Labor-Market Inequality between Blacks and Whites: How Demography, Geography, and History Matter
The presence of racial inequality in labor-market outcomes is well-known.1 Among primary household earners in the U.S. in 2000, for example, blacks took home on average about 72 percent as much as whites.2 But the amount of inequality varies across demographic groups and between cities, and the variation can be substantial. For example, married black women are employed at far higher rates and earn more on average than their white counterparts, whereas married black men receive lower wages on average than comparable whites. Single black males are employed at much lower rates than comparable whites; employment rates are more similar for married men. The more education, the greater the wage gap between otherwise similar black and white men. And wages in Detroit and Milwaukee are much closer to racial parity than wages in Atlanta or Washington, D.C., particularly for men. Yet the former rank among the worst and the latter among the best cities for blacks to live, according to a recent survey by Black Enterprise magazine.3
These variations occur because the degree of labor-market equality is the result of household optimization rather than the objective. I outline below a basic model of wage determination. I then estimate black-white differences in employment rates and wages for various groups using a Heckman two-step approach on microdata from census year 2000, augmented with city-specific information on residential segregation, the proportion of blacks in the population, urbanization rates, population size, union membership, public school quality, public safety, and black-white incarceration rates.4 I suggest that rational household decisions, coupled with historical patterns of neighborhood formation, are responsible for observed differences in labor-market inequality between whites and blacks across demographic and geographic subgroups.5 Awareness of the empirical patterns and the underlying reasons for them is essential for evaluating policies to address racial inequities in the labor market.
Consider a simple model of the labor market, where wages are a function of supply and demand factors. Utility-maximizing individuals supply labor according to their preferences and the constraints they face; profit-maximizing firms generate labor demand as a function of relative prices and productivity. In short,
w = w + wX + w, (1)
where w is typically cast as the natural log of wages and X denotes a vector of individual and firm characteristics.6
Public goods could also appear in X. People might be willing to accept lower wages in exchange for a better public-school system or a lower crime rate, for instance. These relationships are complicated, however, because higher-income neighborhoods can also afford to spend more on local public goods. The sign of a coefficient on a public-good variable in (1) is therefore ambiguous.
Of particular interest is the effect of race on wages. A dummy variable for race could act as a proxy for omitted variables related to preferences of workers or informational signals about productivity, among other things, or it could pick up the presence of labor-market discrimination. Including additional proxies or interactive variables may help differentiate among explanations for labor-market inequality, or for variations in labor-market inequality across subgroups.
One strand of the literature on inequality estimates the effect on wages associated with the interaction of race and residential segregation. Using 1990 census data on individuals aged 20 to 30 years old, Cutler and Glaeser (1997) found that segregation affected blacks negatively but had little effect on whites.7
Spatial mismatch between housing and jobs could explain why blacks living in a segregated area might earn less relative to whites, all else constant, than blacks living in a more integrated area. If some neighborhoods are located far from attractive, high-paying jobs, residents of these areas may rationally choose lower-paying jobs closer to home, or they may drop out of the labor force entirely.8 Suppose, for example, that blacks and whites live in different neighborhoods and jobs are located primarily in the white neighborhood. Because transportation is costly, we might expect whites to earn higher wages and blacks lower wages in this world as compared to one in which jobs are spread about evenly or people of different races live in integrated neighborhoods. This would imply a negative coefficient on an interactive variable “DIblack” (equal to the degree of residential segregation – or “dissimilarity index” -- times a dummy variable equaling one for black persons).
