ArtPubOpinImmig.pdf

Article

Group Size versus Change?
Assessing Americans’ Perception
of Local Inn migration

Political Research Quarterly
2014, Vol. 67(2)293-303
© 2 0 1 4 University of Utah
Reprints and permissions;
sagepuh.com/journalsPermissions.nav
D O I : 10.1177/1065912913517303
prq.sagepuh.com

USAGE

Benjannin J. Newman’ and Yamil Velez^

Abstract
Leading opinion research on Immigration has begun to move from size-based to change-based measures of citizens’
ethnic context. This shift is based on the theoretical assumption that over-time grov^h in immigrant populations is
more likely to capture citizens’ attention than their current size. At present, there is no empirical evidence supporting
this assumption. This article demonstrates that while the size of local immigrant populations exerts virtually no
effect on perceived immigration, over-time growth strongly influences citizens’ perceptions of immigration into their
community. In addition, our analyses illuminate the differential contribution of growth in local Hispanic and Asian
populations to perceived immigration.

Keywords
immigration, public opinion, contextual effect, ethnic change

Recent research on immigration has begun to focus on
over-time growth, rather than contemporaneous size, as
the theoretically and empirically relevant characteristic
of local immigrant minority populations responsible for
driving threat and shaping political behavior and policy
outcomes. Indeed, across several fields of research,
scholarship finds that increases in local immigrant pop-
ulations, rather than their prevailing levels, are respon-
sible for driving behavioral outcomes ranging from
racially motivated hate crime (Green, Strolovitch, and
Wong 1998) and native neighborhood out-migration
(Crowder, Hall, and Tolnay 2011) to xenophobic voting
behavior (Alexseev 2006) and public opinion on immi-
gration more generally (Hopkins 2010;Newman 2013;
Newman and Johnson 2012). At the policy level,
changes in ethnic diversity have been found to trump the
effects of standing levels of heterogeneity in influencing
outcomes such as local government investment in pub-
lic goods (Hopkins 2009) and enactment of anti-immi-
grant ordinances (Hopkins 2010), as well as state
government passage of “official English” language laws
(Citrin et al. 1990) and E-Verify legislation (Newman
et al. 2012).

The shin in focus from the size of immigrant popula-
tions to their over-time growth has largely been moti-
vated at the practical level by the empirical weakness of
immigrant population levels, and thus corresponding
racial or “power threat” hypotheses (Blalock 1967; Key

1949), in predicting political outcomes of interest
(Hainmueller and Hopkins 2014). At the theoretieal level,
the rising focus on the over-time growth, rather than the
size, of immigrant populations is grounded upon the
mounting assertion that drastic changes in local immi-
grant populations are more likely to capture the attention
of local eitizens than the size of these populations
(Hopkins 2010; Newman 2013). We label this assertion
the salience-of-change hypothesis. While this assertion
has received indirect support from extant research dem-
onstrating significant effects for population growth coin-
ciding with null effects for levels (e.g., Hopkins 2009,
2010; Newman et al. 2012), scholarship in the field of
immigration has yet to provide any direct evidence that
growth in local immigrant populations captures citizens’
attention more than their size, or that growth is perceived
at all. In the absence of such evidence, this shift in focus
within the literature, as well as the findings that support
it, are open to uncertainty and criticism regarding whether
a key link in presumed causal processes is in fact true.
Indeed, any disconfirming evidence would throw into

‘University of Connecticut, Stamford. USA
^Stony Brook University, NY, USA

Corresponding Author:
Benjamin J. Newman, Department of Political Science, University of
Connecticut, One University Place, 367, Stamford, CT 06901, USA.
Email: [email protected]

294 Po/;t/co/ Research Quarterly 67(2)

question the validity of the theories underlying these
findings and relegate otir explanation of them to the
“black box.”

Perception of Ethnic Context and
Salience of Change

The case for focusing on over-time changes in immigrant
populations, rather than their present levels, is made most
recently by Hopkins (2010) in a leading study explaining
the sources of public opposition to immigration. Hopkins
reviews the opinion literature on immigration and notes
the inconsistent nature ofthe findings for the power threat
hypothesis (see Blalock 1967), where some studies find
that the residing near large immigrant minority popula-
tions triggers threat (Campbell et al. 2006; Stein et al.
2000; Tolbert and Gmmmel 2003), others find that it
decreases threat (Fetzer 2000; Fox 2004; Hood and
Morris 1997), and most studies find that it exerts no effect
on citizens’ opinions on immigration (Cain et al. 2000;
Citrin et al. 1990; Dixon and Rosenbaum 2004; Frandreis
and Tatolovitch 1997; Taylor 1998). Hopkins suggests
that the inconsistent results might be driven by scholars’
use of size-based rather than change-based measures of
ethnic context. Indeed, Hopkins argues that the empirical
weakness of size-based measures may be due to the fail-
ure of one of the key preconditions for threat to obtain:
citizens ‘perception of their ethnic context.

