AbstractA model is developed that attempts to explain the historical size of the US equity premium by distinguishing between gross and net returns accruing to agents. The model derived by Mehra and Prescott (1985) is augmented with a bid–ask spread, calibrated and simulated. Equity premia in the order of 3–4% are generated for plausible values of the transactions parameters. This contrasts with Mehra and Prescott, who find a maximum equity premium of 0.4% while the historic equity premium has been about 6.2%. Estimates of the bid–ask spread are obtained using GMM and tests of the overidentifying restrictions are not rejected for several lists of instrumental vari