You are on page 1of 3

PADM 2902 Regression and Introduction to Econometrics

Computer Exercise Class 5: Polynomials, Dummy Variables, and Joint F tests

Please download the do-file for this assignment, “Class 5 Computer Exercise do file.do,” from
the course website and perform all the required statistical operations as directed below.

Please submit:
a) A pdf of the Stata log (smcl) file documenting that you’ve completed each of the operations
directed below and showing the results. See the translate command in the downloaded do-
file. (NB: Be sure to run your correct Do-file in its entirety once all your commands are
working properly so your log file does not contain errors.)
b) A word or pdf write-up, which includes typed answers to the questions below.

Reminder: If you need a refresher on Stata, read the Stata Crib Sheet_Brief (on the course
website), or skim the Stata Crib Sheet _Long_ for more advanced programming.

Estimate the following four OLS regressions with total expenditures per pupil as the
dependent variable.

1. Independent variables:
 percent of the school’s students who are black;
 percent of the school’s students who are eligible for free lunches;
 total enrollment in the school.

a. Interpret the results on the coefficient for the percent of a school’s students who are
black.
Answer: holding other variables constant, one unit increase of a school’s students
who are black would decrease total expenditures per pupil by 10.41. This effect is
statistically significant.

2. Independent variables:
 percent of the school’s students who are black;
 percent of the school’s students who are eligible for free lunches;
 total enrollment in the school;
 year dummies (See the Exercise 5 do-file for models of how to do this. If you
remember an alternative way to do this from CORE-GP 1011, you may use that
method. Browse your data to make sure the dummy was properly created; see
cribsheet for a reminder of how to browse.)

a. Interpret the result on the coefficient for the percent of a school’s students who are black.
Does it substantively differ from the previous result? If so, why?
Answer: A school’s students who are black increase by one percent would decrease
total expenditures per pupil by 5.93. In this case, the omitted dummy variable “year”
was included and therefore affecting other variables. This effect is statistically
significant at 5% level.

1
PADM 2902 Regression and Introduction to Econometrics

b. If you include all six year dummies in the model, one of them will be dropped by Stata.
Why?
Answer: because Stata wants to avoid the collinearity which occurs when two or
more variables are exactly correlated.

3. Independent variables:
 percent of the school’s students who are black;
 percent of the school’s students who are eligible for free lunches;
 a polynomial in total enrollment in the school (See the Exercise 5 do-file for how to
create a polynomial, and be sure to use both totreg and totregsq);
 year dummies;
 a dummy variable if a school is a middle school.

a. Interpret the results on the coefficient for the percent of a school’s students who are
black.

Answer: Holding other variables constant, one unit increase of a school’s students who
are black would increase the total expenditures per pupil by 6.49. This effect is
statistically significant.

4. Independent variables:
 percent of the school’s students who are black;
 percent of the school’s student who are eligible for free lunches;
 a polynomial in total enrollment in the school;
 year dummies;
 a dummy variable if the school is a middle school;
 dummy variables for the boroughs of New York City.

a. If you include five borough dummies in the model, one of them will be dropped by Stata.
Why?
Answer: because Stata wants to avoid the collinearity which occurs when
two or more dummy variables are exactly correlated.

b. Interpret the results on the coefficient for the percent of a school’s students who are
black.

Answer: holding other variables constant, one unit increase of a school’s students who are
black would increase total expenditures per pupil by 8.04. This effect is statistically
significant.

c. Use test to see if the polynomial is statistically significant. Is it?

2
PADM 2902 Regression and Introduction to Econometrics

Answer: The p-value of the F-statistic is 0, which is less than any reasonable alpha; we
reject the null hypothesis and conclude that at least one of the coefficients of the polynomial
is statistically significant.

d. Test to see if the borough dummies are statistically significant. Are they?

The p-value of the F-statistic is 0, which is less than any reasonable alpha; we reject the
null hypothesis and conclude that at least one of the coefficients of the borough dummies is
statistically significant.

e. Finally, at what enrollment is total expenditures per pupil minimized? Show your work.
Note: This is not a Stata exercise.

Y=B1 (enrollment)+B2(enrollment)2
=-12.99 (enrollment)+0.004 (enrollment)2

Change in Y/ change in X=-12.99+0.004(enrollment)2=0


Enrollment=56.99

When enrollment is at around 57, the total expenditures per pupil minimized.

You might also like