In addition to citing the literature on citizens’ innumer-
acy with respect to minority populations and noting that
residential and occupational segregation further limits the
visibility of immigrants, Hopkins attempts to resolve the
issue of citizens’ inattention to immigrant population sizes
by issuing the critical assertion that “while levels of ethnic
heterogeneity might escape notice, changes are less likely
to do so” (2010, 42). The assertion that growth in local
immigrant populations will be perceived by residents is
critical to Hopkins’ theory, as his core hypothesis and
findings, as well as those of other scholars focusing on
change rather than size (e.g.. Green, Strolovitch, and
Wong 1998; Newman 2013), are predicated on this asser-
tion, which we label the salience-of-change hypothesis.
This hypothesis predicts that (1) citizens will perceive
over-time growth in local immigrant populations, and that
(2) growth will be more perceptually salient than levels of
local immigrants. The salience-of-change hypothesis, as
advanced by Hopkins (2010), is grounded in prospect
theory, which argues that in the presence of large quanti-
ties of incoming information, people are especially atten-
tive to change (Kahneman and Tversky 1979). According
to Hopkins, in the midst of overwhelming streams of
incoming information from their environments, the lim-
ited attention of citizens will likely be drawn to substantial
in-migrations of immigrants.

This expectation is further supported by existing
research on racial integration and community life, which
suggests that, while processes of enculturation and habit-
uation (Sam and Berry 2010) as well as residential self-
selection (Clark 1992; Oliver and Wong 2003) lead
citizens to be acclimated to baseline levels of diversity in
their community of origin or selection (Hopkins 2009;
Newman 2013), sudden changes in the ethnic composi-
tion of a community constitute a riveting stimulus because
it threatens the displacement of the sociocultural status
quo within a community. As prior ethnographic research
doctiments, sudden ethnic change may undermine exist-
ing social networks and long-term residents’ conceptual-
izations of community identity (Horton 1995; Rieder
1985; Suttles 1972), thus altering expectations about
community life and creating uncertainty about the future
(Hopkins 2009). Taken together, this research provides a
theoretical and empirical foundation for the salience-of-
change hypothesis advanced by Hopkins (2010) because
it casts extant levels of diversity as a presumed quality of
one’s neighborhood while pinpointing the pressure points
stmck by ethnic change. In short, influxes of immigrants
should be locally salient over current population sizes
precisely because change, rather than levels, is what
threatens personal (e.g., tenure in community, property
values) and collective (e.g., social capital, community
identity, neighborhood quality, public services) goods
attached to one’s community in its current form.
Furthermore, sudden ethnic change arouses the potential
for conflict between local ethnic majorities and arriving
immigrant minorities (Green et al. 1998; Horton 1995).
Building on this logic, psychologists have documented
the existence of a negativity bias in humans, whereby
negative information is (1) more potent, (2) associated
with more cognitive elaboration, (3) more capable of
tainting perceptions of objects, and (4) attended to with
more urgency than positive information (see Rozin and
Royzman 2001). Insofar as immigrants are perceived in
negative terms and associated with various threats, we
ought to expect residents to be more attentive to changes
in local immigrant populations than prevailing levels,
especially since these changes ought to signal a signifi-
cant deviation from what residents have come to expect
from their local community.

While intuitively plausible and situated in firm theory,
the core predictions of the salience-of-change hypothesis
have yet to be directly empirically confnmed. To be sure,
leading work in the field has positioned immigration-driven
ethnic change as a central causal factor driving contempo-
rary public opinion and state and local policy outcomes
with respect to immigration, and this work contains shared
embedded hypotheses conceming citizens’ perception of
their ethnic context. What is now needed is to move beyond
assumption and indirect evidence and directly determine

Newman and Velez 295

whether (1) citizens are “receiving the treatmenf (i.e.,
aware of immigrant population growth) and (2) whether the
“treatment effect” of immigrant growth surpasses any
observed for immigrant population size.

