My colleagues’ new paper Implementing Momentum examines the real world applicability of systematic momentum investing. Momentum is one of the strongest, if not the strongest, of the major market “anomalies” documented in the literature. But most academic studies ignore real world costs and other forms of slippage when examining factors, which is likely a larger issue for momentum due to its higher turnover. The concern is that momentum is so costly to trade that its return premium is diminished in the real world.
Some papers try to address this issue in ways we think are frankly odd. Some of them look at the costs of implementing the exact academic portfolios (in other words, rebalancing often and trading immediately to precisely match the factor specifications) which is not how real-world implementers do it. In the real world, some tracking error to the academic versions1 is generally accepted in an attempt to minimize transactions costs (and often taxes, too). Other papers, that are perhaps even odder, look at traditional active stock picking, attempt to infer their factor exposures, and then lay the known failures of traditional active stock picking at the feet of these factor exposures. These papers then make the leap that these failures show that it’s really hard to implement factor investing. How that analysis relates to the ability to efficiently obtain direct factor exposure I just don’t understand.
In contrast, we take a simple approach that we think is far more relevant than either of the above dead-ends. We do a case study of what we think is a reasonably long period for implementing momentum in the real world — 7 years of live data across large, small, and international markets — and find that costs are much lower than other (we’d argue, flawed) studies have claimed.2
Further, we find that our implementation of momentum retains high fidelity to the basic momentum strategy.3 We don’t address this in the paper, but I would argue that this finding — that momentum is implementable at low cost — holds even more strongly when it’s part of a multi-factor portfolio, where factors that complement momentum, like value, will dampen turnover. All considered, we find that momentum is quite implementable in real life.
[1] As a separate issue many real world implementations attempt to improve upon the very simple academic versions, using various measures of momentum and portfolio construction. [2] We discuss this in more detail in the paper (I hope everyone reads it!), but when estimating factor premiums, we need much longer histories than seven years to draw meaningful conclusions about expected returns. On the other hand, seven years is a much more reasonable period for addressing the question of whether implementation costs impact the returns from momentum investing, which is what this paper studies. [3] That is, actions taken to reduce the costs don’t ruin or even seriously impair the ability to get the desired momentum exposure.
Article by AQR
