Performance of Methods Determining Structural Break in Linear Regression Models

In the literature, although many studies are describing the structural break in the linear regression model with time-series data, studies investigating this issue with cross-sectional data are limited. In this study, the performance evaluation of some approaches used to determine the structural break in a linear regression equation based on cross-sectional data was performed. In this context, firstly, the structural break problem is defined. Then, the theoretical expositions of some well-known methods which determine the structural break are given. The methods which used to determine the structural breaks may show performance differences under the effect of some factors. The performances of selected methods were evaluated with a simulation study in the context of the difference of constant terms, the difference of slope coefficients, location of break-point, sample size and homogeneity of error variances. The results of the simulation showed that the performances become different in terms of some structural features from the suggested methods for determination of structural break.