I am analyzing the effect of yearly ln(electricity consumption) on ln(GDP). Both are cointegrated. I need forecasts for around 20 years and am thus just interested in the long-term relationship (or do I need to consider short-term?).
I am confused, as on some pages I read that OLS is fine (over-consistent), some say you need an ECM (error correction model), but now I also read something about FMOLS and DOLS. Are the last two just for multiple independent variables?
Which method is easy to get the long-term relationship?
Depends if the variables you have are I(0) or I(1) and if they cointegrate, the presence of cointegration shows that there is long run, and you can perform ARDL for the short run and ECM or VECM based on the results from the bound test . I hope I was helpful, regards Laura Warnecke
You can examine short-run causality between the variables along with their long-run cointegrating relationship. Estimation of ECM will help you determine the rate of adjustment towards the equilibrium value in the long-run.
20 years is too long for either your simple OLS or Vecm analysis. You can get a form of the forecast using the elasticity from your regression and your estimated growth of GDP over the next 20 years. You then need to add on the effect of switching from carbon-based fuels to electricity. There are also likely to be new industries (e.g. data centers) using large amounts of electricity. You need to be setting out scenarios and evaluating the likely electricity demand under each scenario. I am a little surprised that you have found a stable relationship with your regression.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis