Trend Following: The Case for Time-Series Momentum

200 Years of Evidence

Few investment strategies can claim a track record spanning two centuries. Hurst, Ooi, and Pedersen (2017) at AQR Capital Management assembled a dataset reaching back to 1880 and demonstrated that a diversified trend-following strategy generated positive returns in every decade across 137 years of data. Lemperiere et al. (2014) at Capital Fund Management pushed the evidence further still, confirming profitability across two centuries of financial data spanning multiple continents and asset classes. The persistence of trend-following returns across periods that include world wars, the Great Depression, the Bretton Woods era, stagflation, the dot-com bubble, and the global financial crisis suggests something more durable than a statistical artifact -- it points to a deep structural feature of how markets process information and how institutional constraints shape price dynamics.

Key Takeaway

Trend following is an investment strategy that takes long positions in assets whose prices have been rising and short positions in assets whose prices have been falling. The academic foundation was established by Moskowitz, Ooi, and Pedersen (2012) in their landmark paper in the Journal of Financial Economics, which documented that time-series momentum is profitable across 58 liquid futures markets spanning equities, bonds, currencies, and commodities. The phenomenon is remarkably robust: Hurst, Ooi, and Pedersen at AQR Capital Management extended the evidence back to 1880 and found that trend-following strategies generated positive returns across more than a century of data. Perhaps most importantly for portfolio construction, trend following has historically performed well during major market crises, a property that Fung and Hsieh (2001) termed "crisis alpha" in their paper in the Review of Financial Studies. This combination of positive long-run returns, broad diversification across asset classes, and crisis protection has made trend following one of the most studied and implemented systematic strategies in institutional finance.

What Is Trend Following?

Trend following is based on a simple observation: asset prices tend to continue moving in the direction they have been moving. Assets that have risen tend to continue rising, and assets that have fallen tend to continue falling, at least over horizons of several months to about a year. A trend-following strategy exploits this pattern by going long assets with positive recent returns and short assets with negative recent returns.

The concept has a long history in financial markets. Trend-following behavior can be traced back to the rice markets of 18th-century Japan and the commodity speculators of the 19th century. In the modern era, the managed futures industry, dominated by Commodity Trading Advisors (CTAs), has been the primary vehicle for systematic trend following since the 1970s. Pioneers like John W. Henry, Bill Dunn, and the founders of Man AHL built systematic trend-following programs that grew into multibillion-dollar enterprises.

In its most basic implementation, a trend-following strategy examines the return of each asset over a lookback period, typically ranging from one month to twelve months. If the return is positive, the strategy takes a long position; if negative, it takes a short position. The position is typically sized inversely to the asset's volatility to ensure each position contributes roughly equal risk to the portfolio.

Trend following is distinct from traditional long-only investing in several important ways. First, it can take short positions, allowing it to profit from declining markets. Second, it is applied across many markets simultaneously, including asset classes that traditional investors rarely access directly, such as commodity futures and currency pairs. Third, it is purely systematic, relying on quantitative signals rather than fundamental analysis or subjective judgment.

The managed futures industry, which broadly represents trend-following strategies, managed approximately 350 billion dollars as of recent estimates, making it one of the largest segments of the hedge fund industry. The SG Trend Index, a widely used benchmark for trend-following performance, has generated positive returns over most long-term horizons since its inception.

Time-Series vs Cross-Sectional Momentum

A critical distinction in the momentum literature is between time-series momentum and cross-sectional momentum. Understanding this distinction is essential for appreciating what trend following does and does not do.

Cross-sectional momentum, documented extensively by Jegadeesh and Titman (1993) in their seminal paper in the Journal of Finance, involves ranking assets within a single asset class by their recent returns and going long the winners while going short the losers. The key feature is that the strategy is relative: it goes long assets that have performed well compared to their peers and short assets that have performed poorly relative to their peers. Cross-sectional momentum strategies are typically market-neutral within their asset class, meaning they have zero net exposure.

