PreprintPDF Available

Bridging America's Divide on Abortion, Guns and Immigration: An Experimental Study

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Americans appear increasingly polarized and unable to bridge ideological divides. We study individuals' willingness to engage with others who hold opposite views on polarizing policies. Two thousand five hundred Americans are given the opportunity to listen to recordings of fellow countrymen and women expressing their views on immigration, abortion laws and gun ownership laws. We find that most Americans (more than two-thirds) are willing to listen to a view opposite to theirs, and a small fraction (ten percent) reports changing their views as a result. We also test whether emphasizing common grounds with those who think differently helps bridging views. We identify principles the vast majority of people agree upon: (1) a set of fundamental human rights, and (2) a set of simple behavioral etiquette rules. A random subsample of people are made explicitly aware they share common views, either on human rights or etiquette rules, before they have the opportunity to listen to different views. We find that the treatments induce people to adjust their views towards the center on abortion and immigration, relative to a control group, thus potentially reducing polarization.
Content may be subject to copyright.
Bridging America’s Divide on
Abortion, Guns and Immigration:
An Experimental Study
Mich`ele BelotGuglielmo Briscese
June 29, 2022
Abstract
Americans appear increasingly polarized and unable to bridge ideological divides.
We study individuals’ willingness to engage with others who hold opposite views on
polarizing policies. Two thousand five hundred Americans are given the opportunity
to listen to recordings of fellow countrymen and women expressing their views on
immigration, abortion laws and gun ownership laws. We find that most Americans
(more than two-thirds) are willing to listen to a view opposite to theirs, and a small
fraction (ten percent) reports changing their views as a result. We also test whether
emphasizing common grounds with those who think differently helps bridging views.
We identify principles the vast majority of people agree upon: (1) a set of fundamental
human rights, and (2) a set of simple behavioral etiquette rules. A random subsample
of people are made explicitly aware they share common views, either on human rights
or etiquette rules, before they have the opportunity to listen to different views. We find
that the treatments induce people to adjust their views towards the centre on abortion
and immigration, relative to a control group, thus potentially reducing polarization.
JEL Classification Codes: D83, D91, D72
Keywords
: Polarization, Contact theory, Willingness to listen, Abortion, Immigration, Gun
Laws.
The study was funded by Cornell University (Belot’s research funds). The authors thank Pilar Cardinale
and Sarahi Osuna for excellent research support. This study obtained IRB approval from Cornell University,
protocol no.: 2105010352. The experiment and the analyses were pregistered on the AEA RCT Registry, ID:
AEARCTR-0008029.
Cornell University mb2693@cornell.edu.
University of Chicago, gubri@uchicago.edu.
1
arXiv:2206.13652v1 [econ.GN] 27 Jun 2022
1 Introduction
On June 24th 2022 the Supreme Court overturned the landmark ruling “Roe vs Wade” that
recognized women’s constitutional right to abortion, triggering significant reactions from both
pro-choice and pro-life advocates. In the same week, the US Congress passed a gun control
bill - described as the most significant firearms legislation in nearly 30 years - after two mass
shootings in Buffalo, New York, and a primary school in Uvalde, Texas, that resulted in 31
deaths. The new measures were passed by 234 to 193 votes, a testimony to the polarized
views on gun ownership laws.
These significant developments are taking place at a time when America appears to
be more divided than ever on important policies, reflecting a growing concern among political
scientists on rising political polarization in the US and other Western societies (Mason, 2018).
1
Recent studies point at polarization also becoming “affective”, meaning ordinary Americans
see other fellow citizens holding opposite political ideologies as hypocritical, selfish, and
closed-minded (Iyengar et al., 2019). The concern is that American partisans are speaking
‘different languages’, misunderstanding one another, and distrusting each other on a basic
level (Mason, 2018). In response to these concerns, a number of initiatives have emerged to
bridge the divide. In the US, programs such as “One Small Step” or “Braver Angels” bring
together Americans of different political affiliations to exchange their views in workshops,
debates or one-on-one conversations.
The assumption that interactions and contact could bridge differences is rooted in a
long tradition of work in social and political psychology and the seminal work of Harvard
psychologist Gordon Allport (Allport (1954). Allport’s work inspired numerous studies (see
Pettigrew and Tropp (2006), Pettigrew and Tropp (2005), Zhou et al. (2019) for meta-studies),
which mostly support the hypothesis that ‘contact’ can bridge opposing views and attenuate
prejudice. This wealth of evidence raises the question: why have we not been successful in
bridging the divide? We see two major challenges. First, while engaging with someone who
1
A global Ipsos study, carried out in 27 countries for the BBC in 2018 finds that three in four people on
average across the 27 countries (76%) think society in their country is divided2.
2
thinks differently may help bridging views, social interactions are often endogenous: Most
of us choose with whom to interact, and even if there are opportunities to talk and listen
to individuals from other social groups, people tend to favor interactions with others who
are more similar to them (McPherson et al., 2001). A second reason is that there might be
specific conditions for contact to be beneficial. In his seminal work, Allport wrote: “The
effect [of contact] is greatly enhanced if this contact is sanctioned by institutional supports
(i.e., by law, custom or local atmosphere), and provided it is of a sort that leads to the
perception of common interests and common humanity between members of the two groups.”
(see discussions in Dixon et al. (2005) and McKeown and Dixon (2017)). That is, a key
condition for bridging views may be to first establish common grounds.
This paper aims to address these two challenges. First, in our setting, people will
have the opportunity to engage with others who think differently, rather than being “forced”
into an interaction. Concretely, we conduct an online study where participants are offered
the opportunity to listen to short recordings of others expressing their views on policies
that are known for being polarizing - namely, laws concerning abortion, gun ownership,
and immigration. Participants are aware that these individuals hold different views than
theirs and can choose to listen to them or not. Subsequently, participants are asked how the
recordings affected their views. The voluntary character of the interaction may well be key
in triggering a real willingness to listen. There is indeed a very well-established literature
in Psychology around the concept of “psychological reactance” (Brehm, 1966), which states
that if individuals feel their freedom is reduced or threatened, they may be motivationally
aroused to regain them. In other words, people may be more open to changing their views
if they are choosing whether to engage or not. Importantly, this setting allows us to study
directly the willingness to engage with others who think differently.
Our second contribution is to test an intervention targeting the key condition
outlined in Allport’s work: creating a sense of common humanity. Previous studies attempted
to create common goals or tasks by crafting situations that transform participants’ perceptions
of the memberships from “us” and “them” to a more inclusive “we” (Sherif (1988), Gaertner
et al. (1996)). These studies show that perceptions of a superordinate identity reduces
3
outgroup bias. We propose instead an intervention that attempts to emphasize the “human
identity” more directly. In our view, there is no more direct translation of humanity than
basic human rights. We first elicit views on a selected set of human rights statements as
written in the United Nations Universal Declaration of Human Rights. As expected, the
vast majority of people, independently of their race, gender or political affiliations, agrees
with these basic human rights. We then introduce a treatment where we inform participants
that they share the same views on human rights as those who expressed a view different
than theirs on abortion laws, gun laws or immigration. This offers, in our view, a unique
opportunity to foster a sense of common humanity.
