Suggestions of economic policy stand for the accurate estimation of the relationships among variables. Parameters of the estimated model indicate the effects of policies. If the parameters do not change with respect to policies, it may be suggested that the existing policies are valid in the long-term. This situation is known as Lucas critique in the literature. If the policy change is recognized at an unknown point and slowly in particular time period (as shown in Lucas critique), it may be inferred that the policies are not considered with structural change (Greene, 2002). This circumstance may damage the succes of explanatory power of the model and may cause the invalidity of the results, so the stability of model should be tested by convenient methods.
There were various studies regarding model stability in the literature. These studies stand for Lucas critique (1981), which argues that “the economic policies must change whenever the parameters of the model change”. The common suggestion of the studies refers to the neccessity of model stability against policy changes in order to use the model as an economy policy. Model stability can be tested by package programs with developing technology. In these programs, CUSUM and CUSUMQ packages are generally used.
This study considers specific model stability tests that can be used when the policy change is slowly recognized in a time period at an unknown point. There are certain weaknesses of frequently used CUSUM and CUSUMQ model consistency tests statistically (Andrew, 1993; Yashchin, 1993; Turner, 2010). MOSUM test, which was constructed based on the cumulative sums of the recursive resudials as CUSUM test, enables more robust estimation than COSUM test and therefore gives more information about the model (Chu et al., 1995, Zeileis et al., 2002). For these reasons, the results of two tests were involved in the study.
The paper is organized as follows. Section 2 involves the studies in the literature associated with model stability tests, Section 3 consists of econometric methodology and model form about time series analysis, furthermore after the variables used in the study are introduced, the results of unit root, cointegration, CUSUM and MOSUM tests being obtained by using Eviews-6 and R-2.14.2 package programs are interpreted respectively. Section 4 concludes the study.
For full text: click here
(Author: Emrah Talas, Fatih Kaplan, Ali Kemal Celik