A0992
Title: Statistical arbitrage without arbitrage
Authors: Valentina Raponi - IESE Business School (Spain)
Paolo Zaffaroni - Imperial College London (United Kingdom) [presenting]
Abstract: Statistical arbitrage strategies are quantitative trading strategies based on alpha, i.e., the signal extracted from the residuals when fitting an asset pricing model to return data. A normative theory is developed that combines the insights of mean-variance portfolio choice with statistical arbitrage in the no-arbitrage setting of the APT, specifically studying statistical arbitrage without arbitrage opportunities. A novel two-fund separation result is established, that combines the inefficient statistical arbitrage portfolio and betaportfolio (i.e., the portfolio stemming from factor asset pricing models), showing how their combination can span the efficient frontier. In general, as the number of assets increases, the statistical arbitrage portfolio dominates the $\beta$ portfolio both in terms of the magnitude of its weights and in terms of its SR. Building on insights from mean-variance portfolio choice, it is demonstrated how to construct a special statistical arbitrage portfolio that does not require estimating alpha, contrary to the prevailing view. When statistical arbitrage is combined with factor asset pricing modelling, alpha and the factors premia might not be jointly identified, jeopardizing the possibility of constructing a statistical arbitrage portfolio.
