26
Lecture 16 – Learning to Use Regression Analysis
We are now going to look at the procedure to be followed in doing an econometric study.
While the
general procedure is easy to outline, you should recognize that in practice econometrics is as much an art as
it is a science.
You will appreciate this more after we have encountered some of the problems which
frequently crop up in econometric analysis.
Having said this, let us now look at the general procedure to be
followed in doing regression analysis.
1) The first step in undertaking a regression project is to determine the problem
you are interested in
studying.
It is very important that you select a problem that you are interested in.
This may seem obvious,
but you would be surprised how many people choose independent projects and senior theses in areas in
which they have little interest or knowledge.
Do a literature search.
Once you have selected your area of research, you must undertake a study to see
what work has previously been done.
After all, we don't want to keep reproducing the same thing!
As your
text notes, a good place to begin is with the
Journal of Economic Literature
.
However, many of you have
only taken introductory economics courses up to now, and many of the articles in the
Journal of Economic
Literature
may be quite advanced.
So it is perfectly reasonable for you to rely on the textbooks and
readings you have used in your courses.
2) The second step in regression analysis is also the most important.
This is where we specify the model
to
be analyzed.
Model specification involves four things (I am combining steps 2 and 3 from your text here).
a) First, you must specify which independent variables to include
.
As we will see later on, there are costs
both to including variables that should not be included, and in leaving out variables that should be included.
The fact of the matter, however, is that we usually do not know in advance what the correct theory is, so it
will always be a matter of judgment as to which variables should be included.
When economists make
such judgments, it is said that they are imposing their own prior beliefs.
While it is often impossible to
avoid making Priors, one must attempt to justify these priors, and one must recognize that they may be
incorrect.
b) Second, you must specify the functional form of the equation
.
Is it a linear equation, a quadratic, a log
equation, etc.
We will study this question in more detail later.
c) Third, you must specify the expected signs of the coefficients
.
Do the variables move in the same
direction, in which case we expect a positive sign, or do they move in opposite directions, in which case we
expect a negative sign.