Non-probabilistic surveys and sampling in the human dimensions of fisheries

Abstract

Fisheries management and conservation require consideration of fish, habitat, and people. In fisheries science, a growing body of research on human values, perspectives, and behaviours around fish—known as ‘human dimensions’ research—has emerged from the realization that management and conservation require a better understanding of people. Surveys are a common and versatile tool in human dimensions research, but not all surveys are equal. Large-scale, probabilistic surveys draw random samples from known populations (e.g., all license-holding recreational fishers in a jurisdiction) and represent the ‘gold standard’ in survey research. However, these surveys may fall short of this standard for various reasons. Surveys using non-probabilistic sampling are also common in human dimensions research. Non-probabilistic surveys are attractive to researchers facing time, cost, and other constraints, but differ notably from their probabilistic counterparts: data from non-probabilistic samples are typically unfit for population estimates and other inferences due their uncertain representativeness. Nonetheless, a wealth of research with non-probabilistic data within and outside of fisheries (e.g., in health sciences) suggests that these methods have valid applications and advantages in some contexts. We reviewed the literature on non-probabilistic surveys and sampling in the human dimensions of fisheries, and explored seminal literature from other thematic areas where such methods are common, to better understand their strengths, weaknesses, and applications relative to probabilistic methods. Here, we describe (1) how researchers have used non-probabilistic methods to study the human dimensions of fisheries, (2) how mismatching research questions, objectives, and methods can produce ‘awkward surveys,’ and (3) how researchers can use non-probabilistic surveys in ways that invoke their methodological strengths. While uncertain representativeness may limit the utility of non-probabilistic data in some contexts, non-probabilistic methods are time- and cost-effective, and have distinct advantages in studies of niche groups and phenomena, emergent or understudied phenomena, and in supplementary roles.