Momentum Crashes

Structures and characteristics that underline the periods of poor momentum performance

Bachelor Thesis, 2013

38 Pages, Grade: 1,7


Table of Contents

List of Figures

List of Tables

List of Abbreviations

1. Introduction

2. The Momentum Investment Strategy – a General Overview
2.1 Mode of Operation
2.2 Sources of Momentum
2.3 Momentum in Equities and Other Asset Categories
2.4 Correlation with the Fama French Factors

3. Data, Construction and Application of the Momentum Portfolio
3.1 Data Origin and Portfolio Construction
3.2 Application of the Portfolio

4. Momentum Crashes
4.1 The Major Crash Periods
4.2 Crash Triggers
4.2.1 Skewness and Kurtosis
4.2.2 Bear Markets and Portfolio Betas
4.2.3 Non-Linearity of Market Returns in and after Bear Markets

5. The 2009 Momentum Crash – a Detailed Analysis
5.1 Qualitative Analysis: Companies in the Winner and Loser Portfolio
5.2 Quantitative Analysis: Insights into risk, size, trading volume, and value factors

6. Conclusion



List of Figures

Figure 1: The Portfolio Ranking, Formation, and Holding Process

Figure 2: Cumulative Monthly Returns 1947 –

Figure 3: Zero Investment Strategy 1927 –

Figure 4: The 1930s Momentum Crash

Figure 5: The 2001 Momentum Crash

Figure 6: The 2009 Momentum Crash

Figure7: Frequency Distribution of WML Returns over the 1927 to 2012 Period

Figure 8: Regression of the W and L Portfolio in the After Bear Market Period of the 1930s

Figure 9: Regression of the W and L Portfolio Bull Market Period of 1950 to

Figure I: OLS Regression of the Fama French Factors SMB & HML and Momentum Returns WML

Figure II: Cumulative Monthly Returns 1947 – 2007

Figure III: Zero Investment Strategy 1927 – 2012

Figure IV: The 1930s Momentum Crash

Figure V: The 2001 Momentum Crash

Figure VI: The 2009 Momentum Crash

Figure VII: Separate Regression Analyses for the W Portfolio in the 1930s Period

Figure VIII: Regression of the W and L Portfolio in the In and After Bear Market Period 2007-2009

List of Tables

Table I: Results of the 1996 Benchmark Period

Table II: Results of the 2009 Crash Period

Table III: Relevant Ratios for Performance Evaluation

List of Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

1. Introduction

The momentum strategy is a simple yet powerful trading strategy. Momentum implies that past stock prices can predict future stock price development. According to momentum theory, past winner stocks are likely to continue their good performance while past loser stocks are likely to continue to perform poorly. Hence, applying this strategy, investors buy stocks that have risen in the past the strongest and (short) sell those that have declined in value the most. This very simple decision rule is practically the only important guideline to follow regarding the momentum strategy. Surprisingly and in spite of its simplicity, momentum works and yields high excess returns. Over the 1927 to 2012 period, the portfolio of past winner stocks yields an annualized excess return of 7.157% compared to the market portfolio. Even though momentum usually performs exceptionally well, it does not offer free lunch. In the 1927 to 2012 time frame, there are a few periods of extreme momentum underperformance that could have wiped out some significant wealth. For instance, during the most recent 2009 momentum crash, this strategy would have erased 104.28% of an initial investment in just 3 months.1

This paper focuses on the structures and characteristics that underlie the periods of extremely poor momentum performance and sets a special focus on the latest 2009 momentum crash period. It answers questions regarding the momentum portfolio composition during this period and quantitatively evaluates the momentum portfolio, measuring commonly applied performance indicators. The results are then contrasted with a non-crash benchmark period.

The remainder of this paper is structured as follows: Section 2 will give in depth insights into the momentum strategy regarding its practical application. In addition, Section 2 shows that momentum is not restricted to equities and that momentum returns cannot be explained by the Fama and French (1993) factors. Section 3 is dedicated to the replication of the momentum strategy over the 1927 to 2012 period; Section 4 sets a closer focus on the major momentum crash periods. In this section, the crash periods are first replicated and secondly, common crash triggers identified and investigated. In Section 5, the detailed analysis of the 2009 momentum crash is conducted on a qualitative and quantitative basis. Section 6 concludes.

