Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Frothy Housing Markets and Local Stock-Price Movements Christopher W. Anderson University of Kansas The School of Business, University of Kansas, 1300 Sunnyside Ave., Lawrence, KS, 66045-7585; Tel: 785-864-7340; Email: [email protected]. Eli Beracha East Carolina University College of Business, East Carolina University, 3129 Bate Building, Greenville, NC 27858-4353. Tel: 252-328-5824; Email: [email protected].
Abstract We investigate how conditions in residential real estate markets affect the pricing of stocks for companies with headquarters in metro areas across the United States. We find that stocks of firms headquartered in ‘hot’ residential real estate markets experience higher risk adjusted returns relative to stocks from ‘cold’ markets. We also find that stocks of firms located in hot real estate markets experience stronger return comovement with same-city stocks compared to stocks of firms located in cold markets. These conditional patterns in local stock prices are especially prevalent during the 1999 to 2004 period that coincides with the start of the most recent housing bubble. These findings suggest that shocks to households’ real estate wealth condition habitat effects observable in the pricing of local stocks.
Keywords Residential real estate · Wealth effects · Return comovement · Headquarters city effect · Household finance
1
Introduction
Location matters in portfolio allocation decisions by investors. Specifically, U.S.
investors display a tendency to overweight local stocks when forming investment portfolios
(Coval and Moskowitz 1999; Huberman 2001; Zhu 2003). Additional evidence on this so-
called local bias has been found with respect to liquidity for stocks of firms located in rural
versus urban areas (Loughran and Schultz 2005), trading volume in cities affected by adverse
weather and ethnic holidays (Loughran and Schultz 2004), stock valuation of companies
headquartered in areas with few other local firms (Hong, Kubick, and Stein 2008), and
comovement among returns for firms that are headquartered near each other (Pirinsky and
Wang 2006; Barker and Loughran 2007; Anderson and Beracha 2008). In short, there are
discernible patterns in trading volume, prices, and returns of local stocks consistent with
habitat effects related to social interaction and sharing of information among local investors
who engage in correlated trading (Hong, Kubick, and Stein 2004, 2005; Ozsolyev 2005;
Barberis, Shleifer, and Wurgler 2005).
Our study is one of the few that investigates how observable attributes of securities
pricing are affected by price trends in residential real estate that may condition trading by
local investors. Because investments in housing are often the largest portion of household
wealth for a large segment of the population, this gap in our knowledge appears material.
Furthermore, rising home values in many markets combined with falling mortgage interest
rates were a major source of wealth and liquidity for households over the past several years
(Greenspan and Kennedy 2005). It is widely accepted that housing wealth stimulates
consumption (Case, Quigley, and Shiller 2003, 2005; Bostic, Gabriel, and Painter 2005;
Benjamin and Chinloy 2008). Economic theory suggests that home-related wealth conditions
2
portfolio investment decisions by households (e.g., Palia, Qi, and Wu 2009) and recent
research shows that changes in home prices are positively related to local gross metropolitan
product (Miller, Peng, and Sklarz 2009). Consequently, it seems likely that during the recent
housing bubble rising prices in some areas of the United States and the liquidity provided by
opportunities to refinance and enjoy mortgage equity withdrawals have increased
discretionary investment activity and increased stock market participation.
In this paper we investigate how conditions in residential real estate markets affect
the pricing of local stocks for U.S. metropolitan areas over the fifteen-year period from July
1989 to June 2004. We use the well-established method of classifying stocks by the cities in
which firms are headquartered. We then classify cities according to trends in residential real
estate prices as measured by the repeat-sale home price indexes (HPIs) produced for
metropolitan statistical areas by the Office of Federal Housing Enterprise Oversight
(OFHEO). We then investigate whether extreme changes in home prices condition stock
returns in a manner consistent with local biases in stock trading activity. In particular, we
hypothesize that shocks to household wealth due to home price appreciation may stimulate
demand for stocks of locally headquartered firms and lead to higher than normal returns for
such stocks, a time series manifestation of Hong, Kubick, and Stein’s (2008) “only game in
town” story. We also hypothesize that rising housing wealth will be associated with a greater
degree of return comovement among local stocks to the extent that local investors subject to
shared housing-related wealth shocks invest in local stocks in response to similar information
and at similar times (Pirinsky and Wang 2006). Finally, we expect that the conditional
relations between local housing markets and local stock pricing will be most discernible
during the 1999-2004 portion of our sample period. This period coincides with the beginning
3
of the most recent housing bubble and manifests a greater degree of dispersion of home price
movements across U.S. cities.
We find results consistent with our hypotheses. First, we find that stocks
headquartered in residential real estate markets categorized as ‘hot’ experience higher
monthly returns on average than stocks from ‘cold’ markets. These excess returns persist
after we adjust firm-level returns for exposure to risk factors. Second, we find that stocks of
firms located in hot real estate markets experience stronger return comovement with other
same-city stocks compared to stocks located in cold real estate markets. Finally, consistent
with our expectations these conditional patterns in stock returns are most pronounced during
the frothy housing markets period of 1999-2004. Overall, these findings suggest that
changes in investors’ real estate wealth affect trading activity and pricing of local stocks.
The remainder of this paper is organized as follows. The immediately following three
sections review evidence on the local bias among U.S. investors, section discusses extant
research on housing wealth and its effects on consumer and investor behavior, and motivates
our linking of changes in housing-related wealth to manifestations of local bias. We then
discuss our data and methods. Finally, we present empirical results and conclude.
Housing Wealth and Portfolio Investment Decisions
U.S. residential real estate market is valued at more than $20 trillion and is an asset
class held by about two-thirds of U.S. households. In recent years, residential real estate in
the United States experienced rapid price appreciation in some areas including parts of the
west coast, the east coast, Arizona, and Las Vegas where home prices nearly doubled within
a few years. At the same time, rent prices and personal income increased at a slower pace
4
and thus did not provide fundamental support to housing price appreciation in such areas
(Case and Shiller 2003; Himmelberg, Mayer, and Sinai 2005). The price level of residential
real estate in these markets became a growing economic concern. This issue received
attention from the popular press that extensively referred to the situation as the ‘housing
bubble,’ as well as from the former Federal Reserve chairman, Alan Greenspan, who
described the level of housing prices in some areas as unsustainable in his testimony before
the Joint Economic Committee (Greenspan 2005). On the other hand, housing prices in some
parts of the U.S., e.g., Texas and Tennessee, were almost flat for the most of that time period,
resulting in large geographic dispersion in home price changes across the United States.
Since 2007 these patterns have reversed, as areas in which home prices rose most rapidly
have witnessed rapid price declines (Beracha and Hirschey 2009).
