Faced with a lack of track record for active extension funds, researchers are forced to re-create how these funds would have performed if they had existed for some time. The idea is to select an “alpha model” (an unfortunate term since alpha cannot technically be “modeled”) and run it back in time using real market data to see how it would have performed. The model is run once as a long-only portfolio and then again using any number of 1X0/X0 strategies. Comparing the performance of the models can yield some insight into whether the short extension itself adds value to a given alpha model.
The latest to conduct this analysis are Carl Armfelt and Daniel Somos, graduate students at the Stockholm School of Economics. Armfelt & Somos selected a set of basic Fama/French factors to create their alpha model and ran it all the way back to 1927. Here’s what they found:
As you can see from the last line on this table, the alpha model was true to its name – producing a positive alpha in every decade except 1947-56. We won’t get into the details of the model here, but you can check out his paper for a list of the factors it contained. But what’s interesting is that the alpha-bump from adding the short-extension seems almost random.
Many people quite correctly argue that adding a short extension is like simply “leveraging” the alpha-producing ability of a manager (or lack thereof). While this is conceptually true, the results from this analysis show that a variety of exogenous factors come into play in determining the extent to which alpha actually increases. In other words, a 130/30 fund does not simply have an alpha that is 30% higher (or any predictable amount higher).
For example, the 1930s were relatively kind to the alpha model chosen by Armfelt & Somos, and the 130/30 version produced 60% more alpha. The 1987-1996 period, however, wasn’t so kind to the alpha model – yet the 130/30 version produced almost as much alpha as in the 30’s (almost 5 times as much as the long-only model did in the 1987-1996 period).
More recently, check out how 130/30 suddenly began to blow away the long-only version of the same alpha model in about 2001…
Research by Roger Clarke and his colleagues at Analytic Investors (see posting) and by Manolis Liodakis at Citibank (see posting) have proposed several exogenous factors that might account for time-varying outperformance of short-extensions. But only time will tell whether those studies (or this new one) correctly identify the various factors influencing the success of 1X0/X0 strategies. Even Armfelt & Somos hedge their bets:
[Note: The first 25 pages of this paper provide a good overview of 130/30 and touch on many of the themes familiar to regular readers – quant vs. fundamental, optimal level of shorting, regulatory rationale etc.]
“This should be considered as an observation rather then a formal test, and this data can only give investors an idea of how a sample of active-extension funds have performed, over this short time period. In addition, the empirical data covers a time period with severe market turmoil, and might not be indicative of performance in more normal market conditions. We do not claim that our results are indicative of future performance.”