Hello dear friends!
I am looking for the long relationships between exchange rate fluctuations and balance of payments. I have time series data for the variables of the model. Variables are BOP, REER, Interest rate, M3, Openness, Log Real GDP, Log gov expenditures. All are stationary at 1st order, except interest rate and BOP which are I(0). Bound test shows there is cointegration. In stata, I run ECM to see long run relationship, the problem is that the shown magnitude of the coefficients are very large. I don't know what is the problem? Can anyone help me?
Or more specifically, am I right in running such model? Or I should change the estimation method? Thanks
Look at the size of your dependent variable. I'm guessing it might be 6 or 7 digits in most countries, which would make coefficients in the thousands seem sensible.
I'll add small advice based on Prof. Miller answer to always re-check whether measurement units are consistent. For example, it happens that BoP might be measured in USD while nominal GDP is in the local currency, which might be not appropriate if you have missed controlling the point. It is besides the possibility to be confused by whether a series is measured in mln or bln or units like in your case.
Also, stationarity of the series is not the only indicator whether to take the difference or not. For example, interest rate differential works relatively better from the economics point of view in the particular case imho. Moreover, you might try to juggle with the REER and take smoothed (3months average) gap because the effect of REER change most likely might be belated when we're talking about its effect on BoP.
Hope that you'll find this answer helpful in your research.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running