Between Demographic Optimism and Pessimism?

Abstract The U.S. Census Bureau projects that by 2060, Latinx, African Americans, Asians, and other “minority” groups will together comprise the majority of the country’s population. Past research has found that non-Hispanic Whites, hereafter Whites, find such projections disquieting or threatening. Yet, recent surveys reveal that when given more than binary good/bad choices, most Whites opt for the middle-point response that this development will be “neither good nor bad for the country.” How can we account for this seemingly ambiguous evaluation of projected ethnoracial demographic futures? Using eight waves of nationally representative U.S. survey data collected between 2015 and 2018, this article begins to unpack the “neither” response among Whites, exploring what it might mean and what factors are associated with it, relative to seemingly optimistic and pessimistic stances. Multinomial Logistic Regression analyses and additional descriptive analyses indicate that “neither good nor bad” in this context is a substantive response: White “Neithers” are socio-demographically and attitudinally distinct from their counterparts. Our study demonstrates the value of moving beyond an exclusive focus on expressions of demographic threat and pessimism. Moreover, it invites further investigation into factors that inform and shape how Whites and other ethnoracial populations in the U.S. understand and assess projected population shifts.


Introduction
The United States has become increasingly diverse over the last sixty years, predominantly due to population growth among native-born and immigrant Latinos and Asian Americans and the rise of children with racially mixed parentage (e.g., Frey 2018; Morning and Saperstein, 2018;Pew Research Center 2015a). The U.S. Census Bureau projects that by 2060 non-Hispanic Whites, hereafter Whites, will comprise less than half of the U.S. population. 1 Although they are expected to remain the single largest population for the foreseeable future (Colby and Ortman, 2015), White demographic stagnation and decline has become as of late the subject of extensive media coverage, political discourse, and academic debate (e.g., Chavez 2008;Frey 2018;McConnell 2019;Rodríguez-Muñiz 2021). Within this context, how do U.S. Whites perceive the projected growth of other ethnoracial populations?
Compared with other groups, Whites are generally less optimistic and more pessimistic about demographic change (Wong 2018). From experiments to ethnographic observations, empirical evidence points to powerful strains of demographic fear and anxiety among U.S. Whites and their counterparts in Western Europe (e.g., Abascal 2020; Craig and Richeson et al., 2014;Danbold and Huo, 2015;Hochschild 2016;Otten et al., 2012). This reaction is consequential, contributing to political conservatism, sentiments of victimhood, and decreased support for social welfare programs (e.g., Jardina 2019; Jones and Kiley, 2016;Major et al., 2018;Mutz 2018).
Yet, demographic pessimism-the sense that ethnoracial population trends represent a disconcerting or threatening development-does not tell the whole story, even as it accounts for a significant part. Evidence suggests that some Whites, albeit currently a minority, consider ethnoracial population trends a positive and welcomed development (e.g., Myers and Levy, 2018). Still further, another segment of the White population seems to fall somewhere between demographic pessimism and optimism. A recent spate of surveys has found that-when given more than binary choices-a plurality, if not a majority, of Whites claim that projected ethnoracial demographic trends are "neither good nor bad" for the country (see Arizona State University 2018; Budiman 2020). Some non-White respondents also claim a similar "neither" stance, although at generally lower rates (Budiman 2020). Nevertheless, the longstanding political cultivation of White demographobia (e.g., Alim 2016; Rodríguez-Muñiz 2021) and accompanying rates of pessimism raise unique questions about the White population.
What are we to make of the ambiguous, seemingly middle-of-the-road, "neither good nor bad" response among Whites? This response to questions about future ethnoracial change represents a kind of "neither/nor" response. Although "neither/nor" is a common type of "center" or "middle point" response in social science survey research (e.g., Alamillo 2019; Lemi and Kposowa, 2017;Telles and Torche, 2019), it is the subject and source of ongoing debate among survey researchers and methodologists. 2 Methodological issues aside, there is also a thorny but vital question of interpretation. Like other midpoint responses, it is not at all obvious how to interpret claims that ethnoracial population trends will be "neither good nor bad" for the country. This is compounded by the fact that no systematic analysis has been done on this response. Consequently, scholars have limited knowledge about the views of the potentially largest segment of the White population. We do not yet know whether Whites who select this response are distinct from those that express optimism or pessimism about these shifts. It is premature to surmise what "neither" means, such as, for instance, that it represents demographic pessimism in disguise. Rather than rush to judgement, we instead subject the response to sociological analysis. As an initial but necessary step, we focus our attention on "White Neithers," by which we mean those that claim that increased future ethnoracial diversification in the country is "neither good nor bad." Our study-the first of its kind-draws on eight waves of the nationally-representative Pew Research Center's American Trends Panel (ATP) collected between 2016 and 2018. The first wave of data examined in the analyses was collected before the 2016 election and the last of these waves was collected during the second year of the Trump administration, an administration that embraced narratives of White demographic loss. Equipped with these datasets, we conducted a series of descriptive and multinomial logistic regression analyses to determine whether White Neithers were demographically, ideologically, and attitudinally distinct from their more optimistic and pessimistic counterparts. Using the ATP's extensive set of individual-level variables and questions relevant to this topic, we controlled for a comprehensive set of factors known to shape public opinion and racial attitudes to identify variation in the characteristics of White Neithers, Optimists, and Pessimists. To begin to explore what the "neither good nor bad" response might mean, we also examine other racial attitudes that White Neithers express relative to other White respondents.
Going beyond the almost singular attention on White demographic threat and pessimism, our findings suggest that Whites who choose the "neither good nor bad" response differ from their optimistic and pessimistic contemporaries in sociologically significant ways. The results suggest to us that, in this case, "neither good nor bad" is a substantive (i.e., meaningful) rather than nonsubstantive (i.e., spurious) response (Truebner 2021). Although further analysis and data are needed to fully theorize what this response means, this article cautions against its outright dismissal and invites greater attention to White attitudinal heterogeneity about projected U.S. ethnoracial population change.

