Martha Kropf, Ph.D. Department of Poli4cal Science and Public Administra4on University of North Carolina at CharloAe

Transcription

1 The Impact of Ranked Choice Vo4ng on Elec4on Coopera4on and Civility: Measuring Public Sen4ment through a Content Analysis of Campaign- Related Communica4ons Martha Kropf, Ph.D. Department of Poli4cal Science and Public Administra4on University of North Carolina at CharloAe *Project Funded by The Democracy Fund through Fairvote.org

2 How Does Ranked Choice Vo4ng Affect Democra4c Outcomes? Robb (2012): examina4on of nega4vity in campaigning San Francisco ( ) Content analyzed mailers Team ads led to coordinated aaacks (page 119), though the coordinated aaacks were very limited. Mostly, the RCV contests more posi4ve

3 Nov. 9, 1310:02 PM I wish RCV would change poli4cal behavior and get campaigns to play nice. I fear the old approach of 'trash talk' will con4nue with the excep4on that the current powers will merely be sure 3 nobodies also appear on the ballot.

4 Hypotheses In comparing elec4ons, those using ranked choice vo4ng will have a higher level of coopera4on and civility than those using tradi4onal elec4ons. In comparing elec4ons, those using ranked choice vo4ng will be less nega4ve than those using tradi4onal elec4ons.

5 Methods Part of a larger study using survey methods and examining elec4on data. Ci4es with RCV (Minneapolis, St. Paul & Cambridge) compared to matched ci4es without. Also, a content analysis of campaign discourse (my part) Debates/newspaper ar4cles Comments on newspaper ar4cles Facebook posts Tweets

6 How to analyze text Sen4ment analysis Tweets have been analyzed for a variety of reasons Debates/congressional record analyzed Originated with stock market analysis Not public opinion per se Not an analysis to see if newspaper ar4cles are posi4ve or nega4ve about the elec4on reform change or the candidates Analysis of microblogging is exploding Larsson and Moe (2012): Swedish elec4ons, 2010 Tumasjan et al. (2011) TwiAer feeds can help forecast elec4ons, 2009 German federal elec4on

7 Linguis4c Inquiry and Word Count (Pennybaker, et al., 2007) Creates dic4onaries of terms used, analyzes verb tense and pronoun use, etc. to determine posi4ve and nega4ve tone, but also other affec4ve categories. Pennybaker et al., 2003: LIWC uses a word count strategy whereby it searches for over 2,300 words or word stems within any given text file. The search words have previously been categorized by independent judges into over 70 linguis4c dimensions. Has been used to analyze Congressional debates (Yu, Kaufmann, and Diermeier, 2008) Big Five Personality Dimensions

8 Affec4ve Sen4ment The emo4on language people use can measure campaign coopera4on and civility: posi4ve emo4ons nega4ve emo4ons social communica4on inclusiveness v. exclusiveness Reac4on Language anger anxiety future orienta4on tenta4veness v. certainty

9 Examples of Tweets Checked out our website lately? Request an absentee ballot! Together We Can - hap://t.co/n5scbnot3j #TOGETHERWeCan move our community forward! hap://t.co/ S1kYTU7rs6 #Sus4CR From the Minneapolis Star Tribune: Check out this nut who is running for mayor! Have you voted yet? There's s4ll 4me! The polls are open un4l 8PM - all you need to do is ARRIVE before 8. Rank Betsy #1! # I voted for #MplsMayor and I hope you will too. hap://t.co/hx092ozsiu

10 Comparing The Tweets to 9-11 Speeches Posi4ve Nega4ve Social Inclusive Exclusive Giuliani 9-11 Minneapolis Mayor St. Paul City Council Cambridge City Council

11 Comparing the Tweets to 9-11 Speeches Future Tense Anger Anxiety Tenta4veness Certainty Giuliani 9-11 Minneapolis Mayor St. Paul City Council Cambridge City Council

12 Limita4ons Sarcasm? Irony? Not all local- level candidates are using TwiAer TwiAer is an interes4ng language Which tweets to use? (Is a tweet about the Vikings good game fair game?) How is it ver4cally integrated? (with candidate websites & Facebook) Many viable (and not so viable) candidates are using it The more observa4ons, the beaer the measurement The more tweets, the beaer the measurement The more units of observa4on

13 Conclusions One can detect varia4on among candidates Candidates in these RCV ci4es use more posi4ve language than nega4ve language Candidates use more inclusive than exclusive words Candidates use social words Candidates appear to be more tenta4ve than certain Cannot yet make causal conclusions because of the lack of a counterfactual (in this analysis)

14 Ideally Intensive case study of Minneapolis, MN v. SeaAle, WA Using all the media sources I have tapped Newspaper ar4cles Comments on the newspaper ar4cles Tweets from the public regarding the newspaper ar4cles Debates Facebook posts