News, Sentiment, and Inflation Expectations: Insights From Social Media Data and Experiment (Job Market Paper)
SSRN Link
News media serve as a primary information source for most people, with social media taking on a growing role in the way news is distributed and interacted. This paper investigates how households use information conveyed through media to form their inflation forecasts. Leveraging microdata, social media news data, and machine learning techniques, I show that households dynamically update the news topics they focus on when forming inflation expectations. However, the impact of news media on expectations is time varying — at times, it accounts for a significant portion of the variation in expectations, while at other times, its influence diminishes. Using social media reactions as a proxy for news-induced sentiment, I show that sentiment plays a central role in shaping expectations and can predict the direction of forecast revisions, even when the news is non-economic and unlikely to affect inflation through standard mechanisms. A novel information provision experiment, incorporating a sentiment elicitation method, further confirms the causal importance of sentiment in the expectations formation process. Moreover, I demonstrate that identical policy-related information, when framed with different tones, evokes distinct sentiment that lead to asymmetric effects on forecast revisions. I develop a simple Bayesian learning model in which sentiment distorts signal perception, explaining these results.
Presentations (includes scheduled): ASSA 2026, Indiana University Macroeconomic Seminar (2025), Midwest Macroeconomics Meeting (Spring 2025), Ostrom-Smith Conference in Behavioral and Experimental Economics (2023), Hoosier Economic Conference (2023), Department of Economics Macro Brownbag (2023)
Incentivizing Inflation Expectations with Daniela Puzzello, Ryan Rholes, Alena Wabitsch (Submitted)
SSRN Link
Accurate inflation expectations are central to economic modeling and policy. Yet major surveys elicit them without performance-based marginal incentives, despite their well-established importance to belief data quality in experimental economics. We show that marginal incentives raise effort and fundamentally reshape reported inflation expectations: lowering upward bias by 3.4 percentage points, reducing disagreement by one-third, closing the gender gap, and tripling learning rates in an RCT. Incentivized expectations are more consistent and better predict spending. Calibrating a simple New Keynesian model, we demonstrate that these differences matter: marginal incentives sharpen empirical inference and improve policy guidance -- without increasing participant remuneration.
Presentations (includes scheduled): Sixth Joint BoC – ECB – New York Fed Conference on Expectations Surveys (2025)*, BSE Summer Forum Workshop on Theoretical and Experimental Macroeconomics (2025)*, 3rd Paris Conference on the Macroeconomics of Expectations (2025)*, Ostrom-Smith Conference in Behavioral and Experimental Economics (2025)
* - presented by Co-Author
Forecasting the Future Through a Partisan Lens: Electoral Outcomes and Household Expectations with Nayeon Kang
Link
Using a unique survey experiment fielded around the 2024 U.S. presidential election, we study how politically salient events shape household expectations. We examine both the first and second moments of expectations, as well as the joint updating of inflation and unemployment expectations. We find sharp partisan updating: Republicans revise both inflation and unemployment expectations downward, while Democrats revise both upward. In contrast, subjective uncertainty declines for both groups. We also test whether AI-generated personas can reproduce these patterns. While some models replicate the qualitative direction of updating, Concordance Correlation Coefficients remain low and do not exceed 0.10. These results suggest that LLM-generated responses can only partially capture partisan distortions in expectations, warranting caution in their use as substitutes for human survey data.
A penny for your thoughts? Incentive design and inflation expectations elicitation with Daniela Puzzello, Ryan Rholes, Daniela Valdivia, Alena Wabitsch
Choosing Digital Currency Design with Daniela Puzzello
Misperception or Misconception? Dissonance in Households’ Inflation Forecasts with Nayeon Kang
