Gender and the New Inequality: Explaining the College/Non-College Wage Gap

by Leslie McCall
Gender and the New Inequality: Explaining the College/Non-College Wage Gap
Leslie McCall
American Sociological Review
Start Page: 
End Page: 
Select license: 
Select License


Rutgers University

The new inequality is often characterized by the increasing wage gap between work- ers with a college education and those without. Yet, although the gap in hourly wages between college-educated and non-college-educated women is high and ris- ing, the topic has been overshadowed by research on gender inequality and wage inequality among men. Using the 1990 5-percent Public Use Microdata Samples, independent sources of macro data, and controls for individual human capital char- acteristics, I examine the association between the college/non-college wage gap and key aspects of local economic conditions for women and men. While the college/non- college wage gap among women is comparable in size to the gap among men, sig- nificant gender differences emerge in the underlying sources of high wage gaps in over 500 labor markets across the United States. Compared with men, flexible and insecure employment conditions (e.g., joblessness, casualization, and immigration) are more important in fostering high wage gaps among women than are technology, trade, and industrial composition, the prevailing explanations of rising wage in- equality over time. Based on these gende~ difference~, I reconsider the debate on labor-market restructuring and inequality and discuss a new analytical focus on

differences in within-gender inequality.

Direct all correspondence to Leslie McCall, Department of Sociology and Women's Studies Program, Rutgers University, Piscataway, NJ 08854 (lmccall@rci.rutgers.edn) . This research was supported in part by a faculty grant from the Rutgers Research Council. I thank Eric Parker, Annette Bernhardt, Kathryn Edin, Patricia Roos, ASR editor Glenn Firebaugh, and three anony- mous reviewers for their excellent comments and suggestions.

primary route to upward mobility. Similarly, economists charge that rising earnings in- equality is largely due to skill-biased tech- nological change in which the demand for low-skill workers has fallen relative to workers with computer, information-based, and other technical and high-level skills (Levy 1998; Murnane and Levy 1996). Sharp earnings declines among men with a high school or less education have been at- tributed to this shift in demand, as have the relatively better earnings of college-edu- cated workers. Researchers have therefore focused attention on the collegelnon-college wage gap-the gap in hourly wages between college-educated and non-college-educated workers-which began to widen in 1979 for both men and women (Mishel, Bernstein, and Schmitt 1997: 173-74).'

I For the sake of brevity, I refer to the "col- legelnon-college wage gap" as the "college wage gap." Because the collegelnon-college wage gap


While there is far less certainty about the nature and causes of the collegelnon-college wage gap than this account suggests, the central problem addressed here is the lack of attention to high and rising wage differen- tials between college-educated and non-col- lege-educated women (Elau 1998). In one of the few sociological analyses of female earn- ings inequality, Smith (1991) argued that economic restructuring has "marginally increased the number of opportunities in high paying jobs, but also increased the already heavy concentration of women in jobs with lower earnings" (pp. 133, 122). Morris, Bernhardt, and Handcock (1994) refer to this as a trend toward "polarization" and predict an acceleration in the growth of earnings in- equality among women as conditions con- tinue to worsen at the bottom and/or improve at the top (Bernhardt, Morris, and Handcock 1995). Each of these studies claims that earnings inequality among women is a mat- ter of growing significance.

But despite the major contribution that these and a few other studies have made (e.g., Nelson and Lorence 1988), there are at least three significant gaps in the research literature. First, the specific problem of the collegelnon-college wage gap among women has not been accorded the same prominent attention as the gap has among men. Second, although economic restruc- turing is the main explanatory subtext in these studies, a comprehensive empirical analysis of the impact of recent restructur- ing on earnings disparities among women has yet to be provided. As a result, little is known about whether a wide range of po- tential sources of inequality are the same for men and women (but see DiNardo, Fortin, and Lemieux 1996). And, finally, for both men and women, almost all re- search on the 1980s surge in earnings in- equality has focused on rising inequality rather than levels of inequality (Blau and Kahn 1996). I address these three issues in a comparative study of labor-market restructuring and male and female college1 non-college wage differentials in local la- bor markets in 1990.

is calculated as the difference in logged wages between two educational groups, I also refer to it as a wage differential.

First, even though wage inequality has risen within education groups, I focus on in- equality between the college educated and non-college educated for two main rea~ons.~ To begin with, differences in the wages and mobility of college and non-college educated workers have been at the center of previous research and contemporary policy regarding the impact of economic restructuring. High school-educated men and women lost 16.4 and 3.9 percent of their real hourly wages from 1979 to 1989, while men and women with at least a four-year college degree gained 3.1 and 10.5 percent, re~pectively.~ Although wages deteriorated over the 1980s for all workers at the bottom end of the labor market, not just for the non-college educated, the wage premium for the college edu- cated has been the measure of choice for those who favor education as the key remedy for falling wages and rising inequality. Thus, there has been less attention to within-group inequality because the beneficial role of edu- cation is less obvious in the case of workers with the same formal education. Although a more generalized definition of skill may un- derlie both between-group and within-group inequality (Card and Lemieux 1996), my re- cent investigation of within-group inequality shows that the sources of within- and be- tween-group inequality are sufficiently dif- ferent to require separate treatment (McCall forthcoming).

The second reason I focus on the college wage gap is that it serves as a reasonable in-

Wage inequality between skill groups defined by gender, education, experience, marital status, and race accounted for 57 percent of the increase in the variance in wages from 1979 to 1988, while the remainder of the variance was attribut- able to wage inequality within these detailed skill groups (DiNardo et al. 1996:1033). Over the 1963-1987 period, Katz and Murphy (1992:43) found that about one-third of the change in rela- tive wages occurred among gender, education, and experience groups.

These figures are taken from a study based on the same sample used in Table 2 below (McCall 1998). Data for 1997 indicate that real hourly wages are still below their previous peaks in the 1970s for men and women with a high school degree or less, whereas wages are reach- ing new highs for women with a college degree or more and for men with advanced degrees (Economic Policy Institute 1999).

dicator of stratification or "class" differen- tiation among women, a largely neglected topic in the study of women's economic sta- tus. The longstanding interest in gender in- equality and racial disparities among women has overshadowed the growing divergence in economic opportunities based on access to education-a divergence Teixeira (2000) calls the " 'Great Divide'. . . in today's 'new economy"' (p. 2). As Clement and Myles (1994) argue, one of the more general con- sequences of occupational gender segrega- tion, and specifically the taboo against wo- men managing men, is that women are often supervised by other women. This is increas- ingly the case now that college-educated women from all racial and ethnic groups have advanced into managerial and profes- sional occupations in greater numbers (Bian- chi 1995). Still, gender inequality is not ir- relevant to the present discussion. It has a direct bearing on the interpretation of the college wage gap (and all forms of inequal- ity among women) because economic differ- ences among women are restricted by the glass ceiling and women's segregation into low-paying jobs. Consequently, disparities among women should be understood within the broader context of gender inequality, much like joint analyses of the rise of the black middle class and continuing racial in- equality.

The second major gap in the research lit- erature on women concerns the underlying sources of educational wage differentials and whether they differ for men and women. Determining the sources of inequality a- mong women is of interest in its own right, but a comparative analysis with men is im- portant for both substantive and analytical reasons. Substantively, there is still consid- erable controversy over the weight of differ- ent aspects of economic restructuring in ex- plaining recent trends in inequality. Techno- logical change is the favored explanation of many, but others contend that institutional changes are more important (e.g., deunion- ization and the falling value of the minimum wage); still others highlight international trade competition, immigration, andlor dein- dustrialization (Freeman 1997). The degree of inequality between college-educated and non-college-educated workers is theoreti- cally subject to the same broad social and economic forces for men and for women, but one study has already provided evidence for a much stronger link between the decline in the minimum wage and rising wage inequal- ity among women (DiNardo et al. 1996). Additional evidence on both male and fe- male inequality will advance the broader de- bate by pointing out factors that may be as- sociated with either male or female wage dispersion but not both.

