Experiments suggest that increasing species’ diversity increases productivity over short observation periods. Theory predicts that diversity may increases productivity at equilibrium, such as in communities of coexisting competitors. In effect, experiment and theory are using different response variables. We illustrate how this gap can obscure some of the mechanisms through which diversity shapes productivity. To do this, we propose a formal link to dynamic models of population growth. Using this link, we show that selection (as it is formally defined in dynamic models) does not affect productivity. Instead, selection changes productivity over time. In contrast, niche complementarity can change productivity directly. We illustrate how to measure each of the terms in competition experiments. These results highlight how the effects of diversity on productivity can be obscured by our choice of response variable.