Scientists disagree about the nature of biodiversity change. While there is evidence for widespread declines from population surveys, assemblage surveys reveal a mix of declines and increases. These conflicting conclusions may be caused by the use of different metrics: assemblage metrics may average out drastic changes in individual populations. Alternatively, differences may arise from data sources: populations monitored individually, versus whole-assemblage monitoring. To test these hypotheses, we estimated population change metrics using assemblage data. For a set of 23 241 populations, 16 009 species, in 158 assemblages, we detected significantly accelerating extinction and colonisation rates, with both rates being approximately balanced. Most populations (85%) did not show significant trends in abundance, and those that did were balanced between winners (8%) and losers (7%). Thus, population metrics estimated with assemblage data are commensurate with assemblage metrics and reveal sustained and increasing species turnover.