Methodology

Methodology describes HOW the proposed research is to be carried out. It is a common misunderstanding that “model” means regression model, and methodology means OLS (or some more complex regression estimation technique). In fact methodology is more comprehensive. The research question, which deals with solving some real world problem, must be kept in mind. Methodology refers to HOW we propose to SOLVE this problem? What kind of evidence is needed to resolve the issues raised in the research question. Ideally, we would like both historical and qualitative analysis of the problem, which attempts to answer the research question. Additionally, more formal analysis by statistical and quantitative methods would be useful. It is not necessary to run regressions; many more flexible statistical techniques for using the data to assess the research question are now available.

whereas the Model refers to a framework which will be used for data analysis. A regression model, which assumes a causal chain (dependent and independent variables), and certain types of functional forms (linear, log linear, error correction, etc.) is a framework. However, frameworks can also be more informal and qualitative, and the tendency is shifting in this direction. Methods for applying the model to real world phenomena need to be mentioned here. For regression models, estimation and evaluation techniques are needed – for example OLS estimates, and evaluation for various types of model misspecification. Other types of models may be calibrated against real data, or matched to data via graphical techniques without formal regressions.

There are many common econometric mistakes made in proposals submitted. I will provide a list of these mistakes, and how to recognize them, and how to fix them here. Most textbooks do not provide adequate guidance about these mistakes, which is why they are very common.

Common Econometric Errors:

1. The goal of research is to estimate a regression model.

2. Every question of the type: "what is the effect of X1, X2, X3 on Y?" makes sense and can be researched.

AVOIDING Nonsense Regressions:

This involves tracing the causal chain: If X causes Y which causes Z then a regression of Z on X is not correct. One has to draw a PATH DIAGRAM with the causal chain and use this to run the regression. If some of the variables on the path are not directly observable, then one must use Structural Equation Modelling methods.

Data Sources:

Some mention of data sources to be used for analysis of the research questions should be present within the proposal.