Abstract by Dean Sobczak
A Pairs Trading Approach
Pairs trading is a quantitative method for taking advantage of temporary stock mispricings. The investor is alerted to such mispricings when the spread between two stocks substantially diverges from its average. A pairs trader bets that the spread will eventually return to its equilibrium state and executes trades accordingly. In this presentation, we describe a novel strategy to carrying out pairs trading. Our approach includes a three-stage procedure to choose the ideal stock pair for some trading period. The three stages are: (1) cluster the stocks into candidate pairs using the nearest neighbors algorithm; (2) extract the m stock spreads with the most extreme augmented Dickey-Fuller test statistics; and (3) among the m spreads, find the spread with the optimal Ornstein-Uhlenbeck (OU) process parameter settings. Then the selected stock spread is modeled as an OU process and the optimal threshold to signal pairs trades is estimated via Monte Carlo methods. Finally, the optimal threshold is employed during the subsequent trading period. This process is repeated multiple times over a specified trading horizon before calculating the investor’s earnings.