seeing blue in black and white: race and perceptions of ... · shooting of michael brown in...

! 1

Seeing Blue in Black and White: Race and Perceptions of Officer-Involved Shootings

Hakeem J. Jefferson Fabian G. Neuner

Josh Pasek University of Michigan

Version: April 14, 2017

Abstract: Following racially charged events, individuals often diverge in their perceptions of what happened and how justice should be served. Examining data gathered shortly after the 2014 shooting of Michael Brown in Ferguson, Missouri alongside reactions to a novel officer-involved shooting, this study unpacks the processes by which racial divisions emerge. Compared to Black Americans, Whites disproportionately sought information that supported claims of a justified shooting and regarded that information as important. Conversely, Blacks preferred information that implied that the officer had behaved inappropriately. These differences stemmed from two distinct processes: a form of identity-based motivated reasoning and belief updating based on racially distinct priors. Differences in summary judgments were larger either when individuals identified strongly with their racial group or when expectations about the typical behaviors of Blacks and the police diverged. The findings elucidate processes whereby individuals in different social groups come to accept differing narratives about contentious events.

Hakeem J. Jefferson (corresponding author; [email protected]) and Fabian G. Neuner ([email protected]) are PhD candidates in the department of political science at the University of Michigan. Josh Pasek is an assistant professor of communication studies at the University of Michigan and a faculty associate in the Center for Political Studies at the Institute for Social Research ([email protected]).!

! 2

Recent incidents involving Black Americans’ interactions with law enforcement officials

have garnered international attention. When Florida neighborhood watch coordinator George

Zimmerman was acquitted after killing Trayvon Martin, 49% of White Americans reported being

satisfied with the verdict compared to only 5% of Blacks (Dimock and Doherty 2013). When

Officer Darren Wilson fatally shot Michael Brown in Ferguson, Missouri, 64% of Whites agreed

with the grand jury’s decision not to press charges, whereas 80% of Blacks disagreed (Doherty,

Motel, and Weisel 2014). And when New York City police officers were not charged following

the chokehold death of Eric Garner, 70% of Blacks reported that recent decisions decreased their

confidence in the legal system compared to only 35% of Whites (Marist Poll 2014). These racial

divides in opinion, as we demonstrate in the current project, are not limited to summary or value

judgments, but manifest in response to a question central to the cause of justice: What, given the

evidence, do individuals believe transpired in the interaction between an agent of the state and

the citizen that agent is sworn to protect and serve?

Evidence that Americans diverge in their perceptions of these events calls out for an

explanation. How could Blacks and Whites, simply because of their racial categorization, come

to widely differing conclusions on cases that occupy such a central place in our political

conversation? Two overarching sets of explanations present themselves. The first centers on the

possibility that individuals receive qualitatively different bodies of information as a function of

their race. The other is that, despite similar information environments, Blacks and Whites

process the information they receive differently. And, to the extent that Blacks and Whites are

processing the same information differently, we will try to understand why.1

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1 In this paper, we use “biased processing” to suggest that individuals who occupy different social categories will come to diverging views following exposure to the same set of information. Importantly, in the current case – as, we believe, in much of life – it is difficult to

! 3

To examine these questions, we control the information environment surrounding a

fictitious officer-involved shooting and examine the role race plays in producing diverging

beliefs and opinions among Black and White respondents. If divides emerge in this controlled

setting, we can reject the explanation that racial differences are mere products of exposure to

distinct information environments. Instead, we contend that two distinct processes explain the

patterns we observe. First, building on the theory of partisan motivated reasoning (Kunda 1990;

Lodge and Taber 2013), we explore whether race-based motivated reasoning accounts for the

racial divide. That is, we test whether Blacks and Whites diverge in their responses to officer-

involved shootings because they interpret event-relevant information in a way that protects or

bolsters the image of their racial group.

Alternatively, we examine the possibility that differences in responses among Whites and

Blacks are less about the maintenance of social identities, but instead reflect average differences

in prior beliefs and expectations regarding police bias and the likely culpability of Black victims

(Peffley and Hurwitz 2010). These priors, themselves informed by race, may guide how

individuals process new information involving law enforcement officers and citizens of color.

This explanation, unlike one situated in a motivated reasoning framework, does not require that

Blacks and Whites are motivated by some identity-protective goal, but simply that they have

sufficiently different bases upon which to evaluate new information. Much of the existing work

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!identify some objective truth about the state of the world. Thus, we cannot conclude that some group of people holds “accurate” perceptions whereas some other group’s perceptions are “inaccurate.” We also want to avoid attributing differences in perceptions to any particular psychological process; indeed, one of the principal purposes of this paper is to distinguish between two processes that could ostensibly result in these sorts of diverging perceptions. Notably, our use of the term “bias” is purely technical and should not be regarded as suggesting that any particular viewpoint or interpretation of events is necessarily invalid (cf. Gerber and Green 2003). Indeed, contradictory perceptions of facts may sometimes be equally legitimate interpretations of ambiguous evidence.

! 4

on biased information processing in political science fails to appreciate or at least fully account

for the differences in these two approaches (see Bullock 2009). One of our chief goals is to

carefully adjudicate between these related, though theoretically distinct, explanations.

To set the stage, we first leverage a study fielded shortly after the highly contentious and

internationally covered officer-involved shooting in Ferguson, Missouri to understand the sorts

of racial differences that emerge. We then discuss how these differences might come about,

focusing on the potential roles of (1) motivated reasoning shaped by racial identification2 and (2)

belief updating from racially-informed priors. In so doing, we pay particular attention to

unpacking the precise mechanisms that could underlie the racial divide. We proceed to describe

the study we used to assess how perceptions differ when individuals are presented with identical

information and run tests that allow us to compare competing explanations for how racial

differences manifest in this context. Besides demonstrating clear racial divides in response to a

novel event, our findings shed light on the mechanisms underlying these divides. We conclude

by discussing the implications of this work.

A Motivating Case

In August of 2014, Darren Wilson, a White police officer with the Ferguson, Missouri

Police Department, fatally shot Michael Brown, an 18-year-old Black man. Just as they had

following the 1991 beating of Rodney King by officers with the Los Angeles Police Department

and the 1995 acquittal of O.J. Simpson on charges that he was responsible for the death of his

wife and her friend (Newman et al. 1997), Black and White Americans diverged in their

impressions of what happened and whether justice was served. In a survey conducted by the Pew

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!2 We use “racial identification” to refer to the strength with which one associates with a given racial category. When “racial categorization” is used, we simply mean that an individual belongs to a given racial group.

! 5

Research Center shortly after Michael Brown’s death, 80% of Black Americans asserted that the

event “raise[d] important issues about race,” compared to only 37% of Whites (Drake 2014). In

contrast, the plurality of White Americans (47%) stated that “race [was] getting more attention

than it deserve[d].” Consistent with these perceptual differences, Whites and Blacks also differed

in their confidence in the investigations of Brown’s death: only 18% of Blacks interviewed were

confident in the investigations compared to 52% of Whites.

Table 1 - Perceptions of Ferguson by Racial Categorization

Whites Blacks Difference in Percentage Points

Wilson Should Be Charged (Probably or Definitely)

42.2% 90.9% -48.7 ***

Brown Attacked Wilson (Very or Extremely Likely)

37.3% 11.5% 25.8 ***

Brown Had Weapon (Probably or Definitely) 22.8% 4.4% 18.5 ***

Role of Race In Shooting (Large or Enormous)

35.9% 73.6% -37.7 ***

Average Absolute Difference 32.7 N 2,939-2,946 250-252

Note: Numbers represent the proportion of respondents selecting the top two response categories for each question (listed on left). Variation in Ns is due to nonresponse. *** p<.001 differences, two-tailed t-test.

In original data we collected three weeks after Brown’s death, similar patterns emerged –

Blacks and Whites diverged in both the summary judgments they rendered as well as in beliefs

about what transpired between Brown and Wilson.3 When asked whether Brown had a weapon,

how likely it was that Brown attacked Wilson, how much of a role race played in the shooting,

and whether Wilson should be charged, Blacks and Whites answered differently. When these

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!3 These data were collected between August 29th and September 8th, 2014 in the first wave of a panel study. 3,729 respondents were recruited by Qualtrics from the ClearVoice Surveys panel to complete an online survey. Full details about the sample and question wording can be found in Online Appendix A. For reference, Michael Brown was shot on August 9th of that year.

