Testing the rationality of survey data using the weighted double-bootstrapped method of moments

Conventional tests for rationality of survey data on expectations are not valid in the presence of measurement errors. However, if two or more survey measures of expectations are available on the true unobserved expectational variables, we can devise the appropriate FIML estimation methods and Wald tests for rationality. This paper uses this method for survey data on expectations for the 90-day Treasury bill rates. However, the Wald tests would be based on inaccurate standard errors in the presence of heteroskedasticity, and also be subject to size distortions if asymptotic critical values are used. The present paper corrects these two problems using a weighted double bootstrap method.