1 Introduction

The objective of this paper is to examine the association between racial compositions at the county-level in the USA and the well-being of residents. My hypothesis is simple and straightforward—if people like people of their own race more than they like people of other races, as assumed in Becker’s (1957) theory on discrimination, then I would expect the utility (or disutility) to be reflected in people’s evaluation of their own life. The evolutionary and social psychology literature has extensively documented that “human beings are genetically programmed to form in-group, out-group associations and to prefer members of what they perceive as their own group” (Alesina et al. 2001, p. 227), which suggests that increased interactions across racial lines may have well-being effects.

Previous research suggests that racial prejudice among Whites tends to increase with the percentage of the population that is non-White (Taylor 1998; Enos 2010; Stephens-Davidowitz 2014), and the “racial threat” theory (Key 1949) predicts that Whites, who tend to be the majority group in most areas in the USA, feel worse off as the population of non-Whites increases. Relatively little research has been conducted on how local racial diversity is associated with the well-being of the population in the USA. I also examine the link between the share of the immigrant population and life satisfaction of residents, as immigration is likely to affect local racial compositions given that the majority of foreign-born individuals are either from Asia or Latin America.

I believe that this is a timely topic. During his campaign for president in 2015 and 2016, the Republican presidential nominee Donald Trump, whose victory was described as “part of a global White backlash” (Beauchamp 2016), had called Mexican immigrants criminals and rapists (Lee 2015), criticized the “Black Lives Matter” movement (Sherfinski 2015), and called for a ban on Muslim immigration. His supporters were overwhelmingly White.Footnote 1 Coincidentally, there has been increased academic and public interest in how Whites are feeling status anxiety in the USA in recent years (e.g., Blake 2011; New York Times 2011; Norton and Sommers 2011; Mayrl and Saperstein 2013).Footnote 2 Some even speculate that racial status anxiety is contributing to rising mortality and drug and alcohol abuse among less educated Whites (e.g., Marshall 2015).Footnote 3 This study contributes to the literature by investigating who feels better off or worse off as a result of living in racially homogeneous and heterogeneous areas in the USA.

The magnitudes I find suggest that a ten-percentage-point increase in the share of the non-White population (approximately one half of a standard deviation) is associated with 0.006 and 0.007 points reduction in life satisfaction on a four-point scale for White men and White women, respectively. For White men, this effect appears to be driven mainly by the percentage of the population that is Black. I also find that a ten-percentage-point increase in the percentage of the immigrant population (approximately two standard deviations) is associated with 0.009 and 0.021 points reduction in life satisfaction for White men and White women, respectively. The percentage of the non-White population seems to reduce older Whites’ life satisfaction more than that of younger Whites. Though the scale of the findings relating to the impact of local racial compositions and immigrant population is relatively modest, the findings may pose a challenge in the coming years as the percentage of the population that is non-White rises in the USA.

1.1 The literature review

Previous studies have found that racial heterogeneity is associated with various outcomes, including reduced social solidarity, social capital, altruism, and community cooperation (Putnam 2007), lower participation in social activities (Alesina and La Ferrara 2000), and lower social trust (Alesina and La Ferrara 2002; Putnam 2007; Schmid et al. 2014). Glaeser et al. (2000) document experimentally that people of different races are more likely to cheat one another. DiPasquale and Glaeser (1998) find that racial heterogeneity is a significant determinant of rioting, while poverty in the community is not. Finally, perhaps not surprisingly, racial heterogeneity seems to be an important factor in how local policies are determined. Alesina et al. (2004) show that people prefer to form racially homogeneous political jurisdictions in the USA. Alesina et al. (1999) find that racially heterogeneous areas tend to spend a smaller fraction of their budget on social services and productive public goods, and more on crime prevention in the USA. Alesina et al. (2001) argue that one reason the US redistributes income less than racially homogenous European countries is that the majority of Americans believe that redistribution favors racial minorities. Similarly, Gilens (1999) finds that White Americans who overestimate the percent of the poor population that is Black are less likely to support welfare and view Blacks as lazy and undeserving.

