ECONG301 Econometrics I
Dongwoo Kim
Assignment 4
Analytical Exercise
1. Using data on the GPA and personal characteristics of 500 students from a US university, you
want to estimate the following regression:
GP Ai = β0 +β1 W eeklyStudyHoursi +β2 EconM ajori +β3 W eeklyStudyHoursi ·EconM ajori +ui
where EconM ajori is a binary indicator which takes 1 if student i is econ-major and 0 otherwise.
(a) How do you interpret β2 ?
(b) How do you interpret β3 ?
(c) Suppose you estimated all the values of coefficients, (βˆ0 , βˆ1 , βˆ2 , βˆ3 ). Later, you realize that
students’ family income could also positively affect their college GPA. Students’ family
income is positively correlated with their weekly hours studied. If you include their family
income in your regression and re-estimate it, will the value of β̂1 increase or decrease
compared to the previous regression? Explain clearly why you think so.
(d) Now you decide to include an additional binary indicator for being a native English speaker
in your regression.
N ativei = 1 if student i is native English speaker, N ativei = 0 otherwise.
Being a native English speaker is not correlated with hours studied but you found non-
English speakers are more likely to major in economics. Being a non-native speaker has
a negative effect on the college GPA. If you include the new binary regressor in your
regression and re-estimate it, will the value of β̂2 increase or decrease compared to the
previous regression? Explain clearly why you think so.
2. Suppose that we have data on ECON 301 test scores (Yi ), duration for which student i studies
for exam (Xi ), and the major of the student, call it Di , where
1, if economics major
Di =
0, if non economics major
Consider the following model:
Yi = β0 + β1 Xi + β2 Di + β3 Di Xi + ui (1)
where Assumption 1 holds:
E (ui |Xi , Di ) = 0. (2)
Yi is the score between 0 and 100. Xi is the duration studied in hours, between 1 and 100.
(a) What is the intercept for econ-majors?
1
(b) What is the slope for econ-majors?
(c) What is the interpretation (in words) of β1 ?
(d) What is the interpretation (in words) for β3 ?
Now suppose that the OLS estimate for β1 is 0.60, and its standard error is 0.2.
(e) Test whether β1 = 0, at the 5% and 1% significance level.
Empirical Exercise
3. (Hamermesh and Parker, 2005, Economics of Education Review) Install an R package “AER”.
Type “[Link]("AER")” in your R console. Then type “library(AER)” to use the
package. The data set “TeachingRatings” can be now loaded using “data(TeachingRatings)”.
The data contains course evaluations, course characteristics, and professor characteristics for
463 courses for the academic years 2000-2002 at the University of Texas at Austin. Care-
fully read the data description which can be found by typing “help(TeachingRatings)” or
“?TeachingRatings”.
(a) Run a regression of course evaluation (eval) on instructor’s beauty rating (beauty). Briefly
explain the result.
(b) You may omit important variables that can explain course evaluation. First, compute the
correlation between beauty rating (beauty) and age (age). Suggest whether the effect of
age on course evaluation is positive or negative. Given the correlation and your guess on
the effect of age on course evaluation, explain whether the estimate of beauty rating in (a)
is overestimated or underestimated.
(c) Run a regression of course evaluation (eval) on instructor’s beauty rating (beauty) and age
(age). Does the estimate of the effect of beauty rating change from (a)? Explain why or
why not.
(d) Compare the regression results from (a) and (c). Does it improve the model fit? Do you
want to add age in your regression? Explain.
(e) Run a regression of course evaluation (eval) on instructor’s beauty rating (beauty), age,
gender, minority, native, tenure, division, and credits. Explain the results. Are the esti-
mated coefficients on included regressors in line with your expectation?
(f) Compare the results in (a) and (e). Is the model fit improved when you add more regressors?
Explain why or why not.
(g) The effects of beauty rating on course evaluation may differ between male and female
instructors. Further include the interaction between beauty rating and gender (beauty ∗
gender). Explain the results.