Site-selection bias can drive apparent population declines in long-term studies

Detecting population declines is a critical task for conservation biology. The spatiotemporal variability of populations, along with logistical difficulties in population estimation, makes this task difficult. Here we call attention to a possible bias in estimates of population decline: when study sites are chosen based on abundance of the focal species, for statistical reasons apparent declines are likely even without an underlying population trend. This “site-selection bias” has been mentioned in the literature but is not well known. We show using simulated and real population data that when site-selection biases are introduced, they have substantial impact on inferences about population trends. We use a left-censoring method to show patterns consistent with the operation of the site-selection bias in real population studies. The site-selection bias is, thus, an important consideration for conservation biologists, and we offer suggestions for minimizing or mitigating it in study design and analysis.