Demographic Attitudes: From Explanans to Explanandum
The social and political effects of ethnoracial demographic shifts have long concerned sociologists (e.g., Alba et al., 2005;Maggio 2021;Schlueter and Scheepers, 2010). One prominent approach is group threat theory. While visible as early as Gordon Allport (1954) and others, it was Hubert Blalock (1967) that popularized the theory. Drawing primarily on empirical research on the U.S. south, Blalock examined the relationship between "minority" demographics and perceived threat among White or "majority" populations. His complex theoretical model challenged social psychological explanations of prejudice, famously positing that relative increases to minority population size and growth rate triggered "group threat," which, in turn, increased the "motivation to discriminate" (Blalock 1967, p. 144). Research has since found qualified empirical support for this conclusion (e.g., Fossett and Kiecolt, 1989;Hall and Krysan, 2017;Quillian 1996;Taylor 1998).
Where Blalock and other scholars focused on the attitudinal and behavioral impacts of actual demographic change, a recent wave of mostly experimental research-led primarily by political scientists and psychologists-has considered anticipated changes. This line of inquiry is significant for, as sociologist Ann Mische (2014) reminds, perceptions of the future-demographic or otherwise-can have concrete and measurable impacts on the present (see also Rodríguez-Muñiz 2021). Experimental research has shown that exposure to reports of impending racial diversification has led Whites to express more politically conservative viewpoints and more exclusionary views about immigration and other topics (e.g., Craig and Richeson et al., 2014, 2018a, 2018bDanbold and Huo, 2015;Jardina 2019;Major et al., 2018;Outten et al., 2012Outten et al., , 2018Skinner and Cheadle, 2016). One experiment with White respondents, for instance, found that reading a newspaper article claiming that "racial minorities" were projected to comprise more than half of the U.S. population by 2060 (and that Whites would comprise less than half) stimulated more threat, anger, and fear of those populations than those that read an article stating that the U.S. proportion White would remain the same in 2060 as in 2010 (Outten et al., 2018). A study by Craig and Richeson (2018a) found that information about a coming "majorityminority" future increased Whites' worries about "anti-White discrimination." Another study, which used a simulated game, discovered that compared to African Americans, White participants who were told that Latinos were responsible for the nation's changing diversity were more prone to exhibit pro-White discriminatory behavior (Abascal 2015).
This article shares with the above experimental research an emphasis on anticipated or projected demographic futures. However, it departs in one key respect from both that scholarship and works on group threat, more generally. We shift from the more traditional treatment of demographic stances and sentiments as explanans and move towards their treatment as explanandum. As Lawrence Bobo and Vincent L. Hutchings (1996) once noted, perceptions-including demographic perceptions, we would add-"have meaningful social and psychological underpinnings" (p. 968). Thus, rather than focus on how exposure to projected demographic scenarios affects, for instance, racial identities and policy preferences, we consider what factors influence and shape how U.S. Whites variously perceive said projections in the first place.

Beyond Demographic Threat and Pessimism
To date, studies on White perceptions of ethnoracial population change have emphasized stances and sentiments of anxiety, fear, and threat. The empirical record justifies this attention. As the previous section attests, quantitative and experimental researchers have discovered strong contemporary evidence of demographic pessimism or demographobia among this segment (e.g., Abascal 2020; Bai and Federico, 2021;Baker et al., 2020;Major et al., 2018). Qualitative and ethnographic researchers have, as well (e.g., Gest 2016;Hochschild 2016;Lacayo 2017;Pied 2019). But this is an incomplete picture.
The extant scholarship indicates that demographic pessimism is neither uniform nor universal. Numerous factors have been shown to mediate and modulate White attitudes and feelings about ethnoracial demographics. Some studies have considered demographic factors (e.g., Alba et al., 2005;Richeson, 2018a, 2018b;Hall and Krysan, 2017;Kaufmann 2014;Laurence and Kim, 2021). Other scholars, such as Ashley Jardina (2019), have found that Whites with lower levels of racial identification report less anger after being provided information that Whites would be a minority by 2042 than their more strongly White-identifying contemporaries (see also Major et al., 2018). As with other views, and race and immigration attitudes in particular, partisan affiliation and political ideology seem to have a strong influence on perceptions and responses to demographic change (Abascal 2020;Brown et al., 2021). 3 Using a series of experimental studies, Dowell Myers and Morris Levy (2018; demonstrate that narrative framing can also shape emotional responses to projected population trends. In identifying factors that may intensify or minimize White demographic pessimism, the above scholarship points, by implication, to greater attitudinal heterogeneity than is often assumed. Although it has not been a thematic or theoretical priority, some works reveal stances beyond threat and pessimism. For example, White respondents have at times expressed hopefulness and enthusiasm about projected trends of increasing ethnoracial diversity (e.g., Budiman 2020; Jardina 2019; Levy and Meyers, 2021;Myers and Levy, 2018). This is not all. Growing evidence suggests that most Whites are neither solidly optimistic nor pessimistic about a majority-minority demographic future (Arizona State University 2018; Budiman 2020). But none of these potential alternatives to pessimism has received focused attention. Seeking to address this limitation, we focus on one of these potential sentiments, namely, the claim-captured in several recent surveys-that ethnoracial demographic trends are "neither good nor bad" for the country.

Hypotheses
In what follows, we conduct the first survey-based examination to determine the substantiveness of the "neither good nor bad" claim with regard to this topic. Using multivariate and descriptive analyses, our approach focuses on respondents that chose the "Neither" response category, compared with "Optimists," those that reported "good for the country," and "Pessimists," those who selected "bad for the country." We test the following two hypotheses: Hypothesis 1: White Neithers have different sociodemographic characteristics than White Optimists and White Pessimists.
Hypothesis 2: White Neithers express social attitudes about immigration and race relations that are distinct from White Optimists and White Pessimists.
The results from these tests will help illuminate what the "neither good nor bad" response means among White respondents. For instance, if Neithers are socio-demographically different from Optimists and Pessimists, that provides some evidence that the response is substantive. By substantive, we mean responses that reflect an individual's actual, even if conditional or momentary, position on a given subject. However, if Neithers' profiles are similar to Optimists and Pessimists that could suggest that the "neither" response is nonsubstantive. 4 Likewise, if White Neithers hold distinct racial and immigration attitudes compared to the other two groups, we have further support for the substantive interpretation. In contrast, attitudinal consensus across respondents would instead weaken this conclusion. The strongest empirical support for response substantiveness requires confirmation of both our hypotheses. Indeed, results that demonstrate that White Neithers have both disparate sociodemographic profiles and attitudinal positions as compared to their counterparts indicate to us that "neither good nor bad" is a substantive and sociologically meaningful response. In such a scenario, we believe this response would deserve further investigation and analysis.