2. The Momentum Investment Strategy – a General Overview

2.1 Mode of Operation

The momentum strategy is most successful considering recent, short to medium-term time frames of historical stock prices. Investors usually observe the stock price development of the past 6 to 12 months to determine which stocks to buy and which to sell. This period is called the ranking period of the momentum portfolio. The stocks are then held for only a few months, typically for 1 to 6 months as the momentum effect tends to reverse in the long run. This is shown in Cooper, Guiterrez, and Hameed (2004). After each holding period the stocks are re-evaluated and again, according to their past performance, new portfolios of winner and loser stocks are formed. This is a dynamic process that requires a steady revaluation of the momentum portfolio. Momentum is pervasive and enduring over time and can be observed empirically. Jegadeesh and Titman (1993) first quantifies this investment strategy stating that it yields an average 1% monthly excess return compared to the market in the 1965 to 1989 period. In their analysis Jegadeesh and Titman have considered the US stock market and observed this phenomenon utilizing a 6-months ranking period and a 6-months holding period. Due to the strategy’s generally strong performance it is employed by numerous mutual fund and quantitative traders. However, momentum depends heavily on positive past market development and typically yields excess returns only in bull market states. In fact Cooper, Guiterrez, and Hamed (2004) finds that the premium is low in states of high market volatility and that the average momentum return falls to -0.37% in case of lagged 3-year bear markets. During bull markets however, momentum yields such a high excess return that it remains a profitable investment strategy even after accounting for transaction costs. This is an important factor as transaction costs are relatively high due to the frequent portfolio reconstruction and the generally short holding periods.

2.2 Sources of Momentum

The underlying force that drives momentum is still not clearly identified and further research in this field needs to be conducted. Efficient market theory has failed to explain returns associated with momentum, however, there are a few theories derived from behavioral finance that try to explain the mechanisms behind momentum. One of the most recognized theories has been developed by Daniel, Hirshleifer, and Subrahmanyam (1998). According to their theory, investor’s overconfidence drives momentum. They argue that investors are overconfident with regard to information they acquired privately and then overreact accordingly. In a second step the self attribution bias of investors becomes the driving force behind momentum. If subsequently further information becomes public, these investors will react asymmetrically to confirming information in contrast to disconfirming information. This means that in case of good news, their overconfidence in a certain stock is further reinforced which leads to strong momentum returns. Daniel, Hirshleifer, and Subrahmanyam therefore conclude that momentum should be especially strong if the market overall is strong. In case of a generally bullish market environment, momentum then becomes a self-fulfilling prophecy: If the market rises, momentum investors go long for an increasing number of stocks which results in increasing stock prices which in turn triggers the overconfidence of the investors. A similar approach has been stated in Jegadeesh and Titman (1993) who argues that momentum investors themselves are the source of abnormal momentum returns as they buy past winners and sell past losers and thereby generally cause prices to overreact. Hong and Stein (1998) has developed another behavioral theory that tries to explain momentum returns. In this approach two types of investors exist, namely the newswatchers and the momentum traders. The newswatchers solely rely on contemporaneous information whereas the momentum traders only observe past stock prices to determine their choice of investment. Since a further assumption of this theory states that information diffuses only gradually, stock prices react only gradually to this new information. However, in case of good news, the gradual positive development of the stocks attracts momentum traders who again overreact which leads to a further stock price increase. These theories assume that these overreaction inefficiencies should be reverted in the long run as the market corrects the mispricing. This price correction is observed in Cooper, Guiterrez, and Hameed (2004) which might indicate that these behavioral theories are able to explain momentum returns to some degree. Grinblatt and Han (2005) finally offers a differing approach. It identifies the disposition effect as a possible cause for momentum returns. According to this theory, risk-averse investors are likely to sell past winners too quickly and hold on to past losers for too long as they are averse to recognizing losses.

2.3 Momentum in Equities and Other Asset Categories

The power of momentum is not limited to US stocks. In fact momentum has been proven to exist also outside US equity markets. Rouwenhorst (1998) finds momentum in equity markets of other developed nations. In this study Rouwenhorst uses an internationally diversified portfolio of 12 European countries and is able to show similar results regarding momentum returns for Europe like Jegadeesh and Titman (1993) shows for the US market. Furthermore Rouwenhorst (1999) finds that momentum in emerging markets is qualitatively similar to momentum in developed markets. Moreover, momentum is not limited to equities either but can be found in many other asset categories. Erb and Harvey (2006) and Miffre Rallis (2007) show that momentum is found in commodity markets. Miffre and Rallis further show that commodity momentum returns have low correlations with equity and bond market momentum returns.2 Okunev and White (2003) finds momentum in currencies and shows that a momentum strategy incorporating moving averages can yield significant momentum returns in this asset class. Finally, Asness, Moskowitz, and Pedersen (2008) finds momentum in bonds; Moskowitz, Obi, and Pedersen (2012) in futures.

