Three essays in macroeconomics
OA Version
Citation
Abstract
This dissertation consists of three essays studying firm dynamics and expectation formation. The first essay quantifies a tradeoff associated with lean inventory management. The second essay makes sense of simultaneous over- and underreaction in a noisy information setting with time-varying volatility. The third essay offers a new testable implication as a way to narrow the set of models of belief formation that are consistent with survey data.
The first essay investigates just-in-time production (JIT). I first construct a new measure of JIT at the firm level through a text search. Relative to non-adopters, I document that adopters experience higher sales and smoother outcomes, however, they are also more cyclical and sensitive to weather events. Motivated by these facts, I build and structurally estimate a dynamic general equilibrium model of JIT production. Relative to a counterfactual reflecting the adoption patterns of the 1980s, firms in the estimated economy benefit from a 1% increase in firm value in normal times. Amid a COVID-like disaster, however, the estimated economy experiences a 1.6 percentage point sharper contraction.
The second essay examines the role that volatility can play in generating seemingly non-rational behavior. First, I document that the same professional forecaster over- and underreacts to distinct macroeconomic variables. I then show that such behavior can arise in a noisy information environment with unobserved volatility and costly model adoption. In such a model, forecasters overreact to variables for which they have less precise information and underreact to variables for which they have more precise information. I provide empirical evidence in favor of this explanation and calibrate a version of this model to show that it can replicate meaningful shares of simultaneous over- and underreaction.
The third essay similarly relates to survey expectations. By way of example, I show that rational and non-rational models alike are able to deliver the same linear relationship between forecast errors and revisions. I specifically focus on a rational model of strategic interaction as well as non-rational models of overconfidence and diagnostic expectations. I propose examining the serial correlation of revisions instead as it is able to distinguish between these three models.