S&W, Chapter 6

Solution Exercises
 Exercise 6.1
By equation (6.15) in the text, we know:
𝑛−1
𝑅ത 2 = 1 − 1 − 𝑅2
𝑛−𝑘−1
7440−1
(1) 𝑅ത 2 = 1 − 7440−2−1 1 − 0.162 = 0.162

7440−1
(2) 𝑅ത 2 = 1 − 7440−3−1 1 − 0.180 = 0.180

7440−1
(3) 𝑅ത 2 = 1 − 7440−6−1 1 − 0.182 = 0.181
 Exercise 6.2

(a) Workers with college degrees earn $8.31/hour more, on average, than workers with
only high school degrees, holding the other variable fixed.

(b) Men earn $3.85/hour more, on average, than women holding the other variable
fixed.
 Exercise 6.3
(a) On average, a worker earns $0.51/hour more for each year he ages holding the other
variable(s) fixed.

෣ = 1.87 + 8.32 ∙ 1 − 3.81 ∙ 1 + 0.51 ∙ 29 = 21.17

(b) Sally’s earnings prediction is: 𝐴𝐻𝐸
෣ = 1.87 + 8.32 ∙ 1 − 3.81 ∙ 1 + 0.51 ∙ 34 = 23.72
Betsy’s earnings prediction is: 𝐴𝐻𝐸
=> The expected difference is 2.55
 Exercise 6.4
(a) Workers in the Northeast earn $0.18 more per hour than workers in the West, on
average, controlling for other variables in the regression. Workers in the Midwest
earn $1.23 less per hour than workers in the West, on average, controlling for
other variables in the regression. Workers in the South earn $0.43 less per hour
than workers in the West, on average, controlling for other variables in the
regression.

(b) The regressor West is omitted to avoid perfect multicollinearity. If West is

included, then the intercept can be written as a perfect linear function of the four
regional regressors. Because of perfect multicollinearity, the OLS estimator cannot
be computed.

(c) The expected difference in earnings between Juanita and Jennifer is:
−0.43 − (−1.23) = 0.8
 Exercise 6.5
෣ = 109.7 + 0.567𝐵𝐷𝑅 + 26.9𝐵𝑎𝑡ℎ + 0.239𝐻𝑠𝑖𝑧𝑒 + 0.005𝐿𝑠𝑖𝑧𝑒 + 0.1𝐴𝑔𝑒 − 56.9𝑃𝑜𝑜𝑟
𝑃𝑟𝑖𝑐𝑒
𝑅² = 0.85 𝑆𝐸𝑅 = 45.8

(a) $26900 (recall that Price is measured in $1000s)

(b) In this case ∆𝐵𝑎𝑡ℎ = 1 and ∆𝐻𝑠𝑖𝑧𝑒 = 80. The resulting expected change in price is
26.9 + 0.239 ∙ 80 = 46.02 thousand dollars or $46020.

(c) The loss is $56900.

𝑛−1 200−1
(d) From the text 𝑅ത 2 = 1 − 𝑛−𝑘−1 (1 − 𝑅²), so 𝑅ത 2 = 1 − 200−6−1 1 − 0.85 = 0.845.
 Exercise 6.6
The researcher wants to investigate the effect of having a strong legal system on the
national income per capita. They estimate the following:
෣
𝑁𝑎𝑡 𝐼𝑛𝑐 = 𝛽መ0 + 𝛽መ1 𝑆𝑡𝑟_𝐿𝑒𝑔𝑎𝑙

(a) A large number of factors affect national income, in addition to institutions such as
a strong legal system. Other missing variables could include educational workforce,
the level of capital stock, the level of technological development and so on.

(b) We expect a positive correlation between national income and having a strong legal
system.
1. corr(educational workforce, strong legal system) > 0
 Countries with a very educated workforce will be more likely to
have a strong legal system due to more educated individuals and a
higher level of respect for the rule of law.
2. corr(educational workforce, national income) > 0
 A more educated workforce also increased national income per
capita.
 Exercise 6.6
 POSITIVE BIAS
!! You should look at both correlations to decide the direction of the bias, because the
corr(u,X) is determined by corr(Y,OV) and by corr(X,OV).
So: the direction of bias = sign of corr(u,X) = sign of [corr(Y,OV)*corr(X,OV)]

Corr(X,OV) > 0 Corr(X,OV) < 0

Corr(Y,OV) > 0 Positive bias Negative bias
Corr(Y,OV) < 0 Negative bias Positive bias

There is an overestimation of effect of having a strong legal system on national

income, because omitting educational workforce causes the positive effect of having
a strong legal system to become too positive  𝛽 ෢1 > 𝛽1

𝑝 𝜎𝑢
𝛽መ1 ՜ 𝛽1 + 𝜌𝑢,𝑋
𝜎𝑋
 Exercise 6.9
𝑌𝑖 = 𝛽0 + 𝛽1 𝑋1𝑖 + 𝛽2 𝑋2𝑖 + 𝑢𝑖

𝑌𝑖 , 𝑋1𝑖 , 𝑋2𝑖 satisfy the LSAs (see Key Concept 6.4, p.204)

For omitted variable bias to occur, two conditions must be true: 𝑋1 (the included
regressor) is correlated with the omitted variable, and the omitted variable is a
determinant of the dependent variable. Since 𝑋1 and 𝑋2 are uncorrelated, the
estimator of 𝛽1 does not suffer from omitted variable bias.
 Exercise 6.10
Given:
𝑣𝑎𝑟 𝑢𝑖 𝑋1𝑖 , 𝑋2𝑖 = 4
𝑣𝑎𝑟 𝑋1𝑖 = 6
𝑛 = 400

Formula:
1 1 𝜎²𝑢
𝜎²𝛽෡ = 𝑛
1 1−𝜌²𝑋1 ,𝑋2 𝜎²𝑋1

(a) 𝑐𝑜𝑟𝑟 𝑋1 , 𝑋2 = 0
1 1 4
𝜎²𝛽෡ = 400 = 0.00167
1 1−0² 6

(b) 𝑐𝑜𝑟𝑟 𝑋1 , 𝑋2 = 0.5

1 1 4
𝜎²𝛽෡ = 400 = 0.00222
1 1−0.5² 6
 Exercise 6.10 (ctd.)
(c) The statement correctly says that the larger is the correlation between X1 and X2
the larger is the variance of however the recommendation “it is best to leave X2 out
of the regression” is incorrect. If X2 is a determinant of Y, then leaving X2 out of the
regression will lead to omitted variable bias in 𝛽መ1 .