Abstract
This article examines an idea that is often asserted but rarely tested: that Americans associate big cities with African Americans and that, as a result, racial attitudes influence support for urban policy. Thirty-five years of public opinion data show that cities are in fact a ‘racialised’ concept, and that the relationship between racial attitudes and support for place-based urban policy is as large as that between racial attitudes and support for person-based assistance to the poor. The sources of these racial associations, however, appear to differ. Attitudes about race and cities correlate more closely with attitudes about crime, while attitudes about race and person-based redistribution correlate more with opposition to residential integration. Lastly, the evidence shows that even Americans who do not hold prejudiced views associate urban problems with African Americans, suggesting that social policy, be it person- or place-based, will always need to contend with racial attitudes.
1. Introduction
When Americans think about cities, do they also think about Blacks? If so, how does this connection influence support for urban policy? The idea that Americans associate cities with Blacks is not new. For the past 40 years, American central cities have been not just disproportionately poor, but also disproportionally African American. In 2000, approximately 8 per cent of suburban residents were Black and 9 per cent poor; in the central cities, 21 per cent were Black and 18 per cent were poor. Thus while urban policy may be ‘race-blind’, it may not be ‘race-neutral’. A commitment to helping central cities can be viewed as a commitment, however indirect, to helping African Americans.
In 1996, during the American debate about welfare reform, the Minneapolis Star-Tribune quoted law professor John Powell saying A lot of this discussion is racism in drag. … When you talk about welfare, vouchers, urban strategies, crime, poverty, you’re really talking about race (von Sternberg, 1996, p. 1a).
Martin Gilens, who quoted Powell in his study of how race prejudice influences attitudes about welfare, went on to note that Although most city dwellers are neither poor nor black, discussions of ‘urban policy’ or ‘how to help America’s cities’ are taken to mean ‘what should we do about poor black neighborhoods’ and the crime and poverty with which they are associated (Gilens, 1999, p. 68).
And Susan Fainstein and Dennis Judd wrote Since the Presidential race of 1964, Republican candidates have employed a code language that equates liberal ideology with negative images of cities: terms such as ghetto, welfare, the underclass, crime and inner city have been utilized to play upon the racial fears and resentments of suburbanites and Sunbelt Republicans (Fainstein and Judd, 1999, p. 12).
Such assertions about cities, while often made, are rarely tested. Historians have confirmed that political entrepreneurs used coded rhetoric in an attempt to link cities, poverty, lawlessness and Blacks (Teaford, 1990; Edsall and Edsall, 1992; Alesina and Glaeser, 2004). And researchers have extensively documented the relationship between racial attitudes and opinions about person-based redistribution. People who have negative opinions about Blacks are more likely, all else equal, to have negative opinions about welfare—even though most Blacks are not on welfare and most welfare recipients are not Black (Gilens, 1999; Gilliam, 1999; Alesina and Glaeser, 2004; Kinder and Kam, 2009). Linking Blacks to welfare, however, is not automatically equivalent to linking them with cities and the fact that political entrepreneurs attempted to create (or tap into) a latent association between Blacks and cities does not mean they were successful, or that such an association exists.
This article presents a quantitative test, using representative survey data, of the hypothesis that Americans associate helping cities with helping African Americans. The results have implications for scholars interested in how race shapes public opinion and also for efforts to fight poverty and unemployment. Specifically, understanding how people react to the idea of big cities has implications for the longstanding debate over the relative merits of person- and place-based policy (Winnick, 1966; Bolton, 1992; Glaeser and Gottlieb, 2008). Advocates of person-based policies (direct welfare payments, housing vouchers, food stamps and so on) argue that they deliver benefits directly and exclusively to those who need them. Not everyone in poor places is poor, so channelling aid to distressed cities—through tax incentives, enterprise zones, block grants and other place-based programmes—risks helping some people who need no assistance and overlooking some who do. Critics of placemaking further argue that its record of success is lamentable (Lehman, 1994) and that it discourages the migration necessary to mitigate economic distress (Glaeser, 2000). Person-based policy offers mobility; given resources, poorer people can move to non-poor areas with more jobs, better schools and lower crime rates.
