DUMMY VARIABLES
TANWEER UL ISLAM
BINARY VARIABLES
BINARY VARIABLES
Qualitative factors often come in the form of binary information
• A person is female or male
• A person does or does not own a personal computer
• A firm offers a certain kind of employee pension plan, or it does not
3
BINARY VARIABLES
Define a binary (or dummy) variable: 0 or 1
Choose the name of the variable wisely: e.g., female, not gender
1 𝑖𝑓 𝑓𝑒𝑚𝑎𝑙𝑒
𝑓𝑒𝑚𝑎𝑙𝑒 = ቊ
0 𝑖𝑓 𝑚𝑎𝑙𝑒
4
A SINGLE DUMMY INDEPENDENT VARIABLE
A SINGLE DUMMY INDEPENDENT VARIABLE
How do we incorporate binary information into regression?
In the simplest case, with a single dummy explanatory variable, we
just add it as an independent variable in the equation.
For example:
𝑤𝑎𝑔𝑒 = 𝛽0 + 𝛿0 𝑓𝑒𝑚𝑎𝑙𝑒 + 𝛽1 𝑒𝑑𝑢 + 𝛽2 𝑒𝑥𝑝𝑒𝑟 + 𝑢
In this regression males are the base group (or benchmark group)
6
A SINGLE DUMMY INDEPENDENT VARIABLE
𝛿0 measures the difference in wages between males and females
holding edu and exper constant:
𝛿0 = E 𝑤𝑎𝑔𝑒\female = 1, 𝑒𝑑𝑢, 𝑒𝑥𝑝𝑒𝑟 − 𝐸 𝑤𝑎𝑔𝑒\female = 0, 𝑒𝑑𝑢, 𝑒𝑥𝑝𝑒𝑟
Consider discrimination if 𝛿0 < 0 𝑡𝑒𝑠𝑡 𝐻0 : 𝛿0 = 0 𝑣𝑠 𝐻𝑎 : 𝛿0 < 0 .
The key here is that the level of education & experience is the same
in both expectations; the difference is due to gender only.
7
A SINGLE DUMMY INDEPENDENT VARIABLE
8
A SINGLE DUMMY INDEPENDENT VARIABLE
What is wrong with
𝑤𝑎𝑔𝑒 = 𝛽0 + 𝛿0 𝑓𝑒𝑚𝑎𝑙𝑒 + 𝛿1 𝑚𝑎𝑙𝑒 + 𝛽1 𝑒𝑑𝑢 + 𝛽2 𝑒𝑥𝑝𝑒𝑟 + 𝑢
Including dummy variables for both genders is the so-called dummy
variable trap.
9
TESTING FOR WAGE DISCRIMINATION
Formulate hypothesis and estimate the model:
𝐻0 : 𝛿0 = 0 𝑣𝑠 𝐻𝑎 : 𝛿0 < 0
𝑤𝑎𝑔𝑒
ෟ = −1.57 − 1.81𝑓𝑒𝑚𝑎𝑙𝑒 + 0.57𝑒𝑑𝑢 + 0.025𝑒𝑥𝑝𝑒𝑟 + 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠
The negative intercept, intercept for men, is not meaningful since no
one has close to zero years of edu and exper in the sample.
If we take a woman and a man with same levels of edu & exper the
women earns, on average, $1.81 less per hour than the man.
10
A SINGLE DUMMY INDEPENDENT VARIABLE
Now, estimate:
𝑤𝑎𝑔𝑒 = 𝛽0 + 𝛿0 𝑓𝑒𝑚𝑎𝑙𝑒 + 𝑢
𝑤𝑎𝑔𝑒
ෟ = 7.10 − 2.51𝑓𝑒𝑚𝑎𝑙𝑒
What’s the interpretation of 𝛽0 & 𝛿0 ?
𝛽0 is the average wage for men in the sample ($7.10 per hour).
𝛿0 is the wage differential ($2.51 per hour). Average wage for women
is $4.59=$7.10-$2.51.
Which regression gives us more reliable estimates of wage
differential?