Examining gender differences in these terms also has clear analytical implications for the study of labor market inequality more generally, particularly when family income inequality is used as an indicator of "over- all" inequality or of a labor market's "oppor- tunity structure" (Jargowsky 1997; Nielson and Alderson 1997). Although I do not ex- amine family income inequality here, wage and salary earnings are the primary source of family income, and gender differences in the extent and sources of educational wage differentials suggests that labor markets con- sist of multiple and potentially conflicting opportunity structures concealed by studying family income (Peck 1996). Such differ- ences have consequences for the countless studies using income inequality as a key in- dependent or dependent variable. As an in- dependent variable, family income inequal- ity cannot fully represent the overall struc- ture of labor market inequality if there are multiple and conflicting opportunity struc- tures. And using family income inequality as a dependent variable may be misleading be- cause it is impossible to test for the determi- nants of different types of inequality. Some key aspect of labor market restructuring can be associated with lower inequality of one type (e.g., family income inequality) and higher inequality of another type (e.g., fe- male wage inequality). Therefore, given women's large share of the labor force and high levels of internal differentiation, we should begin thinking of gender analytically in terms of differences in within-gender in- equality.

I address the third major gap (levels of earnings inequality) by testing for gender differences in an analysis of levels of col- lege wage gaps in over 500 regional labor markets in the United States in 1989, a year marking the end of the 1980s surge in in- equality. Surprisingly, there has been little

research on the wide variation in levels of wage inequality (of any kind) across the United States and even less on gender dif- ferences in the determinants of high levels of wage inequality. Given the availability of detailed economic data for counties, the heterogeneous nature of economic condi- tions in regional labor markets, and the large sample sizes afforded by inter-labor-market analysis, spatial research offers a unique op- portunity for examining the relationship be- tween restructuring, educational wage differ- entials, and gender. If certain distinguishing economic characteristics of regional labor markets are associated with higher college wage gaps among women as well as among men, we could conclude that gender differ- ences are relatively unimportant. If signifi- cant gender differences are found, however, much of recent research and debate on this topic may be limited in scope and apply only to certain segments of the workforce.

I improve on earlier spatial analyses by examining these questions using multilevel models and detailed measures of a wide range of key aspects of economic restructur- ing. Previous research has been limited to macro studies of a few measures of broad industry composition (e.g., manufacturing or services employment) and labor market sup- ply and demand (e.g., unemployment and immigration), or it has focused on only one explanatory factor at a time. Immigration is the only factor that has been investigated through extensive area analyses as a cause of rising and high levels of educational wage differentials (Borjas, Freeman, and Katz 1996). Yet, like immigration, many dimen- sions of economic restructuring are geo- graphically concentrated-such as high job- lessness, deindustrialization, and high con- centrations of technology services and man- ufacturing.

Economists DiNardo et al. (1996) have dis- cussed gender differences in the sources of the 1980s rise in wage dispersion, but their approach focused on changes over time in supply, demand, and institutional factors measured in aggregate terms. Nelson and Lorence (1988) provided one of the few spatial analyses of levels of inequality across urban labor markets in 1980. They also found significant differences by gender, but their focus was on the influence of various service sec- tor industries.


Educational wage differentials vary across regions when the same skills are rewarded differently under different social and eco- nomic conditions. This occurs when there is an undersupply or oversupply of workers with certain skills compared with the avail- able jobs requiring such skills in the local labor market (Kain 1992; Wilson 1987). For any particular skill group, an undersupply of workers is expected to result in relative wage premiums, while an oversupply should result in relative unemployment and wage penalties. Contrary to the emphasis on dein- dustrialization, spatiallskill mismatches are not only the consequence of lost unionized manufacturing jobs in selected central cities. They also result from changes or key differ- ences in the broader opportunity structure of the regional economyand are a function of economy-wide transformations taking place across industries (Murphy and Welch 1993) and urban-suburban boundaries (Jargowsky 1997; Kasarda 1995). Regional labor mar- kets offer a strategic setting for the analysis of restructuring and inequality because they capture the net effect of the matching pro- cess between jobs and workers, and demand and supply, that takes place in concrete local labor markets (Granovetter and Tilly 1988; Hanson and Pratt 1995; Peck 1989; Reskin and Roos 1990; Sassen 1995; Waldinger 1996; Wilson 1987).5

A common analytical approach in the im- migration literature (Borjas et al. 1996) and in cross-national studies (Blau and Kahn 1996; Freeman 1994) is to expect that fac- tors responsible for rising aggregate inequal- ity are also responsible for high absolute lev- els of regional inequality. I follow this sim- plifying approach, but two considerations must be kept in mind. First, finding that a given source of rising aggregate inequality is also associated with a high level of local

Blanchflower and Oswald (1994: 3) argue that subnational geographical units represent "mini- economies," while Sassen (1995:115) calls them "economic sub-systems." For a more in-depth dis- cussion of the advantages of studying inequality regionally, see Nielson and Alderson (1997:14- 15) and Xie and Hannum (1996:951-53).

inequality reinforces its explanatory breadth, but finding the opposite says nothing of its influence on rising inequality nationally or even locally. Second, the influence of some factors may not be evident in local concen- trations of inequality but in other spatial dis- tributions. For example, the segregation of high- and low-wage workers into nonspa- tially overlapping markets may lower local levels of inequality but contribute to higher levels in the aggregate.6 As neoclassical equilibrium theory predicts, this outcome would also result if local imbalances in la- bor demand and supply are corrected by per- fect labor and capital mobility in the long- run (Williamson 1996). A more dynamic analysis could address these issues by link- ing changes in the growth, spread, and con- centration of inequality, but that is beyond the scope of this study.

With these considerations in mind, it is relatively straightforward to translate the two leading explanations of rising earnings inequality over time into explanations of variation in levels of college wage differen- tials over space. The first explanation fa- vors a shift in demand to more educated workers brought about by technological change and international trade (Bound and Johnson 1992; Rodrik 1997). The second explanation favors changes in labor market institutions resulting in greater insecurity and flexibility in the low-skill labor markeL7 These changes include deunionization, the decline of the real value of the mini- mum wage and the social wage, deregula- tion, immigration, unemployment, and the growth of alternative work arrangements (e.g., part-time and temporary work, inde- pendent contracting, informal self-employ- ment, etc.) (DiNardo et al. 1996; Kalleberg et al. 1997). For a regional analysis, both explanations identify mechanisms that af-

For example, Harrison (1994) has suggested that the inequality effects of high technology manufacturing are not concentrated in high tech- nology regions, like the Silicon Valley, because the poorest quality assembly jobs are siphoned off to other regions.

I use the terms "flexibility" and "insecurity" interchangeably. Flexibility tends to describe changing employment relationships from the per- spective of employers; insecurity tends to de- scribe them from the perspective of workers.

fect wage levels for broad groups of work- ers cutting across most industries. Regional labor markets with more technology and flexibility should exhibit relatively higher levels of inequality, resembling the national economy as a whole.


Regions with concentrations of high technol- ogy industries in both the service and manu- facturing sectors should trigger a demand for skilled technical workers that raises wages for such workers both inside and outside high technology industries (Blau and Kahn 1994). In addition, nontechnology industries in such regions, because of technological in- fluences, should be more likely to imple- ment technological innovations than their counterparts in other regions. These "spill- over" effects are analogous to those associ- ated with Fordist industrial regions, where union wages for high school-educated men in production jobs had the effect of raising wages for similarly skilled men in nonpro- duction and even nonunionized sectors (Krugman 199 1; Snipp and Bloomquist 1989). The difference is that the advantage in high technology regions is presumably given to the more highly educated workers.