Since residential real estate is a widely held and often highly levered asset -- often the
largest component of an individual’s asset portfolio -- it affects personal consumption and
investment decisions in a material way (Benjamin, Chinloy, and Jud 2008). Bostic, Gabriel,
and Painter (2005) use combined data from the Survey of Consumer Finance and Consumer
Expenditure Survey and show that homeowners’ consumption increases with the value of
their homes. Using data from the United Kingdom, Campbell and Cocco (2005) confirm the
positive relation between housing wealth and consumption, and they show that this relation is
discernible at the regional level within a country. Case, Quigley, and Shiller (2005) find that
housing wealth has a large effect on household consumption in 14 developed countries
including the United States. They argue that the positive relation between housing wealth
and consumption results from relaxation of borrowing constraints through a second mortgage
or a line of credit that facilitate extraction of liquidity from housing equity. Slacalek (2006)
5
shows similar cross-national results, and he finds that the effect of housing wealth on
consumption has increased over the past fifteen years, presumably because of easier access to
mortgage credit. The effect of changes in housing value on consumption behavior is greater
than the effect of changes in stock market wealth, in part because homeowners perceive
housing wealth to be more permanent (Benjamin, Chinloy, and Jud 2004a; Carroll, Otsuka,
and Slacalek 2006; Kishor 2007).
The wealth and risk associated with the volatile and illiquid nature of a house affects
not only consumption but also the level of risk homeowners are willing to bear from other
investments. Consequently, choices about investments in stocks and bonds are conditioned
by the household’s residential real estate holding and its overall risk tolerance (Flavin and
Yamashita 2002; Benjamin, Chinloy, and Jud 2004a; Yao and Zhang 2005; Piazzesi,
Schneider, and Tuzel 2007; Kullman and Siegel 2007). In particular, Cocco (2004) uses a
model of optimal portfolio and consumption decisions, parameterized by data from the Panel
Study of Income Dynamics, and finds that housing price risk and transaction cost crowds out
stockholdings, but he also shows that this effect is larger for low financial net worth. He also
shows that mortgage loan and stockholdings are positively related. Palia, Qi, and Wu (2009)
confirm Cocco’s (2004) result that background risks from housing (as well as from labor
income) decreases stock market participation, but also recognize that housing provides a
borrowing channel through which stock investments can be financed. Because of these
fundamental relations between wealth invested in residential real estate and wealth invested
in financial securities, integration in pricing between the two markets can often be detected
(Quan and Titman 1999; Jud and Winkler 2002; Kullman 2003; Cannon, Miller, and Pandher
2006; Anderson and Beracha 2009)
6
The specific geographic location and pricing characteristics of the homeowner’s
primary residence introduces a natural and unavoidable local bias in his or her asset portfolio.
In the next section we review evidence that shows that even when opportunities for
geographic diversification of financial assets are easy to implement, investors still appear to
form portfolios with a bias toward stocks of locally headquartered companies.
Local Bias in Portfolio Investment
The tendency of investors to deviate from global diversification and instead hold
portfolios of financial assets that disproportionably overweight domestic securities is a
widely observed behavior (French and Porteba 1991; Kang and Stulz 1997; Keloharju and
Grinblatt 2001; Dahlquist, Pinkowitz, Stulz, and Williamson 2003; Chan, Covrig, and Ng
2005). Differences in political and economic systems, tax implications, language, culture,
and limited access to many foreign markets explain at least some of the home bias
phenomenon.
Even among investments within a country, resident investors hold portfolios that are
locally biased. Coval and Moskowitz (1999) find that U.S. money managers hold portfolios
of companies that are located about 10% closer to their offices, on average, than randomly
formed domestic portfolios. This local preference is particularly strong for small, highly
levered firms that produce non-traded goods, suggesting that information asymmetry is the
basis for the local biased behavior. Zhu (2003) observes local bias behavior among
individual investors that increases with firms’ advertising spending, consistent with the
familiarity hypothesis proposed by Huberman (2001).
7
The existence of the local bias phenomenon within the United States appears to affect
stock trading volume, valuation, and patterns in returns. Loughran and Schultz (2004) show
that time zones associated with the city where companies’ headquarters are located affect
intraday trading. Similarly, they show that religious holidays and blizzards, events that affect
specific cities more than others, influence trading volume of companies headquartered in
those cities. Loughran and Schultz (2005) show that patterns in liquidity and price formation
differ for stocks of firms headquartered in rural versus urban areas. Hong, Kubik, and Stein
(2008) show that local bias has implications for stock prices in some regions via an “only
game in town” effect. Specifically, companies located in areas with relatively few firms per
capita are priced at a premium to companies that locate in regions with many firms per
capita.
Pirinsky and Wang (2006) and Anderson and Beracha (2008) show that stocks for
companies headquartered in the same city experience comovement in their returns. The level
of the local comovement of a stock is measured as the time-series sensitivity of its returns to
the return of an index of stocks from the same geographic area. This price comovement is
not explained by the market as a whole, industry classification, or regional macroeconomic
conditions. Similarly, Barker and Loughran (2007) show that pairwise correlations in returns
among stocks in the S&P 500 index increase with proximity of their headquarters locations.
The proximity effect among U.S. stocks may result from an investor habitat effect
manifest when groups of investors who share a literal or virtual common habitat concentrate
their attention on certain classes of securities (Barberis, Shleifer, and Wurgler, 2005).
Correlated trading by investors in a community may result in return comovement among
securities not otherwise attributable to underlying fundamentals or exposure to common risk
8
factors. Correlated trading might be induced by privileged access to locally generated
information or by rumors or noise trading among local traders who share social networks
(Hong, Kubik, and Stein, 2004, 2005; Ozsolyev, 2005). The habitat effect is similar to a
category effect such as that demonstrated by Barberis, Shleifer, and Wurgler (2005), Kumar
and Lee (2006), and Ambrose, Lee, and Peek (2007).
Hypotheses
In spite of many studies on locally biased stock portfolios and the obvious local bias
in residential real estate investments, the effect of the geographically concentrated wealth
invested in housing on the homeowner’s locally biased stocks portfolio has not yet been
studied empirically. Our study expands the findings of Hong, Kubick, and Stein (2008) and
Pirinsky and Wang (2006) and also draws on the evidence provided by the literature on the
subject of housing wealth and investment decisions. We premise our hypotheses on the
assumption that household wealth related to rapid housing appreciation leads to more stock
market participation, in general, and triggers greater trading activity in local stocks, in
particular. We therefore expect that patterns in stock pricing frequently attributed to locally
biased trading activity will be more pronounced in areas with more positive changes in
residential real estate prices.
Our first hypothesis predicts that the returns on stocks of firms headquartered in areas
that experience rapid price appreciation are higher than returns on stocks of firms
headquartered in areas of lower price appreciation. Hong, Kubik, and Stein (2008) argue and
provide evidence that higher demand for local stocks in some areas pushes stock prices to a
9
higher price level. If real estate wealth increases stock market participation and demand for
local stocks then we should observe higher returns for such stocks.
