SSRN Id3966127

Where is the risk in risk factors?
Paul Geertsema∗ and Helen Lu†
First version: 29 May 2020

This version: 18th November 2021
Abstract
The value, profitability, investment and momentum factors are all more profitable in bear
markets than in bull markets. This contradicts the predictions of consumption based models
of risk. We hypothesise that this pattern can be explained by state-dependent exposure
to duration and financial constraint risks in bear markets. We find that equity cash flow
duration subsumes bear market premia of the value, profitability, investment and (to a lesser
extent) momentum factors. Moreover, duration factors are not subsumed by other factors.
In contrast, financial constraint factors only explain the profitability factor in bear markets.
Keywords: Risk factors; Investment CAPM; Bear markets; Cash flow duration; Financial con-
straints; Short-sales constraints
JEL codes: G10; G12; G40
We thank participants in seminars at the University of Auckland, Victoria University of Wellington and Auckland
Centre for Financial Research (Auckland University of Technology) for their valuable comments. We thank Henk
Berkman, Yeguang Chi, Adrian Fernandez-Perez, Griffin Geng, Xing Han, Jinji Hao, Prasad Hegde, John Lee,
Dimitris Margaritis and Peiming Wang for their insights and suggestions.
∗
University of Auckland. E-mail: p.geertsema@auckland.ac.nz
†
University of Auckland. Send correspondence to: helen.lu@auckland.ac.nz
1
Electronic copy available at: https://ssrn.com/abstract=3966127
1 Introduction
We investigate whether duration risk and financial constraint risk can explain the state-dependent
performance of risk factors. We proxy for bad states with the bear market phases of stock markets
(hereafter bear markets, based on Nyberg, 2013). Surprisingly, four of the five non-market factors
from the Fama and French (2018) six-factor model (FF6) make more money in bear markets than
in bull markets. The SMB size factor is the only exception. As shown in Figure 1, these factor
returns are an order of magnitude larger in bear markets than in bull markets. Returns from
investing in the value, profitability, investment and momentum factors (HML, RMW, CMA and
UMD) are highly positive and significant in bad states. By contrast, their average returns in good
states are either close to zero (for HML and CMA) or less than one-third to one-half of their
average returns in bad states (for RMW and UMD). The size factor (SMB) is the only non-market
factor that loses money in bear markets.1
Figure 1: Average monthly returns of factors in bear and bull markets

This figure plots the mean monthly factor returns for the five non-market factors of the FF6 model of Fama and
French (2018) for the full sample, and for bull and bear markets. Bear and Bull markets are defined in Table 1.
(%) Fullsample Bear Bull
1.2 1.10 1.09

