There is much debate over whether agents form rational expectations of variables or whether they suffer from systematic errors in judgment. This paper estimates models for plant-level survey data in order to test rationality for those manufacturing plants that report expectations of capital expenditures. An advantage of using such data is that rationality is tested in markets where agents may not have knowledge of each others' expectations so strategic motives behind purposefully irrational forecasts are minimized. Statistical estimates and test results suggest that expectations may indeed be rational depending on size. That is to say that the larger a plant is, the more resources it can expend on forecasting its future needs. Thus, the statistical results in this paper validate, for the first time, a class of assumptions in the macroeconomic literature.