! 6

questions were dichotomously coded, Black and White respondents diverged by an average of

32.7 percentage points (See Table 1). Moreover, Black and White respondents who reported

having heard the most about Ferguson diverged most in their opinions about whether Brown

attacked Wilson and whether Wilson should be charged. Figure 1 illustrates this divide by

estimating a predicted value for a Black or White respondent depending on how much they

reported hearing about the incident.4

Figure 1

That individuals differed in their reading of Ferguson is not particularly surprising nor

even normatively problematic; we expect differences in opinion among citizens in a diverse

society. Similarly, it is hardly novel to point out that respondents in different racial groups report

different beliefs and attitudes. Previous research has uncovered gaps between Whites and Blacks

in opinions about the fairness of the criminal justice system (Hurwitz and Peffley 2005),

perceptions of biased policing practices (Weitzer and Tuch 2005), support for policies perceived

to benefit African Americans (Hutchings 2009; Kinder and Winter 2001), support for the death

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!4 We also replicated this finding in the first wave of our 2016 study. Blacks and Whites who reported having heard the most about Ferguson continued to have the most discrepant views regarding whether Brown had a weapon, whether Brown committed a crime before his encounter with Wilson, and whether Wilson sent racist emails in the weeks prior to the event. The figures present raw data without controls, though the results are robust to controlling for demographic variables and partisanship. See Online Appendix C.

a) Wilson Should Be ChargedBy Amount Heard About Ferguson and Race

Amount Heard About Ferguson

Shou

ld W

ilson

Be

Cha

rged

(Mea

n af

ter R

ecod

ing

from

0−1

)

Nothing A little Moderate Amt. A lot Great Deal

Def

. Not

Prob

. Not

Prob

ably

Def

inite

ly

WhiteBlackWhiteBlack

b) Brown Attacked WilsonBy Amount Heard About Ferguson and Race

Amount Heard About Ferguson

How

Like

ly B

row

n At

tack

ed W

ilson

(Mea

n af

ter R

ecod

ing

from

0−1

)

Nothing A little Moderate Amt. A lot Great Deal

Not

at a

llA

little

Som

ewha

tVe

ryEx

trem

ely


! 7

penalty (Bobo and Johnson 2004), beliefs about the disproportionate imprisonment of African

American men (Unnever 2008), reactions to racial profiling in airports (Sirin, Villalobos, and

Valentino 2016) and even reactions to natural disasters that have a disparate racial impact

(Huddy and Feldman 2006), among other issues. What makes this scenario important and

distinct, however, is that the differences between Blacks and Whites appeared organically, in real

time, in information environments that should have been relatively similar.5

Unpacking Race

To be sure, political scientists have long recognized that race constitutes an important

variable guiding political behavior. Relatively few studies, however, have sought to unpack why

this is the case (but see Kinder and Winter 2001). In a recent article, Sen and Wasow (2016)

attribute this to the difficulty of identifying causal connections between race and various socio-

political outcomes. Race, the authors argue, is generally considered immutable from birth; it

therefore cannot be manipulated and researchers cannot properly attribute its effects.

Nonetheless, recognizing the powerful role race appears to play in the American political system,

scholars often include a “race dummy” in their empirical models predicting attitudes and

behavior. In the same vein, scholars, media outlets, and survey firms frequently publish evidence

that demonstrates the persistence of “the color line” in American public opinion (Kinder and

Sanders 1996). Although pervasive across domains of American life, the tendency to simply note

the association between race and a given outcome is particularly problematic in discussions of

Americans’ responses to officer-involved shootings, which often bring to the fore questions of

race and justice.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!5 Of course, the fact that the Ferguson study was conducted shortly on the heels of the incident does not preclude the possibility that individuals found out about what happened through very different sources (Garrett and Stroud 2014; Stroud 2011).

! 8

This robust relationship between race and politically relevant outcomes should not be

taken for granted, but we caution against an approach that fails to take seriously the nuanced and

varied ways that the meaningfulness of race can manifest in any given context. “Dummifying”

race not only leads to problematic interpretations of the meaning of this important social

construct; this approach also fails to distinguish the effects of race as an expressive social

identity (Huddy, Mason, and Aarøe 2015) from the effects of race as a social category that,

irrespective of one’s psychological attachment to the group, affects experiences and worldviews.

Indeed, different racial experiences might be expected from individuals embedded in what

sociologist Eduardo Bonilla-Silva (1997) terms a racialized social system. As we argue in more

detail in the pages that follow, these theoretical explanations are not merely two sides of the

same coin; they call out for different interventions that reflect the qualitatively distinct

mechanisms at play in forming the basis for this persistent racial divide.

Race as a Motivating, Expressive Social Identity

If the effects of race in response to officer-involved shootings emerge because of

individuals’ psychological attachments to racial in-groups, the existing literature on motivated

reasoning offers a useful starting place for understanding why Blacks and Whites diverge both in

their summary judgments and beliefs about the particulars of these events. Kunda (1990) argued

that individuals’ perceptions were influenced by both accuracy and directional goals. Scholars

have since shown that group membership, in the form of partisan affiliation, is a principal source

of directional motivation (Lodge and Taber 2013). That is, when presented with new

information, partisans engage in defensive processing, discounting claims that contradict party

orthodoxy (Lebo and Cassino 2007) by arguing against it (Eagly and Chaiken 1995) and even

accepting inaccurate messages that bolster existing viewpoints (Miller, Saunders, and Farhart

! 9

2015). Work on the cognitive processes underlying motivated reasoning suggests that this occurs

as a subconscious affective response to new information (see Westen et al. 2006). People

evaluate new information in line with prior attitudes and group memberships by spontaneously

generating a positive or negative interpretation of a new claim and processing the claim in line

with that interpretation (Lodge and Taber 2005; 2013; Nyhan and Reifler 2010).

Recent scholarship on motivated reasoning suggests that this operates along partisan lines

more than on positions about particular issues. This fits well with the notion that partisanship

facilitates motivated reasoning in its role as a prominent social identity, rather than as a catchall

for some set of political opinions (Green, Palmquist, and Schickler 2002; Greene 2004; Iyengar,

Sood, and Lelkes 2012). Because individuals derive meaning from membership in salient social

groups (Tajfel 1974), group memberships play an important role in shaping how individuals

perceive and engage with the world (D. Abrams and Hogg 1990). This implies, as Lodge and

Taber (2013) found, that the strength of one’s identity and the salience of that identity in a given

context should be key determinants of the extent to which individuals engage in motivated

processes.

Despite a focus on partisan identity for the vast majority of the literature, similar

differences in processing should occur for all sorts of social identities (Kahan 2010a). Although

in the contemporary political environment more and more issues are divided along party lines

(Fiorina and Abrams 2008), there remain many issues for which other identities are more salient

(see e.g., DiMaggio, Evans, and Bryson 1996; Hutchings and Valentino 2004). We argue that

officer-involved shootings constitute issues that divide chiefly not along partisan, but racial lines.

And though the racial divide in America is, as Kinder and Sanders argue, “a divide without peer”

! 10

(1996, 27) , studies of motivated reasoning on the basis of social identities other than

partisanship have been relatively rare (Kahan 2010b; Kunda and Sinclair 1999).

In the context of officer-involved shootings, race, like partisanship, might well act as a

social identity that motivates individuals’ processing of event-relevant information. Such racially

motivated reasoning might be particularly prevalent when situational ambiguity allows

individuals to privilege directional goals (i.e. identity bolstering) over accuracy-oriented ones (cf.

Hodson, Dovidio, and Gaertner 2002; Lawrence 2000). That is, insofar as there are competing

claims about the details of an officer-involved shooting that can be viewed as supporting the

White officer or the Black victim, individuals may discount information that threatens to

denigrate their social group while advantaging information that protects the group’s image.

Likewise, a theory of racially motivated reasoning would expect that, when given the

opportunity to expose themselves to new information, Whites and Blacks will choose

information consistent with the position of their in-group member (cf. Hart et al. 2009; Stroud

2011). Generally, Whites should be expected to select information that supports the officer’s

claims that the shooting was justified, with Blacks choosing, on average, to seek information that

undermines these claims. Importantly, if racially motivated reasoning is responsible for the

divides observed, the tendency to engage in these defensive processes should be more

pronounced when racial identity is made more salient and should emerge more for highly

identified Blacks and Whites than for their less strongly identified counterparts (cf. Leeper and

Slothuus 2014; Lodge and Taber 2013).

An Alternative Explanation: Racially Distinct Priors

Though compelling, this theory of racially motivated reasoning must contend with an

equally plausible alternative explanation that can lead to similar, if not identical divides in Black

! 11

and White responses to officer-involved shootings. As others have argued, much of what we

think of as biased processing based on ego-protective considerations may instead result from

differences in information updating (Bullock 2009; Gerber and Green 1998). In work focused on

partisan responses to the Lewinsky scandal, for example, Fischle (2000) acknowledges that

“individuals may well arrive at ‘congenial’ conclusions not because they were motivated to do

so, and selectively processed the evidence accordingly, but simply because those judgments are

plausible given the individual’s prior beliefs and expectations” (141). Like Fischle, we concede

that it is challenging to distinguish the effects of identity-based motivated processes from ones

that derive from individuals’ different prior beliefs and expectations (Peffley and Hurwitz 2010).

