Sources of Racial Wage Inequality in Metropolitan Labor Markets: Racial, Ethnic, and Gender Differences

by Leslie McCall
Sources of Racial Wage Inequality in Metropolitan Labor Markets: Racial, Ethnic, and Gender Differences
Leslie McCall
American Sociological Review
Start Page: 
End Page: 
Select license: 
Select License


Rutgers University

Research on racial inequality has become increasingly specialized, often focusing on

a single explanation and subgroup of the population. In a diverse society, a broader

comparative framework for interpreting the causes of wage inequality for different

racial, ethnic, and gender groups is called for. The effects of a range of different factors on the wages of Latinos, Asians, and blacks, relative to whites and sepa-

rately for women and men, are examined. New sources of racial wage inequality are

also considered. Significant differences are found in the sources of wage inequality

across race, ethnicity, and gender. Differences are generally greater between racial

and ethnic groups than between men and women. Key findings include a large nega-

tive effect of immigration on the relative wages of Latinos and Asians and only a

small effect on the relative wages of black women (and no effect on black men). In

contrast, the relative wages of blacks remain most affected positively by the presence

of manufacturing employment and unions. New economy indicators of high-skill

services and flexible employment conditions play only a secondary role in explain-

ing metropolitan racial wage inequality.

ince the 1950s, there have been two strands of research on racial earnings inequality in metropolitan labor markets. One strand has examined the impact of a locality's racial composition on economic relations between racial groups, establishing the well-known relationship between black/ white inequality and the size and growth of the black population. Recently, scholars have extended this approach in considering how blacks have fared as a result of changes

Direct all correspondence to Leslie McCall, Department of Sociology, Rutgers University, Piscataway, NJ 08854 ( Early versions of this paper were presented at the 1999 and 2000 annual meetings of the Popula- tion Association of America, where I benefited from the comments of Frank Bean and Reeve Vanneman. I also thank Julie Phillips for helpful conversations, and the ASR Editors and review- ers for helpful comments. The support of the Russell Sage Foundation, where I was a visiting scholar, is gratefully acknowledged.

in the racial, ethnic, and immigrant compo- sition of the United States. The second strand has focused on the role of industrial structure, establishing a relationship be- tween local manufacturing employment and the economic status of blacks. Again, em- phasis has shifted in recent years to consider the impact of changes in the industrial struc- ture, especially the negative impact of deindustrialization on the situation of blacks in the central cities of the Rustbelt. Despite the evolution of research in this area, how- ever, most studies of wage inequality tend to take one or both of these directions. More- over, attention remains focused on black1 white inequality, especially in large-scale studies of metropolitan labor markets.

To better understand contemporary racial inequality in the context of growing racial and ethnic diversity and widespread eco- nomic change, I extend the current research in this area. First, I compare the effects of demographic and industrial factors on the

REVIEW,200 1, VOL.66 (AuG~~T:520-54

relative wages of Latinos and Asians as well as blacks. This requires a sort of meta-analy- sis of competing explanations of multiple forms of racial inequality, whereas the ten- dency in recent research has been to delve into the effect of a single factor, such as im- migration, on a specific subgroup of the population, such as young, low-skilled, na- tive-born blacks (Butcher 1998). In-depth analyses of this kind are valuable for laying the groundwork for comparative analysis, but they cannot tell us whether some factors are more important in fostering some forms of racial inequality but not others. For ex- ample, none of the research on immigration compares immigration's effect on blacks with its effect on Latinos or Asians, nor does it compare the impact of immigration with the impact of other factors. In statistical terms, single-explanation and single-group studies cannot determine whether a given factor remains significant when other forms of racial inequality are considered or other explanations are controlled. Although my approach sacrifices greater depth in the mea- surement of a single form or explanation of wage inequality, it provides for greater breadth in the number of explanations and racial and ethnic groups considered.

Second, I consider gender differences in the underlying sources of racial wage in- equality. Large-scale quantitative research on the intersection of gender and racial wage inequality is rare, especially in the scholar- ship on economic restructuring (but see Browne 1999). This may be due in part to the small absolute levels of wage inequality among women. But even in the existing quantitative research on women, the empha- sis has been on the human capital and indi- vidual-level determinants of wage inequal- ity (England, Christopher, and Reid 1999). Moreover, in those studies that do consider structural factors, some find gender differ- ences and others do not (Bound and Dresser 1999; Browne 1997). Such gender differ- ences could have important consequences for the formulation of anti-inequality poli- cies and for an understanding of the under- lying mechanisms that affect wage inequal- ity more generally (Baker 1999).

Third, I compare "newer" explanations of inequality, such as technological change and increasing flexibility, to "older" explanations (e.g., demographic and industrial fac- tors). Although these new factors cannot be separated entirely from the old factors,' ana- lytical distinctions can be made. For ex- ample, a key tenet of spatiallskill mismatch theory is that many of the new jobs in cen- tral cities require skills that put them beyond the reach of the low-skill blacks living in close proximity to them (Holzer 1998; W. J. Wilson 1987). Because spatiallskill mis- match theory is a variant of the theory of skills-biased technical change, which argues that low-skill workers have lost ground in terms of both employment and wages as a result of the increasing demand for higher skilled workers, it is worth examining its place in metropolitan-level studies of racial wage ineq~ality.~

At the same time, there is a growing number of informal, part-time, and especially temporary jobs, which are disproportionately low-wage and are held primarily by women and minorities (Kalleberg et al. 1997). Should attention now shift to these newer explanations of wage inequality?

Because there is a large research literature on each of the major explanations of racial wage inequality considered here, my review focuses on potential racial, ethnic, and gen- der differences. First, I briefly describe the main argument of each literature, which typically focuses on a single racial-ethnic

Shifts from manufacturing industries to ser- vice industries reflect technological changes that

(1) limit the number of low-skill, high-wage pro- duction jobs and the internal labor markets and unions associated with them; (2) increase the number of technical and computer-intensive jobs in the service sector; and (3) expand the flexible employment relations associated with the service sector (Levy 1998). In addition, demographic changes can be considered an added dimension of flexibility, as immigrants increase the supply of labor in informal and secondary labor markets.

The literature on spatiallskill mismatches is specialized and focuses on employment out- comes for low-skill blacks and not on wages or wage inequality (Ihlanfeldt and Sjoquist 1998). Therefore, I do not test the theory here, which would require detailed commuting and job vacancy data for local areas.

group and tends to overlook gender differ- ences even when they are found in the analy- sis. Next, I discuss reasons why gender might matter, either because the mechanisms differ for men and women or a given set of factors seems to have an impact on only men or women. Finally, I discuss how the argu- ment might apply to inequalities involving other racial-ethnic groups. Because of the lack of previous research, the discussion of Asianlwhite inequality is necessarily more sparse and speculative.

The exodus of manufacturing plants and union jobs from urban areas in the Northeast and Midwest is a major theme of studies of blacwwhite inequalities (Bound and Free- man 1992; Bound and Holzer 1996; Wilson 1987). Such studies initially focused on men, but recent studies have examined the impact of industrial shifts and deunion- ization on women, noting that black women benefited from public sector unionism in the same areas in which black men benefited from industrial unionism (Bound and Dresser 1999; Bound and Holzer 1996). Two comparable studies, one on young men and the other on young women, found that adverse industrial shifts and a decline in union coverage resulted in a significant in- crease in blacWwhite wage differentials over the 1970s and 1980s for both sexes (Bound and Dresser 1999; Bound and Freeman 1992). The relative importance of these fac- tors, however, varies by education and ex- perien~e.~

Changes in unionization explain about 10 percent of the increasing blacWwhite wage dif- ferential among young women from 1973 through 1991, which is more than the effects of industry and occupational shifts combined (Bound and Dresser 1999:73-74). In the Mid- west, the effects of industrial and union shifts to- gether account for about 50 percent of the rise. For young men, industrial shifts are more impor- tant than unionization shifts and once again these factors matter more in the Midwest (Bound and Freeman 1992:213). Bound and Holzer (1996) found that demand shifts over the 1980s had a larger negative effect on the wages for black males than they did for black females in every experience and education category.

In contrast to the clear negative effects of deindustrialization on the wages of blacks, many have argued that Latinos never ben- efited from employment in high-wage manu- facturing jobs in the Midwest because they arrived in the United States during and after the period of economic restructuring in the 1970s. (Puerto Ricans in the Northeast are an exception.) Rather than being displaced from high-wage industrial employment, Latinos filled jobs in low-wage industries in the Sunbelt-a region with a history of boom and bust that often does not follow the chronology and character of Rustbelt (de)industrialization in terms of the degree of unionization (lower), size of manufactur- ing plants (smaller), or technological inten- sity of the production process (higher, as in electronics, and lower, as in garments and textiles). Although the availability of such jobs guards against unemployment, espe- cially for Latinas who are more likely than Latinos to be employed in assembly jobs, such jobs do not offer the high wage premi- ums traditionally found in durable manufac- turing industries (Moore and Pinderhughes 1993; Morales and Bonilla 1993).

