Aspects Concerning Nonlinearity In Econometric Modeling

In this paper, we present certain points of view regarding nonlinearity in the construction of econometric models, as well as its implications on parameter estimation. The research analyzes different scenarios encountered in economic practice and the appropriate methods of approach found in the econometric literature. In this context, we also present the authors' personal viewpoints regarding modeling of nonlinear economic relationships, substantiated by practice in modeling and simulation of economic phenomena and processes. Classifying nonlinear relationships is of importance both for short-term estimation purposes and especially for the analysis of the precision of medium to long term predictions. There is a class of nonlinear models that can be approximated well by linear models, easily approachable using the OLS (Ordinary Least Squares) method. When we opt for such an approach, analyzing how the residual variable (the error) fits within the model is essential (meaning that linearization can be hindered by the additive or multiplicative form in which the error is expressed). Also, practice has shown that medium to long term evolutions can sometimes be nonlinear even though, for short-term predictions, a linear model has proven acceptable.