Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Fear and the Fama-French Factors

Robert B. Durand, Dominic Lim, and J. Kenton Zumwalt∗

Investors’ expectations of market volatility, captured by the VIX (the Chicago Board Options
Exchange’s volatility index, also known as the “investor fear gauge”), affects the expected returns
of US equities. Changes in the VIX drive variations in the expected returns of the factors included
in the Fama and French three-factor model augmented with a momentum factor. The market risk
premium (Rm – Rf ) and the value premium (HML) are especially sensitive to changes in the VIX.
An increase in expected volatility is associated with flights to quality and increases in estimated
required returns.

The volatility index (VIX), introduced by the Chicago Board Options Exchange in 1993, is
known as the “investor fear gauge.” The VIX estimates the expected market volatility of the S&P
500 over the next 30 calendar days based on the implied volatility in the prices of options on
the S&P 500 Index. Whaley (2009) provides a history of the VIX and a summary as to how
prices are calculated. Whaley (2009) notes that “the VIX has been dubbed the ‘investor fear
gauge’ . . . [because] the S&P 500 index option market has become dominated by hedgers who
buy index puts when they are concerned about a potential drop in the stock market . . . [the]
VIX is an indicator that reflects the price of portfolio insurance.” The VIX tends to exhibit a
negative correlation with its underlying index. In fact, Fleming, Ostdiek, and Whaley (1995),
using daily data, report a negative contemporaneous correlation (−0.61) between changes in the
VIX (VIX) and the returns of the S&P 100 index. They also observe that VIX are larger for
negative changes in the market than for positive changes.1
Over a decade before the introduction of the VIX, Merton (1980) pointed out that the market
risk premium [E(Rm ) – Rf ] should be positively related to the variance of the market portfolio
and that greater levels of risk should induce a larger market risk premium.2 Since then, Fama and
French (1993) have added two factors to the market risk premium, size, and value, and Carhart
(1997) has added a momentum factor. While the theorized positive correlation between return
and risk is central to financial economics, Lundblad (2007) has recently noted that empirical
support for Merton’s (1980) proposition has been surprisingly weak.

The authors would like to thank Bill Christie (Editor) and an anonymous reviewer for their helpful comments and
suggestions. We would also like to thank John Elder, John Watson, and seminar participants at the Financial Management
Association/Asian Finance Association Conference (Hong Kong), Colorado State University, and the University of
Queensland for their useful feedback on earlier versions of the paper.


Robert B. Durand is a Professor of Finance at the Department of Finance at Curtin University, Bentley, Western
Australia. Dominic Lim is a Postgraduate Research Student in Accounting and Finance at the University of Western
Australia, Crawley WA, Australia. J. Kenton Zumwalt is a Professor of Finance in the Department of Finance and Real
Estate at Colorado State University, Ft. Collins, CO.

1
They find small, but significantly negative first-order autocorrelation in daily (−0.07) and weekly (−0.26) changes but
also report that the autocorrelation varies substantially over time. They conclude that the VIX forecasts of future volatility
are more accurate than an autoregressive model.
2
Merton (1980) also suggests that if there are changes in the distribution of wealth (or changes in risk preferences) that
coincide with changes in the volatility of market returns, the market risk premium may not change.
Financial Management • Summer 2011 • pages 409 - 426
410 Financial Management r Summer 2011
An issue directly related to changes in volatility is the actions investors take when risk changes.
Abel (1988) and Barsky (1989) model the relationship between stock and bond prices (where
stocks are riskier assets than bonds) and find that an increase (decrease) in the expected risk of
equities increases (decreases) the demand for bonds. In their models, bonds serve as a haven from
risk. Thus, investors “fly to quality” when their expectations of risk increase. Using additional risk
factors, high minus low (HML), small minus big (SMB), and winners minus losers (WML) (Fama
and French, 1993; Carhart, 1997), our findings suggest investors do fly to quality consistent with
the models presented by Abel (1988) and Barsky (1989).
This paper examines how investors’ expectations of the total risk of the market, captured by the
VIX, affects US stocks. Specifically, we investigate whether the VIX has a direct and systematic
effect on equity returns by increasing or decreasing the returns of the systematically priced factors
included in the Fama and French (1993) model augmented with a momentum factor (Carhart,
1997). We find that the VIX drives changes in such factors, notably the market risk premium
(Rm − Rf ) and the value premium, HML. Therefore, the VIX determines investors’ expected
returns as its changes are reflected in the time-varying systematically priced risk premia. Our
findings regarding the association between the market risk premium and the VIX provide clear
evidence in support of Merton’s (1980) proposition and the empirical evidence reported by
French, Schwert, and Stambaugh (1987). The advantage of the VIX is that it is forward looking.
It captures the market’s expectation of future volatility. Thus, our analysis demonstrates a way of
incorporating the market’s expectations in required return estimations.
The paper proceeds as follows. In Section I, we introduce the factors incorporated in the model
and summarize arguments concerning the reasons these factors work. Next, in Section II, we
explain and use Granger (1969) causality tests and vector autoregression (VAR) to demonstrate
empirically how fear flows through the factors. Our conclusions are presented in Section III. Our
analysis suggests that failure to consider the impact of variations in the VIX on the risk premia
may have serious ramifications in the estimation of required returns.

I. The Factors
The capital asset pricing model (CAPM) developed by Sharpe (1964), Lintner (1965), and
Mossin (1966) relates required or expected returns to systematic risk, and has become a standard
(benchmark) for pricing risky capital assets. The CAPM is well accepted for measuring portfolio
and company risk, and investment information providers (such as Morningstar and Value Line)
now report the betas (the systematic risk measure) for individual companies and for mutual
funds.
Fama and French (1993), however, found that small capitalization stocks and high book-to-
market stocks tend to have higher returns than those predicted by the CAPM. Their findings led to
the well-known Fama-French (1993) three-factor model that incorporates a size factor (SMB) and
a value factor (the high-book-value-to-market-price minus the low-book-value-to-market-price,
or HML). The model has become very influential and information providers, such as Morningstar,
now report a three-by-three matrix using size (small, medium, and large) and value (growth, blend,
and value) to describe mutual fund style. Mutual funds and exchange traded funds have been
created based on these factors.
In addition to its importance in the investment and portfolio management arena, the Fama-
French (1993) model is becoming more influential in the corporate finance area. The Fama-
French (1993) three-factor model is now being offered as a method for estimating required
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 411
3
returns. The Cost of Capital Yearbook published by Morningstar provides CAPM and Fama-
French (1993) model required return estimates for industries with five or more companies.
While Morningstar’s Beta Book does not provide estimates for the required returns for individual
companies, it does provide CAPM and Fama-French (1993) model information that can be used
to estimate required returns for individual companies.4
A fourth factor has been added to the Fama-French (1993) model. Jegadeesh and Titman (1993)
found that a factor related to recent performance, momentum, or WML, also helps to explain
asset returns. Carhart (1997) augments the Fama-French (1993) model with a momentum factor.
In addition to its use in asset allocation, a recent text by Berk, DeMarzo, and Harford (2009)
demonstrates the use of the four-factor model (the Fama-French model with WML) for estimating
required returns.
We examine the correlations between market volatility, as measured by the VIX, and the factors
in the familiar Fama-French (1993) model augmented with a momentum factor (Carhart, 1997).
Each of the factors presented in the model below is explained individually in the following
sections:

