Professional Documents
Culture Documents
iew
Hilal Anwar Butta, James W. Kolarib, and Mohsin Sadaqatc
Abstract
Momentum profits collapse and reversal occurs when preceding market volatility is relatively high. Based
ev
on these intertemporal patterns, we implement an investment strategy that switches from momentum to
reversal when volatility is high. The proposed switching strategy has two advantages over scaled
momentum strategies: (1) the leverage factor is constant, and (2) no ex-post information is used to control
for volatility. In U.S. stock market tests across a variety of performance metrics, the switching strategy
r
distinguishes itself from traditional and volatility scaled momentum strategies by eliminating losses due to
momentum crashes. Further evidence confirms that the switching strategy is successful in other developed
and emerging stock markets, especially in Japanese and Chinese stock markets.
___________________
ot
bJPMorgan Chase Professor of Finance, Texas A&M University, Mays Business School, David C. Sinn
Department of Finance, College Station, TX 77843-4218 USA
Email address: j-kolari@tamu.edu
rin
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Momentum, Market Volatility, and Reversal
ed
1. Introduction
A momentum strategy buys previous winners and finances this long position by short selling the previous
iew
losers to earn significant future returns (Jegadeesh and Titman, 1993). This simple investment strategy has
attracted considerable interest in the literature. For example, it is commonly used as a systematic risk factor
in asset pricing models (e.g., Carhart, 1997; and Fama and French, 2018, 2020) as well as investment
ev
strategy by securities companies. Given efficient markets, a vexing problem is explaining the persistence
of momentum profits. Recent work by Guo, Li, and Li (2022) investigated a panoply of competing
r
explanations for momentum, including the anchoring effect, fundamental-related concepts, prospect theory,
firm characteristics related to limits to arbitrage and information uncertainty, and a number of other possible
er
explanatory variables. Unfortunately, they found that these variables only explained 31% of momentum
returns, which leaves 69% unexplained. Another study by Goyal, Jegadeesh and Subrahmanyam (2022) of
pe
international markets concluded that underreaction by investors is the most consistent explanation of
momentum among numerous other possible explanations. Consistent with these studies’ findings, Fama
and French (2008) has observed that momentum is the premier puzzle in financial economics.
ot
Momentum strategies are unique in that they experience frequent crashes. Crashes typically occur in
down market states (DMS) when market volatility is relatively high and market returns contemporaneously
tn
increase (Daniel and Moskowitz, 2016; and Guo, Li, and Li, 2022). This negative relationship between
market volatility and momentum returns was first recognized by Wang and Xu (2015). Well-known studies
rin
by Barroso and Santa-Clara (2015) and Daniel and Moskowitz (2016) showed that crash losses can be
mitigated via scaling momentum returns by the inverse of trailing market volatility.1 Additionally, they
ep
found that scaling improves both Sharpe ratios and momentum returns. Although volatility is used as a
scaling factor on an ex-ante basis, it substantially increases the leverage and overall volatility of the
Pr
1As shown by Barroso and Santa-Clara (2015), lagged market volatility produces the same results as the lagged
volatility of the momentum strategy.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
momentum strategy. Higher leverage and volatility can only be controlled by scaling further through a
ed
constant that is selected on an ex-post basis.2 However, because this approach makes scaled strategies
impractical for investment purposes, Bongaerts, Kang and Dijk (2020) suggested adjustments to make
iew
scaled strategies feasible.3
In this study we propose an investment strategy based on the state of current market volatility in
conjunction with historical volatility to enable a switching strategy between momentum and short-term
reversal (hereafter switching strategy). Short-term reversals have been studied extensively in the U.S.
ev
market, and excess returns associated with this strategy have been documented in different time periods
(e.g., see Lehman, 1990; Jegadeesh, 1990; Bali, Engle, and Murray, 2016; and Medhat and Schmeling,
r
2022).4 The success of the switching strategy relies on the relationship between preceding market volatility
er
and future momentum plus reversal returns. In this regard, historical volatility predicts negative momentum
returns, and this relationship is stronger when the level of volatility is higher. For instance, previous studies
pe
(Barroso and Santa-Clara, 2015; Daniel and Moskowitz, 2016; and Butt, Kolari, and Sadaqat, 2021) have
documented that most momentum crashes occur when market volatility is relatively high.5 One possible
reason is that the demand for liquidity increases at such times, which causes momentum returns to decline
ot
(Jylha, Rinne, and Suominen, 2013; and Ignashkina, Rinne, and Suominen, 2022).
Using market volatility as an indicator of depressed market conditions, Butt, Högholm, and Sadaqat
tn
(2021) for emerging markets and Nagel (2012) for the U.S. market have shown that reversal-related returns
increase when market volatility is relatively high. Some researchers have rationalized this positive
rin
2 Barroso and Santa-Clara (2015) used a constant such that overall volatility of the scaled momentum strategy
remained at 12% on an annual basis. Similarly, Daniel and Moskowitz (2016) used a constant to set the overall
volatility of the scaled strategy equal to the original momentum strategy.
ep
3 The authors proposed a momentum strategy that only scales returns associated with extreme volatility-related states.
4See also studies on long-run reversion by Fama and French (1988), Lo and MacKinlay (1988), Poterba and
Summers (1988), and Kiojen, Rodriguez, and Sbuelz (2009).
Pr
5 For example, Butt, Sadaqat, and Tahir (2022) found that down-market states (with negative cumulative returns over
the last consecutive 24 months) are concentrated in the 5th quintile of the market volatility in which most momentum
crashes occur.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
relationship between the market volatility and reversal returns in the context of liquidity costs, which
ed
increase when the market is depressed (e.g., see Nagel, 2012; Jylha, Rinne, and Suominen, 2013; Cheng,
Hameed, Subrahmanyam, and Titman, 2017; Ignashkina, Rinne, and Suominen, 2022; Medhat and
iew
Schmeling, 2022; among others). More explicitly stated, reversal returns are compensation for providing
market liquidity to those who demand immediacy to trade, such as momentum traders, in adverse market
conditions. The switching strategy capitalizes these intertemporal patterns in momentum and reversal
ev
Extending previous studies, we implement a switching strategy that changes from momentum to
reversal in month t+1 when the volatility of current month t lies in the fourth quartile of historical volatility
r
over the last five years. If the market volatility of the current month belongs to the first three quarters of
er
historical market volatility, the momentum strategy is continued. Two significant advantages of this strategy
are that, unlike scaled strategies, the leverage factor remains constant, and no ex-post information is used
pe
to control for volatility. The switching strategy is a self-financing, zero-investment strategy with one dollar
long and short. Our empirical findings indicate that the strategy outperforms traditional momentum and
reversal strategies in the U.S. stock market. Additionally, the switching strategy outperforms a scaled
ot
momentum strategy based on volatility when the scaling factors are chosen on an ex-ante basis. For
example, in the period 1927:1 to 2020:4, the Sharpe ratios for momentum, reversal, and scaled momentum
tn
strategies are 13.22, 14.81, and 19.72, respectively; by comparison, the Sharpe ratio for the switching
strategy is 22.64. Subsequently, the end-of-period investment size of one dollar is greater for the switching
rin
strategy than other strategies. The main reason for this outperformance is avoidance of momentum crashes
that occur when market-based volatility is relatively high. Further support for the switching strategy is
evidenced by higher alphas in spanning regressions of the switching strategy on all other momentum
ep
strategies.
As already mentioned, volatility scaling substantially decreases the extent of momentum crashes. In
Pr
our sample period, the 15 worst crashes had a monthly average of -36.30%. Scaling by different measures
of volatility reduced the magnitude of these crashes to -14.70%. Strikingly, we find that the switching
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
strategy eliminates these losses by earning on average 2.30% per month in these crash periods. Also, for
ed
the two worst U.S. crash periods, the switching strategy performed substantially better than the scaled
momentum strategy.
iew
Extending our analyses to international stock markets, we test the efficacy of the switching strategy in
five highly capitalized developed and five emerging markets. For all ten stock markets, the switching
strategy performs better than traditional momentum, and for seven countries the switching strategy
performs better than volatility scaled momentum strategy. The results for Japan as a developed country and
ev
China as an emerging country add new insights in view of the fact that previous studies have not detected
momentum in these markets. For these countries, our switching strategy generates monthly returns of 1.19%
r
and of 1.60%, respectively.
er
The rest of the paper is organized as follows. Section 2 describes the relationship between market
volatility and returns on momentum and reversal strategies. Section 3 compares volatility scaled momentum
pe
to the proposed switching strategy. Section 4 documents momentum crashes and the performance of various
momentum strategies. Section 5 presents out-of-sample evidence for other developed markets in addition
According to previous studies by Grundy and Martin (2001), Cooper, Gutierrez, and Hameed (2004), and
others, state dependency plays an important role in explaining momentum returns. After controlling for
rin
market states, Wang and Xu (2015) found that higher market volatility predicts lower momentum returns.
Similarly, Barroso and Santa-Clara (2015) and Daniel and Moskowitz (2016) reported a negative and
predictive relationship between market volatility and momentum returns. They proposed scaling
ep
momentum strategies by the inverse of the volatility to mitigate downside risk (including crashes) as well
as boost Sharpe ratio and alpha performance metrics. Volatility scaling has been shown to enhance
Pr
momentum strategies outside the U.S. also (e.g., see Hanauer and Windmuller, 2023; and Butt, Kolari, and
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Sadaqat, 2021). According to Bongaerts, Kang, and van Dijk (2020), volatility scaling of momentum returns
ed
is particularly beneficial when market volatility is relatively high.
Contrary to momentum returns, higher market volatility predicts higher reversal returns in different
iew
markets.6 There is considerable literature that links higher reversal returns with compensation for providing
liquidity to the market under duress conditions. Demsetz (1968) rationalized liquidity provision
compensation for the person who stands ready to fulfill orders from those who aspire to trade immediately.
Nagel (2012) argued that reversal profits proxying returns for liquidity provision in the market are higher
ev
when expected volatility is high in the U.S. market.7 As discussed by Gromb and Vayanos (2002) and
Brunnermeier and Pedersen (2009), higher volatility tightens funding constraints, which results in higher
r
liquidity provision costs. Similarly, studies by Jylha, Rinne, and Suominen (2013) and Ignashkina, Rinne,
er
and Suominen (2022) have shown that the positive exposure of hedge funds and mutual funds to reversal
returns indicates that they supply liquidity to the market. It is worth mentioning that the linkage between
pe
reversal profits and liquidity provision costs does not necessarily mean that reversal-related profits are
momentum/reversal strategies have important investment-related implications that have not been studied
previously in the literature to our knowledge.9 In this paper we propose a switching strategy that invests in
tn
momentum when volatility is low and reversal when volatility is high. Dobrynskaya (2019) discussed the
6 The short-term reversal strategy goes long losers and short winners in the previous month. Studies by Jegadeesh
rin
(1990), Goyal and Wahal (2015), Bali, Engle, and Murray (2016), and others have shown that reversal strategies are
profitable for the U.S. market. Griffin, Kelly, and Nardari (2010) reported international evidence on the reversal
strategy.
7Higher market volatility is indicated by a higher volatility index (VIX). In emerging markets, Butt, Hogholm, and
ep
Sadaqat (2021) have shown that higher market volatility is linked to higher reversal returns.
8Studies by Kaniel, Saar, and Titman (2008) and Hendershott, Jones, and Menkveld (2011) have discussed the market
making role of individual investors.
Pr
9We should note that, in view of their negative correlation, the potential effectiveness of combining momentum and
value strategies has been discussed by Asness, Moskowitz,and Pedersen (2013).
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
mixing of momentum and reversal strategies but in a narrower setting. She observed that momentum
ed
crashes occur one-to-three months after market crashes. To avoid momentum crashes, investment was
shifted from momentum to reversal strategy after market crashes. This strategy naturally avoids reversal
iew
prone stocks that belong to the short leg of the momentum strategy and therefore mitigates momentum
crashes.
Importantly, Conrad and Yanuz (2017) identified reversal prone stocks on an ex-ante basis by using
stock-related characteristics that are associated with expected returns. They proposed a MAX momentum
ev
strategy that is long winners’ stocks (with small size and high book-to-market ratios) and short losers (with
big size and low book-to-market ratios). They found that MAX momentum did not experience significant
r
long-term reversals.
er
Three important inferences emerge from previous studies on the linkage between momentum and
reversal as well as mixed strategies. First, as in Asness, Moskowitz, and Pedersen (2013), the switching
pe
strategy can be beneficial when the constituent strategies (such as momentum and reversal) are negatively
correlated. Second, enhanced returns can be achieved by excluding reversal prone stocks from the short leg
of the momentum strategy either on an ex-ante basis (Conrad and Yanuz, 2017) or by timing the market
ot
(Dobrynskaya, 2019). Third, some recent studies have employed dual signals for constructing investment
strategies. For example, Medhat and Schmeling (2022) utilized signals based on the previous months’
tn
returns and share turnover. More specifically, given that the previous month’s thinly traded stocks had
short-term reversals and heavily traded stocks exhibited short-term momentum, they found that both a short-
rin
term reversal strategy and a short-term momentum strategy can generate significant negative and positive
returns, respectively. However, they did not combine these two strategies as we do in the present study.
Lastly, Blitz, Hanauer, Honarvar, Huisman, and Vliet (2022) developed investment strategies based on
ep
multiple signals, including short-term reversal, momentum, analyst revisions, monthly seasonality, and risk.
Our study contributes to the literature on switching strategies by clarifying the mechanism that induces
Pr
different patterns in momentum- and reversal-related returns when market volatility is relatively high. We
use market volatility as a signal for enhancing momentum returns. Given that an increase in market
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
volatility indicates a depressed market condition that increases the cost of providing liquidity in the market,
ed
reversal returns tend to increase whereas momentum returns decrease.10 Similar results in the context of
mutual funds are reported by Ignashkina, Rinne, and Suominen (2022). Funds that had positive exposure
iew
to the reversal strategy obtain liquidity provision related premiums, and funds that had positive exposure to
We propose that, when market volatility is relatively high, distinctive intertemporal return and cost
patterns of momentum and reversal strategies exist related to liquidity supply and demand in the market.
ev
Consistent with these patterns, we implement a switching strategy that relies on market volatility as a
persistent indicator.11 In our strategy, relatively low (high) volatility provides an ex-ante signal to invest in
r
momentum (reversal).
2.2
er
U.S. Data and Initial Evidence on the Switching Strategy
U.S. stock market data is downloaded from Kenneth French’s website. Daily and monthly value-weighted
pe
market returns are gathered for momentum and reversal strategies in the sample period from 1927:1 to
2020:4. The momentum strategy is the difference between the monthly returns of the winner and loser
portfolios in month t + 2. The winner and loser portfolios are comprised of the 10% best and 10% worst
ot
performing stocks traded on the NYSE, AMEX, and NASDAQ exchanges in the previous 12 months t – 11
to t. By contrast, on the basis of performance in month t, the reversal strategy is the difference between the
tn
monthly returns of the 10% loser and 10% winner stocks in month t + 1.
Monthly market volatility is estimated using value-weighted daily stock returns. Volatility states
rin
are identified using quartiles in the sample period. We average momentum and reversal returns in these
volatility-related quartiles. The results are shown in Table 1. Our findings confirm that momentum returns
are lower and reversal returns are higher in the 4th quartile of volatility. As such, a switching strategy that
ep
invests in momentum in the first three quartiles and reversal in the 4th quartile yields the highest monthly
10
Reversal returns are also considered as a proxy for the cost of liquidity provision by Nagel (2012) and Butt,
Pr
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
average return of 1.49% with a highly significant t-statistic equal to 7.56. The main advantage of the
ed
switching strategy is that it gains from the reversal strategy in volatile times when the premium associated
with liquidity provision is generally higher. In recessions, it is expected the cost of liquidity provision will
iew
be relatively higher than otherwise (Moreira and Muir 2017). Consequently, in the last column of Table 1,
it is not surprising that momentum as a proxy for liquidity demand is weak and reversal as a proxy for
liquidity provision is strong. Hence, switching from momentum to reversal in recessions12 enhances
momentum returns.
ev
[INSERT TABLE 1 ABOUT HERE]
Can the relationship between market volatility and returns on momentum and reversal strategies be
r
used on an ex-ante basis for investment purposes? Given that many studies have reported persistent
er
volatility, we test whether historical volatility provides a signal to implement the switching strategy. To do
so, we demarcate volatility-related quartiles based on a rolling window of five years.13 If volatility in the
pe
last month within the five-year window14 belongs to one of the quartiles, we record the next month return15
for the momentum /reversal strategy within that quartile. We repeat this process until the end of the sample
period and then average the returns on the momentum/reversal strategies in their respective quartiles. Figure
ot
1 illustrates a clear pattern – that is, momentum returns are higher in first three quartiles but decline in the
4th quartile, whereas reversal returns have opposite patterns. These ex-ante analyses indicate that the
tn
12These recession periods are identified by National Bureau of Economic Research (NBER) for the time period of
rin
13Our results are robust with respect to the estimation of current and historical volatility on the basis of one month to
six months of daily returns as well as an expanding window. Results are reported in Appendix-A.
ep
14An initial five-year window from 1927:1 to 1931:12 for estimating historical volatility and volatility in the last
month of 1931:12 indicates the current state of volatility.
15 In addition to predicting returns for t+1 month, as a robustness check, we also computed the returns for months t +
2 to t + 7 separately. These results are reported in Appendix-A.
Pr
16The success of the switching strategy is not dependent on volatility-related quartiles. For detailed analyses, see
Appendix-A.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
[INSERT FIGURE 1 ABOUT HERE]
ed
3. Momentum Enhancing Strategies
To construct the switching strategy, we define the signal of the current state of market volatility 𝑄𝑖,𝑡 in
iew
connection to the previous five years of historical market volatility. We demarcate four states in the
historical volatility and determine whether the current month’s volatility 𝑄𝑖,𝑡 belongs to one of these quartile
states 𝑖 = 1,2,3,4. The variable 𝑄𝑖,𝑡 takes the value of 1 or 0 to indicate the state of market volatility. Using
ev
this volatility-based signal, the switching strategy generates time series returns defined as 𝑄4,𝑡 ∗ 𝑟𝑒𝑣𝑡 + 1 +
(1 ‒ 𝑄4,𝑡) ∗ 𝑚𝑜𝑚𝑡 + 1. That is, we invest in a reversal strategy when market volatility is higher (𝑄4,𝑡) and
in the momentum strategy otherwise (1 ‒ 𝑄4,𝑡). Table 2 contains descriptive statistics for the switching
r
strategy. Compared to the momentum (MOM) and reversal (REV) strategies, in the sample period 1932:1
er
to 2020:4, our switching strategy (SWITCH) performs better in terms of average monthly returns, Sharpe
ratio, and end-of-period wealth (denoted WI).17 With regard to alphas associated with Capital Asset Pricing
pe
Model (CAPM) of Sharpe (1964) and others, Fama and French (1993) three-factor model, Carhart (1997)
four-factor model, and Fama and French (2018) six-factor model, there are no significant differences.
However, to price the momentum strategy, a momentum-specific factor reduces mispricing, as in the
ot
Carhart (1997) and Fama and French (2018) six factor models.18
As already discussed, volatility scaling improves momentum returns. In its simplest form, volatility scaling
rin
can be implemented as 𝑆𝑀𝑂𝑀 = 𝑀𝑂𝑀𝑡 + 1 𝜎𝑡,126, where 𝜎𝑡,126 is the volatility in the last 126 days of
market returns. This market volatility is available to the investor on an ex-ante basis for making an
ep
investment in the next month. The inverse of 𝜎𝑡,126 corresponds to the amount for which an investor can
17End-of-period wealth is the estimate of the size of one dollar invested in a strategy from the start of the sample
period (1932:1) to the end (2020:4).
