Be sure to check out our guest post over at research supersite Alpha Architect:
Backtests of long-term strategies like tactical asset allocation are usually shown trading at the end of the month, both because it makes the analysis simpler and because monthly asset class data is easier to come by. In our guest post we talk about our unique approach to analyzing what happens when these strategies execute trades on days other than the end of the month. Note: Members have access to this data for all of the monthly strategies that we track.
Generally speaking, strategies have performed better when trading at the end and turn of the month than on other days. We talk about two reasons why: the first positive and to be exploited, and the second negative and to be avoided.
We appreciate the opportunity to be able to share our findings with Alpha Architect’s audience, and we highly encourage our readers to check out our post.
In response to our guest post, Corey Hoffstein, one of the brains over at the very smart Newfound Research, asked an important question that we didn’t cover: Was this outperformance at the end and turn of the month consistent over time, or was it the result of a handful of outliers?
The graph below is our response. Here we show the average outperformance by all of the monthly TAA strategies that we track, trading at the end of the month (DOM 21) versus all other days in orange, and trading near the turn of the month (DOMs 1-3 and 19-21) versus all other days in blue. Positive results indicate that executing monthly trades on those specific days outperformed executing trades on alternative days.
In both cases, outperformance has been fairly consistent, but appears to have accelerated from roughly 2004 – 2011. Outperformance by TAA strategies trading at the end and turn of the month does not appear to be the result of a handful of outlier periods.
While the effect is relatively small, it is widespread. Per our original post, 87% of the strategies tested have performed better when executing trades at the end of the month versus the average day.
Note that both series have flattened out in recent years. The cynical view would say that this might be an indicator of some degree of overfitting to the end of the month in historical pre-published results, as more recent data is mostly out of sample. Both series saw other long flat periods though, and time will tell whether this trend continues.
To reiterate the point made in our Alpha Architect guest post, if this were the case it wouldn’t mean that a given strategy isn’t worthwhile, only that historical results as they are usually shown may be a bit overly optimistic. We presented a simple “back of the envelope” approach in our post for discounting for this end of month overfitting.