Race, arguably more than any other social category, structures and stratifies American

life. Whites and Blacks not only frequently live in different communities and occupy different

social networks (Massey and Denton 1993); they have experiences, both personal and vicarious,

that shape their worldviews and inform their interactions in the social world (Peffley and

Hurwitz 2010). This reality is historically situated and reflects the ways in which race continues

to influence the lived experiences and life outcomes of this country’s citizens. With respect to the

criminal justice system, long-standing divides shape the experiences of Black and White

Americans. Blacks receive harsher sentences for committing the same crimes as Whites

(Steffensmeier, Ulmer, and Kramer 1998), are stopped more frequently by police (Epp,

Maynard-Moody, and Haider-Markel 2014), and are more likely to report that their interactions

with the law enforcement community are negative (Browning et al. 1994; Theobald and Haider-

Markel 2009). Compounding these factors, the most prominent cases of police brutality often

involve Black victims (Lawrence 2000), suggesting a pattern of mistreatment toward Black

Americans not experienced by Whites. Given these differences, Blacks and Whites hold different

! 12

attitudes toward the criminal justice system, with Blacks much less trustful of law enforcement

than Whites (Nunnally 2012). In this manner, racialized social experiences can result in different

perceptions of police bias and, by extension, the likelihood that a law enforcement officer would

shoot a Black victim without justification.

Just as citizens may base their perceptions of an incident on what they expect from the

police, they may also condition their responses on expectations of members of the purportedly

victimized group. Though it may seem strange to treat them as such, racial attitudes are

inherently priors that guide the processing of information. That is, people make inferences about

missing information based in part on expectations derived from stereotypes about a target’s

social group (Hamilton, Sherman, and Ruvolo 1990). When evaluating an officer-involved

shooting, prejudiced individuals may tend to identify fault in the victim’s behavior, thereby

justifying the use of force. Hence, to the extent that Blacks and Whites differ both in their beliefs

about the prevalence of police bias and in beliefs and expectations about the likely culpability of

the victims, they should also differ in the judgments rendered in response to these events. The

question here is whether and to what extent differences in these expectations – rather than more

defensive processes associated with motivated reasoning – account for the racial divide (cf.

Peffley and Hurwitz 2010).

As we stress in the discussion section, distinguishing between these two theoretical

explanations is no mere act of academic pedantry; the distinction, we believe, matters. But given

the nature of the current project and its potential to inform a larger conversation about officer-

involved shootings in the US (as well as state-citizen conflict elsewhere), we want to note from

the outset that, insofar as individuals are processing the same novel information in ways that lead

to racially diverging judgments, either explanation constitutes a social problem that demands

! 13

resolution.

The Current Study

Beyond their contemporary relevance, officer-involved shootings provide a

straightforward case for examining how race shapes information processing. To understand

racial gaps observed in response to such events, we focus primarily on individuals’ treatment of

event-relevant information regarding a novel incident. That is, when presented with information

about an officer-involved shooting for which the facts of the case remain somewhat ambiguous

(as they often are, see Lawrence 2000), how do individuals arrive at their beliefs about whether

the officer’s actions were justified? And, when Blacks and Whites disagree on this question,

what accounts for their disagreement?

In line with our observations following the incident in Ferguson, we expect that beliefs

held by Black and White Americans will differ in response to the scenario we present. On

average, Black respondents are expected to view the officer’s actions as less appropriate than

their White counterparts and, consistent with this view, should support punishing the officer.

Beyond these divides in summary judgments, we expect similar divergence in response to

questions about the facts of the case. Compared to Black respondents, Whites should be more

likely to believe that the victim was armed and to assert that he attacked the officer yet less

willing to adopt the belief that race played a significant role in the shooting.6

H1: Blacks and Whites will differ in both (a) summary judgments about the shooting and

(b) beliefs about what happened during the event.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!6 Admittedly, these outcomes represent at least two classes of evaluations, namely responses that can be deemed subjective judgments as well as responses that speak to more objective assessments. We contend that both are interesting in their own right but acknowledge that divergence in response to questions focused more on factual elements of the case provide more compelling evidence for a motivated reasoning framework, which we discuss in subsequent hypotheses.

! 14

In line with notions that race alters how individuals process this information, we expect to

observe differences in which pieces of information respondents view as most important and

accurate as a function of their race. In general, we hypothesize that:

H2: Whites will, on average, privilege information that defends the officer’s actions,

whereas Blacks will privilege information that appears to support claims of a wrongful

death.

Collectively, evidence for these hypotheses would provide a strong basis for concluding

that individuals are engaging in biased processing. Notably, however, these hypotheses are

agnostic to the mechanism yielding attitude divergence. As we argued previously, differences

between Blacks and Whites – both in summary judgments and assessments of event-related facts

– could emerge as a function of either identity-based motivated reasoning or the different

expectations members of these groups have about interactions between citizens and law

enforcement. To adjudicate between explanations, we propose two sets of hypotheses derived

respectively from an identity-based motivated reasoning process and a process guided principally

by prior expectations and beliefs.

Motivated reasoning. If differences in perceptions of the novel scenario stem from

racially motivated reasoning, the extent to which this process operates should depend on the

salience of racial identity for those evaluating the incident. Racial identity salience could emerge

either as a product of some external cue or as a function of individual-level differences in the

strength of racial identification. This latter expectation builds on prior evidence that motivated

reasoning is moderated by attitude strength (Leeper and Slothuus 2014; Lodge and Taber 2013).

We thus expect:

H3: When racial identity is primed, summary judgments will differ between Black and

! 15

White respondents to a greater degree than when racial identity is not primed.

H4: Racial differences in perceptions will be similarly moderated by measures of racial

group identification including (a) identity strength, (b) group closeness, and (c) linked

fate.7

Moreover, a motivated reasoning explanation further posits that individuals will seek out

information selectively in ways that reinforce their existing views (cf. Garrett and Stroud 2014;

Iyengar and Hahn 2009).

H5: Given the opportunity to seek out further information about the event, Blacks and

Whites will seek different, identity-consonant information.

Finally, the motivated reasoning framework suggests that additional information should not

influence summary judgments once an initial judgment has been rendered. Therefore, evidence

that individuals consider and incorporate new information after rendering an initial judgment

would provide evidence contrary to a purely motivational account.

H6: Respondents’ reactions to witness statements will not be related to final summary

judgments after accounting for initial summary judgments.

Racial priors. If differences in summary judgments result from differing expectations

and experiences across groups, the role of race should diminish once these uncommon priors are

accounted for. We focus on perceptions of police bias and perceptions that Black Americans

violate social norms as scaffolds for interpreting new information about encounters between

members of these groups. This can also be understood in a mediation context, whereby the effect

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!7 Independently, these three measures of racial identification derive from different theoretical and conceptual bases. Here, however, we treat them each as a proxy of how attached an individual is to her racial group. For a more complete discussion of identity and its role in shaping individuals’ responses in political contexts, we recommend Huddy (2001; 2003).

! 16

of race operates through differences in these underlying beliefs. Thus, the influence of racial

categorization (i.e. the “race dummy”) on summary judgments as well as reactions to

information would be expected to diminish when we account for prior beliefs and expectations.

H7: Respondents’ (a) perceptions of police bias and (b) racial attitudes should mediate

the relationship between racial identification and summary judgments as well as

evidence weights.

Methods

Data

To test these hypotheses, we fielded a two-wave study using Qualtrics Panels. Qualtrics

recruited respondents to complete an online survey using targeted emails sent to members of the

ROI Rocket online panel (see Online Appendix D). In the first wave, fielded between June 15

and June 20, 2016, 1,430 U.S. respondents completed a survey that queried them about their

demographics, perceptions of the criminal justice system, experiences with racial bias, racial

attitudes, and recall about the incident in Ferguson in 2014 (see Table 2). Quotas were used to

gather a sample that would be approximately 50% Black and 50% White at the end of wave 2.

Respondents to the first wave were re-contacted one week later to take part in a second study,

which ran from June 27 to August 5, 2016. The vast majority of the participants responded

within one week, though the survey was kept open with the hopes of re-contacting 1,000

respondents. In total, 895 respondents completed the second wave of the study, for a re-contact

rate of 62.6%.

For the current study, we only report data from respondents who completed the second

wave before July 5, 2016, when Alton Sterling was shot by a police officer in Baton Rouge,

Louisiana. This incident – along with subsequent widely publicized shootings in St. Paul,

! 17

Question Description of measures used in study Administration Notes / RandomizationWave 1 (N=1430) June 15-20 2016Demographics sex; race; age; educationIdeology/Partisanship party identificationFairness of Police and Criminal Justice System criminal justice system fair; police treatment of Blacks vs Whites (5)

Questions About Ferguson -[Note: these were a replication of the results from the 2014 Ferguson study]

Racial ResentmentOthers overcame prejudice without special favors; generations of slavery created current conditions; Blacks gotten less than deserve; Blacks not trying hard enough

Wave 2 (N=895; 726 before cutoff) June 27-August 5 2016 (July 5 cutoff used)Introduction to ScenarioNon-Racial Demographics -Racial Identification Measures importance of race; linked fate; group closeness Groups A and B OnlyExperiences with Police - Group B OnlyPolice Chief Statement Pro-Officer - Overview of Events

Mr. Davis Statement Pro-Taylor - Was in a heated argument with Mr. Taylor when events transpiredOfficer Silver Statement Pro-Officer - Statement from the officer involved in shootingPartway Summary Judgments Officer's actions appropriate; should officer be charged Randomly assigned to 1/2 of respondents

Mrs. Walker Statement Pro-Officer - Woman who called police to report argument, did not see shootingMrs. Thomas Statement Pro-Taylor - Character witness for Mr. Taylor, did not see shootingMrs. Williams Statement Pro-Taylor - Was walking dog and saw events transpireMr. Anthony Statement Pro-Officer - Was driving home and stopped to watch events transpire

Selective Exposure TaskIdentify which additional statements respondent would like to read from one-sentence excerpts

1/2 of respondents see all 8 statements, 1/2 see a randomly assigned set of 4

Summary Judgments Officer's actions appropriate; should officer be chargedSummary Beliefs Taylor attacked officer; officer had weapon; race role in shootingRacial Identification Measures importance of race; linked fate; group closeness Group C OnlyExperiences with Police - Groups A and C Only

Notes - Scenario statements are bolded; each was followed with an open-ended response and three questions about how the statement should be weighted by a jury, how accurate the statement was, and how biased/objective the statement was. Only measures used in the current study are described. Descriptions with "-" are not evalutaed in the current study. See Online Appendix F for question wording, response options, and coding of all measures.

