Trend following: riding each asset's own trend

Advanced9 min read · 7 sources

Trend following bets that markets that have risen will keep rising for a while and that falling markets will keep falling. Unlike classic momentum, it does not compare stocks with one another but each asset with its own past (time-series momentum). Professional trend followers, known as managed futures funds or CTAs, trade futures on stock indices, bonds, currencies and commodities, both long and short.

How trend following differs from momentum

With cross-sectional momentum, stocks are ranked by their return over the past few months: the winners are bought and the losers sold. Jegadeesh and Titman (1993) documented this effect for US stocks, with holding periods of three to twelve months. Whether the overall market rises or falls plays no role. Only relative strength counts.

With time-series momentum, the position depends only on an asset's own past: if its return over the last twelve months was positive, it is bought; if it was negative, it is sold or avoided. If all markets fall, a trend follower is short in all of them or in cash. This makes the strategy a potential source of protection in long downturns. Strategies based on relative strength between entire markets are explained in the article momentum strategies.

Signal and position size

Scaling by volatility is essential: without it, volatile markets such as commodities would dominate the result, and calm markets such as government bonds would barely count. With it, each market carries roughly the same amount of risk. This is the same idea as in risk parity. Instead of the sign of the 12-month return, many trend followers use moving averages or several time windows at once.

Conversely, trend following loses when a trend reverses abruptly: after a long rise, a market falls sharply while the strategy is still long; after a crash, the market recovers quickly while the strategy is still short. Such turning points, and sideways phases with many false signals, are the typical periods of losses.

The simple version: a moving average

For individual investors without access to futures, a long-only version is common: if the price is above its moving average of the last ten months, you are invested; if it is below, the money goes into the money market. Faber (2007) studied this rule for five asset classes.

This is how Faber describes the result of his allocation since 1972. Importantly, the rule mainly reduces the large declines in long bear markets. In fast crashes with a quick recovery, such as in spring 2020, it often gets out late and back in late.

What research shows

Moskowitz et al. (2012) found time-series momentum in 58 liquid futures on stock indices, currencies, commodities and bonds, in every single one of the instruments studied. Returns persist over one to twelve months and partially reverse over longer horizons, consistent with theories in which investors first underreact and later overreact. A portfolio diversified across all asset classes delivered substantial excess returns, had little relation to the standard factors and performed best in extreme market phases.

Hurst et al. (2017) study trend following over more than a century and across many markets to test whether the results also hold outside the more recent futures data. Asness et al. (2013) show that (cross-sectional) momentum exists in eight very different markets and asset classes and is negatively correlated with value. That is one reason to combine the two.

A warning comes from Daniel and Moskowitz (2016): cross-sectional momentum strategies suffer rare but severe crashes, namely in panic states after market declines, when volatility is high and the market rebounds sharply. Scaling by expected volatility considerably mitigated these crashes in their data. Something similar applies to trend following: the turning points after long trends are the most dangerous moments.

Advanced: the building blocks of a trend-following system

• Signals: the sign of the return over one or more time windows (1, 3, 6, 12 months), the distance from the moving average, the crossing of two moving averages, a breakout from a trading range.
• Position size: scaling to a target volatility per market and for the overall portfolio; correlations between markets are often taken into account so that markets moving in tandem are not counted twice.
• Universe: the more independent markets, the more stable the results. A single market with a trend rule is a risky undertaking; many markets together make a strategy.
• Trading costs and rolling: futures have to be rolled into the next contract before expiration. The return therefore also includes the roll yield and the interest on the posted collateral.
• Linking to volatility: Moreira and Muir (2017) show more generally that portfolios that take less risk when volatility is high improve the Sharpe ratio of many factors.

Trend following as a portfolio addition

Hardly any investor holds trend following on its own. It is interesting as a building block alongside stocks and bonds, because its results have little to do with either and it tends to gain in prolonged crises, when the other building blocks lose. What counts in judging it is therefore less its own return than its contribution to the overall portfolio: does a 10 to 20% allocation reduce volatility and the maximum drawdown without depressing returns much?

This question can only be answered cautiously from the past. A few crises dominate the result, and between them there are often long years of weak returns in which the allocation demands patience. If you use trend following, you should decide in advance how long you will stick with it and reset its share regularly through rebalancing. Otherwise, investors typically sell it right before the next crisis.

Limitations and common mistakes

• Sideways markets: without clear trends, false signals pile up and every reallocation costs money. Dry spells lasting several years are normal.
• Abrupt reversals: after long trends, the strategy loses until the signal flips.
• Overfitting: window length, thresholds and universe invite experimentation. A window that worked best in the past is not a good guide to the future (backtest overfitting).
• Long-only is not the same: without short positions, a trend rule can only retreat into the money market in falling markets. It then protects you but does not profit from the decline.
• Fund costs: managed futures funds often have high ongoing costs and sometimes performance fees, which can eat up a large part of the premium.

How Portfolio Lab calculates

Tactical offers trend following as long-only rules on a monthly basis: “Moving average (trend following)” holds each asset as long as its price is above its own average (10 months by default, adjustable from 2 to 24); otherwise its share goes into cash or a chosen fallback asset. “Dual momentum” combines relative strength with a filter against the money market (12 months by default), and “Relative strength” averages several windows (1, 3, 6 and 12 months). Signals are based only on prices up to the month end and are executed at the closing price of that same month end (or, optionally, one month later); trading costs per order can be set. Short positions and futures are not available.

For the professional version, the Asset classes include the class “Trend following (managed futures)” (@TRENDFOLGE), chained from managed futures funds since 2010 and converted into euros. With it, you can test in the Backtest how an allocation would have changed a stock-bond portfolio in crisis years. Details are in the Tactical methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.