Stacked against these economic arguments is a cold political fact: power tends to be place-based. Thus person-based policy, however desirable in the abstract, might be unrealistic in practice. Delivering aid to troubled places might simply be more feasible than delivering it to distressed people. In the US, moreover, the political obstacles to person-based policy are compounded by poverty’s entanglement with race. Because a disproportionate share of America’s poor are African American, and animus towards Blacks remains prevalent among Whites, prejudice can erode support for person-based redistribution (Gilens, 1999; Luttmer, 2001). And race-based opposition could be exacerbated if person-based assistance helps poor minorities move to non-poor White areas (see Briggs et al., 2010, ch. 4).
Investing in troubled cities, then, could sidestep political landmines associated with racial attitudes, including White suburbanites’ fear about the arrival of racial minorities. Yet investing in cities only offers this advantage if racial attitudes do not influence support for place-based urban policy, or if they influence it less than they influence support for person-based policy.
The next section of this article describes the data and empirical approach I take in addressing this question. Section 3 presents a three-part analysis, which first establishes that race and racial attitudes are strongly associated with people’s support for helping cities. Further, the effect of racial attitudes on support for helping cities is comparable in size to the effect on supporting person-based redistribution, suggesting that helping places will not allow policy-makers to avoid racial politics. From there, the analysis shows that the magnitude of these racial associations has changed over time, and presents suggestive (but not definitive) evidence that the racial associations with person- and place-based policies spring in part from different sources: cities and racial attitudes exhibit a stronger correlation with concerns about crime, while opposition to person-based assistance is correlated with opposition to residential integration. Lastly, the analysis demonstrates that even people who hold progressive racial views associate poverty and cities with African Americans; that racialisation is not merely a function of racial animus.
2. Data and Empirical Approach
I measure the effect that attitudes about African Americans have on support for place-based investment in cities and compare the size of that effect with the effect racial attitudes have on support for person-based assistance to the poor. In doing so, I follow the standard procedure for analysing the influence of racial attitudes on support for social policy (Sears et al., 1997; Gilens, 1999; Kinder and Kam, 2009). This approach generally uses linear or binomial regressions to analyse survey data and evaluates the impact that attitudes about a particular group have on attitudes about particular government programmes. There are limits to this approach. First, isolating causality, as opposed to correlation, is difficult in statistical tests of this sort, because attitudes can be endogenous. For example, people’s opinions about a racial group might be influenced by their attitudes about cities, not the other way around. For my purposes, however, it is enough to demonstrate a strong association between attitudes about Blacks and attitudes about cities, and to show that this association remains strong in the presence of control variables that could also plausibly influence opinions about cities.
A related and conceptually more difficult question lies in the origins of the attitudes themselves. Regressions can document correlations between different attitudes, but these correlations rest on the perhaps-uneasy assumption that such attitudes are divisible. Thoughts about Blacks might influence thoughts about cities, or both might arise from deeper, unobserved psychological predisposition, be it towards ethnocentrism (Kinder and Kam, 2009) or authoritarianism (Hetherington and Weiler, 2009). This is an important question, but not one I can resolve in this article.
I examine attitudes about Blacks, rather than other racial minorities, for three reasons. First, opinion surveys have asked Americans about their attitudes towards Blacks for a longer time, and in more ways, than they have about other minorities. Secondly, the existing literature about racialised social policy emphasises perceptions about Blacks (for example, Sears et al., 1997). Lastly, without minimising the difficulties faced by other racial minorities in the US, there are strong arguments, rooted in the character and duration of American slavery and its aftermath, that Blacks are the racial minority Whites most fear and resent (see Glazer, 1992).
The statistical tests I conduct require a dataset that asks a representative sample of Americans questions about cities, about Blacks, and about person-based assistance to the poor. For this reason, I use the National Opinion Research Center (NORC’s) General Social Survey (GSS). NORC administered the first GSS in 1972 and completed its 27th round in 2008. Researchers who study racial attitudes frequently use both the GSS and the National Election Survey (NES). 1 The NES, however, does not ask questions about assisting cities. 2
Since 1973, the GSS has asked a representative sample of Americans a number of questions about social problem spending, all of which start with the following preamble We are faced with many problems in this country, none of which can be solved easily or inexpensively. I’m going to name some of these problems, and for each one I’d like you to tell me whether you think we’re spending too much money on it, too little money, or about the right amount
The preamble is followed by different topics, of which three are
— Solving the problems of the big cities.
