The finite sample properties of break point estimated in Linear Models

In a recent paper, Bai and Perron(1998) present a comprehensive treatment of issues related to the estimation of linear models with multiple structural changes. They consider the properties of the estimators, including the estimates of the break dates, and the construction of the tests that allow inference to be made about the presence of structural change and the number of breaks. Following their theoretical results, they consider the finite sample properties of those estimators in their companion paper (2000, Feb). In the companion paper, they investigate the relative merits of different methods to select the number of structural breaks and suggest the method based on the sequential application of the sup FT(r+1|r) test. They also present the results of simulation experiments (2000, Apr) to analyze the size and power of the tests, the coverage rates of the confidence intervals for the break dates. A wide variety of data generating processes are considered allowing different variances for the residuals and different distributions for the regressors across segments as well as serial correlation. But their simulation study is limited to discrete change models. In the real world the economic structure evolves continuously over time. Thus the continuous change models seems to be more realistic. This paper investigates how Bai and Perron’s method performs if the true model involves continuous structural change. For that purpose. I consider the smooth transition regression model and markov switching model.