Using Matched Samples to Test for Differences in Trade Execution Costs

This paper provides guidance on how to use matched samples to test for differences in trade execution costs (e.g. quoted and effective spreads). Based on extensive simulation results, we conclude that the best practice is to match firms one-to-one based on market capitalization and share price, and to test for differences between the matched pairs using a Wilcoxon signed rank test. We demonstrate that pre-sorting by industry groups or discarding apparent poor matches may reduce test power. We show that, in general, tests based on one-to-one nearest neighbor matching have comparable power and less size distortion than alternatives that place more weight on distant firms. We find that matching without replacement can reduce size distortion when the control sample is relatively small. We highlight conditions under which matched sample estimation may be preferred to the corresponding event study