Panel Error Correction model forecating
Posted: Thu Oct 31, 2019 9:57 am
Hi,
I estimated a panel ECM with fixed effects, using FMOLS for the long run dynamics, and then using OLS for the short term dynamics. I use eviews 9.5, therefore I estimated the long run relation using the following estimation command: COINTREG(PANMETHOD=GROUPED) LOG_NPL_CONS LOG_LAG_LEV_CONS LOG_PIB.
Then, I went to Proc, and generated a residual series. This residual is the one I use to estimate the EMC equation of the following estimation command:
LS(CX=F) D(LOG_NPL_CONS) C D(LOG_NPL_CONS(-1)) D(LOG_LAG_LEV_CONS) D(LOG_PIB) RESID_FOLS_PRUEBA(-1).
The restricted ECM could be written as:
D(LOG_NPL_CONS) = 0.00197766354072 + 0.269590701578*D(LOG_NPL_CONS(-1)) + 0.539345237869*D(LOG_LAG_LEV_CONS) - 0.975875372066*D(LOG_PIB) - 0.0424886489025*[ LOG_NPL_CONS - (0.537641531923*LOG_LAG_LEV_CONS - 0.652596322867*LOG_PIB )]
Up to here is fine. But now I'd like to do out of sample forecast, say for example 8 periods ahead. Is there any easy ways to build it?
kind regards.
I estimated a panel ECM with fixed effects, using FMOLS for the long run dynamics, and then using OLS for the short term dynamics. I use eviews 9.5, therefore I estimated the long run relation using the following estimation command: COINTREG(PANMETHOD=GROUPED) LOG_NPL_CONS LOG_LAG_LEV_CONS LOG_PIB.
Then, I went to Proc, and generated a residual series. This residual is the one I use to estimate the EMC equation of the following estimation command:
LS(CX=F) D(LOG_NPL_CONS) C D(LOG_NPL_CONS(-1)) D(LOG_LAG_LEV_CONS) D(LOG_PIB) RESID_FOLS_PRUEBA(-1).
The restricted ECM could be written as:
D(LOG_NPL_CONS) = 0.00197766354072 + 0.269590701578*D(LOG_NPL_CONS(-1)) + 0.539345237869*D(LOG_LAG_LEV_CONS) - 0.975875372066*D(LOG_PIB) - 0.0424886489025*[ LOG_NPL_CONS - (0.537641531923*LOG_LAG_LEV_CONS - 0.652596322867*LOG_PIB )]
Up to here is fine. But now I'd like to do out of sample forecast, say for example 8 periods ahead. Is there any easy ways to build it?
kind regards.