Immigration is a second major source of earnings inequality. While demand has shifted away from unskilled labor, interna- tional trade and immigration have risen, in- creasing the supply of low-skilled workers (Borjas, Freeman, and Katz 1996). Despite the widespread assumption that immigrants displace native minorities from jobs by ac- cepting lower wages, the available evidence tends to exonerate immigration as a leading cause of real wage decline and rising wage inequality among native workers. Some studies reveal only a small negative effect of immigration on the wages of low-skilled blacks (Hamermesh and Bean 1998), al- though the effects are somewhat worse for black women than they are for black men (Butcher 1998).4 These studies also show

Butcher (1998:172) finds a significant in- crease in the white malelblack female hourly wage gap among unskilled workers in cities with high Hispanic immigration from 1980 to 1990,

that the main beneficiaries of the influx of new immigrant laborers are high-skill work- ers and owners of capital, who are over- whelmingly white (Hamermesh and Bean 1998; Smith and Edmonston 1997).

A remaining gap in this literature encom- passes the question of whether Latinos and Asians (native-born and immigrant) are dis- advantaged by immigration. The few large- scale econometric studies that include His- panics indicate a larger negative effect of immigration on immigrant wages as opposed to the wages of natives (Altonji and Card 199 1; Baker 1999), although inequality per se is not investigated. In fact, Baker (1999) argues that the presence of Mexican immi- grant women in Southwestern cities in 1990 raised the wages of black women and white women. Baker (1999) suggests that immi- grant women are more likely to work in oc- cupations like domestic service that neither black women nor white women fill in large numbers any longer. In addition, immigrants with low English proficiency are barred from female-dominated occupations involv- ing interaction with customers, which also reduces competition. In short, immigrants work in immigrant-dominated segments of the labor market, competing with other im- migrants and their families and lowering their own group's wages.

There is, however, less consensus in the case-study literature on immigrant-rich cit- ies. Some scholars highlight the positive as- pects of a racially divided labor market. Eth- nic enclaves offer mobility opportunities to skilled immigrants and may even generate high-skilled jobs for natives by keeping cer- tain industries afloat with an influx of low-skilled immigrant labor (Portes and Zhou 1992; Waldinger 1996). On the other hand, employers express preferences for immi- grant workers over black workers, and blacks express their sense of competition with immigrants over job opportunities (Kirschenman and Neckerman 199 1). Still others argue that overall levels of racial and

while the white maleiblack male hourly wage gap was unaffected. In addition, Altonji and Card (1991) found that black women were the most concentrated in high-immigrant-share industries, and black men were the least concentrated in these industries.

class polarization have increased, with im- migrants concentrated in "casual" jobs and native whites concentrated in professional jobs (Frey and Liaw 1998; Mollenkopf and Castells 199 1; Sassen 1991). Unfortunately, there has been no systematic accounting of earnings inequality between different racial and ethnic groups in these cities. Attention to gender differences also has been lacking, except to note that Latino and Asian women are the most likely to end up in the worst jobs at the bottom of the urban labor market (but see Fernandez-Kelly and Garcia 1989).

The third relevant stream of literature ex- plores the relationship between the size of the black population, and discrimination and inequality. According to an early argument in this field, whites became more discrimi- natory as they felt threatened by increased economic competition from a large and in- creasing number of minorities (Blalock 1956; Jiobu and Marshall 1971). Alternative accounts suggest that elite whites gain from exploiting minorities while working-class whites are left worse off (or unaffected), re- gardless of any perceived competition over resources (Glenn 1966; Tienda and Lii 1987).

There are now many excellent summaries of this literature (e.g., Beggs, Villemez, and Arnold 1997; Tomaskovic-Devey and Roscigno 1996). However, only Beggs et al. (1997) and Grant and Parcel (1990) examine gender differences, finding that the effect of black population concentration is stronger


1 for black women. In some models, Grant and I Parcel (1990) find that the effect is nonsig- nificant for black men. Beggs et al. (1997) argue that black women have stronger fam- ily and social ties than black men that pre- vent them from moving to search for better jobs. Similarly, Grant and Parcel (1990) sug- gest that "traditional" factors such as black population concentration are less important in determining black men's relative labor market outcomes than are structural economic changes because of black men's greater integration in the industrial economy. Although increased blacklwhite inequali- ties in labor markets with high concentra-

tions of blacks has been well documented, the effect of immigrant composition on wage inequality involving racial and ethnic groups with the largest shares of recent immigrants (i.e., Latinos and Asians) has not been con- sidered. This is true even though the moti- vation to include other racial-ethnic groups is clear-Tienda and Lii (1987) argue that "the altered racial-ethnic composition of the

U.S. population as a result of contemporary immigration flows warrants the consider- ation of these other groups [Asians and Latinos]" (p. 142). Studies that include other racial and ethnic groups (but not immigrants as such) generally find a positive association between minority concentration and racial earnings inequality (Frisbie and Neidert 1977; Tienda and Lii 1989). I therefore con- sider the population density of immigrants, as well as of blacks, as a possible source of high levels of wage inequality for all racial-ethnic groups.

Increasing wage inequality between black men and white men over the 1980s has been linked not only to the decline of high-paying manufacturing jobs but also to the growth of high-skill service industries. These new in- dustries are part of a general technological shift in relative demand toward skilled labor that has made existing skills differentials be- tween blacks and whites more costly in terms of earnings and employment dispari- ties (Juhn, Murphy, and Pierce 1991). In a study of hiring decisions in four cities, Holzer (1998) found that "the apparent growth in the demand for many of these tasks (such as computer use) may well have contributed to the improved relative wages of less educated women in the past decade and to the deteriorating relative wages of less educated and especially minority [black and Hispanic]) males" (p. 241). In contrast, Holzer (1998) found little effect of increas- ing skill requirements on Asian employment and wages.

Employment conditions have undergone other significant changes. The core postwar ideal employment relationship included a full-time job with benefits, job mobility, and long-term job security, all within a full-em- ployment context. Now jobs are seen as less secure with the rise of temporary work, in- dependent contracting, and involuntary part- time work (Kalleberg et al. 1997; Schmidt 1999). Moreover, the current employment context is uncertain, with a federal commit- ment to full employment taking second place to fighting inflation (Galbraith 1998), al- though this was less true in the late 1990s. Because these transformations have oc- curred within the context of declining wages at the bottom of the labor market, the as- sumption has been that those groups who are overrepresented among the less-skilled have been most affected. Thus, I expect that these new economic factors will negatively affect the relative positions of blacks and Latinos.

Although I pursue a broad comparative ap- proach, there are several important issues beyond the scope of this analysis. First, I fo- cus on only one outcome-wage inequality-and examine levels of wage inequality rather than changes in wage inequality over time.

Second, I examine the influence of con- temporaneous labor market conditions rather than changes in those conditions (except for deindustrialization, which cannot be mea- sured in any other way).5

Third, spatial studies often encounter dif- ficulties when it comes to testing the pos- sible mechanisms that underlie the relation- ship between particular labor market condi- tions and wage inequality. For example, my analysis cannot pinpoint whether a particu- lar labor market characteristic fosters wage inequality in the aggregate because of dy- namics that are specific to particular indus- tries, occupations, skill groups, or demo- graphic groups. This methodological prob- lem can only be solved through detailed and

For example, labor market outcomes can be altered as a result of either an inflow of immi- grants over time or a high overall stock of immi- grants at one point in time, although the two are often related because new immigrants tend to lo- cate in large cities where past immigrants have already concentrated. Both measures are com- mon in the literature, although gross flows seem to be less important than flows within the par- ticular industries in which immigrants work (F. Wilson and Jaynes 2000).

disaggregated analyses of a single labor mar- ket factor.

Fourth, there is considerable debate over whether wage disparities should be inter- preted mainly as an indicator of discrimina- tion, market (supply and demand) forces, or segmented labor markets (with no direct role for discrimination). Some argue that dis- crimination should be measured only after controlling for observable human capital dif- ferences among workers, while others con- tend that such differences are themselves the result of discrimination. I include controls for human capital and suggest that the ad- justed wage gaps can be interpreted as lower-bound estimates of discrimination. However, note that the adjusted wage gaps could be higher-bound estimates if group differences in certain omitted characteristics are linked to wage disparities but not to dis- crimination. My analysis cannot adjudicate among these different theories.

