Research
Working Papers
“Cross-Fitting-Free Debiased Machine Learning with Multiway Dependence”
with Harold D. Chiang
first circulation - 2026 - arXiv:2602.11333
arXiv
Abstract
TThis paper develops an asymptotic theory for two-step debiased machine learning (DML) estimators in generalised method of moments (GMM) models with general multiway clustered dependence, without relying on cross-fitting. While cross-fitting is commonly employed, it can be statistically inefficient and computationally burdensome when first-stage learners are complex and the effective sample size is governed by the number of independent clusters. We show that valid inference can be achieved without sample splitting by combining Neyman-orthogonal moment conditions with a localisation-based empirical process approach, allowing for an arbitrary number of clustering dimensions. The resulting debiased GMM estimators are shown to be asymptotically linear and asymptotically normal under multiway clustered dependence. A central technical contribution of the paper is the derivation of novel global and local maximal inequalities for general classes of functions of sums of separately exchangeable arrays, which underpin our theoretical arguments and are of independent interest.“Inference in High-Dimensional Panel Models: Two-Way Dependence and Unobserved Heterogeneity”
Best Student Paper Award at 2024 Midwest Econometrics Group Meeting
first circulation - 2025 - arXiv:2504.18772
arXiv latest version code & replication
Abstract
Panel data allows for the modeling of unobserved heterogeneity, significantly raising the number of nuisance parameters and making high dimensionality a practical issue. Meanwhile, temporal and cross-sectional dependence in panel data further complicates high-dimensional estimation and inference. This paper proposes a toolkit for high-dimensional panel models with large cross-sectional and time sample sizes. To reduce the dimensionality, I propose a variant of LASSO for two-way clustered panels. While being consistent, the convergence rate of LASSO is slow due to the cluster dependence, rendering inference challenging in general. Nevertheless, asymptotic normality can be established in a semiparametric moment-restriction model by leveraging a clustered-panel cross-fitting approach and, as a special case, in a partial linear model using the full sample. In an exercise of estimating multiplier using panel data, I demonstrate how high dimensionality could be hidden and the proposed toolkit enables flexible modeling and robust inference.“Identification of Average Responses with Endogenous Controls”
with Kyoo il Kim
first circulation - 2024 - arXiv:2401.14395
arXiv code & replication
Abstract
Control variables are routinely treated as exogenous, yet in many empirical settings they are themselves endogenous. This creates a dilemma: omitting controls may leave the treatment endogenous, while including them may contaminate identification. The problem is not resolved by instrumental variables when they are only conditionally valid. We show that average responses to the treatment remain identified under a rank condition called measurable separability, which accommodates endogenous controls. For parametric models, our approach amounts to estimating a nonparametric model that nests the parametric specification. For nonparametric models, our results imply that endogenous controls are generally innocuous under standard identification conditions, except in the presence of "bad controls". We further propose a test for endogenous controls. Simulation results and an empirical application demonstrate this prevalent issue and provide practical implications of our methods.“Another Look at the Linear Probability Model and Nonlinear Index Models”
with Robert S. Martin and Jeffrey M. Wooldridge (Revise & Resubmit at Econometric Reviews)
first circulation - 2023 - arXiv:2308.15338
arXiv latest version code & replication
Abstract
We reassess the linear probability model (LPM) for binary responses from the perspective of average partial effects (APEs). Under certain conditions, the LPM recovers APEs even when the true response probability is nonlinear. Outside these scenarios, we demonstrate in simulations that having a large fraction of fitted values in [0, 1] is neither necessary nor sufficient for OLS to approximate the APEs well, which contradicts the common rationale. We further investigate and refine an ad hoc trimming procedure to reduce bias for LPM, and establish its statistical properties by relating it to a numerically equivalent nonlinear least squares estimator.Publication
“Fixed-b Asymptotics for Panel Models with Two-Way Clustering”
with Timothy J. Vogelsang (Journal of Econometrics, 2024, 244(1): 105831)
first circulation - 2023 - arXiv:2309.08707
arXiv published version
STATA command: xtregtfb. Installation: type net from https://kaichengchen.github.io/statafile/ in STATA