Read, Seen or Heard A Text-Analytic Approach to Campaign Dynamics Gallup, Michigan, Georgetown Working Group: Frank Newport, Lisa Singh, Stuart Soroka, Michael Traugott, Andrew Dugan Slides presented at the Center for Political Studies, Institute for Social Research, University of Michigan, October 5 2016. Background There is a good deal of research capturing trends in government/candidate approval, and vote intentions. We are often able to connect these trends to campaign events, and/or shifts in media content. We are rarely able to directly capture the information that voters remember. Background Whereas most data sources focus on the effects of campaign information, i.e., vote intentions, or the provision of campaign information, i.e., media coverage, we capture the reception of information, i.e., what respondents recall having read, seen or heard about the candidates. Data The Gallup U.S. Daily* interviews roughly 500 people by per night, and asks what respondents have read, seen or heard about Hillary Clinton / Donald Trump in the past several days. Verbatim responses are recorded by interviewers. * The Gallup U.S. Daily interviews adults aged 18 and older living in all 50 states and the District of Columbia using a dual-frame design, which includes both landline and cellphone numbers. Gallup samples landline and cellphone numbers using random-digit-dial methods. % Having Read, Seen and Heard Something about Clinton or Trump % Non−Empty Responses, 3−day Mean 100 90 RNC convention DNC convention 9-11 80 70 First debate 60 50 ~41,000 interviews 40 July 10 Sept 1 Sept 26 (debate) Rolling 3-day (lagged) averages, unweighted. Words Distinguishingclinton Read, Seen and Heard about Clinton and Trump ads children untrustworthy facebook department mostly congress corruption deleted done campaigning private found questions criminal classified response saw ella investigation timkaine time jail her trail berniesanders problems playshe released dishonestsupporters got memorial information bengazi regarding server crook corrupt debate secretary truth benghazi issues charges interview money situation still billclinton state well fun mikepence tryingbaby flood putin trip wentbuilding border war mouth bornstatement idiot black wants gold made church soldier negative people man flip calling build russia speech isis racist plan remarks meeting making family obama wall son like rally mind foundervisitsaid going tax wife paid twitter parents Top 160 words that are changed president policy talking most uniquely associated illegal changing amendment doesnt second with responses about louisiana american bigot manager african Clinton or Trump, font size missuniverse paulryan heart tedcruz make america women weighted by frequency, hack veterans military stance change based on all responses, mccain statements business July 11-Oct 3, using states lost rnc way star wifes take limited synonym recoding arizona americans and lemmatization. scandal dnc fbi liar health email get stupid detroit muslim mexico country immigration media makeamericagreat khan economy foundation Weekly Trends in Read, Seen and Heard Jul 10−16 email berniesanders Jul 17−23 email convention liar Jul 24−30 Jul 31−6 Aug 7−13 Aug 14−20 Aug 21−27 Aug 28−3 Sep 4−10 Sep 11−17 Sep 8−24 Sep 25−1 Oct 2− fbi liar scandal speech going convention email speech email liar dnc president conventionspeech campaign email liar email tax foundation scandal foundation liar health going email foundation liar email email scandal health foundation liar fbi scandal health health liar scandal fbitalking email issues campaign liar debate email liar newyork debate debate email tax liarcampaign email liar health night Top 5 words, font size weighted by frequency, based on all responses, July 11-Oct 3, using limited synonym recoding and lemmatization. Weekly Trends in Read, Seen and Heard Jul 10−16 mikepence dallas going shooting pick Jul 17−23 convention speech Jul 24−30 russia email convention speech Jul 31−6 muslim family son paulryan convention Aug 7−13 isis amendment obama second people Aug 14−20 campaign Aug 21−27 immigration wife national going speech isis people louisiana campaign people louisiana trying Aug 28−3 mexico immigration Sep 4−10 mexico immigration speech Sep 11−17 Sep 8−24 Sep 25−1 Oct 2− going speech president going going president health obama born talking speech talkingdebatenewyork obamapeople debate tax missuniverse women people going debate paid missuniverse returns Top 5 words, font size weighted by frequency, based on all responses, July 11-Oct 3, using limited synonym recoding and lemmatization. The Changing Tone of Read, Seen or Heard Responses −> Trump Advantage −> 6 4 2 Differences in the tone of what respondents report having “read, seen or heard” about Donald Trump, versus what they report having “read, seen or heard” about Hillary Clinton. REPUBLICAN RESPONDENTS ONLY 0 Sept 2 Sept 26 (debate) Rolling 3-day (lagged) averages, unweighted. Tone estimated using the Lexicoder Sentiment Dictionary The Changing Tone of Read, Seen or Heard Responses −> Clinton Advantage −> 6 4 2 Differences in the tone of what respondents report having “read, seen or heard” about Donald Trump, versus what they report having “read, seen or heard” about Hillary Clinton. DEMOCRATIC RESPONDENTS ONLY 0 Sept 2 Sept 26 (debate) Rolling 3-day (lagged) averages, unweighted. Tone estimated using the Lexicoder Sentiment Dictionary Motivated Reasoning in Recollections of the First Debate Based on respondents interviewed on Sept 27 and 28, 61% believe that Clinton won, while 27% believe that Trump won.