You are on page 1of 21

Business Econometrics

Session VII-VIII
Dr Tutan Ahmed
IIT Kharagpur
February 2020
Topics we cover today (Dummy Variable)…..

• Interpretation of dummy variable in a regression equation

• Concept of intercept dummy

• Some properties and usages of dummy variables

• Dummy variable trap

• Slope dummy

• Reduced form equation

• Chow Test
References

• Dummy Variable: Text- Chapter 5 (5.1,5.2,5.3,5.4) of Christopher Dougherty


and reference - Chapter 3 ( 3.1, 3.2, 3.3) of Gujarati
Several Steps to understand Dummy Variable

• Dummy variable with binary categories (Gender, Urban-Rural)

• Dummy variable with more than two categories (region,


religion, social group etc)

• Multiple Dummy variables in a regression equation (all the


above dummies in one equation)

• Interaction of the Dummy variables (females in rural vs. females


in urban)

• Interaction of dummy and other non-qualitative explanatory


variables
Qualitative explanatory variable: Basic characteristics

• First, if an intercept is included in the model and if a qualitative variable has m categories, then
introduce only (m – 1) dummy variables.

• the category that gets the value of 0 is called the reference, benchmark or comparison category

• if there are several dummy variables, the reference category must be ; otherwise, it will be difficult
to interpret the results.

• since dummy variables take values of 1 and 0, we cannot take their logarithms.

• if the sample size is relatively small, do not introduce too many I dummy variables. Remember that
each dummy coefficient will cost one degree of freedom.
Mincerian Wage Regression Equation and
Dummy Variables

• Male specific wage regression

• Wage regression with gender dummy

• Female Specific wage regression

• Example using National Sample Survey Data in Bengal

• Example using CPS data in the United States (from Gujarati dataset)

Refer to the codes


Intercept Dummy

• Concept of intercept dummy: These dummy coefficients are often called


differential intercept dummies, for they show the differences in the
intercept values of the category that gets the value of 1 as compared to the
reference category.

• Why do we have a negative intercept value? - mechanical interpretation of


intercept
Dummy Variable Trap

• What would happen if you included a dummy variable for the reference category?

• - (i.e. all m categories of dummy instead of m-1 categories).

• Two possible outcome

• Two ways of avoiding DV Trap: dropping the constant term or having m-1 categories o
dummy. Which one we should choose and why?

See Dougherty for details


Dummy Variable Trap

• m-1 dummy rule is in order to avoid dummy variable trap

• Dummy variable trap is a situation of perfect collinearity

• When all the dummies are included, their sum is 1 which


is exactly leads to a relationship D1 = 1 - D2

• The concept of reference dummy and the importance of


comparison

• The role of intercept in the equation form of dummy variable


Types of usages of dummy in a regression equation:
Explanations

• Case 1: For example, gender has only two categories; hence we introduce
only one dummy variable for gender
• Case 2: Social Group has 4 categories: we can introduce three dummy for
social group
• Case 3: In the regression equation, there will be one dummy for gender and
three dummies for social groups
• Case 4: Introduce interaction terms for the dummies
• Case 5: Introduce interaction terms between dummy variables and other
explanatory variables
Multiple sets of dummy variables (from US
Population Survey Data) - Case II
Multiple sets of dummy variables (from US
Population Survey Data)

• What would be the wage equation for male nonwhite non-union workers?

• W(estimated) = 1.37*Edu + .167*exper + 0 (union) + 0 (female) + - 1.57


(nonwhite) - 7.2 (constant)

• W (estimated) = 1.37*Edu + .167*exper - 8.77

• What would be the wage equation for female white union workers?
Dummy with more than two categories (case
III)

• Base dummy (reference dummy) and separate dummy for each of the
categories within the dummy

• SC, ST, OBC (with general as the base dummy), Political view (Right,
Center) with Left as base dummy

• Which one is to chosen as the base dummy? - choice of the researcher but
in general a well known category

• Can we incorporate all the different categories of a given dummy?


Slope Dummy or Interactive Dummy Variable
(Case IV)
We have implicitly assumed that the slope of the regression line is the same for
each category of the qualitative variable with only difference in the intercept as if
there is a translational movement of the regression lines

• However, we have also seen the coefficients of other explanatory variables are
not the same in all types of equations (male specific, female specific, equation
with dummy variable)

• To do that, we introduce a product of a dummy variable and another explanatory


variable: Gender and Education (in this case) or Gender and Experience

• Corresponding rotational movement of the regression line (diagram - next slide)

• Source: Dougherty
Slope Dummy (Interactive Dummy Variable)

Source: Introduction to Econometrics, Stock & Watson, pp-280


Slope Dummy (Interactive Dummy Variable)

• Can we have a slope dummy with two dummy variables? (Case V)

• Illustration
Slope Dummy (Interactive Dummy Variable)
from Dougherty

Can we explain the previous differences in the coefficients with the concept of
slope dummy?
Reduced form Equations

• Different examples of the reduced form equations

• reduced form equation with only the intercept dummy -


examples

• reduced form equation using both slope and intercept


dummy - examples
Joint explanatory power of the intercept and
slope dummy variables

• Use of the previous variant of F-test: if we introduce


more variables and see if the explanatory power
The Chow Test

• Which subsample to be taken into account?

• Sample to population estimation - we did some filtering (dropping some


observations),

• you may further ask whether to drop observations which are below 15
years of age

• or whether to run it only for males (you know that female LFPR in India is
one of the lowest in the World),

• whether to drop observations above 59 years of age (because 59 is


considered to be the max age for people to be in the LF]

• or you may think of omitting the rural people Very few of them are salaried
The Chow Test

IMPROVEMENT = (RSS(P) - RSS(A) - RSS(B) )/ K / (RSS(A) + RSS (B) )/ (2n - 2K)

You might also like