Pr
In unreported results, we find that asset pricing models that do not explicitly contain a momentum factor by Hou,
18
Xue, and Zhang (2015) and Stambaugh and Yuan (2017) suppressed the momentum anomaly also.
10
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
go short in previous losers and long in previous winners. Such scaled momentum strategies have excessive
ed
leverage in comparison to one dollar long-short strategies.19 For instance, the minimum leverage of going
short in losers is 28.55 and maximum is 366.60. Figure 2 traces this leverage for the period from 1932:1 to
iew
2020:4. The performance of scaled momentum strategy SMOM is shown in the fourth row of Table 2. As
shown there, the attractiveness of this strategy is a higher Sharpe ratio. However, the minimum return and
value-at-risk (VaR) equal to 5% are so large that such a strategy cannot be realistically implemented.
ev
Interestingly, the higher Sharpe ratio of scaled momentum is directly proportional to leverage. Figure
3 depicts this relationship by plotting Sharpe ratios on the Y-axis for momentum estimated for a given
r
leverage on the X-axis. As the maximum leverage limit increases, the Sharpe ratio increases also. The
er
scaled strategy requires leverage up to 140 times higher to achieve a Sharpe ratio equal to the switching
strategy. After leverage of 170, there is little or no increase in Sharpe ratios. These analyses suggest that a
pe
target level of volatility for scaling momentum must be chosen to make it a feasible investment choice.
Although trailing market volatility is available to investors on an ex-ante basis, due to high leverage and
increased volatility of the scaled momentum, it is not practically useful to investors. The common
tn
adjustment for target volatility is by means of some constant. For instance, if the target volatility (𝜎𝑡𝑎𝑟𝑔𝑒𝑡)
is the full sample volatility of the momentum strategy (e.g., 7.72%), the constant is a simple fraction of
rin
target volatility to the volatility of the scaled momentum strategy (𝜎𝑠𝑚𝑜𝑚).20 This constant is then
𝜎𝑡𝑎𝑟𝑔𝑒𝑡 𝜎
𝑠𝑚𝑜𝑚. The constant 𝐶 controls the volatility of the scaled strategy but is not available on an ex-
ep
19 Several studies have discussed potential issues (including leverage) in implementing volatility-scaled strategies. See
Caderburg, Doherty, Wang, and Yan (2019), Liu, Tang and Zhou (2019), Bongaerts, Kang, and van Dijk (2020), and
Pr
others.
20 Daniel and Moskowitz (2016) used the same procedure. Barroso and Santa-Clara (2015) chose a target volatility
of 12%.
11
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
ante basis (Liu, Tang, and Zhou, 2019; and Bongaerts, Kang, and van Dijk, 2020). In the present study,
ed
based on the volatilities, we set 𝐶 = 0.01. This look ahead bias is important to mention as it is directly
iew
We conducted a simple exercise of simulating 50,000 times the end-of-period wealth index (WI) for
investing in the scaled momentum strategy with the range of 𝐶 = [0.001 , 0.01]. Figure 4 gives the
frequency distribution of WI. Almost 75% of the time, WI is less than $558,829, which corresponds to the
switching strategy’s WI in Table 2. Different ranges of 𝐶 yield different distributions of WI. At times, the
ev
odds by chance favor the scaled momentum strategy; however, the important choice of 𝐶 cannot be
determined on an ex-ante basis with surety of ex-post superiority over the switching strategy.
r
In the fifth row of Table 2, we report the performance of SMOM(C) for which the volatility of the scaled
er
strategy is set equal to the volatility of the traditional momentum strategy. As expected, the Sharpe ratio
and WI for the scaled strategy are higher than the switching strategy. It is apparent that scaling by 𝐶 does a
pe
good job of managing the risk of scaled momentum. As in Barroso and Santa-Clara (2015), the range of the
leverage is now between a minimum of 0.18 and a maximum of 2.35. Even so, as discussed previously, this
An ex-ante scaling factor 𝐶 can be estimated using the volatility history of both the momentum and
tn
scaled momentum strategies. The resultant scaling is accomplished on a rolling basis using the rolling
constant 𝑉𝐶𝑡 = σ𝑡𝑎𝑟𝑔𝑒𝑡,𝑡 ‒ 6 σ𝑠𝑚𝑜𝑚,𝑡 ‒ 6. For the first period, 𝑉𝐶𝑡 is estimated for the time series from 1927:2
rin
to 1931:12; subsequently, scaled momentum returns for 1932:1 are multiplied by 𝑉𝐶𝑡. This process is
repeated to the end of the sample period. The time series of scaled momentum returns from 1932:1 to 2020:4
is calculated using both volatility and the constant on an ex-ante basis, or 𝑆𝑀𝑂𝑀(𝑉𝐶) = (𝑀𝑂𝑀𝑡 + 1 𝜎𝑡,126)
ep
𝑋 𝑉𝐶𝑡. In the sixth row of Table 2, we report the performance of scaled momentum SMOM(VC). This
strategy outperforms MOM with both higher Sharpe ratio and WI in addition to other risk management
Pr
12
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
indicators such as minimum return and skewness with slightly higher model-based alphas. These findings
ed
indicate that, on an ex-ante basis, volatility scaling improves the performance of the momentum strategy.
Importantly, the switching strategy has better performance than SMOM(VC) with respect to the Sharpe
iew
ratio and WI as well as risk management with reduced minimum return, positive skewness, and lower VaR
at 5%. Notice also that the leverage factor remains constant for this strategy.
In addition to using market volatility for scaling purposes, we use the volatility of momentum returns
as a scaling factor. When target volatility for scaled momentum is chosen on an ex-post basis, this strategy
ev
outperforms others. For instance, the Sharpe ratio is almost two times higher than the momentum strategy
MOM. Also, WI is very large, and the risk of the strategy is well managed with somewhat larger alphas.
r
However, look ahead bias casts doubt on the feasibility of this strategy. When target volatility is calculated
er
on an ex-ante basis, the last line in Table 2 shows that the performance of SMOM(VC-M) is no better than
others, including SWITCH, MOM, SMOM(VC), and SMOM(VC-M). Based on time series of monthly
returns, Table 3 shows that SWITCH has the largest and most significant alpha term equal to 0.009
ot
compared to other strategies. Also, the lower R-squared value of only 26.4 percent for SWITCH compared
to 82.9 percent or higher for the other strategies suggests that SWITCH is quite different from other
tn
strategies.
As mentioned earlier, the momentum strategy is prone to large declines in returns known as momentum
crashes. Kothari and Shanken (1992) have documented that, when the market is downward trending for
ep
some time, stocks with higher/lower returns are likely to be lower/higher beta stocks. This scenario causes
the momentum strategy to yield lower returns when the market rebounds, as higher beta stocks are
Pr
associated with the short leg (losers) of the momentum strategy. Findings in Daniel and Moskowitz (2016)
13
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
have documented momentum crashes in down market states21 when market volatility is relatively high and
ed
the market rebounds.
According to our findings in Table 1 and Figure 1, momentum crashes occur when market volatility is
iew
relatively high.22 Hence, it is not surprising that, by reducing the investment in the momentum strategy in
these high states, the magnitudes of crashes are reduced. Volatility scaling works to achieve this benefit.
Extending previous work, we propose to mitigate momentum crashes via volatility-related signals. When
the current month’s volatility is in the 4th quartile of historical market volatility, it is advantageous to switch
ev
the investment from a momentum strategy to a reversal strategy in the next month. Table 4 provides results
on the efficacy of this switching strategy compared to other momentum strategies. For the 15 largest
r
momentum crashes with an average monthly return of -36.30%, the switching strategy remarkably avoids
er
losses and instead earns on average 2.30%. Comparatively, other volatility-based scaling strategies reduce
the effects of these crashes but not as dramatically. When volatility scaling is carried out with look ahead
pe
bias using SMOM(C) and SMOM(C-M), momentum crashes are reduced to -15.20% and -14.70%,
respectively. Also, upon using realistic volatility scaling strategies SMOM(VC) and SMOM(VC-M), crashes
It has been reported in prior studies that, once crashes occur, it takes almost a decade for an investor to
tn
recover their initial investment. Barroso and Santa-Clara (2015) identified the two time periods 1930:01 to
1939:12 and 2000:01 to 2009:12 in which traditional momentum did not recover the initial amount of one
rin
dollar but scaled momentum did provide some recovery. We replicated their analyses with the addition of
the switching strategy. Figure 5 shows the results for both time periods. It is apparent that the performance
ep
21 The down-market state is roughly defined as negative cumulative market index returns for the last 24 months. Some
authors have used 36 months for this definition.
Pr
22 Most momentum crashes occur when market volatility is in the 4th quartile (Butt, Sadaqat, and Tahir, 2022).
14
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
of the switching strategy is much better than scaled momentum. We infer that the switching strategy is more
ed
efficacious as a remedy for momentum crashes.
iew
We further investigate the periods 1930:01 to 1939:12 and 2001:01 to 2019:12 to better understand
the contribution of the REV strategy to boosting the performance of SWITCH. We compare the performance
of MOM, REV, and SWITCH strategies in the full sample period as well as low volatility and high volatility
sample periods. For instance, in full sample period 1930:01 to 1939:12, as shown in Panel A of Table 5,
ev
SWITCH yields positive returns due to higher REV returns (i.e., MOM returns are negative). In Panel B, the
MOM strategy has higher returns when market volatility is relatively lower compared to high volatility
r
times, and the opposite effect is obvious for the REV strategy. Out of 96 months in the period of 1930:01
er
to 1939:12, the MOM strategy switched to REV strategy for 31 months on the basis of the market volatility
related indicator of higher volatility. For the period of 2001:01 to 2019:12, the intertemporal patterns of
pe
returns are more interesting, as in this sample period overall returns for the MOM and REV strategies are
lower than the SWITCH strategy. These results are attributable to the REV strategy providing positive
returns when market volatility is relatively high. There are 41-out-of-120 months when the REV strategy is
ot
mixed with the MOM strategy to enhance SWITCH returns. More importantly, the mixing of REV with
MOM in conjunction with volatility-related indictors helps to mitigate the tail risk of SWITCH strategy.
tn
We further tested the switching strategy for 200 U.S. industry-based portfolios. Data are downloaded
from Refinitiv for the period from1978:1 to 2020:4. As shown in Figure B.1 of Appendix B, industry-
related momentum returns increase for the first three quartiles of market volatility; however, in the fourth
ep
quartile, when market volatility is higher, momentum returns decline. For reversal-based industry returns,
there is no obvious pattern, except for the fourth quartile in which reversal returns are highest. Subsequently,
Pr
the results reported in Table B.1 for industry-based momentum strategies are very similar to the results
reported in Table 2 for individual stocks. For instance, momentum strategy MOM has economically and
15
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
statistically significant returns for the industry-related portfolio as previously reported by Moskowitz and
ed
Grinblatt (1999); however, the reversal strategy is not significant for industry-based portfolios. More
importantly, as shown in Table B.1, the switching strategy outperformed the scaled momentum strategies
iew
SMOM(C) and SMOM(VC). As a further out-of-sample test, we implemented the SWITCH strategy based
on recently proposed short-term momentum (STMOM) and short-term reversal (STREV) strategies by
Medhat and Schmeling (2022). Based on data generously shared by the authors, Appendix C contains the
results. In general, as shown in Figure C.1 and Table C.1, SWITCH improves upon the performance of
ev
STMOM better than volatility scaling and has relatively lower crash risk also.
5. Robustness Tests
r
In this section we test whether the switching strategy improves the performance of unscaled and scaled
er
momentum strategies in other developed and emerging market countries. For these markets to conserve
space, momentum returns are only scaled by market volatility, and the volatility of the momentum strategy
pe
is not examined.
Here we extend the analyses to other developed markets, including Canada, France, Germany, Japan, and
ot
the United Kingdom (U.K.). Data are downloaded from Refinitiv for the period from 1981:1 to 2020:4.
Following previous studies by Ince and Porter (2006), Griffin, Kelly, and Nardari (2010), and others, we
tn
implement data cleaning procedures in each market. The process of demarcating the volatility quartiles
indicating the current state of volatility and applying the switching strategy is the same as in the U.S. market.
rin
The initial window for demarcating quartiles of realized volatility in each market is from 1981:01 to
1985:12. Based on the positioning of volatility in 1985:12 among the volatility quartiles, the investment for
the month of 1986:01 is implemented using either the momentum or reversal strategy.
ep
Pr
16
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Average returns for the momentum and reversal strategies across the five developed markets are shown
ed
in Figure 6.23 The overall results are similar to those for the U.S. When market volatility is in the 4th quartile,
momentum returns decline (bar with vertical lines), and reversal gains occur (bar with horizontal lines). In
iew
lower volatility quartiles, momentum returns are always higher than reversal returns. We next construct the
switching strategy for each market and compare the results to the unscaled and scaled momentum strategies.
As reported in Table 6, except for France, the switching strategy has a higher Sharpe ratio than momentum,
ev
[INSERT FIGURE 6 ABOUT HERE]
In Panel A of Table 6, momentum returns are higher than reversal returns in the Canadian market.
r
However, due to the opposite relationship between these strategies with market volatility, the switching
er
strategy exhibits stronger performance metrics – namely, a higher Sharpe ratio with less tail risk as indicated
by lower minimum returns (i.e., VaR equal to 5%) and therefore higher wealth index (WI) of $23,627.
pe
These metrics are improved for the switching strategy compared to the scaled momentum SMOM(VC)
strategy. In Panel B for France, the switching strategy does not outperform other strategies; nevertheless,
in unreported analyses, when market volatility is in the 4th quartile, the reversal strategy outperforms the
ot
momentum strategy. As such, the overall relationship between market volatility and future returns based
on reversal and momentum strategies remains intact. Additionally, in Panel C for Germany, the switching
tn
Panel D reports the interesting case of Japan. As previously noted by Chaves (2016), momentum returns
are not strong in the Japanese market, but reversal is significant. Importantly, the switching strategy
combining MOM and REV outperforms both of the latter strategies. Particularly for Japan, market volatility
ep
is strongly linked with future returns on momentum and reversal strategies. For instance, in the 4th quartile
Pr
23To conserve space, results across all markets are shown. Results for individual markets are available upon the
request.
17
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
of volatility, average monthly momentum returns equal -1.62% and reversal returns equal 2.53%.
ed
Contrarily, momentum returns are improved in lower volatility-related quartiles more than reversal returns.
For these reasons, the switching strategy is a very successful strategy in Japan. The Sharpe ratio and WI are
iew
noticeably higher for the switching strategy than both versions of scaled momentum strategies.
Finally, in Panel E for the U.K., given positive/negative momentum/reversal returns, the switching
strategy outperforms both strategies in terms of higher Sharpe ratio and WI. Its performance is slightly
ev
We should mention that our results in Table 6 are robust for strategies constructed based on both
pentiles and deciles using equal-weighted returns. For value-weighted returns in Germany and the U.K.,
r
we find that the switching strategy performs worse than the simple momentum strategy. This outcome can
er
be explained by reversal that is more common among smaller and illiquid stocks (Avramov, Chordia and
Goyal, 2006). Therefore, using a value-weighted scheme, the impact of small and illiquid stocks is reduced.
pe
In this respect, the results for the Japan are consistent irrespective of investment weighing schemes and
percentiles used for the construction of portfolios. The main reason is that the reversal for Japan is strong
The five largest capitalization emerging market countries of Brazil, China, India, South Korea, and Taiwan
tn
are selected to further investigate the performance of the switching strategy. Data are downloaded from
Refinitiv for the period from 1995:1 to 2020:4. As in the U.S. and other developed markets, the rolling
rin
window is comprised of five years. Since the first rolling window is in the period from 1995:1 to 1999:12,
Average returns for momentum and reversal strategies across the five emerging markets24 are shown in
ep
Figure 7. In line with previous evidence in Rouwenhorst (1998), Griffin, Kelly, and Nardari (2010), Butt,
Pr
24To conserve the space, the results across all markets are shown. Results for individual markets are available upon
request from the authors.
18
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Kolari, and Sadaqat (2021), and others, momentum returns (bar with vertical lines) are lower on average in
ed
emerging markets compared to the U.S. and other developed markets. Consistent with our earlier findings,
notice that momentum returns decline and reversal gains occur (bar with horizontal lines) when market
iew
volatility is in the 4th quartile. Subsequently, we repeat the process for constructing the switching strategy
for each emerging market and compare the results to the unscaled momentum and scaled momentum
strategies.
ev
In Panel A of Table 7 for Brazil, unscaled momentum returns are negatively significant, and there is a
strong reversal effect. The switching strategy improves upon momentum returns but understandably
r
performs below the reversal strategy. Nevertheless, the switching strategy materially outperforms the scaled
er
momentum strategies. In Panel B for China, weak momentum and strong reversal are obvious. Again, the
switching strategy yields better results than both scaled momentum strategies. The results for India in Panel
pe
C are similar. Only for Taiwan in Panel E and partially for South Korea in Panel D, the scaled momentum
6. Conclusion
This study documented evidence that market volatility affects the returns on momentum and reversal
tn
strategies differently and that these patterns can be used as investment signals. Following high market
volatility periods, the returns on the momentum/reversal strategy are lower/higher. High market volatility
rin
indicates stress in the market (for example) due to possible recession. At such times, it is plausible to assume
that the cost of liquidity provision will increase, such that a momentum/reversal strategy experiences
losses/gains. When market volatility is relatively high, previous studies have found that returns on the
ep
reversal strategy increase in U.S. and emerging stock markets. Conversely, it is well known that market
volatility is negatively correlated with future momentum returns. These opposite patterns for momentum
Pr
and reversal strategies provide investors with an opportunity to invest in momentum when market volatility
19
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Confirming this conjecture, empirical evidence showed that implementing a switching strategy between
ed
momentum and reversal outperforms traditional and volatility scaled momentum strategies. For the U.S.
stock market in the period 1932 to 2020, the switching strategy produced average monthly returns, Sharpe
iew
ratio, and wealth index (WI) equal to 1.46%, 22.64, and $558,829, respectively. By comparison, the
momentum (volatility scaled momentum) strategy produced 1.02% (1.35%), 13.22 (17.96), and $957
($70,262), respectively. Additionally, crash risk was considerably lower for the switching strategy
compared to other momentum strategies. For the switching strategy, the minimum return in any month,
ev
skewness, and value-at-risk (VaR) at 5% were -33.20%, 0.76, and -8.22% respectively. Comparatively, the
momentum (volatility scaled momentum) strategy produced -77.02% (-47.60%), -2.48 (-0.41), and -9.93%
r
(-10.70%), respectively. Particularly noteworthy, the monthly average returns of the 15 worst crashes for
er
the momentum strategy was -36.63%, which volatility scaling reduced to -14.70%, but the switching
Corroborating our main U.S. market findings, the proposed switching strategy dominated various
momentum strategies in these tests. Among developed countries, the case of Japan was particularly
ot
interesting with average monthly return, Sharpe ratio, and WI for the switching strategy in our sample
period equal to 1.19%, 18.78%, and $62, respectively. In this regard, previous studies have reported no
tn
momentum effect in Japan. In line with those studies, we found that the average monthly return for the
momentum strategy was only 0.07% in Japan. Even with volatility scaling, average monthly returns
rin
improved negligibly to only 0.30%. Among different emerging market countries, traditional momentum
returns were absent in China also (i.e., an average monthly return of only -0.20% in our sample period).
Again, volatility scaling momentum improved matters little to only 0.26%. By contrast, the switching
ep
strategy provided an average monthly return equal to 1.60%, Sharpe ratio at 26.34%, and WI at $31.