Of course, it is possible that the mere emphasis of common attributes or interests, no
matter how futile they may be, could trigger a higher willingness to listen to others, as shown
by an extensive literature in social psychology and behavioral economics. The “minimal
group paradigm” shows that even attributing irrelevant group identity tokens to individuals
create a sense of group belonging and in and out-group dynamics (Tajfel et al. (1971), with a
recent review of experimental findings in Balliet et al. (2014)). To have a broader sense of
how the impact of raising awareness about commonality may affect willingness to engage, we
introduce a treatment where we inform participants that they share views on other, more
futile, attitudes, specifically basic behavioral etiquette rules.
We conduct the experiment with a representative sample of about 2,500 US citizens,
and collected audio recordings from a separate sample. We find a very high baseline level
of Americans’ willingness to listen: two-thirds of respondents are willing to listen to the
recording of someone holding views opposite to their own, with the treatments not significantly
shifting this tendency. Similarly, in the aggregate, views don’t change after having had the
opportunity to listen to someone holding opposing beliefs, but at the individual level 10%
of people do change their views. Emphasizing common grounds contributes to reducing
polarization on the topic of abortion laws and to some extent immigration, with more extreme
opinions shifting towards the centre. These treatments are however ineffective in reducing
polarization on gun laws.
4
The rest of the paper is organized as follows. Section 2 describes the experimental
design and Section 3 presents the results. Section 5 concludes.
2 Experimental Design
2.1 Providing an Opportunity to Engage
A key feature of our setting is to allow people to choose to engage with others who think
differently. We chose to operationalize this in the following way: Participants have the
opportunity to listen to a short recording of someone expressing a view opposite to their own
on a policy topic known to be polarizing. Listening is obviously a specific form of engaging,
but in our view, it is a crucial one. It is a prime mode of interaction in the current world, with
people choosing what material to read or listen to off or on-line. We chose to use recordings
because previous studies have shown that listening to another person’s voice might be a more
effective medium of communication compared to for example reading (Schroeder et al., 2017).
Further, we want to know whether listening to someone expressing their (opposite) opinion
on a policy leads to a change in views.
Note that we chose to give a choice to engage with someone with a different view or
not, and did not offer the alternative of engaging with someone with a similar view instead
(or any other alternative). The reason for not giving an alternative is because it is presumably
easy for participants to access people who share their views in real life, while it is presumably
much harder to access people who don’t.
The next section describes how we chose the common grounds used as treatments,
and the policies that were more likely to be polarizing. Section 2.4 describes the procedure
for collecting recordings of Americans expressing their views about the policies (immigration,
gun laws, and abortion). Finally, Section 2.5 describes the set up of the main experiment
and the methodology for measuring the outcomes of interest.
5
2.2 Common Values
Our treatments involve creating a sense of common interests and humanity. To do that, we
first needed to identify values and principles people most agree on. We began by shortlisting
14 statements about human rights from the United Nations Human Rights Declaration
(UNHRD) and 10 behavioral etiquette. The UNHRD, first proclaimed by the United Nations
General Assembly in 1948, sets out fundamental human rights to be universally protected
and is recognized to this date as the inspiration for human rights treaties around the world.
The UNHRD compromises a total of 30 articles, from which we selected the shortest 14 to
meet standard survey completion time constraints.
3
The behavioral etiquette rules were
selected from a private website providing information and education for refugees, asylum
seekers, immigrants and welcoming communities living in the United States.4
We then recruited a representative sample of 325 Americans and asked them to
rank, on a slider scale from 0 to 10, how much they agreed or disagreed with each of the
24 statements.
5
Note that we did not mention that these statements were derived from the
United Nations Declaration of Human Rights. We then selected the statements that had the
highest level of agreement as defined by the highest share of respondents who gave a score
above 7 or below 3, as per preregistered trial protocol. In addition, we performed a series of
tests that confirmed there was indeed no significant difference in the share of agreement on
these statements between (self-declared) Democrats and Republicans or across socio-economic
and demographic variables that were collected at the end of the survey (see Online Appendix
for survey instructions and balance tests).
The shortlist of human rights selected for the main experiment includes: (1) No one
shall be held in slavery or servitude; slavery and slave trade shall be prohibited in all their
forms; (2) No one shall be arbitrarily deprived of his property; (3) Everyone has the right to
3
For a copy of the full declaration, see:
https://www.un.org/en/about-us/universal-declaration-
of-human-rights
4See: https://usahello.org/life-in-usa/culture/be-polite/
5
Respondents took on average 4 minutes to complete the survey, and they were compensated for their
time as per Qualtrics panel payment agreement. All statements were shown randomly to avoid fatigue or
anchoring effects.
6
freedom of peaceful assembly and association; (4) Everyone has the right to life, liberty and
security of person; (5) Everyone has the right to freely to participate in the cultural life of the
community, to enjoy the arts and to share in scientific advancement and its benefits.
The shortlist of etiquette rules selected for the main experiment includes: (1) “Wait
your turn in a waiting line”; (2) “Say ‘please’ when you ask for something”; (3) “Say ‘thank
you’ to acknowledge service, kindness, or the receipt of something”; (4) “Arrive on time.
Not being late for classes, intimate gatherings, or appointments.”; (5) “Refrain from talking
loudly in quiet settings (or on your cell phone in movies, plays, or other quiet or focused
communal settings)”.
2.3 Polarizing Policies
We selected three policies that have been shown in several polls to be polarizing among
Americans, and that are often debated by politicians and media outlets: abortion, immigration
and gun ownership laws.
6
For consistency, we looked at the level of polarization on these
policies based on the most recent IPSOS polls available at the time of the design of the
experiment. A 2019 IPSOS poll found that while most Americans believed abortion should
be legal in most or all cases, but among Republicans only 35% believed so, compared to 75%
of Democrats.
7
A 2021 survey from the same company found the immigration was an equally
divisive topic.
8
Even before the COVID-19 pandemic, polls showed that 62% of Republicans
believed the number of immigrants allowed in the country should be decreased, compared to
about 25% of Democrats
9
. On gun laws, a 2017 IPSOS poll found that 65% of Democrats
6
In a first phase, we also included free trade as a fourth topic, as pollsters and commentators argued it
became a polarizing topic in recent American politics. The question read: “Free trade agreements between the
US and other countries have generally been [A bad thing-neither good nor bad-a good thing scale]”. However,
we were unable to reach a minimum sample of audio files of respondents who were against free trade, and
therefore decided to drop it from the main survey. We updated the pre-analysis plan in the trial registry
prior to the launch of the main survey.
7
Source:
https://www.ipsos.com/sites/default/files/ct/news/documents/2019-06/ipsos_usat_
abortion_topline_053119.pdf last accessed May 2022.
8
Source:
https://www.ipsos.com/en-us/news-polls/immigration-americans-favor-both-
restrictions-and-reforms last accessed May 2022.