— Welfare.
— Improving the conditions of Blacks.
Figure 1 shows, by year, the share of Americans who said the country was spending “too little” on each of these problems (the trend line is probability-weighted to correct for sampling error in the GSS). Across all years, investing in cities is substantially more popular than directly investing in the poor (represented by “welfare”). In most years, between 45 and 60 per cent of Americans said too little was spent on cities, while only 12–25 per cent thought too little was spent on welfare. 3 “Solving the problems of big cities” is also more popular than “improving the conditions of Blacks”, but the trend lines in support for all policies exhibit strong similarities, for the most part rising and falling in sync. Across all years, the coefficient of correlation between spending on cities and spending on Blacks is a moderately strong (for survey data) 0.32, which is essentially equal to the correlation between support for spending on Blacks and support for spending on welfare (0.3). One notable exception is the period immediately following 1992—among other things, the aftermath of the Los Angeles riots. In that period, support for increased aid to cities rises, while support for spending on Blacks and welfare falls. These trend lines therefore lend some initial albeit superficial credence to the idea that helping places could be politically tractable when helping people is not.
Distribution of support for increased spending on select social problems, 1973–2008.
A well-known problem with attitudinal surveys, however, is that the phrasing of a question can bias its response (Kalton and Schuman, 1982; Goetz, 2006). In the early 1980s, GSS administrators began to suspect that some questions about social problem spending were framed in a way that unduly influenced responses. For example, two of the three questions in Figure 1 use wording that implies success: the national government would not just spend money on cities or Blacks, it would solve the problems of cities and improve the conditions of Blacks. Further, asking about “welfare” may not elicit accurate responses about person-based policy, because “welfare” carries with it (rightly or wrongly) connotations not just of redistribution, but also of unwieldy bureaucracy, intrusive and ineffective government, and fraud (Wright, 1977; Smith, 1987; Rasinski, 1989; Smith 1984, 2009). Thus two questions are phrased in a way that could needlessly imply success, while the third could needlessly imply failure.
To address concerns about question wording, in 1984 the GSS began to randomly split its sample when asking about social problem spending. Between one-half and two-thirds of GSS respondents were asked the questions shown in Figure 1, while the remainder were asked neutrally worded variations. These variations have the same preamble as the original questions (“We are faced with a number of problems …”), but each social problem is phrased differently
— Assistance to cities.
— Assistance to the poor.
— Assistance to Blacks.
Figure 2 shows probability-weighted trends in support for increased spending when these alternative wordings are employed. The results are striking. While “welfare” was the least popular form of social spending in Figure 1, “assistance to the poor” is the most popular; substantial majorities consistently favour increased efforts at person-based redistribution. Spending on cities, by contrast, is now the least popular policy. Support for spending on Blacks changes only slightly. Notably, while the altered wording changes the absolute levels of support, it appears to have less influence on the year-to-year variation in support. Opinions about urban policy still vary closely with opinions about helping Blacks and the correlations between the responses remain around 0.3. The ‘riot effect’, however, appears to shrink. After 1992, support for assistance to Blacks falls, but support for both assistance to the poor and assistance to cities rises. 4
American support for increased spending on select social problems (alternative wording), 1983–2008.
The large differences in absolute levels of support that result from changing question wording suggest the importance of controlling for wording when analysing survey responses on these topics. The small changes in variance, however, suggest that question phrasing may have few practical implications. People might react to a word when they answer a survey, but react to a concept (including the variety of words used to describe it) when they ponder a policy’s actual implementation. To sort this out, in my analysis I examine both sets of GSS respondents: those who answered the original questions between 1973 and 2008, and those who answered the neutrally worded variations from 1984 to 2008.