The Differential Salience of Different
Immigrant Groups

In addition to the question of whether citizens perceive
immigrant growth, a substantively interesting question is
whether the attention given to growth is heterogeneous
across different immigrant groups. For example, are per-
ceptions of growth more strongly tied to changes in local
Hispanic populations than to Asian or other immigrant
populations? At present, the literature provides a sound
basis for the expectation that citizens will pay more atten-
tion to local Hispanic populations.

Hispanics are the largest and fastest growing nonblaek
ethnic minority group, constitute the majority of the for-
eign-bom population, and are consistently the largest group
among new immigrants into the country (Passel, Cohn, and
Lopez 2011 ). News stories mentioning immigration refer to
Hispanics nearly twice as often as the next most mentioned
group (East Asians), and more stories spotlight Hispanic
immigration than immigration from all other regions com-
bined (Brader, Valentino, and Suhay 2008; Valentino et al.
2013). In contrast to other immigrant groups, Hispanies
have been argued to pose a unique and unprecedented threat
to American society and culture (Chavez 2008; Huntington
2004). The link between Hispanics and threat is stronger
than with Asians, for example, who are often stereotyped as
“model minorities” who successfully assimilate into
Ameriean society (Kim 1999). This is reinforced by several
studies finding that while residential proximity to large
Asian populations reduces Anglo opposition to immigra-
tion, proximity to large Hispanic populafions augments
anti-immigrant sentiment (Ha 2010; Hood and Morris
2000). All of these considerations together lend support to a
Hispanic threat hypothesis, which predicts that citizens’
perception of local immigration will be more responsive to
growth in local Hispanic populations than to growth in
Asian or other immigrant groups.

Analysis

To test these hypotheses, we draw upon the 2006 Poll on
Immigration conducted by the Pew Research Center for
the People & the Press and the Pew Hispanic Center. This
poll is a nationally representative survey of the adult
American population conducted by telephone between
Febmary 8 and March 7, and contains a total sample size
of A =̂ 2,000. For the purposes of the present analyses, we
focus our tests on the perceptions of the 1,499 non-His-
panic whites included in the national sample.’

Measurement

To measure citizens’ awareness of changes in the local
immigrant population, we rely on a survey item asking
respondents, “How many recent immigrants would you
say live in your area?” There are four ed response
options for this question, ranging from (1) “Many” and
(2) “Some,” to (3) “Only a few,” and (4) “None.” This
item will serve as the main dependent variable for all of
our analyses. In contrast to the standard “local commu-
nity” item in other numeracy research that asks respon-
dents to guess the percentage of various racial/ethnic
groups in their community (e.g., Wong 2007), this ques-
tion specifically refers to immigrants, and moreover,
recent immigrants. In the world of opinion data, it is rare
to encounter questions tapping citizens’ perceptions of
their local racial/ethnic context, let alone an item with
this degree of specificity. As such, this item represents the
best available measure of perceptions of local immigra-
tion in existing survey data and provides a unique oppor-
tunity to test our hypotheses. While not as fine grained as
a continuous measure, this ordinal item does provide a
coarse measure of variation in citizens’ perception of the
amount of recent immigrants in their proximate environ-
ment. Finding that actual growth in local immigrant pop-
ulations corresponds to an increase in the probability of
reporting “Many” as compared with “Some” recent
immigrants in one’s local environment would provide the
literature with some important and unprecedented—
albeit rough—evidence that changes in local immigrant
populations do register on the perceptual radar of
citizens.

To measure actual immigration, we rely upon county-
level data from the 1990 Decennial Census and the 2005-
2009 American Community Survey conducted by the U.S.
Census Bureau.^ We created two measures pertaining to
the foreign-bom population in each respondent’s county.
First, we obtained the percentage of the county population
that was foreign bom in 2005-2009 {percentage immi-
grant, X = 9.0, â = 8.4, minimum = 0.08, maximum =
47.1), and then subtracted the percentage foreign bom in
1990 from the 2005-2009 value to generate a percentage
point change variable {immigrant growth, x =3.7, a =
3.30, minimum = -1.5, maximum = 18.1). This process
was repeated with the Hispanic and Asian population^ in
each county to produce corresponding measures of levels
{percentage Hispanic, x = 10.6, â = 12.5, minimum = 0,
maximum = S9.3; percentage Asian, x = 3 . 1 , à = 3 . 9 ,
minimum = 0, maximum = 31.2) and pereentage point
change in each of these two specific minority populations
{Hispanic growth, x = 4.9, â = 4.6, minimum = – I . I ,
maximum = 2.6; Asian growth, x = 1.3, â = 1.7, min =
-1.2, max = 13.0). Figure 1 displays perceived local
immigrafion and actual levels of immigrant growth for the

296 Politícal Research Quarterly 67(2)

Perceived Local Immigration Actual Immigrant Gronih

Figure I. County-level mapping of perceived and actual immigration.

