Estimating the Recession Risk Exposure of Stocks
Jingling Guan∗†
November 2010
Job Market Paper
Abstract
In this paper, I develop a new method to estimate the recession risk exposure of
stocks using nancial statement information. I show that rm nancial ratios such
as protability and leverage are useful measures of an individual stock's recession risk
exposure, as given by recession premium: the stock's return dierence between NBER
recessions and expansions. I construct a composite recession score based on common
nancial ratios and nd that rms with scores in the safest decile earn 1.31% per month
more during recessions than rms in the most risky decile. Sorting on recession score
generates portfolios with larger dispersion in recession premium than sorting based on
a host of standard systematic risk measures and rm characteristics. These recession
score portfolios highlight exposure to recession risk and make good candidates to test
linear asset pricing models. I nd that while both the Fama-French three-factor model
and the consumption-based CAPM can explain the cross section of returns on these
recession score portfolios, the CAPM cannot. This method provides a simple way to
capture stocks' exposure to recession risk in real time.
I am greatly indebted to my committee members Ravi Jagannathan (co-chair), Jonathan Parker (cochair), Jules van Binsbergen, Larry Christiano, and Paola Sapienza. I am also grateful for helpful suggestions
and comments from Sneehal Benerjee, Vineet Bhagwat, Kevin Crotty, Andrea Eisfeldt, Kathleen Hagerty,
In Gu Khang, Jiro Kondo, Arvind Krishnamurthy, David Matsa, Brian Melzer, Dermot Murphy, Dimitris
Papanikolaou, Sergio Rebelo, Costis Skiadas, Linda Vincent, Annette Vissing-Jørgensen, Beverly Walther,
and Mary Zaki. All errors are mine.
†
Department of Finance,
Kellogg School of Management,
Northwestern University.
Email: [email protected], web site: http://www.kellogg.northwestern.edu/faculty/guan/.
∗
1
This paper investigates the relationship between macroeconomic risk and the returns on
dierent stocks. To capture the bad macroeconomic states, I focus on recessions, since they
are times when aggregate conditions are severe and deteriorating. I measure the recession
risk exposure of a stock by its recession premium, the average return during recessions minus
the average return during expansions. Stocks that perform better during recessions, i.e., with
higher recession premiums, are safer and more valuable to investors since they provide better
insurance for periods of declining output and rising unemployment.
To estimate the recession risk exposure of stocks in real time, I characterize stocks by
their fundamental attributes, since it is dicult to estimate a single stock's recession risk
exposure from its historical recession premium. This is because recessions occur infrequently,
and few stocks live through several recessions. Even for stocks that have lived through
recessions, their recession risk exposure might change due to investment decisions or mergers
and acquisitions. To make the problem tractable, I construct a composite recession score
for each stock by using its nancial ratios to capture four risk relevant attributes: nancial
health, cyclicality, size, and availability and sustainability of cash ows. My conjecture is
that stocks with similar nancial ratios (e.g., protability, leverage, and accounting liquidity)
have similar exposure to recession risk. This is because nancial ratios have been found
to contain information of a rm's risk, such as probability of bankruptcy (Beaver 1966,
Altman 1968), CAPM beta (Beaver, Kettler, and Scholes 1970), cash ow and discount
factor betas (Campbell, Polk, and Vuolteenaho 2010), and equity risk (Morningstar 2004).
The recession score is constructed in a simple way and in real time, i.e., using nancial
statement information that is available at the time of portfolio formation.
I have three main ndings. First, the recession score provides a better estimate of the
recession risk exposure of stocks as compared to other rm characteristics and standard
systematic risk measures. Stocks with recession scores in the safest decile earn 1.31% per
month more than stocks with scores in the most risky decile during the 83 recession months
covering 7 NBER (National Bureau of Economic Research) recessions. In contrast, sorting
2
on historical factor and consumption betas, size, book-to-market ratio, past returns, and
industry product durability, does not generate portfolios with such a large dispersion in recession risk. Compared to the above-mentioned alternatives, the t -statistic for the dispersion
in recession premium is most signicant for stocks sorted on recession score constructed with
nancial ratios.
Second, I nd that rms with stronger cash ows and in better nancial health are in a
better position to withstand potential economic downturns. The strength of a rm's cash
ows is captured by rm protability (gross prots to total assets) and receivable turnover
(accounts receivable to sales). Firms with protability in the highest decile and receivable
turnover in the lowest decile earn 1.09% and 1.31% per month more during recessions respectively, than rms with lowest protability and highest receivable turnover. A rm's nancial
health is measured by rm leverage (debt to book equity) and accounting liquidity (current
assets over current liabilities). Firms in better nancial health have a larger buer against
bankruptcy and do better during recessions: rms with leverage in the lowest decile within
their industries and with accounting liquidity in the highest decile earn 0.67% and 0.88%
per month more during recessions respectively than rms in the highest decile of leverage
and lowest decile of liquidity.
Third, I nd that the CAPM is not very successful at capturing macroeconomic risk at
business cycle frequencies. The recession score portfolios have very small dispersion in their
CAPM betas, but considerate amount of dispersion in their SMB, HML, and consumption
betas: the CAPM beta of the safest decile is only 13% lower than the beta of the most
risky decile sorted on recession score. In contrast, the SMB, HML, and consumption betas
of the safest decile are 57%, 97%, and 48% lower than the betas of the most risky decile
respectively. Both the Fama-French three-factor model and the consumption-based CAPM
(CCAPM) can explain the cross section of returns on these recession score portfolios, but
the CAPM cannot.
This paper contributes to the literature by constructing portfolios that highlight recession
3
risk exposure of stocks to evaluate the CAPM, CCAPM, and the Fama-French three-factor
model. Macroeconomic risk is the economy wide pervasive systematic risk in most macroasset pricing models. The state of the economy that is relevant for measuring macroeconomic
risk is captured by the returns on the aggregate wealth portfolio in the standard CAPM
(Sharpe 1964 and Lintner 1965) and the growth in aggregate per capita consumption in the
standard consumption-based asset pricing models (Breeden 1979 and Lucas 1978), and a
combination of the two in models where the representative investor prefers earlier resolution
of uncertainty (Epstein and Zin 1989). As a comparison, I follow a model free approach
and capture the state of the economy by NBER expansions and recessions. Recession risk
represents macroeconomic risk at business cycle frequencies and is a major component of
macroeconomic risk.
To evaluate macro-based asset pricing models, it is ideal to use portfolios that highlight
macroeconomic risk, and the recession score portfolios make good candidates. It is common
to test asset pricing models using the cross section of returns on various characteristicsorted portfolios, such as market value of equity, market-to-book ratio, past returns on
the stock, earnings yield, dividends yield, idiosyncratic volatility of the stock, unexpected
trading volume, accruals, analysts' earnings forecast dispersion, etc. The cross section of
returns on such characteristic-sorted portfolios may depend on macroeconomic risk as well
as other types of risk caused by market frictions (He and Krishnamurthy 2010, Acharya and
Pedersen 2005, Brunnermeier and Pedersen 2009). Hansen and Jagannathan (1997) points
out that all models are potentially wrong, and the question is how wrong a model is. That
cannot be answered by examining any arbitrarily chosen portfolios of assets. That is because,
as Jagannathan and Wang (1996) demonstrates, it is possible to form portfolios of assets
to either mask or highlight what is missing in a given model. Since standard asset pricing
models try to capture macroeconomic risk, we want to use portfolios that highlight such risk
to evaluate these models. The portfolios sorted on recession score developed in this paper
amplify dispersion in recession risk, which is a signicant component of macroeconomic risk,
4
and thus serve as good candidates to test standard asset pricing models.
The paper is organized as follows. Section 1 provides a more detailed review of relevant
literature. Section 2 describes the method used to estimate the recession risk exposure of
stocks using nancial statement information. Section 3 examines alternative methods of
estimating systematic risk for comparison. Section 4 conducts asset pricing tests using the
10 portfolios sorted on recession score. Section 5 analyzes the information content of each
individual nancial ratio and provides economic intuition behind the relations between individual nancial ratios and recession risk exposure of stocks. Section 6 performs robustness
checks. Section 7 concludes.
1
Relevant Literature
In this paper, I develop a measure of recession risk exposure of individual stocks in real time.
In contrast, the literature has focused on characterizing the aggregate stock index returns
over the business cycle. Ferson and Harvey (1991), Chauvet (1998), Harrison and Zhang
(1999), Chauvet and Potter (2000), Campbell and Diebold (2009), and Backus et al. (2010)
have studied the expected equity excess returns conditional on the business cycle. Brandt
and Kang (2004), Ludvigson and Ng (2007), Lettau and Ludvigson (2009), and Lustig and
Verdelhan (2010) focus on the dynamics of the Sharpe ratio over the business cycle. Schwert
(1989) and Hamilton and Lin (1996) investigate the relation between stock volatility and
macroeconomic conditions, while Perez-Quiros and Timmermann (2001) focus on the higher
moments of the stock index returns. Recent papers investigate the returns of a small set
of portfolios (size, book-to-market, and momentum portfolios) over the business cycle as a
way to verify whether the portfolios are exposed to systematic risk (e.g., Perez-Quiros and
Timmermann 2000, Liew and Vassalou 2000, Chordia and Shivakumar 2002, Scheurle and
Spremann 2010). Instead of the verication purpose, the objective of this paper is to estimate
the recession risk exposure of individual stocks. I investigate the relation between stock
5
return performance during recessions and a broad set of rm fundamental attributes, and
develop an estimate of recession risk exposure of stocks using real time nancial statement
and market information.
This research extends the literature that links nancial ratios and rm risk by showing
that in addition to predicting failure probability and equity betas, nancial ratios also predict
a stock's exposure to recession risk. Prior literature has shown that nancial ratios are
useful in predicting a rm's probability of distress (Beaver 1966, Altman 1968, Ohlson 1980,
Demers and Joos 2007, Bhattacharya et al. 2010, Shumway 2001). Furthermore, nancial
ratios have been a useful input in forming bond ratings for corporate loans (Altman et al.
1977) and in predicting a stock's CAPM beta (Beaver et al. 1970, Melicher 1974, Rosenberg
and McKibben 1973, Rosenberg 1974, Rosenberg and Marathe 1975, Breen and Lerner 1973,
Thompson 1976) and cash ow and discount rate betas (Campbell, Polk, and Vuolteenaho
2010). Consistent with prior ndings, I show that rm nancial ratios are also useful in
predicting a stock's exposure to recession risk.
This research also studies the incremental power of nancial ratios in identifying systematic risk exposure beyond the size and book-to-market characteristics. Size and the
book-to-market ratio have been found to be related to systematic risk, even after controlling
for betas. Daniel and Titman (1997) argues that it is a rm's size and book-to-market ratio,
rather than the covariances between the rm's stock returns and the size and book-to-market
factors (SMB and HML), that measure systematic risk. It shows that after controlling for
rm size and book-to-market ratio, rms with dierent historical size betas (the covariance
between a rm's stock returns and the SMB factor) and historical B/M betas (the covariance
between a rm's stock returns and the HML factor) have the same expected returns. The
results indicate that rm size and B/M characteristics are better measures of systematic
risk than betas. Campbell, Polk, and Vuolteenaho (2010) nds that the dierent risk levels
in value and growth stocks are mostly driven by their dierences in rm protability and
leverage. The results suggest that there is a link between rm fundamentals and systematic
6
risk. However, it is not clear whether the eect of fundamentals is incremental to the size and
book-to-market characteristics. This paper extends the literature by showing that nancial
ratios including protability, leverage, accounting liquidity, and receivable turnover ratio,
provide additional information about a rm's recession risk exposure besides the rm's size
and book-to-market characteristics.
This paper is also linked to the literature that explores the relation between stock returns
and characteristics of the production side. Cochrane (1991) develops a production-based asset pricing model that links stocks returns to investment returns. Berk et al. (1999), Carlson
et al. (2004), Zhang (2005), and Gomes and Schmid (2010) model the relations between
rm investment activities and risk characteristics in a theoretical framework. Empirically,
researches have studied the relations between various rm fundamentals of the production
side and stock expected returns (e.g. Dichev 1998, Campbell et al. 2008, Titman et al.
2004, Fama and French 2006). Chen, Novy-Marx, and Zhang (2010) and Novy-Marx (2010)
exploit this link to form pricing factors based on rm fundamentals such as investment and
protability. In contract, instead of stock expected returns or risk in general, I focus on
the recession risk exposure of stocks and develop an estimate of such risk based on rm
fundamentals.