Table 1 shows summary statistics from both samples for the variables used in the analysis. The dependent variables are questions about helping cities and the poor. The independent variables are questions that ask about racial attitudes. The control variables are demographic attributes generally associated with attitudes towards social policy—opposition to social programme spending is more prevalent among Whites, Southerners, the less-educated, the elderly, males and conservatives (Sears et al., 1997).
Summary statistics for variables used in analysis
Only 1983–2008 responses used.
Question only asked 1990–2008.
Responses from 1990–2008 used. Question has two wordings: original question asks about “halting the rising crime rate”; alternative question asks about “law enforcement”.
Note: Tabulations are weighted to account for sampling error.
Sources: NORC GSS 1973–2008.
Demographically, the two samples are identical. About 80 per cent of respondents in both groups are White and just under half are male. Between 20 and 25 per cent have a college education or more and about 9 per cent are over 65. Roughly a third live in the South. I also control for residence in a central city: about 17 per cent of each sample live in central cities (the survey’s smallest central city has about 257 000 people, the largest—New York—just over 8 million).
Just under half of each sample is politically conservative. Controlling for conservatism is important because opposition to programmes that help minorities might represent a preference for small government, rather than racial animus. (Although, again, the extent to which these attidues are divisible is an open question.) The GSS asks respondents to self-identify as conservative or liberal on a 1–7 scale, but I use voting data, rather than ideological self-reporting, to place respondents in political categories. Self-reported ideological data are troublesome because ‘conservative’ and ‘liberal’ are broad terms. One could identify as conservative out of a preference for small government, but also out of religiosity or a desire to maintain existing traditions and authority. For instance, a 2009 Gallup poll found that Blacks were the most conservative American ethnic group (Jones, 2010). Yet few Blacks hold policy opinions suggesting a desire for small government. To avoid this sort of confusion, I rely on revealed preferences for candidates who actively espoused limited government. I classify respondents as conservative if they voted Republican in the most recent presidential election, or—for those who did not vote—if they say they would have voted Republican (I count people who voted for George Wallace in 1972 as conservatives). 5
The table includes no data on income, because about 40 per cent of GSS respondents decline to reveal their yearly earnings and this non-response is more prevalent among Blacks, women, political conservatives and people with less education. Systematically missing data have no good solution, so I do not use a direct measurement of income, and instead use income data only to validate other results. This decision is less problematic than it may at first seem. Although one might expect income to determine preferences about national government spending, a substantial body of empirical work suggests this is not the case (see Caplan, 2007, for a discussion). Income also has little demonstrated correlation with racial attitudes, once educational attainment is held constant (Sears et al., 1997).
3. Analysis
3.1 Support for Spending on Blacks and Support for Spending on Cities
Table 2 presents the results of four logit regressions. In models 1 and 3, the dependent variable is coded 1 if a respondent said the nation spends “too little” on “solving the problems of big cities” or “assistance to cities”, and zero otherwise. In models 2 and 4, the dependent variable is coded 1 if the respondent said the country spends “too little” on “welfare” or “assistance to the poor”. The regression output thus reveals the odds that a person favours increased federal spending for cities or the poor. All regressions include fixed effects for each year. The controls are, again, dichotomous variables for gender, conservatism, being age 65 or older, having a college education, living in the South and living in a central city. I use dichotomous variables primarily for ease of interpretation; the results do not change substantially if the variables are left continuous. The results are also unchanged if the central-city variable is restricted to include only those respondents who live in the central city of one of the country’s 10 largest metropolitan areas.
Logit regressions of support for increased spending on cities and the poor
Notes: * p<0.05; ** p<0.01; *** p<0.001. Standard errors in parentheses, all regressions include year fixed effects. Samples are probability-weighted to correct for sampling error.
There are two independent variables of interest. The first is a binary variable indicating whether a respondent is Black. If Americans tend to associate cities with African Americans, then Blacks should be more willing to spend money on cities, all else equal. The second is a variable indicating whether the respondent thinks the government already spends “too much” helping Blacks. (This is a recoded version of the survey question about federal spending on Blacks.) If Americans tend to associate cities and poverty with African Americans, then opposition to spending on Blacks should, even in the presence of controls, also lead people to oppose spending on cities and the poor.
