As long as total effort cannot be completely controlled, a more thorough understanding of fishers' supply response decisions will be beneficial for fisheries managers. In this paper, we present a model of fishers' gear choice, which is empirically estimated on a panel of Swedish demersal trawlers. The approach allows for heterogeneity both in production technology and in risk preferences. Stochastic revenue functions with fixed effects are estimated and used to predict expected revenue and standard deviation for each trip. We employ a linear utility function in the mean-standard deviation framework and then analyze the gear choices, using the predicted values together with vessel capacity and lagged variables for the previous trip in a random parameters- logit model, which allow for heterogeneous preferences. The results indicate that fishers have a strong tendency to choose the same gear used on the previous trip, while in general they react to changes in economic and biological conditions by responding positively to increases in expected landing and negatively to increases in the variability of the expected landing values, indicating risk aversion.