Dissecting the Positive Relation between Cash Holdings and Stock Returns ☆ F.Y. Eric C. Lam Department of Finance and Decision Sciences Hong Kong Baptist University Kowloon Tong, Hong Kong Email: [email protected]Tel: (852)-3411-5218; Fax: (852)-3411-5585 Tai Ma Department of Finance National Sun Yat-sen University Kaohsiung, Taiwan Email: [email protected]Tel: (886)-7-525-2000 ext. 4810; Fax: (886)-7-525-1523 Shujing Wang Department of Finance The Hong Kong University of Science and Technology Clear Water Bay, Kowloon, Hong Kong Email: [email protected]Tel: (852)-2358-7666; Fax: (852)-2358-1749 K.C. John Wei * Department of Finance The Hong Kong University of Science and Technology Clear Water Bay, Kowloon, Hong Kong Email: [email protected]; Tel: (852)-2358-7676 Fax: (852)-2358-1749 This version: January, 2015 _______________________________ ☆ We appreciate the helpful comments from Alex Barinov, Jay Cao (FMA Asian discussant), Charles Clarke, John Crosby (AFBC discussant), Gerald Garvey, Gianluca Marcato (EFMA discussant), Clinton Newman, Anna-Leigh Stone (Eastern FA discussant), and conference participants at the 2014 European Financial Management Association (EFMA) Annual Meeting in Rome, the 2014 Financial
Transcript
Dissecting the Positive Relation between Cash Holdings and Stock Returns☆
F.Y. Eric C. LamDepartment of Finance and Decision Sciences
Hong Kong Baptist University Kowloon Tong, Hong Kong
Department of FinanceThe Hong Kong University of Science and Technology
Clear Water Bay, Kowloon, Hong KongEmail: [email protected]; Tel: (852)-2358-7676
Fax: (852)-2358-1749
This version: January, 2015
_______________________________☆We appreciate the helpful comments from Alex Barinov, Jay Cao (FMA Asian discussant), Charles Clarke, John Crosby (AFBC discussant), Gerald Garvey, Gianluca Marcato (EFMA discussant), Clinton Newman, Anna-Leigh Stone (Eastern FA discussant), and conference participants at the 2014 European Financial Management Association (EFMA) Annual Meeting in Rome, the 2014 Financial Management Association Asian (FMA Asian) Conference in Tokyo, the 2014 Eastern Finance Association (Eastern FA) Annual Meeting in Pittsburgh, and the 2013 Australasian Finance and Banking Conference (AFBC) in Sydney. The previous version of the paper carried the title “Cash Holdings and Stock Returns: Risk or Mispricing?” All errors are ours.
*Corresponding author: K.C. John Wei, Department of Finance, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong. Tel: (852)-2358-7676; Fax: (852)-2358-1749. Email: [email protected].
2012). This phenomenon is referred to as the cash holdings effect. Archarya, Davydenko, and
Strebulaev (2012) and Palazzo (2012) study rational models based on precautionary savings to
explain this positive relation.1 However, the effect still exists even after controlling for various
risk factors, such as the Fama and French (1993) three factors or the Hou, Xue, and Zhang
(2014) four factors motivated by the q-theory. As the literature has not yet examined any
alternative explanations based on the mispricing channel, our paper fills this gap and provides
new findings to understand the effect from a different angle.
We hypothesize that investors overreact to the salient agency problem of high cash holdings
but underreact to the implicit real illiquidity concern of low cash holdings. On one hand, the
agency theory of Jensen and Meckling (1976) and Jensen (1986) argues that self-interested
managers might overspend cash reserves for their own benefits at the expense of outside
shareholders. For example, Harford (1999) find that cash rich firms are more likely to bid for bad
targets in the takeover market. Titman, Wei, and Xue (2004) document that the negative relation
between capital investment and future stock returns is stronger for firms with greater agency
problems of free cash flow, suggesting that managers of high cash firms have a tendency to
overinvest. Harford, Mansi, and Maxwell (2008) further show that poorly governed firms spend
cash reserves quickly on value-destroying acquisitions and capital expenditures. If investors
excessively discount high cash-holdings firms due to overly emphasizing the salient agency
costs, firms with high cash holdings might be temporarily undervalued.
1 An opposite rational view under no agency problems is that since cash is more liquid and less risky as an asset, everything being equal, the expected return for firms with high cash holdings should be lower. See, for example, Ortiz-Molina and Phillips (2014).
1
On the other hand, low cash reserves could be detrimental to firms, especially for those
lacking liquidity. Bolton, Chen, and Wang (2011) show that corporate liquidity endogenously
arises in a model of dynamic investment for financially constrained firms. Bolton, Chen, and
Wang (2013) further show that when financing conditions are stochastic, firms optimally save
cash due to both market-timing and precautionary-savings motives. Harford, Klasa, and Maxwell
(2014) provide evidence that cash holdings can help firms mitigate refinancing risk. If investors
insufficiently discount low cash-holdings firms due to neglecting the implicit real illiquidity
costs, firms with low cash holdings might be temporarily overvalued. Taken together, the
misreactions would generate a positive relation between cash holdings and future stock returns.
We measure misvaluation by combining a battery of widely documented cross-sectional
stock return anomalies into an ex-ante signal to indicate the relative degree of overvaluation. Our
argument predicts that cash holdings should be negatively correlated with relative overvaluation.
High cash-holdings firms that are ex-ante identified as relatively undervalued should have higher
future return than low cash-holdings firms that are overvalued. Conversely, high cash-holdings
firms that are not undervalued should not subsequently outperform low cash-holdings firms that
are not overvalued. In addition, we expect the positive relation to vanish once ex-ante
misevaluation is properly controlled for. Another mispricing prediction is that the relation should
be stronger when arbitrage is limited (e.g., Shleifer and Vishny, 1997) and should disappear
when arbitrage is easy.
If investors overreact to the salient agency problem of high cash holdings, then the
undervaluation of high cash-holdings firms should be more severe for those with low leverage as
these firms tend to have stronger agency conflicts (e.g., Jensen, 1986). Hence we expect the
position relation to be stronger among firms with low leverage. Moreover, if investors underreact
2
to the implicit real illiquidity concern of low cash holdings, then the overvaluation of low cash-
holdings firms should be more severe for those that are unprofitable as these firms tend to have
poorer liquidity. Consequently we expect the position relation to be stronger among unprofitable
firms as well.
Using a sample of stock returns on U.S. listed firms from July 1960 to December 2011, we
find that future average size and book-to-market adjusted stock returns are significantly higher
for high cash-holdings firms than for low cash-holdings firms by 0.51% per month. Cash
holdings are negatively correlated with our measure of ex-ante relative overvaluation identified
with 11 cross-sectional stock return anomalies well documented in the literature (Stambaugh,
Yu, and Yuan, 2012; 2014). Such correlation indicates that higher (lower) cash holdings are
associated with a higher degree of undervaluation (overvaluation). Consistent with our
prediction, high cash-holdings firms that are ex-ante undervalued generate significantly higher
future stock returns than low cash-holdings firms that are overvalued. By contrast, high cash-
holdings firms that are not undervalued do not particularly outperform low cash-holdings firms
that are not overvalued. Furthermore, the positive relation becomes insignificant when ex-ante
misevaluation is controlled for in an interactive regression.
We construct a measure of limits to arbitrage based on three common dimensions of
arbitrage barriers. We do not find a relation between cash holdings and returns when arbitrage is
easy, while a significant positive relation emerges and becomes stronger when limits to arbitrage
are more serve. In addition, we find that the future stock returns of high cash-holdings firms with
low leverage are higher than those with high leverage and the position relation is also stronger
among firms with low leverage than among firms with high leverage. Moreover, we find that the
future stock returns of low cash-holdings firms that are unprofitable, indicated by non-positive
3
return on assets, are lower than those that are profitable and the position relation is also stronger
among unprofitable firms.
Palazzo (2012) argues that systematically riskier firms optimally hold more cash to hedge
against future cash flow shortfall to avoid costly external financing. In a related precautionary
savings study, Archarya, Davydenko, and Strebulaev (2012) predict that firms subject to higher
default risk optimally hold more cash to buffer against cash flow shortfall in the future. As such,
these riskier firms should be priced with higher expected stock returns. We examine the
exposures of firms with different levels of cash holdings to the five macroeconomic risks put
forth by Chen, Roll, and Ross (1986). We find that high cash-holdings firms indeed have
significantly higher exposure to the default risk factor than low cash-holdings firms. However,
high cash-holdings firms still deliver significantly higher risk-adjusted returns than low cash-
holding firms even after controlling for exposures to the five macroeconomic risks. Furthermore,
our findings based on relative overvaluation, limits to arbitrage, and leverage or profitability
classification hold with the returns adjusted for exposures to the macroeconomic risks.
We document that the positive relation drops substantially over time. Even with the
substantial increase in cash holdings for high cash-holdings firms, these firms no longer provide
significantly higher future returns than low cash-holdings firms in the most recent subperiod. It
seems that investors’ pricing of cash holdings has become much less systematically erroneous.
There is also a clustering of cash holdings across some industries.2 For example, Bates,
Kahle, and Stulz (2009) report that the average cash ratio of high-tech firms is greater than that
of manufacturing firms. We do find that many more firms in medical equipment, pharmaceutical
products, business services, computer hardware, and computer software industries are among the
high cash holders. In contrast, many more firms in utilities, petroleum and natural gas,
2 We thank Gerald Garvey and Clinton Newman for pointing this out.
4
communication, and wholesale industries are among low cash holders. After controlling for the
industry effect, we still find a positive relation between cash holdings and stock returns. We also
find that around half of the positive relation is due to the industry effect and another half is due
to the stock-level effect.
We provide the following major contribution to the literature. Previous studies have
provided rational explanations through the precautionary savings channel for the positive relation
between cash holdings and future stock returns. In this paper, we are the first to behaviorally
dissect the positive relation through the mispricing with limits to arbitrage channel. Specifically,
we argue that investors overreact to the salient agency problem of high cash holdings but
underreact to the implicit real illiquidity concern of low cash holdings. We find that the positive
relation is subsumed by ex-ante misvaluation identified from 11 well documented stock return
anomalies. The positive relation also disappears when limits to arbitrage are easy. The position
relation is also stronger among firms with low leverage or among unprofitable firms.
The paper proceeds as follows. Section 2 reviews the literature and develops our
hypotheses. Section 3 describes the variables and our sample. Section 4 presents the positive
relation between cash holdings and stock returns. Section 5 tests the role of ex-ante misvaluation
on the cash holdings effect. Section 6 tests the role of limits to arbitrage on the effect. Section 7
examines the effect across leverage and profitability. Section 8 tests the role of macroeconomic
risks on the effect. Section 9 conducts the robustness checks by examining the effect over time
and by controlling for the industry effect. Finally, Section 10 concludes the paper.
2. Literature Review and Hypothesis Development
5
Simutin (2010) and Palazzo (2012) find that high cash-holdings firms earn higher average
future stock returns than low cash-holdings ones. The positive relation cannot be explained by
the CAPM or the Fama and French (1993) three-factor model consisted of the market, size, and
value factors. Furthermore, the relation cannot be explained by the Hou, Xue, and Zhang (2014)
four-factor model motivated by the q-theory consisted of the market, size, investment, and
profitability factors either.
As we mention at the outset, we test whether investors’ misreactions to the implications of
cash holdings generate the positive relation between cash holdings and stock returns.
Specifically, investors might excessively discount high cash-holding firms due to overly
emphasizing the salient agency conflict associated with high cash holdings. Harford (1999),
Titman, Wei, and Xue (2004), and Harford, Mansi, and Maxwell (2008) show that, due to agency
problem, firms with high cash holdings spend cash reserves quickly on value-destroying
acquisitions and capital expenditures. Investors might be overly concerned that a substantial part
of the cash reserves of high cash-holding firms might be wasted one way or the other by self-
interested managers. Therefore investors undervalue firms with high cash holdings. In addition,
investors might insufficiently discount low cash-holding firms due to neglecting the implicit real
illiquidity concern of these firms. Bolton, Chen, and Wang (2011), Bolton, Chen, and Wang
(2013), and Harford, Klasa, and Maxwell (2014) argue that financially constrained firms are
subject to corporate liquidity hence cash holdings help mitigate refinancing risk and satisfy
market-timing and precautionary-savings motives. Investors might overlook the implicit
potential detriments to real liquidity associated with low cash holdings, therefore overvalue low
cash-holdings firms.
6
We measure ex-ante misvaluation by combining multiple cross-sectional stock return
anomaly variables into a measure of relative overvaluation. According to our misvaluation
argument, cash holdings should be negatively correlated with the measure. In particular, high
cash-holding firms have high future stock returns only if they are undervalued and low cash-
holding firms have low returns only if they are overvalued. The above discussion leads to our
first hypothesis.