Other influences could also yield a negative coefficient on “DIblack.” Residential segregation may act as a proxy for labor-market discrimination. Lack of contact between races could lead to statistical discrimination among employers, for example.9 Or greater residential segregation could imply fewer positive role models for blacks.10
Theory is ambiguous as to the expected sign on “DIblack,” however. Preferences could generate a positive coefficient on the interactive variable. This might occur if blacks prefer to live in a racially mixed neighborhood but would be willing to accept more residential segregation if they also receive relatively higher income.11
Social networks might matter as well. Neighborhoods could yield a sense of community, thriving internal markets, or superior job contacts.12 These effects may differ across races.13 An observed positive coefficient on “DIblack” could reflect better availability and use of social networks for blacks relative to whites in more-segregated cities; an observed negative coefficient could indicate the reverse. Notably, slight differences in initial conditions could lead to large differences in outcomes – if network members have relatively worse starting positions, remaining in a network has comparatively fewer benefits, which could lead to greater network (and labor-market) dropout.14
A city is simply a neighborhood writ large. Just as residential segregation could affect contact between races, role-model availability, and social networking, so might the proportion of black persons within a given Metropolitan Statistical Area (MSA).15 A larger black presence within a city could imply greater familiarity between the races or greater networking opportunities for blacks. This suggests a positive coefficient on an interactive variable “propbblack” (equal to the proportion of city population that is black times a dummy variable equaling one for black persons).
Again, theory is ambiguous. A “blacker” city might also imply greater antagonism between races. Alternatively, non-labor-market benefits for individual blacks associated with a larger percentage of blacks in an MSA could yield a negative coefficient on “propbblack.” That is, blacks might be willing to accept more labor-market inequality if they also have family members close by, a strong religious community, or less of a sense of isolation.
Wage rates are not the only labor-market outcome that matters to people, of course – having a job is a prerequisite to earning wages at all. Spatial mismatch, discrimination, social networks, and tradeoffs could affect relative employment rates as well as wage rates. Including “DIblack” and “propbblack” variables in regressions using employment status as the dependent variable could indicate whether these forces influence the relative probability of employment as well as relative wage.16
One final point: interrelationships among residential segregation, urban sprawl, public transportation, and car ownership can complicate matters. Glaeser and Kahn (2003) suggest, for instance, that cities with more sprawl may actually have greater integration yet lower well-being for poor people – particularly poor blacks -- because they cannot afford cars. Stoll et al. (2000) note that many low-skilled jobs in metropolitan areas simply cannot be reached via public transportation. Fortunately, my data allow me to control for both the degree of sprawl in the city of residence and vehicle ownership by the household.17
Race and Region of Residence
Patterns of migration throughout the last century, as well as differences in the history of the South and the North, suggest that distinct relationships among residential segregation, black presence in an MSA, and labor-market outcomes might emerge across regions. Ninety percent of blacks lived in the South at the turn of the century, and more than half remained there as of the year 2000.18 Mass migration northward, typically to urban areas, occurred after both World Wars. Movement north slowed once the civil rights revolution took hold, and migration to and from metro areas has varied across regions over the last decade.19 Most recently, blacks (particularly college graduates) have begun to move back to the South.20
The reasons for migration and neighborhood formation patterns are complex; they include the desire to live near family and friends, different regional conditions for certain types of workers, and institutional constraints such as redlining. Moreover, the relatively longer history in the South of large numbers of blacks living alongside whites could imply greater (or lesser) comparability in the labor market. The huge shifts in the black population over the past several decades thus may mean that cross-sectional relationships among relative employment and wage rates, housing segregation by race, and proportion of MSA residents who are black could differ substantially across regions.
Race, Sex, and Marital Status
Both theory and empirical evidence suggest that sex and marital status matter for labor-force outcomes.21 Market outcomes may differ for women and men because of differences in non-market activities (housework and childcare, for example), preferences, employer attitudes, or social networks. Occupational segregation might also play a role. Mate selection could mean that married people differ fundamentally from single people: people who are less attractive in the labor market may also be less attractive in the marriage market. What is more, married people arguably make decisions jointly whereas single persons enjoy more autonomy.
One question is whether these factors matter in considering the effects of race in the labor market.22 Suppose, for example, that more segregation in housing tends to affect black men negatively via greater spatial separation from jobs or inferior networks. Black men able to overcome these obstacles could be more attractive in the marriage market. We might therefore observe different coefficients – even different signs -- on “DIblack” for single and married men.