Contributing to the literature on what drives momentum, we conclude that momentum is explained in
Pr
part by state dependency and associated trade to immediacy costs linked to demand for market liquidity.
By implication, momentum and other investment strategies can be enhanced by considering multiple market
20
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
signals. We earlier cited studies by Medhat and Schmeling (2022) and Blitz, Hanauer, Honarvar, Huisman,
ed
and Vliet (2022), who utilized multiple market signals in their investment strategies. We find that mixing
strategies not only can boost profits but can provide potential diversification benefits that mitigate tail risk.
iew
Importantly, it is possible that certain market conditions exist in which opposite intertemporal patterns in
various zero-investment strategies can be exploited for investment purposes. We found that market
volatility is such a condition with respect to the momentum/reversal switching strategy. Other market
r ev
er
pe
ot
tn
rin
ep
Pr
21
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
References
ed
Asness, C. S., Moskowitz, T. J., and Pedersen, L. H. (2013). Value and momentum everywhere. Journal
iew
Avramov, D., Chordia, T., and Goyal, A. (2006). Liquidity and autocorrelations in individual stock returns.
Bali, T. G., Engle, R. F., and Murray, S. (2016). Empirical Asset Pricing: The Cross Section of Stock
ev
Barroso, P., and Santa-Clara, P. (2015). Momentum has its moments. Journal of Financial Economics 116,
111-120.
r
Blitz, D., Hanauer, M. X., Honarvar, I., Huisman, R., and van Vliet, P. (2022). Beyond Fama-French
er
factors: Alpha from short-term signals. Working paper, Robeco Quantitative Investments, available at
SSRN 4115411.
pe
Bongaerts, D., Kang, K., and van Dijk, M. (2020). Conditional volatility targeting. Financial Analysts
Brunnermeier, M. K., and Pedersen, L. H. (2009). Market liquidity and funding liquidity. Review of
ot
Butt, H. A., Högholm, K., and Sadaqat, M. (2021). Reversal returns and expected returns from liquidity
tn
provision: Evidence from emerging markets. Journal of Multinational Financial Management 59,
Butt, H. A., Kolari, J. W., and Sadaqat, M. (2021). Revisiting momentum profits in emerging markets.
Butt, H. A., Sadaqat, M., and Tahir, M. (2022). Revisiting the performance of the scaled momentum
ep
Carhart, M. M. (1997). On persistence in mutual fund performance, Journal of Finance 52, 57-
Pr
82.
22
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Cederburg, S., O’Doherty, M. S.,Wang, F., and Yan, X. S., (2019). On the performance of volatility
ed
managed portfolios. Journal of Financial Economics 138, 95-117.
Chaves, D. B. (2016). Idiosyncratic momentum: US and international evidence. Journal of Investing 25,
iew
64-76.
Cheng, S., Hameed, A., Subrahmanyam, A., and Titman, S. (2017). Short-term reversals: The effects of
past returns and institutional exits. Journal of Financial and Quantitative Analysis 52, 143-173.
Conrad, J., and Yavuz, M. D. (2017). Momentum and reversal: Does what goes up always come down?
ev
Review of Finance 21, 555-581.
Cooper, M. J., Gutierrez, R. C. Jr., and Hameed, A. (2004). Market states and momentum. Journal of
r
Finance 59, 1345-1365.
er
Daniel, K., and Moskowitz, T. J. (2016). Momentum crashes. Journal of Financial Economics 122, 221-
247.
pe
Demsetz, H. (1968). The cost of transacting. Quarterly Journal of Economics 82, 33-53.
Dobrynskaya, V. (2019). Avoiding momentum crashes: Dynamic momentum and contrarian trading.
Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United
tn
Fama, E. F., and French, K. R. (1988). Permanent and temporary components of stock prices. Journal of
rin
Fama, E. F., and French, K. R. (1993). Common risk factors in the returns on stocks and bonds.
Fama, E. F., and French, K. R. (2008). Dissecting anomalies. Journal of Finance 63, 1653-1678.
Fama, E. F., and French, K. R. (2018). Choosing factors. Journal of Financial Economics 128,
Pr
234-252.
23
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Fama, E. F., and French, K. R. (2020). Comparing cross-section and time-series factor models.
ed
Review of Financial Studies 33, 1892-1926.
Goyal, A., Jegadeesh, N. and Subrahmanyam, A. (2022). Determinants of momentum: A perspective from
iew
international data. Working paper, University of Lausanne, Emory University, and the University of
Goyal, A., and Wahal, S. (2015). Is momentum an echo? Journal of Financial and Quantitative Science
ev
50, 1237-1267.
Griffin, J. M., Kelly, P. J., and Nardari, F. (2010). Do market efficiency measures yield correct inferences?
r
A comparison of developed and emerging markets. Review of Financial Studies 23, 3225-3277.
Gromb, D., and Vayanos, D. (2002). Equilibrium and welfare in markets with financially constrained
er
arbitrageurs. Journal of Financial Economics 66, 361-407.
pe
Grundy, B. D., and Martin, J. S. M. (2001). Understanding the nature of the risks and the source of the
Guo, J., Li, P., and Li, Y (2022). What can explain momentum? Evidence from decomposition.
Hanauer, M. X., and Windmüller, S. (2023). Enhanced momentum strategies. Journal of Banking and
tn
Hendershott, T., Jones, C. M., and Menkveld, A. (2011). Does algorithmic trading improve liquidity?
Hou, K., Xue, C., and Zhang, L. (2015). Digesting anomalies: An investment approach. Review of Financial
Ignashkina, A., Rinne, K., and Suominen, M. (2022). Short-term reversals, returns to liquidity provision
and the cost of immediacy. Journal of Banking and Finance 138, Issue C, available online, 106430.
Ince, O. S. and Porter, R. B. (2006). Individual equity return data from Thomson DataStream: Handle with
Pr
24
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Jegadeesh, N. (1990). Evidence of predictable behavior of security returns. Journal of Finance 45, 881-
ed
898.
Jegadeesh, N., and Titman, S. (1993). Returns to buying winners and selling losers: Implications for stock
iew
market efficiency. Journal of Finance 48, 65-91.
Jylha, P., Rinne, K., and Suominen, M. (2013). Do hedge funds supply or demand liquidity? Review of
Kaniel, R., Saar, G., and Titman, S. (2008). Individual investor trading and stock returns. Journal of
ev
Finance 63, 273-310.
Koijen, R. S., Rodriguez, J. C., and Sbuelz, A. (2009). Momentum and mean reversion in strategic asset
r
allocation. Management Science 55, 1199-1213.
er
Kothari, S. P., and Shanken, J. (1992). Stock return variation and expected dividends: A time-series and
28.
Liu, F., Tang, X., and Zhou, G. (2019). Volatility-managed portfolio: Does it really work? Journal of
ot
Lo, A. W., and MacKinlay, A. C. (1988). Stock market prices do not follow random walks: Evidence from
tn
Medhat, M., and Schmeling, M. (2022). Short-term momentum. Review of Financial Studies 35, 1480-
rin
1526.
Moreira, A., and Muir, T. (2017). Volatility‐managed portfolios. Journal of Finance 72, 1611-1644.
Moskowitz, T. J., and Grinblatt, M. (1999). Do industries explain momentum? Journal of Finance 54, 1249-
ep
1290.
Poterba, J. and Summers, L. (1988). Mean reversion in stock returns: Evidence and implications.
25
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Rouwenhorst, K. G. (1998). International momentum strategies. Journal of Finance 53, 267-284.
ed
Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. Journal
iew
Stambaugh, R. F., and Yuan, Y. (2017). Mispricing factors. Review of Financial Studies 30, 1270-1315.
Wang, K. Q., and Xu, J. (2015). Market volatility and momentum. Journal of Empirical Finance 30, 79-
91.
r ev
er
pe
ot
tn
rin
ep
Pr
26
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 1: Returns in Market Volatility Quartiles
ed
This table provides the average monthly returns for market volatility-based quartiles for momentum
(MOM) and reversal (REV) and their combined switching-related strategies in the period January 1927
to April 2020. Market volatility is estimated on a monthly basis, and volatility-related quartiles are
demarcated over the full sample. The switching strategy (SWITCH) invests in momentum for quartiles
iew
Q1, Q2, and Q3 and in reversal for the last quartile Q4. Associated t-statistics are provided in parentheses.
In the last column, the monthly returns for the strategies are averaged in the recession period as identified
by the National Bureau of Economic Research (NBER) for the time period of January 1927 to April
2020.
Strategies Average Q1 Q2 Q3 Q4 Recession
1.169 1.463 1.191 1.488 0.534 0.604
ev
MOM (4.99) (5.83) (3.94) (4.16) (0.69) (0.671)
0.901 0.602 0.664 0.525 1.813 2.015
REV (4.99) (2.84) (2.74) (1.91) (3.11) (2.892)
1.489 1.463 1.191 1.488 1.813 1.420
SWITCH (7.56) (5.83) (3.94) (4.16) (3.11) (6.719)
r
er
pe
ot
tn
rin
ep
Pr
27
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Market Volatility and Returns for 1932:1 to 2020:4
ed
0.02
0.018
Average Monthly Returns
0.016
iew
0.014
0.012
0.01 MOM
0.008 REV
0.006
0.004
ev
0.002
0
Q1 Q2 Q3 Q4
r
Figure 1. Market volatility quartiles and next month returns on momentum and reversal strategies
er
pe
ot
tn
rin
ep
Pr
28
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 2: Performance of Different Momentum Strategies
ed
This table provides the performance of various types of the momentum strategies for the period January 1932 to April 2020.
A number of performance metrics are shown, including average monthly returns (Avg), standard deviation of returns (SD),
Sharpe ratio (SR), minimum return (Min), maximum return (Max), skewness (Skew), end-of-period wealth of holding a
strategy (WI), and value-at-risk (VaR) at 5%. In the last four columns, estimated alphas are reported for the Capital Asset
iew
Pricing Model (CAPM) of Sharpe (1964) and others, Fama and French (1993) three-factor model (FF3), Carhart (1997) four-
factor model (CF4), and Fama and French (2018) six-factor model (FF6). Descriptive statistics and alphas are calculated on a
monthly basis except for WI which indicates holding period return of an initial investment of $1 that is compounded to the
end of the sample period. MOM is a momentum strategy that goes long 10th decile winners’ stocks and short 1st decile losers’
stocks based on the performance over the last 11 months. REV indicates the reversal strategy that goes long 1st decile losers
in the previous month and short 10th decile winners. SWITCH is a combination of momentum and reversal strategies based on
the previous volatility-related indicator. The switching strategy implements the momentum strategy for the next month if the
ev
current month’s market volatility is less than the 0.75 percentile of the previous 5 years of market volatility, otherwise it
chooses the reversal strategy. SMOM is the momentum strategy scaled by the volatility of daily market returns over the last
126 days. The strategy SMOM(C) is further scaled by a constant that equates its volatility with the switching strategy on an
ex-post basis. The strategy SMOM(VC) is scaled by a constant that equates the volatility of SMOM(C) with the SWITCH
r
strategy for every 5 years. In the strategy SMOM(VC), the constant is used on an ex-ante basis for scaling. For strategies
SMOM(C) and SMOM(VC), the volatility of daily returns of the momentum strategy over the last 126 days is used as in
Barroso and Santa-Clara (2015).
MOM
Strategies Avg
1.02
(4.52)
SR
13.22
Min
-77.02
Max
26.16
er
Skew WI ($)
-2.48 957
VaR
(5%)
-9.93
CAPM FF3
1.41
(7.68)
1.62
(8.45)
CF4
0.29
FF6
0.00
(3.51) (0.31)
pe
REV 0.86 14.81 -28.53 58.63 1.36 1640 -6.95 0.67 0.67 0.92 0.00
(4.92) (3.81) (3.73) (5.09) (0.40)
SWITCH 1.46 22.64 -33.2 58.63 0.76 558829 -8.22 1.41 1.45 1.09 0.01
(7.92) (7.33) (7.72) (5.28) (1.10)
180.21 - 197.65 211.16 90.61 1.18
ot
SMOM(VC) 1.35 17.96 -47.6 49.44 -0.41 70262 -10.7 1.60 1.76 0.60 0.01
(6.32) (7.98) (8.74) (4.35) (0.72)
SMOM(C-M) 2.04 26.36 -31.17 29.09 -0.3 6393590 -8.67 2.21 2.35 1.15 0.01
(7.94) (8.92) (9.47) (5.85) (1.46)
rin
29
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 3: Strategy-Based Alphas
ed
Using monthly returns from January 1932 to April 2020, columns provide estimates of the linear regression
𝑟𝑒𝑡𝑖,𝑡 = 𝛼 + 𝑟𝑒𝑡𝑗,𝑡 + 𝜀𝑡 for each strategy on other strategies, including SWITCH, MOM, SMOM(VC), and
SMOM(VC-M). Estimated coefficients (t-statistics in parentheses) are reported as well as the number of
observations and adjusted R-squared value.
iew
Strategies SWITCH MOM SMOM(VC) SMOM(VC-M)
MOM -0.445 0.241 0.231
(-8.64) (12.29) (11.59)
SMOM(VC) 0.927 0.518 0.715
(12.76) (12.29) (32.65)
SMOM(VC-M) -0.162 0.488 0.703
ev
(-2.10) (11.59) (32.65)
SWITCH -0.149 0.144 -0.026
(-8.64) (12.76) (-2.10)
Alpha (α) 0.009 -0.001 -0.001 0.002
(5.06) (-1.23) (-0.76) (2.67)
r
Adjusted R-squared 0.264 0.829 0.917 0.909
er
pe
ot
tn
rin
ep
Pr
30
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
370.00
ed
320.00
Weights on MOM
270.00
iew
220.00
170.00
120.00
70.00
ev
20.00
32
36
40
44
48
52
56
60
64
68
72
76
80
84
88
92
96
00
04
08
12
16
20
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
20
20
20
20
20
r
Figure 2: The scaling factor for the momentum strategy
er
pe
ot
tn
rin
ep
Pr
31
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Momentum and Sharpe Ratios
ed
0.260
0.240
140, [Y VALUE]
0.220
iew
Sharpe Ratios
0.200
0.180
SMOM
0.160
SWITCH
0.140
ev
0.120
0.100
20 70 120 170 220 270 320 370
r
Leverage
er
Figure 3: The leverage factor and Sharpe ratio of the scaled momentum strategy SMOM(C)
pe
ot
tn
rin
ep
Pr
32
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Distribution of Wealth Index
ed
16000 120.00%
14000
100.00%
95.54%
iew
12000
80.00%
10000 75.97%
Frequency
8000 60.00%
50.43%
6000
40.00%
ev
4000 30.86%
20.00%
2000 15.43%
0 0.00%
r
68 753 13,740 558,829 4,694,680 7,820,857
Frequency Cumulative %
er
Figure 4: The wealth index for momentum strategy SMOM(C) using a constant between 0.001 to 0.01
pe
ot
tn
rin
ep
Pr
33
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 4: Momentum Crashes
ed
This table shows the 15 most extreme momentum crashes for the momentum strategy observed by Daniel and
Moskowitz (2016). For months in which these crashes occurred, returns for the different momentum strategies are
shown. MOM is the momentum strategy that goes long 10th decile winners’ stocks and short 1st decile losers’ stocks
based on the performance over the last 11 months. REV indicates the reversal strategy that goes long in 1st decile
iew
losers in the previous month and short 10th decile winners. SWITCH is a combination of momentum and reversal
strategies based on the previous volatility-related indicator. The switching strategy implements the momentum
strategy for the next month if the current month’s market volatility is less than 0.75 percentile of the previous 5
years of market volatility, otherwise it chooses the reversal strategy. SMOM is the momentum strategy scaled by
the volatility of daily market returns over the last 126 days. The strategy SMOM(C) is further scaled by a constant
that equates the volatility of SMOM(C) with the switching strategy on an ex-post basis. The strategy SMOM(VC)
is scaled by a constant that equates its volatility with the switching strategy for every 5 years. In the strategy
ev
SMOM(VC), the constant is used on an ex-ante basis for scaling. For strategies SMOM(C-M) and SMOM(VC-M),
the volatility of the daily returns of the momentum strategy over the last 126 days is used as in Barroso and Santa-
Clara (2015).
Date MOM REV SWITCH SMOM(C) SMOM(VC) SMOM(C-M) SMOM(VC-M)
r
8:1932 -0.770 -0.180 -0.180 -0.304 -0.476 -0.303 -0.512
7:1932 -0.602 -0.078 -0.078 -0.242 -0.321 -0.257 -0.379
4:2009
9:1939
1:2001
4:1933
-0.456
-0.452
-0.420
-0.419
0.010
0.239
0.214
0.068
0.010
0.239
0.214
0.068
er -0.132
-0.348
-0.302
-0.157
-0.182
-0.438
-0.316
-0.296
-0.104
-0.374
-0.221
-0.125
-0.151
-0.665
-0.219
-0.242
pe
3:2009 -0.398 0.172 0.172 -0.115 -0.124 -0.098 -0.112
6:1938 -0.332 0.272 -0.332 -0.188 -0.226 -0.152 -0.244
4:2020 -0.287 0.384 0.384 -0.114 -0.087 -0.191 -0.226
5:1933 -0.269 -0.099 -0.099 -0.102 -0.197 -0.117 -0.232
8:2009 -0.254 -0.039 -0.254 -0.127 -0.212 -0.058 -0.106
11:2002 -0.201 0.004 0.004 -0.105 -0.135 -0.124 -0.175
ot
34
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Wealth Index: 1932:1 to 1939:12
ed
10.00
9.00
8.00
7.00
6.00
iew
5.00
4.00
3.00
2.00
1.00
0.00 $0.06
ev
-1.00
1932 1933 1934 1935 1936 1937 1938 1939
WI(MOM) WI(SWITCH) WI(SMOM)
r
Wealth Index: 2000:1 to 2009:12
5.00
4.00 er
pe
3.00
2.00 $2.20
1.00
ot
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
35
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 5: Performance of Momentum and Reversal in Crisis Periods.
ed
This table provides the performance of momentum (MOM), reversal (REV) and SWITCH strategies for the
sample periods 1932-1939 and 2000-2009. MOM is a momentum strategy that goes long 10th decile winners’
stocks and short 1st decile losers’ stocks based on the performance over the last 11 months. REV indicates
the reversal strategy that goes long 1st decile losers in the previous month and short 10th decile winners.
iew
SWITCH is a combination of momentum and reversal strategies based on the previous volatility-related
indicator. The switching strategy implements the momentum strategy for the next month if the current
month’s market volatility is less than the 0.75 percentile of the previous 5 years of market volatility,
otherwise it chooses the reversal strategy. In Panels A and C, we report average monthly returns (Avg),
standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum return (Max),
skewness (Skew), end-of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. In Panels
B and D, the same information is provided for momentum and reversal strategies in the cases of low
ev
volatility, or MOM(LV) and REV(LV), and high volatility, or MOM(HV) and REV(HV).
