The Autoregressive Distributed Lag Extensions And Applications
The autoregressive distributed lag (ARDL) framework is comprehensive with both flexible lag orders of dependent and independent variables to explain the dynamic of the responding variable. This thesis extends the ARDL methods to provide practical applications in economic analysis. The extensions inc...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/59126/1/SAM%20CHUNG%20YAN%20-%20TESIS%20cut.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The autoregressive distributed lag (ARDL) framework is comprehensive with both flexible lag orders of dependent and independent variables to explain the dynamic of the responding variable. This thesis extends the ARDL methods to provide practical applications in economic analysis. The extensions include providing bounds of critical values of the additional testing on lagged level independent variables coefficients for the ARDL cointegration test, innovating the ARDL framework for a multivariate unit root test, and proposing ARDL model for Taylor rule studies. Users who are unfamiliar with programming could use the provided critical values to perform the familiar bounds testing procedure to run the cointegration test proposed by McNown et al. (2018). The dynamic and flexible model in the multivariate ARDL unit root test with related time series information helps to gain more statistical power than the univariate framework unit root tests. Besides that, the limitation in the covariate Dickey-Fuller framework that rules out the possibility of cointegration is covered by the ARDL model to avoid power loss caused by model misspecification. Many Taylor rule empirical studies do not follow proper econometric procedures such as unit root, cointegration, and diagnostic tests. Besides, many pieces of evidence show that the Taylor rule regression is unbalanced with mixed integration order variables. Therefore, level estimates without cointegration reported in the studies could give spurious results. The robust standard error is commonly used in empirical studies, but it is helpless to deal with the residual autocorrelation problem, especially if the model includes lagged dependent variable. |
|---|