Let's add another control to your regression in an attempt to obtain more reliable estimates of the effect of campaign spending. As you may know, 2018 was a good year for Democrats in House elections.

a) Is the overall good performance of Democrats in 2018 a potential confounder in your regression? Why or why not?

b) Create a new variable indicating whether the incumbent is a Republican - call it republicanincumbent. It should take a value of 1 if the incumbent is a Republican and a value of 0 if the incumbent is a Democrat. Now re-run your regression, but include that variable as a control. What is the coefficient on this variable? Interpret what the coefficient means.

c) Does including this control variable meaningfully change your estimated coefficients of interest (i.e. the coefficients on incumbent and challenger spending? Why or why not, do you think?

Display keyboard shortcuts for Rich Content Editor