Encompassing

FAILURE TO ENCOMPASS: I want to find determinants of Inflation. Previous researchers have used variables V,W, but no one has used T. Therefore I will now run a regression of Inflation on V,W, and T.

This is not correct. We have to show that there is some DEFECT in earlier models which can be fixed by the use of T. In other words, putting in T IMPROVES performance of earlier models. A comparison is required to show that of all the models proposed, the one that I am proposing is the best, in some sense.

A good illustration of the encompassing approach is a thesis on Determinants of Inflation by my student M. Rehman:

SEE Also Lecture 12: Introduction to Hendry Methodology, in my course on Applied Econometrics at PIDE.

WRONG APPROACH: I wish to find out the contribution of Energy Prices to inflation, therefore I will run a regression with inflation as dependent variable and energy prices, together with other suitable exogenous variables on the right hand side. I will use very fancy estimation methods like GMM to get required results.

PROBLEM WITH THIS APPROACH: There is a huge literature on determinants of inflation which this approach neglects. The idea of encompassing is that our research should take into account, and be able to explain, ALL EXISTING RESEARCH.

ENCOMPASSING: Rehman first surveys ALL monetary models that have been estimated for Pakistan. From this, he gets the list of all the variables that have ever been used in explaining inflation. Then he builds a big model which includes all of the existing models as special cases. With this big model, it is possible to compare the performance of all existing models. Finally he comes up with a best model, and shows that it is better than all existing models in terms

SIMILARLY -- Munazza Jabeen has use an encompassing approach in her thesis on Modelling Exchange Rate Volatility. Again, she reviews all the models in the literature, and then creates an encompassing model which allows for comparisions of all existing models and selection of a best one.

For more material on the encompassing approach to econometrics, see my lectures on the DHSY article, which basically originated the approach as a new methodology for doing econometrics.