Abstract
Biodiversity‐ecosystem functioning experiments have established that species richness and composition are both important determinants of ecosystem function in an experimental context. Determining whether this result holds for real‐world ecosystem services has remained elusive, however, largely due to the lack of analytical methods appropriate for large‐scale, associational data. Here, we use a novel analytical approach, the Price equation, to partition the contribution to ecosystem services made by species richness, composition and abundance in four large‐scale data sets on crop pollination by native bees. We found that abundance fluctuations of dominant species drove ecosystem service delivery, whereas richness changes were relatively unimportant because they primarily involved rare species that contributed little to function. Thus, the mechanism behind our results was the skewed species‐abundance distribution. Our finding that a few common species, not species richness, drive ecosystem service delivery could have broad generality given the ubiquity of skewed species‐abundance distributions in nature.