Two Essays on Behavioral Economics


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date21 Jun 2023


This thesis discusses both theoretical and empirical evidence for deviation from full information rational expectations, with the focus on the application of diagnostic expectations (DE) in theory and the empirical examination of full information rational expectations.

The first chapter investigates the implications of incorporating diagnostic expectations (DE) into macroeconomic models. The study first explores whether DE generates additional volatility for endogenous variables and under what conditions. The results reveal that DE generates extra volatility only under certain conditions, such as price rigidity and agency frictions. When these conditions are absent, DE does not generate additional volatility. The paper also examines whether DE improves the fit of macroeconomic models compared to rational expectations (RE). To answer this question, the study employs a Bayesian estimation of the financial friction model. The results demonstrate that DE outperforms RE in terms of model fit. In addition, the paper analyzes the effects of DE on various economic variables, such as the natural rate, zero lower bound, and marginal propensity of consumption. The diagnosticity parameter as well as the persistence of the TFP shock, determines the degree of extrapolation of the natural interest rate. With the introduction of diagnosticity, consumers exhibit a high marginal propensity to consume, which aligns with recent empirical evidence. The study shows that DE can have significant effects on these variables, which has important implications.

The second chapter applies three approaches to assess the predictability of forecast errors in order to test the full-information rational expectations using professional forecast data in China and the Euro area. I observe simultaneous underreaction and overreaction to forecast revisions as well as recent outcomes, which cannot be fully explained by any of the three expectation formation models. To capture key features of my findings, I adopt a state space model widely used in the literature. The model includes two behavioral biases: information frictions and misspecified beliefs. In line with most of the literature, I find evidence of over-extrapolation, where professional forecasters overestimate the persistence of shocks. However, unlike previous studies that found overconfidence to be prevalent, my results demonstrate that both overconfidence and underconfidence exist, and the level of overconfidence is determined by the volatility of the actual data-generating process.