We next hypothesize that return comovement among same-city stocks, as
documented by Pirinsky and Wang (2006), is positively related to local residential real estate
market performance. Specifically, we expect to observe more positive return comovement
between stocks of firms headquartered in cities that experience rapid residential real estate
appreciation.
Finally, we hypothesize that the effect of residential real estate on portfolio returns
and return comovement is more pronounced during time periods with rapid appreciation in
the national residential real estate prices and greater dispersion in appreciation across cities,
such as the beginning of the housing bubble observed during the early 2000s. Our third
hypothesis is based on the fact that higher average national price appreciation is generally
associated with larger dispersion in price appreciation across U.S. cities. Consequently, we
anticipate home-price conditioned habitat effects to be more distinct during booming real
estate periods.
Data and Methods
Stock Price Data
We identify all firms with common stock price data available from the University of
Chicago’s Center for Research in Security Prices (CRSP) that also have company
headquarters data available from Standard & Poor’s COMPUSTAT database. As has
become the convention in the emerging literature on economic geography and finance, we
allocate each stock to a metropolitan statistical area (MSA) based on its headquarters location
10
(Loughran and Schultz 2004; Hong, Kubick, and Stein 2004; Pirinsky and Wang 2006). We
discard stocks that cannot be assigned reliably to an MSA by headquarters location. In the
current version we also exclude all ADRs, REITs, and financial firms, although in earlier
versions we found that inclusion or exclusion of financial firms did not materially affect our
results. Finally, we exclude stocks without at least 36 months of return data available from
CRSP within a five-year period, and we also discard stocks associated with cities for which
the local set does not contain at least five firms throughout a relevant period of analysis.
Our classification of stocks to cities with HPI data results in samples of 85 cities and
1,927 headquartered stocks for the July 1989-June 1994 period, 98 cities and 2,563
headquartered stocks for the July 1994-June 1999 period, and 100 cities and 3,058
headquartered stocks for the July 1999-June 2004 period,. Table 1 summarizes the number
of stocks and associated headquarters cities, as well as the distribution of the number of
stocks per city, over each of our three five-year sample periods. Clearly, there are some
heavily firm-populated cities, but most cities have a sufficiently small number of firms that
local investors are likely to be familiar with these stocks from daily newspaper reading and
socializing in the community.
For each city we calculate the monthly return on an equally weighted portfolio of
locally headquartered firms (i=1,….N) as follows:
∑=
=N
i
titHQcity N
RR
1
,, (1)
We categorize these city-specific portfolios based on trends in local residential real
estate prices. We then investigate whether real estate trends condition excess stock returns.
In particular, these data also allow us to see if local price pressures might induce abnormal
11
returns, consistent with the findings of a local value effect in Hong, Kubick, and Stein
(2008).
Also, as per Pirinsky and Wang (2006), for each stock assigned to a specific city we
compute the return on an equally-weighted portfolio of all other stocks for that city,
excluding the firm in question. In other words, each stock is assigned a unique local
portfolio of other stocks located within the same city, and for any month that portfolio’s
return is computed as follows:
∑≠=
− −=
N
jii
titjHQcity N
RR
1
,, 1 (2)
where Ri,t is the return in month t on another stock in the same city as stock j, and there are a
total of N such stocks (including stock j) in the city.
For each sample firm we also identify its industry using the industry classification
scheme of Kenneth French and obtain the corresponding monthly industry return in excess of
the risk-free rate (RIND-RF) from his data library. We also rely on French to obtain the
monthly market return over the risk free rate (RM-RF), high-minus-low book to market
portfolio return (HML), small-minus-big portfolio return (SMB), momentum factor return
(UMD), and the monthly risk-free rate (RF). These data allow us to follow the methods of
Pirinsky and Wang (2006) and Anderson and Beracha (2008) to investigate stock return
comovement among proximately headquartered firms. These studies document that
sensitivity to movements in other local stocks is strong and positive, a result suggestive of
local factors in price formation attributable to local bias.
Home Price Data
12
To identify trends in local real estate prices, we rely on the home price indexes (HPI)
estimated and published by the Office of Federal Housing Enterprise Oversight (OFHEO).
OFHEO systematically collects all sale price data for homes from Freddie Mac and Fannie
Mae. OFHEO applies the weighted repeat sales method developed by Case and Shiller
(1989) to produce these indexes on a quarterly basis for 379 metropolitan statistical areas
(MSAs), including the cities for which we were able to assign firm’s on the basis of their
headquarters location. We subsequently refer to these MSAs as “cities” even though some
MSAs comprise multiple communities, including some MSAs that cross state borders, such
as Kansas City or the Quad Cities. Calhoun (1996) provides a description of the OFHEO’s
procedures in calculating the HPI, and there are several methodological issues surrounding
the construction and use of repeat sales price home price indexes (e.g., Clapham, Englund,
Quigley, and Redfearn 2005). For our purposes, we merely assume that city-specific price
patterns manifest in the HPIs are discernible by local investors or otherwise proxy for
perceptions about home equity values.1
We next classify cities by house price changes as measured by the HPIs. Specifically,
for three five-year periods – July 1989 through June 1994, July 1994 to June 1999, and July
1999 to June 2004 – we sort cities by the cumulative change in the city-specific HPIs.
Specifically, within each five-year period we categorize cities as “hot” (the top 30% by HPI
growth), “medium” (the middle 40% by HPI growth), and “cold” (the bottom 30% by HPI
growth).
Table 2 provides some summary statistics on HPI growth across sample MSAs for
each sample period. Mean and median HPI growth increases over time, as does dispersion
1 Bucks and Pence (2005) compare household survey data to mortgage transaction data and find that homeowners, on average, have accurate perceptions of home values.
13
across MSAs as measured by standard deviation. In particular, the spread between HPI
growth between the hot markets and cold markets increases, with five-year growth rates in
hot markets exceeding that of cold markets by more than 50% during 1999-2004.
Conditioning Stock-Price Movements on Housing Market Trends
We first investigate whether local housing market trends affect performance of local
stocks. For each sample stock j we estimate the following four-factor model of monthly
returns across time:
( ) t,jtjtjtjt,Ft,Mjjt,Ft,j UMDmHMLhSMBsRRRR ε++++−β+α=− (3)
where for each month t
Rj, – RF = return on stock j in excess of the short-term risk-free rate,
RM–RF = return on the value-weighted market return in excess of the short-term risk-free rate,
SMB = the ‘small minus big’ factor, i.e., returns to small firms in excess of returns to large firms,
HML = the ‘high minus low’ factor, i.e., the returns on ‘value’ firms in excess of ‘growth’ firms,
UMD = the ‘momentum’ (‘up minus down’) factor, i.e., returns on previously high returning stocks minus low returning stocks
We compare the mean coefficient estimates from equation (3) across sample stocks
associated with hot, medium, and cold real estate markets in each five-year sample period.
