Abstract/Summary

Testing hypotheses regarding how individual survey respondents form their expectations is susceptible to the multiple testing problem. The probability of falsely rejecting the null hypothesis for one or more respondents will exceed the nominal single-hypothesis significance level. The Bonferroni correction and related approaches control the family-wise error rate, but are conservative and result in low power when the null hypotheses are false. We compare controlling the family-wise error rate with the effect of controlling the false discovery rate and the false discovery proportion, in terms of the conclusions we draw about forecaster behaviour. The effects of adjustments for multiple testing are investigated for tests of weak efficiency and the over-reaction hypothesis, for beliefs about the persistence of shocks to output growth, and for the accuracy of survey respondents’ perceptions of the uncertainty they face.