Unveiling the Relevancy of Momentum Strategies
A study on the Swedish Equity Market
Oscar Bodin & Pär Börjeson
A thesis presented for the degree of
Master of Science in Finance
University of Gothenburg
Sweden
Spring 2023
Title Momentum Strategies
Authors Oscar Bodin and Pär Börjeson
Supervisor Ming Zeng
Department Economics
Keywords Momentum, Residual Momentum, Factor Exposure
Abstract
This study investigates the performance of the traditional return momentum strategy
and the residual momentum strategy on the Swedish market over the period 1990 to
2022. The residual momentum strategy show higher risk-adjusted return compared
to the traditional return momentum strategy in equally weighted portfolios, and the
opposite in value-weighted portfolios. A key finding is that the residual momen-
tum strategies experience notably lower volatility overall. In addition, we find the
momentum strategies to be size-dependent and perform significantly better in the
medium-sized companies. In the end, it is still difficult to say whether strategies like
this will generate positive return in real life since momentum investing is plagued by
high turnover, which implies high transaction costs.
1
Acknowledgements
We would like to thank our supervisor Ming Zeng at the University of Gothen-
burg’s Division of Economics, for his assistance and advice throughout this semester.
Additionally, we would like to thank our friends and family for their encouragement
as well as discussions regarding the results and thesis design.
2
Contents
1 Introduction 4
2 Literature Review 9
3 Data and Methodology 11
3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4 Empirical Results 18
4.1 Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Performance difference over the calendar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Sector effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.4 Market size effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Additional Results 26
5.1 Winsorizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.2 Microstructure Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.3 Fama and French five-factor model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6 Conclusion 32
List of Figures
1 Number of eligible stocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Market performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3 Market performance after winsorize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4 FF5 vs FF3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
List of Tables
1 Factor Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Factor persistence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3 Market performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4 Calendar effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5 Sector effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
6 Size effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
7 Market after winsorize . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
8 Microstructure niose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
9 Factor persistence - Fama and French five-factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
10 Market performance - Fama and French five-factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3
1 Introduction
The primary objective of most portfolio managers is to discover a profitable invest-
ment strategy that generates excess return. This involves identifying undervalued
assets and predicting price movements. However, the market efficiency hypothesis
suggests that stock prices already incorporate all relevant information, implying that
it should not be possible to consistently outperform the market with fundamental
and technical analysis (Fama (1970)). Nevertheless, the notion of market efficiency
has faced challenges over the years due to the observation of various anomalies. Je-
gadeesh and Titman (2011) mention that numerous studies present evidence of stock
return predictability using firm-specific variables. Out of these anomalies, the price
momentum effect stands out as the most challenging to explain within the framework
of the conventional risk-based asset pricing model. If stock prices do not always react
efficiently to new information, it creates opportunities for trading strategies that use
past returns to be profitable.
A return momentum strategy is an investment strategy that involves buying stocks
that have shown strong return in the recent past (winners) and selling stocks that
have shown weak return (losers). This strategy is based on the idea that stocks
that have performed well in the past are likely to continue to perform well, while
those that have performed poorly are likely to continue to perform poorly. Blitz
et al. (2011) mention a drawback of the traditional momentum strategy, the so-called
”momentum crashes”. The crashes occur because the momentum strategy exposes
itself to factors that performed well during the formation period. This exposure
positively affects the strategy’s performance if there is persistence in the factor return.
However, when the factor return (which the strategy was exposed to) reverts, there
is a backlash in the strategy’s performance. Blitz et al. (2011) proposed an alternate
approach to momentum investing in their paper ”Residual Momentum”. This residual
momentum strategy aims to neutralize the dynamic exposure to factors and isolate
the performance of individual stocks based on their own merit. This would, in theory,
4
mitigate the potential backlash of a reverting factor. This approach involves ranking
stocks based on the size of the residuals from a rolling regression with factors as
dependent variables. What these factors are and how they are constructed is explained
in the methodology. The residuals are seen as idiosyncratic returns and are used to
rank the stocks from highest to lowest, and similar to the return momentum strategy,
it involves buying the recent winners and selling the recent losers.
Despite the fact that plenty of previous research examine momentum strategies,
e.g., Fama and French (1993); Jegadeesh and Titman (2001); Korajczyk and Sadka
(2004); Blitz et al. (2011) they all have done their studies on the US market and in
comparison, there is just of fraction of that research that explore the Swedish market.
The paper of Rouwenhorst (1998) replicates Jegadeesh and Titman (1993) analysis for
12 European countries between 1978 and 1995 and discover that the profits observe
are similar to those in the United States. The evidence of profitable momentum
strategies suggest inefficiencies in the market. Interestingly enough, Sweden is the
only country where the excess return was found not to be statistically significant.
It would therefore be interesting to examine the relevancy of the return momentum
strategy and see how the residual momentum strategy compare on data from 1990 to
2022 on the Swedish stock market.
In this study, we explore the conventional momentum strategy approach by Je-
gadeesh and Titman (1993) together with the residual momentum strategy using
Fama and French three- and five-factor model. Our findings suggest that investors
are more likely to achieve higher risk-adjusted return by adopting a residual mo-
mentum strategy rather than a traditional return momentum strategy in an equally
weighted setting. Over the period May 1994 to December 2022, the equally weighted
Residual momentum strategy generate an annual return of 8.24%, whereas the equally
weighted Return momentum strategy generate 8.46%. However, since the annualized
volatility is significantly lower in the residual strategy, 19.90% compared to 27.39%,
the Sharpe ratio is higher, 0.41 to 0.31. A possible explanation to this can be the time-
5
varying exposure to the Fama and French factors of the return momentum strategy.
Specifically, during the formation period, the momentum strategy exhibits positive
(negative) loading on systematic factors when these factors experience positive (neg-
ative) returns. Consequently, a return momentum strategy experience losses when
the sign of factor returns during the holding period opposes that of the formation
period. In contrast, the design of residual momentum minimizes time-varying fac-
tor exposures, which can be an explanation to the lower volatility. Considering the
value-weighted setting, we notice the opposite result. Over the same period, the re-
turn strategy generates a significantly higher return than the residual strategy. Even
though the residual strategy experience lower volatility, it does not make up for the
loss in return, which result in a lower risk-adjusted return. Once the value-weighted
strategies are winsorized at the 80th percentile of the market equity, we observe a
significant improvement in both of the strategies’ performance. Furthermore, our
findings suggest that the momentum strategies are not sector-dependent but instead
size-dependent. Specifically, among the small and medium-sized companies, the resid-
ual momentum strategies displayed a higher Sharpe-ratio than the return momentum
strategies. Furthermore, in line with the findings of Blitz et al. (2011), the equally
weighted residual strategy experience a mitigated seasonal trend compared to the
return strategy.