The regressions lend support to both hypotheses. Blacks are far more likely than others to support increased spending on both cities and the poor. The odds that a Black person will favour increased spending on “solving the problems of big cities” are 60 per cent higher than the odds that a non-Black will, and the odds that a Black person will favour increased spending on “welfare” are 150 per cent the odds that a non-Black will. These effects are even more pronounced when alternative question wordings are employed: the odds that an African American will favour increased “assistance to cities” are twice as large as the odds that a non-Black will favour such an increase, and the odds that a Black American will favour increased “assistance to the poor” are three times the odds that a non-Black will. These two findings—that Blacks are much more likely to support social spending than non-Blacks, and that the effect is magnified by the more neutral question wording—are present in every regression in the article.
Models 1 and 3 further suggest that people who oppose current spending on Blacks also oppose increased spending on cities. A person who believes that the government already spends too much helping Blacks has odds of supporting increased aid to cities that are about 50 per cent lower than a person who believes the government spends “too little” or “about the right amount” on Blacks. Models 2 and 4 show that these effects are roughly equal in magnitude to the racial effect on opinions about person-based assistance. People who oppose current spending on Blacks have odds of supporting increased aid to the poor that are 45–55 per cent lower than other people.
To help clarify these results, the bottom rows of the table convert the coefficients into probabilities. The “probability shifts” compare the likelihood that two hypothetical people will support increased spending on cities or the poor. The first stylised respondent is a non-Black male under age 65 who has not finished college, lives in neither the South nor a central city, and does not vote conservatively. He also does not believe the government spends too much helping Blacks. The second stylised respondent is exactly the same, but does believe the government already spends too much helping Blacks. Creating such hypothetical individuals, who are alike in every way except for the independent variable in question, is a common approach in statistical simulation, as it allows researchers to isolate the effect of that variable (for an overview, see King et al., 2000; Kinder and Kam, 2010, p. 83). In this case, then, the probability shift isolates the effect of attitudes about spending on Blacks have on attitudes towards spending on cities.
The effect is substantial, resulting in a 10–20 percentage point change in the probability that a respondent favours increased spending. In models 1 and 2, for example, the respondent who does not believe the US already spends too much helping Blacks has a 54 per cent chance of supporting increased spending on cities and a 24 per cent chance of supporting increased spending on welfare. His counterpart who thinks that the US already devotes too many resources to Blacks, however, has only a 34 per cent chance of supporting increased spending on cities and only a 13 per cent chance of supporting increased spending on welfare. Expressed as percentage changes, opposition to spending on Blacks reduces the likelihood of supporting increased aid to cities by about 40 per cent. The results are less consistent with respect to person-based aid: opposition to assisting Blacks reduces the probability of a stylised respondent supporting “welfare” by 45 per cent, but “assistance to the poor” by only 19 per cent—a testament to the apparently toxic (and racially suggestive) power of the word “welfare”.
These results are robust to randomly adding and dropping different years, as well as to including an income variable (although non-response means that including income greatly reduces the sample size). They are also robust to substituting a different question about helping Blacks as an independent variable. The regression outcomes in sum suggest that Americans do in fact associate helping cities with helping Blacks, and that the size of this racial association is comparable with the association between helping Blacks and support for person-based assistance to the poor.
3.2 Sources of the Racial Association
It is worth examining whether results for the 35-year period conceal significant temporal variation. Racial attitudes have liberalised considerably since the 1970s (Bobo and Charles, 2009) and their influence on support for social policy may have changed as well. Re-estimating the regressions for subsets of years suggests this is the case. If the regression examining support for “solving the problems of the big cities” is confined to 1973–77, the coefficient associated with the belief that America already spends too much on Blacks is essentially unchanged (0.74). After 2000, however, it shrinks dramatically, to 0.41. Similarly, when the question analysing support for welfare is re-analysed using only the 1970s responses, the coefficient is unchanged. Yet after 2000 the racial attitude coefficient grows—almost doubles—to 0.99. 6 Thus after 2000 the racial association with cities appears to moderate, while the association with poverty appears to grow.