1.03
1
0.8 0.66
0.57 0.54
0.6
0.4 0.28 0.27 0.27
0.21 0.25
0.16
0.2
0.05 0.03
0
-0.04
-0.2
SMB HML RMW CMA UMD
1
For the sake of brevity, in this paper we use the term factors to refer specifically to the five non-market factors
of the FF6 model, namely SMB, HML, RMW, CMA and UMD, unless otherwise specified.
2
The above-average profitability of factors in bad states is puzzling in the context of classical
consumption based models of risk. If a risk factor proxies for the consumption state as proposed
by Fama and French (1992), it should yield lower payoffs on average in bad states. An asset with
sub-par returns in bad states of the world is undesirable and investors therefore require a premium
to hold assets of this nature.2 Contrary to the predictions of consumption-based asset-pricing
models, four of the five factors “pay out” more during bad consumption states, despite earning a
positive premium in the full sample.
The expected market risk premium (ERP) tends to increase dramatically in bear markets (see
Table 1).3 The ERP increased in all of the 11 bear markets, with increases ranging from 0.62 and
3.88 percentage points per annum. By contrast, the ERP often stays at lower levels in bull markets
than in bear markets. The persistent increases in risk premia in bear markets reduce the value
of long-duration stocks more than short-duration stocks, because long-duration stocks are more
sensitive to changes in discount rates. We posit that duration risk explains the counter-cyclical
risk premia of the value, profitability, investment and momentum factors. Our findings support
this hypothesis.
We find that the duration of stocks on the long side of SMB, HML and CMA are significantly
shorter than those on the short side. This difference is consistent with a duration-based explanation
of the bear market profits earned by these factors. To test whether duration does indeed explain
the bear market premia earned by these factors, we construct a duration based factor following the
same approach used to construct the Fama-French CMA factor in Fama and French (2018) (refer
to Section 5 for the detail of portfolio construction). DURG and DUR refer to the duration factors
based on duration measures from Gonçalves (2021) and Dechow et al. (2004), respectively. Similar
to HML, RMW, CMA and UMD, the DURG and DUR factors also almost all of their profits
in bear markets. Monthly returns from DURG and DUR average 1.77% and 1.06% each in bear
market. However, their profits in bull markets average around zero. We find that the DURG factor
2
Cochrane (2009) puts it this way: Other things equal, an asset that does badly in states of nature like a recession,
in which the investor feels poor and is consuming little, is less desirable than an asset that does badly in states of
nature like a boom in which the investor feels wealthy and is consuming a great deal. The former asset will sell for
a lower price; its price will reflect a discount for its “riskiness,” and this riskiness depends on a co-variance, not a
variance. (page 3)
3
ERP is sourced from Damodaran’s database https://pages.stern.nyu.edu/∼adamodar/New_Home_Page/datafile/histimpl.ht
3
subsumes the bear-market profits of the value (HML), profitability (RMW), investment (CMA)
and momentum (UMD) factors. The DUR factor also renders almost all factors insignificant and
explains a substantial portion of the bear-market profits from CMA and UMD factors. Importantly,
the factors themselves do not subsume DURG in bear markets. The bear-market profit from the
DUR factor is only subsumed by the CMA factor.
Credit spreads tend to widen in bear markets and stay at lower levels in bull markets (see Table
1). Credit spreads widened in 10 out of the past 11 bear markets. The increases in credit spreads
range from 0.05 to 1.96 percentage points. Such changes in external financing conditions dispropor-
tionately affect financially constrained firms (Jansen & Tsai, 2010; Livdan et al., 2009). Widening
credit spreads might force financially constrained firms to scale back their operations more aggress-
ively. For example, financially constrained firms that face tightening external financing conditions
may have to delay expansion plans or curtail sales to limit working capital demands.
We find limited evidence for the mechanism described above. We construct two financial constraint
indices, namely, the WW Index based on Whited and Wu (2006) and the HP Index based on
Hadlock and Pierce (2010).4 We refer to the financial constraint factors based Whited and Wu
(2006) and Hadlock and Pierce (2010) as FCWW and FCHP, respectively. Consistent with Farre-
Mensa and Ljungqvist (2016), we show that firms in the small, value, weak (less-profitable) and
aggressive (high-investment) portfolios are more financially constrained on average than firms in
the large, growth, strong and conservative portfolios. Financial constraint factors (which buy
financially constrained stocks and sell unconstrained stocks) are profitable in bull markets but
loss-making in bear markets; this stands in contrast to the duration factors because duration
factors make all their profits in bear markets. Financial constraint factors (FCWW and FCHP)
only subsume the RMW factor in bear markets and leave the other factors intact. Moreover, the
financial constraint factors are themselves subsumed by the HML factor and the CMA factor, as
well as by the duration factors. This suggests that financial constraints at most plays a minor role
4
Farre-Mensa and Ljungqvist (2016) show that these two measures are consistent with other important financial
constraint indicators, including non-dividend-paying status and low credit ratings. “’Constrained’ firms according
to the dividend, ratings, HP, or WW measures are smaller and younger, carry more cash on their balance sheets,
have fewer tangible assets, lower return on assets, and lower marginal tax rates, and are more likely to face a zero
marginal tax rate. They are also less leveraged and rely more on short-term debt” (Farre-Mensa and Ljungqvist
(2016), p 277)
4
in explaining bear market factor profitability.
The 11 bear markets from July 1963 to March 2020 in our study is dated following the approach in
Nyberg (2013) using the market excess return (MKTRF) from Ken French’s data library.5 We also
generate alternative proxies for bad states. These include bear markets dated using the algorithm
of Lunde and Timmermann (2004), down markets (months with negative MKTRF), crash markets
(months where the MKTRF is below −10%) and NBER recessions. The overall pattern observed
(except for NBER recessions) is that HML, RMW, CMA and UMD are all more profitable in bad
states. The HML, RMW, CMA and UMD factors are not alone in generating most of their profits
in bear markets. In our robustness tests, we show that a large number of well-known factors are
also more profitable in bear markets than in bull markets. These factors include the I2A, ROE and
EG factors of the five-factor model of Hou et al. (2020), the long-term reversal factor (LTREV) of
De Bondt and Thaler (1985), the HLMdevil factor of Asness and Frazzini (2013) and the betting
against beta factor (BAB) of Frazzini and Pedersen (2014).
We examine returns to the long- and short-legs of the FF6 factors in bull and bear markets. The
full sample premia generated by HML, RMW, CMA and UMD are mostly the result of large
negative returns on the short legs of these strategies during bear markets. In bull markets the long
leg returns exceed the short leg returns for all five factors, but the difference is typically small –
of the order of 3 to 27 bp/month (except for UMD where the difference is 55 bp/month). In bear
markets the excess of short leg returns over long leg returns is 104 bp/m for HML, 57 bp/m for
RMW, 110 bp/m for CMA and 109 bp/m for UMD, all statistically significant at the 1% level (5%
for RMW). Because short-selling is particularly costly and difficult in bear markets, this finding
suggests that it could be difficult to realise the apparent profits earned by factors in bear markets.
Another possible explanation is that the profitability of factors reflect behavioural biases rather
than priced risks. If factor returns are the result of biased investor expectations, then we would
expect earnings announcement returns to be a substantial fraction of factor returns (as biases
are revealed to be in error by the release of actual earnings). Following the approach of Porta
et al. (1997) we construct our own versions of the factors, but with actual returns replaced by
5
https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
5
[-1,1] cumulative returns centred on earnings announcement days. In bear markets the earnings
announcement returns for HML, RMW, CMA and UMD make up between 8% and 15% of full
returns while accounting for 5% of the days. On the whole the evidence suggests that biased
investor expectations explain only a small part of realised factor premia.
In summary, our study contributes to the literature investigating the drivers of factor returns.
First, we add to the literature investigating risk-based explanations for factor premia by uncovering
the strong and state-dependent explanatory power of duration and (to a lesser extent) financial
constraint risks. In this sense, our work is most closely related to Zhang (2005). Zhang (2005)
shows that costly reversibility in investments and time-varying exposure to risk can explain the
counter-cyclical value premium. We show that many risk factors in addition to the value factor
also exhibit strong counter-cyclical risk premia. In addition, the fact that duration risk can explain
these counter-cyclical premia appears to be consistent with the hypotheses of Zhang (2005).
Second, we contribute to the recent literature on the role of duration risk in asset-pricing. A
recent paper by Weber (2018) considers the duration measure of Dechow et al. (2004) and finds
that long-duration stocks earn high returns in periods with high levels of investor sentiment.
Gonçalves (2021) motivates a short-duration premium with an Intertemporal Consumption Asset
Pricing Model and finds that the duration factor explains the profitability and investment premia.
More recently, Gormsen and Lazarus (2021) propose a duration-based explanation for risk-premia
for the value, profitability, investment, low-risk and payout factors. We complement Gonçalves
(2021) and Gormsen and Lazarus (2021) in that we show that the duration premium is strongly
counter-cyclical and that its explanatory power for factors is most evident in bear markets.
2 Factor premia
This section briefly summarises the literature motivating the five non-market factors of the Fama
and French (2018) six-factor model, in this paper simply referred to as factors.
6
2.1 Size and value
Fama and French (1993) motivate the size and value factors in their seminal paper as follows: “one
of our central themes is that if assets are priced rationally, variables that are related to average
returns, such as size and book-to-market equity, must proxy for sensitivity to common (shared and
thus undiversified) risk factors in returns.”
The size factor originates from the size anomaly uncovered by Banz (1981) and it is the second
most widely studied factor by citations (Alquist et al., 2018). The size anomaly led to the inclusion
of the SMB factor in the Fama and French (1993) three-factor model. Since then academics have
put forward a large number of risk-based explanations for the size premium including liquidity
risk, default risk, distress risk, cash flow risk and innovation risk (for example, see Vassalou and
Xing (2004), Campbell and Vuolteenaho (2004), Campbell and Vuolteenaho (2004) and Stoffman
et al. (2019)). Recent papers by Alquist et al. (2018) and Vassalou and Xing (2004) provide
detailed summaries of risk-based explanations for the size premium. Alquist et al. (2018) point
out that small stocks face tighter limits to arbitrage (for example, lower institutional ownership and
higher trading costs), thus the size premium is consistent with both behavioural interpretations
and risk-based explanations.
The value factor is by far the most studied factor by citations (Alquist et al., 2018). Early studies
use risks related to consumption and wealth to explain the value premium (for example, see Fama
and French, 1992; Panton et al., 1976). Lakonishok et al. (1994) and Porta et al. (1997) argues for
a behavioural explanation for the value premium on the basis that value is more profitable in bad
times and thus cannot be a compensation for risk exposure. The work by Zhang (2005) is among
the first to propose time-varying investment risk of firms (the production side) as an explanation
for the value premium. As Zhang (2005) explains, “In bad times, value firms are burdened with
more unproductive capital, finding it more difficult to reduce their capital stocks than growth firms
do. The dividends and returns of value stocks will hence covary more with economic downturns.”
These predictions are supported by Petkova and Zhang (2005) who show that time-varying betas
and expected market risk premia explain the value premium. Golubov and Konstantinidi (2019)
7
conduct a comprehensive investigation of a large number of explanations for the value premium
documented in the literature. Risk-based explanations investigated in their study include firms’
investment risk (Zhang, 2005), operating leverage (Novy-Marx, 2011), duration of the stock cash
flows (Dechow et al., 2004), duration risk (Lettau & Wachter, 2007), exposure to technology shocks
(Kogan & Papanikolaou, 2014), analyst risk ratings (Lui et al., 2012), cash flow risk (Campbell
& Vuolteenaho, 2004) and consumption risk (Panton et al., 1976). Golubov and Konstantinidi
(2019) finds evidence that cash flow risk and consumption risk are associated with the value
premium. In addition, irrational expectations and limits to arbitrage explanations also partially
explain the value premium. More recently, Gonçalves (2021) constructs a new duration measure
based on Dechow et al. (2004) and shows that duration subsumes the value and profitability
premia. Gormsen and Lazarus (2021) proposed a duration-based explanation for a wider range of
risk factors and find results consistent with their hypotheses using expected returns implied from
futures on dividend strips.
2.2 Investment and profitability
Berk et al. (1999), Cochrane (1991) and Zhang (2005) are among the first to model the link between
firm characteristics and stock returns from the production side of financial assets. Subsequent
factor models (e.g. the q-factor model by Hou et al. (2015) and the Fama-French five-factor model
by Fama and French (2015)) use sorted portfolios to proxy for the investment factor and the
profitability factor.
Hou et al. (2015) build on the q-theory of investment and derive relations between investment,
expected profitability and expected return within an economic framework. The intuition under-
pinning the investment factor is that for a fixed cash flow forecast, a higher expected return (or
discount rate) will result in a lower net present value of the project and thus lower investment.
Conversely, lower investment reflects higher expected returns. The same model shows that future
profitability is positively related to expected return.
Fama and French (2015) motivate the investment factor and profitability factor by way of a di-
8
vidend discount model. Because the market value of equity is the present value of future earnings
less future investment, they argue that both the profitability and investment factors are natural
choices to augment the three-factor model of Fama and French (1993) (p2, Fama and French
(2015)).
2.3 Momentum
Fama and French (2018) add the momentum factor to the Fama and French (2015) five-factor
model to “satisfy insistent popular demand” (p237, Fama and French (2018)). Indeed, according
to Alquist et al. (2018), momentum is the third most studied factor by citations. While researchers
put forward many behavioural explanations for momentum returns (for example, see Barberis et
al., 1998; Daniel et al., 1998; Grinblatt and Han, 2005; Hong and Stein, 1999), there are also a
number of risk-based explanations (for example, see Johnson, 2002; Sagi and Seasholes, 2007).
We examine the factors of the Fama and French (2018) six-factor model in the main body of our
paper (for brevity referred to simply as factors hereafter). In the Appendix we provide results
relating to the factors of the Hou et al. (2020) five-factor model (see Section 9.2 in Appendix), as
well as a number of other well known factors that are not part of the Fama and French (2018) or
the Hou et al. (2020) models – see Section 9.3 in the Appendix.
3 Data
We use factor time-series data from Ken French’s data library to generate our main results and
factor returns from Lu Zhang’s global-q library and the AQR insights data library for ancillary
analysis.6 Table 10 in the Appendix contains a summary description of the factors and the sorting
variables used in their construction.
To construct portfolios we rely on the Compustat quarterly and annual data files, as well the
CRSP monthly file. Our sample starts in July 1963 and ends in March 2020. The portfolios
6
See Sections 9.2 and 9.3 in the Appendix for detail
9
are constructed using value-weighted returns with NYSE breakpoints. We only include common
domestic US stocks (CRSP share codes 10 or 11) listed on the NYSE, Amex or NASDAQ (CRSP
exchange codes 1, 2 or 3). Returns are delisting adjusted. Where delisting returns are missing we
follow the approach in Beaver et al. (2007) which entail replacing missing delisting returns with
the average delisting return over the prior 60 months for firms with the same exchange and 2-letter
delisting code. If still missing we replace missing delisting returns with -30% for NYSE stocks and
with -55% otherwise, following Shumway (1997) and Shumway and Warther (1999).
To test whether factor returns during bear markets may be driven by biased expectations, we
calculate stock-level earnings announcement returns for each factor during both bull and bear
markets. The earnings announcement returns are calculated over a [-1,1] window entered on the
quarterly earnings announcement date (Compustat quarterly variable RDQ) and are set to zero
for all days outside the window. We then recreate the SMB, HML, CMA and RMW factors using
the earnings announcement returns instead of normal returns.
[Insert Table 1 here]
Our classification of bear markets prior to the COVID-19 pandemic is detailed in Table 1 and
follows Table 1 of Nyberg (2013). We add an additional bear market for the period covering
January 2020 to March 2020 to reflect the initial preriod of the COVID-19 pandemic of 2020.
