Ethnic Change and Polarization over Immigration
In the American Public
This article explores the interplay between ethnic change and individual psychology in shaping mass opinion on immigration. Recent research highlights a “personality cleavage” underlying the left-right divide in American politics. Extending this into the domain of immigration, we argue this cleavage should be an important factor shaping citizens’ reactions to ethnic change. Using national survey data and a national survey experiment, we demonstrate that uncertainty aversion moderates the effect of perceived and actual ethnic change on citizens’ immigration attitudes. Our analysis reveals that ethnic change polarizes citizens by personality as those averse to uncertainty feel heightened cultural threat from ethnic change, while those open to uncertainty and novelty feel less threatened. As such traits are associated with left-right identifications, our results suggest that polarization of the American public over immigration, rather than being a mere product of top-down elite influence, is significantly driven by the interaction of citizen and context.
Word Count: 8,460
Keywords: immigration, public opinion, ethnic change, polarization, personality, ideology
The United States is growing more ethnically diverse each year, with steady immigration and a rapidly growing Hispanic population altering the sociocultural landscape surrounding many American citizens. What are the consequences of ethnic change for the politics of immigration in the United States? More specifically, what are the effects of ethnic change on citizens’ opinions on immigration? At present, the opinion literature on immigration comes up surprisingly short in providing a satisfactory answer to these questions. The lion’s share of studies in the opinion research analyzing Americans’ immigration policy preferences focus on the effects of the size of, rather than the change in, the immigrant populations surrounding citizens. Beyond the evident limitation of not directly addressing ethnic change, the results from studies addressing the effects of the size of immigrant populations are notoriously inconclusive. While lay intuition may suggest that immigration-driven ethnic change across the nation may explain observable patterns of nativist resentment and anti-immigrant sentiment among the American public, the empirical research has yet to establish a strong connection between growing immigrant populations and individual opposition to immigration.
In this article, we engage the question of the impact of ethnic change on mass opinion on immigration and offer two key innovations over the standard approach taken in past research. First, we explore the effect of over-time growth in, rather than the size of, immigrant populations as the key feature of these populations responsible for driving public opinion. Second, in contrast to previous research which largely assumes that the effects of ethnic context are uniform across all citizens, we offer a novel framework that explores heterogeneity in the effects of ethnic change on citizens’ attitudes toward immigration. Recent research demonstrates that a significant portion of the variance in ideological orientations in the contemporary American public can be explained by differences on a small set of core personality traits (e.g. Hetherington and Weiler 2009; Jost et al. 2003; Mondak 2010). Extending this perspective to immigration, we argue that this “personality cleavage” should be highly influential in shaping how citizens respond to ethnic change with important consequences for polarization of opinion across the left-right divide in American politics.
This article develops the differential adaptation hypothesis, which argues that substantial ethnic change should serve as an environmental determinant of opinion on immigration, but that citizens should react differently to ethnic change conditional on their relative aversion or attraction to epistemic uncertainty and novelty. The rapid influx of members of racial and ethnic out-groups can be viewed as threatening to existing cultural institutions, and implies uncertainty in one’s environment and one’s interaction with fellow citizens. As Jost and Hunyady (2005) note, “There is a good match between needs to reduce uncertainty” and the experience of institutional change as threatening, because “preserving the status quo allows one to maintain what is familiar while rejecting the uncertain prospect of social change” (p. 262). Such change could, however, also be viewed in a positive light, as diversity and opportunity enhancing. As Gerber et al. (2010) argue with respect to institutional change generally, “It follows that this attraction to novelty and tolerance for complexity encourage not only overall liberalism, but also support for liberal social and economic policies, which typically involve new programs or interventions that overturn existing practices” (p. 116). According to the differential adaptation hypothesis, there should be differential responsiveness to rapid demographic change across personality types, such that the uncertainty averse should see ethnic change as threatening, while those comfortable with uncertainty and attracted to novelty should find such changes desirable.
