We study differently framed incentives in dynamic laboratory buyerseller relationships with multi-tasking and endogenous matching. The experimental design tries to mitigate the role of social preferences and intrinsic motivation. Absent explicit incentives, effort is low in both tasks. Their introduction boosts efficiency substantially increasing effort in the contractible task, mildly crowding it out in the non-contractible one, and increasing buyer surplus. Bonuses and penalties are equivalent for efficiency and crowding-out, but different in distributional effects: sellers’ surplus increases with bonuses as buyers’ offers become more generous. Buyers tend to prefer penalties, which may explain why they are dominant in procurement.