Managing Momentum and Negative Skew; a Dual Approach

July 31, 2014

Robert A. Gillam, CFA Chief Investment Officer McKinley Capital Management, LLC
Gregory S. Samorajski, CFA Director of Investments McKinley Capital Management, LLC

Abstract

The McKinley Capital Management, LLC (“McKinley Capital”) investment process relies on momentum and earnings acceleration factors. Underreaction to news is asserted as the rationale for the efficacy of these factors, especially at the single stock level. Data is presented to show the positive and significant long-term results of a naïve long-short momentum strategy in the United States. The usefulness of the momentum factor is tempered by negatively skewed returns which occasionally lead to large losses. Negative skew is demonstrated to be a market-wide overreaction phenomenon which is most acute during major market inflection or turning points. McKinley Capital analyzes two methods which can be used to limit the negative skew of momentum based returns and significantly improve performance. The first method, referred to as dynamic momentum, seeks to limit exposure to the momentum factor following significant down market moves thereby decreasing the likelihood of momentum exposure during market inflections from bear to bull markets. The second method, referred to as residual momentum, seeks to adjust the momentum indicator to focus on stock specific rather than market-wide momentum. McKinley Capital demonstrates how both techniques, when combined, improve the effectiveness of the momentum factor. The results are extended to EAFE markets and global markets.

McKinley Capital is a global growth equity specialist. The firm uses a quantitative investment process with a qualitative overlay.
This process drives all of its mandates: global, emerging markets, non-U.S., nonU.S. developed, non-U.S. developed 130/30, non-U.S. small cap, U.S. large cap, U.S. small cap, global dividend growth, and U.S. equity-income. The quantitative model is used to create portfolios that are systematically exposed to momentum and earnings acceleration (referred to by McKinley Capital as “growth”). The quantitative risk control models are used to limit exposure to other risk factors and the transaction cost analysis (TCA) model is used to mitigate the cost of executing the strategy.

Momentum and earnings acceleration factors work because the market, investors, and the analysts who cover stocks systematically underreact to news. Therefore, McKinley Capital holds that its factors perform due to behavioral rather than risk-based reasons. For example, one component of the earnings acceleration factor is earnings forecasts at the consensus level. Among other characteristics, the firm buys stocks which have consensus upward analyst revisions. The idea is that when good news is uncovered, the average analyst increases his or her earnings estimates. However, evidence indicates that such upward revisions systematically underestimate the news. Analysts tend to be “anchored” to their previous forecasts and slow to fully incorporate new news. Therefore, as the news becomes better understood, upward revisions tend to be followed by further upward revisions and positive earnings surprise. Both of these lead, on average, to excess returns.

Momentum is explained by the same phenomenon. As news is uncovered, the market reacts in the right direction, but systematically underestimates the magnitude of the information. Again, investors like analysts, tend to be “anchored” to their previous opinions and slow to incorporate new and relevant information. Therefore, stocks with positive news based momentum tend to continue to rise as analysts increase their earnings estimates and as the market begins to price the true impact of the good news. The risk to momentum investing is overreaction, sometimes referred to as the “herd mentality”. As good news becomes better received by the market, more and more participants jump on the bandwagon. This “herd” effect can cause the market to overreact. When played out, the stock in question can be subject to a sharp downward spike. Thus, a pure momentum strategy often shows negative skew characteristics. Skew is a statistical term that measures the asymmetry of the probability distribution of a real-valued random variable about its mean. The task of a momentum manager is to separate stocks which have underreacted to news from those which have overreacted: buying the former and avoiding the latter. At the stock level, McKinley Capital seeks to control this risk by using valuation criteria. When a stock becomes significantly overvalued, compared to similar stocks with a similar level of growth and which fully meet McKinley Capital’s buy discipline, the stock is sold.

Still, there remains a market-wide systematic risk of overreaction. This tends to occur at major market inflection points. Other techniques are required to address this market-wide risk. Conceptually, consider the situation near a market top. It is likely that high beta stocks and some high volatility stocks will show up as best in naïve momentum screens. This occurs even without stock specific news. Similarly, low beta portfolios will dominate a naïve momentum strategy near market bottoms. In either case, when the market direction reverses, the relative return of a naïve momentum portfolio can be negative and significant. In statistical terms, it has negative skew. As will be demonstrated, the situation near market bottoms is the most troublesome for a momentum manager.