H1a: The positive relation between cash holdings and expected stock returns should be strong
when high cash-holding firms are undervalued and low cash-holding firms are overvalued and
the relation should be weak otherwise.
H1b: The positive relation should disappear after controlling for ex-ante misvaluation.
If extreme cash holders are misvalued, the resulting profit opportunities would attract
arbitrage activities that correct the mispricing. When such opportunities are riskless and costless
to exploit, arbitrageurs should correct the mispricing quickly. However, when arbitrage is risky
and costly, the correction might slow down. De Long, Shleifer, Summers, and Waldmann (1990)
suggest that noise trading would cause prices to diverge from fundamental values, causing
arbitrage to be risky. Shleifer and Vishny (1997) argue that arbitrageurs are typically capital
constrained and might have to prematurely close arbitrage positions due to margin calls and
suffer significant losses. Liu and Longstaff (2004) show that even optimized trades can still
experience loss before prices converge when arbitrage is risky.
Pontiff (1996, 2006) shows that arbitrageurs prefer to hold fewer stocks with higher
idiosyncratic volatility. Arbitrageurs are typically under-diversified and hence the idiosyncratic
risk adds substantially to the total risk of their overall positions. The added risk should be of
great concern to arbitrageurs as there is still a debate whether they would be compensated with
7
higher expected returns.3 Transaction costs would be another barrier to arbitrage. Trading
expenses obviously reduce the profitability of arbitrage trades, which reduces their attractiveness
to arbitrageurs. Finally, a lack of liquidity might further make arbitrage opportunities technically
harder to exploit.
When arbitrage is riskier, transaction costs are higher, and/or liquidity is lower, arbitrage
opportunities provided by the misvaluation of extreme cash holders become more difficult and
hence less attractive to exploit. It follows that arbitrageurs would only sparingly correct the
mispricing. Hence our second hypothesis is as follows.
H2: The positive relation between cash holdings and stock returns should be stronger when
limits to arbitrage are more severe while the relation should be weaker when arbitrage is easier.
As we discussed at the outset, we hypothesize that investors overreact to the salient agency
problem due to high cash holdings. It follows that the undervaluation of high cash-holdings firms
should be more severe for those with low leverage as these firms tend to have stronger agency
conflicts. Thus our third hypothesis is as follows.
H3: The positive relation between cash holdings and stock returns should be stronger among
firms with low leverage.
Furthermore we hypothesize that investors underreact to the implicit real illiquidity concern
of low cash holdings. It follows that the overvaluation of low cash-holdings firms should be
more severe for those that are unprofitable as these firms tend to have poorer liquidity. Hence
our fourth hypothesis is as follows.
H4: The positive relation between cash holdings and stock returns should be stronger among
firms that are unprofitable.
3 For example, while Fu (2009) shows that stock returns are positively associated with idiosyncratic risk, Ang, Hodrick, Xing, and Zhang (2006, 2009) find the opposite.
8
Palazzo (2012) models the optimal corporate cash holdings policy when a firm’s cash flow is
correlated with a source of priced aggregate risk and faces costly external financing. As cash
savings allow the firm to avoid costly financing to fund the exercise of valuable growth options
in the future when the firm experiences a shortfall in cash flow, there is a hedging need of
precautionary savings. On one hand, riskier firms save more cash. On the other hand, riskier
firms are priced to provide higher expected returns. As cash holdings are positively correlated
with this systematic risk, it is positively correlated with expected returns.
In a related study, Archarya, Davydenko, and Strebulaev (2012) argue that a firm’s asset
composition, especially cash, depends on the liability structure. When the financially-constrained
firm faces higher default likelihood and expects lower future cash flow but higher returns, it
holds more cash to raise the liquidity of its assets to buffer the potential shortfall in cash flow in
the future. As the higher liquidity does not completely overcome the higher default risk, the
distressed firm remains more risky. This hedging need of precautionary savings predicts that
cash holdings are positively correlated with the default risk and therefore expected returns. The
above discussion leads to our final hypothesis.
H5: If the positive relation between cash holdings and stock returns is consistent with the risk-
based rational explanation, it should be explained by macroeconomic risks, especially the
default risk.
3. Variables and Sample Selection
This section summarizes the variables in our analysis and describes the sample data. Appendix
1 provides the detailed definitions of the variables. We measure a firm’s cash holdings (CHt) by its
cash-to-assets ratio, i.e., cash and short-term investments scaled by total assets at the end of
9
fiscal year t. It measures the proportion of total assets that the firm holds as cash and cash
equivalents. The higher the ratio, the more intensive the firm hoards cash but less of the asset
base is productive.
3.1. The measure of ex-ante misvaluation
Our measure of relative overvaluation (RO) is as follow. Like Stambaugh, Yu, and Yuan
(2012, 2014), we identify the dispersion of abnormal valuations in the cross section by
combining the following 11 stock return anomalies. (1) Total asset growth (TAG): Cooper,
Gulen, and Schill (2008) find that firms that increase their total assets have lower future
abnormal stock returns. They suggest that it is due to investors’ overreaction to total asset
expansions or contractions. (2) Accruals (Acc): Sloan (1996) documents that firms with higher
accruals have lower abnormal returns. He suggests that it is due to investors’ overestimation of
the persistence of the accruals component of earnings. (3) Net operating assets (NOA):
Hirshleifer, Hou, Teoh, and Zhang (2004) show that firms with higher net operating assets have
lower abnormal returns. They suggest that it is due to limited attention to accounting profitability
and to the neglecting of cash profitability. (4) Capital investment (I/A): Titman, Wei, and Xie
(2004) report that firms with higher capital investment have lower future abnormal returns. They
suggest that it is due to underreaction to the overinvestment by empire-building managers. (5)
External financing (XF): Bradshaw, Richardson, Sloan (2006) document that firms that increase
their overall external funding earn lower abnormal returns. They suggest that managers time the
market by opportunistically issuing overvalued securities and retiring undervalued securities.
(6) Net share issuance (NSI): Daniel and Titman (2006) find that firms that issue more
shares have lower abnormal returns. They suggest that managers tend to issue (retire) shares in
10
response to favorable (unfavorable) intangible information, which might reflect overvaluation
(undervaluation). (7) Financial distress (O): Dichev (1998) documents that firms with higher
bankruptcy risk earn lower abnormal returns. We use Ohlson’s (1980) bankruptcy risk score (O)
to measure financial distress.4 (8) Gross profitability (GP): Novy-Marx (2013) shows that firms
with higher gross profitability earn higher abnormal returns. (9) Earnings on assets (ROA): Fama
and French (2006) document that firms with higher earnings earn higher abnormal returns.5 (10)
Piotroski and So’s (2012) misvaluation score (MSCORE): The MSCORE is based on comparing
the market-to-book equity ratio (M/B) to Piotroski’s (2000) financial strength score (F). Piotroski
and So (2012) document that firms with higher book-to-market equity ratios but stronger
fundamentals have higher abnormal returns, while firms with higher book-to-market equity ratios
but weaker fundamentals do not. They suggest that it is due to biased expectations. (11)
Momentum (PRet): PRet is measured as the previous year’s cumulative stock return. Jegadeesh
and Titman (1993, 2001) show that firms with higher cumulative stock returns in the previous
one year have higher abnormal returns. They suggest that it is due to misreaction to firm
information.6
While each anomaly serves as a proxy for abnormal valuation itself, we combine them in
order to diversify away the measurement error in each individual effect and produce a more
precise measure. The combination also provides a comprehensive measure that summarizes
abnormal valuation due to the various behavioral reasons described. We independently sort
4 Campbell, Hilscher, and Szilagyi (2008) further show that the effect is stronger among firms with more informational frictions and suggest that it is due to misvaluation.5 Lam, Wang, and Wei (2014) find that these GP and ROA effects exist, only when market valuations are inconsistent with the profitability and during high sentiment periods. They suggest that the effects are due to expectation errors.6 Barber, Shleifer, and Vishny (1998) theorize that is due to conservatism and slow updating in face of new evidence. Daniel, Hirshleifer, and Subrahmanyam (1998) propose that it is also due to overconfidence and self-attribution bias. Chui, Titman, and Wei (2010) find that the momentum profits are higher in countries with a higher individualism index, which is consistent with the prediction of the momentum behavioral model of Daniel, Hirshleifer, and Subrahmanyam (1998).
11
stocks into terciles based on each of the above 11 variables reflecting the anomalies. The first 10
variables are measured at the end of fiscal year t and the last one is measured at the end of June
of calendar year t+1. For each of the sorts we assign a tercile rank to each stock such that the
highest rank is associated with the lowest future abnormal stock return, i.e., the highest degree of
relative overvaluation according to the given anomaly. Relative overvaluation (RO) is then the
simple average of these 11 rankings on each stock. Stocks with higher (lower) RO are associated
with higher (lower) relative overvaluation in the cross section and would earn lower (higher)
future abnormal stock returns.
3.2. The measure of limits to arbitrage
Like Pontiff (1996) and others, we use idiosyncratic stock return volatility (IVol) to measure
arbitrage risk.7 Our measure of transaction costs is the inverse of stock price (1/Price), which is
related to the bid-ask spread and brokerage commission (Bhardwaj and Brooks, 1992). Ball,
Kothari, and Shanken (1995) also use stock price as an inverse proxy for the bid-ask spread.
Furthermore, Stoll (2000) shows that recent stock prices are inversely related to the relative bid-
ask spread. Our measure of liquidity is the dollar trading volume (DVol), which is inversely
related to price pressure and time required to fill an order or to trade a large block of shares
(Bhushan, 1994). We consider stocks with higher arbitrage risk, higher transaction costs, and
lower liquidity to have more severe limits to arbitrage. We compute these variables at the end of
June of calendar year t+1. Again, we combine these three variables in order to diversify away the
measurement errors and produce a more precise measure as well as a cleaner presentation of the
results. We independently sort stocks into terciles based on each of these three variables. We
7 See, also, Wurgler and Zhuravskaya (2002), Ali, Hwang, and Trombley (2003), Mashruwala, Rajgopal, and Shevlin (2006), Duan, Hu, McLean (2010), McLean (2010), Lam and Wei (2011), and Lipson, Mortal, and Schill (2011).
12
then take the simple average of these three rankings on each stock to measure limits to arbitrage
(LTA).
3.3. Measure of leverage and identification of unprofitable and profitable firms
We measure leverage by the debt-to-asset ratio (D/A), which is long term debt scaled by the
book value of total assets. We identify firms with low (high) leverage as those with leverage
below (above) the cross-sectional median at the end of fiscal year t. We classify firms into
unprofitable ones versus profitable ones with return on assets (ROA). Specifically, we identify
unprofitable (profitable) firms as those that have non-positive (positive) ROA at the end of fiscal
year t.
3.4. Macroeconomic risks
We measure the exposures to macroeconomic risks in Chen, Roll, and Ross (1986) (hereafter
CRR) with the following time-series regression:
Ret p , t−Ret ft=α p+β p , MP Ret MP ,t+ βp , UI Ret UI ,t+β p , UTS Ret UTS, t
+β p , DEI Ret DEI , t+β p ,UPR RetURP ,t+ϵ p ,t , (1)
where Retp,t is the monthly return on a cash holdings portfolio p and Retft is the risk-free rate in
month t. RetMP, RetUI, RetUTS, RetDEI, and RetURP are the macroeconomic risk factors proposed by
CRR and are related to the growth rate of industrial production (MP), unexpected inflation (UI),
the term structure of interest rate (UTS), changes in expected inflation (DEI), and default risk
(URP), respectively. The slope coefficient (β) measures the exposure of a cash holdings portfolio
to a particular risk factor. The intercept (α) measures the average excess return on a cash
holdings portfolio after controlling for the macroeconomic risks.
13
The original factors are constructed as follows. MP is the growth rate of industrial production
and is defined as MPt = log(IPt/IPt–1), where IPt is the index of industrial production in month t.
MPt is led by a month to synchronize with the timing of the stock return. UI is unexpected
inflation and is the change in expected inflation (DEI) as calculated in CRR and is derived from
the total seasonally-adjusted consumer price index (CPI). UTS is the term premium and is the
yield spread between the long-term (10-year) and the one-year Treasury bonds. Finally, URP is
the default premium and is the yield spread between Moody’s Baa and Aaa bonds. All the
variables are from the Federal Reserve Bank of St. Louis.8
As the explanatory variables in equation (1) are returns while the factors MP, UI, and DEI
are not traded assets, we follow the standard asset pricing literature to employ mimicking
portfolios to track these factors. Although the factors UTS and URP are traded assets, as in Chan,
Karceski, and Lakonishok (1998) and Cooper and Priestley (2011), we also employ mimicking
portfolios to track them for consistency. In order words, we construct mimicking portfolios for
all the five CRR macroeconomic risks and use the returns on the mimicking portfolios to
estimate equation (1). The basis assets of the mimicking portfolios consists of 10 equal-weighted
portfolios, and 10 equal-weighted cash-holdings portfolios. The book-to-market, size, and
momentum portfolios are from French’s data library and we form our own cash holdings
portfolios based on the cash-to-assets ratio.