Finally, over the long term, inequality has been rising within racial groups and declin- ing between racial groups, making wage in- equality between racial groups seem less pressing. Nonetheless, wage inequality be- tween racial groups remains high, especially among men, and it is extremely uneven spa- tially. A key objective of this analysis is to determine whether the "old" explanations of racial wage inequality still carry weight in this new environment.

Local labor markets have been the preferred unit of analysis in studying the structural de- terminants of racial wage inequality at the subnational level, but beginning in the late 1980s, the primary statistical unit of analy- sis changed from labor markets to individu- als. Labor market variables were merged with individual-level variables to assess the effect of local conditions on individual out- comes (Tienda and Lii 1989). Although pref- erable in the measurement of individual pro- ductivity-related differences-assuming such differences are a cause rather than con- sequence of racial inequality-this strategy is not appropriate for the measurement of la- bor market effects, the significance of which is overestimated because of correlation error within-labor markets.

To correct for these and other problems, I use a two-level hierarchical linear model with detailed data on individuals and labor markets. In level 1, adjusted racial wage gaps are estimated using individual-level data for each labor market. In level 2, variation in adjusted racial wage gaps across la- bor markets is modeled as a function of lo- cal labor market conditions, measured using labor market data. This two-level approach includes random errors that control for cor- relation error among individuals in the same labor market. In the unbalanced hierarchical data used in studies of this kind, the ordi- nary-least-squares (OLS) assumption that random errors are independent and have constant variance is thus violated. Instead, generalized-least-squares techniques are used to obtain efficient estimates of the nonvarying coefficients, and iterative maxi- mum-likelihood techniques estimate the more complex error structure in the two-level model. Of particular interest in the study of racial groups that have small sample sizes in some areas is the use of "shrunken estimatesv-shrunken toward the mean co- efficient-of the within-labor market coeffi- cients that vary across labor markets (i.e., the racial wage gaps).6 Such estimates are an optimal combination of labor market-spe- cific and pooled labor market estimates.

The model is estimated separately for men and women, but for simplicity I discuss the equations in generic terms. A standard indi- vidual-level wage-determination equation is specified for each labor market:

where Wageyis the hourly wage of individual i in labor market j calculated from annual earnings divided by the product of number of weeks worked and usual hours per week worked in 1989. Blacky, Asiany, and Latinoy

1 use the HLM program to conduct these analyses (Bryk and Raudenbush 1992). I experimented with several different thresholds for in- clusion of minority groups in the labor market sample, including a restriction of at least 50 cases of each gender-by-race group. These analyses yielded many of the same substantive conclu- sions as I report below, but they also restricted the sample to a very unique set of "multiracial" labor markets with some distinctive dynamics.

are binary variables for individual i in labor market j, and Pj is a corresponding vector of coefficients for each labor market j (i.e., they are randomly varying coefficients). Whites are the omitted category. The Py, P2j, and P3j coefficients represent adjusted wage gaps (also called wage differentials)-calculated as the log difference between the hourly wages of whites and the hourly wages of each of the three other racial-ethnic groups. I include a standard vector of Rij individual hu- man capital variables (which are assumed to be fixed across j) with their associated coef- ficients p. These variables include three edu- cation dummy variables based on completed years of schooling (less than 12 years, 12 years, and between 13 and 15 years, with 16- and-more years as the omitted category), marital status (married = l), number of own children, immigrant status (foreign born = l), potential employment experience (age years of education -6) and its square, and hours worked (full-time, year-round = 1). The individual error term is denoted by E~ and is assumed to have a normal distribution, mean of zero, and constant variance within labor markets.

An important feature of this model is that the wage gaps are adjusted for both observed and unobserved differences across labor markets when the individual-level variables are centered around their labor market means. That is, the wage gaps are estimated net of differences across labor markets in the distribution of the observed individual-level variables (i.e., education, race, experience, immigrant status, etc.) as well as the distri- bution of any relevant unobserved character- istics. The differencing from labor market means in the estimation of within-labor mar- ket regressions provides for this "fixed-ef- fect" result. Moreover, additional controls for unobserved differences across labor mar- kets are provided for with random effects in the labor-market-level equations (to be de- scribed below). Finally, immigration and black population concentration are measured at both levels: They are controlled at the in- dividual level and then measured again in the aggregate at the labor market level in or- der to estimate an additional contextual effect on wage inequality

Variation in wages and wage gaps across labor markets is estimated in labor-market-

level equations 2 through 5:

The adjusted average hourly wage (In) in la- bor market j is represented by Pojin equation 2, and the adjusted racial-ethnic wage gaps are once again represented by Py ,bj,and P3jin equations 3 through 5. The level-2 er- ror terms indicate that a separate variance component is estimated for each wage gap. This random spatial variation in wages and racial-ethnic wage gaps is partially explained by a vector of Zj variables describing the de- mographic and economic conditions of each labor market j. The associated vectors of co- efficients in equations 3 through 5, x, y2, y3, represent the effect of labor market charac- teristics on the adjusted relationship between race-ethnicity and wages-the racial wage gap-and therefore should be properly un- derstood as interaction effects, as is clear when equations 2 through 5 are substituted into equation 1 to form the full mixed model.

The coefficients and variance statistics from only equations 3 through 5 are pre- sented below because the focus is on the sources of relative wages and not wage lev- els. This means that I cannot assess whether inequality results from gains among whites and/or from losses among minorities, only whether the labor market variables are asso- ciated with relative wages for the three ra- cial-ethnic groups (relative to whites). Nega- tive-signed coefficients indicate a larger log wage gap with whites and lower relative wages for minorities, whereas positive- signed coefficients indicate a smaller log wage gap with whites and higher relative wages for minorities.

Two types of data are used that correspond to the two levels of the model. The first type, individual-level data, includes every vari- able entered in equation 1 and is obtained from the Census of Population Public Use Microdata Samples (5 percent) for 1990 (PUMS-A), the largest available sample of individuals. I restricted the sample to adults aged 25 to 64 who work either part-time or
Table 1. Description of Labor Market Variables

Variable Description

Control Variables

Population size (In) Natural log of residential population, aged 16 to 64

Region South, Northeast, Midwest, West

Demographic Variables

Percent black Percentage of African Americans aged 16 to 64 in the residential

Percent immigrant a Percentage of foreign-born aged 16 to 64 in the residential population

De/industrialization Variables

Manufacturing growth, Average annual employment growth in manufacturing, 1969-1989

Percent durable manu- Percentage of total employment in durable manufacturing industries, facturing employment 1989

Percent nondurable manu- Percentage of total employment in non-durable manufacturing
facturing employment industries, 1989

Percent union coverage Percentage of workers covered by a union contract, 1989

Percent high-tech servicesC Percentage of total employment in research and development
intensive service industries

Flexible/lnsecure Employment Condtions Variables

Percent unemployed Percentage of civilian labor force that is unemployed, 1989

Percent casual employment Percentage of workers in casual employment, including part-timers (less than 35 hours per week or 30 weeks per year), personnel supply industry workers, and the unincorporated self-employed

Sources:Data for manufacturing growth come from the Regional Economic Information System (REIS). Data for Union coverage come from the Current Population Survey. All other data come from the Census of Public Use Microdata Samples (5-Percent) for 1990 (1990 PUMS).

a "Percent immigrant" includes only individuals who completed census questionnaires and identified them- selves as foreign-born, thus undercounting undocumented immigrants. However, regional concentrations of documented and undocumented immigrants are probably highly correlated.

