This paper offers an improvement to the trade-to-trade model for event studies. While the trade-to-trade model of Maynes and Rumsey (1993) addresses the problem of thin trading by eliminating periods in which no trading is recorded, the proposed improvement addresses the influence of zero-value returns resulting from liquidity trading. This entails segmentation by the sign of company returns (positive, negative, zero). The approach allows for all levels of thinness in security trading. It is evaluated against the trade-to-trade methodology developed by Maynes and Rumsey (1993) and the Market Model using Monte Carlo simulations developed from the method of Brown and Warner (1980 and 1985). The improved trade-to-trade model is better at picking up the presence of very small levels of abnormal performance.