We test the differential adaptation hypothesis in two steps. In Study 1, using the 2005 CID national survey and data from the U.S. Census Bureau, we demonstrate that traits related to uncertainty aversion moderate the effect of local ethnic change on the perception that immigrants pose a cultural threat. In Study 2, we report the results of an internet-based national survey experiment that manipulated perceptions of ethnic change. The data from this study strongly reinforce the cross-sectional findings from Study 1—namely, that citizens with a strong aversion to uncertainty are significantly more culturally threatened than their uncertainty-tolerant counterparts by induced perceptions of ethnic change. In addition to this core finding, we also demonstrate the political relevance of the interplay between personality and ethnic change through the estimation of a structural equation model linking perceptions of cultural threat to support for restrictive immigration policies. The results from our analyses demonstrate that the differential experience of cultural threat, across personality, in response to perceived or actual local ethnic change ultimately results in a substantial deepening of polarization on immigration policy within the American public.
In total, this article makes several important contributions. First, it contributes to the opinion literature on immigration. We move beyond group-size-based measures of ethnic context that tend to dominate the contextual research and instead focus on ethnic change as the principle feature of citizens’ ethnic context driving their opinions on immigration. Further, we reconcile tension between intergroup threat and contact theories by demonstrating heterogeneity in the effects of ethnic change, with change showing opposite effects across citizen types. And last, we demonstrate that the effects of ethnic change on citizens’ policy preferences are mediated by cultural threat perceptions, thus offering the opinion literature a mediated-moderated effects model of opinion formation on immigration. Beyond these specific contributions, we believe our paper also makes a more general contribution to the study of public opinion and political behavior. By assessing the interplay of large-scale, objective contextual processes and individual psychological factors, we fuse the macro with the micro into an integrated approach for understanding political behavior.
IMMIGRATION, ETHNIC CONTEXT, AND PUBLIC OPINION
The ethnic composition of citizens’ residential environment has long stood as a primary factor hypothesized to account for public opinion on immigration. Underlying the research on the contextual sources of opinion on immigration rests the issue of identifying which aspect of immigrant populations is responsible for driving public opinion. A substantial body of opinion research exists that explores the effect of the size of the immigrant population surrounding citizens on their immigration policy preferences. The racial or power threat hypothesis (Blalock 1967; Key 1949), when translated from White-Black relations to the case of immigration, argues that anti-immigrant sentiment and policy support will be greater among citizens residing in more immigrant heavy areas (Hopkins 2010). This line of opinion research, however, has generated notoriously mixed results, with some studies finding limited evidence in support of the power threat hypothesis (Campbell, Wong, and Citrin 2006; Tolbert and Grummel 2003), other studies finding that residing near large immigrant populations reduces anti-immigrant sentiment and policy support (Fetzer 2000; Hood and Morris 1997), and the bulk of the research finding that the size of local immigrant populations exerts no significant effect on citizens’ immigration policy preferences (Cain, Citrin, and Wong 2000; Citrin et al. 1990; Citrin, Reingold, Walters, and Green 1990; Dixon and Rosenbaum 2004; Taylor 1998).
While several factors have been proposed to explain the inconsistency of results for group-size based measures of ethnic context, and the relative empirical weakness of the power threat hypothesis (e.g. degree of contact, Hood and Morris 2000; residential segregation, Rocha and Espino 2008), they ignore perhaps its most defining aspect, which is its focus on the size, rather than the growth, of immigrant populations. Hopkins (2010) argues that American citizens are surprisingly unaware of their demographic surroundings, and that occupational and residential segregation limit the visibility of immigrants to American citizens. Citing a principal axiom of prospect theory (Kahneman and Tversky 1979), Hopkins reasons that while the average citizen may filter out vast quantities of information streaming in from their environment, significant changes in their environment are more likely to capture attention. Given these considerations, Hopkins concludes that while contemporary levels of ethnic diversity may elude citizens’ attention, significant changes in the level of ethnic diversity are less likely to evade notice. Hopkins (2010) argument for focusing on immigrant growth, rather than population size, is strongly supported by the results from his analysis of time series cross sectional survey data, as well as by other recent research on immigration policy and opinion (Alexseev 2006; Citrin et al. 1990; Newman 2012; Newman et al. 2012; Newman and Johnson 2012).
Beyond the limitations of the group size framework, extant research implicitly assumes that context exerts a uniform effect across all citizens. While the literature has seen the emergence of a new line of contextual studies exploring the conditional effects of ethnic context (Branton and Jones 2005; Hood and Morris 2000; Hopkins 2010; Oliver and Mendelberg 2000; Rocha and Espino 2008), this work restricts its focus to how the effects of local minority populations may be conditional upon other contextual-level or broader environmental factors. What is missing from the opinion research is an attempt to move the literature forward by engaging the issue of whether important differences across citizens shape how they react to factors operative within their ethnic context.
