This thesis is comprised by three chapters that are not necessarily connected or linked by causality between them, since each one is individually motivated and self-contained. The first chapter evaluates the hypothesis on meeting rational expectations on the interest rate surveys of the central banks of Brazil, Chile and Colombia. Based on a quantitative analysis, it concludes that there is no evidence to support the fulfillment of the rational expectations hypothesis in the surveys. The second chapter discusses whether the implementation of adaptive learning in a standard business cycle model has a differentiated effect on output volatility and persistence depending on the market (labor or financial) of origin of such learning, gauged for the observed U.S. figures. The study shows the importance of learning in the financial (capital) market as the factor that contributes the most to output volatility by way of increased investment volatility, as well as its ability to adjust to the observed figures. The third chapter analyzes the effects of learning expectations on two models with financial frictions, which present a specific vision which is closer to how the financial market actually works. In particular, it asks whether learning expectations adds greater volatility to the output in the models with financial frictions of Carlstrom (1997) and Carlstrom (1998) without altering the persistence generated by these models. Moreover, the results show the relevance of learning expectations in a financial market with frictions arising from information asymmetry. Lastly, a better adjustment to the observed output volatility for the U.S. is achieved, associated to a better adjustment in the volatility seen in both investment and rate of return.
