Homework 2 Questions

Note: Please respect intellectual property and do not make this document available in the
public domain

1. You would like to know how the years of education attained by individuals is
determined. In particular, you would like to know whether the average school fees
charged by schools in a person’s neighbourhood influences the years of education
attained by the person. To investigate this, you collect data on peoples’ education,
average fees charged by schools in the person’s neighbourhood, the distance of the
individual’s residence (home) to the nearest tertiary institution, the combined
income earned by the individual’s parents, the education of the individual’s father,
and the individual’s gender and race among others. The variable education
represents the total number of months of education received by the individual; fees
represents the average school fees charged per year by schools in the individual’s
residential neighbourhood, measured in thousands of dollars (In other words, a
value of $1,000 shows up as 1 in the dataset); fees2 denotes the square of fees;
high_income represents a binary variable equal to one if the individual has high
income parents (his/her parents have a combined income greater than or equal to
$80,000 per year) and equal to zero if the individual has low income parents (his/her
parents have a combined income of less than $80,000 per year); dad_grad
represents a binary variable equal to one if the individual’s father is a university
graduate and equal to zero if the individual’s father is not a university graduate;
female represents a binary variable equal to one if the individual is female and equal
to zero if the individual is male. minority represents a binary variable indicating
whether the student belongs to a minority race (equal to 1 if the individual belongs
to a minority race and equal to zero if the individual belongs to the majority race).
distance is a continuous variable representing the distance from an individual’s
home to the nearest tertiary institution, measured in kilometres.

You log transform distance (using natural logs) and you call this variable lndistance.
Further, you interact the variables female and high_income and you call this variable
int_fhighinc.

You then run the regression below:

1
. regress education fees fees2 high_income dad_grad female minority lndistance int_fhighinc, robust

Linear regression Number of obs = 2,473

F(8, 2464) = 40.75
Prob > F = 0.0000
R-squared = 0.1139
Root MSE = 20.237

Robust
education Coef. Std. Err. t P>|t| [95% Conf. Interval]

fees 37.12135 9.735449 3.81 0.000 18.03084 56.21186

fees2 -19.67065 5.288026 -3.72 0.000 -30.04009 -9.301219
high_income 3.46022 1.380232 2.51 0.012 .7536847 6.166755
dad_grad 13.62843 1.142116 11.93 0.000 11.38883 15.86804
female -.0233956 .9559493 -0.02 0.980 -1.897943 1.851151
minority -3.665875 .9077404 -4.04 0.000 -5.445888 -1.885862
lndistance -.7379352 .3640536 -2.03 0.043 -1.451818 -.0240526
int_fhighinc 2.358989 1.842266 1.28 0.200 -1.253561 5.971539
_cons 145.3011 4.240155 34.27 0.000 136.9865 153.6157

You then performed a test using the command “test high_income int_fhighinc” and
obtained the following result:

. test high_income int_fhighinc

( 1) high_income = 0
( 2) int_fhighinc = 0

F( 2, 2464) = 11.71
Prob > F = 0.0000

Use this information to answer Questions 1 to 12. Be careful of units.

Person A is male, with high income parents, whose father is a university graduate, who
belongs to the majority race, who lives 0.2km from the nearest tertiary institution, and
whose average school fees charged per year by schools in the residential neighbourhood
amounts to $889.15. Person B is female, with low income parents, whose father is not a
university graduate, who belongs to the majority race, who lives 0.5km from the nearest
tertiary institution, and whose average school fees charged per year by schools in the
residential neighbourhood amounts to $900. Round your final answer to the following
question to 1 decimal place and choose the closest option.

Complete the statement. Person A’s education is predicted to be:

2. Trump is male, with low income parents, whose father is a university graduate, who
belongs to the majority race, who lives 0.6km from the nearest tertiary institution, and
whose average school fees charged per year by schools in the residential neighbourhood
amounts to $931.87. Trump’s has 192 months of education. What is the residual
specific to Trump? Round your final answer to 2 decimal places and choose the closest
option.

2
3. Which of the following statements is incorrect? The regression output implies that:

4. Refer to the regression in question 1. What would the estimated coefficient on the
variables fees and fees2 be if the variable fees were now representing the average
school fees charged per year by schools in the individual’s residential neighbourhood,
and measured in hundreds of dollars?

5. What is the adjusted R-squared for the regression in Question 1? Round your answers to
4 decimal places. Pick the closest option.

