One-to-many matching: An alternative trading cost comparison technique

We compare the relative merits of different matching techniques used in cross-market studies. Using the latest November 2007 data from both NASDAQ and the NYSE, we conduct simulations in which the firm characteristic distributions differ on the two matching sides. We keep the sample size small to create a difficult matching environment and highlight the relative strength of the different matching approaches. We propose a one-to-many matching method that has not been used previously in cross-market studies; each NASDAQ stock in our sample is matched with several comparable stocks from the NYSE. This method yields small matching errors than the widely used one-to-one matching process. Our simulation shows that choosing the best-matching technique actually matters. The one-to-many method consistently produces the correct result while the standard one-to-one without replacement method generates wrong answers under difficult matching environments.