The category of high technology employ- ment must be broken down, however, to de- termine whether the college/non-college wage gap should in fact be any higher in such environments. For example, the wage gap is likely to narrow for men in areas with a concentration of high technology, export- competitive manufacturing industries that tend to be unionized and predominantly male (e.g., defense, aerospace, and chemi- cals). Owing to women's exclusion from premium jobs, whether or not those jobs re- quire a college education, inequality should be lower among women as well (Colcough and Tolbert 1992; Jenson 1989). In contrast, educational wage differentials should not be any lower among men in areas with high technology, import-sensitive manufacturing industries that tend to be nonunionized and female dominated (e.g., electronics and computer-related equipment). The female differential may decline if non-college-edu- cated women are able to earn relatively higher wages in some of these manufactur-

ing industries than in the services.

A high male wage gap and a low female wage gap is also likely in high technology service regions where there are more jobs for both college-educated and non-college-edu- cated women in professional, technical, and clerical occupations and considerable up- grading among college-educated men in the advanced producer service industries. At least one national-level study has found that the effect of technology's spread to the ser- vices has been to fill in the middle of the fe- male wage distribution (Mishel and Bernstein 1996). However, there is no com- parable upgrading expected among non-col- lege-educated men (Brint 199 1 ; Nelson and Lorence 1988). For various reasons, then, high technology regions are expected to raise the wage gap among men but may con- tribute to a lower wage gap among women.8

Institutional explanations of rising inequal- ity in the United States have been based al- most entirely on changes in unionization and the minimum wage. It is difficult to in- vestigate these factors spatially because union coverage rates are only measured at the state and metropolitan levels, and mini- mum wage laws are typically the province of states, not cities or counties. As an alter- native, I examine other key developments associated with employers' search for flex- ibility and an overall erosion of bargaining power among low-skill workers. ~ogether, these developments in effect "deinstitu- tionalize" the local labor market and lead to lower wages at the bottom end. Three such developments have been identified in previ- ous research: (1) unemployment (Blanch- flower and Oswald 1994; Galbraith 1998),

(2) immigration (Borjas et al. 1996; Tope1 1994), and (3) casualization, which is de- fined as temporary work, part-time work, in- dependent contracting, and informal self-

This is consistent with studies by Blau and Kahn (1994) and Katz and Murphy (1992), who argue that technical change favored women rela- tive to men at lower skill levels and men relative to women at higher skill levels.

employment (Kalleberg et al. 1997; Spalter- Roth and Hartmann 1995).

Most previous intermetropolitan area re- search on the effects of unemployment and immigration has been concerned with out- comes related to wage levels rather than dis- persion. A few studies have found wages for males to be more adversely affected by un- employment and immigration than wages for females; wages for high school dropouts are more adversely affected than wages for workers with a high school diploma (Blanchflower and Oswald 1994; Borjas et al. 1992). In terms of casualization, national studies show that women, and especially less-educated women, are concentrated in low-quality nonstandard work arrangements, whereas white men are concentrated in high- quality positions. Some research has also shown that part-time work may offer higher wages among a select group of more highly skilled women (Ferber and Waldfogel 1996; Tilly 1996). It is therefore an open question as to whether these broad economic and in- stitutional factors will affect male and fe- male wage differentials equally. Because low-skill women workers, especially those who also are mothers, are more likely to be concentrated in the low-wage labor market (Waldfogel 1997), and college-educated women are more likely to have made wage gains than college-educated men (according to trends over time), I expect inequality to be greater among women than among men and especially between the top and bottom of the education hierarchy.

To examine the effects of economic restruc- turing on regional collegelnon-college wage differentials, I use a two-level model with detailed data on individuals and regional la- bor markets. In general terms, the model consists of an individual-level and a region- level equation with error terms at each level and uncorrelated across levels. Random (un- observed) and systematic variation is speci- fied for both individual and region equa- tions. This two-level approach has two main advantages over the typical specification of an individual-only or region-only model. First, the model uses detailed information on individuals to guard against aggregation bias (which leads to inflated coefficients and standard errors when regions are the only unit of analysis). Second, the model accounts for unobservable similarities among individuals within the same region and re- gional variation in the degree of homogene- ity (and sample sizes in unbalanced data). Otherwise, standard errors are underesti- mated for region-level variables when indi- viduals are the only unit of analysis. Because the ordinary-least-squares (OLS) assumption that random errors are independent and have constant variance is violated, generalized- least-squares and iterative maximum-likeli- hood techniques are used to estimate the co- efficients and the error component^.^

This type of model was not chosen to compare within-region and between-region variation in wage levels (which is the indi- vidual-level dependent variable). Rather, the discussion centers on the individual-level wage relationship between educational background, an independent variable, and the log of hourly wages, the dependent vari- able. This two-level model offers a better es- timate of the education-wage relationship within regions before analyzing its variation and determinants across regions. In statisti- cal terms, this translates into an investiga- tion of how slopes vary rather than how in- tercepts vary across regions.

The models are run separately for men and women, but for simplicity, the equations are presented below in generic terms. The indi- vidual-level equation is estimated with the natural log of hourly wages as the outcome for individuals in labor markets:

where Y, is the hourly wage of individual i in labor market j calculated from annual earnings data, number of weeks worked, and usual hours worked per week in 1989; Xijis a vector of three binary education variables for individual i in labor market j; and pj is a corresponding vector of randomly varying coefficients, one for each of the j labor mar- kets. The education variable consists of four

I use the HLM program to conduct these analyses. See Bryk and Raudenbush (1992), DiPrete and Forrestal (1994), and Hox and Kreft (1994) for a detailed description of HLM's esti- mation algorithms.

1 categories based on completed years of 1 schooling: less than high school graduate, high school graduate, some college, and col- lege graduate or more. The latter is the ex- cluded category and is estimated by the in- tercept pojwhen all RUvariables are cen- tered around their (gender-specific) labor market means. The PU, P2], and P3jcoefficients measure the log difference between the hourly wage for college grads (Poj) and the hourly wages of each of the other three education groups. These measure the col- legelnon-college wage gaps for each of the j labor markets. A standard vector of RUindividual human capital variables is also in- cluded, and their effects are assumed to be fixed across j. These variables are marital status (married = I), number of own chil- dren, immigrant status (foreign born = l), potential work experience and its square, hours worked (full-time, year-round = I), three binary variables for racelethnicity (black, Asian, Latinalo), and nine binary variables for ten broad industries of employ- ment. The individual error term is denoted by cijand is assumed to have a normal dis- tribution, mean of 0, and constant variance. Variation across regions in the hourly wages of the college educated and the wage gaps between the college educated and non- college educated is estimated by equations 2 through 5:

Each of the four randomly varying coeffi- cients from the micro equations (Poj through /?3j) is a function of the mean across j (yoo~30) and an error term representing the ran- dom deviation of each labor market from the grand mean (voj -v~~).

This random varia- tion is partially explained by a vector of Zj variables describing the economic condi- tions of each labor market j. The associated vectors of fixed coefficients in equations 3 through 5-yl, y2, and y3-represent the ef- fect of labor market characteristics on the portion of wage gaps left over after control- ling for the distribution of observable and unobservable characteristics within regions.

When the full two-level model is specified by substituting equations 2 through 5 into equation 1, it becomes evident that the yo through y3coefficients actually measure the interaction between individual-level (Xij) and region-level variables (Zj) and represent the influence of region-level variables on the relationship between education and wages. The region-level error terms (voj -vdj)are also shared by individuals within the same labor market and therefore interact with in- dividual-level terms as well.lo For purposes of interpretation, the form suggested by equations 3 through 5 is preferred because the variation of micro-adjusted wage gaps across regions is the prime interest. Discus- sion centers on the gender-specific y,, y2, and y3 coefficients, which represent the ef- fect of key dimensions of labor market re- structuring on college/non-college wage gaps separately for men and women. To sim- plify the presentation and to substantively highlight the college/non-college divide, I focus on only two of the three wage differ- entials, those between college graduates (16 or more years of school) and the non-col- lege educated (high school graduate or high school dropout). Hereafter I refer to these as the high school graduate/college graduate wage gap and the high school dropout/col- lege graduate wage gap.

