Parameter non-constancy in linear regression: applications of threshold regression and structural change models

Abstract

This dissertation provides an overview of recent developments in the identification and estimation of unknown change-points. It explores the applicability and efficacy of the new approaches to the analysis of economic relationships in agricultural marketing, agricultural production, and food markets. It demonstrates via empirical analyses how the methods can be implemented and the results interpreted. In particular, this dissertation includes three empirical essays, focusing on two families of parameter-nonconstancy models: threshold regression and structural change models. In addition, given the breadth of the topics (generic dairy promotion program, rice production efficiency, and vertical price relationship) and the countries of data collected (the United States, Bangladesh, and Japan), the three empirical essays address a broad spectrum of important issues in different commodity sectors under different economic environments.