9Source:https://www.ipsos.com/en-us/news-polls/americans-views-immigration-policy
7
believed that they should be a lot more strict compared to a lower 29%.10
2.4 Recordings of Views on Polarizing Policies
We collected audio recordings from Americans about the selected policies with a second
on-line survey, using the Qualtrics Survey Panel. Respondents were first asked to state their
level of agreement with each of the shortlisted set of human rights and etiquette rules from the
first study phase. Then, they were asked to indicate their views on the abortion, immigration,
and gun laws. The policy questions were shown on a slider scale 0 to 10, with text prompts on
each extreme of the sliding bar, and the survey was then programmed to identify the policy
that the respondent felt most strongly about as the one they assigned the highest or lowest
score of the three (and randomly to break ties).
11
Respondents were then shown a screen
reminding them of their views on the selected policy, and were asked to share an audio file
of yourself expressing, in about 60 seconds or less, your views about the topic. How would
you explain your point of view about this topic to someone else?”. Thus, each respondent
was asked to record their views only about one policy. To help them with the recording and
uploading of the audio files, we also showed respondents a page with user-friendly instructions
depending on the device they were using. After uploading the audio file, respondents were
asked a short set of questions about their socio-economic and demographic characteristics, as
well as their political views (see Online Appendix for a copy of the survey).
Respondents who completed the survey and correctly submitted an audio file received
a larger compensation for their time compared to standard surveys.
12
This second survey
was conducted over the Summer and early Fall 2021. We received around 100 audio files
that we screened to shortlist the ones that were at least 30 seconds long, were of sufficiently
good sound quality that could be listened from any device, and did not contain profanity or
10
Source:
https://www.ipsos.com/en-us/news-polls/npr-gun-control-2017-10
last accessed May
2022
11
The policy slider questions asked whether: Abortion should be [illegal-legal scale] in most cases; Current
gun laws in the United States are [too strict-about right-too lenient scale]; Legal immigrants in the United
States today [Burden the country by taking jobs, healthcare, and housing-strengthen the country through hard
work and talent scale]
12Compensation was based on Qualtrics panel payment agreement.
8
swearing. No other exclusion criteria were applied. This selection processes gave us 47 audio
files: 10 (7) against (pro) abortion, 4 (9) against (pro) immigration, and 9 (8) against (pro)
stricter gun laws.
2.5 Main Experiment
We launched the main survey online through Qualtrics on April 4 and closed it on April 8, 2022,
after reaching the pre-specified sample size of approximately 2,500 Americans. Respondents
who took the survey first saw a participant information statement (PIS) and consent form,
which instructed them that they would be asked to answer a number of question and be
given the opportunity to listen to audio recordings (without specifying the substance of these
recordings). The first survey questions asked respondents to rank, on a 0-to-10 slider scale,
their agreement with the human rights and behavioral etiquette statements selected from the
previous study phase, as well as their views on the three policies (immigration, abortion and
gun laws). The subsequent part of the survey informed participants that they would have
the opportunity to listen to recordings from people who have views that differed from theirs
on the three policies. Participants were explicitly reminded that they could choose to listen
to none, all, or only some of these files. The message also reminded respondents to make
sure their volume or headphones were on. The purpose of this additional instruction page
was to make sure that engagement with the audio file was not affected by any technical issue
(e.g., malfunctioning sound) and that respondents understood that their engagement with
the audio files was completely voluntary.
Respondents were then randomly allocated to control or one of the treatment groups.
The control group only saw a blank page asking them to click a button to continue to the next
page. The “Etiquette Rules” treatment group (hereafter, T1 group) saw a page reminding
them of the etiquette questions they were just asked, and were told: We asked these exact
same questions to the people who recorded the audio files you will be shown in the next pages
and each of them had a rate of agreement above 9 out of 10.”. We then showed them again
the full list of etiquette rules as an additional reminder. Respondents who were randomly
9
allocated to the Human Rights treatment group (hereafter, T2 group) saw an almost identical
page reminding them of the human rights questions they just replied to, and informing
them that the people who recorded the audio files they were about to be shown had a rate
of agreement above 9 out of 10 on those statements (see Appendix
??
for details of the
treatment messages). On average, respondents in the control group spent 8 seconds reading
this instruction page, compared to about 24 seconds in T1 group and 19 seconds in the T2
group.13
After reading the instructions, respondents landed on a page with the first recording.
They had to click on the Play Button to listen to it. They were automatically matched
with a recording of an American who expressed views opposed to theirs (as measured by
their response to the same policy questions). For example, if someone stated to be against
stricter gun laws at the beginning of the survey, we would show them an audio file of someone
explaining why they thought gun laws in the United States were too lenient. As we had
multiple audio files for each policy view, the recording was randomly selected from the
relevant group to (e.g., recordings in favor of stricter gun laws). Further, we also randomized
the order in which respondents saw the policies - that is, whether they first saw an audio
file about immigration, gun laws, or abortion, since the audio files were shown one per page,
sequentially.
While we used a hidden timer to track how many seconds respondents spent on each
page with an audio file, we also wanted to collect self-reported levels of engagement. Following
the three pages displaying each of the recordings, respondents were asked to report, for each
policy, (a) whether they listened to all, some, or none of the audio file, and (b) whether
listening to the recordings changed their views about that policy (conditional on listening
to it). If the respondent stated their views changed on a policy, they were asked again the
same policy view question as at the beginning of the survey; if they said their views didn’t
change, they were asked to explain why.
14
The last survey module collected respondents’
13
As explained in greater detail in the Results section, the difference between the two treatment groups
was not significant; t=1.10, Satterthwaite’s df=911, p=0.267
14The question asked: “Please indicate why your views on (policy) have not changed: (i) I already knew
and considered the arguments presented, (ii) I didn’t find the arguments convincing, (iii) other (free text)”
10
socio-economic and demographic characteristics, as well as their political ideology. We
purposely left these questions at the end to avoid priming respondents about their political
identity.
3 Descriptive statistics
3.1 Basic demographics
We first present basic descriptives the sample participating in the main experimental study.
The 2,507 Americans recruited through the Qualtrics Survey Panel were randomly allocated
across control and treatment groups. The randomization ensured an almost perfect balance
across all key demographic and socio-economic characteristics, in particular for political party
affiliation which is perhaps the most important correlate of views on the policies we selected.
Table A1 in Appendix A.1 reports the averages of the main socio-economic and demographic
characteristics, and the relative balance checks across groups.
3.2 Baseline levels of agreement
Figure 1 shows the baseline distribution of attitudes towards the three selected policies. As
expected, views on these topics are spread. On immigration, around 23% of the sample
gave a score between 0 and 4 out of 10, thus believing that legal immigrants in the United
States are more of a burden than a strength; conversely, 25% of respondents gave a score
of exactly 10, strongly agreeing that legal immigrants strengthen the country. Views on
abortion laws appear to the most polarized, with more than 40% of the sample answering at
the very extremes (0 or 10), with an almost equal split of 18% giving a score of 0, believing in
most cases abortion should be illegal, and about 25% giving a score of 10 (i.e., in most cases
abortion should be legal). Gun laws appears to be the relatively least polarizing policy of
the three, with about 16% of respondents giving a score of 5 out of 10, although about 23%
11
giving a score between 0 and 4 (‘too strict’), and the remaining 60% giving a score between 6
and 10 (‘too lenient’).