Or consider this possibility: suppose greater residential segregation or a larger proportion of blacks in a city tends to reduce employment or wage rates for black men relative to white men. To counteract the expected negative relative effect on family income, married black women might seek better-paying jobs than comparable white wives as housing segregation or MSA black presence increases.23 We could therefore observe opposite signs on a “DIblack” or “propbblack” coefficient for married men and married women.
Residential segregation and black presence in an MSA could affect women and men differently in a more general way. Black women might enjoy relatively better job prospects in segregated (or heavily black) cities if they also form stronger networks in these cities, for example, whereas black men might suffer more from spatial separation between home and work in highly segregated areas. The degree to which whites feel threatened by black population concentration might play out differently for men and women in the labor market. Or stronger family ties for, say, women, in a predominently black community might make seeking better job opportunities elsewhere more costly for women than for men, thus generating different coefficients on a “propbblack” variable for males and females.24
Race, Education, Age, and Occupational Category
Education, age, and occupational category are typically included in the vector X in equation (1), often as proxies for productivity. Yet focusing on inequality across educational backgrounds, age groups, and occupational categories could offer useful additional information.
Suppose college graduates experience a larger racial wage gap than high-school graduates, for instance. This might indicate diverse degrees of discrimination for the two groups, or it could imply something about the perceived relative quality of schools attended. City characteristics could interact with educational status as well, perhaps because schooling level helps determine opportunities for migration.
Likewise, differences in labor-market inequality for people of different ages could emerge. Suppose we observe greater racial inequality in wages for older individuals. This could indicate bigger pre-market racial differences – for example, in the relative quality of schools attended – for the older people in a cross-section.25 Or it might imply that blacks enjoy fewer advancement prospects or greater discrimination on the job as they age. Age could interact with city characteristics as well. For instance, residential segregation could have different labor-market impacts for younger and older individuals because of generational differences in the formation of social networks.
Inequality might differ as well across broad occupational categories. Employers might be less able to discriminate in heavily unionized occupations, for example, or social networks might be better established for whites in the professions. But inequality within a broad occupational category might also reflect occupational segregation, so an investigation of specific occupations by race could be fruitful.
Correcting for Selection Bias
In moving from model to estimation, recall that employment is the essential prerequisite to earning wages. Researchers can only observe actual wages, as Heckman (1979) pointed out in his seminal research, so appropriate estimation includes a variable in equation (1) to capture potential selection bias:26
w = + X + + (2)
Although Cutler and Glaeser (1997) analyze the degree of “idleness” present in different races, they do not explicitly account for selection issues in their wage analysis. Instead, they simply exclude individuals who earned nothing. Large racial differences in employment rates suggest at least two things: selection may matter in estimating the racial wage gap, and race-related variables such as residential segregation and percent of the population that is black could influence relative employment rates as well as wages.27
Individual and household data are from the census-2000 5-percent public use micro samples organized at the Minnesota Population Center.28 I consider only people who identify themselves as black or as white. Although cross-racial differences for other groups are interesting and important, they are beyond the scope of this study.
By focusing my analysis on household heads and their spouses, I also exclude potential wage earners with relatively less attachment to the labor force: children, and adults living with their parents or other adults.29 Some information for these possible earners remains, because I include a measure of other household income in employment and wage regressions. My results nevertheless pertain only to primary wage earners. Still, it is these individuals who are chiefly responsible for the wellbeing of the household, so my findings on racial inequality for household heads and spouses say something about racial inequality across families as well.