2.4 Correlation with the Fama French Factors

The correlation of monthly momentum returns with the Fama French Factors small (market capitalization) minus big (SMB) and high (book to market ratio) minus low (HML) over the 1927 to 2012 period equals -0.2229 and -0.3705, respectively. An OLS regression over the same period yields beta values of -0.0931 (R² = 0.0497) for the SMB regression and -0.1663 (R² = 0.1373) for the HML regression.3 Thus, there is (especially in the case of the value factor) some persistent negative correlation between momentum returns and the Fama French Factors. However, Carhart (1997) shows that the Fama French 3-factor model’s errors are highly positive for the past year momentum winners and highly negative for the past year momentum losers. Thus, the 3-factor model systematically fails to explain momentum returns and a 4-factor model extended by a momentum factor is needed such as proposed by Carhart (1997).

3. Data, Construction and Application of the Momentum Portfolio

3.1 Data Origin and Portfolio Construction

In this section, the paper aims to replicate the momentum crashes observed in Daniel and Moskowitz (2012) (DM). For this purpose, the momentum portfolio’s returns found in Kenneth French Data Library (2013) were employed for replicating the value- 2A commodity momentum portfolio is therefore a good ingredient for a diversified momentum portfolio. 3See Appendix Figure Figure I for details weighted momentum period returns.4 These portfolios are very similar to those constructed in DM. The data comprises monthly stock data from January 1927 to December 2012 for all US firms listed on NYSE, AMEX, and NASDAQ. Figure 1 displays the ranking, formation and holding periods of the momentum portfolio graphically. The cumulative returns of all relevant stocks are calculated over an 11- month ranking period. Based on their respective 11-month cumulative returns, the stocks are assigned to one of the 10 decile portfolios. Each decile comprises the same number of stocks. Portfolio 10 contains the firms that performed the best in the past 11- month ranking period; this portfolio is called the Winner (W) portfolio. Portfolio 1 lists the stocks which yielded the lowest ranking period return. This portfolio is denoted the Loser (L) portfolio. After a 1-month gap the stocks in the W portfolio are bought at the formation date of the portfolio. The return of the 1-month holding period is the value- weighted return of the firms in the portfolio. In general, the momentum strategy can be expanded by short selling the L portfolio. Furthermore this strategy allows for a self- financing zero investment strategy, where the W portfolio is bought and the L portfolio is short sold at an equal amount. This portfolio is called the Winner minus Loser (WML) Portfolio. Note, that the 1-month gap between the ranking period and holding period is implemented to minimize the bid-ask bounce and to avoid exposure to market microstructure events. The usefulness of this approach is documented in Jegadeesh (1989) and is also applied by DM. When constructing the portfolios, the following data requirements must be met: The stocks must have valid closing prices at the formation date of the portfolio. Also, only returns of common shares are considered. Stock data is adjusted for dividend payments and stock splits; dividend pay outs are reinvested in the respective stock. The decile breakpoints are set that there are an equal number of NYSE stocks in each of the 10 portfolios. All of these requirements are consistent with the data derivation and portfolio construction of DM. Finally, for the analysis found in Section 4.2 and Section 5 of this paper, the Wharton Research Data Services was conducted in terms of The Center for Research in Security Prices (CRSP) and Compustat databases.

Abbildung in dieser Leseprobe nicht enthalten

Figure1: The Portfolio Ranking, Formation, and Holding Process

3.2 Application of the Portfolio

Abbildung in dieser Leseprobe nicht enthalten

Figure 2: Cumulative Monthly Returns 1947 - 2007

Figure 2 plots the cumulative monthly returns of alternative investment strategies over the 1947 to 2007 time frame; the scale is in log10-format. The blue line represents the cumulative return of the W portfolio, the red line the cumulative return of the L portfolio. The green and the violet line represent the cumulative investments in 1-month US Treasury bills (risk-free investment) and the CRSP value-weighted index, respectively. All investments are long positions with an initial investment outlay of $1 per investment alternative.

It becomes eminent that the momentum premium over this 50-year time-frame is strong and enduring and highly outperforms the investment in the market portfolio. The compound annual growth rates of the portfolios are: 23.8% for the W portfolio, 14.12% for the market portfolio, 5.65% for the risk-free investment, and 0.60% for the L portfolio. The annualized excess returns equal 15.35% per year for the W portfolio, 7.50% per year for the market portfolio, and -1.33% per year for the L portfolio. These numbers indicate that the momentum strategy worked exceptionally well over this 50- year period.

