Temporary deviations of trade prices from fundamental values impart bias to estimates of mean returns to individual securities, to differences in mean returns across portfolios, and to parameters estimated in return regressions. We consider a number of corrections, and show them to be effective under reasonable assumptions. In an application to the Center for Research in Security Prices monthly returns, the corrections indicate significant biases in uncorrected return premium estimates associated with an array of firm characteristics. The bias can be large in economic terms, for example, equal to 50% or more of the corrected estimate for firm size and share price.

We study the Lucas asset pricing model in a controlled setting. Participants trade two long‐lived securities in a continuous open‐book system. The experimental design emulates the stationary, infinite‐horizon setting of the model and incentivizes participants to smooth consumption across periods. Consistent with the model, prices align with consumption betas and comove with aggregate dividends, particularly so when risk premia are higher. Trading significantly increases consumption smoothing compared to autarky. Nevertheless, as in field markets, prices are excessively volatile. The noise corrupts traditional generalized method of moment tests. Choices display substantial heterogeneity, with no subject representative for pricing.