Rjt − Rft = a j0 + b j1 (E(Rmt ) − Rft ) + b j2 SMBt + b j3 HMLt + b j4 MOMt + ε jt . (1)

A. The Market Risk Premium [E(Rm ) – Rf ]


After noting that the market risk premium [E(Rm ) – Rf ] should be positively related to the
variance of the market portfolio, Merton (1980) generates estimates of the variance of the market
and reports that the variance of the market’s returns varies substantially over time. He concludes
that the market risk premium should also vary over time. French et al. (1987) argue that observed
market variance is composed of ex ante (expected) variance, as well as unexpected changes in
variance. They indicate that the expected market risk premium is positively related to expected
volatility and negatively related to unexpected changes in volatility. Their empirical results support
their expectations. We examine how VIX impacts the factors in the Fama-French (1993)/Carhart
(1997) model and our findings are consistent with those of French et al. (1987).
Bakshi and Kapadia (2003) use S&P 500 index options to examine the volatility risk premium.
They report a negative volatility risk premium that is consistent with positive volatility shocks
causing equity prices to react negatively. They conclude that the volatility risk premium is negative
because volatility is priced in capital markets.
More recently, Ang et al. (2006) use VIX to examine the correlation between market volatil-
ity, idiosyncratic and total risk, and cross-sectional firm-level returns. They report a negative
relationship between returns and VIX. Furthermore, when the residuals from the Fama-French
(1993) model are used as measures of idiosyncratic risk, they find that high idiosyncratic volatility
is also negatively related to returns. Jiang and Lee (2006) report significant associations between
idiosyncratic volatility and excess returns. They also find that the market reacts negatively to
innovations in market volatility.

3
See Berk et al. (2009) for a textbook presentation and Pratt and Grabowski (2008) and Koller, Goedhart, and Wessels
(2005) from McKinsey and Company for presentations designed for practitioners. All three sources provide examples of
required return calculations.
4
See Pratt and Grabowski (2008). They caution about use of the model and point out that the Fama-French (1993) model
“. . . has not proven to provide a consistently reliable estimate for the cost of equity capital.”
412 Financial Management r Summer 2011
In summary, finance theory suggests that the market risk premium is related to the variance of
the market. A number of empirical studies report that the volatility of market returns impacts the
market risk premium and that volatility varies over time.

B. The Size Factor (SMB)


Fama and French (1993) argue that two factors, the first capturing the size premium (SMB) and
the second a value-to-growth premium (HML), supplement the market risk premium to provide
a satisfactory model of a cross-section of US returns. The size premium, where firms with
lower market capitalizations generate higher risk-adjusted returns than firms with larger market
capitalizations, was one of the first empirical irregularities documented in the literature (Banz,
1981). The small-firm premium is highest in January. Initially, the tax-loss selling hypothesis
was suggested as a possible explanation, although it turned out to be far from satisfactory (Keim,
1983; Reinganum and Shapiro, 1987). Some researchers argue that the small-firm premium is
an artifact. Roll (1981) posits that small-firm premium results from difficulties in estimating
risk while Roll (1983) argues that the premium may result from difficulties in estimating return.
Shumway (1997) and Shumway and Warther (1999) contend that the small-firm premium arises
due to the problematic nature of the data.
Huij and Verbeek (2009) use mutual funds to examine the four-factor model. They find no
evidence of the SMB factor, but find support for the WML and HML premiums. They suggest
the evaluation of mutual fund managers should be based on factors developed from mutual fund
returns. Behavioral explanations have also been offered for the size effect. For example, Kumar
and Lee (2003) argue that the small firm effect may be due to fluctuating investor sentiment.
Durand, Juricev, and Smith (2007), studying the Australian size premium (one of the highest
size premiums in the world), argue that fluctuations in SMB are a positive function of investors’
misreaction (either over- or underreaction) and their emotional arousal.
Conversely, Liew and Vassalou (2000) present evidence that both the SMB and HML convey
information regarding the future state of the economy in a number of countries. Petkova (2006)
finds that the variation in SMB (and HML) can be captured using a number of state variables that
describe current investment opportunities, consumption preferences, and the future state of the
economy. Sohn (2009) reports that volatility is related to both trading and macroeconomic factors
and suggests these factors are priced because they provide information about future volatility.
Finally, Campbell, Hilscher, and Szilagyi (2008) argue that both SMB and HML capture, to some
extent, a degree of default risk. Even after accommodating these factors, returns of stocks with
high default risk are unusually low.

C. The Value Factor (HML)


In addition to the size premium, Fama and French (1993) add HML, a factor spanning the
value-to-growth premium, to their model of returns. “Value” stocks are those with a higher ratio
of book value of equity to their market capitalization while “growth” stocks are those with a lower
ratio.
Early work on the book-to-market anomaly suggested that the value-growth premium is a
function of systematic misvaluation. Rozeff and Zaman (1998) found that insiders seem to trade
on the book-to-market anomaly, selling glamour (low book-to-market ratio) stocks and buying
value stocks. La Porta et al. (1997) argue that the value-to-growth premium is a quasi-rational
phenomenon. Investors overreact to bad (good) information about earnings, thereby driving the
prices of these stocks below (above) their “correct” values. Therefore, higher (lower) book-to-
market ratios are simply a function of this overreaction and the lower (higher) returns to growth
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 413
(value) stocks represent the correction of prices to this overreaction. When investors’ expectations
are directly observed using analysts’ estimates of future earnings, no overreaction can be detected
(Doukas, Kim, and Pantzalis, 2002). Houge and Loughran (2006) use the augmented Fama-French
(1993) model factor to examine the HML factor. They report similar returns for growth and value
mutual funds from 1975 to 2002 and found no evidence of a significant value premium for a
number of style indexes and large cap stocks.
Some recent studies that examine the value-growth premium suggest that it has an economic,
rather than a behavioral, rationale. For example, Vassalou and Xing (2004) argue that HML
impounds expectations of the likelihood of bankruptcy, while Campbell et al. (2008) suggest that
both HML and SMB reflect default risk. Furthermore, as noted previously in our discussion of
SMB, Liew and Vassalou (2000) report that HML impounds economic expectations. Petkova and
Zhang (2005) argue that, in part, the return of value stocks is driven by economic conditions.
Petkova (2006) finds that variation in HML and SMB can be modeled using variables capturing
economic expectations. As mentioned above, Sohn (2009) confirms that macroeconomic factors
are directly related to volatility. Thus, the bulk of the evidence indicates that HML reflects
investors’ expectations of economic well-being in the future.
Zhang (2005) contends that the value anomaly is a function of the nature of the assets held by
individual firms. He finds that assets in place cause value firms to have greater risk than growth
firms. Thus, the observation that value firms provide greater returns than growth firms reflects
the appropriate return for taking greater risk. Chen, Petkova, and Zhang (2008) find the value
premium to be countercyclical and suggest that value firms are more risky than growth firms
during poor economic conditions. These studies raise the possibility that the correlation between
volatility and the value-growth factor may vary over the economic cycle.