Information on MSA total and black populations, residential segregation as measured by dissimilarity indices (DI), and degree of urban sprawl comes from the U.S. Census Bureau.30 The total number of MSAs for which these data are available is 206. I obtained MSA unionization rates from the Bureau of Labor Statistics.31 For a smaller group of 44 MSAs, I also obtain median public school quality, median public safety rating, and the statewide relative black-white incarceration rate for men.32
Table 1 shows average annual earnings, weeks and usual hours worked, and age for employed urban residents by sex, race, and marital status.33 It also gives average employment rates. Although these figures do not control for underlying differences in individual and MSA characteristics, they are nonetheless striking. Despite comparable mean ages, hours worked, and weeks worked, black male household heads earn only two-thirds as much as white men.34 Black women are employed at much higher rates than white women; married women earn about the same but single black female household heads earn only 82 percent as much as their white counterparts.35
Table 2 reports averages for several variables associated with employed urban household heads and their spouses by sex, race, and region of residence. Black-white ratios by region of these averages reveal several interesting (though not always surprising) results, as do more complex ratios measuring regional differences. Blacks are far more likely to be responsible for grandchildren and to live in the central city, particularly outside the South. Blacks have relatively less education on average, smaller amounts of other household income, and fewer vehicles. A greater proportion of blacks report a work disability, a much larger proportion of blacks work in service or laborer occupations, and a larger proportion of blacks (particularly females outside the South) have less than a high-school education. Blacks (particularly women) work relatively more in public-sector jobs.36
Table 2 offers only an “average” picture. Here are some additional details pertaining to MSAs in 2000: the proportion black (propb) in an MSA ranged from 0.5 percent to 26.3 percent outside the South and from 0.5 percent to 45.9 percent in the South. The degree of segregation ranged from 0.198 to 0.846 outside the South and 0.359 to 0.992 in the South. Of the ten MSAs with the largest black population, only one – Dallas – had a DI less than 0.6, which is the threshold commonly used to denote “highly segregated.”37 The five most segregated urban areas lie outside the South: Detroit, Milwaukee, New York-Northern New Jersey, Chicago-Gary, and Cleveland-Akron.38
One individually based statistic is worth noting: the well-known difference in marriage rates between blacks and whites.39 Table 2 reports figures for employed persons. Among all urban male household heads, 29 percent of whites and 46 percent of blacks were single in 2000. The discrepancy for females is even larger: 34 percent of whites and 67 percent of blacks were single. The percent single is slightly smaller in the South for whites and 6 to 7 percentage points smaller for blacks.
Following Oaxaca (1973), decompositions can help determine the extent to which diverse underlying characteristics explain wage gaps.40 Table 3 lists variables included; these variables are also used in the race-combined regressions discussed below.41
Using regressions for whites as the reference, I estimated the portion of the wage gap attributable to racial differences in observed characteristics. About 95 percent of the gap for single females in 2000 is explained by differences in observed characteristics, 87 percent for single males, and only 74 percent for married males. Married black females in the sample used for the decomposition actually earned slightly more on average than their white counterparts; 48 percent of the tiny difference is explained by disparate observed characteristics.
These results suggest a reasonable next step: combining the races for regression analysis and attempting to ascertain possible sources for the unexplained portion of the wage gaps. I analyze results for both sexes; the inequality investigation for males is arguably more relevant because the gaps for men are either larger or less well-explained than those for women. Recall, however, that “explained” differences could still result from discrimination (perhaps pre-market or non-labor-market) and “unexplained” differences might be due to unmeasured taste or ability dissimilarities.42
The following sections outline my findings when the data are partitioned according to marital status, educational background, and city of residence. Table 4 summarizes the signs of coefficients on proportion black and residential segregation by region for stage 1 and stage 2 regressions.43 I briefly discuss below the results associated with age groups and occupational classifications as well.