Data on union coverage were compiled by Hirsch and Macpherson (1993). The typology of research and development intensive industries was compiled by Hallock, Hecker, and Gannon (1991) and includes computer and data processing, engineering and architectural services, research and testing, management and public relations, miscellaneous professional services, and financial investment and securities.

full-time,7 are not self-employed or farm To measure local labor market conditions, workers, and earn hourly wages between I use weighted PUMS-A data aggregated to $1.00 and $250.00 . In addition, this the MSA level, and MSA-level data from the microdata sample includes residents of met- Regional Economic Information System ropolitan statistical areas (MSAs) only and (REIS) and from the Current Population Sur- cannot be used to draw generalizations about veys (CPS). Most of these labor market mea- the nation as a whole. Descriptive statistics sures are straightforward or have been used for the individual-level variables are pre- in previous studies; they are summarized in sented in Appendix Table A. Table 1. Because the MSAs included in the PUMS-A, REIS, and CPS differed some- what, the number of metropolitan areas in

'In similar models with different wage gap

the final sample was reduced to 181 from an

outcomes, samples of full-time workers only

original 272 in the PUMS-A. I also include

yielded similar conclusions. I have not performed similar tests here because of the small number of in Zj a set of four control variables for un- minorities in many of the labor markets. measured fixed differences across j. Popula

Table 2. Mean Racial Wage Gaps across 181 U.S. Metropolitan Statistical Areas, 1989 (in 1995 In Dollars): PUMS, 1990

Unadjusteda Adjustedb

MeanC Range MeanC Range Sex and Wage Gap (S.D.) Maximum Minimum (S. D.) Maximum Minimum
Blacklwhite     -.lo9     -.297     .057     -.040     -.I83     ,084
    (.069)            (.059)         
Blacklwhite     -.284     -.470     -.067     -.I62     -.274     -.020
    (.069)            (.055)         
Asianlwhite     -.I12     -.521     .202     -.040     -.389     ,168
    (.105)            (.106)         
Latinotwhite     .260     -.614     -.022     -.088     -.300     ,069
    (.108)            (.074)         

Note: Numbers in parentheses are standard deviations. Number of women = 1,193.1 19; number of men = 1,261,767. a Estimates taken from the ylo- 730 coefficients in equations 3 through 5 when all micro and macro vari- ables are omitted except for the racial-ethnic dummy variables (i.e., all R and Z variables are set to zero). Same as in note a above, except estimates are adjusted for the full set of individual characteristics repre- sented by R in equation 1. This is a weighted mean with weights inversely proportional to the sample size of labor markets.

tion size controls for the tendency of wages ethnic groups separately for men and for to be higher in large cities; and binary vari- women. These estimates should be inter- ables for the Northeast, West, and Midwest preted as weighted means across the sample control for differences in broad regional of 18 1 labor market^.^ As expected, the wage wage and wage inequality levels. Together, gaps narrow once all the individual-level these variables provide assurances that un- variables are included in the model. The observed labor market characteristic~ corre- Asianlwhite gap for women is eliminated lated with wage inequality are not biasing (declining from -.070 to .010) and the the results. Although these variables are of Latinatwhite gap narrows by .13 log points substantive interest in their own right, a de- from -. 159 to -.026. When converted to per- tailed consideration of their effects is not centages by taking the antilog and multiply- provided here. Descriptive statistics for the ing by 100, these estimates indicate that labor market variables are presented in Ap- Latina wages are 85.3 percent of white pendix Table B. women's wages before controlling for edu- cational and other differences between the groups, and 97.4 percent after such controls RESULTS are added. Similar declines are evident in the

The unadiusted figures are the coefficients

-ylo-~30from a baseline model that contains the

It is well documented that wage inequality

three racial-ethnic variables (black, Latino, and

varies by region, race, ethnicity, and gender.

Asian) and no other individual-level or labor Table 2 presents unadjusted and adjusted es- market variables (i.e., all R and Z are set to zero). timates of the mean log hourly wage gaps For the adjusted figures, the individual-level con- between whites and the three other racial- trol variables (R) are included.

equations for men, although the adjusted wage inequality for men remains substantial for blacks and Latinos compared with whites

(85.0 and 91.6 percent, respectively). Note that although inequality is lowest between Asians and whites, the average hourly wage level of Asian men does not exceed that of white men, a pattern often found when com- paring family income levels in the general population.

Table 2 also reveals substantial variation in wage gaps across labor markets. Regard- ing the adjusted gaps, the share of white male wages earned by black men ranges from 76 percent to 98 percent, and the share earned by Asian men ranges from 67.8 to

118.3 percent. Similarly, the female wage gaps each span a 25 to 30 percentage-point range across labor markets. The estimated standard deviations of the adjusted wage gaps are listed in parentheses in Table 2 (the variances are listed at the bottom of Table 6 in the row labeled "random m~del").~

These estimates indicate a somewhat greater de- gree of spatial variation for women than for men in the blackfwhite wage gap despite women's much lower level of wage inequal- ity. The variance statistics also show a gen- erally higher degree of variation in Asian/ white and Latinofwhite differentials, com- pared with those between blacks and whites. This greater variation is surely due to mea- surement error given the smaller sample sizes of Asians, but differences across cities in ethnic composition probably plays a part as well. A more disaggregated analysis could discern differences in the absolute and rela- tive standing of these subgroups (e.g., Japa- nese and Vietnamese workers), but small sample sizes preclude this type of analysis for many of the labor markets.

Do labor markets with high levels of one type of racial wage inequality also have high levels of other types? If this were the case, differences in the patterns, and even sources, of wage inequality by gender, race, and eth- nicity presumably would be small. If this is not the case, some types of wage inequality

These are the variances for each of the ad- justed wage gaps before any labor market vari- ables are entered into the model and the adjusted racial wage gaps are permitted to vary randomly (i.e., the variances of m).

might be more extreme than other types of wage inequality in a given region and thus might have different sources. Tables 3a and 3b present the matrix of bivariate correla- tions between all of the outcome variables, both adjusted and unadjusted. All of the cor- relations are positive, significant, and at least moderate in strength, suggesting a de- gree of convergence in the spatial distribu- tion of racial wage inequality that is much greater than that found for other measures of wage inequality (McCall 2000a, 2000b).

Observe, however, that the strength of the correlations varies by race, ethnicity, and gender. The lowest correlations between ad- justed wage gaps (r = .23) suggest that high levels of blacWwhite and Asianlwhite wage inequality are the least likely to be found in the same labor market, whereas the highest correlations (r = .75) suggest that high lev- els of Latinofwhite and Asianlwhite inequal- ity are the most likely to be found in the same labor market. There also appears to be a significant overlap in the location of black1 white and Latinofwhite wage inequality for men. Next in strength are correlations be- tween the wage gaps for men and women in the same racial-ethnic group, especially those between Latinos and Latinas, followed by correlations between wage gaps for men of one race or ethnicity and women of a dif- ferent race or ethnicity. Thus, similarities in the location of wage inequality are greater within gender groups than they are within racial-ethnic groups, suggesting that gender differences could be more important than ra- cial-ethnic differences. Nevertheless, the correlations are high in both cases.


The fact that different types of racial wage inequality are all not necessarily high or low in the same regions suggests that there might be differences in the sources of wage in- equality for different groups. The reduced models provide two pieces of information to evaluate this question. First, for each sub- stantive group of labor market variables, the effect on each wage gap is estimated after controlling for all individual-level variables and the control variables for region and population size (see Table 4). Second, Table

Table 3a. Correlations between the Racial Wage Gaps among Women vs. among Men across 181 U.S. Metropolitan Statistical Areas, by Race and Ethnicity, 1989: PUMS, 1990

Wage G~D amone Men
------------    -
Wage Gap     BlacWhite     AsianIWhite     LatinotWhite
among Women     Unadjusted     Adjusted     Unadjusted     Adjusted     Unadjusted     Adjusted
Unadjusted     .62     .56     .40     .28     .43     .49
Adjusted     .5 1     .52     .34     .23     .32     .43
Unadjusted     .46     .43     .57     .46     .52     .53
Adjusted     .42     .37     .49     .54     .48     .45
Latinolwhi te:         
Unadjusted     .45     .45     .57     .54     .74     .66
Adjusted     .50     .50     .48     .48     .57     .61

Note: Number of women = 1,193,119; number of men = 1,261,767.

Table 3b. Correlations between Racial Wage Gaps across 181 U.S. Metropolitan Statistical Areas, by Gender, 1989: PUMS, 1990

Wage Gap

BlacWhite AsianlWhite

Gender and Race-Ethnicity Unadiusted Adiusted Unadjusted Adjusted

Women Asianlwhite: Unadjusted




Men Asianlwhite: Unadjusted




Note: Number of women = 1,193.1 19; number of men = 1,261,767.