I use two types of data sources, which corre- spond to the micro and macro levels of the model. Individual-level data includes every variable entered in equation 1 and is ob-

lo Following the notation above. and letting q index the three randomly varying education co- efficients (also the number of level-two equations predicting wage gaps), the full two-level equa- tion is:

lnYij= YOO+ Xijl/yo + ZjYo + XrjZjYq + Rij~

+ xijvqj+ voj + Ejj,

where the first term is the intercept, the next two terms are the main effects, the fourth term is the interaction effect of substantive interest, the fifth term is the level-one vector of fixed effects, and the last three terms are the random error compo- nents representing the macro-level (weighted by individual-level characteristics) and individual- level residuals.

tained from the Census of Population Public Use Microdata Samples (5 percent) for 1990 (PUMS-A). The sample includes adults aged 25 to 64 who are not self-employed or farm industry workers and earn hourly wages be- tween $1 .OO and $250.00. Part-time workers are included in the sample, which boosts the sample sizes of labor markets as 40 percent of the female sample works part-time (less than 35 hours per week or less than 50 weeks per year). In addition to this statistical rea- son, there are substantive reasons for includ- ing part-time workers."

Smith (1991) and Bernhardt et al. (1995) note that findings based on full-time work- ers are potentially biased against low-wage women workers (their studies use full-time samples). Prior analyses of income inequal- ity tended to include part-time as well as full-time workers, and most revealed higher levels of inequality among women. But re- cent studies yield mixed results depending on the measure used (DiNardo et al. 1996: 1041; Karoly 1993:90-93; Levy and Mur- nane 1992: 1344-45; Mishel and Bernstein 1994: 144-45). Among recent studies on col- lege wage differentials, those restricted to full-time workers show higher levels of in- equality among men than among women (Gittleman 1994; Katz and Murphy 1992). Combining full-time and part-time workers, however, Card and Lemieux (1996:331-32) present estimates showing higher 1989 edu- cational wage differentials among women in three of four age categories. As these differ- ences in levels of inequality indicate, the in- clusion of part-time workers often provides a different, and perhaps a truer, representa- tion of the workforce. To control for impor- tant differences between full-time and part- time workers, however, I include hours sta- tus in the individual-level equations. An ad- ditional contextual effect is controlled by in- cluding casualization in the region-level equations (described below)

The microdata sample is also restricted to individuals reporting a place-of-work code in order to group individuals by their area of

l1 However, the main substantive conclusions were the same when the analysis was replicated with a sample of full-time, year-round workers. (Coefficients from the full-time samples are available from the author on request.)
Table la. Descriptive Statistics for the Micro-Level Analysis Variable Name Female Sample Mean Standard Deviation     Mean Male Sample Standard Deviation

Log Hourly Wage 2.19 .59 2.54 .61

College graduate or more     .25     .43     .28     .45
Some college     .32     .46     .28     .45
High school graduate     .32     .47     .29     .45
High school dropout     .12     .32     .15     .36

Huma~z Capital

Married Number of own children .76 1.04 .84 1.13
Foreign-born     .10     .30     .ll     .3 1
Work experience (in years)     21.86     10.83     21.78     10.79
Full-time employee     .60     .49     .80     .40



Industry of Employment
Agriculture, fishing, forestry     .OO     .06     .O1     .07
Mining     .OO     .05     .O1     .10
Construction     .O1     .ll     .10     .29
Manufacturing     .15     .35     .27     .44
Transportation. communications,     .05     .22     .I1     .32
and other public utilities                 
Wholesale trade     .03     .17     .06     .24
Retail trade     .14     .35     .ll     .32
Finance, insurance and real estate     .10     .30     .05     .22
Services     .46     .50     .2 1     .41
Government     .05     .23     .07     .25

Note: The samples include adults, ages 25 to 64, working in 554 regional labor markets. The N for the female sample is 1,5 16,73 1; the N for the male sample is 1,7 10,139. Data are derived from the Census of Population Public Use Microdata Samples (5 percent) for 1990.

employment (rather than by residence). This differences in economic development (e.g., facilitates matching with region-level data suburbanization and ruralization of manu- from county employer reports on the work- facturing plants), divide the largest metro- force (rather than with residents) of an area. politan areas into smaller units, and ensure a The regional labor markets are derived from larger sample of labor markets. In size, the smallest geographical units in the PUMAs fall somewhere between large PUMS-A, the Public Use Microdata Areas MSAs and counties, both of which have (PUMAs). PUMAs are state planning dis- been used in other geographical studies (e.g., tricts composed of county parts, intact coun- Abrahamson and Sigelman 1987; Nielson ties, or groups of counties with populations and Alderson 1997). Greater technical detail of 100,000 or more. I use PUMAs rather on the construction and sample of PUMAs, than MSAs because they capture urbanlrural which includes a final total of 554 areas, and

Variable Name

Insecure Employment Conditions

Percent immigrant a
Percent casualizedd
Unemployment ratea

Trade and Technology

Percent high-technology

Percent import-sensitive

Percent high-technology services "

Ind~rstrial Cornposition

Ratio of manufacturing to services

Manufacturing employment growth, 1979-1989C

Mean Standard Deviation Minimum Maximum









Region and Demographic Control Variables
Population (In) a     12.44
Urban "     .56
Northeast a     .20
Westa     .14
Midwesta     .29

The sample includes adults, ages 18 to 64, working in 554 regional labor markets. Data are derived from the fully weighted Census of Population Public Use Microdata Samples (5 percent) for 1990. Data are derived from the 1987 Economic Census Location of Manufacturing Plants for counties. Data are derived from the Regional Economic Information System for counties.

the county-level variables that were matched to the PUMAS is provided elsewhere (Mc- Call 1998). Descriptive statistics on micro and macro variables are presented in Tables la and lb.

The region-level data come from the PUMS-A and independent sources of coun- ty data. Included in Z, is a set of five con- trol variables for unmeasured price differ- ences and other area fixed effects across j: Population size and urban area control for the tendency of wages to be higher in large cities; binary variables for the Northeast, West, and Midwest regions control for dif- ferences in broad regional wage levels and dispersion.

Three additional groups of variables are of substantive interest. First, three measures of insecure employment conditions were con- structed from the entire weighted PUMS-A sample for each PUMA. A composite mea- sure of casualized employment includes the percentage (3-digit) temporary service indus- try workers, part-time and part-year workers, and the self-employed in unincorporated businesses as an indicator of informal self- employment (Belous 1989).12 The two other measures of insecure employment conditions are the percentage of unemployed and the percentage of immigrant workers in the la- bor force. Deunionization is another impor- tant source of increasing job insecurity, but unfortunately there are no data on union cov- erage at the county or PUMA level.

Second, three measures of trade and tech- nology were constructed from the PUMS-A as well as the 1987 Economic Censuses. From the latter data set, which provides population coverage of the location and em-

l2 Other common measures of casualization from other data sources were not available, such as involuntary part-time labor and independent contracting.

ployment size of manufacturing establish- ments by 4-digit SIC groups, two variables were constructed to measure the number of high-technology and import-sensitive manu- facturing establishments as a share of total manufacturing establishments in each county

(U.S. Department of Commerce 1994a). High technology and import sensitive indus- tries were selected based on levels of re- search and development (gathered by the National Science Foundation) and import1 export ratios in 4-digit SIC industries (Bednarzik 1993; Hallock, Hecker, and Gannon 1991; Riche, Hecker, and Burgan 1983; Schoepfle 1982). Strober and Arnold (1987) and Castells (1989), among others, use Riche et al.'s (1983) typology of high technology industries, which also includes service industries. The third variable, em- ployment share in (3-digit) high technology service industries, was derived from the full weighted PUMS-A sample and should be considered a measure of R&D-intensive pro- ducer services linked to the global economy.