D. The Momentum Factor (WML)


Since the size and value factors in the Fama-French (1993) model are empirically determined,
additional factors have been proposed to explain cross-sectional returns. Jegadeesh and Titman
(1993) report that momentum, in the form of a WML portfolio, adds to the explanation of cross-
sectional returns beyond that explained by the Fama-French (1993) model. Jegadeesh and Titman
(2001) revisit the momentum issue and report that the sources of the excess return are split about
equally between the winning portfolio (0.56%) and the losing portfolio (−0.67%) relative to
an equal-weighted index. Agarwal and Taffler (2008) examine the momentum factor using the
augmented Fama-French (1993) model and an Altman-type (1968) z-score proxy for financial
distress risk. They report no correlation between the z-score proxy and either the HML or SMB
factor, but conclude the momentum effect is a proxy for a financial distress factor.
A behavioral interpretation has also been proposed as the cause of the momentum effect.
Barberis, Shleifer, and Vishny (1998) point out that the observed momentum is consistent with
underreaction to new information in the short term causing momentum but overreaction in the
long term causing reversals. Grinblatt and Han (2005) indicate that investors with unrealized
losses or gains are more reluctant to sell the losers than the winners, and the “perturbed” demand
functions can cause an increase in the spread between the fundamental value of the firm and the
market price. Grinblatt and Moskowitz (2004) argue that both “consistency” and tax loss selling
are strongly related to momentum profitability and the turn-of-the-year effect. Momentum can
partially be explained when investors expect firms that have had a series of positive returns to
continue to generate positive returns in the future, although no correlation is found for consistent
losers.
414 Financial Management r Summer 2011
Momentum has also been examined using mutual funds in the Fama-French (1993) context.
Carhart (1997) augments the three-factor Fama-French (1993) model with the Jegadeesh and
Titman (1993) momentum factor. He reports that persistence for a few funds in the top performing
decile seems to be due to chance, but operating expenses appear to drive many persistent funds
in the bottom decile.
Finally, Campello, Chen, and Zhang (2008) test the Fama-French (1993)/Carhart (1997) models
using expected returns instead of actual historical returns. They generate the expected equity
returns based on the bond yield spreads that are a function of the probability of default and the
default loss rate. They find that the market beta has greater explanatory power with expected
returns than with historical returns, and that SMB and HML returns are countercyclical and
positive but that the momentum factor is not significant.

II. How Fear Flows through the Factors

This section considers the influence of changes in expected volatility, VIX, on the risk premia
captured by Rm − Rf , SMB, HML, and WML. We find that changing expectations of total risk,
captured by VIX, affects the four risk premia in the model.
We examine daily data from February 1, 1993 to July 30, 2007. The data for Rm − Rf ,
SMB, HML, and WML were downloaded from Kenneth French’s website (http://mba.tuck.
dartmouth.edu/pages/faculty/ken.french/) and data on the VIX were obtained from the CBOE.
The end date for our analysis has been chosen to avoid confounding the analysis with the “Global
Financial Crisis.” Our sample ends prior to August 2007, a date that has been referred to as the
start of the Global Financial Crisis. Subsequent to this date, on August 9, 2007, the European Cen-
tral Bank and the US Federal Reserve injected US$90 billion into the financial markets. The VIX
reached 26.48 at the end of the day, almost 2.4 standard deviations more than the average value
during the base case period. Similarly, on September 25, 2008, Washington Mutual was seized
by the Federal Deposit Insurance Corporation and its banking assets sold to JP MorganChase.
The closing level of the VIX on that day was 32.82, around 3.5 standard deviations more than the
average value of 12.9 during the estimation period. It is obvious that investors’ expectations of
total risk, captured by movements in the VIX, varied substantially during the financial crisis.
Before analyzing the data in detail, it is useful to examine summary statistics for the data.
Summary statistics are reported in Table I and are presented in percentage terms. For example,
the mean value for Rm − Rf is 0.031%. It should be noted that the mean values for all of the
variables are positive and the four factors exhibit negative skewness, while the VIX displays
positive skewness. All variable distributions exhibit more kurtosis than the normal distribution.
VIX is the most volatile as shown by the greater range and larger standard deviation. The
Ljung-Box Q-statistics (which test the null hypothesis that the variables are not autocorrelated)
are statistically significant in all cases except Rm − Rf . That is to say the values of SMB, HML,
WML, and VIX are predictable, to varying extents, simply by using historical values. Such a
finding may be consistent with these variables being thinly traded, although in the case of VIX,
this might be an unlikely explanation given that it is calculated using “. . . prices from a deep and
active index option market” (Whaley, 2009). We might expect VIX to be autocorrelated given the
literature on generalized autoregressive conditional heteroskedasticity models (Bollerslev, 1986)
in the finance literature. The autocorrelations in SMB, HML, and WML reported in Table I may
be a function of the portfolios containing thinly traded securities. Campbell, Lo, and MacKinlay
(1997) summarize the effects of thin trading on portfolios. It may be the case that nontrading,
or thin trading, induces autocorrelation in some securities that is reflected in autocorrelation in
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 415
Table I. Summary Statistics

This table presents summary statistics for the variables used in this study. The variables are Rm −Rf , small
minus big (SMB), high minus low (HML), momentum (WML), and the volatility index (VIX). The analysis
uses daily data from February 1, 1993 to July 30, 2007. Rm −Rf is the market risk premium. SMB the
return on the mimicking portfolio for the common size factor in stock returns, is the difference each day
between the simple average of the percentage returns on the three small stock portfolios and the simple
average of the returns on the three big stock portfolios. HML, the return on the mimicking portfolio for
the common book-to-market equity factor in returns, is the simple average of the returns on the two high
BE/ME portfolios and the average of the returns on the two low BE/ME portfolios. WML is the average of
the returns on high prior return portfolios minus the average of the returns on low prior return portfolios.
Big means a firm is above the median market cap on the NYSE at the end of the previous day; small
firms are below the median NYSE market cap. VIX measures the daily percentage change of the Chicago
Board Options Exchange’s VIX. Autocorrelation is measured by the Ljung-Box Q-statistic (where the null
hypothesis is that there is no autocorrelation).

