Statistical models are fitted for various purposes. One key purpose is to observe relationships between variables. Another significant reason is to facilitate predictions, often used in selection processes. After fitting a regression model from a sample of observations, one may focus on predicting the value of the dependent variable for a specific value of an independent variable. This specific value of the independent variable might lie within the range of sample values, or more frequently, it might lie outside the sample observations. A crucial criterion for an estimated regression equation is its relevance to data outside the sample used for estimation. This criterion is captured by the concept of parameter constancy, which means that the parameter vector should apply both within and outside the sample data. Parameter constancy can be assessed by testing predictive accuracy. This research paper proposes tests for parameter constancy and predictive accuracy using different parameter vectors in the forecast period