Working Papers
LLM Survey Framework: Coverage, Reasoning, Dynamics, Identification (with Cynthia Wu and Shihan Xie) [link]
Abstract: We propose a new LLM-based survey framework that enables retrospective coverage, economic reasoning, dynamic effects, and clean identification. We recover human-comparable treatment effects in a multi-wave randomized controlled trial of inflation expectations surveys, at 1/1000 the cost. To demonstrate the framework’s full potential, we extend the benchmark human survey (10 waves, 2018–2023) to over 50 waves dating back to 1990. We further examine the economic mechanisms underlying agents’ expectation formation, identifying the mean-reversion and individual-attention channels. Finally, we trace dynamic treatment effects and demonstrate clean identification. Together, these innovations demonstrate that LLM surveys enable research designs unattainable with human surveys.
LLM Measurement of Open-Ended Reasoning(with Pedro Siman, Cynthia Wu and Shihan Xie) [link]
We propose a framework that uses large language models to measure open-ended reasoning through four stages: embedding and clustering, LLM-assisted category definition, probabilistic scoring of category membership, and validation against human-coded benchmarks. Two design choices are central to the framework: it separates category construction from measurement, and it represents category membership as a continuous probability rather than a binary indicator. We apply the framework in three settings. First, in a survey on the Inflation Reduction Act, our probability measures align closely with independent human coding, while a naive LLM baseline produces unstable categories. Second, moving beyond external validation, we use reasoning about the Federal Reserve's perceived political affiliation to show that continuous probabilities recover subjective uncertainty and overlap across categories that binary coding discards. Third, extending the framework to LLM-generated reasoning across survey question formats (distributional bins and point forecasts), we show that it recovers shared economic content as well as format-specific differences. Together, these results show how open-ended reasoning can be converted into scalable, interpretable, and comparable quantitative measures.
Higher-Order Bias-Corrected Marginal Tests for Predictive Quantile Regressions with Highly Persistent Regressors (with Xiaosai Liao, Xinjue Li and Qingliang Fan)
Abstract: This paper proposes a robust inference procedure for predictive quantile regression with highly persistent predictors, with a focus on marginal tests in multivariate settings. Existing methods either fail to perform marginal inference when control variables have non-zero coefficients, suffer from severe size distortion in tail quantiles due to inaccurate density estimation, or require prior knowledge of predictor stationarity. To address these limitations, we develop a two-step instrumental variable quantile regression estimator that remains consistent under arbitrary coefficient values, enabling valid marginal tests. We further eliminate the higher-order finite-sample bias inherent in the IVX framework using a sample-splitting and weighting scheme, while proposing a simulation-based density estimator that circumvents residual estimation. A combined test statistic is constructed to simultaneously achieve strict size control under the null and high power under the alternative. Monte Carlo simulations demonstrate that the proposed method delivers substantially better size control and power than the mainstream methods in the literature across the quantile range, particularly in tail quantiles and when the number of predictors is large. An application to U.S. equity-premium predictability uncovers significant predictive relationships that vary across quantiles, despite the well-documented difficulty of predicting mean returns.
Machine Learning using Nonstationary Data [paper]
Abstract: Machine learning offers promising tools for forecasting with large datasets, but its theoretical guarantees can fail under nonstationarity. We show that LASSO applied directly to integrated data can either miss predictive signals or select spurious predictors. To address these failures, we propose a simple detrending procedure that does not require prior knowledge of the nature of nonstationarity. For LASSO and adaptive LASSO applied to automatically detrended data, we establish variable-selection and out-of-sample risk consistency. We also derive oracle bounds for LASSO and show that detrending strictly reduces asymptotic forecast risk in an integrated design with stationary predictive signals. Applications to U.S. inflation and industrial production growth using a broad class of machine learning techniques show that the method substantially improves forecasting accuracy over standard practice or performs comparably.
Publications
Principal Component Analysis for a Mix of Stationary and Nonstationary Variables (with James D. Hamilton and Xinwei Ma). Accepted by Journal of Econometrics (2026). [link].
Abstract: This paper develops a procedure for uncovering the common cyclical factors that drive a mix of stationary and nonstationary variables. The method does not require knowing which variables are nonstationary or the nature of the nonstationarity. An application to the FREDMD macroeconomic dataset demonstrates that the approach offers similar benefits to those of traditional principal component analysis with some added advantages.
Model Selection for Multivalued-Treatment Policy Learning in Observational Studies (with Yue Fang and Haitian Xie). Journal of Business & Economic Statistics (2025). [link]
Abstract: This study investigates the policy learning problem in observational studies, where the treatment variable can be multivalued and the propensity scores are unknown. We approximate the optimal policy in a global policy class with infinite complexity (VC/Natarajan) dimension, using a sequence of sieve policy classes with finite complexity dimension. The optimal policy within each sieve class is estimated by maximizing the empirical welfare, constructed through the doubly robust moment condition and cross-fitting method. To select the suitable sieve space, we maximize the penalized empirical welfare, with the penalty determined by either the Rademacher complexity or a holdout method. We establish oracle inequalities that demonstrate the bias and variance tradeoff achieved by the data-driven policy estimator. We also investigate two specific sieve selections: (a) a monotone single index model and (b) a systematic discretization method, which uses conventional sieve results for smooth functions such as linear sieves and deep neural networks. In the empirical study, we apply our method to examine the policy of assigning individuals to job training of different lengths.
Strength in numbers: robust mechanisms for public goods with many agents (with Haitian Xie). Social Choice and Welfare (2023). [link]
Abstract: This study examines the mechanism design problem for public goods provision in a large economy with n independent agents. We propose a class of dominant-strategy incentive compatible and ex-post individually rational mechanisms, which we call the adjusted mean-thresholding (AMT) mechanisms. We show that when the cost of provision grows slower than the square root n rate, the AMT mechanisms are both eventually ex-ante budget balanced and asymptotically efficient. When the cost grows faster than the square root n rate, in contrast, we show that any incentive compatible, individually rational, and eventually ex-ante budget balanced mechanism must have provision probability converging to zero and hence cannot be asymptotically efficient. The AMT mechanisms have a simple form and are more informationally robust when compared to, for example, the second-best mechanism. This is because the construction of an AMT mechanism depends only on the first moment of the valuation distribution.
Work in Progress
A Synthetic Business Cycle Approach to Counterfactual Analysis with Nonstationary Macroeconomic Data (with Zhan Gao, Zhentao Shi and Haitian Xie)
Consumer Subsidy and Household Behavior: Evidence from LLM-Generated Survey Data (with Xiaoan Ma, Yue Shen and Ji Zhang)
Testing for Long-Horizon Predictability: A Reverse Regression Approach with Refined Innovation Orthogonalization (with Xiaosai Liao and Qingliang Fan)