* Within partisan groups: 92% of Democrats believe that Clinton won the debate, and 3% believe that Trump won. 53% of Republicans believe that Trump won the debate, and 28% believe that Clinton won. * Based on Gallup analyses (of the same survey data used here). Words Distinguishing Democrats and Democrat Republicans’ Read, Seen and Heard about Clinton, Post-Debate ella going great missuniverse helping live next sobre debt night won saw response really presidential reaction plans bring taxissues living done immunity opinion prison performance good debate polls remember gets putting deleted money crook much foundation got crooked server everything fbi thats negative media reading problems health berniesanders Republican year supporters cant Identifiers information involved dishonest unfavorable criminal truth parents specifically one bad scandal benghazi hear anything email liar article twitter positive country young saturday times talking poll recent college economy states Democratic prepared women responses trail newyork Identifiers better carolina president campaigning strong statements children campaign today education rally read election help Top 100 words that are most uniquely associated with responses about Clinton or Trump, font size weighted by frequency, based on all responses after the first debate, using limited synonym recoding and lemmatization. internet campaign fat said ads trying veterans plans election different situation recall country ago much results brought ive gonna goingcountries michigan great florida last polls jobs well right really america Republican Identifiers remember Words Distinguishing Republicans and Republican Democrats’ Read, Seen and Heard about Trump, Post-Debate everyone foundation newyork liar people racist cuba tax paid fact russia talks twitter night poor performanceweight minorities made way remarks embargo many disrespect lady new debate women returns young pig called pay presidential years Democratic Identifiers hasnt regarding read fox idiot using charity saw deal good makespeech wants states email kind doesnt got plan issues rally facebook makeamericagreat immigration whatever economy media better talking rallies back little contest Top 100 words that are most uniquely associated with responses about Clinton or Trump, font size weighted by frequency, based on all responses after the first debate, using limited synonym recoding and lemmatization. Summary of Findings (thus far) Republicans’ read, seen or heard responses have been more systematically positive about Trump than Democrats have been about Clinton. Clinton’s read, seen or heard responses were dominated by ‘email’ until Sept 11. Recalled information about Clinton has been more positive since that time. Summary of Findings (thus far) A majority of respondents believe that Clinton won the first debate. But there are real differences in what partisans report having read, seen or heard since then. Clinton: Democrats cite her debate performance, Republicans continue to cite news on email and other scandals Trump: Republicans cite his debate performance, while Democrats cite his complex record with women, particularly Alicia Machado. Trump debate rather refuses like called seeing mismo mainly year brought para kids recall put lost nafta read shared health sources muchtelevision four left bengazi facebook black around change live the experience helping cartoon situation dishonest government corruption daughter saw crooked saturday foundations womens Top 100 words email 0 Differences in the tone of what respondents report having “read, seen or heard” about Donald Trump, versus what they report having “read, seen or heard” about Hillary Clinton. INDEPENDENT RESPONDENTS ONLY −4 anything back Clinton −2 showed fat fact paid everyday even support 2 veterans sending children foundation feed let send scandal bad think want smart tax missuniverse returns twitter federal honest towards morning women life 4 −> Trump Advantage −> doesnt contestant liar 6 negative racist people drink trying forth two good last girl performance Amongst Independents, 59% believe that Clinton won the debate and 30% believe that Trump won the debate. money What About Independents? wants weight loss mexico flint complaining mrs handled hasnt uniquely associated with Clinton or Trump, amongst Independents only, font size weighted by frequency, based on all responses after the first debate, using limited synonym recoding and lemmatization. Rolling 3-day (lagged) averages, unweighted. Tone estimated using the Lexicoder Sentiment Dictionary −6 Sept 2 Sept 26 (debate)
© Copyright 2024 Paperzz
