Question

Question 2 Using the difference in differences (DID) methodology In 1998 there was a phased-in introduction to electronic voting (EV) in Brazil, and 10% of Brazilian municipalities used the new voting system at the time. 1 The remaining municipalities had their voters casting paper ballots, which required voters to write the name of their candidates for legislative positions. As many voters couldn’t clearly write, nor read ballot instructions, the number of valid votes to turnout ratio for federal representatives in places not using EV was close to 70% in 1998.

In contrast, places using EV had this ratio close to 92% in 1998.2 Using the DID methodology, you can verify if this change was significant and caused by EV (it could be driven by unparalleled trends). First download the data at Blackboard, which contains seven columns. The first two columns are identifying the municipalities and states. Notice that each municipality has two observations, one for each year we analyze. Next, there is a description of each variable: Year_D, which is a dummy equal to one if year=1998 (when EV was introduced) and equal to zero if year=1994 (when no municipality used EV) EV_D, which is a dummy equal to one if the municipality used EV in 1998 and equal to zero otherwise Interaction, which is a variable multiplying Year_D and EV_D Illiterates is the illiteracy rate among adults in the municipality 1 After 1998 all municipalities started using EV in Brazil. 2 Voters using EV had to type their candidates number in a machine and verify their faces before pushing the confirmation button. vv_turnout measures the valid votes to turnout ratio (dependent variable).

a) First, explain why EV usage identifies the treatment and control group.

b) Second, calculate the DID estimation without running a regression. That is, get the mean of valid votes to turnout ratio for the treatment group before the treatment and subtract it from the mean of valid votes to turnout ratio for the treatment group after the treatment (i.e. municipality using EV in 1998 – municipality using EV in 1994, let’s call this number “value 1”).

Then, get the mean of valid votes to turnout ratio for the control group before the treatment and subtract it from the mean of valid votes to turnout ratio for the control group after the treatment (i.e. municipality not using EV in 1998 – municipality not using EV in 1994, let’s call this number “value 2”).

Finally, subtract value 2 from value 1 (value 1 – value 2) and get your DID estimation.

c) Test if your findings are statistically significant by running a regression as the following one:

d) Is your estimation for beta 4 (letter c) equal to the number you found in letter b? Why? Interpret your beta 4 coefficient.

e) Add share of illiterates to your previous estimation (letter c) and interpret its coefficient. f) Write down a semester paper idea that allows you to use DID. Can you get the data for your idea (i.e. is it feasible)?

Answer 1