2
2.1
Estimation Method
Data
I use annual nancial statement data from the COMPUSTAT database from 1951 to 2009. I
only include rms with at least three years of data in COMPUSTAT to avoid selection bias,
as indicated in Banz and Breen (1986). I only include rm-year observations with available
data on the set of selected nancial ratios (accounting liquidity, leverage, gross protability
and its variance, net prot margin, and receivable turnover). I exclude nancial rms from
the sample (with SIC code between 6000 and 6999). I use annual instead of quarterly
7
nancial statements because quarterly nancial statements are likely to reect seasonality
in rm nancial ratios. Because dierent industries have dierent seasonal patterns, crosssectional comparisons at each quarter might be mostly driven by seasonality and not be very
informative about the cross-sectional dispersion of rms' exposure to recession risk.
Monthly stock returns are obtained from CRSP from 1926 to 2009. I adjust for delisted
returns as suggested in Shumway (1997). That is, if a rm delists for performance reasons,
and the delisting return and the delisting price is missing, the delisting return is assumed
to be 30%. This paper analyzes monthly returns instead of annual returns mainly for one
reason: it is easier to study a portfolio's recession premium (average return during NBER
recessions minus average return during expansions) with monthly returns, since the NBER
denes recessions within monthly time periods. The nal sample includes 13,241 rms,
131,118 rm-year observations of nancial statements from COMPUSTAT, and 1,509,406
rm-year-monthly return observations from 1965 to 2009.
2.2
Financial Ratios Used and the Composite Recession Score
To investigate rm attributes that are related to a rm's exposure to recession risk, I study
nancial ratios that capture the key risk-relevant dimensions of a rm's operations. Financial
practitioners have acknowledged the usefulness of rm attributes in assessing the risk of
equity. For example, the nancial rm Morningstar favors the use of rm fundamentals
instead of betas when evaluating the cost of equity from a long term perspective. Its analysts
estimate equity risk from four particular categories:
When assigning a cost of equity to a stock, our analysts score a company in the
following areas:
Financial leverage: The lower the debt, the better.
Cyclicality: The less cyclical the rm, the better.
Size: We penalize very small rms.
8
Free cash ows: The higher as a percentage of sales and the more sustainable,
the better.
(Morningstar 2004, p3).
Following this guide, I select nancial ratios in the above four categories to estimate stock
recession risk.
To capture nancial leverage, I include rm leverage ratio (debt-to-equity ratio) as a
measure of overall nancial leverage and liquidity ratio as a measure of short-term solvency.
Debt is computed as the sum of short term debt (COMPUSTAT annual item 34, DLC) and
long term debt (item 9, DLCC), and equity is book equity (item 216, SEQ). The liquidity
ratio is current assets (item 4, ACT) over current liabilities (item 5, LCT).
To capture cyclicality, I use the volatility of a rm's gross protability, measured as the
standard deviation of a rm's gross protability in the past ve years. Gross protability is
measured as the dierence between sales (item 12, SALE) and direct production costs (costs
of goods sold, item 41, COGS) scaled by the total book value of assets (item 6, AT) at the
beginning of the year. Volatility of a rm's cash ows has been used in previous researches,
such as Beaver, Kettler, and Scholes (1970) and Campbell, Polk, and Vuolteenaho (2010),
as an accounting-based risk measure.
In the size category, I include a rm's market capitalization as a measure of size. It is
calculated as close price times total numbers of shares outstanding at the time of portfolio
formation.
To measure the availability and sustainability of a rm's free cash ows, I include a
rm's net prot margin, gross protability, and receivable turnover ratio. Net prot margin
is calculated as net income (EAIT, item 18, IB) over sales (item 12, SALE). The gross
protability measure does not deduct away sales, advertising, and R&D expenditures from
prots and is potentially a better measure of a rm's sustainable prots than earnings and
free cash ows (Novy-Marx 2010).
9
Receivable turnover ratio is calculated as the ratio of accounts receivable (item 2, RECT)
over sales (item 12, SALE). By maintaining accounts receivable, rms are indirectly extending interest-free trade credits to their customers. A high ratio indicates that a rm is lending
extensively to its customers and its collection of debt is inecient. As a result, a rm with
higher receivable turnover has a smaller proportion of cash ows available since more cash
has been lent to customers.
In summary, I estimate the stock recession premium using six nancial ratios (leverage
ratio, liquidity ratio, volatility of gross protability, net prot margin, gross protability, and
receivable turnover ratio) and rm size. Firms with low leverage, high accounting liquidity,
low volatility of gross prots, big size, high net prot margin, high gross protability, and
low receivable turnover should have lower exposure to recession risk. The economic intuition
behind these relations are explained in Section 5.2. The summary statistics of the nancial
ratios used are shown in Table 1.
Overall and Within-Industry Ranks of Financial Ratios
In forming portfolios, I rank each nancial ratio within the whole sample and also within
each industry. The overall ranking (among all rms in the same year) of nancial ratios
captures the industry component of a rm's recession risk. Industry conditions, such as
product market competition, production capital intensity, regulations and entry barriers,
and cyclicality of sales revenues aect the protability, leverage, and accounting liquidity of
an industry. For example, Bradley et al. (1984), Fries et al. (1997), Kovenock and Phillips
(1997), MacKay and Phillips (2005), Miao (2005), and Rauh and Su (2010) nd that
industry factors have a signicant eect on a rm's leverage. Researches show that industry
conditions also signicantly aect rm protability (Gordon and Hoberg 2010) and margins
(Campello 2003). Because the industry eect can be very strong, the overall ranking of ratios
might mainly reect industry characteristics which are associated with recession risk. It is
common wisdom that dierent industries perform well during dierent stages of a business
10
cycle. For example, technology stocks perform well during expansions while utility stocks
perform well during recessions. This phenomenon is known as sector rotation. Financial
practitioners have used this conventional wisdom to try to time the market (e.g., Salsman
1997, Money 2006, Reuters 2008, Stangl et al. 2009). Since the recession risk exposure of
stocks might be largely related industry factors, the overall ranking of nancial ratios that
capture industry characteristics might contain information about recession risk.
On the other hand, the within-industry ranks of nancial ratios capture a rm's relative
position within an industry and are associated with the recession risk exposure of stocks.
I capture the intra-industry variations across rms by ranking each rm's nancial ratios
within its respective industry (3-digit SIC code) each year. Researchers have found that industry adjusted ratios are more informative for investors in evaluating rm value, predicting
corporate bankruptcy, and predicting rm performance than unadjusted ratios. While many
studies nd no relationship between rm leverage and rm value, Aggarwal and Zhao (2007)
nds that rm leverage is negatively related to rm value after controlling for industry leverage. Studies also nd that using industry adjusted nancial ratios improves the predictions
of corporate bankruptcy (Platt and Platt 1990, 1991) and predictions of future protability
(Soliman 2004) compared to using unadjusted nancial ratios. Because the total recession
risk exposure of a rm consists of a rm-specic component, I rank nancial ratios within
its respective industry in addition to the overall ranking.
In each year, based on each nancial ratio, I rank all rms into 10 deciles, both within
the whole sample and within its industry respectively. In each year, each rm has 7 overall
ranks and 7 within-industry ranks based on the 6 nancial ratios and size. Each individual
rank is a number between 1 and 10: 1 if the nancial ratio is in the lowest decile, and 10 if
the nancial ratio is in the highest decile.
11
Composite Measure: Recession Score
Inspired by Piotroski (2000)1 and Mohanram (2005), I dene the composite measure, recession score, as a simple aggregation of the individual ranks on leverage, accounting liquidity,
volatility of gross protability, size, net prot margin, gross protability, and receivable
turnover. The recession score for a rm at year
t
is the sum of its ranks in the nancial
ratios that are expected to be negatively related to the stock's recession risk exposure (accounting liquidity, size, net prot margin, and gross protability) minus its ranks in the
ratios that are positively related to the stock's recession risk exposure (leverage, volatility
of gross protability, and receivable turnover).
Regarding the choice of overall versus within-industry ranks in constructing the recession
score, I apply the following rule: For a nancial ratio, if the recession premium spread
between the top and the bottom decile portfolios based on the overall ranking is larger than
the spread based on the within-industry ranking, I include the overall rank of the ratio
in the calculation of recession score, and vice versa. If both rankings generate statistically
signicant spreads, I include both ranks in the recession score. Details on portfolio formation
and estimation of recession premiums are explained in the subsequent subsections.
This simple scoring and aggregation method is easily implementable. It might work
better out-of-sample than nonlinear multivariate estimation methods because it has less
estimation error. DeMiguel et al. (2009) shows that a simple portfolio allocation strategy
has the best out-of-sample performance compared to sophisticated methods because the
gain from optimal diversication is outweighted by increased estimation error. Using the
same intuition, I apply this simple method to avoid overtting the in-sample data and avoid
potential estimation error associated with the more sophisticated models.
1I
thank Ravi Jagannathan for suggesting the use of nancial ratios as in Piotroski (2000) for assessing
business cycle risk.
12
2.3
Portfolio Formation
To analyze the relations between rm nancial ratios and recession risk exposure, I construct
portfolios based on each nancial ratio as well as on recession score and study the portfolio
returns. The portfolios are constructed similarly to the way the 25 size and book-to-market
portfolios are constructed in Fama and French (1993). I select stocks based on their nancial
ratios, the values of the nancial ratios within their respective industries, and their recession
scores. The portfolios are constructed as follows: in September of each year t, I compute
nancial ratios and recession scores using annual nancial statements with scal year-end
from June of year
t-1
to May of year
t
(scal year
t-1 ).
The four-month lag period ensures
that the nancial statement data used is available to the public at the time of portfolio
formation (Banz and Breen 1986). Based on the values of nancial ratio
i
or recession
scores, I rank all rms into 10 deciles: 10 being the decile with the highest values in ratio
i
or recession score, and 1 being the decile with the lowest values in ratio
i
or recession
scores. The stocks in each decile portfolio are equally weighted, held for 12 months, and
then rebalanced in September of year
2.4
t+1.
Returns of Portfolios Sorted on Individual Financial Ratios and
Recession Score
In this section, I analyze the monthly returns and recession premiums of portfolios sorted on
the individual nancial ratios as well as the recession score that aggregates the information
in all these ratios. A portfolio with high recession premium, that is, low excess returns
during expansions and high excess returns during recessions, has low exposure to recession
risk, since its payos are high during bad times when returns are more valuable to investors.
13
2.4.1 Estimation of Recession Premium
To estimate the recession premium of a portfolio, I run an OLS regression of the portfolio
returns on a constant and a recession dummy (Recession Dummyt equals 1 if month
in an NBER recession and 0 if month
t
t
is
is in an NBER expansion) over the whole sample
period:
rit = Intercepti + Recession P remiumi ∗ Recession Dummyt + εit ,
where rit is the return of portfolio i in month t. The intercept is an estimate of the average
return of portfolio
i
during expansions, and the Recession P remiumi coecient captures
the average return dierence of portfolio
i
between recessions and expansions. The sum of
the intercept and the recession premium coecient captures the average return of portfolio i
during recessions. A larger recession premium coecient implies lower recession risk exposure
of portfolio i.
2.4.2 Recession Premium Spreads on Individual Financial Ratios
To assess the ability of the individual nancial ratios to predict recession premium, I form
a high-minus-low portfolio based on each ratio. For example, the high-minus-low portfolio
based leverage ratio (debt-to-equity) holds stocks with high leverage ratios (top 10%) and
shorts stocks with low leverage ratios (bottom 10%). The estimate of recession premium
on the high-minus-low portfolio based on a ratio measures the spread in recession premium
between the high and the low portfolios. A large spread in the recession premium means
that the high and low portfolios based on the ratio are very dierent in their exposures to
recession risk, indicating that the ratio is informative of recession risk exposure of stocks.
The estimates of recession premium of the high-minus-low portfolios sorted on the individual nancial ratios are shown in Table 2. When sorted on the overall ranks of nancial
ratios, portfolios with high gross protability, high liquidity, and low receivable turnover do
increasingly better during recessions. The recession premiums for the 3 high-minus-low port-
14
folios are economically and statistically signicant (1.37%, 0.85%, and -1.24% per monthly
for gross protability, liquidity, and receivable turnover respectively). The recession premium spread generated by size is large (0.92%) but not signicantly, indicating that size
potentially contains a lot of noise as a measure of recession risk exposure. When ranked
within each industry, portfolios with high gross protability, high liquidity, and low leverage do increasingly well during recessions. The recession premiums for the high-minus-low
portfolios sorted within each industry on gross protability, liquidity and leverage are 0.68%,
0.73%, and -0.68% per month respectively, and all are statistically signicant. Overall, rm
leverage, liquidity, gross protability, and receivable turnover are found to have the most
signicant relations with stock recession premiums.
2.4.3 Returns of the Recession Score Portfolios
The recession score aggregates information in individual nancial ratios and is computed as:
Recession Score = overall&industryRank(liquidity)/2 − industryRank(leverage)
−industryRank(σ[gross prof itability]) + industryRank(size) + overallRank(prof it margin)
+overall&industryRank(gross prof itability)/2 − overallRank(receivable turnover).
