Essay in Bayesian Econometrics

Abstract

Bayesian Econometrics is not as popular as the more common frequentists methods. However, there are problems that Bayesian methods can handle that are not amenable to frequentists methods. The essays to follow show two such applications where Bayesian methods offer solutions to problems posed in widely different areas of economics. In the first essay, the predictive posterior distribution is used to calculate a number of economic quantities that often require extensive work arounds using frequentist methods. In the second essay, it is shown that Bayesian estimation techniques can handle situations wherein an independent variable exhibits no variability, a situation that would lead frequentists methods to fail. In the first essay, A new model is proposed for estimating demand systems based on the Dirichlet and Beta Regression work that has been developed. This model can overcome some of the criticisms aimed at other models used to estimate demand system, in particular the heterogeneity of preferences and the relationship of those preferences to other predictors. By modeling the preferences explicitly, the model can be given a causal interpretation of consumers decisions. Finally, it will be shown that how to aggregate the model up to the market demands. In the second essay, Nurse lead primary care is a relatively rare model for primary care, but given nurses more holistic approach to medicine can prove effective in both reducing costs and disparities in access and outcomes. To compare the cost adjusted quality of care between the traditional physician and a nurse lead primary care model the outcomes from a population of patients that visited each are compared with respect to three complications of hypertension (stroke, heart attack, coronary heart disease). It is found that the population that visited the nurse lead primary care facility had better cost adjusted outcomes than the population that visited the physician lead facility.