Following Lehmann and Modest (1988) and Cooper and Priestley (2011), we construct the
pure factor mimicking portfolios as follows. Firstly, we project the monthly returns (in the excess
of the risk-free rate) on each of the 40 basis assets on the five CRR factors. That is, we perform
40 time-series regressions to estimate a 40×5 matrix B of the slope coefficients on the five
8 These observed CRR factors are from Liu and Zhang (2008). We thank Laura Liu for sharing the updated data.
14
factors. Let V be the 40×40 covariance matrix of error terms for these regressions, which are
imposed to be orthogonal. It follows that the portfolio weights to track the five factors are given
by the 5×40 matrix w = (B’V-1B)-1B’V-1 and the mimicking portfolios are given by wR’, where R
is a T×40 matrix with each column containing the time-series returns on a basis asset over the
sample period. The product wR’ gives a 5×T matrix, in which each row represents the returns on
a mimicking portfolio for a CRR factor over the sample period. The mimicking portfolio
constructed this way for a specific factor has a sensitivity of one with respect to that factor and
zero sensitivity with respect to other factors. (Why do we need to use Laura’s data? This reduces
our sample period.)
3.5. Industry classification
Each year we classify firms into industries according to the 49 industry groupings in Fama
and French (1997). We obtain the updated groupings from Kenneth French’s data library. As
described later we exclude financial firms from our sample, and therefore groups 45 to 48 are
excluded. Our sample includes a total of 45 industries.
3.6. Sample selection
Our data contain firms traded on the NYSE, Amex, and Nasdaq exchanges. Their annual
financial statements and monthly stock data are from the Compustat and the Center for Research
in Security Prices (CRSP), respectively. Like Fama and French (1992, 1993), certificates,
American depositary receipts (ADRs), shares of beneficial interest (SBIs), unit trusts, closed-end
funds, real estate investment trusts (REITs), and financial firms are excluded. We delete firms on
which we do not have the information to compute the necessary variables in a year. Following
15
the literature, delisting returns are further used to mitigate the survivorship bias.9 The baseline
sample and the sample involving limits to arbitrage (LTA) covers annual firm characteristics
from fiscal year 1959 to year 2011, and monthly stock returns from the end of July of 1960 to the
end of December of 2012. The return data of the sample involving relative overvaluation (RO)
begins at the end of July 1973, when the data needed to calculate the Piotroski and So (2012)
misvaluation score (MSCORE) and external financing (XF) becomes more widely available. As
return on assets (ROA) does not have sufficient variation around zero until the fiscal year 1962,
the sample used to examine the cash holdings effect across profitable versus non-profitable firms
starts at the end of July 1963. As the factors are available till the end of November 2011, the
sample used to examine the macroeconomic risks also ends there.
4. The Baseline Results on the Positive Relation between Cash Holdings and Stock Returns
Panel A of Table 1 reports summary statistics of cash holdings (CH) in our baseline sample. The
average firm holds around 15% of its total assets as cash. The standard deviation is about 17% and
hence there is a meaningful variation in cash holdings in the cross section. While the 10th percentile
of cash holdings is only 2%, the 90th percentile of cash holdings is 39%. Consistent with the
previous literature (e.g., Simutin, 2010; Palazzo, 2012), Panel B of Table 1 shows that cash
holdings are negatively correlated with market capitalization (Size) at the end of June of calendar
year t+1 and the book-to-market equity ratio (B/M) at the end of fiscal year t.10 Therefore, when we
characterize the relation between cash holdings and future stock returns, we control for these
variables as previous studies (e.g., Fama and French, 1992) find that they are associated with
future stock returns.
9 See, for example, Shumway (1997) and Shumway and Warther (1999).10 Size is stock price multiplied by number of shares outstanding at the end of June of year t+1 while B/M is the Fama and French (1993) book value of equity divided by the market value of equity at the end of fiscal year t.
16
We first identify the relation between cash holdings and stock returns by forming decile
portfolios at the end of June every calendar year t+1 using cash holdings at the end of fiscal year
t.11 We measure risk-adjusted returns as follows. We first sort all available stocks into quintiles
based on their market capitalization (Size) and book-to-market equity (B/M) at the end of year t
independently. The intersection of these 55 portfolios forms our benchmark portfolios. We
measure the monthly risk-adjusted returns on a stock (aRet) between July of calendar year t+1 and
June of calendar year t+2 as the monthly raw stock returns minus the monthly returns on the
benchmark portfolio matched to the stock with the same ranks in Size and B/M. We then
compute the monthly risk-adjusted returns for each cash holdings decile and rebalance the
portfolios annually.
Panel C of Table 1 presents summary statistics for cash holdings deciles. There are about 328
firms in each decile per year in the baseline sample. Firms in decile 1 (low) hold around 1% of its
total assets in cash and those in decile 10 (high) hold around 52% of its total assets in cash. The
difference in cash holdings between these two extreme groups amounts to a substantial 51%. The
average risk-adjusted return on the lowest cash-holdings decile is –0.20% per month and is
significant at the 5% level. The return monotonically increases as cash holdings increase. The
average risk-adjusted return on the highest cash-holdings decile is 0.31% and is significant at the
5% level. High cash-holdings firms outperform low cash-holdings firms by 0.51% and the
difference is significant at the 1% level.
Second, we identify the relation between cash holdings and stock returns using the slope of
lagged cash holdings (b1) estimated from the following Fama and MacBeth (1973) cross-sectional
regression:
11 The portfolio approach allows us to address econometric issues such as overlapping observations, nonlinearities, and the “bad model issue” discussed in Fama (1998) and Mitchell and Stafford (2000).
17
Ret i , t+1=a+b1CH i ,t+b2 ln ¿ (2)
where Rett+1 is the monthly raw stock return from July of calendar year t+1 and June of calendar
year t+2 and Ln(Size) is the natural logarithm of market capitalization.12 Panel D of Table 1
presents the estimated slope coefficients. The slope of cash holdings (b1) is 0.692 and is significant
at the 5% level. Consistent with the previous studies, we identify a significantly positive relation
between cash holdings and subsequent stock returns.
5. Ex-ante Misvaluation and the Relation between Cash Holdings and Stock Returns
5.1. Summary statistics for the relative overvaluation measure
Panel A of Table 2 presents the descriptive statistics of relative overvaluation (RO) and the
constituent variables. The mean and median are similar at about 3.00 (it should be 2.00 since you
use tercile ranking?), suggesting that the measure is not specifically skewed. Panel B of Table 2
reports the correlations between cash holdings (CH) and RO as well as the constituent variables.
Cash holdings are negatively correlated with RO (correlation = –0.12), suggesting that firms with
higher cash holdings tend to have a higher degree of undervaluation. In addition, cash holdings
are negatively correlated with net operating assets (NOA) with a correlation of –0.38, suggesting
that high (low) cash-holdings firms are more operationally inactive (active).
5.2. The role of ex-ante misvaluation in the cash holding effect
For reference, Panel A of Table 3 reproduces Panel A of Table 2 using the sample excluding
firms with missing RO. Although the number of firms in each decile per year drops to 128 from
328 (it drops too much. Can we allow some of them missing? For example, firms with at least 7
anomalies available to compute the RO score will be included in the sample), cash holdings and
12 We take the natural logarithm of Size to reduce its skewness.
18
F.Y. Eric C. Lam, 01/12/15,
The variable RO is the average of 11 anomaly quintile rankings. A RO tercile ranking is used later in the portfolios and FM regressions.
stock returns on the portfolios remain similar. Firms in decile 1 (low) hold almost none of its total
assets in cash and those in decile 10 (high) hold around 46% of its total assets in cash. The
difference in cash holdings between these two extreme groups amounts to a substantial 46%. The
average risk-adjusted return on the low cash-holdings decile is –0.24% per month and is significant
at the 5% level. The return almost monotonically increases as cash holdings increase. The risk-
adjusted return on the high cash-holdings decile is 0.47% and is significant at the 5% level. High
cash-holdings firms outperform low cash-holdings firms by 0.71% and the difference is significant
at the 5% level. We again observe a significant positive relation between cash holdings and
subsequent stock returns in this subsample.
Panel B of Table 3 presents portfolios independently sorted by terciles of relative
overvaluation and deciles of cash holdings. By construction, the distribution of cash holdings
across cash holding deciles is very similar across RO. When RO is high, the average risk-adjusted
return on the highest cash-holdings decile is only 0.15% (t-stat = 1.09) per month and is
insignificant, suggesting that high cash-holdings firms do not necessarily earn higher returns. When
RO is medium, the risk-adjusted return on the highest cash-holdings decile increases to 0.48% per
month and is significant at the 1% level (please check the significance levels for the whole paper. In
the paper, we use three levels, 1%, 5%, and 10%). Furthermore, when RO is low, the risk-adjusted
return on the highest cash-holdings decile further rises to 0.51% and is significant at the 1% level (t-
stat = 4.20). High cash-holding firms have high future stock returns only if they are not
overvalued. Although the risk-adjusted return for the lowest cash-holding decile is significantly
negative at –0.47% (t-stat = –3.78) when RO is high, it turns positive albeit insignificant at 0.06%
when RO is low. These suggest that low cash-holdings firms do not necessarily earn lower returns.
Low cash-holding firms have low future stock returns only if they are overvalued.
19
The strategy of longing high cash-holdings firms with low RO and shorting low cash-
holdings firms with high RO ([low,10]–[high,1]) produces risk-adjusted return of 0.98% (t-stat =
4.50). By contrast, the strategy of longing high cash-holdings firms with high RO and shorting
low cash-holdings firms with low RO produces a nearly zero risk-adjusted return of 0.09% (t-stat
= 0.50). The findings indicate that the cash holdings effect strongly depends on ex-ante
misvaluation. Consistent with our hypothesis H1a, the positive relation between cash holdings and
stock returns is strong when high cash-holding firms are undervalued and low cash-holding firms
are overvalued but weak when high cash-holding firms are overvalued and low cash-holding
firms are undervalued.
We employ the Fama and MacBeth (1973) regression to further examine the dependence of
the cash holdings effect on relative overvaluation. Panel C of Table 3 reports the estimated slope
coefficients of the following cross-sectional regression:
Ret i , t+ 1=a+b1CH hii ,t +b2 CH loi ,t +b3 ROh ii ,t+b4 ROloi ,t
+b5CH hii ,t × RO loi ,t+b6CH loi ,t × ROhii ,t
+b7 ln¿ (3)
where CH_hi (CH_lo) is a dummy variable that equals one if the firm’s cash holdings are in the
top (bottom) tercile of cash holdings and zero otherwise. RO_hi (RO_lo) is a dummy variable
that equals one if the firm is in the top (bottom) tercile of relative overvaluation and zero
otherwise. Compared to Equation (2), the dummies in this specification set the cash holdings and
relative overvaluation variables in similar footings.
Model 1 of Panel C in Table 3 reports that the estimated slope on CH_hi (b1) is 0.193 and is
significant at the 10% level (t-stat = 1.66). The positive sign shows that firms with high cash
holdings provide higher average stock returns than the average firm. The estimated slope on
20
CH_lo (b2) is –0.157 and is significant at the 5% level (t-stat = –2.62). The negative sign
indicates that firms with low cash holdings provide lower average stock returns than the average
firm. These results are consistent with a positive relation between cash holdings and stock
returns. In Model 2 the estimated slope on RO_hi (b3) is –0.554 and is significant at the 1% level
(t-stat = –7.08). The negative sign shows that firms with high relative overvaluation provide
lower average stock returns than the average firms. The estimated slope on RO_lo (b4) is 0.225
and is significant at the 1% level (t-stat =4.21). The positive sign indicates firms with low
relative overvaluation provide higher average stock returns than the average firms.