Panel A: Switch Strategy for 1932-1939
Strategies Avg SD SR Min Max Skew WI ($) VaR (5%)
MOM -0.98 15.64 -6.26 -77.02 24.99 -2.15 0.06 -28.48
r
REV 3.85 10.14 38.00 -17.97 58.63 1.84 24.87 -7.96
SWITCH 2.97 11.10 26.79 -33.20 58.63 1.12 9.78 -9.87
Panel B: Contribution of MOM and REV
MOM(LV)
MOM(HV)
1.42
-6.01
8.84
23.85
16.08
-25.21
er -33.20
-77.02
24.99
24.96
-0.58
-1.35
1.93
0.03
-9.84
-52.68
pe
REV(LV) 2.72 7.19 37.88 -10.21 27.24 0.70 4.92 -7.51
REV(HL) 6.22 14.39 43.27 -17.97 58.63 1.55 5.05 -9.16
Panel C: Switch Strategy for 2000-2009
Strategies Avg SD SR Min Max Skew WI VaR
MOM 0.22 11.19 1.95 -45.58 26.16 -1.33 0.55 -16.70
ot
36
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Market Volatility and Returns for
ed
Developed Markets
0.02
0.018
iew
Monthly Average Returns
0.016
0.014
MOM
0.012
0.01 REV
0.008
0.006
0.004
ev
0.002
0
-0.002 Q1 Q2 Q3 Q4
-0.004
r
Figure 6: Market volatility quartiles and average next month returns on the momentum and reversal
strategies for developed markets in the period 1986:1 to 2020:4
er
pe
ot
tn
rin
ep
Pr
37
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 6: Performance of Different Momentum Strategies in Developed Markets
ed
This table provides the performance of various momentum strategies for Canada, France, Germany, Japan,
and the United Kingdom (U.K.). A number of performance metrics are shown, including average monthly
returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum
return (Max), end-of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. The sample
iew
period is from January 1986 to April 2020. MOM indicates the momentum strategy that goes long in 10th
decile winners’ stocks and short in 1st decile losers’ stocks based on the performance over the last 11
months. REV indicates the reversal strategy that goes long in 1st decile losers of the previous month and
short 10th decile winners. The switching strategy is a combination of momentum and reversal strategies
based on previous volatility related indicator. SWITCH chooses the momentum strategy for the next month
if the current month’s market volatility is less than 0.75 percentile of previous 5 years of market volatility,
ev
otherwise it chooses the reversal strategy SMOM is the volatility scaled momentum strategy, which is very
similar to the one used in Barroso and Santa-Clara (2015), except that the market volatility over the last
126 is used instead of the volatility of the momentum strategy. SMOM is the momentum strategy scaled by
the volatility of the daily market returns over the last 126 days. The strategy SMOM(C) is further scaled by
r
a constant that equates its volatility with the switching strategy on an ex-post basis. The strategy
SMOM(VC) is scaled by a constant that equates the volatility of SMOM(C) with the switching strategy for
every 5 years. In the strategy SMOM(VC), the constant is used on an ex-ante basis for scaling. All results
except WI($) are in percentages.
Panel A: Canada
Strategies Avg SD
er
SR Min Max WI ($) VaR (5%)
pe
MOM 1.95 8.80 22.10 -30.17 40.39 617 -12.53
REV 0.87 8.17 10.63 -54.99 43.33 9 -11.37
SWITCH 2.80 8.52 32.91 -23.99 43.33 23,627 -10.79
SMOM 351.99 1298.47 27.11 -3587.69 5359.95 -1551.14
SMOM(C) 2.39 8.80 27.11 -24.32 36.33 3,955 -10.51
ot
38
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Tab6, continued
ed
Panel D: Japan
MOM 0.07 6.35 1.06 -49.72 22.16 1 -9.62
REV 0.83 5.83 14.28 -11.26 48.78 16 -7.16
SWITCH 1.19 6.32 18.78 -19.28 48.78 62 -8.15
iew
SMOM 36.86 700.68 5.26 -3614.53 3095.76 -945.56
SMOM(C) 0.33 6.35 5.26 -32.75 28.05 2 -8.57
SMOM(VC) 0.30 5.88 5.10 -29.63 23.88 2 -8.76
Panel E: U.K.
MOM 1.01 8.45 11.92 -34.16 40.63 14 -12.41
ev
REV -0.63 7.15 -8.88 -39.64 35.44 0 -10.52
SWITCH 1.24 8.01 15.49 -39.64 35.44 43 -11.21
SMOM 221.09 1229.06 17.99 -6402.59 6787.49 -1432.85
SMOM(C) 1.52 8.45 17.99 -44.04 46.69 118 -9.86
r
SMOM(VC) 1.29 8.31 15.47 -58.72 33.08 43 -9.91
er
pe
ot
tn
rin
ep
Pr
39
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
ed
Market Volatility and Returns for
Emerging Markets
iew
0.025
0.02
0.015
0.01 MOM
REV
ev
0.005
0
Q1 Q2 Q3 Q4
-0.005
r
-0.01
er
Figure 7: Market volatility quartiles and average next month returns on the momentum and reversal strategies for
emerging markets in the period 2000:1 to 2020:4
pe
ot
tn
rin
ep
Pr
40
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 7: Performance of Different Momentum Strategies in Emerging Markets
ed
This table provides the performance of various momentum strategies for Brazil, China, India, South Korea,
and Taiwan. A number of performance metrics are shown, including average monthly returns (Avg),
standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum return (Max), end-
of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. The sample period is from
iew
January 2000 to April 2020. MOM is the momentum strategy that goes long 10th decile winners’ stocks and
short 1st decile losers’ stocks based on performance over the last 11 months. REV indicates the reversal
strategy that goes long 1st decile losers of the previous month and short 10th decile winners. The switching
strategy is a combination of momentum and reversal strategies based on the previous volatility-related
indicator. SWITCH chooses the momentum strategy for the next month if the current month’s market
volatility is less than 0.75 percentile of the previous 5 years of market volatility, otherwise it chooses the
reversal strategy. SMOM is the volatility scaled momentum strategy, which is very similar to the one used
ev
in Barroso and Santa-Clara (2015), except that the market volatility over the last 126 is used instead of the
volatility of the momentum strategy. SMOM is the momentum strategy scaled by the volatility of the daily
market returns over the last 126 days. The strategy SMOM(C) is further scaled by a constant that equates
its volatility with the switching strategy on an ex-post basis. The strategy SMOM(VC) is scaled by a constant
r
that equates the volatility of SMOM(C) with the switching strategy for every 5 years. In the strategy
SMOM(VC), the constant is used on ex-ante basis for the scaling. All results except for WI($) are in
percentages.
Panel A: Brazil
Strategies
MOM
Avg
-0.55
SD
8.39
erSR
-6.57
Min
-44.90
Max
23.04
WI ($)
0
VaR (5%)
-12.52
pe
REV 1.09 7.81 14.01 -25.56 33.35 8 -9.50
SWITCH 0.64 8.50 7.58 -32.19 33.35 2 -12.52
SMOM -18.60 692.27 -2.69 -3206.78 2088.33 -1042.74
SMOM(C) -0.23 8.39 -2.69 -38.85 25.30 0 -12.63
SMOM(VC) 0.29 7.88 3.66 -28.69 30.88 1 -11.77
ot
Panel B: China
MOM 0.09 6.05 1.51 -18.08 22.45 1 -9.34
REV 1.48 5.54 26.71 -20.21 27.90 32 -7.65
tn
Panel C: India
MOM 0.79 9.38 8.41 -68.11 30.18 2 -16.76
REV 2.17 7.84 27.70 -14.56 33.80 134 -8.17
SWITCH 1.48 8.79 16.85 -68.11 29.02 14 -11.50
ep
41
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 7, continued
ed
Panel D: South Korea
MOM -0.08 7.96 -0.96 -29.35 24.93 0 -12.56
REV 1.77 9.46 18.71 -39.69 57.15 32 -10.07
SWITCH 0.86 7.73 11.09 -27.16 30.30 4 -10.82
iew
SMOM 41.72 518.50 8.05 -1564.40 1559.00 -824.31
SMOM(C) 0.64 7.96 8.05 -24.02 23.93 2 -12.65
SMOM(VC) 1.05 9.21 11.35 -37.37 46.91 5 -11.48
Panel E: Taiwan
MOM 0.16 7.47 2.08 -50.65 25.43 1 -11.54
ev
REV 0.09 5.82 1.51 -13.38 45.77 1 -8.21
SWITCH 0.37 6.83 5.42 -50.65 25.43 1 -8.54
SMOM 58.67 558.42 10.51 -3494.10 1582.54 -788.38
SMOM(C) 0.79 7.47 10.51 -46.77 21.18 4 -10.55
r
SMOM(VC) 0.74 6.80 10.94 -24.81 23.98 4 -10.21
er
pe
ot
tn
rin
ep
Pr
42
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Appendix A
ed
In the text, daily returns are used to estimate market volatility in the current month as well as historical
volatility over five years. As a robustness check, we estimate market volatility over two, three, four, five, and
iew
six months and then expand the window to the end of the sample period. Using various measures of market
volatility, we find in Panel A of Table A.1 that the Sharpe ratios of different versions of the SWITCH strategy
remain very close to those for the original SWITCH strategy shown in the first row of the Panel A.
Similarly, in Panel B, we conduct robustness checks for the original SWITCH strategy using different
ev
percentiles than the 75th percentile. Starting with the 50th percentile, when the current month’s market volatility
is higher than the 50th percentile of market volatility in the previous five years, the SWITCH strategy invests
r
in reversal instead of momentum. These percentiles are incrementally increased by 0.05 until the 80th percentile
er
is reached. As shown in the Panel B, SWITCH strategy results remain fairly consistent as volatility-related
percentiles are increased. We infer that the success of the SWITCH strategy is not dependent on volatility-
pe
related percentiles.
Lastly, in Panel C, we test for how long the volatility-based indicator remains profitable for the SWITCH
strategy. Our main SWITCH strategy focuses on the next month t + 1 given the current month’s volatility is
ot
higher than 75th percentile of five years of the market volatility. By contrast, SWITCHt+2 is based on month
t + 2 (with one month lag), and other strategies increase the time horizon to t + 7 (with six month lag). The
tn
results show that the volatility indicator remains profitable. For instance, MOM, SMOM(VC) and SMOM(VC-
M) based momentum strategies have Sharpe ratios equal to 13.22,17.96 and 19.72 respectively (Table 2),
whereas the SWITCH strategy with increasing time lags normally have higher Sharpe ratios. Nevertheless, the
rin
Sharpe ratio for the SWITCH strategy gradually decreases as the time between the availability and
43
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table A.1: Volatility and Robustness of SWITCH Strategy
ed
This table provides the performance results fdor various SWITCH based strategies in terms of average monthly
returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum return
(Max), skewness (Skew), end-of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. In Panel
A, SWITCH is a combination of momentum and reversal strategies based on the previous volatility-related
iew
indicator. The switching strategy implements the momentum strategy for the next month if the current month’s
market volatility is less than the 0.75 percentile of the previous 5 years of market volatility, otherwise it chooses
the reversal strategy. By contrast, the subscripts m2 to m6 indicate that the previous 5 years volatility is measured
on the basis of rolling windows of 2 to 6 month, respectively. The investment criteria for SWITCHm2 to
SWITCHm6 are same as for SWITCH. Lastly, for SWITCHexp, volatility is expanded until the end of the sample
period. In Panel B, the SWITCH strategy is implemented using percentiles that increase by 0.05; that is, the next
month investment in the momentum strategy is made when the recent month’s volatility is less than 0.5 percentile
ev
(p.5) of the previous 5 years of market volatility, otherwise it chooses the reversal strategy. This process is
repeated for other percentiles such as 0.55 percentile (p.55) to the 0.80 percentile (p.80). In Panel C, the decision
to switch investment from momentum to reversal is delayed by one month. For example, SWITCHt+2 implements
the momentum strategy with one month’s lag if the current month’s market volatility is less than the 0.75
percentile of the previous 5 years of market volatility, otherwise it chooses the reversal strategy. Similarly, for
r
SWITCHt+3 to SWITCHt+7, the lag between the available information and investment decision incrementally
increases by one month.
Strategies
SWITCH
Avg
1.46
SD
6.44
SR
22.64
er
Panel A: Different Horizon of Market Volatility for SWITCH
Min
-33.20
Max
58.63
Skew
0.76
WI ($)
558,829
VaR (5%)
-8.22
pe
SWITCHm2 1.40 6.56 21.35 -45.19 58.63 0.48 270,376 -8.48
SWITCHm3 1.40 6.52 21.45 -45.19 58.63 0.59 274,097 -8.48
SWITCHm4 1.54 6.54 23.48 -45.19 58.63 0.37 1,119,475 -8.29
SWITCHm5 1.41 6.46 21.82 -45.19 58.63 0.41 318,827 -8.38
SWITCHm6 1.42 6.44 21.97 -45.19 58.63 0.45 342,078 -8.29
ot
SWITCH (p.60) 1.40 6.32 22.12 -25.04 58.63 1.07 333,346 -7.98
SWITCH (p.65) 1.45 6.36 22.86 -33.20 58.63 0.90 572,223 -7.98
SWITCH (p.70) 1.48 6.37 23.21 -33.20 58.63 0.87 731,761 -7.99
SWITCH (p.80) 1.43 6.63 21.55 -45.19 58.63 0.38 346,628 -8.38
rin
44
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Appendix B
ed
Market Volatility and Returns for
Industry Portfolios
0.019
iew
0.014
Average Monthly Returns
0.009 MOM
ev
REV
0.004
r
-0.001 Q1 Q2 Q3 Q4
-0.006
er
pe
Figure A.1: Market volatility quartiles and next month returns on momentum and reversal strategies for U.S.
industry-based portfolios
ot
tn
rin
ep
Pr
45
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table B.1: Performance of Different Momentum Strategies for U.S. Industry Portfolios
ed
This table reports the performance of various types of the momentum strategies for the period January 1978
to April 2020 for 200 U.S. industry-based portfolios. A number of performance metrics are shown,
including average monthly returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR), minimum
return (Min), maximum return (Max), end-of-period wealth of holding a strategy (WI), and value-at-risk
iew
(VaR) at 5%. MOM is the momentum strategy that goes long 10th decile winners and short 1st decile losers
among industry-based portfolios over the last 11 months. REV indicates the reversal strategy that goes long
1st decile losers of the previous month and short 10th decile winners. Switching is a combination of
momentum and reversal strategies based on the previous volatility-related indicator. SWITCH implements
the momentum strategy for the next month if the current month’s market volatility is less than 0.75
percentile of the previous 5 years of market volatility, otherwise it chooses the reversal strategy. SMOM is
the momentum strategy scaled by the volatility of the daily market returns over the last 126 days. The
ev
strategy SMOM(C) is further scaled by a constant that equates its volatility with the switching strategy on
an ex-post basis. The strategy SMOM(VC) is scaled by a constant that equates the volatility of SMOM(C)
with the switching strategy for every 5 years. In the strategy SMOM(VC), the constant is used on an ex-ante
basis for the scaling. All the results except for WI($) are in percentages.
r
Strategies Avg SD SR Min Max WI ($) VaR (5%)
MOM 0.85 6.08 13.96 -47.57 22.56 29 -8.01
REV
SWITCH
SMOM
SMOM(C)
0.06
1.00
117.75
0.88
4.62
5.68
815.08
6.06
er
1.32
17.55
14.45
14.45
-20.47
-47.57
20.11
22.56
-7243.88 2846.47
-53.87 21.17
1
68
32
-7.12
-7.37
-1074.61
-7.99
pe
SMOM(VC) 0.84 5.93 14.09 -47.89 21.38 25 -8.04
ot
tn
rin
ep
Pr
46
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Appendix C
ed
Medhat and Schmeling (2022) constructed separate short-term reversal and momentum strategies.
Departing from previous studies, the authors double sort stock returns on the previous month’s return and
iew
share turnover. In the sample period July 1963 to December 2018, the strategy goes long in previous month
winners and short the losers within the highest turnover decile and produces momentum profits (denoted
STMOM) of 16.4% per annum; conversely, the same strategy in lowest turnover decile generates reversal
profits (denoted STREV) of -16.9% per annum. Compared to the momentum strategies used in our study,
ev
their momentum and reversal strategies are different and therefore provide another opportunity to test our
switching strategy.
r
As another out-of-sample robustness test, we contacted the authors, who generously shared their
er
data. To test our switching strategy, the initial window for demarcating quartiles of realized volatility is
from 1958:07 to 1963:6. Based on the positioning of volatility in 1963:6 among the volatility quartiles, the
pe
investment for the month of 1963:07 is implemented using either the short-term momentum or short-term
reversal strategy. Average returns for STMOM and STREV strategies are shown in Figure B.1. Given that
market volatility is in the 4th quartile, momentum returns decline (bar with vertical lines) and reversal
ot
returns increase (bar with horizontal lines). The results are very similar to those in Figure1 in which
47
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
ed
Market Volatility, STMOM, and STREV
0.020
iew
0.018
0.016
Monthly Average Returns
0.014
0.012
0.010 MOM
ev
0.008 REV
0.006
0.004
0.002
r
0.000
Q1 Q2 Q3 Q4
er
Figure C.1: Market volatility quartiles and next month returns on STMOM and STREV strategies for the
period 1963:07 to 2018:12.
pe
We next replicated Table 1 using STMOM and STREV as the main testing strategies for the period
July 1963 to December 2018. Since daily series for STMOM are unavailable, we did not reproduce the last
ot
two strategies in Table 1. As shown in Table B.1, the switching strategy has better performance metrics in
comparison to short-term momentum strategy STMOM and scaled versions of STMOM including
tn
STSMOM, STSMOM(C) and STSMOM(VC).25 For instance, the Sharpe ratio of the switching strategy
(SWITCH) is 22.06, which is higher than STMOM and other volatility scaled STMOM strategies. More
importantly, cumulative returns captured by the wealth index (WI) are highest for SWITCH relative to other
rin
strategies including STREV. In terms of crash risk, the monthly average of the 10 worst returns from July
1963 to December 2018 is -25.96% for the traditional momentum strategy MOM, -23.65% for STMOM, -
ep
25.83% for SSTMOM(C), and -22.33% for SSTMOM(VC). Improving upon these losses, for SWITCH we
get -17.15%. These results suggest that SWITCH can reduce the tail risk of STMOM.
Pr
25 The short-term momentum strategy STMOM is scaled in similar way as the traditional momentum strategy as
discussed Subsections 3.1 and 3.2.
48
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table C.1: Performance of Short-term Momentum Strategies for the U.S. Market
ed
This table provides the performance of various types of short-term momentum strategy proposed by Medhat and
Schmeling (2022) for the U.S. market for the period July 1968 to December 2018. Several performance metrics
are shown, including average monthly returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR),
minimum return (Min), maximum return (Max), skewness (Skew), end-of-period wealth of holding a strategy
iew
(WI), and value-at-risk (VaR) at 5%. STMOM is a momentum strategy that goes long for previous month winners
and shorts the losers within the highest turnover decile. STREV goes long in previous month losers and short the
winners within lowest turnover decile. SWITCH is a combination of momentum and reversal strategies based on
the previous volatility-related indicator. The switching strategy implements the momentum strategy for the next
month if the current month’s market volatility is less than the 0.75 percentile of the previous 5 years of market
volatility, otherwise it chooses the reversal strategy. SSTMOM is the momentum strategy scaled by the volatility
ev
of daily market returns over the last 126 days. The strategy SSTMOM(C) is further scaled by a constant that
equates its volatility with the switching strategy on an ex-post basis. The strategy SSTMOM(VC) is scaled by a
constant that equates the volatility of SMOM(C) with the SWITCH strategy for every 5 years.
