Empirical Inference of Related Trading Between Two Securities: Detecting Pairs Trading, Merger Arbitrage, and Strategy Rules*
Empirical inference of related trading between two securities: Detecting pairs trading, merger arbitrage, and strategy rules* Keith Godfrey The University of Western Australia Working paper: 5 September 2013 The traditional approach to studying pairs trading is to simulate profitability using ex-post historical prices. I study the actual trades reported anonymously in security pairs and build statistical inferences of related trading. The approach is based on the time differences between trades. It can distinguish intrinsically related securities from pseudo-random sets, find stocks involved in merger arbitrage in massive sets of paired index constituents, and infer dominant trading rules of mean reversion algorithms. Empirical inference of related trading can enable further studies into pairs trading, strategy rules, merger arbitrage, and insider trading. Keywords: Inferred trading, empirical inference, pairs trading, merger arbitrage. JEL Classification Codes: G00, G10, C10, C40, C60 The availability of intraday trading or “tick” data with time resolution of a millisecond or finer is opening many avenues of research into financial markets. Analysis of two or more streams of tick data concurrently is becoming increasingly important in the study of multiple-security trading including index tracking, pairs trading, merger arbitrage, and market-neutral strategies. One of the greatest challenges in empirical trading research is the anonymity of reported trades. Securities exchanges report the dates, times, prices, and volumes traded, without identifying the traders. In studies of a single security, this introduces uncertainty of whether each market order that caused a trade was the buy or sell order, and there are documented approaches of inference such as Lee and Ready (1991).
[Show full text]