Description
The goal of this assignment is to provide a simple test of whether the effects of air quality
regulations are the same across locations with different racial mix. To this end you will test if the
NOx Budget Program, a cap-and-trade market for nitrogen oxides (NOx) emissions from power
plants lead to similar effects in counties that are predominantly white versus counties that are
predominantly African American.
The data are a subset of data sample I used in the following
paper: https://olivierdeschenes.weebly.com/uploads/1/3/5/0/135068654/defensive-investmentsand-the-demans-for-air-quality.pdf. You can also get more information on the NOx Budget
Program, here: https://www.epa.gov/airmarkets/nox-budget-trading-program
The data included in the file NBP.xls, which is available on Gauchospace, are: fips (fips code
identifying each county), NBP (indicator =1 if the county was regulated under the NOx Budget
Program), PctBlack (fraction of the county population that is African American), and
Dnox_masstons (change in annual NOx emissions from all power plants in a county between
2000 and 2008 (in tons)).
Note that the NBP market was in effect in 212 of the 485 counties in
the sample from 2003 to 2008, so the 2008-2000 change give us a sense of the program’s
effect on emissions. If emissions of NOx from power plants declined in a county, then
Dnox_masstons should be negative.
Finally, it may be useful to review the RegressionSpecification.pdf notes on Gauchospace
before doing the assignment.
(a) Make a histogram depicting the distribution of Dnox_masstons.
(b) Create an indicator =1 if the county has PctBlack above the sample median, and =0
otherwise (in the rest of the assignment, I refer to this variable as ‘D’). What is the average of
PctBlack for counties above the median, i.e. counties for which D=1?
(c) Estimate a regression of Dnox_masstons on NBP. Interpret the estimated intercept and the
coefficient on NBP.
(d) Create an interaction between the variables NBP and D. Estimate a regression of
Dnox_masstons on NBP, D, and this interaction. Interpret each estimated regression coefficient,
including the intercept.
(e) What is the predicted Dnox_masstons in a county that was not regulated under NBP and
where PctBlack is above the sample median (i.e., where D=1)? Report the 95% confidence
interval for this prediction. Make sure to use “heteroskedasticity-robust” standard errors.