Chart 1 shows the growth of $1 invested in a U.S. based naïve high momentum (winners) portfolio or alternatively, in a low momentum (losers) portfolio. The data covers 1927 to 2013. A high momentum portfolio is defined as the top decile of U.S. stocks when ranked by their past 12-month performance (skipping the most recent month). A low momentum portfolio is defined as the bottom decile of U.S. stocks when ranked by their past 12-month performance (skipping the most recent month). Returns are measured monthly, with deciles computed each month. As is clear from the chart, either buying high momentum stocks, shorting low momentum stocks or doing both would have been wildly successful strategies. However, as expected, the naïve strategy returns are subject to occasional large losses; statistically, the returns are negatively skewed.

1407 Australia-01 Chart1

Chart 2 illustrates the point. It shows large negative returns to a long/short naïve momentum strategy in 1932, 1939, 1974, 2001, and most recently in 2009 (major market bottoms occurred in each of these years). While these events are rare, when they occur they are highly troublesome. Even though average monthly returns are significantly positive (+1.21%), the shorter-term performance of an investor that entered the market just prior to a major down move would be very poor at best. Are investors strong enough to wait for the benefit of the longer-term? While the long-term trend is clearly positive, there are certainly periods of sustained underperformance as well.

1407 Australia-02 Chart2

Chart 3 sheds some further light on the problem. The beta of a long/short naïve momentum portfolio was very negative around the same years: 1932, 1939, 1974, 2001, and 2009. Near the end of a major market decline, the high relative momentum stocks were large, defensive, low beta equities. As markets recovered, these stocks significantly underperformed high beta, high risk stocks causing large short-term relative losses. The chart also shows that long/short naïve momentum betas never seemed to reach extreme positives. For that reason, losses to a naïve momentum strategy have never been as large around bull to bear inflection points as they have been around bear to bull inflection points.

1407 Australia-03 Chart3

Table 1 reflects the fourteen worst monthly returns for a naïve long/short momentum strategy in the U.S. Leading up to the poor momentum performance months, in 13 of the 14 prior two year periods the overall market declined significantly. This data indicates that poor performance for naïve momentum strategies is the most extreme following a sustained market decline, as the risk of a market reversal to the strategy becomes more likely.

1407 Australia-04 Tab1

This effect is not limited to only U.S portfolios. Chart 4 shows the returns of a naïvelong/short global momentum portfolio from 2004 to 2014. While on average, returns are positive there is an large negative return in the beginning of 2009.

1407 Australia-05 Chart4

Due to the behavioral effect of underreaction, momentum strategies seem to work well when applied to individual stock selection. Good stock selection translates into a successful portfolio strategy. Two valuable tools for limiting large losses due to market level overreaction around major turning points – particularly bear market to bull market are dynamic momentum and residual momentum. Dynamic momentum seeks to reduce overall exposure to momentum in periods of significant risk to the factor. Residual momentum seeks to isolate stock specific momentum and limit market wide effects.

Dynamic momentum is a brute force procedure which is used to avoid the risk of underperformance at major market inflection points, particularly upturns, for a momentum strategy. Consider this simple rule for a long/short momentum strategy: if over the past six months the market is down by -15% or more, the Market State is defined as adverse; otherwise, the Market State is defined as normal. In a normal state, buy the top decile of momentum stocks and short the bottom decile. In an adverse state, invest the entire portfolio in cash (alternatively, an investor could invest the entire portfolio in an index). This strategy will, almost by definition, avoid momentum investing during a V-shaped recovery since such a recovery is only possible following a major market decline (an adverse state). The cost, however, is being out of the momentum factor during the final phases of a bear market (often a good relative return period for the factor). Chart 5 and Table 2 show the result of this dynamic strategy for the 1927 to 2013 U.S. equity market. The large negative drawdowns of the naïve strategy are mitigated – especially in the early 1930’s Depression Era and the most recent 2009 Global Financial Crisis. The simulated long/short annualized return improves from 9.8% to 14.9% with a significant decline in volatility. Apparently, the benefit to a long/short momentum strategy of avoiding the major V-shaped turning points outweighed the cost of some missed opportunity. While directly applicable through McKinley Capital’s dynamic momentum long/short strategy, this technique could also be used in a long-only portfolio at the request of a client. During periods of high risk to momentum, model weights could be shifted to emphasize earnings acceleration with less weight on momentum, or even shifted to a lower tracking error portfolio, or use other tools to mitigate the coming inflection point. Derivatives use would be an excellent tool to use to aid in managing this process. Among the various methods used to manage dynamic momentum, Market State is a simple and effective way to accomplish that goal.