Both Jegadeesh and Titman (1993) and Jegadeesh and Titman (2011) find that
the return momentum strategy produce positive and significant abnormal return dur-
ing their study period. They attribute this to the phenomenon that investors tend
to underreact to new information. Alternatively, the delayed reaction is actually an
overreaction by investors who react with a delay or simply like to invest in past win-
ners. This allow momentum investors to capitalize on the market’s slow adjustment
to new information and potentially earn excess returns. It is important to note that
underreaction is not the sole source of momentum profits, other factors can also con-
tribute to momentum profits. Since the publication of Jegadeesh and Titman (1993)’s
6
paper, the return momentum strategy has become a widely studied investment strat-
egy. Several studies confirm its effectiveness in various markets and time periods
(see e.g., Herberger et al. (2011);Blitz et al. (2011);Korajczyk and Sadka (2004);Gong
(2017)).
For instance, Bondt and Thaler (1985)’s paper shows that stocks that have per-
formed poorly over three to five year periods tend to outperform those that have
performed well, over the subsequent three to five years. Similarly, Jegadeesh (1990)
and Lehmann (1990) find that stocks that have performed poorly over the past week
to one month tend to outperform those that have performed well over the next week
to one month. These studies, which cover long-term and short-term returns, suggest
that contrarian trading strategies can be profitable. They generally support the idea
that stock prices overreact to new information. The study of Jegadeesh and Titman
(1993) contradict these studies. They find that, with formation and holding periods
of three to twelve months, buying stocks that have performed well and selling stocks
that have performed poorly in the past generate significant positive return in the US
market. Furthermore, the studies by Griffin et al. (2003) and Chui et al. (2010) inves-
tigate momentum profits globally and find that the momentum strategy is profitable
in most major markets, although there are some exceptions in Asia (such as Japan).
7
With the use of the return and residual momentum strategies and the given the
existing body of research, we intend to answer the following research question:
Research question:
• How does the residual momentum strategy, based on the Fama and French three-
and five-factor model, compare to the traditional return momentum strategy in
the Swedish Market? Also, are the strategies’ performance dependent on sector,
size, calendar months?
The subsequent content is structured as follows: Section 2 provides an overview
of the existing literature. Section 3 describes our data and the methodologies used.
Section 4 and 5 document the empirical- and additional results, respectively. Lastly,
Section 6 concludes.
8
2 Literature Review
Blitz et al. (2011) compared the performance of a residual momentum strategy to
a return momentum strategy on the US market during the period January 1930 to
December 2009. They find that, using monthly stock returns on the Fama and French
three factor model, the residual momentum strategy generate a risk-adjusted return
(Sharpe ratio) that is approximately twice as large as the one of the return momentum
strategy. Not only do the residual momentum yield higher return than the return
momentum, but also lower volatility. One possible explanation to the lower volatility
in the residual momentum strategy is the significantly lower exposure to the factors
in the Fama and French three-factor model. Blitz et al. (2011) also show that the
residual momentum perform more consistently across diverse economic conditions;
it does not exhibit a systematic bias towards small-cap stocks with higher levels of
firm-specific risk; it is not plagued to the same degree by seasonal trends, such as the
January effect. Grundy and Martin (2001) show that the return momentum strategy
documented by Jegadeesh and Titman (1993) is subject to exposure in the Fama
and French factors. They investigate the drivers of risk and profits of momentum
strategies. Their research reveal that while the momentum strategy exhibit dynamic
exposure to these factors and cross-sectional variability in returns, these alone are
insufficient to explain the average profitability of the strategy. Instead, they find that
overreaction to new information is a key driver behind the observed returns.
Herberger et al. (2011) find evidence of industry-dependent momentum effects
in the Swiss stock market, between December 1979 and February 2009. They also
find that momentum strategies yield significant return, even when accounting for
market-adjustments and transaction costs. This holds for various combinations of
formatting- and holding periods and the momentum strategies perform best in the
high-technology industry.
9
Furthermore, Griffiths and White (1993) provide evidence supporting the tax-loss
selling hypothesis using intra-day data over the Canadian and American stock market.
They find abnormal selling pressure before year-end for stocks with large capital
losses, especially small-cap stocks. As a result of the selling activity, market liquidity
decreases, and large investors capitalize on the drop in prices to buy capital loss stocks
before the year-end and capital gain stocks in January. This creates high returns in
December, followed by a large decline in January. Since residual momentum is more
neutral toward the size factor than the return momentum strategy, this calendar effect
should in theory not be as large for the residual momentum strategy and thus the
strategy should be more consistent over calendar year than the return momentum
strategy.
10
3 Data and Methodology
3.1 Data
We retrieve the unadjusted daily close stock prices for all companies listed on the
Swedish market from Wharton Research Database Services (Compustat). The daily
data is converted to monthly data by only considering the close price of the last trad-
ing day of the month for each stock. The sample covers the period from 1990 up until
the end of 2022. Only primary issues and domestic stocks are considered. In contrast
to Blitz et al. (2011), we include stocks regardless of its price. Blitz et al. (2011) tem-
porarily exclude stocks that fell below 1$ to reduce the effect of microstructure noise.
However, with our sample, we find that we omit a significant part of the data set
when we temporarily exclude stocks that fell below the equivalent of 1$. Therefore,
we keep all stocks in the data set, regardless of its price. The figure 1 below shows
the number of available stocks for each trading strategy during the sample period.
The difference in the number of available stocks is because of the differences in the
ranking procedure. For instance, the residual momentum requires a complete history
of 36 months of returns to be available in that period whereas the return momentum
only requires 12 months of return history.
Figure 1: Number of available stocks during the sample period
11
The monthly factor data is retrieved from ’Global Factor Data’, Jensen et al. (2022).
The considered factors are the RMRF , the SMB and the HML and they are con-
structed in line with Fama and French (1993). RMRF is the excess return of the
market, i.e., the value weighted return of all Swedish firms minus the one-month
Swedish treasury bill. SMB and HML stands for ”Small Minus Big” and ”High Mi-
nus Low”, respectively. SMB is calculated by sorting the sample of stocks by market
equity and divided it into deciles. Thereafter, a long position is taken in the bottom
50% (smallest companies) and a short position in the top 50% (largest companies).
The return of this portfolio is the value of the SMB-factor. Similarly, the HML-
factor is calculated by sorting the sample of stocks by their book-to-market ratio, and
then divided into deciles. A long position is taken in the top 30% (value stocks) and
a short position in the bottom 30% (growth stocks). The return of this portfolio is
the value of the HML-factor. How the sample of stocks is divided is showed by Table
1 below.
Book-Market-Ratio (BE/ME)
Low 30% Mid 40% High 30%
Low 50% Small Growth Small Neutral Small Value
Market Equity
High 50% Big Growth Big Neutral Big Value
Table 1: Mutually exclusive portfolios divided by Market Equity and Book-Market-ratio
Further, the market data return is also retrieved from ’Global Factor Data’, Jensen
et al. (2022). The market data covers the period 1990 up until December 2022. The
market return proxy is calculated by compiling the value weighted raw return of ev-
ery stock listen on the Swedish market. The ”risk-free” rate is the 1-month Swedish
Treasury bill (from ”riksbank.se”).
