Sources of inadvertent bias in the stock assessment prioritization and review process
Will Satterthwaite, Southwest Fisheries Science Center
Tuesday October 4th, 1:00 PM PST
Fishery Management Councils and Science Centers are tasked with managing and assessing a large suite of stocks, too many to routinely assess them all. I have identified various examples in the Pacific Fishery Management Council process where it appears that assessments yielding low-status outcomes may be subjected to extra scrutiny during the review process and/or prioritized for re-asssessment on a faster timeline than assessments yielding high-status outcomes. Simple models suggest that even if done with the best of intentions, and even if there is some success at screening out assessments with errors in the direction of poor stock status, this has the potential to introduce bias into outcomes at the FMP-wide level. This has broad similarities to the reproducibility crisis in science and instances of “p-hacking”. Simple numeric simulations suggest that this can lead to higher probabilities of overfishing than implied by the Council’s choice of P*. It may be possible to modify the P*/sigma framework to account for this.