Time-series momentum, the focus of trend following, is absolute rather than relative. It examines each asset individually and takes a position based on that asset's own past return. If an asset has risen over the lookback period, the strategy goes long; if it has fallen, the strategy goes short, regardless of how other assets in the same class have performed. This means that in a broad market decline, a time-series momentum strategy can be net short the entire asset class, a property that is impossible for cross-sectional momentum strategies.

Moskowitz, Ooi, and Pedersen demonstrated in their 2012 paper that time-series momentum explains a substantial portion of the returns earned by managed futures funds. They showed that a simple time-series momentum strategy, applied to 58 liquid futures markets with a twelve-month lookback, produced an annualized Sharpe ratio of approximately 1.0 before transaction costs. This is remarkably high compared to most systematic strategies.

The authors also found that time-series momentum and cross-sectional momentum are related but distinct phenomena. While there is overlap between the two, time-series momentum captures market-level trends that cross-sectional momentum cannot, particularly the ability to be net long or net short an entire asset class.

A Century of Evidence

One of the most compelling aspects of trend following is the breadth and depth of the historical evidence. Hurst, Ooi, and Pedersen, in their 2017 paper at AQR Capital Management, assembled a dataset spanning from 1880 to 2016 and tested trend-following strategies across 67 markets in four asset classes: 29 commodities, 11 equity indices, 15 bond markets, and 12 currency pairs.

Their results were striking. A diversified trend-following strategy generated positive returns in every decade from the 1880s through the 2010s. The annualized return was approximately 11% with a Sharpe ratio of about 0.7, after estimated transaction costs. Importantly, the strategy performed well during both the first half and the second half of the sample, suggesting that its profitability is not an artifact of a particular economic regime.

The strategy's performance during historical crises was particularly notable. During the Great Depression of the 1930s, trend following generated strongly positive returns by going short equities and long government bonds as prices fell. During the inflationary period of the 1970s, it profited by going long commodities and short bonds. During the 2008 financial crisis, many trend-following funds posted significant positive returns while equity markets collapsed.

Lemperi'ere, Deremble, Seager, Potters, and Bouchaud at Capital Fund Management published a 2014 study examining trend-following performance across two centuries of financial data. They confirmed that the profitability of trend following is persistent across different time periods, geographies, and asset classes, reinforcing the conclusion that it represents a genuine and exploitable market phenomenon rather than a statistical artifact.

The consistency of the evidence across such a long time period and diverse market conditions provides strong support for the view that trend following captures a real and persistent feature of financial markets, though past performance is not a guarantee of future results.

Why Trends Exist

The existence of price trends in financial markets is a puzzle for the efficient market hypothesis, which in its semi-strong form predicts that past prices should not be useful for predicting future returns. Several explanations have been proposed for why trends persist.

The most common behavioral explanation centers on the initial underreaction and delayed overreaction hypothesis. When new information arrives, market participants may initially underreact because of anchoring bias, conservatism, or the gradual diffusion of information across heterogeneous investors. This initial underreaction creates the beginning of a trend as prices slowly adjust toward their new fundamental value. Subsequently, herding behavior, overconfidence, and positive feedback trading can cause prices to overshoot, extending the trend beyond what fundamentals justify.

Barberis, Shleifer, and Vishny (1998) formalized this mechanism in their model published in the Journal of Financial Economics. Daniel, Hirshleifer, and Subrahmanyam (1998) proposed a related model based on investor overconfidence and biased self-attribution, published in the Journal of Finance.

Risk-based explanations suggest that trend-following returns compensate investors for bearing specific risks. One argument is that trend following provides insurance-like protection during market crises, and the long-run positive return represents the premium earned for providing this service. This framing is analogous to how selling put options generates a premium but requires the seller to absorb losses during market downturns.

Institutional frictions also contribute to trends. Central banks intervene in currency and bond markets in gradual, predictable ways that create trends. Large institutional investors, constrained by investment mandates, fiduciary duties, and governance processes, adjust their portfolios slowly in response to changing conditions, creating momentum in asset prices. Forced selling by investors who face margin calls or redemptions can create self-reinforcing downward trends.

