Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
The Booker Effect: The Sequel
An Investigation of Race-of-Interviewer Effects in the2013 and 2014 New Jersey U.S. Senate Elections
David P. Redlawsk, DirectorAshley Koning, Asst. Director
Eagleton Center for Public Interest PollingRutgers, The State University of New Jersey
[email protected]@rutgers.edu
Eagleton Institute of Politics
Eagleton Institute of PoliticsThe Eagleton Centerfor Public Interest Polling
Timeline of a Possible Booker Effect?Part I: 2013 NJ Special U.S. Senate Election
Eagleton Institute of Politics
October 16, 2013Cory Booker (D)
vs.Steve Lonegan (R)
June 4, 2013Gov. Chris Christie
schedules special election(August 13 primary)
June 3, 2013Sen. Frank Lautenberg
dies; U.S. senate seat inNJ becomes vacant
Rutgers-Eagleton Pre-election PollOctober 7 – 13, 2013
Eagleton Institute of Politics
WithoutLeaners
(unweighted)
WithLeaners
(unweighted)
Reported(weighted)
ActualResult
Booker 54% 56% 58% 55% -3%Lonegan 31% 33% 36% 44% +8%Someone else 2% 2% (Likely
voters)Will not vote 2% 2% N=513Don’t know 11% 6% MOE=+/- 4.3
Eagleton Institute of Politics
Why so far off in 2013? Some possibilities …
• Question Order
• Interviewer effects
Call CenterRace/Ethnicity
N Percentage
Asian 54 46.2%White 31 26.5%African-American 21 17.9%Hispanic 11 9.4%
Predicted Probability of Completing WhiteRespondent by Interviewer Race
(October 2013)
Eagleton Institute of Politics
Race-of-Interviewer Effects on Vote for BookerOctober 2013
N=678 RVs for whom we have race/ethnicityWithout leaners
Predicted Probability of Vote for Booker
Race of Interviewer
White Black Asian Hisp.Booker 50.3% 60.1% 50.5% 65.2%
Lonegan 32.9% 26.4% 33.0% 27.3%
Other 1.3% 0.7% 3.5% 0%
Will not vote 2.7% 1.4% 3.2% 0%
Don’t know 12.8% 11.5% 9.8% 7.6%
149 148 315 66
Vote by Race of Interviewer(October 2013)
Eagleton Institute of Politics
Part I: Oct. 2013 Special Election Initial ConclusionDoes it explain all of the error?No - small Booker under-estimate, large Lonegan underestimate.But if we substitute Lonegan vote for Booker vote, there is nointerviewer effect.
Is anything going on in terms of race of interviewer withBooker?Yes – after controlling for a wide range of other effects, “race ofinterviewer x respondent” has effects on likelihood of Booker vote.
The primary effect seems to be to reduce Booker vote among non-whites speaking with white interviewers.
Eagleton Institute of Politics
Part II: 2014 NJ U.S. Senate Pre and Post Election
Respondent Interviewer Race Interviewer Race AccuracyRace Actual Perceived Accurate Inaccurate DK
White 76.8% 22.4% 23.9% 29.6% 18.3% 52.1%Black 8.5% 23.2% 7.2% 19.8% 26.4% 53.8%Hispanic 8.0% 13.5% 8.3% 19.7% 31.0% 49.3%Asian/Other 4.1% 40.8% 14.1% 27.1% 36.7% 36.2%Don’t Know 43.6%Refused 2.6% 2.8% Overall 23.7% 29.9% 46.4%
Key additional variable: Respondent’s perception of interviewer race(Thanks to Celinda Lake for this suggestion)
October 2014 Rutgers-Eagleton Pre-Election Poll
Eagleton Institute of Politics
Pre-election Take 2, but Back to Square 1:Is It Recruitment?
Propensity of interviewer to talk to Whites Variance in Booker vote?
Eagleton Institute of Politics
Pre-election Booker vs. Bell Vote:Perception of Interviewer Race is Everything
All R’sBlack White Hispanic Asian
Don’tKnow Total
InterviewerActual Race
Booker 43.2% 48.8% 45.2% 48.6% - 46.9%Bell 24.8% 24.8% 24.7% 19.5% 22.6%
Perceived RaceBooker 59.0% 48.8% 40.0% 51.5% 42.6% 46.3%
Bell 7.7% 24.0% 33.3% 18.2% 24.7% 23.2%
Eagleton Institute of Politics
Pre-Election Model Predicting Vote for BookerBy Perceived Race of Interviewer
(All Respondents)
PerceivedInterviewer Race
Mean PredictedProbability
T-test vs. WhiteInterviewer
White (n=127) .515 (.343) - -Black (n=38) .605 (.314) 1.453 n.s.Hispanic (n=43) .372 (.324) -2.386 p<.05Asian (n=64) .515 (.344) .017 n.s.DK/Ref (n=224) .424 (.339) -2.396 p<.05
(Standard Deviation in parentheses)
Eagleton Institute of Politics
Respondents by Race – A Different StoryR’s Race Perceived
Interviewer RacePredicted Prob of
Booker VoteT-test vs. White
InterviewerWhite
White (n=104) .464 (.343) - - Black (n=25) .640 (.331) 2.319 p<.05 Hispanic (n=31) .322 (.309) -2.067 p<.05
Asian (n=48) .479 (.350) -0.249 n.s.Non-White White (n=23) .742 (.240) - -
Black (n=13) .538 (.280) -2.307 p<.05 Hispanic (n=12) .500 (.340) -2.450 p<.05
Asian (n=16) .625 (.339) -1.264 n.s.
Eagleton Institute of Politics
Pre-Election: White Respondents OnlyPredicted Probability of Vote for
Booker by Perceived Interviewer Race
Eagleton Institute of Politics
Post-Election (December 2015): No EffectsPredicted Probability of Reported Vote for
Booker by Perceived Interviewer Race
Eagleton Institute of Politics
Post Election: Ferguson & Staten Island Police Cases
We asked split samples about either the Ferguson killing of Michael Brownor the Staten Island killing of Eric Garner (both by police)
Question was whether the grand jury decisions to not indict were right orwrong
Given the racially-charged nature of the cases, could responseto these questions be subject to race of interviewer effects?
Ferguson Staten Island
Right decision 49% 26%Wrong decision 34% 58%
Eagleton Institute of Politics
Post Election – Staten Island ModelPredicted Probability of Answering Decision was Wrong
by Perceived Interviewer Race
Eagleton Institute of Politics
Post Election – Ferguson ModelPredicted Probability of Answering Decision was Wrong
by Perceived Interviewer Race
Eagleton Institute of Politics
Summary2013: Results unexpectedly suggested no race-of-interviewereffects for white respondents – instead non-whites reducedplanned vote for Booker when talking to a white interviewer
2014: Additional question on perception of interviewer racePerception is everything; no effects of actual race of interviewerPre-election: Strong effects of perception in expected directionPost-election: No effects of perception in reported vote
But when we ask about a racially-charged issue, the effects return
TAKE-HOME: Perceptions of interviewer race can influencerespondent reports on planned voting for an African-Americancandidate, as well as on racially charged issues
Visit the Rutgers-Eagleton Poll Online
http://eagletonpoll.rutgers.edu for our website
Facebook at https://www.facebook.com/RutgersEagletonPollTwitter @EagletonPoll.
Eagleton Institute of Politics
David P. Redlawsk, DirectorAshley Koning, Asst. Director
Eagleton Center for Public Interest PollingRutgers, The State University of New Jersey
[email protected]@rutgers.edu