Performance pay in general amounts to only a small fraction of total pay. In this paper, we show that performance pay is nevertheless important for level and dynamics of wages over the life cycle because of the incentives it indirectly provides for human capital acquisition and because of its impact on the variability of total pay. We articulate this argument in the context of a model that combines three key mechanisms for wage growth and dispersion: employer learning about workers’ ability, human capital acquisition, and performance incentives. We use this model to account for the experience profile of wages, their dispersion, and their composition in terms of fixed and variable (performance) pay. The model admits an analytical decomposition of performance pay into four terms that capture (i) the trade-off between risk and incentives characteristic of settings of moral hazard; (ii) the insurance that firms provide against the wage risk due to the uncertainty about ability; (iii) incentives for effort arising from this uncertainty (career concerns); and (iv) incentives for effort generated by the prospect of human capital acquisition. We prove the model is identified under standard assumptions. Despite its parsimony, the model fits the data very well, including the empirical finding that performance pay as a share of total pay first increases and then decreases with experience. This feature of performance pay, which we are the first to document, runs contrary to the prediction of standard models of performance incentives that the ratio of performance pay to total pay increases with experience, especially at the end of the life cycle. Our estimates imply that effort to produce output augments human capital. Also, human capital acquisition and insurance against uncertainty about ability are quantitatively the main determinants of performance pay. Career-concerns incentives, on which the theoretical literature has focused, and the strength of the contemporaneous trade-off between risk and incentives—the primary determinant of variable pay in static moral-hazard models—are instead much less relevant. Importantly, we find that through the cumulative impact of effort on the job on human capital acquisition and the contribution of variable pay to the variance of total pay, performance incentives are a crucial source of wage growth and dispersion over the life cycle.