Life history theory makes two contrasting predictions about the population consequences of variation in age structure: over long time periods life history diversity should spread risk across different segments of a population, thereby buffering it from environmental variation, but at smaller time scales simple life histories should grow faster than complex life histories because of heterogeneity in vital rates associated with a diverse population.  While these predictions have long been accepted on a theoretical level, empirical tests of these ideas are almost nonexistent because they require large data sets of many generations and population data with significant life history variation.  We tested the hypothesis that life history variation is related with population productivity over long but not short time periods in nine Bristol Bay sockeye salmon (Oncorhynchus nerka) populations that have been monitored since the 1950s.  Our findings indicate that population growth rates were negatively correlated with life history diversity over short time scales (one year), but were positively correlated at longer time scales (10 to 40 years).  Using simulations, we found that annual variation and interpopulation differences in the distribution of life history diversity were insufficient to explain the empirical trend.  These empirical findings provide important insight into the role of life history diversity in populations by linking two measures which have sometimes been treated as separate aspects of population status, and suggest that expanding the life history portfolio of small populations may be a crucial element for recovery.