Results show that recession score is a good estimate of the recession risk exposure of
stocks. As shown in Table 3, rms with higher recession scores have higher returns during
recessions and lower recession risk exposure. The high recession score portfolio, which consists of stocks with recession scores in the highest decile, earns 1.3% per month more during
recessions than the low recession score portfolio, which contains stocks with recession scores
in the lowest decile. The high-minus-low portfolio, which holds the high recession score portfolio and shorts the low recession score portfolio, generates a signicant recession premium
of 1.72% per month. The high recession score portfolio is safer than the low recession score
portfolio in the sense that it has higher returns during economic downturns.
15
Results are also consistent with market eciency and the recession risk been priced by
investors. During expansions, the safe high recession score portfolio underperforms the risky
low recession score portfolio by 0.40% per month. When evaluated over all time periods, the
safe portfolio earns 0.13% per month less than the risky portfolio. In Figure 1, I present a
scatter plot of the average excess returns versus the recession premiums of the 10 recession
score portfolios. The relation is negative and fairly linear, and the variation in portfolio
recession premiums accounts for 52.6% of the variation in portfolio average returns of the
10 portfolios sorted on recession score. The fact that the safer portfolios have lower average
returns indicates that investors are willing to pay a higher price for the safer portfolios that
oer them protection during recessions.
The estimate of systematic risk based on recession score is also correctly aligned with
other standard measures of systematic risk. Portfolios with higher recession scores also have
lower CAPM betas, Fama-French three-factor betas and consumption betas as shown in
Table 4 (CAPM and FF3) and in Table 5 (CCAPM).
3
Alternative Methods
This section compares the use of the recession score with alternative methods of estimating
systematic risk, including sorting on historical CAPM beta, historical consumption beta, and
rm characteristics such as size, book-to-market ratio, momentum, and product durability.
The portfolios sorted on recession score generate the largest spread in recession premium
compared to the alternative methods.
3.1
Portfolios Sorted on Historical CAPM Betas
Sortings based on historical CAPM betas does not provide satisfactory results. The dierence
in recession returns between the decile portfolios with the highest and lowest betas is -0.50%,
much smaller than the dierence in recession returns based on recession score (1.31%), and
16
is not statistically signicant as shown in Table 6.
Historical CAPM betas are calculated from rolling regressions of weekly individual stock
excess returns on weekly excess market returns. At the end of each month, I compute the
historical CAPM betas for all the stocks in the CRSP universe using returns in the past
one-year window. I require each stock to have at least 26 weekly observations in the past
year. I then sort stocks into ten deciles based on their historical CAPM betas. Stocks in
each portfolio are equally weighted and held for one month. The portfolios are rebalanced
at the end of each month.
Historical CAPM betas are not good estimates of recession premiums. Table 6 shows
that stocks with high historical betas earn only 0.50% per month less than stocks with low
historical betas during recessions, and the dierence is not signicant. The lack of power
of CAPM betas to predict recession premiums is potentially due to either instability in the
CAPM betas or large measurement errors in estimating individual stock betas. The failure
of the historical CAPM betas to predict recession premiums might also be caused by capital
constraints of investors. Frazzini and Pedersen (2010) shows that when funding constraints
tighten (which is likely to happen in recessions), low beta stocks earn lower returns than high
beta stocks as levered investors have to delever and sell their low beta stocks. As a result,
even if historical beta is a perfect estimate of true beta, low beta stocks might still not have
higher returns during recessions than high beta stocks. Overall, my results show that the
recession score is a better estimate of recession risk exposure of stocks than historical CAPM
betas.
3.2
Portfolios Sorted on Historical Consumption Betas
Since recessions are states where marginal utility is high, stock consumption betas might
be good estimates of recession premiums. However, I nd that it is not true: Stocks with
high historical consumption betas do not have signicantly lower returns during recessions
as compared to stocks with low historical consumption betas.
17
Historical consumption betas are calculated from rolling regressions of quarterly stock
excess returns on quarterly real per capita consumption growth in nondurables and services.
The estimation window is 15 years (60 data points). At the end of each quarter, I compute
historical consumption betas for all stocks in the CRSP universe with at least 30 data
points. Then I sort stocks into ten decile portfolios. Stocks in each portfolio are held for one
quarter and the portfolios are rebalanced at the end of each quarter. The returns are equally
weighted in each portfolio. To compute quarterly real per capita consumption growth, I use
quarterly seasonally adjusted aggregate nominal consumption expenditure on nondurables
and services from 1966 to 2009 from National Income and Product Accounts (NIPA) Table
2.3.5. Nominal consumption is divided by price deator (NIPA Table 2.3.4) and population
(NIPA Table 2.1) to construct a series of per capita real consumption.
Historical consumption betas are not good estimates of recession premiums. Table 6
shows that stocks with high historical consumption betas earn 0.30% per month more than
stocks with low historical consumption betas during recessions, but the dierence is not
signicant. The inability of historical consumption betas to capture recession premiums
might be caused by large measurement errors in estimating individual stock consumption
betas.
To reduce measurement errors, I compute historical industry consumption betas similarly
as computing stock consumption betas and sort stocks based on their industry betas. Stocks
in the same industry are likely to have similar consumption betas, and portfolio betas are
estimated with smaller measurement errors than individual betas. Historical industry consumption betas are not good estimates of recession premiums. As shown in Table 6, there is
no statistical dierence between recession returns of portfolios with high historical industry
betas and those with low betas.
18
3.3
Other Cyclical Portfolios
In addition to historical betas, I also compare the recession premium spread generated by
recession score with the spread generated by a set of portfolios that are known to have
varying returns across the business cycle. In this set, I include the market portfolio, the
small size minus big size portfolio (SMB, Fama and French 1993), the value minus growth
portfolio (HML, Fama and French 1993), the momentum portfolio (Jegadeesh and Titman
1993), the 30 day Treasury-bill rate, the long-term (10-year) bond rate, a nondurable goods
portfolio, a durable goods portfolio, and a long nondurable short durable goods portfolio
(Yogo 2006, Gomes, Kogan, and Yogo 2009)2 .
The results are shown in Table 6. The recession premium spread generated by recession
score (1.72% per month) is larger than the spread generated by SMB (0.03%), HML (0.10%),
long term bonds (0.54%), durables and nondurable stocks (0.88%), and all other portfolios
mentioned earlier. Overall, the results provide evidence that rm nancial ratios contain
more information about recession risk exposure of stocks as compared to historical CAPM
and consumption betas as well as rm characteristics such as size, B/M, momentum, and
product durability.
There are several possible reasons that nancial ratios work better than these alternative measures. First, compared to historical betas, nancial ratios provide a broader and
more timely estimate of a stock's recession risk exposure. Firms' operations are constantly
changing over time and their risk exposure changes as well. Thus, betas appear to be timevarying (Blume 1971, 1975). For example, a rm's recession risk exposure might change
dramatically after a major acquisition or a spin-o. Such a change will be reected in the
rm's nancial statement within a year. However, the estimated historical beta will only
recognize the change in the rm's risk exposure gradually over time since it has to be estimated over a longer horizon. Another example is that rm investment decisions aect a
2I
thank João F. Gomes, Leonid Kogan, and Motohiro Yogo for making the durable and nondurable
industry classication data available.
19
rm's systematic risk (Berk et al. 1999, Carlson et al. 2004, Zhang 2005, and Gomes and
Schmid 2010). Firm attributes such as protability, leverage, and book-to-market ratio will
reect such changes in recession risk exposure while it takes longer for the change to show
up in historical betas. Second, individual stock betas are known to be measured with large
estimation errors (Alexander and Chervany 1980). Financial ratios of individual rms might
contain less noise since its measuring process is straightforward. Therefore, sorting based on
noisy individual stock betas might not generate portfolios with a large spread in recession
risk. Third, nancial ratios capture the key dimensions of a rm's operations and provide a
potentially more comprehensive picture of a rm's risk exposure than single measures such
as size, book-to-market, and past returns.
4
Asset Pricing Tests of Linear Factor Pricing Models
One important implication of this new estimate is to construct portfolios to test asset pricing
models. The 10 recession score portfolios highlight recession risk and make good candidates
to test whether the linear factor pricing models capture systematic risk at business cycle
frequencies. In this section, I test linear factor pricing models (the CAPM, the Fama-French
three-factor model Fama and French 1993, and the consumption-based CAPM) using the 10
recession score portfolios. As shown in Section 2, the 10 portfolios generate a large spread in
returns during recessions. Portfolios that have higher recession scores also have lower CAPM
betas, Fama-French three-factor betas and consumption betas. Test results show that the
CAPM model is rejected at the 5% level. However, the Fama-French three-factor model and
the consumption-based CAPM are not rejected
To test the ability of the CAPM and the Fama-French three-factor model to price the
ten portfolios, I use the GRS test (Gibbons, Ross, and Shanken 1989). The GRS test is a
statistical test of the hypothesis that the alphas of the ten portfolios are jointly zero. The
GRS test results are shown in Table 7. The CAPM model is rejected, but the Fama-French
20
three-factor model is not. This result implies that the Fama-French three-factor model is
more successful at capturing recession risk as compared to the CAPM model.
I test the consumption-based CAPM with consumption growth at two horizons: (1)
quarterly consumption growth and (2) annual consumption growth from the fourth quarter
to the next fourth quarter (Q4-Q4). Jagannathan and Wang (2007) shows that the Q4-Q4
consumption CAPM is successful at pricing the 25 size and book-to-market portfolios because
investors tend to make investment and consumption decisions simultaneously at the end of
each year. Consumption growth is computed using two consumption series: (1) real per
capita consumption growth in nondurables and services and (2) real per capita consumption
growth in only nondurables.
I consider a linear version of the CCAPM following Breeden et al. (1989) and Jagannathan
and Wang (2007):
E[Ri,t+j ] = λcj βicj ,
where
βicj
)
cov(Ri,t+j , ct+j
ct
.
=
ct+j
var( ct )
βicj is the consumption beta and λcj is the market price of consumption risk. I test the
above specication using the cross-sectional regression (CSR) method3 of Black et al. (1972)
and Fama and MacBeth (1973).
Test results are shown in Table 8. In general, the consumption CAPM does well in
pricing the 10 recession score portfolios. When using consumption growth in nondurables
and services, the CCAPM with quarterly consumption growth has an insignicant intercept.
However, the CCAPM with Q4-Q4 consumption growth has a signicantly positive intercept.
The weak performance of the Q4-Q4 CCAPM should not come as a surprise. It uses annual
returns while the quarterly CCAPM uses quarterly returns. There are signicantly fewer
observations used in testing the Q4-Q4 CCAPM (43) than in testing the quarterly CCAPM
3I
thank Ravi Jagannathan and Wong Wang for providing the Matlab code for the CSR test used in their
paper.
21
(173). The sample period only covers 7 recessions, and each recession lasted for around one
year. Thus, there are only about 7 annual observations reecting recession years. The test
result in such a small sample might not be indicative of the true pricing power of the Q4-Q4
CCAPM.
Using the consumption growth in only nondurables improves the pricing power of both
CCAPM. Both quarterly and Q4-Q4 specications have smaller and insignicant intercepts
as compared to using consumption growth in nondurables plus services. The quarterly and
Q4-Q4 consumption betas also have larger R2 in explaining excess returns as compared to
using consumption growth in nondurables plus services. The results suggest that compared
to nondurables plus services, consumption in nondurables alone might be a better proxy for
the consumption process of the representative agent in the CCAPM.
5
Information Content of the Individual Financial Ratios
5.1
Regression Analysis
This section analyzes the information content of each individual nancial ratio. I nd that
most nancial ratios used in constructing the composite recession score contribute to the
ability of recession score to dierentiate the recession risk exposure of stocks.
Because nancial ratios are inter-correlated, in order to assess the additional information
of recession risk contributed by each nancial ratio, I estimate a pooled OLS regression
of monthly individual stock returns on a recession dummy (Rect = 1 when month
t
is in
a recession and 0 otherwise), the lagged ranks of the nancial ratios (from 1 to 10), and
the interactions between the recession dummy and the lagged ranks of nancial ratios. For
rm nancial ratios computed using nancial statements with scal year ends in scal year
t
(June of calendar year
t
to May of calendar year
t+1 ),
the respective stock returns in
the regression are the monthly returns from October of calendar year
calendar year
t+1.
t+1
to September of
There is a four months lag period to ensure that the nancial statement
22
information has been made available to the public during the corresponding return period
as suggested in Banz and Breen (1986).
The parameters of interest are the coecients on the interactions. The coecient on the
interaction term between the recession dummy and a nancial ratio measures the relation
between a stock's recession-expansion return dierence and the nancial ratio. If the coefcient on the interaction term is negative, it means that for rms with high levels in this
nancial ratio, their stock returns during recessions drop a lot more from their returns during
expansions compared to rms with low levels in this ratio. That is, a negative coecient
means that the nancial ratio is positively associated with recession risk exposure of stocks.