This divergence suggests that the racial association with attitudes about helping cities and helping the poor do not spring entirely from the same source. Fainstein and Judd (1999), quoted earlier in this article, suggested that Americans mentally connected Blacks, cities, crime and poverty. Crime is a spatial phenomenon and began to fall dramatically in the 1990s. Hence it is possible that attitudes about crime correlate more powerfully with cities than with person-based assistance to the poor. Similarly, mobility strategies became a more prominent aspect of federal anti-poverty policy in the late 1990s and early 2000s, particularly with the implementation of the federal Section 8 housing voucher programme and the smaller albeit more ambitious Moving to Opportunity programme (Briggs et al., 2010). An emphasis on mobility for the poor may therefore have increased racial associations with anti-poverty policy.
The GSS offers only indirect ways to test these hypotheses. From 1990–2008, GSS respondents have been asked two variations of a question about crime: a random sample of respondents is asked if the nation should spend more on “halting the rising crime rate”, while the remainder are asked if the nation should spend more on “law enforcement”. The first variation of this question is more provocative, as it mentions “crime” and makes the assertion (not always true) that crime is rising.
In addition, beginning in 1990, survey respondents were asked, on a 1–5 scale, if they favoured or opposed living in a half-Black neighbourhood, with 1 being “strongly in favor” and 5 being “strongly opposed”.
We can hypothesise the following: first, that attitudes about fighting crime will have a stronger association with urban policy than they will with person-based redistribution. Secondly, the attitudinal association between cities and crime will be weaker after 2000 than between 1990 and 2000. Thirdly, attitudes about “crime” will have a more powerful association with attitudes about cities than will attitudes about the more neutral “law enforcement”. Fourthly, attitudes about racial integration will correlate more strongly with attitudes about person-based redistribution than with attitudes about helping cities.
Simple correlations lend some support to these ideas. Between 1990 and 2000, the coefficient of correlation between “halting the rising crime rate” and “solving the problems of big cities” is 0.32, while the correlation between slowing crime and “improving the condition of Blacks” is 0.21. During that same period, however, the correlation between “assistance to cities” and “law enforcement” was only 0.07 and the correlation between law enforcement and “assistance to Blacks” was only 0.05. In survey responses from 2000 on, the links between attitudes about cities, Blacks and crime weaken considerably. The correlation between crime and cities falls to 0.15 and between crime and Blacks to 0.15.
Table 3 presents a series of logit regressions that make these relationships more explicit. These regressions, like those earlier, examine the willingness to help cities and the poor, and account for differences in question wording. Each question, however, is also analysed separately for the periods 1990–99 and 2000–08. There are two independent variables of interest. The first is a binary variable indicating if the respondent thinks the US should spend more money on “crime” or “law enforcement”. The second is a binary variable indicating if the respondent is opposed to living in a half-Black neighbourhood. (This is a recoded version of the integration question described earlier—with a ‘1’ indicating that the respondent “opposed” or “strongly opposed” residential integration.)
Sources of evolving attitudes toward support for cities and the poor
Notes: Standard errors in parentheses. All regressions include year fixed effects. All samples probability-weighted to correct for sampling error. In original question wording, “wants more spending to fight crime” is response to a question about “halting the rising crime rate”. In alternative question wording, it is a response to a question about “law enforcement”. Probability shifts refer to effect of change in opinion about racial integration.
Source: GSS, 1990–2008
Attitudes about spending on “law enforcement” have no effect in any of the equations. Attitudes about “halting the rising crime rate”, by contrast, are strongly associated with support for helping cities, but after 2000 the association is only half the size it was the decade before. Attitudes about crime have no statistically significant association with support for helping the poor.
Opposition to racial integration is more consistently and powerfully associated with person-based policy than with place-based aid to cities, particularly after 2000. In the 1990s, a stylised White suburban male who opposed living in a half-Black neighbourhood was 10 percentage points less likely to support more spending on “solving the problems of the big cities”. In the 2000s, however, the coefficients on questions about helping cities are statistically insignificant and the associated probability shifts are small. With spending for the poor, the story is different. The coefficients associated with racial integration are, in the 2000s, large and statistically significant and are associated with a 5–10 percentage point reduction in the likelihood that a stylised respondent would support increased spending on the poor. The effect of “welfare” is, again, much stronger than the effect of “assistance to the poor”.