1407 Australia-06 Chart5

Residual momentum seeks to take the market effect out of momentum calculations at the stock specific level. At a minimum, momentum is calculated on a market beta adjusted basis. Such a calculation has the effect of limiting extreme portfolio betas into the definition of momentum at market turning points, and allows a focus on company specific underreaction to news. Other systematic betas could similarly be considered, (i.e. sectors, industries, market cap, etc.). Charts 6 through 9 illustrate the impact of McKinley Capital’s proprietary approach to residual momentum in the U.S. and MSCI EAFE markets. In both cases, the extreme down periods for naïve momentum – especially in 2009 – were mitigated.

1407 Australia-07 Chart6

1407 Australia-08 Chart8

While this approach to mitigating beta exposures to a momentum portfolio at major market inflection points is both novel and powerful, there are several other working papers and research approaches that address this problem and they should be acknowledged. There is an interesting method recently developed by Geczy and Samonov (2013)1 that not only examines the consistent in- and out-of-sample power of the phenomenon of momentum over the longest backtest ever  examined in academia, 212 years, but it also includes a manner in which to manage this aforementioned problem of beta exposure. Geczy and Samonov’s method relies on the average number of months in which the beta of a momentum portfolio differs from the market direction and hedges away that risk, whereas McKinley Capital’s method extracts the beta of the market from each stock within Momentum calculation individually. McKinley Capital expects that its method is more precise, but further investigation is warranted to determine which of the methods actually provides more benefit or if they could somehow be used together.

Chart 10 shows the simulated impact of combining dynamic momentum and residual momentum in a single strategy across eight developed countries. This combined approach is employed in McKinley Capital’s stand-alone dynamic momentum strategy and in a portion of its global dividend stripping strategy: dividend growth. In each case the Sharpe Ratio of a naïve long/short momentum strategy is improved by adding dynamic and residual momentum.

1407 Australia-09 Chart10

Disclosure

McKinley Capital Management, LLC (“McKinley Capital”) is a registered investment adviser under the U.S. Investment Advisers Act or 1940. All information contained herein is believed to be acquired from reliable sources but accuracy cannot be guaranteed. This presentation is for informational purposes only, and was prepared for academic, financially sophisticated, and institutional investors. This material may contain confidential and/or proprietary information and may only be relied upon for this report.

Charts, graphs and other visual presentations and text information were derived from internal, proprietary, and/or service vendor technology sources and/or may have been extracted from other firm data bases. As a result, the tabulation of certain reports may not precisely match other published data. Data may have originated from various sources including but not limited to Bloomberg, MSCI/Barra, Russell Indices, APT, ClariFI, Zephyr, and/or other systems and programs. With regard to any materials accredited to MSCI/Barra: Neither MSCI nor any other party involved in or related to compiling, computing or creating the data have any liability for any direct, indirect, special, punitive, consequential or any other damages (including lost profits) even if notified of the possibility of such damages. No further distribution or dissemination of the MSCI data is permitted without MSCI’s express written consent. Please refer to the specific service provider’s web site for complete details on all indices. McKinley Capital makes no representation or endorsement concerning the accuracy or propriety of information received from any other third party.

 

1Greczy, Christopher C. and Samonov, Mikhail. “212 Years of Price Momentum (The World’s Longest Backtest: 1801-2012). A University of Pennsylvania Wharton School of Business working paper. August, 1, 2013.

Author’s note: I must admit to a small bias and conflict to my opinions of this paper. Dr. Christopher C. Geczy is the Academic Director of the Jacobs Levy Equity Center for Quantitative Financial Research at The Wharton School, University of Pennsylvania. I am both a graduate of Wharton (W’94) as well as a member of the Advisory Board of that Center