The results of this regression (Table 9) indicate that most of the ratios are signicantly
related to returns during recessions and have the expected signs. The volatility of gross
protability does not signicantly predict recession-expansion return dierences. Also, while
controlling for gross protability, rms with a higher net prot margin have lower returns
during recessions relative to expansions.
5.2
Economic Intuition
This section discusses the economic intuition behind the relations between individual nancial ratios and the recession risk exposure of stocks.
Leverage Ratio
Researches such as Hamada (1972), Black and Scholes (1973), Galai and Masulis (1976),
Hecht (2000), and Charoenrook (2004) have indicated that rms with higher leverage (debtto-equity ratio) have higher exposure to systematic risk. The eect is two-fold. First,
stockholders bear a larger proportion of variation in cash ows for rms with higher leverage.
Increased leverage increases the equity share of cash ows in the good state of the world and
decreases the equity share of cash ows in the bad state of the world. Second, rms with
higher leverage generally have higher risk of bankruptcy. Such risk can be systematic if rms
23
face higher risk of bankruptcy during recessions. This scenario is very likely because during
recessions most rms' cash ows are negatively aected, which makes it harder for them to
service their debt. Hence, the action of taking more debt increases a rm's risk by amplifying
the risk exposure of its equityholders and increasing its probability of default.
Even when a rm's capital structure is endogenously chosen, rm leverage could still
be positively related to stock systematic risk. Some may argue that rms with higher risk
will nance with more equity than debt, so that rms with lower leverage might be rms
with higher risk exposure. This argument is not contradictory with leverage being positively
associated with systematic risk. One instance in which rms with higher leverage also have
higher recession risk is when rms choose their capital structures according to their total
risk, which consists of both systematic risk and idiosyncratic risk. Two rms could have
the same systematic risk in their cash ows but dierent idiosyncratic risk, such as risk
exposure to hurricanes due to their dierent locations. The rm in the region with higher
probability of hurricanes is more likely to choose a lower level of debt since its cash ows are
more volatile. During a recession, the two rms will have the same negative shock to their
cash ows since the systematic component of cash ow risk is the same for the two rms.
However, the rm with higher leverage will have lower returns during recessions since it has
less of a buer against bankruptcy. Thus, leverage is positively related to the recession risk
exposure of stocks.
Accounting Liquidity
The liquidity ratio is an indicator of a rm's short term solvency. It is used to determine a
company's ability to pay o its short term obligations. The liquidity ratio is calculated as
current assets over current liabilities. Current liabilities include short term debt, accounts
payable, and other short term liabilities. The amount of accounts payable measures the
extent of how much trade credits a rm uses. Trade credits have been found to be an
important source of nancing, especially for small rms (Petersen and Rajan 1994, 1997).
24
Petersen and Rajan (1997) nds that rms use more trade credits when bank loans are not
available. If we assume that rms have similar levels of current assets and short term debt,
then a high liquidity ratio means that a rm has a low level of trade credits, which means
that it has high trade credit capacity. Compared to a rm in the same industry which has
already used a high amount of trade credits, a rm with a lower level of accounts payable
will be able to borrow more from its suppliers during bad economic times, when it is hard
to get bank loans. High current assets and low current liabilities provide a safety net during
bad economic times. Therefore, rms with high liquidity ratios should have low levels of
recession risk exposure.
Size
The regression result suggests that small rms have lower returns during recessions. According to Bernanke and Gertler (1989), Gertler and Gilchrist (1994), Kiyotaki and Moore
(1997), Perez-Quiros and Timmermann (2000), one reason that rm size contains information about recession risk is that the amount of external nance available to small and big
rms are aected asymmetrically by recessions. Due to agency costs, there is asymmetry
in rm information available to creditors, so it is necessary for rms to use collateral when
borrowing in the credit markets. Small rms have increasingly less collateral than big rms
during recessions and are therefore more adversely aected.
Protability
Both the portfolio and regression analysis show that a rm's gross protability is negatively
related to recession risk exposure; that is, rms with higher gross prots have lower recession
risk exposure. Firm protability has found to be persistent. Fama and French (2006)
nds that current protability is the best predictor of future protability. Furthermore,
Novy-Marx (2010) nds that gross prot is a better predictor of future protability than
other protability measures, such as earnings and free cash ows. Unlike earnings and
25
free cashows, the gross protability measure does not deduct sales, advertising, and R&D
expenditures, which are positively related to future prots. Hoberg and Phillips (2010)
nds that rms use advertising to dierentiate themselves from competitions and increase
protability. Therefore, gross protability is a better measure of the true economic rents that
a rm collects from its customers. For example, consider two rms that have the same net
income, but rm A has higher gross prots than rm B because it spends more on building
its brand through sales promotions and advertising. Firm A is likely to have a more loyal
customer base and more persistent revenue, which is especially valuable during bad economic
times. The gross prot measure is informative of this dierence, while net income is not.
Firms' persistent prots are related to their market power, and market power insulates a
rm from aggregate economic uctuations. The dierences in market power could be caused
by dierences in good-specic habit levels (van Binsbergen 2007), product dierentiation
(Hoberg and Phillips 2010), product market competition (Peress 2010), and entry barriers.
Through the above-mentioned channels, market power insulates rms from aggregate shocks
in the economy (van Binsbergen 2007, Peress 2010). During bad economic times, when
consumer income decreases, highly protable rms still have stable revenues because their
market power stays strong, while the income for unprotable rms might dwindle a lot due
to their low market power. In this case, a rm's gross protability measure is negatively
associated with its exposure to recession risk.
This intuition also explains the regression result that net prot margin is positively
related to recession risk exposure when gross protability is controlled for. When gross
protability is held the same, rms with higher net prot margin have lower non-production
costs, including sales, advertising, and R&D expenditures. Low spending on these items
decreases a rm's advantage over its competitors and might result in low market power in
the long run, which in turn increases the rm's exposure to recession risk.
26
Receivable Turnover Ratio
Evidence shows that rms with higher receivable turnover ratios have lower returns during
recessions. The receivable turnover ratio measures the accounts receivable as a percentage of
total sales. By maintaining accounts receivable, rms are indirectly extending interest-free
trade credits to their customers. A high ratio indicates that a rm is lending extensively
to its customers and its collection of accounts receivable is inecient. It implies that there
is a smaller amount of cash ows available, since cash has been lent to customers. During
bad times, the customers might rely more on trade credits as a source of funding since
bank lending might not be available (Petersen and Rajan 1997). In this case, rms that
are inecient in collecting trade credits from their customers will observe even less available
cash ow during bad times.
6
Robustness Checks
6.1
Adjust for Size, Book-to-Market and Momentum Benchmarks
Since rm characteristics such as size, book-to-market, and momentum are known to be
associated with systematic risk, it is important to identify the incremental information of recession risk provided by nancial ratios. To control for the eect of size, book-to-market, and
momentum characteristics, I use characteristics-benchmark-adjusted excess returns in the
portfolio analysis (size and book-to-market benchmarks, Daniel and Titman 1997; size, bookto-market, and momentum benchmarks, Daniel, Grinblatt, Titman, and Wermers 1997). The
characteristic-adjusted excess return of a stock in month
benchmark portfolio return in month
t
t
is computed by subtracting the
from the raw stock return in that month. For each
stock, I nd its benchmark portfolio by matching its size, book-to-market ratio, and past
twelve-month return quintiles.
The size, book-to-market, and momentum benchmark portfolios are constructed following
27
Fama and French (1993), Daniel and Titman (1997), and Daniel, Grinblatt, Titman, and
Wermers (1997). In calculating book-to-market ratios, the book equity used is obtained from
the annual nancial statements with scal year ends in calendar year
is from the last trading day of year
t-1.
t-1,
and market equity
In calculating size, I use the market equity on the
last trading day of June of year t. Past returns are calculated as the twelve-month return
from the end of June in year
t-1
to the end of May in year t. Firms with less than two years
history on COMPUSTAT are eliminated from the sample. All the stocks with sucient data
are rst sorted into quintiles based on rm size, then on book-to-market ratio, and lastly on
past twelve-month returns. The breakpoints for the size sort are the quintile breakpoints of
the NYSE stocks. The size, book-to-market, and past return quintiles of the rms are then
applied to rms from July of year
t
through June of year
t+1.
The previous results are still persistent. As shown in Table 10, after adjusting for the
size and book-to-market benchmarks, stocks with high recession scores still have signicantly
higher returns (1.23%) during recessions than stocks with lower recession scores. Table 11
shows that after controlling for the size, book-to-market, and momentum benchmarks, stocks
with high recession scores earns 1.26% more during recessions. This return is signicant at
the 1% level. The results indicate that rm nancial ratios have power in estimating recession
risk for reasons other than picking up the eect of big, growth, and momentum rms that
do well during recessions.
6.2
Returns during Each NBER Recession and Expansion
Table 12 shows the average returns earned by the high, low, and high-minus-low portfolios
sorted on recession score during each NBER recession and expansion in the sample period
(1966 to 2009). Stocks with high recession scores (top decile) have higher returns than
stocks with low recession scores (bottom decile) in six of the seven recessions. The oil shock
recession from 1973 to 1975 is the only exception, during which the high recession score
stocks earned 1.65% per month less than low recession score stocks.
28
7
Conclusion
This paper adds to the literature by constructing portfolios that highlight recession risk for
the purpose of testing standard asset pricing models. The paper develops a better estimate
of recession risk exposure of stocks than would be possible with historical betas, size, bookto-market, momentum, and product durability. I show that nancial ratios can help identify
stocks that outperform other stocks during recessions, even after controlling for rm historical
betas and characteristics such as size, book-to-market ratio, and past returns. I nd that
during recessions, rms with high gross protability, high accounting liquidity, low receivable
turnover, and low industry-adjusted leverage earn 1.10%, 0.88%, 1.31%, and 0.67% per month
respectively more than rms with low protability, low liquidity, high receivable turnover,
and high industry-adjusted leverage. Using a parsimonious composite recession score based
on six nancial ratios and size, rms with recession scores in the safest decile earn 1.31% per
month more than rms with recession scores in the most risky decile during recessions. The
return dierences are smaller but signicantly negative during expansions, suggesting that
the ability of nancial ratios to pick up the recession risk exposure of stocks is not an anomaly.
When tested with the 10 portfolios sorted on recession score, the CAPM model is rejected at
the 5% level, while the Fama-French three-factor model and the consumption-based CAPM
are not.
Also, the ndings of this paper would be of particular interest to long term investors
who care about protection in economic downturns. Using rm characteristics derived from
nancial ratios, which are readily available to the public, long term investors can avoid stocks
with a high level of recession risk and favor stocks with a low level of recession risk when
choosing their optimal portfolio mix of securities.
29
References
Acharya, V., Pedersen, L. H., 2005. Asset pricing with liquidity risk. Journal of Financial
Economics 77, 375410.
Aggarwal, R., Zhao, X., 2007. The leverage-value relationship puzzle: an industry eects
resolution. Journal of Economics and Business 59, 286297.
Alexander, G., Chervany, N. L., 1980. On the estimation and stability of beta. Journal of
Financial and Quantitative Analysis 15, 123137.
Altman, E., Haldeman, R. G., Narayanan, P., 1977. Zeta analysis: A new model to identify
bankruptcy risk of corporations. Journal of Banking and Finance 1, 2954.
Altman, E. I., 1968. Financial ratios, discriminant analysis and the prediction of corporate
bankruptcy. The Journal of Finance 23, 589609.
Backus, D. K., Routledge, B. R., Zin, S. E., 2010. The cyclical component of us asset returns.
Working paper.
Banz, R. W., Breen, W., 1986. Sample-dependent results using accounting and market data:
Some evidence. The Journal of Finance 41, 779793.
Beaver, W., Kettler, P., Scholes, M., 1970. The association between market determined and
accounting determined risk measures. The Accounting Review 45, 654682.
Beaver, W. H., 1966. Financial ratios as predictors of failure. Empirical Research in Accounting: Selected Studies 4, 71111.
Berk, J. B., Green, R. C., Naik, V., 1999. Optimal investment, growth options and security
returns. The Journal of Finance 54, 11531607.
Bernanke, B., Gertler, M., 1989. Agency costs, net worth, and business uctuations. American Economic Review 79, 1431.
Bhattacharya, N., Demers, E., Joos, P., 2010. The relevance of accounting information in
a stock market bubble: Evidence from internet ipos. Journal of Business Finance and
Accounting 37, 291321.
Black, F., Jensen, M., Scholes, M., 1972. Studies in the Theory of Capital Markets. Praeger,
New York, Ch. The capital asset pricing model: some empirical tests, pp. 79121.
Black, F., Scholes, M., 1973. The pricing of options and corporate liabilities. Journal of
Political Economy 81, 637654.
