ABSTRACT
Regression models derived from economic theory are long run-equilibrium models. Based on theory, they show functional relationships among a number of variables in the long run. The time series data that are used in regression models are, however, dynamic, showing the magnitude of economic variables over specified intervals of time. There is no guarantee that these observed magnitudes are long-run values for the variables under consideration, or that interactions among economic variables are completed over one time period. There could be some lagged responses between the dependent variable and the regressors specified in the model, due to slow adjustment processes. The data captures the lagged responses, but the regression model excludes them. There is, therefore, inconsistency between long-run static econometric models and the data used for estimation. This is detected by the diagnostic tests, the DW and the RESET, as a specification problem. This chapter continues with the specific to general (SG) methodology and explains how this methodology attempts to introduce a dynamic lag structure into the regression model in response to the specification error problem indicated by the diagnostic tests. Closely linked to this correction procedure is the phenomenon of the spurious regression and the way the traditional (SG) approach attempts to deal with this problem. This chapter also includes additional diagnostic tests used in practice for the evaluation of regression models. Key topics
The phenomenon of the spurious regression
The partial adjustment and the autoregressive short-run dynamic model
Testing for the structural break and parameter stability: the Chow test
Testing for measurement error: the Hausman test
