AQR on momentum

Cliff Asness, Andrea Frazzini, Ronen Israel and Toby J. Moskowitz have a lovely SSRN paper "`Fact, Fiction and Momentum Investing."

The whole momentum literature is so huge, it's hard to know where to sum up, and this is a great place, especially if you're teaching an MBA class. Since they're obviously a bit conflicted (momentum + value is one of AQR's core strategies), their emphasis on the simple facts, which you (or your students) can replicate from Ken French's databse, is great.

Some "Myths,"

Myth #1: Momentum returns are too “small and sporadic”. Like any factor, it's not an aribtrage opportunity.

The return premium is evident in 212 years (yes, this is not a typo, two hundred and twelve years of data from 1801 to 2012) of U.S. equity data

So, to sum up, who you calling small and sporadic?

A writing style more bold than my own.


Myth #2: Momentum cannot be captured by long-only investors as “momentum can only
be exploited on the short side.

(evidence that the long side works as well)

Myth #3: Momentum is much stronger among small cap stocks than large caps.

Putting it starkly: in-sample, out-of-sample, calculated in Greenwich Connecticut, Chicago, Boston, Palo Alto, Santa Monica, Austin, or in the library with a candlestick, wherever or however you want to look, along any dimension, those who make the claim that momentum fails for large caps, while being supporters of value investing, are not simply mistaken, they have it backwards.

Myth #4: Momentum does not survive, or is seriously limited by, trading costs.

Frazzini, Israel,and Moskowitz (2013), “FIM”, which uses trades from a large institutional investor [hmm, I wonder who that could be] over a long period of time. . Their conclusion is that per dollar trading costs for momentum are quite low... 

... Trading patiently (by breaking orders up into small sizes and setting limit order prices that provide, not demand, liquidity) and allowing some tracking error to a theoretical style portfolio can significantly reduce trading costs... 

Where did this myth come from? ... First, the studies that find much larger trading costs do so because they estimate costs for the average investor...which turn out to be about ten times larger than the costs of a large institutional manager,...Second,...these other studies examine portfolios that do not consider  transactions costs in their design, which can significantly reduce turnover and therefore trading costs further. 

This really is the DFA sales pitch. Momentum, a simple strategy may work. But if you try to do it, you'll get swamped by trading costs.


Myth #5: Momentum does not work for a taxable investor.

and so on.

This isn't a 100 percent endorsement. I've seen other cuts of the data that suggest other results, particularly Fama and French's finding that momentum didn't exist 1926-1963, the big left tail of the financial crisis, the question whether it's finally being arbitraged away, and Fama and French's evidence that it's weaker in large stocks, though not absent. I need to reconcile all the evidence, and it's still a huge project that I've put off for another day.

Also, momentum is either large or small depending on how you look at it. Suppose returns have a 0.1 autocorrelation, and thus a 0.01 R2 in R_t+1 = a + b R_t + error_t+1. The top 1/10 of stocks in the previous year went up about 50%. The bottom 10% of stocks in the previous year went down about 50%. So, a 100% previous year return spread, and a 0.01 R2 implies a 10% return spread next year -- about what we see in the 1-10 momentum portfolios. You might say that a 0.01 R2 is a tiny bit of forecastability. You might say that a 10% return spread is a huge phenomenon. They are the same.

But if you want one great paper to read and assign your students about momentum, this is it.