The sector code for each company, according to the Global Industry Classification
Standard (GICS), is retrieved from Wharton Research Database Services (Compu-
stat). The sector code is a two-digit code that corresponds to the following eleven
12
sectors: Energy, Materials, Industrials, Consumer Discretionary, Consumer Staples,
Health Care, Financials, Information Technology, Communication Services, Utilities
and Real Estate. Furthermore, the market equity for each company in each time
period was also retrieved from Wharton Research Database Services (Compustat).
This is used to calculate the value weighted returns. Lastly, because the data does
not provide the bid-ask spread we are unable to consider transaction costs.
3.2 Methodology
This study follows the common approach in the analysis of the momentum strategies,
the traditional return momentum strategy is in line with Fama and French (1993) and
the residual momentum strategy is in line with Blitz et al. (2011). The methodology
involves creating portfolios based on past returns. Then, the ex-post returns of the
portfolios are used as a dependent variable in a regression where the Fama and French
three factors are used as explanatory variables to identify the underlying risk and the
return characteristics of the portfolios.
This study focus on a formation period of 12 months, a skip-period of 1 month
(from here on denoted 12 - 1M). According to Blitz et al. (2011) this setup is most
broadly used. Also, the only holding period considered in this study is 1 month. The
returns/residuals are recorded during a 12 months period and portfolios are formed,
i.e., formation period. Then a skip-period of 1 month is implemented to avoid the
potential short-term reversal effects. The portfolios formed in the formation period is
held in 1 month. In other words, the stocks are ranked in time t based on the return
in period t − 12 to t − 1, then the investment is implemented in t + 1 and held up
until t+ 2.
13
A traditional return momentum strategy ranks stocks based on their prior cu-
mulative return. More specifically, a return momentum strategy buys stocks that
perform well under the formation period and sells stocks that perform poorly under
the formation period. For each time period t, and each stock i, the raw cumulative
return in the prior 12 months is calculated by dividing the price in the t − 1 time
with the price in t− 12 subtracted by one, as shown in equation (1) below:
Pi,t−1
Ri,t = − 1 (1)
Pi,t−12
The cumulative return in the prior period, Ri,t, is then used in time t to rank the
stocks. The return is then standardized for the stocks in each time period as shown
in equation (2):
Ri,t − R̄t
ReturnMomentumi,t = √ ∑ (2)
1 N
− n=1(R
2
N 1 n,t
− R̄t)
for n=1,..., No. of stocks
for i=1,..., No. of stocks
for t=37,..., No. of months
Based on the standardized prior return, the stocks are ranked from highest to lowest
and divided into deciles. Then, a zero investment strategy is implemented by going
long in the top two deciles (20%) of prior return, which is funded by a short position
in the bottom two deciles (lowest prior return). This procedure is done once every
month, which means that the portfolio is rebalanced every month. Lastly, in order
for the strategy to be valid for that period, there needs to be at least five stocks in
each of the long leg (long position) and short leg (short position). Or else, the return
is temporarily omitted.
The residual momentum strategy is a development of the return momentum strat-
egy. Instead of rank stocks based on prior return, a residual momentum strategy ranks
the stocks based on the magnitude of the residuals from a rolling linear regression.
14
This study uses the three-factor model from Fama and French (1996) to calculate the
residuals. As shown in equation (3):
Ri −Rf = αi + bi(RM −Rf ) + siSMB + hiHML+ εi (3)
Ri−Rf stands for the individual stock’s excess return. RM −Rf stands for the excess
market return, from here on denoted RMRF . SMB and HML are the factor returns
and bi, si and hi are the factor sensitivities, or loadings. The factor sensitivities, bi,
si and hi, are the individual asset’s exposure to each factor. It implies to which
degree the individual asset correlates with the respective factor return. That is, a
high exposure of a stock to a factor in a certain period would imply that the stock’s
return in that period is likely explained by that factor. For instance, if an exposure
of a stock to the SMB-factor is high (in absolute values) for a certain period, then
it would imply it is likely that the return of that stock is explained by its size.
As Blitz et al. (2011), we estimate the regression over a rolling window of 36
months, i.e., from t − 36 for every t to t − 1. 36 months is used to have a sufficient
amount of observations to get accurate estimates of the exposures to the different
factors. With that said, stocks that do not have a complete history of 36 months
return are omitted from that time period. Equation (4) is used for the OLS time
series regression:
         
ε ei,t−36
.. .  
Ri,t−36 RMRFt−36 SMBt−36 HMLt−36 .= .. −  .  αi,t− bi,t  .. − si,t  .. −  .. hi,t ..  (4)
ε ei,t−1 Ri,t−1 RMRFt−1 SMBt−1 HMLt−1
for i=1,..., No. of stocks
for t=37,..., No. of months
15
From the rolling regression in equation (4), there are 36 residuals for each t.
However, only the mean of the residuals from t − 12 to t − 1 is used in time period
t, and this is denoted εi,t. This mean residual for each stock is standardized for each
time period with equation (5) below. The residual is standardized in order to reduce
noise in the estimate, as shown by Gutierrez and Prinsky (2007).
εi,t − ε̄t
ResidualMomentumi,t = √ ∑ (5)
1 N 2
N−1 n=1(εn,t − ε̄t)
for n=1,..., No. of stocks
for i=1,..., No. of stocks
for t=37,..., No. of months
Based on the standardized prior return, the stocks are ranked from highest to lowest
and divided into deciles. A zero investment strategy is then implemented by going
long in the top two deciles (20%) of prior return, the long position is funded by a short
position in the bottom two deciles (lowest prior return). This portfolio is rebalanced
once every month. And again, in order for the strategy to be valid for that period,
there needs to be at least five stocks in each of the long leg and short leg. Or else,
the return is temporarily omitted.
To analyze whether the returns of the strategies are sector-dependent, we divide
the sample into their respective sector according to the Global Industry Classification
Standard (GICS). This division reduce the sizes of the legs significantly. In order to
get a sufficient number of stocks in each leg and each sector we change two things;
the long and short are increased to the top and bottom tercile; the minimum required
number of stocks in each leg is changed to four. If this requirement is not fulfilled
the return is temporarily omitted.
To analyze whether the returns of the strategies are size-dependent, we divide
the sample into terciles of market equity and label the groups ”Large”, ”Medium”
and ”Small” based on its size. Again, to get sufficient number of stocks in each leg
16
and each market size group the long and short leg is the top and bottom tercile,
respectively. Also, the required number of stocks in each leg is four for the strategy
to be valid.