Our primary hypothesis is with respect to excess risk adjusted stock returns as measured by
14
the intercept coefficient for equation (3). Specifically, we hypothesize that αHOT > αMEDIUM
> αCOLD. For completeness, we also report how the non-intercept factor sensitivities differ
across real estate market conditions, allowing us to see if stocks in hot real estate markets
have differential betas or factor sensitivities with respect to the size, value, and momentum
factors.
We next investigate whether local real estate conditions affect return comovement
among local stocks. As per Anderson and Beracha (2008), for each sample stock j we
estimate equations such as the following:
( ) tjtFtMjjtFtj SMBsRRRR +−+=− ,,,, βα
( ) tjtFtjHQcityjtjtj RRUMDmHMLh ,,, ελ +−+++ − (4)
where λj is a measure of firm j’s return comovement with returns on the portfolio of other
stocks in the same headquarters city as stock j (Rcity-j,t as defined in eq. (2)) while controlling
for market, size, book-to-market, and momentum factors.2 We predict greater local
comovement, on average, among stocks in cities which have experienced unusually large
increases in local home prices. Specifically, we hypothesize that on average λHOT
>λMEDIUM>λCOLD.
Empirical Results
Housing Market Conditions and Local Stock Returns
2 Anderson and Beracha (2008) demonstrate that estimates of λj are materially lower when conditioned on a multiple factor model that includes size, value, and momentum effects. Estimates for λj based on a market model or a model that also includes industry portfolio returns, as used in Pirinsky and Wang (2006), are higher.
15
Table 3 documents monthly raw returns on stocks associated sorted by whether they
are headquartered in hot, medium, or cold real estate markets for each of our five-year
sample periods. In the latter two sample periods higher average return is associated with
stocks headquartered in the hot areas, while stocks headquartered in the cold areas yield
lower average returns. Specifically, monthly returns on stocks from hot cities exceed those
from cold cities by twelve basis points per month, or a cumulative amount of about 1.4% per
year for both these latter periods. In contrast, housing market conditions do not appear to
condition raw returns on local stocks for the first 1989-1994 sample period, and the
difference in returns between hot market firms and cold market firms is a statistically
insignificant six basis points per month.
Results in Table 3 are for raw returns. One might expect that risk factors differ
systematically for firms across locations, for example due to industry clustering of growth
stocks in areas such as California (Anderson and Beracha 2008). We next attempt to adjust
for equity risk factors. Specifically, Table 4 shows the mean coefficient estimates from
security-specific factor regressions as per equation (3). In this table, we concentrate mainly
on the mean α coefficient estimates, which capture the excess return for each sample stock
after controlling for the Fama-French three risk factors and the Carhart momentum factor.
Nevertheless, the coefficient estimates for the equity risk factors suggest some systematic
differences across cities characterized by residential real estate trends.3
3 Specifically, in the 1989-1994 period sensitivity to the market risk factor and the small-minus-big (SMB) factor both appear to be inversely related to real estate price trends. In the 1994-1999 period stocks of firms from hot real estate markets appear to have negative exposure to the high-minus-low (HML) book-to-market factor. Finally, in the 1999-2004 period hot market firms appear to have more positive exposure to market risk, more positive exposure to the size factor, and more negative exposure to the book-to-market factor. Curiously, in no sample period does headquarters city home price appreciation appear to condition exposure to the up-minus-down (UMD) momentum factor.
16
Consistent with average raw returns reported in Table 3, Table 4 reports no material
difference in risk adjusted returns for the 1989-1994 sample period. In contrast, for the latter
two periods we find that risk adjusted monthly returns -- as measured by α -- increase from
cold to medium to hot residential real estate markets of headquarters cities. In particular, for
1994-1999 difference between mean HOTα and COLDα is nearly 24 basis points per month, or
nearly 2.8% annually. For 1999-2004 the difference between mean HOTα and COLDα is
nearly 61 basis points per month, or about 7.3% annually. Again, the latter period is the one
associated with the highest average rates of home price appreciation and the greatest degree
of cross-sectional dispersion across sample cities.
In Table 5 we regress estimates of excess return measured by α , as per equation (3),
on a series of city-specific economic indicators and change in home prices. We run four
different regression specifications for each of the three predetermined five-year time periods.
In the first specification we control for population, number of firms, investment and non
investment per capita income, and change in home prices. The second specification controls
for the same factors except that the population and number of firms factors are replaced by
their ratio, which serves as a proxy for local firm density. The third and fourth specifications
are similar to the first and second specifications, respectively, but also control for changes in
population and per capita income through the subsample period. By regressing α on local
economic factors and change in local housing prices we are able to isolate the effect of home
prices on excess returns from other economic factors.
The results of Table 5 show that while each of the economic factors is statistically
significant at least during one of the three sample periods, none of the factors consistently
demonstrate statistical significance during the three time periods. The %Δ HPI coefficient,
17
while mostly not statistically significant, is gradually increasing in magnitude through time.
The higher value of the %Δ HPI coefficient during the periods associated with rapid home
prices supports our results from Table 4 and provides additional evidence that excess risk
adjusted return is positively related to change in home prices even after controlling for local
economic factors.
Housing Market Conditions and Return Comovement among Local Stocks
Table 6 presents the estimated results for equation (4), which tests for local
comovement while controlling for the Fama-French three risk factors and Carhart momentum
factor. Specifically, for each firm-specific return series we extend the factor model of returns
as per Pirinsky and Wang (2006) to include the monthly return on each firm’s unique
headquarters city portfolio. The coefficient on this headquarters city factor measures the
degree of comovement in returns among same-city firms.
Overall, the results in Table 6 provide evidence that supports our second hypothesis
for the latter two sample periods. The λ coefficients for the July 1999 to June 2004 and the
July 1994 to June 1999 periods satisfy the HOTλ > MEDIUMλ > COLDλ condition, suggesting that
local comovement is indeed stronger in areas that experience greater residential real estate
appreciation. The difference between the λ coefficients is also statistically significant. The t-
statistics associated with the test COLDHOT λλ = are 6.59 and 3.60 for the later and the middle
period respectively. In contrast, for the period of July 1989 to June 1994 when home price
appreciation is lower on average and less widely dispersed across cities we find no
significant difference between the λ coefficients for the different real estate market
conditions. It is also important to note from Table 6 that inclusion of the headquarters city
18
return factor results in attenuation of the average alpha coefficients in all sample periods.
This suggests that returns are indeed higher among firms located in hot real estate markets
(as shown in Table 4), but not after controlling for the comovement in returns experienced by
other firms in the same city.
Table 7 provides additional evidence on the determinants of return comovement
among same-city firms by regressing estimates of λ (from equation (4)) on a series of city-
specific economic indicators and change in home prices. We run four different regression
specifications for each predetermined five-year time period, which are identical to the
specifications presented in Table 5. Regressing λ on local economic factors and change in
housing prices allows us to observe the effect of home prices on local stock-price
comovement while controlling for other economic factors that may contribute to the return
comovement among same-city firms.