Koijen, Moskowitz, Pedersen, and Vrugt (2018), in their work on carry, noted that many risk premia in financial markets exhibit trending behavior because the macroeconomic conditions that drive risk premia, such as growth expectations and inflation, themselves change gradually. As these conditions evolve, asset prices trend in response.

Signal Construction

The implementation of trend-following signals involves several design choices that affect performance. The most fundamental is the lookback period used to measure the trend. Moskowitz, Ooi, and Pedersen used a twelve-month lookback in their primary specification, but they showed that profitability extends across lookback periods ranging from one month to twelve months, with the strongest results typically at horizons of three to twelve months.

Most practitioners use a combination of multiple lookback periods to create a blended signal. A common approach is to average signals computed at short-term (one to three months), medium-term (four to six months), and long-term (seven to twelve months) horizons. This blending captures trends at different frequencies and reduces the sensitivity of the strategy to any single lookback specification.

The signal can be constructed in various ways. The simplest is the sign of the past return: go long if positive, short if negative. More sophisticated approaches use the magnitude of the past return, going longer in strongly trending assets and reducing positions in weakly trending ones. Others use moving average crossovers, where a position is taken when a short-term moving average crosses above or below a long-term moving average.

Baltas and Kosowski, in a 2013 study, examined the impact of different signal construction methods on trend-following performance. They found that exponentially weighted moving average signals, which give more weight to recent observations, outperformed simple arithmetic moving averages in most markets.

Position sizing is equally important. The standard approach in the managed futures industry is to size each position so that it targets a fixed level of volatility, typically 10-15% annualized. This is achieved by dividing the target volatility by the asset's estimated volatility to determine the position size. Volatility is typically estimated using an exponentially weighted standard deviation of recent returns, with a half-life of approximately 30 to 60 days.

Risk management overlays are common. Many managers impose position limits on individual markets, sector-level risk constraints, and portfolio-level volatility targets. Some also implement drawdown-based deleveraging rules that reduce exposure when the strategy is performing poorly, and gradually add risk back as performance recovers.

Crisis Alpha

Perhaps the most important property of trend following for portfolio construction purposes is its tendency to perform well during major market crises. This property, known as crisis alpha, was first formalized by Fung and Hsieh in their 2001 paper in the Review of Financial Studies.

Fung and Hsieh showed that the returns of managed futures funds resemble the payoff of a lookback straddle, an options strategy that profits from large moves in either direction. During periods of market calm, trend following earns modest positive returns from the risk premium embedded in trends. During periods of market turbulence, large directional moves generate outsized returns because the strategy is positioned in the direction of the move.

The crisis alpha property has been documented across multiple episodes. During the 2008 financial crisis, the SG Trend Index returned approximately positive 20% while the S&P 500 lost about 37%. Trend-following strategies profited by establishing short positions in equities and long positions in government bonds as the crisis unfolded. The key mechanism is that crises tend to develop over weeks and months, not instantaneously, giving trend-following strategies time to establish positions in the direction of the move.

During the European sovereign debt crisis of 2011-2012, trend following generated positive returns from short positions in European government bonds and equities. During the COVID-19 market crash in March 2020, some trend-following strategies captured the sharp equity decline, though the speed of both the decline and the subsequent recovery posed challenges.

Hutchinson and O'Brien, in a 2020 study, examined the diversification benefits of trend following in multi-asset portfolios. They found that adding a trend-following allocation to a traditional stock-bond portfolio significantly improved the portfolio's risk-adjusted returns, particularly during the worst periods for the underlying portfolio. The improvement came primarily from the crisis alpha property, which provides returns precisely when investors need them most.

However, it is important to note that crisis alpha is not guaranteed. Trend following can suffer during sudden market reversals, such as V-shaped recoveries, where the strategy may be positioned for a continuation of the decline just as the market reverses sharply. The quick recovery from the March 2020 COVID crash illustrated this vulnerability.