Furthermore, we consider the post-formation returns of the zero investment strate-
gies over the period March 1994 to December 2021. We analyze the performance of
the strategies in terms om return, volatility and Sharpe ratio. Also, we estimate the
respective strategies’ alpha and exposure to the Fama and French factors, similar to
Grundy and Martin (2001)’s approach. This is done with equation (6) where ri,t is
the ex-post return of each strategy:
ri,t = αi + β1,iRMRFt + β2,iSMBt + β3,iHMLt
(6)
+β4,iRMRF UPt + β5,iSMB UPt + β6,iHML UPt + εi
The variables RMRF UPt, SMB UPt and HML UPt are interaction variables that
equal the return of the factors RMRF , SMB and HML when the premium of the
respective factor are positive over the prior period, t− 12 up until t− 2.
Lastly, in the additional result-part, we winsorize the market equity for the com-
panies at the 80th percentile of the market by each time period. This measure is
implemented in order to create more balanced value weighted portfolios to ensure
that not one or few stocks dominate the portfolio.
17
4 Empirical Results
4.1 Main Results
First off, we investigate if there is persistence in the factor returns. It is defined in
the same way as Blitz et al. (2011), i.e., what is the probability that the sign of the
return during the formation period and the holding period is the same. Significant
persistence would suggest that the return in the factors auto-correlate. In other words,
returns can be predicted to some extent. Persistence in the factors can arguably
enhance the performance of the return momentum strategies. Table 2 below shows
the percentage of times the sign of the return during the formation period and the
holding period is the same. The sign of the return during the formation period and
the holding period will be the same 50% of the time if there is no persistence, i.e., it
is random. Therefore, the null hypothesis is that the frequency of each factor equal
50%, and the alternative hypothesis is that it is different from 50%.
Table 2: The table shows the persistence in the Fama and French factors: market (RMRF), size (SMB), and value
(HML). The sample period is January 1990 to December 2022. A format- and skip period of 12 - 1M and a holding
period of 1 month is used. The t-statistic is shown in the parentheses and reports if the persistences are different from
50%.
RMRF TREND SMB TREND HML TREND
48.57% (-0.56) 48.83% (-0.46) 46.75% (-1.28)
Our findings suggest that persistence in factor returns might not exist in our
sample, if anything, the reversal effect seem stronger than the persistence since all
of the factors are less than 50%. Given the lack of persistence in the Fama French
factors it is likely that the factor exposures contribute negatively to the return mo-
mentum strategy’s profitability. However, the magnitude of the contribution to the
performance is still unknown, and what the potential impact of neutralizing these
exposures is also unknown.
18
Furthermore, we analyze the contribution of dynamic factor exposures to the
performance of the return momentum and the residual momentum strategy. We
examine this for an equally weighted and value weighted portfolio for each strategy.
The result for each strategy is presented in Table 3.
Table 3: The table compare the residual momentum and the return momentum strategy in both a equally- and a value weighted setting.
The table show the returns, volatilities, sharpe ratios for each strategy. As well as the alphas, betas and R-squared from the ex-post
regression for each strategy. The sample period is between May 1994 to December 2022. The long and short leg contains the top and
bottom two deciles (20%), respectively, and at least 5 stocks in each leg. All values are annualized and displayed in percentage. The
t-statistic is shown in parentheses and those who are greater or equal to two (≥|2|) are bold.
Return Volatility Sharpe Alpha RMRF SMB HML RMRF UP SMB UP HML UP Adj,RSQ
Residual Momentum Strategy
Equally Weighted 8.24 19.90 0.41 9.51 -0.41 -0.20 0.06 0.30 0.25 -0.11 0.07
(2.56) (-5.02) (-1.65) (0.75) (2.66) (1.39) (-0.99)
Value Weighted 2.82 25.73 0.11 6.36 -0.69 0.20 -0.06 0.47 0.04 -0.53 0.20
(1.43) (-7.10) (1.35) (-0.68) (3.48) (0.19) (-3.94)
Return Momentum Strategy
Equally Weighted 8.46 27.39 0.31 8.68 -0.85 -0.66 -0.11 0.85 0.45 0.08 0.20
(1.83) (-8.28) (-4.24) (-1.06) (5.87) (2.00) (0.59)
Value Weighted 8.51 35.24 0.24 10.21 -1.31 -0.34 -0.76 1.27 0.49 0.05 0.41
(1.95) (-11.47) (-1.99) (-6.74) (7.92) (1.96) (0.31)
The result is in line with the findings of Blitz et al. (2011), and the equally weighted
residual momentum strategy generates a higher risk adjusted return than the return
momentum strategy. For the value weighted strategies, it is the other way around. It
is also shown that return momentum strategies exhibit higher dynamic exposure to the
factors than the residual momentum. The return momentum is statistically significant
exposed to the market (RMRF ) and size (SMB), while the residual momentum is
only exposed to the market (RMRF ). Both strategies are negatively exposed to
the factors after negative return and less negatively exposed when the returns are
positive. For instance, when the market return is negative during the formatting
period, the market beta for the equally weighted return momentum strategy is -0.85,
and 0.00 (-0.85 + 0.85) after positive returns. While for equally weighted residual
momentum, the market beta is -0.41 after positive return and -0.11 (-0.41 + 0.30)
after negative return. The strategies exhibit greater exposure to the market when
19
the portfolio formation is based on value rather than equal weights. Specifically, the
value weighted return momentum strategy has a market beta of -1.31 and -0.04 (-1.31
+ 1.27) compared to the residual momentum strategy, which has a market beta of
-0.69 and -0.22 (-0.69 + 0.47). The adjusted R-squared demonstrates that 20% and
41% of the total return momentum’s variance can be explained by the factors, and
7% and 20% for the residual momentum. These results indicate that the factors have
a higher explanatory power of the return in the return momentum strategies than
in the residual momentum strategy. This is expected since the residual momentum
strategy aims to minimize the exposure to the factors.
We can conclude that a ranking stocks based on residuals is decreasing the ex-
posure to the factors significantly, in our case by two to three times, which is one
way to reduce volatility. Even though the equally weighted strategies yield similar
returns, the residual momentum strategies experience lower volatility. The residual
momentum strategy has a annualized volatility of 19.90% compared to 27.39% for
the return momentum. Hence, the lower volatility results in higher Sharpe ratio for
the residual momentum (0.41) compared to the return momentum (0.31).
Furthermore, one way of examine the factor exposure’s contribution to the strate-
gies’ profitability is to neutralize the factors. Our analysis shows that the factor
exposure contribute negatively to all strategies. For the equally weighted return mo-
mentum portfolio, the alpha is 0.22 percentage points (pp) higher than the raw return,
while the alpha for the respective residual momentum portfolio is 1.27 pp higher than
the raw return. In the case of the value-weighted portfolios, alphas are notably higher
than the raw returns. The residual momentum strategy has an alpha 3.54 pp higher
and the return momentum 1.70 pp higher than the raw return. The negative factor
contribution can possibly be explained by the lack of persistence in the factors.
