#Polar Scores for Members of Congress

My final project is a replication study of a forthcoming paper by Libby Hemphill, et al. titled #Polar Scores: Measuring Partisanship Using Social Media Content. The paper uses CASM Lab’s purpletag repository to collect, parse, and analyze tweets from our congressional representatives’ official Twitter accounts. Members of congress are given a #polar score based on the liklihood that the hashtags they have used come from Democrats or Republicans. While the existing paper uses data from the 2012 presidential election, I will use data previously collected by CASM Lab from the 2014 midterm election as well as data I collected myself from the 2016 election. So far I have completed the following steps:

  • Locate original #Polar Scores paper (thanks, scihub bot!)
  • Launch and configure an AWS instance for 2016 election purpletag data
  • Collect, parse, score and serve 2016 election data using purpletag
  • Create a table from existing polar scores showing members of congress scores by date

PurpleTag data: 2016 election

Next steps include:

  • Clean up code to generalize table of member of congress scores from small working sample to larger data set, in separate script
  • Run new script on 2016 data
  • Find 2014 data on CASM Lab’s server & run new script on that data
  • Compare #polar scores of members of congress 6 weeks prior to 2014 election (i.e. six 7-day windows) with similar period prior to 2016 election
  • Create a table from existing #polar scores showing hashtag scores by date
  • Generalize hashtag code into separate script and run hashtag script on 2014 and 2016 data
  • Compare hashtag scores with member of congress scores
  • Write scripts to perform additional analysis: which members of congress are using which hashtags, how many times per date is a hashtag used

I will finish by writing a paper that summarizes my findings and compares them to the original paper results. I will begin with only #polar scores of members of congress in 2016 and 2014 for six 7-day periods prior to the election. If time permits I will add analysis of hashtag scores, and possibly additional detail showing which members of congress use which hashtags, and how many times those hashtags are used.

Written on November 26, 2016 by Carol Schmitz