We compare turnout under proportional power-sharing electoral systems and winner-take-all elections. The effect of such institutional differences on turnout depends on the distribution of voter preferences. If the two parties have relatively equal support, turnout is higher in a winner-take-all system; the result is reversed when there is a clear underdog. We report findings from a laboratory experiment that was designed and conducted to explore this theoretical hypothesis and several other secondary hypotheses that are also implied by the theoretical model. The results are broadly supportive of the theoretical predictions on comparative turnout, the partial underdog compensation effect and the competition effect.