Given the empirical evidence, it is plausible that people might be less happy in racially fragmented areas, but well-being of Whites may be particularly affected by racial heterogeneity in the area of residence. Sociologists have suggested that members who hold positions of power are motivated to maintain their position of privilege and more likely to favor individuals who share their demographic characteristics (e.g., Reskin et al. 1999; Smith 2002). In aggregate, Whites are the majority group members and hold positions of power, while non-Whites are accustomed to being in small numbers in work environments and other social contexts throughout US society. Thus, if Whites feel that their status at the top of the American racial hierarchy is under threat by non-Whites, their racial status anxiety may lead to lower life satisfaction. Additionally, in the USA, many Whites seem to view non-Whites as a fiscal burden (Gilens 1999), which might make Whites feel less happy about the presence of non-Whites.

In his seminal work, Blalock (1967) argues that as the numerical size of a minority group begins to approach the size of the majority group, increased interactions across racial lines induce a sense of competition among the majority group, who will feel increasingly threatened and often engage in discriminatory acts to protect their resources and advantages. This view is called the “racial threat” hypothesis (Key 1949), which predicts racial animus tends to increase with the percentage of the population that is non-White. Empirical evidence seems to support the hypothesis.Footnote 4 Taylor (1998) finds that Whites’ prejudice tends to increase with the local Black population share (though concentrations of local Asian American and Latino population do not engender White antipathy toward these groups). Enos (2010) finds that White support for Obama has a negative relationship with the size of the Black population. Stephens-Davidowitz (2014), using Google search data, finds that racially charged search rate is higher in areas with higher proportions of Black residents.

Though the literature largely suggests that most people may prefer living in a racially homogeneous area, it is possible that living in a racially heterogeneous area leads to higher levels of cross-racial interactions, which in turn may lead to more understanding and less prejudice. One suggestive piece of evidence can be seen in people’s opinions toward immigrants. SurveyUSA’s survey in 2005 revealed that people in states with more immigrants tend to have more favorable views toward immigration than people who live in areas with few immigrants (SurveyUSA 2005). Caplan (2006, 2016) argues that, when people directly observe many immigrants, they can easily see that most of them do hard, dirty jobs few Americans want, while people who rarely see an immigrant find it easy to scapegoat them for social and economic problems. In the UK, people in areas with many immigrants, such as Londoners, were more likely to prefer to remain in the European Union, though the country as a whole made the decision to leave the European Union.Footnote 5 Caplan (2014) also points out that, when the Swiss passed a referendum to restrict immigration from the EU, Swiss anti-immigration voting was highest in the places with the least immigrants.

Consistent with these statistics on people’s attitudes toward immigrants and voting patterns, two studies find a positive link between immigration and residents’ well-being.Footnote 6 Betz and Simpson (2013) find a positive correlation between immigration and subjective well-being in the 26 European countries,Footnote 7 and Akay et al. (2014) find that natives experience higher life satisfaction from living in areas with more immigrants in Germany. On a similar topic, Akay et al. (2017) find that ethnic diversity is also associated with higher life satisfaction in Germany. However, Longhi (2014) finds that White British people living in racially diverse areas tend to report lower levels of life satisfaction than those living in areas where diversity is low, while she finds little evidence that diversity affects life satisfaction of non-White British people and foreign-born people. Thus, evidence seems to be somewhat mixed among the existing studies on the link association between immigration/ethnic diversity and residents’ well-being. This study, to my knowledge, is the first study to examine the association in the USA.

1.2 Data and methodology

The dataset I use is the Behavioral Risk Factor Surveillance System Survey (BRFSS), which is a household-level repeated cross-sectional survey collected throughout the USA by the U.S. Government’s National Center for Chronic Disease Prevention and Health. The measure of life satisfaction is the response, on a four-point scale ranging from “Very satisfied” to “Very dissatisfied” to the question, “In general, how satisfied are you with your life?” The life satisfaction question has been asked since 2005, except in 2011 and 2012. But due to the changes in weighting methodology and the addition of the cell phone sampling frame, the BRFSS 2011–2015 are not comparable to the BRFSS 2005–2010.Footnote 8 Thus, I use 2005, 2006, 2007, 2008, 2009, and 2010. Among those who answered the life satisfaction question during the period 2005–2010, 46.2% of the sample reported “Very satisfied” and 48.3% reported “Satisfied.” Only 4.5 and 1.0% of the sample reported “Dissatisfied” and “Very dissatisfied,” respectively.