Model 3 partially studies the interaction terms and the cash holding dummies. The estimated
slopes on CH_hi (b1) and CH_lo (b2) are 0.032 (t-stat = 0.19) and 0.088 (t-stat = 1.26),
respectively, and both are insignificant. The estimated slopes on CH_hi×RO_lo (b5) and
CH_lo×RO_hi (b6) are 0.395 (t-stat = 3.28) and -0.587 (t-stat = –7.47), respectively, and both are
significant at the 5% level. When relative overvaluation is low (RO_lo = 1), the slope on CH_hi
increases to 0.427 ( = 0.032+0.395), suggesting that high cash-holdings firms (CH_hi) with low
relative overvaluation earn a significantly higher return of 0.427% per month than the average
firm. However, when relative overvaluation is not low (RO_lo = 0), the slope on CH_hi reduces
to 0.032, suggesting that high cash-holdings firms (CH_hi = 1) without low relative
overvaluation do not earn a significantly higher return than the average firm. When relative
overvaluation is high (RO_hi = 1), the slope on CH_lo decreases to –0.499 ( = 0.088–0.587),
suggesting that low cash-holdings firms (CH_lo = 1) with high relative overvaluation earn a
significantly lower return of –0.499% per month than the average firm. However, when relative
overvaluation is not high (RO_hi = 0), the slope on CH_lo is 0.088, indicating that low cash-
holdings firms (CH_lo = 1) without high relative overvaluation do not earn a significantly lower
21
return than the average firm. The results suggest that the cash holding effect is strongly
influenced by ex-ante misvaluation as expected.
Model 4 partially examines the interaction terms and the relative overvaluation dummies.
The estimated slopes on RO_hi (b3) and RO_lo (b4) are –0.490 and 0.159, respectively, and both
are significant at the 5% level. The estimated slopes on CH_hi×RO_lo (b5) and CH_lo×RO_hi
(b6) are 0.123 and –0.136, respectively, and both are insignificant. When cash holdings are low
(CH_lo = 1), the slope on RO_hi decreases to –0.626 ( = –0.490–0.136), indicating that high
overvaluation firms (RO_hi = 1) with low cash holdings earn a significantly lower return of –
0.626% per month than average firms. Even when cash holdings are not low (CH_lo = 0), the
slope on RO_hi is –0.490, suggesting that high RO firms (RO_hi = 1) still earn a significantly
lower return than the average firm. When cash holdings are high (CH_hi = 1), the slope on
RO_lo increases to 0.282 ( = 0.159+0.123), indicating that low overvaluation firms (RO_lo = 1)
with high cash holdings earn a significantly higher return of 0.282% per month than the average
firm. Even when cash holdings are not high (CH_hi = 0), the slope on RO_lo is 0.159, indicating
that low relative overvaluation firms (RO_lo = 1) without high cash holdings still generates a
significantly higher return than the average firm.
In the full interactive model (Model 5), the estimated slopes on CH_hi (b1) and CH_lo (b2)
are 0.125 and –0.091, respectively, and both are insignificant. The estimated slopes on two
interaction terms CH_hi×RO_lo (b5) and CH_lo×RO_hi (b6) are –0.008 and 0.004, respectively,
and both are insignificant. The estimated slopes on RO_hi (b3) and RO_lo (b4) are –0.526 and
0.203, respectively, and both are significant at the 5% level. These results suggest that once ex-
ante misvaluation is controlled for, the cash holdings effect vanishes. Overall our findings are
consistent with our hypothesis H1b.
22
6. Limits to Arbitrage and the Relation between Cash Holdings and Stock Returns
6.1. Summary statistics for the limits to arbitrage measure
Panel A of Table 4 presents the descriptive statistics of limits to arbitrage (LTA) and the
constituent variables. The mean and median of LTA are similar at 2.00, suggesting that the
measure is not particularly skewed. Panel B of Table 4 reports the correlations between cash
holdings (CH) and limits to arbitrage as well as the LTA constituent variables. Cash holdings are
positively correlated with limits to arbitrage (correlation = 0.11) and the correlation seems to be
mainly driven by higher idiosyncratic volatility (IVol) and lower stock price (Price) of firms with
higher cash holdings.
6.2. The role of limits of arbitrage in the cash holding effect
For reference, Panel A of Table 5 reproduces Panel C of Table 1 using the sample excluding
firms with missing limits to arbitrage. Although the number of firms in each decile per year drops
to 219 from 328, cash holdings and risk-adjusted returns on the portfolios remain similar. In
particular, the average risk-adjusted return on the lowest cash-holdings decile is –0.12% per month
and is significant at the 10% level. The return is almost monotonically increasing as cash holdings
increases. The average risk-adjusted return on the highest cash-holdings decile is 0.44% and is
significant at the 5% level. High cash-holdings firms outperform low cash-holdings firms by 0.57%
and the difference is significant at the 5% level. Panel B of Table 5 shows that the slope of cash
holdings (b1) in Equation (2) is 0.713 and is significant at the 5% level. We also observe a
significant positive relation between cash holdings and subsequent stock returns in this subsample.
Panel C of Table 5 presents portfolios independently sorted by tercile of limits to arbitrage
and decile of cash holdings. By construction, the spread in cash holdings across limits to
23
arbitrage remains very similar. The results indicate that the cash holding effect is substantially
affected by limits to arbitrage. When limits to arbitrage are low, cash holdings is no longer
related to future stock returns. The average risk-adjusted returns on the lowest and highest cash-
holdings deciles are 0.03% and 0.11% per month, respectively, and both are insignificant. The
difference in the risk-adjusted return between the highest cash-holdings decile and the lowest cash-
holdings decile is merely 0.08% and is insignificant. When limits to arbitrage is medium, the risk-
adjusted return on the lowest cash-holdings decile is –0.20% and is insignificant while the risk-
adjusted return on the highest cash-holdings decile is 0.21% and is significant at the 10% level. The
difference in risk-adjusted return between the highest cash-holdings decile and the lowest cash-
holdings decile rises to 0.41% but it is still insignificant. When limits to arbitrage are high, the risk-
adjusted return on the lowest cash-holdings decile is –0.27% and is significant at the 5% level while
the risk-adjusted return on the highest cash-holdings decile is 0.88% and is also significant at the
5% level. High cash-holdings firms now outperform low cash-holdings firms by 1.15% and the
difference is significant at the 5% level. The difference in the return spread between the highest and
the lowest cash-holdings firms across high and low limits to arbitrage deciles ([high–low] of [10–1])
is 1.07% and is significant at the 5% level as well.
Panel D of Table 5 presents the estimated slope coefficients of Equation (2) across
subgroups sorted by the limits-to-arbitrage measure. When limits to arbitrage are low, the slope
of cash holdings (b1) is 0.537 and is insignificant. When limits to arbitrage are medium, the slope
decreases slightly to 0.495 and remains insignificant. When limits to arbitrage are high, the slope
increases substantially to 1.452 and is significant at the 5% level. The increase in the magnitude of
the slope ([high–low]) is 0.795 and is also significant at the 5% level. Both the portfolio and
regression approaches show that a relation between cash holdings and future stock returns does
24
not appear in the low limits-to-arbitrage environment. The magnitude of the positive relation gets
larger in the medium limits-to-arbitrage environment and becomes significant in the high limits-
to-arbitrage scenario. Overall, these findings are consistent with our second hypothesis H2.
7. The Cash Holding Effect: The Roles of Leverage and Profitability
7.1. The role of leverage
For reference, Panel A of Table 6 replicates Panel C of Table 1 using the sample having
non-missing debt-to-asset ratio. The average risk-adjusted return on the lowest cash-holdings
decile is –0.20% per month and is significant at the 1% level. The return is monotonically
increasing as cash holdings increases. The average risk-adjusted return on the highest cash-holdings
decile is 0.31% and is significant at the 5% level. High cash-holdings firms outperform low cash-
holdings firms by 0.51% per month and the difference is significant at the 1% level. Panel B of
Table 6 shows that the slope of cash holdings (b1) in Equation (2) is 1.199 and is significant at the
1% level. We again observe a significant positive relation between cash holdings and subsequent
stock returns in this subsample.
Panel C of Table 6 presents portfolios constructed by first sorting firms into low leverage
versus high leverage ones and then by decile of cash holdings. The spread in cash holdings is
much higher among low leverage firms (0.60) than that among high leverage firms (0.27).
Consistent with our agency cost argument, high cash-holdings firms having low leverage hold
substantially more cash (0.62) than high cash-holdings firms having high leverage (0.28). As a
result, the cash holdings effect is stronger among low leverage firms than among high leverage
firms as predicted. Among low leverage firms, the average risk-adjusted returns on the lowest and
the highest cash-holdings deciles are –0.19% and 0.36% per month, respectively, and both are
25
significant at the 5% level. The difference in risk-adjusted returns between the highest cash-
holdings decile and the lowest cash-holdings decile is 0.55% and is significant at the 1% level.
Among high leverage firms, the risk-adjusted return on the highest cash-holdings decile is only
0.07% and is insignificant while the risk-adjusted return on the lowest cash-holdings decile is –
0.19% and is significant at the 5% level. The return spread between high cash-holdings firms and
low cash-holdings firms drops to 0.26% and is significant at the 10% level.
The difference in the return spread between high and low cash-holdings firms across low
leverage and high leverage firms ([low–high] of [10–1]) is 0.29% and is significant at the 5% level.
In addition, the difference in the return between low cash-holdings firms having low leverage and
those having high leverage ([low,1]–[high–1]) is 0.29% and is significant at the 5% level. The
difference in the return between high cash-holdings firms having low leverage and those having
high leverage ([low,10]–[high–10]) is 0.00% and is insignificant. This is consistent with investors
overreacting to the salient agency costs associated high cash holdings and the undervaluation of
high cash-holdings firms is more severe for those with low leverage as these firms are perceived
to have stronger agency conflicts.
Panel D of Table 6 presents the estimated slope coefficients of Equation (2) across low
leverage and high leverage firms. Among low leverage firms, the slope on cash holdings (b1) is
0.672 and is significant at the 5% level. Among high leverage firms, the slope is 0.640 but is no
longer significant while the difference in the magnitude of the slope ([low–high]) is 0.032 and is
insignificant. This echoes that the weaker cash holding effect among high leverage firms is due to
weaker agency conflict and lower cash holding. Overall, these findings are consistent with our
third hypothesis H3.
26
7.2. The role of profitability
For reference, Panel A of Table 7 replicates Panel C of Table 1 using the sample starting
from the end of July 1963 with firms having non-missing return on asset. The average risk-
adjusted return on the lowest cash-holdings decile is –0.21% per month and is significant at the 5%
level. The return is almost monotonically increasing as cash holdings increases. The average risk-
adjusted return on the highest cash-holdings decile is 0.32% and is significant at the 5% level. High
cash-holdings firms outperform low cash-holdings firms by 0.53% per month and the difference is
significant at the 5% level. Panel B of Table 7 shows that the slope on cash holdings (b1) in
Equation (2) is 1.209 and is significant at the 1% level. We also observe a significant positive
relation between cash holdings and subsequent stock returns in this subsample.
Panel C of Table 7 presents portfolios constructed by first sorting firms into unprofitable
versus profitable ones then by decile of cash holdings. The spread in cash holdings is higher
among unprofitable firms (0.62) than that among profitable firms (0.43). The findings indicate
that the cash holdings effect is stronger among unprofitable firms than among profitable firms as
expected. Among unprofitable firms, the average risk-adjusted returns on the lowest and the
highest cash-holdings deciles are –0.49% and 0.42% per month, respectively, and both are
significant at the 5% level. The difference in risk-adjusted return between the highest cash-holdings
decile and the lowest cash-holdings decile is 0.91% and is significant at the 1% level. Among
profitable firms, the risk-adjusted return on the lowest cash-holdings decile is –0.12% and is
insignificant while the risk-adjusted return on the highest cash-holdings decile is 0.25% and is
significant at the 1% level. Although the return spread between high cash-holdings and low cash-
holdings firms remains significant at the 1% level, it drops to 0.37%. The difference in the return
27
spread between high and low cash-holdings firms across unprofitable and profitable firms
([unprofitable–profitable] of [10–1]) is 0.54% and is significant at the 5% level.
In addition, the difference in the return between low cash-holdings firms that are unprofitable
and those that are profitable ([unprofitable,1]–[profitable–1]) is –0.37% and is significant at the 5%
level. The difference in returns between high cash-holdings firms that are unprofitable and those
that are profitable ([unprofitable,10]–[profitable–10]) is 0.17% and is insignificant. This is
consistent with investors underreacting to the implicit real illiquidity concern associated low cash
holdings and the overvaluation of low cash-holdings firms is more severe for those that are
unprofitable as these firms have poorer liquidity.
Panel D of Table 7 presents the estimated slope coefficients of Equation (2) across
unprofitable and profitable firms. Among unprofitable firms, the slope of cash holdings (b1) is
1.750 and is significant at the 1% level. Among profitable firms, the slope is significant at the 5%
level but drops to 0.629. The difference in the magnitude of the slope ([unprofitable–profitable]) is
1.122 and is significant at the 10% level. Both the portfolio and regression analyses show that the
positive relation between cash holdings and stock returns is stronger among unprofitable firms
than among profitable firms. Overall, these findings are consistent with our fourth hypothesis
H4.
