Statistical pairs trading and analyst recommendations

Abstract

PURPOSE OF THE STUDY Previous literature has documented statistical pairs trading with stocks yielding positive abnormal returns. This phenomenon violates the weak form of the efficient market hypothesis and cannot be explained by known risk factors. An emerging field in pairs trading studies focused on information flow has provided interesting results on the relation of statistical trading and information events, but without being able to explain the abnormal returns. This thesis aims to elaborate this relation by studying analyst recommendations which have been shown to be related to both stock price movements and information events, but have not been previously studied in connection to statistical pairs trading.

DATA I study a daily return data of all U.S. utility sector stocks listed on the NYSE, AMEX and NASDAQ exchanges during years 2000 to 2010. Of this sample I form 20 stock pairs every 6 months for statistical pairs trading. I match the 20 pairs with newly issued analyst recommendations during trading. The full data consists of approximately 150 stocks forming on average 6 602 potential pairs every 6 months, and 3 126 matched analyst recommendations.

RESULTS My results support previous findings that statistical pairs trading yields positive abnormal returns, though smaller in recent years, and that relative analyst recommendations hold valuable information to investors. However, I find that stock prices do not evidence post-recommendations drift in pairs after being issued divergent recommendations. Divergent recommendations are a signal of a statistical stock pair breaking up. The spread may move to the direction recommendations indicate or irrationally to the opposite direction and hence the direction of the spread between a pair cannot be inferred from recommendations. Pairs breaking up causes negative return to pairs trading and thus the aggregate abnormal positive return to pairs trading is driven by pairs that are not issued divergent recommendations.