Data and Methods
Our examination of White Neithers, as compared to Optimists and Pessimists, employs and exploits a total of eight waves of Pew Research Center's American Trends Panel (ATP) data, collected between 2015 and 2018. ATP is a probability-based national survey of non-institutionalized U.S. adults over eighteen years old initially recruited from two national landline and cellphone Pew surveys in 2014 and 2015 (Abt SRBI et al., 2016). ATP respondents complete surveys in English or Spanish, predominantly online, with non-internet users completing computer-assisted telephone interviews or mail surveys. Self-administered surveys, especially online surveys, are more widespread, have increased respondent participation, and exhibit less social desirability bias than live telephone interviews (Keeter 2015;. Although ATP waves cover a varying set of substantive topics, each wave includes a core set of demographic variables about respondents. As some ATP respondents complete earlier waves and then drop out and new respondents join the panel in later waves, each wave includes a stable unique identifier to link panel respondents across waves. The Appendix provides more information about the ATP waves used in the study, ancillary analyses, and the data sources used in the Figures.
We test Hypotheses 1 and 2 using a multivariate regression framework with a merged data set of three ATP waves conducted between 2015-2016. The sample used in the multivariate analyses is limited to those who self-identified as non-Hispanic White in Wave 10 and who also completed Waves 15 and 16. The dependent variable is based on the following Wave 16 question: According to the U.S. Census Bureau, in the next 25 to 30 years African Americans, Latinos, and people of Asian descent will make up a majority of the population. In general, do you think that this is… In addition to "Don't know," the question provides three exclusive responses: "Good for the country," "Bad for the country," or "Neither good nor bad for the country" (which we code as Optimism, Pessimism and Neither, respectively). Less than 1 percent of White respondents in the analytic sample gave a "don't know" response or refused to answer the question. 5 Table 1 describes this and other variables in more detail and Table 2 provides descriptives of the analytic sample used in the regression analyses. As Table 2 shows, 69.8% of the analytic sample of White panel respondents reported that a future in which African Americans, Latinos, and Asians are the majority of the U.S. population is "neither good nor bad for the country." About 9% and 21% reported optimism or pessimism about this projection, respectively.
The multivariate analyses incorporate independent variables to test Hypothesis 1, that there are differences in the sociodemographic characteristics between Whites reporting Neither, Optimism, or Pessimism regarding future ethnoracial diversification. Independent variables from Wave 10 include respondent age, educational attainment, gender, married/cohabitating or not, total family income, political partisanship, has health insurance, connections to immigrants via generational status, having friends or relatives that are immigrants, and the respondent's U.S. census region of residence. 6 Such indicators are routinely included in quantitative analyses of racial and immigration attitudes (e.g., Abrajano and Hajnal, 2017;Bobo et al., 2012;Forman and Lewis, 2015;Krysan and Couper, 2003;Quillian 1996;Schildkraut and Marotta, 2018;Schuman et al., 1997).
We also consider an additional set of factors potentially associated with an optimistic, pessimistic, and "neither" response. These include variables from Wave 10 on cognition, knowledge, and media consumption: the level of thought given to presidential candidates in the upcoming 2016 election, the overall level of contemporary news knowledge (also see Pew Research Center 2015b), and whether a respondent prefers complex problems requiring a lot of thought rather than simple problems. 7 Along with substantive reasons, the inclusion of these variables may reduce specification error.
The multivariate analyses also are used to test Hypothesis 2, that Whites who evaluate future ethnoracial demographic change as "neither good nor bad" express different social attitudes, in this case about immigration and race, than Optimists and Pessimists. 8 The first attitude examined is a binary indicator that immigration to the U.S. should be decreased, a common immigration attitude (e.g., Alba et al., 2005;Jardina 2019;Schildkraut and Marotta, 2018). Aligning with scholarship on immigration as a perceived threat (e.g., Herda 2010; Jardina 2019), another indicator captures whether respondents believe immigrants are making things worse. Based on a mean scale, this indicator taps into perceptions about immigrants' linguistic and cultural assimilation (e.g., Berg 2013; Paxton and Mughan, 2006). Views about immigrants from different world regions also was included, as it could reveal how views about racialized immigrant groups shape views about population trends.
Past research also suggests that Whites' out-group and in-group racial attitudes are likely associated with their evaluations of future ethnoracial change and may differ between Neithers and other groups. Scholars routinely use feeling thermometers about different race groups as measures of in-group or out-group identity or attachment (e.g., Abrajano and Hajnal, 2017;Jardina 2019;Kinder and Kam, 2009;Outten et al., 2012;Schildkraut and Marotta, 2018;Valentino et al., 2013). 9 Thermometers thus provide a means to capture Mean scale based on 4 items: Immigrants coming to the United States make American Society worse in the long run; Immigrants in the U.S. are making things worse in crime, the economy, and/or social and moral values. For each item -1 = making things better, 0, not having much effect, 1=making things worse. Mean values for scale variable ranges from -1 to 1. Internal consistency for this scale is α=0.82.

Recent immigrants don't learn English
In your opinion, do most recent immigrants learn English within a reasonable amount of time, or don't they? Yes, they do OR No, they don't. 1=No, they don't learn English within a reasonable amount of time. 0=Yes they do. Immigrants don't want to Assimilate Immigrants in our country today generally want to adopt American customs and way of life OR Immigrants in our country today (Continued) "racialized emotions" (Bonilla-Silva 2019). For the respondents in our study, the White thermometer overlaps with their strength and attachment to their own racial identity and is a reasonable alternative measure of their group identity (Jardina 2019). In contrast, thermometers about African Americans, Latinos, and Asian Americans tap into Whites' out-group attitudes.