The results of Table 7 show that population is negatively related to λ while the
number of local firm has positive relation with λ. These two coefficients are statistically
significant in all the three five-year sample periods and suggest that return comovement
among same-city stocks is stronger in cities with more firms and weaker in cities with larger
population. The firm density coefficient, defined as the ratio between population and number
of firms, is also negative and statistically significant throughout the sample period. On the
other hand, the coefficients of investment and noninvestment per capita income as well as
changes in population and per capita income do not demonstrate consistent relation with λ
that has statistical significance throughout the sample period. %Δ HPI, which is the main
coefficient of interest in this table, is mainly positive, but not statistically significant during
the July 1989-June 1994 period. However, during the July 1994-June 1999 period the %Δ
19
HPI coefficient is positive and statistically significant. The %Δ HPI coefficient demonstrates even
higher magnitude and statistical significance during the July 1999-June 2004 period. The
gradual increase in the value of the %Δ HPI coefficient during an increasingly positive
housing market supports our results from Table 6. It also provides additional evidence that
return comovement among same-city firms is stronger during periods of rapid housing
appreciation even after controlling for local economic factors.
Conclusion
We investigate how conditions in local residential real estate markets affect the
pricing of local stocks for U.S. metropolitan areas over the fifteen-year period from July
1989 to June 2004. We classify stocks by the city in which the firm is headquartered and
then partition cities according to trends in residential real estate prices as measured by the
repeat-sale home price indexes (HPIs) produced for metropolitan statistical areas by the
Office of Federal Housing Enterprise Oversight (OFHEO). We then investigate whether
extreme changes in home prices in these headquarters cities affect stock returns in a manner
consistent with local bias in stock trading and investment. In particular, we hypothesize that
large increases in household wealth due to home price appreciation may stimulate demand
for local stocks and lead to higher than normal returns. We also hypothesize that rising
housing wealth and habitat-effect trading will be associated with a greater degree of
comovement among returns of local stocks. We expect such effects to be most prevalent
during the 1999-2004 period characterized by greater dispersion in home price appreciation
across U.S. cities.
20
We find results consistent with our hypotheses, especially during the July 1999 to
June 2004 period that has been characterized both in the popular press and by scholars as the
start of the most recent housing bubble. We find that stocks located in ‘hot’ residential real
estate markets experience excess returns relative to stocks from ‘cold’ markets. We also find
that stocks of firms located in hot real estate markets experience stronger return comovement
with other same-city stocks compared to stocks located in cold real estate markets. Our
findings suggest that changes in investors’ real estate wealth affect trading activity and
pricing of local stocks.
21
References
Ambrose, B., Lee, D.W., & Peek, J. (2007). Comovement after joining an index: Spillovers of nonfundamental effects. Real Estate Economics, 35(1), 57-90.
Anderson, C.W. & Beracha, E. (2008). Robustness of the headquarters-city effect in stock returns. Journal of Financial Research, 31(3), 271-300.
Anderson, C.W. & Beracha, E. (2009). Home-price sensitivity to capital market factors: Analysis of zip code level data. Forthcoming, Journal of Real Estate Research.
Barberis, N., Shleifer, A., & Wurgler, J. (2005). Comovement. Journal of Financial Economics, 75(2), 283-317.
Barker, D. & Loughran, T. (2007). The geography of S&P 500 stock returns. Journal of Behavioral Finance, 8(4), 177-190.
Benjamin, J.D. & Chinloy, P. (2008). Home equity, household savings, and consumption. Journal of Real Estate Research, 37(1), 21-32.
Benjamin, J.D., Chinloy, P., & Jud, G.D. (2004a). Why do households concentrate their wealth in housing? Journal of Real Estate Research, 26(4), 329-343.
Benjamin, J.D., Chinloy, P., & Jud, G.D. (2004b). Real estate versus financial wealth in consumption. Journal of Real Estate Finance and Economics, 29(3),341-354.
Beracha, E. & Hirschey, M. (2009). When will housing recover? Financial Analysts Journal, 65(2), 36-47.
Bostic, R., Gabriel, S., & Painter, G. (2005). Housing wealth, financial wealth, and consumption: New evidence from micro data. Working paper, Lusk Center for Real Estate, University of Southern California.
Bucks, B. & Pence, K. (2005). Do homeowners know their house values and mortgage terms? Working paper, Federal Reserve Board of Governors.
Calhoun, C.A. (1996). OFHEO house price indexes: HPI technical description. Office of Federal Housing Enterprise Oversight.
Campbell, J. & Cocco, J. (2005). How do house prices affect consumption? Evidence from micro data. Working paper.
Cannon, S., Miller, N.G., & Pandher, G. (2006). Risk and return in the U.S. housing market: A cross-sectional asset-pricing approach. Real Estate Economics, 34(4), 519-552.
Carroll, C.D., Otsuka, M., & Slacalek, J. (2006). How large is the housing wealth effect? A new approach. Working paper, John Hopkins University, Baltimore, MD.
Case, K.E. & Shiller, R.F. (1989). The efficiency of the market for single family homes. American Economic Review, 79(1), 125-137.
22
Case, K.E. & Shiller, R.F. (2003). Is there a bubble in the housing market? Brookings Papers on Economic Activity, 2, 299-342.
Case, K.E., Quigley, J., & Shiller, R.F. (2003). Home buyers, housing, and the macroeconomy. Working paper presented at the 2003 Reserve Bank of Australia’s Conference on Asset Prices and Monetary Policy.
Case, K.E., Quigley, J., & Shiller, R.F. (2005). Comparing wealth effects: The stock market versus the housing market. Working paper No. W01-004, Berkley Program on Housing and Urban Policy, Institute of Business and Economic Research, University of California-Berkley.
Chan, K., Covrig, V., & Ng, L. (2005). What determines the domestic bias and foreign bias? Evidence from mutual fund equity allocations worldwide. Journal of Finance, 60(3), 1495-1534.
Clapham, E., Englund, P., Quigley, J., & Redfearn, C. (2005). Revisiting the past and settling the score: Index revision for house price derivatives. Working paper No. W04-005, Berkley Program on Housing and Urban Policy, Institute of Business and Economic Research, University of California-Berkley.
Cocco, J. F. (2004). Portfolio choice in the presence of housing. Review of Financial Studies, 18(2), 535-567.
Coval, J. D. & Moskowitz, T. J. (1999). Home bias at home: Local equity preference in domestic portfolios. Journal of Finance 54(6), 2045-2073.
Dahlquist, M., Pinkowitz, L., Stulz, R.M., & Williamson, R. (2003). Corporate governance, Investor protection, and the home bias. Journal of Financial and Quantitative Analysis, 38(1), 87-110.
Flavin, M. & Yamashita, T. (2002). Owner occupied housing and the composition of the household portfolio. American Economic Review, 92(1), 345-362.
French, K.R. & Poterba, J.M. (1991). Investor diversification and international equity markets. American Economic Review, 81(2), 222-226.