Blume, M. E., 1971. On the assessment of risk. The Journal of Finance 26, 110.
Blume, M. E., 1975. Betas and their regression tendencies. The Journal of Finance 30, 785
795.
30
Bradley, M., Jarrell, G., Kim, E., 1984. On the existence of an optimal capital structure:
Theory and evidence. Journal of Finance 39, 857881.
Brandt, M. W., Kang, Q., 2004. On the relationship between the conditional mean and
volatility of stock returns: A latent var approach. Journal of Financial Economics 72,
217257.
Breeden, D., 1979. An intertemporal asset pricing model with stochastic consumption and
investment. Journal of Financial Economics 7, 265296.
Breeden, D., Gibbons, M., Litzenberger, R., 1989. Empirical tests of the consumptionoriented capm. The Journal of Finance 44, 231262.
Breen, W. J., Lerner, E. M., 1973. Corporate nancial strategies and market measures of
risk and return. The Journal of Finance 28, 339351.
Brunnermeier, M., Pedersen, L. H., 2009. Market liquidity and funding liquidity. The Review
of Financial Studies 22, 20012238.
Campbell, J. Y., Hilscher, J., Szilagyi, J., 2008. In search of distress risk. The Journal of
Finance 63, 28992939.
Campbell, J. Y., Polk, C., Vuolteenaho, T., 2010. Growth or glamour? fundamentals and
systematic risk in stock returns. Review of Financial Studies 23, 305344.
Campbell, S. D., Diebold, F. X., 2009. Stock returns and expected business conditions: Half
a century of direct evidence. Journal of Business and Economic Statistics 27, 266278.
Campello, M., 2003. Capital structure and product markets interactions: evidence from
business cycles. Journal of Financial Economics 68, 353378.
Carlson, M., Fisher, A., Giammarino, R., 2004. Corporate investment and asset price dynamics: Implications for the cross-section of returns. The Journal of Finance 59, 25772603.
Charoenrook, A., 2004. The role of capital structure in cross-sectional tests of equity returns.
Working Paper.
Chauvet, M., 1998. Stock market uctuations and the business cycle. Journal of Economic
and Social Measurement 25, 235257.
Chauvet, M., Potter, S., 2000. Coincident and leading indicators of the stock market. Journal
of Empirical Finance 7, 87111.
Chen, L., Novy-Marx, R., Zhang, L., 2010. An alternative three-factor model. Working
Paper.
Chordia, T., Shivakumar, L., 2002. Momentum, business cycle, and time-varying expected
returns. The Journal of Finance 57, 9851019.
31
Cochrane, J. H., 1991. Production-based asset pricing and the link between stock returns
and economic uctuations. The Journal of Finance 46, 209237.
Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997. Measuring mutual fund performance with characteristic-based benchmarks. The Journal of Finance 62, 10351058.
Daniel, K., Titman, S., 1997. Evidence on the characteristics of cross sectional variation in
stock returns. The Journal of Finance 52, 133.
Demers, E., Joos, P., 2007. Ipo failure risk. Journal of Accounting Research, 45, 333371.
DeMiguel, V., Garlappi, L., Uppal, R., 2009. Optimal versus naive diversication: How
inecient is the 1/n portfolio strategy? The Review of Financial Studies 22, 19151953.
Dichev, I. D., 1998. Is the risk of bankruptcy a systematic risk? The Journal of Finance 53,
11311147.
Epstein, L. G., Zin, S. E., 1989. Substitution, risk aversion, and the temporal behaviour of
consumption and asset returns: A theoretical framework. Econometrica 57, 937969.
Fama, E., MacBeth, J., 1973. Risk, return, and equilibrium: Empirical tests. Journal of
Political Economy 71, 607636.
Fama, E. F., French, K. R., 1993. Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics 33, 356.
Fama, E. F., French, K. R., 2006. Protability, investment and average returns. Journal of
Financial Economics 82, 491518.
Ferson, W. E., Harvey, C. R., 1991. The variation of economic risk premiums. Journal of
Political Economy 99, 385415.
Frazzini, A., Pedersen, L., 2010. Betting against beta. Working Paper.
Fries, S., Miller, M., Perraudin, W., 1997. Debt in industry equilibrium. Review of Financial
Studies 10, 3967.
Galai, D., Masulis, R. W., 1976. The option pricing model and the risk factor of stock.
Journal of Financial Economics 3, 5381.
Gertler, M., Gilchrist, S., 1994. Monetary policy, business cycles, and the behavior of small
manufacturing rms. Quarterly Journal of Economics 109, 309340.
Gibbons, M. R., Ross, S. A., Shanken, J., 1989. A test of the eciency of a given portfolio.
Econometrica 57, 11211152.
Gomes, J. F., Kogan, L., Yogo, M., 2009. Durability of output and expected stock returns.
The Journal of Political Economy 117, 941986.
Gomes, J. F., Schmid, L., 2010. Levered returns. The Journal of Finance 65, 467494.
32
Gordon, P., Hoberg, J., 2010. Real and nancial industry booms and busts. The Journal of
Finance 65, 4596.
Hamada, R. S., 1972. The eect of the rm's capital structure on the systematic risk of
common stocks. The Journal of Finance 27, 435452.
Hamilton, J., Lin, G., 1996. Stock market volatility and the business cycle. Journal of Applied
Econometrics 11, 573593.
Hansen, L. P., Jagannathan, R., 1997. Assessing specication errors in stochastic discount
factor models. The Journal of Finance 52, 557590.
Harrison, P., Zhang, H. H., 1999. An investigation of the risk and return relation at long
horizons. Review of Economics and Statistics 81, 399408.
He, Z., Krishnamurthy, A., 2010. Intermediary asset pricing. Working Paper.
Hecht, P. A., 2000. The cross section of expected rm (not equity) returns. AFA
2001 New Orleans; HBS Finance Working Paper No. 03-044. Available at SSRN:
http://ssrn.com/abstract=235108 or doi:10.2139/ssrn.235108.
Hoberg, G., Phillips, G., 2010. Dynamic product-based industry classications and endogenous product dierentiation. Working Paper.
Jagannathan, R., Wang, Y., 2007. Lazy investors, discretionary consumption, and the crosssection of stock returns. The Journal of Finance 62, 16231661.
Jagannathan, R., Wang, Z., 1996. The conditional capm and the cross-section of expected
returns. The Journal of Finance 51, 353.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: Implications
for stock market eciency. Journal of Finance 48, 6591.
Kiyotaki, N., Moore, J., 1997. Credit cycles. Journal of Political Economy 105, 211248.
Kovenock, D., Phillips, G., 1997. Capital structure and product market behavior. Review of
Financial Studies 10, 767803.
Lettau, M., Ludvigson, S., 2009. Handbook of Financial Econometrics,. Elsevier, Ch. Measuring and Modeling Variation in the Risk-Return Tradeo.
Liew, J., Vassalou, M., 2000. Can book-to-market, size and momentum be risk factors that
predict economic growth? Journal of Financial Economics 57, 221245.
Lintner, J., 1965. The valuation of risk assets and the selection of risky investments in stock
portfolios and capital budgets. Review of Economics and Statistics 47, 1337.
Lucas, R. E. J., 1978. Asset prices in an exchange economy. Econometrica 46, 14291445.
Ludvigson, S. C., Ng, S., 2007. The empirical risk-return relation: a factor analysis approach.
Journal of Financial Economics 83, 171222.
33
Lustig, H., Verdelhan, A., 2010. Business cycle variation in the risk-return trade-o. Working
paper. Available at SSRN: http://ssrn.com/abstract=1670715.
MacKay, P., Phillips, G. M., 2005. How does industry aect rm nancial structure? The
Review of Financial Studies 18, 14331466.
Melicher, R. W., 1974. Financial factors which inuence beta variations within an homogeneous industry environment. Journal of Financial and Quantitative Analysis 9, 231241.
Miao, J., 2005. Optimal capital structure and industry dynamics. Journal of Finance 60,
26212659.
Mohanram, P. S., 2005. Separating winners from losers among low book-to-market stocks
using nancial statement analysis. Review of Accounting Studies 10, 133170.
Money, C., 2006. Stocks that can ride a pause. Cable News Network. A Time Warner Company.
Morningstar, 2004. Morningstar equity research methodology. Morningstar Research Report.
URL http://quicktake.morningstar.com/err/erm/researchmethod.pdf
Novy-Marx, R., 2010. The other side of value: Good growth and the gross protability
premium. Working Paper.
Ohlson, J., 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of
Accounting Research 18, 109131.
Peress, J., 2010. Product market competition, insider trading and stock market eciency.
The Journal of Finance 65, 143.
Perez-Quiros, G., Timmermann, A., 2000. Firm size and cyclical variations in stock returns.
The Journal of Finance 55, 12291262.
Perez-Quiros, G., Timmermann, A., 2001. Business cycle asymmetries in stock returns: Evidence from higher order moments and conditional densities. Journal of Econometrics 103,
259306.
Petersen, M. A., Rajan, R. G., 1994. The benets of lending relationships: Evidence from
small business data. The Journal of Finance 49, 337.
Petersen, M. A., Rajan, R. G., 1997. Trade credit: Theories and evidence. Review of Financial Studies 10, 661691.
Piotroski, J., 2000. Value investing: The use of historical nancial statement information to
separate winners from losers. Journal of Accounting Research 38, 141.
Platt, H. D., Platt, M. B., 1990. Development of a class of stable predictive variables: the
case of bankruptcy prediction. Journal of Business Finance and Accounting 17, 3151.
34
Platt, H. D., Platt, M. B., 1991. A note on the use of industry-relative ratios in bankruptcy
prediction. Journal of Banking and Finance 15, 11831194.
Rauh, J., Su, A., 2010. Explaining corporate capital structure: Product markets, leases, and
asset similarity. Working Paper, Available at SSRN: http://ssrn.com/abstract=1592463.
Reuters, 2008. Goldman sachs sees recession in 2008. Thomson Reuters.
Rosenberg, B., 1974. Extra-market components of covariance in security returns. Journal of
Financial and Quantitative Analysis 9, 263274.
Rosenberg, B., Marathe, V., 1975. The prediction of investment risk: Systematic and residual
risk. In: Proceedings of the Seminar on the Analysis of Security Prices. Center for Research
on Security Prices, University of Chicago, pp. 85226.
Rosenberg, B., McKibben, W., 1973. The prediction of systematic and specic risk in common stocks. Journal of Financial and Quantitative Analysis 8, 317333.
Salsman, R. M., 1997. Using market prices to guide sector rotation. AIMR Conference Proceedings, Economic Analysis for Investment Professionals, 4857.
Scheurle, P., Spremann, K., 2010. Size, book-to-market, and momentum during the business
cycle. Review of Managerial Science 4, 201215.
Schwert, G., 1989. Why does stock market volatility change over time? The Journal of
Finance 44, 11151153.
Sharpe, W. F., 1964. Capital asset prices: A theory of market equilibrium under condi- tions
of risk. Journal of Finance 19(3), 425442.
Shumway, T., 1997. The delisting bias in crsp data. Journal of Finance 52, 327340.
Shumway, T., 2001. Forecasting bankruptcy more accurately: A simple hazard model. Journal of Business 74, 101124.
Soliman, M., 2004. Using industry-adjusted dupont analysis to predict future protability
and returns. Ph.D. thesis, University of Michigan.
Stangl, J., Jacobsen, B., Visaltanachoti, N., 2009. Sector rotation over business cycles. Working Paper. Available at SSRN: http://ssrn.com/abstract=1467457.
Thompson, D. J., 1976. Sources of systematic risk in common stocks. The Journal of Business
49, 173188.
Titman, S., Wei, K. C., Xie, F., 2004. Capital investments and stock returns. Journal of
Financial and Quantitative Analysis 39, 677700.
van Binsbergen, J., 2007. Good-specic habit formation and the cross-section of
expected returns. AFA 2009 San Francisco Meetings Paper. Available at SSRN:
http://ssrn.com/abstract=1101456.
35
Yogo, M., 2006. A consumption-based explanation of expected stock returns. The Journal
of Finance 61, 539580.
Zhang, L., 2005. The value premium. The Journal of Finance 60, 67 103.
36
Figure 1: Recession Score Portfolio Average Returns on Portfolio Recession Premiums.
This gure plots the average monthly excess returns (over the risk free rate, Y-axis) of the 10 recession score
portfolios against their recession premiums (X-axis, average returns during recessions minus average returns
during expansions). A higher recession premium means that the portfolio has lower return shortfalls during
recessions, indicating that the portfolio has relatively low exposure to recession risk.