8. Macroeconomic Risks and the Relation between Cash Holdings and Stock Returns
8.1. Summary statistics for macroeconomic risks
For reference, Panel A of Table 8 reproduces Panel C of Table 1 using the sample with
returns ending at the end of November 2011. The number of firms in each decile per year remains
at around 328, cash holdings and risk-adjusted returns on the portfolios also remain similar. Firms in
28
decile 1 (low) hold around 1% of its total assets in cash and those in decile 10 (high) hold 51% more
of its total assets in cash. The average risk-adjusted return on the lowest cash-holdings decile is –
0.20% per month and is significant at the 5% level. The return is monotonically increasing as cash
holdings increase. The risk-adjusted return on the highest cash-holdings decile is 0.32% and is
significant at the 5% level. High cash-holdings firms outperform low cash-holdings firms by 0.52%
and the difference is also significant at the 5% level.
Panel B of Table 8 reports the average premiums on the five CRR macroeconomic risks or
essentially the average return on the mimicking portfolios tracking the five factors. Like Liu and
Zhang (2008) and Cooper and Priestley (2011), we find a significantly positive premium on the
industrial production factor (MP). Similar to Cooper and Priestley (2011), we also find
insignificant premiums on the unexpected inflation (UI) and the change in expected inflation
(DEI) factors, a significantly positive premium on the term structure (UTS) factor, and a
significantly negative premium on the default risk (URP) factor. The average premiums on the
MP, UTS, and URP factors are 1.22%, 1.10%, and –0.25% per month, respectively. The
magnitudes of these premiums are very close to those reported in Cooper and Priestley (2011).13
8.2. Macroeconomic risk exposures and the cash holdings effect
Panel C of Table 8 presents estimated parameters of Equation (1) across the cash holdings
deciles. The exposures of the lowest and highest cash-holdings deciles to the MP factor (βMP) are
–0.003 and 0.046, respectively. The difference in the exposure [10–1] is 0.049 and is
insignificant. It seems that the macroeconomic risk in terms of the MP factor is not able to
explain the relation between cash holdings and returns. For the UI factor, the exposures of the
13 The negative premium on the default factor is indeed consistent with the distress risk puzzle, i.e., firms with higher default risk have lower average future stock returns, documented by Dichev (1998), Campbell, Hilscher, and Szilagyi (2008), and many others.
29
lowest and highest cash-holdings deciles (βUI) are 0.337 and -0.010, respectively. The difference
in the exposure is –0.347 and is significant at the 10% level. For the UTS factor, the exposures of
the lowest and highest cash-holdings firms (βUTS) are 0.000 and –0.062, respectively. The
difference in the exposure is –0.062 and is significant at the 5% level. As high cash-holdings
firms are indeed less risky in terms of the UI and UTS factors, these dimensions of
macroeconomic risks do not seem to be able to explain the positive relation between cash
holdings and stock returns. For the DEI factor, the exposures of the lowest and highest cash-
holdings deciles (βDEI) are –1.304 and 1.061, respectively. The difference in the exposure is 2.365
and is significant at the 10% level. Although high cash-holdings firms are indeed riskier in terms
of the DEI factor, the premium on this factor is insignificant and hence this dimension of
macroeconomic risk does not seem to be able to explain the relation either.
Finally, for the URP factor, the exposures of the lowest and highest cash-holdings deciles
(βURP) are –0.335 and 0.454, respectively. The difference in the exposure is 0.789 and is
significant at the 5% level. High cash-holdings firms are indeed riskier in terms of the covariance
with the URP factor, which echoes the finding in Panel B of Table 2 that cash holdings and the
Ohlson (1980) bankruptcy risk score (O) are positively correlated (correlation = 0.14). The
hedging need of precautionary savings does seem to drive firms to hoard more cash to buffer
future bad states. However, the premium on the default factor is significantly negative and hence
neither this dimension of macroeconomic risk seems to be able to explain the relation. In fact, the
significant difference in the exposure to the default risk between the highest and lowest cash-
holdings deciles in fact worsens the cash holding effect. Furthermore, after controlling for all the
CRR macroeconomic risks, the average stock return on the lowest cash-holdings decile remains
negative at –0.28% per month and is significant at the 5% level. The return still monotonically
30
increases as cash holdings increase. The return on the highest cash-holdings decile is again
positive at 0.45% per month and is significant at the 5% level. The difference in the return
between the two extreme cash-holdings deciles is 0.73% and is also significant at the 5% level.
All these findings are inconsistent with our last hypothesis H5 and do not support the prediction
that high cash holdings carries a higher expected return due to higher exposures to
macroeconomic risks, especially the default risk.14
8.3. Further results on mispricing after controlling for macroeconomic risk
We examine whether the key results from Tables 3, 5, 6, and 7 survive after controlling for
the exposures to the CRR macroeconomic risks. Panel A of Table 9 reproduces Panel B of Table
3 using the estimated intercept of equation (1). The dependence of the cash holdings effect on ex-
ante misvaluation remains similar. The risk-adjusted return for the highest cash-holding decile is
insignificant at 0.53% (t-stat = 0.74) per month when relative overvaluation is high and it rises to
a significant 0.74% (t-stat = 4.63) when relative overvaluation is low. The risk-adjusted return
for the lowest cash-holding decile is insignificant –0.02% (t-stat = –0.21) when relative
overvaluation is low and drops to a significant –0.55% (t-stat = –5.16) when relative
overvaluation is high. The strategy of longing high cash-holdings firms with low relative
overvaluation and shorting low cash-holdings firms with high relative overvaluation ([low,10]–
[high,1]) produces the high risk-adjusted return of 1.29% (t-stat = 5.40). Conversely, the strategy 14 The finding is similar when we simultaneously control for the five macroeconomic risk factors and the Fama and French (1993) three factors. We also test two other potential rational explanations. First, we examine the explanation based on the q-theory of investment, i.e., whether high cash-holdings firms provides higher returns as they might have lower capital investments and/or higher true economic profitability. However, the slope of cash holdings (b1) in equation (2) remains significantly positive when we add to the set of control variables: the investment-to-capital ratio, total asset growth, growth in cash, or growth in noncash asset to proxy for capital investment as well as return on assets (ROA) and research-and-development intensity (R&D) to proxy for true economic profitability. Second, we examine whether high cash-holdings firms provide higher returns as they tend to increase future dividends, which conveys good news to the market. However, we find that high cash-holdings firms have higher returns in the subgroup in which firms do not raise dividends as well as the other subgroup. These untabulated results are available upon request.
31
of longing high cash-holdings firms with high relative overvaluation and shorting low cash-
holdings firms with low relative overvaluation produces an insignificant risk-adjusted return of
0.55% (t-stat = 0.77).
Panel B of Table 9 reproduces Panel C of Table 5. Similar as before, when limits to
arbitrage are low, cash holdings is no not related to future stock returns. The relation becomes
stronger when arbitrage is limited. The risk-adjusted return on the lowest cash-holdings decile is
0.09% per month and is insignificant while the return on the highest cash-holdings deciles is 0.29%
and is significant at the 5% level. The difference in the risk-adjusted return between the highest and
the lowest cash-holdings deciles is 0.10% and is insignificant. When limits to arbitrage is medium,
the risk-adjusted return on the lowest and highest cash-holdings deciles are –0.24% and is 0.27%
and both are significant at the 5% level. The difference in risk-adjusted return between the two
deciles rises to 0.51% and is significant at the 1% level. When limits to arbitrage are high, the risk-
adjusted return on the lowest cash-holdings decile is –0.65% and is significant at the 1% level while
the return on the highest cash-holdings decile is 0.77% and is significant at the 5% level. High cash-
holdings firms now outperform low cash-holdings firms by 1.42% and the difference is significant
at the 1% level. The difference in the return spread between high and low cash-holdings firms
across high and low limits to arbitrage deciles ([high–low] of [10–1]) is 1.32% and is significant at
the 1% level.
Panel C of Table 9 reproduces Panel C of Table 6. The cash holdings effect across low
leverage and high leverage firms remains similar. Among low leverage firms, the risk-adjusted
returns on the lowest and highest cash-holdings deciles are –0.30% and 0.47% per month,
respectively and are significant at the 1% level. The difference in the risk-adjusted return between
the highest cash-holdings decile and the lowest cash-holdings decile is 0.78% and is significant at
32
the 1% level. Among high leverage firms, the risk-adjusted return on the lowest cash-holdings
decile is –0.23% and is significant at the 1% level while the return on the highest cash-holdings
decile is 0.18% and is significant at the 5% level. The return spread between high cash-holdings and
low cash-holdings firms remains significant at the 1% level but it drops to 0.41%. The difference in
the return spread between high and low cash-holdings firms across low leverage and high leverage
firms ([low–high] of [10–1]) is 0.37% and is significant at the 5% level. Moreover, the difference in
the return between low cash-holdings firms having low leverage and those having high leverage
([low,1]–[high–1]) is –0.07% and is insignificant. The difference in the return between high cash-
holdings firms having low leverage and those having high leverage ([low,10]–[high–10]) is 0.30%
and is significant at the 5% level.
Panel D of Table 9 reproduces Panel C of Table 7. The cash holdings effect across
unprofitable and profitable firms remains similar. Among unprofitable firms, the risk-adjusted
return on the lowest cash-holdings decile is –0.87% per month and is significant at the 1% level
while the return on the highest cash-holdings decile is 0.25% and is insignificant. The difference in
the risk-adjusted return between the highest cash-holdings decile and the lowest cash-holdings
decile is 1.12% and is significant at the 1% level. Among profitable firms, the risk-adjusted return
on the lowest cash-holdings decile is –0.10% and is insignificant while the return on the highest
cash-holdings decile is 0.47% and is significant at the 1% level. The return spread between high
cash-holdings and low cash-holdings firms remains significant at the 1% level but it drops to 0.57%.
The difference in the return spread between high and low cash-holdings firms across unprofitable
and profitable firms ([unprofitable–profitable] of [10–1]) is 0.54% and is significant at the 5% level.
Furthermore, the difference in the return between low cash-holdings firms that are unprofitable and
those that are profitable ([unprofitable,1]–[profitable–1]) is –0.76% and is significant at the 1%
33
level. The difference in the return between high cash-holdings firms that are unprofitable and those
that are profitable ([unprofitable,10]–[profitable–10]) is –0.22% and is insignificant. Overall these
findings are again consistent with our mispricing hypotheses H1 to H5.
9. Robustness Checks
9.1. The relation between cash holdings and stock returns over time
Panel A of Table 10 reports the estimated slopes of Equation (2) across three subperiods of
our full sample. During the earlier fiscal years 1959 to 1976, the slope of cash holdings (b1) is
1.199 and is significant at the 5% level. During the fiscal years 1977 to 1994, the slope decreases
by almost 42% ((1.199–0.699)/1.199) to 0.699 but remains significant at the 5% level. During the
most recent fiscal years 1995 to 2011, the slope further drops by about 85% ((0.699–0.130)/0.699)
to 0.130 and is insignificant. The relation between cash holdings and future stock returns has
declined substantially by around 89% ((1.199–0.130)/1.199) over the three subperiods and there
seems to no longer have a relation between cash holdings and returns in the most recent
subperiod.
Panel B of Table 10 presents the cash-holdings decile portfolios across the three subperiods.
During the earlier fiscal years 1959 to 1976, the number of firms in each decile per year is about
191. Firms in decile 1 (low) hold around 2% while firms in decile 10 (high) hold around 30% of its
total assets in cash. The spread in cash holdings is 28% and is smaller than the spread of 51% in the
full period reported in Panel C of Table 1. During the fiscal years 1977 to 1994, the number of
firms in each decile per year rises to 408. Firms in decile 1 (low) hold almost none while firms in
decile 10 (high) hold about 52% of its total assets in cash. The spread in cash holdings rises to 51%.
During the recent fiscal years 1995 to 2011, the number of firms in each decile per year drops
34
slightly to 388. Firms in decile 1 (low) hold around 1% while firms in decile 10 (high) hold about
76% of its total assets in cash. The spread in cash holdings further increases to 75%. The substantial
rise in cash holdings of high cash-holdings stocks echoes the findings in Bates, Kahle, and Stulz
(2009) that listed U.S. industrial firms’ average cash-to-assets ratio increase substantially from
10% in 1980 to 23% in 2006.
Even with the massive rise in the spread of cash holdings, the strategy based on longing high
cash-holdings stocks and shorting low cash-holdings stocks does not generate monotonically more
reliable profits. During the early fiscal years 1959 to 1976, the average risk-adjusted return on the
lowest cash-holdings decile is –0.06% per month and is insignificant. The risk-adjusted return on
the highest cash-holdings decile is 0.24% and is significant at the 5% level. High cash-holdings
firms outperform low cash-holdings ones by 0.30% and the difference is significant at the 5% level.
