Bias, rationality and asymmetric loss functions

•Survey forecasts are often found to be biased at the individual level.•Raises question if they are irrational or have asymmetric loss functions.•I show that the finding is largely due to the pattern of missing observations.

In the literature, it is a common empirical finding that survey based expectations are biased at the individual level. This has sparked a large debate if forecasters have asymmetric loss functions or the rationality assumption is violated. In this paper, I will show that the bias can in large part be explained by the pattern of missing observations in the survey. Thus the assumption of asymmetric loss functions is not required to satisfy the rationality assumption.