Hi Community, I am trying to estimate Foreign trade elasticities using two approaches for cointegration relationships: DOLS and ARDL. Yet, the results of the coefficients are very VERY different. Coefficients from ARDL are half in the magnitude of those from DOLS. Supposedly, ARDL is more accurate than DOLS because there is no need that all variables to be I(1), but the results that I get using ARDL are really rare (magnitude and significance level). On the other hand, DOLS results are quite convincing in magnitude, sign and significance level. May does someone help me to understand why this is happening? Sample 1997q1-2016q4, dependent variables (2 different equations): real exports and real imports, explanatory variables: trade partners gdp (for exports), domestic gdp (for imports) and reer. All variables are I(1). Thank you for your time!
DOLS and ARDL methods employ different estimation models. For example,DOLS permits leads as well as lags. Hence, getting the same parameter estimates from both methods is very unlikely.
Thank you very much for your answer. 1) So, if I have all the variables I(1), do you recommend to employ DOLS or ARDL? Because theoretically, both of them are used for the same kind of relationship (cointegration) among variables. 2) Another issue that arises here is that DOLS commands (Stata and Eviews), do not facilitate the error correction term coefficient. Could you suggest some solution to that?
1) Florencia asks "DOLS or ARDL?". I recommend first examining the variables individually to assess whether they are, or are not, all I(1). This should inform the choice of method.
2) Florencia asks "How to get the error correction coefficient in DOLS?". DOLS is designed to accurately estimate only the coefficients in the cointegrating equation. DOLS does not attempt to estimate the complete error-correction model.
- Badges
- Science topic
- Similar topics
- Econ
- Manufacturing
- Industry