Table 2 - Overview of Relevant Measures from 2016 Study

! 18

Minnesota and Dallas, Texas – had the potential to alter attitudes on issues measured in the

study.8 Among the 726 respondents who completed both the first and second waves of the study

before July 5, 370 identified as Black and 356 identified as White.9 Outside of the goal of

ensuring racial balance, no quotas were used for the current study.

Procedure

In the second wave of the study, which was designed to appear unrelated to the first

wave, respondents were told that they were being contacted “on behalf of a local municipality

where a police officer was recently involved in a controversial incident.” They were asked to

“put [themselves] in the role of a local citizen who may be selected to serve on a grand jury

deciding whether the officer should be indicted for acting inappropriately or whether he acted in

good faith.” Unbeknownst to respondents, this was a fictitious scenario that existed only in the

context of this study. To aid in deception, respondents were also told that we had anonymized

details of the event, including the name of the city where the event had taken place, and provided

pseudonyms for relevant actors and all witnesses (see Online Appendix E for the full

language).10

In a recent article focused on partisan motivated reasoning, Leeper and Slothuus argue

that “experiments where motivations are primed are the best—and perhaps only—way to clearly

distinguish the effects and mechanisms of motivated reasoning” (2014, 149). Therefore, before

reading the scenario, respondents were assigned randomly to one of three conditions aimed at

priming racial self-identification (to assess H4; cf. Steele and Aronson 1995). Respondents in

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!8 Including data collected beyond this period yields slightly stronger results, though we decided a priori to exclude these data. See Online Appendix H for results. 9 Individuals who identified as both Black and White (N=13) were excluded from all analyses. 10 At the conclusion of the study all respondents were fully debriefed and informed about the fictitious nature of the case.

! 19

group A were asked to identify their racial identities and to assert how close they felt to members

of their racial group at the start of the survey to prime their racial identities (Identity Prime; see

Table 2). Respondents in group B were asked these questions alongside an additional battery of

questions that gauged whether they had had various encounters with the criminal justice system

(Identity and Experience Prime). The remaining respondents proceeded directly to the scenario

and answered all identity-related questions at the end of the study (Control; group C).

Scenario. The controlled information environment began when respondents were

presented with seven statements related to the incident.11 To mirror the kind of information

typically available to the public when these events occur, participants received statements that

either supported or called into question the actions of Officer Silver, who was responsible for

shooting Mr. Taylor in our scenario. Collectively, the statements also generate a sense of

ambiguity that so often characterizes these events (see Lawrence 2000). All respondents

encountered the same series of seven statements about the hypothetical event, and these

statements were always presented in the same order (bolded items in Table 2). The first

statement, from the city’s police chief, provided an overview and introduction to the events. In

that statement, respondents were told that Mr. Taylor was an “African-American male in his

mid-twenties.” The race of the police officer and other witnesses was never stated.

Across the statements, respondents learned that a neighbor, Mrs. Walker, called police

when two men, Mr. Taylor and Mr. Davis, were arguing on the sidewalk. As a patrol car

responded to the scene, Mr. Taylor fled and was pursued by Officer Silver. After a short

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!11 Although it was important that our scenario was plausible, creating a fictitious event allowed us to constrain the information environment from which respondents could gather information. Differences in perceptions and attitudes about the case thus cannot be attributed to differences in media exposure or exposure to different characterizations of the event. Instead, they must be attributed to some element of information processing.

! 20

distance, Mr. Taylor turned around and was shot by the officer. In addition to Mr. Davis and

Officer Silver, respondents read statements from five additional individuals, two of whom also

witnessed the events: Mrs. Williams, who was walking her dog, and Mr. Anthony, who had been

driving down the street. These eye-witnesses disagreed about the circumstances under which Mr.

Taylor turned toward the officer. The other three statements came from the chief of police, Mrs.

Walker, and Mrs. Thomas – a character witness for Mr. Taylor (see Online Appendix E for the

full text of all statements).

After the first three statements, half the respondents were asked to provide an initial

judgment on two summary judgement variables described below (This manipulation is used to

assess H7).12 Respondents then read each of the remaining four statements. Collectively, the

seven statements provided evidence that could be used to support the notion that the officer’s

actions were appropriate or that the actions were inappropriate.

Upon reading all seven statements, respondents were presented with one-sentence

excerpts from either four or eight additional statements (Selective Exposure Task in Table 2).13

They were then asked which additional statements they would be interested in reading, though

the full statements were never provided. Finally, respondents were asked about their overall

perceptions of the case.

Key Measures

Statement evaluations. Respondents were asked a series of questions about each

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!12 Only half of respondents were given this manipulation because we wanted to test whether that the act of rendering an initial judgment altered the later judgments that respondents conferred. We did not find significant differences in later judgments based on respondent condition. 13 Respondents were randomly asked to evaluate either eight statements or only four statements (randomly sampled from the eight) for a separate analysis on whether providing additional information might alter perceptions. Limiting the results of the current study to the subset of respondents who were asked all eight statements did not alter any conclusions.

! 21

statement. First, they provided an open-ended reaction.14 Second, they were asked how much

weight they think jurors should place on the statement, using a 5-point scale ranging from “None

at all” (coded 0) to “A great deal” (coded 1). Respondents were then asked to assess both how

accurate and how objective or biased they thought the statement was. Full question wordings,

response options, and coding for all measures is provided in Online Appendix F.

Summary judgments. For half of respondents after the third statement and for all

respondents following the reading of the seven statements and the selective exposure task,

respondents answered a series of questions evaluating the officer’s actions. First, they were

asked, “Given what you have read, how appropriate do you think Officer Silver’s actions were?”

Response options ranged from “Not at all appropriate” (0) to “Completely appropriate” (1) on a

5-point scale. Next, respondents were asked, “Given what you have read, do you think that

Officer Silver should be charged with a crime?” with response options ranging from “He

definitely should NOT be charged” (0) to “He definitely should be charged” (1) on a 4-point

scale.

Summary beliefs. Beyond these questions focused on summary judgments, respondents

were asked questions related to the facts of the case. Here, respondents were first asked, “Given

what you have read, how likely do you think it is that Mr. Taylor attacked Officer Silver?” with

response options ranging from “Not at all likely” (0) to “Extremely likely” (1) on a 5-point scale.

They were then asked, “Do you happen to recall whether Mr. Taylor had a weapon?” Response

options ranged from “He definitely did NOT have a weapon” (0) to “He definitely did have a

weapon” (1) on a 4-point scale. Lastly, respondents were asked, “How much of a role do you

think race played in the shooting?” Here, responses ranged from “No role at all” (0) to “An

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!14 This paper relies on responses to closed-ended questions only.

! 22

enormous role” (1) on a 5-point scale.

Racial identification measures. In addition to these case-related measures, we rely on

several other measures to examine the processes underlying the racial divide. To assess whether

the divide is a function of identity-based motivated reasoning, we generated three measures that

tap racial identification: group importance, group closeness, and linked fate. These measures

were coded to range from 0, for the least racially identified, to 1 for the most strongly identified.

To ascertain whether the divide emerges as a consequence of different prior beliefs and

expectations, we rely on a battery of five questions meant to capture beliefs about police bias15 as

well as the standard 4-item measure of racial resentment (Kinder and Sanders 1996; Kinder and

Sears 1981).

Results

Racially Biased Processing

Our first set of hypotheses posited the existence of a racial divide in response to our

fictitious scenario. Evidence of such a divide would suggest that Black and White individuals

were indeed processing information differently, not simply encountering different information.

To assess this, we compared Black and White respondents in the answers they gave to the

questions about summary judgments and beliefs (H1) as well as in the importance and accuracy

they accorded to the various scenario statements (H2). We also we explored whether their ratings

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!15 The questions that made up this index were as follows: 1) In general, do the police treat Whites better than Blacks, treat Blacks better than Whites, or treat them both the same? 2) Do the police stop Whites more than Blacks, Blacks more than Whites, or do they stop them both equally? 3) How often do you think the police use more force than is necessary under the circumstances when dealing with White people? 4) How often do you think the police use more force than is necessary under the circumstances when dealing with Black people? 5) How common do you think racial or ethnic prejudice is among police officers? The resulting index variable is rescaled 0-1, such that higher values indicate a greater sense that the criminal justice system is biased.