This period covers the incubation, outbreak and fever periods of COVID-19 studied by Ramelli
and Wagner (2020). The Bull and Bear indicator variables reflect these bull and bear markets
respectively.
The bull and bear market turning points in Nyberg (2013) are constructed using the Bry and
Boschan (1971) dating rule with modifications proposed by Pagan and Sossounov (2003) for de-
tecting cycles in stock markets. The Pagan and Sossounov (2003) algorithm first recognises local
peaks and troughs within a fixed window. Then it applies a set of censoring rules to determine
whether these peaks and troughs qualify as bull or bear market turning points. The censoring
rules include a minimum bear and bull market length (six months unless the change is more than
20% in absolute terms), minimal full cycle length (16 months) and some additional heuristics.7
7
Appendix B in Pagan and Sossounov (2003) describe the steps to recognise and filter the peaks and troughs.
10
Nyberg (2013) applied the Pagan and Sossounov (2003) algorithm and manually deleted two mild
bear markets in 1971 and 1994, because the market decreased by less than 10% in those episodes.
As a robustness check, we also use the Lunde and Timmermann (2004) algorithm to identify bull
and bear markets (indicated by the BullLT and BearLT indicator variables). The Lunde and
Timmermann (2004) approach only specifies the trough to peak increase (15%) and decease (15%)
and does not impose any limit on durations; thus it requires fewer parameters than either the Bry
and Boschan (1971) or the Pagan and Sossounov (2003) approaches.
We also use NBER business cycles (Expansion and Recession) and stock market crash months
(where the monthly return is lower than -10%) as additional state variables to investigate factor
returns in different states of the world. Stock market crashes (based on the market premium
MKTRF) are indicated by Crashmarket, with Stablemarket representing any month that is not a
Crashmarket month. Likewise we identify positive and negative market premium months by the
Upmarket and Downmarket indicator variables, respectively.
4 Factor performance in bear and bull markets
We find that most factors perform better in bear markets than in bull markets. Table 2 details
the performance of factors (arranged in columns) under different market states (arranged in rows).
Panel A reports averages of monthly factor returns in the full sample (first row) and in good and
bad states, where state variables are defined in the five different ways as described in Section 3.
Panel B summarises the means of factor returns for each of the 11 bear markets we consider.
Only SMB consistently conforms to the behaviour expected of a consumption-based risk factor.
SMB produces lower returns in bear markets than bull markets, while generating a positive return
– a premium – over the full sample. All the other factors (HML, RMW, CMA and UMD) produce
lower returns in bull markets than in bear markets. For HML and CMA the difference is significant
at the 5% level. The pattern is the same when we use an alternative approach by Lunde and
11
Timmermann (2004) to date bull and bear markets. SMB continues to be the only factor that
generates losses in bear markets (BearLT ) and gains in bull markets (BullLT ). HML, RMW, CMA
and UMD all generate economically higher returns in BearLT months, significant at the 5% level
for all but UMD.
In terms of month-by-month variation, all factors (again apart from SMB) are more profitable in
Upmarket months (positive market excess return months) compared to the remaining Downmarket
months. This is also the case when comparing Crashmarket months (market excess returns below
-10%) with Stablemarket months (the remaining months).
NBER expansions and recessions give rise to different results than states based on the stock market.
This difference occurs because business cycles and stock markets are not perfectly synchronised,
since stock markets are leading indicators of the real economy (Fama, 1981; Harvey, 1989). Both
SMB and HML generate slightly higher returns in expansions, while RMW and CMA continue to
be more profitable in recessions (though only marginally so in the case of RMW). On the other
hand, UMD generates the bulk of its premium during NBER expansions (0.73% in expansions vs
0.17% in recessions).
Panel B of Table 2 details the performance of factors in each of the 11 bear markets from the
Vietnam war to the COVID-19 pandemic. As expected, average monthly MKTRF returns are
all significantly negative, ranging from -11.20% for the 1987 crash (Crash1987 ) to -1.08% for the
second oil shock (Oilshock2 ). Overall SMB is the most consistently negative (7 of 11 bear markets).
All other factors produce more positive than negative bear market returns. For example, HML is
negative in only three bear markets and positive in eight bear markets. CMA generated positive
returns in 10 out of 11 bear markets; the only negative return occurred during the COVID bear
market.
Many other well-known factors are also more profitable in bear markets. These factors include
the I2A, ROE and EG factors in the five-factor model of Hou et al. (2020), the long-term reversal
factor (LTREV) of De Bondt and Thaler (1985), the HLMdevil factor ofAsness and Frazzini (2013)
and the betting against beta factor (BAB) of Frazzini and Pedersen (2014) (refer to Tables 11 and
12 in the Appendix).
12
Examining each bear market in turn we find that positive factor returns outnumber negative factor
returns in all bear markets apart from Crash87 and COVID. During the COVID bear market,
SMB, HML, RMW and CMA made the largest average monthly losses in the bear market history
we cover. Only UMD has performed well during the COVID bear market, producing a monthly
profit of 4.59% on average. Factors generally behaved differently in the COVID bear market. This
difference may be a result of fund outflow from factors during the COVID period. Active funds
experienced fund outflows that exceeded pre-crisis trends in the early stage of pandemic (Pástor &
Vorsatz, 2020). In addition, the COVID pandemic shutdown drastically reduced immediate cash
flows thus severely impacted firms with short cash flow durations, which are typically value firms
(Dechow et al., 2021).
[Insert Figure 2 here]
Figure 2 provides a visual description of the performance of factors in bull and bear markets.
By construction the MKTRF market premium (Panel A) is perfectly aligned with bear markets.
The bull markets of 1991-1999 and 2010-2020 are substantially longer, and achieve much higher
cumulative returns, than other bull markets. By comparison the recent COVID bear market
(cumulative return of -20.5% over three months) seems small and short. The SMB factor (Panel
B) is initially in synchrony with bear markets, but after the mid-1970’s this association breaks
down. The HML value factor seems to be particularly counter-cyclical, often generating high
returns in bear markets but producing mixed results in bull markets. It did particularly well
during the Dotcom crises as lofty (decidedly non-value) internet stocks tumbled back to earth.
Both CMA and RMW exhibits relative low volatility, with the exception of the Dotcom crisis,
when they both perform well. The UMD momentum factor is fairly subdued in the early part of
our sample, but does very well during the 1990’s bull market. By contrast it suffers large losses in
the early stages of the most recent two bull markets. UMD is the only factor to produce positive
returns during COVID.
If factors do not generate low returns during classical bad consumption states, then when do they
generate low returns? To investigate we apply the bear market dating approach of Lunde and
13
Timmermann (2004) to cumulative factor returns. These factor bear markets indicate states when
factors are generally losing money, exactly analogous to the bear markets identified by analysing
the aggregate market.
Figure 3 illustrates factor bear markets visually. The shaded grey areas in Figure 3 correspond
to the bear markets listed in Table 1. The bottom half of Figure 3 identifies all 11 individual
(MKTRF-based) bear markets, for ease of reference. NBER recessions are indicated directly above
the individual bear markets. As expected NBER recessions and bear markets correspond loosely,
although the turning points do not always line up. Above NBER recessions we indicate MKTRF
bear markets identified using the Lunde and Timmermann (2004) approach, which corresponds
closely to the bear markets in Table 1.
The SMB bear markets initially line up with the second and third bear markets (Nixon and
Oilshock1 ), but thereafter produces long stretches that appear to be mostly aligned with bull
markets (1984-91, 1994-99 and 2012-2020). While SMB is the only factor that performs worse
during bear markets, the long losing streaks of SMB during apparently tranquil times make it
hard to interpret the SMB premium as compensation for classical “bad state” risk. HML bear
markets are shorter and reasonably evenly distributed over time, but only occasionally correspond
with bear markets or recessions. There are only three RMW bear markets. The first bridges
the two oil shocks while the latter two bookends the Dotcom crisis. The CMA bear markets are
concentrated in the latter part of our sample, particularly after the Dotcom crisis and continuing
unbroken for seven years starting in 2013 – during the longest bull market in history. UMD
bear markets are of far shorter duration than the other factor bear markets and seem to be loosely
clustered around bear markets without actually aligning with them. This suggests some connection
with the formation and ultimate resolution of bear markets. The results in Figure 3 suggests that
the performance of factors during bear markets reflect the unique circumstances of each factor
instead of systematic risk across business cycles. While economic textbooks treat “bad” states as
homogeneous and interchangeable, real bear markets all seem to be different in their own way.
14
5 Duration and financial constraints
5.1 Duration and financial constraints in factor edge portfolios
We hypothesize that factor premia are associated with cross-sectional differences in duration and
financial constraints. Changes in expected market risk premium affect high-duration stocks dis-
proportionality. In addition, financially constrained firms are more sensitive to variations in credit
spreads. As a first step, we examine the difference in duration and financial constraints between
stocks allocated to the long and short legs of factors.
We use two duration measures. The first measure is constructed following the approach in Dechow
et al. (2004) and Weber (2018). It is estimated for each firm annually from 1963 to 2020 with
Compustat/CRSP annual financial data for US common stocks excluding firms in the utilit-
ies (4900≤SIC≤4949) and financials (6000≤SIC≤6999) sectors.8 In addition, we also use the
Gonçalves (2021) duration measure from his database which runs from 1973 to 2018.9
As for financial constraints, we rely on two commonly used financial constraint indices – the WW
Index (Whited & Wu, 2006) and the HP Index (Hadlock & Pierce, 2010). The construction of
these indices are detailed in section 9.4 in the Appendix.
Both the duration measures and financial constraint indices are estimated in June of year t relying
on information no later than December of calendar year t − 1. Then, these duration measures are
used to form portfolios from July of year t to June of year t + 1.
Average duration differ significantly between the long and short legs of factors (refer to LMS in
Panels A and B of Table 3) in seven out of 10 cases across the two different measures of duration.
8
Following Dechow et al. (2004), we use a cost of equity capital of 12% and a long-run growth rate of 6%. The
number of projection years used to estimate equity duration in our study is 15 years, as in Weber (2018). The
auto-correlation coefficient for growth in our panel data is 17%.
9
In our sample, the Dechow et al. (2004) duration average about 20 years which is shorter than the 50 years
using the Gonçalves (2021) measure. In addition, the Gonçalves (2021) duration factor exhibits stronger pricing
power than the factor constructed based on the duration measure from Dechow et al. (2004). These attributes of
two measures are consistent with findings in Gonçalves (2021).
15
For example, the average duration of value stocks (the long leg of HML in Table 3) is shorter than
that of growth stocks by 7 years based on the Dechow et al. (2004) duration measure (Duration
in Panel A of Table 3) and by 24 years based on the Gonçalves (2021) duration measure.
To mitigate the impact from extreme duration estimates, we also examine the difference in average
tercile numbers of duration. Stocks are sorted into terciles based on their duration in each year.
The results are similar (see Panels C and D in Table 3).
Panels E and D of Table 3 report the difference in financial constraint indices between the firms
allocated to the long and short legs of different factors. Value and high-investment firms are
more financially constrained than growth and low-investment firms and the difference is highly
significant at a 0.1% level. In contrast, small firms tend to be more financially constrained than
large firms based on both the WW Index and the HP Index.
We also consider the time-series dynamics of duration and financial constraints of firms allocated
to the long and short sides of different factors (Figure 4)
The average duration tercile based on the measure from Dechow et al. (2004) is higher for the short
leg than for the longt leg for the size, value and investment factors for most of the sample period
(Figure 4). The value factor stands out. Throughout the sample period, the duration tercile for
growth stocks average about 3 while that for value stocks average about 1 (refer to Panel B of
Figure 4). Small and big firms have similar duration on average but the duration of big firms (the
short leg) peaked and far exceeded that of small firms around 1980s and 2000s. The duration gap
between low-investment (conservative) and high-investment (aggressive) stocks disappeared after
2000.
The time-series of Gonçalves (2021) duration displays a similar pattern, except for two noticeable
differences (refer to Figure 6 in the Appendix). First, based on the Gonçalves (2021) measure
the duration gap between value and growth stocks narrows after 1990 and disappears around
2018; while this gap is highly persistent when considering the duration measure of Dechow et al.
(2004). Second, the Gonçalves (2021) duration of weak profitability stocks exceeds that of robust
16
profitability stocks after 2000; but based on the measure in Dechow et al. (2004) robust profitability
stocks tend to have longer duration than weak profitability stocks throughout the sample period.10
Difference in financial constraint indices persists throughout the sample period for the SMB, HML
and CMA factors (refer to Panels A, B and D in Figure 5). In contrast, weak profitability stocks
were more financially constrained than robust profitability stocks after the mid-1980s only. The
time-series difference in financial constraints remains similar if we use the HP Index (Refer to
Figure 7 in the Appendix).
5.2 Duration and financial constraint factors
We construct duration and financial constraint factors to formally test whether the exposure to
these risks can explain the observed state-dependent risk premia of factors.
When constructing duration factors and financial constraint factors, we follow Fama and French
(2018) by sorting stocks into six size-duration portfolios (small and big times short-, medium- and
long-duration). The returns to a hedge duration portfolio is the value-weighted returns to short
duration stocks minus those of long duration stocks (or, 1/2 (small and short duration + big and
short duration) minus 1/2(small and long duration + big and long duration)). At the end of June
in year t, we use accounting information ending on or before December year t − 1 to form portfolios
to be held from July in year t to June in year t + 1. Financial constraint factors are constructed
using the same procedure where we take long positions in financially constrained stocks and short
financially unconstrained stocks.
As predicted by the mechanisms through which expected market premium and credit spreads
affect firms differently in bad and good times, we find that duration premia are strongly counter-
10
This might be related to the different methods of forecasting cash flow in two methods. While Dechow et al.
(2004) forecast cash flow as the future earnings less the change in book equity, Gonçalves (2021) forecast cash
flow as payout (or the sum of dividends and repurchases less equity issuance). As a result, the duration measure
based on Gonçalves (2021) is shorter for high issuance firms, all else equal. Because firms with less tangible assets
(or, high book-to-value firms) and profitable firms are high-issuance firms [[[ citation? ]]] recent decades, the
Gonçalves (2021) duration of growth firms and robust profitability firms shortens relative to that for value and
weak profitability firms.
17
cyclical while financial constraint premia are strongly pro-cyclical (Table 4). Duration and financial
constraint factors exhibit strong state-dependent risk premia. The Dechow et al. (2004) duration
factor (DUR) and the Gonçalves (2021) duration factor (DURG) realise monthly returns of 1.10%
and 1.79% on average in bear markets, respectively; while their average monthly profits in bull
markets are close to zero and insignificant. In sharp contrast, the two financial constraint factors
(FCWW and FCHP) are profitable in bull markets and highly unprofitable in bear markets.
Duration and financial constraint factors are strongly and negatively correlated with each other,
as well as strongly correlated with some factors (Table 5). Duration factors are most strongly
correlated with the HML factor at 0.82 (DUR) and 0.57 (DURG), followed by the CMA factor at
0.58 (DUR) and 0.51 (DURG). This pattern holds in both bull and bear markets.
Financial constraint factors are strongly negatively correlated with the HML, CMA and RMW
factors. By contrast, financial constraint factors are positively correlated with the SMB factor.
The relationship between financial constraints and UMD is weakly positive in bull markets and
weakly negative in bear markets.
To disentangle the relationship between the FF6 non-market risk factors and our newly constructed
duration and financial constraint factors, we turn to factor-on-factor regressions. The regression
alphas are detailed in Table 6, for regression betas see Section 4 in the Appendix.
Consistent with the costly reversibility hypothesis of Zhang (2005), we find that duration factors
