how do voters decide? preliminary results from a field experiment g. michael weiksner stanford...
TRANSCRIPT
How do voters decide?Preliminary results from a field experiment
G. Michael WeiksnerStanford University
April 2, 2008
Agenda
• Why local primaries• Theory & Predictions• Methodology• Preliminary Results– A short-term effect– Why?
Why local primaries
• Not well understood• In low information environment, information
treatment should have a larger impact• Regularly occurring – More opportunities to research– Possibility to generalize to large class of elections
Theory & Predictions
• Dahl: “A person’s interest or good is whatever that person would choose with fullest attainable understanding of the experiences resulting from that choice and its most relevant alternatives”– Our information conditions are close to a practical operationalization
of “full-information” voting—does it make a difference? Are citizens competent under better circumstances?
• Popkins: “Low information rationality,” or using heuristics or shortcuts to the information to make decisions– Information should reduce reliance on heuristics, like voting for the
candidate who shares my gender
Methodology
A Randomized Field Experiment• Mundane realism: the subjects are in a realistic setting • Internal validity: causal inference appropriate through
random assignment• Generalizable: ideally, random sampling• Cheap & easy to administer
E.g., Iyengar’s MAPSS talk, 2/6/07 and Green’s talk 2/13/07
Methodology (cont’d)Procedure
• Orlando primaries, Tues Sept. 5, 2006– Orlando Sentinel uses theVoterGuide.org to collect candidate data on
22 races including Governor, Senate, Congress, non-partisan judges• 514 participants recruited from online panel on Thurs Aug 31
– Skews towards more education, politically aware, male– Randomly assigned to condition differing by kind and amount of
election information– 307 participants responded to follow up survey (Sept 6-7)
• Surveyed on vote intentions, vote choice, attitudes and general and campaign-specific knowledge
Research design
R X1 O1 O2
R X2 O1 O2
R X3 O1 O2
R X4 O1 O2Party,
Vote Choice,Attitudes,
Demographics
Turnout, Vote Choice,Bio & Issue Knowledge,
Random Assignment
Contact Information OnlyBiographical InformationIssue information
All Information
Condition
T1 Vote Choice-Executive
Likelihood ratio test
Issue Bio
Governor 0.024 * 0.732
FL Attorney General 0.047 * 0.134
FL Chief Financial Officer 0.046 * 0.683
Orange County Mayor 0.028 * 0.006 **
T1 Vote Choice-Legislative
Likelihood ratio test
Issue Bio
US Senate 0.158 0.085
US Congress 5th 0.246 0.238
US Congress 8th 0.655 0.759
US Congress 15th 0.814 0.697
US Congress 24th 0.008** 0.313
Florida Senate - 8th 0.424 0.772
Florida house - 36 0.963 0.204
Florida House - 41 0.015* 0.072
T1 Vote Choice - Judicial
Likelihood ratio test
Issue Bio
County Judge – 17 0.380 0.040 *
County Judge – 6 0.477 0.012 *
County Judge – 7 0.778 0.768
County Judge – 5 0.726 0.289
Circuit Judge 5th Group 7 0.027 0.124
Circuit Judge 9th Group 5 0.304 0.005 **
Circuit Judge 18thF 0.796 0.836
Percent who vote forcandidate of same gender
0.59
0.510.51
0.61
0.46
0.48
0.50
0.52
0.54
0.56
0.58
0.60
0.62
No Biographical Information Biographical Information
Male
Female
Summary of results
• No long –term effects• Election information makes a difference in local primary vote
choice– Issue information changes choices in many executive races– Issue information changes some choices in legislative races– Biographical information changes choices in judicial
• Some evidence that information affects vote choice through gender- No consistent story (yet?) for why issue information changes vote choice- Among males, biographical information reduces gender-based voting- Among females, biographical information increases gender-based voting
Parting thoughts…
• Memory aids are really important and interesting potential impact of mail & internet voting
• What is the point of local primaries?• Future research:
– Can we replicate the gender results in a lab experiment?– How would these results differ in a general election?– Would deliberation make a difference?– What impact does social information (i.e., personal endorsements)
have?– Would edited information make a difference?– Does one-sided information make a difference?
Random Assignment (cont’d)
Age FemaleGen'l Political
KnowledgeN M SD M SD M SD
No issue infoNo Bio 133 47.9 14.4 0.368 0.484 0.882 0.193Bio 124 48.1 14.4 0.339 0.475 0.866 0.203
Issue InfoNo Bio 112 48.0 15.1 0.402 0.492 0.860 0.244Bio 105 48.9 13.7 0.505 0.502 0.902 0.190
T2 Vote Choice
Likelihood ratio test Likelihood ratio test
Issue Bio Issue Bio
Governor 0.128 0.621
Attorney General 0.734 0.444 Cty Judge 17 0.664 0.087
CFO 0.567 0.570 Cty Judge 6 0.982 0.082
OC Mayor 0.691 0.307 Cty Judge 4 0.128 0.492
U.S. House 8Th 0.193 0.243 Cty Judge 5 0.781 0.206
U.S. House 13Th 0.144 0.581 Crct 5 Judge 0.139 0.092
U.S. House 24Th 0.259 0.994 Crct 9 Judge 0.780 0.148
State Senate 8 0.165 0.441 Crct 18 Judge 0.890 0.326
State House 41 0.083 0.253
T1 Results
Source df F Eta pIssue Information
Endorsement Knowl. Item 1 3.385 0.082 0.066Candidates on Issues 1 8.666 0.130 0.003
Biographical InformationEndorsement Knowl. Item 1 22.310 0.205 0.000Candidates on Issues 1 1.391 0.052 0.239
Issue x Biographical InformationEndorsement Knowl. Item 1 0.349 0.026 0.555Candidates on Issues 1 0.367 0.027 0.545
ErrorEndorsement Knowl. Item 506 0.170Candidates on Issues 506 0.055
T1 Vote Choice – GovernorIssue Information Bio Information
No Yes No YesDemocrats
Carol Castagnero 1.4 0.8 1.2 1.2 Glenn Burkett 1.8 1.7 1.2 2.4
Jim Davis 22 21.1 20.8 22.4 John M. Crotty 2.9 0.4 1.9 1.6
Rod Smith 11.9 17.3 12.4 16.5Republicans
Charlie Crist 36.1 29.5 35.9 30.2 Michael W. St. J 1.1 0.4 0.8 0.8
Tom Gallagher 11.9 11.4 10.4 12.9 Vernon Palmer 1.8 0.4 1.5 0.8
Likelihood ratio test .024 * .732
T1 Results
Endorsement Knowledge Item
Understand candidates on the
issuesN M SD M SD
No Issue Info
No Bio Info 139 0.165 0.373 0.268 0.168
Bio Info 136 0.360 0.482 0.305 0.188
Issue Info
No Bio Info 117 0.120 0.326 0.342 0.204
Bio Info 118 0.271 0.446 0.354 0.191
T2 Turnout, Knowledge
T2 TurnoutT2 Issue
KnowledgeT2 Biographical
Knowledge
N M SD M SD M SD
No issue info
No Bio Info 90 64.4 48.1 0.157 0.217 0.419 0.351
Bio Info 76 61.8 48.9 0.202 0.23 0.482 0.320
Issue Info
No Bio Info 66 63.6 48.5 0.25 0.251 0.523 0.327
Bio Info 74 64.8 48.1 0.207 0.221 0.421 0.346