! 23

of statements were correlated with their summary judgments and evaluations of the event.

Table 3 - Perceptions of Novel Incident by Racial Self-Identification Whites Blacks Difference Officer's Actions Appropriate .52 .28 .25 *** Officer Should Be Charged .35 .67 -.32 *** Taylor Attacked Silver .31 .15 .16 *** Taylor Had Weapon .36 .22 .14 *** Role of Race in Shooting .32 .59 -.27 ***

Average Absolute Difference .23 N 355-356 368-370 Note: Numbers represent mean values for respondents on each outcome. Variation in Ns is due to nonresponse. All variables were coded to range from 0 to 1. *** p<.001 differences, two-tailed t-test.

Blacks and Whites, who encountered identical information in our scenario, diverged

significantly in their summary judgments about whether the officer’s actions were appropriate

and whether he should face criminal charges; they also differed in perceptions of what occurred

during the encounter between the Officer Silver and Mr. Taylor (H1; Table 3). Whites were more

likely than Blacks to believe that Officer Silver’s actions were appropriate (difference=.25,

t=11.0, p<.001), that Mr. Taylor attacked Officer Silver (difference=.16, t=7.8, p<.001) and that

Mr. Taylor had been armed during the encounter (difference=.14, t=7.6, p<.001). In contrast,

Blacks were more likely than Whites to think that Officer Silver should be charged (difference=-

.32, t=-13.9, p<.001) and that race played a role in the encounter (difference=-.27, t=-11.4,

p<.001). The average measure differed between Blacks and Whites by .23 on a 0-1 scale. These

results mirror the discrepancies observed shortly after the events in Ferguson. Differences

between Blacks and Whites thus appear to be largely attributable to differential processing of the

same information.

! 24

Table 4 - Statement Ratings by Racial Categorization Statement Weight Statement Accuracy Statement Bias Whites Blacks Difference Whites Blacks Difference Whites Blacks Difference Police Chief (Pro-Officer) .62 .54 .08 *** .64 .46 .18 *** .37 .58 -.21 *** Mr. Davis (Pro-Taylor) .51 .66 -.15 *** .44 .59 -.15 *** .51 .43 .08 *** Officer Silver (Pro-Officer) .61 .51 .10 *** .59 .40 .20 *** .43 .62 -.19 *** Mrs. Walker (Pro-Officer) .41 .46 -.05 + .56 .54 .02 .41 .47 -.06 * Mrs. Thomas (Pro-Taylor) .36 .55 -.19 *** .45 .59 -.14 *** .61 .50 .11 *** Mrs. Williams (Pro-Taylor) .62 .76 -.15 *** .58 .73 -.15 *** .41 .33 .07 ** Mr. Anthony (Pro-Officer) .52 .43 .09 *** .51 .39 .12 *** .37 .53 -.16 ***

Average Absolute Difference .12 .14 .13 N 356 368-370 355-356 366-370 356 368-370 Note: Numbers represent mean values for respondents on each outcome. Variation in Ns is due to nonresponse. Statements are listed in order of presentation. All variables were coded to range from 0 to 1. + p <.10 | * p<.05 | ** p<.01 | *** p<.001 differences, two-tailed t-test.

! 25

White and Black Americans also diverged in the value that they accorded to the various

statements (H2; Table 4). Across all measures, Whites placed more value on information

suggesting the officer’s actions were appropriate than did Blacks. In contrast, Blacks put more

weight on statements critical of the officer and reported that these measures were less accurate

and more biased. For example, when Black and White respondents read a statement from the

victim’s neighbor (Mrs. Thomas) suggesting he “was a good boy” who would “never hurt

anybody,” Blacks accorded the statement the same amount of weight they gave the police chief’s

statement. Whites, however, thought it was the least relevant among the statements provided.

The only statement that did not demonstrate this pattern was from Mrs. Walker, who called

police to the scene of the argument, but did not witness subsequent events (Table 4, row 4). With

Mrs. Walker’s statement again serving as the sole exception, the weight respondents associated

with the various statements predicted their summary judgments in the expected direction (see

Online Appendix G).

Racially Motivated Reasoning

A second set of hypotheses examined whether racial differences were the product of

identity-motivated processes. To test this, we first compared respondents in experimental groups

A, B, and C (see Table 2), to determine whether priming racial identity and experiences with

police would exacerbate the perceptual divide (H3). We also examined whether self-reports of

identity strength, group closeness, and linked fate served a similar moderating role (H4). Because

motivated reasoning is expected to result in selective exposure, we next assessed whether Blacks

and Whites privileged different pieces of information in the selective exposure task (H5).

Finally, in line with the notion that motivated reasoning is goal-oriented, we tested whether

individuals ceased to incorporate new information once they had heard enough to render a

! 26

summary judgment. To do this, we limited our sample to the randomly selected set of individuals

who rendered summary judgments twice, once after the first three statements and once at the end

of the study. Here, we examined whether the weights associated with statements 4 through 7

accounted for changes in the summary judgments individuals made between these two time

points. Evidence that these statements were uniquely related to final judgments even controlling

for initial judgments would suggest that individuals were continuing to rely on new information

and thus were not acting as pure motivated reasoners (H6).

To the extent that respondents were behaving with the goal of bolstering their identities,

we should find that priming racial identity will exacerbate racial divides (H3). Column 1 of

Figure 2 shows the effect of our primes on all outcomes. Contrary to expectations, answering

additional questions about race did not lead respondents to answer summary judgement or belief

questions differently.

If racially biased processing results from motivated reasoning, more strongly identified

individuals should diverge the most in their perceptions of the incident (H4). Here we found

mixed results. Compared to respondents who reported that their racial identities were not very

important, Blacks and Whites who reported that their racial identities were important diverged

more in their summary judgments (Figure 2, column 2, rows 1-2), but not in their perceptions of

what happened in the incident (rows 3-5). Similar patterns emerged when racial identification

was measured as perceived closeness to one’s racial group (column 3) or as linked fate with

other group members (column 4). Collectively, these results imply that more highly identified

individuals diverged most on subjective measures, but not more factual ones.

Work on motivated reasoning and selective exposure suggests that motivated individuals

will seek additional information that reinforces their prior beliefs when given the opportunity to!

! 27

!Figure 2 – Summary Judgments and Beliefs by Condition and Racial Identification

a) Priming ConditionF=.27(2,701)

Racial Priming Condition

Control Identity ID + Exp

Not a

t all

Littl

eSo

mew

hat

Very

Com

plet

ely


Offi

cer's

Act

ions

App

ropr

iate

b) Identity StrengthF=4.92(1,698)*

Identity Importance

Not at all A little Moderately Very Extremely

Not a

t all

Littl

eSo

mew

hat

Very

Com

plet

ely


c) Group ClosenessF=7.42(1,703)**

Closeness to Racial Group

Not at all A little Somewhat Very Extremely

Not a

t all

Littl

eSo

mew

hat

Very

Com

plet

ely


d) Linked FateF=.91(1,700)

Sense of Linked Fate

None Not much Some A lot

Not a

t all

Littl

eSo

mew

hat

Very

Com

plet

ely


e) Priming ConditionF=.75(2,699)


Control Identity ID + ExpDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


Offi

cer S

houl

d Be

Cha

rged

f) Identity StrengthF=4.52(1,696)*

Identity Importance

Not at all A little Moderately Very ExtremelyDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


g) Group ClosenessF=9.06(1,701)**


Not at all A little Somewhat Very ExtremelyDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


h) Linked FateF=10.92(1,698)***


None Not much Some A lotDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


i) Priming ConditionF=2.05(2,699)



Not a

t all

Littl

eSo

mew

hat

Very

Extre

mel

y


Taylo

r Atta

cked

Silv

er

j) Identity StrengthF=.64(1,696)

Identity Importance


Not a

t all

Littl

eSo

mew

hat

Very

Extre

mel

y


k) Group ClosenessF=.79(1,701)



Not a

t all

Littl

eSo

mew

hat

Very

Extre

mel

y


l) Linked FateF=.24(1,698)



Not a

t all

Littl

eSo

mew

hat

Very

Extre

mel

y


m) Priming ConditionF=.23(2,698)


Control Identity ID + ExpDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


Taylo

r Had

Wea

pon

n) Identity StrengthF=1.18(1,695)

Identity Importance

Not at all A little Moderately Very ExtremelyDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


o) Group ClosenessF=.39(1,700)


Not at all A little Somewhat Very ExtremelyDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


p) Linked FateF=1.74(1,697)


None Not much Some A lotDefin

itely

not

Prob

ably

not

Prob

ably

Defin

itely


q) Priming ConditionF=.22(2,698)



No a

t all

Littl

eM

oder

ate

Larg

eEn

orm

ous


Role

of R

ace

In S

hoot

ing

r) Identity StrengthF=.85(1,695)

Identity Importance


No a

t all

Littl

eM

oder

ate

Larg

eEn

orm

ous


s) Group ClosenessF=3.51(1,700)+



No a

t all

Littl

eM

oder

ate

Larg

eEn

orm

ous


t) Linked FateF=1.19(1,697)



No a

t all

Littl

eM

oder

ate

Larg

eEn

orm

ous


Perceptions by Race and

! 28

do so (H5). When presented with additional excerpts in our selective exposure task, Black and

White respondents indeed expressed interest in reading different accounts. For four of the eight

excerpts provided, this divide was significant (Table 5). For example, 49 percent of White

respondents wanted to read the following statement: “If he was trying to surrender to the officer,

the way he turned around is not the way to do it” (Excerpt 3) whereas only 33 percent of Blacks

expressed a similar interest. In contrast, given the opportunity to read Excerpt 7, which suggested

that Officer Silver had a history of engaging in racially biased policing, 55 percent of Black

respondents wanted to read the statement, compared to only 42 percent of Whites. These results

provide evidence of selective exposure by race, consistent with the motivated reasoning

framework.

Table 5 - Desire to Read Additional Excerpts by Racial Self-Identification Whites Blacks Difference sig

Excerpt 1 Pro-Officer 0.33 0.30 -.02 (.04) Excerpt 2 Pro-Taylor 0.43 0.49 .06 (.04) Excerpt 3 Pro-Officer 0.49 0.33 -.15 (.04) *** Excerpt 4 Pro-Officer 0.20 0.18 -.02 (.03) Excerpt 5 Pro-Officer 0.37 0.25 -.12 (.04) ** Excerpt 6 Pro-Taylor 0.15 0.25 .10 (.03) ** Excerpt 7 Pro-Taylor 0.42 0.55 .14 (.04) ** Excerpt 8 Pro-Officer 0.41 0.33 -.06 (.04)

N 251-264 263-291 Note: Numbers represent mean values for respondents on each outcome. Difference column is calculated using a logistic regression controlling for whether respondents randomly saw four or eight statements. Variation in Ns is due to nonresponse and due to the fact that half of respondents saw four excerpts and half saw eight. All variables were coded to range from 0 to 1.** p<.01 | *** p<.001 differences, two-tailed.

Finally, a pure motivated reasoning account would suggest that, after rendering an initial

judgment, subsequent information should have little to no additional influence on future

! 29

judgments on the same question. Alternatively, we could imagine that individuals continue to

update their beliefs even after rendering an initial judgment. If this were the case, we would

instead find that the weights respondents accorded to new pieces of information (after they

rendered their first judgment) would explain unique variance in their eventual summary

judgments over and above the explanatory power of their initial judgments.16

When individuals were asked to render two sets of summary judgments – one partway

through the statements and one at the end – information provided after their initial evaluations

continued to influence their assessments (H6, Table 6). In particular, respondents’ ratings of the

statements by the two bystanders – Mrs. Williams and Mr. Anthony – continued to have a strong

and statistically significant influence on eventual judgments about whether the Officer’s actions

were appropriate and whether he should be charged. We see this as some evidence that

respondents were not merely rendering a snap judgment but rather, contrary to H6, subsequent

information still has a significant bearing on their views.

Collectively, the evidence presented here provides mixed support for the claim that

individuals engage in motivated reasoning along racial lines in response to officer-involved

shootings. We found that individuals whose identities were more salient were more likely to

polarize in their subjective assessments, but not on more factual questions. We also found that

individuals tended to seek additional information that might reinforce their predispositions. Yet,

we had no success attempting to experimentally manipulate the salience of racial identities and

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!16 To test this explanation, causal ordering is important. From a motivated reasoning perspective, we would expect individuals to report that summary-judgment-confirming information was more important than disconfirming information. But these reports should be driven by the summary judgments, rather than the credibility of the new information. This means that the new information would not have a unique influence on later summary judgments. Instead, the weights for the information would serve as an expression of the already-rendered judgments. In contrast, evidence that new information influenced summary judgments over and above initial judgments would instead imply that the information was changing people’s views.

! 30

found that individuals continued to respond to new information even after they had rendered a

summary judgment. It therefore seems likely that at least some other mechanism accounts for the

biased perceptions we observed.

Table 6 - Do Respondents Update Beliefs with New Information? Officer's Actions Appropriate Should Be Charged

All Weights

Including Partway

Evaluation

All Weights

Including Partway

Evaluation

Coef. s.e.

Coef. s.e.

Coef. s.e.

Coef. s.e.

Intercept .36 (.05) *** .07 (.04) + .44 (.05) *** .06 (.04) Partway Evaluation

.75 (.04) ***

.80 (.04) ***

Police Chief (Pro-Officer)

.06 (.05)

--

-.09 (.06)

-- Mr. Davis (Pro-Taylor)

-.25 (.06) ***

--

.35 (.07) ***

--

Officer Silver (Pro-Officer)

.43 (.06) ***

--

-.39 (.06) ***

-- Mrs. Walker (Pro-Officer)

.02 (.05)

.03 (.04)

.00 (.05)

-.03 (.04)

Mrs. Thomas (Pro-Taylor)

-.10 (.05) +

-.01 (.04)

.29 (.06) ***

.11 (.04) **

Mrs. Williams (Pro-Taylor)

-.26 (.06) ***

-.09 (.04) *

.21 (.06) ***

.10 (.04) *

Mr. Anthony (Pro-Officer) .27 (.05) *** .17 (.04) *** -.27 (.06) *** -.10 (.04) * N

349

350

349

350

R-squared !! .44 !! .66 !! .47 !! .72 Note: + p <.10 | * p<.05 | ** p<.01 | *** p<.001 differences, two-tailed.

Prior Beliefs and Expectations

To explain the biased updating we observed, we test the possibility that differences in

responses between Blacks and Whites result from a process that is not necessarily identity-

driven, but rather emanates from the fact that Blacks and Whites encounter officer-involved

shootings with differing expectations based on sets of uncommon priors (H7). Two sets of priors

seemed likely to alter the weights that respondents accorded to the various statements as well as

their summary judgments: their expectations from Blacks and their expectations from police

officers. These were measured using racial resentment and perceptions of police bias,

! 31

respectively. To assess whether these variables mediated the effect of racial category on

outcomes and statement weights, we compared the coefficient for the racial dummy variable

across two sets of regressions – one controlling for demographics alone and one controlling for

demographics and priors (Table 7). Coefficients for the racial dummy measures are presented in

columns 1 (demographics only) and 2 (controlling for moderators). Coefficients for the two

mediators are shown in columns 3 and 4. And a Monte Carlo test of the difference between the

two racial dummy measures is shown in column 5.

Prior beliefs about Blacks and police bias largely mediated relations between racial

categorization and outcome measures. When race was simply included alongside demographic

controls (Table 7, column 1), we observed significant relations between race and all outcome

variables. When prior beliefs were included, however, the magnitude of the coefficients for race

decreased substantially, losing significance in many cases (column 2). This decrease was usually

itself statistically significant (column 5). Further, the coefficients on prior beliefs were often

strong and significant unique predictors of the outcomes (columns 3 and 4). Indeed, for seven of

13 outcomes, racial categorization was no longer a significant predictor of outcome measures

once racial resentment and perceptions of police fairness were controlled. These results suggest

that racial differences in perceptions stem from attitudes and expectations informed by lived

experiences that themselves differ across racial lines.

Moreover, if we insert prior beliefs and the average of the three racial identification

measures simultaneously into a set of regressions similar to those in Table 7, we find no residual

difference between Blacks and Whites for any of the outcomes we examined. This implies that

the motivated reasoning and prior belief models we examined collectively fully explain the

differences in information processing (Online Appendix G).

! 32

Table 7 - Explanatory Power of Prior Beliefs for Predicting Summary Judgments and Evidence Weights Model Demographics

Only Regression Regression Controlling for Demographics and Prior Beliefs Difference in

Black Coefficients Variable Black Black Racial

Resentment Police Fairness (Monte Carlo)

Outcome Coef s.e. Coef s.e. Coef s.e. Coef s.e.

Officer's Actions Appropriate -.17 (.03) *** -.05 (.03) .21 (.05) *** -.62 (.09) *** .12 ** Officer Should Be Charged .23 (.03) *** .13 (.03) *** -.29 (.05) *** .37 (.09) *** -.11 ** Taylor Attacked Silver -.11 (.02) *** .00 (.03) .14 (.05) ** -.58 (.08) *** .10 ** Taylor Had Weapon -.11 (.02) *** -.04 (.02) + .08 (.04) + -.41 (.08) *** .07 * Role of Race in Shooting .20 (.03) *** .06 (.03) * -.27 (.06) *** .68 (.09) *** -.14 ***

Police Chief (Pro-Officer) -.05 (.02) * .01 (.03) .19 (.05) *** -.18 (.09) * .06 + Mr. Davis (Pro-Taylor) .11 (.02) *** .09 (.03) *** -.07 (.05) .11 (.08) -.03 Officer Silver (Pro-Officer) -.06 (.02) * .01 (.03) .22 (.05) *** -.23 (.09) ** .07 * Mrs. Walker (Pro-Officer) .06 (.03) * .11 (.03) *** .00 (.06) -.34 (.10) ** .05 Mrs. Thomas (Pro-Taylor) .16 (.03) *** .13 (.03) *** -.12 (.05) * .08 (.09) -.03 Mrs. Williams (Pro-Taylor) .10 (.02) *** .03 (.03) -.07 (.05) .43 (.09) *** -.07 * Mr. Anthony (Pro-Officer) -.07 (.03) ** .00 (.03) .12 (.05) * -.31 (.09) *** .06 + Note: Difference column tests whether the influence of the Race Dummy differed significantly between the demographics only model and one where prior beliefs were controlled. Significance for these comparisons was tested using a Monte Carlo simulation with 100,000 resamples. All models control for age, education, sex, income, and partisanship. Analyses were limited to the 696 cases with no missing data on any of these measures. + p<.10 | * p<.05 | ** p<.01 | *** p<.001 two-tailed.

