The paper introduces the Covariate Balancing Propensity Score (CBPS) methodology, which models treatment assignment while optimizing covariate balance. The CBPS leverages the dual characteristics of the propensity score as both a covariate balancing score and a conditional probability of treatment assignment. The estimation of the CBPS is conducted within the generalized method-of-moments (GMM) or empirical likelihood (EL) framework. The CBPS significantly improves the empirical performance of propensity score matching and weighting methods, which are often sensitive to slight misspecification of the propensity score model. The CBPS can be extended to handle non-binary treatments and generalize experimental estimates to a target population. Open-source software is available for implementing the CBPS. The paper also discusses related methods and provides simulation and empirical studies to demonstrate the effectiveness of the CBPS.The paper introduces the Covariate Balancing Propensity Score (CBPS) methodology, which models treatment assignment while optimizing covariate balance. The CBPS leverages the dual characteristics of the propensity score as both a covariate balancing score and a conditional probability of treatment assignment. The estimation of the CBPS is conducted within the generalized method-of-moments (GMM) or empirical likelihood (EL) framework. The CBPS significantly improves the empirical performance of propensity score matching and weighting methods, which are often sensitive to slight misspecification of the propensity score model. The CBPS can be extended to handle non-binary treatments and generalize experimental estimates to a target population. Open-source software is available for implementing the CBPS. The paper also discusses related methods and provides simulation and empirical studies to demonstrate the effectiveness of the CBPS.