100 Years of Trend Following

Investors like a good track record. No matter how many times they’re told “Past performance is not necessarily indicative of future results,” no matter how many times a study shows that Morningstar’s top-performing mutual funds go on to underperform their peers, people want to know that something has worked in the past before they’ll risk their money on it. (Except when they’re piling into the hot new trend… Yes, we know it’s contradictory – as Josh Brown would say, good luck with that.)

It’s why the most common selling point for managed futures tends to be someone pointing at the industry’s returns during the 2008 financial crisis: “It worked before, don’t miss out next time!” But for some, the desire for a track record goes way beyond the last few years, or even the last few decades. They aren’t the kind that seeks out the latest fad; they want to know that an investment strategy has withstood the test of time, to know that a hot 2008 wasn’t just a fluke, but part of a larger pattern of solid performance.

And for managed futures, this presents something of a problem. You see, it’s a fairly young investing strategy: the Newedge CTA Index only goes back to 2000. The BarclayHedge CTA Index? 1980. This is certainly not a criticism of the indices – after all, you can’t build an index out of something that doesn’t exist, and the famous “Turtle Traders” didn’t get started until the mid-1980s. But it can be a problem for those who want more performance history.

And that’s where $40+ billion hedge fund AQR Capital Management comes in, with their recent research piece “A Century of Evidence on Trend-Following Investing.” Their paper has “for investment professional use only” stamped on it and some language that it is not to be redistributed, so we’ll abstain from linking to it here – but if you’re so inclined, a simple Google search is likely to find a publicly available version.

In light of the continuing conversation we’ve been having over on LinkedIn on “the death of trend following,” their piece couldn’t be more timely. Of course, this is all hypothetical backtested data, so all of the regular disclaimers apply. This is not representative of actual trading profits, but rather a tool to see how a simple trend following strategy might have performed in another time. Their model is to analyze the past 1, 3 and 12 month performance of markets in their portfolio, and if the performance is positive, consider that an uptrend and initiate a hypothetical long position, and vice versa if the performance negative.

You should definitely read the piece yourself, but here are our main takeaways:

1. They find “trend following” has had a -0.05 correlation to both equities and bonds over 100 years – impressive to say the least. It’s hard to be much less correlated than that.
2. This is the proper way to see effect of fees (versus critics we’ve seen in the past, like the author of “The Hedge Fund Mirage”). After fees, their model shows the client retaining 71.5% of the gross returns over time.
3. The model’s return for the 2003 to 2012 decade turns out to be the second-lowest on record (after 1933-42), showing just how difficult an environment it’s been (despite great performance in 2008). This also helps explain some of the recent reduction in returns of some programs.
4. When they plot their model’s return against the return of the S&P 500, it forms a nice “smile” curve, showing that managed futures likes big moves up or down, not just market crisis periods. The largest returns for their model were during the largest S&P 500 moves – both positive and negative.
5. Speaking of crisis periods, they show 10 of the biggest market crisis periods of the past 100 years. The trend following model remained positive in all but one (the 1987 crash).
6. They tackle the “industry is too big” argument – making estimates that trend following could at most represent 0.2% of the size of the underlying equity markets, 3% of the underlying bond markets, 5% of the underlying commodity markets, and 0.2% of the underlying currency markets. They conclude that it seems unlikely that trend following’s trading activity would have a material effect on the markets’ trend dynamics.
7. They also address the problem of high correlations (risk on/risk off), looking at the average pair-wise correlation between every market in their model going back 100 years. The data shows that we’re in a rapidly rising correlation environment, although it is still smaller than past such moves in the 1920s and 1930s.
8. With a 100 year lens – the current drawdown and poor performance doesn’t look so bad. Of course, not many investors have a 100-year investment time horizon… but it certainly makes the recent struggles look more like an anomaly than a “new normal.”

We had just one point of criticism for the piece. Despite the title, they are only using 3 markets back the full 100 years, and for the first 60 of those years their market data includes only stock and bond markets. The commodity and currency exposure in the model is only there for the past 40 years. Don’t get us wrong, that’s an impressive test in and of itself – but there were corn, wheat, and cotton futures going back at least 100 years (although the data is surely more difficult to come by), and getting gold and silver prices shouldn’t be too hard for what is likely a research team of dozens . Nevertheless, it’s a great piece, and definitely worth a read if you can find a copy.