During the fiscal years 1977 to 1994, the risk-adjusted return on the lowest cash-holdings decile is
–0.21% and is significant at the 5% level. The risk-adjusted return on the highest cash-holdings
decile is 0.47% and is significant at the 5% level. High cash-holdings firms outperform low cash-
holdings ones by 0.68% and the difference is also significant at the 5% level. As Figure 1 shows, the
higher average risk-adjusted return spread seems to be heavily influenced by the spike in the spread
during October 1987. Finally, during the fiscal years 1995 to 2011, the risk-adjusted return on the
lowest cash-holdings decile is –0.34% and is significant at the 10% level. The return on the highest
cash-holdings decile is 0.23% and is insignificant. The difference in the average risk-adjusted return
between the highest and lowest cash-holdings deciles is 0.56% but is insignificant due to a high
standard error of the estimate. High cash-holdings firms do not outperform low cash-holdings ones
unambiguously as in the past. Consistently Figure 1 shows that the monthly risk-adjusted return
spread becomes much more volatile during this period. Despite a few very high return spreads
35
between the end of 1999 and the beginning of 2000, the strategy suffers from numerous negative
returns afterwards. It seems that investors overvalue low cash-holdings firms to a much lesser
extent and no longer consistently undervalue high cash-holdings firms in the recent period. The
strategy might arguably still be economically profitable but is undoubtedly much riskier.
9.2. The relation between cash holdings and stock returns: The industry effect
As discussed at outset, we find that there is substantial clustering of cash holdings across
some industries. As a result, our results may be driven by the industry effect. To exclude this
possibility, we conduct a robust check by controlling for the industry effect. Table 11 reports the
results of the industry effect on the relation between cash holdings and stock returns. Panel A of
Table 11 presents the decile portfolios sorted by industry-adjusted cash holdings, which is a
firm’s cash holdings minus the average cash holdings of the industry to which the firm belongs.
Firms in decile 1 (low) hold about 3% of its total assets in cash and 17% less than its industry
average holdings. Firms in decile 10 (high) hold around 51% of its total assets in cash and 30%
more than its industry average holdings. The difference in industry-adjusted cash holdings or cash
holdings between these two extreme groups is 48%.
The average risk-adjusted return on the lowest industry-adjusted cash-holdings decile is –
0.02% per month and is insignificant. This is 90% ( = (0.20–0.02)/0.20) less than the risk-adjusted
return based on raw cash holdings reported in Panel C of Table 1. The return on the highest
industry-adjusted cash-holdings decile is 0.24% and is significant at the 5% level. This is about 23%
( = (0.31–0.24)/0.31) less than the return based on raw cash holdings. The spread in the risk-
adjusted return between the two extreme deciles is 0.26% and it is significant at the 5% level. This
is about 49% (= (0.51–0.26)/0.51) less than the spread based on raw cash holdings. After controlling
36
for the industry effect, we still observe a significant positive relation between cash holdings and
subsequent stock returns. Each of the stock-level effect and the industry effect contributes around
50% to the positive relation between raw cash holdings and returns.
For a robustness check, we replace cash holdings in Equation (2) with the industry-adjusted
cash holdings. Panel B of Table 11 reports that the slope coefficient on industry-adjusted cash
holdings is positive and significant at the 5% level. Panel C of Table 10 shows that the slope on
cash holdings remains significantly positive when we add industry dummies to the set of control
variables in Equation (2).
10. Conclusions
We dissect the positive relation between cash holdings and future stock returns through a
mispricing with limits to arbitrage channel, which has not yet been explored in the literature. We
find that the relation is tightly linked to ex-ante relative misvaluation, limits to arbitrage, as well as
to leverage or profitability classification but is not captured by the exposures to macroeconomic risk
factors. The relation weakens over time and high cash-holdings firms no longer precisely provide
higher returns than do low cash-holdings ones in the recent period. The industry and stock-level
effects are equally important in the relation.
37
References
Ali, Ashiq, Lee-Seok Hwang, and Mark A. Trombley, 2003. Arbitrage risk and the book-to-market anomaly. Journal of Financial Economics 69, 355-373.
Ang, Andrew, Robert J. Hodrick, Yuhang Xing, and Xiaoyan Zhang, 2006. The cross-section of volatility and expected returns. Journal of Finance 61, 259-299.
Ang, Andrew, Robert J. Hodrick, Yuhang Xing, and Xiaoyan Zhang, 2009. High idiosyncratic volatility and low returns: international and further U.S. evidence. Journal of Financial Economics, 91, 1-23.
Archarya, Viral, Sergei A. Davydenko, and Ilya A. Strebulaev, 2012. Cash holdings and credit risk. Review of Financial Studies, 25, 3572-3609.
Ball, Ray, S.P. Kothari, and Jay Shanken, 1995. Problems in measuring portfolio performance: an application to contrarian investment strategies. Journal of Financial Economics 38, 79-107.
Bates, Thomas W., Kathleen M. Kahle, and René M. Stulz, 2009. Why do U.S. firms hold so much more cash than they used to ? Journal of Finance 65. 1985-2021.
Bhardwaj, Ravinder K., and Leroy D. Brooks, 1992. The January anomaly: effects of low share price, transaction costs and bid-asked bias. Journal of Finance 47, 553-575.
Bhushan, Ravi, 1994. An informational efficiency perspective on the post-earnings announcement drift. Journal of Accounting and Economics 18, 45-66.
Bradshaw, Mark T., Scott A. Richardson, and Richard G. Sloan, 2006. The relation between corporate financing activities, analysts’ forecasts and stock returns. Journal of Accounting and Economics 42, 53-85.
Bolton, Patrick, Hui Chen Hui, and Neng Wang, 2011. A unified theory of Tobin’s q, corporate investment, financing, and risk management. Journal of Finance 66, 1545-1578.
Bolton, Patrick, Hui Chen, Neng Wang, 2013. Market timing, investment, and risk management. Journal of Financial Economics 109, 40-62.
Campbell, John Y., Jens Hilscher, and Jan Szilagyi, 2008. In search of distress risk. Journal of Finance 63, 2899-2939.
Chan., Louis K.C., Jason Karceski, and Josef Lakonishok, 1998. The risk and returns from factors. Journal of Financial and Quantitative Analysis 33, 159-188.
Chen, Nai-Fu, Richard Roll, and Stephen A. Ross, 1986. Economic forces and the stock market. Journal of Business 59, 383-403.
38
Chui, Andy C.W., Sheridan Titman, and K. C. John Wei, 2010, Individualism and momentum around the world, Journal of Finance 65, 361-392.
Cooper, Michael J., Huseyin Gulen, and Michael J. Schill, 2008. Asset growth and the cross-section of stock returns. Journal of Finance 63, 1609-1651.
Cooper, Ilan, and Richard Priestley, 2011. Real investment and risk dynamics Journal of Financial Economics 101. 182-205.
Daniel, Kent, David Hirshleifer, and Avanidhar Subrahmanyam, 1998. Investor psychology and security market unders- and overreactions. Journal of Finance 53, 1839-1885.
Daniel, Kent, and Sheridan Titman, 2006. Market reactions to tangible and intangible information. Journal of Finance 61, 1605-1643.
De Long, J. Bradford, Andrei Shleifer, Lawrence H. Summers, and Robert J. Waldmann, 1990. Noise trader risk in financial markets. Journal of Political Economy 98, 703-738.
Dichev, Ilia D., 1998. Is the risk of bankruptcy a systematic risk? Journal of Finance 53 1131-1147.
Duan, Ying, Gang Hu, and R. David McLean, 2010. Costly arbitrage and idiosyncratic risk: evidence from short sellers. Journal of Financial Intermediation 19, 564-579.
Fama, Eugene F., 1998. Market efficiency, long-term returns, and behavioral finance. Journal of Financial Economics 49, 283-306.
Fama, Eugene F., and Kenneth R. French, 1992. The cross-section of expected stock returns. Journal of Finance 47, 427-465.
Fama, Eugene F., and Kenneth R. French, 1993. Common risk factors in the returns and bonds. Journal of Financial Economics 33, 3-56.
Fama, Eugene F., and Kenneth R. French, 1997. Industry costs of equity. Journal of Financial Economics 43, 153-193.
Fama, Eugene F., and Kenneth R. French, 2006. Profitability, investment and average returns. Journal of Financial Economics 82, 491-518.
Fama, Eugene F., and James D. MacBeth, 1973. Risk, return and equilibrium: empirical tests. Journal of Political Economy 81, 607-636.
Fu, Fangjian, 2009. Idiosyncratic risk and the cross-section of expected stock return. Journal of Financial Economics, 91, 24-37.
Harford, Jarrad, 1999, Corporate cash reserves and acquisitions, Journal of Finance 54, 1969-
39
1997.
Harford, Jarrad, Sattar Mansi, and William Maxwell, 2008, Corporate governance and firm cash holdings in the US, Journal of Financial Economics 87, 535-555.
Harford, Jarrad, Sandy Klasa, and William F. Maxwell, 2014. Refinancing risk and cash holdings. Journal of Finance 69, 975-1012.
Hirshleifer, David, Kewei Hou, Siew Hong Teoh, and Yinglei Zhang, 2004. Do investors overvalue firms with bloated balance sheets? Journal of Accounting and Economics 38, 297-331.
Hou, Kewei, Chen Xue, and Lu Zhang, 2014. Digesting anomalies: an investment approach. Review of Financial Studies, forthcoming.
Jegadeesh, Narasimhan, and Sheridan Titman, 1993. Return to buying winners and selling losers: implications for stock market efficiency. Journal of Finance 48, 65-91.
Jegadeesh, Narasimhan, and Sheridan Titman, 2001. Profitability of momentum strategies: an evaluation of alternative explanations. Journal of Finance 56, 699-720.
Lam, F.Y. Eric. C., and K.C. John Wei, 2011. Limits-to-arbitrage, investment frictions, and the asset growth anomaly. Journal of Financial Economics 102, 127-149.
Lam, F.Y. Eric C., Shujing Wang, and K.C. John Wei, 2014. The profitability premium: macroeconomic risks or expectation errors? Working paper.
Lehmann, Bruce N., and David M., Modest, 1988. The empirical foundations of the arbitrage pricing theory. Journal of Financial Economics 21, 213-154.
Lipson, Marc L., Sandra Mortal, and Michael J. Schill, 2011. On the scope and drivers of the asset growth effect. Journal of Financial and Quantitative Analysis 46, 1651-1682.
Liu, Jun, and Francis A. Longstaff, 2004. Losing money on arbitrage: optimal dynamic portfolio choice in markets with arbitrage opportunities. Review of Financial Studies, 17, 611-641.
Liu, Xiaolei Laura, and Lu Zhang, 2009. Momentum profits, factor pricing, and macroeconomic risk. Review of Financial Studies 21, 2417-2448.
Mashruwala, Christina, Shivaram Rajgopal, and Terry Shevlin, 2006. Why is the accrual anomaly not arbitraged away? The role of idiosyncratic risk and transaction costs. Journal of Accounting and Economics 42, 3-33.
McLean, R. David 2010. Idiosyncratic risk, long-term reversal, and momentum. Journal of Financial and Quantitative Analysis, 45, 883-906.
40
Mitchell, Mark L., and Erik Stafford, 2000. Managerial decisions and long-term stock price performance. Journal of Business 73, 287-329.
Newey, Whitney K., and Kenneth D. West, 1987. A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55, 703-708.
Novy-Marx, Robert, 2013. The other side of value: the gross profitability premium. Journal of Financial Economics 108, 1-28.
Ohlson, James A., 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research 18, 109-131.
Oritz-Molina, Hernan, and Gordon M. Phillips, 2014, Real asset illiquidity and the cost of capital, Journal of Finance and Quantitative Analysis 48, 1-32.
Palazzo, Berardino, 2012. Cash holdings, risk, and expected returns. Journal of Financial Economics 104, 162-185.
Piotroski, Joseph D., 2000. Value investing: the use of historical financial statement information to separate winners from losers. Journal of Accounting Research 38, 1-41.
Piotroski, Joseph D., and Eric C. So, 2012. Identifying expectation errors in value/glamour strategies: a fundamental analysis approach. Review of Financial Studies 25, 2841-2875.
Pontiff, Jeffrey, 1996. Costly arbitrage: evidence from closed-end funds. Quarterly Journal of Economics 111, 1135-1152.
Pontiff, Jeffrey, 2006. Costly arbitrage and the myth of idiosyncratic risk. Journal of Accounting and Economics 42, 35-52.
Shleifer, Andrei, and Robert W. Vishny, 1997. The limits of arbitrage. Journal of Finance 52, 35-55.
Shumway, Tyler, 1997. The delisting bias in CRSP data. Journal of Finance 52, 327-340.
