Investigating record linkage for combining voluntary catch reports with a probability sample

Abstract

Incorporating multiple samples for estimation is becoming increasingly popular across industries and disciplines. Indeed, various fisheries agencies combine samples when estimating catch totals. In this paper, we examine a probabilistic matching technique to link a non-probability sample of electronic reports of fishing trips with a dockside probability sample. We use 2017 data from a capture-recapture pilot study in the Gulf of Mexico and compare it with current methodologies. We also examine a key parameter used in record linkage—a threshold score for the algorithm—and study its effect in this setting. We show that combining non-probability data with a probability sample is useful for estimating fish catch and argue record linkage is a useful way to link the data.