6. What would the root mean squared error of the regression from Question 1 be if we
divided all values of education by 2.5, before running a regression of this modified
education variable on all the regressors that appear in the regression in Question 1?
Round your final answer to 3 decimal places.

7. Refer to the regression in Question 1 again. Had you removed the binary variable
minority from that regression, your estimated regression would have been:
. regress education fees fees2 high_income dad_grad female lndistance int_fhighinc, robust

Linear regression Number of obs = 2,473

F(7, 2465) = 43.58
Prob > F = 0.0000
R-squared = 0.1084
Root MSE = 20.295

Robust
education Coef. Std. Err. t P>|t| [95% Conf. Interval]

fees 36.85774 9.766442 3.77 0.000 17.70646 56.00902

fees2 -19.05966 5.302802 -3.59 0.000 -29.45807 -8.661255
high_income 3.769938 1.383674 2.72 0.006 1.056655 6.48322
dad_grad 14.11593 1.136272 12.42 0.000 11.88779 16.34408
female -.1517545 .9596826 -0.16 0.874 -2.033622 1.730113
lndistance -.5243943 .36021 -1.46 0.146 -1.23074 .1819511
int_fhighinc 2.49199 1.849202 1.35 0.178 -1.13416 6.118139
_cons 143.794 4.233083 33.97 0.000 135.4933 152.0948

Which of the following statements is correct?

8. Refer to the regression in Question 7, where the variable minority is excluded as a

regressor. Suppose that the true causal effect of being of a minority race is to decrease
educational attainment (perhaps because of racial discrimination). Then what do the
regression results in Question 1 and the regression results in Question 7 together imply?

3
9. Refer to the regression output in Question 1 again. Suppose you are now thinking of
creating another binary variable which is equal to one if the individual has low income
parents (his/her parents have a combined income of less than $80,000 per year) and
equal to zero if the individual has high income parents (his/her parents have a combined
income of greater than or equal to $80,000 per year). You call this binary variable
low_income. If you ran the same regression in question 1, but replace the variable
high_income with the variable low_income, and replace the variable int_fhighinc with
the variable int_flowinc, where int_flowinc is obtained by interacting the variables
female and low_income. what would the coefficients and standard errors on
low_income and int_flowinc be?

10. Refer to the regression output in Question 1 again. Which of the following statements is
true?

11. Refer to the regression output in Question 1 again. Rounded to 2 decimal places, what is
the 79.6% confidence interval for the coefficient on the variable dad_grad?

12. Refer to the regression output in Question 1 again. Suppose you include a binary
variable mon_grad which is equal to one if the individual’s mother is a university
graduate and equal to zero if the individual’s mother is not a university graduate as a
regressor. And you obtain the following regression output:

. regress education fees fees2 high_income dad_grad female minority lndistance int_flowinc mom_grad, robust

Linear regression Number of obs = 2,473

F(9, 2463) = 41.55
Prob > F = 0.0000
R-squared = 0.1241
Root MSE = 20.125

Robust
education Coef. Std. Err. t P>|t| [95% Conf. Interval]

fees 34.75205 9.678221 3.59 0.000 15.77376 53.73034

fees2 -18.34573 5.266235 -3.48 0.001 -28.67244 -8.019029
high_income 2.709981 1.385492 1.96 0.051 -.0068692 5.426831
dad_grad 11.09158 1.254174 8.84 0.000 8.632234 13.55092
female 2.626507 1.558911 1.68 0.092 -.4304041 5.683418
minority -3.583195 .903216 -3.97 0.000 -5.354337 -1.812054
lndistance -.6673087 .3618306 -1.84 0.065 -1.376832 .0422148
int_flowinc -2.737482 1.829484 -1.50 0.135 -6.324969 .8500042
mom_grad 7.102339 1.376777 5.16 0.000 4.402579 9.802099
_cons 145.9511 4.20086 34.74 0.000 137.7135 154.1886

Suppose that you now conduct the following test:

4
 . test dad_grad = mom_grad

( 1) dad_grad - mom_grad = 0

F( 1, 2463) = 3.21
Prob > F = 0.0733

What do your results imply?

13. Please see Canvas (under Files -> Homework -> Homework 2). The file
pisa_2015_apr2023.dta contains data collected from students across the World through
the OECD’s Programme for International Student Assessment study. This is a large-scale
study conducted in 2015 comprising 15-year-old students from all across the world.