Rm −Rf SMB HML WML VIX


Mean 0.031 0.003 0.020 0.039 0.172
Median 0.060 0.020 0.010 0.060 −0.180
Maximum 5.310 2.910 3.850 5.120 64.215
Minimum −6.650 −4.580 −4.920 −7.270 −25.909
Std. dev. 0.994 0.565 0.578 0.760 5.703
Skewness −0.131 −0.440 −0.038 −0.915 1.034
Kurtosis 7.120 6.658 9.142 12.631 9.995
Autocorrelation
Q-statistic (1 lag) 0.026 0.088 0.134 0.230 −0.066
p-value 0.112 0.000 0.000 0.000 0.000

Table II. Correlations

This table presents the correlations among the variables used in this study. The analysis uses daily data from
February 1, 1993 to July 30, 2007. The variables are described in depth in the text accompanying Table I.

Rm −Rf SMB HML WML VIX


Rm −Rf 1
SMB −0.078 1
HML −0.622 −0.211 1
WML −0.027 0.148 0.006 1
VIX −0.200 −0.021 0.027 −0.116 1

any portfolio containing these securities. While we cannot differentiate between autocorrelation
due to thin trading and autocorrelation as a result of investors responding to past information,
the use of VAR controls for autocorrelation by including lagged observations of the variables we
study. We are able to analyze the extent to which the results are driven by autocorrelation when
we discuss variance decomposition.
Table II demonstrates that, consistent with Merton (1980) and Fleming, Ostdiek, and Whaley
(1995), there is a negative correlation (−0.200) between Rm −Rf and VIX. That is, increasing
fear is associated with a falling market (and vice versa). The market risk premium also ex-
hibits negative contemporaneous correlations with HML (−0.622), SMB (−0.078), and WML
416 Financial Management r Summer 2011
(−0.027). The correlation between HML and Rm −Rf appears high when compared to the others
5

reported in Table II. It may be the case that this high, but negative, correlation is indicative of
a contemporaneous flight-to-quality effect. This would be consistent with the models of Abel
(1988) and Barsky (1989), as well as the empirical analysis of lagged, rather than contempo-
raneous, relations we present below. VIX exhibits negative correlations with SMB (−0.021)
and WML (−0.116). The correlation between VIX and SMB might also be consistent with
a flight-to-quality interpretation as increasing (decreasing) expected risk may lead to investors
being less (more) willing to hold small stocks. It is difficult to argue, however, that such an inter-
pretation might apply to WML (where the negative correlation coefficient suggests that investors
prefer past losers when risk increases). There is a small positive correlation with HML that is also
difficult to interpret. HML and SMB also show a negative correlation (−0.211), while a positive
correlation (0.148) is found between size and momentum.
We use VAR to consider the proposition that VIX influences the risk premia captured by
Rm −Rf , SMB, HML, and WML. (See Hamilton, 1994). VAR facilitates the analysis of any
interrelation among variables. It is a particularly useful technique when, as we see from the
data on autocorrelation reported in Table I, variables exhibit a degree of time dependency. VAR
combines two features. The first is autoregression where the value of a variable today (e.g., VIXt
where t is a subscript denoting time) is a function of past observations (continuing our example,
VIXt−1 where t −1 is a subscript denoting the value of VIX yesterday). Standard regression
can be used to estimate α and β in the equation VIXt = α + βVIXt−1 . If β is found to be
statistically significant, we would conclude that the past values of VIX were associated with
current values. While only one past value is used in this example, any number of past values (lags)
may be used. The second feature of VAR involves running autoregressions for other variables
(say Rm −Rf ). However, when a second autoregression is estimated, it is necessary to include
lagged values of the dependent variable as well as lagged values of the dependent variables in the
other equations. Using the example of a VAR for VIX and Rm −Rf , two regression equations
would be estimated (where the coefficients α 1 , β 1 , and γ 1 for the first equation and α 2 , β 2 , and
γ 2 for the second are obtained using standard regression analysis)

VIXt = α1 + β1 VIXt−1 +γ1 (Rm − Rf ,t−1 ) + ε1t , (2)

Rm − Rf ,t = α2 + β2 VIXt−1 +γ2 (Rm − Rf ,t−1 ) + ε2t . (3)

In this example, both VIX and Rm −Rf are endogenous (determined jointly by the past
observations in the system). VAR allows us to examine the variables we are interested in without
imposing a structural model. The dependent variables in the VAR analysis reported below are
VIX and the returns of Rm −Rf , SMB, HML, and WML. All of these variables are treated as
endogenous. In addition, to consider seasonal and day-of-the-week effects, we include dummy
variables for January, Monday, Tuesday, Wednesday, and Thursday and treat these variables as
exogenous to the model. An examination of the values for Akaike’s information criterion and
likelihood-ratio statistics indicate that the use of two lags is optimal.
The VAR analysis provides information regarding Granger (1969) causality. Granger (1969)
causality demonstrates whether variations in factor X “cause” variations in factor Y , or if variations
in factor Y cause variations in factor X , or if variations in X and Y cause each other. Granger (1969)

5
An anonymous reviewer suggests that the high correlation between Rm −Rf and HML may be spurious as Rm is value-
weighted.
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 417
causality tests are silent regarding the direction of causality (whether the correlation is positive or
negative), but the direction of causality may be analyzed using impulse response Functions that
are discussed below. To extend the two variable example of the preceding paragraph, replacing
X and Y , we can ask if a change in VIX causes a change in Rm −Rf , if a change in Rm −Rf
causes a change in VIX, or if changes in VIX and Rm −Rf cause variations in each other. For
example, using Equations (2) and (3), we might find that β 2 is statistically significant while γ 1
is not statistically significant. Such a finding would suggest that VIX Granger (1969) causes
changes in Rm −Rf , but that changes in Rm −Rf do not Granger (1969) cause VIX. The notion
of causality is one where the lagged observations of one variable have a statistically significant
relationship to another variable. The results of the Granger (1969) causality tests of the null
hypothesis that X does not cause Y are reported in Table III for all of the endogenous variables
included in the VAR analysis: Rm −Rf , SMB, HML, and WML. Panel A of Table III reports that
VIX drives changes in all of the risk premia. The F-statistics reported are all highly significant,
so we can reject the null hypothesis of no causation. The results relating VIX to Rm −Rf provide
support for Merton (1980) and French et al. (1987). Changes in volatility affect the market risk
premium. Changes in volatility impact the Fama-French (1993)/Carhart (1997) factors as well,
and the implications of these results are presented later.
Panel B demonstrates that some causality runs both ways as changes in Rm −Rf , SMB, and HML
cause VIX. In summary, VIX affects all the other risk premia and three of the risk premia,
the market risk premium, the size premium, and the value-growth premium, affect VIX. Panel
C presents all of the remaining causality results and indicates the complexity of the markets.
Causality is evident among nearly all of the comparisons; however, momentum does not cause
the market risk premium and size does not cause momentum.
The analysis of Granger (1969) causality allows us to determine if the variables are related. To
analyze how VIX impacts the risk premia and how changes in the risk premia impact VIX,
we examine the impulse response functions derived from the VAR. In the subsequent analysis,
we focus on the effect of VIX as VIX captures investors’ expectations and allows us to focus
on how changing expectations affect investors’ expected returns. These results are illustrated in
Figure 1 and depict how Rm −Rf , SMB, HML, and WML respond to a one standard deviation
innovation in VIX. The forecasted, or most likely response, is depicted by the unbroken line and
confidence intervals (two standard errors) are depicted using broken lines.6 If the broken lines
contain zero (they cross the horizontal axis), we cannot conclude that the effect is statistically
significant using conventional levels of confidence.
In each impulse response function graph, the vertical axis represents the percentage change
in the variable in the days following a one standard deviation increase in VIX. The horizontal
axis refers to the days relative to the shock. Day 1 is the day of the shock, Day 2 is the first
day after the shock, etc. The shape of the impulse response functions is affected by both the
sequence in which the variables are entered and by the relationships among the contemporaneous
and lagged variables. Because the VIX is entered last, the four factors are unaffected on the
day of the shock, Day 1, and their initial values are zero. The contemporaneous and lagged

