Improved Testing of Distributed Lag Model in Presence of Heteroscedasticity of Unknown Form

Abstract

The finite distributed lag models (DLM) are often used in econometrics and statistics. Application of the ordinary least square (OLS) directly on the DLM for estimation may have serious problems. To overcome these problems, some alternative estimation procedures are available in the literature. One popular method to estimate these models is the Almon technique, proposed by Almon (1965). However, testing of the DLM has not attained the attention of researchers. The present study covers this gap and proposes some methods for testing the DLM combined with the Almon technique. Furthermore, the testing of DLM in presence of heteroscedasticity may be invalid due to adverse consequences of heteroscedasticity when applying the OLS procedure combined with the Almon technique. The present article suggests to use heteroscedasticity-consistent covariance matrix (HCCM) estimators to draw valid inference for the parameters of the DLM with heteroscedastic errors. The HCCM estimators based confidence intervals, t- and F-tests are proposed and the performance of these tests and confidence intervals is evaluated through the Monte Carlo simulations by computing empirical null rejection rate (NRR) and coverage.

Keywords: Almon technique; coverage; Finite distributed lag model; heteroscedasticity-consistent covariance matrix estimator; Null rejection rate.