This article examines the accuracy of earnings forecasts made by companies in Hong Kong, China in their new issue prospectuses between 1992 and 1995. In order to obtain a better picture of the initial public offering (IPO) profit forecast accuracy, prediction errors are compared with the errors from using a random walk model and with the errors from using a growth model. The forecasts made in prospectuses are found to be superior to those made by time series models. Bias and rationality tests give conflicting outcomes as to the efficacy of IPO forecasts. While the results are consistent with forecasts being made rationally in the sense of researcher W. De Bondt and R. Thaler, and J. K. Capstaff et al., they also indicate a pessimistic bias where forecasts are invariably less than actual.