Strategies Avg SD SR Min Skew Max WI ($) VaR (5%)
STMOM 1.37 8.26 16.53 -38.42 0.134 39.97 896 -11.61
r
STREV 1.41 5.60 25.26 -22.34 0.736 40.18 4,259 -7.26
SWITCH 1.65 7.47 22.06 -25.25 0.723 40.18 9,251 -9.94
STSMOM
STSMOM(C)
STSMOM(VC)
220.72 1231.56 17.92
1.48
1.46
8.26
8.57
17.92
17.09
er-6837.30
-45.87
-42.01
0.060
0.060
0.489
6291.46
42.21
40.17
1,848
782
-1426.54
-9.57
-10.82
pe
ot
tn
rin
ep
Pr
49
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
ed
Momentum, Market Volatility, and Reversal
iew
Abstract
Momentum profits collapse and reversal occurs when preceding market volatility is relatively high. Based
ev
on these intertemporal patterns, we implement an investment strategy that switches from momentum to
reversal when volatility is high. The proposed switching strategy has two advantages over scaled
momentum strategies: (1) the leverage factor is constant, and (2) no ex-post information is used to control
for volatility. In U.S. stock market tests across a variety of performance metrics, the switching strategy
r
distinguishes itself from traditional and volatility scaled momentum strategies by eliminating losses due to
momentum crashes. Further evidence confirms that the switching strategy is successful in other developed
and emerging stock markets, especially in Japanese and Chinese stock markets.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Momentum, Market Volatility, and Reversal
ed
1. Introduction
A momentum strategy buys previous winners and finances this long position by short selling the previous
iew
losers to earn significant future returns (Jegadeesh and Titman, 1993). This simple investment strategy has
attracted considerable interest in the literature. For example, it is commonly used as a systematic risk factor
in asset pricing models (e.g., Carhart, 1997; and Fama and French, 2018, 2020) as well as investment
ev
strategy by securities companies. Given efficient markets, a vexing problem is explaining the persistence
of momentum profits. Recent work by Guo, Li, and Li (2022) investigated a panoply of competing
r
explanations for momentum, including the anchoring effect, fundamental-related concepts, prospect theory,
firm characteristics related to limits to arbitrage and information uncertainty, and a number of other possible
er
explanatory variables. Unfortunately, they found that these variables only explained 31% of momentum
returns, which leaves 69% unexplained. Another study by Goyal, Jegadeesh and Subrahmanyam (2022) of
pe
international markets concluded that underreaction by investors is the most consistent explanation of
momentum among numerous other possible explanations. Consistent with these studies’ findings, Fama
and French (2008) has observed that momentum is the premier puzzle in financial economics.
ot
Momentum strategies are unique in that they experience frequent crashes. Crashes typically occur in
down market states (DMS) when market volatility is relatively high and market returns contemporaneously
tn
increase (Daniel and Moskowitz, 2016; and Guo, Li, and Li, 2022). This negative relationship between
market volatility and momentum returns was first recognized by Wang and Xu (2015). Well-known studies
rin
by Barroso and Santa-Clara (2015) and Daniel and Moskowitz (2016) showed that crash losses can be
mitigated via scaling momentum returns by the inverse of trailing market volatility.1 Additionally, they
ep
found that scaling improves both Sharpe ratios and momentum returns. Although volatility is used as a
scaling factor on an ex-ante basis, it substantially increases the leverage and overall volatility of the
Pr
1As shown by Barroso and Santa-Clara (2015), lagged market volatility produces the same results as the lagged
volatility of the momentum strategy.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
momentum strategy. Higher leverage and volatility can only be controlled by scaling further through a
ed
constant that is selected on an ex-post basis.2 However, because this approach makes scaled strategies
impractical for investment purposes, Bongaerts, Kang and Dijk (2020) suggested adjustments to make
iew
scaled strategies feasible.3
In this study we propose an investment strategy based on the state of current market volatility in
conjunction with historical volatility to enable a switching strategy between momentum and short-term
reversal (hereafter switching strategy). Short-term reversals have been studied extensively in the U.S.
ev
market, and excess returns associated with this strategy have been documented in different time periods
(e.g., see Lehman, 1990; Jegadeesh, 1990; Bali, Engle, and Murray, 2016; and Medhat and Schmeling,
r
2022).4 The success of the switching strategy relies on the relationship between preceding market volatility
er
and future momentum plus reversal returns. In this regard, historical volatility predicts negative momentum
returns, and this relationship is stronger when the level of volatility is higher. For instance, previous studies
pe
(Barroso and Santa-Clara, 2015; Daniel and Moskowitz, 2016; and Butt, Kolari, and Sadaqat, 2021) have
documented that most momentum crashes occur when market volatility is relatively high.5 One possible
reason is that the demand for liquidity increases at such times, which causes momentum returns to decline
ot
(Jylha, Rinne, and Suominen, 2013; and Ignashkina, Rinne, and Suominen, 2022).
Using market volatility as an indicator of depressed market conditions, Butt, Högholm, and Sadaqat
tn
(2021) for emerging markets and Nagel (2012) for the U.S. market have shown that reversal-related returns
increase when market volatility is relatively high. Some researchers have rationalized this positive
rin
2 Barroso and Santa-Clara (2015) used a constant such that overall volatility of the scaled momentum strategy
remained at 12% on an annual basis. Similarly, Daniel and Moskowitz (2016) used a constant to set the overall
volatility of the scaled strategy equal to the original momentum strategy.
ep
3 The authors proposed a momentum strategy that only scales returns associated with extreme volatility-related states.
4See also studies on long-run reversion by Fama and French (1988), Lo and MacKinlay (1988), Poterba and
Summers (1988), and Kiojen, Rodriguez, and Sbuelz (2009).
Pr
5 For example, Butt, Sadaqat, and Tahir (2022) found that down-market states (with negative cumulative returns over
the last consecutive 24 months) are concentrated in the 5th quintile of the market volatility in which most momentum
crashes occur.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
relationship between the market volatility and reversal returns in the context of liquidity costs, which
ed
increase when the market is depressed (e.g., see Nagel, 2012; Jylha, Rinne, and Suominen, 2013; Cheng,
Hameed, Subrahmanyam, and Titman, 2017; Ignashkina, Rinne, and Suominen, 2022; Medhat and
iew
Schmeling, 2022; among others). More explicitly stated, reversal returns are compensation for providing
market liquidity to those who demand immediacy to trade, such as momentum traders, in adverse market
conditions. The switching strategy capitalizes these intertemporal patterns in momentum and reversal
ev
Extending previous studies, we implement a switching strategy that changes from momentum to
reversal in month t+1 when the volatility of current month t lies in the fourth quartile of historical volatility
r
over the last five years. If the market volatility of the current month belongs to the first three quarters of
er
historical market volatility, the momentum strategy is continued. Two significant advantages of this strategy
are that, unlike scaled strategies, the leverage factor remains constant, and no ex-post information is used
pe
to control for volatility. The switching strategy is a self-financing, zero-investment strategy with one dollar
long and short. Our empirical findings indicate that the strategy outperforms traditional momentum and
reversal strategies in the U.S. stock market. Additionally, the switching strategy outperforms a scaled
ot
momentum strategy based on volatility when the scaling factors are chosen on an ex-ante basis. For
example, in the period 1927:1 to 2020:4, the Sharpe ratios for momentum, reversal, and scaled momentum
tn
strategies are 13.22, 14.81, and 19.72, respectively; by comparison, the Sharpe ratio for the switching
strategy is 22.64. Subsequently, the end-of-period investment size of one dollar is greater for the switching
rin
strategy than other strategies. The main reason for this outperformance is avoidance of momentum crashes
that occur when market-based volatility is relatively high. Further support for the switching strategy is
evidenced by higher alphas in spanning regressions of the switching strategy on all other momentum
ep
strategies.
As already mentioned, volatility scaling substantially decreases the extent of momentum crashes. In
Pr
our sample period, the 15 worst crashes had a monthly average of -36.30%. Scaling by different measures
of volatility reduced the magnitude of these crashes to -14.70%. Strikingly, we find that the switching
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
strategy eliminates these losses by earning on average 2.30% per month in these crash periods. Also, for
ed
the two worst U.S. crash periods, the switching strategy performed substantially better than the scaled
momentum strategy.
iew
Extending our analyses to international stock markets, we test the efficacy of the switching strategy in
five highly capitalized developed and five emerging markets. For all ten stock markets, the switching
strategy performs better than traditional momentum, and for seven countries the switching strategy
performs better than volatility scaled momentum strategy. The results for Japan as a developed country and
ev
China as an emerging country add new insights in view of the fact that previous studies have not detected
momentum in these markets. For these countries, our switching strategy generates monthly returns of 1.19%
r
and of 1.60%, respectively.
er
The rest of the paper is organized as follows. Section 2 describes the relationship between market
volatility and returns on momentum and reversal strategies. Section 3 compares volatility scaled momentum
pe
to the proposed switching strategy. Section 4 documents momentum crashes and the performance of various
momentum strategies. Section 5 presents out-of-sample evidence for other developed markets in addition
According to previous studies by Grundy and Martin (2001), Cooper, Gutierrez, and Hameed (2004), and
others, state dependency plays an important role in explaining momentum returns. After controlling for
rin
market states, Wang and Xu (2015) found that higher market volatility predicts lower momentum returns.
Similarly, Barroso and Santa-Clara (2015) and Daniel and Moskowitz (2016) reported a negative and
predictive relationship between market volatility and momentum returns. They proposed scaling
ep
momentum strategies by the inverse of the volatility to mitigate downside risk (including crashes) as well
as boost Sharpe ratio and alpha performance metrics. Volatility scaling has been shown to enhance
Pr
momentum strategies outside the U.S. also (e.g., see Hanauer and Windmuller, 2023; and Butt, Kolari, and
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Sadaqat, 2021). According to Bongaerts, Kang, and van Dijk (2020), volatility scaling of momentum returns
ed
is particularly beneficial when market volatility is relatively high.
Contrary to momentum returns, higher market volatility predicts higher reversal returns in different
iew
markets.6 There is considerable literature that links higher reversal returns with compensation for providing
liquidity to the market under duress conditions. Demsetz (1968) rationalized liquidity provision
compensation for the person who stands ready to fulfill orders from those who aspire to trade immediately.
Nagel (2012) argued that reversal profits proxying returns for liquidity provision in the market are higher
ev
when expected volatility is high in the U.S. market.7 As discussed by Gromb and Vayanos (2002) and
Brunnermeier and Pedersen (2009), higher volatility tightens funding constraints, which results in higher
r
liquidity provision costs. Similarly, studies by Jylha, Rinne, and Suominen (2013) and Ignashkina, Rinne,
er
and Suominen (2022) have shown that the positive exposure of hedge funds and mutual funds to reversal
returns indicates that they supply liquidity to the market. It is worth mentioning that the linkage between
pe
reversal profits and liquidity provision costs does not necessarily mean that reversal-related profits are
momentum/reversal strategies have important investment-related implications that have not been studied
previously in the literature to our knowledge.9 In this paper we propose a switching strategy that invests in
tn
momentum when volatility is low and reversal when volatility is high. Dobrynskaya (2019) discussed the
6 The short-term reversal strategy goes long losers and short winners in the previous month. Studies by Jegadeesh
rin
(1990), Goyal and Wahal (2015), Bali, Engle, and Murray (2016), and others have shown that reversal strategies are
profitable for the U.S. market. Griffin, Kelly, and Nardari (2010) reported international evidence on the reversal
strategy.
7Higher market volatility is indicated by a higher volatility index (VIX). In emerging markets, Butt, Hogholm, and
ep
Sadaqat (2021) have shown that higher market volatility is linked to higher reversal returns.
8Studies by Kaniel, Saar, and Titman (2008) and Hendershott, Jones, and Menkveld (2011) have discussed the market
making role of individual investors.
Pr
9We should note that, in view of their negative correlation, the potential effectiveness of combining momentum and
value strategies has been discussed by Asness, Moskowitz,and Pedersen (2013).
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
mixing of momentum and reversal strategies but in a narrower setting. She observed that momentum
ed
crashes occur one-to-three months after market crashes. To avoid momentum crashes, investment was
shifted from momentum to reversal strategy after market crashes. This strategy naturally avoids reversal
iew
prone stocks that belong to the short leg of the momentum strategy and therefore mitigates momentum
crashes.
Importantly, Conrad and Yanuz (2017) identified reversal prone stocks on an ex-ante basis by using
stock-related characteristics that are associated with expected returns. They proposed a MAX momentum
ev
strategy that is long winners’ stocks (with small size and high book-to-market ratios) and short losers (with
big size and low book-to-market ratios). They found that MAX momentum did not experience significant
r
long-term reversals.
er
Three important inferences emerge from previous studies on the linkage between momentum and
reversal as well as mixed strategies. First, as in Asness, Moskowitz, and Pedersen (2013), the switching
pe
strategy can be beneficial when the constituent strategies (such as momentum and reversal) are negatively
correlated. Second, enhanced returns can be achieved by excluding reversal prone stocks from the short leg
of the momentum strategy either on an ex-ante basis (Conrad and Yanuz, 2017) or by timing the market
ot
(Dobrynskaya, 2019). Third, some recent studies have employed dual signals for constructing investment
strategies. For example, Medhat and Schmeling (2022) utilized signals based on the previous months’
tn
returns and share turnover. More specifically, given that the previous month’s thinly traded stocks had
short-term reversals and heavily traded stocks exhibited short-term momentum, they found that both a short-
rin
term reversal strategy and a short-term momentum strategy can generate significant negative and positive
returns, respectively. However, they did not combine these two strategies as we do in the present study.
Lastly, Blitz, Hanauer, Honarvar, Huisman, and Vliet (2022) developed investment strategies based on
ep
multiple signals, including short-term reversal, momentum, analyst revisions, monthly seasonality, and risk.
Our study contributes to the literature on switching strategies by clarifying the mechanism that induces
Pr
different patterns in momentum- and reversal-related returns when market volatility is relatively high. We
use market volatility as a signal for enhancing momentum returns. Given that an increase in market
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
volatility indicates a depressed market condition that increases the cost of providing liquidity in the market,
ed
reversal returns tend to increase whereas momentum returns decrease.10 Similar results in the context of
mutual funds are reported by Ignashkina, Rinne, and Suominen (2022). Funds that had positive exposure
iew
to the reversal strategy obtain liquidity provision related premiums, and funds that had positive exposure to
We propose that, when market volatility is relatively high, distinctive intertemporal return and cost
patterns of momentum and reversal strategies exist related to liquidity supply and demand in the market.
ev
Consistent with these patterns, we implement a switching strategy that relies on market volatility as a
persistent indicator.11 In our strategy, relatively low (high) volatility provides an ex-ante signal to invest in
r
momentum (reversal).
2.2
er
U.S. Data and Initial Evidence on the Switching Strategy
U.S. stock market data is downloaded from Kenneth French’s website. Daily and monthly value-weighted
pe
market returns are gathered for momentum and reversal strategies in the sample period from 1927:1 to
2020:4. The momentum strategy is the difference between the monthly returns of the winner and loser
portfolios in month t + 2. The winner and loser portfolios are comprised of the 10% best and 10% worst
ot
performing stocks traded on the NYSE, AMEX, and NASDAQ exchanges in the previous 12 months t – 11
to t. By contrast, on the basis of performance in month t, the reversal strategy is the difference between the
tn
monthly returns of the 10% loser and 10% winner stocks in month t + 1.
Monthly market volatility is estimated using value-weighted daily stock returns. Volatility states
rin
are identified using quartiles in the sample period. We average momentum and reversal returns in these
volatility-related quartiles. The results are shown in Table 1. Our findings confirm that momentum returns
are lower and reversal returns are higher in the 4th quartile of volatility. As such, a switching strategy that
ep
invests in momentum in the first three quartiles and reversal in the 4th quartile yields the highest monthly
10
Reversal returns are also considered as a proxy for the cost of liquidity provision by Nagel (2012) and Butt,
Pr
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
average return of 1.49% with a highly significant t-statistic equal to 7.56. The main advantage of the
ed
switching strategy is that it gains from the reversal strategy in volatile times when the premium associated
with liquidity provision is generally higher. In recessions, it is expected the cost of liquidity provision will
iew
be relatively higher than otherwise (Moreira and Muir 2017). Consequently, in the last column of Table 1,
it is not surprising that momentum as a proxy for liquidity demand is weak and reversal as a proxy for
liquidity provision is strong. Hence, switching from momentum to reversal in recessions12 enhances
momentum returns.
ev
[INSERT TABLE 1 ABOUT HERE]
Can the relationship between market volatility and returns on momentum and reversal strategies be
r
used on an ex-ante basis for investment purposes? Given that many studies have reported persistent
er
volatility, we test whether historical volatility provides a signal to implement the switching strategy. To do
so, we demarcate volatility-related quartiles based on a rolling window of five years.13 If volatility in the
pe
last month within the five-year window14 belongs to one of the quartiles, we record the next month return15
for the momentum /reversal strategy within that quartile. We repeat this process until the end of the sample
period and then average the returns on the momentum/reversal strategies in their respective quartiles. Figure
ot
1 illustrates a clear pattern – that is, momentum returns are higher in first three quartiles but decline in the
4th quartile, whereas reversal returns have opposite patterns. These ex-ante analyses indicate that the
tn
12These recession periods are identified by National Bureau of Economic Research (NBER) for the time period of
rin
13Our results are robust with respect to the estimation of current and historical volatility on the basis of one month to
six months of daily returns as well as an expanding window. Results are reported in Appendix-A.
ep
14An initial five-year window from 1927:1 to 1931:12 for estimating historical volatility and volatility in the last
month of 1931:12 indicates the current state of volatility.
15 In addition to predicting returns for t+1 month, as a robustness check, we also computed the returns for months t +
2 to t + 7 separately. These results are reported in Appendix-A.
Pr
16The success of the switching strategy is not dependent on volatility-related quartiles. For detailed analyses, see
Appendix-A.
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
[INSERT FIGURE 1 ABOUT HERE]
ed
3. Momentum Enhancing Strategies
To construct the switching strategy, we define the signal of the current state of market volatility 𝑄𝑖,𝑡 in
iew
connection to the previous five years of historical market volatility. We demarcate four states in the
historical volatility and determine whether the current month’s volatility 𝑄𝑖,𝑡 belongs to one of these quartile
states 𝑖 = 1,2,3,4. The variable 𝑄𝑖,𝑡 takes the value of 1 or 0 to indicate the state of market volatility. Using
ev
this volatility-based signal, the switching strategy generates time series returns defined as 𝑄4,𝑡 ∗ 𝑟𝑒𝑣𝑡 + 1 +
(1 ‒ 𝑄4,𝑡) ∗ 𝑚𝑜𝑚𝑡 + 1. That is, we invest in a reversal strategy when market volatility is higher (𝑄4,𝑡) and
in the momentum strategy otherwise (1 ‒ 𝑄4,𝑡). Table 2 contains descriptive statistics for the switching
r
strategy. Compared to the momentum (MOM) and reversal (REV) strategies, in the sample period 1932:1
er
to 2020:4, our switching strategy (SWITCH) performs better in terms of average monthly returns, Sharpe
ratio, and end-of-period wealth (denoted WI).17 With regard to alphas associated with Capital Asset Pricing
pe
Model (CAPM) of Sharpe (1964) and others, Fama and French (1993) three-factor model, Carhart (1997)
four-factor model, and Fama and French (2018) six-factor model, there are no significant differences.
However, to price the momentum strategy, a momentum-specific factor reduces mispricing, as in the
ot
Carhart (1997) and Fama and French (2018) six factor models.18
As already discussed, volatility scaling improves momentum returns. In its simplest form, volatility scaling
rin
can be implemented as 𝑆𝑀𝑂𝑀 = 𝑀𝑂𝑀𝑡 + 1 𝜎𝑡,126, where 𝜎𝑡,126 is the volatility in the last 126 days of
market returns. This market volatility is available to the investor on an ex-ante basis for making an
ep
investment in the next month. The inverse of 𝜎𝑡,126 corresponds to the amount for which an investor can
17End-of-period wealth is the estimate of the size of one dollar invested in a strategy from the start of the sample
period (1932:1) to the end (2020:4).
Pr
In unreported results, we find that asset pricing models that do not explicitly contain a momentum factor by Hou,
18
Xue, and Zhang (2015) and Stambaugh and Yuan (2017) suppressed the momentum anomaly also.
