Resumen de Essays in learning uncertainty
- català
Els agents prenen decisions sota incertesa. No només estan incerts sobre les realitzacions de les variables d’interès, sinó també sobre el seu grau d’incertesa. Utilitzant l’enquesta d’incertesa empresarial, estudio la incertesa subjectiva de les empreses i demostro que les variacions en la incertesa subjectiva de les empreses poden ser impulsades a través de l’aprenentatge. A continuació, estudio les implicacions de la incertesa en l’aprenentatge en els patrons d’inversió de les empreses. Per fer-ho, primer construeixo un senzill model d’inversi´o de les empreses amb incertesa en l’aprenentatge i demostro que aquesta t´e tres implicacions principals. Utilitzant les dades de Compustat, demostro que les tres implicacions de la incertesa d’aprenentatge per als patrons d’inversi´o de les empreses són observables a les dades. Atès que la incertesa sobre els segons moments afecta les decisions dels agents, també té implicacions importants per a la conducta òptima de polítiques com la política monetària. Finalment, construeixo un model amb incertesa sobre la dispersi´o de la productivitat i estudio la resposta òptima de la política monetària al xoc de dispersió de la productivitat.
- English
Subjective uncertainty plays an important role in the decision-making process. Agents are uncertain about the realizations of many variables of interest, such as sales revenue, exchange rates, income, productivity, and so forth. The degree of agents' uncertainty varies over time; sometimes, they become more certain and sometimes more uncertain. How does the degree of uncertainty vary over time? Which factors affect and drive agents' uncertainty? We can think of subjective uncertainty as the second moment of the subjective probability density function of the variables of interest. By observing the realizations of the variables of interest, agents can learn the whole distribution. Throughout the three chapters of my thesis, I concentrate on variables that are assumed to be normally distributed with unknown second moments. By observing the history of the realizations of those variables, agents can learn the second moments or variances of the data-generating distributions. Learning second moments or, in other words, learning uncertainty, is a potential driver of agents' subjective uncertainty.
In chapter 1, using firm-level survey data, I study firms' subjective uncertainty. I show that subjective uncertainty is time varying. Moreover, subjective uncertainty responds significantly positively to realized uncertainty, and the conditional responsiveness to realized uncertainty decreases over time. Given these findings, I propose learning uncertainty as a mechanism for mapping realized uncertainty to subjective uncertainty.
After validating learning uncertainty as a possible driver of subjective uncertainty, in chapter 2, I study the implications of learning uncertainty for firms' investment patterns. To do this, I first build a partial equilibrium model of firms' investment with learning uncertainty. I show learning uncertainty results in 1) lower investment responses to the idiosyncratic TFPR shocks for firms that experience more volatile productivity in their lifetime, 2) lower investment responses to larger idiosyncratic TFPR shocks, and 3) asymmetric responses to symmetric positive and negative idiosyncratic TFPR shocks (asymmetric S-shaped response). Next, using Compustat data, I estimate TFPR at firm-level and study the dynamism of firms' investment rate response to idiosyncratic TFPR shocks. I show that three implications of learning uncertainty for the investment patterns are observable in the data. Finally, based on the findings from Survey of Business Uncertainty about the drivers of subjective uncertainty, I assume Compustat firms are Bayesian learners who are uncertain about true realizations of the variance of the idiosyncratic TFPR shocks' data-generating process. After building their time-varying posteriors' uncertainty (variance of posterior beliefs) regarding idiosyncratic TFPR shocks, I study the impact of firms' posteriors' uncertainty on the investment response to TFPR shocks. I show that firms' posteriors' uncertainty about idiosyncratic TFPR shocks negatively affects their investment response to the shock, which verifies learning uncertainty as a potential driver of the three mentioned investment patterns in the data.
In chapter 3, I present a theoretical framework that features a contractionary productivity dispersion shock, which is a result of the interaction between the substitutability of supplied labor and demanded goods. I introduce information friction as a source of nominal rigidity to study the impact of the dispersion shock on the conduct of the monetary policy. In particular, I assume firms have incomplete information about the productivity dispersion when they set the price. I show that, in an environment with nominal rigidity, replicating the full-information flexible price equilibrium is always feasible and optimal. The optimal monetary policy is the policy that eliminates the dependence of the idiosyncratic nominal variables on the unknown productivity dispersion and, as a result, removes information friction.
