Groupiness and Polarization 1 / 29
Lecture 3: Groupiness and Polarization
Mateusz Stalinski
EC340: Topics in Applied Economics (3a)
University of Warwick
Autumn 2023
Groupiness and Polarization 2 / 29
Introduction
Groupiness
▶ Today, we will formalize the notion of groupiness and political
polarization.
▶ Papers: Iyengar and Westwood (2015) , Iyengar et al. (2019).
▶ Outgroup is a group to to which a person does not belong.
▶ Ingroup is a group to which a person does belong.
▶ Identifying with a group may lead to negative evaluations of
the outgroup (outgroup bias).
▶ Or an excessively positive view of the ingroup (ingroup
favoritism).
▶ Groupiness can arise wrt to many characteristics.
▶ Race, ethnicity, nationality, gender, political identity, team.
Groupiness and Polarization 3 / 29
Introduction
Political Polarization
▶ Political identity creates clear ingroups and outgroups.
▶ Most focus on party affiliation (Democrats vs. Republicans).
▶ Ideological polarization: difference in (policy) positions
between (members of) political parties.
▶ Polarizing content is such that it increases the distance
between the groups.
▶ Media, especially social media (SM), play a key role.
▶ Today’s presentation will cover the role of SM algorithms
(Levy, 2021).
Groupiness and Polarization 4 / 29
Introduction
Affective Polarization
▶ The other type of polarization, affective polarization, is an
example of an outgroup bias in political context.
▶ Iyengar and Westwood (2015): “tendency of people identifying
as Republicans or Democrats to view opposing partisans
negatively and copartisans positively.”
▶ It amounts to disliking and distrusting members of the
opposing political party.
▶ Ideological and affective polarization appear linked, with the
former hypothesized to contribute to the latter.
Groupiness and Polarization 5 / 29
Introduction
Key Questions
1. How to measure affective polarization?
▶ Survey self-reports, implicit measures, behavioral measures.
2. What are welfare effects of affective polarization?
▶ Dating market, workplace discrimination, lack of cooperation.
3. Is affective polarization on the rise and what are the
contributing factors?
4. Is affective polarization stronger than other forms of outgroup
bias?
▶ Such as race-based outgroup bias.
Groupiness and Polarization 6 / 29
Introduction
Important Details
▶ Why out of all outgroup biases, affective polarization is
considered a distinct category?
▶ Lack of social norms against hostility along the party lines.
▶ After all, being able to voice disagreements is a key element of
the democratic process.
▶ Strong social norms against race-based discrimination.
▶ Political identity is less salient than race or gender.
▶ When is it optimal for people to reveal it?
Groupiness and Polarization 7 / 29
Empirical Evidence
Survey Measures
▶ The most common self-reported survey measure is the feeling
thermometer.
▶ On a scale from 0 (coldest) to 100 (warmest) how do you feel
about the following people and groups?
▶ Republicans and Democrats.
▶ Using standardized questions allows for easy comparison of
affective polarization over time (Iyengar et al., 2019) and
across countries (Boxell et al., 2022).
▶ However, many concerns:
▶ Not incentivized.
▶ Risk of exaggeration (comparison, experimenter demand).
▶ Risk of suppression (awareness of norm violations).
▶ Norm differences complicate benchmarking (politics vs. race).
Groupiness and Polarization 8 / 29
Empirical Evidence
Affective Polarization in the United States
Groupiness and Polarization 9 / 29
Empirical Evidence
Affective Polarization Across Countries
Groupiness and Polarization 10 / 29
Empirical Evidence
Trends: Results
▶ Affective polarization in the United States has been rising
since 1990s.
▶ Possible small reverse trend in 2012-2016.
▶ This is driven by outgroup bias (out-party feeling) rather than
ingroup favoritism (in-party feeling).
▶ The rise in affective polarization is particularly strong in the
U.S. in comparison to other countries.
▶ In some countries, we observe the reverse trend (Germany,
Sweden).
Groupiness and Polarization 11 / 29
Empirical Evidence
Implicit Measures
▶ Iyengar and Westwood (2015) used a Brief Implicit Association
Test, or BIAT (Study 1).
▶ Participants are shown pairs of words and images.
▶ Words can be either good or bad.
▶ Positive : Wonderful, Best, Superb, Excellent.
▶ Negative : Terrible, Awful, Horrible, Worst.
▶ Images relate to either different political parties or races.
▶ Democrats vs. Republicans.
▶ Texas/California, party logos, “partisan” NGOs.
▶ African Americans vs. European Americans.
▶ Young males.
▶ Neutral facial expressions, similar hairstyles and attractiveness.
Groupiness and Polarization 12 / 29
Empirical Evidence
BIAT Study
▶ Participants have to correctly classify words as positive or
negative (by pressing keys).
▶ Asked to proceed fast: “as quickly as you can while making as
few mistakes as possible.”
▶ Outcome: reaction time.
▶ Easier if good/bad word combined with ingroup/outgroup?
▶ Implicit measures address some of the concerns about the
self-reported survey measures.
▶ Fast speed makes it hard to go beyond natural reaction.
▶ Random order of race BIAT and politics BIAT.
▶ H: stronger effect for political affect than race-based affect.
▶ This is due to presence of social norms against racial
discrimination.
