Abstract One of the criteria used by the International Union for Conservation of Nature (IUCN) to assess threat status is the rate of decline in abundance over 3 generations or 10 years, whichever is longer. The traditional method for calculating generation length ( T ) uses age‐specific survival and fecundity, but these data are rarely available. Consequently, proxies that require less information are often used, which introduces potential biases. The IUCN recommends 2 proxies based on adult mortality rate, = α + 1/ d , and reproductive life span, = α + z * RL, where α is age at first reproduction, d is adult mortality rate, RL is reproductive life span, and z is a coefficient derived from data for comparable species. We used published life tables for 78 animal and plant populations to evaluate precision and bias of these proxies by comparing and with true generation length. Mean error rates in estimating T were 31% for and 20% for , but error rates for were 16% when we subtracted 1 year ( ), as suggested by theory; also provided largely unbiased estimates regardless of the true generation length. Performance of depends on compilation of detailed data for comparable species, but our results suggest taxonomy is not a reliable indicator of comparability. All 3 proxies depend heavily on a reliable estimate of age at first reproduction, as we illustrated with 2 test species. The relatively large mean errors for all proxies emphasized the importance of collecting the detailed life‐history information necessary to calculate true generation length. Unfortunately, publication of such data is less common than it was decades ago. We identified generic patterns of age‐specific change in vital rates that can be used to predict expected patterns of bias from applying .