Hence, people who want to devote more resources to fighting crime also want to spend more solving the problems of cities, and these attitudes vary closely with attitudes about helping Blacks. They vary less closely, however, with attitudes about helping the poor.
3.3 Attitudes about Blacks and Attitudes about Urban Policy
The equations presented earlier only measure the impact of racial animus on support for assistance to cities and the poor. Yet concepts can be racialised even in the absence of racial animus. People who have progressive views about African Americans might nevertheless associate them with cities or poverty. Moreover, different racial views may have asymmetrical effects: the increase in support for social policy associated with positive feelings toward Blacks might be larger than the decrease in support for social policy associated with negative feelings toward Blacks.
To test this idea, I introduce independent variables drawn from four questions that more directly measure attitudes about African Americans. The GSS asked these questions consistently from 1983 to 2008. All four questions begin with the same preamble On the average, Blacks have worse jobs, income, and housing than White people. Do you think these differences are:
Respondents are then presented with four options
— Because most Blacks have less inborn ability to learn?
— Because most Blacks just do not have the motivation or willpower to pull themselves up out of poverty?
— Mainly due to discrimination?
— Because most Blacks do not have the chance for education that it takes to rise out of poverty?
These questions are designed to elicit different racial attitudes. The first question measures ‘Jim Crow’ or ‘old-fashioned’ racism—the belief that Blacks are genetically inferior to Whites. The influence this belief would have on support for social spending is theoretically ambiguous. Considering some people innately inferior might lead to less support for helping them (if such support would be futile, or if inferior status made them unworthy of aid) but might also lead to greater support (if there is an obligation to help those who cannot help themselves). The second question measures ‘laissez-faire’ or ‘symbolic’ racism—the belief that most obstacles have been removed from Black progress, but that Blacks prefer to rely on the government rather than work hard to get ahead (Bobo and Smith, 1995; Sears et al., 1997). People holding this belief should be more likely to oppose programmes they believe will help Blacks; if a group is capable but lazy, government should not intervene on its behalf. The third and fourth questions, by contrast, measure racial progressivity: the belief that low Black socioeconomic status (SES) results primarily from continued prejudice and/or structural obstacles that are the legacy of institutionalised racism. People who hold progressive beliefs, and who believe as a result that Blacks continue to face large obstacles through no fault of their own, should be more likely to support social programmes that disproportionately help Blacks.
Figure 3 shows probability-weighted trends in the responses to these questions. The share of Americans stating that Blacks have an inborn disability fell from over 20 per cent in the 1970s and 1980s to about 10 per cent in 2008. The belief that Black social problems stem from lack of will also declined, but remains more prevalent, and in fact in 2008 was the most common answer given. The other views have fluctuated over time.
American opinions about sources of African-American socioeconomic status.
All the responses, and particularly the progressive responses, have considerable overlap. In most years, over 65 per cent of people who said low Black SES was the result of discrimination also said that it was the result of restricted educational opportunities. Similarly, over 20 per cent of respondents who attributed low Black SES to lack of effort also said that Blacks had innate disabilities. About 10 per cent of respondents blamed both discrimination and an inborn disability, a combination that might at first seem incongruent. It is not inconsistent, however, to believe that a group is innately inferior and that it suffers discrimination for being so.
Table 4 presents logit regression results where responses to these questions are used as independent variables. The dependent variables, as was the case in the prior regressions, are the willingness to spend more assisting cities or the poor. The first models evaluate only negative racial attitudes. ‘Old-fashioned’ racism, which holds that Blacks are genetically inferior, has no statistical or practical association with attitudes towards social policy. The belief that Blacks lack willpower, by contrast, is consistently associated with lower odds of supporting both urban policy and redistribution. The odds of supporting increased federal spending on cities are 17 to 20 per cent lower for an individual who thinks that Blacks lack willpower. For spending on the poor, the odds are 17–30 per cent lower. The final row of the table again shows changes in probabilities. The probability shifts here are relatively small: believing that Blacks refuse to work hard reduces a stylised respondent’s chances of supporting increased spending on urban policy by between 3 and 6 percentage points, and reduces his support for further helping the poor by between 1 and 6 percentage points.
