Welcome to my website! I am a Ph.D. candidate in Economics at the University of British Columbia. My research interests are Monetary Economics, Macroeconomics, and Time-Series Econometrics. My work focuses on the identification of macroeconomic shocks, in particular high-frequency instruments and how to combine several imperfect instruments in structural VARs. I am on the job market in 2026–2027.
Job Market Paper
Presentations: Canadian Economics Association (2026); Student T2M Conference, HEC Montréal (2026); Banco de México (2026)
Impulse responses to monetary policy shocks identified from high-frequency interest-rate surprises are large: larger than any conventional identification scheme delivers on the same reduced form, and larger still when the surprise is first orthogonalised on pre-announcement information. Bauer and Swanson (2023) read that enlargement as the removal of an attenuation bias. I argue that it is residual contamination magnified by a weaker first stage. The correction is necessary, since the surprise is predictable from pre-announcement data, but at meeting frequency it removes proportionally more of the first stage than of the contamination. Because a proxy-SVAR divides whatever contamination survives by the first stage, the bias can be larger after the correction than before it, on impact and at every horizon. The constructive response is to widen the instrument window. The shock is the part of the policy decision not owed to economic conditions, and the announcement window reveals only the slice the market had not already priced. Instruments built from the decision itself, an extended Romer and Romer (2004) shock and a Cholesky-based instrument, carry contamination of their own but keep their first stage under the same purge, and their responses are not enlarged. When no single proxy is both strong and clean, a pool-within, annihilate-across estimator replaces a clean instrument with a restriction on the structure of the contamination: pooling within families of proxies that share a contamination direction buys strength, annihilating across families buys validity, and the direction is identified by the dispersion of relevance within families. On a panel of sixteen U.S. monetary proxies in five families, the estimator delivers a first stage of 24 and a standard monetary contraction.
Working Papers
High-frequency monetary policy surprises are predictable from pre-announcement information, and orthogonalizing them on that information has become standard practice. I show that this purge is necessary but not sufficient, through three results. First, what the purge delivers is conditional on the control set. The raw surprise implies that output rises after a tightening; orthogonalization corrects the sign, but the magnitude remains set-identified: the peak output response varies by a factor of two across equally defensible sets of pre-announcement controls. Second, predictability is not spread evenly across announcements. I document a high-frequency footprint of the Fed put: the predictable component is an easing response to financial deterioration, so it concentrates in easing surprises, while tightenings, which typically follow ordinary good-news data, are already close to clean. A symmetric projection fitted largely on the easings therefore over-cleans the near-clean tightenings. Third, even the recommended purging order is not innocuous at meeting frequency: the pre-announcement controls are measured over multi-week windows that can overlap the previous announcement, so they have already moved in response to the previous policy shock, and purging can remove genuine monetary variation, weakening the first stage.