Comparing predictions and outcomes: Theory and application to income changes

Household surveys often elicit respondents' intentions or predictions of fu ture outcomes. The survey questions may ask respondents to choose among a s election of (ordered) response categories. If panel data or repeated cross- sections are available, then predictions may be compared with realized outc omes. The categorical nature of the predictions data complicates this compa rison, however. Generalizing previous findings on binary intentions data, w e derive bounds on features of the empirical distribution of realized outco mes under the "best-case" hypothesis that respondents form rational expecta tions and that reported expectations are best predictions of future outcome s. These bounds are shown to depend on the assumed model of how respondents form their "best prediction" when forced to choose among (ordered) categor ies. An application to data on income change expectations and realizations illustrates how alternative response models may be used to test the best-ca se hypothesis.