The multitask-agency problem is examined empirically using contracts b
etween private, integrated oil companies and their service stations in
the city of Vancouver. The empirical tests assess how variations in t
he characteristics of one task affect the choice of agent-compensation
scheme for another. Comparative statics from the model predict that h
igher-powered incentives will be offered for gasoline sales when the s
econdary activity is not highly complementary with gasoline retailing,
where complementarity is measured by the cross-price demand effect, t
he covariation in uncertainty, and the degree of effort substitutabili
ty.