Abstract

Knowing the history of your topic of interest is important: It teaches what happened in the past, helps to understand the present, and allows one to look ahead in the future. Given my interest in the development of Bayesian econometrics, this thesis starts with a description of its history since the early 1960s. My aim is to quantify the increasing popularity of Bayesian econometrics by performing a data analysis in the sense of measuring both publication and citation records in major journals. This will give a concrete idea about where Bayesian econometrics came from and in which journals its papers appeared. With this information, one will be able to predict some future patterns. Indeed, the analysis indicates that Bayesian econometrics has a bright future. I also look at how the topics and authors of the papers in the data set are connected to each other using the bibliometric mapping technique. This analysis gives insight in the most important topics examined in the Bayesian econometrics literature. Among these, I find that a topic like unobserved components models and time varying patterns has shown tremendous progress. Finally, I explore some issues and debates about Bayesian econometrics. Given that the analysis of time varying patterns has become an important topic, I explore this issue in the following two chapters. The subject of Chapter 3 is twofold. First, I give a basic exposition of the technical issues that a Bayesian econometrician faces in terms of modeling and inference when she is interested in forecasting US real GDP growth by using a time varying parameter model using simulation based Bayesian inference. Having observed particular time varying patterns in the level and volatility of the series, I propose a time varying parameter model that incorporates both level shifts and stochastic volatility components. I further try to explain the GDP growth series using survey data on expectations. Doing posterior and predictive analyses, the forecasting performances of several models are compared. The results of this chapter may become an input for more policy oriented models on growth and stability. In addition to output growth stability, price stability is also an important policy objective. Both households and businesses are interested in the behavior of prices over time and follow the decisions of policymakers in order to be able to make sound decisions. Moreover, policymakers are interested in making inflation forecasts to be able to make sound policy decisions and guide households and businesses. Therefore, inflation forecasting is important for everybody. I deal with this topic in Chapter 4. In this chapter, I explore forecasting of US inflation via the class of New Keynesian Phillips Curve (NKPC) models using original data. I propose various extended versions of the NKPC models and make a comparative study based on posterior and predictive analyses. I also show results from using models that are misspecified and from using survey inflation expectations data. The latter is done since most macroeconomic series do not contain strong data evidence on typical patterns and using survey data may help strengthening the information in the likelihood. The results indicate that inflation forecasts are better described by the proposed class of extended NKPC models and this information may be useful for policies such as inflation targeting. Section 1.2 summarizes the contributions of this thesis. Section 1.3 presents an outline of the thesis and summarizes each chapter.

, , ,
H.K. van Dijk (Herman)
Thela Thesis, Amsterdam , Erasmus University Rotterdam
hdl.handle.net/1765/76037
Tinbergen Instituut Research Series
Erasmus School of Economics

Ceyhan, P. (2014, September 4). Essays on Bayesian Analysis of Time Varying Economic Patterns (No. 590). Tinbergen Instituut Research Series. Retrieved from http://hdl.handle.net/1765/76037