Table I. Correlations between Key Contextual Variables.

% Immigrant
Immigrant growth
% Hispanic

Hispanic grov/th
% Asian
Asian growth

Median household income
% Bush vote
Population density

Total population

1.0
.72
.82
.54
.65
.55
.38

-.09
-.09

.29

1.0
.60
.68
.64
.77
.41

-.40
.29
.51

1.0
.77
.26
.24
.02
.06

-.18
.48

1.0
.14
.23

-.07
-.01

.04

.62

1.0
.90
.63

-.18
.10
.05

1.0
.62

-.33
.22
.18

1.0
-.10

.07
-.11

1.0
-.45
-.20

1.0
.15 1.0

Eastem part of the United States.”* Each grey-scale gradi-
ent represents a sample quintile. Both maps look strik-
ingly similar. Many of the counties that experienced low
immigrant growth rates in the period examined, such as
those adjacent to Lake Erie, also appear to score low on
perceived immigration. In high immigration states like
Florida, there appears to be a strong correspondence
between actual and perceived immigrant growth, such that
residents of high growth counties also tend to have per-
ceptions that strongly map onto the objective features of
their environment. While these pattems are not perfect
and some counties show a misalignment between per-
ceived and aetual immigrant growth, the two maps are

alike in many respects, thus hinting at a possible system-
atic relationship.

Our analyses included a variety of controls. First, to
control for the potential effects of variation in affluence,
political culture, urbanicity, and size across the eounties in
our data on perceived levels of immigration, our analyses
include measures of median household income, % Bush
vote 2004, population density, and total population for
each county.^ The correlations between our contextual
variables are presented in Table 1. The table reveals that
over-time growth in foreign-bom, Hispanic, and Asian
populations are strongly correlated with their correspond-
ing size measures in 2006. At the individual level, controls

Newman and Velez 297

were included for education, income, age, gender (1 =
male), homeownership, and place of birth (1 = born in the
United States). To control for the potential role of eco-
nomic concerns in shaping attention to immigrant popula-
tions, all models include measures of employment status
(1 = unemployed), sociotropic and pocketbook economic
evaluations, and respondents’ assessment of the job oppor-
tunities in their local community. To control for respon-
dents’ political and symbolic leanings, all models include
controls for party identification, ideological self-identifi-
cation, and negative affect toward Hispanics and Asians.*
These latter two measures were included to control for the
tendency of citizens to use their feelings toward groups as
a heuristic for making judgments about a group (e.g.,
Brady and Sniderman 1985). Last, as existing research
demonstrates that opposition to immigration is higher
among citizens residing in the United States-Mexico bor-
der states (e.g.. Bums and Gimpel 2000; Hood and Morris
1997), we included a control for b state (1 = b
stated to account for potential differences in attention to
local immigrant populations between citizens residing in
b and nonb states. For ease of interpretation, all
variables, except age, were recoded to range from 0 to 1.
For more information about variable measurement and
question wording, please see the online appendix.^ Given
the ordinal nature of the perceived immigration dependent
variable and our use of county-level demographic predic-
tors, we estimate ed logistic regression models with
standard errors clustered by county.’