Figure 1: Baseline policy views
Notes.
The figure plots the share of respondents who gave a score between 0 and
10 to each of the three policy statements. The title and the x-axis label show the
exact wording shown to respondents on the slider questions.
In contrast, views on human rights and etiquette rules are, as expected, much more
aligned. For each of the questions corresponding to either category, more than 80% of the
sample reports a rate of agreement of 8 or more on a scale from 0 to 10 (see Figures 2 and 3).
12
Figure 2: Attitudes on human rights values
Notes.
The figure plots the share of respondents who gave a score between 0
(strongly disagree) to 10 (strongly agree) to each of the five human rights values
statements.
13
Figure 3: Attitudes on behavioral etiquette
Notes.
The figure plots the share of respondents who gave a score between 0
(strongly disagree) to 10 (strongly agree) to each of the five behavioral etiquette
statements.
14
4 Results
4.1 Willingness to engage
A key feature of our design is that participants are given an opportunity to listen to others,
knowing they have a different view, but are not forced to do so. Our first outcome of interest
is their willingness to engage (WTE). We defined two variables in our registered pre-analysis
plan to capture this outcome. The first outcome is the number of recordings listened to (0,
1, 2 or 3). A recording is considered as ’listened to’ if the participant listened to at least 5
seconds of it and indicated having listened to “some of it” or “all of it” in the post-treatment
questions. The second outcome is the percentage of recordings listened to, measured as the
time spent on the pages with the audio recordings divided by the total time length of the
recordings.
Result 1. Levels of engagement are high.
Our first result is that both measures indicate very high levels of engagement. The first
measure indicates that 69% of respondents listened to all three files, and about 18% to two
files. The second measure indicates that 56% of the sample listened to the recordings for an
amount of time that was equal to or greater than their total length (mean 1.03, s.d. 1.006).
(see Figures 6 and 7 in Appendix A.1 for the distribution graphs).
One concern is that participants had “nothing else to do” and therefore listened to
the recordings for that reason. We made sure the instructions clearly stated that they had the
choice to listen or not. Obviously, they could finish the survey faster if they would not spend
time listening to recordings. So while it is true that we cannot be sure that participants really
listened to the actual substance of the recording, it is notable to see such level of engagement
when it was clearly communicated to them that it was voluntary and there were no incentives
to do so.
Result 2. Emphasizing common grounds does not further increase will-
ingness to engage.
15
Our second result evaluates the effects of the two treatments on the WTE outcomes. For
the first outcome we report results from a negative binomial regression, and for the second
outcome we report the results of a Tobit regression since the outcome variable is truncated
at 100%, and 56% of our sample has a value of 100%.
15
Table 1 presents the estimated
treatment effects. We find precisely null effects, that is, we do not find any indication that
emphasizing common grounds increases people’s willingness to listen. Knowing that the
other person shares the same views on behavioral etiquette rules (T1/etiq) increases the
willingness to listen compared to the control group by a small and non-significant 0.007 points
on our first outcome measure, and increases also in a non-significant manner the percentage of
audio recordings listened to by around one percent compared to the control group. Knowing
that the other person shares the same views on human rights (T2/values) decreases in a
non-significant way respondents’ willingness to listen by 0.005 points, and increases, also
non-significantly, by around one percent the length of audios they listen to. These results
indicate that the high levels of engagement are not driven by the treatments. Even with no
other information about the other person, participants show a remarkably high willingness to
engage.
It is possible that these average null treatment effects mask underlying heterogeneity.
We examine this possibility by replicating the regressions by policy and by considering the
initial position of the respondents’ views on each policy. These additional analyses were
not pre-registered and are therefore exploratory. See the results in Table A2 and Table A3.
The regressions splitting by policy topic confirm the precisely null effects. The regressions
examining each policy topic and conditioning on respondents’ initial stand on such policies
instead shows significant heterogeneous effects. Specifically, emphasizing shared support for
behavioral etiquette rules increases a person’s WTE on abortion and gun laws among people
who had more conservative views on these policies. However, we also find that having more
conservative views on abortion and gun laws are correlated with stronger beliefs on behavioral
etiquette rules; this is instead not the case for immigration and this is also where we do not
15
In the pre-analysis plan we announced we would run a linear regression, but did not anticipate that
more than half of our sample would be truncated at the top. Given the data structure, a Tobit model is more
appropriate.
16
find an effect. Thus, it could be that agreeing with the basic etiquette rules is correlated with
more conservative attitudes, and for this reason emphasizing common agreement on these
rules might affect conservatives more.
Table 1: WTE outcomes
(1) (2)
Willingness
to engage
Percentage total
audios length
listened to
T1/etiq 0.00709 0.0120
(0.0308) (0.0335)
T2/values -0.00517 0.00136
(0.0308) (0.0333)
lnα-36.07
var(e.outcome) 0.365***
(0.0178)
Constant 0.925*** 1.068***
(0.0217) (0.0250)
Observations 2,507 2,507
Notes.
Model (1) is a negative binomial regression, with likelihood-ratio test of
α
=0 is chibar2(01) = 0.00. Model (2) is a Tobit regression T1/etiq and T2/values
refer to the two treatment groups. The pseudo-R
2
is less than 0 for both models.
∗∗∗ p < 0.01; ∗∗ p < 0.05; p < 0.1.
4.2 Changes in views
The second outcome of interest is the change in views. We are interested in the magnitude of
the change and the direction (towards more or less extreme values).
Result 3. Ten percent of participants report changing their views after
listening to the recordings.
The vast majority of participants does not change their view after having had the opportunity
to listen to the recordings. Remarkably though, approximately 10% of the respondents do
report changing their views. This is true across all three policies. Given the short duration
17
of the recordings, one would not expect views on such fundamental and polarizing policies to
shift so rapidly, indicating that listening to others even for a short time might be a powerful
intervention.
Respondents who reported not having changed their views, were asked to explain
why by selecting on of the following three options: I already knew and considered the
arguments”, I didn’t find the arguments convincing ”, or Other ”. We see roughly the
same split between the two main reasons across all three policies. About 51% of those who
didn’t change their views on gun ownership laws said that they already knew the arguments,
compared to 44% of said they didn’t find the arguments convincing. On immigration laws,
48% already considered the arguments, while 46% didn’t find them convincing. Lastly, about
52% of those who didn’t change their views on abortion laws said that they already knew the
arguments, compared to 40% of said they didn’t find the arguments convincing.
Result 4. In the aggregate the distribution of views remains unchanged
after listening to the recordings.