strongly explain the counter-cyclical bear market premia of factors. The Gonçalves (2021) duration
factor (DURG) subsumes all of the positive bear-market premia from the HML, RMW, CMA and
UMD factors. In Panel A of Table 6, the monthly average alphas of these factors in bear markets
(“A. Bear”) are −0.11% to 0.30% per month (all insignificant at conventional confidence levels)
when regressed on the DURG factor. The Dechow et al. (2004) duration factor (DUR) does not
render the CMA and UMD factors insignificant, but reduce the sizes of their average monthly
18
returns from 1.10% and 1.09% to 0.58% (with a p-value of 1%) and 0.76% (with a p-value of 10%),
respectively. By contrast, none of the factors subsume the DURG duration factor (Panel B of
Table 6). In bear markets, DURG makes an average monthly profit of 1.79% with a p-value of
1% (“A. Bear” in Panel B of Table 6). Controlling for the factors marginally reduces the DURG
bear-market alphas, but leaves its statistical significance intact. Only the CMA factor subsumes
the bear-market premium of DUR.
Financial constraint factors (FCWW and FCHP) appear to explain the RMW factor where they
reduce the alphas by around half and render them insignificant (in “A. Bear” of Panel A of Table
6). In addition, both financial constraint factors are very robust to the RMW factor. In “A. Bear”
of Panel B of Table 6, the RMW factor only reduces the bear-market premia of FCWW and FCHP
from −0.66% and −0.67% (both significant at 1%) to −0.45% and −0.44% (significant at 5% and
10%), respectively. However, both financial constraint factors are highly correlated with duration
factors so that duration factors completely explain financial constraint factors’ premia in bear
markets (see Table 7). Thus, financial constraint factors do not offer much additional explanatory
power in addition to duration factors for the bear-market premia of the RMW factor.
6 Short-sale constraints
Another potential explanation for the profitability of factors in bear markets could be limits to
arbitrage, particularly short sales. Shorting stocks can be costly in normal times and even more
so in bad times. If the short legs of factor strategies account for the bulk of factor returns, it may
be difficult for an investor to actually realise the full factor strategy return. We find that bear
market factor profits originate primarily from the short side.
We repeat the analyses in Panel A of Table 2 with the long and short legs of each non-market
factor. Table 8 reports the results. Both long and short legs are profitable in good states and loss-
19
making in bad states for all factors, following overall market movements. For example, monthly
returns to the long leg of SMB (SMB_L) average 2.04% in bull markets (Bull) and -2.20% in
bear markets (Bear). This remains the case when we use alternative definitions of states. In good
states the long and short legs of factor portfolios tend to have similar average returns. In contrast,
in bad states the short legs drop more than the long legs. For example, in the Bear state the
HML short leg (HML_S) produces -2.74% while the long leg (HML_L) produces -1.70%. The
difference of 1.03% accounts for the bear market return of HML (Panel A of Table 2). Thus the
short leg contributes disproportionately to the the overall profit of the HML strategy during bear
markets. Similarly, in bear markets RMW, CMA and UMD also generate somewhere between
two-thirds to almost all of their monthly profits from shorting stocks. The only exception is the
SMB size factor. SMB consistently lose more money from its long leg than its short leg during
bear markets (Bear, BearLT, Downmarket and Crashmarket). Similar to our findings in Section
4, the overall magnitude of the difference between short and long legs is lower in NBER recessions,
but the overall pattern of short legs disproportionately contributing to factor returns during bad
states remain intact.
7 Biased expectations
In the presence of limits to arbitrage biased expectations can also lead to return predictability
and hence factor profitability. We pursue this potential explanation by examining the reaction of
stocks around earnings announcements. Earnings announcements are often regarded as important
events that correct biased expectations (for example, see Engelberg et al. (2018) and Porta et al.
(1997)). We calculate factor returns using stock-level returns around earnings announcement dates
to test whether factor profits in bad states are consistent with biased expectations.
In the spirit of Porta et al. (1997), we generate earnings announcement returns for each factor,
which we then compare with normal factor returns in different states of the market. Earnings
announcement returns for each stock are cumulative returns over the three-day window surrounding
20
the earnings announcement date. We replace a constituent stock’s monthly return with its three-
day cumulated returns surrounding an earnings announcement date if this three-day window falls
in the calendar month. If the constituent stock does not have an earnings announcement in a
given month we set the monthly earnings announcement return to zero. We then re-construct risk
factor returns following the methodology of Fama and French (2018)11 . Table 9 report the factors
returns and earnings announcement only returns (indicated by the suffix “_rea”).
Table 9 shows that the value and momentum factors (HML and UMD) generate significant returns
around earnings announcement dates over the full sample. HML_rea and UMD_rea in Table 9
are 0.04 and 0.07 percentage points and significant at 5% and 1% levels respectively. The average
earning announcement return of HML is 15% of the average HML factor return and that of UMD
is 11% of its average factor return.
The value, profitability and momentum factors (HML, RMW and UMD) usually generate signi-
ficant returns around earnings announcement dates in the four bad market states (Bear, BearLT,
Downmarket and Crashmarket). UMD earnings announcement returns are significant in all four
bad states, while RMW is significant in all except Downmarket. HML earnings announcement
returns are significant in two bad states (Downmarkets and BearLT ). The magnitude of earnings
announcement returns are around 10% of full returns but reflect only 5% of trading days12 , sug-
gesting that factors earn twice as much on earnings announcement days in bear markets than on
other days in bear markets. These results suggest that the profits of RMW and UMD in bad states
may be partially attributed to the correction of biased expectations.
By contrast there is little to suggest that CMA earnings announcement returns are different from
other days, both in the full sample and during bad market states. SMB earning announcement
returns are close to zero in the full sample, and around 8% of normal returns during bad markets.
On the whole biased expectations does not seem to offer an explanation for these two factors.
11
Our replicated factors (SMB, HML, RMW, CMA and UMD) using conventional returns all have correlations
above 0.97 with the factors obtained from Ken French’s data library.
12
There are 3 earnings announcement days per quarter and around 62 trading days per quarter. This means 3/62
= 4.8% of days are included in earnings announcement returns.
21
8 Conclusion
We investigate risk, limits to arbitrage and biased expectations as explanations for the large returns
produced by factors during bear markets from the Vietnam War to the COVID-19 pandemic. We
find that all factors (except SMB) generate substantially higher average monthly returns in bear
markets than in bull markets. Such bear market profitability can not be readily explained by
consumption based models because high payoffs in bad states should be associated with negative
risk premia. This inconsistency with risk-based explanations might be related to limits to arbitrage
since the bear market profitability of factors primarily comes from the short side. We find evidence
that the bear market profitability of HML, CMA, RMW and UMD is explained by duration; in
particular, the duration factors we construct largely subsumes HML, CMA, RMW and UMD, while
the converse does not hold. The role of financial constraints appear to be subsidiary to that of
duration and is limited to explaining the bear market profits of RMW only. Finally, we considered
whether factor profitability in bear markets might be related to biased expectations, but on the
whole we find little support for this hypothesis.
22
References
Alquist, R., Israel, R. & Moskowitz, T. (2018). Fact, fiction, and the size effect. The Journal of
Portfolio Management, 45 (1), 34–61.
Asness, C. & Frazzini, A. (2013). The devil in hml’s details. Journal of Portfolio Management,
39 (4), 49.
Asness, C., Frazzini, A. & Pedersen, L. H. (2017). Quality minus junk.
Banz, R. W. (1981). The relationship between return and market value of common stocks. Journal
of Financial Economics, 9 (1), 3–18.
Barberis, N., Shleifer, A. & Vishny, R. (1998). A model of investor sentiment. Journal of financial
economics, 49 (3), 307–343.
Beaver, W., McNichols, M. & Price, R. (2007). Delisting returns and their effect on accounting-
based market anomalies. Journal of Accounting and Economics, 43 (23), 341–368.
Berk, J. B., Green, R. C. & Naik, V. (1999). Optimal investment, growth options, and security
returns. The Journal of Finance, 54 (5), 1553–1607.
Bry, G. & Boschan, C. (1971). Cyclical analysis of time series: Selected procedures and com-
puter programs. Cyclical analysis of time series: Selected procedures and computer programs
(pp. 13–2). NBER.
Campbell, J. Y. & Vuolteenaho, T. (2004). Bad beta, good beta. American Economic Review,
94 (5), 1249–1275.
Cochrane, J. H. (1991). Production-based asset pricing and the link between stock returns and
economic fluctuations. The Journal of Finance, 46 (1), 209–237.
Cochrane, J. H. (2009). Asset pricing: Revised edition. Princeton university press.
Daniel, K., Hirshleifer, D. & Subrahmanyam, A. (1998). Investor psychology and security market
under- and overreactions. The Journal of Finance, 53 (6), 1839–1885.
De Bondt, W. F. & Thaler, R. (1985). Does the stock market overreact? The Journal of Finance,
40 (3), 793–805.
Dechow, P. M., Erhard, R. D., Sloan, R. G. & Soliman, M. T. (2021). Implied equity duration: A
measure of pandemic shutdown risk. Journal of Accounting Research, 59 (1), 243–281.
Dechow, P. M., Sloan, R. G. & Soliman, M. T. (2004). Implied equity duration: A new measure of
equity risk. Review of Accounting Studies, 9 (2), 197–228.
Engelberg, J., McLean, R. D. & Pontiff, J. (2018). Anomalies and news. The Journal of Finance,
73 (5), 1971–2001.
Fama, E. F. (1981). Stock returns, real activity, inflation, and money. The American economic
review, 71 (4), 545–565.
Fama, E. F. & French, K. R. (1992). The cross-section of expected stock returns. The Journal of
Finance, 47 (2), 427–465.
Fama, E. F. & French, K. R. (1993). Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics, 33 (1), 3–56.
Fama, E. F. & French, K. R. (2015). A five-factor asset pricing model. Journal of Financial
Economics, 116 (1), 1–22.
Fama, E. F. & French, K. R. (2018). Choosing factors. Journal of Financial Economics, 128 (2),
234–252.
Farre-Mensa, J. & Ljungqvist, A. (2016). Do measures of financial constraints measure financial
constraints? The Review of Financial Studies, 29 (2), 271–308.
23
Frazzini, A. & Pedersen, L. H. (2014). Betting against beta. Journal of Financial Economics,
111 (1), 1–25.
Golubov, A. & Konstantinidi, T. (2019). Where is the risk in value? evidence from a market-to-book
decomposition. The Journal of Finance, 74 (6), 3135–3186.
Gonçalves, A. S. (2021). The short duration premium. Journal of Financial Economics.
Gormsen, N. J. & Lazarus, E. (2021). Duration-driven returns. Available at SSRN 3359027.
Grinblatt, M. & Han, B. (2005). Prospect theory, mental accounting, and momentum. Journal of
financial economics, 78 (2), 311–339.
Hadlock, C. J. & Pierce, J. R. (2010). New evidence on measuring financial constraints: Moving
beyond the kz index. The Review of Financial Studies, 23 (5), 1909–1940.
Harvey, C. R. (1989). Forecasts of economic growth from the bond and stock markets. Financial
Analysts Journal, 45 (5), 38–45.
Hong, H. & Stein, J. C. (1999). A unified theory of underreaction, momentum trading, and over-
reaction in asset markets. The Journal of finance, 54 (6), 2143–2184.
Hou, K., Mo, H., Xue, C. & Zhang, L. (2020). An augmented q-factor model with expected growth.
Review of Finance.
Hou, K., Xue, C. & Zhang, L. (2015). Digesting anomalies: An investment approach. Review of
Financial Studies, 28, 650–705.
Jansen, D. W. & Tsai, C.-L. (2010). Monetary policy and stock returns: Financing constraints and
asymmetries in bull and bear markets. Journal of Empirical finance, 17 (5), 981–990.
Johnson, T. C. (2002). Rational momentum effects. The Journal of Finance, 57 (2), 585–608.
Kogan, L. & Papanikolaou, D. (2014). Growth opportunities, technology shocks, and asset prices.
The Journal of Finance, 69 (2), 675–718.
Lakonishok, J., Shleifer, A. & Vishny, R. W. (1994). Contrarian investment, extrapolation, and
risk. The Journal of Finance, 49 (5), 1541–1578.
Lettau, M. & Wachter, J. A. (2007). Why is long-horizon equity less risky? a duration-based
explanation of the value premium. The Journal of Finance, 62 (1), 55–92.
Livdan, D., Sapriza, H. & Zhang, L. (2009). Financially constrained stock returns. The Journal of
Finance, 64 (4), 1827–1862.
Lui, D., Markov, S. & Tamayo, A. (2012). Equity analysts and the market’s assessment of risk.
Journal of Accounting Research, 50 (5), 1287–1317.
Lunde, A. & Timmermann, A. (2004). Duration dependence in stock prices: An analysis of bull
and bear markets. Journal of Business & Economic Statistics, 22 (3), 253–273.
Novy-Marx, R. (2011). Operating leverage. Review of Finance, 15 (1), 103–134.
Nyberg, H. (2013). Predicting bear and bull stock markets with dynamic binary time series models.
Journal of Banking & Finance, 37 (9), 3351–3363.
Pagan, A. R. & Sossounov, K. A. (2003). A simple framework for analysing bull and bear markets.
Journal of applied econometrics, 18 (1), 23–46.
Panton, D. B., Lessig, V. P. & Joy, O. M. (1976). Comovement of international equity markets: A
taxonomic approach. Journal of Financial and Quantitative Analysis, 11 (03), 415–432.
Pástor, L. & Vorsatz, M. B. (2020). Mutual fund performance and flows during the covid-19 crisis.
The Review of Asset Pricing Studies, 10 (4), 791–833.
Petkova, R. & Zhang, L. (2005). Is value riskier than growth? Journal of Financial Economics,
78 (1), 187–202.
Piger, J. M. & Chauvet, M. (2020). Smoothed u.s. recession probabilities [recprousm156n], retrieved
from fred, federal reserve bank of st. louis.
24
Porta, R. L., Lakonishok, J., Shleifer, A. & Vishny, R. (1997). Good news for value stocks: Further
evidence on market efficiency. The Journal of Finance, 52 (2), 859–874.
Ramelli, S. & Wagner, A. F. (2020). Feverish stock price reactions to covid-19.
Sagi, J. S. & Seasholes, M. S. (2007). Firm-specific attributes and the cross-section of momentum.
Journal of Financial Economics, 84 (2), 389–434.
Shumway, T. (1997). The delisting bias in CRSP data. The Journal of Finance, 52 (1), pp. 327–340.
Shumway, T. & Warther, V. A. (1999). The delisting bias in CRSP’s Nasdaq data and its implic-
ations for the size effect. The Journal of Finance, 54 (6), 2361–2379.
Stoffman, N., Woeppel, M. & Yavuz, M. D. (2019). Small innovators: No risk, no return. Kelley
School of Business Research Paper, (19-5).
Vassalou, M. & Xing, Y. (2004). Default risk in equity returns. The Journal of Finance, 59 (2),
831–868.
Weber, M. (2018). Cash flow duration and the term structure of equity returns. Journal of Financial
Economics, 128 (3), 486–503.
Whited, T. M. & Wu, G. (2006). Financial constraints risk. The Review of Financial Studies,
19 (2), 531–559.
Zhang, L. (2005). The value premium. The Journal of Finance, 60 (1), 67–103.
25
Table 1: Bear markets
This table summarises bear market periods. For each bear market, we report its starting month (Start), end-
ing month (End), length in number of months (N ), cumulative excess market return (Total Drop), change in
implied ERP (Chg in ERP), change in credit spread (Chg in Spread) and change in sentiment (Chg in Sent).
The dating rule for bear markets prior to the COVID-19 pandemic follows Nyberg (2013) which is based on
Bry and Boschan (1971) and Pagan and Sossounov (2003). The COVID-19 pandemic includes the first three
months in 2020 (Ramelli & Wagner, 2020). Market Drop is the cumulative market excess return (MKTRF)
in each bear market in percentage points. Chg in ERP is the change in the levels of implied equity risk
premium in each bear market (for example, the ERP at the end of 2009m2 minus that at the end of 2007m10
for GFC). Chg in Spread and Chg in Sent are the changes in the levels of credit spread and in sentiment, re-
spectively. ERP is based on the implied ERP for the S&P500 portfolio (column “ERP(T12m)”) from Damodaran
(http://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/implpr.html); when monthly ERP is unavail-
able, we estimate ERP as ERPt = ERPt−1 × SP 500t−1
SP 500t . Credit spreads are from FRED, Federal Reserve Bank of
St. Louis. The sentiment index is from Wurgler’s website (http://people.stern.nyu.edu/jwurgler/). The sample
period runs from July 1963 to March 2020 inclusive.
Bear Market N Start End Market Drop Chg in ERP Chg in Spread Chg in Sent
Vietnam 8 1966m2 1966m9 -18.02 0.70 0.35 0.40
Nixon 19 1968m12 1970m6 -40.18 1.89 0.18 0.21
Oilshock1 21 1973m1 1974m9 -53.02 3.88 0.46 -1.50
Oilshock2 14 1977m1 1978m2 -14.61 1.92 -0.37 0.29
Stagflation 20 1980m12 1982m7 -33.30 1.60 0.68 1.32
Correction83 11 1983m7 1984m5 -17.99 0.62 0.26 1.83
Crash1987 3 1987m9 1987m11 -31.04 1.14 0.05 -0.06
Gulfwar 5 1990m6 1990m10 -19.72 0.65 0.36 0.04
Dotcom 30 2000m9 2003m2 -47.45 2.30 0.33 -1.72
GFC 16 2007m11 2009m2 -51.51 3.87 1.96 -0.79
COVID 3 2020m1 2020m3 -20.52 0.96 0.87 NA
26
Cumulative Gross Return Cumulative Gross Return Cumulative Gross Return
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
0.00
1.00
2.00
3.00
4.00
5.00
6.00
1963m1 1963m1 1963m1
1964m1 1964m1 1964m1
1965m1 1965m1 1965m1
1966m1 1966m1 1966m1
1967m1 1967m1 1967m1
1968m1 1968m1 1968m1
1969m1 1969m1 1969m1
1970m1 1970m1 1970m1
1971m1 1971m1 1971m1
1972m1 1972m1 1972m1
1973m1 1973m1 1973m1
1974m1 1974m1 1974m1
1975m1 1975m1 1975m1
1976m1 1976m1 1976m1
1977m1 1977m1 1977m1
1978m1 1978m1 1978m1
1979m1 1979m1 1979m1
bull and bear market.
1980m1 1980m1 1980m1
1981m1 1981m1 1981m1
1982m1 1982m1 1982m1
1983m1 1983m1 1983m1
1984m1 1984m1 1984m1
1985m1 1985m1 1985m1
1986m1 1986m1 1986m1
Panel C: HML (Value factor)

