Problem Set 6

Due by 11:59 PM on Tuesday, March 16, 2021


You must submit all assignments as a Word document. Answer all questions carefully and neatly. Use section headings to distinguish what question you are answering. If the question involves a graph, copy/paste the graph into your document and make sure the graph has a title and axis labels.

Task 1: List Experiment

We talked in class about how studying clientelism is hard because people may not want to tell you that they’ve sold their vote (because they’re embarrassed, afraid, etc.). One way researchers try to get around this issue is by asking about the behavior indirectly, through a list experiment.

Read this cached link since WB took the whole page down for some reason tutorial on what list experiments are and how they work. Then, skim this, paying special attention to the section What about the innocuous questions?. I also added slides on list experiments to the week’s slides that will be helpful!

Using what you learned, make your own list experiment where you would want to measure vote-buying. Make sure to justify your choice of non-sensitive items.

Task 2: Vote-buying in Latin America

Use the following survey data from LAPOP fielded in 2010.

Let’s look at the vote-buying variable, clien1. First, filter out responses like “Not Applicable” or “Don’t know”, or “No response”; we don’t want those. Then, use IFS to create a new variable called “vote_sold” that takes the value of 1 if a person answered “Frecuentemente” or “Rara vez” to clien1, and 0 otherwise.1

  1. Take the average of vote_sold for each country and plot the results using a suitable graph (remember: the average of a variable that only takes on 0 or 1 values is the proportion of respondents who answered 1). Discuss which countries have highest and lowest levels of vote-buying according to the survey, and whether this is surprising to you based on your priors about those countries.

  2. Use pivot tables to take the average of vote_sold for 2-3 socio-demographic variables (e.g., gender on rows, vote_sold on values). Report the results using either tables or plots. Are certain people more likely to be targeted for vote-buying than others? What does the data say?

  3. Create a new variable called “activist” that is the average of the “cp” variables (Use a formula and the AVERAGE function, or just do it manually). This variable will measure how politically engaged a person is: a person who scores low on activist will tend to do more of the stuff described in the “cp” variables than someone who scores high (see variable dictionary below).

Use pivot tables to find the average of vote_sold across levels of activist and plot them with activist on the x-axis. What does the relationship tell you about who is most likely to be targeted for vote-buying? Why might that be?

original labels
pais Country
ur Urban/Rural
q2 Age
q1 Sex
clien1 Have you been offered a bribe to vote in a particular way?
ed Years of Schooling
tamano Size of Location
cp5 Helped Solve a Problem in the Community (1 = very often, 4 = never)
cp8 Attendance at Meetings of Community Improvement Group (1 = very often, 4 = never)
cp9 Attendance at Meetings of Professionals or Merchants Associations (1 = very often, 4 = never)
cp13 Attendance at Meetings of Political Movements or Parties (1 = very often, 4 = never)
q10 Income categories (0 = very low income, 10 = very high income)

  1. Be careful with spelling, capitalization, etc.!↩︎