PERSONALITY AND DIFFERENTIAL ADAPTATION
We pursue individual-level heterogeneity in reactions to ethnic change with a practical purpose, namely, with respect to its implications for the politics of immigration. The extent to which individual differences in response matter for American politics is conditional on the degree to which such differences align with politically relevant divisions. More specifically, they will matter if they align with the left-right dimension of American politics, and thus map onto elite-level political conflict. We begin by considering how citizens who identify with the left and right in American politics differ psychologically with respect to their stable traits, and how these personality differences should shape interpretation and responses to ethnic change.
Personality Differences in American Politics
Recent research converges on the proposition that citizens identifying with the left and right in American politics differ with respect to their relative aversion or attraction to epistemic uncertainty and novelty. The idea is that conservatism, in its emphasis on institutional stability, is palliative for individuals who find uncertainty and change aversive, as it lends stability and predictability to one’s social environment. In contrast, liberalism’s emphasis on institutional change and diversity is appealing to citizens who are comfortable with uncertainty, and who seek out novelty and new experiences (Jost et al. 2003; Jost, Federico and Napier 2009). In political science, supportive evidence comes from studies of the “Big Five” personality traits (Carney et al. 2008; Gerber et al. 2010; Mondak 2010; Mondak and Halperin 2008). Such work suggests that, of these five traits, “openness to experience” and “conscientiousness” most reliably distinguish the political right from the left. As Mondak (2010) explains, “the openness and conscientiousness hypotheses are best understood in terms of traditional views in which liberalism corresponds with a willingness to see government tackle new and varied problems, while conservatism implies a more cautious approach in which presumption favors the status quo” (p. 127). As Gerber et al. (2010) explain, openness to experience corresponds with a general attraction to novelty, while conscientiousness is associated with rule and norm-following, and “socially prescribed impulse control’ (p. 115; see also Carney et al. 2008).
This work largely converges with other recent research in psychology proper. Jost et al. (2003) report the results of a meta-analysis of over eighty studies, and find that conservative political orientations are strongly associated with several indicators related to epistemic needs for certainty, such as “intolerance of ambiguity” (Frenkel-Brunswik 1949). Federico and Goren (2009) find that an aversion to uncertainty is associated with conservative self-identifications. Hetherington and Weiler (2009) find that needs for order and certainty constitute a highly influential basis for party identification in American politics since at least 2004.
Overall, this body of work suggests that left and right orientations in contemporary American mass politics can be distinguished in terms of relative attraction or aversion to epistemic uncertainty, and thus a general dislike or preference for novelty and change.
Implications for Responses to Ethnic Change
The psychological divide between the left and right in the American mass public is important as it implies the potential for differential responses to immigration-driven ethnic change. At its most basic level, immigration engenders intercultural contact. Such contact can lead to a process of large-scale cultural change labeled acculturation (Redfield, Linton, and Herskovits 1936; Castro 2003), where the original cultural patterns of either or both groups become permanently altered by the transmission and fusion of culture. When triggered by immigration, the process of acculturation can be characterized by the displacement of the ethno-cultural status quo of the host community and the emergence of a more ethnically and culturally diverse sociocultural landscape.
Psychological research on acculturation focuses on how individuals residing within environments undergoing cultural change adapt to the over-time dislocation and replacement of their habituated ethnic context. According to this literature, adaptation to one’s environment involves both psychological and sociocultural components, with the former pertaining to feelings of belonging to one’s community, social trust, and satisfaction with life (Berry and Sam 1997; LaFromboise, Coleman, and Gerton 1993), and the later pertaining to the ability to effectively interact and communicate with cultural outgroups (i.e., sociocultural competence) (Castro 2003; La Fromboise et al. 1993; Ward and Rana-Deuba 1999). The acculturation literature contends that individuals are susceptible to the experience of “culture shock” (Furnham and Bochner 1986; Oberg 1960) or “acculturative stress” (Berry 1970; Berry 1997) as their habituated environment changes and they (potentially) fail to adapt to heightened levels of ethnic and cultural diversity.