Subjective well-being (SWB), such as self-reported happiness and life satisfaction, has been extensively used by economists despite justifiable concerns that people’s moods at the time of the survey can bias their subjective well-being.Footnote 9 Recent notable studies include Stevenson and Wolfers (2009) who recorded a declining female happiness over time, Sacks et al. (2010) who showed a robust relationship between subjective well-being and income, and Oswald and Wu (2010) who demonstrated that there is a close match between US life satisfaction scores and objective well-being indicators.Footnote 10

I restrict my analyses to those between 18 and 85 years old, not residing in unincorporated US territories, and exclude respondents who refused or were unsure of their response, or whose response is missing, for any of the variables included in my analyses. I match people who were surveyed in a particular county and year with the population statistics, which is obtained from the U.S. Census Bureau.Footnote 11 Data on the foreign-born population is also obtained from the U.S. Census Bureau, but unfortunately the only data available at the county-level is the 2005–2009 American Community Survey, which provides the average share of foreign-born population over the 5-year period of time.Footnote 12 In this paper I define foreign-born population as immigrant population, as the foreign-born population includes anyone who was not a US citizen at birth, that is, those who are US citizens by naturalization or not US citizens. As control variables, yearly county-level median income is also obtained from the U.S. Census Bureau,Footnote 13 and yearly unemployment rates from the Bureau of Labor Statistics.Footnote 14 Table 8 in the appendix shows summary statistics for the county-level variables.Footnote 15

Figures 1, 2, 3, 4, and 5 show the distributions of non-White, Black, Hispanic, Asian, and immigrant populations, respectively, at the county level in the USA. Figure 1 reveals that non-Whites tend to live on the Pacific Coast, the East Coast, East South Central, South Atlantic, and southern Border States.Footnote 16 Figure 2 shows that Black population are concentrated in the South. Figure 3 shows that Hispanic population are concentrated in southern Border States. Figure 4 shows that Asian population heavily concentrate in California and New York as well as other states on the West Coast and East Coast.Footnote 17 Figure 5 shows that immigrants tend to live in California, New York, the southern part of Florida, border counties in Texas, Arizona, and New Mexico as well as large cities such as Chicago, Seattle, and Las Vegas.

Fig. 1
figure 1

Percent of non-White population by county, 2005–2010. Source: Intercensal Estimates of the Resident Population for Counties: April 1, 2000 to July 1, 2010

Full size image
Fig. 2
figure 2

Percent of Black population by county, 2005–2010. Source: Intercensal Estimates of the Resident Population for Counties: April 1, 2000 to July 1, 2010

Full size image
Fig. 3
figure 3

Percent of Hispanic population by county, 2005–2010. Source: Intercensal Estimates of the Resident Population for Counties: April 1, 2000 to July 1, 2010

Full size image
Fig. 4
figure 4

Percent of Asian population by county, 2005–2010. Source: Intercensal Estimates of the Resident Population for Counties: April 1, 2000 to July 1, 2010

Full size image
Fig. 5
figure 5

Percent of foreign-born population by county, 2005–2009. Source: The 2005–2009 American Community Survey 5-year estimates

Full size image

Table 1 shows summary statistics for the BRFSS 2005–2010 respondents by race and gender. It shows that Whites in the sample on average live in counties where more than 75% of the population is White. Blacks, Hispanics, and Asians are more likely than others to live in counties where the own-race population share is larger. That is, we observe residential racial segregation. The county-level variables also reveal differences in local area characteristics across racial groups. Whites tend to live in areas with a smaller population size. Even though both Asians and Hispanics tend to live in areas with a large population size, the median household income is about $10,000 higher and unemployment rate one-percentage-point lower in the areas Asians tend to live than in the areas Hispanics tend to live, on average. Finally, Blacks on average are more likely to live in counties where median income is much lower and unemployment higher than are Whites, Hispanics, and Asians.

