Trend-following is the dominant alternative investment strategy. Its historical return profile has been attractive on its own and for diversification purposes. It is suitable for rising and falling prices, albeit not for range-bound and “gapping” markets. A basic trend-following algorithm is easy to build. Trend-following commands over USD300 billion in dedicated assets and a lot more are managed by informal trend-followers. The style is itself a major force of price trends, with no direct ties to fundamental asset value.
Societe Generale Cross Asset Research (2016), “Why you need trend systems in your portfolio – A multi-asset approach”, 22 June 2016.
There is no public link to the paper. Please contact Societe Generale for a copy.
The documented success of simple trend following ties in with other evidence that markets are not information efficient, as summarized here.
The below are excerpts from the paper. Headings and cursive text have been added.
What is trend-following?
“Trend-following is the investment strategy of exploiting the direction of moves in asset prices and profiting from the persistence of those moves…There is ample empirical evidence of the existence of a trend effect in financial prices… The persistence of trend-following has been well documented.”
“Over the past 25 years, a fund invested in the ten biggest trend-following CTAs would have outperformed the MSCI World and be roughly on par with the S&P. And that would have been achieved with much lower volatility and a more controlled drawdown. This good track record explains the growth of the CTA industry over recent years. CTAs currently manage more than USD320bn.”
N.B.: CTA means Commodity Trading Advisor, a hedge fund that runs investment strategies in futures markets. According to Morningstar, roughly two thirds of these funds deploy systematic strategies, mainly forms of trend following.
“Trend following is one of the few investment concepts that has survived the test of time. Two hundred years ago, economist and trader David Ricardo (credited as the richest economist in history) came up with the statement ‘cut short your losses; let you profits run’. Later, several generations of aspiring traders grew under the adage ‘the trend is your friend’. The famous turtle experiment by Richard Dennis showed that even novice traders can be successful if they abide to trend-following rules in a disciplined manner.”
“Why trend-following exists has been an area of active research interest. Numerous explanations have been put forward: behavioural patterns, institutional funds flows, or the existence of a premium, to cite a few. No consensus exists about the dominant factor, and each explanation has its own merits.”
“There is a real abundance of approaches on how to measure trends…from simple to highly exotic ones. A recent academic publication has shown that there is equivalence among many of the commonly accepted trend measures. Moving averages, moving average crossovers, and even a few filtering methods, like the HP filter and the Kalman filter, can be unified in a common framework.”
N.B.: An HP (Hodrick-Prescott) filter is a data-smoothing technique designed to estimate the trend a time series by discounting short-term fluctuations. A Kalman filter is a multi-equation algorithm that estimates the unobservable state of a process by minimizing the implied mean squared error of observable data.
How to build your own trend-following system
“We propose our own way to build a successful trend-following system…[and] show that it can match a CTA’s performance over recent decades.”
“Our investment universe consists of 56 future liquid markets that span various markets and asset classes – commodities, equities, currencies and fixed income. Every market segment has a fixed allocation in the total strategy. This allocation is designed to ensure prudent representation of the various regions and asset classes.”
“We derive a trend signal that is kept simple and straightforward: we consider whether the asset’s average risk-adjusted return has gone up or down over a given time window. If the average return is positive for a given lookback period, the signal is +1. If the average return is negative, the signal is -1. If the return is zero, the signal is zero.”
“We chose lookback windows of 1.5 months, three months, six months and 12 months…We average the signals of all lookback periods using… equal risk contribution…as the most robust from a theoretical viewpoint.”
“Once the position in each market is computed, we aggregate the strategies over each asset class…This allows us to obtain a single trend-following strategy for each asset class…Finally, we aggregate the strategies per asset class with an inverse volatility rule,”
A stylized trend-following return profile
“Trend-following strategies have a very different return profile than those of standard assets. They take positions opportunistically and with decisions solely determined by the trend. As the trend changes, so does the position of the trend-following system. If we assume that upward and downward trends appear with equal chances, we can expect trend-following to exhibit a low correlation to the asset class…When combined with a traditional portfolio, trend-following strategies tend to introduce useful diversification. With a 20% allocation to trend-following, the Sharpe ratio of a diversified portfolio increases by 40% and drawdown decreases by 8%.”
“We split past market data into four distinct regimes. A crisis is a period when the S&P falls by 10% or more; a gap event occurs when the S&P falls by 10% or more in one week; the market is range-bound when the S&P crosses its one-year average more than ten times; the market rises or recovers the rest of the time.”
“[Stylized historical performance features are:]
- Trend-following systems have a defensive risk profile due to their positive returns in most market downturns. They can also generate returns when the market rises.
- Trend systems benefit from markets rising or falling substantially… As the momentum gets stronger, the expected P&L increases for all lookback windows.
- In a range-bound market, a trend-following system is more likely to be wrong and to lose money.
- Trend-following strategies may fail if markets fall or rise abruptly. Only put options would have effectively hedged an investment in the US stock market in October 1987…Trend-following can face headwinds when markets abruptly recover after a sharp sell-off. This was the case during Q4 1998 and March 2009, just after the Lehman crisis.”
“In a trendless environment, a trend-following strategy will have zero expected profit from trading and will only incur transaction costs… transaction costs are an increasing function of the ratio of the execution and replication costs to volatility, and inversely proportional the length of the lookback window….In a trendless market, a short-term, one-day strategy costs around 3.4% per year for a liquid equity market. In similar setup, a one-year strategy costs just around.30bp per year. Hence, in a trendless market, it is much more cost-efficient to trade on longer periods.”