Research

Job market paper

Cross-Validation for Ridge Regularization in Debiased Machine Learning

2026 · Latest Draft

This paper studies cross-validation in debiased machine learning (DML) for linear functionals of a regression function γ0, where γ0 and a debiasing nuisance α0 are estimated with ridge regularization. To ensure valid inference, DML places a joint rate requirement on the estimation errors for γ0 and α0. The feasibility of attaining the required rate depends on the joint complexity of the two. We study cross-validation for ridge-regularized estimators of γ0 and α0 on a shared, ill-conditioned dictionary of growing size. Based on our learning-rate bounds, we characterize the boundary at which the rate condition is achievable in terms of how fast the dictionary can approximate the nuisances. We then propose theory-driven feasible candidate sets for both the dictionary dimensions and the ridge penalties, and formally justify the use of separate cross-validation in DML based on these sets. In particular, we develop an oracle inequality for the resulting nuisance estimation risk. Our simulation results confirm these findings.

Working papers

Debiased Machine Learning: Identification, Estimation, and Shape Constraints

With Qihui Chen and Zheng Fang.

2026 · arXiv

We develop a general framework of identification and estimation for automatic debiased machine learning (DML) where the parameter of interest θ0 is identified by a moment condition involving a nuisance γ0 that may be high dimensional. We establish conditions under which the Riesz representer α0, which is at the core of DML, is identified, and show that the identification occurs precisely when α0 uniquely optimizes a quadratic functional. This characterization enables us to develop a general estimation procedure for α0 that allows for generic γ0 including those defined by models with endogeneity and encompasses both classical sieves and modern architectures such as deep neural networks. To improve estimation precision and mitigate the curse of dimensionality, we incorporate shape constraints on γ0 by embedding them into a possibly nonlinear parameter space. We illustrate our estimation procedure through simulations and empirical applications.

Work in progress

Intergenerational Mobility Through the Lens of φ-Divergence

With Esfandiar Maasoumi.

Doubly Robust Bounds for Network Spillovers under Sample Selection

With Pablo Estrada.

Robust Inference on Variance under First-Order Degeneracy

With Alejandro Sánchez-Becerra.

Distributional Difference in Differences

With Wonjun Choi and Florian Gunsilius.

Publications

Survey-Based Forecasting: To Average or Not to Average

With Naijing Huang and Zhentao Shi · Behavioral Predictive Modeling in Economics, Studies in Computational Intelligence, vol. 897, Springer, 2021, pp. 87–104

Forecasting inflation rate is of tremendous importance for firms, consumers, as well as monetary policy makers. Besides macroeconomic indicators, professional surveys deliver experts’ expectation and perception of the future movements of the price level. This research studies survey-based inflation forecast in an extended recent sample covering the Great Recession and its aftermath. Traditional methods extract the central tendency in mean or median and use it as a predictor in a simple linear model. Among the three widely cited surveys, we confirm the superior forecasting capability of the Survey of Professional Forecasters (SPF). While each survey consists of many individual experts, we utilize machine learning methods to aggregate the individual information. In addition to the off-the-shelf machine leaning algorithms such as the Lasso, the random forest and the gradient boosting machine (GBM), we tailor the standard Lasso by differentiating the penalty level according to an expert’s experience, in order to handle for participants’ frequent entries and exits in surveys. The tailored Lasso delivers strong empirical results in the SPF and beats all other methods except for the overall best performer, GBM. Combining forecasts of the tailored Lasso model and GBM further achieves the most accurate inflation forecast in both the SPF and the Livingston Survey, which beyonds the reach of a single machine learning algorithm. We conclude that combination of machine learning forecasts is a useful technique to predict inflation, and averaging should be exercised in a new generation of algorithms capable of digesting disaggregated information.