When brought together, we believe that the research on personality and politics and the psychological research on acculturation hold an important implication for the politics of immigration and ethnic change. Namely, these works suggest that the politically relevant personality traits underlying the left-right divide will be relevant for the politics of immigration because they will shape how individuals adapt to residing in environments undergoing acculturation. The reactions of primary theoretical concern revolve around the feelings and perceptions of cultural threat in response to immigration, which should be highly politically consequential given that cultural threat is a pre-potent source of citizens’ immigration policy preferences (Citrin et al. 1997; Espenshade and Calhoun 1993; Ha 2008; Hood and Morris 1997; Sides and Citrin 2007; Sniderman et al. 2004).
The personality and acculturation frameworks suggest openness to, and indeed affinity for, novelty and change, should lead citizens to find immigration-driven sociocultural changes less culturally threatening and should thus engender positive adaptation to ethnic change and heightened diversity. In contrast, dispositions associated with aversion to ambiguity and uncertainty and a corresponding desire for stability and predictability in one’s social environment should lead citizens to find ethnic changes aversive, thus fixating on the displacement of the sociocultural status quo and negative adaptation to ethnic change. Indeed, extant research on ethnic change argues that one principle consequence of the influx of ethnic minorities for white residents, especially those who reside in previously white dominated areas, is that it creates uncertainty about the identity and future of one’s community (Green, Strolovitch, and Wong 1998). In sum, we offer the differential adaptation hypothesis:
(H1) Ethnic change and citizen personality will interact to predict perceived cultural
threat from immigrants such that citizens low in needs for certainty will become
less culturally threatened by immigration as a function of ethnic change, while
those high in such needs will become more culturally threatened.
Furthermore, we argue that differential adaptation by personality should impact policy preferences indirectly through cultural threat perceptions. To be sure, while our theoretical approach departs from theories of racial or power threat by focusing on changes in the ethnic composition of citizens’ environment as the principle feature of ethnic context driving opinion, we take a cue from these theories by focusing on threat as a perceptual intermediary between citizens’ context and their policy preferences. Applying this standing theoretical logic to our theory, we offer the following second prediction concerning policy preferences over immigration:
(H2) Ethnic change induced polarization of perceived cultural threat by personality will
mediate polarization of support for restrictive immigration policies, such that
citizens low in needs for certainty will become less supportive of restrictive policies,
while those high in such needs will become more supportive.
Finally, we consider theoretically and empirically the broader implications of these dynamics, and argue that personality-driven partisan and ideological sorting in contemporary U.S. politics, combined with hypotheses one and two, entails exacerbated partisan and ideological polarization over immigration as a function of continuing ethnic change. If correct, then the influx of immigrant populations into the United States, rather than leading to convergence of opinion over the correct approaches to dealing with this issue, will further divide the public along already existing lines of political conflict. Such polarization incentivizes politicians to move to the extremes to score political points with core constituencies, forcing out pragmatic approaches and solutions, and thus lowering the probability of genuine bipartisanship and compromise across party lines, even as these issues increase in importance as a function of such change. We represent our model and its potential political implications graphically in Figure 1.
To test our hypotheses regarding differential adaptation and polarization we rely upon secondary analysis of nationally representative survey data and analysis of an original national survey experiment. We address these studies in turn.
Figure 1. Connecting Personality Processes to Polarization
Notes: The dashed, two-headed arrow is intended to represent an induced bivariate association between political orientations and immigration policy preferences (i.e. “partisan polarization”), not the residual covariance of these constructs within the context of the structural model.
Our first study relies upon the 2006 United States Citizenship, Involvement, Democracy Survey, conducted for the Center for Democracy and Civil Society at Georgetown University by International Communications Research (Howard, Gibson and Stolle 2005). The overall survey consists of approximately 1,000 interviews collected door-to-door using a cluster sample design intended to represent the adult population residing in occupied residential housing units. Respondents were matched to county-level context data (e.g. change in the Hispanic population) which we describe further below. In keeping with prior opinion research on immigration (e.g. Brader, Valentino, and Suhay 2008; Campbell et al. 2006; Citrin et al. 1990), the present analysis restricts its focus to N=905 non-Hispanic respondents in the survey.
Perceived Cultural Threat. In line with our theory, we expect ethnic change to influence perceptions of the cultural threat posed by immigrants. To measure individual perceptions of cultural threat, this analysis relied upon an item in the CID tapping whether respondents believe that “America’s cultural life is undermined or enriched by people coming to live here from other countries.” This item is comparable to measures of cultural threat in leading opinion research (e.g., Citrin et al. 1997; Sniderman et al. 2004). This item has 11 response options, ranging from 0 (“cultural life undermined”) to 10 (“cultural life enriched”), and was recoded to range from 0 to 1 with higher values indicating greater perceived threat.
