The hypothesis that New Zealand 90-day bank bill futures rates are an unbiased predictor of 90-day bank bill rates is tested by applying the single-equation method of Stock and Watson (1993) to quarterly data from 1989 to 1997. The results do not reject the unbiasedness hypothesis for the one and two-quarter-ahead horizons tested. However, the estimated residuals are found to contain significant serial correlation in both cases, which suggests that some degree of information inefficiency might exist. The relative forecasting performance of futures rates against the random-walk is also investigated for weekly horizons up to 26 weeks. The results indicate that futures rates outperform the random-walk over all horizons tested, with the improvement statistically significant for all horizons greater than 1 week.