The paper introduces a new method called "Coarsened Exact Matching" (CEM) for improving causal inferences in observational studies. CEM is derived from the "Monotonic Imbalance Bounding" (MIB) class of matching methods, which guarantees that the imbalance between treated and control groups will not exceed a user-chosen level. Unlike other matching methods, CEM does not require assumptions about the data generation process and is easy to understand and use. The authors derive and illustrate several desirable statistical properties of CEM, including its ability to bound model dependence, reduce estimation error, and improve balance. They also propose extensions to make CEM more widely applicable and provide open-source software for R, Stata, and SPSS. The paper discusses the theoretical properties of CEM and its practical applications, emphasizing the connection between these aspects.The paper introduces a new method called "Coarsened Exact Matching" (CEM) for improving causal inferences in observational studies. CEM is derived from the "Monotonic Imbalance Bounding" (MIB) class of matching methods, which guarantees that the imbalance between treated and control groups will not exceed a user-chosen level. Unlike other matching methods, CEM does not require assumptions about the data generation process and is easy to understand and use. The authors derive and illustrate several desirable statistical properties of CEM, including its ability to bound model dependence, reduce estimation error, and improve balance. They also propose extensions to make CEM more widely applicable and provide open-source software for R, Stata, and SPSS. The paper discusses the theoretical properties of CEM and its practical applications, emphasizing the connection between these aspects.