technical issues with a VECM for testing the J-curve
Posted: Mon Dec 21, 2015 5:12 am
Hello everyone,
I am attempting to test for the existence of a "J-curve" effect (deterioration of the trade balance in the short-term after the depreciation of the currency via more expensive imports, followed by an improvement in the medium/long-run via an increase in exports) for South Africa and am running into some difficulty.
Regarding the data I am using:
Quaterly, 1994Q1-2015Q3 (87 observations)
Trade balance defined as the ratio of Exports over Imports, natural logged
Real effective exchange rate
some Index for OECD income: I summed the real GDP of Japan, UK, USA, Germany, all from the same source
real GDP of SA
real GDP of China
All series are natural logged.
I am using Eviews 9
All my variables are I(1) so I'm following a pretty standard Johansen method to check for co-integration.
So: First when I estimated the VAR, I put all series as endogenous and used 4 lags. However I am not sure whether this is the right thing to do. Intuitively, I would think that only the trade balance and the exchange rate "feed" into each other, and there are even reasons to think that the exchange rate could be exogenous for South Africa.
Should I include all the series as endogenous or should I put the ones I think are endogenous as endo and the others as exogenous, for the lag selection process?
Also someone told me to estimate the VAR in first difference, but I read somewhere else that it shouldn't matter for lag selection, which option is more correct?
I chose the AIC index that indicated 2 lags (still a bit quick in my opinion).
I am also hesitant between using an VECM and UECM. The way I understand, the VECM is useful if I also want to check the effects of the depreciation on the GDP series that I have via the trade balance, but if I only want to observe the effects of the change in REER on the trade balance, then UECM should be enough.
Is this correct?
In the literature, I have seen both being used. I am also not sure how to run an UECM on eviews, and have only found tutorials for VECMs on youtube.
Last question, I estimated the VECM to try and put all the series as endogenous once again. Is this the correct way to do it?
When I tried to copy paste the equation for the Trade balance, to "estimate the equation" on eviews, I got an error message saying: (illegal lag or index specification for series A(1,1), any ideas why?
Thanks in advance for anyone's time, and if my questions are not clear I'll try to specify them asap.
I am attempting to test for the existence of a "J-curve" effect (deterioration of the trade balance in the short-term after the depreciation of the currency via more expensive imports, followed by an improvement in the medium/long-run via an increase in exports) for South Africa and am running into some difficulty.
Regarding the data I am using:
Quaterly, 1994Q1-2015Q3 (87 observations)
Trade balance defined as the ratio of Exports over Imports, natural logged
Real effective exchange rate
some Index for OECD income: I summed the real GDP of Japan, UK, USA, Germany, all from the same source
real GDP of SA
real GDP of China
All series are natural logged.
I am using Eviews 9
All my variables are I(1) so I'm following a pretty standard Johansen method to check for co-integration.
So: First when I estimated the VAR, I put all series as endogenous and used 4 lags. However I am not sure whether this is the right thing to do. Intuitively, I would think that only the trade balance and the exchange rate "feed" into each other, and there are even reasons to think that the exchange rate could be exogenous for South Africa.
Should I include all the series as endogenous or should I put the ones I think are endogenous as endo and the others as exogenous, for the lag selection process?
Also someone told me to estimate the VAR in first difference, but I read somewhere else that it shouldn't matter for lag selection, which option is more correct?
I chose the AIC index that indicated 2 lags (still a bit quick in my opinion).
I am also hesitant between using an VECM and UECM. The way I understand, the VECM is useful if I also want to check the effects of the depreciation on the GDP series that I have via the trade balance, but if I only want to observe the effects of the change in REER on the trade balance, then UECM should be enough.
Is this correct?
In the literature, I have seen both being used. I am also not sure how to run an UECM on eviews, and have only found tutorials for VECMs on youtube.
Last question, I estimated the VECM to try and put all the series as endogenous once again. Is this the correct way to do it?
When I tried to copy paste the equation for the Trade balance, to "estimate the equation" on eviews, I got an error message saying: (illegal lag or index specification for series A(1,1), any ideas why?
Thanks in advance for anyone's time, and if my questions are not clear I'll try to specify them asap.