As performance-based contracting in social welfare services continues to expand, concerns about potential unintended effects are also growing. We analyze the incentive effects of high-powered, performance-based contracts and their implications for program outcomes using panel data on Dutch cohorts of unemployed and disabled workers that were assigned to private social welfare providers in 2002 to 2005. We employ a difference-in-differences design that takes advantage of the fact that contracts gradually moved from partial performance-contingent pay to full (100 percent) performance-contingent contracting schemes. We develop explicit measures of selection into the programs and find evidence of cream skimming and other gaming activities on the part of providers, but little impact of these activities on program outcomes. Moving to a system with contract payments fully contingent on performance appears to increase job placements, but not job duration, for more readily employable workers.