5 presents the amount of spatial variance in women of the same race or ethnicity. Among each of the wage gaps accounted for by each the demographic factors, percent black is as- of the substantive groups of variables. These sociated with low relative wages for blacks, variances are net of individual-level vari- while immigrant-rich labor markets have ables and the controls for region and popu- only a small negative effect on the relative lation size. wages of black women and no significant ef-

These reduced models indicate that nearly fect on the relative wages of black men. In all of the expected relationships are signifi- contrast, the share of immigrants in the local cant and in the same direction for men and population is associated with significantly

Table 4. Reduced-Model Effects on Micro-Adjusted Racial Wage Gaps, by Gender, across 181 U.S. Metropolitan Statistical Areas, 1989 (in 1995 In dollars): PUMS, 1990


Independent Variable Women Men
Panel I: Controls

Population (In) .005 (.042)

Northeast .959** (. 142)

Midwest .933** (.102)

West ,195 (.121)

Panel 2: Demographic Concentration

Percent black -.023** -.017** (.005) (.006)

Percent immigrant -.011* -.008 (.005) (.005)

Panel 3: De/industrialization

Manufacturing growth, 1969-1989

Percent durable manufacturing employment

Percent nondurable manufacturing employment

Percent union coverage

.082** (.026)

.029** (.008)

-.004 (.013)

.025** (.008)

Panel 4: High Technology

Percent high-tech ,050 services (.035)

Panel 5: Flexible/lnsecure Employment Conditions
Percent unemployed     -.044     -.030
    (.028)    (.030)
Percent casual     .014     .065**
employment     (.020)     (.021)

Asianwhite Latinowhite Women Men Women Men
-.029** (.011) -.024** (.007)     -.028* (.013) -.049** (.O 10)     .001 (.007) -.028** (.004)     -.001 (.009) -.038** (.006)
,007 (.045) -.003 (.030)     -.070 (.059) .006 (.039)     -.088** (.029) .004 (.022)     -. 1 12** (.036) ,002 (.026)

Notes: Numbers in parentheses are standard errors. All standard errors and coefficients are multiplied by 10 to simplify presentation. Number of women = 1,193,119; number of men = 1,26 1,767.

Unstandardized coefficients are derived from a reduced multilevel model. The reduced models include all of the variables in equation 1 and selected variables in equations 2 through 5 as listed in each panel. Each reduced model includes the listed variables and controls, although the coefficients for the controls are not

presented, except in the first panel. *pS .05 **p S .O1 (two-tailed tests)

lower relative wages for both Latinos and Asians. Unexpectedly, percent black also widens the wage gap between Asians and whites, but these effects weaken in the full model. This association probably stems from the coresidence of low-income Asians in cit- ies with large black populations, and the coresidence of high-income Asians in cities without large black populations, rather than any direct causal relationship between the two. Similarly, percent black has no signifi- cant effect on the Latinolwhite wage gap.

Table 5. Percentage of the Variance in Racial Wage Gaps Explained by Labor Market Factors across 181 U.S. Metropolitan Statistical Areas, 1989: PUMS, 1990
    BlackIWhite    AsianlWhite     LatinolWhite
Variable     Women     Men     Women     Men     Women     Men
Controls Regionslpopulation (reduced model)     54.8     42.2     13.4     28.1     39.2     27.8
Net of Controls (Reduced Models) Demographic concentration Delindustrialization     1 1.7 12.2     3.6 9.8     19.9 5.7     20.3 8.2     29.3 2.5     23.3 4.5

Percent high-tech services 0 0 9.6 0 0 0
Flexibility 1.5 0 0 0 10.5 8.9

Total variance explained 73.5 50.7 34.8 47.7 82.2 57.9 (full model)

Source: For reduced models, see Table 4; for the full model, see Table 6.

Among the delindustrialization factors, the results also are in the expected direction for men and women of the same race or eth- nicity. High durable manufacturing employ- ment and unionization are both associated with high relative wages for blacks. Interest- ingly, manufacturing employment growth lifts the relative wages of black women but has no effect on black men, suggesting that for men levels of manufacturing employ- ment matter more than changes in them. As expected, these variables have no significant effect on the relative wages of Latinos, ex- cept for a small positive effect of percent durable manufacturing employment on the relative wages of Latinas, confirming the importance of delindustrialization in deter- mining blacklwhite labor market outcomes. Finally, and unexpectedly, durable manufac- turing employment concentration narrows the wage gap between Asians and whites. This may reflect better relative opportunities for Asians in high-technology and capital- goods manufacturing industries, especially in ethnic-enclave industrial districts or trans- planted capital-goods manufacturing com- plexes. Although the amount of spatial vari- ance in relative wages explained by all four delindustrialization variables is greatest for blacks (12.2 and 9.8 percent for women and men, respectively), the amount explained for Asian men is almost as large (8.2 percent).

The impact of the high-technology services and flexible employment variables- the new economy variables-is much less consistent and conclusive. The notion that black men have been harmed by the new economy is still captured best by the delin- dustrialization variables. The percent high- technology service industry employment has a significant negative effect on black men's relative wages, but this effect becomes non- significant in the full model (see Table 6), and the variance explained is zero in the re- duced model. Surprisingly, even the impact of joblessness on the wage gap for blacks is not significant. Finally, the impact of percent casual employment improves the relative wages of black men, an effect that probably says more about low wages among white men than it does about high wages among black men. This effect is persistent, though, and resurfaces in the full models. These vari- ables have no significant effects on the rela- tive wages of black women.

For Asians and Latinos, the new economy factors are next in importance to immigra- tion in explaining relative wages for all but Asian men. First, percent unemployed has a significant negative effect on the relative wages of Latinos, explaining roughly 10percent of the spatial variance. However, the strength of the effect for Latinos diminishes somewhat in the full model when other vari- ables are added (Table 6).This may indicate that high joblessness in these reduced mod- els is describing areas with other character- istics, such as high immigration. Because the

same dynamic does not appear for Asians, who are also affected by immigration, it is clear that unemployment has an important independent effect on the relative wages of Latinos. Because evidence suggests that un- employment reduces the wages of low- skilled workers, a strong negative effect of percent unemployed was also expected for blacks. Perhaps this would surface if we looked at low-skilled blacks.

Second, in areas that have a disproportion- ate share of high-technology service em- ployment-computer and data processing, engineering, research and testing, manage- ment and public relations, and financial in- vestments-Asian women are relatively worse off. The effect is significant, explains 10 percent of the spatial variance in the rela- tive wages of Asian women, and persists in the full model. Given that high-tech service industries are often concentrated in areas with high-tech manufacturing industries, this finding could reflect the employment of many low-skilled Asian women in low-wage assembly jobs. Alternatively, these results could be picking up the fact that high-tech service centers, such as New York City, have large Asian enclaves with low-wage service and manufacturing employment dominated by women. Regardless, this is one of the few cases in which a significant gender differ- ence appears in the source of inequality within the same racial-ethnic group and the only case in which technology has a strong independent effect.

My focus thus far has been on the substan- tive variables, but the controls are clearly important. Nearly half the overall variance in blacklwhite wage gaps across labor mar- kets is accounted for by differences in black/ white wage gaps across broad regions. Com- pare this with between 13percent (men) and 25 percent (women) of the variance that is explained by the demographic and delindus- trialization factors combined. Although less consistent, there is also considerable varia- tion across regions in the relative wages of Latinos and Asians. The relative wages of blacks are highest in the Northeast and Mid- west and lowest in the South, while the rela- tive wages of Latinos, Latinas, and Asian men are highest in the Midwest, a region with low immigration and fewer recent im- migrants. Beyond these general patterns, the regional effects vary by gender and race-eth- nicity. For example, the Latinalwhite wage gap is significantly lower in both the West and Northeast, relative to the South, a sign perhaps of greater relative inequities for women in the southern regions of the United States that have large Latino native and im- migrant populations.


Most of the patterns described above hold up in the full model that includes all individual- level and labor market variables (see Table 6). First, Latinos and Asians still have sig- nificantly lower wages relative to whites in areas with a disproportionate share of immi- grants, and black women have lower relative wages in these areas as well. Second, areas with relatively large black populations con- tinue to have larger wage gaps between blacks and whites, but the effect is not sig- nificant for men in the full model. Third, du- rable manufacturing employment and union- ization continue to significantly raise the relative wages of black men and women. Fourth, unionization and manufacturing em- ployment growth significantly improve the relative wages of Latinas, whereas these variables were only marginally significant in the reduced models. The industrial character of the labor market has little effect on the relative wages of Latino men. Broadly speaking, then, there are more similarities than differences among women and men of the same racial-ethnic group, although sev- eral important exceptions to this overall pat- tern persist in the full model.

