Variance-Ratio Tests: Small-Sample Properties with an Application to International Output Data.

Two aspects of statistical inference using variance-ratio statistics are studied, (1) the accuracy of asymptotic approximations in small samples and (2) the size distortion arising from searching over many horizons in deciding whether to reject a model. A joint test combining variance-ratio statistics at various horizons is proposed and a Monte Carlo procedure for conducting exact inference is provided. The real output data of nine countries are used to discuss these issues.