6
Impulse response functions (and also the variance decompositions) are influenced by the order in which the variables
are incorporated into the analysis. Examination of different combinations of the ordering of variables in the analysis
did not lead to materially different interpretations of the economic impact of VIX on Rm −Rf , SMB, HML, WML and
of Rm −Rf , SMB, HML, and WML on VIX from those presented here. The ordering of the variables in the reported
analysis is Rm −Rf , SMB, HML, WML, and VIX. A practical consequence of this ordering, however, is that the estimated
responses of the variables to VIX will occur with a one-period lag. The last variable in the ordering is “allowed” to
respond contemporaneously to all other variables, but no other variable is allowed to respond contemporaneously to any
of the other variables.
418 Financial Management r Summer 2011
Table III. Granger Causality: Fama and French’s (1993) Three Factors, the
Momentum Factor, and the VIX

This table presents tests of the null hypotheses that changes in variable X , Granger cause changes in variable
Y , and vice versa. The variables are Rm −Rf , SMB, HML, WML, and VIX. The analysis uses daily data
from February 1, 1993 to July 30, 2007. The variables are described in depth in the text accompanying
Table I.

Null Hypothesis F-Statistic

Panel A

VIX does not Granger cause Rm −Rf 401.51∗∗∗


VIX does not Granger cause SMB 18.988∗∗∗
VIX does not Granger cause HML 140.31∗∗∗
VIX does not Granger cause WML 11.286∗∗∗

Panel B

Rm −Rf does not Granger cause VIX 6.0414∗∗∗


SMB does not Granger cause VIX 4.2322∗∗∗
HML does not Granger cause VIX 3.9868∗∗∗
WML does not Granger cause VIX 0.5532

Panel C

Rm −Rf does not Granger cause SMB 38.913∗∗∗


SMB does not Granger cause Rm −Rf 2.6898
Rm −Rf does not Granger cause HML 12.323∗∗∗
HML does not Granger cause Rm −Rf 4.0364∗∗
Rm −Rf does not Granger cause WML 2.691
WML does not Granger cause Rm −Rf 1.4688
SMB does not Granger cause HML 1.3409
HML does not Granger cause SMB 42.197∗∗∗
SMB does not Granger cause WML 1.3231
WML does not Granger cause SMB 21.076∗∗∗
HML does not Granger cause WML 1.7605
WML does not Granger cause HML 6.4515∗∗∗

∗∗∗
Significant at the 0.01 level.
∗∗
Significant at the 0.05 level.

correlations estimated in the model cause the variables’ primary responses to occur over the two
days following the shock.
The figure in the top left-hand corner depicts how a one standard deviation increase in VIX
impacts the market risk premium Rm −Rf . (Since the relation is symmetric, we can readily infer
the effect of a one standard deviation decrease in VIX by reversing our descriptions.) The
figure reveals that an increase in VIX is associated with a statistically significant six basis point
decrease in Rm −Rf the day after the shock and a much larger, statistically significant drop of over
40 basis points on the second day after the shock. After these downward movements, the figure
suggests a slight recovery in Rm −Rf as the impulse response function is above the horizontal line
and the standard error lines do not contain zero.
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 419
Figure 1. Impulse Response Functions for the Fama and French (1993) Factors,
the Momentum Factor, and VIX

The figures below present the impulse response functions for the market risk premium (Rm −Rf ), size
premium (SMB), value premium (HML), and momentum (WML) following a one standard deviation
innovation in VIX. The unbroken line represents the unique response for each variable each day following
the innovation, Day 1 (reported on the horizontal axis of each table). The magnitude of the accumulated
response, measured as percentage change, is recorded on the vertical axis of each figure. The broken
lines represent confidence intervals at two standard errors. The variables are described in depth in the text
accompanying Table I.

The response of Rm-Rf The response of SMB

.4 .4

.2 .2

% change
% change

.0 .0

-.2 -.2

-.4 -.4

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Days after shock Days after shock
The response of HML The response of WML

.4 .4

.2 .2
% change

% change

.0 .0

-.2 -.2

-.4 -.4

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Days after shock Days after shock

