Back to Portfolio for the Future™

Risk-Adjusted Time Series Momentum Strategies

August 21, 2016

The name is awkwardly long, and the standard abbreviation, “RAMOM,” sounds like what one says when cheering on one’s mother as she nears a finish line.

Still, risk adjusted time series momentum strategies have something to be said for them, in comparison to cross-sectional momentum (MOM without preface), or even to unadjusted time-series momentum (TSMOM) strategies.

A study published late last year in The Journal of Alternative Investments by Martin Dudler, Bruno Gmuer, and Semyon Malamud, goes into the numbers, testing both RAMOM and TSMOM against a universe of 64 liquid futures contracts.

The study began life as a Swiss Finance Institute Research Paper. Dudler, a former UBS quant, is responsible for portfolio management and trading at Quantica Capital AG.  Gmuer is a founding partner at Quantico. Malamud is a scholar affiliated with the Swiss Finance Institute and the Centre for Economic Policy Research.  The issues they address in the course of this study have a long academic past, reaching back at least to some of John Maynard Keynes’ writings in the 1920s.

Background of Suspicion

Dudler et al have to make their case against a background of suspicion. After all, momentum trading on its face sounds like an all-too-common and losing casino gambler’s strategy. If the wheel has been stopping on red a lot lately, bet on red! Another school of thought among traders and portfolio managers, contrarianism, sounds the same although contrary: if the wheel has been stopping on red a lot lately, bet on black! Neither strategy does anyone, other than the house, much good at casinos. Shouldn’t futures traders be wary of their analogs?

From a more technical perspective, a paper by Yao Hua Ooi, Tobias J. Moscowitz, and Lasse Heje Pederson not long ago critiqued the inefficiency of momentum seeking strategies. Ooi et al found (in the words of a paraphrase of their work by Dudler et al.) that “averaging past realizations of highly heteroskedastic returns may produce a very noisy estimate of the true expected return.”

Vocabulary refresher: a population of returns is heteroskedatic if it consists of a lot of subpopulations that have differing variances from one another.

So Dudler at al gave now say to Ooi et al, “we take your point, and we’ll work on the numbers to account for that.” That explains their distinction between TSMOM and the new better thing, RAMOM. By a simple risk-adjustment procedure one removes heteroskedasticity from the trading signals, thereby significantly improving the strategy’s performance.

Ignoring Cross-Sectional Comparisons

The paper by Ooi et al had another important feature. It abstracted from the issue of “cross-sectional” momentum. Until that paper, studies of momentum had defined the issue by the performance of different instruments measured against one another.  This is a bit like reasoning, “Red has been coming up a lot lately on this wheel, compared to the red/black percentages at that other wheel across the room – so I’ll bet red on this one and black on that one.” Ooi et al defined momentum by a single wheel, by the performance of a single instrument over a given look-back period, hence the “TS” part of “TSMOM.”

Some of the scholars who have written on the subject of late have been more sweeping in their criticisms of TSMOM than are either the Dudler or the Ooi groups. For example, in a recent paper by Abby Kim, a financial economist with the U.S. Securities and Exchange Commission, Kim and two associates took the view that without scaling by volatility, “time series momentum and a buy-and-hold strategy offer similar cumulative returns, and their alphas are not significantly different.”

But Dudler et al seek to continue and build upon the time series aspect of the Ooi group’s work, and thus on TSMOM itself. They build by creating RAMOM, and then they crunch numbers to announce that RAMOM lets an investor gain exposure to three important risk factors: momentum itself, as well as market and value risk (taking each in the Fama-French sense) without having to trade a large number of assets.

Another bit of good news concerns what we might, reverting to the above analogy, call the casino’s cut. The dollar turnover of RAMOM strategies is about 40% lower than the same figure for TSMOM.