This article presents a principal-agent model in the context of public schools to help explain the use of merit pay for teachers. The model considers how both the nature of teaching and the political costs of union resistance affect school district merit pay decisions. Our results support the idea that merit pay is more likely in environments where there is more performance information and less likely where teachers are unionized. The negative effect from unions, however, appears stronger than the positive effect of performance information. We also find that teachers in merit pay districts earn more than their counterparts in non-merit pay districts. (Contains 5 tables and 25 footnotes.)