Table 1 Summary statistics
Full size table

My empirical strategy involves using reported satisfaction with life as a proxy measure for individual utility and regressing life satisfaction on county-level racial compositions/immigrant population and an extensive collection of covariates and indicator variables (e.g., month and year dummies and state fixed effects). The idea for the empirical test is captured in the following three regression equations:

$$ {LS}_{ict}={\alpha}_1\%{\mathrm{OwnRace}}_{ct}+{\beta X}_{ict}+{\gamma Z}_{ct}+{\theta}_s+{\delta}_t+{\varepsilon}_{ict} $$
(1)
$$ {LS}_{ict}=\sum_{j=2}^5\ {\alpha}_j\%{\mathrm{OtherRace}}_{j ct}+{\beta X}_{ict}+{\gamma Z}_{ct}+{\theta}_s+{\delta}_t+{\varepsilon}_{ict} $$
(2)
$$ {LS}_{ict}={\alpha}_6\%{\mathrm{Immigrant}}_c+{\beta X}_{ict}+{\gamma Z}_{c t}+{\theta}_s+{\delta}_t+{\varepsilon}_{ict} $$
(3)

where LS ict is life satisfaction for the individual i in county c in year t. %OwnRace ct is the share of own-race population in the county of residence. %OtherRace jct is the share of race group j other than the respondent’s own group. For example, for Whites, county-level population shares for Blacks, Hispanics, Asians, and “other” race (American Indian, Alaska Native, Native Hawaiian and Other Pacific Islander, and multiracial) are included in the regression. %Immigrant c is the share of the foreign-born population in the county of residence. X ict is demographic and socioeconomic controls (age, marital status, number of children in household, education, and employment status) as well as month of interview. Z ct is county-level controls (log median income, unemployment, and log county population).

Finally, θ s and δ t are state fixed effects and year dummies, respectively. I use a state fixed model, rather than a county fixed model, for Eqs. (1) and (2), because it is inconceivable that local racial compositions vary substantially over a short period of time at the county-level. Since fixed effects absorb all factors that do not change over time, identifying the effect of local racial compositions would be difficult with county fixed effects. Therefore, coefficients of interest are identified from variation in racial compositions across counties within a state over time for Eqs. (1) and (2). For Eq. (3), coefficients of interest are identified from variation in immigrant population across counties within a state, as the immigration variable does not vary over time.

Despite the wide range of controls included in the regressions above, there still remains the possibility of nonrandom selection. If people who strongly prefer to live in proximity to people of the same race are more likely to move to, or stay in, racially homogenous areas, the association would not be necessarily causal due to selection bias. Therefore, it is important to keep in mind that this nonrandom selection of people into different areas may bias the results.

2 Results

Table 2 shows the results for men. Though life satisfaction is measured on an ordinal scale and is discontinuous, I use a linear model for ease of interpretation, but similar results are obtained from ordered probit or logit models. The BRFSS-provided weights are used to adjust for sampling and nonresponse, and standard errors are clustered at the county-level. In order not to overload the table, I report the coefficients on personal characteristics and county-level controls in Table 9 the appendix.Footnote 18 Columns (1)–(4) show when the share of own-race population is used, and columns (5)–(8) show when the shares of race groups other than one’s own race group are used. For White men, a ten-percentage-point decrease in the White population (approximately one-half of a standard deviation) is associated with a 0.006 points decrease on a four-point scale in life satisfaction. This seems to come mainly from the effect of the Black population, as shown in column (5); a ten-percentage-point increase in the Black population (approximately two thirds of a standard deviation) is associated with 0.007 points decrease in life satisfaction for White men. Paradoxically, Asian men’s life satisfaction decreases with the share of the Asian population. A one-percentage-point increase in the Asian population (more than 2 standard deviations) is associated with approximately 0.003 points reduction in their life satisfaction, as shown in column (4). Perhaps, Asian men living in areas where there are many Asians, such as the Bay Area, are often expected to be successful more than men of other races, and the pressure to succeed contributes to lower life satisfaction. Of course, this is highly speculative and beyond the scope of this study. Finally, columns (9)–(12) show the results when the share of the immigrant population is used. A ten-percentage-point increase in the share of the immigrant population (approximately 2 standard deviations) is associated with a 0.009 points decrease in life satisfaction for White men (column 9).