The expected effects of demographic and industrial conditions on racial wage differ-


entials hold up in the full model, but the same cannot be said of the effects of the new economy variables. In terms of the wage gap between Asians and whites, the unexpected I contribution of black population concentra- tion in widening the gap and the unexpected contribution of durable manufacturing employment in narrowing the gap were both weakened to nonsignificance for two of the four wage gaps. On the other hand, there were several unexpectedly strong relation- ships that remained significant-the effect of high-tech service economies in lowering

Table 6. Full-Model Effectsa on Micro-Adjusted Racial Wage Gaps in 1995 Dollars, by Race- Ethnicity and Gender, across 181 U.S. Metropolitan Statistical Areas, 1989: PUMS, 1990
BlackWhite     AsianlWhite     LatinolWhite
Variable     Women     Men     Women     Men     Women     Men
Intercept     -.a81 (.935)     -2.08 1 (1.096)     .57 1 (1.520)     2.745 (2.046)     ,836 (.969)     2.030 (1.332)
Region and population size     Yes     Yes     Yes     Yes     Yes     Yes
Demographic Concentration Percent black -.020** (.005)     -.011 (.006)     -.024* (.O 12)     -.029 (.016)     ,012 (.008)     ,002 (.010)
Percent immigrant     -.010* (.005)     -.008 (.006)     -.023** (.008)     -.049** (.011)     -.027** (.005)     -.030** (.007)
De/industrialization Manufacturing growth, 1969-1989     .043
Percent durable manufacturing employment
Percent nondurable manufacturing employment
Percent union coverage     .023**
High Technology Percent high-tech services     .066 (.035)     -.043 (.042)     -.133** (.054)     .028 (.076)     -.035 (.032)     -.080 (.048)
Flexible/lnsecure Employment Conditions Percent unemployed ,006 -.032 (.03 1) (.036)     .03 1 (.050)     .04 1 (.067)     -.054* (.028)     -.089* (.039)
Percent casual employment     -.009 (.020)     .062** (.023)     -.020 (.029)     ,012 (.038)     ,010 (.019)     -.001 (.025)
Level-2 Variance

Random modelb .00343 .00306 .00366 ,01118 ,003 14 ,00553

Full modelC ,0009 1 .00151 .00193 ,00585 ,00056 ,00233

Variance explained 73.5 50.7 34.8 47.7 82.2 57.9 (percent)

Note: Numbers in parentheses are standard errors. Number of women = 1,193,119; number of men = 1,26 1,767.

a Unstandardized coefficients are derived from the full multilevel model as expressed in equations 3, 4, and 5. Equation 1 and 2 coefficients are not presented. All standard errors and coefficients are multiplied by 10 to simplify presentation.

Total level-2 variance taken from baseline random coefficients model in which all Z are set to zero in equations 2, 3, 4, and 5.

Unexplained variance in full multilevel models when all Z are included.


*p5 .05 p 5 .O1 (two-tailed tests)

the relative wages of Asian women, the ef- fect of casual employment in raising the relative wages of black men, and the effect of unemployment in lowering the relative wages of Latino men and women. Last, the expected negative effect of high-tech service economies on the relative wages of black men, which was significant in the reduced models, was nonsignificant in the full model.

Drawing on the amount of variance explained in the reduced models, five main patterns are evident (see Table 5). First, not only is immigration the main source of rela- tive wage disparities with whites for Asians and Latinos but also the amount of spatial variance it explains is large in absolute terms (between 19.9 percent and 29.3 percent) compared with the explained variance due to other substantive factors. And the variance explained is fairly comparable in magnitude for men and women. Second, the demo- graphic and industrialization variables ex- plain about the same amount of spatial vari- ance in levels of blacklwhite wage inequal- ity, especially among women (11.7 and 12.2 percent among women and 3.6 and 9.8 per- cent among men). Moreover, no other set of substantive factors has much of an additional effect on wage inequality between blacks and whites. Third, the industrialization vari- ables also have a significant effect on the wage gaps between Asians and whites (ex- plaining 5.7 percent among women and 8.2 percent among men). Fourth, the new eco- nomic factors are of secondary importance in explaining spatial variation in wage in- equality. Although secondary, they explain more of the variation in Asianlwhite and Latinolwhite inequality than industrializa- tion factors do and, in absolute terms, job- lessness has a substantial effect on the rela- tive wages of Latinos (explaining roughly 10 percent of the variation) and high-tech ser- vice economies have a substantial effect on the relative wages of Asian women (explain- ing 9.6 percent of the variation). Casual em- ployment, perhaps the best indicator of flex- ibility, had the lowest overall impact. Fi- nally, broad unobservable regional differ- ences in levels of wage inequality are quite large.

I have examined a wide range of explana- tions of racial wage inequality in metropoli- tan labor markets. These include leading ex- planations that date back to the 1950s, such as black population concentration and the lack of manufacturing employment, as well as newer explanations emerging from the re- cent period of economic restructuring, such as the spread of high technology and flex- ible employment conditions. My focus has been on whether certain explanations matter more for some groups than for others and, secondarily, whether the conventional expla- nations of wage inequality still matter in today's new economy. Thus, my approach has been comparative: First, in comparing the sources of wage inequality for different gender, race, and ethnic groups, and, second, in comparing the relative importance of competing explanations of wage inequality. Recent research has been less concerned with broad comparative frameworks of this sort, focusing instead on increasingly de- tailed studies of particular subgroups, espe- cially on black men in the large-scale stud- ies of metropolitan labor markets. Both ap- proaches are necessary and in principle should be complementary rather than op- posed. At their best, broad frameworks can point to systematic patterns that need to be explicated in further detail. Several of the findings of this study, especially those con- cerning the sources of wage inequality among women, could benefit from further in-depth research of this kind.

The findings indicate that the sources of wage inequality vary across racial, ethnic, and gender groups. The findings also offer unambiguous evidence that industrial and demographic structures have a significant impact on wage inequality in metropolitan I labor markets as of 1990. Industrial structure 1 remains one of the most important sources ' of blacklwhite wage inequality, while demo- graphic structure emerges as the most impor- tant source of Latinolwhite and Asianlwhite inequality for men and women. On balance, the sources of racial wage inequality were more similar for men and women of the same race-ethnicity than they were for women of different races or ethnicities or for men of different races or ethnicities. Never-

theless, there were several important gender differences.

Overall, the results offer the first system- atic support for much of the case-study lit- erature on the different underlying condi- tions of racial earnings inequality for blacks and Latinos. Summarizing case studies on the changing economic fortunes of Latinos, Moore and Pinderhughes (1993) argue that:

The "new poverty" described so effectively by Wilson for the black population of deindustrialized Chicago is directly appli- cable only to the New York and Chicago Puerto Rican communities in this volume: the deindustrialization framework simply does not work in cities that were never in- dustrialized to begin with. Nevertheless, these studies indicate that national economic restructuring has affected all cities, even though most are peripheral to mainstream trends. Immigration is of major importance because most of the new jobs are in low- wage manufacturing and service occupa- tions, and these jobs are easily filled by ex- ploitable immigrants. (P. xxxvii)

Although the relative wages of Latinas (and not Latinos) get a boost in regions with high rates of manufacturing growth and unioniza- tion, low immigration and low unemploy- ment are much more important factors. In contrast, manufacturing employment and unionization are more likely than any other factor to raise the relative wages of blacks in metropolitan areas (see Table 7 for a list of the cities with the lowest blacklwhite wage gaps). For black men in particular, unionization is the strongest source of high relative wages, even after considering a wide range of other labor market characteristics. This not only supports the enduring impor- tance of industrial-era institutions for black Americans, it also exposes the limits of fo- cusing on new factors like technology. This does not mean that technology is of no con- sequence, for it may have an effect on rising inequality within labor markets, between la- bor markets, and across the nation as a whole. Rather, technology's impact is often mediated by more fundamental characteris- tics of the local economy-its industrial and demographic structure and regional location (McCall 2000b).

Although industrial structure remains the most important source of blacklwhite wage

1 inequality, immigration is by far the most important source of Latinolwhite and Asian1 white wage inequality, among men and women. After adjusting wage gaps for dif- ferences in the racial and immigrant compo- 1 sition of labor markets-as well as differ- , ences in other compositional factors that

1 would affect relative wage levels-Latinos

I and Asians still have significantly lower wages relative to whites in immigrant-rich labor markets. Note that this is true for the relative wages of native as well as immigrant Latinos and Asians, as immigrant status is controlled at the individual level. This means that the wage gap between whites on one hand and Latinos and Asians on the other hand is wider than would be expected based simply on the existence of a large pool of relatively low-paid immigrants and rela- tively high paid whites in these metropolitan areas.

These results can be interpreted in at least three ways. First, they may call to mind theories previously applied to the relation- ship between black population concentration and an increase in white discrimination against blacks. The theories that predict white managerial or elite exploitation of black workers as a source of cheap labor are perhaps the most apt. They provide some justification for linking immigration to dis- crimination against the racial-ethnic groups most likely to be composed of immigrants. A second interpretation, which is consistent with the first interpretation, notes that immi- grants settle in labor markets where other immigrants live and thus compete among themselves in segmented labor markets to their detriment and to the relative benefit of white workers and managers (Baker 1999; Hammermesh and Bean 1998). Both per- spectives counter the thrust of most previ- ous econometric research and popular atten- tion that has focused on immigrants as a po- tential source of cheap labor that is respon- sible for the declining economic opportuni- ties of native non-Latino whites and blacks. Although there was some evidence for lower relative wages among black women (and not black men) in immigrant-rich labor markets, the effects were relatively small.