Logit regressions examining how opinions on Black socioeconomic status affect willingness to spend on cities and the poor, 1983–2008
Notes: Standard errors in parentheses. All regressions include year fixed effects. Samples are probability-weighted to correct for sampling error.
Source: GSS 1983–2008.
When progressive racial attitudes are introduced into the models, however, the results change. Believing that low Black SES is a result of discrimination or unequal opportunity is associated with substantially increased support for spending on both cities and the poor. In addition, the coefficients associated with Black laziness shrink dramatically and are now statistically significant at only the 80 per cent level. Across both topics (cities and the poor) and both question wordings, belief that low Black SES is a result of discrimination and unequal education is associated with a 10 percentage point increase in the probability that a stylised respondent supports further social spending.
These results are robust to a variety of alternative specifications. Adding an income variable reduces the sample size but does not change the results; the income variable is statistically insignificant. Substituting different independent variables that ask about the lower socioeconomic status of Blacks also does not materially change the results. Importantly, these results do not change when the years are restricted to 2000–08.
4. Discussion and Conclusion
This article has shown strong and consistent evidence that race permeates American attitudes about cities, and that beliefs about African Americans influence the willingness to support place-based urban policy. Thus, the article confirms an idea that has been long asserted but rarely tested. The article also shows that the racial association with support for helping cities is comparable in size to the racial association with support for person-based assistance to the poor, even if the two effects spring in part from different sources.
These findings discourage the idea that policy-makers can dodge racial obstacles to social policy by investing in places rather than people; both cities and poverty appear to elicit strong racial associations. One possibility is to ‘reframe’ urban and anti-poverty policies—to present them in a manner that no longer taps into prejudices. Goetz (2006), for example, cites survey evidence showing that ‘lifecycle housing’ is more popular among suburbanites than the racially tinged term ‘affordable housing’. However, he also notes the weakness in this strategy Any such [replacement] term might have a finite shelf-life, after which planners and advocates will have to search for the next acceptable label (Goetz, 2006, p. 228).
The evidence presented here reinforces Goetz’s concern. “Welfare” is certainly less popular than “assistance to the poor” but both elicit strong racial associations, even from people who are not antagonistic towards Blacks. The same is true of differently worded phrases for urban policy. Moreover, no group can completely dominate the discourse about controversial policies, so any carefully crafted agenda for “assisting the poor” could be undone by an opponent framing it as “welfare”. It seems there is no escaping race when it comes to making social policy, regardless of the wording used and regardless of whether the issue is framed as one of people or one of places.
It also seems, however, that progressive racial views are associated with greater support for social policy than prejudiced views are associated with opposition to it. It is tempting, in light of this fact, to suggest that policy-makers focus less on reframing ‘cities’ and more on improving attitudes about African Americans. Yet this is no small task. If racial attitudes continue to liberalise, urban and social policy might well become more popular, but the process of racial liberalisation has been labourious and painful, and policy-makers have few direct levers to instill tolerance and racial progressivism.
Perhaps a more promising approach is to make plain the extent to which all Americans, including affluent White suburbanites, benefit from government social programmes. Perceptions of American social policy are distorted both because many White Americans believe the majority of the poor are non-White (they are not) and because programmes that benefit the affluent—such as employer-paid health insurance subsidies and the mortgage interest tax deduction—are hidden as tax expenditures, rather than visible in the budget as direct outlays (Mettler, 2010). These tax expenditures dwarf spending on welfare and aid to cities, but they are also largely unseen. The affluent majority that resents government assistance for poor minorities might shift its opinions if it understood the degree to which the government helps it as well.
Footnotes
Acknowledgements
The author would like to thank Alex Beata, Matt Drennan, Tom Kemeny, Erzo Luttmer, Amanda Sheely and Don Treiman for valuable help and comments.
Notes
Funding Statement
The UCLA Lewis Center for Regional Policy Studies generously funded a portion of this research.