Table 5 shows average employment and wage gaps between whites and blacks, separately by sex and marital status. At the mean, for example, single black women in the South were employed at nearly the same rate as their white counterparts; outside the South, single women achieved virtual parity in wages.44 More striking, however, are these four findings: (1) single black male household heads were employed at much lower rates than comparable whites (particularly outside the South) whereas employment rates for married male household heads are more similar across races, (2) married black women were employed at much higher rates than comparable whites, (3) married black women out-earned similar married white women – at the mean, 6.1 percent more outside the South and 2.3 percent more in the South, and (4) black men earn up to 10 percent less on average than comparable white men and the relative wage gap was more than twice as large at the mean for married as compared to single male household heads.45
One very tentative explanation for these findings is this: Blacks, especially females, are much more likely to remain single than whites. When they do marry, black women in particular may select a certain (more employable) type of spouse. Alternatively, employers (who may be prejudiced or engage in statistical discrimination, especially against black men) might use marital status as a signal of responsibility or stability. Employment rates thus appear more comparable across race for married men than for single men. Yet black men generally earn less than white men, regardless of marital status. If similar black and white families have like aspirations and desires, a plausible response for black wives is to work and earn more than their white counterparts. But this relatively larger cushion of family income might in turn permit employed married black men to accept a relatively larger wage gap than employed single black men.46
Table 5 also offers regression coefficients for both stages separately by sex and marital status. Greater residential segregation reduced black employment rates relative to white rates for all groups except married females in the South.47 It also lowered black wages relative to white wages for single persons (zero effect for single males outside the South) and married Southern males, but raised relative wages for married black men outside the South and married black women generally.48 A larger proportion of blacks in the population raised relative employment rates for blacks in the North (or had zero effect in the case of single males) but lowered them in the South. It lowered relative wages for blacks or had zero effect, except for single females outside the South.49
These patterns suggest several things. Spatial mismatch, discrimination, and other factors that affect blacks negatively relative to whites manifest themselves primarily via lower relative employment rates. To the extent inferior networks affect relative outcomes, single blacks seem to suffer more. And black couples living outside the South appear to accept segregated neighborhoods in exchange for relatively higher wages. The regional difference in the effect of proportion black could imply that blacks gravitate to Northern cities with better employment prospects but are willing to stay in Southern cities for reasons other than labor-market benefits.50 Education
People over age 25 are split (separately by sex) into three groups in Table 6: those with at most a high-school degree, those with some college education, and those with at least a B.A. degree.51The averages show that, among males, the largest wage gap corresponds to those with the most education. At the mean, college-educated black men earned 10.6 percent less than comparable white men outside the South; in the South, the figure is 14.5 percent.52 The average wage gap is larger for men in the South across all educational backgrounds, but the average employment gap is much larger outside the South for men with a high-school degree or less. At the mean, Southern black men with some college and college-educated black men everywhere were employed at higher rates than comparable white men.53
Black women on average generally were employed at greater rates than white women, particularly in the South, and the gap increases with education. The least-educated black women earned on average 6.1 percent more than similar white women outside the South; the wage gap is near zero in the South.54 Interestingly, black women with some college education earned slightly more on average outside the South in 2000 but somewhat less on average in the South relative to comparable white women. College-educated black women earned slightly more on average than similar white women.55
The regression coefficients help us detect the sources of these gaps. Perhaps the most notable feature for Southern males is that the negative impact of residential segregation on relative employment and wages is generally stronger with greater educational attainment. Residential segregation actually had a positive effect on relative wages for the least-educated black men outside the South, although it had a negative impact on relative employment. For more-educated men outside the South, the most striking aspect of these regression coefficients is the negative effect of the proportion of blacks in the population -- on relative employment for all men with more than a high-school education and on relative wages for college-educated men. The only significant coefficient in the wage regression for non-Southern men with some college is the intercept term.56
A tentative explanation consistent with these patterns is this: Professional contacts yielding higher-paying jobs, as well as greater comfort with diversity, might arise in the South more easily in integrated residential areas. These contacts could be especially important for more-educated men given the significance of historically black colleges and universities in the South.57 That is, a lack of networks due to matriculation in different institutions could be more easily overcome if people are neighbors. Outside the South, segregated neighborhoods grew up around manufacturing hubs; these industries brought opportunities for less-educated men, black and white.58 The fortunes of both races rise and fall with economic prospects in these areas, leading to less inequality in wages for the less-educated employed here as compared to men living in less-segregated MSAs outside the South. But spatial mismatch might especially have affected the job prospects of less-educated black men left behind when plants moved out of the central city to the suburbs or overseas, or simply shut their doors, causing greater inequality in employment rates in highly segregated non-Southern MSAs.59
Explaining the “propb” coefficients for men outside the South is a bit more difficult. One possibility is that white-collar employers may find it easier to engage in discrimination in cities with a larger proportion of blacks. Another is that non-South MSAs with a greater proportion of educated blacks attract other educated blacks, despite some degree of labor-market inequality between races.