1987m1 1987m1 1987m1
1988m1 1988m1 1988m1
1989m1 1989m1 1989m1
1990m1 1990m1 1990m1
1991m1 1991m1 1991m1
1992m1 1992m1 1992m1
Panel E: CMA (Investment factor)

1993m1 1993m1 1993m1
1994m1 1994m1 1994m1
1995m1 1995m1 1995m1
1996m1 1996m1 1996m1
1997m1 1997m1 1997m1
1998m1 1998m1 1998m1
1999m1 1999m1 1999m1
2000m1 2000m1 2000m1
2001m1 2001m1 2001m1
2002m1 2002m1 2002m1
Panel A: MKTRF (Market premium factor)
2003m1 2003m1 2003m1

2004m1 2004m1 2004m1
2005m1 2005m1 2005m1
2006m1 2006m1 2006m1
2007m1 2007m1 2007m1
2008m1 2008m1 2008m1
2009m1 2009m1 2009m1
2010m1 2010m1 2010m1
2011m1 2011m1 2011m1
2012m1 2012m1 2012m1
2013m1 2013m1 2013m1
2014m1 2014m1 2014m1
2015m1 2015m1 2015m1
2016m1 2016m1 2016m1
2017m1 2017m1 2017m1
2018m1 2018m1 2018m1
2019m1 2019m1 2019m1
2020m1 2020m1 2020m1
27
Cumulative Gross Return Cumulative Gross Return Cumulative Gross Return
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
0.00
0.50
1.00
1.50
2.00
2.50
3.00
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
1963m1 1963m1 1963m1

1964m1 1964m1 1964m1
1965m1 1965m1 1965m1
1966m1 1966m1 1966m1
1967m1 1967m1 1967m1
1968m1 1968m1 1968m1
1969m1 1969m1 1969m1
1970m1 1970m1 1970m1
1971m1 1971m1 1971m1
1972m1 1972m1 1972m1
1973m1 1973m1 1973m1
1974m1 1974m1 1974m1
1975m1 1975m1 1975m1
1976m1 1976m1 1976m1
1977m1 1977m1 1977m1
1978m1 1978m1 1978m1
1979m1 1979m1 1979m1
1980m1 1980m1 1980m1
1981m1 1981m1 1981m1
1982m1 1982m1 1982m1
1983m1 1983m1 1983m1
Figure 2: Cumulative benchmark factor returns
1984m1 1984m1 1984m1

Panel B: SMB (Size factor)
1985m1 1985m1 1985m1

1986m1 1986m1 1986m1
1987m1 1987m1 1987m1
1988m1 1988m1 1988m1

1989m1 1989m1 1989m1
1990m1 1990m1 1990m1
1991m1 1991m1 1991m1
1992m1 1992m1 1992m1
Panel F: UMD (Momentum factor)

1993m1 1993m1 1993m1
1994m1 1994m1 1994m1
Panel D: RMW (Profitability factor)
1995m1 1995m1 1995m1

1996m1 1996m1 1996m1
1997m1 1997m1 1997m1
1998m1 1998m1 1998m1
1999m1 1999m1 1999m1
2000m1 2000m1 2000m1
2001m1 2001m1 2001m1
2002m1 2002m1 2002m1
2003m1 2003m1 2003m1
2004m1 2004m1 2004m1
2005m1 2005m1 2005m1
2006m1 2006m1 2006m1
2007m1 2007m1 2007m1
2008m1 2008m1 2008m1
2009m1 2009m1 2009m1
2010m1 2010m1 2010m1
2011m1 2011m1 2011m1
2012m1 2012m1 2012m1
2013m1 2013m1 2013m1
2014m1 2014m1 2014m1
2015m1 2015m1 2015m1
2016m1 2016m1 2016m1
2017m1 2017m1 2017m1
2018m1 2018m1 2018m1
2019m1 2019m1 2019m1
2020m1 2020m1 2020m1
in grey indicate bear markets (see Table 1). Small red dots mark the cumulative gross return at the end of each
each bull and bear market. Cumulative returns reset to $1 at the start of each bear and bull market. Areas shaded
These figures depict cumulative gross returns from $1 investments in each of the Fama and French (2018) factors in
Figure 3: Factor bear market states
This figure plots bear market states for the factors of the Fama and French (2018) six-factor model, along with 11
individual bear markets (see Table 1). The factor bear markets are indicator variables identified by applying the
approach of Lunde and Timmermann (2004) to the cumulated returns of the individual factors.
UMD Bear phase

CMA Bear phase
RMW Bear phase
HML Bear phase
SMB Bear phase
MKTRF Bear phase
NBER Recession
COVID
GFC
Dotcom
Gulfwar
Crash1987
Correction83
Stagflation
Oilshock2
Oilshock1
Nixon
Vietnam
1963m1
1964m1
1965m1
1966m1
1967m1
1968m1
1969m1
1970m1
1971m1
1972m1
1973m1
1974m1
1975m1
1976m1
1977m1
1978m1
1979m1
1980m1
1981m1
1982m1
1983m1
1984m1
1985m1
1986m1
1987m1
1988m1
1989m1
1990m1
1991m1
1992m1
1993m1
1994m1
1995m1
1996m1
1997m1
1998m1
1999m1
2000m1
2001m1
2002m1
2003m1
2004m1
2005m1
2006m1
2007m1
2008m1
2009m1
2010m1
2011m1
2012m1
2013m1
2014m1
2015m1
2016m1
2017m1
2018m1
2019m1
2020m1
28
Figure 4: Value-weighted average durations of edge portfolios
These figures plot the value-weighted means of equity duration terciles across time for the long legs (Long) and
short legs (Short) of five non-market benchmark factors (Fama & French, 2018). Duration is estimated using the
method approach of Dechow et al. (2004). Stocks are ranked by their durations in each month and allocated to
three terciles. Tercile one (three) contains stocks with the shortest (longest) durations.
Panel A: SMB (Size factor) Panel B: HML (Value factor)

2.6
3
2.4
2.5
2.2
2
2
1.5
1.8
1.6
1970m1 1980m1 1990m1 2000m1 2010m1 2020m1 11970m1 1980m1 1990m1 2000m1 2010m1 2020m1
mth mth
Long Short Long Short
Panel C: RMW (Profitability factor) Panel D: CMA (Investment factor)

2.6
2.6
2.4
2.4
2.2
2.2
2
2
1.8
1.8
1.6
1.6
1970m1 1980m1 1990m1 2000m1 2010m1 2020m1 1970m1 1980m1 1990m1 2000m1 2010m1 2020m1
mth mth
Panel E: UMD (Momentum factor)

25
20
15
10
1970m1 1980m1 1990m1 2000m1 2010m1 2020m1

mth
Long Short
29
Figure 5: Value-weighted average financial constraints of edge portfolios
These figures plot the value-weighted means of equity financial constraint terciles across time for the long legs
(Long) and short legs (Short) of five non-market benchmark factors (Fama & French, 2018). We use the WW Index
(Whited & Wu, 2006) to measure financial constraints. Stocks are ranked by their financial constraint indices in
each month and allocated to three terciles. Tercile one (three) contains stocks with the lowest (highest) constraint
indices.

2.5
2
1.8
2
1.5
1.6
1.4
1
1960m1 1980m1 2000m1 2020m1 1960m1 1980m1 2000m1 2020m1

mth mth

2.2
2
1.9
2
1.8
1.8
1.7
1.6
1.6
1.5
1.4
1960m1 1980m1 2000m1 2020m1 1960m1 1980m1 2000m1 2020m1

mth mth

2
1.9
1.8
1.7
1.6
1.5
1960m1 1980m1 2000m1 2020m1

mth
Long Short
30
Table 2: FF6 factor returns in different states
Panel A reports averages of monthly factor returns for the six factors of Fama and French (2018) (in columns)
for the full sample and for different states. Returns are stated in percentage points. Bear and Bull markets are
defined in Table 1. BearLT and BullLT refer to bear and bull markets dated using the approach of Lunde and
Timmermann (2004). Upmarket and Downmarket refers to positive and negative market excess returns (measured
by MKTRF). Stablemarket is any month that is not a Crashmarket (when monthly market excess returns are below
-10%.) Recession and Expansion periods are from NBER. In addition we classify the first three months of 2020 as
a recession period because the smoothed U.S. recession probability (Piger & Chauvet, 2020) jumped to a level of
25.98 on January 1, 2020, which is similar to the 24.15 level in the first month of the GFC . Panel B reports averages
of monthly factor returns during each bear market. Significance levels are indicated by * (significant at the 10%
level), ** (significant at the 5% level) and *** (significant at the 1% level) based on Newey-West HAC adjusted
standard errors at a lag of 6 months.
N MKTRF SMB HML RMW CMA UMD
Panel A
Fullsample 681 0.51*** 0.21* 0.27** 0.25*** 0.27*** 0.66***
Bull 531 1.44*** 0.28** 0.05 0.16* 0.03 0.54***

Bear 150 -2.78*** -0.04 1.03*** 0.57** 1.10*** 1.09***
(Bull-Bear) 4.22*** 0.32 -0.98** -0.41 -1.07*** -0.54
BullLT 524 1.54*** 0.41*** 0.07 0.08 0.03 0.53***

BearLT 157 -2.95*** -0.47 0.91** 0.81*** 1.07*** 1.12***
(BullLT-BearLT) 4.49*** 0.88*** -0.84** -0.73** -1.04*** -0.60
Upmarket 407 3.29*** 0.80*** -0.16 -0.10 -0.23** 0.41

Downmarket 274 -3.62*** -0.67*** 0.90*** 0.77*** 1.01*** 1.03***
(Upmarket-Downmarket) 6.91*** 1.47*** -1.06*** -0.88*** -1.25*** -0.62
Stablemarket 667 0.79*** 0.29** 0.23** 0.21** 0.21*** 0.63***

Crashmarket 14 -12.97*** -3.83*** 1.90 2.04 3.00*** 2.12
(Stablemarket-Crashmarket) 13.76*** 4.12*** -1.67 -1.82 -2.79*** -1.49
Expansion 595 0.71*** 0.22* 0.28** 0.24** 0.21** 0.73***

