Random versus self-selected samples of Arkansas anglers: An empirical comparison

Objective: Different sampling designs can influence survey results and potentially lead to management decisions that are deemed unacceptable by anglers. This study compared survey responses from a random sample of licensed Arkansas anglers with responses from a nonrandom self-selected sample of individuals who voluntarily completed the same survey on the Arkansas Game and Fish Commission (AGFC) social media platforms. Methods: Both instruments were online surveys, completed by Arkansas resident license holders. The independent variable was the type of sample (i.e., random vs. nonrandom self-selected). We compared angler responses to questions related to angler demographics, fishing behaviors, centrality of fishing, fishing satisfaction, and use of AGFC information resources. The random sample and nonrandom self-selected sample were compared relative to (1) response rates, (2) item nonresponse, and (3) differences in response distributions. Results: Similar to other recent research, the response rate for the random sample was 13% (n=1,312). A response rate for the self-selected sample (n=1,105) could not be calculated because the population of visitors to the AGFC social media platforms could not be determined. The random sample consistently had higher item nonresponse than the self-selected sample. For example, for the nine satisfaction questions, the random sample respondents had 30% missing, whereas the self-selected respondents had 11% missing. The response distribution results indicated that the self-selected respondents were more avid anglers than those in the random sample. For example, the self-selected anglers fished more often (within and among years), fished a greater variety of locations (e.g., out of state), and placed a greater importance on fishing. The self-selected sample was also older, more likely male, and more likely to receive information from all AGFC sources than the random sample. Those in the self-selected sample were generally less satisfied with their fishing experiences in Arkansas (e.g., overall, number of fish caught) than the random sample. Conclusions: Of the 30 comparisons, the two samples differed statistically on 27 tests (90%). The results highlight important differences between the two methods. Both methods have trade-offs in costs and biases. If the goal of the research is to understand the general angler population, a survey of randomly selected anglers is recommended. On the other hand, if the research focuses on the beliefs and behaviors of avid anglers, the self-selected anglers may be acceptable.