Table 2 and Table 3 present summary statistics of the views and their firmness (absolute
distance from 5) for each of the policy. Figure 4 shows histograms of the distribution of
views on each topic before and after having the opportunity to engage. We see that the
distributions remain almost identical before and after. Among those who change view, the
size of the change varies from 2 points on the scale from 1 to 10 (for the control group and
gun laws) to 2.7 (for the Etiquette treatment and immigration). These changes are sizable
and correspond to two-thirds of the standard deviation in the initial views.
18
Table 2: Views and Firmness of Views - Before and After Opportunity to Engage
Views (0-10)
Means (standard deviations)
Firmness of views
(distance from 5)
Means
Immigration Gun laws Abortion Immigration Gun laws Abortion
All - before 6.6 (3.2) 6.4 (3.1) 5.6 (3.8) 3.2 2.8 3.5
All - after 6.6 (3.2) 6.4 (3.1) 5.5 (3.8) 3.2 2.8 3.5
p-value, t-test before= after 0.77 0.66 0.7 0.00 0.69 0.77
(1) Control - before 6.6 (3.2) 6.4 (3.1) 5.5 (3.7) 3.2 2.8 3.3
(2) Control - after 6.6 (3.2) 6.3 (3.1) 5.4 (3.8) 3.2 2.8 3.4
p-value, t-test (1)=(2) 0.95 0.45 0.70 0.69 0.94 0.63
(3) T1/etiq - before 6.5 (3.3) 6.4 (3.1) 5.6 (3.8) 3.3 2.8 3.5
(4) T1/etiq - after 6.4 (3.3) 6.4 (3.1) 5.6 (3.8) 3.1 2.9 3.4
p-value, t-test (3)=(4) 0.51 0.96 0.86 0.74 0.69 0.65
(5) T2/values - before 6.8 (3.2) 6.4 (3.1) 5.6 (3.8) 3.3 2.9 3.5
(6) T2/values - after 6.8 (3.3) 6.4 (3.1) 5.6 (3.8) 3.3 2.9 3.5
p-value, t-test (5)=(6) 0.93 0.98 0.67 0.81 0.83 0.77
p-value, DD T1/etiq vs Control 0.09 0.12 0.69 0.06 0.32 0.01
p-value, DD T2/values vs Control 0.94 0.08 0.27 0.70 0.68 0.03
Notes.
The table reports the average and standard deviation (in parenthesis) for
views and firmness of views (distance from 5) before and after the opportunity to
listen to the recordings.
19
Table 3: Change in Views (Absolute values) - Overall and Conditional on Changing Views
Absolute change in views
|after - before|
Absolute change in views
|after - before|
(conditional on changing)
Immigration Gun laws Abortion Immigration Gun laws Abortion
All
(1) Control 0.3 (1.3) 0.3 (1.2) 0.3 (1.3) 2.3 (2.5) 2.0 (2.3) 2.6 (2.9)
(2) T1/etiq 0.4 (1.4) 0.3 (1.2) 0.3 (1.3) 2.7 (2.8) 2.4 (2.7) 2.6 (2.8)
(3) T2/values 0.4 (1.4) 0.4 (1.4) 0.3 (1.3) 2.4 (2.6) 2.5 (2.7) 2.4 (2.9)
p-value, t-test (1)=(2) 0.84 0.53 0.87 0.24 0.29 0.97
p-value, t-test (1)=(3) 0.9 0.33 0.43 0.66 0.16 0.51
Notes.
The table reports the average and standard deviation (in parenthesis) for
absolute changes in views following the opportunity to listen to the recordings.
20
Figure 4: Changes in views
21
Result 5. Emphasizing common grounds does not affect views on aver-
age.
We now turn to the treatment effects of emphasizing common grounds. We pre-registered the
analysis of how treatments might change views on the selected policies as a simple before-after
comparison.
In the first row of Table 4 we report the average and standard deviation of each
outcome variable. The next rows show the estimated coefficients. Here again, we do not
find significant effects, but the estimates are less precise. From models (1) to (3) in Table 4
we can see that the average changes in views is about 0.35 for each policy. Compared to
the control group, the T1 behavioral etiquette rules treatment increases changes in views
on immigration and abortion by a small and insignificant 0.01 points approximately, and it
decreases it by around 0.03 points, also insignificantly, on views about gun ownership. The
T2 human rights values treatment has a slightly larger effect in increases views changes on
immigration and gun laws, and a very small negative effect on views about abortion laws,
although all coefficients are statistically insignificant.
Thus, so far, we find evidence that engagement is large and that a significant fraction
of participants changes their views after having listened to others’ views. In view of these
very high baseline levels, it is perhaps not surprising that emphasizing common grounds does
not have much effect.
As per the previous outcomes, we conduct additional heterogeneous analysis consid-
ering the respondents’ initial views. We see that emphasizing shared beliefs on human rights
values has an impact on views on gun laws for those who started with a more conservative
initial position, but all the other estimated coefficients are not statistically significant (see
Table A4 for the regression results).
Result 6. Emphasizing common grounds leads to less polarized views on
abortion and immigration.
The last pre-registered outcome of interest is individuals’ firmness of views that is, how
polarized to the extremes individuals are on a scale 0 to 10 for each policy.
22
Recall that the firmness of views is calculated as the absolute distance from 5 (the
middle point on our 0 to 10 scale) for each policy. The difference in firmness before and after
takes values of -5 to 5.
We report the results in models (5) to (7) in Table 4. Looking at the constant term
first, we find that polarization increases marginally in the control group for views on abortion.
There is no significant change in firmness of views on the two other topics. We also find
compelling evidence that the treatments have an impact in reducing respondents’ firmness of
views. Knowing that the person who recorded the audio files shared the same behavioral
etiquette preferences (T1 treatment) significantly decreases the listeners’ firmness of their
views on immigration and abortion, although not gun ownership laws. Knowing that the
person who recorded the audio files shared the same human rights values (T2 treatment)
significantly decreases the listeners’ firmness of views on abortion. We find no significant
treatment effects for views on gun laws.
23
Table 4: Changes in views and firmness of views
(1) (2) (3) (4) (5) (6)
in views
Immigration
in views
Abortion
in views
Gun laws
in firmn.
of views
Immigration
in firmn.
of views
Abortion
in firmn.
of views
Gun laws
Avg 0.362 0.329 0.347 0.008 - 0.0071 0.0207
(1.337) (1.345) (1.301) (0.6214) (0.6188) (0.6002)
T1/etiq 0.0110 0.0132 -0.0398 -0.0575* -0.0791*** 0.0291
(0.0655) (0.0659) (0.0637) (0.0304) (0.0303) (0.0294)
T2/values 0.0518 -0.00798 0.0613 -0.0118 -0.0642** 0.0120
(0.0653) (0.0657) (0.0635) (0.0303) (0.0302) (0.0293)
Constant 0.342*** 0.328*** 0.340*** 0.0309 0.0404* 0.00713
(0.0461) (0.0464) (0.0449) (0.0214) (0.0213) (0.0207)
Observations 2,507 2,507 2,507 2,507 2,507 2,507
R20.000 0.000 0.001 0.002 0.003 0.000
Notes.
