It has been said by many that stock picking is dead or that it hasn’t been “paying off” recently. While correlations remain high and the environment for stock picking is clearly more difficult than in the past, we do not believe that it’s entirely lost in the investment process (that’s coming from a top-down biased group!). In addition to a highly-correlated environment, the business cycle has been shorter than normal and more volatile than in past decades in which most backtests are constructed. Moreover, an enormous M&A cash build-up on corporate balance sheets as well as ongoing stimulus rumors have made the short side of investing even riskier. Surely, it is a bitter experience when a shorted stock rips to the upside because of rumors of QE3 or the like.  While we can’t prevent the rumor mill from circling, we can take steps to help increase the hit rate on the short side.  While much of our work in the past has been about selecting stocks within the context of the business cycle, this report focuses on a short-selection model that we believe is fit for all seasons and for all investors … no regressions necessary!

A good model starts with strong building blocks. When creating a short model, too often investors simply take the “bad” side of a long-only model. While this can be moderately effective, it may not be the best solution possible. As we’ve seen with many factors (e.g., accruals, momentum, etc.), performance is not always symmetric across fractiles. We built this model (last published in 2010) with a short-only framework in mind from the bottom up. Over half of the factors in our model display great efficacy on the short side, but not on the long side. Too often a factor displaying this profile would be tossed out by stock pickers.  Since we’re only looking for shorts, we’ve kept them. Since 1994, the model has outperformed the S&P 1500 over a 12-month period only 13% of the time (or underperformed 87% of the time!). Note the performance in 2012 (and since 1994) has been a consistent downward trend, hardly displaying a market-like profile (i.e., beta) that so many models look like today.