The impulse response functions summarize the correlation between two variables holding all
other variables constant, but the relationship is itself a function of the interplay between all of the
variables. While a one-day change of 40 basis points in the risk premium might appear relatively
small given the shock to VIX, such a change can have a much larger impact on the overall
market movement as it is influenced by SMB, HML, and WML in addition to VIX. It appears to
take five days after a shock to VIX for all of the relationships among the variables to fully play
out in the shock’s effect on Rm −Rf holding all other variables constant. The impulse response
function demonstrating the effect of a shock to VIX on Rm −Rf is consistent with a negative
420 Financial Management r Summer 2011
relationship between the market risk premium and unexpected changes in expected volatility.
That VIX, the change in the expected volatility of the market, must be unexpected follows from
the law of iterated expectations. Overall, these results are consistent with Merton (1980), French
et al. (1987), and Copeland and Copeland (1999).
An examination of the impulse response functions showing the effect of VIX on HML,
SMB, and WML suggests a flight-to-quality effect similar to Abel (1988) and Barsky (1989). The
positive effect of VIX on the value-growth premium (HML) is consistent with investors moving
to value stocks from growth stocks when volatility is expected to increase. As indicated earlier,
this effect must be interpreted with caution as Zhang (2005) and Chen et al. (2008) suggest value
firms may be more risky than growth firms during recessions. The National Bureau of Economic
Research data show a recession occurred from March 2001 to November 2001. Our statistical
results consider the entire period and do not allow for different responses during a recession.
Furthermore, in the figure depicting the effect of VIX on WML (second figure in the right
column), we see a positive effect of VIX on the momentum premium. Investors appear to move
to proven stocks (past winners) and value, rather than glamour stocks. The effect of VIX on
SMB (illustrated in the plot in the top right-hand corner) indicates that investors’ response is
mixed (and is not as clear as those observed for HML and WML). After four days, the size
premium falls and, with this passage of time, is consistent with investors preferring larger stocks
over smaller stocks.
While Figure 1 depicts the response of (Rm −Rf ), SMB, HML, and WML to a one standard
deviation innovation in VIX on the days following an initial shock, Figure 2 presents the
accumulated effect of a one standard deviation innovation over the following 10 days. The figure
in the top left-hand corner of Figure 2 confirms that an increase in the VIX is associated with an
initial decrease in Rm −Rf of about 0.50% or 50 basis points and remains at about 0.38% over the
following 10 days.
The initial positive effects of VIX on the value-growth premium, HML, and the momen-
tum premium, WML, also do not reverse in the 10 days following the initial shock. While in
Figure 1, the effect of the VIX on SMB appears to be mixed, the accumulated response depicted
in Figure 2 suggests that following an increase in the VIX, investors initially prefer larger stocks
over smaller stocks. After five days have passed, however, we find that zero is contained within
two standard errors of the estimated effect (although barely). Such a result indicates that there
is insufficient evidence to conclude that VIX has a lasting effect on SMB using conventional
confidence intervals.
To summarize, the results illustrated in Figure 1 confirm our contention that the response of
Rm −Rf is consistent with French et al. (1987) who find that unexpected changes in volatility are
negatively associated with the risk premium. Given that the VIX represents expected volatility,
VIX must be unexpected. The analysis depicted in Figure 2 supports our contention based on
Figure 1 that the response of the HML and WML is also suggestive of a flight-to-quality effect
(Abel, 1988; Barsky, 1989) as investors move from a more volatile portion of the market to a
safer asset class and, at the same time, move to proven stocks (past winners) and value rather than
glamour stocks.
Impulse response functions show each variable’s influence on the other variables in the system,
one variable at a time. Variance decomposition facilitates the examination of the interrelations
between variations in the risk premia and VIX by determining the proportion of the movement
of each variable that is a function of its own changes as well as changes in the other variables
in the system. Thus, variance decomposition acts as a check on the inferences made utilizing
impulse response functions. If a factor’s variance is fully explained by its own variance, none of
the lagged observations of the other variables have a role in explaining the variable. Examining
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 421
Figure 2. Cumulative Impulse Response Functions for the Fama and French
(1993) Factors, the Momentum Factor, and VIX

The figures below present the cumulative impulse response functions for the market risk premium (Rm −Rf ),
size premium (SMB), value premium (HML), and momentum (WML) following a one standard deviation
innovation in VIX. The unbroken line represents the accumulated response for each variable from the day
of the innovation, Day 1 (reported on the horizontal axis of each table). The magnitude of the accumulated
response, measured as percentage change, is recorded in the vertical axis of each figure. The broken lines
represent confidence intervals at two standard errors. The variables are described in depth in the text
accompanying Table I.

The accumulated response of Rm-Rf The accumulated response of SMB


.4 .4

.2 .2
% change

% change
.0 .0

-.2 -.2

-.4 -.4

-.6 -.6
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Days after shock Days after shock
The accumulated response of HML The accumulated response of WML
.4 .4

.2 .2
% change

% change

.0 .0

-.2 -.2

-.4 -.4

-.6 -.6
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Days after shock Days after shock

the variance decomposition for this variable will confirm that 100% of its variance is explained
by its own past movements.
The variance decompositions of Rm −Rf , SMB, HML, WML, and VIX are reported in Table
IV. The “Period” column indicates the number of days after the system has been “shocked,” while
the other columns indicate the proportion of total variation explained over the 10-day period.
Panel A in Table IV reveals that there are two dominant influences on the variance of Rm −Rf :
itself and the VIX. After a lag of three days, almost 20% of the variance of the market risk
premium is accounted for by the variance of VIX (we see the proportion explained by VIX
rising very quickly after the second day and, at the same time, the proportion of the variation in
422 Financial Management r Summer 2011
Rm −Rf explained by falling). Virtually all of the remaining variance is endogenous to the market
risk premium. The other variables’ contribution to the variance of Rm −Rf is minimal. Clearly,
VIX plays an important role in determining the market risk premium.
The variation in HML is the most complex of those examined. Variation in Rm −Rf accounts
for over 30% of the variance of HML at the first lag and remains at nearly 30% by the end of the

Table IV. Variance Decomposition: Fama and French’s (1993) Three Factors, the
Momentum Factor, and VIX

This table presents variance decompositions of the forecast errors, denoted “s.e.,” of Rm −Rf , SMB, HML,
WML, and VIX at forecast horizons of one to ten days. The columns headed Rm −Rf , SMB, HML, WML,
and VIX give the percentage variation in s.e. due to variations in each of the variables. The variables are
described in depth in the text accompanying Table I.

Period s.e Rm −Rf SMB HML WML VIX


Panel A. Variance Decomposition of Rm −Rf
1 0.89 100.00 0.00 0.00 0.00 0.00
2 0.89 99.43 0.00 0.02 0.05 0.50
3 0.99 81.07 0.13 0.09 0.12 18.60
4 0.99 80.98 0.14 0.09 0.17 18.62
5 1.00 80.50 0.18 0.10 0.17 19.05
6 1.00 80.50 0.18 0.10 0.17 19.05
7 1.00 80.49 0.18 0.10 0.17 19.06
8 1.00 80.49 0.18 0.10 0.17 19.06
9 1.00 80.49 0.18 0.10 0.17 19.06
10 1.00 80.49 0.18 0.10 0.17 19.06
Panel B. Variance Decomposition of SMB
1 0.55 0.06 99.94 0.00 0.00 0.00
2 0.56 1.77 96.32 0.70 1.20 0.00
3 0.56 1.83 94.97 0.69 1.22 1.29
4 0.57 1.84 94.60 0.69 1.24 1.62
5 0.57 1.84 94.53 0.69 1.24 1.70
6 0.57 1.84 94.52 0.69 1.24 1.70
7 0.57 1.84 94.52 0.69 1.24 1.71
8 0.57 1.84 94.52 0.69 1.24 1.71
9 0.57 1.84 94.52 0.69 1.24 1.71
10 0.57 1.84 94.52 0.69 1.24 1.71
Panel C. Variance Decomposition of HML
1 0.54 33.47 7.58 58.95 0.00 0.00
2 0.55 32.51 7.37 59.57 0.05 0.50
3 0.58 29.80 6.86 54.58 0.36 8.41
4 0.58 29.86 6.86 54.50 0.36 8.42
5 0.58 29.78 6.86 54.29 0.36 8.72
6 0.58 29.77 6.86 54.28 0.36 8.73
7 0.58 29.77 6.86 54.28 0.36 8.73
8 0.58 29.77 6.86 54.28 0.36 8.73
9 0.58 29.77 6.86 54.28 0.36 8.73
10 0.58 29.77 6.86 54.28 0.36 8.73