Groupiness and Polarization 13 / 29
Empirical Evidence
BIAT Study: Choice Screen
Groupiness and Polarization 14 / 29
Empirical Evidence
BIAT Study: Political Images
Groupiness and Polarization 15 / 29
Empirical Evidence
BIAT Study: Results (Politics)
Positive D-scores: participants respond faster to Republican-good
than to Democrat-good pairings.
Groupiness and Polarization 16 / 29
Empirical Evidence
BIAT Study: Results (Benchmarking Against Race)
Positive D-scores: affective preferences for Republicans or European
Americans
Groupiness and Polarization 17 / 29
Empirical Evidence
Explicit vs. Implicit Partisan Affect
Net feeling thermometer vs. partisan D-scores
Groupiness and Polarization 18 / 29
Empirical Evidence
BIAT Study: Summary
▶ Implicit measures confirm the prevalence of outgroup bias wrt
political affiliation and ideology.
▶ Less clear whether intensity of belonging moderates the effect.
▶ As expected, implicit measures are harder to move than
explicit ones (distributions).
▶ D-scores suggest that party-based polarization is greater than
race-based polarization.
▶ This is consistent with the social norms explanation.
Groupiness and Polarization 19 / 29
Empirical Evidence
Behavioral Evidence
▶ Iyengar and Westwood (2015) further use games to elicit
affective polarization (Study 3 and 4).
▶ Dictator and trust games.
▶ These can be used as incentivized measures.
▶ Participants are randomly assigned to play either four rounds
of the dictator game or the trust game.
▶ Dictator game: Player 1 splits $10 with Player 2 in any way
they want.
▶ Trust game: Player 1 receives $10 and can pass any amount
from $0-$10 to Player 2. That amount would triple. Then.
Player 2 has a chance to give back money to Player 1.
Groupiness and Polarization 20 / 29
Empirical Evidence
Player 2’s Characteristics
▶ Providing multiple traits is important.
▶ Age, gender, income, race/ethnicity, party affiliation.
▶ Reduces demand effects.
▶ Wiggle room encourages honesty.
▶ Placebo characteristics:
▶ Age randomized between 32 and 38
▶ Income from $39,000 to $42,300.
▶ Gender always male.
▶ Key variation:
▶ Partisanship: Democrat or Republican.
▶ Race: white or African American.
▶ Within-subjects design.
▶ Four rounds (one per characteristic) in random order.
Groupiness and Polarization 21 / 29
Empirical Evidence
Dictator Game Screen
Groupiness and Polarization 22 / 29
Empirical Evidence
Trust Game Screen
Groupiness and Polarization 23 / 29
Empirical Evidence
Results
Behavioral evidence of negative partisan affect both in the trust
game and the dictator game. No race-based outgroup bias.
Groupiness and Polarization 24 / 29
Empirical Evidence
Study 4: Adding Control and Independents
Outgroup bias and ingroup favoritism both play a role, with the
former being a bit stronger (especially in the trust game).
Groupiness and Polarization 25 / 29
Empirical Evidence
Study 2: Workplace Discrimination
Groupiness and Polarization 26 / 29
Empirical Evidence
Results
Copartisan candidates favored even if less qualified. Much weaker
effects for race.
Groupiness and Polarization 27 / 29
Empirical Evidence
Conclusion
▶ Understanding political polarization is an active research area.
▶ We need to continue documenting welfare impacts of
polarization.
▶ Lack of natural, large-scale studies.
▶ Typical settings: labor market, dating. Can we think of more?
▶ Politically motivated reasoning may have profound effects
when it implies ignoring science.
▶ Vaccine hesitancy, climate change, antibiotic resistance.
▶ Strength of politically motivated reasoning is likely a function
of ideological polarization.
▶ Lastly, depolarization efforts are hard.
▶ Exploring more potential solutions remains a priority.
Groupiness and Polarization 28 / 29
Presentations:
Presentations
▶ Presentation 1: Levy (2021)
▶ Social media contribute to creating echo chambers by
tailoring content, which can increase polarization.
▶ What is the role of social media algorithms?
▶ Can exposure to counter-attitudinal content reduce
affective/ideological polarization?
▶ Presentation 2: Schwardmann et al. (2022)
▶ Field experimental evidence in favor of self-persuasion.
▶ Goals affect beliefs.
▶ Staying loyal to one’s political party can be such a goal.
▶ This way, self-persuasion contributes to polarization.
Groupiness and Polarization 29 / 29
References
References
Boxell, L., M. Gentzkow, and J. M. Shapiro (2022): “Cross-country trends in affective
polarization,” Review of Economics and Statistics, 1–60.
Iyengar, S., Y. Lelkes, M. Levendusky, N. Malhotra, and S. J. Westwood (2019): “The
origins and consequences of affective polarization in the United States,” Annual
review of political science, 22, 129–146.
Iyengar, S. and S. J. Westwood (2015): “Fear and loathing across party lines: New
evidence on group polarization,” American journal of political science, 59, 690–707.
Levy, R. (2021): “Social media, news consumption, and polarization: Evidence from a
field experiment,” American economic review, 111, 831–870.
Schwardmann, P., E. Tripodi, and J. J. Van der Weele (2022): “Self-persuasion:
Evidence from field experiments at international debating competitions,” American
Economic Review, 112, 1118–1146.
