Quantifying Noise

Expectations affect economic decisions, and therefore inaccurate expectations are costly. Expectations can be wrong in ways that are systematic (bias) or unsystematic (noise). We provide a general method for quantifying the noise component. The method is based on the insight that theoretical models of expectation formation predict a factor structure for individual expectations. This insight leads to a widely applicable factor-based measurement procedure. Using data from professional forecasters, we find that noise is large and pervasive. Our findings have implications for forecast combination, macro models with incomplete information, and empirical research using micro data on expectations