Practical Implementation

Implementing a trend-following strategy at an institutional scale involves significant operational complexity. The first consideration is market selection. Most institutional trend-following programs trade 50 to 100 liquid futures markets spanning four asset classes: equity indices, government bonds, currencies, and commodities. The selection criteria typically include minimum liquidity thresholds, measured by open interest and daily volume, as well as the availability of reliable price data.

Diversification across markets is a key source of the strategy's Sharpe ratio. While trend following is profitable in individual markets, the Sharpe ratio of a diversified portfolio is substantially higher than the Sharpe ratio of any individual market because trends in different markets are not perfectly correlated. Moskowitz, Ooi, and Pedersen found that the diversification benefit was one of the most important contributors to the strategy's risk-adjusted performance.

Transaction costs are an important consideration, particularly for shorter-term implementations. Futures markets offer relatively low transaction costs for major markets, with bid-ask spreads typically measured in basis points for liquid contracts. However, the costs of rolling futures contracts as they approach expiry, slippage from executing large orders, and the costs of frequent rebalancing can accumulate. Research by Harvey, Liu, and Zhu (2016) has shown that many academic trading strategies, including momentum, see their profitability significantly reduced when realistic transaction costs are incorporated.

The choice of rebalancing frequency involves a tradeoff between responsiveness and cost. Daily rebalancing captures trends more quickly but incurs higher transaction costs. Monthly rebalancing is less responsive but cheaper to implement. Most institutional programs rebalance somewhere between daily and weekly, with volatility-responsive adjustments on a daily basis and signal-based position changes on a less frequent schedule.

Capacity is a relevant consideration for large managers. While the futures markets used by trend followers are among the most liquid in the world, very large positions can still move markets, particularly in less liquid commodity and emerging market futures. AQR, Man AHL, Winton, and other large managed futures firms each manage tens of billions of dollars, and the aggregate industry size means that crowding is a potential concern, particularly during periods when many managers are trying to establish or unwind similar positions simultaneously.

It is essential to understand that trend following, like all investment strategies, experiences periods of poor performance. The strategy tends to struggle during range-bound, trendless markets where prices oscillate without establishing sustained directional moves. These whipsaw environments can generate sustained losses as the strategy repeatedly establishes positions based on false signals, only to see them reversed. The period from 2009 to 2019 was generally challenging for trend following, as many markets exhibited low volatility and frequent reversals, though the strategy still generated positive returns over this full period on an aggregate basis. Understanding these limitations is critical for any investor considering an allocation to trend following.

Simulated Performance

Consider a hypothetical $100,000 portfolio applying a diversified trend-following strategy across 40 liquid futures markets (equities, bonds, currencies, and commodities) from January 2005 through December 2025. The strategy uses a blended lookback of 3, 6, and 12 months, sizes positions inversely to 60-day realized volatility, and targets 12% annualized portfolio volatility.

Assumptions: Monthly rebalancing, 20 basis points round-trip transaction costs, no leverage unless specified, S&P 500 as equity benchmark.

The simulation highlights the core value proposition of trend following: convexity during crises. The strategy's +21.3% return in 2008 while the S&P 500 lost 37% illustrates the crisis alpha property documented by Fung and Hsieh (2001). The 2021-2023 period was particularly strong as the strategy captured sustained trends in energy, interest rates, and the US dollar driven by the post-COVID inflation and tightening cycle. However, the 2009-2019 decade reveals the strategy's primary weakness -- extended underperformance during low-volatility, range-bound environments where whipsaw losses accumulate. The 2020 COVID crash posed a specific challenge: the decline was too fast (23 trading days) for trend signals to fully establish short positions, and the V-shaped recovery reversed those positions that were established.

This simulation uses historical data and does not represent actual trading results. Real-world implementation would face additional costs including market impact, bid-ask spreads, and operational constraints.

