1: \begin{abstract}

2: Weighting methods in causal inference have been widely used to achieve a desirable level of covariate balancing. However, the existing weighting methods have desirable theoretical properties

3: only when a certain model, either the propensity score or outcome regression model,

4: is correctly specified. In addition, the corresponding estimators

5: do not behave well for finite samples due to large variance even when the model is correctly specified.

6: In this paper, we consider to use the

7: integral probability metric (IPM), which is a metric between two probability measures, for covariate balancing. Optimal weights are determined so that weighted empirical distributions for the treated and control groups have the smallest IPM value for a given set of discriminators. 

8: %We look for the weights such that the two weighted empirical distributions for the treated and control groups have a small values of the IPM. 

9: We prove that the corresponding estimator can be consistent without correctly specifying any model (neither the propensity score nor the outcome regression model).

10: In addition, we empirically show that our proposed method outperforms

11: existing weighting methods with large margins for finite samples. 

12:   

13: %We also show through simulation studies that our proposed method performs the best in almost every simulation scenario.

14: 

15: \end{abstract}

16: