Problem Set 7: Experiments

This week’s homework involves both a conceptual part and a coding part. The homework is based on the following paper:

Gerber, Alan S., Donald P. Green, and Christopher W. Larimer. 2008. “Social Pressure and Voter Turnout: Evidence from a Large-scale Field Experiment.” American Political Science Review 102(1):33-48.

Conceptual questions (Part 1 of 2)

  1. What is the unit of analysis in this study (the unit for which the outcome is measured)?

    1. A voter
    2. A household
    3. An election

  2. What is the unit that was randomized to an experimental condition?

    1. A voter
    2. A household
    3. An election

  3. In this study, how many potential outcomes does each person have? Include the outcome under control as one of the potential outcomes.

  4. Which of the following is the clearest example of exchangeability as it applies in this study?

    1. Whether a person received the “Neighbors” treatment is independent of whether they would vote if given the “Neighbors” treatment
    2. Whether a person received the “Neighbors” treatment is independent of whether their neighbor received the “Neighbors” treatment
    3. Whether a person received the “Neighbors” treatment is independent of their past history of voting in elections
    4. Whether a person received the “Neighbors” treatment is independent of the potential outcome of the election

Coding question (Part 2 of 2)

The following code will load the data from the study.

library(tidyverse)
library(haven)
social_pressure <- read_dta("https://soc114.github.io/assets/data/social_pressure.dta")

The relevant variables are:

  • treatment is the randomized treatment
  • voted is the outcome variable

Write an R script that uses these data to reproduce the point estimates from Table 2 of the paper.

  • Your result should be a data frame or tibble named results.
  • There should be two columns: treatment and a column voted, where voted is a proportion between 0 and 1 for the mean outcome under each treatment value.
  • You do not need to produce the number of individuals. Just the proportion.

You might use group_by and summarize. It is possible to write this code in 4 or fewer lines (though longer is also fine).

Back to top