We use the principal-agent model as a focal theoretical frame for synthesizing what we know, both theoretically and empirically, about the design and dynamics of the implementation of performance management systems in the public sector. In this context, we review the growing body of evidence about how performance measurement and incentive systems function in practice and how individuals and organizations respond and adapt to them over time, drawing primarily on examples from performance measurement systems in public education and social welfare programs. We also describe a dynamic framework for performance measurement systems that takes into account strategic behavior of individuals over time, learning about production functions and individual responses, accountability pressures, and the use of information about the relationship of measured performance to value added. Implications are discussed and recommendations derived for improving public sector performance measurement systems. (Contains 13 footnotes.)