Shumway, Tyler, and Vincent A. Warther, 1999. The delisting bias in CRSP's Nasdaq data and its implications for the size effect. Journal of Finance 54, 2361-2379.
Sloan, Richard G., 1996. Do stock prices fully reflect information in accruals and cash flows about future earnings. Accounting Review 3, 289-315.
Stambaugh, Robert F., Jianfeng Yu, and Yu Yuan, 2012. The short of it: investor sentiment and anomalies. Journal of Financial Economics 104, 288-302.
41
Stambaugh, Robert F., Jianfeng Yu, and Yu Yuan, 2014. Arbitrage asymmetry and the idiosyncratic volatility puzzle. Journal of Finance, forthcoming.
Stoll, Hans R., 2000. Friction. Journal of Finance 55, 1479-1514.
Titman, Sheridan, K.C. John Wei, and Feixue Xie, 2004. Capital investments and stock returns. Journal of Financial and Quantitative Analysis 39, 677-700.
Wurgler, Jeffrey, and Ekaterina Zhuravskaya, 2002. Does arbitrage flatten demand curves for stocks? Journal of Business 75, 583-608.
42
Appendix 1Variable definitions
CH: Cash holdings or cash-to-assets ratio, calculated as cash and short-term investments (item CHE) scaled by total assets (item AT).at the end of a fiscal year. Data source: annual Compustat.
Size: Market capitalization, calculated as the closing stock price multiplied by the number of shares outstanding at the end of June of calendar year t+1. Data source: CRSP.
B/M: Book-to-market equity ratio, calculated as the book value of equity divided by the market value of equity (Size) at the end of fiscal year t. Book equity is total assets minus liabilities (item LT), plus balance sheet deferred taxes (item TXDB) and investment tax credits (item ITCI), minus preferred stock liquidation value (item PSTKL) if available, or redemption value (item PSTKRV) if available, or carrying value (item PSTK) if available. Data source: Compustat Annual and CRSP.
TAG: Growth in book value of total assets, calculated as the change in total assets (item AT) over a fiscal year scaled by beginning total assets. Data source: Compustat Annual.
Acc: Accounting accruals, calculated as the change in non-cash assets (item AT less item CHE) less the change in non-debt liabilities (item LT less item DLTT less item DLC) over a fiscal year scaled by beginning total assets. Data source: Compustat Annual.
NOA: Net operating assets, calculated as the change in operating assets and operating liabilities over a fiscal year scaled by beginning total assets. Operating assets is total assets minus cash and short-term investments (item CHE). Operating liabilities is total assets less current liabilities (item DLC), long-term debt (item DLTT), minority interests (item MIB), preferred stocks (item PSTK), and common equity (item CEQ). Data source: Compustat Annual.
I/A: Investment-to-assets ratio, calculated as the change in inventories (item INVT) and gross property, plant, and equipment (item PPEGT) over a fiscal year scaled by beginning total assets. Data source: Compustat Annual.
XF: Net cash flow from external financing, calculated as the sum of ∆E and ∆D. ∆E is net cash flow from equity financing, measured as the cash proceeds from sales of common and preferred stocks (COMPUSTAT item SCSTKC plus item SPSTKC) less the cash payments for purchases of common and preferred stocks (item PRSTKCC plus PRSTKPC) less cash payments for dividends (item CDVC) over a fiscal year scaled by beginning total assets. ∆D is net cash flow from debt financing, measured as the cash proceeds from issuance of long-term debt (Compustat item DLTIS) less the cash payments for long-term debt reductions (item DLTR) plus changes in current debt (item DLCCH, set to zero if it is missing) over a fiscal year scaled by beginning total assets.15 Data source: Compustat Annual.
15 Setting a missing value in item DLCCH to zero provides us with a much larger sample.
43
NSI: Net share issuance, calculated as the natural logarithm of the ratio of split-adjusted shares (item CSHO multiplied by item ADJEX_C) outstanding at the end of a fiscal year to that at the beginning of the year. Data source: Compustat Annual.
O: Bankruptcy risk score suggested by Ohlson (1980), which is calculated as
where Ln(A) is the natural logarithm of total assets, L is liabilities CA is current assets (item ACT), and CL is current liabilities (item LCT) at the end of a fiscal year. NI is net income (item NI) for the lagged fiscal year. Loss is equal to one if net income for both a fiscal year and the lagged fiscal year is negative and zero otherwise. NegBook is equal to one if L is greater than A and zero otherwise. ΔNI is the change in net income between a fiscal year and the lagged fiscal year scaled by the sum of the absolute values of the net income for the two years. Op, funds from operations, is defined as that in FSCORE. Data source: Compustat.
GP: Gross profitability, calculated as the gross profit (item GP) over a fiscal year scaled by beginning total assets. Data source: Compustat Annual.
ROA: Return on assets or earnings profitability, calculated as operating income before extraordinary items (item IB) over a fiscal year scaled by beginning total assets. Data source: Compustat Annual.
F: A financial strength score of Piotroski (2000) and is calculated as the sum of nine dummies, each equals one if a given condition holds and zero otherwise.16 The conditions are: (1) income before extraordinary items (Compustat item IB) for a fiscal year is positive; (2) cash flow from operations for a fiscal year as defined below is positive; (3) the change in return on assets, defined as income before extraordinary items over a fiscal year divided by beginning total assets, is positive; (4) cash flow from operations exceeds income before extraordinary items for a fiscal year; (5) the change in leverage, defined as long-term debt (items DLTT and DD1) divided by assets at the end of a fiscal year, is negative; (6) the change in liquidity, defined as current assets (item ACT) divided by current liabilities (item LCT) at the end of a fiscal year, is positive; (7) the change in gross margin, defined as one minus the ratio of the cost of goods sold (item COGS) to sales (item REVT) for a fiscal year, is positive; (8) the change in asset turnover, defined as sales for a fiscal year divided by beginning total assets, is positive; and (9) the company has a positive cash flow from
16 The nine financial binary signals measure three different aspects of a firm’s financial condition, namely profitability, changes in financial leverage/liquidity, and changes in operational efficiency It measures the overall improvement or deterioration in a firm’s recent financial health. Firms with a low FSCORE of 3 or below are considered to have deteriorated the most in fundamentals or weak financial performance. Firms with a high FSCORE of 7 or above are considered to have improved the most or strong financial performance. Finally, firms with a medium FSCORE falling between 4 and 6 are considered to have no substantial deterioration or improvement hence medium financial performance.
44
the sale of common and preferred stock (item SSTK) for a fiscal year. The changes above are measured between a fiscal year and the lagged fiscal year. If Compustat indicates that the company reports a statement of cash flows (format code 7), cash flow from operations is net cash from operating activities (OANCF). If the company reports a statement of working capital (format code 1), cash flow from operations equals funds from operations (Op), minus other changes in working capital (item WCAPC), if available. For other format codes, cash flow from operations is funds from operations plus other changes in working capital, if available. Op is income before extraordinary items (item IB) plus income statement deferred tax (item TXDI), if available, plus equity’s share of depreciation expenses for a fiscal year, which is depreciation expenses (item DP) multiplied by market value of equity and divided by total assets minus book value of equity plus market value of equity at the end of a fiscal year. Data source: Compustat.
M/B: Market-to-book equity ratio, calculated as the market value of equity divided by the book value of equity at the end of a fiscal year.17 As in Fama and French (1993), book equity is total assets (Compustat item AT) minus liabilities (item LT), plus balance sheet deferred taxes (item TXDB) and investment tax credits (item ITCI), minus preferred stock liquidation value (item PSTKL) if available, or redemption value (item PSTKRV) if available, or carrying value (item PSTK) if available. Data source: Compustat and CRSP.
MSCORE: A misevaluation score in Piotroski and So (2012) and is calculated by comparing the rank of the financial strength measure FSCORE with the rank of the market-to-book equity ratio M/B. When a firm’s financial performance is strong (high FSCORE) and its earnings expectation is weak (low M/B), expectation error is set to 1 and indicates a high potential for undervaluation. When financial performance is strong (high FSCORE) and earnings expectation is medium (medium M/B) or when financial performance is medium (medium FSCORE) and earnings expectation is weak (low M/B), MSCORE is set to 2 and indicates a mild potential for undervaluation. Likewise, with medium FSCORE and high M/B or low FSCORE and medium M/B, MSCORE is set to 4 and indicates a mild potential for overvaluation. With low FSCORE and high M/B, MSCORE is set to 5 and indicates a high potential for overvaluation. For the rest, MSCORE is set to 3 and indicates a low potential for misvaluation. Data source: Compustat and CRSP.
PRet: Prior-one-year stock return at the end of June, calculated by compounding the 11 monthly raw stock return since the end of previous June. Data source: CRSP.
IVol: Idiosyncratic stock return volatility, measured as the standard deviation of the residual values from the following time-series market model:Ri ,t=bi0+bi 1RM ,t+ei , t ,
17 According to Fama and French (2003), this ratio measures the valuation of a firm’s equity and could proxy for investor expectation of the firm’s future earnings. Earnings expectation is regarded as weak, medium, and strong if the tercile ranking of M/B is low, medium, and high, respectively.
45
where Ri is the monthly stock return and RM is the monthly return on S&P 500 index. The model is estimated with 36 months of return ending in June and requires a full 36-month history. Data source: CRSP.
Price: Share price, measured as the closing stock price (the average of bid and ask prices if the closing price is not available) at the end of June. Data source: CRSP.
DVol: Dollar trading volume, defined as the time-series average of monthly share trading volume multiplied by the monthly closing price over the past one year ending in June. Data source: CRSP.
D/A: Leverage or debt-to-asset ratio, calculated as long term debt (item DLTT) scaled by total assets at the end of a fiscal year. Data source: Compustat Annual.
46
Table 1. Descriptive statistics, sample correlations, and the relation between cash holdings and stock returns
Panel A reports descriptive statistics of cash-to-assets ratio or cash holdings (CH) at the end of fiscal year t. Stdev is the standard deviation. 10%, 25%, 75%, and 90% refer to the 10 th, 25th, 75th, and 90th percentiles, respectively. Panel B reports the time-series averages of the cross-sectional correlations between CH and Size or B/M. Size is the market value of equity at the end of June of calendar year t+1. B/M is the book-to-market equity ratio using Fama and French (1993) book value at the end of fiscal year t. Panel C reports time-series averages of firm characteristics at portfolio formation and adjusted monthly stock returns in % from July of year t+1 to June of year t+2 on decile portfolios sorted and rebalanced annually by CH. N is the number of firms. CHm is the median cash holdings. Ret is the equal-weighted characteristic-adjusted return, which is the stock return minus the return on a 5-by-5 benchmark portfolio matched to the stocks by Size and B/M. [10–1] is the difference in CHm or Ret between the high (10) and the low (1) CH deciles. t is the t-statistic of the average return. Panel D reports the estimated slope coefficients (Coeff.) of the following Fama and MacBeth (1973) cross-sectional regression:
Ret i , t=a+b1 CH i , t+b2 ln¿where Rett is the unadjusted monthly stock return and Ln(Size) is the natural logarithm of Size. t is the t-statistic of the estimate. All the t-statistics are based on Newey and West (1986) standard error with autocorrelations up to 12 lags.
Panel A. Descriptive statistics of cash holdingsMean Stdev 10% 25% Median 75% 90%0.15 0.17 0.02 0.03 0.08 0.21 0.39
Panel B. Sample correlations with cash holdingsSize B/M
–0.03 –0.17
Panel C. Decile portfolios sorted by cash holdingsN CHm aRet t-stat
Table 2. Summary statistics for the relative overvaluation measure
Panel A reports descriptive statistics of relative overvaluation (RO) and its constituents. The constituents are total asset growth (TAG), accounting accruals (Acc), net operating assets (NOA), the capital-investment-to-assets ratio (I/A), external financing (XF), net share issuance (NSI), Ohlson’s (1980) bankruptcy risk score (O), gross profitability (GP), return on assets (ROA), market-to-book equity ratio (M/B), and Piotroski (2000) financial strength score (F) at the end of fiscal year t as well as prior year stock return ending at the end of May of calendar year t+1 (PRet). Panel B reports the time-series averages of the cross-sectional correlations between cash holdings (CH) and RO or its constituents. Panel C reproduces the decile portfolios sorted by CH using the sample containing non-missing RO observations. The detailed definitions of these variables are described in the Appendix.