The variables in the dataset are defined as follows:

science is the student’s Science test score, immigrant is a binary variable equal to 1 if the
student was not born in the country of test and is equal to 0 if the student was born in the
country of test, outhours is a continuous variable measuring the number of hours that a
student spends per week doing self-study outside of school, escs is a continuous
composite variable measuring a student’s economic, social and cultural (i.e.
socioeconomic) status. The higher the level of escs, the higher a person’s socioeconomic
status. Finally, pared is a continuous variable measuring the years of education received
by the student’s father. Use this information to answer Questions 13 to 16.

Note that you may have to create new variables in stata before you can run some of the
regressions in the questions that follow.

Run a regression of the natural logarithm of Science test score, ln(science), on immigrant,
outhours, and escs. Holding immigration status and socioeconomic status constant, if
number of hours that a student spends per week doing self-study outside of school
increases from 20 to 23, which of the following is true?

14. Now, run a regression of the natural logarithm of Science test score, ln(science), on
outhours, immigrant, escs, as well as the natural logarithm of pared (i.e. ln(pared)).
Holding immigration status, socioeconomic status constant, and out of school self-study
constant, if parental education increases from 15 years to 16 years, which of the following
is true?

5
15. Now, run a regression of the natural logarithm of Science test score, ln(science), on
outhours, immigrant, escs, as well as pared and pared2. Conduct additional tests to
examine the following statements, if necessary. Which of the following statements is
correct?

16. Interact the variables immigrant and outhours and include this interaction term in the
regression specified in Question 15. What do the results tell you?

17. In this question (and question 18), we build on what we learnt about fixed effects in
Tutorial 5. Specifically, we are once again interested in the effects of alcohol taxes on
traffic fatalities. We address this question using data on traffic fatalities, alcohol taxes,
and other related variables for the 48 contiguous U.S. states for each of the five years
from 1982 to 1986. These data are contained in the dataset fatality_apr2023.dta, which
can be located in Canvas (under Files -> Homework -> Homework 2), where our main
variables of interest, fatalityrate is the number of annual traffic deaths per 10,000
people in the population in the state and beertax is the real tax on a case of beer in the
state, in dollars. year represents the year in which the fatality rate and the beer tax is
observed. Note that the dataset fatality_apr2023.dta is slightly different from the one
we saw in the tutorial.

Focus only on the data from the years 1983 and 1984. Create the first differenced
fatalityrate and beertax variables. Run a regression of the first differenced fatality rate
on the first differenced beer tax. Cluster your standard errors by state. What do the
results tell you? Choose the best option.

18. Return to the dataset mentioned in question 17. Instead of using data only from the
years 1983 and 1984, use data on all the years available (i.e. data on all years, from 1982
to 1986). Declare the dataset to be a panel. Use the xtreg command to run a state fixed
effects regression with the main dependent variable being fatalityrate and the main
regressor being beertax. Include the variables perinc (measuring per capita income in a
state in a year) and unrate (measuring unemployment rate in a state in a year) as
regressors. What do the results tell you?

19. In questions 19 and 20, we build on what we learnt about IV regression in Tutorial 5.
Specifically, you want to know how much a woman's labour supply changes when she
has an additional child. To learn about this, you will examine the data contained in the
dataset fertility_apr2023.dta, which can be found in Canvas (under Files -> Homework ->
Homework 2). The dataset contains information on married women aged 21-35 with at
least 2 children. Note that the dataset fertility_apr2023.dta is slightly different from the
one we saw in the tutorial.

6
 Your variables of interest are weeksworked, which represents the number of weeks
worked by a woman in a year, morekids which is a binary variable equal to 1 if a woman
had more than 2 children and equal to zero if a woman had only 2 children, and
samesex, which is a binary variable equal to 1 if the first two children are of the same
sex (boy-boy or girl-girl) and equal to 0 otherwise. Your dataset consists of people
belonging to 4 mutually exclusive and exhaustive race categories – white, black,
Hispanic, and “all other races”. The variable black is a binary variable equal to 1 if the
person is black and equal to zero otherwise. The other race variables are defined
similarly. agem1 represents the age of the woman.

Run a regression of weeksworked on morekids, using samesex as an instrument for

morekids. Include controls for race (black, hispan, othrace) and age of the woman. What
do your results tell you?

20. Refer to the regression in question 19. What is the value of the F-statistic from the first
stage regression?