I also add measures of broad industrial composition and change, both as controls and as a third group of variables of second- ary substantive interest. Drawing from the Regional Economic Information System (REIS), which provides population employ- ment data for broad industrial categories by county and year (U.S. Department of Com- merce 1994b), I constructed a measure of deindustrialization by calculating the aver- age annual rate of manufacturing employ- ment change between 1979 and 1989. I also include a measure of employment in manu- facturing industries as a share of employ- ment in service industries in 1989 to (1) con- trol for the relative size of each industry in regional labor markets, and (2) identify la- bor markets where manufacturing still exerts a significant presence. Most temporal stud- ies have found industrial shifts to be of mi- nor significance in the explanation of rising wage inequality (Murphy and Welch 1993).

Using a different data set than most previ- ous studies of wage inequality, this analysis is similar in some respects but notably dif- ferent in others. Because my concern is not with how inequality has changed over time, I am not restricted to the annual Current Population Survey (CPS) files, the staple of wage inequality research.13 On the one hand, despite a number of differences in sample restrictions (on valid values for hourly wages, for example), my estimates of educational wage differentials are similar to those found by the few CPS studies that combine part-time and full-time workers and report actual levels, rather than changes in levels, of educational wage differentials (Mishel and Bernstein 1994:26). On the other hand, my estimates of collegelnon-col- lege wage differentials differ in that they are comparable or higher for women than they are for men even in full-time samples.

Table 2 presents unadjusted estimates of collegelnon-college wage differentials for two age groups and for full-time workers and for all workers. The median and mean high school graduatelcollege graduate differential among full-time 18- to 64-year-old workers is about the same for men and women. At roughly -.49, this translates into a ratio of .61 (after taking the antilog)-this indicates that high school-educated full-time workers earn about 61 percent of the wages of college graduates. In contrast, differentials of both median and mean log wages for the three other groups are substantially higher among women. Men's differentials are especially sensitive to age restrictions, declining once lower-waged young men are excluded be- cause they experienced large declines in real wages over the 1980s. For example, among men, the high school dropout/college gradu- ate log wage gap of -.736 for all workers narrows to -.630 when 18- to 24-year-olds are dropped from the sample (a change from

47.9 to 53.3 percent). In contrast, age restric- tions are less pivotal among women because

l3 The advantage of using the decennial cen- sus files is that they are the largest and most ac- curate samples available from the U.S. Census Bureau. The disadvantage relative to the CPS is that hourly wages must be constructed from data on annual earnings and hours worked in the pre- vious year, whereas the CPS provides actual hourly and/or weekly wages at the time of the survey.

Table 2. Median Wage Gap (In, 1995 Dollars) between Non-College Graduates and College Graduates, by Age and Sex, 1989

SexIEducation Group

Women College graduates compared with: High school graduates (In) High school dropouts (In)

Men College graduates compared with: High school graduates (In) High school dropouts (In)

Ages 18 to 64 Ages 25 to 64 All" Full-~ime~ All" Full-Timeb Median (Mean) Median (Mean) Median (Mean) Median (Mean)

-.607 (-,558) -.489 (-,496) -.586 (-,542) -.493 (-,481) -.811 (-,743) -.698 (-,689) -.798 (-,720) -.720 (-,690)

-.513 (-,510) -.484 (-,498) -.434 (-,444) -.416 (-,449) -.736 (-,718) -.664 (-.673) -.630 (-,632) -.603 (-,636)

Note: The pooled sample includes U.S. nonfarm and non-self-employed adults, ages 18 to 64, who are working and earning hourly wages between $1.00 and $250.00; N = 5,270,502. "Includes part-time and full-time workers. Includes full-time, year-round workers who work 35 or more hours per week for 50 or more weeks per


their wages are still quite low at age 25 (com- pare the female high school dropout/college gaps of -.811 and -.798 for all workers aged 18 to 64 and all workers aged 25 to 64, re- spectively). In sum, in the tables and analy- ses to follow, the combination of including part-time workers and excluding young workers (to eliminate the confounding ef- fects of schooling) produces estimates of the college wage gap that are higher for adult women.

Although wage differentials are higher among women at the national level, any single labor market may exhibit greater wage differentials among men. That is, la- bor markets with high female differentials are not necessarily those with high male differentials. But the prior empirical ques- tion is whether the differentials vary spa- tially at all. Table 3 presents the range of variation in both unadjusted and adjusted wage gaps across the 554 regional labor markets. The adjusted gaps reflect controls within each labor market for the full set of individual-level characteristics expressed in equation 1. These adjusted figures are OLS estimates of equations 4 and 5 when all re- gion-level (Zj)variables are set to 0 and the education coefficients vary randomly across

'"ecause variation across regions in wage gaps is at issue, not wage levels. the results for

labor markets.14 The total inter-labor-mar- ket variance for each adjusted wage gap is also provided in Table 6. The range of variation is substantial for both unadjusted and adjusted figures and for both men and women. Adjusted wage gaps remain greater for women and the variance across labor markets is also greater in the female equa- tions (.00656 and .00715) than in the male equations (.00364 and .00691). However, the range of the wage gaps is considerable for both men and women, with a low-end adjusted gap of -.250 for the male high school graduate/college graduate differen- tial and a high-end adjusted gap of -.928 for the female high school dropout/college graduate differential. After taking the anti- logs, the labor market with the highest dif- ferential has a non-college/college wage ra- tio of 39.5 percent, and the labor market with the lowest differential has a ratio al- most twice that (77.9).

While the college wage gap varies sub- stantially across labor markets, its spatial

equation 1 and the baseline one-way ANOVA model are not presented (they are available from the author on request). Of the within-labor mar- ket variation in wages, between one-fifth (for women) and one-fourth (for men) is explained by differences in individual characteristics (be- tween-group), and the rest occurs within detailed demographic groups.

Table 3. Variation in Mean Non-College/College Wage Gaps (In, 1995 Dollars) across 554 Labor Markets, 1989

Unadjusted" Adjustedb

Sex/ Standard Education Grouu MeanC Deviation


College graduates compared with:

High school -.506 ,094 graduates

High school -.620 .I14 dropouts


College graduates compared with:

High school -.406 ,092 graduates

High school -.535 ,124 dropouts

Talculations are based on raw data. Range Range

Standard Minimum Maximum MeanC Deviation Minimum Maximum

-.225 -.845 -.531 ,089 -.301 -.812
-.I62 -.937 -.654 ,097 -.336 -.928

-.090 -.736 -.440 .070 -.250 -.708
-.I20 -.874 -.603 ,093 -.339 -.915

Calculations are adjusted for individual characteristics represented in equation 1.
This is an unweighted mean of the average college wage gap for each of the 554 labor markets.

distribution is not identical for men and women. Table 4 presents bivariate correla- tion coefficients between the male and fe- male wage differentials. Some coefficients are extremely low, such as those between the female high school graduate/college graduate wage gap and both male wage gaps (ranging from .013 to .154). In samples of full-time employees, these cor- relations are higher (between .200 and .367) but still weak, indicating that the part-time female workforce exerts at least some influ- ence on the distinctiveness of the wage structure for females. The female high school dropout/college graduate differential shows more of a correlation with male dif- ferentials, but it is still only weak to moder- ate, ranging from .I95 to .433. The highest correlations are between male and female high school dropout/college graduate wage gaps, suggesting that there may be a general labor market effect for the lowest skilled workers regardless of gender. Overall, though, the weakness of these correlations points to potential gender differences in how this form of inequality is distributed across regions. The question now is whether the economic conditions that foster high educational wage differentials also dif- fer for men and women.

