Estimation of Commercial Fishing Trip Costs Using Sea Sampling Data

Abstract

When estimating commercial fishing costs, selection bias can impact any model derived from non-census sampling methodologies. In the northeastern United States, commercial fishing operating cost models may suffer from selection bias, as they are often estimated using data collected for biological, rather than economic, purposes. We investigate the effects of sampling bias on trip cost model estimations using weighted/unweighted least squares and Heckman sample selection models. Results suggest that (1) the propensity for a trip to carry an observer is not random with respect to costs and that (2) selection bias exists in the majority of cost models investigated. To gauge the magnitude of selection bias, we compare results of the unweighted least squares and Heckman models. The differences between models can lead to erroneous conclusions at the subfleet level and in estimating trip cost maxima. Results suggest that assessing and correcting for selection bias is necessary when using sampled fishing cost data.