The purpose in this paper is to examine the properties of time series of expectations data derived from survey data, when explicit allowance is made for the time horizon of the forecasts and the information set which agents possess. Our empirical results suggest that three of the five series examined exhibit ex-post bias.