As shown in Figure 2, the return momentum strategies are more volatile than
the residual momentum strategies. Although it yields slightly higher return than the
residual momentum, it is exposed to these ”momentum crashes” when the factors
20
Figure 2: The figures compare the cumulative return of the residual momentum and the return momentum strategy
in both a equally- and a value weighted setting. The subplot above displays the performance of the equally weighted
strategies and the one below the value weighted. The period is May 1994 to December 2022. The long and short leg
contains the top and bottom two deciles (20%), respectively, and there are at least 5 stocks in each leg. The Y-axis
displays the return, where 1 = 100%.
revert, and thus perform worse over time. The equally weighted strategies are not as
affected by these backlashes as the value weighted strategies, hence, the performance
are better over time. Furthermore, it looks like all of the strategies seem to be affected
by the dot-com bubble around 2002, especially the return momentum strategies. The
value-weighted strategies never really recover from that steep decline.
One of our key findings, in an equally weighted setting, is that ranking stocks
based on their residual return, rather than total return, lower the volatility without
penalizing the return. This results in a 32% higher Sharpe ratio. This is in line with
Blitz et al. (2011), however, the Sharpe ratio do not increase in the same extent. In
21
contrast to Blitz et al. (2011), we find no persistence in the factors; lower explanatory
power of the factors in the return momentum strategy; the factors have negative
impact on the strategies’ return. All these differences can be the reason for why the
Sharpe ratio do not increase in the same extent as Blitz et al. (2011).
4.2 Performance difference over the calendar
To examine potential seasonal patterns, we look at how both strategies performed on
average for each calendar month.
Table 4: The table compare the seasonal effect of the residual momentum and the return momentum strategy in both a equally- and a
value weighted setting. The mean return of each strategy is displayed for each month over the sample period January 1990 to December
2022. The long and short leg contains the top and bottom two deciles (20%), respectively, and there are at least 5 stocks in each leg.
The significance of the return is shown by the t-statistic in parentheses and those who are greater or equal to two (≥|2|) are bold. All
values are annualized and displayed in percent.
Jan Feb Mar Apr Maj Jun Jul Aug Sep Okt Nov Dec
Residual Momentum Strategy
Equally Weighted -1.28 0.69 -0.29 0.70 -0.84 1.37 3.33 2.12 0.93 1.41 -0.55 0.56
(-1.09) (1.06) (-0.41) (0.79) (-1.17) (2.51) (1.66) (2.43) (0.89) (1.38) (-0.50) (0.46)
Value Weighted -0.02 -0.30 -1.58 -1.58 -0.23 0.91 2.02 2.39 2.50 -0.46 -1.77 0.78
(-0.01) (-0.27) (-1.29) (-1.06) (-0.21) (1.11) (1.70) (2.32) (1.58) (-0.18) (-1.63) (0.87)
Return Momentum Strategy
Equally Weighted -7.23 3.08 0.68 1.07 1.55 0.50 2.71 1.32 1.33 1.81 -1.64 3.10
(-3.71) (2.41) (0.61) (0.74) (2.18) (0.48) (1.84) (1.32) (0.94) (1.08) (-0.83) (2.35)
Value Weighted -0.14 4.03 -0.46 -3.18 0.64 3.88 0.86 0.25 0.95 -0.78 -0.29 2.66
(-0.09) (1.98) (-0.30) (-1.45) (0.58) (3.32) (0.63) (0.19) (0.43) (-0.20) (-0.19) (2.22)
According to previous research conducted by Griffiths andWhite (1993), Moskowitz
and Grinblatt (1999) and Jegadeesh and Titman (2001), momentum returns exhibit
a seasonal pattern, with a strong negative performance for total return momentum in
January due to the tax-loss selling effect. This effect occurs because fund managers
tend to sell poorly performing stocks, particularly small-cap stocks, in December,
resulting in a negative impact on returns in January. However, Blitz et al. (2011)
mention that residual momentum is less concentrated in small-cap stocks, therefore,
22
we expect the January effect to have a smaller impact on its performance. To high-
light this previous findings, we displays the average returns in each calendar month
over the period between January 1990 to December 2022.
Our results, presented in Table 4, confirm the strong performance of return mo-
mentum strategies in December with significant positive return of 3.10% and 2.66%
for the equally- and value-weighted respectively. This effect is mitigated in the case
of the residual momentum strategies, with no significant returns. The positive return
in December is followed by negative return in January for all strategies, however, it
is only significantly negative for the equally weighted return momentum. Whether it
is the tax-loss selling effect that Blitz et al. (2011) observe or another seasonal trend,
we do not know. Furthermore, we observe that the performance vary more for the
return momentum strategy for each month, while the residual momentum strategy is
more consistent. This suggest that the residual momentum strategy is less sensitive
to calendar effects.
4.3 Sector effect
The sample has been divided into sectors according to the Global Industry Classifi-
cation Standard (GICS) to see if the return of the strategies are sector-dependent.
Table 5 displays the return, volatilities, and Sharpe ratios for the four strategies in
each sector. The sector analysis begin in February 2001 because of the limited sample
size. It is by that time all of the sectors fulfill the requirement of minimum number
of stocks in each leg. It is noteworthy that the dot-com bubble is relatively close in
time to the beginning of our sector analysis, which can be related to the fact that the
”Information Technology” and ”Communication Services” sectors report relatively
low returns.
Considering the equally weighted strategies, in all the sectors where residual mo-
mentum perform well and report a relatively high Sharpe ratio - the return strategy do
as well. Except in Consumer Staples, Health Care and Financials, where the residual
23
Table 5: The table compare the sector-dependency of the residual momentum and the return momentum strategy in both a equally-
and a value weighted setting. The returns, volatilities and Sharpe ratios for the four strategies in each sector is displayed. The sample
period is between February 2001 to December 2022. The long and short leg contains the top and bottom terciles, respectively, and there
are at least 4 stocks in each leg. The significance of the return is shown by the t-statistic in parentheses. All values are annualized and
displayed in percent.
