Date post: | 02-Jan-2016 |
Category: |
Documents |
Upload: | sophia-dean |
View: | 215 times |
Download: | 0 times |
Why Groups Divide: Understanding the Mechanisms that Lead to Polarization in Deliberative Polling™
Shawn TutejaB.S. Mathematics ’15
M.S. Statistics ’15
Polarized Society
• Congress: divided societies are unproductive• Studies show that inter-party marriages are
increasingly frowned upon (~9% of couples are Democrat-Republican)• Democrats and Republicans have stated
that they tend to avoid each other in social contexts.
But Is Polarization Bad?
• Polarization is not inherently bad– It often leads to debate
• In the United States, divisions are often not based on factual or researched opinions.
Can Education/Deliberation be the Solution?
• Studies show that people choose news that they agree with.
• A recent Pew Research Center showed that 47% of conservatives get their news from Fox News, whereas liberals get their news from a blend of CNN, MSNBC, and the New York Times.
• The same study shows that people discuss news with like-minded individuals.
Research Question
• What explains the shifts in group opinions that occur during deliberations?– Radicalization – the shift of a group in the
direction of the pre-deliberation views; indicated by change in median).
– Group-Splitting – the division of the group following deliberation (e.g. radicalization when there are multiple factions); indicated by increase in variance
Extrema A Extrema B
Deliberation
Spectrum of opinion
Deliberation on issues often leads to radicalization of opinion in groups
Extrema A Extrema B
Deliberation
Spectrum of opinion
In groups with multiple factions, deliberation drives each faction to extreme opinions.
Hypothesis/Model
• I hypothesize the more balanced group discussions are, the more unpredictably groups shift.
• Dependent Variable– Two different groups shifts
• Independent Variable– Balance of Argumentation
Testing by Deliberative Polling
• I test the research question in deliberative polling—a particular type of deliberative forum.
• Other types of deliberative forums include juries, focus groups, and crowdsourcing campaigns.
Relevant Literature
• Radicalization has been shown to happen in certain types of deliberative forums regularly
• Luskin et al (2007) showed that deliberative polling does not have systematic shifts
Sunstein/Manin Explanation
• Briefing materials help push larger majorities into smaller ones prior to the poll.
• Unprecedented diversity of opinions/representativeness in small groups.
• “The empirical answer is not available”(Manin 8).
Dependent Variable
• Radicalization and Group-Splitting on three topics: – Budget Stability– Greater Equity– Revisiting Proposition 13
Dependent Variable:Index Formation
12345678910111213
Index 1: Budget Stability [0-10]
Index 2: Greater Equity [0-10]
Index 3: Proposition 13 [0-10]
Dependent Variable:Index Formation
Index/Topic 1 (Budget Stability)
Index 2(Greater Equity)
Index 3(Prop 13)
Number of questions
5 3 5
Questions 2aa-2ae 2af-2ah 2ai-2am
Observations 338 329 297
Alpha 0.79 0.62 0.86
Independent Variable
• Balance of Argumentation (Warranted and Unwarranted)– Obtained through a coding process of all the
transcripts• Knowledge (Control)
Results
• Exploratory Analysis: Dependent Variable• Exploratory Analysis: Independent Variable• Regression– Dependent Variable: Radicalization– Dependent Variable: Group-splitting
Exploratory Analysis: Median
Index 1 Index 2 Index 3
Average shift of mean toward radicalization
0.113 -0.039 (toward midpoint!)
-0.563 (toward midpoint!)
% of groups that became more radicalized
66.67% 37.5% 29.17%
Exploratory Analysis: Variance
Index 1 Index 2 Index 3
Average Increase in Variance
-0.390 -0.585 -1.546
% of groups with increase in Variance
29.17% 37.5% 25%
Exploratory Analysis: Independent VariableIndex 1 Index 2 Index 3
Imbalance in Warranted Arguments
35 15 -15
Imbalance in Unwarranted Arguments
8 0 -11
Total Warranted Arguments
277 258 198
Total Unwarranted Arguments
93 94 74
Difference (p-value)
0.459 0.926 0.904
Exploratory Analysis: Independent Variable
Index 1 Index 2 Index 3
Persuasive Argument Hypothesis (Median)
75% 33.33% 41.67%
Groups with Prior Leanings
95.83% 66.67% 83.33%
Regression:Radicalization Index 1
Coefficient Std. Error T value p-value
(Intercept) -0.1388 0.2363 -0.587 0.5635
Warranted Arguments
1.1256 0.5277 2.133 0.0455*
Unwarranted Arguments
-0.2891 0.3363 -0.860 0.4001
Knowledge Gain
1.1237 1.5822 0.710 0.4857
Multiple R-squared
0.2541
p-value 0.1114
Regression:Radicalization Index 2
Coefficient Std. Error T value p-value
(Intercept) 0.2811 0.3448 0.815 0.4245
Warranted Arguments
1.2397 0.5801 2.137 0.0451*
Unwarranted Arguments
0.2833 0.3603 0.786 0.4408
Knowledge Gain
-1.2473 2.1988 -0.567 0.5768
Multiple R-squared
0.2485
p-value 0.1191
Regression: Radicalization Index 3
Coefficient Std. Error T value p-value
(Intercept) 0.5365 0.4526 1.185 0.250
Warranted Arguments
-0.6047 0.5966 -1.014 0.323
Unwarranted Arguments
-0.1323 0.4475 -0.296 0.771
Knowledge Gain
0.8500 2.7699 0.307 0.762
Multiple R-squared
0.05545
p-value 0.7605
Regression:Group-Splitting Index 1
Coefficient Std. Error T value p-value
(Intercept) 0.8425 0.6013 1.401 0.176
Warranted Arguments
1.2113 1.3427 0.902 0.378
Unwarranted Arguments
0.5124 0.8558 0.599 0.556
Knowledge Gain
-6.1393 4.0259 -1.525 0.143
Multiple R-squared
0.1266
p-value 0.4278
Regression:Group-Splitting Index 2
Coefficient Std. Error T value p-value
(Intercept) -1.1327 0.6455 -1.755 0.0946
Warranted Arguments
2.5309 1.0862 2.330 0.0304*
Unwarranted Arguments
0.3126 0.6746 0.463 0.6480
Knowledge Gain
3.2608 4.1168 0.792 0.4376
Multiple R-squared
0.266
p-value 0.1547
Regression:Group-Splitting Index 3
Coefficient Std. Error T value p-value
(Intercept) -0.9006 0.8970 -1.004 0.3274
Warranted Arguments
-2.6239 1.1825 -2.219 0.0382*
Unwarranted Arguments
0.5821 0.8871 0.656 0.5192
Knowledge Gain
-0.5393 5.4902 -0.098 0.9227
Multiple R-squared
0.2421
p-value 0.1284
Deliberative Democracy
• Additional Support for Merits to Deliberative Polling
• Sunstein/Manin Question– “What leads to group shifts?”
• Role of Moderators in Deliberation
Policy Recommendations
• Classroom Simulation• Congressional Committees• Federal Incentives to present multiple views
Acknowledgments
• Professor James Fishkin and Dr. Alice Siu at the Center for Deliberative Democracy (CDD)
• Professor Francis Fukuyama and Dr. Didi Kuo at the Center on Democracy, Development, and the Rule of Law (CDDRL)