! 33

Discussion

The goal of the current study was to examine the mechanisms that underlie a persistent

divide in Americans’ responses to officer-involved shootings. First, we show in both a study

conducted shortly after Michael Brown’s death in Ferguson, Missouri as well as in a novel

scenario that attitudes pertaining to these cases diverged along racial lines. The fact that

differences emerged in the second study, where all respondents were presented with the same

information, suggests that this divergence was not simply the result of Blacks and Whites

occupying different information environments. Instead, Americans in different racial groups

perceived the same evidence differently.

In light of this conclusion, we tested two explanations for the racial divide. First, we

examined the possibility that social identity – in the form of racial salience – shapes how

individuals make sense of a racialized incident. We found mixed evidence for this suggestion

that these perceptual differences represented a form of racially motivated reasoning. In support

of this notion, we found that individuals expressed an interest in acquiring new information

likely to bolster their racial identities in a process akin to partisan selective exposure (Stroud

2011). We also found that racial identification measures moderated relations between racial

categorization and some summary judgments. Challenging this explanation, we found that

priming racial identity and experiences did not exacerbate racial divides,17 that racial

identification did not account for racial differences in factual beliefs about the incident, and that

individuals continued to update their beliefs even after forming an initial summary judgment. It

therefore appears that racial salience can motivate processing, but that this comprised at best a

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!17 While it is possible that the battery of additional items failed to prime race, earlier studies have found success with similar techniques (Steele and Aronson 1995). It is therefore likely either that race was inherently primed by the scenario or that the processes observed simply were not driven by motivational goals.

! 34

partial explanation of the diverging perceptions we observed.

We also found support for a second explanation, whereby individuals in different racial

groups interpreted novel information in ways that were shaped by preexisting beliefs and

expectations (Peffley and Hurwitz 2010). Here, we found that differences in summary judgments

and beliefs across racial groups corresponded with respondents’ racial attitudes and perceptions

of police bias. Combined with results suggesting that individuals were updating their beliefs in

line with new evidence, this implies that some of the racial divide is a product of information

updating from racially divergent priors (cf. Bullock 2009; Gerber and Green 2003). That is,

insofar as there are racial differences in response to officer-involved shootings, these differences

largely reflect average differences in prior beliefs and expectations related to the fairness of

policing and the culpability of Black victims.

Results from this study not only inform on-going conversations about officer-involved

shootings in the United States, they also complicate scholarly discussions related to biased

information processing. For one, the current project takes race seriously as a social construct (cf.

Sen and Wasow 2016) and argues that previous work on motivated reasoning in the political

domain should be extended to social identities beyond partisan identification. Yet, we cannot

attribute all differences in evaluations and attitudes between social groups to identity preserving

goals. That is, it appears that racial differences are informed not only by directional motivations

(cf. Kunda 1990), but also by different pre-existing beliefs and expectations.18

As we state at the outset, the implications and remedies differ depending on whether we

ascribe these differences to a psychological tendency to arrive at identity-consistent conclusions

or whether they stem from expectation-guiding priors that themselves differ across racial lines.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!18 Notably, however, these different priors could themselves have resulted from earlier motivated processes.

! 35

To address the former, we would need to somehow diminish the salience of race as a

consideration when evaluating officer-involved shootings. Not only is this difficult to

accomplish, but it precludes a range of potential interventions. The suggestion that lived

experiences and associated beliefs are underlying causes of the divide, however, yields more

tangible remedies. Though perceptions of police bias and racial resentment are clearly linked to

race, they may themselves be manipulable. Specifically, efforts to promote positive interactions

between police officers and Black community members could improve intergroup expectations,

resulting in less-polarized perceptions when incidents occur (cf. Pettigrew 1998).

As a final note, the reader will likely recognize that we have been careful throughout not

to expect differentially biased information processing from members of one racial group and not

the other. Indeed, there are good reasons to believe that both Whites and Blacks are similarly

susceptible to the tendencies to engage in identity-protective strategies or to update based on

existing beliefs and expectations. But the presence of similar psychological processes should not

be construed to imply that the effects of biased perceptions are somehow racially neutral. To the

contrary, the presence of the sorts of biases we identify can instead serve to exacerbate racial

tensions and to induce additional inequities.

Left unchecked, the perceptual biases we identify would be expected to result in a vicious

cycle of distrust between Blacks and the criminal justice system. When officer-involved

shootings are adjudicated in court, convictions may be structurally unlikely. As White jurors and

Black jurors consider the evidence, we would expect them to seek out information that confirms

their existing perspectives and build on pre-existing beliefs to arrive at diverging judgments. This

may make it difficult to achieve a unanimous conviction and may leave Black communities, in

particular, feeling that justice has not been served. And these differences in perceptions could

! 36

also have larger implications in a society stratified by race, where Whites maintain greater

influence over many of our social and political institutions.

Conclusion

It is no secret that White and Black Americans hold different attitudes on a range of

political issues. In this paper, we do more than merely note the presence of this racial divide in

the context of officer-involved shootings. Using data collected in response to an actual shooting

and data from a fictitious event, we not only show that Whites and Blacks differ in their

judgments and beliefs, but that these differences emerge as a consequence of biased information

processing. Our results offer two explanations for how this biased processing occurs. First, we

find some evidence that Whites and Blacks arrive at their opinions as a function of identity-

motivated processes. Yet, this identity-driven explanation does not capture the full scope of the

racial divide. We also find that Black-White differences emerge as a function of different prior

beliefs and expectations, that are themselves informed by race.

Our findings have important implications for debates both within and outside the

academic realm. Specifically, these results suggest that when individuals come to deliberate

about officer-involved shootings and other racially charged events, the information they

prioritize and draw upon will differ along racial lines. By attending to differences in beliefs,

expectations, and motivations when assessing the effects of race, scholars and policymakers can

more effectively understand the work that race is doing in the criminal justice world and beyond.

Whether the goal is to further explicate the meaning of race for an academic audience or to

understand its complexities to design effective interventions that aid in the administration of

justice, future scholarship must do more than simply note that race matters; it must take up the

much more difficult task of understanding why.

! 37

References

Abrams, Dominic, and Michael A Hogg. 1990. “Social Identification, Self-Categorization and

Social Influence.” European Review of Social Psychology 1(1): 195–228.

Bobo, Lawrence D, and Devon Johnson. 2004. “A Taste for Punishment: Black and White

Americans' Views on the Death Penalty and the War on Drugs.” Du Bois Review: Social

Science Research on Race 1(01): 151–80.

Bonilla-Silva, Eduardo. 1997. “Rethinking Racism: Toward a Structural Interpretation.”

American Sociological Review 62(3): 465.

Browning, Sandra Lee et al. 1994. “Race and Getting Hassled by the Police: a Research Note.”

Police Studies 17(1): 1–11.

Bullock, John G. 2009. “Partisan Bias and the Bayesian Ideal in the Study of Public Opinion.”

The Journal of Politics 71(3): 1109–24.

DiMaggio, Paul, John Evans, and Bethany Bryson. 1996. “Have American's Social Attitudes

Become More Polarized?” American Journal of Sociology 102(3): 690–755.

Dimock, Michael, and Carroll Doherty. 2013. Big Racial Divide Over Zimmerman Verdict.

Washington, DC: Pew Research Center.

Doherty, Carroll, Seth Motel, and Rachel Weisel. 2014. Sharp Racial Divisions in Reactions to

Brown, Garner Decisions. Washington, DC: Pew Research Center.

Drake, Bruce. 2014. Ferguson Highlights Deep Divisions Between Blacks and Whites in

America. Washington, DC: Pew Research Center.

Eagly, Alice H, and Shelly Chaiken. 1995. “Attitude Strength, Attitude Structure, and Resistance

to Change.” In Attitude Strength, eds. Richard E Petty and Jon A Krosnick. New York:

Lawrence Erlbaum.

! 38

Epp, Charles R, Steven Maynard-Moody, and Donald Haider-Markel. 2014. Pulled Over.

Chicago, IL: Chicago University Press.

Fiorina, Morris P., and Samuel J. Abrams. 2008. “Political Polarization in the American Public.”

Annual Review of Political Science 11(1): 563–88.