Table 5 presents the effects of the region- level variables separately for men and wo- men and on the micro-adjusted log wage dif- ferentials. Higher than average differentials are indicated by negative coefficients: The log college wage gap widens by the value of the coefficient for every unit change in the associated labor market variable. For example, the high school dropout/college graduate differential for females widens by -.0157 log points for every percentage point increase in the unemployment rate, net of all other labor market variables. To put this in perspective, when all variables are substi- tuted into the region-level equation at their means, the mean high school dropout/col- lege graduate wage gap is -.653. Because the unemployment rate varies from 2.0 to

14.0 percent in the data (see Table lb), the gap narrows to -.589 in a labor market with the lowest unemployment rate and widens to -.777 in a labor market with the highest un- employment rate, nearly a 20-percent swing as a result of regional variation in unemploy- ment rates. Rather than present these ranges for every variable and wage gap combina-

Table 4. Correlation Coefficients between Mean Male and Female Non-College/College Wage Gaps, across 554 Labor Markets, 1989



College graduates compared with: High school graduates High school dropouts


College graduates compared with: High school graduates High school dropouts

Unadjusted" Adjustedb
College Graduates Coinpared with:
High School High School High School High School

Graduate Dropout Graduate Dropout

,023 .0 13 ,121 ,154

.2 12 ,273 ,256 ,309

.031 .03 1 ,184 ,243

,195 .235 ,342 ,433

"Calculations are based on raw data.

Calculations are adjusted for individual characteristics as represented in equation 1 and in equations 4 and 5 when all macro variables (Z) are set to equal 0.

tion, I discuss the direction and significance of the region effects shown in Table 5. I also discuss the percentages of variance explained by each of the three main groups of variables shown in Table 6.

The models examine whether key dimen- sions of labor market restructuring have spa- tially concentrated effects and, if so, whether the effects are the same for men and women. There is considerable evidence, for women especially, that several aspects of labor mar- kets significantly alter the wage gap between college-educated and non-college-educated workers. Table 5 reveals that about half the coefficients for the eight variables of sub- stantive interest are significant at the p < .05 level or lower. Table 6 indicates that up to 33 percent of the variance in wage differentials is explained by these eight variables. This figure is derived by subtracting the percent- age explained by the region and demographic controls in the adjusted female high school graduate/college graduate equation (30.3 percent) from the total variance explained in the full equation (64.0 percent). The compa- rable equations for men explain a much smaller percentage of the variance-at most only 10 percent. In addition to these gender differences, there is only one variable for which the coefficients are significant and in the same direction for both men and women.

These results provide initial evidence for two broad claims. First, differences in the coefficient signs and their significance lev- els indicate that key aspects of economic re- structuring do not have the same impact on wage differentials among women and among men. Second, these diverse measures of re- structuring are better at explaining variation in the female wage differential. The amount of variation explained by area fixed ef- fects-the region and demographic con- trols-is about the same for men and women and is greater than that for any of the three substantive groups of variables. Importantly, though, these control variables make up a much greater portion of total explained variation for men. Thus, other underlying patterns responsible for broad regional var- iation and unexplained within-region varia- tion are central to understanding variation in the wage gap among men and should be the subject of future research. Potential sources of both between-region and within-region variation may be captured by using more de- tailed measures of industrial structure or measures of changes over time in local eco- nomic structure.

While the overall results are robust for women, they are strongest for the group of variables defining flexiblelinsecure labor markets. All three variables measuring inse-

Table 5. Unstandardized Coefficients from the Regression of the Non-College/College Wage Gap (In, 1995 Dollars) on Selected Labor Market Characteristics, 1989

Women College Grads Men College Grads Compared with: Compared with:

High School High School High School High School Labor Market Variable Graduates Dropouts Graduates Dropouts
Intercept Coefficient     -4.621'** (.832)
Insecure Employnzent Conditi Percent unemployed ons     

Percent casualized

Percent immigrant

Trade and Technology

Percent high-technology manufacturing

Percent import-sensitive manufacturing

Percent high-technology services

Industrial Conzposition

Ratio of manufacturing to -.485"*" services (.087)

Percent manufacturing ,016 employment (.O 16)

Region and Demographrc Control Variables

Population (In) -.04 1 (.059)

Urban ,102 (.086)

Northeast .393* * (.100)

West 1.186"*' (.115)

Midwest .766'*' (.088)

Note: Standard errors are in parentheses. Coefficients are derived from the full multilevel model as ex- pressed in equations 4 and 5. All standard errors and coefficients are multiplied by 10 to simplify presenta- tion.

"x < ,ol **"


p I ,001 (two-tailed tests)

cure employment conditions are associated top and bottom of the educational hierarchy with significantly greater log wage differen- is greater under these conditions, as one tials between high school dropouts and col- would expect if the least educated workers lege grads. The negative coefficients for the are assumed to be the most vulnerable to local unemployment rate, the share of im- downward wage pressure in a competitive migrant workers, and the share of and deregulated labor market for low-skill casualized workers are all highly significant workers. However, this vulnerability (p < .01). Inequality between those at the reaches both groups of non-college-edu-

Table 6. Percentage of Variance in Non-CollegelCollege Wage Gaps Explained by Labor Market Characteristics

Percentage of Variance Explained Total Residual

Net of Controlsb

Variance" Variance Total Models (Adiusted) [A1 [Bl [(A-B)lA] All Full-TimeC


College graduates compared with ,00656 high school graduates (adjusted): Region and demographic (RID) controls RID and insecure employment conditions RID and technology and trade RID and industrial composition

RID and all variablesd College graduates compared with ,007 15 high school dropouts (adjusted):

Region and demographic (R1D)controls RID and insecure employment conditions RID and technology and trade RID and industrial composition RID and all variablesd


College graduates compared with high school graduates (adjusted): Region and demographic (RID) controls RID and insecure employment conditions RID and technology and trade RID and industrial composition

RID and all variablesd College graduates compared with ,00691 high school dropouts (adjusted):

Region and de~nographic (RID) controls RID and insecure employment conditions RID and technology and trade RID and industrial composition RID and all variablesd

Taken from the micro-adjusted models represented in equations 4 and 5 when all macro variables (Z) are set to equal 0. The percentage of variance explained by the region and demographic controls alone subtracted from the total percentage of variance explained. For full-time, year-round workers only. Based on the full multilevel models presented in Table 5.

cated workers since the high school gradu- explained by these three variables is similar atelcollege graduate differential is also sig- in the equations for the two wage gaps and nificantly greater in labor markets with high at 21.5 to 22.7 percent is comparable in levels of joblessness and casualization. The magnitude to the explanatory power of in- percentage of variance in wage differentials stitutions in time-series analyses.

In contrast, the amount of regional varia- tion in the male wage gap explained by inse- cure employment conditions is virtually nil, and the substantive conclusions are less con- sistent. On the one hand, the unemployment rate significantly increased the male high school graduatelcollege graduate gap (and increased the high school dropoutlcollege graduate gap at p < .lo) making it the only factor associated with higher inequality for both men and women. This is particularly noteworthy in light of recent declines in in- equality during the tight labor markets of the late 1990s. On the other hand, the male high school graduatelcollege graduate wage gap is actually lower in immigrant-rich regions. Because immigrant status as a compositional effect is controlled at the individual level, one interpretation of this contextual effect is that there is greater competition even among the college educated, at least among men. If true, this would represent a form of across- the-board downward wage compression re- gardless of education that is not as evident among women.I5

Gender differences are also clearly appar- ent in the influence of technology and trade on the college wage gap. Previous research has focused on the increase in demand for higher skilled workers resulting from tech- nological changes, shifts out of import-sen- sitive manufacturing, and other unobserv- able competitive pressures. Based on the size of residual effects rather than on direct evidence, technology in particular is alleged to be the major factor increasing earnings inequality over time. These conclusions per- tain mainly to men, however, while less edu- cated women may have fared better (at least relative to less educated men) because of their concentration in clerical jobs in post- industrial industries and because of the op- portunity to earn marginally better wages in import-sensitive manufacturing industries. Tables 5 and 6 suggest that high school-edu- cated women are doing better relative to fe- male college graduates in labor markets spe- cializing in high-technology service indus- tries and, with less certainty, in labor mar-

l5 This would also explain why casualization lowers the wage gaps among male full-time workers (results available on request from the author).

kets with a disproportionate share of import- sensitive manufacturing establishments (p < .lo). Moreover, the group of technology and trade variables explains 15.3 percent of the variation in the adjusted high school gradu- atelcollege graduate wage differential.16 Un- like for women, these two indicators of trade and technology concentration result in sig- nificantly greater wage differentials among men, but they explain only 4.6 percent of the variation. If technological changes are large- ly responsible for higher inequality among either men or women, they may be more discernable in inequalities between regions rather than within regions. Alternatively, their impact may be felt in rising rates of in- equality within regions.