Objective change in local immigrant population. We measure the objective change in the immigrant population of the respondent’s environment as the percentage change in the Hispanic population in the respondent’s county of residence.1 Drawing upon data from the U.S. Census Bureau, we obtained the percent Hispanic in each respondent’s county of residence in 1990 and 2000. To obtain the measure of change in this population in each county, we subtracted the percentage of the county population which was Hispanic in 1990 from the percentage in 2000.2 The usage of a 10 year time frame is consistent with extant research analyzing the effects of ethnic change on opinion and behavior (Alexseev 2006; Citrin et al. 1990; Green et al. 1998; Hopkins 2010).
Personality. To measure uncertainty aversion, we rely on two items measuring respondents’ levels of authoritarianism (r=.48). Both items asked respondents how much they agree or disagree with a specific statement (5-points, from “Strongly agree” to “Strongly disagree”). The first statement read, “What young people need most of all is strict discipline by their parents.” The second statement read, “In preparing children for life, it is extremely important that they learn to be obedient.” In recent work, scholars have theorized that authoritarian attitudes and behaviors are derivative of more basic psychological needs for a well-ordered and predictable environment (e.g. Duckitt 2001; Feldman 2003; Hetherington and Weiler 2009; Jost et al. 2003; Stenner 2005). In a meta-analysis of over 80 previous studies, Jost et al. (2003) find strong associations between epistemic needs and authoritarianism (see also Hetherington and Weiler 2009; Van Hiel, Pandelaere and Duriez 2004).In addition to their similarity to recent operationalizations of the construct (see, e.g., Hetherington and Suhay 2011; Hetherington and Weiler 2009), these items should distinguish respondents on the basis of epistemic needs. Each item considers potential solutions to dealing with a dangerous or uncertain world, more specifically, by adhering to established norms, rules and institutions, and respecting legitimate authorities. As Duckitt (2001) argues, “A view of the world as dangerous, unpredictable, and threatening…would activate the motivational goal of social control and security. This motivational goal would be expressed in the collectivist sociocultural values of conformity and traditionalism and in…authoritarian social attitudes” (p. 50).
Controls. We control for three additional indicators measured at the county-level. To account for the predictions of power and economic threat hypotheses, we control for levels of immigrants as the percent Hispanic of the county population in 2000, as well as data from the U.S. Department of Labor’s Bureau of Labor Statistics to obtain the unemployment rate within each county in 2005. Finally, we control for the political culture of citizens’ counties of residence with a variable measuring the proportion of the vote won by George W. Bush in each county in the 2004 Presidential Election. This latter variable has been identified as a relevant factor shaping opinion on policies concerning racial and ethnic minorities (Campbell et al. 2006). At the individual level, we control for several additional factors, including age, gender, black self-identification, educational attainment, income, nationalism, personal economic retrospections, whether the respondent has friends and/or family members who are immigrants, birth in the U.S., employment status, partisanship (higher values = more Republican), and ideology (higher values = more conservative). All variables were recoded to range from zero to one prior to analysis.
We estimate the following regression for perceived cultural threat (Y) via restricted maximum likelihood:
In this model both the intercept of the individual-level equation and the marginal effect of authoritarianism are modeled as a function of county-level predictors and normally distributed, random disturbances. With respect to the marginal effect of authoritarianism, we can think of this specification in terms of two cross-level interactions, one with the change in the county-level Hispanic population from 1990 to 2000, and one with the percent Hispanic at the county level in 2000. We include the latter as a control to ensure that the dynamic we observe is due to change per se, and not simply the overall percentage of Hispanics in a given county. We expect a negative and significant coefficient on the Hispanic change variable (, indicating that at low levels of authoritarianism change in the ethnic composition of one’s social environment decreases perceptions of cultural threat. Conversely, we expect a large and positive interaction term between change and authoritarianism () such that , indicating that at high levels of authoritarianism, ethnic change entails an increase in perceptions of cultural threat. These dynamics are conceptually equivalent to polarization of threat by personality as a function of ethnic change.