Analytic Strategy
As the dependent variable has three possible categorical responses, multinomial logistic regression models are appropriate (Hoffman 2004). Multinomial logistic regressions were estimated to calculate relative risk ratios (RRRs) indicating the risk, or probability, that a respondent with a particular characteristic is more likely to select one response category relative to the reference category. Identical models were specified with the same baseline comparison group of the dependent variable, Neithers, compared with the other two groups: Neither versus Good followed by Neither versus Bad. The RRRs in multivariate results for these contrasts represent the likelihood of selecting "neither good nor bad" versus expressing optimism or pessimism, which along with statistical significance, indicates the characteristics that distinguished Neithers from the other two groups. 10 Two specifications were estimated. The preliminary specification with sociodemographic characteristics was used to test Hypothesis 1, and a second specification that added immigration and racial attitudes was used to test Hypothesis 2. For each contrast, diagnostics indicated mean Variance Inflation Factors for each model are below 1.74, suggesting that multicollinearity was not a problem affecting the results. Stata 17's multiple imputation for chained equations algorithm was used to address missing data for independent and dependent variables in the analytic models with any missing data (m=25 imputed data sets). 11 After imputation, the merged data yielded 2,076 White respondents, a larger sample size than other related survey-based analyses addressing similar topics (e.g., Jardina 2019; Schildkraut and Marotta, 2018;Valentino et al., 2013). As this specific analysis followed Wave 10 respondents over time, the multivariate analyses were weighted using weights provided in the Wave 10 ATP data created to address differential probabilities of selection into the ATP and non-response (Abt SRBI et al., 2015). Feeling thermometers for Blacks, Whites, Hispanics and Asians based on the statement, "We'd like to get your feelings toward a number of groups in the U.S. on a "feeling thermometer." A rating of zero degrees means you feel as cold and negative as possible. A rating of 100 degrees means you feel as warm and positive as possible. You would rate the group at 50 degrees if you don't feel particularly positive or negative toward the group." To impute missing data on the thermometer variables successfully, the possible responses were reduced from 100 possible values into ten ordinal categories (0-10), such that an original thermometer value of 0=0, 1-9=1, 10-19=2…90-100=10.
Note: All independent and control variables are from Wave 10 except for the racial thermometers, which are from Wave 15. The dependent variable is from Wave 16. In addition to the above multivariate tests, we used a series of descriptive analyses based on a different set of ATP waves to further test Hypothesis 2, i.e., whether White Neithers hold different attitudes than Optimists and Pessimists. The first of these analyses is presented in a data visualization that compares the attitudes that White Optimists, Pessimists, and Neithers categorized using Wave 16 data (and the outcome of the above multivariate analyses) expressed in four subsequent survey waves (Waves 18, 20, 22, and24.5, collected between June 2016 andMarch 2017). 12 We focus on attitudes that are commonly studied in survey-based racial attitudes scholarship (e.g., Bobo and Kluegel, 1997;Jardina 2019): general racial attitudes, awareness of structural racism and discrimination, perceived level of discrimination that different racial groups experience, and recognition of White privilege. For simplicity, each of these attitudes was coded as a binary variable in the direction of less racially progressive views equaling 1 (described in the Appendix). Tests of statistical significance identified whether the racial attitudes differed among Neithers, Optimists, and Pessimists at less than the .05 level of significance. Finally, we turn to more recent data, ATP Wave 41 (December 2018) that asked a similar question soliciting views about future ethnoracial demographic change and two new questions about the likely effects of these expected ethnoracial changes. 13 These cross-sectional Wave 41 data were used to further explore whether White Neithers, Optimists, and Pessimists hold distinct attitudinal profiles.

Are White Neithers Sociodemographically Distinct from White Optimists and Pessimists?
Our first hypothesis tested whether there are sociodemographic and ideological differences among White Neithers, Optimists, and Pessimists. 14 The results-presented in Table 3offer insight into the substantiveness of White respondents' claims that projected ethnoracial trends were "neither good nor bad" for the United States. Multinomial regression analyses reveal significant differences in some sociodemographic characteristics between Neithers and their counterparts. For example, compared to Optimists (individuals that selected "good for the country"), Neithers were older: nearly 30% less likely to be eighteen to twenty-nine years old and about 40% less likely to be thirty to forty-nine years old than over sixty-five (RRRs=0.310 and 0.405, p>.01, Model 2, Table 3). In addition, Neithers were half as likely as Optimists to live in the West than the South (RRR for the West=0.551, p>.05, Model 2). Partisan ideology in Wave 10 also strongly shaped a Neither versus an Optimistic response in Wave 16. Indeed, Whites identifying as Republican were about 2.5 times more likely than non-Republicans to report that future ethnoracial diversification was "Neither good nor bad" than "Good for the country" (RRR=2.460, p>.05, Model 2). Neithers and Optimists were, however, similar in their level of education, controlling for other variables.
The contrast between Neithers and Pessimists (individuals that responded that the future ethnoracial growth would be "bad for the country") in the baseline model indicate that Neithers were more likely to be younger (eighteen to twenty-nine; fifty to sixty-four) than over sixty-five, more likely to be college graduates than high school graduates, and less likely to be Republican (Model 3, Table 3). For instance, White college graduates were nearly two times as likely to report a "Neither" response than pessimism (RRR=1.921, p> .001, Model 3). This specification also indicates that, relative to Pessimists, Neithers were more likely to prefer complicated problems to simpler ones and had more news knowledge (RRRs of 1.208 and 1.086, p>.05 or .01, Model 3).  Source: ATP Waves 10, 15, and 16, linked data, limited to Non-Hispanic White respondents present in all three waves. Note: All specifications also control for respondent gender, married/cohabitating, income, has health insurance, second generation in the U.S. or more, has immigrant friends or relatives, thought about the 2016 campaign. The second specification also includes indicators regarding perceptions regarding whether immigrants want to assimilate and whether recent immigrants learn English. †p < .10. *p < .05. **p < .01. ***p < .001 Some sociodemographic differences between Neithers and Pessimists remained significant once immigration and racial attitudes were added to the specification (Model 4). For example, age continued to differentiate Neithers from Pessimists, net of all variables (e.g., RRR of 1.565 for age fifty to sixty-four, Model 4, Table 3). However, the significant partisanship difference between Neithers and Pessimists in the baseline specification disappeared once immigration and racial attitudes were added to the analyses (Model 4). 15 Prior research has found strong associations between identifying Republican and holding negative views about immigration among Whites (e.g., Abrajano and Hajnal, 2017;Jardina 2019). This could explain that result. Net of all variables, the inclusion of variables on race and immigration also rendered baseline educational differences between Neithers and Pessimists to insignificance in the complete specification (Model 4).