The recession score is an aggregate of information contained in seven nancial ratios. It is calculated
as: Recession Score=overall and within-industry-rank(liquidity)/2 - within-industry-rank(leverage) - withinindustry-rank(volatility of gross protability) + within-industry-rank(size) + overall-rank(prot margin) +
overall and within-industry-rank(gross protability)/2 - overall-rank(receivable turnover). For each nancial
ratio, all rms are assigned a rank between 1 (in the lowest decile) and 10 (in the highest decile). All
rms are sorted into 10 deciles in September of year t based on their recession scores (formed with nancial
statements in scal year t-1 ). The 10 decile portfolios are held from October of year t to September of year
t+1 and rebalanced annually.
37
Table 1: Summary Statistics of Financial Ratios Used
This table provides summary statistics of nancial ratios capturing four risk-relevant aspects of
rm fundamentals: (1) nancial leverage (accounting liquidity and leverage ratio), (2) cyclicality
(σ (gross protability)), (3) size (market capitalization), and (4) availability and sustainability of
cash ows (net prot margin, gross protability, and receivable turnover ratio). The nancial ratios
are computed with annual COMPUSTAT nancial statement data as follows:
Accounting liquidity = current assets (COMPUSTAT annual item 4, ACT) / current liabilities
(item 5, LCT).
Leverage ratio=debt/book equity. Debt is computed as the sum of short term debt (item 34, DLC)
and long term debt (item 9, DLCC), and equity is book equity (item 216, SEQ).
σ (gross protability) is measured as the standard deviation of a rm's gross protability in the past
ve years.
Gross protability =(sales (item 12, SALE) - direct production costs (costs of goods sold, item 41,
COGS) )/ total book value of assets (item 6, AT) at the beginning of the year.
Net prot margin = as net income (EAIT, item 18, IB) / sales (item 12, SALE).
Receivable turnover ratio = accounts receivable (item 2, RECT) / sales (item 12, SALE).
Ratio
Accounting Liquidity
Leverage Ratio
σ (Gross Protability)
Size (Market Cap)
Net Prot Margin
Gross Protability
Receivable Turnover Ratio
Mean
2.84
0.73
0.09
1190433.09
-0.43
0.65
0.2
Min
0
-3505.21
0
49.5
-5678.5
-69.17
-13.97
First
Quartile
1.48
0.08
0.03
16775.1
0
0.27
0.11
Median
2.14
0.38
0.06
70748.81
0.03
0.43
0.16
Third
Quartile
3.13
0.85
0.11
365591.88
0.07
0.65
0.21
# of observations
1509406
131118
13241
517
Total rm-scal year-monthly return
Total Financial Statements
Total # of rms
Total Months
38
Max
1719.25
3096.64
32.09
513361997.2
1861
18501
597.08
Table 2: Recession Premium Spread Generated by the Individual Financial Ratios
This table shows the recession premiums (average returns during recessions minus average returns during
expansions) of the high-minus-low portfolios sorted on the individual nancial ratio. They measures the
recession premium spread in the high (top decile) portfolio and the low (bottom decile) portfolio sorted on
each nancial ratio. A positive recession premium spread means that the high portfolio has lower return
shortfalls during recessions than the low portfolio, indicating that the high portfolio has lower exposure to
recession risk than the low portfolio.
The high-minus-low portfolios are formed as follows: In September of year t , all annual nancial statements
with scal year end from June of year t-1 to May of year t (scal year t-1 ) are put together. For each
nancial ratio, all rms are ranked into 10 deciles according to the level of the ratio and also ranked within
each industry (10 being the highest decile and 1 being the lowest decile). Then I form an equally-weighted
portfolio for each decile based on each ratio. The portfolio that holds stocks in decile 10 is dened as the high
portfolio and the portfolio that holds stocks in decile 1 is dened as the low portfolio. The high-minus-low
portfolio holds stocks in the high portfolio and short stocks in the low portfolio.
For the time series of the high-minus-low portfolio monthly returns sorted on each nancial ratio, I run
the time series regressions of
rit = Intercepti + Recession P remiumi ∗ Recession Dummyt + εit ,
where rit is the high-minus-low portfolio return sorted on nancial ratio i in month t . The Recession Dummy
is a dummy variable which equals 1 if month t is in an NBER recession and 0 otherwise. The intercept
term captures the average return of the high-minus-low portfolio return during expansions. The Recession
Premium coecient captures the average return during recessions minus average returns during expansions
of the high-minus-low portfolios. It is a measure of the recession premium spread generated by the high and
low portfolios.
Sorted within all rms
Sorted within each industry
Intercept Recession Premium
Intercept Recession Premium
Spread
Spread
0.03%
0.85%
-0.05%
0.73%
-0.21
(1.99)**
(-0.38)
(2.17)**
leverage ratio
-0.25%
-0.28%
0.01%
-0.68%
( -1.63)
(-0.63)
-0.1
(-1.74)*
σ (gross protability)
0.03%
-0.22%
0.09%
-0.52%
-0.13
( -0.39)
-0.55
(-1.35)
size
-1.71%
0.92%
-1.73%
1.36%
(-5.19)
(0.99)
(-5.41)
(1.59)
net prot margin
-0.06%
0.16%
-0.13%
0.47%
( -0.21)
(0.2)
(-0.50)
(0.65)
gross protability
-0.27%
1.36%
-0.25%
0.68%
(-1.50)
(2.75)***
(-1.46)
(1.65)*
receivable turnover
-0.07%
-1.24%
-0.47%
-0.23%
(-0.36)
(-2.46)**
(-4.80)
(-0.90)
t -statistics adjusted for white standard errors are in parentheses, *** p<0.01, ** p<0.05, * p<0.1
High-minus-Low Portfolio
Sorted on
liquidity
39
40
avg
Portfolio
return
High-minus-Low portfolio
-0.41%
Low Recession Score (High Risk) 1.19%
2
1.15%
3
1.24%
4
1.11%
5
1.03%
6
1.01%
1.01%
7
8
1.00%
9
0.82%
High Recession Score (Low Risk) 0.79%
*** p<0.01, ** p<0.05, * p<0.1. t -statistics
All periods
(519 months)
Recession Premium
(avg recession ret avg expansion ret)
avg
avg
std dev
t
return std dev
t
return
t
6.08% (1.97)** -0.13% 5.56%
(-0.53)
1.72%
(2.40)**
11.41%
(-0.71)
0.86% 8.95% (2.19)** -2.09%
(-1.59)
10.47%
(-0.21)
0.93% 7.80% (2.72)*** -1.40%
(-1.17)
9.99%
(-0.41)
0.97% 7.31% (3.02)*** -1.69%
(-1.48)
9.53%
(-0.24)
0.89% 6.82% (2.97)*** -1.36% ( -1.25)
9.02%
(-0.23)
0.82% 6.53% (2.87)*** -1.26%
(-1.22)
9.22%
(-0.02)
0.85% 6.27% (3.07)*** -1.04%
(-0.99)
8.91%
(-0.20)
0.82% 6.15% (3.03)*** -1.21%
(-1.19)
8.30%
(-0.07)
0.83% 5.88% (3.20)*** -1.07% ( -1.13)
8.46%
(0)
0.69% 5.88% (2.66)*** -0.82%
(-0.85)
8.40%
(0.45)
0.73% 5.90% (2.82)*** -0.37%
(-0.38)
the recession premiums are adjusted for white standard errors.
NBER recessions
(83 months)
avg
std dev
t
return
5.42%
(-1.56)
1.31%
8.38% (2.97)*** -0.89%
7.17% (3.36)*** -0.25%
6.66% (3.89)*** -0.45%
6.15% (3.75)*** -0.25%
5.93% (3.60)*** -0.23%
5.53% (3.81)*** -0.03%
5.46% (3.86)*** -0.20%
5.29% (3.94)*** -0.07%
5.24% (3.25)*** 0.00%
5.31% (3.10)*** 0.42%
in parentheses. t -statistics for
NBER expansions
(436 months)
The monthly returns for the 10 portfolios are excess returns over the risk free rate. The monthly returns for the low risk high risk (high recession
score - low recession score) portfolio is not adjusted for the risk free rate since it is a long-short portfolio. To estimate the recession premium for
each portfolio, I run the time series regressions of rit = Intercepti + Recession P remiumi ∗ Recession Dummyt + εit , where Recession Dummyt is a
dummy variable which equals 1 if month t is in an NBER recession and 0 otherwise. The Recession Premium coecient captures the average return
during recessions minus average returns during expansions of the portfolios.
The portfolios are formed as follows: In September of year t , all annual nancial statements with scal year end from June of year t-1 to May
of year t (scal year t-1 ) are put together. For each nancial ratio, all rms are assigned a rank between 1 (in the lowest decile) and 10 (in the
highest decile). The recession score is an aggregate of information contained in seven nancial ratios. It is calculated as: Recession Score =overall and
within-industry-rank(liquidity)/2 - within-industry-rank(leverage) - within-industry-rank(volatility of gross protability) + within-industry-rank(size)
+ overall-rank(prot margin) + overall and within-industry-rank(gross protability)/2 - overall-rank(receivable turnover). All rms are sorted into
10 deciles in September of year t based on their recession scores. The 10 decile portfolios are held from October of year t+1 to September of year t+2
and rebalanced annually. The high-minus-low portfolio holds stocks with the highest (top decile, low risk) recession scores and shorts stocks with the
lowest (bottom decile, high risk) recession scores. Returns in each portfolio are equally weighed.
This table shows the average excess returns (over the risk free rate) during NBER expansions, recessions, and over all periods of the high-minus-low
and the 10 portfolios sorted on recession score, as well as their recession premiums (average returns during recessions minus average returns during
expansions). A higher recession premium means that the portfolio has lower return shortfalls during recessions, indicating that the portfolio has
relatively low exposure to recession risk.
Table 3: Monthly Excess Returns of the 10 Recession Score Portfolios during NBER Expansions, Recessions, and all periods.
Table 4: Risk Adjusted Returns and Betas (CAPM and Fama-French three-factor Model) of
the Recession Score Portfolios.
This table shows the intercept and betas (CAPM and FF3) of the high-minus-low and the 10 portfolios
sorted on recession score.
To analyze the risk adjusted returns of the 10 portfolios using the CAPM, I run time-series regressions of
rit = α0 + βmkt (Rmt − rft ) + it , where rit is the monthly excess return of portfolio i sorted on recession
score. To analyze the risk adjusted returns of the 10 portfolios using the Fama-French three-factor model, I
run time-series regressions of rit = α0 + βmkt (Rmt − rft ) + βSM B SM Bt + βHM L HM Lt + it , where rit is
the monthly excess return of portfolio i sorted on recession score.
The % change of beta from the high risk portfolio to the low risk portfolio is calculated as (beta of high risk
portfolio beta of low risk portfolio)/beta of high risk portfolio.
Portfolio
Low Recession Score
(High Risk)
2
Estimate
t
t
3
-statistic
Estimate
t
High Recession Score
(Low Risk)
High-minus-Low
-statistic
Estimate
t
9
-statistic
Estimate
t
8
-statistic
Estimate
t
7
-statistic
Estimate
t
6
-statistic
Estimate
t
5
-statistic
Estimate
t
4
-statistic
Estimate
-statistic
Estimate
t
-statistic
Estimate
t
-statistic
CAPM
Intercept
beta
0.29%
1.32
1.02
18.81
0.37%
1.28
1.73
22.78
0.43%
1.25
2.24
24.31
0.36%
1.22
2.2
26.46
0.30%
1.2
2.05
28.53
0.34%
1.17
2.51
29.05
0.31%
1.16
2.44
31.06
0.34%
1.13
2.9
35.27
0.19%
1.14
1.7
35.11
0.23%
1.16
2.1
43
-0.06%
-0.17
-0.24
-2.86
Fama-French Three-Factor Model
Intercept MKT beta SMB beta HML beta
0.00%
1.15
1.23
0.37
0
12.31
5.59
2.36
0.09%
1.15
1.07
0.38
0.57
16.79
6.4
3.31
0.14%
1.13
1.02
0.39
1.04
19.97
7.49
4.01
0.11%
1.1
0.92
0.34
1.05
21.4
7.26
4.05
0.10%
1.09
0.81
0.28
0.97
23.18
6.95
3.64
0.12%
1.08
0.78
0.3
1.34
29.45
8.95
4.87
0.12%
1.07
0.75
0.27
1.47
28.77
8.19
4.36
0.20%
1.04
0.66
0.18
2.67
26.81
6.53
2.84
0.09%
1.03
0.64
0.12
1.23
25.61
6.22
1.85
0.18%
1.05
0.54
0.01
2.45
27.01
5.35
0.23
0.19%
-0.1
-0.7
-0.36
0.81
-1.48
-5.19
-2.92
% change of beta from
the high risk portfolio
12.64%
8.88%
to the low risk portfolio
t -statistics are adjusted for white standard errors to account for heteroscedasticity.
41
56.57%
96.27%
Table 5: Risk Adjusted Returns and Betas (Consumption CAPM) of the Recession Score
Portfolios
This table shows the intercept and betas (consumption-based CAPM) of the high-minus-low and 10 portfolios sorted on recession
score.
