ARDL Modeling in Eviews
Posted: Wed Apr 04, 2018 11:50 pm
Hi, my colleague and I are undertaking a study on exchange rate pass-through for the Philippines by employing an ARDL model. I have a few questions/clarifications.
1. In specifying the unrestricted ECM equation for the PSS bounds test, is it proper to declare of the one-period lagged level variables under the list of fixed regressors of the ARDL model?
2. Is it accurate to say that the PSS bounds test just affirms whether cointegration exists but does not constrain the final specification to be just in the first difference form (growth rate, since level is in log)?
3. If we need to break the full sample into two sub-periods (i.e, Pre-IT and IT), is it necessary to establish cointegration per sub-period as well?
4. If individual coefficients in the unrestricted ECM format are insignificant but the F-test exceeds the upper bound of the PSS critical value, can we proceed to paring down the equation to log first difference form?
5. If the final equation is in log difference form, does the level equation reported after clicking the long-run and bounds test refer to short-run coefficients only, considering that the log level variables have been dropped?
6. If the answer to 5 is not short-run but long-run coefficients, does this mean that LR coefficients are not strictly refer to equation on level variables?
1. In specifying the unrestricted ECM equation for the PSS bounds test, is it proper to declare of the one-period lagged level variables under the list of fixed regressors of the ARDL model?
2. Is it accurate to say that the PSS bounds test just affirms whether cointegration exists but does not constrain the final specification to be just in the first difference form (growth rate, since level is in log)?
3. If we need to break the full sample into two sub-periods (i.e, Pre-IT and IT), is it necessary to establish cointegration per sub-period as well?
4. If individual coefficients in the unrestricted ECM format are insignificant but the F-test exceeds the upper bound of the PSS critical value, can we proceed to paring down the equation to log first difference form?
5. If the final equation is in log difference form, does the level equation reported after clicking the long-run and bounds test refer to short-run coefficients only, considering that the log level variables have been dropped?
6. If the answer to 5 is not short-run but long-run coefficients, does this mean that LR coefficients are not strictly refer to equation on level variables?