Models (1) to (3) report changes in views, while models (4) to (6) report
changes in firmness of views. The first row reports the average and standard
deviation of each outcome variable, which can take values between -5 and +5 for
models (5) to (7). All models are OLS.∗∗∗ p < 0.01; ∗∗ p < 0.05; p < 0.1.
24
We also conduct a herogeneity analysis, this time distinguishing between people
who have more extreme values to start with (i.e. 8 or above, or 2 and below) from people
who are more neutral (initial views between 3 and 7). We find that the treatment effects
are driven by those who are more extreme on abortion laws, and become significantly less
extreme when common grounds are emphasized.
5 Conclusions
Political polarization is a major concern in American and other Western societies. In a world
where citizens with different views do not talk and listen to one another online or offline,
the opportunity to bridge gaps in views on important policies appears to be a challenging
goal. A major obstacle in increasing contact between partisans is that people choose whom
they interact with and could deliberately avoid engaging with others who thing differently.
Numerous efforts have been undertaken to increase contact between individuals with different
backgrounds and political affinities, but these initiatives may not be very effective on a larger
scale if people are not willing to listen to others who do not share their views without any
incentive to do so.
In this paper, we present evidence based on a representative sample of 2,500
respondents that Americans’ willingness to engage with others on polarizing policies (abortion
laws, gun ownership laws and immigration) is in fact high. We also find that a non-negligible
fraction of people ( 10 percent) reports changing their views after having listened to someone
expressing their opposite views. We examined whether emphasizing common grounds, such
as shared agreement on basic human rights or on simple behavioral etiquette rules, further
raises the willingness to engage. This is not the case overall, although we have suggestive
evidence that for more conservative Americans a shared view on behavioral etiquette rules
might increase their willingness to listen to someone they disagree with. We find that listening
to others who hold opposite views doesn’t change the distribution of views on average but
emphasizing common grounds does contribute to reduce polarization in views on abortion
25
laws and immigration.
The paper is the first to examine directly the willingness to engage with others who
do not share the same views on important policies. We also confirm Allport’s hypothesis that
contact is effective in altering views when common grounds are emphasized, both when these
are common views on human rights and on basic behavioral etiquette rules.
Overall, our take-away message from this paper is an encouraging one for those
advocating a reduction in political polarization: most people are willing to listen to others
with opposite views, and a small but crucial fraction of people is willing to change their views
after listening to others they disagree with. Most importantly, overall views become less
polarized after listening to others when common grounds are emphasized. These findings
suggest that interventions aimed at bridging the partisan gap can be effective, particularly
those that offer opportunities to voluntarily engage with others and emphasize common
grounds.
References
Allport, G. (1954). The Nature of Prejudice. Cambridge,Mass.: Addison-Wesley Pub.
Balliet, D., J. Wu, and C. K. De Dreu (2014). Ingroup favoritism in cooperation: a meta-
analysis. Psychological bul letin 140 (6), 1556.
Dixon, J., K. Durrheim, and C. Tredoux (2005). Beyond the optimal contact strategy: A
reality check for the contact hypothesis. American psychologist 60 (7), 697.
Gaertner, S. L., J. F. Dovidio, and B. A. Bachman (1996). Revisiting the contact hypothesis:
The induction of a common ingroup identity. International Journal of Intercultural
Relations 20 (3), 271–290. Prejudice, Discrimination and Conflict.
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 (1), 129–146.
Mason, L. (2018). Losing common ground: Social sorting and polarization. The Forum 16 (1),
47–66.
McKeown, S. and J. Dixon (2017). The “contact hypothesis”: Critical reflections and future
directions. Social and Personality Psychology Compass 11 (1), e12295.
26
McPherson, M., L. Smith-Lovin, and J. M. Cook (2001). Birds of a feather: Homophily in
social networks. Annual Review of Sociology 27 (1), 415–444.
Pettigrew, T. F. and L. R. Tropp (2005). Allport’s intergroup contact hypothesis: Its history
and influence. On the nature of prejudice: Fifty years after Allport, 262–277.
Pettigrew, T. F. and L. R. Tropp (2006). A meta-analytic test of intergroup contact theory.
Journal of personality and social psychology 90 (5), 751.
Schroeder, J., M. Kardas, and N. Epley (2017). The humanizing voice: Speech reveals,
and text conceals, a more thoughtful mind in the midst of disagreement. Psychological
science 28 (12), 1745–1762.
Sherif, M. (1988). The robbers cave experiment: Intergroup conflict and cooperation.[Orig.
pub. as Intergroup conflict and group relations]. Wesleyan University Press.
Tajfel, H., M. G. Billig, R. P. Bundy, and C. Flament (1971). Social categorization and
intergroup behaviour. European Journal of Social Psychology 1 (2), 149–178.
Zhou, S., E. Page-Gould, A. Aron, A. Moyer, and M. Hewstone (2019). The extended contact
hypothesis: A meta-analysis on 20 years of research. Personality and Social Psychology
Review 23 (2), 132–160.
27
A Appendix A
A.1 Additional tables and figures
Table A1: Sample characteristics & balance checks
(1) (2) (3) (4) (5) (6)
Control T1/etiquette T2/values T1 vs C T2 vs C T1 vs T2
Female 0.499 0.475 0.484 -0.024 -0.014 -0.009
(0.500) (0.500) (0.500) (0.024) (0.024) (0.025)
Age 18-34 0.305 0.276 0.261 -0.030 -0.044** 0.014
(0.461) (0.447) (0.440) (0.022) (0.022) (0.022)
Age 55-70 0.297 0.289 0.310 -0.008 0.013 -0.021
(0.457) (0.454) (0.463) (0.022) (0.022) (0.022)
Age 35-54 0.279 0.307 0.297 0.028 0.018 0.010
(0.449) (0.462) (0.457) (0.022) (0.022) (0.023)
Age 70+ 0.118 0.126 0.130 0.008 0.012 -0.004
(0.322) (0.332) (0.337) (0.016) (0.016) (0.016)
HS educ or more 0.982 0.985 0.971 0.003 -0.011 0.014**
(0.132) (0.120) (0.167) (0.006) (0.007) (0.007)
Income 31K or less 0.385 0.393 0.391 0.008 0.007 0.002
(0.487) (0.489) (0.488) (0.024) (0.024) (0.024)
Income 31-70K 0.420 0.393 0.407 -0.027 -0.014 -0.014
(0.494) (0.489) (0.492) (0.024) (0.024) (0.024)
Income 70K+ 0.195 0.214 0.202 0.019 0.007 0.012
(0.396) (0.410) (0.401) (0.020) (0.019) (0.020)
Democrat 0.398 0.366 0.405 -0.031 0.007 -0.038
(0.490) (0.482) (0.491) (0.024) (0.024) (0.024)
Republican 0.397 0.417 0.407 0.020 0.010 0.010
(0.489) (0.493) (0.492) (0.024) (0.024) (0.024)
Indep/Centrist/others 0.205 0.216 0.189 0.011 -0.017 0.028
(0.404) (0.412) (0.391) (0.020) (0.019) (0.020)
Observations 842 827 838 1,669 1,680 1,665
Notes.
