Harvest efficiency in the production functions of fisheries should increase over time with the introduction of new technology and with increasing knowledge of the biology and behaviour of the target species. This implies that the estimated catchability coefficients of Gordon-Schaefer or Fox models should also increase over time. We estimated both models using annual catch and effort data for four lobster fisheries and compared the regression results for different time periods. We also estimated modified forms of the models with a time trend in the harvest equation. Both approaches show that the estimates of the catchability coefficients have decreased significantly over time in all four fisheries. We suggest that one important source of this anomaly is the use of annual aggregate catch and effort data. To test this hypothesis, we examined trends in the distribution of effort over the season, and their effect on annual catch per unit effort. The results show that methods based on aggregate annual catch and effort data can introduce spurious trends in annual catch per unit effort, and hence in catchability estimates. Estimation of production functions for bioeconomic modelling should use catch and effort data that is disaggregated both temporally and spatially.