Research

This study develops a two-step procedure to identify and quantify fiscal news shocks. First, we augment a narrative identification strategy using Large Language Model searches to compile events (2001–2023) that altered the expected path of U.S. defense expenditure. Second, for each event, we estimate market-implied shifts in expected defense spending with cross-sectional regressions of contractors’ stock returns on their reliance on military revenues. We show that this approach statistically validates each event, quantifies each shock in an intuitive, model-consistent fashion, and readily generalizes to other macroeconomic contexts. Employing the estimated shocks in a shift-share analysis yields a two-year MSA-level GDP multiplier of approximately 1 for U.S. military build-ups.

This study proposes a novel approach to interpreting conventional tests of the Full Information Rational Expectations (FIRE) hypothesis, based on the distinctions between forecasts and expectations. First, I argue that these two objects coincide under highly restrictive conditions that are unrealistic in the context of professional forecasting. Evidence from reduced-form analysis suggests that forecasters pursue strategic incentives when responding to surveys, contaminating their responses and making their use as measures of expectations misleading. Second, I leverage this distinction to introduce a new parsimonious model of forecast formation that is related to rational inattention and sparsity-based models of bounded rationality, but that is exempt from their complications, including the dependence on Gaussian information. The proposed framework employs a global game structure featuring public and private information, as well as the strategic behavior of forecasters, and demonstrates that strategy can explain the anomalies in Coibion and Gorodnichenko (2015)-type regressions. Finally, I exploit the model’s transparency to develop and apply a formal test that validates the strategic channel, advising caution in using surveys to elicit expectations.

High-frequency surprises are a popular instrument for monetary policy shocks, and yet they are predictable from information already public at the time of the FOMC announcement. Orthogonalizing them against pre-announcement macro-financial predictors has become the state-of-the-art way of reinstating instrument validity. This paper shows that when surprises are aggregated to monthly, the timing of orthogonalization becomes part of the identifying design. I illustrate this in the orthogonalized monetary surprises of Bauer and Swanson (2023a): their series sums within-month surprises and purges the total using first-announcement predictors. I show that the obtained instrument retains a predictable component, therefore identifying a combination of shocks and news effects. Instead, I orthogonalize each surprise against its own pre-announcement information set and then aggregate. Under this method, first-stage strength collapses and the large macroeconomic responses reported are no longer statistically detectable. I rationalize the stark differences: the original instrument’s power is concentrated in multi-announcement months — exactly where a first-announcement purge cannot remove later-announcement information. A simple model explains the asymmetry and formalizes orthogonalization timing as part of the identification design.

We explore two major surveys of macroeconomic forecasts to examine patterns of heterogeneity in the cross-section of forecasting performance. We identify a range of new stylized features of forecast accuracy that, we show, are robust across horizons, variables, and surveys. We argue that many of these properties cannot be accommodated by existing models of noisy information or sticky expectations. Instead, we develop a parsimonious heterogeneous-agent hybrid model that features both noisy information and sticky expectations. We calibrate the parameters of this model and demonstrate through simulations that this model comes closer to fitting the data.

What portion of the movements in real estate prices cannot be rationalized by the most commonly imputed economic factors? How much should we care about the “animal spirits”, intended as fluctuations in agents' beliefs, when studying the origination of financial crises? The goal of this paper is to assess the role of confidence in understanding the boom-bust dynamics of credit and house prices in the United States, with a focus on the Great Financial Crisis. We identify confidence shocks in a structural VAR as exogenous innovations to the University of Michigan Sentiment Index. A positive shock to confidence increases significantly and prolongedly house prices, and accounts for almost 30% of their variance after 12 months, remaining significant after five years. We embed confidence shocks in a DSGE model with households’ heterogeneity and financial frictions. Confidence shocks are modeled as exogenous shifts in beliefs that are unrelated to current and future economic fundamentals, and that can generate waves of optimism or pessimism about current economic activity. Results from a Bayesian estimation of the model show that innovations to confidence account for half of the volatility of observed house prices. Our estimated confidence series shows a strong positive correlation with the Sentiment Index, suggesting that belief fluctuations in the model resemble empirical measures of confidence.

Economic models with input-output networks assume that firm or sector growth is driven by a combination of trade partners’ growth and idiosyncratic shocks. This assumption generates unrealistic restrictions on network weights. Allowing for correlated shocks exposes units to additional risk that captures their ability to substitute away from supply and demand shocks. I provide evidence that substitutability between trade partners is related to technological and product dispersion that is not captured by standard firm and industry definitions. I propose a production-based asset pricing model in which supply chain substitutability depends on product/technology dispersion and shock correlation driven by shared suppliers and customers. The model predicts that assets positively exposed to upstream and downstream shocks are useful hedges and earn lower average risk premia than less exposed peers. This is confirmed by estimated return spreads of -11.4% and -4.2% and a negative association with aggregate growth.