Greenspan, A. (2005). The economic outlook: Testimony before the Joint Economic Committee, U.S. Congress, June 9, 2005: http://www.federalreserve.gov/boarddocs/testimony/2005/.
Greenspan, A. & Kennedy, J. (2005). Estimates of home mortgage originations, repayments, and debt on one-to-four-family residences. Working paper, Federal Reserve Board, Washington, DC.
Himmelberg, C., Mayer, C., &.Sinai, T. (2005). Assessing high house prices: Bubbles, fundamentals, and misperceptions. Journal of Economic Perspectives, 19(4), 67-94.
Hong, H., Kubick, J.D. & Stein, J.C. (2004). Social interaction and stock market participation. Journal of Finance, 59(1), 137-163.
Hong, H., Kubick, J.D. & Stein, J.C. (2005). Thy neighbor’s portfolio: Word of mouth effects in the holdings and trades of money managers. Journal of Finance, 60(6), 2801-2824.
Hong, H., Kubick, J.D. & Stein, J.C. (2008). The only game in town: Stock-price consequences of local bias. Journal of Financial Economics, 90(1), 20-37.
Huberman, G. (2001). Familiarity breeds investment. Review of Financial Studies 14(3), 659-680.
23
Kang, J.K. & Stulz, R.M. (1997). Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan. Journal of Financial Economics, 46(1), 3-28.
Keloharju, M. & Grinblatt, M. (2001). How distance, language, and culture influence stockholdings and trades. Journal of Finance, 56(3), 1053-1073.
Kishor, N.K. (2007). Does consumption respond more to housing wealth than to financial market wealth? If so, why? Journal of Real Estate Finance and Economics, 35(4), 427-428.
Kullman, C. (2003). Real estate and its role in asset pricing. Working paper, University of British Columbia.
Kullman, C. & Siegel, S. (2007). Real estate and its role in household portfolio choice. Working paper, University of Washington.
Kumar, A. & Lee, C.M.C. (2006). Retail investor sentiment and return comovements. Journal of Finance, 61(5), 2451-2486.
Le Blanc, D. & Lagarenne, C. (2004). Owner-occupied housing and the composition of the household portfolio: The case of France. Journal of Real Estate Finance and Economics, 29(3), 259-275.
Loughran, T. & Schultz, P. (2004). Weather, stock returns, and the impact of localized trading behavior. Journal of Financial and Quantitative Analysis, 39(2), 343-364.
Loughran, T. & Schultz, P. (2005). Liquidity: urban versus rural firms. Journal of Financial Economics, 78(2), 341-374.
Miller, N., Peng, L., & Sklarz, M. (2009). House prices and economic growth. Forthcoming, Journal of Real Estate Finance and Economics.
Ozsolyev, H.N. (2005). Asset pricing implications of social networks. Working paper, Oxford University.
Palia, D., Qi, Y., & Wu, Y. (2009). The importance of background risks. Working paper, available at SSRN: http://ssrn.com/abstract=986489.
Piazzesi, M., Schneider, M., & Tuzel, S. (2007). Housing, consumption, and asset pricing. Journal of Financial Economics, 83(3), 531-569.
Pirinsky, C., & Wang, Q. (2006). Does corporate headquarters location matter for stock returns? Journal of Finance, 61(4), 1991-2015.
Quan, D. & Titman, S. (1999). Do real estate prices and stock prices move together? An international analysis. Real Estate Economics, 27(2), 183-207.
Slacalek, J. (2006). What drives personal consumption? The role of housing and financial wealth. Working paper, German Institute for Economic Research.
Yao, R. & Zhang, H. (2005). Optimal consumption and portfolio choices with risky housing and borrowing constraints. Review of Financial Studies, 18(1), 197-239.
Zhu, N. (2003). The local bias of individual investors. Working paper, Yale School of Management.
24
Table 1 Attributes of headquarters-city stock portfolios
Headquartered stocks per city Time period Stocks HQ cities Mean 25% Median 75% July 1989 to June 1994 1,927 85 22.7 9 15 34 July 1994 to June 1999 2,563 98 26.2 8 16 35 July 1999 to June 2004 3,058 100 30.6 8 16 41 This table shows the number of common stocks for CRSP/COMPUSTAT listed firms that are identified with their respective headquarters cities (metropolitan statistical areas) for each of three five-year time periods. We exclude financial firms, REITs, and ADRs, as well as stocks with fewer than 36 valid monthly returns within each five-year period. We also exclude cities with fewer than five headquartered firms.
Table 2 Residential real estate price trends across U.S. cities
Five-year change in home price index (HPI)
Time period HQ cities Mean Median Stdev. Hot markets
Cold markets
July 1989 to June 1994 85 15.6% 15.5% 12.1% 30.9% 2.3% July 1994 to June 1999 98 19.9% 19.6% 16.0% 37.0% 3.4% July 1999 to June 2004 100 40.3% 22.6% 24.9% 70.5% 19.2% This table shows the means, medians, and standard deviations of five-year cumulative changes in home price indices (HPI) for U.S. metro areas as published by OFHEO, as well as the mean changes for markets the hottest markets and the coldest markets. Hot (cold) market MSAs are those in the top (bottom) 30% of sample MSAs sorted by change in HPI within each five-year period. The summary statistics in this table represents only the MSAs for which we identify at least five companies with 36 or more months of data within each five-year time period.
25
Table 3 Raw monthly stock returns sorted by headquarter city HPI growth
Time period Cities (Stocks)
Mean (stdev.) % monthly stock returns
July 1989 to June 1994 Hot markets Medium markets Cold markets Hot minus cold (t-test)
25 (448) 35 (631) 25 (848)
1.23 (4.88) 1.29 (5.25) 1.17 (5.21)
0.06
(1.57)
July 1994 to June 1999 Hot markets Medium markets Cold markets Hot minus cold (t-test)
29 (942)
40 (1102) 29 (519)
1.70 (5.70) 1.45 (4.83) 1.58 (4.70)
0.12
(2.84)
July 1999 to June 2004 Hot markets Medium markets Cold markets Hot minus cold (t-test)
30 (1613) 40 (1088)
30 (357)
1.83 (8.34) 1.75 (6.98) 1.71 (6.26)
0.12
(2.12)
This table reports mean monthly stock returns for stocks whose headquarters can be identified from COMPUSTAT, for which there are at least five same-city stocks, and when there are at least 36 months of return data within a five-year period. Headquarters cities are categorized as hot, medium, or cold portfolios based on the cumulative change in the home price index (HPI) for each city for each five-year time period. Hot markets are among the top 30% of markets according to cumulative change in the city-specific housing price index (HPI), cold markets are the lowest 30%, and medium markets are the middle 40%.
