Double/Debiased/Neyman Machine Learning of Treatment Effects

By Victor Chernozhukov, D. Chetverikov, Mert Demirer, Esther Duflo, Christian Hansen, Whitney K. Newey. 2017

415 citations recorded by the source corpus.

Chernozhukov et al. (2016) provide a generic double/de-biased machine learning (ML) approach for obtaining valid inferential statements about focal parameters, using Neyman-orthogonal scores and cross-fitting, in settings where nuisance parameters are estimated using ML methods. In this note, we illustrate the application of this method in the context of estimating average treatment effects and average treatment effects on the treated using observational data.

Subfields: Statistics and Probability, Causal Inference & Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

In Causal Inference & Discovery

Citations in the corpus

Using Wasserstein Generative Adversarial Networks for the design of Monte Carlo simulations (2019, neutral): “[1999], Chernozhukov et al. [2017]:”
Estimating dynamic treatment effects in event studies with heterogeneous treatment effects (2018, methodology): “The above analysis contributes to a recent literature analyzing linear two-way fixed effects regressions (de Chaisemartin and D’Haultfœuille, 2018; Imai and Kim, 2017). We show the similar result that the FE estimand is a linear combination of average treatment effects in each cohort, however due to”
Locally Robust Semiparametric Estimation (2016, background): “been constructed by Robins, Rotnitzky, and Zhao (1994, 1995), Robins and Rotnitzky (1995), Scharfstein, Rotnitzky, and Robins (1999), Robins, Rotnitzky, and van der Laan (2000), Robins and Rotnitzky (2001), Graham (2011), and Firpo and Rothe (2017).”
Machine Learning Advances for Time Series Forecasting (2020, background): “Recently, Chernozhukov et al. (2017, 2018) consider the case where the number of covariates diverge as the sample size increases in a very general setup.”
Poorly Measured Confounders are More Useful on the Left than on the Right (2017, background): “One recent strand of this literature is interested in the selection of control variables in a causal regression and inference for the parameter of interest after such an initial variable selection step (Belloni, Chernozhukov, and Hansen, 2014a,b; Chernozhukov et al. 2017; Chernozhukov et al. 2018).”
Predictably Unequal? The Effects of Machine Learning on Credit Markets (2020, neutral): “Academic economists also increasingly rely on such techniques (e.g., Belloni et al., 2014; Varian, 2014; Kleinberg et al., 2017; Mullainathan and Spiess, 2017; Chernozhukov et al., 2017; Athey and Imbens, 2017).”
Incremental causal effects (2019, background): “For general functionals the orthogonalization step can be more involved, see for example Chernozhukov et al. (2018). The intuition behind the orthogonalization step is as follows: If we define the transformed functions b̃k = bk(t, x) − tαk with αk = 1 n ∑n i=1 ∂tbk(ti, xi) for k > 1 and b̃k = t for ”
Estimation of Conditional Average Treatment Effects With High-Dimensional Data (2019, methodology): “We use the local linear smoother studied by Fan (1992, 1993) and Fan and Gijbels (1992) to estimate τ0(x1) in the second stage. While it is possible to extend our results to local polynomial estimators with an extra degree of smoothness, we focus on the linear case for simplicity.8 Our second-stage ”