Finally, a third interpretation is that the immigrant composition of a labor market- and black composition as well-stand in for

Table 7. Cities with the Lowest and Highest Micro-Adjusted Racial Wage Gaps, 1989
Wage Gap     Women     Men
Lowest Wage Gaps         
BlacWwhite     Rochester, NY     Lansing, MI
    Flint, MI    SaginawlBay City, MI
    Ann Arbor, MI    Flint, MI
    Peoria, IL    Canton, OH
    Benton Harbor, MI    Ann Arbor, MI
Asianlwhite     Rochester, NY     Canton, OH
    Peoria, IL    Pittsburgh, PA
    Canton, OH    Buffalo, NY
    Huntington, WV    Lexington, KY
    SaginawlBay City, MI    Gainesville, FL
Latinolwhite     Modesto, CA     Lansing, MI
    Cleveland, OH    Canton, OH
    Rochester, NY    Gary, IN
    Peoria, IL    Bloomington, IL
    Kansas City, MO    Pittsburgh, PA
Highest Wage Gaps         
Blacklwhite     Monroe, LA     Dallas, TX
    Baton Rouge, LA    Houston, TX
    Macon, GA    Jackson, MS
    Shreveport, LA    Brownsville, TX
    New Orleans, LA    Galveston, TX
Asianlwhite     New York City, NY     New York City, NY
    Jersey City, NJ    Norfolk, VA
    Los Angeles, CA    San Francisco, CA
    San Francisco, CA    Jersey City, NJ
    Atlanta, GA    Chicago, IL
Latinolwhite     Los Angeles, CA     Los Angeles, CA
    New York City, NY    New York City, NY
    Miami, FL    Brownsville, TX
    San Antonio, TX    El Paso, TX
    Dallas, TX    Mission, TX

unobservable human capital differences man capital differences in greater detail, I among racial and ethnic groups. Urban areas cannot rule out this possibility (England et with large populations of immigrants and al. 1999; Neal and Johnson 1996). However, blacks may be areas in which the quality of the strength of the labor market effects of schools is poor for these groups (i.e., Asians, immigration, which explain up to 30 percent Latinos, and blacks) relative to the quality of the variance after controlling for observed of schools for whites. Because my microdata and unobserved differences across labor do not allow me to measure individual hu- markets with fixed-effects and random-ef- fects specifications, suggests that unob- served human capital differences cannot ac- count for the entire relationship. Therefore, the greater inequality between whites and Asians and Latinos in immigrant-rich labor markets appears to be a result, at least to some extent, of all three forces-increased discrimination, increased competition among immigrant racial-ethnic groups that lowers their wages, and low levels of unob- servable skills among Asians and Latinos as a group.

Although the patterns found in this analy- sis seem to overwhelm gender-related differ- ences, the gender differences are more than simply differences of degree-they represent significant differences of kind. For each of the three minority groups, there is some fac- tor that has a significant effect on the rela- tive wages of women but not of men. As a result, more of the coefficients were signifi- cant in the female sample, and the substan- tive factors better explained the spatial variation in wage inequality among women than they did among men. Among blacks and Latinos, the total explained variation, in- cluding regional controls, was also greater for women than for men, and the amount ex- plained was high in each case (73.5 percent and 82.2 percent, respectively). Thus, de- spite lower levels of wage inequality among women and lower spatial variation in in- equality among women (among Latinos and Asians), the sources of inequality for them are more identifiable. However, I emphasize that the additional factors that help explain wage inequality among women are not the same for each racial-ethnic group.

Given the lack of previous research on the structural sources of racial wage inequality among women, some of the findings for women were unexpected-for example, the strong association between high-tech service economies and low relative wages for Asian women. At first glance, the cities with the highest levels of Asianlwhite wage inequal- ity (see Table 7) suggest that this effect might be spurious. There appears to be an overlap between cities with major Asian im- migrant populations and cities that special- ize in management, professional, and finan- cial services, such as San Francisco, Los Angeles, and New York City. Closer inspec- tion reveals that Atlanta is on the list for women and not for men, suggesting a genu- ine independent effect of high-tech services apart from immigration. Many case studies of such cities have suggested that they are highly polarized because high-paid service workers create a demand for personal and other low-end service workers (Sassen 1991). Although researchers have not singled out Asian immigrant women from Latina immigrant women as more likely to fill these low-end service jobs, the findings here suggest a fairly well-defined division of labor between Asian and Latina women in these types of labor markets.

Among Latinas, the degree of unionization and manufacturing growth has a distinct im- pact on their relative wages. Although this seems consistent with the over-representa- tion of Latinas in factory jobs, Latinas are not often thought of as significantly better off in midwestern cities with a strong manu- facturing and union presence (see Table 7). If anything, they are expected to be worse off in southwestern economies dominated by low-wage, nondurable manufacturing. Once again, it could be argued that these findings are spurious-that the low-wage-gap cities in the Midwest shown in Table 7 are simply areas with low immigration levels. However, this interpretation would not fully account for the independent effects of the industrial- ization and unionization variables even when immigration is included in the full models. The strong overlap in the cities with low ra- cial wage gaps for all groups suggest there is a potent combination of forces that reduces wage inequality in midwestern cities-low immigration, high-wage manufacturing em- ployment, and greater unionization. Al- though the latter two factors have declined over the past several decades, contributing to an increase in inequality over time, their lev- els remain higher than in other parts of the country and thus inequality is lower.

Finally, the female-specific findings among blacks are perhaps the least surpris- ing, although the mechanisms involved are little understood. As Beggs et al. (1997) and Grant and Parcel (1990) found, a large con- centration of blacks hurts women's relative economic standing more than it does men's. There is no overlap in the list of cities with the highest blacklwhite wage gaps for men and for women, and my inspection of the data reveals that the cities listed for women all have larger black populations than do the cites listed for men. Beggs et al. (1997) sug- gest that black women might be more "tied to place" than men because they have stron- ger family and social ties that make them less able or willing to migrate from labor markets with few opportunities. Certainly this possibility should be explored in future research on the migration and mobility pat- terns of black women versus black men.

There are several other avenues of research to pursue on the basis of the findings reported here. First, although broad measures of manufacturing employment, high-tech service employment, and flexibility still matter, important changes are taking place within sectors and within detailed industries and these should be taken into consideration in future research. There also is considerable variation in the economic standing of differ- ent ethnic subgroups within the four broad racial-ethnic groups, as well as among dif- ferent cohorts of immigrants (Myers and Cranford 1998). But this type of analysis is possible only for cities that have large enough populations that can be disaggre- gated into finer categories and cannot be car- ried out with a large sample of metropolitan labor markets. Additional case-study research is clearly needed. Finally, the fact that there were many similarities in the sources of wage inequality for men and women should not gloss over continuing inequalities between them and the unique mechanisms that perpetuate low relative wages for women of color. Although recent studies have been careful to compare the effects of different sources of inequality on men and women (Butcher 1998), the findings here suggest a need to do so on a more routine basis. Thus, we would benefit from research that is more broadly comparative while attending to the increasing complexity of racial inequality.

Leslie McCall is Associate Professor of Sociol- ogy and Gender and Women's Studies at Rutgers University. She is interested in the areas of eco- nomic change and social inequality, and her book on these topics, Complex Inequality: Gen- der, Class, and Race in the New Economy

(Routledge), has just been published. Her other research interests include political inequality, social theory, and methodology.

Appendix Table A. Means and Standard Deviations for Individual- Level Variables: PUMS, 1990

Variable Women Men Married Number own children Foreign-born College graduate Some college High school graduate Less than high school graduate Experience Full-time Black Asian Latino Hourly wage (log)

Note: Numbers in parentheses are standard devia- tions. Number of women = 1,193,119; number of men = 1,261,767.

