Team members: Kalki Seksaria, Gary Burnett, Michael Drachkovitch, Argyro Nicolaou
The data say that the top topic choices for political TV ads in Iowa did not always match the issues voters considered to be the most important in that state. We want to tell this story because it provides insight into the logic of political TV ads during a primary, where the emphasis is less on pitting the voters against the rival political party but more about differentiating same-party candidates and getting voters to the polls.
We mainly worked with the Political TV ad data set but also used data from the CNN entrance polls from the Iowa primaries.
We first had to clean up the data set, giving each topic its own column, since the ‘topic’ cells were populated with every topic mentioned in each ad. After aggregating the number of times each topic came up, we picked the top five issues for each party. In order to attribute ad affiliation as either Republican or Democrat we worked with the Sponsor column and not the candidate column since the latter included every candidate mentioned in the ad. We made a list of each of the Sponsors and researched their affiliation. Our charts exclude unaffiliated, non-profit donors (there were only two such organizations that advertised in Iowa anyway).
Having done this work, we used Tableau to create a line graph per issue per party (10 graphs total) mapping topic against time.
The CNN entrance polls on the Iowa primaries were used to make a bar chart of the timing of Republican and Democrat voters. Since the timing offered by the exit poll survey was under categorical values: ‘Today’; ‘last week’; ‘last month’; we had to decide which range of dates to include under each of these terms, to make sure that a relationship existed between the two datasets. Kalki describes the process: I first converted the categories into dates. Today = 2/1/16 (Iowa Caucuses Date). Last few days = the 2 days before the caucuses. For before last month, I assumed it meant between 1 and 3 months ago. I then assumed that the number of people who decided in a time window were evenly distributed over that time window. For example, if 30% of people decided last month, and “last month” included 23 days (30 day month – last 7 days are “last week” or shorter), then 30% / 23 = 1.3% decided each day.
Our choice to present the most popular ad topics and most important topics according to voters as a table aims at pointing to the unexpected discrepancies between the two sets of information. What other reasons could there be for pushing a specific ad topic, even if voters don’t think it is important? To try to understand this, we plotted each party’s top-5 ad topics as a line graph against time and superimposed an area chart that maps the timing of voter decisions in order to see whether there is a correlation between certain ad topics being blasted out to voters and when the voters made up their mind.