This paper discusses recent developments in econometrics that are crucial for empirical researchers working on policy evaluation questions. The authors focus on three main areas: identification strategies in program evaluation, supplementary analyses to enhance credibility, and advances in machine learning methods for causal effects. They highlight specific techniques such as synthetic control methods, regression discontinuity, external validity, and the causal interpretation of regression methods. The paper also emphasizes the importance of supplementary analyses, including placebo analyses and sensitivity and robustness analyses, to support the credibility of primary analyses. Additionally, it reviews recent advances in machine learning, particularly in handling high-dimensional data and estimating heterogeneous treatment effects. The authors provide recommendations for applied work, emphasizing the use of local linear or local quadratic methods in regression discontinuity designs, the importance of bandwidth selection, and the need for supplementary analyses to assess the credibility of the design. They also discuss the synthetic control approach and nonlinear difference-in-difference models, and the extension of methods for estimating average treatment effects under unconfoundedness to settings with multivalued treatments. The paper concludes by highlighting the ongoing research in causal effects in networks and social interactions.This paper discusses recent developments in econometrics that are crucial for empirical researchers working on policy evaluation questions. The authors focus on three main areas: identification strategies in program evaluation, supplementary analyses to enhance credibility, and advances in machine learning methods for causal effects. They highlight specific techniques such as synthetic control methods, regression discontinuity, external validity, and the causal interpretation of regression methods. The paper also emphasizes the importance of supplementary analyses, including placebo analyses and sensitivity and robustness analyses, to support the credibility of primary analyses. Additionally, it reviews recent advances in machine learning, particularly in handling high-dimensional data and estimating heterogeneous treatment effects. The authors provide recommendations for applied work, emphasizing the use of local linear or local quadratic methods in regression discontinuity designs, the importance of bandwidth selection, and the need for supplementary analyses to assess the credibility of the design. They also discuss the synthetic control approach and nonlinear difference-in-difference models, and the extension of methods for estimating average treatment effects under unconfoundedness to settings with multivalued treatments. The paper concludes by highlighting the ongoing research in causal effects in networks and social interactions.