Perhaps the most intriguing pattern for women evident in Table 6 is that greater residential segregation implies higher relative employment for college-educated black women in the South and higher relative wages for these women outside the South, and higher relative wage rates for black women outside the South with at most a high-school degree. The latter parallels the experience for black men. I speculate that the former may be due in part to positive assortative mating plus family dynamics. More highly educated women tend to marry more highly educated men, and most people marry within their own race.60 We know that more-educated black men earn relatively less than their white counterparts, and that residential segregation widens both the employment and the wage gap for these men in the South. To attain a certain standard of living, then, more-educated black wives living in highly segregated MSAs would be more likely to seek employment and higher wages than similar white wives.
MSAs and Public Goods
This section focuses on 44 MSAs that contain about 21.1 million blacks out of a total 36 million residing in the U.S.61 The 44 include all cities that ranked high or low in several surveys conducted by Black Enterprise magazine. The magazine’s 2007 survey listed Washington (DC), Atlanta, Houston, Nashville, Dallas, Charlotte, Columbus (OH), Raleigh-Durham, Indianapolis, and Jacksonville as the best cities for blacks. The last three replaced Birmingham, Baltimore, and Memphis from the 2004 survey. Nashville and Columbus replaced Chicago and Philadelphia from the 2001 survey. Cleveland, Detroit, and Milwaukee rank as the three worst cities for blacks.62
In the tables below, I report below the results of regressions on the 44 cities together that include the expanded set of independent variables -- median public school quality for the MSA, median public safety rating, and the relative (black-to-white) state incarceration rate for men — and of separate regressions for each MSA.63 Characteristics such as median school quality and safety undoubtedly indicate something about the attractiveness of an MSA and interact with labor-market outcomes for people regardless of race. Because blacks are more likely to attend public school64 and live in more dangerous neighborhoods,65 however, these attributes may be particularly relevant for them. Specifically, we could observe a negative relationship between relative wages and higher-quality public goods if blacks are willing to trade off wage equity for other amenities. One salient community feature is race-specific: relative incarceration rates. Admittedly a crude measure of the “warmth” of racial relationships, nevertheless a large difference in incarceration rates by race could signal a disamenity for blacks. In other words, blacks who live in a state that disproportionately puts them in jail might require compensation in the form of higher wages.
Table 7 gives coefficients for regressions that include blacks and whites from all 44 cities; table 8 offers results for regressions on blacks only. As expected, greater safety corresponds to lower relative (but higher absolute) wages for blacks, and a larger relative incarceration rate yields higher relative and absolute wages for blacks. Better median school quality implies lower absolute wages for black men and lower relative (but higher absolute) wages for black women.66
Regional effects of residential segregation and black representation in the population are quite distinct. Greater housing segregation decreases absolute and relative employment rates everywhere, but it increases relative wages (and black male absolute wages) outside the South and decreases relative (and absolute) wages in the South. A larger proportion of blacks in the population increases relative employment outside the South and decreases it in the South, decreases relative wages everywhere, and increases black absolute wages outside the South (and black male absolute wages in the South). As before, these patterns suggest that spatial mismatch, network effects, and other factors negatively affecting blacks relative to whites take their toll mostly on employment rates, that blacks outside the South accept residential segregation if they also gain greater wage parity, and that non-wage benefits of living in a “blacker” city may counterbalance wage inequities for blacks.
