This paper examines the impact of the teacher certification program in Indonesia on student and teacher outcomes. I use rule-based two-sample instrumental variables strategy. The thresholds that lead to discontinuities are determined empirically using the universe of Indonesian teachers. The discontinuities are generated by the rule of assigning teachers into certification program. Teachers are ranked based on several criteria and those whose ranks are below the threshold are more likely to be certified. These thresholds are not officially reported and therefore have to be empirically determined from the data. In the first stage, estimated in an Indonesian teacher population, having rank below the estimated thresholds are significant predictor for a teacher’s certification status. In the second stage, I find that teachers who were predicted to be certified do not increase student learning outcomes as measured by standardized tests and there is no evidence that these teachers perform better in teacher’s tests, have no side jobs, and have lower probability of being absent last week. The estimated effects are informative for those teachers whose certification status can be predicted by the instruments. This limits the generalizability of the study to the teacher population at large. Field: development.