Sample selection bias in fisheries technical efficiency studies using stochastic frontiers; presence and correction for an artisanal fishery in Northwest Spain

Technical efficiency (TE) is a key topic in fishery economics due to its relevance in guiding fishing regulations and policy measures. In some territories, the presence of artisanal and small-scale fisheries (SSFs) is clearly decreasing but is still substantial, and regulators are concerned about their productivity and TE. A distinguishing feature of SSFs when compared to industrial fishing is that multi-gear artisanal vessels do not participate in certain fisheries: climate and other technical and non-technical determinants condition the decision to take part in a fishery. Consequently, data gathered from samples of multi-gear artisanal vessels may not be purely random (sample selection), leading to biased estimations and erroneous conclusions for management purposes. In contrast to other fields of economic and social analysis, this bias has not been generally considered when studying TE and productivity in artisanal fisheries and SSFs. With the aim to improve TE modelling in fisheries and propose a correction, this article tackles sample selection bias, its sources and its consequences on stochastic frontiers (SF)-based TE studies on fishing. To this end, as has been applied in non-fishery studies, an alternative method is proposed to test and, when applicable, correct an SF function affected by selection bias and that takes into account the estimates resulting from a connected participation probability model. An empirical exercise was conducted to estimate TE for an artisanal octopus fishery in Asturias (NW Spain) for the 2008–2009 season. The results show that participant vessels had undergone self-selection and that the SF adjusted for sample-selection produced marked differences in the estimates of the contributions of vessel length and fishing effort to productivity, together with smaller technical efficiency scores. In terms of policy implications, these results led to the identification of improved on-board conditions for crew members and a market-based policy of eco-labelling as preferable sustainable measures, among others, and allowed the most efficient local fleets to be correctly identified to apply for an eco-label. The method described herein offers an alternative when non-technical determinants (fishermen's attributes) are not subject to analysis, and it can be extended to other TE studies dealing with fisheries that use multi-gear artisanal and small-scale vessels. [Display omitted] •Sample selection bias has not been explored in fisherie