Strategies
Equally weighted Value Weighted
Residual Momentum Return Momentum Residual Momentum Return Momentum
Sectors Return Vola Sharpe Return Vola Sharpe Return Vola Sharpe Return Vola Sharpe
Energy 61.57 347.83 0.18 74.04 349.98 0.21 -7.03 75.07 -0.09 23.70 74.84 0.32
Materials 12.76 40.00 0.32 11.22 37.67 0.30 11.00 32.22 0.34 21.57 43.09 0.50
Industrials 6.18 18.15 0.34 7.54 19.50 0.39 -2.04 17.08 -0.12 -2.75 20.25 -0.14
Concumer Discretionary 9.44 22.69 0.42 23.81 24.75 0.96 5.56 27.21 0.20 6.44 32.36 0.20
Consumer Staples 8.98 30.42 0.30 3.43 32.81 0.10 6.66 29.78 0.22 -2.03 29.84 -0.07
Health Care 6.42 40.24 0.16 -1.98 32.93 -0.06 8.63 34.36 0.25 0.22 33.09 0.01
Financials 10.92 25.70 0.43 8.10 29.82 0.27 1.50 22.28 0.07 7.97 27.37 0.29
Information Technology -4.57 35.62 -0.13 -3.01 37.16 -0.08 -0.30 34.69 -0.01 10.11 41.26 0.24
Communication Services -2.96 49.29 -0.06 -4.24 58.37 -0.07 -8.27 55.05 -0.15 18.37 56.07 0.33
Utilities 17.83 61.22 0.29 30.68 59.14 0.52 15.13 59.83 0.25 34.35 59.68 0.58
Real Estate 1.35 23.82 0.06 8.79 26.79 0.33 -3.44 20.07 -0.17 -2.13 22.33 -0.10
strategy perform better than the return strategy. Other than that, the Sharpe ratio is
relatively similar between the two strategies. The volatility is however smaller in the
residual momentum compared to the return momentum in the majority of the sec-
tors, but this seem to have come to a price of smaller returns as well. Considering the
value-weighted strategies, the return momentum have higher Sharpe ratio than the
residual momentum in all sectors except Consumer Staples and Health Care. Once
again, the volatility in the residual momentum strategy is smaller than the return
momentum strategy in the majority of the sectors, and the smaller volatility seem to
have come to a price of lower return.
A common denominator between the two strategies is that residual momentum
in both an equally- and value weighted setting, performed better than the return
momentum in the Consumer Staples and Health Care sector in terms of Sharpe ratio.
However, unlike the conclusion drawn by Herberger et al. (2011), we cannot draw any
conclusion from this and the chance that this result is random cannot be dismissed.
24
4.4 Market size effect
Finally, we analyze whether the strategies are size-dependent. To do this, we divided
the sample into terciles based on the market equity of each stock and denote the groups
”Small”, ”Medium”, and ”Large”. In Table 6 we display the returns, volatilities and
Sharpe ratio for the four strategies in different size categories.
Table 6: The table compare the size-dependency of the residual momentum and the return momentum strategy in both a equally- and
a value weighted setting. The returns, volatilities and Sharpe ratios for the four strategies in each size group is displayed. The sample
period is between October 1992 to December 2022. The long and short leg contains the top and bottom terciles, respectively, and there
are at least 4 stocks in each leg. All values are annualized and displayed in percent.
Size
Small Medium Large
Weights Equally Value Equally Value Equally Value
Strategy Residual Return Residual Return Residual Return Residual Return Residual Return Residual Return
Return 4.66 -11.30 8.71 4.82 10.73 4.29 8.59 5.24 1.14 2.62 1.30 2.96
Volatility 27.42 33.43 18.03 24.08 23.91 21.61 16.54 20.00 11.15 18.42 18.93 23.13
Sharpe 0.17 -0.34 0.48 0.20 0.45 0.20 0.52 0.26 0.10 0.14 0.07 0.13
Considering the Small group, all strategies generate positive return, except for the
equally weighted return momentum that yield significant negative return. The take-
away from the small group is that the volatility is above average and that the reversal
effect seem strong in small stocks. The medium sized group is the one performing
the best among all groups overall, with the highest Sharpe ratio for all portfolios.
In this group, the residual momentum strategy reports twice as large Sharpe ratios
compared to the return momentum strategy. Lastly, considering the large group, it
has considerably lower overall return than the other size groups. This could either
mean that the large companies have low auto-correlation in its returns or that large
companies overall has lower average return. It is worth noting that by ”capping”
the value weighted strategies within the small and medium group, they both perform
better than the equally weighted in the respective group. Given this result, the
momentum strategies seem to be size-dependent, where the performance is the best
among the medium-sized companies.
25
5 Additional Results
5.1 Winsorizing
In this final section, we conduct a market analysis similar to the main results, by com-
paring the performance between residual momentum and return momentum. How-
ever, we address the potential problem with value-weighting, which is that poor per-
formance of the strategies may come from unbalanced portfolios. Value-weighting
assigns weights to companies based on their market equity, a relatively large com-
pany will therefore have a substantial impact on the portfolio’s performance. In
this section we examine whether introducing ”value-cap” will affect our result, e.i.,
make sure that not any one company dominate the portfolio. The market equity for
each company is capped at the 80th percentile of the market at each time period.
This will reduce the assigned weight of relatively large companies and increase the
assigned weight of smaller companies, resulting in a more balanced portfolio in the
value-weighted strategies. The new result for each strategy is shown in the Table 7.
Note that the equally weighted results are the same as Table 3.
Table 7: The table compare the residual momentum and the return momentum strategy in both a equally- and a value weighted setting.
The market equity for each company is capped at the 80th percentile of the market at each time period. The table show the returns,
volatilities, sharpe ratios for each strategy. As well as the alphas, betas and R-squared from the ex-post regression for each strategy.
The sample period is between May 1994 to December 2022. The long and short leg contains the top and bottom two deciles (20%),
respectively, and at least 5 stocks in each leg. All values are annualized and displayed in percentage. The t-statistic is shown in
parentheses and those who are greater or equal to two (≥|2|) are bold.
Return Volatility Sharpe Alpha RMRF SMB HML RMRF UP SMB UP HML UP Adj.RSQ
Residual Momentum Strategy
Equally Weighted 8.24 19.90 0.41 9.51 -0.41 -0.20 0.06 0.30 0.25 -0.11 0.07
(2.56) (-5.02) (-1.65) (0.75) (2.66) (1.39) (-0.99)
Value Weighted 6.33 16.84 0.38 8.26 -0.40 -0.09 0.03 0.26 0.25 -0.20 0.12
(2.71) (-6.10) (-0.86) (0.50) (2.76) (1.74) (-2.10)
Return Momentum Strategy
Equally Weighted 8.46 27.39 0.31 8.68 -0.85 -0.66 -0.11 0.85 0.45 0.08 0.20
(1.83) (-8.28) (-4.24) (-1.06) (5.87) (2.00) (0.59)
Value Weighted 13.36 27.24 0.49 14.45 -1.18 -0.36 -0.37 1.08 0.31 0.17 0.38
(3.48) (-13.12) (-2.65) (-4.17) (8.52) (1.58) (1.38)
26
Both the value-weighted strategies improved significantly from the winsorizing.
The residual strategy improved its return to 6.28% (2.45%) and lowered its volatility
to 16.84% (25.73%). The return strategy improved its return to 13.36% (6.64%) and
volatility to 27.24% (35.24%). Furthermore, we notice a reduction in the exposures
to the factors for the residual momentum strategy, with a lower R-squared of 0.12
(0.20). However, the factors still negatively contribute to the return since the alpha is
1.93 pp higher than the raw returns, but to a lower extent than before. Also, for the
return momentum, the factors contribute more to the strategy’s return than before.