Recession 86 -0.86 0.09 0.17 0.31 0.68* 0.17
(Expansion-Recession) 1.57* 0.13 0.11 -0.06 -0.47 0.56
Panel B
Vietnam 8 -2.41*** -0.01 -0.18 0.15 0.24 0.68
Nixon 19 -2.56*** -1.38** 0.58 0.23 1.08 0.81***
Oilshock1 21 -3.43*** -0.50 2.08*** -0.64 1.63*** 2.02***
Oilshock2 14 -1.08** 2.11*** 0.81 0.04 0.17 1.30***
Stagflation 20 -1.94*** 0.44 1.99*** -0.28 1.20*** 0.32
Correction83 11 -1.76*** -0.77*** 2.85*** 0.59* 1.47*** -0.93
Crash1987 3 -11.20*** -1.63* 2.55*** -0.27 1.66*** -2.77***
Gulfwar 5 -4.24*** -2.92*** 0.12 0.20 1.79*** 4.48***
Dotcom 30 -1.97*** 1.07** 1.78* 2.30*** 1.80*** 0.96
GFC 16 -4.26*** 0.05 -1.11 1.68*** 0.12 1.79***
COVID 3 -7.21*** -4.31*** -8.13*** -1.40*** -1.22** 4.59***
31
Table 3: Duration and financial constraints of edge portfolios
This table reports the value-weighted means of equity duration measures and financial constraint measures for the
long legs (Long), short legs (Short) and their difference (LMS) of five benchmark factors. Duration measures include
duration estimated based on Dechow et al. (2004) or Gonçalves (2021) and their terciles in each month (Duration
tercile and Goncalves duration tercile). Financial constraint indices include the WW Index based on Whited and
Wu (2006) and the HP Index based on Hadlock and Pierce (2010). Following the literature, firms are sorted into
terciles based on a financial constraint index in each year. Firms in the highest tercile are financially constrained
and those in the lowest tercile are unconstrained. p-values for H0 : LM S 6= 0 are enclosed in parentheses.
SMB HML RMW CMA UMD

A. Duration
Long 17.82 14.00 19.44 17.83 18.70
Short 18.51 21.13 17.76 19.60 18.64
LMS -0.69 -7.13 1.68 -1.77 0.06
( 0.006 ) (< 0.001 ) (< 0.001 ) (< 0.001 ) ( 0.800 )
B. Goncalves duration
Long 52.76 40.66 51.43 55.17 53.75
Short 50.63 64.69 61.72 56.19 54.00
LMS 2.14 -24.02 -10.29 -1.02 -0.25
( 0.040 ) (< 0.001 ) (< 0.001 ) ( 0.551 ) ( 0.863 )
C. Duration tercile
Long 1.96 1.25 2.30 1.99 2.14
Short 2.07 2.63 1.94 2.29 2.09
LMS -0.10 -1.38 0.36 -0.30 0.05
( 0.001 ) (< 0.001 ) (< 0.001 ) (< 0.001 ) ( 0.257 )
D. Goncalves duration tercile

Long 2.04 1.75 2.26 2.08 2.24
Short 2.27 2.49 2.20 2.35 2.15
LMS -0.23 -0.75 0.06 -0.28 0.10
(< 0.001 ) (< 0.001 ) ( 0.316 ) (< 0.001 ) ( 0.003 )
E. WW Index
Long 2.07 1.59 1.65 1.60 1.75
Short 1.28 1.78 1.77 1.83 1.71
LMS 0.81 -0.19 -0.07 -0.25 0.04
(< 0.001 ) (< 0.001 ) (< 0.001 ) (< 0.001 ) ( 0.136 )
F. HP Index
Long 1.82 1.26 1.47 1.37 1.49
Short 1.03 1.64 1.47 1.52 1.45
LMS 0.75 -0.33 0.03 -0.13 0.02
(< 0.001 ) (< 0.001 ) ( 0.865 ) (< 0.001 ) ( 0.036 )
32
Table 4: Duration and financial constraint factor returns in different states
Panel A reports averages of monthly factor returns for duration and financial constraint factors (in columns) for the
full sample and for different states. Returns are stated in percentage points. DUR (Dechow et al., 2004) and DURG
(Gonçalves, 2021) refer to strategies that buy short duration stocks and sell long duration stocks. FCWW (Whited
& Wu, 2006) and FCHP (Hadlock & Pierce, 2010) are strategies that buy financially constrained stocks and sell
financially unconstrained stocks. Following Fama and French (2018), we use a two-by-three sort to control for size.
Panel B reports averages of monthly factor returns during each bear market. Refer to Table 2 for definitions of the
different states (Panel A) and bear markets (Panel B). Significance levels are indicated by * (significant at the 10%
level), ** (significant at the 5% level) and *** (significant at the 1% level) based on Newey-West HAC adjusted
standard errors at a lag of 6 months.
N DUR DURG FCWW FCHP
Panel A
Fullsample 681 0.26* 0.46*** 0.10 0.03
Bull 531 0.02 0.11 0.30*** 0.25**

Bear 150 1.10*** 1.79*** -0.59*** -0.74***
(Bull-Bear) -1.08*** -1.68*** 0.88*** 1.00***
BullLT 524 0.05 0.09 0.32*** 0.25**

BearLT 157 0.95** 1.77*** -0.62*** -0.69***
(BullLT-BearLT) -0.91** -1.67*** 0.93*** 0.95***
Upmarket 407 -0.19 -0.22* 0.70*** 0.62***

Downmarket 274 0.92*** 1.49*** -0.78*** -0.84***
(Upmarket-Downmarket) -1.11*** -1.71*** 1.48*** 1.46***
Stablemarket 667 0.24** 0.38*** 0.15* 0.10

Crashmarket 14 1.05 4.21*** -2.08*** -2.99**
(Stablemarket-Crashmarket) -0.81 -3.83*** 2.23*** 3.09**
Expansion 595 0.26* 0.37*** 0.13 0.05

Recession 86 0.22 1.07*** -0.10 -0.07
(Expansion-Recession) 0.05 -0.70* 0.23 0.12
Panel B
Vietnam 8 -0.46 0.00 0.19 0.69*
Nixon 19 0.62 0.00 -0.60 -0.42
Oilshock1 21 2.05*** 1.92*** -0.66 -1.65***
Oilshock2 14 0.96* 0.53 0.59*** 0.45
Stagflation 20 1.38*** 1.61*** -1.07*** -1.11***
Correction83 11 2.44*** 2.50*** -1.52*** -2.21***
Crash1987 3 2.09*** 1.15*** -2.18*** -1.45***
Gulfwar 5 -1.31*** 0.44* -0.67*** 0.58***
Dotcom 30 2.53** 2.43*** -1.21*** -1.25
GFC 16 -0.42 1.85** 0.27 -0.42
COVID 3 -7.55*** 0.00 2.48*** 3.93***
33
Table 5: Factor correlations
The table below presents time-series correlations between factors in bear markets (Panel A), bull markets (Panel B) and the full sample (Panel C). Refer to
Tables 2 and 4 for the construction of factors. Table 1 describes bull and bear markets.
A. Bear
MKTRF SMB HML RMW CMA UMD DUR DURG FCWW FCHP
MKTRF 1.00
SMB 0.37*** 1.00
HML -0.28*** 0.01 1.00
RMW -0.35*** -0.18* 0.16* 1.00
CMA -0.40*** -0.10 0.71*** 0.23** 1.00
UMD -0.37*** -0.24** 0.14 0.43*** 0.28*** 1.00
DUR -0.29*** 0.06 0.78*** 0.33*** 0.64*** 0.29*** 1.00
DURG -0.48*** 0.05 0.67*** 0.37*** 0.64*** 0.19** 0.68*** 1.00
FCWW 0.50*** 0.35*** -0.55*** -0.46*** -0.60*** -0.27*** -0.54*** -0.49*** 1.00
FCHP 0.41*** 0.32*** -0.58*** -0.42*** -0.61*** -0.23** -0.66*** -0.59*** 0.63*** 1.00
B. Bull
MKTRF 1.00
34
SMB 0.24*** 1.00
HML -0.19*** -0.09* 1.00
RMW -0.17*** -0.48*** 0.10** 1.00
CMA -0.32*** -0.04 0.65*** -0.09* 1.00
UMD -0.04 0.05 -0.37*** -0.10** -0.14*** 1.00
DUR -0.19*** -0.15*** 0.83*** 0.37*** 0.52*** -0.40*** 1.00
DURG -0.33*** -0.07 0.48*** 0.23*** 0.38*** -0.06 0.50*** 1.00
FCWW 0.28*** 0.43*** -0.56*** -0.43*** -0.52*** 0.27*** -0.61*** -0.38*** 1.00
FCHP 0.23*** 0.49*** -0.63*** -0.48*** -0.48*** 0.36*** -0.72*** -0.33*** 0.71*** 1.00
C. Fullsample
MKTRF 1.00
SMB 0.23*** 1.00

HML -0.27*** -0.05 1.00
RMW -0.25*** -0.38*** 0.14*** 1.00
CMA -0.40*** -0.05 0.69*** 0.05 1.00
UMD -0.14*** -0.03 -0.19*** 0.09** 0.01 1.00
DUR -0.28*** -0.08* 0.82*** 0.37*** 0.58*** -0.17*** 1.00
DURG -0.44*** -0.02 0.57*** 0.30*** 0.51*** 0.04 0.58*** 1.00
FCWW 0.38*** 0.40*** -0.57*** -0.45*** -0.56*** 0.10** -0.60*** -0.44*** 1.00
FCHP 0.32*** 0.44*** -0.63*** -0.47*** -0.53*** 0.18*** -0.71*** -0.43*** 0.70*** 1.00
Table 6: Alphas from factor-on-factor regressions
Panel A reports alphas from from univariate regressions of the non-market FF6 factors (in columns) on one of the
duration and financial constraint factors (in rows). Panel B reports the alphas from univariate regressions where
we switch the dependent variable and independent variable in Panel A, so that duration and financial constraint
factors are each regressed separately on the FF6 non-market factors. Column headings are the dependent variables.
Row headings are the explanatory variables. Refer to Table 2 for definitions for bear markets. Significance levels
are indicated by * (significant at the 10% level), ** (significant at the 5% level) and *** (significant at the 1% level)
based on Newey-West HAC adjusted standard errors at a lag of 6 months.
Panel A: FF6 non-market factors regressed on duration and financial constraint factors
SMB HML RMW CMA UMD

A. Bear
Mean -0.04 1.03 *** 0.57 ** 1.10 *** 1.09 ***
DUR -0.18 0.03 0.32 0.58 *** 0.76 *
DURG 0.37 -0.11 0.07 0.30 0.24
FCWW 0.24 0.50 0.29 0.73 *** 0.80 **
FCHP 0.19 0.44 * 0.31 0.71 *** 0.84 **
B. Bull
Mean 0.28 ** 0.05 0.16 * 0.03 0.54 ***
DUR 0.28 ** 0.04 0.16 * 0.03 0.55 ***
DURG 0.19 -0.06 0.13 0.02 0.54 **
FCWW 0.08 0.26 ** 0.27 *** 0.17 ** 0.40 **
FCHP 0.13 0.21 ** 0.24 *** 0.12 * 0.41 **
C. Fullsample
Mean 0.21 * 0.27 ** 0.25 *** 0.27 *** 0.66 ***
DUR 0.22 * 0.05 0.20 ** 0.16 ** 0.73 ***
DURG 0.25 ** -0.03 0.15 0.11 0.57 ***
FCWW 0.14 0.34 *** 0.29 *** 0.32 *** 0.65 ***
FCHP 0.19 0.29 *** 0.26 *** 0.28 *** 0.65 ***
Panel B: Duration and financial constraint factors regressed on FF6 non-market factors
DUR DURG FCWW FCHP

A. Bear
Mean 1.10 *** 1.79 *** -0.59 *** -0.74 ***
SMB 1.11 *** 1.76 *** -0.58 *** -0.73 **
HML 0.33 * 1.03 *** -0.22 -0.24
RMW 0.92 ** 1.52 *** -0.40 ** -0.53 *
CMA 0.14 0.85 *** 0.01 0.05
UMD 0.97 ** 1.70 *** -0.48 ** -0.63 **
B. Bull
Mean 0.02 0.11 0.30 *** 0.25 **
SMB 0.05 0.12 0.21 *** 0.13
HML -0.03 0.11 0.32 *** 0.28 ***
RMW -0.04 0.08 0.36 *** 0.34 ***
CMA -0.01 0.09 0.31 *** 0.28 **
UMD 0.16 0.13 0.23 *** 0.13
C. Fullsample
Mean 0.26 ** 0.46 *** 0.10 0.03
SMB 0.27 ** 0.46 *** 0.04 -0.05
HML 0.04 0.33 *** 0.21 *** 0.19 **
RMW 0.17 0.37 *** 0.20 ** 0.15
CMA 0.03 0.27 *** 0.25 *** 0.23 **
UMD 0.34 *** 0.45 *** 0.07 -0.04
35
Table 7: Two-way regressions: duration and financial constraint factors
Panel A (B) reports alphas from univariate regressions of duration factors on financial constraint factors (from
regressing financial constraint factors on duration factors). Panel C (D) reports betas from regressing duration
factors on financial constraint factors (from regressing financial constraint factors on duration factors). Column
headings are the dependent variables. Row headings are the explanatory variables. Refer to Table 2 for definitions
for bear markets. Significance levels are indicated by * (significant at the 10% level), ** (significant at the 5% level)
and *** (significant at the 1% level) based on Newey-West HAC adjusted standard errors at a lag of 6 months.
Panel A: Alphas: duration factors regressed on fin- Panel B: Alphas: financial constraint factors re-
ancial constraint factors gressed on duration factors
DUR DURG FCWW FCHP

