How does momentum trading work?

An increasing amount of capital is devoted to mechanical trading strategies designed to exploit market anomalies such as ‘value’ or ‘momentum’. The latter strategy – momentum investing – is one of the oldest stock-selection strategies still in practice. At least since the time of David Ricardo – who advised, “cut short your losses, let your profits run” – the strategy of buying recent winners and selling recent losers has been a favorite of stock market advisors. During the nineteenth and early twentieth century investment manuals and stock guides carried advice columns recommending momentum strategies and included information on stock highs and lows over various horizons for readers who wished to use recent momentum to select stocks. The enduring appeal of momentum as a stock selection strategy is presumably due to its consistent profitability.

One might have expected the profitability of the momentum strategy (momentum premium) to disappear after it became well known and attracted both the attention and capital of professional money managers. Indeed, the momentum premium appeared to vanish around the early 2000s, only to reappear in recent years. What explains the persistence and reoccurrence of momentum profits? How does a strategy as simple as “buy winners and sell losers” not become crowded with followers who bid away any abnormally high risk-adjusted returns?

Historical perspective

We combine new self-collected historical data with existing data to take a long look at the risks and returns to momentum investing in the CRSP-era United States and Victorian-era London. Across both periods a consistently applied momentum strategy generated abnormally high risk-adjusted returns (alpha), but was prone to periodic crashes (see Table 1). Thus, while investors enjoyed high average returns by buying winners and selling losers, the momentum strategy exposed investors to large losses with some regularity. It is tempting to argue that the abnormally high risk-adjusted returns are an illusion. Investors’ aversion to large losses may not be adequately captured by standard asset pricing models and momentum’s measured alpha may simply reflect inadequate risk adjustment, and is actually compensation for exposure to crash risk.  However, these crashes are predictable and the hazards of these crashes vary in the same way across both eras.

The fact that momentum returns show many similarities across two different eras suggests that the underlying drivers have remained the same.  This would be consistent with a popular explanation put forward in the behavioural finance literature that abnormal momentum returns are due to behavioural biases in investors’ decision making.  But there are sophisticated professional investors who are not subject to such biases, and they should be able drive away any persistent abnormal returns.

We argue that the periodic crashes play an important role in sustaining momentum through limits to arbitrage examined in models such as Schleifer and Vishney (1997). The persistent historical pattern across both eras is consistent with a market where sophisticated investors who had the necessary skills to efficiently execute momentum strategies did not have sufficient capital of their own, and had to rely on other people’s money. Indeed, the momentum strategy does require frequent trading and therefore requires skill in executing trades at minimal cost. This separation of brains from capital can be overcome by professional managers endowed with trading skills offering their services to investors with capital in exchange for a percentage of the profits (in the latter part of the CRSP sample) or through leverage financed by risky borrowings at suitable interest rates (in 19th century London).

A number of papers have suggested a link between investor behaviour, leverage, crashes, and average returns. When an anomaly exists but only a limited number of investors have both the capital and skill to exploit it, they may not have enough collective resources to drive the anomalous returns to zero. Typically, skilled investors will seek to expand their profitable positions either through trading leverage (short positions funding long positions) or by soliciting outside capital from less skilled investors. However, in theoretical models such as Gromb and Vayanos (2002), Geanakopolis (2003), Fostel and Geanakopolis (2008), Brunnermeier and Pedersen (2009), and Kondor (2009) leverage constraints can result in sudden reversals if idiosyncratic declines lower the value of collateral and force correlated liquidations. Fears of forced liquidations allow anomalies to persist by limiting the leverage arbitrageurs are willing to devote to eliminating anomalies. Moreover, when arbitrageurs raise equity from return-chasing investors, poor returns can result in contagious liquidations. In models such as Shleifer and Visney (1997), Liu and Longstaff (2004), and Acharya et al. (2010) idiosyncratic volatility combines with the fear of outside investor flight to prevent arbitrageurs from devoting sufficient capital to high alpha investments.

The role of crashes

By itself a separation of brains from capital does not explain why momentum profits did not get eliminated. For a complete understanding we must look to an additional feature of momentum returns – large crashes predictable by sophisticated investors occur at times when momentum is attractive to return-chasing capital.

While we are not the first to notice periodic crashes in momentum returns, to the best of our knowledge, the existing models do not feature the predictable time-varying crash likelihood we observe in the long historical data. One of our contributions is to introduce a hazard model that successfully captures the time-varying likelihood of momentum crashes. We introduce a new methodology based on modeling profit duration dynamics as a function of the risk-free rate, the past stock market return, and the past performance of the momentum strategy, which we interpret as a proxy for the scarcity of capital available to skilled momentum traders. Consistent with models of leverage-induced crashes, we find strong statistical evidence that our proxy for capital available to momentum traders predicts sharp downturns in momentum profits across both the CRSP and Victorian eras.