(Continued)
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 423
Table IV. Variance Decomposition: Fama and French’s (1993) Three Factors, the
Momentum Factor, and VIX (Continued)

Period s.e Rm −Rf SMB HML WML VIX


Panel D. Variance Decomposition of WML
1 0.74 0.00 1.47 0.09 98.43 0.00
2 0.76 0.09 1.40 0.09 98.42 0.00
3 0.76 0.13 1.39 0.31 97.71 0.45
4 0.76 0.13 1.39 0.34 97.68 0.45
5 0.76 0.13 1.39 0.34 97.68 0.45
6 0.76 0.13 1.39 0.34 97.68 0.46
7 0.76 0.13 1.39 0.34 97.68 0.46
8 0.76 0.13 1.39 0.34 97.68 0.46
9 0.76 0.13 1.39 0.34 97.68 0.46
10 0.76 0.13 1.39 0.34 97.68 0.46
Panel E. Variance Decomposition of VIX
1 5.55 6.85 0.10 2.16 1.29 89.60
2 5.57 7.18 0.10 2.14 1.37 89.20
3 5.60 7.30 0.34 2.15 1.36 88.84
4 5.60 7.30 0.34 2.15 1.36 88.84
5 5.60 7.30 0.34 2.15 1.36 88.84
6 5.60 7.30 0.34 2.15 1.36 88.84
7 5.60 7.30 0.34 2.15 1.36 88.84
8 5.60 7.30 0.34 2.15 1.36 88.84
9 5.60 7.30 0.34 2.15 1.36 88.84
10 5.60 7.30 0.34 2.15 1.36 88.84

10-day period. Both VIX and SMB play prominent roles in the HML decomposition. VIX
explains over 8% of the variation after a lag of three days and SMB accounts for between 6%
and 8% over the 10 days. Their influence diminishes only slightly with the passage of time. The
variance decompositions of SMB, WML, and VIX, by contrast, reveal that the variation in these
variables may be predominately accounted for by their own variation. Any influence on these
variables by the others in the system is, at best, marginal.

A. Summary
The determinants and magnitude of the equity premium, Rm −Rf , and the other risk premia,
SMB and HML (Fama and French, 1993) and WML (Carhart, 1997) have attracted the attention
of financial economists for some time. When factors are determined empirically, we cannot know
if their significance is simply serendipitous or a function of economically rational expectations or
quasi-rational regularities. Without understanding why the factors work, we cannot be confident
that they will continue to function in the future. The analysis in this section indicates that there is
an economically rational influence on the magnitude of the risk premia. The risk premia varies
with changes in expected risk captured by VIX. The market risk premium, Rm −Rf , has a
negative association with VIX and is, arguably, consistent with Merton (1980) and in keeping
with the empirical findings of French et al. (1987). VIX drives the value-growth premium,
HML, and the momentum premium, WML, in a way that is consistent with theories of investors
flying to quality when expected risk increases. The evidence that VIX has a lasting effect on
424 Financial Management r Summer 2011
the size premium, SMB, is, at best, marginal. Overall, the results of our analysis point to an
economically rational underpinning for the determinants of the risk premia.

III. Conclusion

The Chicago Board Options Exchange’s volatility index, the VIX, reflects investors’ expec-
tations of volatility of the S&P 500 over the next 30 calendar days. The VIX has been called
the “investor fear gauge” as its highs are associated with financial turmoil. The analysis in this
paper has found that the expectations of market volatility captured by the VIX affect equities by
acting on the risk premia in the augmented Fama-French (1993) three-factor model. Investors’
fear and confidence, captured by their expectations of the total risk of the market encapsulated
by the VIX, affects securities through the changing of the other risk premia factors. Specifically,
changes in the VIX (VIX) resulted in variations in the expected returns for the factors included
in the four-factor model, most notably the market risk premium Rm −Rf and the value premium,
HML. The risk premia studied here are determined empirically. Debate rages as to whether their
significance is a function of economically rational expectations or quasi-rational behavior. Con-
sistent with Merton (1980) and French et al. (1987), the market risk premium, Rm −Rf , has a
negative association with VIX. VIX drives the value-growth premium, HML, the size pre-
mium, SMB, and the momentum premium, WML, in a way predicted by flight-to-quality theories
(Abel, 1988; Barsky, 1989). Thus, our analysis confirms an economically rational underpinning
for the determinants of the risk premia. 

