Inferring maximum lifespan from maximum recorded longevity in the wild carries substantial risk of estimation bias

When comparing lifespan (longevity) between species, it is common practice to take the maximum recorded longevity value within each species as a proxy of maximum lifespan. Whether maximum recorded longevity is a reliable proxy of species' maximum longevity remains unclear. Some researchers correct for previously documented life history correlates of maximum recorded longevity before analysing new predictors of lifespan across species in the context of their current, specific hypotheses. At present there is no certainty that all relevant statistical, phenotypic, or ecological biases are accounted for by such corrective measures. Here, we employ Monte Carlo simulation to investigate the effect of differences in recapture numbers, recapture types (the point in life at which individuals are initially captured or recaptured), and actuarial population decay structure of simulated species on their maximum recorded longevities. We show that maximum recorded longevities differ in response to all three of these variables, as well as all of their two- and three-way interactions. We then investigate empirical avian band-recapture data for evidence of biases caused by recapture number and recapture type, predicted by the Monte Carlo analysis, and confirm the predicted biases as major sources of variance. Finally, we investigate the relationship between recapture type, recapture number, and a selection of ecological and life-history variables previously documented to correlate with maximum recorded longevity, and find significant correlations between the biasing variables and those published correlates. Our results call into question the validity of using maximum recorded longevity as a proxy for different species' maximum longevities in comparative studies investigating the evolution of lifespan.