26
Table 4 Mean coefficient estimates from factor models of monthly returns α
(t-stat) β
(t-stat) s
(t-stat) h
(t-stat) m
(t-stat) July 1989-June 1994 Hot RE Markets 0.472 0.855 0.887 0.325 -0.106 (5.04) (24.97) (14.77) (5.96) (-2.71) Medium RE Markets 0.540 0.889 0.949 0.323 -0.019 (7.23) (28.01) (16.62) (6.742) (-0.63) Cold RE Markets 0.501 0.956 1.112 0.247 -0.090 (6.81) (33.41) (22.49) (5.44) (-3.22) Hot minus Cold
0.029
(-0.23)
-0.101 (-2.16)
-0.225 (-2.78)
0.078 (1.04)
-0.016 (-0.35)
July 1994-June 1999 Hot RE Markets 0.700 0.773 0.796 -0.382 -0.255 (8.27) (28.87) (24.00) (-7.51) (-8.11) Medium RE Markets 0.565 0.849 0.673 0.085 -0.240 (6.43) (33.13) (22.70) (2.00) (-7.84) Cold RE Markets 0.461 0.783 0.718 0.100 -0.204 (5.75) (22.72) (17.72) (1.66) (-4.89) Hot minus Cold
0.239 (2.02)
-0.010 (-0.23)
0.078 (1.45)
-0.482 (-5.89)
-0.051 (-0.97)
July 1999-June 2004 Hot RE Markets 1.428 1.101 0.882 -0.090 -0.274 (22.16) (43.84) (33.63) (-3.04) (-14.86) Medium RE Markets 1.046 0.913 0.656 0.340 -0.220 (15.21) (34.77) (25.27) (11.69) (-12.00) Cold RE Markets 0.819 0.810 0.607 0.492 -0.217 (7.41) (21.09) (15.94) (9.51) (-7.23) Hot minus Cold
0.609 (4.16)
0.291 (5.16)
0.275 (4.69)
-0.582 (-8.65)
-0.570 (-1.38)
For stocks whose headquarters can be identified from COMPUSTAT, for which there are at least five same-city stocks, and when there are at least 36 months of return data for specified three five-year periods from July 1989 to June 2004, we estimate coefficients for the following stock return factor model (equation (3)):
( ) t,jtjtjtjt,Ft,Mjjt,Ft,j UMDmHMLhSMBsRRRR ε++++−β+α=− , where Rj is the monthly return on stock j, RF is the risk-free rate, RM is the value-weighted market return, SMB is the ‘small minus big’ size factor, HML is the ‘high minus low’ value factor, and UMD is the ‘up minus down’ momentum factor. Monthly stock returns are from CRSP. The remaining variables are obtained from Kenneth French’s data library available on the internet. Means of coefficient estimates and their respective t-statistics are reported in the table below according to whether a firm’s headquarters city is classified as a hot, medium, or cold residential real estate market for the specified five-year periods. Hot markets are among the top 30% of markets according to cumulative change in the city-specific housing price index (HPI) over the five-year period, cold markets are the lowest 30%, and medium markets are the middle 40%.
27
Table 5 Regressions of excess returns (α) on local economic factors including housing market conditions 1989-1994 1994-1999 1999-2004 (1a) (2a) (3a) (4a) (1b) (2b) (3b) (4b) (1c) (2c) (3c) (4c) Intercept 7.19
(3.00) 7.42
(3.17) 4.67
(1.51) 4.79
(1.56) 2.28
(0.90) 1.98
(0.78) 1.60
(0.56) 1.30
(0.45) 6.03
(2.37) 6.40
(2.53) 1.95
(0.64) 2.17
(0.71) Log(number of firms) 0.25
(2.21) 0.24
(2.07) 0.20
(1.61) 0.22
(1.78) 0.35
(3.00) 0.38
(3.18)
Log(population) -0.28 (-2.81)
-0.27 (-2.66)
-0.15 (-1.26)
-0.16 (-1.30)
-0.43 (-3.81)
-0.46 (-4.00)
Log(pop/no. of firms) -0.28 (-2.78)
-0.27 (-2.64)
-0.17 (-1.42)
-0.18 (-1.50)
-0.40 (-3.58)
-0.43 (-3.81)
Log(Noninvestment PCI) -1.45 (-2.00)
-1.60 (-2.50)
-1.35 (-1.71)
-1.52 (-2.22)
-0.48 (-0.78)
-0.24 (-0.44)
-1.27 (-1.72)
-0.89 (-1.36)
-0.24 (-0.39)
-0.65 (-1.20)
-0.32 (-0.52)
-0.71 (-1.32)
Log(Investment PCI) 0.53 (1.42)
0.57 (1.56)
0.61 (1.46)
0.65 (1.65)
0.65 (1.93)
0.61 (1.83)
0.92 (2.51)
0.84 (2.34)
0.35 (1.00)
0.54 (1.68)
0.65 (1.74)
0.85 (2.46)
%Δ Population 2.07 (1.56)
2.08 (1.57)
-0.09 (-0.07)
-0.12 (-0.10)
3.28 (2.43)
3.26 (2.42)
%Δ PCI -0.17 (-0.12)
-0.06 (-0.04)
2.37 (1.92)
2.18 (1.79)
0.61 (0.74)
0.78 (0.97)
%Δ HPI -0.10 (-0.28)
-0.07 (-0.20)
-0.27 (-0.67)
-0.26 (-0.64)
0.45 (0.90)
0.39 (0.78)
0.22 (0.40)
0.15 (0.28)
0.52 (1.96)
0.35 (1.48)
0.47 (1.69)
0.30 (1.19)
N 1927 1927 1927 1927 2561 2561 2561 2561 3057 3057 3057 3057 R-square 0.62% 0.39% 0.75% 0.74% 0.69% 0.66% 0.85% 0.80% 1.85% 1.78% 2.07% 2.00% For stocks whose headquarters can be identified from COMPUSTAT, for which there are at least five same-city stocks, and when there are at least 36 months of return data for specified three five-year periods from July 1989 to June 2004, we estimate coefficients for the following stock return factor model (equation (3)):
( ) t,jtjtjtjt,Ft,Mjjt,Ft,j UMDmHMLhSMBsRRRR ε++++−β+α=− , where Rj is the monthly return on stock j, RF is the risk-free rate, RM is the value-weighted market return, SMB is the ‘small minus big’ size factor, HML is the ‘high minus low’ value factor, and UMD is the ‘up minus down’ momentum factor. Individual stock returns are from CRSP. The remaining variables were obtained from Kenneth French’s data library available on the internet. We regress the stock-level estimates of excess return as measured by the individual jα̂ coefficients on a number of city-specific economic factors including housing-price data.
