminate dependence on initial conditions) toobtain the sequences of Ki,t, Li,t, and βi,t.
47