Robust Specification Testing in Regression: The FRESET Test and Autocorrelated Disturbances

This study investigates the degree of size-distortion of the RESET, FRESETL and FRESETS tests, and their ability to reject falsely specified models in terms of an omitted variable, in the presence of autocorrelation. Specifically, in the presence of AR(1) and MA(1) processes, respectively. We also explore the properties of the corresponding tests when the robust error covariance matrix estimator of Newey and West (1987) is used in the construction of the tests. Our results indicated that the FRESET tests performed equally as well as, and in many cases yield higher "power" than the comparable RESET test. The dominance of the FRESET tests was even more pronounced when a Newey-West correction for autocorrelation was used as evidenced by higher rejection rates.