This data is based on a survey of voting behavior from the 2000 presidential election – votingdata.csv. The
codebook is important and also available for download with the assignment on Blackboard.
1) Generate a new variable
a. Generate a new variable high_income =1 if the person’s family income is above the 68th
percentile (income=4 or income=5), high_income=0 otherwise
2) Run a regression of voting Republican on your new high_income variable. How do you interpret
this coefficient?
3) Re-run this regression but add age, female, and race (use white as the reference group) to your
model. Interpret these coefficients and note at what level they are significant.
4) Create an interaction term using this data by multiplying two of the variables together and include
it in the regression. If you use one of the other variables that we haven’t used so far, be sure to
add the main effects.
a. Interpret the interaction term qualitatively and quantitatively. What does it tell us?
b. Is the interaction significant? How do you know?
Create a table that shows these results. Each regression should be a different column – so you
should have 3 columns of results. Put standard errors in parentheses. Put appropriate stars on
the coefficients to indicate statistical significance (*** for p<0.01, **="" for=""><0.05, *="" for=""><>
Include the number of observations. You need to add an appropriate title and table notes. Table
should have actual word NOT variable names.
Part II: Determinants of the State Suicide Rate (50 points)
The following questions are based on a state level panel data on the suicide rates from 1990-2000 and
was originally used in the paper “Are Mental Health Insurance Mandates Effective? Evidence from
Suicides” Health Economics (2006) by Jonathan Klick and Sara Markowitz. The codebook is important
and available to download with this assignment from Blackboard.
1) Run a regression of the adult suicide rate on percent unemployed, per capita state income,
percent living in rural areas, percent of population with a college degree, and percent of
uninsured
a. How do you interpret the coefficient on the unemployment rate?
b. Are you surprised by the sign of the coefficient?
2) Re-run the regression but include both state and year fixed effects.
a. Which of your coefficients changed sign and significance from the regression in (1)?
b. What are potential explanations for these changes?
3) Re-run your regression but now also include the available religion variables.
a. How do you interpret the coefficient on percent southern Baptist?
b. Are the religion variables jointly significant? How do you know? (what is the value of the
statistic you used? And the p-value?)
4) Use this dataset to tell me something else interesting about the suicide rate, be sure to include
state and year fixed effects. Make this a 4th regression and interpret 1 or more relevant
coefficients. Do you think that this effect you estimated is causal? Do we have more confidence
in the causal interpretation than if we ran a cross-sectional regression that used data in a single
year only? Why or why not?
5) You used heteroskedasticity-robust standard errors, do you have reason to believe these
standard errors may be incorrect? Why or why not?
Create a table that shows these results. This table does not need to be created in R/Stata. You may use
the output to create the table by hand. Each regression should be a different column – so you should
have 4 columns of results. Put heteroskedasticity-robust standard errors in parenthesis under the
coefficients. Do NOT show the coefficients on the fixed effects. Instead, include a row called “State
Fixed Effects” and indicate either No or Yes in each column for whether those are included, same for
year fixed effects. Put appropriate stars on the coefficients to indicate statistical significance (*** for
p<0.01, **="" for=""><0.05, *="" for=""><0.1). include="" the="" adjusted="" r^2="" and="" the="" number="" of="" observations.="" you="">
to add an appropriate title and table notes. Table should have actual words, NOT variable names.
Additional Notes:
- Your code should be saved in a .R/.do file as good practice. This code should have
comments at the top that include your name, and that this is code for Econ 436
Homework 3. Please turn in this .R/.do file and upload as a SEPARATE file onto Blackboard.
This .R /.dofile should have all the commands you actually used to run your homework.
I.e. if I opened the .R/.do file and ran it (after of course changing the directory) then I
would be able to see all the variables you created, regressions you ran, and tests you did.
A CORRECT R/DO FILE WILL BE PART OF THE GRADE THIS TIME. IT MUST INCLUDE THE
read.csv (for R)/insheet (for Stata) COMMAND AT THE BEGINNING and have appropriate
comments