Results

Change versus Levels

Figure 2 (Panel A) plots the results from our ed
logistic regression analysis (for full results tables, see
supplemental Appendix B). The results in Panel A reveal
that whites residing in counties experiencing high growth
in the immigrant population were significantly more
likely to report the highest level of perceived recent
immigration into their local area (B = 2.96, 5 ^ = .672,
p < .000). In addition to supporting the hypothesis by demonstrating a strong correspondence between actual immigrant growth and the amount of perceived immigra- tion, the results also indicate that the size of the immi- grant population in 2000 in whites' county of residence exerted no effect on their perceptions of the degree of recent local immigration (B = 1.07, SE= .03, not signifi- cant). To provide a sense of the magnitude of the effect of immigrant growth, and how strongly perceived immigra- tion tracks with actual immigrant growth. Figure 3 (Panel A) presents the predicted probability of perceiving "many" recent immigrants in one's local area across min- imum, 5th percentile, mean, 95th percentile, and maximum values oí immigrant growth. For example, the change in the predicted probability of reporting the high- est level of recent immigration when moving from whites in counties at the 5th to the 95th percentile values of immigrant growth is roughly .35, which reveals that in addition to exerting a highly significant efîect on percep- tions, actual immigrant growth exerted a substantively large effect on perceived immigration. Furthermore, this effect is the largest in magnitude among all other signifi- cant variables in the equation, such as age, where a move- ment from the 5th to 95th percentile (i.e., 25-79 years old) is associated with a more modest decrease of .127 in the probability of perceiving high recent immigration. Taken together, these results establish a finding of critical importance for the contextual opinion literature on immi- gration, which is that citizens, while arguably innumerate with respect to group size, are perceptually astute when it comes to grovi^h in the local immigrant population.'" Hispanic and Asian Population Changes The results thus far reveal that citizens do perceive the growth in local immigrant populations. Here, we con- sider the results ft'om our analyses of how changes in the population size of the two most numerous immigrant groups—Hispanics and Asians—factor into citizens' perceptions of local immigration." Consistent with the Hispanic threat hypothesis, the results presented in Figure 2 (Panel B) reveal that while local Asian popula- tions factor little into citizens' evaluations of local immigration, local Hispanic populations are indeed a significant force in shaping citizens' perception of recent immigration in their local context. However, con- sistent with the salience of change hypothesis, only Hispanic growth emerged significant. Moving from counties with minimal to maximal growth in the Hispanic population is associated with a significant increase in the probability of reporting the highest level of perceived recent immigration into one's local area (B = 3.54, SE = .775, p < .000). In addition to being statistically significant, the size of the effect, displayed in Figure 3 (Panel B), is substantively meaningful; for example, the change in predicted probability of report- ing the highest level of perceived immigration when moving from the 5th to 95th percentile of Hispanie growth is .29—again, an effect that dwarfs all other sig- nificant variables in the equation. Robustness Check Using ZIP-code-level Data So far, our core results have been that growth matters while size does not, and that citizens' perceptions are highly responsive to Hispanic population shifts but rather insensi- tive to Asian population shifts. To assess the robustness of 298 Political Research Quarterly 67(2) Panel A Peicent tamigtant Median Household Income Percent Bush Vote (2004) Population Density Total Population Education Income Age Oend« Homeowner Nativt Bom unemployed Sodotropic Evaluations Pocketbook Evaluations Local lob Mafeet Perceptions Patty Identification Id Hispaiác Affect Asian Affect B State .4 Panel B I I I -2 0 2 Ordered Logistic Regresaon CoefiSck»* Hispanic Growth Percent Hispanic Asian Growth Percent Asian Median Household Income Percent Bush Vote (Mi) Peculation Density Total Population Education Income Age Gender Native Bom Unemployed Sociotropic Evaluations Pocketbook Evaluations L

1

5
Coefficient

1

10

Figure 4. Effect of ZIP-code-level Hispanic and Asian growth and levels on perceived immigration.
Note. To view full results’ tables, see the online appendix.

empirical confirmation for the logic underlying this shift
in focus and should help allay scholars’ concems about the
validity of this presumed perceptual mechanism linking
context to policy opinion.

Our results are additionally important given recent
research calling into question the validity of using cen-
sus-defined geographies in survey research as measures
of citizens’ residential context. For example, Wong et al.
(2012) argue for an approach that allows citizens to define
their own community context and find that these contexts
rarely overlap with census geographies. While the mea-
surement strategy pioneered by Wong et al. represents an
innovation relative to existing approaches, we believe
that census geographies are useful as proxies for racial
and ethnic context so long as (1) the requisite perceptual
assumptions hold and (2) scholars show that their results
do not depend on the use of specific contextual units. The
results from this article meet these criteria, as they reveal
that variation in citizens’ pereeptions of immigration into
their “community” maps onto variation in objective
immigration across census-defined local units.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.

Funding

The author(s) received no financial support for the research,
authorship, and/or publication of this article.

Notes

1. These non-Hispanic white respondents are spread aeross
808 …