Appendix Table B. Means and Standard Deviations for Labor Market Variables
Mean         Range
Variable     (S.D.)     Maximum     Minimum
Population size (In) 12.56     11.01     15.57
    (1 .OO)        
Northeast     .12     0     1.00
West     .21     0     1.00
Midwest     .26     0     1.00
Percent black     10.95     .03     48.44
Percent immigrant     8.04     .90     56.52


(Continued on next page)
(Auuendix Table B continued) I

Mean Range Variable (S.D.) Maximum Minimum

Manufacturing .89 -4.65 7.57 growth, 1969-1989 (2.42)

Percent manufac- turing employment

Percent durable manufacturing employment

Percent nondurable manufacturing employment

Percent union coverage

Percent high-tech services

Percent unemployed

Percent casual employment

Sources: See Table 1 Note: Numbers in parentheses are standard devia-


Altonji, Joseph and David Card. 199 1. "The Ef- fects of Immigration on the Labor Market Out- comes of Less-Skilled Natives." Pp. 201-34 in Immigration, Trade, and the Labor Market, edited by J. Abowd and R. Freeman. Chicago, IL: University of Chicago Press.

Baker, Susan Gonzalez. 1999. "Mexican-Origin Women in Southwestern Labor Markets." Pp. 244-69 in Latinas and African American Women at Work, edited by I. Browne. New York: Russell Sage Foundation.

Beggs, John, Wayne Villemez, and Ruth Arnold. 1997. "Black Population Concentration and Black-White Inequality: Expanding the Con- sideration of Place and Space Effects." Social Forces 76:65-91.

Blalock, Hubert M. 1956. "Economic Discrimi- nation and Negro Increase." American Socio- logical Review 21 :612-33.

Borjas, George, Richard Freeman, and Lawrence Katz.1996. "Searching for the Effect of Immi- gration on the Labor Market." AEA Papers and Proceedings 86:246-5 1.

Bound, John and Laura Dresser. 1999. "Losing Ground: The Erosion of the Relative Earnings of African American Women During the 1980s." Pp. 61-104 in Latinas and African American Women at Work, edited by I. Browne. New York: Russell Sage Foundation.

Bound. John and Richard Freeman. 1992. "What Went Wrong? The Erosion of Relative Earn- ings and Employment among Young Black Men in the 1980s." Quarterly Journal of Eco- nomics 107:201-32.

Bound, John and Harry Holzer. 1996. "Demand Shifts, Population Adjustments, and Labor Market Outcomes During the 1980s." Working Paper No. 5685, National Bureau of Economic Research, Cambridge, MA.

Browne, Irene. 1997. "Explaining the Black- White Gap in Labor Force Participation among Women Heading Households." American So- ciological Review 62: 236-52. ,ed. 1999. Latinas and African American Women at Work. New York: Russell Sage Foundation.

Bryk, Anthony and Stephen Raudenbush. 1992. Hierarchical Linear Models for Social and Be- havioral Research: Applications and Data Analysis Methods. New York: Sage.

Butcher, Kristin. 1998. "An Investigation of the Effect of Immigration on the Labor-Market Outcomes of African Americans." Pp. 149-81 in Help or Hindrance? The Economic Implica- tions of Immigration for African Americans, edited by D. Hamermesh and F. Bean. New York: Russell Sage Foundation.

England, Paula, Karen Christopher, and Lori Reid. 1999. "Gender, Race, Ethnicity, and Wages." Pp. 139-82 in Latinas and African American Women at Work, edited by I. Browne. New York: Russell Sage Foundation.

Fernandez-Kelly, M. Patricia and Anna Garcia. 1989. "Informalization at the Core: Hispanic Women, Homework, and the Advanced Capi- talist State." Pp. 247-64 in The Informal Economy, edited by A. Portes, M. Castells, and L, Benton. Baltimore, MD: Johns Hopkins University Press.

Frey, William and Kao-Lee Liaw. 1998. "The Impact of Recent Immigration on Population Redistribution within the United States." Pp. 388-488 in The Immigration Debate, edited by

J. Smith. Washington, DC: National Academy.

Frisbie, W. Parker and Lisa Neidert. 1977. "In- equality and the Relative Size of Minority Populations: A Comparative Analysis." American Journal of Sociology 82: 1007-30.

Galbraith, James K. 1998. Created Unequal: The Crisis in American Pay. New York: Twentieth Century Fund.

Glenn, Norval. 1966. "White Gains from Negro Subordination." Social Problems 14: 159-78.

Grant, D. Sherman and Toby Parcel. 1990. "Re- visiting Metropolitan Racial Inequality: The Case for a Resource Approach." Social Forces 68:1121-42.

Hallock, Paul, Daniel Hecker, and Joseph Gannon. 1991. "High Technology Employ-

ment: Another View." Monthly Labor Review 1 14(7):26-30.

Hamermesh, Daniel and Frank Bean, eds. 1998. Help or Hindrance? The Economic Implica- tions of Immigration for African Americans. New York: Russell Sage Foundation.

Holzer, Harry J. 1998. "Employer Hiring Deci- sions and Antidiscrimination Policy." Pp. 223- 57 in Generating Jobs: How to Increase De- mand for Less-Skilled Workers, edited by R. Freean and P. Gottschalk. New York: Russell Sage Foundation

Ihlanfeldt, Keith and David Sjoquist. 1998. "The Spatial Mismatch Hypothesis: A Review of Recent Studies and Their Implications for Welfare Reform." Housing Policy Debate 9:849-92.

Jiobu, Robert and Harvey Marshall. 1971. "Ur- ban Structure and the Differentiation between Blacks and Whites." American Sociological Review 36:638-49.

Juhn, Chinhui, Kevin Murphy, and Brooks Pierce. 1991. "Accounting for the Slowdown in Black-White Wage Convergence." Pp. 107- 43 in Workers and Their Wages, edited by M. Kosters. Washington, DC: American Enter- prise Institute. Kalleberg, Arne, Edith Rasell, Kenneth Hudson, David Webster, Barbara Reskin, Naomi Cassirer, and Eileen Appelbaum. 1997. Nonstandard Work, Substandard Jobs: Flexible Work Arrangements in the U.S. Washington D.C.: Economic Policy Institute. Kirschenman, Joleen and Kathryn Neckerman. 1991. "'We'd Love to Hire Them, But . . .': The Meaning of Race for Employers." Pp. 203-32 in The Urban Underclass, edited by P. Peterson and C. Jencks. Washington, DC: Brookings Institution. Levy, Frank. 1998. The New Dollars and Dreams. New York: Russell Sage Foundation. McCall, Leslie. 2000a. "Gender and the New In- equality: Explaining the CollegeINon-College Wage Gap." American Sociological Review 65:234-55. McCall, Leslie. 2000b. "Explaining Levels of Within-Group Wage Inequality in U.S. Labor Markets." Demography 37:415-30. Mollenkopf, John and Manuel Castells, eds. 1991. Dual City: Restructuring New York.

New York: Russell Sage Foundation.

Moore, Joan and Raquel Pinderhughes, eds. 1993. In the Barrios: Latinos and the Underclass Debate. New York: Russell Sage Foundation.

Morales, Rebecca and Frank Bonilla, eds. 1993. Latinos in a Changing U.S. Economy. London, England: Sage.

Myers, Dowel1 and Cynthia Cranford. 1998. "Oc- cupational Mobility of Immigrant and Native- Born Latinas." American Sociological Review 63:68-92.

Neal, Derek and William Johnson. 1996. "The Role of Pre-Market Factors in Black-White Wage Differences." Journal of Political Economy 1042369-95.

Portes, Alejandro and Min Zhou. 1992. "Gaining the Upper Hand: Economic Mobility among Immigrant and Domestic Minorities." Ethnic and Racial Studies 15:49 1-521.

Sassen, Saskia. 1991. The Global City. Princeton, NJ: Princeton University Press.

Schmidt, Stefanie. 1999. "Long-Run Trends in Workers' Beliefs about Their Own Job Secu- rity: Evidence from the General Social Sur- vey." Journal of Labor Economics 17(supp.): S127-S41.

Smith, James and Barry Edmonston, eds. 1997. The New Americans: Economic, Demographic, and Fiscal Effects of Immigration. Washington DC: National Academy.

Tienda, Marta and Ding-Tzann Lii. 1987. "Mi- nority Concentration and Earnings Inequality: Blacks, Hispanics, and Asians Compared." American Journal of Sociology 93: 141-65.

Tomaskovic-Devey, Donald and Vincent Roscigno. 1996. "Racial Economic Subordina- tion and White Gain in the U.S. South." American Sociological Review 61:565-89.

Waldinger, Roger. 1996. Still the Promised City? African Americans and New Immigrants in Postindustrial New York. Cambridge, MA: Harvard University Press.

Wilson, Franklin and Gerald Jaynes. 2000. "Mi- gration and the Employment and Wages of Native and Immigrant Workers." Work and Occupations 27: 135-67.

Wilson, William Julius. 1987. The Truly Disad- vantaged. Chicago, IL: University of Chicago Press.

  • Recommend Us