Are White Neithers Attitudinally Distinct from White Optimists and Pessimists?
Our second hypothesis shifted from differences in sociodemographic characteristics between Neithers and others to investigating intra-group variation in social attitudes. Specifically, it tested whether White Neithers hold race and immigration attitudes distinct from their optimistic and pessimistic peers. As previously noted, we conducted both multinominal regression and descriptive analyses to test this hypothesis.
Beginning with the multinominal regression analyses, the results-as seen in Table 3largely confirmed our expectations. White Neithers and Optimists, for example, differed on general immigration attitudes, controlling for other variables. Specifically, respondents who wanted decreased levels of immigration were two times as likely to report that the future ethnoracial shifts are "Neither Good nor Bad for the country" than "Good for the country" (RRR=2.027, p>.10, Model 2). 16 Similarly, Whites who reported that immigrants make things worse in the country were 2.4 times as likely to perceive that increasing diversification is "neither good nor bad" instead of expressing optimism (RRR=2.432, p>.05, Model 2). Importantly, the full model indicated that Neithers and Optimists held similar views about the national impacts of different immigrant groups, such as Latin American immigrants, and have similar racial thermometer scores (Model 2).
White Neithers and Pessimists varied on both immigration and racial attitudes, net of other variables (Model 4). Specifically, those who earlier claimed that immigrants make things worse were less likely to select the "Neither" response than to express pessimism about future ethnoracial diversification (RRR=.279; p>.001, Model 4, Table 3). This pattern also held for those who expressed the belief that immigrants from Latin America and Africa have had negative impacts on the country. For example, Whites who perceived Latin American immigrants in this manner were less than half as likely to report "Neither" over pessimism compared to those who reported that these immigrants have had positive or neither positive or negative impacts on the country (RRR= 0.476 respectively, Model 4). 17 Moreover, White respondents who expressed warmer feelings about African Americans (higher Black thermometer values) and colder feelings about their own group (lower White thermometer values) were significantly more likely to be Neithers than Pessimists, controlling for the full set of variables. 18 Indeed, with each one-unit increase in Black thermometer scores, respondents were nearly 40% more likely to select "neither good nor bad" than "bad." In contrast, each one-unit increase in White thermometer scores was linked with an approximately 25% decrease in the probability of selecting Neither rather than Bad for the country (RRRs of 1.395 and 0.754, respectively, Model 4, Table 3). Notably, thermometer values for Latinos were not independently linked with selecting Neither rather than optimism or pessimism, net of views about African Americans, Whites, and the social impacts of Latin American immigrants (Models 2 and 4, Table 3).
Additional descriptive analyses complement and extend the multinominal regression results to further examine whether White Neithers are attitudinally alike or different from those expressing demographic optimism or pessimism. In these next analyses, we investigated how panel respondents who reported that projected future ethnoracial diversification was "neither," "good," or "bad" in Wave 16 (April-May 2016) responded to a series of questions on racial and demographic attitudes asked in later waves. Figure 1 presents a data visualization using data from Wave 16 and Waves 18-24.5 (collected between June-July 2016 through February-March 2017). 19 Figure 1 reveals a consistent pattern: Whites who said "Neither" in Wave 16 later expressed views that were consistently in between the more progressive racial attitudes of White Optimists and the comparatively more regressive attitudes of White Pessimists. Mean group comparisons further indicate that differences among the three segments were statistically significant at the .05 level on nearly all racial attitudes. For example, about 25% of Neithers later reported that they did not consider racism to be a major problem, compared to 8% of Optimists and 37% of Pessimists. Similarly, nearly half of all Neithers (45%) reported that too much attention is paid to race and racial issues, compared to 12% of Optimists and 71% of Pessimists. White Neithers also sat between the other two sets of respondents on whether the murders of African Americans at the hands of the police were isolated incidents rather than a systemic problem and whether White people have benefits and advantages not afforded to Black people (Figure 1, differences significant at the .05 level). The one exception was attitudes about racial intermarriage as a "bad thing for society." Nearly all Optimists and Neithers rejected that statement (only 1.2% and 2.5% agreed), while a much larger proportion of Pessimists agreed with that view (23.7%).

Fig. 1. Racial Attitudes for White Optimists, Neithers, and Pessimists
Finally, we exploited a more recent ATP wave to examine why respondents might report being "neither" as opposed to "good" and "bad." Wave 41 not only asked respondents about their views regarding projections that a "majority of the population will be made up of Blacks, Asians, Hispanics, and other racial minorities" but also asked two follow-up questions about the consequences of these shifts on racial conflict and American customs and values. As Figures 2 and 3 show, Whites who said that projected changes were "neither good nor bad" in Wave 41 report a distinct pattern of response about the impacts of future ethnoracial demographic shifts relative to Optimists and Pessimists. 20 For instance, White Neithers were divided between whether these future ethnoracial dynamics eventually will lead to "more conflicts between racial and ethnic groups" (42.8%) and "not much impact" (33.7%, Figure 2). They were the most likely of the three segments to expect "not much" impact on racial and ethnic conflict. Turning to effects on "American customs and values," White Neithers were significantly more likely to report "not much impact" (47.9%) than their optimistic and pessimistic contemporaries (27.4% and 4.0%, respectively, Figure 3, p>0.001). These results suggest that Neithers report more mixed views about the likely effects of future ethnoracial population growth and perceive them to be less impactful than their more uniformly optimistic or pessimistic counterparts. 21

