2016/11/28 by Christian Hansen, Yuan Liao, Hansen, Christian +1
Economics, Econometrics and Finance · Mathematics · #Advanced Causal Inference Techniques #Econometrics (econ.EM) #Economic Policies and Impacts #FOS: Computer and information sciences #FOS: Economics and business #Methodology (stat.ME) #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.1611.09420
openalex publication_date 2016/11/28 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We consider inference about coefficients on a small number of variables of\ninterest in a linear panel data model with additive unobserved individual and\ntime specific effects and a large number of additional time-varying confounding\nvariables. We allow the number of these additional confounding variables to be\nlarger than the sample size, and suppose that, in addition to unrestricted time\nand individual specific effects, these confounding variables are generated by a\nsmall number of common factors and high-dimensional weakly-dependent\ndisturbances. We allow that both the factors and the disturbances are related\nto the outcome variable and other variables of interest. To make informative\ninference feasible, we impose that the contribution of the part of the\nconfounding variables not captured by time specific effects, individual\nspecific effects, or the common factors can be captured by a relatively small\nnumber of terms whose identities are unknown. Within this framework, we provide\na convenient computational algorithm based on factor extraction followed by\nlasso regression for inference about parameters of interest and show that the\nresulting procedure has good asymptotic properties. We also provide a simple\nk-step bootstrap procedure that may be used to construct inferential statements\nabout parameters of interest and prove its asymptotic validity. The proposed\nbootstrap may be of substantive independent interest outside of the present\ncontext as the proposed bootstrap may readily be adapted to other contexts\ninvolving inference after lasso variable selection and the proof of its\nvalidity requires some new technical arguments. We also provide simulation\nevidence about performance of our procedure and illustrate its use in two\nempirical applications.\n