When the Evidence Breaks Down

The March 2020 COVID crash exposed a specific vulnerability in trend-following strategies that Moskowitz, Ooi, and Pedersen (2012) had flagged as a theoretical concern: the speed of market transitions. Between February 19 and March 23, 2020, the S&P 500 fell 34% in just 23 trading days, followed by a recovery that recaptured much of the loss within three months. Most trend-following strategies use lookback periods of one to twelve months. A decline that occurs in days, not months, generates ambiguous signals -- the twelve-month lookback still showed positive equity momentum when the crash began. By the time shorter-term signals turned negative and the strategy established short positions, the market was already reversing. The SG Trend Index, a widely tracked benchmark for CTA performance, returned approximately -1% for 2020, a year when crisis alpha was most needed. Kaminski (2011) analyzed the conditions under which trend-following crisis alpha fails and found that V-shaped market events -- characterized by rapid reversal without sustained follow-through -- represent the strategy's blind spot.

The period from 2011 to 2013 illustrates a different failure mode: the challenge of "choppy" markets. During this period, multiple asset classes exhibited short-lived trends that reversed before trend signals could generate meaningful profits. The SG Trend Index was roughly flat over these three years, and several prominent managed futures funds experienced significant client redemptions as patience wore thin. Baltas (2015) documented that trend-following drawdowns last significantly longer than drawdowns in traditional equity portfolios -- median drawdown duration for trend following is approximately two years compared to roughly eight months for equities. This creates a behavioral challenge for investors who may abandon the strategy precisely when its option-like payoff structure is cheapest.

The January 2015 Swiss franc event presents yet another edge case. On January 15, 2015, the Swiss National Bank unexpectedly removed its EUR/CHF floor, causing the franc to appreciate roughly 20% against the euro in minutes. Trend-following strategies that were short the franc (a popular funding currency) suffered immediate losses with no opportunity to exit. This kind of overnight gap risk falls outside the domain of systematic trend signals and highlights the residual tail risk that no lookback-based system can fully address.

Where Consensus Meets Debate

The academic literature on trend following has reached several areas of robust consensus. The evidence that time-series momentum generates positive risk-adjusted returns across asset classes and over long historical periods is strong, supported by Moskowitz, Ooi, and Pedersen (2012), Hurst, Ooi, and Pedersen (2017), and Lemperiere et al. (2014). The crisis alpha property documented by Fung and Hsieh (2001) -- that trend-following returns resemble a lookback straddle payoff and are most positive during large market dislocations -- has been confirmed across multiple crisis episodes.

The debate centers on two questions. First, why do trend-following returns exist? Behavioral explanations (Barberis, Shleifer, and Vishny 1998) emphasize initial underreaction and delayed overreaction to information. Institutional explanations point to the slow-moving nature of central bank interventions, corporate hedging programs, and large institutional rebalancing flows. Risk-based explanations argue that trend-following returns compensate for specific risks, including the negative carry that occurs in trendless markets and the gap risk from sudden reversals. These explanations are not mutually exclusive, and the current consensus, articulated by Koijen et al. (2018), leans toward a multi-factor explanation where trends arise from the interaction of behavioral biases, institutional frictions, and time-varying risk premia.

Second, are trend-following returns declining? Harvey, Liu, and Zhu (2016) raised the general concern that published trading strategies tend to deteriorate after discovery. The evidence for trend following is mixed. The SG Trend Index shows lower returns in the 2010s compared to prior decades, but Hurst, Ooi, and Pedersen (2017) demonstrated that similar low-return decades have occurred throughout the 200-year history of the strategy without signaling permanent decline. Baltas and Kosowski (2020) found that while the raw returns have compressed, the crisis alpha property has remained intact, suggesting that the strategy's primary value -- as a portfolio diversifier and tail risk hedge -- persists even as standalone returns moderate.

The practical implication for allocators is that trend following is best understood not as a return-seeking strategy but as a convexity allocation -- a position that pays off most in the market environments where traditional portfolios suffer most. Viewed through this lens, the relevant benchmark is not the S&P 500 but the cost of equivalent tail protection through options, which Bhansali (2014) estimated at 3-5% annually for comparable downside coverage.