Monetary Communication Rules

(with Amy Handlan)

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 and regularized regressions, we find strong evidence for systematic communication rules that vary over time, with changes in the rule often being associated with changes in the economic environment. We also find that shifts in communication rules increase disagreement among professional forecasters and correlate with monetary policy surprise measures. Our method is general and can be applied to investigate systematic communication in a wide variety of settings.

Paper; Online Appendix

NBER Monetary Economics Fall Meeting  November 10, 2023 (see the livestream here)

NBER Summer Institute Workshop on Methods and Applications for Dynamic Equilibrium Models July 11-12, 2024 (see the livestream here)

Reputation for Confidence

(with Amy Handlan)

How does reputation affect the announcement strategy for central bankers? This paper provides an answer with a novel combination of theory and computational text analysis. We study a communication game between a central bank and the public where  the central bank makes announcements about their noisy forecasts of shocks and sets policy to balance its inflation and output targets. Meanwhile, the public updates beliefs about policy conditional on (1) the bank’s announcement and (2) the public's belief about the central bank's ability to accurately forecast, called reputation for confidence.  A tension between the central bank’s ability and their reputation induces nonmonotonic effects on how granularly the central bank communicates. Finally, we use a state-of-the-art text analysis model on internal drafts of alternative Fed policy statements to validate the model.

Monetary Policy & Anchored Expectations - An Endogenous Gain Learning Model

Journal of Monetary Economics, 2023

This paper analyzes monetary policy in a model with a potential unanchoring of inflation expectations. The degree of unanchoring is given by how sensitively the public's long-run inflation expectations respond to inflation surprises. I find that optimal policy moves the interest rate aggressively when expectations unanchor, allowing the central bank to accommodate inflation fluctuations when expectations are well-anchored. Furthermore, I estimate the model-implied relationship that determines the extent of unanchoring. The data suggest that the expectations process is nonlinear and asymmetric: expectations respond more sensitively to large or downside surprises than to smaller or upside ones.

Published version; Online Appendix; Originally submitted version (substantially longer)

VMACS 2021 talk (10 minutes)

NBER Inflation Expectations 2022 talk (with discussion by Karthik Sastry, available until June 2022)

NBER Summer Institute 2022, Behavioral Macro Workshop talk (30 minutes, starting at 6:00:30, with discussion by Bruce Preston, available until August 2022)

Talking Over Time - Dynamic Central Bank Communication

Journal of Money, Credit & Banking, 2022

This paper studies the optimal dynamic communication strategy of central banks using a Bayesian persuasion game framework. In a dynamic environment, financial market participants and the general public have misaligned interests because the present and future have different relevance in their optimization problems, leading to a novel tradeoff for the monetary authority. Compared to the static benchmark, I show that the central bank’s optimal dynamic communication policy should put a higher weight on talking about the present state than the future. In addition, the central bank should strategically send more noisy signals than in the static benchmark.

Published version; Originally submitted version

Investment and Communication Technologies and Medium-Run Fluctuations 

(with Marco Brianti)

Journal of Economic Dynamics and Control, 2023

This paper explores the possibility that productivity improvements in information and communication technologies (ICT) are a source of medium-run fluctuations in total factor productivity (TFP) at the aggregate level. We first document in a structural VAR setting that innovations in ICT investment are followed by gradual, hump-shaped increases in TFP that peak after about six years. Following the literature on ICT, we then use a two-sector model to suggest a possible mechanism behind the hump-shaped TFP response: that ICT may be a general-purpose technology (GPT). Capturing the notion of a GPT as a spillover from the stock of ICT capital, we estimate the spillover elasticity via impulse-response matching. We show that a model augmented with a spillover from ICT capital is able to match the hump-shaped TFP response, implying that the size and dynamics of medium-run fluctuations in response to ICT innovations crucially hinge on the extent to which new ICT technologies diffuse in the economy.

Published version

The Feeling of the Age: A Quantitative Analysis of the Correlation between Novelistic and Economic Sentiment

(with literature scholar Daniella Gáti)

What predictive information do works of fiction contain for long-run economic fluctuations? We investigate this question in a two-step procedure. First, we conduct a sentiment analysis of prize-winning novels of the Pulitzer, National Book Award, PEN Faulkner, PEN Hemingway and Book Critics Circle awards from the years 1948-2018. We assign the novels quantitative sentiment scores. Second, we explore the correlation between the sentiment scores and macroeconomic aggregates such as GDP, total factor productivity, R&D expenditures, and so on. Once we have identified the frequencies where most of the comovement occurs via frequency domain methods, we analyze the cointegration relationships and the effects of innovations to literary sentiment using a structural Vector Error Correction model (VECM). 

Do Long-Horizon Expectations Matter for New Keynesian Models?

In the literature on New Keynesian (NK) models with statistical learning, common practice is to write down the log-linearized first order conditions of the NK model and replace the objective with a subjective expectations operator ("Euler-equation approach"). An alternative approach following Preston (2005) involves obtaining model equations in which long-horizon expectations are explicitly spelled out ("long-horizon approach"). I investigate numerically what implications the two approaches have for model dynamics. While both lead to oscillatory impulse responses, the long-horizon approach results in much more volatility because the term structure of interest rate expectations, absent in Euler-equation learning, responds very sensitively to shocks.