Fischle, Mark. 2000. “Mass Response to the Lewinsky Scandal: Motivated Reasoning or

Bayesian Updating?” Political Psychology 21(1): 135–59.

Garrett, R Kelly, and Natalie Jomini Stroud. 2014. “Partisan Paths to Exposure Diversity:

Differences in Pro" and Counterattitudinal News Consumption.” Journal of Communication

64(4): 680–701.

Gerber, Alan S, and Donald P Green. 1998. “Rational Learning and Partisan Attitudes.”

American Journal of Political Science 42(3): 794–818.

Gerber, Alan, and Donald P Green. 2003. “Misperceptions About Perceptual Bias.” Annual

Review of Political Science 2(1): 189–210.

Green, Donald P, Bradley Palmquist, and Eric Schickler. 2002. Partisan Hearts and Minds:

Political Parties and the Social Identities of Voters. New Haven, CT: Yale University Press.

Greene, Steven. 2004. “Social Identity Theory and Party Identification.” Social Science

Quarterly 85(1): 136–53.

Hamilton, David L, Steven J Sherman, and Catherine M Ruvolo. 1990. “Stereotype"Based

Expectancies: Effects on Information Processing and Social Behavior.” Journal of Social

Issues 46(2): 35–60.

Hart, William et al. 2009. “Feeling Validated Versus Being Correct: a Meta-Analysis of

Selective Exposure to Information..” Psychological Bulletin 135(4): 555–88.

Hodson, Gordon, John F Dovidio, and Samuel L Gaertner. 2002. “Processes in Racial

! 39

Discrimination: Differential Weighting of Conflicting Information.” Personality and Social

Psychology Bulletin 28(4): 460–71.

Huddy, Leonie. 2001. “From Social to Political Identity: a Critical Examination of Social

Identity Theory.” Political Psychology 22(1): 127–56.

Huddy, Leonie. 2003. “Group Identity and Political Cohesion.” In Emerging Trends in the Social

and Behavioral Sciences, Hoboken, NJ, USA: John Wiley & Sons, Inc., 1–14.

Huddy, Leonie, and Stanley Feldman. 2006. “Worlds Apart: Blacks and Whites React to

Hurricane Katrina.” Du Bois Review: Social Science Research on Race 3(01): 97–113.

Huddy, Leonie, Lilliana Mason, and Lene Aarøe. 2015. “Expressive Partisanship: Campaign

Involvement, Political Emotion, and Partisan Identity.” American Political Science Review

109(01): 1–17.

Hurwitz, Jon, and Mark Peffley. 2005. “Explaining the Great Racial Divide: Perceptions of

Fairness in the U.S. Criminal Justice System.” The Journal of Politics 67(3): 762–83.

Hutchings, Vincent L. 2009. “Change or More of the Same? Evaluating Racial Attitudes in the

Obama Era.” Public Opinion Quarterly 73(5): 917–42.

Hutchings, Vincent L, and Nicholas A Valentino. 2004. “The Centrality of Race in American

Politics.” Annual Review of Political Science 7: 383–408.

Iyengar, Shanto, and Kyu S. Hahn. 2009. “Red Media, Blue Media: Evidence of Ideological

Selectivity in Media Use.” Journal of Communication 59(1): 19–39.

Iyengar, Shanto, Gaurav Sood, and Yphtach Lelkes. 2012. “Affect, Not Ideology: a Social

Identity Perspective on Polarization.” Public Opinion Quarterly 76(3): 405–31.

Kahan, Dan. 2010a. “Fixing the Communications Failure.” Nature 463(7279): 296–97.

Kahan, Dan M. 2010b. “Culture, Cognition, and Consent: Who Perceives What, and Why, in

! 40

Acquaintance-Rape Cases.” University of Pennsylvania Law Review 158(3): 729–813.

Kinder, Donald R, and David O Sears. 1981. “Prejudice and Politics: Symbolic Racism Versus

Racial Threats to the Good Life..” Journal of Personality and Social Psychology 40(3): 414–

31.

Kinder, Donald R, and Lynn M Sanders. 1996. “Divided by Color: Racial Politics and

Democratic Ideals.” University of Chicago Press.

Kinder, Donald R, and Nicholas Winter. 2001. “Exploring the Racial Divide: Blacks, Whites,

and Opinion on National Policy.” American Journal of Political Science 45(2): 439–56.

Kunda, Ziva. 1990. “The Case for Motivated Reasoning..” Psychological Bulletin 108(3): 480.

Kunda, Ziva, and Lisa Sinclair. 1999. “Motivated Reasoning with Stereotypes: Activation,

Application, and Inhibition.” Psychological Inquiry 10(1): 12–22.

Lawrence, Regina G. 2000. The Politics of Force. Berkeley, CA: University of California Press.

Lebo, Matthew J, and Daniel Cassino. 2007. “The Aggregated Consequences of Motivated

Reasoning and the Dynamics of Partisan Presidential Approval.” Political Psychology 28(6):

719–46. http://onlinelibrary.wiley.com/doi/10.1111/j.1467-9221.2007.00601.x/full.

Leeper, Thomas J, and Rune Slothuus. 2014. “Political Parties, Motivated Reasoning, and Public

Opinion Formation.” Political Psychology 35(S1): 129–56.

Lodge, Milton, and Charles S Taber. 2005. “The Automaticity of Affect for Political Leaders,

Groups, and Issues: an Experimental Test of the Hot Cognition Hypothesis.” Political

Psychology 26(3): 455–82.

Lodge, Milton, and Charles S Taber. 2013. The Rationalizing Voter. Cambridge University

Press.

Marist Poll. 2014. Race Shapes Americans’ Attitudes About Decisions in Ferguson.

! 41

Massey, Douglas S, and Nancy A Denton. 1993. American Apartheid. Cambridge, MA: Harvard

University Press.

Miller, Joanne M., Kyle L Saunders, and Christina E Farhart. 2015. “Conspiracy Endorsement as

Motivated Reasoning: the Moderating Roles of Political Knowledge and Trust.” American

Journal of Political Science: n/a–n/a.

Newman, Leonard S et al. 1997. “Reactions to the O. J. Simpson Verdict: ‘Mindless Tribalism’

or Motivated Inference Processes?” Journal of Social Issues 53(3): 547–62.

http://doi.wiley.com/10.1111/0022-4537.00034.

Nunnally, Shayla C. 2012. Trust in Black America. New York: New York University Press.

Nyhan, Brendan, and Jason Reifler. 2010. “When Corrections Fail: the Persistence of Political

Misperceptions.” Political Behavior 32(2): 303–30.

Peffley, Mark, and Jon Hurwitz. 2010. Justice in America: the Separate Realities of Blacks and

Whites. New York, NY: Cambridge University Press.

Pettigrew, Thomas F. 1998. “Intergroup Contact Theory.” Annual Review of Psychology 49(1):

65–85. http://www.annualreviews.org/doi/abs/10.1146/annurev.psych.49.1.65.

Sen, Maya, and Omar Wasow. 2016. “Race as a Bundle of Sticks: Designs That Estimate Effects

of Seemingly Immutable Characteristics.” Annual Review of Political Science 19(1): 499–

522.

Sirin, Cigdem V, José D Villalobos, and Nicholas A Valentino. 2016. “Group Empathy Theory:

the Effect of Group Empathy on US Intergroup Attitudes and Behavior in the Context of

Immigration Threats.” The Journal of Politics 78(3): 893–908.

Steele, Claude M, and Joshua Aronson. 1995. “Stereotype Threat and the Intellectual Test

Performance of African Americans..” 69(5): 797–811.

! 42

http://psycnet.apa.org.proxy.lib.umich.edu/journals/psp/69/5/797.html.

Steffensmeier, Darrell, Jeffery Ulmer, and John Kramer. 1998. “The Interaction of Race, Gender,

and Age in Criminal Sentencing: the Punishment Cost of Being Young, Black, and Male.”

Criminology 36(4): 763–98.

Stroud, Natalie Jomini. 2011. Niche News: the Politics of News Choice. New York: Oxford

University Press.

Tajfel, Henri. 1974. “Social Identity and Intergroup Behaviour.” Social Science Information

13(2): 65–93.

Theobald, Nick A, and Donald P Haider-Markel. 2009. “Race, Bureaucracy, and Symbolic

Representation: Interactions Between Citizens and Police.” Journal of Public Administration

Research and Theory 19(2): 409–26.

Unnever, James D. 2008. “Two Worlds Far Apart: Black-White Differences in Beliefs About

Why African-American Men Are Disproportionately Imprisoned.” Criminology 46(2): 511–

38.

Weitzer, Ronald, and Steven A Tuch. 2005. “Racially Biased Policing: Determinants of Citizen

Perceptions.” Social Forces 83(3): 1009–30.

Westen, Drew et al. 2006. “Neural Bases of Motivated Reasoning: an fMRI Study of Emotional

Constraints on Partisan Political Judgment in the 2004 U.S. Presidential Election.” Journal of

Cognitive Neuroscience 18(11): 1947–58.

seeing blue in black and white: race and perceptions of ... · shooting of michael brown in...

Documents