Debiased Machine Learning; establishment of valid causal inferences

The Debiased Machine Learning (DML) methods are only recently introduced in the field of econometrics. DML can be used with multiple Machine Learning (ML) methods for the estimation of a dataset. ML methods suffer from the regulation and the over-fitting bias when applied to causal inference problems. DML removes these biases through the use of the Neyman-orthogonal restriction and sample splitting. In this paper the causality between parameters established in three different papers, is investigated by applying the DML methods to the datasets. These papers used standard estimation methods to obtain their results. This paper shows that the causality between parameters established in two papers can be held unreliable. The results of this paper display that standard methods perform poor when applied to causal inference problems, especially when there is a large set of covariates used in the estimation of a regression. This displays the relevance of this paper and gives the researchers of the articles a reason to review their results.