Individual MSA regressions allow the ranking of MSAs according to the size of average employment and wage gaps. Table 9 shows that Atlanta, for example, ranks 38th for men and 30th for women on the wage gap – black men earn 14 percent less than comparable white men, and black women earn 1.9 percent less than comparable white women. These figures are 19.6 percent and 1.6 percent for those with at least a B.A. degree.67 Of the ten highest-ranked cities in the Black Enterprise survey for 2007, only Indianapolis and Nashville scored well in terms of wage equity in 2000.68 The formerly high-ranked cities of Baltimore, Chicago and Philadelphia also ranked fairly well, particularly for women.69 Houston, Dallas, Charlotte, and Atlanta had large wage gaps – on the order of 12 to 14 percent for men (20 to 25 percent for college-educated men) and 2 to 5 percent for women who were comparable aside from race.70 Houston and Jacksonville also had hefty employment gaps. In contrast, the lowest ranked MSAs scored very well in terms of wage equity, with black women earning 4 to 10 percent more than comparable white women and black men earning only 0 to 4 percent less than their white counterparts.71
Given these figures on wage parity, one might question the survey results. Yet cities, like people, are bundles of characteristics, and wage equity is only one of them. What is more, individual workers can do very little about their skin color and thus are more likely to seek the best living conditions available to them, ceteris paribus, than to minimize the racial wage gap. The gap itself certainly may matter to individuals, both as a matter of fairness and because it might act as a proxy for underlying social conditions; it is certainly a concern for policy makers. But wage gaps might usefully be considered an outcome of household optimization rather than part of the objective function. To the extent people have choices, we might expect to observe greater racial wage inequality in MSAs that offer more of other sorts of amenities – higher real wages, for example, or a higher probability of employment or greater equity in employment rates, or better public schools, or a larger percentage of blacks in the population, or a lower relative incarceration rate. How an MSA ranks to blacks therefore depends on how people value wage parity relative to these other qualities.72
Take another look at Table 9 and consider the Black Enterprise survey rankings. As well as ranking labor-market gaps, the table reports various MSA attributes that matter to people, including MSA ranking of real wages for blacks by sex. Cleveland does well on wage gaps but miserably on employment gaps. Milwaukee is at the very bottom of the employment gap ranking; it also has the largest relative incarceration rate aside from Washington, DC. Detroit scores well for real wages and is just below the middle of the employment-gap ranking, but it has the worst public schools and among the worst safety records. Detroit and Milwaukee both have highly segregated housing as well. Detroit and Cleveland are on worst-city list for many surveys, regardless of race.73 Despite good records on racial wage parity, these cities apparently offer little else to blacks.
What about the highly ranked cities? Atlanta, Washington, Dallas, and Nashville offer fairly high real wages for blacks. Atlanta and Charlotte rank well on the employment gap for both sexes, and Washington, Nashville, Raleigh, and Jacksonville rank highly for one sex. Raleigh has good public schools and safety. Charlotte and Raleigh often score high on surveys for livability, regardless of the race of respondents.74 Even though racial wage inequity (sometimes quite large) persists in the highly ranked cities, other features can reasonably make them attractive to blacks.
Some cities, by virtue of the statistics, seem to belong on the “best of” list for blacks – Boston and Los Angeles, for example – but are not cited. Yet Boston has a long-held reputation for racism, including virulent fights against busing to integrate schools, the hunt for the non-existent black killer in the Charles Stuart case, and Henry Louis Gates’s recent dustup with the police.75 Los Angeles is notorious for the 1992 Rodney King incident and subsequent race riots, as well as for purported racism within its police department.76 That these cities did not make the top ten in livability for black residents is perhaps not that surprising after all.
Segmenting the data by age group highlights some troubling patterns.77 Outside the South, the largest employment gap occurs among male household heads aged 25 and younger. Residential segregation appears to be a primary factor explaining this pattern, with an additional factor being skin color for single black men. The underlying source may be discrimination, lack of networks, or some other factor; whatever the reason, the size of the employment gap for the youngest black male household heads is notable – and alarming.
Residential segregation also contributes heavily to relatively low employment rates for single black men over age 40 outside the South, again perhaps indicating a lack of job contacts or role models. But more segregation also means higher relative wages for older black males outside the South – once employed, these men experience less wage inequality than similar men living in less-segregated areas. This suggests that black men in their prime working years outside the South face a difficult tradeoff – they can live in highly segregated areas and earn wages fairly comparable to those of whites if they can get a job, or they can live in less-segregated MSAs with more wage inequality but less employment inequality.
