Trend-following strategies work great – until they don’t. This is called a trend break. Managers often focus their time on deciding what static momentum speed to choose. Choosing a slow speed (which means a longer lookback) produces smooth signals but runs the risk of missing out of turning points – or trend breaks. Choosing a fast speed (which means a shorter lookback) produces a choppy signal that is more responsive to current information – however, what might be a turning could also just be noise. Rather than choosing a static speed, why not consider a dynamic speed that depends on market conditions. This is the contribution of Breaking Bad Trends.

Trend-following strategies typically buy assets that have shown an upward trend in price and sell short those in a downward trend. . The literature on time-series momentum, which forms the basis of trend-following strategies, has dedicated little attention to the performance impacts of reversals or turning points in trends, also known as trend breaks. After these turning points, trend-following investors can place bad bets because their decisions are based on trailing returns that reflect an older and no longer active trend direction. To quantify these bad bets and explain the challenges trend-following strategies face, the authors investigate the impact of increasing trend breaks on asset prices. They do this by analyzing the performance of traditional monthly trend-following methods across various assets and asset classes and present three main findings from their analysis. First, they document and quantify the impact of turning-point frequency on the profitability of trend following. Second, they discover that the number of breaking points can help explain the deterioration of trend-following performance in the expansion period (2009–2019) following the GFC. And third, they offer new trend-following strategies that speed up and slow down dynamically to harvest extra performance around turning points. .

Building on their recently published research in the Journal of Financial Economics, the authors define a turning point for an asset as a month in which its slow (longer lookback horizon) and fast (shorter lookback horizon) momentum signals differ in their indications to buy or sell. If the average return of an asset over a shorter period is pointing in a different direction than the average return over a longer period, then the market may have encountered a break in trend. Their analysis provides concrete examples of how agreement or disagreement between trend breaks and other simple momentum signals of different horizons carries predictive information for future returns across a variety of assets and asset classes.

To test their hypothesis of the negative impacts of trend breaks, the authors analyze a sample period from 1990-2023, using monthly returns for 43 futures markets across three major asset classes: equities, bonds and commodities. For each market, the consider two static strategies: a  SLOW strategy with a 12-month lookback and a FAST strategy with a one or two month lookback. Turning points are situations where the SLOW and the FAST disagree.  

Plotting the annual Sharpe ratios of these static strategies against the number of turning points in a year, the authors find a strong negative relationship between the number of turning points a single asset trend following strategy experiences and its risk-adjusted performance. For assets with six or more turning points within a year, median returns are negative while for assets with eight or more turning points, the vast majority of returns are negative with annualized Sharpe ratios below −1.25 for the median asset.

Such findings apply both to single asset trend-following strategies as well as to strategies of multi-asset portfolios. To analyze multi-asset trend following strategies, the authors compute the weighted average number of turning points across all assets by allocating equal weight to each asset’s value within its asset class and equal weight to each asset class across the three asset classes. They then calculate a multi-asset static trend portfolio return as the equally weighted average of individual asset static trend-following returns. Again, the negative impacts of trend breaks are found to be meaningful: for multi-asset trend-following portfolio normalized to 10% annualized volatility over the 33-year period, a one–standard deviation increase in the average number of breaking points per year is associated with a decrease of about 8.9 percentage points in its annual portfolio return.

Periods that experience elevated levels of trend breaks have historically created performance headwinds for trend-following portfolios. The expansion period following the global financial crisis was one such period which helps explain the deterioration of trend-following performance during that period. Tracking the distribution of turning points across the 43 assets surveyed in the study period, the distribution of turning points from 2009 to 2019 exhibits an upward shift relative to the distribution in other years. In fact, nine of the 11 post-GFC expansion years have a weighted average number of turning points that rank at or above the median for the 33-year period.

To counteract the negative performance impacts of turning points, the authors propose a dynamic trend-following methodology. The approach first  partitions an asset’s return history into four observable states—bull, correction, bear, and rebound. A bull state is where FAST and SLOW are positive and a bear state is when they are negative. Correction means that the SLOW is positive and the FAST is negative. Finally, a rebound is when SLOW is negative and FAST is positive.  The authors show that after bull states there are positive average returns and after bear states there are negative excess returns.

The dynamic trend strategy return for each asset in each month blends the fast and slow returns in a way that can vary after observing different market states using mixing parameters. The mixing parameter determines how much weight is placed on the FAST vs. SLOW signals. Behavior after bull and bear states mimics the static strategy where there is a strong separate of both subsequent returns and volatility. The authors estimate mixing parameters from historical returns in months following corrections and rebounds prior to portfolio formation. Mixing parameters are estimated ex ante and do not use data from the future.

Decomposing the returns of this dynamic, multi-asset trend-following approach compared to the static multi-asset approach, the authors find that the dynamic strategy outperforms the static strategies (1-month lookback or FAST, 12-month lookback or SLOW and the average of one, three and 12 month lookbacks)  both the full 33-year study and the post-GFC expansion from 2009-2019. The dynamic trend portfolio not only generates returns after bull or bear phases in similar magnitude to the static strategy but also generates returns in months after turning points. While average returns of both methods decreased in post-GFC expansion years, the dynamic trend portfolio generated 3.4% average returns, well above the 0.3% generated of the static 12-month trend portfolio. Moreover, nearly half of the dynamic gains are from returns harvested after turning points. By contrast, faster static trend strategies or static blends of static trend strategies struggle to generate returns after turning points, particularly following the GFC where the turning points states led to negative performance.

The authors’ analysis suggests that observed market corrections and rebounds carry predictive information about subsequent returns and that such breaks can be utilized to enhance trend-following strategies’ performance. A multi-asset dynamic trend portfolio that allows momentum speed (fast or slow) to vary through time can harvest returns after turning points that might have been lost under standard static trend following. As a result, this dynamic approach may offer a more effective solution for financial markets especially in times of heightened uncertainty.