The alpha is 1.09 pp higher than the raw returns compared to before where the alpha
was 1.70 pp higher.
Figure 3: The figure compare the cumulative return of the residual momentum and the return momentum strategy in
both a equally- and a value weighted setting. The market equity for each company is capped at the 80th percentile
of the market at each time period. The period is May 1994 to December 2022. The long and short leg contains the
top and bottom two deciles (20%), respectively, and there are at least 5 stocks in each leg. The Y-axis displays the
return, where 1 = 100%.
27
As seen in Figure 7, the value-weighted portfolios do not experience the same drop
in return as the non-winsorized strategies do. By winsorizing the weights, the value-
weighted portfolios got more balanced and less dependent on a single or a few stocks.
The strategies experience an increased return and reduced volatility.
5.2 Microstructure Noise
Blitz et al. (2011) temporarily exclude stocks that fell below 1$ to reduce the effect
of microstructure noise. However, if we were to exclude the equivalent of 1$ we omit
a large part of the data. To achieve the same effect and to remove illiquid stocks, we
instead temporarily omit the smallest 5% of stocks based on market equity.
The result after removing the lowest percentiles of market equity is shown in Table
8 below.
Table 8: The table compare the residual momentum and the return momentum strategy in both a equally- and a value weighted
setting. The lowest 5% of the stocks in terms of market equity are removed to reduce microstructure noise. The table show the returns,
volatilities, sharpe ratios for each strategy. As well as the alphas, betas and R-squared from the ex-post regression for each strategy.
The sample period is between May 1994 to December 2022. The long and short leg contains the top and bottom two deciles (20%),
respectively, and at least 5 stocks in each leg. All values are annualized and displayed in percentage. The t-statistic is shown in
parentheses and those who are greater or equal to two (≥|2|) are bold.
Return Volatility Sharpe Alpha RMRF SMB HML RMRF UP SMB UP HML UP Adj.RSQ
Residual Momentum Strategy
Equally Weighted 8.24 19.90 0.41 9.51 -0.41 -0.20 0.06 0.30 0.25 -0.11 0.07
(2.56) (-5.02) (-1.65) (0.75) (2.66) (1.39) (-0.99)
Value Weighted 2.82 25.74 0.11 6.36 -0.69 0.20 -0.06 0.47 0.04 -0.53 0.20
(1.43) (-7.10) (1.35) (-0.68) (3.48) (0.19) (-3.93)
Return Momentum Strategy
Equally Weighted 8.46 27.39 0.31 8.68 -0.85 -0.66 -0.11 0.85 0.45 0.08 0.20
(1.83) (-8.28) (-4.24) (-1.06) (5.87) (2.00) (0.59)
Value Weighted 8.60 35.27 0.24 10.30 -1.31 -0.34 -0.76 1.27 0.49 0.05 0.41
(1.96) (-11.46) (-1.97) (-6.73) (7.91) (1.95) (0.31)
As we can see when comparing Table 8 to Table 3, the strategies practically yield
the same result. It appears that the alleged microstructure noise from the smallest
stocks do not have any impact on the performance of the strategies during our sample
period. Another explanation is that the smallest companies in our sample tend to
not be included in the portfolios.
28
5.3 Fama and French five-factor model
In this final section we extend the analysis with the Fama and French five-factor
model. This is done as a robustness check to investigate if the additional factors help
to explain the individual stock return, or if the three-factor model is sufficient. The
five-factor model looks like the following:
Ri −Rf = αi + biRMRF + siSMB + hiHML+ riRMW + ciCMA+ εi (7)
The additional factors are the RMW (Robust Minus Weak) and CMA (Conser-
vative Minus Aggressive). The former is the profitability-factor and represents the
average return of robust operating profitability stocks minus the average return of
weak operating profitability stocks. The latter is the investment-factor, which is the
average return of conservative investment stocks minus the average return of aggres-
sive investment stocks. Blitz et al. (2018) show in their study that the explanatory
power of the cross-sectional variance is improved by adding two more variables.
We start to investigate whether there is persistence in the new factors or not.
The result is shown in Table 9 below. The sign of the return during the formation
Table 9: The table shows the persistence in the Fama and French factors: market (RMRF), size (SMB), value (HML),
profitability (RMW) and investment (CMA). The sample period is January 1990 to December 2022. A format- and
skip period of 12 - 1M and a holding period of 1 month is used. The t-statistic is shown in the parentheses and reports
if the persistences are different from 50%.
RMRF TREND SMB TREND HML TREND RMW TREND CMA TREND
48.57% (-0.56) 48.83% (-0.46) 46.75% (-1.28) 46.38% (-1.40) 51.21% (0.47)
period and the holding period is the same 46.38% and 51.21% of the times for the
additional factors respectively. The reversal effect is stronger in the RMW -factor, like
the other factors in the three-factor model. CMA is the only factor that shows signs
of slight persistence. However, the result lacks statistical significance and therefore
cannot any persistence nor reversal effect of the factors be inferred. Furthermore, we
examine how the contribution of dynamic factor exposure changes when using the
29
five-factor model rather than the three-factor model. The results for the Fama and
French five-factor model are shown in Table 10 below.
Table 10: The table compare the residual momentum and the return momentum strategy in both a equally- and a value weighted setting.
This time the residual momentum strategy is developed using Fama and French 5-factor. The table show the returns, volatilities, sharpe
ratios for each strategy. As well as the alphas, betas and R-squared from the ex-post regression for each strategy. The sample period is
between May 1994 to December 2022. The long and short leg contains the top and bottom two deciles (20%), respectively, and at least
5 stocks in each leg. All values are annualized and displayed in percentage. The t-statistic is shown in parentheses and those who are
greater or equal to two (≥|2|) are bold.
Return Volatility Sharpe Alpha RMRF SMB HML RMW CMA RMRF UP SMB UP HML UP RMW UP CMA UP Adj.RSQ
Residual Momentum Strategy
Equally Weighted 7.66 19.40 0.39 8.66 -0.30 -0.23 0.05 -0.07 0.09 0.19 0.22 -0.05 0.16 0.06 0.04
(2.34) (-3.22) (-1.80) (0.49) (-0.32) (0.50) (1.51) (1.23) (-0.43) (0.69) (0.27)
Value Weighted 3.02 21.90 0.14 5.54 -0.31 -0.07 -0.24 -0.12 0.13 0.07 0.02 0.04 0.11 0.16 0.05
(1.33) (-2.92) (-0.46) (-2.15) (-0.46) (0.66) (0.49) (0.09) (0.29) (0.41) (0.63)
Return Momentum Strategy
Equally Weighted 8.46 27.39 0.31 7.09 -0.68 -0.57 0.10 0.25 -0.31 0.64 0.40 -0.01 0.27 0.69 0.22
(1.50) (-5.69) (-3.51) (0.80) (0.88) (-1.41) (4.04) (1.77) (-0.07) (0.91) (2.32)
Value Weighted 8.51 35.24 0.24 6.02 -0.81 -0.03 -0.32 0.74 -0.34 0.70 0.37 -0.15 0.98 1.40 0.52
(1.27) (-6.79) (-0.18) (-2.57) (2.53) (-1.53) (4.43) (1.64) (-1.03) (3.23) (4.70)
The residual momentum strategy shows similar result compared with the three-
factor model. The strategies is only statistically significant to the market (RMRF ).