References

Abel, A.B., 1988, “Stock Prices under Time-Varying Dividend Risk: An Exact Solution in an Infinite-
Horizon General Equilibrium Model,” Journal of Monetary Economics 22, 375-393.
Agarwal, V. and R. Taffler, 2008, “Does Financial Distress Risk Drive the Momentum Anomaly?” Financial
Management 37, 461-484.
Altman, E.I., 1968, “Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy,”
Journal of Finance 23, 589-609.
Ang, A., R.J. Hodrick, Y. Xing, and X. Zhang, 2006, “The Cross-Section of Volatility and Expected Returns,”
Journal of Finance 61, 259-299.
Bakshi, G. and N. Kapadia, 2003, “Delta-Hedged Gains and the Negative Market Volatility Risk Premium,”
Review of Financial Studies 16, 527-566.
Banz, R., 1981, “The Relationship between Returns and Market Values of Common Stocks,” Journal of
Financial Economics 9, 3-18.
Barberis, N., A. Shleifer, and R. Vishny, 1998, “A Model of Investor Sentiment,” Journal of Financial
Economics 49, 307-343.
Barsky, R.B., 1989, “Why Don’t the Prices of Stocks and Bonds Move Together?” American Economic
Review 79, 1132-1145.
Berk, J., P. DeMarzo, and J. Harford, 2009, Fundamentals of Corporate Finance, New York, NY, Pearson
Prentice Hall.
Bollerslev, T, 1986, “Generalized Autoregressive Conditional Heteroskedasticity,” Journal of Econometrics
31, 307-327.
Durand, Lim, & Zumwalt r Fear and the Fama-French Factors 425
Campbell, J.Y., J. Hilscher, and J. Szilagyi, 2008, “In Search of Distress Risk,” Journal of Finance 63,
2899-2939.
Campbell, J.Y., A.W. Lo, and A.C. MacKinlay, 1997, The Econometrics of Financial Markets, Princeton,
NJ, Princeton University Press.
Campello, M., L. Chen, and L. Zhang, 2008, “Expected Returns, Yield Spreads, and Asset Pricing Tests,”
Review of Financial Studies 21, 1297-1338.
Carhart, M.M., 1997, “On Persistence in Mutual Fund Performance,” Journal of Finance 52, 57-82.
Chen, L., R. Petkova, and L. Zhang, 2008, “The Expected Value Premium,” Journal of Financial Economics
87, 269-280.
Copeland, M.M. and T.E. Copeland, 1999, “Market Timing: Style and Size Rotation Using the VIX,”
Financial Analysts Journal 55, 73-81.
Doukas, J.A., C. Kim, and C. Pantzalis, 2002, “A Test of the Errors-in-Expectations Explanation of the
Value/Glamour Stock Returns Performance: Evidence from Analysts’ Forecasts,” Journal of Finance 57,
2143-2165.
Durand, R.B., A. Juricev, and G. Smith, 2007, “SMB–Arousal, Disproportionate Reactions and the Size-
Premium,” Pacific-Basin Finance Journal 15, 315-328.
Fama, E.F. and K.R. French, 1993, “Common Risk Factors in the Returns on Stocks and Bonds,” Journal of
Financial Economics 33, 3-57.
Fleming, J., B. Ostdiek, and R.E. Whaley, 1995, “Predicting Stock Market Volatility: A New Measure,”
Journal of Futures Markets 15, 265-302.
French, K., G.W. Schwert, and R. Stambaugh, 1987, “Expected Stock Returns and Volatility,” Journal of
Financial Economics 19, 3-30.
Granger, C.W.J., 1969, “Investigating Causal Relations by Econometric Models and Cross-Spectral Meth-
ods,” Econometrica 37, 424-438.
Grinblatt, M. and B. Han, 2005, “Prospect Theory, Mental Accounting, and Momentum,” Journal of
Financial Economics 78, 311-339.
Grinblatt, M. and T.J. Moskowitz, 2004, “Predicting Stock Price Movements from Past Returns: The Role
of Consistency and Tax-Loss Selling,” Journal of Financial Economics 71, 541-579.
Hamilton, J.D., 1994, Time Series Analysis, Princeton, NJ, Princeton University Press.
Houge, T. and T. Loughran, 2006, “Do Investors Capture the Value Premium?” Financial Management 35,
5-19.
Huij, J. and M. Verbeek, 2009, “On the Use of Multifactor Models to Evaluate Mutual Fund Performance,”
Financial Management 38, 75-102.
Jegadeesh, N. and S. Titman, 1993, “Returns to Buying Winners and Selling Losers: Implications for Stock
Market Efficiency,” Journal of Finance 48, 65-92.
Jegadeesh, N. and S. Titman, 2001, “Profitability of Momentum Strategies: An Evaluation of Alternative
Explanations,” Journal of Finance 56, 699-720.
Jiang, X. and B-S. Lee, 2006, “The Dynamic Relation between Returns and Idiosyncratic Volatility,”
Financial Management 35, 43-65.
Keim, D., 1983, “Size Related Anomalies and Stock Return Seasonality (Further Empirical Evidence),”
Journal of Financial Economics 25, 75-97.
426 Financial Management r Summer 2011
Koller, T., M. Goedhart, and D. Wessels, 2005, Valuation: Measuring and Managing the Value of Companies,
4th Ed., Hoboken, NJ, John Wiley and Sons.
Kumar, A. and C.M.C. Lee, 2003, “Individual Investor Sentiment and Comovement in Small
Stock Returns,” Cornell University Department of Economics Working Paper. Available at SSRN:
http://ssrn.com/abstract=328980.
La Porta, R., J. Lakonishok, A. Shleifer, and R. Vishny, 1997, “Good News for Value Stocks: Further
Evidence of Market Efficiency,” Journal of Finance 52, 859-874.
Liew, J. and M. Vassalou, 2000, “Can Book-to-Market, Size and Momentum Be Risk Factors that Predict
Economic Growth?” Journal of Financial Economics 57, 221-245.
Lintner, J., 1965, “Security Prices, Risk and Maximal Gains from Diversification,” Journal of Finance 20,
587-615.
Lundblad, C., 2007, “The Risk Return Tradeoff in the Long Run: 1836–2003,” Journal of Financial
Economics 85, 123-150.
Merton, R. 1980, “On Estimating the Expected Return on the Market: An Exploratory Investigation,”
Journal of Financial Economics 8, 323-361.
Mossin, J., 1966, “Equilibrium in a Capital Asset Market,” Econometrica 34, 768-783.
Petkova, R., 2006, “Do the Fama-French Factors Proxy for Innovations in Predictive Variables?” Journal of
Finance 61, 581-612.
Petkova, R. and L. Zhang, 2005, “Is Value Riskier than Growth?” Journal of Financial Economics 78,
187-202.
Pratt, S.P. and R.J. Grabowski, 2008, Cost of Capital: Applications and Examples, 3rd Ed., Hoboken, NJ,
John Wiley and Sons.
Reinganum, M. and A. Shapiro, 1987, “The Anomalous Stock Market Behavior of Small Firms in January:
Empirical Tests for the Tax-Loss Selling Effects,” Journal of Business 60, 281-295.
Roll, R., 1981, “A Possible Explanation of the Small Firm Effect,” Journal of Finance 36, 879-888.
Roll, R., 1983, “On Computing Mean Returns and the Small Firm Premium,” Journal of Financial Economics
12, 371-386.
Rozeff, M.S. and M.A. Zaman, 1998, “Overreaction and Insider Trading: Evidence from Growth and Value
Portfolios,” Journal of Finance 53, 701-716.
Sharpe, W.S., 1964, “Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk,”
Journal of Finance 19, 425-442.
Shumway, T., 1997, “The Delisting Bias in CRSP Data,” Journal of Finance 52, 327-340.
Shumway, T. and V.A. Warther, 1999, “The Delisting Bias in CRSP’s Nasdaq Data and Its Implications for
the Size Effect,” Journal of Finance 54, 2361-2379.
Sohn, B., 2009, “Cross-Section of Equity Returns: Stock Market Volatility and Priced Factors” (March 18).
Available at SSRN: http://ssrn.com/abstract=1364834.
Vassalou, M. and Y. Xing, 2004, “Default Risk in Equity Returns,” Journal of Finance 59, 831-868.
Whaley, R.E., 2009, “Understanding the VIX,” Journal of Portfolio Management 35, 98-105.
Zhang, L., 2005, “The Value Premium,” Journal of Finance 60, 67-103.

You might also like