Abstract
This dissertation investigates the role of dispersed information and higher-order beliefs in macroeconomics and finance. In the first chapter, I study the implica- tions of dispersed information in determining the joint behavior of business cycle fluctuations and equity market movements. Consistent with the empirical evi- dence, I construct a real business cycle model where lagged public information regarding economic fundamentals and instantaneous financial information (the equity price) are available to agents. The model features the “forecasting the fore- casts of others” problem in which the higher-order expectations (HOEs) constitute additional state variables that serve the role of “missing fundamentals.” Due to the presence of HOEs, the variance bound for stock price volatility does not need to hold. To generate sufficient volatility, I introduce a dividend shock. This shock poses direct effect on the equity price and induces correct comovement among output, investment, labor, and consumption. The quantitative analysis shows the model accounts for 55% of the stock price volatility and generates data-consistent movements in the equity return, the dividend, and the dividend-price ratio. In ad- dition, the model features plausible information dynamics witnessed in the data. The degree of information frictions is well-matched to empirical observations from the Survey of Professional Forecasters (SPF) in terms of the forecasting error mag- nitude, autocorrelations, and the expectation dispersions. Using the Kalman gain decomposition, I show the importance of equity price in shaping the agents’ HOEs, even though it deviates from the conventional valuation of true fundamentals.
In the second chapter, I study a general class of linear models with incomplete information, strategic complementarity, and decentralized market. I introduce a higher-order belief shock (sentiment) that is assumed to be tied to the aggregate pay-off relevant shock. Using a simple stylized model, I show (1) the effect of higher-order beliefs diminishes exponentially (2) the impact response to the sentiment shock is strictly bounded by the response of pay-off relevant shock regard- less of the degree of information frictions (3) increasing the degree of strategic
complementarity monotonically amplifies the effect of sentiment. I discuss the robustness of these results when endogenous information or heterogenous prior are incorporated in the model. In contrast with the previous literature, different formulations of learning and belief shocks do matter for quantitative evaluations of belief-driven fluctuations, and there is a theoretical difference between higher- order belief shocks and the first-order noise shocks. As an example, I compute a business cycle model with persistent information frictions and capital accumula- tion, in which the sentiment shock is outperformed by the TFP shocks in terms of persistence and volatility