Discussion
Drawing on eight waves of nationally representative panel survey data collected between 2016 and 2018, this article investigated the substantiveness of the most common survey response that U.S. Whites give when given the option: a projected ethnoracial majority of comprised of non-Whites is "neither good nor bad for the country." Multinomial regression and descriptive analyses were used to test two hypotheses that explore what a "neither good nor bad" response might mean, who these respondents are, and what other social attitudes and demographic traits they might hold. The first hypothesis posited that White Neithers had distinct sociodemographic characteristics from those that reported explicitly pessimistic and optimistic sentiments. The second asserted that White Neithers and their optimistic and pessimistic counterparts were attitudinally dissimilar on issues of immigration and race. Results confirm both hypotheses. Our results thus offer support for the interpretation that, overall, "neither good nor bad" is a substantive rather than spurious response to this question among Whites.
Our findings indicate that demographic threat and pessimism-the overwhelming focus of past scholarship-is not the whole story. Despite decades of alarmist discourse about ethnoracial population growth and change (e.g., Chavez 2008Chavez , 2021McConnell 2011McConnell , 2019Rodríguez-Muñiz 2021), a significant portion of Whites in recent surveys appear not to be convinced that ethnoracial demographic change will be "bad" for the country. White Neithers and White Optimists, the smallest segment in the study, offer strong evidence of White attitudinal heterogeneity about projected ethnoracial futures. Our regression analyses uncovered some of the factors that distinguish between Neithers and their contemporaries, such as political partisanship, their embrace of "White racial identity" (Jardina 2019), and broader social attitudes. For instance, younger people, those who held less negative attitudes about immigrants in general and specifically about Latin American and African immigrants, and who express more warmth for African Americans and less warmth for Whites were more likely to report that the projected demographic changes were "neither good nor bad" than "bad for the country." Descriptive analyses further revealed that White Neithers also expressed racial attitudes that were consistently in between their optimistic and pessimistic counterparts, such as holding comparatively less regressive racial attitudes than White pessimists. We are now confident that "neither good nor bad" represents a substantive and socially meaningful response, one which does not appear to reflect "pessimism" in disguise. Yet more work is needed to uncover what this stance fully means. As past research has shown, midpoint responses may mean that respondents are neutral, undecided, or ambivalent (Klopfer and Madden, 1980;Truebner 2021). None of these stances should be interpreted as an "empty attitude" (Baka et al., 2012). Each is meaningful, albeit in different ways. Our data and analyses do not allow us to make a conclusive determination but do show that White Neithers-like White respondents, more generally-are not monolithic. These respondents were internally divided. Nearly half of White Neithers, for instance, claimed that ethnoracial diversification will "strengthen American customs and values," compared with the 34% that foresaw cultural weakening (Figure 3). Reversing the modal order, more White Neithers reported that population shifts produce "more conflict between racial and ethnic groups" than lessen conflict, 43% to 22%, respectively (Figure 2). Such distributions may offer one starting point to further interrogate the meaning of "neither good nor bad" about ethnoracial diversification.
Nonetheless, our results do caution against concluding that the existence of "neither good nor bad" responses among White Neithers proffers straightforward evidence of greater inclusivity, tolerance, and otherwise racially liberal attitudes. Although sharing some characteristics with Optimists and being comparatively more liberal than their demographically pessimistic counterparts, White Neithers were much less likely than Optimists to acknowledge White privilege and structural racism (Figure 1). For instance, over 44% of Neithers claimed that "too much attention" is paid to race and racial issues, compared to just 12% of Optimists. Perhaps more telling, nearly 60% of White Neithers expressed that "Blacks who can't get ahead in this country are most responsible for their own condition." Thus, although between the attitudes held by Pessimists and Optimists, many White Neithers variously exhibit traits of colorblind (Bonilla-Silva 2006) and laissezfaire racism Denis 2015).
Ultimately, we believe that this study demonstrates the value of shifting from demographic perceptions as explanans to explanandum. Recent scholarship has focused on the effects of demographic perceptions about race and ethnicity on policy preferences, partisan affiliation, and group identification (e.g., Craig et al., 2014Craig et al., , 2018bDanbold and Huo, 2015;Major et al., 2018;Skinner and Cheadle, 2016). Even if this scholarship were to become more attentive to non-pessimistic responses, we still need accounts that explain what factors inform these perceptions to begin with. As cultural sociologists and sociologists of knowledge have long maintained, imagined futures are socially conditioned and cultivated (Auyero and Swistun, 2009;Mische 2014;Rodríguez-Muñiz 2021;Schütz 1959). Perspectives and feelings about the future-demographic or otherwise-thus do not arise out of thin air. With our focus on White Neithers, we illustrate the purchase of prioritizing the sources and underpinnings of demographic imaginaries.

