December 8, 2011

Is Momentum Really Dead?

There have been some working papers out that purport to show the demise of momentum. Financial blogs have picked up on this information recently and passed it along to a wider audience. But as we shall soon see, momentum may just be going through a Mark Twain moment in which rumors of its death are greatly exaggerated.  Let’s take a closer look and see what’s really going on. 

“Are Momentum Strategies Still Profitable Work for U.S. Equity?”  came out in October 2010.  It showed that the mean monthly return from momentum during the five years ending in 2009 was -.163%. In contrast to this, the mean monthly momentum return from 1965 through 1998 was 1.14%, by the author’s calculations. Momentum returns were determined by going long the strongest and short the weakest 10% of stocks during the preceding six months. Positions were held for six months then readjusted. However, there are no statistical criteria, such as t statistics or p values, which might indicate any significance to these results. Nor are there any alternative momentum calculations, such as the use of different holding or formation periods. Finally, there is no exploration as to why momentum results may have changed over time.  So there isn’t much to go on here.

The second paper that challenges momentum came out in April of this year. It  is called  “Momentum Crashes”.  In it, the author looks at momentum based on a twelve month formation period lagged one month.  The holding period is one month before portfolio reformation. Top and bottom deciles are again used to sort stocks into momentum longs and shorts.  Here the author identifies what has been going on under the surface. He shows there have been two major “momentum crashes” that have depressed momentum returns. The first occurred in July-August of 1932.  The second was in Mar-May 2009.


We see from this chart that if your analysis of momentum performance is for a short period of time (say the last five years that ends in 2009), results look lackluster. A single, rare event can seriously distort performance over a short evaluation period.  Over a longer time frame, however, momentum still looks good. Furthermore, momentum returns are depressed during these two “crash” periods only because momentum is looked at from both a long and short point of view.  The “crash” loses occurred on the short side during strong market rebounds following market declines with high volatility. During the 1932 “crash” period, past momentum losers rose by 236%, causing heavy losses as shorts, while past winners were up only 30%.  Similarly, during the 2009 crash, past losers were up 156%, while past winners were up only 6.5%.  However, most actual momentum investing is done only on the long side. If we look at how long only momentum did during these “crash” periods, we get an entirely different story.

Here are the monthly returns from the above winner minus loser (WML) momentum portfolio versus the AQR long only momentum equity index (top one-third stocks based on 12 month momentum lagged one month) during the worst WML momentum months of 2009:
                                                                                          WML                  AQR
Aug 09             -24.8                    2.3
Apr 09              -46.0                    1.8
Mar 09             -39.6                     5.3

Long only momentum looks good here.  For a longer term view, the folks at Dorsey Wright recently reported in a blog post called  Momentum Over Multiple Cycles  some long only momentum results using data off the Kenneth French website from 1930 through October 2011. Portfolios were selected based on being in the top half in market capitalization and the top third in momentum, which is similar to the AQR index methodology. Tracking ten year rolling returns from 1940 onwards, they discovered that the average ten year return of the momentum portfolios was 405%, versus 216% for the S&P 500 index. Momentum earned nearly double the rate of return of the market. In every ten year period, including the most recent one, the momentum portfolio outperformed the market portfolio.

But long only momentum has downside in the form of high drawdowns during equity bear markets. For example, the AQR equity index was down 49% from October 2007 through February 2009.  However, as those who have read my research paper know, momentum applied to only equities is not the best way to proceed. During this same time period, our momentum portfolio using multiple asset classes achieved a profit of 18%. Momentum is still very much alive and well. Long live momentum!

June 9, 2011

Key Momentum Factors

There are three key factors in structuring a momentum program. They are all known through momentum research, so it's not hard to get them right. 

The first is the formation look back period, i.e., how many months back do you measure momentum.  In 1937, Cowles and Herbert came up with 12 months. In 1967, Levy used 6 months. In 1993, Jegadeesh & Titman showed that momentum works well using anywhere from 3 to 12 months, with the best results being 6 to 12 months.  Literally hundreds of research papers since then have reconfirmed the 3 to 12 month time window. It has held up consistently when applied to data from the 1890s until now.  It works with nearly all markets and asset classes.

There is a sharp drop off in momentum profitability once you extend beyond a 12-month formation period. As you reach 3 to 5 years of past data, mean reversion becomes strong, giving you the opposite effect of momentum. 

Anyone needing convincing of the 3 to 12 month formation window can look at the key momentum research papers on AQR's Annotated Bibliography. You can also go to the Social Science Research Network website and do a search on the word "momentum."   You will find dozens of papers that all successfully use a 3 to 12 month time frame.
  
The second key momentum factor is the chosen universe of available investment opportunities. Currently, most investment programs that use momentum apply it to either individual stocks or, occasionally, industry groups. But as we know from my research paper, momentum is more powerful when it is applied across a group of non-correlated assets. The paper Momentum and Value Everywhere also illustrates this point.

It's also desirable for one's momentum portfolio to include a way to retreat to the safety and stability of cash or short term fixed income securities.  By not including these in your momentum portfolio, you forgo the risk reduction that comes from adapting to market conditions and opting out of risky assets early during market regime changes. Absolute, or time series, momentum deals with this issue explicitly. It requires a positive trend as a prerequisite to momentum investing.

The third and final key momentum factor is how many assets from your investment universe that you decide to use. We all know that diversification is a good thing. But more is not always better with momentum. Profits from long positions decline as you go down the ladder of assets ranked by momentum. 

                                     Volatility
                                                                             Number of Assets


There is a dramatic fall off in volatility as one goes from 1 to 2 asset classes, then a lower drop in volatility when going from 2 to 3 assets, and so on. Momentum profits, on the other hand, drop in a more linear fashion as you add additional assets. The following example shows what happens if you use too many asset classes.

The SGI Global Momentum index universe consists of ten exchange traded funds in diversified assets including developed market equities (EU, US, Japan), emerging equities (Asia/Pacific, Russia, Latin America), listed private equity, bonds (Euro zone Government Bonds and Inflation-linked) and commodities. Equities funds make up 7 out of the 10 assets. The five best performing shares at any one time make up the index. The SGI index began in late 2007. Over the past 3 years, it has outperformed the MSCI World index by a little less than 2%, while having a maximum drawdown of 46%, versus 53% for MSCI World. Too many assets not only lowered the return of the SGI Index; it also did little to reduce the maximum drawdown of the momentum portfolio. With 5 assets in play at all times, there is nowhere safe to fully retreat to under adverse market conditions.