11
LOG REGRESSION
= 0.417 − 0.297𝑓𝑒𝑚𝑎𝑙𝑒 + 0.08𝑒𝑑𝑢 + 0.029𝑒𝑥𝑝𝑒𝑟 + 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠
𝑙𝑛𝑤𝑎𝑔𝑒
What’s the interpretation of 𝛿0 ?
The coefficient on 𝛿0 implies that for the same levels of edu, exper,
etc., women earn about 100(0.297)=29.7% less than the men.
For the log-regression, the exact percentage differential can be
computed as:
100*[exp(𝛿0 )-1]=100*[exp(-0.297)-1] = -25.7%
12
DUMMY VARIABLE WITH CATEGORIES
DUMMY VARIABLE WITH CATEGORIES
Consider:
𝑙𝑛𝑤𝑎𝑔𝑒
= 0.321 + 0.213𝑚𝑎𝑙𝑒𝑚 − 0.198𝑓𝑒𝑚𝑎𝑙𝑒𝑚 − 0.11𝑓𝑒𝑚𝑎𝑙𝑒𝑠 + 0.079𝑒𝑑𝑢
+ 0.027𝑒𝑥𝑝𝑒𝑟 + 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠
Four groups: married men, married women, single men, single women
Base group: single men
14
DUMMY VARIABLE WITH CATEGORIES
𝑙𝑛𝑤𝑎𝑔𝑒
= 0.321 + 0.213𝑚𝑎𝑙𝑒𝑚 − 0.198𝑓𝑒𝑚𝑎𝑙𝑒𝑚 − 0.11𝑓𝑒𝑚𝑎𝑙𝑒𝑠 + 0.079𝑒𝑑𝑢
+ 0.027𝑒𝑥𝑝𝑒𝑟 + 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠
Married men earn 21.3% more than the single men.
Married women earn 19.8% less than the single men.
The differential between married and single female is -0.11-(-0.198)=8.8%.
15
ORDINAL DUMMY VARIABLES
ORDINAL DUMMY VARIABLES
Hamermesh & Biddle (1994) used physical attractiveness in a wage equation. For
regression, they used three categories: average, below average & above average.
For Men:
= 𝛽መ0 − 0.164𝐴𝑣𝑔𝑏𝑒𝑙𝑜𝑤 + 0.016𝐴𝑣𝑔𝑎𝑏𝑜𝑣𝑒 + 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠
𝑙𝑛𝑤𝑎𝑔𝑒
For Women:
= 𝛽መ0 − 0.124𝐴𝑣𝑔𝑏𝑒𝑙𝑜𝑤 + 0.035𝐴𝑣𝑔𝑎𝑏𝑜𝑣𝑒 + 𝑜𝑡ℎ𝑒𝑟 𝑓𝑎𝑐𝑡𝑜𝑟𝑠
𝑙𝑛𝑤𝑎𝑔𝑒
ORDINAL DUMMY VARIABLES
For men, those with below average looks are estimated to earn 16.4%
less than an average looking man (t=-3.57).
For men, those with above average looks are estimated to earn 1.6%
more than an average looking man (t=0.5).
A women with below average looks earns about 12.4% less than an
otherwise comparable average looking woman (t=-1.88).
INTERACTIVE DUMMY VARIABLES
INTERACTIVE DUMMY VARIABLE
Consider
= 0.321 − 0.11𝑓𝑒𝑚𝑎𝑙𝑒 + 0.213married − 0.301female ∗ married + …
𝑙𝑛𝑤𝑎𝑔𝑒
There is a statistically significant interaction between gender and
marital status.
What’s the base group?
What’s the wage differential for married & single men?
SLOPE DUMMY
SLOPE DUMMY
Dummy variables may interact with explanatory variables which are
not dummy to allow for differences in slopes
= 0.9 − 0.23𝑓𝑒𝑚𝑎𝑙𝑒 + 0.08𝑒𝑑𝑢 − 0.0056edu ∗ female+. . .
𝑙𝑛𝑤𝑎𝑔𝑒
The estimated return to education for men is 8.0%.
For women, its 7.6% (=0.8-0.0056=0.076).
The differential of 0.56% is insignificant (t=-0.43). Thus, return to
education are same for both men and women.
THANK YOU
tanweer@[Link]
[Link]