A third possible explanation for variation in levels of wage inequality is the industrial composition of employment and the decline of manufacturing in particular. Surprisingly, the ratio of manufacturing to service em- ployment resulted in a significantly greater wage gap among women for both wage gaps (high school dropoutlcollege graduate and high school graduatelcollege graduate), and the coefficient for the male high school dropoutlcollege graduate gap was in the same direction (p < .lo). In addition, manu- facturing growth resulted in a greater high school graduatelcollege graduate gap among men. Thus, by 1990, relatively strong manu- facturing presence and growth do not result in greater wage equality between college- educated and non-college-educated men as past research on the positive spillover effects of manufacturing has suggested. Yet looking at the inverse of these measures suggests that manufacturing decline advances high school graduatelcollege graduate equality among men, an effect that probably is pick- ing up the heavily deindustrialized cities of the North Central region. Overall, however, these variables have virtually no explanatory

l6 Table 6 indicates that the group of trade and technology variables explains 8.2 percent of the variation in the female high school dropout/col- lege graduate gap as well. In the reduced equa- tions (including region and demographic controls and trade and technology variables), both tech- nology variables were positive and significant, presenting further evidence of lower iequality among women in these regions.

power in the equations for men and only 5 to 6 percent in the equations for women. Thus, in keeping with the temporal evidence, I find that broad industrial differences are not critical spatial determinants of levels of educational wage inequality, although a sig- nificant negative effect among women should be noted.


According to the descriptive data, the col- legelnon-college wage gap among adults aged 25 to 64 is at least as high among women as among men in both full-time and combined part-timelfull-time samples. Al- though this is not significantly at odds with earlier research, recent studies of full-time workers have generally found wage inequal- ity to be greater among men. As a first step, highlighting the high degree of educational wage inequality among women may draw greater attention to it as an emerging social issue.

More important, though, is a second step: The need to move beyond the gender wage gap as the central indicator of women's eco- nomic status. The emphasis on the gender wage gap has obscured growing stratifica- tion among women in at least two contradic- tory ways. From one perspective, women appeared to be doing relatively well-real wages have fallen more for men than for women and disparities between them have declined. In effect, an emphasis on relative advances among women as a whole, and vis- ible absolute gains among more highly edu- cated women in particular, came at the ex- pense of the worsening situation of low- skilled women, whose real wages have been falling. From another perspective, the aver- age gender wage gap was still considered unacceptably high as women were entering the labor force in unprecedented numbers. This not only kept the gender wage gap on the research agenda, but maintained a cer- tain continuity with studies predating eco- nomic restructuring. As a result, the key role of labor market restructuring in women's working lives was often overlooked. Be- cause these perspectives treated women more or less as a single group relative to men, economic restructuring and its relation to emerging forms of stratification among

women appeared either less visible or less

pressing from both viewpoints.

Given this background, it is not a trivial

matter to argue for the analytical importance

of gender in studies of what Wilson (1997)

calls the "new social inequality," construed

broadly to include all forms of increasing in-

equality (which by definition excludes the

gender wage gap), though rarely referring to

inequality among women as a particular con-

cern. If there are gender differences in the

sources of educational wage differentials

and women constitute nearly half the

nation's labor force, it would seem costly to

ignore inequalities among women or to sub-

sume them under the rubric of male wage

inequality or family income inequality. The

evidence of gender differences in the effects

of flexiblelinsecure employment conditions

and technology on wage differentials chal-

lenges the representativeness of these stan-

dard measures of "overall" inequality. While

these findings point to the need for a new

analytical emphasis on gender differences in

within-gender structures of inequality, they

also provide new evidence to evaluate sub-

stantive claims about the underlying causes

of the college wage gap.

For women, insecure employment condi-

tions were the most pivotal factors fostering

inequality between college graduates and the

non-college educated (especially those with-

out a high school degree). Joblessness, cas-

ualization, and immigration explained about

one-fifth of regional variation in wage gaps

among women, more than was explained by

technology, trade, and broad industrial struc-

ture (combined) and net of regional and de-

mographic controls. These insecure employ-

ment conditions are associated with in-

creased competition and lower wages among

less educated workers and should be consid-

ered defining characteristics of a "dein

stitutionalized" labor market. So far, how-

ever, they have received less empirical sup-

port as explanations for rising inequality

over time, but they are consistent with recent

trends: Tighter labor markets and rising

minimum wages seem to have reduced in-

equality somewhat and reinforced support

for institutional explanations (Galbraith

1998). Indeed, high joblessness alone sig-

nificantly increased college wage differen-

tials for both men and women, although the adverse effects of a deinstitutionalized labor market are concentrated among less edu- cated women. These results are also consis- tent with benefits for college-educated wo- men, either through the enhanced flexibility that valued professional and managerial women can demand (Tilly 1996) or the greater substitutability of college-educated women for more highly paid college-edu- cated men. This analysis cannot distinguish between these possibilities, but other re- search suggests that both influences are at work (McCall 1998).

For men, the role of technology and trade takes on greater importance than for women, yet its influence must be qualified in several respects. On the one hand, male collegelnon-college wage gaps were signifi- cantly wider in labor markets with a dispro- portionate share of high technology service and import-sensitive manufacturing indus- tries. ~echnolo~y

and trade variables also explained more of the spatial variation in wage gaps than did insecure employment conditions and broad industrial structure (combined). These findings seem to corroborate claims about skill-biased techno- logical change and the intensifying dualism of postindustrial economies that provide fewer low-skill, high-wage jobs while bid- ding up the earnings of highly skilled work- ers in technology- and knowledge-intensive service industries (Harrison and Bluestone 1988; Sassen 1991; Levy and Murnane 1992; Bound and Johnson 1992). On the other hand, only a small portion of the re- gional variance in men's inequality was ex- plained by these factors (roughly 5 percent) because almost all of the explained varia- tion was absorbed by fixed region and de- mographic controls. Moreover, the female high school graduatelcollege graduate wage differential is significantly lower in high technology service and import-sensitive re- gions, and the amount of variation explained by these factors is both substantial and greater than in the equations for men (8 to 15 percent). To better interpret these gen- der differences, we need to know whether lower inequality among women is a conse- quence of the glass ceiling among college- educated women-and greater gender in- equality-or a genuine upgrading among high school-educated women. Because pre-

vious research suggests that, again, both in-

fluences may be at work (Blau and Kahn

1997; McCall 1998), gerzeralizations about

the structure of inequality in postindustrial

economies, commonly based on men's ex

periences only, would be difficult to sub-


In sum, the method employed in this study capitalizes on regional variation in economic conditions and inequality to investigate whether generalizations of this sort are plau- sible, that is, whether given underlying con- ditions produce the same or different out- comes for men and women. A comparative framework of this kind could be extended to include several other forms of inequality not considered here. Currently, however, research on inequality is typically limited to the analysis of only one aspect of inequality or restructuring at a time. There is already a large body of research on male inequality, family income inequality, and gender in- equality. Although each form is important in its own right, the findings of this study re- veal significant differences in levels as well as sources of educational wage inequality that are best uncovered in a comparative framework. These differences need to be taken into consideration in future studies of restructuring and inequality at both the mi- cro and macro levels. While a macro-level analysis such as this one can identify simi- larities and differences in the sources of various types of inequality, micro studies are needed to determine more precisely how multiple and potentially conflicting opportu- nity structures develop, persist, and change in specific local labor markets. The dynam- ics of how inequality is structured within re- gions and relative to other regions, and how levels of inequality are related to changes in inequality, are important questions to ad- dress in future comparative case-study and macro-level research.

