The Chow test is not robust when the errors are heteroscedastic as well as dependent. The presence of heteroscedasticity and dependency will affect level of significance as well as power of the test, especially when the sizes of the samples are small. The present paper not only resolves the problem of simultaneous existence of heteroscedasticity and dependency in the error terms, but also extends the existing method of comparing two regression equations to many equations in order to make comparisons of the successive coefficients to be possible, thus enabling one to detect structural changes, if any. The procedure is then illustrated through detection of structural change, by comparing the decadal growth rates of population, using State level data of India.