10
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
go short in previous losers and long in previous winners. Such scaled momentum strategies have excessive
ed
leverage in comparison to one dollar long-short strategies.19 For instance, the minimum leverage of going
short in losers is 28.55 and maximum is 366.60. Figure 2 traces this leverage for the period from 1932:1 to
iew
2020:4. The performance of scaled momentum strategy SMOM is shown in the fourth row of Table 2. As
shown there, the attractiveness of this strategy is a higher Sharpe ratio. However, the minimum return and
value-at-risk (VaR) equal to 5% are so large that such a strategy cannot be realistically implemented.
ev
Interestingly, the higher Sharpe ratio of scaled momentum is directly proportional to leverage. Figure
3 depicts this relationship by plotting Sharpe ratios on the Y-axis for momentum estimated for a given
r
leverage on the X-axis. As the maximum leverage limit increases, the Sharpe ratio increases also. The
er
scaled strategy requires leverage up to 140 times higher to achieve a Sharpe ratio equal to the switching
strategy. After leverage of 170, there is little or no increase in Sharpe ratios. These analyses suggest that a
pe
target level of volatility for scaling momentum must be chosen to make it a feasible investment choice.
Although trailing market volatility is available to investors on an ex-ante basis, due to high leverage and
increased volatility of the scaled momentum, it is not practically useful to investors. The common
tn
adjustment for target volatility is by means of some constant. For instance, if the target volatility (𝜎𝑡𝑎𝑟𝑔𝑒𝑡)
is the full sample volatility of the momentum strategy (e.g., 7.72%), the constant is a simple fraction of
rin
target volatility to the volatility of the scaled momentum strategy (𝜎𝑠𝑚𝑜𝑚).20 This constant is then
𝜎𝑡𝑎𝑟𝑔𝑒𝑡 𝜎
𝑠𝑚𝑜𝑚. The constant 𝐶 controls the volatility of the scaled strategy but is not available on an ex-
ep
19 Several studies have discussed potential issues (including leverage) in implementing volatility-scaled strategies. See
Caderburg, Doherty, Wang, and Yan (2019), Liu, Tang and Zhou (2019), Bongaerts, Kang, and van Dijk (2020), and
Pr
others.
20 Daniel and Moskowitz (2016) used the same procedure. Barroso and Santa-Clara (2015) chose a target volatility
of 12%.
11
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
ante basis (Liu, Tang, and Zhou, 2019; and Bongaerts, Kang, and van Dijk, 2020). In the present study,
ed
based on the volatilities, we set 𝐶 = 0.01. This look ahead bias is important to mention as it is directly
iew
We conducted a simple exercise of simulating 50,000 times the end-of-period wealth index (WI) for
investing in the scaled momentum strategy with the range of 𝐶 = [0.001 , 0.01]. Figure 4 gives the
frequency distribution of WI. Almost 75% of the time, WI is less than $558,829, which corresponds to the
switching strategy’s WI in Table 2. Different ranges of 𝐶 yield different distributions of WI. At times, the
ev
odds by chance favor the scaled momentum strategy; however, the important choice of 𝐶 cannot be
determined on an ex-ante basis with surety of ex-post superiority over the switching strategy.
r
In the fifth row of Table 2, we report the performance of SMOM(C) for which the volatility of the scaled
er
strategy is set equal to the volatility of the traditional momentum strategy. As expected, the Sharpe ratio
and WI for the scaled strategy are higher than the switching strategy. It is apparent that scaling by 𝐶 does a
pe
good job of managing the risk of scaled momentum. As in Barroso and Santa-Clara (2015), the range of the
leverage is now between a minimum of 0.18 and a maximum of 2.35. Even so, as discussed previously, this
An ex-ante scaling factor 𝐶 can be estimated using the volatility history of both the momentum and
tn
scaled momentum strategies. The resultant scaling is accomplished on a rolling basis using the rolling
constant 𝑉𝐶𝑡 = σ𝑡𝑎𝑟𝑔𝑒𝑡,𝑡 ‒ 6 σ𝑠𝑚𝑜𝑚,𝑡 ‒ 6. For the first period, 𝑉𝐶𝑡 is estimated for the time series from 1927:2
rin
to 1931:12; subsequently, scaled momentum returns for 1932:1 are multiplied by 𝑉𝐶𝑡. This process is
repeated to the end of the sample period. The time series of scaled momentum returns from 1932:1 to 2020:4
is calculated using both volatility and the constant on an ex-ante basis, or 𝑆𝑀𝑂𝑀(𝑉𝐶) = (𝑀𝑂𝑀𝑡 + 1 𝜎𝑡,126)
ep
𝑋 𝑉𝐶𝑡. In the sixth row of Table 2, we report the performance of scaled momentum SMOM(VC). This
strategy outperforms MOM with both higher Sharpe ratio and WI in addition to other risk management
Pr
12
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
indicators such as minimum return and skewness with slightly higher model-based alphas. These findings
ed
indicate that, on an ex-ante basis, volatility scaling improves the performance of the momentum strategy.
Importantly, the switching strategy has better performance than SMOM(VC) with respect to the Sharpe
iew
ratio and WI as well as risk management with reduced minimum return, positive skewness, and lower VaR
at 5%. Notice also that the leverage factor remains constant for this strategy.
In addition to using market volatility for scaling purposes, we use the volatility of momentum returns
as a scaling factor. When target volatility for scaled momentum is chosen on an ex-post basis, this strategy
ev
outperforms others. For instance, the Sharpe ratio is almost two times higher than the momentum strategy
MOM. Also, WI is very large, and the risk of the strategy is well managed with somewhat larger alphas.
r
However, look ahead bias casts doubt on the feasibility of this strategy. When target volatility is calculated
er
on an ex-ante basis, the last line in Table 2 shows that the performance of SMOM(VC-M) is no better than
others, including SWITCH, MOM, SMOM(VC), and SMOM(VC-M). Based on time series of monthly
returns, Table 3 shows that SWITCH has the largest and most significant alpha term equal to 0.009
ot
compared to other strategies. Also, the lower R-squared value of only 26.4 percent for SWITCH compared
to 82.9 percent or higher for the other strategies suggests that SWITCH is quite different from other
tn
strategies.
As mentioned earlier, the momentum strategy is prone to large declines in returns known as momentum
crashes. Kothari and Shanken (1992) have documented that, when the market is downward trending for
ep
some time, stocks with higher/lower returns are likely to be lower/higher beta stocks. This scenario causes
the momentum strategy to yield lower returns when the market rebounds, as higher beta stocks are
Pr
associated with the short leg (losers) of the momentum strategy. Findings in Daniel and Moskowitz (2016)
13
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
have documented momentum crashes in down market states21 when market volatility is relatively high and
ed
the market rebounds.
According to our findings in Table 1 and Figure 1, momentum crashes occur when market volatility is
iew
relatively high.22 Hence, it is not surprising that, by reducing the investment in the momentum strategy in
these high states, the magnitudes of crashes are reduced. Volatility scaling works to achieve this benefit.
Extending previous work, we propose to mitigate momentum crashes via volatility-related signals. When
the current month’s volatility is in the 4th quartile of historical market volatility, it is advantageous to switch
ev
the investment from a momentum strategy to a reversal strategy in the next month. Table 4 provides results
on the efficacy of this switching strategy compared to other momentum strategies. For the 15 largest
r
momentum crashes with an average monthly return of -36.30%, the switching strategy remarkably avoids
er
losses and instead earns on average 2.30%. Comparatively, other volatility-based scaling strategies reduce
the effects of these crashes but not as dramatically. When volatility scaling is carried out with look ahead
pe
bias using SMOM(C) and SMOM(C-M), momentum crashes are reduced to -15.20% and -14.70%,
respectively. Also, upon using realistic volatility scaling strategies SMOM(VC) and SMOM(VC-M), crashes
It has been reported in prior studies that, once crashes occur, it takes almost a decade for an investor to
tn
recover their initial investment. Barroso and Santa-Clara (2015) identified the two time periods 1930:01 to
1939:12 and 2000:01 to 2009:12 in which traditional momentum did not recover the initial amount of one
rin
dollar but scaled momentum did provide some recovery. We replicated their analyses with the addition of
the switching strategy. Figure 5 shows the results for both time periods. It is apparent that the performance
ep
21 The down-market state is roughly defined as negative cumulative market index returns for the last 24 months. Some
authors have used 36 months for this definition.
Pr
22 Most momentum crashes occur when market volatility is in the 4th quartile (Butt, Sadaqat, and Tahir, 2022).
14
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
of the switching strategy is much better than scaled momentum. We infer that the switching strategy is more
ed
efficacious as a remedy for momentum crashes.
iew
We further investigate the periods 1930:01 to 1939:12 and 2001:01 to 2019:12 to better understand
the contribution of the REV strategy to boosting the performance of SWITCH. We compare the performance
of MOM, REV, and SWITCH strategies in the full sample period as well as low volatility and high volatility
sample periods. For instance, in full sample period 1930:01 to 1939:12, as shown in Panel A of Table 5,
ev
SWITCH yields positive returns due to higher REV returns (i.e., MOM returns are negative). In Panel B, the
MOM strategy has higher returns when market volatility is relatively lower compared to high volatility
r
times, and the opposite effect is obvious for the REV strategy. Out of 96 months in the period of 1930:01
er
to 1939:12, the MOM strategy switched to REV strategy for 31 months on the basis of the market volatility
related indicator of higher volatility. For the period of 2001:01 to 2019:12, the intertemporal patterns of
pe
returns are more interesting, as in this sample period overall returns for the MOM and REV strategies are
lower than the SWITCH strategy. These results are attributable to the REV strategy providing positive
returns when market volatility is relatively high. There are 41-out-of-120 months when the REV strategy is
ot
mixed with the MOM strategy to enhance SWITCH returns. More importantly, the mixing of REV with
MOM in conjunction with volatility-related indictors helps to mitigate the tail risk of SWITCH strategy.
tn
We further tested the switching strategy for 200 U.S. industry-based portfolios. Data are downloaded
from Refinitiv for the period from1978:1 to 2020:4. As shown in Figure B.1 of Appendix B, industry-
related momentum returns increase for the first three quartiles of market volatility; however, in the fourth
ep
quartile, when market volatility is higher, momentum returns decline. For reversal-based industry returns,
there is no obvious pattern, except for the fourth quartile in which reversal returns are highest. Subsequently,
Pr
the results reported in Table B.1 for industry-based momentum strategies are very similar to the results
reported in Table 2 for individual stocks. For instance, momentum strategy MOM has economically and
15
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
statistically significant returns for the industry-related portfolio as previously reported by Moskowitz and
ed
Grinblatt (1999); however, the reversal strategy is not significant for industry-based portfolios. More
importantly, as shown in Table B.1, the switching strategy outperformed the scaled momentum strategies
iew
SMOM(C) and SMOM(VC). As a further out-of-sample test, we implemented the SWITCH strategy based
on recently proposed short-term momentum (STMOM) and short-term reversal (STREV) strategies by
Medhat and Schmeling (2022). Based on data generously shared by the authors, Appendix C contains the
results. In general, as shown in Figure C.1 and Table C.1, SWITCH improves upon the performance of
ev
STMOM better than volatility scaling and has relatively lower crash risk also.
5. Robustness Tests
r
In this section we test whether the switching strategy improves the performance of unscaled and scaled
er
momentum strategies in other developed and emerging market countries. For these markets to conserve
space, momentum returns are only scaled by market volatility, and the volatility of the momentum strategy
pe
is not examined.
Here we extend the analyses to other developed markets, including Canada, France, Germany, Japan, and
ot
the United Kingdom (U.K.). Data are downloaded from Refinitiv for the period from 1981:1 to 2020:4.
Following previous studies by Ince and Porter (2006), Griffin, Kelly, and Nardari (2010), and others, we
tn
implement data cleaning procedures in each market. The process of demarcating the volatility quartiles
indicating the current state of volatility and applying the switching strategy is the same as in the U.S. market.
rin
The initial window for demarcating quartiles of realized volatility in each market is from 1981:01 to
1985:12. Based on the positioning of volatility in 1985:12 among the volatility quartiles, the investment for
the month of 1986:01 is implemented using either the momentum or reversal strategy.
ep
Pr
16
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Average returns for the momentum and reversal strategies across the five developed markets are shown
ed
in Figure 6.23 The overall results are similar to those for the U.S. When market volatility is in the 4th quartile,
momentum returns decline (bar with vertical lines), and reversal gains occur (bar with horizontal lines). In
iew
lower volatility quartiles, momentum returns are always higher than reversal returns. We next construct the
switching strategy for each market and compare the results to the unscaled and scaled momentum strategies.
As reported in Table 6, except for France, the switching strategy has a higher Sharpe ratio than momentum,
ev
[INSERT FIGURE 6 ABOUT HERE]
In Panel A of Table 6, momentum returns are higher than reversal returns in the Canadian market.
r
However, due to the opposite relationship between these strategies with market volatility, the switching
er
strategy exhibits stronger performance metrics – namely, a higher Sharpe ratio with less tail risk as indicated
by lower minimum returns (i.e., VaR equal to 5%) and therefore higher wealth index (WI) of $23,627.
pe
These metrics are improved for the switching strategy compared to the scaled momentum SMOM(VC)
strategy. In Panel B for France, the switching strategy does not outperform other strategies; nevertheless,
in unreported analyses, when market volatility is in the 4th quartile, the reversal strategy outperforms the
ot
momentum strategy. As such, the overall relationship between market volatility and future returns based
on reversal and momentum strategies remains intact. Additionally, in Panel C for Germany, the switching
tn
Panel D reports the interesting case of Japan. As previously noted by Chaves (2016), momentum returns
are not strong in the Japanese market, but reversal is significant. Importantly, the switching strategy
combining MOM and REV outperforms both of the latter strategies. Particularly for Japan, market volatility
ep
is strongly linked with future returns on momentum and reversal strategies. For instance, in the 4th quartile
Pr
23To conserve space, results across all markets are shown. Results for individual markets are available upon the
request.
17
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
of volatility, average monthly momentum returns equal -1.62% and reversal returns equal 2.53%.
ed
Contrarily, momentum returns are improved in lower volatility-related quartiles more than reversal returns.
For these reasons, the switching strategy is a very successful strategy in Japan. The Sharpe ratio and WI are
iew
noticeably higher for the switching strategy than both versions of scaled momentum strategies.
Finally, in Panel E for the U.K., given positive/negative momentum/reversal returns, the switching
strategy outperforms both strategies in terms of higher Sharpe ratio and WI. Its performance is slightly
ev
We should mention that our results in Table 6 are robust for strategies constructed based on both
pentiles and deciles using equal-weighted returns. For value-weighted returns in Germany and the U.K.,
r
we find that the switching strategy performs worse than the simple momentum strategy. This outcome can
er
be explained by reversal that is more common among smaller and illiquid stocks (Avramov, Chordia and
Goyal, 2006). Therefore, using a value-weighted scheme, the impact of small and illiquid stocks is reduced.
pe
In this respect, the results for the Japan are consistent irrespective of investment weighing schemes and
percentiles used for the construction of portfolios. The main reason is that the reversal for Japan is strong
The five largest capitalization emerging market countries of Brazil, China, India, South Korea, and Taiwan
tn
are selected to further investigate the performance of the switching strategy. Data are downloaded from
Refinitiv for the period from 1995:1 to 2020:4. As in the U.S. and other developed markets, the rolling
rin
window is comprised of five years. Since the first rolling window is in the period from 1995:1 to 1999:12,
Average returns for momentum and reversal strategies across the five emerging markets24 are shown in
ep
Figure 7. In line with previous evidence in Rouwenhorst (1998), Griffin, Kelly, and Nardari (2010), Butt,
Pr
24To conserve the space, the results across all markets are shown. Results for individual markets are available upon
request from the authors.
18
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Kolari, and Sadaqat (2021), and others, momentum returns (bar with vertical lines) are lower on average in
ed
emerging markets compared to the U.S. and other developed markets. Consistent with our earlier findings,
notice that momentum returns decline and reversal gains occur (bar with horizontal lines) when market
iew
volatility is in the 4th quartile. Subsequently, we repeat the process for constructing the switching strategy
for each emerging market and compare the results to the unscaled momentum and scaled momentum
strategies.
ev
In Panel A of Table 7 for Brazil, unscaled momentum returns are negatively significant, and there is a
strong reversal effect. The switching strategy improves upon momentum returns but understandably
r
performs below the reversal strategy. Nevertheless, the switching strategy materially outperforms the scaled
er
momentum strategies. In Panel B for China, weak momentum and strong reversal are obvious. Again, the
switching strategy yields better results than both scaled momentum strategies. The results for India in Panel
pe
C are similar. Only for Taiwan in Panel E and partially for South Korea in Panel D, the scaled momentum
6. Conclusion
This study documented evidence that market volatility affects the returns on momentum and reversal
tn
strategies differently and that these patterns can be used as investment signals. Following high market
volatility periods, the returns on the momentum/reversal strategy are lower/higher. High market volatility
rin
indicates stress in the market (for example) due to possible recession. At such times, it is plausible to assume
that the cost of liquidity provision will increase, such that a momentum/reversal strategy experiences
losses/gains. When market volatility is relatively high, previous studies have found that returns on the
ep
reversal strategy increase in U.S. and emerging stock markets. Conversely, it is well known that market
volatility is negatively correlated with future momentum returns. These opposite patterns for momentum
Pr
and reversal strategies provide investors with an opportunity to invest in momentum when market volatility
19
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Confirming this conjecture, empirical evidence showed that implementing a switching strategy between
ed
momentum and reversal outperforms traditional and volatility scaled momentum strategies. For the U.S.
stock market in the period 1932 to 2020, the switching strategy produced average monthly returns, Sharpe
iew
ratio, and wealth index (WI) equal to 1.46%, 22.64, and $558,829, respectively. By comparison, the
momentum (volatility scaled momentum) strategy produced 1.02% (1.35%), 13.22 (17.96), and $957
($70,262), respectively. Additionally, crash risk was considerably lower for the switching strategy
compared to other momentum strategies. For the switching strategy, the minimum return in any month,
ev
skewness, and value-at-risk (VaR) at 5% were -33.20%, 0.76, and -8.22% respectively. Comparatively, the
momentum (volatility scaled momentum) strategy produced -77.02% (-47.60%), -2.48 (-0.41), and -9.93%
r
(-10.70%), respectively. Particularly noteworthy, the monthly average returns of the 15 worst crashes for
er
the momentum strategy was -36.63%, which volatility scaling reduced to -14.70%, but the switching
Corroborating our main U.S. market findings, the proposed switching strategy dominated various
momentum strategies in these tests. Among developed countries, the case of Japan was particularly
ot
interesting with average monthly return, Sharpe ratio, and WI for the switching strategy in our sample
period equal to 1.19%, 18.78%, and $62, respectively. In this regard, previous studies have reported no
tn
momentum effect in Japan. In line with those studies, we found that the average monthly return for the
momentum strategy was only 0.07% in Japan. Even with volatility scaling, average monthly returns
rin
improved negligibly to only 0.30%. Among different emerging market countries, traditional momentum
returns were absent in China also (i.e., an average monthly return of only -0.20% in our sample period).
Again, volatility scaling momentum improved matters little to only 0.26%. By contrast, the switching
ep
strategy provided an average monthly return equal to 1.60%, Sharpe ratio at 26.34%, and WI at $31.
Contributing to the literature on what drives momentum, we conclude that momentum is explained in
Pr
part by state dependency and associated trade to immediacy costs linked to demand for market liquidity.
By implication, momentum and other investment strategies can be enhanced by considering multiple market
20
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
signals. We earlier cited studies by Medhat and Schmeling (2022) and Blitz, Hanauer, Honarvar, Huisman,
ed
and Vliet (2022), who utilized multiple market signals in their investment strategies. We find that mixing
strategies not only can boost profits but can provide potential diversification benefits that mitigate tail risk.
iew
Importantly, it is possible that certain market conditions exist in which opposite intertemporal patterns in
various zero-investment strategies can be exploited for investment purposes. We found that market
volatility is such a condition with respect to the momentum/reversal switching strategy. Other market
r ev
er
pe
ot
tn
rin
ep
Pr
21
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
References
ed
Asness, C. S., Moskowitz, T. J., and Pedersen, L. H. (2013). Value and momentum everywhere. Journal
iew
Avramov, D., Chordia, T., and Goyal, A. (2006). Liquidity and autocorrelations in individual stock returns.
Bali, T. G., Engle, R. F., and Murray, S. (2016). Empirical Asset Pricing: The Cross Section of Stock
ev
Barroso, P., and Santa-Clara, P. (2015). Momentum has its moments. Journal of Financial Economics 116,
111-120.
r
Blitz, D., Hanauer, M. X., Honarvar, I., Huisman, R., and van Vliet, P. (2022). Beyond Fama-French
er
factors: Alpha from short-term signals. Working paper, Robeco Quantitative Investments, available at
SSRN 4115411.
pe
Bongaerts, D., Kang, K., and van Dijk, M. (2020). Conditional volatility targeting. Financial Analysts
Brunnermeier, M. K., and Pedersen, L. H. (2009). Market liquidity and funding liquidity. Review of
ot
Butt, H. A., Högholm, K., and Sadaqat, M. (2021). Reversal returns and expected returns from liquidity
tn
provision: Evidence from emerging markets. Journal of Multinational Financial Management 59,
Butt, H. A., Kolari, J. W., and Sadaqat, M. (2021). Revisiting momentum profits in emerging markets.
Butt, H. A., Sadaqat, M., and Tahir, M. (2022). Revisiting the performance of the scaled momentum
ep
Carhart, M. M. (1997). On persistence in mutual fund performance, Journal of Finance 52, 57-
Pr
82.
22
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Cederburg, S., O’Doherty, M. S.,Wang, F., and Yan, X. S., (2019). On the performance of volatility
ed
managed portfolios. Journal of Financial Economics 138, 95-117.
Chaves, D. B. (2016). Idiosyncratic momentum: US and international evidence. Journal of Investing 25,
iew
64-76.
Cheng, S., Hameed, A., Subrahmanyam, A., and Titman, S. (2017). Short-term reversals: The effects of
past returns and institutional exits. Journal of Financial and Quantitative Analysis 52, 143-173.
Conrad, J., and Yavuz, M. D. (2017). Momentum and reversal: Does what goes up always come down?
ev
Review of Finance 21, 555-581.
Cooper, M. J., Gutierrez, R. C. Jr., and Hameed, A. (2004). Market states and momentum. Journal of
r
Finance 59, 1345-1365.
er
Daniel, K., and Moskowitz, T. J. (2016). Momentum crashes. Journal of Financial Economics 122, 221-
247.
pe
Demsetz, H. (1968). The cost of transacting. Quarterly Journal of Economics 82, 33-53.
Dobrynskaya, V. (2019). Avoiding momentum crashes: Dynamic momentum and contrarian trading.
Engle, R. F. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United
tn
Fama, E. F., and French, K. R. (1988). Permanent and temporary components of stock prices. Journal of
rin
Fama, E. F., and French, K. R. (1993). Common risk factors in the returns on stocks and bonds.
Fama, E. F., and French, K. R. (2008). Dissecting anomalies. Journal of Finance 63, 1653-1678.
Fama, E. F., and French, K. R. (2018). Choosing factors. Journal of Financial Economics 128,
Pr
234-252.
23
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Fama, E. F., and French, K. R. (2020). Comparing cross-section and time-series factor models.
ed
Review of Financial Studies 33, 1892-1926.
Goyal, A., Jegadeesh, N. and Subrahmanyam, A. (2022). Determinants of momentum: A perspective from
iew
international data. Working paper, University of Lausanne, Emory University, and the University of
Goyal, A., and Wahal, S. (2015). Is momentum an echo? Journal of Financial and Quantitative Science
ev
50, 1237-1267.
Griffin, J. M., Kelly, P. J., and Nardari, F. (2010). Do market efficiency measures yield correct inferences?
r
A comparison of developed and emerging markets. Review of Financial Studies 23, 3225-3277.
Gromb, D., and Vayanos, D. (2002). Equilibrium and welfare in markets with financially constrained
er
arbitrageurs. Journal of Financial Economics 66, 361-407.
pe
Grundy, B. D., and Martin, J. S. M. (2001). Understanding the nature of the risks and the source of the
Guo, J., Li, P., and Li, Y (2022). What can explain momentum? Evidence from decomposition.
Hanauer, M. X., and Windmüller, S. (2023). Enhanced momentum strategies. Journal of Banking and
tn
Hendershott, T., Jones, C. M., and Menkveld, A. (2011). Does algorithmic trading improve liquidity?
Hou, K., Xue, C., and Zhang, L. (2015). Digesting anomalies: An investment approach. Review of Financial
Ignashkina, A., Rinne, K., and Suominen, M. (2022). Short-term reversals, returns to liquidity provision
and the cost of immediacy. Journal of Banking and Finance 138, Issue C, available online, 106430.
Ince, O. S. and Porter, R. B. (2006). Individual equity return data from Thomson DataStream: Handle with
Pr
24
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Jegadeesh, N. (1990). Evidence of predictable behavior of security returns. Journal of Finance 45, 881-
ed
898.
Jegadeesh, N., and Titman, S. (1993). Returns to buying winners and selling losers: Implications for stock
iew
market efficiency. Journal of Finance 48, 65-91.
Jylha, P., Rinne, K., and Suominen, M. (2013). Do hedge funds supply or demand liquidity? Review of
Kaniel, R., Saar, G., and Titman, S. (2008). Individual investor trading and stock returns. Journal of
ev
Finance 63, 273-310.
Koijen, R. S., Rodriguez, J. C., and Sbuelz, A. (2009). Momentum and mean reversion in strategic asset
r
allocation. Management Science 55, 1199-1213.
er
Kothari, S. P., and Shanken, J. (1992). Stock return variation and expected dividends: A time-series and
28.
Liu, F., Tang, X., and Zhou, G. (2019). Volatility-managed portfolio: Does it really work? Journal of
ot
Lo, A. W., and MacKinlay, A. C. (1988). Stock market prices do not follow random walks: Evidence from
tn
Medhat, M., and Schmeling, M. (2022). Short-term momentum. Review of Financial Studies 35, 1480-
rin
1526.
Moreira, A., and Muir, T. (2017). Volatility‐managed portfolios. Journal of Finance 72, 1611-1644.
Moskowitz, T. J., and Grinblatt, M. (1999). Do industries explain momentum? Journal of Finance 54, 1249-
ep
1290.
Poterba, J. and Summers, L. (1988). Mean reversion in stock returns: Evidence and implications.
25
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Rouwenhorst, K. G. (1998). International momentum strategies. Journal of Finance 53, 267-284.
ed
Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of risk. Journal
iew
Stambaugh, R. F., and Yuan, Y. (2017). Mispricing factors. Review of Financial Studies 30, 1270-1315.
Wang, K. Q., and Xu, J. (2015). Market volatility and momentum. Journal of Empirical Finance 30, 79-
91.
r ev
er
pe
ot
tn
rin
ep
Pr
26
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 1: Returns in Market Volatility Quartiles
ed
This table provides the average monthly returns for market volatility-based quartiles for momentum
(MOM) and reversal (REV) and their combined switching-related strategies in the period January 1927
to April 2020. Market volatility is estimated on a monthly basis, and volatility-related quartiles are
demarcated over the full sample. The switching strategy (SWITCH) invests in momentum for quartiles
iew
Q1, Q2, and Q3 and in reversal for the last quartile Q4. Associated t-statistics are provided in parentheses.
In the last column, the monthly returns for the strategies are averaged in the recession period as identified
by the National Bureau of Economic Research (NBER) for the time period of January 1927 to April
2020.
Strategies Average Q1 Q2 Q3 Q4 Recession
1.169 1.463 1.191 1.488 0.534 0.604
ev
MOM (4.99) (5.83) (3.94) (4.16) (0.69) (0.671)
0.901 0.602 0.664 0.525 1.813 2.015
REV (4.99) (2.84) (2.74) (1.91) (3.11) (2.892)
1.489 1.463 1.191 1.488 1.813 1.420
SWITCH (7.56) (5.83) (3.94) (4.16) (3.11) (6.719)
r
er
pe
ot
tn
rin
ep
Pr
27
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Market Volatility and Returns for 1932:1 to 2020:4
ed
0.02
0.018
Average Monthly Returns
0.016
iew
0.014
0.012
0.01 MOM
0.008 REV
0.006
0.004
ev
0.002
0
Q1 Q2 Q3 Q4
r
Figure 1. Market volatility quartiles and next month returns on momentum and reversal strategies
er
pe
ot
tn
rin
ep
Pr
28
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 2: Performance of Different Momentum Strategies
ed
This table provides the performance of various types of the momentum strategies for the period January 1932 to April 2020.
A number of performance metrics are shown, including average monthly returns (Avg), standard deviation of returns (SD),
Sharpe ratio (SR), minimum return (Min), maximum return (Max), skewness (Skew), end-of-period wealth of holding a
strategy (WI), and value-at-risk (VaR) at 5%. In the last four columns, estimated alphas are reported for the Capital Asset
iew
Pricing Model (CAPM) of Sharpe (1964) and others, Fama and French (1993) three-factor model (FF3), Carhart (1997) four-
factor model (CF4), and Fama and French (2018) six-factor model (FF6). Descriptive statistics and alphas are calculated on a
monthly basis except for WI which indicates holding period return of an initial investment of $1 that is compounded to the
end of the sample period. MOM is a momentum strategy that goes long 10th decile winners’ stocks and short 1st decile losers’
stocks based on the performance over the last 11 months. REV indicates the reversal strategy that goes long 1st decile losers
in the previous month and short 10th decile winners. SWITCH is a combination of momentum and reversal strategies based on
the previous volatility-related indicator. The switching strategy implements the momentum strategy for the next month if the
ev
current month’s market volatility is less than the 0.75 percentile of the previous 5 years of market volatility, otherwise it
chooses the reversal strategy. SMOM is the momentum strategy scaled by the volatility of daily market returns over the last
126 days. The strategy SMOM(C) is further scaled by a constant that equates its volatility with the switching strategy on an
ex-post basis. The strategy SMOM(VC) is scaled by a constant that equates the volatility of SMOM(C) with the SWITCH
r
strategy for every 5 years. In the strategy SMOM(VC), the constant is used on an ex-ante basis for scaling. For strategies
SMOM(C) and SMOM(VC), the volatility of daily returns of the momentum strategy over the last 126 days is used as in
Barroso and Santa-Clara (2015).
MOM
Strategies Avg
1.02
(4.52)
SR
13.22
Min
-77.02
Max
26.16
er
Skew WI ($)
-2.48 957
VaR
(5%)
-9.93
CAPM FF3
1.41
(7.68)
1.62
(8.45)
CF4
0.29
FF6
0.00
(3.51) (0.31)
pe
REV 0.86 14.81 -28.53 58.63 1.36 1640 -6.95 0.67 0.67 0.92 0.00
(4.92) (3.81) (3.73) (5.09) (0.40)
SWITCH 1.46 22.64 -33.2 58.63 0.76 558829 -8.22 1.41 1.45 1.09 0.01
(7.92) (7.33) (7.72) (5.28) (1.10)
180.21 - 197.65 211.16 90.61 1.18
ot
SMOM(VC) 1.35 17.96 -47.6 49.44 -0.41 70262 -10.7 1.60 1.76 0.60 0.01
(6.32) (7.98) (8.74) (4.35) (0.72)
SMOM(C-M) 2.04 26.36 -31.17 29.09 -0.3 6393590 -8.67 2.21 2.35 1.15 0.01
(7.94) (8.92) (9.47) (5.85) (1.46)
rin
29
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 3: Strategy-Based Alphas
ed
Using monthly returns from January 1932 to April 2020, columns provide estimates of the linear regression
𝑟𝑒𝑡𝑖,𝑡 = 𝛼 + 𝑟𝑒𝑡𝑗,𝑡 + 𝜀𝑡 for each strategy on other strategies, including SWITCH, MOM, SMOM(VC), and
SMOM(VC-M). Estimated coefficients (t-statistics in parentheses) are reported as well as the number of
observations and adjusted R-squared value.
iew
Strategies SWITCH MOM SMOM(VC) SMOM(VC-M)
MOM -0.445 0.241 0.231
(-8.64) (12.29) (11.59)
SMOM(VC) 0.927 0.518 0.715
(12.76) (12.29) (32.65)
SMOM(VC-M) -0.162 0.488 0.703
ev
(-2.10) (11.59) (32.65)
SWITCH -0.149 0.144 -0.026
(-8.64) (12.76) (-2.10)
Alpha (α) 0.009 -0.001 -0.001 0.002
(5.06) (-1.23) (-0.76) (2.67)
r
Adjusted R-squared 0.264 0.829 0.917 0.909
er
pe
ot
tn
rin
ep
Pr
30
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
370.00
ed
320.00
Weights on MOM
270.00
iew
220.00
170.00
120.00
70.00
ev
20.00
32
36
40
44
48
52
56
60
64
68
72
76
80
84
88
92
96
00
04
08
12
16
20
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
20
20
20
20
20
r
Figure 2: The scaling factor for the momentum strategy
er
pe
ot
tn
rin
ep
Pr
31
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Momentum and Sharpe Ratios
ed
0.260
0.240
140, [Y VALUE]
0.220
iew
Sharpe Ratios
0.200
0.180
SMOM
0.160
SWITCH
0.140
ev
0.120
0.100
20 70 120 170 220 270 320 370
r
Leverage
er
Figure 3: The leverage factor and Sharpe ratio of the scaled momentum strategy SMOM(C)
pe
ot
tn
rin
ep
Pr
32
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Distribution of Wealth Index
ed
16000 120.00%
14000
100.00%
95.54%
iew
12000
80.00%
10000 75.97%
Frequency
8000 60.00%
50.43%
6000
40.00%
ev
4000 30.86%
20.00%
2000 15.43%
0 0.00%
r
68 753 13,740 558,829 4,694,680 7,820,857
Frequency Cumulative %
er
Figure 4: The wealth index for momentum strategy SMOM(C) using a constant between 0.001 to 0.01
pe
ot
tn
rin
ep
Pr
33
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 4: Momentum Crashes
ed
This table shows the 15 most extreme momentum crashes for the momentum strategy observed by Daniel and
Moskowitz (2016). For months in which these crashes occurred, returns for the different momentum strategies are
shown. MOM is the momentum strategy that goes long 10th decile winners’ stocks and short 1st decile losers’ stocks
based on the performance over the last 11 months. REV indicates the reversal strategy that goes long in 1st decile
iew
losers in the previous month and short 10th decile winners. SWITCH is a combination of momentum and reversal
strategies based on the previous volatility-related indicator. The switching strategy implements the momentum
strategy for the next month if the current month’s market volatility is less than 0.75 percentile of the previous 5
years of market volatility, otherwise it chooses the reversal strategy. SMOM is the momentum strategy scaled by
the volatility of daily market returns over the last 126 days. The strategy SMOM(C) is further scaled by a constant
that equates the volatility of SMOM(C) with the switching strategy on an ex-post basis. The strategy SMOM(VC)
is scaled by a constant that equates its volatility with the switching strategy for every 5 years. In the strategy
ev
SMOM(VC), the constant is used on an ex-ante basis for scaling. For strategies SMOM(C-M) and SMOM(VC-M),
the volatility of the daily returns of the momentum strategy over the last 126 days is used as in Barroso and Santa-
Clara (2015).
Date MOM REV SWITCH SMOM(C) SMOM(VC) SMOM(C-M) SMOM(VC-M)
r
8:1932 -0.770 -0.180 -0.180 -0.304 -0.476 -0.303 -0.512
7:1932 -0.602 -0.078 -0.078 -0.242 -0.321 -0.257 -0.379
4:2009
9:1939
1:2001
4:1933
-0.456
-0.452
-0.420
-0.419
0.010
0.239
0.214
0.068
0.010
0.239
0.214
0.068
er -0.132
-0.348
-0.302
-0.157
-0.182
-0.438
-0.316
-0.296
-0.104
-0.374
-0.221
-0.125
-0.151
-0.665
-0.219
-0.242
pe
3:2009 -0.398 0.172 0.172 -0.115 -0.124 -0.098 -0.112
6:1938 -0.332 0.272 -0.332 -0.188 -0.226 -0.152 -0.244
4:2020 -0.287 0.384 0.384 -0.114 -0.087 -0.191 -0.226
5:1933 -0.269 -0.099 -0.099 -0.102 -0.197 -0.117 -0.232
8:2009 -0.254 -0.039 -0.254 -0.127 -0.212 -0.058 -0.106
11:2002 -0.201 0.004 0.004 -0.105 -0.135 -0.124 -0.175
ot
34
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Wealth Index: 1932:1 to 1939:12
ed
10.00
9.00
8.00
7.00
6.00
iew
5.00
4.00
3.00
2.00
1.00
0.00 $0.06
ev
-1.00
1932 1933 1934 1935 1936 1937 1938 1939
WI(MOM) WI(SWITCH) WI(SMOM)
r
Wealth Index: 2000:1 to 2009:12
5.00
4.00 er
pe
3.00
2.00 $2.20
1.00
ot
0.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
35
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 5: Performance of Momentum and Reversal in Crisis Periods.
ed
This table provides the performance of momentum (MOM), reversal (REV) and SWITCH strategies for the
sample periods 1932-1939 and 2000-2009. MOM is a momentum strategy that goes long 10th decile winners’
stocks and short 1st decile losers’ stocks based on the performance over the last 11 months. REV indicates
the reversal strategy that goes long 1st decile losers in the previous month and short 10th decile winners.
iew
SWITCH is a combination of momentum and reversal strategies based on the previous volatility-related
indicator. The switching strategy implements the momentum strategy for the next month if the current
month’s market volatility is less than the 0.75 percentile of the previous 5 years of market volatility,
otherwise it chooses the reversal strategy. In Panels A and C, we report average monthly returns (Avg),
standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum return (Max),
skewness (Skew), end-of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. In Panels
B and D, the same information is provided for momentum and reversal strategies in the cases of low
ev
volatility, or MOM(LV) and REV(LV), and high volatility, or MOM(HV) and REV(HV).
