The Accuracy of Trade Classification Rules for the Selected CEE Stock Exchanges

Abstract

We assess the accuracy of the popular trade classification rules—the tick (T), the reverse tick (RT), the quote (Q), the Lee and Ready (LR) and the Ellis, Michaely and O’Hara (EMO) rules—for the selected Central and Eastern Europe stock exchanges: the Warsaw Stock Exchange (WSE), the Prague Stock Exchange (PSE) and the Budapest Stock Exchange (BSE). We employ the transaction data on the most liquid stocks included in the large cap indices, namely WIG20, PX and BUX, from 9th of January till 14th of March, 2019, and discover the Q rule to be the most successful. It correctly classifies 100% of trades initiated both by the buyers and the sellers in the case of the BSE and 84.35% (82.78%) in the case of the WSE (the PSE). The results obtained for the LR rule are similarly good (96.32% for the BSE, 84.34% for the WSE and 82.86% for the PSE). The third one is the EMO rule with the rate of success of 83.69% for the BSE, 80.40% for the WSE and 67.86% for the PSE. The T and the RT rules are characterized by a considerably low level of accuracy (ranging from 19.91 to 40.65% for the T rule and from 13.04 to 38.37% for the RT rule). The modifications of the T and the RT rules that take into account the preceding and the following transaction price changes allow to obtain the distinctively higher estimates of accuracy statistics.