Columns 1 to 3 report the (mean) share of respondents in each of the three
experimental groups across key demographic and socio-economic characteristics,
with standard deviation in parenthesis. Columns 3 to 6 report the results of
between sub-sample balance tests. We record a marginally smaller share (4%) of
respondents aged 18 to 34 in the T2 group compared to control, and a marginally
larger share (1%) of more educated respondents in T1 compared to T2. None
of these differences affect our main results in any meaningful way.
∗∗∗ p <
0
.
01;
∗∗ p < 0.05; p < 0.1.
28
Figure 5: Time spent on page as fraction of length of recordings
Notes.
The figure plots the frequency of the time spent on the three audio files
as a fraction of the total length of the recordings.
Figure 6: Time spent on page as fraction of length of recordings,
censored
Notes.
The figure plots the frequency of the time spent on the three audio files
as a fraction of the total length of the recordings. Any values above 1 is equalized
to 1, as per Tobit regressions in the main text.
29
Figure 7: Self-reported engagement
Notes.
The figure plots the self-reported engagement with the recordings for each
policy.
Table A2: WTE by policy
(1) (2) (3)
WTE
Immigration
WTE
Abortion
WTE
Gun laws
T1/etiq -0.121 0.131 0.0706
(0.159) (0.130) (0.124)
T2/values -0.106 0.0268 -0.0431
(0.159) (0.128) (0.122)
Constant 2.187*** 1.521*** 1.404***
(0.114) (0.0898) (0.0866)
Observations 2,507 2,507 2,507
Notes.
Models (1) to (3) are Logit regressions. T1/etiq and T2/values refer to
the two treatment groups.∗∗∗ p < 0.01; ∗∗ p < 0.05; p < 0.1.
30
Table A3: WTE by policy and pre-treatment views
(1) (2) (3) (4) (5) (6)
WTE
Immigration
WTE
Immigration
WTE
Abortion
WTE
Abortion
WTE
Gun laws
WTE
Gun laws
T1/etiq 0.130 -0.311 0.461** -0.0605 0.505* -0.0629
(0.249) (0.213) (0.218) (0.165) (0.259) (0.142)
T2/values -0.0293 -0.214 0.112 -0.0323 0.0965 -0.0870
(0.246) (0.215) (0.200) (0.166) (0.239) (0.142)
Constant 1.326*** 2.651*** 1.379*** 1.616*** 1.191*** 1.473***
(0.172) (0.160) (0.139) (0.118) (0.170) (0.101)
Observations 597 1,910 943 1,564 601 1,906
Notes.
All models are Logit regressions. Models (1), (3), and (5) only consider
the subset of respondents who gave a score strictly less than 5 out of 10 on each
of the policies at the beginning of the experiment. Models (2), (4) and (6) only
consider the subset of respondents who gave a score 5 or above out of 10 on each
of the policies at the beginning of the experiment. T1/etiq and T2/values refer to
the two treatment groups.∗∗∗ p < 0.01; ∗∗ p < 0.05; p < 0.1.
Table A4: Changes in views conditional on priors
(1) (2) (3) (4) (5) (6)
in views
Immigration
∆in views
Immigration
in views
Abortion
in views
Abortion
in views
Gun laws
in views
Gun laws
T1/etiq -0.154 0.0641 0.0232 0.00771 0.0780 -0.0774
(0.174) (0.0662) (0.107) (0.0836) (0.170) (0.0644)
T2/values 0.174 0.0203 -0.0571 0.0225 0.331* -0.0279
(0.176) (0.0656) (0.105) (0.0841) (0.169) (0.0643)
Constant 0.522*** 0.284*** 0.331*** 0.326*** 0.342*** 0.339***
(0.122) (0.0466) (0.0743) (0.0594) (0.121) (0.0451)
Observations 597 1,910 943 1,564 601 1,906
R-squared 0.006 0.001 0.001 0.000 0.007 0.001
Notes.
All models are OLS regressions. Models (1), (3), and (5) only consider
the subset of respondents who gave a score strictly less than 5 out of 10 on each
of the policies at the beginning of the experiment. Models (2), (4) and (6) only
consider the subset of respondents who gave a score 5 or above out of 10 on each
of the policies at the beginning of the experiment. T1/etiq and T2/values refer to
the two treatment groups.∗∗∗ p < 0.01; ∗∗ p < 0.05; p < 0.1.
31
Table A5: Changes in firmness of views conditional on priors
(1) (2) (3) (4) (5) (6)
Changes firmn.
of views
Immigration
[Neutral]
Changes firmn.
of views
Immigration
[Extreme]
Changes firmn.
of views
Abortion
[Neutral]
Changes firmn.
of views
Abortion
[Extreme]
Changes firmn.
of views
Gun laws
[Neutral]
Changes firmn.
of views
Gun laws
[Extreme]
T1/etiq 0.117 0.0783 -0.0482 -0.195** 0.268 -0.000787
(0.156) (0.0934) (0.0300) (0.0957) (0.173) (0.0994)
T2/values 0.0549 0.0441 -0.0131 -0.265*** -0.0277 0.148
(0.157) (0.0923) (0.0297) (0.0982) (0.175) (0.0983)
Constant 1.577*** -0.263*** -0.0307 0.350*** 2.336*** -0.237***
(0.110) (0.0654) (0.0212) (0.0660) (0.123) (0.0696)
Observations 823 2,057 2,079 428 742 1,765
R-squared 0.001 0.000 0.001 0.019 0.005 0.002
Notes.
All models are OLS regressions. Models (1), (3), and (5) only consider
the subset of respondents who gave a score between 3 and 7 (included) on each
of the policies at the beginning of the experiment. Models (2), (4) and (6) only
consider the subset of respondents who gave a score of o to 2, or on each of the
policies at the beginning of the experiment. T1/etiq and T2/values refer to the
two treatment groups.∗∗∗ p < 0.01; ∗∗ p < 0.05; p < 0.1.
32
B Appendix B
B.1 Treatment messages
Figure 8: Treatment instructions text
33
B.2 Surveys
See Online Appendix Link for the full surveys conducted for (1) eliciting views on Human
Rights and Basic Etiquette rules, (2) collecting recordings on views on Immigration, Abortion
and Gun laws, (3) the main experiment.
34
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
A person’s speech communicates his or her thoughts and feelings. We predicted that beyond conveying the contents of a person’s mind, a person’s speech also conveys mental capacity, such that hearing a person explain his or her beliefs makes the person seem more mentally capable—and therefore seem to possess more uniquely human mental traits—than reading the same content. We expected this effect to emerge when people are perceived as relatively mindless, such as when they disagree with the evaluator’s own beliefs. Three experiments involving polarizing attitudinal issues and political opinions supported these hypotheses. A fourth experiment identified paralinguistic cues in the human voice that convey basic mental capacities. These results suggest that the medium through which people communicate may systematically influence the impressions they form of each other. The tendency to denigrate the minds of the opposition may be tempered by giving them, quite literally, a voice.
