Interesting white paper that suggests that spikes in short selling prior to downgrades are indicative of analysts tipping off their clients before issuing reports during 2000 and 2001. Based on a few things I saw while trading options during that span, I can't say that I'm surprised by this result at all. Nonetheless, it's good to see statistical evidence to support something I've always suspected.
By examining levels of short selling during the days prior to an analyst downgrade, the authors show a steep increase in shorts during the 3 days just before the downgrade is announced. This spike in shorting is demonstrated to be significant even after correcting for earnings releases, corporate announcements, "me-too" downgrades, and other biases. I admit I'm not a statistician, but the results look robust to me.
I have only one minor quibble with this paper. The authors should have taken into account the reason for the downgrades, and eliminated so-called downgrades on valuation. These downgrades, which occur when a stock has reached the analyst's price target, often do not result in a significant market reaction, and are also somewhat predictable. I suspect that eliminating downgrades that were a result of a stock reaching its target would increase the efficacy of the study, and may also show that analysts were more eager to tip off their customers about highly impactful downgrades.
Full pdf version of the paper can be found here. H/t Paul Kedrosky.