One noteworthy pattern associated with “propbblack” is its consistently negative coefficient in the employment regression for persons aged 26-40 living outside the South. For other age groups outside the South, the coefficient is positive or zero. All else constant, non-Southern blacks in the early to middle part of their working years experience much lower employment rates than comparable whites if they live in a “blacker” MSA.
Greater residential segregation in the South yields lower relative wages for blacks in nearly every age group; for one age group, its effect on employment stands out. Generally, Southern residence tends to mitigate the negative effect of residential segregation on relative employment. This is not true, however, for married men and all women of ages 26 to 40. What this may indicate is that old regional patterns may be eroding -- where residential segregation once took its toll primarily on relative employment outside the South and relative wages in the South, the impact may become less regionally distinct in the future.78 Occupation
Occupational categories used in the regressions are quite broad and mask possible occupational segregation as a possible source of wage inequity. As Altonji and Blank (1999, p. 3153) note, occupational segregation could be a form of discrimination.
Male professionals exhibit a large racial wage gap at the mean, for example -- 12.9 percent outside the South and 15.1 percent in the South.79 Thirty-five percent of white men fall into this occupational category, but only 22 percent of black men outside the South and 19 percent in the South. Some of the wage discrepancy is undoubtedly due to a divergence in the type of professional job held: a detailed analysis of occupational codes reveals that white male professionals are much more likely to be lawyers, scientists, engineers, and CEOs, whereas black male professionals are much more likely to be teachers, social workers, therapists, and postmasters.80 The wage gap for male administrators is likewise attributable in part to the specific occupations blacks and whites have: in this category, blacks are much more likely to be clerks, messengers, postal workers, meter readers, and licensed practical nurses, whereas whites are much more likely to be supervisors, computer software designers, pilots, surveyors, and sales engineers.81
For men, one notable feature in separate occupational wage regressions is the unique outcome for service workers.82 The signs on “propbblack” and “DIblack” are positive and negative, respectively, implying that relative wages for black male service workers are higher when a greater proportion of the population is black but lower when housing is more segregated, ceteris paribus. The opposite is true for all other occupational classifications outside the South and for skilled workers in the South. Residential segregation also has a strong negative effect on relative employment rates for black service workers outside the South.83
How can we explain these patterns? As with other occupational groups, white and black service workers hold distinct jobs. Blacks are more likely than whites to be watchmen, nursing aides, housekeepers and butlers, child care workers, and baggage porters, and less likely to be policemen and firefighters. Roughly speaking, the former earn less than the latter.84 History, plus social and professional networks, may help explain the occupational crowding and thus the regression coefficients. Blacks have long worked in the former group of occupations, and openings for particular jobs could easily become known by word of mouth. So a greater proportion of blacks in the MSA population could reasonably strengthen the networks available to individual blacks for finding these sorts of jobs and therefore lessen inequality. Residential segregation, on the other hand, reduces daily contact between the races; discord thus might more easily arise when the races do meet. Whites in segregated cities might be particularly hostile to entrusting public protection to blacks. In fact, racial tension about hiring and promotion among police and firefighters in highly segregated Northern cities is commonplace.85 Consequently, more residential segregation could well generate greater racial inequality among service workers.
Racial discrepancies within occupational groups are not as notable for women as for men. More than 90 percent of female workers are in the administrative, professional, or service occupations.86 Blacks are disproportionately represented as licensed practical nurses, mail and postal clerks, social workers, vocational counselors, nursing aides, and guards. Whites work disproportionately as real estate sales workers, dental hygienists and assistants, advertisers, CEOs, therapists, waitresses, and hairdressers.
One interesting result, however, is that greater residential segregation lowers relative wages for black female professionals outside the South but raises them for service workers outside the South. The opposite is true in the South. The effects are not large, compared to those for men, except for female Southern service workers. For these women, a greater proportion black also implies lower relative wages. What this may suggest is that black Southern women in service-oriented jobs stay in segregated neighborhoods in heavily black MSAs – and tolerate labor-market inequality – for reasons outside the workplace. Moving could be difficult financially, or amenities like nearby family and friends may compensate for job inequities.