To analyze the risk adjusted returns of the 10 portfolios using the CCAPM , I run time-series regressions of ri,t+j =
α0 + βcij (Ct+j /Ct ) + it+j , where ri,t+j is the excess return from month t to month t+j of the i th decile portfolio sorted
on recession score. Ct+j /Ct is the consumption growth from quarter t to quarter t+j . I estimate the CCAPM excess returns
and betas with consumption growth at two horizons: (1) quarterly consumption growth and (2) annual consumption growth
from the fourth quarter to the fourth quarter (Q4-Q4). I use quarterly portfolio returns corresponding to quarter consumption
growth, and annual returns corresponding to annual (Q4-Q4) consumption growth. The consumption growth are computed
using two consumption series: (1) real per capita consumption growth in nondurables and services and (2) real per capita
consumption growth in nondurables alone.
The % change of beta from the high risk portfolio to the low risk portfolio is calculated as (beta of high risk portfolio beta
of low risk portfolio)/beta of high risk portfolio.
Consumption growth in
Consumption growth
non-durables and services
in non-durables
C-CAPM with
C-CAPM with
C-CAPM with
C-CAPM with
quarterly
Q4-Q4
quarterly
Q4-Q4
consumption
consumption
consumption
consumption
growth
Portfolio
growth
growth
growth
Intercept
cbeta
Intercept
cbeta
Intercept
cbeta
Intercept
cbeta
Estimate
6.94
10.11%
2.92
2.14%
3.65
7.53%
6.25
(High Risk)
t -statistic
0.00%
0
1.71
0.59
0.43
1.2
1.5
0.84
1.83
2
Estimate
0.56%
5.72
9.27%
2.55
2.16%
3.49
6.52%
5.83
0.24
1.59
0.67
0.47
1.43
1.73
0.93
2.18
0.88%
5.2
10.06%
2.61
2.27%
3.39
7.33%
5.91
0.41
1.56
0.71
0.47
1.62
1.84
1.04
2.19
Estimate
0.66%
5.02
7.25%
3.11
2.01%
3.23
5.86%
5.61
0.34
1.67
0.62
0.7
1.53
1.9
0.94
2.27
Estimate
0.58%
4.63
6.48%
2.78
1.84%
2.94
5.19%
5.04
0.31
1.61
0.6
0.66
1.49
1.89
0.92
2.35
0.90%
4.06
6.82%
2.65
1.89%
2.91
5.38%
4.98
0.49
1.47
0.61
0.61
1.59
1.97
0.95
2.36
Estimate
0.57%
4.53
4.81%
3.47
1.77%
2.99
4.70%
5.17
0.32
1.7
0.46
0.87
1.52
2.08
0.87
2.5
Estimate
0.68%
4.31
4.83%
3.32
1.89%
2.62
5.48%
4.36
0.41
1.73
0.53
0.94
1.72
1.93
1.09
2.3
0.32%
4.17
3.05%
3.22
1.48%
2.56
3.60%
4.3
0.2
1.73
0.35
0.96
1.35
1.93
0.72
2.25
0.70%
3.61
4.02%
2.84
1.68%
2.29
4.29%
3.95
0.43
1.54
0.47
0.89
1.55
1.79
0.89
2.21
0.69%
-3.33
-6.10%
-0.08
-0.45%
-1.36
-3.24%
-2.3
0.45
-1.33
-0.63
-0.02
-0.42
-0.85
-0.6
-1.11
Low Recession Score
3
4
5
6
7
8
9
High Recession Score
t -statistic
Estimate
t -statistic
t -statistic
t -statistic
Estimate
t -statistic
t -statistic
t -statistic
Estimate
t -statistic
Estimate
(Low Risk)
t -statistic
High-minus-Low
Estimate
t -statistic
% change of beta from
the high risk portfolio
47.97%
2.71%
to the low risk portfolio
t -statistics are adjusted for white standard errors to account for heteroscedasticity.
42
37.29%
36.74%
43
1.19%
-0.09%
0.29%
High Risk (Low Recession Score)
High historical CAPM beta stocks
Low historical CAPM beta stocks
8.38%
5.31%
7.61%
6.81%
4.93%
5.97%
5.60%
0.95%
1.00%
0.61%
0.81%
0.90%
Small Stocks
Value Stocks
Growth Stocks
Durables
0.09%
Long term (10year) bonds
2.14%
0.22%
(0.83)
(42.98)***
(0.71)
(3.34)***
(2.82)***
(2.44)**
(4.79)***
(1.54)
(3.68)***
(3.79)***
(4.42)***
(3.27)***
(-0.25)
(3.61)***
(2.47)**
(0.54)
(3.27)***
( 2.66)***
(-1.29)
(1.33)
(-0.19)
0.63%
0.54%
0.30%
3.19%
0.34%
8.16%
9.96%
7.87%
7.54%
8.41%
6.41%
(1.80)*
(14.30)***
(0.33)
(-0.61)
(-0.75)
(1.26)
(-0.19)
(-0.33)
(0.67)
(-0.81)
(0)
(-0.91)
(-0.32)
(-0.15)
(-0.25)
(0.59)
(-0.26)
(0.04)
(-0.53)
(-1.46)
(-1.10)
(-0.71)
0.17%
0.46%
0.80%
0.57%
0.41%
(43.11)***
(1.67)*
2.34%
(1.89)*
( 3.00)***
(1.92)*
(1.63)
(3.71)***
(1.68)*
(2.85)***
0.24%
6.07%
6.78%
5.74%
5.02%
5.99%
(2.61)***
(3.56)***
(2.14)**
(-0.37)
(2.94)***
(1.90)*
(0.76)
(2.55)**
(2.25)**
(-1.39)
(0.31)
(-0.75)
(-0.53)
(2.19)**
(2.82)***
t
0.54%
0.09%
-0.60%
(1.49)
(2.36)**
(-0.64)
(-1.31)
(0.25)
( -1.41)
(-1.36)
(-1.32)
(0.06)
(-1.77)*
(-1.10)
(-1.73)*
0.88% (2.19)**
-1.48%
-1.27%
-1.16%
-1.26%
0.03%
-1.29%
-0.83%
-1.31%
(-1.16)
(-0.91)
(0.39)
(-1.13)
( -0.71)
-0.12% (-0.22)
-0.98%
-1.10%
0.21%
-1.13%
-0.92%
(-1.77)*
(-0.99)
(-1.59)
-0.12% (-0.11)
-1.41%
-1.52%
-2.09%
(-0.38)
t
1.72% (2.40)**
-0.37%
avg ret
avg expansion ret)
(avg recession ret -
0.41% 3.03% (3.06)*** 0.10%
0.82%
0.75%
4.44%
4.49%
4.62%
0.24% 3.25%
0.51%
0.70%
0.43%
5.40%
7.54%
-0.07% 4.03%
0.70%
0.63%
6.24%
8.31%
0.12% 3.70%
0.70%
0.82%
4.87%
9.56%
8.95%
5.90%
-0.40% 6.28%
0.07%
-0.33%
0.86%
0.73%
std dev
0.97% 3.48% (2.54)** 0.23% 2.76%
-0.67%
-0.65%
-0.16%
-0.31%
6.65%
6.67%
0.26% 3.55%
-0.57%
0.01%
-0.66%
7.39%
10.59%
-0.17% 4.69%
-0.12%
-0.29%
8.83%
11.31%
0.30% 4.63%
0.05%
-0.25%
6.67%
12.81%
-0.50% 8.08%
-1.11%
-1.61%
11.41%
(0.45)
avg ret
*** p<0.01, ** p<0.05, * p<0.1. t -statistics in parentheses. t -statistics for the recession premiums are adjusted for white standard errors.
0.45%
0.09% 2.59%
5.23%
-0.89%
8.40%
t
1.31% 6.08% (1.97)** -0.13% 5.56%
(2.97)***
(-1.56)
0.42%
(3.10)***
std dev
0.39% 2.91% (2.79)*** 0.50% 3.58%
4.37%
5.40%
3.93%
0.24% 3.20%
t-bill
Non-durables- Durables
Non-Durables
HML
SMB
0.72%
Big Stocks
4.10%
Momentum
3.94%
0.64%
0.84%
Market Excess Return
High-Low Industry consumption beta stocks -0.05% 3.89%
0.81%
0.85%
Low historical Industry consumption beta stocks
5.60%
0.09% 3.50%
High historical Industry consumption beta stocks
High-Low consumption beta stocks
0.97%
0.88%
Low historical consumption beta stocks
-0.38% 5.89%
High historical consumption beta stocks
High-Low CAPM beta Stocks
4.42%
8.80%
Low-High Risk (High-Low Recession Score) -0.41% 5.42%
0.79%
Low risk (High Recession Score)
avg ret
(519 months)
t
(83 months)
std dev
(436 months)
avg ret
Portfolios Sorted on
All periods
NBER recessions
NBER expansions
Recession Premium
This table shows the average monthly excess returns (over the risk free rate) of various cyclical portfolios during NBER expansions, recessions, and over all periods, as well as
their recession premiums (average returns during recessions minus average returns during expansions). A higher recession premium means that the portfolio has lower return
shortfalls during recessions, indicating that the portfolio has relatively low exposure to recession risk.
The returns of the long-short portfolios are not adjusted for the risk free rate. Historical CAPM betas are computed each month using a rolling window of 1-year. The
10 decile portfolios sorted on historical CAPM betas are held for 1 month and rebalanced at the end of each month. Historical consumption betas are computed each
quarter with a rolling window of 15 years. The 10 decile portfolios sorted on historical consumption betas (individual and industry) are held for 3 months and rebalanced
at the end of each quarter. The monthly market returns, momentum returns, big, small, SMB, value, growth, HML returns are obtained from Ken French's data library
(http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html). To compute returns of durable and nondurable stocks, I use the durable/non-durable industry
classications from Gomes, Kogan, and Yogo (2009). I compute monthly equally-weighted excess returns of all stocks in the durable and non-durable categories.
Table 6: Average Monthly Excess Returns of the Comparison Portfolios
Table 7: Time Series Regressions and GRS Test Results
Panel A reports pricing errors (α) of the CAPM and the Fama-French (1993) three-factor model.
Test portfolios are the 10 recession score portfolios. Pricing errors are estimated by the time-series
regression
Ri,t = αi + βi ft + i,t ,
where Ri,t =excess return of portfolio i , ft =Rm,t (market excess return) for the CAPM, and ft
=[Rm,t ,SMB,HML] for the Fama-French three-factor model (FF3). Test portfolios are the 10
recession score portfolios. The monthly excess returns in the sample are from October 1966 to
December 2009.
Panel B reports the Gibbons, Ross, and Shanken (1989) test statistics and p -values using the 10
recession score portfolios. The GRS test is a statistical test of the hypothesis that the alphas of the
portfolios are jointly zero. The GRS test statistic is calculated as:
GRS =
i−1
T −N −K h
α̂0 Σ̂−1 α̂ ∼ FN,T −N −K ,
1 + ET (f )0 Ω̂−1 ET (f )
N
where T =the length of the time series, N =number of test portfolios, K =number of factors, Ω̂ is an
unbiased estimate of the factors' covariance matrix, α̂ is a Nx1 vector of estimated intercepts, and
Σ̂ is an unbiased estimate of the residual covariance matrix.
Panel A. Pricing Errors
Low Recession Score Portfolio
CAPM alpha 0.29% 0.37% 0.43% 0.36%
t-value
1.02
1.73
2.24
2.2
FF3 alpha 0.00% 0.09% 0.14% 0.11%
t-value
0
0.57
1.04
1.05
Panel B. GRS Test
GRS
p -value
CAPM
1.86
0.048
44
High Recession Score Portfolio
0.30% 0.34% 0.31% 0.34% 0.19% 0.23%
2.05
2.51
2.44
2.9
1.7
2.1
0.10% 0.12% 0.12% 0.20% 0.09% 0.18%
0.97
1.34
1.47
2.67
1.23
2.45
Fama-French Three-Factor Model
1.37
0.19
Table 8: Cross-Sectional Regression Test Results.
This table reports the Fama-MacBeth (1973) cross-sectional regression estimation results for the
consumption-based asset pricing model:
E [Ri,t ] = λ0 + λ1 βc .
Consumption betas are estimated by the time-series regression of excess returns on consumption
growth. Test portfolios are the 10 portfolios sorted on recession score. For quarterly consumption
growth, quarterly portfolio percentage returns are used. For the fourth quarter-to-fourth quarter
(Q4-Q4) consumption growth, annual portfolio percentage returns are used. Consumption growth
is computed using two consumption series: (1) real per capita consumption growth in nondurables
and services and (2) real per capita consumption growth in nondurables alone.