If crash risk is hardwired into momentum returns, we should expect periodic rare crashes in both the CRSP and Victorian era in the data to occur at similar times. This is what we find; crashes are more likely to occur when momentum has performed well over the recent past (see Table 1). These are periods when capital is easily available to investors employing the momentum strategy with other people’s money. In Victorian London, momentum investors leveraged with margin loans collateralized by the value of their portfolio, and momentum crashes were more likely to occur when traders had more collateral following large momentum strategy gains and periods of easy borrowing as evidenced by low interest rates. In the CRSP era, when many momentum investors access other people’s money through active funds established to invest in momentum, crashes are more likely to occur when the recent performance of momentum has compared well with the overall stock market. These are times when momentum fund managers are likely to find it easy to access the blind capital of return-chasing investors.

Table 1.   Summary statistics and regression results

Predictable crashes imply that momentum investors could increase the Sharp Ratio of their investments by taking account of the time varying crash hazard and exiting momentum whenever the relative performance of momentum becomes too frothy (Figure 1). The fact that fund managers employing algorithmic momentum strategies suffer occasional dramatic losses suggests these managers either cannot anticipate momentum crashes or lack the incentives to take actions to minimize exposure to large losses.

Figure 1. Sharp ratios of CRSP-era managed portfolio

Why don’t skilled portfolio managers exit momentum whenever the trade becomes too ‘crowded’ and crash risk rises? The momentum strategy is most likely to crash when past returns are high – exactly the time when a manager is most able to attract funds and have a high proportion of assets under management above the fee generating high-water mark. The risk-shifting convex payout structure of incentive fees combines with the return-chasing behaviour of future investors to provide incentives for fund managers who have a finite horizon to remain in a crowded momentum strategy.

We introduce a theoretical model to illustrate why rational managers will commit other people’s money to momentum when crashes are more likely – even when they will not commit their own capital. In fact, times when large losses are likely are times when momentum has done better than the market and managers will find it easy to attract the capital of return-chasing investors, i.e. blind capital. This behaviour exposes blind capital to crash losses. One consequence is also that return-chasing capital available to sophisticated investors will be scarce following such crash losses, thereby letting momentum persist.

Policy implications

What are the policy implications, if any? Our findings suggest that regulators who wish to monitor the buildup of crash risk in the financial system should pay particular attention to asset classes and strategies that become crowded by inflows of capital following recent success. These are the times when the interests of even sophisticated money managers may start diverging from the interests of those whose money they manage when it comes to deciding how much risk to take.

References

Acharya V, H Shin, and T Yorulmazer (2010), “A theory of slow-moving capital and contagion”, Working Paper, New York University.

Brunnermeier M and L Pedersen (2009), “Market liquidity and funding liquidity”, Review of Financial Studies 22, 2201-2238.

Fostel A and J Geanakoplos (2008), “Leverage cycles and the anxious economy”, American Economic Review 98, 1211-1244.

Geanakoplos J (2003), “Liquidity, default and crashes: endogenous contracts in general equilibrium”, in Advances in Economics and Econometrics: Theory and Applications II, Econometric Society Monographs: Eighth World Congress, ed. M Dewatripont, L Hansen, S Turnovsky, 170-205. Cambridge, UK: Cambridge University Press.

Gromb D and Vayanos D (2002), “Equilibrium and Welfare in Markets with Financially Constrained Arbitrageurs”, Journal of  Financial Economics 66, 361-407

Kondor P (2009), “Risk in dynamic arbitrage: price effects of convergence trading”, Journal of Finance, 64, 638-658.

Liu J and F Longstaff (2004), “Losing money on arbitrage: optimal dynamic portfolio choice in markets with arbitrage opportunities”, Review of Financial Studies 17, 611-641.

Lunde, A and A Timmerman (2004), “Analysis of bull and bear markets”, Journal of Business and Economic Statistics, 22, 253-273.

Shleifer, A and R W Vishny (1997), “The limits of arbitrage”, Journal of Finance, 52, 35 -56.

This article is published in collaboration with VoxEU. Publication does not imply endorsement of views by the World Economic Forum.

To keep up with the Agenda subscribe to our weekly newsletter.

Author: Benjamin Chabot is a Financial Economist at the Federal Reserve Bank of Chicago. Eric Ghysels is the Bernstein Distinguished Professor of Economics at the University of North Carolina. Dr. Ravi Jagannathan is the Chicago Mercantile Exchange/John F. Sandner Professor of Finance at Northwestern University’s Kellogg School of Management.

Image: A man walks past buildings at the central business district of Singapore. REUTERS/Nicky Loh.