Abbildung in dieser Leseprobe nicht enthalten

Figure 3: Zero Investment Strategy 1927 - 2012

Figure 3 plots an alternative self-financing zero investment strategy already introduced in section 3.1. Utilizing this strategy, a potential investor goes long the W portfolio and, at the same time, goes short the L portfolio. Repeatedly, the investment is conducted each month by buying and short selling the W and the L portfolio at the amount of $1. The excess returns (losses) generated are the differences between the gains (losses) in the long and short investment. The returns are not reinvested but accumulate as the strategy is profitable or are reduced as losses are incurred. Thus, this investment is a self-financing strategy and is called the WML strategy. As Figure 3 displays, this strategy yields an excess return of $12.29 over the 85-year period. As the WML strategy seems to offer a free lunch opportunity in the long-run, there are certain periods during which this WML strategy performed poorly and incurred heavy losses. For instance, on June 30th 1932, this strategy had accumulated excess profits of $2.46. By July 31st 1933 however, the WML’s excess profits were down to $0.46 which equals a net loss of 2 monthly investment inputs.5 Only by November 30th 1951 the losses incurred during the 1930s crash were recovered and the excess profits were back on the level of June 1932. A more aggressive, high input zero investment at this point of time would have been a costly endeavor. Other important periods of poor performance applying the WML strategy are found from December 2000 to January 2001 where 41.97% of the monthly investment input was erased and from March 2009 to March 2011 which erased 101.89% of the monthly investment input. These crash periods are now examined in more detail.

4. Momentum Crashes

4.1 The Major Crash Periods

Figures 4-6 display the 3 time frames that have witnessed the worst momentum crashes in the 1927 to 2012 period. The cumulative returns of each investment alternative are displayed together with the corresponding zero-investment WML strategy. The methodology remains the same as introduced in Section 3 of this paper: Figures 4-6 plot the cumulative returns of a $1-long investment in each of the 4 possible alternatives; the W portfolio, the L portfolio, an investment in 1-month US T-bills (the risk-free investment), and the investment in the CRSP value-weighted index (the market investment). The WML investment again is carried out by a $1 long-short investment.

In Figure 4 the momentum performance during and after the Great Depression is examined. The L portfolio continuously outperforms the W portfolio with minor exemptions from July 1932 to March 1943. However, after the USA entered World War 2, the stock market recovers and the W portfolio regains its strength and permanently outperforms the L portfolio from November 1943 on.6 As it is displayed in the corresponding WML plot, the WML strategy suffers severe losses especially during the months of July and August 1932 as well as April and May 1933. For instance, July and August 1932 combined wipe out 139.07% of the monthly investment input, the losses of the months April and May 1933 amount to 70.72%.

Abbildung in dieser Leseprobe nicht enthalten

Figure 4: The 1930s Momentum Crash

Figure 5 displays the 2001 momentum crash which can be attributed to the burst of the dotcom bubble. Even though this crash seems a lot more moderate than the other two crash periods it must be noted that January 2001 wipes out 41.97% of the monthly investment input which is also a remarkable loss. However, the WML plot indicates,that the losses incurred during January 2001, are almost fully recovered by June 2001, ifthe WML strategy is continued to be carried out. In contrast to the other major crashes,this is a comparatively fast recovery of losses incurred during the crash period.


1 Requires that short side of the investment is applied also.

2 A commodity momentum portfolio is therefore a good ingredient for a diversified momentum portfolio

3 See Appendix Figure Figure I for details

4 For robustness purposes, equal-weighted momentum portfolios were constructed and can be found in Figures II through VI of the Appendix.

5 A net loss of 2 monthly investment inputs signifies, that the amount lost during this crash equals the amount of investment committed on both sides, the long and the short side of the monthly investment (here $1). Hence, a net loss of 2 monthly investment inputs equals in this case a loss of $2.

6 The results regarding this specific period differ quite significantly from the results found in DM. The overall pattern of L and W returns cannot be shown over the entire period as results deviate strongly for the early 1940s. However, this paper also shows that the WML strategy would have suffered severe losses if it was applied. See Daniel and Moskowitz (2012) for details.

Excerpt out of 38 pages


Momentum Crashes
Structures and characteristics that underline the periods of poor momentum performance
University of Mannheim
Catalog Number
ISBN (eBook)
ISBN (Book)
Momentum Investig, Momentum Crashes, Momentum Strategy, Factor Investing, Momentum Factor, Momentum Factor Crash
Quote paper
Heinrich Stilling (Author), 2013, Momentum Crashes, Munich, GRIN Verlag,


  • No comments yet.
Read the ebook
Title: Momentum Crashes

Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free