Matching and the Estimated Impact of Inter-listing (updated July 2003)

sing 1998 trade and quote data for securities listed on the Toronto Stock Exchange, this paper employs nonparametric estimation to measure the effect of being interlisted on a US exchange on: (i) the daily number of trades, trading volume, and dollar trading volume; (ii) the number of inside quote revisions and the percentage bid-ask spread; (iii) registered trader gross trading revenues; (iv) composition of order flow between orders submitted for client, non-client, and registered trader accounts. Unlike previous studies, I use kernel-based matching estimates in addition to variants of the standard nearest-neighbor approach for constructing matched samples of interlisted stocks and non-interlisted stocks. I explore the sensitivity of results to: (i) using different bandwidth parameters and caliper-matching criteria; (ii) using different matching characteristics; (iii) the exclusion/inclusion of firms. I highlight instances when kernel-based and nearest-neighbor matching estimation techniques produce significantly different results and thereby argue that results based on standard matching techniques commonly employed in the finance literature should be interpreted cautiously.