Research
Working Papers
-
“Weak Identification in News-Jump Regressions” (Job Market Paper)
Abstract. News-jump regressions are high-frequency event-study regressions under infill asymptotics. When news-jumps in explanatory variables over narrow observation windows are of the same magnitude as diffusive background noise, the regression coefficient is weakly identified, invalidating least-squares estimation and inference. I develop identification-robust inference based on an Anderson–Rubin test and derive, in a Gaussian limit experiment arising from the jump-diffusion process, a minimal sufficient and complete statistic for jumps, thereby enabling powerful conditional inference. In applications, I invert the proposed robust tests to revisit the empirical questions of Nakamura and Steinsson (2018) using minute-by-minute data, assess the effects of oil shocks triggered by the Russia–Ukraine and Iran wars, and examine the dependence of news-driven ETF jumps on market-index jumps.
-
“Identifying Monetary Policy Shocks with Large Language Models”
Abstract. We use large language models to measure sentiment indicators from pre-FOMC briefing materials and construct aggregate-level control variables for Federal Reserve staff expectations about the U.S. economy. Our low-dimensional, AI-based measures extract economically meaningful information from the semantic context of the briefing materials without relying on high-dimensional regularized regressions, thereby avoiding the associated regularization bias. As a complement to the existing narrative approaches, our AI-based monetary policy shocks provide a credible alternative for subsequent impulse response analysis.
-
“Honest Inference for High-Dimensional Linear Models”
Abstract. We propose a heteroskedasticity-robust and dimension-adaptive test for multiple linear restrictions in sparse high-dimensional linear regression models. It combines desparsified LASSO estimation with a ridge-regularized quadratic form statistic and critical values that adapt to the restriction dimension. Under stated regularity conditions, our approach is asymptotically valid under a fixed or diverging number of restrictions without specifying the asymptotic regime in advance. Our random-design theory allows independent non-identically distributed covariate rows with marginal sub-Weibull tails and heteroskedastic errors satisfying finite polynomial moments. Monte Carlo simulations demonstrate improved finite-sample size control relative to conventional Wald tests, particularly as the number of restrictions increases.
Work-in-Progress
-
“A Fixed-k Bipower Jump Test”
-
“Evaluating AI Forecasts on Future Sport Events,”