Table 2 Ordinary least squares life satisfaction equation: men
Full size table

Table 3 shows the results for women. The coefficients on personal characteristics and county-level controls are reported in the appendix, in Table 10. For White women, a ten-percentage-point decrease in the county-level White population is associated with 0.007 points decrease in life satisfaction, and a ten-percentage-point increase in the Black, Hispanic, and Asian population in the county of residence is associated with a reduction in life satisfaction of 0.004 points, 0.008 points, and 0.018 points on a four-point scale, respectively.Footnote 19 A ten-percentage-point increase in the share of the immigrant population (approximately two standard deviations) is associated with 0.021 points reduction in White women’s life satisfactionFootnote 20 and 0.035 points reduction in Asian women’s life satisfaction. Black women’s life satisfaction and Hispanic women’s life satisfaction do not seem to be affected by the racial compositions or immigrant population of the county of residence. Overall, these results show that, for both White men and women, life satisfaction is negatively correlated with the population that is non-White and immigrants.

Table 3 Ordinary least squares life satisfaction equation: women
Full size table

Next, I partition the sample across one’s educational attainment (high school dropout, high school graduate, college graduate) and age (<35, 35–50, 50<). Panel A in Table 4 shows the results for men with different levels of education when the explanatory variable is the share of the own-race population. White male high school graduates and college graduates are happier if they live in areas with higher percentages of the White population, and their life satisfaction seems to decrease with the share of the Black population, as shown in panel B. White college graduates in areas with a large Asian population and a large immigrant population are also less happy, perhaps because they experience more competition in the labor market with Asians, who also tend to be college graduates. Similarly, perhaps for the same reason, Black college graduates are less satisfied with their life in areas with a large Asian population (panel B) and a large immigrant population (panel C).

Table 4 Ordinary least squares life satisfaction equation: men by education
Full size table

Among high school dropouts, Black men’s life satisfaction seems to decrease as the share of the Hispanic population increases. This may be due to the perceptions of the negative labor market effects of Hispanic immigrants, as found in Borjas (2003), among low-skill Black workers. However, in panel C, the effect of the immigrant population is not statistically significant for Black high school dropouts. Perhaps, Black high school dropouts perceive that a large Hispanic population has negative effects on their labor market outcomes even in areas where the Hispanic population may consist of mostly native-born Hispanics.Footnote 21 It is somewhat surprising that a higher share of the immigrant population is associated with lower life satisfaction for college-educated Whites, but not other Whites, as it seems to contradict with the finding that more educated people tend to have favorable views toward immigration in Europe (Card et al. 2005).Footnote 22

Table 5 shows the results for men for three age categories. The percentage of the population that is White seems to increase life satisfaction for White men aged 35–50 and those aged over 50 (panel A), and both groups are negatively affected by the share of the Black population in the county of residence (panel B). Those aged over 50 also seem to feel worse off as the share of the Hispanic population and Asian population increase. Finally, the share of immigrant population is associated with lower life satisfaction for White and Black men age over 50 (panel C). These results are consistent with the finding that younger people tend to have favorable views toward immigration (Card et al. 2005).