Panel A: Switch Strategy for 1932-1939
Strategies Avg SD SR Min Max Skew WI ($) VaR (5%)
MOM -0.98 15.64 -6.26 -77.02 24.99 -2.15 0.06 -28.48
r
REV 3.85 10.14 38.00 -17.97 58.63 1.84 24.87 -7.96
SWITCH 2.97 11.10 26.79 -33.20 58.63 1.12 9.78 -9.87
Panel B: Contribution of MOM and REV
MOM(LV)
MOM(HV)
1.42
-6.01
8.84
23.85
16.08
-25.21
er -33.20
-77.02
24.99
24.96
-0.58
-1.35
1.93
0.03
-9.84
-52.68
pe
REV(LV) 2.72 7.19 37.88 -10.21 27.24 0.70 4.92 -7.51
REV(HL) 6.22 14.39 43.27 -17.97 58.63 1.55 5.05 -9.16
Panel C: Switch Strategy for 2000-2009
Strategies Avg SD SR Min Max Skew WI VaR
MOM 0.22 11.19 1.95 -45.58 26.16 -1.33 0.55 -16.70
ot
36
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Market Volatility and Returns for
ed
Developed Markets
0.02
0.018
iew
Monthly Average Returns
0.016
0.014
MOM
0.012
0.01 REV
0.008
0.006
0.004
ev
0.002
0
-0.002 Q1 Q2 Q3 Q4
-0.004
r
Figure 6: Market volatility quartiles and average next month returns on the momentum and reversal
strategies for developed markets in the period 1986:1 to 2020:4
er
pe
ot
tn
rin
ep
Pr
37
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 6: Performance of Different Momentum Strategies in Developed Markets
ed
This table provides the performance of various momentum strategies for Canada, France, Germany, Japan,
and the United Kingdom (U.K.). A number of performance metrics are shown, including average monthly
returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum
return (Max), end-of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. The sample
iew
period is from January 1986 to April 2020. MOM indicates the momentum strategy that goes long in 10th
decile winners’ stocks and short in 1st decile losers’ stocks based on the performance over the last 11
months. REV indicates the reversal strategy that goes long in 1st decile losers of the previous month and
short 10th decile winners. The switching strategy is a combination of momentum and reversal strategies
based on previous volatility related indicator. SWITCH chooses the momentum strategy for the next month
if the current month’s market volatility is less than 0.75 percentile of previous 5 years of market volatility,
ev
otherwise it chooses the reversal strategy SMOM is the volatility scaled momentum strategy, which is very
similar to the one used in Barroso and Santa-Clara (2015), except that the market volatility over the last
126 is used instead of the volatility of the momentum strategy. SMOM is the momentum strategy scaled by
the volatility of the daily market returns over the last 126 days. The strategy SMOM(C) is further scaled by
r
a constant that equates its volatility with the switching strategy on an ex-post basis. The strategy
SMOM(VC) is scaled by a constant that equates the volatility of SMOM(C) with the switching strategy for
every 5 years. In the strategy SMOM(VC), the constant is used on an ex-ante basis for scaling. All results
except WI($) are in percentages.
Panel A: Canada
Strategies Avg SD
er
SR Min Max WI ($) VaR (5%)
pe
MOM 1.95 8.80 22.10 -30.17 40.39 617 -12.53
REV 0.87 8.17 10.63 -54.99 43.33 9 -11.37
SWITCH 2.80 8.52 32.91 -23.99 43.33 23,627 -10.79
SMOM 351.99 1298.47 27.11 -3587.69 5359.95 -1551.14
SMOM(C) 2.39 8.80 27.11 -24.32 36.33 3,955 -10.51
ot
38
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Tab6, continued
ed
Panel D: Japan
MOM 0.07 6.35 1.06 -49.72 22.16 1 -9.62
REV 0.83 5.83 14.28 -11.26 48.78 16 -7.16
SWITCH 1.19 6.32 18.78 -19.28 48.78 62 -8.15
iew
SMOM 36.86 700.68 5.26 -3614.53 3095.76 -945.56
SMOM(C) 0.33 6.35 5.26 -32.75 28.05 2 -8.57
SMOM(VC) 0.30 5.88 5.10 -29.63 23.88 2 -8.76
Panel E: U.K.
MOM 1.01 8.45 11.92 -34.16 40.63 14 -12.41
ev
REV -0.63 7.15 -8.88 -39.64 35.44 0 -10.52
SWITCH 1.24 8.01 15.49 -39.64 35.44 43 -11.21
SMOM 221.09 1229.06 17.99 -6402.59 6787.49 -1432.85
SMOM(C) 1.52 8.45 17.99 -44.04 46.69 118 -9.86
r
SMOM(VC) 1.29 8.31 15.47 -58.72 33.08 43 -9.91
er
pe
ot
tn
rin
ep
Pr
39
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
ed
Market Volatility and Returns for
Emerging Markets
iew
0.025
0.02
0.015
0.01 MOM
REV
ev
0.005
0
Q1 Q2 Q3 Q4
-0.005
r
-0.01
er
Figure 7: Market volatility quartiles and average next month returns on the momentum and reversal strategies for
emerging markets in the period 2000:1 to 2020:4
pe
ot
tn
rin
ep
Pr
40
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 7: Performance of Different Momentum Strategies in Emerging Markets
ed
This table provides the performance of various momentum strategies for Brazil, China, India, South Korea,
and Taiwan. A number of performance metrics are shown, including average monthly returns (Avg),
standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum return (Max), end-
of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. The sample period is from
iew
January 2000 to April 2020. MOM is the momentum strategy that goes long 10th decile winners’ stocks and
short 1st decile losers’ stocks based on performance over the last 11 months. REV indicates the reversal
strategy that goes long 1st decile losers of the previous month and short 10th decile winners. The switching
strategy is a combination of momentum and reversal strategies based on the previous volatility-related
indicator. SWITCH chooses the momentum strategy for the next month if the current month’s market
volatility is less than 0.75 percentile of the previous 5 years of market volatility, otherwise it chooses the
reversal strategy. SMOM is the volatility scaled momentum strategy, which is very similar to the one used
ev
in Barroso and Santa-Clara (2015), except that the market volatility over the last 126 is used instead of the
volatility of the momentum strategy. SMOM is the momentum strategy scaled by the volatility of the daily
market returns over the last 126 days. The strategy SMOM(C) is further scaled by a constant that equates
its volatility with the switching strategy on an ex-post basis. The strategy SMOM(VC) is scaled by a constant
r
that equates the volatility of SMOM(C) with the switching strategy for every 5 years. In the strategy
SMOM(VC), the constant is used on ex-ante basis for the scaling. All results except for WI($) are in
percentages.
Panel A: Brazil
Strategies
MOM
Avg
-0.55
SD
8.39
erSR
-6.57
Min
-44.90
Max
23.04
WI ($)
0
VaR (5%)
-12.52
pe
REV 1.09 7.81 14.01 -25.56 33.35 8 -9.50
SWITCH 0.64 8.50 7.58 -32.19 33.35 2 -12.52
SMOM -18.60 692.27 -2.69 -3206.78 2088.33 -1042.74
SMOM(C) -0.23 8.39 -2.69 -38.85 25.30 0 -12.63
SMOM(VC) 0.29 7.88 3.66 -28.69 30.88 1 -11.77
ot
Panel B: China
MOM 0.09 6.05 1.51 -18.08 22.45 1 -9.34
REV 1.48 5.54 26.71 -20.21 27.90 32 -7.65
tn
Panel C: India
MOM 0.79 9.38 8.41 -68.11 30.18 2 -16.76
REV 2.17 7.84 27.70 -14.56 33.80 134 -8.17
SWITCH 1.48 8.79 16.85 -68.11 29.02 14 -11.50
ep
41
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table 7, continued
ed
Panel D: South Korea
MOM -0.08 7.96 -0.96 -29.35 24.93 0 -12.56
REV 1.77 9.46 18.71 -39.69 57.15 32 -10.07
SWITCH 0.86 7.73 11.09 -27.16 30.30 4 -10.82
iew
SMOM 41.72 518.50 8.05 -1564.40 1559.00 -824.31
SMOM(C) 0.64 7.96 8.05 -24.02 23.93 2 -12.65
SMOM(VC) 1.05 9.21 11.35 -37.37 46.91 5 -11.48
Panel E: Taiwan
MOM 0.16 7.47 2.08 -50.65 25.43 1 -11.54
ev
REV 0.09 5.82 1.51 -13.38 45.77 1 -8.21
SWITCH 0.37 6.83 5.42 -50.65 25.43 1 -8.54
SMOM 58.67 558.42 10.51 -3494.10 1582.54 -788.38
SMOM(C) 0.79 7.47 10.51 -46.77 21.18 4 -10.55
r
SMOM(VC) 0.74 6.80 10.94 -24.81 23.98 4 -10.21
er
pe
ot
tn
rin
ep
Pr
42
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Appendix A
ed
In the text, daily returns are used to estimate market volatility in the current month as well as historical
volatility over five years. As a robustness check, we estimate market volatility over two, three, four, five, and
iew
six months and then expand the window to the end of the sample period. Using various measures of market
volatility, we find in Panel A of Table A.1 that the Sharpe ratios of different versions of the SWITCH strategy
remain very close to those for the original SWITCH strategy shown in the first row of the Panel A.
Similarly, in Panel B, we conduct robustness checks for the original SWITCH strategy using different
ev
percentiles than the 75th percentile. Starting with the 50th percentile, when the current month’s market volatility
is higher than the 50th percentile of market volatility in the previous five years, the SWITCH strategy invests
r
in reversal instead of momentum. These percentiles are incrementally increased by 0.05 until the 80th percentile
er
is reached. As shown in the Panel B, SWITCH strategy results remain fairly consistent as volatility-related
percentiles are increased. We infer that the success of the SWITCH strategy is not dependent on volatility-
pe
related percentiles.
Lastly, in Panel C, we test for how long the volatility-based indicator remains profitable for the SWITCH
strategy. Our main SWITCH strategy focuses on the next month t + 1 given the current month’s volatility is
ot
higher than 75th percentile of five years of the market volatility. By contrast, SWITCHt+2 is based on month
t + 2 (with one month lag), and other strategies increase the time horizon to t + 7 (with six month lag). The
tn
results show that the volatility indicator remains profitable. For instance, MOM, SMOM(VC) and SMOM(VC-
M) based momentum strategies have Sharpe ratios equal to 13.22,17.96 and 19.72 respectively (Table 2),
whereas the SWITCH strategy with increasing time lags normally have higher Sharpe ratios. Nevertheless, the
rin
Sharpe ratio for the SWITCH strategy gradually decreases as the time between the availability and
43
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table A.1: Volatility and Robustness of SWITCH Strategy
ed
This table provides the performance results fdor various SWITCH based strategies in terms of average monthly
returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR), minimum return (Min), maximum return
(Max), skewness (Skew), end-of-period wealth of holding a strategy (WI), and value-at-risk (VaR) at 5%. In Panel
A, SWITCH is a combination of momentum and reversal strategies based on the previous volatility-related
iew
indicator. The switching strategy implements the momentum strategy for the next month if the current month’s
market volatility is less than the 0.75 percentile of the previous 5 years of market volatility, otherwise it chooses
the reversal strategy. By contrast, the subscripts m2 to m6 indicate that the previous 5 years volatility is measured
on the basis of rolling windows of 2 to 6 month, respectively. The investment criteria for SWITCHm2 to
SWITCHm6 are same as for SWITCH. Lastly, for SWITCHexp, volatility is expanded until the end of the sample
period. In Panel B, the SWITCH strategy is implemented using percentiles that increase by 0.05; that is, the next
month investment in the momentum strategy is made when the recent month’s volatility is less than 0.5 percentile
ev
(p.5) of the previous 5 years of market volatility, otherwise it chooses the reversal strategy. This process is
repeated for other percentiles such as 0.55 percentile (p.55) to the 0.80 percentile (p.80). In Panel C, the decision
to switch investment from momentum to reversal is delayed by one month. For example, SWITCHt+2 implements
the momentum strategy with one month’s lag if the current month’s market volatility is less than the 0.75
percentile of the previous 5 years of market volatility, otherwise it chooses the reversal strategy. Similarly, for
r
SWITCHt+3 to SWITCHt+7, the lag between the available information and investment decision incrementally
increases by one month.
Strategies
SWITCH
Avg
1.46
SD
6.44
SR
22.64
er
Panel A: Different Horizon of Market Volatility for SWITCH
Min
-33.20
Max
58.63
Skew
0.76
WI ($)
558,829
VaR (5%)
-8.22
pe
SWITCHm2 1.40 6.56 21.35 -45.19 58.63 0.48 270,376 -8.48
SWITCHm3 1.40 6.52 21.45 -45.19 58.63 0.59 274,097 -8.48
SWITCHm4 1.54 6.54 23.48 -45.19 58.63 0.37 1,119,475 -8.29
SWITCHm5 1.41 6.46 21.82 -45.19 58.63 0.41 318,827 -8.38
SWITCHm6 1.42 6.44 21.97 -45.19 58.63 0.45 342,078 -8.29
ot
SWITCH (p.60) 1.40 6.32 22.12 -25.04 58.63 1.07 333,346 -7.98
SWITCH (p.65) 1.45 6.36 22.86 -33.20 58.63 0.90 572,223 -7.98
SWITCH (p.70) 1.48 6.37 23.21 -33.20 58.63 0.87 731,761 -7.99
SWITCH (p.80) 1.43 6.63 21.55 -45.19 58.63 0.38 346,628 -8.38
rin
44
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Appendix B
ed
Market Volatility and Returns for
Industry Portfolios
0.019
iew
0.014
Average Monthly Returns
0.009 MOM
ev
REV
0.004
r
-0.001 Q1 Q2 Q3 Q4
-0.006
er
pe
Figure A.1: Market volatility quartiles and next month returns on momentum and reversal strategies for U.S.
industry-based portfolios
ot
tn
rin
ep
Pr
45
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table B.1: Performance of Different Momentum Strategies for U.S. Industry Portfolios
ed
This table reports the performance of various types of the momentum strategies for the period January 1978
to April 2020 for 200 U.S. industry-based portfolios. A number of performance metrics are shown,
including average monthly returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR), minimum
return (Min), maximum return (Max), end-of-period wealth of holding a strategy (WI), and value-at-risk
iew
(VaR) at 5%. MOM is the momentum strategy that goes long 10th decile winners and short 1st decile losers
among industry-based portfolios over the last 11 months. REV indicates the reversal strategy that goes long
1st decile losers of the previous month and short 10th decile winners. Switching is a combination of
momentum and reversal strategies based on the previous volatility-related indicator. SWITCH implements
the momentum strategy for the next month if the current month’s market volatility is less than 0.75
percentile of the previous 5 years of market volatility, otherwise it chooses the reversal strategy. SMOM is
the momentum strategy scaled by the volatility of the daily market returns over the last 126 days. The
ev
strategy SMOM(C) is further scaled by a constant that equates its volatility with the switching strategy on
an ex-post basis. The strategy SMOM(VC) is scaled by a constant that equates the volatility of SMOM(C)
with the switching strategy for every 5 years. In the strategy SMOM(VC), the constant is used on an ex-ante
basis for the scaling. All the results except for WI($) are in percentages.
r
Strategies Avg SD SR Min Max WI ($) VaR (5%)
MOM 0.85 6.08 13.96 -47.57 22.56 29 -8.01
REV
SWITCH
SMOM
SMOM(C)
0.06
1.00
117.75
0.88
4.62
5.68
815.08
6.06
er
1.32
17.55
14.45
14.45
-20.47
-47.57
20.11
22.56
-7243.88 2846.47
-53.87 21.17
1
68
32
-7.12
-7.37
-1074.61
-7.99
pe
SMOM(VC) 0.84 5.93 14.09 -47.89 21.38 25 -8.04
ot
tn
rin
ep
Pr
46
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Appendix C
ed
Medhat and Schmeling (2022) constructed separate short-term reversal and momentum strategies.
Departing from previous studies, the authors double sort stock returns on the previous month’s return and
iew
share turnover. In the sample period July 1963 to December 2018, the strategy goes long in previous month
winners and short the losers within the highest turnover decile and produces momentum profits (denoted
STMOM) of 16.4% per annum; conversely, the same strategy in lowest turnover decile generates reversal
profits (denoted STREV) of -16.9% per annum. Compared to the momentum strategies used in our study,
ev
their momentum and reversal strategies are different and therefore provide another opportunity to test our
switching strategy.
r
As another out-of-sample robustness test, we contacted the authors, who generously shared their
er
data. To test our switching strategy, the initial window for demarcating quartiles of realized volatility is
from 1958:07 to 1963:6. Based on the positioning of volatility in 1963:6 among the volatility quartiles, the
pe
investment for the month of 1963:07 is implemented using either the short-term momentum or short-term
reversal strategy. Average returns for STMOM and STREV strategies are shown in Figure B.1. Given that
market volatility is in the 4th quartile, momentum returns decline (bar with vertical lines) and reversal
ot
returns increase (bar with horizontal lines). The results are very similar to those in Figure1 in which
47
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
ed
Market Volatility, STMOM, and STREV
0.020
iew
0.018
0.016
Monthly Average Returns
0.014
0.012
0.010 MOM
ev
0.008 REV
0.006
0.004
0.002
r
0.000
Q1 Q2 Q3 Q4
er
Figure C.1: Market volatility quartiles and next month returns on STMOM and STREV strategies for the
period 1963:07 to 2018:12.
pe
We next replicated Table 1 using STMOM and STREV as the main testing strategies for the period
July 1963 to December 2018. Since daily series for STMOM are unavailable, we did not reproduce the last
ot
two strategies in Table 1. As shown in Table B.1, the switching strategy has better performance metrics in
comparison to short-term momentum strategy STMOM and scaled versions of STMOM including
tn
STSMOM, STSMOM(C) and STSMOM(VC).25 For instance, the Sharpe ratio of the switching strategy
(SWITCH) is 22.06, which is higher than STMOM and other volatility scaled STMOM strategies. More
importantly, cumulative returns captured by the wealth index (WI) are highest for SWITCH relative to other
rin
strategies including STREV. In terms of crash risk, the monthly average of the 10 worst returns from July
1963 to December 2018 is -25.96% for the traditional momentum strategy MOM, -23.65% for STMOM, -
ep
25.83% for SSTMOM(C), and -22.33% for SSTMOM(VC). Improving upon these losses, for SWITCH we
get -17.15%. These results suggest that SWITCH can reduce the tail risk of STMOM.
Pr
25 The short-term momentum strategy STMOM is scaled in similar way as the traditional momentum strategy as
discussed Subsections 3.1 and 3.2.
48
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008
Table C.1: Performance of Short-term Momentum Strategies for the U.S. Market
ed
This table provides the performance of various types of short-term momentum strategy proposed by Medhat and
Schmeling (2022) for the U.S. market for the period July 1968 to December 2018. Several performance metrics
are shown, including average monthly returns (Avg), standard deviation of returns (SD), Sharpe ratio (SR),
minimum return (Min), maximum return (Max), skewness (Skew), end-of-period wealth of holding a strategy
iew
(WI), and value-at-risk (VaR) at 5%. STMOM is a momentum strategy that goes long for previous month winners
and shorts the losers within the highest turnover decile. STREV goes long in previous month losers and short the
winners within lowest turnover decile. SWITCH is a combination of momentum and reversal strategies based on
the previous volatility-related indicator. The switching strategy implements the momentum strategy for the next
month if the current month’s market volatility is less than the 0.75 percentile of the previous 5 years of market
volatility, otherwise it chooses the reversal strategy. SSTMOM is the momentum strategy scaled by the volatility
ev
of daily market returns over the last 126 days. The strategy SSTMOM(C) is further scaled by a constant that
equates its volatility with the switching strategy on an ex-post basis. The strategy SSTMOM(VC) is scaled by a
constant that equates the volatility of SMOM(C) with the SWITCH strategy for every 5 years.
Strategies Avg SD SR Min Skew Max WI ($) VaR (5%)
STMOM 1.37 8.26 16.53 -38.42 0.134 39.97 896 -11.61
r
STREV 1.41 5.60 25.26 -22.34 0.736 40.18 4,259 -7.26
SWITCH 1.65 7.47 22.06 -25.25 0.723 40.18 9,251 -9.94
STSMOM
STSMOM(C)
STSMOM(VC)
220.72 1231.56 17.92
1.48
1.46
8.26
8.57
17.92
17.09
er-6837.30
-45.87
-42.01
0.060
0.060
0.489
6291.46
42.21
40.17
1,848
782
-1426.54
-9.57
-10.82
pe
ot
tn
rin
ep
Pr
49
This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4342008