The estimates for this model are shown in the first column of Panel B of Figure 2 with their associated 95% confidence bounds. They show strong support for theoretical expectations. With respect to significance tests of our key directional hypotheses, we utilize one-tailed tests. First, the coefficient for change in the county-level Hispanic population from 1990-2000 is in the expected negative direction and statistically significant (=-.28, p<.05). In addition, as expected, the coefficient for the interaction of change with authoritarianism is positive, larger in absolute value than the coefficient for change, and statistically significant (=.49, p<.05), indicating a reversal in direction of the effect of Hispanic change from negative to positive as authoritarianism increases from low to high.
To better interpret the substance of these results, we generated predicted values of perceived cultural threat as a function of Hispanic change and authoritarianism in Panel A of Figure 2. The black line represents the predicted values of threat, moving from the 5th to the 95th percentile of Hispanic change, for citizens at the 5th percentile of authoritarianism. The grey line represents the predicted values for citizens at the 95th percentile of authoritarianism. This graph illustrates nicely the theorized conditional relationship between ethnic change and perceived threat. For citizens low in authoritarianism, and thus open to change, uncertainty and novelty, changes in the ethnic composition of one’s county entail a decrease in the perception that immigrants pose a cultural threat to the United States. Conversely, at high levels of authoritarianism, and thus for citizens averse to uncertainty and novelty, changes in ethnic composition entail an increase in the perception that immigrants pose a cultural threat.
We will additionally comment briefly on the estimates for the model’s control variables. With respect to Hispanic population levels in 2000, we find no evidence of an interaction, but some suggestion that levels may have an influence on the attitudes of low authoritarians. At the lowest levels of authoritarianism, increases in the percentage Hispanic at the county-level in 2000 are associated with decreased perceptions of cultural threat (p<.10, two-tailed). For every additional 10% of the county-level population that is Hispanic, we predict a decline in perceived cultural threat among low authoritarians of about four percentage points. The political context, operationalized as the percent voting for Bush in 2004, had no influence on threat, while the unemployment rate had a positive effect, but did not attain conventional levels of statistical significance. With respect to individual-level controls, education matters a great deal. Moving from the lowest to the highest levels of educational attainment entails a decrease in perceived cultural threat of about 17 percentage points. Males were about 3 points more threatened than females on average, Black citizens were about 5 points more threatened than other groups, and conservatives were about 7 points more threatened. Finally, citizens who perceive their own financial situation to be tenuous were more likely to feel that immigrants threaten American culture. As seen in Figure 1, however, the strongest determinant of perceptions of threat are real changes in the ethnic composition of the local environment, but in distinct ways conditional on the personality traits of citizens themselves.
The Potential for Bias Due to Selection
We turn now to a brief consideration of the well-known problem of selection bias, or in other words, the possibility that respondents select into geographical areas on the basis of the presumed causal variable as a function of their attitudes or political dispositions. In our case, this would entail citizens selecting into counties whose demographic profile with respect to ethnicity fits with their dispositional levels of perceived cultural threat. While we address the issue of causality rigorously through a survey experiment (described further below), we believe there are reasons to reject the selection bias story in the present case.
While we do not deny that citizens may select into environments on the basis of attitudes and traits which may be relevant to the immigration attitudes examined herein, we believe that our hypothesized dynamic of interest, namely, the interactive effects of ethnic change with personality, should be less subject to claims of selection-induced spuriousness. To see why, consider the logic of the selection bias argument. This argument would claim that citizens select into geographic regions as an interactive function of both the region’s level of expected and/or ongoing ethnic change and the citizen’s own traits and attitudes. Thus, citizens who are most negative toward immigrants would select into relatively homogenous regions, and those most positive would select into regions with greater extant or expected diversity. But our hypothesis is not a simple direct effect of change on attitudes, but rather a conditional one. Specifically, we expect the impact of ethnic change to be exactly opposite across personality orientations, with those low in needs for epistemic certainty becoming more positive and vice versa.
While the selection bias argument could potentially hold water for the former hypothesis, it simply fails for the latter. Specifically, for the selection argument to work, one would have to posit that those with personality traits and attitudes most averse to such social uncertainty would be most likely to move to regions undergoing substantial change. This makes little sense. Thus, while we cannot rule out the possibility of these selection biases in one of the two cases, confirmation of our hypothesis for citizens with high needs for certainty in their social and cultural environs speaks strongly in favor of our causal model.