A. Bear A. Bear
Mean 1.10 *** 1.79 *** Mean -0.59 *** -0.74 ***
FCWW 0.62 ** 1.40 *** DUR -0.16 -0.09
FCHP 0.51 ** 1.24 *** DURG 0.08 0.13
B. Bull B. Bull
Mean 0.02 0.11 Mean 0.30 *** 0.25 **
FCWW 0.24 ** 0.23 ** DUR 0.30 *** 0.26 ***
FCHP 0.19 ** 0.18 * DURG 0.35 *** 0.33 ***
C. Fullsample C. Fullsample
Mean 0.26 ** 0.46 *** Mean 0.10 0.03
FCWW 0.33 *** 0.52 *** DUR 0.21 *** 0.20 ***
FCHP 0.28 *** 0.47 *** DURG 0.30 *** 0.27 **
Panel C: Betas: duration factors regressed on fin- Panel D: Betas: financial constraint factors re-
ancial constraint factors gressed on duration factors
DUR DURG FCWW FCHP

A. Bear A. Bear
FCWW -0.82 *** -0.61 *** DUR -0.38 *** -0.59 ***
FCHP -0.80 *** -0.60 *** DURG -0.40 *** -0.59 ***
B. Bull B. Bull
FCWW -0.74 *** -0.38 *** DUR -0.43 *** -0.70 ***
FCHP -0.71 *** -0.25 *** DURG -0.38 *** -0.42 ***
C. Fullsample C. Fullsample
FCWW -0.78 *** -0.49 *** DUR -0.43 *** -0.67 ***
FCHP -0.74 *** -0.37 *** DURG -0.40 *** -0.50 ***
36
Table 8: Long and short leg returns in different states
This table reports means of monthly returns to portfolios on the long legs (suffix _L) and short legs (suffix _S) of five benchmark factors in different states.
These portfolio returns are constructed from the 2 by 3 sorted portfolio returns for each factor as sourced from Ken French’s data library. Refer to Table 2
for the definitions of states. Significance levels are indicated by * (significant at the 10% level), ** (significant at the 5% level) and *** (significant at the
1% level) based on Newey-West HAC adjusted standard errors at a lag of 6 months.
N SMB_L SMB_S HML_L HML_S RMW_L RMW_S CMA_L CMA_S UMD_L UMD_S
Fullsample 681 1.11*** 0.90*** 1.15*** 0.88*** 1.09*** 0.84*** 1.13*** 0.87*** 1.30*** 0.64**
Bull 531 2.04*** 1.77*** 1.95*** 1.90*** 2.00*** 1.84*** 1.97*** 1.94*** 2.25*** 1.70***
Bear 150 -2.20*** -2.16*** -1.70*** -2.74*** -2.13*** -2.70*** -1.82*** -2.92*** -2.04*** -3.13***
(Bull-Bear) 4.25*** 3.93*** 3.66*** 4.64*** 4.13*** 4.54*** 3.79*** 4.86*** 4.29*** 4.83***
37
BullLT 524 2.27*** 1.86*** 2.12*** 2.05*** 2.13*** 2.05*** 2.13*** 2.10*** 2.41*** 1.89***
BearLT 157 -2.76*** -2.29*** -2.10*** -3.01*** -2.38*** -3.20*** -2.20*** -3.27*** -2.39*** -3.52***
(BullLT-BearLT) 5.02*** 4.14*** 4.22*** 5.06*** 4.51*** 5.24*** 4.33*** 5.37*** 4.80*** 5.40***
Upmarket 407 4.39*** 3.59*** 4.00*** 4.16*** 4.06*** 4.16*** 4.04*** 4.27*** 4.43*** 4.01***
Downmarket 274 -3.77*** -3.10*** -3.09*** -3.99*** -3.33*** -4.10*** -3.18*** -4.19*** -3.34*** -4.37***
(Upmarket-Downmarket) 8.16*** 6.70*** 7.09*** 8.15*** 7.39*** 8.26*** 7.21*** 8.46*** 7.77*** 8.39***
Stablemarket 667 1.46*** 1.17*** 1.45*** 1.22*** 1.40*** 1.18*** 1.43*** 1.22*** 1.63*** 0.99***
Crashmarket 14 -15.73*** -11.90*** -13.24*** -15.15*** -13.68*** -15.72*** -13.01*** -16.01*** -14.08*** -16.21***
(Stablemarket-Crashmarket) 17.19*** 13.07*** 14.69*** 16.37*** 15.08*** 16.91*** 14.44*** 17.23*** 15.71*** 17.20***

Expansion 595 1.30*** 1.08*** 1.34*** 1.06*** 1.27*** 1.03*** 1.29*** 1.08*** 1.52*** 0.78***
Recession 86 -0.23 -0.32 -0.18 -0.35 -0.20 -0.50 0.08 -0.60 -0.18 -0.36
(Expansion-Recession) 1.53 1.40* 1.52 1.41 1.47 1.53 1.20 1.68 1.70* 1.14
Table 9: Factor earnings announcement returns in different states
This table reports monthly averages of factor earnings announcement returns (indicated by _rea) in different states along with the means of replicated
factors. In order to identify constituent stocks in factor portfolios, we replicate all five non-market factors following the approach in Fama and French (2018).
The correlations between our constructed factors and those from Ken French’s data library are all above 0.97. Refer to Table 2 for the definitions of states.
Significance levels are indicated by * (significant at the 10% level), ** (significant at the 5% level) and *** (significant at the 1% level) based on Newey-West
HAC adjusted standard errors at a lag of 6 months.
N SMB SMB_rea HML HML_rea RMW RMW_rea CMA CMA_rea UMD UMD_rea
Fullsample 681 0.21* 0.01 0.27** 0.04** 0.25*** 0.02 0.27*** 0.02 0.66*** 0.07***
Bull 531 0.28** 0.01 0.05 0.02 0.16* -0.00 0.03 0.01 0.54*** 0.05**
Bear 150 -0.04 -0.01 1.03*** 0.09 0.57** 0.08** 1.10*** 0.04 1.09*** 0.13***
(Bull-Bear) 0.32 0.02 -0.98** -0.06 -0.41 -0.08* -1.07*** -0.03 -0.54 -0.08*
38
BullLT 524 0.41*** 0.02 0.07 0.02 0.08 -0.00 0.03 0.01 0.53*** 0.05**
BearLT 157 -0.47 -0.02 0.91** 0.11* 0.81*** 0.08* 1.07*** 0.04 1.12*** 0.13***
(BullLT-BearLT) 0.88*** 0.04 -0.84** -0.09 -0.73** -0.08* -1.04*** -0.03 -0.60 -0.08*
Upmarket 407 0.80*** 0.04*** -0.16 0.01 -0.10 0.01 -0.23** -0.00 0.41 0.04*
Downmarket 274 -0.67*** -0.04** 0.90*** 0.07* 0.77*** 0.03 1.01*** 0.04* 1.03*** 0.11***
(Upmarket-Downmarket) 1.47*** 0.08*** -1.06*** -0.06 -0.88*** -0.03 -1.25*** -0.04 -0.62 -0.07*
Stablemarket 667 0.29** 0.02 0.23** 0.04** 0.21** 0.01 0.21*** 0.02 0.63*** 0.06***
Crashmarket 14 -3.83*** -0.34** 1.90 -0.04 2.04 0.33* 3.00*** -0.00 2.12 0.42*
(Stablemarket-Crashmarket) 4.12*** 0.36** -1.67 0.08 -1.82 -0.32* -2.79*** 0.02 -1.49 -0.36

Expansion 595 0.22* 0.00 0.28** 0.04*** 0.24** 0.00 0.21** 0.02 0.73*** 0.07***
Recession 86 0.09 0.06* 0.17 -0.02 0.31 0.10* 0.68* -0.01 0.17 0.08
(Expansion-Recession) 0.13 -0.06* 0.11 0.07 -0.06 -0.10 -0.47 0.03 0.56 -0.02
9 Appendix
9.1 Factor time-series description
39
Table 10: Factor time-series definitions
The table below details factor time series used in this paper. Panel A contains the factors sourced from Ken French’s data library, Panel B contains factors
sourced from Lu Zhang’s data library and Panel C contains data sourced from the AQR data library. All series are monthly returns. Data source URL’s are
provided at the bottom of the table.
Panel A: From Ken French’s data library

Factor Description Sorted on
RF Risk free rate (t-bills) n/a
MKTRF Market return less risk free rate n/a
SMB Small - minus - Big size factor size = #shares x share price
HML High - minus - Low book-to-market value factor book2market = book value / market value, and size
RMW Robust - minus - Weak profitability factor profitability = net income / book value, and size
CMA Conservative - minus - Aggressive investment factor investment = change in assets / lagged assets, and size
UMD Up - minus - Down momentum factor momentum = past t-2 to t-12 month cumulative returns
STREV Short term reversal factor prior month return
LTREV Long term reversal factor historical returns = returns from t-60 to t-13
Panel B: From Lu Zhang’s data library
40
MKT Market return less risk free rate n/a
ME Market Equity size factor #shares x share price
I2A Investment to Asset factor i2a = annual change in assets / total assets
ROE Return On Equity factor roe = net income / book value (using most recent monthly data)
EG Expected Growth factor expected growth from cross-Sectional predictive regressions
Panel C: From AQR’s data library


HMLdevil High - minus - Low book-to-market value factor book2market = book value / market value (calculated monthly)
BAB Betting Against Beta factor beta (with leverage overlay), see Frazzini and Pedersen (2014)
QMJ Quality - minus - Junk factor quality measure, see Asness et al. (2017)
Sources:
Ken French’s data library: https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html

Lu Zhang’s data library: http://global-q.org/factors.html
AQR’s data library: https://www.aqr.com/Insights/Datasets
9.2 Hou et al. (2020) factor results
Table 11: Hou et al. (2020) benchmark factor returns in different states
Panel A reports averages of monthly factor returns for the five factors of Hou et al. (2020) (in columns) for the
full sample and for different states. Returns are stated in percentage points. The number of months corresponding
to each state is indicated in the first column (heading N). Bear and Bull markets are defined in Table 1. BearLT
and BullLT refer to bear and bull markets dated using the approach of Lunde and Timmermann (2004). Upmarket
and Downmarket refers to positive and negative market excess returns (measured by MKTRF). Stablemarket is any
month that is not a Crashmarket (when monthly market excess returns are below -10%.) Recession and Expansion
periods are from NBER. In addition we classify the first three months of 2020 as a recession period because the
smoothed U.S. recession probability (Piger & Chauvet, 2020) jumped to a level of 25.98 on January 1, 2020, which
is similar to the 24.15 level in the first month of the GFC . Panel B reports averages of monthly factor returns
during each bear market. Significance levels are indicated by * (significant at the 10% level), ** (significant at the
5% level) and *** (significant at the 1% level) based on Newey-West HAC adjusted standard errors at a lag of 6
months.
N MKT ME I2A ROE EG
Panel A
Fullsample 681 0.53*** 0.27** 0.36*** 0.54*** 0.81***
Bull 531 1.44*** 0.30** 0.13* 0.46*** 0.58***

Bear 150 -2.71*** 0.17 1.18*** 0.83*** 1.64***
(Bull-Bear) 4.15*** 0.14 -1.05*** -0.37 -1.06***
BullLT 524 1.56*** 0.43*** 0.14* 0.39*** 0.53***

BearLT 157 -2.90*** -0.26 1.12*** 1.04*** 1.76***
(BullLT-BearLT) 4.46*** 0.69** -0.99*** -0.65** -1.22***
Upmarket 407 3.35*** 0.86*** -0.09 0.24 0.26**

Downmarket 274 -3.66*** -0.59*** 1.04*** 0.98*** 1.64***
(Upmarket-Downmarket) 7.01*** 1.44*** -1.13*** -0.73*** -1.38***
Stablemarket 667 0.82*** 0.35*** 0.31*** 0.51*** 0.75***

Crashmarket 14 -13.03*** -3.32** 3.05*** 2.00* 3.81***
(Stablemarket-Crashmarket) 13.85*** 3.67*** -2.74*** -1.49 -3.06***
Expansion 595 0.71*** 0.26* 0.30*** 0.56*** 0.74***