Article
Full-text available
Research on intergroup contact has grown exponentially over the past decade. Such research has typically extolled the benefits of positive interaction between members of historically divided communities, particularly on outcomes related to prejudice reduction. Emerging work in the field, however, has qualified this optimistic picture by identifying three gaps in the existing literature. First, in everyday life, contact may be construed as a negative experience that increases rather than decreases responses such as prejudice, anxiety, and avoidance. Second, in real-life settings, contact is often circumscribed by informal practices of (re)segregation that are easily overlooked if researchers rely primarily on examining structured contact and explicit processes using primarily laboratory and questionnaire methods. Third, positive contact may have “ironic” effects on the political attitudes and behaviors of the historically disadvantaged, undermining their recognition of social injustice and decreasing their willingness to engage in collective action to challenge the status quo. Although it is now a truism that intergroup contact can reduce intergroup prejudice, these developments emphasize the importance of maintaining a critical perspective on the “contact hypothesis” as a model for promoting social change in historically divided and unequal societies. They also lay the foundations for future developments in the field.
Article
Full-text available
Although theory suggests individuals are more willing to incur a personal cost to benefit ingroup members, compared to outgroup members, there is inconsistent evidence in support of this perspective. Applying meta-analytic techniques, we harness a relatively recent explosion of research on intergroup discrimination in cooperative decision making to address several fundamental unresolved issues. First, summarizing evidence across studies, we find a small to medium effect size indicating that people are more cooperative with ingroup, compared to outgroup, members (d = 0.32). Second, we forward and test predictions about the conditions that moderate ingroup favoritism from 2 influential perspectives: a social identity approach and a bounded generalized reciprocity perspective. Although we find evidence for a slight tendency for ingroup favoritism through categorization with no mutual interdependence between group members (e.g., dictator games; d = 0.19), situations that contain interdependence result in stronger ingroup favoritism (e.g., social dilemmas; d = 0.42). We also find that ingroup favoritism is stronger when there is common (vs. unilateral) knowledge of group membership, and stronger during simultaneous (vs. sequential) exchanges. Third, we find support for the hypothesis that intergroup discrimination in cooperation is the result of ingroup favoritism rather than outgroup derogation. Finally, we test for additional moderators of ingroup favoritism, such as the percentage of men in the sample, experimental versus natural groups, and the country of participants. We discuss the implications of these findings for theoretical perspectives on ingroup favoritism, address implications for the methodologies used to study this phenomenon, and suggest directions for future research. (PsycINFO Database Record (c) 2014 APA, all rights reserved).
Chapter
Full-text available
This chapter contains section titled:
Article
Full-text available
Similarity breeds connection. This principle--the homophily principle--structures network ties of every type, including marriage, friendship, work, advice, support, information transfer, exchange, comembership, and other types of relationship. The result is that people's personal networks are homogeneous with regard to many sociodemographic, behavioral, and intrapersonal characteristics. Homophily limits people's social worlds in a way that has powerful implications for the information they receive, the attitudes they form, and the interactions they experience. Homophily in race and ethnicity creates the strongest divides in our personal environments, with age, religion, education, occupation, and gender following in roughly that order. Geographic propinquity, families, organizations, and isomorphic positions in social systems all create contexts in which homophilous relations form. Ties between nonsimilar individuals also dissolve at a higher rate, which sets the stage for the formation of niches (localized positions) within social space. We argue for more research on: (a) the basic ecological processes that link organizations, associations, cultural communities, social movements, and many other social forms; (b) the impact of multiplex ties on the patterns of homophily; and (c) the dynamics of network change over time through which networks and other social entities co-evolve.
Article
While previously polarization was primarily seen only in issue-based terms, a new type of division has emerged in the mass public in recent years: Ordinary Americans increasingly dislike and distrust those from the other party. Democrats and Republicans both say that the other party’s members are hypocritical, selfish, and closed-minded, and they are unwilling to socialize across party lines. This phenomenon of animosity between the parties is known as affective polarization. We trace its origins to the power of partisanship as a social identity, and explain the factors that intensify partisan animus. We also explore the consequences of affective polarization, highlighting how partisan affect influences attitudes and behaviors well outside the political sphere. Finally, we discuss strategies that might mitigate partisan discord and conclude with suggestions for future work. Expected final online publication date for the Annual Review of Political Science Volume 22 is May 11, 2019. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.
Article
The alignment between partisan and other social identities has generated a rift between Democrats and Republicans that is deeper than any seen in recent American history. Without the cross-cutting identities that have traditionally stabilized the American two-party system, partisans in the American electorate are now seeing each other through prejudiced and intolerant eyes. In this article, partisan social divisions are examined in the 2016 electorate, with an eye to the co-occurrence between these social divisions and contemporaneous resistance to compromise and intolerance of social outgroups. In particular, as Republicans and Democrats grow more socially distinct, they like outgroups less and privilege victory over the national greater good. This effect is particularly visible among Republicans, whose social makeup is particularly homogeneous, even in comparison with Democrats. Some potential solutions are briefly addressed.
Article
According to the extended contact hypothesis, knowing that in-group members have cross-group friends improves attitudes toward this out-group. This meta-analysis covers the 20 years of research that currently exists on the extended contact hypothesis, and consists of 248 effect sizes from 115 studies. The aggregate relationship between extended contact and intergroup attitudes was r = .25, 95% confidence interval (CI) = [.22, .27], which reduced to r = .17, 95% CI = [.14, .19] after removing direct friendship’s contribution; these results suggest that extended contact’s hypothesized relationship to intergroup attitudes is small-to-medium and exists independently of direct friendship. This relationship was larger when extended contact was perceived versus actual, highlighting the importance of perception in extended contact. Current results on extended contact mostly resembled their direct friendship counterparts, suggesting similarity between these contact types. These unique insights about extended contact and its relationship with direct friendship should enrich and spur growth within this literature.
Article
This paper reviews evidence pertaining to the Common Ingroup Identity Model for reducing intergroup bias. This model proposes that intergroup bias and conflict can be reduced by factors that transform members cognitive representations of the memberships from two groups to one more inclusive social entity. Theoretically, a common ingroup identity extends or redirects the cognitive and motivational processes that produce positive feelings toward ingroup members to former outgroup members. It is proposed that the prerequisite features specified by the contact Hypothesis (Allport, 1954; Cook, 1985), such as equal status between the memberships, cooperative interdependence, opportunity for self-revealing interactions and egalitarian norms, successfully reduce bias, in part, because they help transform members' perceptions of the memberships from “Us” and “Them” to a more inclusive “We”. Evidence from a laboratory experiment, two survey studies involving students attending a multi-ethnic high school and executives who have experienced a corporate merger, and a field experiment involving fans attending a college football game are summarized. In general, across these diverse settings, greater perceptions of a superordinate identity predicted lower levels of intergroup bias toward original outgroup members. In particular, this presentation of our research discusses the promise of exploring the role of a dual identity, in which both sub-group and superordinate group identities exist simultaneously, and how contextual features of the situational context may moderate the relation between the strength of a dual identity and intergroup attitudes.