The equally weighted strategy yields slightly worse risk adjusted return, while it is
the opposite for the value weighted. For the return momentum strategy, the equally
weighted strategy is significant to market (RMRF ), size (SMB) and investments
(CMA) while the value weighted is significantly exposed to all factors, except size
(SMB).
Furthermore, we observe that the factor exposures contribute positively to the
performance of the return momentum strategy. This is evident from the lower alpha
compared to the raw return, which is not the case from the ex-post regression of the
three-factor model. However, the findings remain consistent with the previous results
for the residual momentum strategies, where the alpha continues to surpass the raw
returns. Considering the R-squared for the return momentum strategy, the result
align with the findings of Blitz et al. (2018). Specifically, the return momentum
exhibits an increase in the adjusted R-squared. The equally-weighted and value-
weighted strategies increased to 0.22 (0.20) and 0.52 (0.38), respectively. This suggests
that part of the return in the momentum strategy can be attributed and explained
30
by the additional factors.
In Figure 4, we compare the cumulative performance of the residual momentum
strategy based on the three- and the five-factor model, in an equally weighted and
value-weighted setting. The differences between the two strategies are not substantial
Figure 4: The figure compare the cumulative return of the residual momentum based on the five- and the three-factor
model. This is shown in both an equally- and value-weighted setting. The period is May 1994 to December 2022.
The long and short leg contains the top and bottom two deciles (20%), respectively, and there are at least 5 stocks in
each leg. The Y-axis displays the return, where 1 = 100%.
in absolute values. However, in relative terms, the additional factors significantly
increase the performance in the value-weighted setting.
31
6 Conclusion
Our analysis of momentum strategies in the Swedish stock market reveals that the
strategies do generate positive return between 1990 and 2022. Considering the equally
weighted strategies, the residual momentum strategy experience a higher risk-adjusted
return than the traditional return momentum strategy. For the value weighted strate-
gies, it is the opposite. Once the market equity for each stock is winsorized at the
80th percentile, the value-weighted portfolios perform significantly better. This indi-
cate that the strategies suffer from unbalanced portfolios. Furthermore, we observe
that the residual strategy experience a significantly mitigated seasonal trend effect,
especially less affected by the January effect; momentum strategies are not sector-
dependent but instead size-dependent; the strategies do not suffer from microstructure
noise; the residual momentum strategy using the Fama and French five-factor rather
than the three-factor model yield similar result. One important detail is that we do
not consider any transaction costs nor quoted spreads throughout the study. If we
were to consider it, it would likely result in a significant drop in the return since mo-
mentum strategies in general suffer from high turnover (see e.g., Korajczyk and Sadka
(2004)). Another important point that speaks for why the momentum strategies are
inefficient is the fact that the R-squares in the ex-post regressions for the return mo-
mentum strategies are relatively low (see e.g., Blitz et al. (2011)). The reason can
be related to the fact that we observe no significant persistence in the Fama and
French factors in our sample. With that said, even though the strategies managed to
generate significant positive return it is still difficult to say whether these strategies
will generate positive return in real life since momentum investing is plagued by high
turnover, which implies high transaction costs.
32
References
Blitz, D., Hanauer, M., Vidojevic, M., and Van Vliet, P. (2018). Five concerns with
the five-factor model. The Journal of Portfolio Management, 44(4):71–78.
Blitz, D., Huij, J., and Martens, M. (2011). Residual momentum. Journal of Empirical
Finance, 18(3):506–521.
Bondt, W. F. M. D. and Thaler, R. (1985). Does the stock market overreact? The
Journal of Finance, 40(3):793–805.
Chui, A. C., Titman, S., and Wei, K. J. (2010). Individualism and momentum around
the world. The Journal of Finance, 65(1):361–392.
Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work.
The Journal of Finance, 25(2):383–417.
Fama, E. F. and French, K. R. (1993). Common risk factors in the returns on stocks
and bonds. Journal of Financial Economics, 33(1):3–56.
Fama, E. F. and French, K. R. (1996). Multifactor explanations of asset pricing
anomalies. The Journal of Finance, 51(1):55–84.
Gong, Y. (2017). Does the momentum strategy work at the industry level? evidence
from the chinese stock market. Emerging Markets Finance and Trade, 53(5):1072–
1092.
Griffin, J. M., Ji, X., and Martin, J. S. (2003). Momentum investing and business
cycle risk: Evidence from pole to pole. The Journal of Finance, 58(6):2515–2547.
Griffiths, M. D. and White, R. W. (1993). Tax-induced trading and the turn-of-the-
year anomaly: An intraday study. The Journal of Finance, 48(2):575–598.
33
Grundy, B. D. and Martin, J. S. (2001). Understanding the nature of the risks and
the source of the rewards to momentum investing. The Review of Financial Studies,
14(1):29–78.
Gutierrez, R. C. and Prinsky, C. A. (2007). Momentum, reversal, and the trading
behaviors of institutions. Journal of Financial Markets, 10(1):48–75.
Herberger, T., Kohlert, D., and Oehler, A. (2011). Momentum and industry-
dependence: An analysis of the swiss stock market. Journal of Asset Management,
11:391–400.
Jegadeesh, N. (1990). Evidence of predictable behavior of security returns. The
Journal of Finance, 45(3):881–898.
Jegadeesh, N. and Titman, S. (1993). Returns to buying winners and selling losers:
Implications for stock market efficiency. The Journal of Finance, 48(1):65–91.
Jegadeesh, N. and Titman, S. (2001). Profitability of momentum strategies: An
evaluation of alternative explanations. The Journal of Finance, 56(2):699–720.
Jegadeesh, N. and Titman, S. (2011). Momentum. Annual Review of Financial
Economics, 3(1):493–509.
Jensen, T. I., Kelly, B. T., and Pedersen, L. H. (2022). Is there a replication crisis in
finance? Journal of Finance, Forthcoming. https://jkpfactors.com/.
Korajczyk, R. A. and Sadka, R. (2004). Are momentum profits robust to trading
costs? The Journal of Finance, 59(3):1039–1082.
Lehmann, B. N. (1990). Fads, martingales, and market efficiency. The Quarterly
Journal of Economics, 105(1):1–28.
Moskowitz, T. J. and Grinblatt, M. (1999). Do industries explain momentum? The
Journal of Finance, 54(4):1249–1290.
34
Rouwenhorst, K. G. (1998). International momentum strategies. The Journal of
Finance, 53(1):267–284.
35