UC Irvine

This dissertation uses Bayesian methods to understand how expectations are formed and their role in macroeconomic fluctuations. All three essays study expectations formation through different perspectives and econometric tools.

The first chapter analyzes the impact of central bank transparency on the evolution of agents’ expectations for the Mexican case via a New Keynesian model with adaptive learning and survey forecasts. Among multiple scenarios, the data prefer the observed transparency degree followed by the Mexican central bank, where the central bank credibly communicatesthe inflation target and discloses relevant information about its policy rule. The results show that agents exhibit a faster learning speed than the U.S. and a declining perceived inflation persistence. Plus, the model-implied learning mechanism can match the empirical inflation expectations from the Survey on Expectations of Private Sector Specialists. Moreover, there is evidence suggesting that higher degrees of transparency increase the effectiveness of monetary policy in stabilizing the economy.

Chapter 2 assesses the role of economic conditions in inflation expectation formation using a Bayesian latent class ordinal model and qualitative survey data from the Michigan Survey of Consumers. The results show evidence that inflation expectations have been formed distinctly depending on the economic conditions faced by individuals. Furthermore, the effect of demographic indicators, such as age, gender, education, and income, on inflation expectations varies with the level of distress of the economy.

Lastly, the final chapter develops and estimates a model with informational frictions. Agents are inattentive and form subjective expectations using an economic model. The proposed expectation formation mechanism is estimated using Bayesian methods and tested against rational expectations. The paper yields three novel results. First, the model embedding inattention à la Mankiw and Reis (2002) and subjective expectations under adaptive learning provides the best fit of the data. Secondly, the degree of inattention is susceptible to how the expectation formation process is modeled. In particular, the level of inattention is considerably reduced when I depart from the rational expectation assumption. Finally, this result remains unchanged when tested using real-time macroeconomic series and expectations data from the U.S. Survey of Professional Forecasters.