Recession 86 -0.66 0.39 0.76** 0.41 1.33***
(Expansion-Recession) 1.37* -0.13 -0.46 0.15 -0.59**
Panel B
Vietnam 8 0.00 0.00 0.00 0.00 0.00
Nixon 19 -2.54*** -1.05** 1.30* 0.86** 1.73***
Oilshock1 21 -3.42*** -0.16 1.41*** 0.34 2.12***
Oilshock2 14 -1.08** 2.17*** 0.34** 0.68** 1.31***
Stagflation 20 -2.03*** 0.75 1.33*** 0.26 1.43***
Correction83 11 -1.75*** -0.98*** 1.27*** 0.25 1.49***
Crash1987 3 -11.09*** -1.15 1.95*** 0.15 1.75***
Gulfwar 5 -4.06*** -2.72*** 2.16*** 1.08*** 1.37***
Dotcom 30 -1.94*** 1.00* 1.65*** 1.37* 1.63***
GFC 16 -4.36*** -0.07 -0.09 1.67*** 1.64***
COVID 3 0.00 0.00 0.00 0.00 0.00
41
9.3 Other factor results
Table 12: Other factor returns in different states

Panel A reports averages of monthly factor returns for a variety of other factors (in columns) for the full sample and
for different states. Returns are stated in percentage points. The number of months corresponding to each state is
indicated in the first column (heading N). Bear and Bull markets are defined in Table 1. BearLT and BullLT refer
to bear and bull markets dated using the approach of Lunde and Timmermann (2004). Upmarket and Downmarket
refers to positive and negative market excess returns (measured by MKTRF). Stablemarket is any month that is
not a Crashmarket (when monthly market excess returns are below -10%.) Recession and Expansion periods are
from NBER. In addition we classify the first three months of 2020 as a recession period because the smoothed U.S.
recession probability (Piger & Chauvet, 2020) jumped to a level of 25.98 on January 1, 2020, which is similar to
the 24.15 level in the first month of the GFC . Panel B reports averages of monthly factor returns during each bear
market. Significance levels are indicated by * (significant at the 10% level), ** (significant at the 5% level) and ***
(significant at the 1% level) based on Newey-West HAC adjusted standard errors at a lag of 6 months.
N RF STREV LTREV HMLdevil BAB QMJ
Panel A
Fullsample 681 0.38*** 0.48*** 0.18 0.22 0.39*** 0.80***
Bull 531 0.34*** 0.52*** -0.01 0.04 0.15 0.84***

Bear 150 0.51*** 0.32 0.86*** 0.88** 1.25*** 0.66
(Bull-Bear) -0.17*** 0.20 -0.86*** -0.84* -1.10*** 0.18
BullLT 524 0.33*** 0.61*** 0.08 0.12 0.03 0.86***

BearLT 157 0.52*** 0.02 0.53* 0.59 1.58*** 0.61
(BullLT-BearLT) -0.19*** 0.60* -0.44 -0.47 -1.55*** 0.24
Upmarket 407 0.35*** 0.91*** 0.17 -0.09 -0.38*** 0.55***

Downmarket 274 0.42*** -0.17 0.20 0.69*** 1.53*** 1.17***
(Upmarket-Downmarket) -0.07*** 1.09*** -0.02 -0.79** -1.90*** -0.62*
Stablemarket 667 0.38*** 0.54*** 0.17 0.20 0.29*** 0.83***

Crashmarket 14 0.48*** -2.69* 0.99 1.60 5.05*** -0.61
(Stablemarket-Crashmarket) -0.10 3.24** -0.82 -1.40 -4.76*** 1.44
Expansion 595 0.36*** 0.45*** 0.09 0.17 0.33*** 0.91***

Recession 86 0.52*** 0.66 0.81** 0.60 0.80** 0.00
(Expansion-Recession) -0.17** -0.20 -0.71* -0.43 -0.47 0.91*
Panel B
Vietnam 8 0.38*** 0.58 0.53** 0.08 0.24 -0.69***
Nixon 19 0.54*** 0.14 1.39 0.51 1.59*** 0.39
Oilshock1 21 0.60*** 1.00** 1.17** 1.96*** 0.24 -0.67**
Oilshock2 14 0.43*** 0.63*** 0.62* 0.74 -0.05 1.27***
Stagflation 20 1.10*** 0.15 2.05*** 1.97*** 0.61*** 1.80***
Correction83 11 0.75*** 1.12*** 0.59** 2.96*** 0.97*** 1.26***
Crash1987 3 0.47*** -0.38*** -1.02 3.47*** 1.43*** -2.76**
Gulfwar 5 0.65*** -2.57*** -0.92*** -1.24*** 2.83*** -0.86
Dotcom 30 0.25*** 1.21 1.42*** 1.63 2.03*** 3.35***
GFC 16 0.14*** -0.96 -0.15 -1.83** 2.72*** -2.23***
COVID 3 0.12*** -3.81** -5.08*** -8.66*** 1.63 -3.03*
42
9.4 Construction of financial constraint indices
We follow Farre-Mensa and Ljungqvist (2016) in constructing the two financial constraint indices.
Refer to Farre-Mensa and Ljungqvist (2016), p305.
The WW index is constructed as –0.091[(ib + dp)/at]–0.062[indicator set to one if dvc + dvp is

positive, and zero otherwise]+0.021[dltt/at]–0.044[log(at)] + 0.102[average industry sales growth,
estimated separately for each three-digit SIC industry and each year, with sales growth defined as
above]–0.035[salesgrowth].
The HP index is constructed as –0.737Size + 0.043Size2 − 0.040Age, where Size equals the log of
inflation-adjusted Compustat item at (in 2004 dollars), and Age is the number of years the firm is
listed with a non-missing stock price on Compustat. In calculating the index, we follow Hadlock
and Pierce (2010) and cap Size at (the log of) $4.5 billion and Age at 37 years.
Following convention, firms are sorted into terciles based on their index values in the previous
year. Firms in the top tercile are coded as constrained and those in the bottom tercile are coded
as unconstrained.
43
9.5 Value-weighted average durations of edge portfolios
Figure 6: Value-weighted average durations of edge portfolios

These figures plot the value-weighted means of equity duration terciles across time for the long legs (Long) and
short legs (Short) of five non-market benchmark factors (Fama & French, 2018). The duration measure is from
Gonçalves (2021). Stocks are ranked by their durations in each month and allocated to three terciles. Tercile one
(three) contains stocks with the shortest (longest) durations.
3
2.6
2.5
2.4
2
2.2
1.5
2
1.8
1970m1 1980m1 1990m1 2000m1 2010m1 2020m1 1970m1 1980m1 1990m1 2000m1 2010m1 2020m1
mth mth

2.6
2.6
2.4
2.4
2.2
2.2
2
2
1.8
1.8
1.6
1970m1 1980m1 1990m1 2000m1 2010m1 2020m1 1970m1 1980m1 1990m1 2000m1 2010m1 2020m1
mth mth

2.6
2.4
2.2
2
1.8
1.6
1970m1 1980m1 1990m1 2000m1 2010m1 2020m1

mth
Long Short
44
9.6 Value-weighted average financial constraints of edge portfolios
Figure 7: Value-weighted average financial constraints of edge portfolios

These figures plot the value-weighted means of equity financial constraint terciles across time for the long legs
(Long) and short legs (Short) of five non-market benchmark factors (Fama & French, 2018). We use the HP Index
(Hadlock & Pierce, 2010) to measure financial constraints. Stocks are ranked by their financial constraint indices in
each month and allocated to three terciles. Tercile one (three) contains stocks with the lowest (highest) constraint
indices.

2.5
2
1.8
2
1.6
1.5
1.4
1.2
1
1960m1 1980m1 2000m1 2020m1 1960m1 1980m1 2000m1 2020m1

mth mth
Panel C:RMW (Profitability factor) Panel D: CMA (Investment factor)

1.8
1.8
1.6
1.6
1.4
1.4
1.2
1.2
1960m1 1980m1 2000m1 2020m1 1960m1 1980m1 2000m1 2020m1

mth mth

1.8
1.7
1.6
1.5
1.4
1.3
1960m1 1980m1 2000m1 2020m1

mth
Long Short
45

SSRN Id3966127

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SSRN Id3966127

Uploaded by

Copyright:

Available Formats

Where is the risk in risk factors?

Paul Geertsema∗ and Helen Lu†

First version: 29 May 2020

Figure 1: Average monthly returns of factors in bear and bull markets

(%) Fullsample Bear Bull

1.2 1.10 1.09

2.2 Investment and profitability

[Insert Table 1 here]

4 Factor performance in bear and bull markets

[Insert Table 2 here]

[Insert Figure 2 here]

[Insert Figure 3 here]

5.1 Duration and financial constraints in factor edge portfolios

[Insert Table 3 here]

[Insert Figure 4 here]

[Insert Figure 7 here]

5.2 Duration and financial constraint factors

[Insert Table 4 here]

[Insert Table 5 here]

[Insert Table 6 here]

[Insert Table 7 here]

[Insert Table 8 here]

[Insert Table 9 here]

Panel C: HML (Value factor)

Panel E: CMA (Investment factor)

2003m1 2003m1 2003m1

1963m1 1963m1 1963m1

1984m1 1984m1 1984m1

1985m1 1985m1 1985m1

Electronic copy available at: https://ssrn.com/abstract=3966127

Panel F: UMD (Momentum factor)

1995m1 1995m1 1995m1

UMD Bear phase

Panel A: SMB (Size factor) Panel B: HML (Value factor)

Long Short Long Short

Panel C: RMW (Profitability factor) Panel D: CMA (Investment factor)

Long Short Long Short

Panel E: UMD (Momentum factor)

1970m1 1980m1 1990m1 2000m1 2010m1 2020m1

Panel A: SMB (Size factor) Panel B: HML (Value factor)

1960m1 1980m1 2000m1 2020m1 1960m1 1980m1 2000m1 2020m1

Long Short Long Short

Panel C: RMW (Profitability factor) Panel D: CMA (Investment factor)

1960m1 1980m1 2000m1 2020m1 1960m1 1980m1 2000m1 2020m1

Long Short Long Short

Panel E: UMD (Momentum factor)

1960m1 1980m1 2000m1 2020m1

Bull 531 1.44*** 0.28** 0.05 0.16* 0.03 0.54***

BullLT 524 1.54*** 0.41*** 0.07 0.08 0.03 0.53***

Upmarket 407 3.29*** 0.80*** -0.16 -0.10 -0.23** 0.41

Stablemarket 667 0.79*** 0.29** 0.23** 0.21** 0.21*** 0.63***

Expansion 595 0.71*** 0.22* 0.28** 0.24** 0.21** 0.73***

SMB HML RMW CMA UMD

D. Goncalves duration tercile

Bull 531 0.02 0.11 0.30*** 0.25**

BullLT 524 0.05 0.09 0.32*** 0.25**

Upmarket 407 -0.19 -0.22* 0.70*** 0.62***

Stablemarket 667 0.24** 0.38*** 0.15* 0.10

Expansion 595 0.26* 0.37*** 0.13 0.05

Electronic copy available at: https://ssrn.com/abstract=3966127

SMB HML RMW CMA UMD

DUR DURG FCWW FCHP

DUR DURG FCWW FCHP

Bull 531 1.44* 0.28 0.05 0.16* 0.03 0.54***

BullLT 524 1.54* 0.41* 0.07 0.08 0.03 0.53***

Upmarket 407 3.29* 0.80* -0.16 -0.10 -0.23** 0.41

Stablemarket 667 0.79* 0.29 0.23 0.21 0.21* 0.63*

Expansion 595 0.71*** 0.22* 0.28 0.24 0.21 0.73*

Bull 531 0.02 0.11 0.30* 0.25

BullLT 524 0.05 0.09 0.32* 0.25

Upmarket 407 -0.19 -0.22* 0.70* 0.62*

Stablemarket 667 0.24 0.38* 0.15* 0.10

Bull 531 1.44* 0.30 0.13* 0.46* 0.58*

BullLT 524 1.56* 0.43* 0.14* 0.39* 0.53*

Upmarket 407 3.35* 0.86* -0.09 0.24 0.26**

Stablemarket 667 0.82* 0.35* 0.31* 0.51* 0.75***

Expansion 595 0.71*** 0.26* 0.30* 0.56* 0.74***

Bull 531 0.34* 0.52* -0.01 0.04 0.15 0.84***

BullLT 524 0.33* 0.61* 0.08 0.12 0.03 0.86***

Upmarket 407 0.35* 0.91* 0.17 -0.09 -0.38* 0.55*

Stablemarket 667 0.38* 0.54* 0.17 0.20 0.29* 0.83*

Expansion 595 0.36* 0.45* 0.09 0.17 0.33* 0.91*