28
Table 6 Mean coefficient estimates for models of returns, including HQ- city factor α
(t-stat) β
(t-stat) s
(t-stat) h
(t-stat) m
(t-stat) λ
(t-stat) July 1989-June 1994 Hot RE Markets 0.378 0.696 0.715 0.247 -0.090 0.305 (3.90) (13.70) (10.98) (4.32) (-2.38) (7.27) Medium RE Markets 0.441 0.620 0.655 0.223 -0.011 0.323 (5.59) (13.20) (10.12) (4.53) (-0.37) (8.50) Cold RE Markets 0.368 0.636 0.750 0.147 -0.052 0.339 (4.93) (12.79) (12.11) (3.13) (-1.84) (7.68) Hot minus Cold
0.010 (0.08)
0.057 (0.78)
-0.035 (-0.36)
0.100 (1.31)
-0.038 (-0.84)
-0.034 (-0.49)
July 1994-June 1999 Hot RE Markets 0.403 0.371 0.415 -0.141 -0.144 0.494 (4.78) (8.14) (9.87) (-3.09) (4.50) (11.88) Medium RE Markets 0.440 0.586 0.483 0.047 -0.196 0.300 (5.98) (13.44) (12.85) (1.07) (-6.03) (7.27) Cold RE Markets 0.457 0.608 0.574 0.082 -0.183 0.242 (4.34) (11.18) (10.90) (1.35) (-4.25) (4.29) Hot minus Cold
-0.054 (-0.39)
-0.237 (-3.22)
-0.159 (-2.31)
-0.223 (-2.92)
0.039 (0.73)
0.252 (3.60)
July 1999-June 2004 Hot RE Markets 0.483 0.414 0.321 -0.088 -0.061 0.739 (6.82) (10.00) (9.47) (-2.84) (-3.13) (18.42) Medium RE Markets 0.546 0.520 0.374 0.168 -0.120 0.492 (7.73) (13.22) (12.22) (4.91) (-6.09) (13.30) Cold RE Markets 0.679 0.694 0.501 0.400 -0.182 0.164 (6.05) (14.63) (12.12) (7.28) (-6.04) (4.25) Hot minus Cold
-0.196 (-1.23)
-0.280 (-3.10)
-0.180 (-2.47)
-0.488 (-6.88)
0.121 (2.75)
0.575 (6.59)
For stocks whose headquarters can be identified from COMPUSTAT, for which there are at least five same-city stocks, and when there are at least 36 months of return data for specified three five-year periods from July 1989 to June 2004, we estimate coefficients for the following stock return factor model (equation (4)):
( ) ( ) tjtFtjHQcityjtjtjtjtFtMjjtFtj RRUMDmHMLhSMBsRRRR ,,,,,,, ελβα +−++++−+=− − , where Rj is the monthly return on stock j, RF is the risk-free rate, RM is the value-weighted market return, SMB is the ‘small minus big’ size factor, HML is the ‘high minus low’ value factor, UMD is the ‘up minus down’ momentum factor, and Rcity-j is the return on the equally-weighted portfolio of other stocks located in stock j’s headquarters city. Individual stock returns are from CRSP. The remaining variables were obtained from Kenneth French’s data library available on the internet. Means of coefficient estimates and their respective t-statistics are reported in the table below according to whether the headquarters city is classified as a hot, medium, or cold residential real estate market for the specified five-year periods. Hot markets are among the top 30% of markets according to cumulative change in the city-specific housing price index (HPI) over the five-year period, cold markets are the lowest 30%, and medium markets are the middle 40%.
29
Table 7 Regressions of local comovement (λ) on local economic factors including housing market conditions 1989-1994 1994-1999 1999-2004 (1a) (2a) (3a) (4a) (1b) (2b) (3b) (4b) (1c) (2c) (3c) (4c) Intercept 2.57
(1.93) 1.28
(0.98) 2.88
(1.68) 2.24
(1.31) 2.72
(1.91) 2.03
(1.43) 2.03
(1.26) 1.48
(0.93) -0.33
(-0.22) -0.88
(-0.59) 1.28
(0.71) 0.97
(0.54) Log(number of firms) 0.30
(4.73) 0.27
(4.21) 0.27
(3.89) 0.28
(3.99) 0.23
(3.37) 0.21
(3.00)
Log(population) -0.14 (-2.49)
-0.13 (-2.27)
-0.16 (-2.37)
-0.16 (-2.39)
-0.10 (-1.56)
-0.09 (-1.31)
Log(pop/no. of firms) -0.16 (-2.90)
-0.15 (-2.59)
-0.20 (-2.96)
-0.20 (-2.99)
-0.16 (-2.43)
-0.13 (-2.02)
Log(Noninvestment PCI) -0.32 (-0.79)
0.53 (1.49)
0.13 (0.44)
0.99 (2.60)
-0.32 (-0.93)
0.23 (0.75)
-0.56 (-1.37)
0.12 (0.32)
0.64 (1.78)
1.24 (3.90)
0.68 (1.87)
1.25 (3.93)
Log(Investment PCI) -0.30 (-1.46)
-0.51 (-2.50)
-0.50 (-2.16)
-0.74 (-3.36)
-0.17 (-0.92)
-0.26 (-1.41)
-0.06 (-0.31)
-0.20 (-0.98)
-0.56 (-2.74)
-0.85 (-4.46)
-0.69 (-3.13)
-0.98 (-4.80)
%Δ Population 0.79 (1.08)
0.74 (1.01)
0.43 (0.61)
0.37 (0.53)
-1.10 (-1.39)
-1.07 (-1.35)
%Δ PCI -1.97 (-2.46)
-2.57 (-3.26)
0.70 (1.02)
0.37 (0.55)
-0.46 (-0.97)
-0.72 (-1.52)
%Δ HPI 0.02 (0.08)
-0.15 (-0.81)
0.16 (0.71)
0.08 (0.37)
0.83 (2.90)
0.67 (2.39)
0.70 (2.31)
0.58 (1.94)
0.55 (3.57)
0.79 (5.65)
0.60 (3.66)
0.85 (5.84)
N 1927 1927 1927 1927 2561 2561 2561 2561 3057 3057 3057 3057 R-square 2.27% 1.19% 2.66% 1.83% 1.90% 1.41% 1.97% 1.44% 3.46% 3.02% 3.55% 3.15% For stocks whose headquarters can be identified from COMPUSTAT, for which there are at least five same-city stocks, and when there are at least 36 months of return data for specified three five-year periods from July 1989 to June 2004, we estimate coefficients for the following stock return factor model (equation (4)):
( ) ( ) t,jt,Ft,jcityjtjtjtjt,Ft,Mjjt,Ft,j RRUMDmHMLhSMBsRRRR ε+−λ++++−β+α=− − , where Rj is the monthly return on stock j, RF is the risk-free rate, RM is the value-weighted market return, SMB is the ‘small minus big’ size factor, HML is the ‘high minus low’ value factor, UMD is the ‘up minus down’ momentum factor, and Rcity-j is the return on the equally-weighted portfolio of other stocks located in stock j’s headquarters city. Individual stock returns are from CRSP. The remaining variables were obtained from Kenneth French’s data library available on the internet. We regress the stock-level estimates of comovement with same-city stocks as measured by the individual jλ coefficient estimates on a number of city-specific economic factors including housing-price data.