RESEARCH
RESEARCH
Links: [Published Paper] [Draft of working paper]
Abstract: We assess whether men and women are treated differently when presenting their research in economics seminars. We collected data on every interaction between presenters and audience members across thousands of seminars, job market talks and conference presentations, leveraging both human judgment and audio processing algorithms to measure the number, tone and type of interruptions. Within a seminar series, women are interrupted more than men, and this finding holds when controlling for characteristics of the presenter and their paper topic. This differential treatment appears to reflect in part greater engagement with female speakers - resulting in larger attendance and seminar engagement, and in part greater hostility toward female speakers - resulting in more negative, mid-sentence, and declarative interruptions.
Links: [Updated Draft, 2026][CEPR Working Paper] [Online Appendix]
Abstract: In a cheap-talk communication game, we model how a sender communicates their noisy forecasts while taking into account their own uncertainty (confidence) and the receiver's perception of the sender's uncertainty (reputation for confidence). This creates a mismatch between the sender's and receiver's interpretation of the announcement. This misunderstanding friction induces the sender to communicate with partial transparency and deliberate imprecision. Moreover, with higher confidence (lower reputation) announcements are more precise. To test the theory, we leverage unique data on Federal Reserve communication deliberations to create new text-based measures as direct counterparts to the model. We find communication patterns are largely consistent with the model except the Fed’s communication strategy underreacts to reputation compared to the model.
Links: [Updating Draft (to be posted soon)] [Slides]
Abstract: Does the Federal Reserve follow a communication rule? We propose a simple framework to estimate communication rules, which we conceptualize as a systematic mapping between the Fed’s expectations of macroeconomic variables and the words they use to talk about the economy. Using text analysis with Fed forecasts, we find strong evidence for a systematic communication rule. However, we find strong evidence for a break in the rule at the end of 2008, over and above a break in policy decision-making. Post-2008 communication loads more on financial conditions and economic forecasts and overall more systematically moves with internal forecasts. Meanwhile pre-2008 communication looks more like direct communication of a Taylor rule. To test market reactions, we first decompose changes of announcements into systematic (communication rule fitted values) and non-systematic (deviations from the rule). We find both systematic and non-systematic communication account for more variation in monetary surprises post-2008. Our interpretation is that systematic communication brought more investor attention, which increased investor reactions to both systematic and non-systematic communication.
Abstract: This paper shows that the wording of Federal Reserve communication affects expectations and other economic variables over and above the effects of setting the federal funds rate. Adapting neural network methods for text analysis from the computer science literature, I analyze how the wording in statements of the Federal Open Market Committee (FOMC) impacts fed funds futures (FFF) prices when these statements are announced. Using text analysis on FOMC statements and internal meeting materials, I create a new monetary policy “text shock” series for 2005-2014 that isolates the variation of FFF prices induced by the FOMC’s forward guidance, not their current assessment of the economy. I find that the wording of FOMC statements accounts for four times more variation in FFF prices than direct announcements of changes in the target federal funds rate. I also find that the impact of forward guidance on real interest rates is twice as large when using text shocks over other measures, like changes in FFF prices. Furthermore, the text shock produces responses in output and inflation that are qualitatively consistent with workhorse macroeconomic models, whereas changes in FFF prices do not.
Links: [Updating Draft (to be posted soon)]
Abstract: The Federal Open Market Committee (FOMC) claims that their post-meeting statements shift market expectations of future monetary policy. In this paper, I provide evidence supporting this claim. I apply a methodology from computational text analysis to produce a pairwise-statement similarity measure that compares wording between two FOMC statements. This similarity measure documents that FOMC statements have become more similar over time. With an event-study approach, I find that a decrease in the similarity of sequential FOMC statements is correlated with an increase the variation of federal funds rate expectations, calculated from high-frequency fed funds futures prices. This relationship persists even after controlling for changes in the target federal funds rate and Federal Reserve Chair. Standard monetary regressions omit any measure of policy statement texts and are thus biased. Adding the sequential statement similarity measure to a regression of federal funds rate expectations on the target rate accounts for 1.5 times the variation in market expectations. This paper suggests that more detailed text analysis on FOMC statements will improve modeling of monetary policy expectations.
"Government-Directed R&D and Innovation: Evidence from NASA Budget Requests,'' with Ethan Ilzetzki and Réka Juhász
"FedSpeak Indicators," with Sara Canilang and Amanda Michaud
"Hot off the Press: Newspaper Sentiment and Economic Fluctuations,'' with Jennifer La'O
"Joint Taxation and Labor Supply," with Ross Batzer
"Gender and Tone in Recorded Economics Presentations: Audio Analysis with Machine Learning," with Haoyu Sheng
Link: [Draft, May 2025]
Abstract: This paper measures seminar dynamics using a replicable, scalable, machine-learning approach and finds a gender-tone gap in economics presentations. We train a deep convolutional neural network to impute labels for gender and tone-of-voice. We apply this to recorded presentations from the 2022 NBER Summer Institute to measure tone at a high frequency, which allows us to provide novel results on how economists interact with each other in talks. We find that female economists are less likely to be spoken to in a positive tone and more likely to be addressed with a serious and stern tone. Female economists are also more likely to speak in a positive tone. Overall, we show that gender differences in economics presentations exist across fields and presentation formats.
"How do Stimulating Policies Move Firms' Expectations? Evidene from Chinese Micro and Small Enterprises" by Huo, LI, Song, Yang, and Zhang
Conference: Fridays at the Boston Fed Workshop, May 2026
"ECB Communication and its Impact on Financial Markets" by Istrefi, Odendahl, & Sestieri
Link to slides: [Discussion Slides]
Conference: Central Bank Communications: Theory and Practice Conference 2024 (FRB Cleveland, OH)
"Estimating the Effects of Political Pressure on the Fed: A Narrative Approach with New Data," by Drechsel
Link to slides: [Discussion Slides]
Conference: NBER Monetary Economics Spring Workshop 2024 (Chicago, IL)
"Policymakers' Uncertainty," by Cieslak, Hansen, McMahon & Xiao
Link to slides: [Discussion Slides]
Conference: Western Finance Association 2023 (San Francisco, CA)