Table 5 Ordinary least squares life satisfaction equation: men by age
Full size table

Turning to examining women’s life satisfaction, panel A in Table 6 shows that, regardless of education levels, White women in areas with a higher share of the White population tend to be more satisfied with their life, with the association being higher for high school dropouts. Panel C shows similar results when the share of the immigrant population is used as the explanatory variable. However, when the non-own-race population are used (panel B), White high school graduates are affected by the shares of the Black and Hispanic population, while White college graduates are affected by the Hispanic and Asian population. Black female high school graduates are less happy if they live in areas with a higher share of the White population, and Black female college graduates are happier if they live in areas with a higher share of the Asian population. Asian high school graduates are less satisfied with their life in areas with many immigrants.

Table 6 Ordinary least squares life satisfaction equation: women by education
Full size table

Finally, panel A in Table 7 shows that White women aged 35–50 and those aged over 50 are more satisfied with their life if they live in a higher share of the White population. Panel B shows that life satisfaction of young White women (<35) decreases with a share of the Hispanic population, that of White women aged 35–50 decreases with a share of the Asian population, and that of White women aged over 50 decreases with shares of the Black, Hispanic, and Asian population. Panel C reveals that the immigrant population seem to decrease life satisfaction for White women aged 35–50 and aged over 50, young Asian women, and older Hispanic women, while life satisfaction of young Black women tends to increase with a share of the immigrant population for reasons that are not well understood.

Table 7 Ordinary least squares life satisfaction equation: women by age
Full size table

Overall, the results demonstrate that the local racial composition/immigrant population effects differ for Whites and non-Whites. The results are mixed for non-Whites, but Whites tend to feel worse off when (1) the share of White population declines, and (2) the share of immigrants increases.

3 Conclusions

The main purpose of this paper is to examine if one’s life satisfaction is associated with the racial compositions and immigration population in the county of residence. I find that a larger percentage of the population that is non-White lowers Whites’ life satisfaction. The finding is consistent with the view that Whites feel heightened status anxiety as they are not accustomed to the notion that they are in smaller numbers. Younger Whites seem to have favorable views toward racial minorities and immigrants, as I find that older Whites are less happy in racially diverse areas than their counterparts in more racially homogenous areas. Somewhat surprisingly, own-race preference increases with education for White men, and there is little evidence that White male high school dropouts in racially diverse areas feel worse off.

One caveat must be stated. As mentioned above, nonrandom selection of people into areas with different racial compositions and immigrant population makes the coefficients difficult to interpret as the causal effect. People may tend to move to, or stay in, areas where they can find more own-race residents, and the decision to stay or move may be correlated with one’s life satisfaction. However, the results found in this study are not inconceivable, given that the previous studies find that racial heterogeneity is associated with various negative outcomes such as trust. Also, the findings are in line with the study by Longhi (2014), who finds a negative well-being effect of racial diversity for Whites in the UK. Furthermore, the findings are in line with the recent US presidential election, of which racial issues, namely white backlash against multiculturalism, were a constant feature.

The author acknowledges that the magnitudes found are relatively small compared to other personal characteristics such as marital status and employment status. For example, the coefficients on divorce are approximately 0.21 for men and 0.18 for women, and the coefficients on unemployment are approximately 0.22 for men and 0.23 for women. I find that a ten-percentage-point decrease in the White population (approximately one half of a standard deviation) is associated with a 0.006–7 points decrease on a four-point scale in life satisfaction for Whites, while a ten-percentage-point increase in the percentage of the immigrant population (approximately 2 standard deviations) is associated with 0.009–0.021 points reduction in life satisfaction for Whites.

Nevertheless, the results found in this study are in sharp contrast with those of Betz and Simpson (2013), who find that immigrants tend to increase well-being of residents in Europe, and those of Akay et al. (2014) and Akay et al. (2017), who find that immigrants and racial diversity tend to be associated with higher subjective well-being in Germany. It may be dispiriting to some readers and policymakers in the USA, where the percentage of non-White population is expected to increase, to learn that, though seemingly trivial in magnitude, negative well-being effects of racial diversity and immigration are found in this study. As the current demographic trend indicates that White people will no longer make up a majority of Americans by 2043,Footnote 23 the findings of this study may pose a challenge in the coming years for Whites and policymakers in the USA.