Sensitivity of the Ramsey’s Regression Specification Error Term Test on the Degree of Nonlinearity of the Functional Form

Abstract

The present paper aims to demonstrate that the Ramsey’s Regression Specification Error Term Test (RESET) is very sensitive to the degree of nonlinearity between the variables of the under-specification functional form. This widely used test, for testing the functional specification of a model, is based on the notion that if nonlinear combinations of the explanatory variables have any power in explaining the predictor, the model is mis-specified and the data generating mechanism might be approximated by a nonlinear functional form. Using Monte Carlo techniques, we find that the power of the Ramsey’s RESET test is highly influenced and related with the degree of nonlinearity between the dependent and the independent variables of the under-specification functional form.