□ VAR summary
- Unrestricted VAR models suffer from the problem of overparametrization. The Bayesian version (BVAR) has found to give better results and has a good forecasting record.
- Generally, there have been many tests suggested fr unit roots. In most of these tests, the null hypothesis is that there is a unit root, and it is rejected only when is a strong evidence against it.
Using these tests, most economic time series have been found to have unit roots. However, some tests have been devised that use stationary as the null and unit root as the alternative hyotheses. The evidence of unit roots in econimic time series appears to be fragile.
- The theory of co integration tries to study the interrelationships between the long run movements in econnomic time series. Most economic theories about the long run behavior, and thus much important information relevant for testing these theories is lost if the time series is de-trended or differenced, as in the Box-Jenkins approach, before any analysis is done.
- Co integration implies the existence of an error correction model (ECM). It also implies some restrictions on the VAR model.
- Tests for the co integration specify the null hypothesis as no co integration. This is because the unit root tests have the null hypothesis of unit root.
- Co integration theory has been used to get tge rational expectations hypothesis (REH) and the market efficiency hypothesis (MEH). The former rests on rejecting the hypothesis of no co integration and the latter on acceptance of this hypothesis. The results of these tests are sensitive to whether we consider bivariate or multi variate relationships. For ex, x and y may nkt be co integrated, but x, y and z may be co integrated.
Thank You
Aditya Pokhrel
MBA, MA Economics, MPA
No comments:
Post a Comment