The major objective of this paper is to demonstrate, theoretically and empirically, the test of a single structural break/change. Failure to address a structural break can lead to forecasting errors and the general unreliability of a model. Three approaches of testing for structural change are discussed using data from Johnston et al. (1997, p.130) on Stata 14 software. The first approach assesses whether there is a structural break in parameters (slope and intercept) while the second and third assess whether there is a break in slope and intercept respectively. The Residual Sum of Squares (RSS) for the restricted and unrestricted models are established to necessitate the use of an F-test in making inferences. According to the first approach, a structural break exists at 5% level of significance. This result is confirmed by the Chow test. The second and third approaches establish that the structural break is from the intercept and not the slope. These results are also affirmed by the Chow test. Furthermore, all these results, from the first to the third approach, are confirmed by an alternative approach which relies on the knowledge that . Therefore, the dependent variable is not affected by the policy change on the explanatory variable but it is mainly affected by the basic unobserved qualitative characteristics of the two sub-periods. For further analysis, it is recommended that a unit root test be conducted using the Zivot-Andrews test. This test has been established as the panacea for the interplay between unit root and structural changes.
Structural Break chow test Zivot-Andrews Test Unit Root