Introduction to Econometrics
101
Technical Notes
Introduction to Econometrics
Lecture 10: Model design and specification
Introduction
This lecture is an illustration of the process that an investigator might go through in trying to specify and
estimate a simple econometric model using multiple regression techniques to analyse crosssection data. It
is important to stress that there is no single set of rules for carrying out this process which will produce the
right answer in every case. However it is important to design any investigation in a systematic way, to use
economic theory and common sense to suggest the hypotheses which you wish to test, and to make sure
that you have a complete record of the regressions which you ran and of their results. That means that if
you later have a bright idea about how to extend your original investigation, or if you get some additional
data, you do not need to go back to the beginning. You can keep this record in a notebook (as a scientist
would record details of a series of experiments), but it’s often easier just to ensure that any results which
you generate are written to a computer file before you go on to the next regression. Eviews makes it easy to
do this, since you can save the specification of any regression (as an
Equation
object) or the results (as a
Table
object) in your Eviews workfile.
The example used in the lecture is the problem of explaining individual earnings, using a dataset on a
random sample of 1000 US workers in 1988
1
. Although this is the same problem as illustrated in earlier
lectures, I have used a different dataset which includes older workers, and has more information on
exogenous variables. All the workers in this new dataset are male, so obviously we cannot use it to
investigate the differences between male and female wage levels. Instead the illustration focusses on a more
detailed investigation of the role of schooling and educational attainment on wage determination. The
variables available in the dataset are
AGE
Individual’s age
WAGE
Hourly wage (in $)
LNWAGE
Log of wage
OCC1
Categorical variable for occupational category (see below)
IND1
Categorical variable for industrial category (see below)
UNION
1 if union member, 0 otherwise
GRADE
highest educational grade completed (112 are high school grades)
MARRIED
1 if married, 0 otherwise
PARTT
1 if parttime worker, 0 otherwise
POTEXP
Years of potential experience
EXPSQ
POTEXP squared
WEIGHT
Sampling weight
GRADSESQ
GRADE squared
FINHIGH
1 if graduated from high school, 0 otherwise (implies GRADE >= 12)
STCOL
1 if started college, 0 otherwise (implies GRADE > 12)
FINCOL
1 if completed college, 0 otherwise (implies GRADE >= 16)
DROPOUT
1 if dropped out of college, 0 otherwise (implies STCOL = 1, FINCOL = 0)
HSGRADE
GRADE for those who did not complete high school, 0 otherwise
HPOCC
1 if in highpaid occupation, 0 otherwise (see below)
LPIND
1 if in lowpaid industry, 0 otherwise (see below)
The categories for the occupational variable, OCC1 are:
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
This is the end of the preview.
Sign up
to
access the rest of the document.
 Spring '10
 Cowell
 Economics, Econometrics, Regression Analysis, dependent var, Akaike info criterion, technical notes, potexp

Click to edit the document details