Panel A. Descriptive statistics of relative overvaluation and its constituentsMean Stdev 10% 25% Median 75% 90%
Table 3. Relative overvaluation and the relation between cash holdings and stock returns
Panel A reproduces the median cash holdings (CHm) and average risk-adjusted returns (aRet) for decile portfolios sorted by cash holdings (CH) reported in Table 1 using the sample containing non-missing relative overvaluation (RO) observations. Panel B reports time-series averages of firm characteristics at portfolio formation and risk-adjusted monthly stock returns in % from July of year t+1 to June of year t+2 on portfolios independently sorted and rebalanced annually by terciles of RO and deciles of CH. [low,10]–[high,1] is the difference in median cash holdings or returns between firms with low relative abnormal valuation (RO=low) and high cash holdings and firms with high relative overvaluation (RO=high) and low cash holdings. The other differences are defined analogously. Panel C reports the estimated slope coefficients of the following cross-sectional regression:
Ret i , t+ 1=a+b1CH hii ,t +b2 CH loi ,t +b3 ROh ii ,t+b4 ROloi ,t
+b5CH h ii , t × ROloi ,t +b6 CH loi , t × ROhii ,t+b7 ln ¿where Reti , t+ 1 is the raw return on stock i in moth t+1, CH_hi (CH_lo) is a dummy variable which equals one if the firm’s cash holdings are in the top (bottom) tercile of cash holdings and zero otherwise, while RO_hi (RO_lo) is a dummy variable which equals one if the firm is in the top (bottom) tercile of relative overvaluation and zero otherwise. Ln(Size) is the logarithm of firm market capitalization and B/M is book-to-market equity. The sample starts from July 1973.
Panel A. Decile portfolios sorted by cash holdings (sample without missing RO or returns)N CHm aRet t-stat
Table 4. Summary statistics for the limits to arbitrage measure
Panel A reports descriptive statistics for the limits to arbitrage (LTA) measure and its constituents. The constituents are idiosyncratic volatility (IVol), stock price (Price), and dollar trading volume (DVol) at the end of June of calendar year t+1. Panel B reports the time-series averages of the cross-sectional correlations between CH and LTA or its constituents.
Panel A. Descriptive statistics of limits to arbitrage and its constituentsMean Stdev 10% 25% Median 75% 90%
Panel B. Sample correlations with cash holdingsLTA IVol Price DVol0.11 0.14 –0.03 0.00
51
Table 5. Limits to arbitrage and the relation between cash holdings and stock returns
Panel A reproduces the median cash holdings (CHm) and average risk-adjusted returns for decile portfolios sorted by cash holdings (CH) reported in Table 1 using the sample containing non-missing LTA observations. Panel B reproduces the estimated slope coefficients of the following cross-sectional regression:
Ret i , t=a+b1 CH i , t+b2 ln¿using the sample containing non-missing LTA observations. Panel C reports time-series averages of firm characteristics at portfolio formation and risk-adjusted monthly stock returns in % from July of year t+1 to June of year t+2 on portfolios independently sorted and rebalanced annually by tercile of LTA and decile of CH. “[High–Low] of [10–1]” is the difference in the [10–1] spread of median cash holdings or risk-adjusted return between firms with high limits to arbitrage (LTA=high) and firms with low limits to arbitrage (LTA=low). Panel D reports the estimated slope coefficients of the above cross-sectional regression across the three LTA terciles. [high–low] is the difference in the slope estimate between firms with high limits to arbitrage and firms with low limits to arbitrage.
Panel A. Decile portfolios sorted by cash holdings (sample with non-missing LTA)N CHm aRet t-stat
Panel D. Limits to arbitrage and the slopes of returns against cash holdings and controlsLTA CH (b1) Ln(Size) (b2) B/M (b3)low Coeff 0.537 –0.083 0.121
Table 6. The relation between cash holdings and stock returns across low and high leverage
Panel A reproduces the median cash holdings (CHm) and average risk-adjusted returns for decile portfolios sorted by cash holdings (CH) reported in Table 1 using the sample containing non-missing debt-to-asset ratio (D/A) observations. Panel B reproduces the estimated slope coefficients of the following cross-sectional regression:
Ret i , t=a+b1 CH i , t+b2 ln¿using the sample containing non-missing ROA observations. Panel C reports time-series averages of firm characteristics at portfolio formation and risk-adjusted monthly stock returns in % from July of year t+1 to June of year t+2 on portfolios first grouped into firms with low leverage (D/A below median) and firms with high leverage (D/A above median) at the end of fiscal year t then by decile of CH. “[low–high] of [10–1]” is the difference in the [10–1] spread of median cash holdings or risk-adjusted return between low leverage firms and high leverage firms. [low,1]–[high,1] is the difference in the median cash holdings or risk-adjusted return between low leverage firms with low cash holdings and high leverage firms with low cash holdings. [low,10]–[high,10] is the difference in the median cash holdings or risk-adjusted return between low leverage firms with high cash holdings and high leverage firms with high cash holdings. Panel D reports the estimated slope coefficients of the above cross-sectional regression across the low leverage firms and high leverage firms. [low–high] is the difference in the slope estimate between low leverage firms and high leverage firms.
Panel A. Decile portfolios sorted by cash holdings (sample from fiscal year 1959 and non-missing D/A)N CHm aRet t-stat
[low–high] of [10–1] 0.29 (2.25)[low,1]–[high,1] 0.29 (2.53)[low,10]–[high,10] 0.00 (0.03)
Panel D. The slopes of returns against cash holdings and controls across low and high leverageLeverage CH (b1) Ln(Size) (b2) B/M (b3)low Coeff 0.672 –0.136 0.357
Table 7. The relation between cash holdings and stock returns across unprofitable and profitable firms
Panel A reproduces the median cash holdings (CHm) and average risk-adjusted returns for decile portfolios sorted by cash holdings (CH) reported in Table 1 using the sample starts from fiscal year 1962 and containing non-missing return on assets (ROA) observations. Panel B reproduces the estimated slope coefficients of the following cross-sectional regression:
Ret i , t=a+b1 CH i , t+b2 ln¿using the sample starts from fiscal year 1962 and containing non-missing ROA observations. Panel C reports time-series averages of firm characteristics at portfolio formation and risk-adjusted monthly stock returns in % from July of year t+1 to June of year t+2 on portfolios first grouped into firms with non-positive ROA (unprofitable) and firms with positive ROA (profitable) in fiscal year t then by decile of CH. “[unprofitable–profitable] of [10–1]” is the difference in the [10–1] spread of median cash holdings or risk-adjusted return between unprofitable firms (ROA≤0) and profitable firms (ROA>0). [unprofitable,1]–[profitable,1] is the difference in the median cash holdings or risk-adjusted return between unprofitable firms with low cash holdings and profitable firms with low cash holdings. [unprofitable,10]–[profitable,10] is the difference in the median cash holdings or risk-adjusted return between unprofitable firms with high cash holdings and profitable firms with high cash holdings. Panel D reports the estimated slope coefficients of the above cross-sectional regression across the unprofitable firms and profitable firms. [Unprofitable–Profitable] is the difference in the slope estimate between unprofitable firms and profitable firms.
Panel A. Decile portfolios sorted by cash holdings (sample from fiscal year 1962 and non-missing ROA)N CHm aRet t-stat
Table 8. Macroeconomic risks and the relation between cash holdings and stock returns
Panel A reproduces the decile portfolios sorted by cash holdings (CH) using the sample that ends on November 2011. Panel B reports the average monthly return in % on five traded portfolios mimicking the Chan, Roll, and Ross (1986) macroeconomic risk factors. RetMP is the return on the portfolio that tracks the growth rate of industrial production (MP). RetUI is the return on the portfolio that tracks the unexpected inflation (UI). RetUTS is the return on the portfolio that tracks the term premium (UTS). RetDEI is the return on the portfolio that tracks the change in expected inflation (DEI). RetURP is the return on the portfolio that tracks the default premium (URP). Panel C reports the estimated parameters of the following time series regressions:Ret p , t−Ret ft=α p , CRR+β p , MP Ret MP ,t+β p , UI RetUI , t+β p ,UTS Ret UTS,t +β p , DEI Ret DEI ,t+β p , URP RetURP , t+ϵ p , t ,where Retp is the subsequent characteristics-adjusted monthly return on a cash-holdings decile portfolio or the return spread [high–low] between the highest and the lowest cash-holdings deciles while Retf is the risk-free rate. Bolded intercept estimates are significant at the 5% level.
Panel A. Decile portfolios sorted by cash holdings (sample ending November 2011)N CHm aRet t-stat
Table 9. The relation between cash holdings and macroeconomic risk-adjusted stock returns across relative overvaluation, limits to arbitrage, low versus high leverage, and unprofitable versus profitable firms
Panels A, B, C, and D reproduce key results of Panel B of Table 3, Panel C of Table 5, Panel C of Table 6, and Panel C of Table 7, respectively, using the estimated intercept of the following time series regressions described in Table 7:Ret p , t−ℜ t ft=α p ,CRR+ βp , MP Ret MP ,t +β p ,UI Ret UI ,t+ βp , UTS RetUTS, t+β p , DEI Ret DEI , t+β p ,URP Ret URP ,t+ϵ p ,t .
Panel A. Relative overvaluation and the return spread on decile portfolios sorted by cash holdingsRO =low αCRR t-stat1 (low CH) –0.02 (–0.21)10 (high CH) 0.74 (4.63)RO =medium 1 (low CH) –0.14 (–1.32)10 (high CH) 0.63 (3.76)RO =high 1 (low CH) –0.55 (–5.16)10 (high CH) 0.53 (0.74)
Panel B. Limits to arbitrage and the return spread on decile portfolios sorted by cash holdingsLTA =low 1 (low CH) 0.09 (1.51)10 (high CH) 0.29 (2.20)[10–1] 0.10 (1.28)LTA =medium 1 (low CH) –0.24 (–2.57)10 (high CH) 0.27 (2.50)[10–1] 0.51 (2.93)LTA =high 1 (low CH) –0.65 (–4.49)10 (high CH) 0.77 (2.57)[10–1] 1.42 (3.86)
[high–low] of [10–1] 1.32 (3.27)
Panel C. The return spread on decile portfolios sorted by cash holdings across low and high leverageLeverage = low1 (low CH) –0.30 (–3.98)10 (high CH) 0.47 (3.58)[10–1] 0.78 (4.51)Leverage = high1 (low CH) –0.23 (–3.35)10 (high CH) 0.18 (1.95)[10–1] 0.41 (3.20)
[low–high] of [10–1] 0.37 (2.54)[low,1]–[high,1] –0.07 (–0.74)[low,10]–[high,10] 0.30 (2.53)
59
Table 9 - continued
Panel D. The return spread on decile portfolios sorted by cash holdings across unprofitable and profitable firmsROA ≤0 (unprofitable) 1 (low CH) –0.87 (–5.27)10 (high CH) 0.25 (1.22)[10–1] 1.12 (4.54)ROA >0 (profitable) 1 (low CH) –0.10 (–1.52)10 (high CH) 0.47 (5.62)[10–1] 0.57 (4.62)
[profitable–unprofitable] of [10–1] 0.54 (2.08)[unprofitable,1]–[profitable,1] –0.76 (–4.16)[unprofitable,10]–[profitable,10] –0.22 (–1.05)
60
Table 10. The relation between cash holdings and stock returns over time
Panel A reports the estimated slope coefficients of the following cross-sectional regression:ℜ t i ,t=a+b1CH i ,t+b2 ln ¿
across three subperiods: fiscal year 1959 to 1976, 1977 to 1994, and 1995 to 2011. Panel B reports the decile portfolios sorted by CH across the three subperiods.
Panel A. Slopes of returns against cash holdings and controls across subperiods1959-1976 CH (b1) Ln(Size) (b2) B/M (b3)Coeff 1.199 –0.132 0.397t-stat (3.50) (–1.49) (3.05)1977-1994Coeff 0.699 –0.095 0.299t-stat (2.10) (–1.46) (3.01)1995-2011Coeff 0.130 –0.086 0.139t-stat (0.15) (–1.16) (2.44)
Table 11. Industry and the relation between cash holdings and stock returns
Panel A reports time-series averages of firm characteristics at portfolio formation and adjusted monthly stock returns in % from July of year t+1 to June of year t+2 on decile portfolios sorted and rebalanced annually by industry adjusted cash holdings (aCH) at the end of fiscal year t. Every year we compute the average cash holdings by Fama and French (1997) 49 industries as defined in French’s data library. Firm level industry adjusted cash holdings is the firm’s cash holdings minus the average cash holdings of the industry to which the firms belong. aCHm is the median industry-adjusted cash holdings in %. Panel B reports the estimated slope coefficients of the following cross-sectional regression:
Ret i , t=a+c1 aCH i , t+c2 ln ¿Panel C reports the estimated slope coefficients of the following cross-sectional regression:
Ret i , t=∑j=1
49
a j IndDV i , j+b1CH i ,t+b2 ln ¿
where IndDV is the set of 49 industry dummies with each equals to one if a firm belongs to industry j and zero otherwise.
Panel A. Decile portfolios sorted by industry-adjusted cash holdingsN aCHm CHm aRet t-stat
