Current reforms in the public sector are characterized by the introduction of businesslike incentive structures, in particular the introduction of “pay for performance” schemes in public institutions. However, the public sector has some specific characteristics, which might restrict the naive adoption of pay for performance. Our article analyzes whether the impact of pay for performance on performance is bound to conditions, and if this is the case, under which conditions pay for performance has a positive or a negative effect on performance. We explore this contingency in a meta-analytic review of previous experimental studies on the effects of pay for performance on performance. We further show why pay for performance sometimes negatively affects personal efforts. With an experimental vignette study we demonstrate (a) that motivation is likely to be a key influence on the effect of performance-related pay on performance, and (b) that pay for performance is generally more costly as it appears because it almost always produces hidden costs of rewards. Our findings help to explain the modest success of pay for performance in the public sector.