I'm estimating an ECM for panel data and I found X and Y are cointegrated (from various panel cointegration tests and the significant error correction coefficient in ECM). However, when estimating the long-run relationship between X and Y, the FMOLS, DOLS and the long-run coefficient in ECM were insignficant. Is there a possible reason for such results?
Thank you.
What is the indication of the (adequate) unit root tests first?
Issues of heterogeneity and cross sectional dependence have to be considered initially and then to proceed to the adequate (new generation) cointegration tests.
Hi George,
Thank you for your response.
Panel unit root test (Pesaran 2003) results suggest both X and Y are integrated of order 1. For panel cointegration tests, I used both Pedroni (1999) and Westerlund (2007) tests.
ECM insignificant means no long run relationship but if the coefficient of the first difference term of the explanatory variable is positive and statistical significant a short run relationship exists. Also look into the cointegration results. Full rank, partial rank etc.
- Badges
- Science topic
- Similar topics
- Correction