The estimation method is the Fama-MacBeth cross-sectional regression procedure. The rst row
reports the coecient estimates (λ̂). Fama-MacBeth t -statistics are reported in the second row.
Estimate
t-value
R2
adj. R2
Real per capita consumption growth
in nondurables and services
quarterly
Q4-Q4
consumption
consumption
growth
growth
λ̂0
λ̂1
λ̂0
λ̂1
1.38
0.31
21.36
-3.06
1.29
1.09
2.12
-1.55
67%
19.66%
63%
9.62%
45
Real per capita consumption growth
in nondurables
quarterly
Q4-Q4
consumption
consumption
growth
growth
λ̂0
λ̂1
λ̂0
λ̂1
0.52
0.78
-1.83
2.76
0.38
1.3
-0.28
1.44
87.76%
92.15%
86.23%
91.16%
Table 9: Regression Analysis
This table shows the pooled predictive regression of monthly excess stock returns (over the risk free rate)
on lagged ranks of nancial ratios and the interaction terms of nancial ratios times the recession dummy
Rect , which equals 1 if month t is in an NBER recession and 0 otherwise.
For nancial ratios computed using nancial statements with scal year ends in scal year t (June of
calendar year t to May of calendar year t+1 ), the respective returns in the regression are the monthly
returns from October of calendar year t+1 to September of calendar year t+1 . There is a four months lag
period to ensure that the nancial statement information has been made available to the public during the
corresponding return period.
Independent Variables
Rect *liquidity
Coecient
0.00101***
(5.271)
Rect *leverage
-0.000330*
(-1.754)
Rect *σ (gross protability) -0.000150
(-0.828)
Rect *size
0.00152***
(8.068)
Rect *net prot margin
-0.000719***
(-3.876)
Rect *gross protability
0.00111***
(5.741)
Rect *receivable turnover
-0.000501***
(-3.119)
liquidity
-0.000107
(-1.066)
leverage
-0.000259***
(-2.780)
std(gross protability)
-0.000326***
(-3.866)
size
-0.00810***
(-58.87)
net prot margin
-0.000735***
(-7.736)
gross protability
-0.000395***
(-3.491)
receivable turnover
-0.000470***
(-4.324)
recession
-0.00494**
(-2.174)
Observations
1,509,406
R-squared
0.026
Firm FE
YES
Year FE
YES
*** p<0.01, ** p<0.05, * p<0.1. t -statistics in parentheses
46
47
avg
Portfolio
return
High-minus-Low portfolio
-0.15%
Low Recession Score (High Risk) 0.26%
2
0.21%
3
0.28%
4
0.17%
5
0.14%
6
0.14%
7
0.13%
8
0.18%
9
0.07%
High Recession Score (Low Risk) 0.10%
*** p<0.01, ** p<0.05, * p<0.1. t -statistics
All periods
(519 months)
Recession Premium
(avg recession ret avg expansion ret)
avg
avg
std dev
t
return std dev
t
return
t
4.47%
(2.51)**
0.07%
4.12%
(0.36)
1.38% ( 2.62)***
3.93%
(-1.36)
0.12%
3.71%
(0.74)
-0.84%
(-1.81)*
2.74%
(-0.47)
0.15%
2.41%
(1.46)
-0.35%
( -1.09)
2.02%
(-1.01)
0.20%
1.93%
(2.35)**
-0.51% (-2.11)**
1.72%
(-0.37)
0.14%
1.44%
(2.14)**
-0.24%
(-1.22)
1.55%
(-0.17)
0.11%
1.31%
(1.92)*
-0.17%
(-0.92)
1.29%
(0.75)
0.14%
1.06% ( 2.94)*** -0.04%
(-0.23)
1.11%
(-0.33)
0.10%
0.94%
(2.48)**
-0.17%
(-1.32)
1.16%
(1.14)
0.17%
0.93%
(4.15)*** -0.03%
(-0.22)
1.36%
(1.77)*
0.10%
0.99%
(2.23)**
0.20%
(1.27)
1.43% (4.10)*** 0.19%
1.19%
(3.61)***
0.54%
(3.26)***
the recession premiums are adjusted for white standard errors.
NBER recessions
(83 months)
avg
std dev
t
return
4.02%
(-0.79)
1.23%
3.66%
(1.46)
-0.59%
2.34%
( 1.88)*
-0.14%
1.90% ( 3.08)*** -0.23%
1.38%
(2.64)*** -0.07%
1.26%
( 2.28)** -0.03%
1.01%
(2.94)***
0.11%
0.90%
(3.00)*** -0.04%
0.89%
(4.12)***
0.15%
0.90%
(1.51)
0.26%
1.12%
( 1.91)*
0.64%
in parentheses. t -statistics for
NBER expansions
(436 months)
This table shows the returns adjusted for size and book-to-market benchmarks (Daniel and Titman 1997, Daniel, Grinblatt, Titman, and Wermers
1997) during NBER expansions, recessions, and over all periods of the high-minus-low and 10 portfolios sorted on recession score, as well as their
recession premiums (average returns during recessions minus average returns during expansions). A higher recession premium means that the portfolio
has lower return shortfalls during recessions, indicating that the portfolio has relatively low exposure to recession risk.
The size and book-to-market benchmark returns are calculated as follows. In calculating the book-to-market ratios, the book equity used is from
any point in calendar year t-1 , and the market equity if from the last trading day of year t-1 . In calculating size, I use the market equity on the last
trading day of June of year t . All the stocks with sucient data are rst sorted into quintiles based on rm size, then based on book-to-market ratio.
The breakpoints for the size sort are the quintile breakpoints of the NYSE stocks. The size and book-to-market quintiles of the rms are then applied
to rms from July of year t through June of year t+1 . The characteristics benchmark returns are calculated as value-weighted returns within each
size and b/m quintile. Stock raw returns are adjusted for size and b/m characteristics by subtracting the benchmark returns of corresponding size
and b/m quintile.
To estimate the recession premium for each portfolio, I run the time series regressions of rit = Intercepti + Recession P remiumi ∗
Recession Dummyt + εit , where the Recession Dummyt is a dummy variable which equals 1 if month t is in an NBER recession and 0 otherwise. The
Recession Premium coecient captures the portfolio average return during recessions minus average return during expansions.
Table 10: Returns Adjusted for Size and Book-to-Market Characteristics Benchmarks of Portfolios Sorted on Recession Score
48
avg
Portfolio
return
High-minus-Low portfolio
-0.17%
Low Recession Score (High Risk) 0.32%
2
0.24%
3
0.30%
4
0.18%
5
0.15%
6
0.16%
7
0.13%
8
0.19%
9
0.10%
High Recession Score (Low Risk) 0.15%
*** p<0.01, ** p<0.05, * p<0.1. t -statistics
All periods
(519 months)
Recession Premium
(avg recession ret avg expansion ret)
avg
avg
std dev
t
return std dev
t
return
t
3.90% (2.94)*** 0.06%
3.60%
(0.37)
1.42%
(3.10)***
3.53%
(-1.73)*
0.16%
3.20%
(1.12)
-0.99% ( -2.38)**
2.50%
(-0.47)
0.18%
2.07%
(2.00)**
-0.37%
(-1.28)
1.85%
(-1.06)
0.21%
1.74% ( 2.80)*** -0.51% (-2.34)**
1.67%
(-0.35)
0.14%
1.30%
( 2.43)** -0.24%
(-1.26)
1.52%
(-0.16)
0.12%
1.22%
(2.25)**
-0.18%
(-1.00)
1.19%
(0.76)
0.15%
0.96%
(3.50)*** -0.06%
(-0.41)
1.06%
(-0.37)
0.11%
0.91%
(2.65)*** -0.18%
(-1.44)
1.08%
(1.31)
0.19%
0.87%
(4.86)*** -0.04%
(-0.28)
1.23%
(1.88)*
0.12%
0.93%
(2.96)***
0.16%
(1.12)
1.24% (4.32)*** 0.22%
1.05% ( 4.69)*** 0.44% ( 3.04)***
the recession premiums are adjusted for white standard errors.
NBER recessions
(83 months)
avg
std dev
t
return
3.50%
( -1.00)
1.26%
3.12%
(2.10)**
-0.67%
1.97%
(2.55)**
-0.13%
1.71% ( 3.61)*** -0.21%
1.22%
(3.04)*** -0.07%
1.16%
(2.69)*** -0.03%
0.91%
(3.59)***
0.10%
0.87%
(3.20)*** -0.04%
0.82%
(4.84)***
0.16%
0.86%
(2.32)**
0.25%
1.00%
(3.06)***
0.58%
in parentheses. t -statistics for
NBER expansions
(436 months)
This table shows the returns adjusted for size, book-to-market, and momentum benchmarks (Daniel and Titman 1997, Daniel, Grinblatt, Titman,
and Wermers 1997) during NBER expansions, recessions, and over all periods of the high-minus-low and 10 portfolios sorted on recession score, as
well as their recession premiums (average returns during recessions minus average returns during expansions). A higher recession premium means
that the portfolio has lower return shortfalls during recessions, indicating that the portfolio has relatively low exposure to recession risk.
The size, book-to-market, and momentum benchmark returns are calculated as follows. In calculating the book-to-market ratios, the book equity
used is from any point in calendar year t-1 , and the market equity if from the last trading day of year t-1 . In calculating size, I use the market
equity on the last trading day of June of year t . The past returns are calculated as the twelve-month return from the end of June in year t-1 to the
end of May in year t . All the stocks with sucient data are rst sorted into quintiles based on rm size, then based on book-to-market ratio, and
lastly on past return. The breakpoints for the size sort are the quintile breakpoints of the NYSE stocks. The size, book-to-market, and momentum
quintiles of the rms are then applied to rms from July of year t through June of year t+1 . The characteristics benchmark returns are calculated as
value-weighted returns within each size, b/m, and momentum quintile. Stock raw returns are adjusted for size, b/m, and momentum characteristics
by subtracting the benchmark returns of corresponding quintile.
To estimate the recession premium for each portfolio, I run the time series regressions of rit = Intercepti + Recession P remiumi ∗
Recession Dummyt + εit , where the Recession Dummyt is a dummy variable which equals 1 if month t is in an NBER recession and 0 otherwise. The
Recession Premium coecient captures the portfolio average return during recessions minus average return during expansions.
Table 11: Returns Adjusted for Size, Book-to-Market, and Momentum Characteristics Benchmarks of Portfolios Sorted on
Recession Score
49
All periods
NBER
Expansions
NBER
Recessions
to
to
to
to
to
to
to
to
1966/10
1970/12
1975/04
1980/08
1982/12
1991/04
2001/12
2009/07
1969/12
1973/11
1980/01
1981/07
1990/07
2001/03
2007/12
2009/12
1970/11
1975/03
1980/07
1982/11
1991/03
2001/11
2009/06
1966/10 to 2009/12
to
to
to
to
to
to
to
1970/01
1973/12
1980/02
1981/08
1990/08
2001/04
2008/01
Portfolios
1.32%
2.90%
-0.05%
3.69%
2.70%
0.33%
1.66%
1.21%
6.69%
1.19%
1.48%
0.40%
1.90%
2.48%
1.04%
1.20%
0.86%
4.14%
-0.13%
-1.42%
0.45%
-1.79%
-0.22%
0.71%
-0.46%
-0.36%
-2.56%
average monthly return
Low
High
Highrecession
recession
Low
score
score
recession
(high risk) (low risk)
score
-2.41%
-0.36%
2.05%
2.31%
0.66%
-1.65%
0.38%
2.07%
1.68%
-0.68%
1.96%
2.64%
1.43%
2.31%
0.88%
-0.11%
2.95%
3.06%
-2.34%
-0.72%
1.61%
cumulative returns
Low
High
Highrecession
recession
Low
score
score
recession
(high risk) (low risk)
score
-32.24%
-6.28%
25.96%
22.78%
4.35%
-18.43%
-0.63%
10.02%
10.64%
-13.89%
28.37%
42.26%
7.28%
15.61%
8.33%
-4.16%
19.46%
23.61%
-61.20%
-20.73%
40.47%
Total
98.16%
53.60%
-44.57%
-14.47%
7.49%
21.96%
193.04%
101.67%
-91.36%
30.73%
28.68%
-2.05%
9.22%
80.64%
71.42%
145.44%
124.72%
-20.72%
65.71%
53.68%
-12.02%
36.77%
23.65%
-13.11%
Total
482.53%
524.91%
42.38%
11
16
6
16
8
8
18
83
39
36
58
12
92
120
73
6
436
519
number
of
months
This table shows the average returns and cumulative returns of the low (bottom decile, high risk), high (top decile, low risk) and
high-minus-low portfolios sorted on recession score during each of the NBER recessions and expansions after 1965.
Table 12: High, Low, and High-Minus-Low Recession Score Portfolio Returns in Each of the 7 Recessions and 8 Expansions
after 1965.