Conclusion
Over the past several decades, ethnoracial demographic change has become a major focus of academic and public discussion in U.S. society (e.g., Abascal 2020; Alba 2020; Chavez 2008; Richeson, 2014, 2018b;McConnell 2011McConnell , 2019Outten et al., 2012). A growing body of social science research has sought to understand how the White population-currently the most numerous and politically powerful-understands, evaluates, and responds to projected population trends. Experimental, survey-based, and qualitative studies have uncovered widespread White demographobia and pessimism.
While unsurprising given histories and contemporary manifestations of racialized population politics (Rodríguez-Muñiz 2021), research shows that such sentiments are contributing to increased levels of conservatism and racial animus among Whites. Yet, as this study has demonstrated, Whites are not of one mind about what ethnoracial change bodes for the future of the country. A substantial portion of the White population-larger than both pessimists and optimists-affirms that these future trends as "neither good nor bad." While further data and analysis are needed to arrive at definitive conclusions, the results of our study suggest that "neither good nor bad" is a substantive response that differs from optimism and pessimism.
We believe the current study suggests several future lines of research. Three lines seem especially pertinent to us. As previously noted, more research is needed to ascertain whether this response reflects uncertainty, neutrality, or ambivalence. We believe that qualitative research could help us unpack the ideas, associations, and emotions that underpin this midpoint response provided on quantitative surveys. With such knowledge, analysts would be better positioned to theorize the meaning and implications of this sentiment and its relation to Whites' sense of group position (Blumer 1958).
We also need more longitudinal research on the "neither good nor bad" response to questions about ethnoracial diversification. Although not the focus of our study, sentiments about the country's future ethnoracial composition can change (Budiman 2020). We do not yet know how fluid or fixed this middle-point response is, relative to expressions of demographic optimism and pessimism. Of particular importance is identifying factorsindividual, ideological, or institutional-that may move White Neithers toward demographic optimism or pessimism. Research shows, for instance, that historical events can profoundly influence how cohorts perceive social issues and policies (Schuman and Scott, 1989). These data were collected before and after the 2016 U.S. Presidential Election, a period that seems to have further polarized Whites' political and racial attitudes (Doherty et al., 2019) and perhaps their views of the nation's demographic future, as well. Research on change over time-both for individuals over the life course and at the aggregate levelcould improve our understanding about how people perceive shifts in U.S. demographics.
Finally, as other scholars have argued, we need more research on how African Americans, Latinos, Asians, and other populations perceive demographic trends. With few exceptions (e.g., Abascal 2015; Craig and Richeson, 2018b), most research on this topic has focused exclusively on Whites. Our research suffers from this limitation. Although it is beginning to change, ATP data has traditionally generated small sample sizes for non-White panel respondents. This has hindered the development of a more comprehensive and comparative understanding based on this timely, frequently collected, and nationally representative data source. However, available data does suggest that non-White respondents also claim "neither good nor bad" at high levels (Budiman 2020). But we know even less about what motivates this response and what it means for members of other populations.
In conclusion, we must note that any of the above lines for future research rest on the inclusion of middle point categories (see Truebner 2021;Wang and Krosnick, 2020) as response categories. Yet, most surveys on demographic perceptions continue to employ a binary design. Our analysis demonstrates that the omission of "neither good nor bad" or other intermediate categories runs the risk of painting a polarized picture of White demographic attitudes uncorroborated by our study.
at the University of California-Riverside for their helpful comments on previous versions of the article. We thank Radhardhika Utama for his help with the data visualization and Nick Bertoni at Pew Research Center for information about the American Trends Panel. The opinions expressed herein, including any policy implications, are those of the authors and not of Pew Research Center.

Notes
1 As with all statistical knowledge, demographic projections are based on a series of assumptions and conventions (Keyfitz 1987;Rodríguez-Muñiz 2021), including their underlining racial conceptualization (Morning 2011). 2 Extensive empirical testing (e.g., Baka et al., 2012;Presser and Schuman, 1980;Truebner 2021;Velez and Ashworth, 2007) has failed to generate consensus. Some contend that the inclusion of middle point response categories lowers the data reliability (Alwin et al., 2018), while others have found that their absence may induce "forced directional" responses (Sturgis et al., 2014) and increase non-response rates (Revilla et al., 2014). Nora Cate Schaeffer and Stanley Presser (2003) note that midpoint response categories may also connote indifference. Patrick Sturgis and colleagues (2014) consider this a form of social desirability bias. 3 To the contrary, Brenda Major and colleagues (2018) find that partisanship was not a major predictor, as news of diversification moved both Democrats and Republicans towards the right. 4 Survey researchers have highlighted several reasons to doubt the substantiveness of this response category.
Respondents may opt for midpoint responses as an act of "satisficing" to minimize cognitive burden, perhaps due to response fatigue or topical disinterest (Krosnick 1991). They may instead choose "neither/nor" out of discomfort with admitting ignorance on a subject, what Sturgis and colleagues (2014) call "face-saving don't knows." Scholars also posit that such responses may reflect an individual's "response style" (Hurley 1998). Finally, an individual may choose a "neither/nor" to obscure their actual position, out of social desirability or some other reason. 5 Although some recent research (e.g., Alexander 2018) has found that White respondents employ "don't know" and item refusals to avoid discussing race-related topics on surveys, these data indicate very low non-response on this variable. Most White respondents in Wave 16 provide a "neither" response irrespective of the four modes of ATP data (mobile phone, tablet, desktop, and mail). None of these modes involved interaction with a live person, which have been found to increase social desirability bias compared with self-administered surveys (Keeter 2015;Krysan 1998;Krysan and Couper, 2003;Morning et al., 2019). Respondents who completed the survey by mail were the most likely to report that ethnoracial diversification was "bad for the country" and the least likely to say "Neither" relative to those who completing the survey by a mobile phone, tablet, or desktop, at the .05 level of significance. Although the difference between mail respondents and the three other modes implies something related to mode, it is more likely related to the demographics of whites responding by mail, such as their older ages, which has long been related to more negative racial attitudes (e.g., Bobo et al., 2012;Jardina 2019;Quillian 1996). 6 The only publicly available geographic information in the survey is the respondent's census region, included to tap into larger contextual factors. The racial attitudes literature commonly uses region to control for larger spatial and sociohistorical context when more detailed data is unavailable (e.g., Schildkraut and Marotta, 2018). While unusually rich, the ATP waves we use exclude direct indicators or perceptions about current racial group size, concentration, and intergroup contact. Prior studies have found that these variables are relevant to evaluations of ethnoracial change (e.g., Alba et al., 2005;Enos 2014;Quillian 1996). Their absence here is a limitation. 7 These Wave 10 data were collected before all candidates had announced their candidacy for president and the 2016 selection of the Republican Presidential nominee. 8 Race and immigration attitudes are commonly studied together (e.g., Alba et al., 2005;Bobo et al., 2012;Jardina 2019;Quillian 1996;Valentino et al., 2013); and appropriate for this study given that national ethnoracial shifts are due in part to post-1965 Asian and Latin American migration (e.g., Pew Research Center 2015a). 9 Thermometers allow respondents to offer unconstrained evaluations about the strength and direction of their feelings and affect about a particular group (Jardina 2019, p. 83). 10 Another potential contrast is between Whites who are optimistic about future ethnoracial diversification versus pessimistic; results for the two identical specifications are provided in the Appendix. 11 There was less than 13.0% missing data on any single variable, as was the case for an item used in the creation of the news knowledge variable from Wave 10. Less than 1% of White respondents had missing data on the dependent variable. If listwise deletion was used, approximately 18.2% of the total sample would be missing.
Multiple imputation can partly correct the biases of listwise or pairwise deletions, while offering more statistical power and/or less biased results than listwise or casewise deletion (van Ginkel et al., 2020, p. 302). Data were imputed using a customized imputation method related to each variable type consistent with recommendations (van Ginkle et al., 2020, p. 305;von Hippel 2008;White et al., 2010;White et al., 2011).
Additional analyses indicate that these results are nearly identical to alternative approaches of imputing only missing values on the independent variables or using complete-case analysis with unimputed data. 12 These analyses are based on Whites in Wave 16 who participated in any of these later waves, independently.
We did not impose the requirement that respondents had to complete Wave 16 and all four other waves as that would unduly decrease the sample size for the less-common "good" and "bad" responses. See the Appendix for more information about these waves. 13 Wave 41 asked respondents, "According to the U.S. Census Bureau, by the year 2050, a majority of the population will be made up of Blacks, Asians, Hispanics, and other racial minorities." Five response choices about whether the impact on the country of this change ranged from "a very good thing" to "Neither a good nor bad thing." 14 Results with identical analyses with Pessimists as the reference group relative to Optimists are provided in the Appendix. 15 Additional analyses indicate that the addition of variables on immigration attitudes reduces the Republican variable to insignificance. 16 Ancillary analyses using an alternative coding with three binary options (want immigration decreased, want immigration increased, want immigration kept at its present level) suggests that White respondents who say that they want increased immigration (compared to those wanting decreased immigration) have lower RRRs of providing a "Neither Good nor Bad" response about future ethnoracial diversification rather a response of Good. There is no significant difference in the outcome (Neither versus Optimism) between those who want immigration kept at its present level versus decreased. 17 Additional analyses indicate that White respondents who report that Latin American immigrants negatively impacted society are less likely to be Neither (and more likely to be pessimistic) than Whites who think that Latin American immigrants have positive impacts or those who think that Latin American immigrants have "neither positive nor negative impacts." Ancillary analyses suggest that the relationship between Whites' views about the social impacts of African immigrants and their opinions about future ethnoracial diversification applies to only one set of contrasts. Specifically, Whites who think the impacts of African immigrants are "neither positive nor negative" are more likely to be "Neither" than pessimistic about the increasing ethnoracial diversification of the country overall. 18 The independent variables that are the most highly correlated with each other are the racial thermometer variables (between .6503 and .7999). Ancillary multinomial regression analyses were conducted with the racial thermometers removed; all statistically significant results presented in Table 3 regarding other characteristics (e.g., age, immigrants make things worse, etc.) had the same level of significance and similar Relative Risk Ratios in this alternative model. 19 See the Appendix Tables for more information about these survey waves and attitudes. 20 Ancillary analyses suggest that "Neither" responses to this particular demographic shift do not appear to stem from a general uncertainty about anticipated future events. Whites in Wave 41 were also asked about whether "the number of people 65 and over will outnumber people younger than 18" would be good, bad, or neither good nor bad for the country. Most White respondents indicated that an aging country would be a bad thing for the country. 21 For example, more than 90% of Pessimists expect the changes will cause "more conflicts" and "weaken" American customs and values (90.3% and 91.6%, Figures 2 and 3).  Source: ATP Waves 10, 15, and 16, linked data. Note: All models also control for respondent gender, married/cohabitating, income, has health insurance, second generation in the U.S. or more, has immigrant friends or relatives, and thought about the 2016 campaign. The second specification also includes indicators regarding perceptions regarding whether immigrants want to assimilate and whether recent immigrants learn English. †p < .10. *p < .05. **p < .01. ***p < .001  Responses of those who refused to answer either question are dropped from the sample.

N=1641
Notes: The analyses use non-imputed data and respondents with missing data on any item are dropped from the analyses. Table A4. Description of Variables Used in Figure 1 Wave 18 Not more difficult to be Black How much more difficult, if at all, is it to be Black in this country than it is to be White? 1=No more difficult, 0=A lot more difficult or a little more difficult. White people don't benefit from advantages How much, if at all, do White people benefit from advantages in society that Black people do not have? 1=Not too much or Not at all; 0=A great deal or a fair amount.

Wave 20
Country made changes to give Blacks equal rights with Whites Which of these two statements comes closer to your own viewseven if neither is exactly right? 1=Our country needs to continue making changes to give Blacks equal rights with Whites; 0=Our country has made the changes needed to give Blacks equal rights with Whites.

Death of African Americans are isolated incidents
Do you think the deaths of African Americans during encounters with police in recent years are: 1=Isolated incidents; 0=Signs of a broader problem between African Americans.

Wave 22
Racism is a "small problem" or "not a problem at all" How much of a problem do you think racism is in the country today? 1=A small problem or not a problem at all; 0=A very big problem or a moderately big problem. Wave 24.5 Too much attention paid to race and racial issues in our country these days In general, do you think there is too much, too little, or about the right amount of attention paid to race and racial issues in our country these days? 1=Too much attention; 0=Too little attention or about the right amount. Blacks who can't get ahead in this country are mostly responsible for their own condition Please choose the statement that comes closer to your own viewseven if neither is exactly right. 1=Blacks who can't get ahead in this country are mostly responsible for their own condition. 0=Racial discrimination is the main reason why many black people can't get ahead these days. More people of different races marrying each other is a bad thing for society Please indicate if you think the following trend is generally a good thing for our society, a bad thing for our society, or doesn't make much difference. 1=More people of different races marrying each other is a bad thing for our society; 0=A good thing, doesn't make much difference. "Only a little" or "none at all" discrimination against Blacks Please tell us how much discrimination there is against each of these groups in our society today. Blacks: 1=Only a little or none at all; 0=A lot or some "Only a little" or "none at all" discrimination against Hispanics Please tell us how much discrimination there is against each of these groups in our society today. Hispanics: 1=Only a little or none at all; 0=A lot or some Some or a lot of discrimination against Whites Please tell us how much discrimination there is against each of these groups in our society today. Whites: 1=A lot or some; 0=Only a little, none at all.