Leslie McCall is an Assistant Professor of Soci- ology and Women's Studies at Rutgers Univer- sity. She is interested in the areas of economic change and social inequality and is completing a forthcoming book (from Routledge) on the struc- ture and politics of gender, class, and racial wage inequality in U.S. labor markets. Her other research interests include social theory and methodology. Next year, she will be a visiting

scholar at the Russell Sage Foundation.

Abrahamson, Mark and Lee Sigelman. 1987. "Occupational Sex Segregation in Metropoli- tan Areas." American Sociological Review 52: 588-97.

Bednarzik, Robert. 1993. "An Analysis of U.S. Industries Sensitive to Foreign Trade, 1982- 1987." Monthly Labor Review 2: 15-31.

Belous, Richard. 1989. The Contingent Economy. Washington, DC: National Planning Associa- tion.

Bernhardt, Annette, Martina Morris, and Mark Handcock. 1995. "Women's Gains or Men's Losses? A Closer Look at the Shrinking Gen- der Gap in Earnings." American Journal of So- ciology 101 :302-28.

Bianchi, Suzanne. 1995. "Changing Economic Roles of Women and Men." Pp. 107-54 in State of the Union: America in the 1990s, vol. 1, edited by R. Farley. New York: Russell Sage Foundation.

Blanchflower, David and Andrew Oswald. 1994. The Wage Curve. Cambridge, MA: MIT Press.

Blau, Francine. 1998. "Trends in the Well-Being of American Women, 1970-1995." Journal of Economic Literature 36: 112-65.

Blau, Francine and Lawrence Kahn. 1994. "Ris- ing Wage Inequality and the U.S. Gender Gap." American Economic Review 84(2):23-8.

. 1996. "International Differences in Male Wage Inequality: Institutions versus Market Forces." Journal of Political Economy 104: 791-837.

. 1997. "Swimming Upstream: Trends in the Gender Wage Differential in the 1980s." Journal ofLabor Economics 15(1):1-42.

Borjas, George, Richard Freeman, and Lawrence Katz. 1992. "On the Labor Market Effects of Immigration and Trade." Pp. 213-44 in Immigration and the Work Force, edited by G. Borjas and R. Freeman. Chicago, IL: Univer- sity of Chicago Press.

. 1996. "Searching for the Effect of Immi- gration on the Labor Market." AEA Papers and Proceedings 86(2):246-5 1.

Bound, John and George Johnson. 1992. "Changes in the Structure of Wages in the 1980's: An Evaluation of Alternative Explana- tions." American Economic Review 82:37 1-92.

Brint, Stephen. 1991. "Upper Professional: A High Command of Commerce, Culture, and Civic Regulation." Pp. 153-204 in Dual City: Restructuring New York, edited by J. Mollen- kopf and M. Castells. 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.

Card, David and Thomas Lemieux. 1996. "Wage Dispersion, Returns to Skills, and Black-White Wage Differentials." Journal of Econometrics 74:319-61.

Castells, Manuel. 1989. The Informational City. Cambridge, MA: Basil Blackwell.

Clement, Wallace and John Myles. 1994. Relations of Ruling: Class and Gender in Post- industrial Societies. Buffalo, NY: McGill-Queens Univesity Press.

Colcough, Glenna and Charles Tolbert. 1992. Work in the Fast Lane. Albany, NY: State Uni- versity of New York Press.

DiNardo, John, Nicole M. Fortin, and Thomas Lemieux. 1996. "Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach." Econometrica 64: 1001-44.

DiPrete, Thomas and Jerry Forristal. 1994. "Mul- tilevel Models: Methods and Substantive Ap- plications." Annual Review of Sociology 20: 331-57.

Economic Policy Institute. 1999. "Average Real Hourly Wages of Male and Female Workers by Education, 1973-1997." Economic Policy In- stitute:Washington, D.C. Retrieved May 26, 1999 ( .

Ferber, Marianne and Jane Waldfogel. 1996. "'Contingent' Work: Good, Bad, or Indiffer- ent?" Cambridge, MA: Radcliffe Public Policy Institute.

Freeman, Richard, ed. 1994. Working Under Dif- ferent Rules. New York: Russell Sage Founda- tion.

. 1997. When Earnings Diverge: Causes, Consequences, and Cures for the New Inequal- ity in the U.S. Report #284, National Planning Association, Washington, DC.

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

Gittleman, Maury. 1994. "Earnings in the 1980s: An Occupational Perspective." Monthly Labor Review 117(7): 16-27.

Granovetter, Mark and Charles Tilly. 1988. "In- equality and Labor Processes." Pp. 175-219 in The Handbook of Sociology, edited by N. Smelser. Newbury Park, CA: Sage.

Hallock, Paul, Daniel Hecker, and Joseph Gannon. 1991. "High Technology Employ- ment: Another View." Monthly Labor Review 114(7):26-30.

Hanson, Susan and Geraldine Pratt. 1995. Gender; Work, and Space. New York: Routledge. Harrison, Bennett. 1994. Lean anrl Mean. New

York: Basic Books.

Harrison, Bennett and Barry Bluestone. 1988. The Great U-Turn. New York: Basic Books.

Hox, Joop and Ita G. G. Kreft. 1994. "Multilevel Analysis Methods." Sociological Methods and

Research 22:283-99.

Jargowsky, Paul. 1997. Poverty and Place: Ghet- tos, Barrios, and the American City. New York: Russell Sage Foundation.

Jenson, Jane. 1989. "The Talents of Women, the Skills of Men: Flexible Specialization and Women." Pp. 141-55 in The Transformation of Work, edited by S. Wood. Winchester, MA: Unwin Hyman Inc.

Kain, John. 1992. "The Spatial Mismatch Hy- potheses: Three Decades Later." Housing Pol- icy Debate 3:371-460.

Kalleberg, Arne, Edith Rasell, Ken Hudson, David Webster, Barbara Reskin, Naomi Cass- irer, and Eileen Appelbaum. 1997. Nonstandard Work, Substandard Jobs: Flexible Work Arrangements in the U.S. Washington, DC: Economic Policy Institute.

Karoly, Lynn. 1993. "The Trend in Inequality among Families, Individuals, and Workers in the United States: A Twenty-Five Year Per- spective." Pp. 19-97 in Uneven Tides: Rising Inequality in America, edited by S. Danziger and P. Gottschalk. New York: Russell Sage Foundation.

Kasarda, John. 1995. "Industrial Restructuring and the Changing Location of Jobs." Pp. 215- 68 in State ofthe Union: America in the 1990s, vol. 1, edited by R. Farley. New York: Russell Sage Foundation.

Katz, Lawrence and Kevin Murphy. 1992. "Changes in Relative Wages, 1963-1987: Sup- ply and Demand Factors." Quarterly Journal of Economics 107(1):35-78.

Krugman, Paul. 199 1. Geography and Trade. Cambridge, MA: MIT Press. Levy, Frank. 1998. The New Dollars and Dreams. New York: Russell Sage Foundation.

Levy, Frank and Richard Murnane. 1992. "U.S. Earnings Levels and Earnings Inequality: A Review of Recent Trends and Proposed Expla- nations." Journal of Economic Literature 30: 1333-81.

McCall, Leslie. 1998. "Spatial Routes to Gender Wage (1n)equality." Economic Geography 74: 379-404.

. Forthcoming. "Explaining Within-Group Wage Inequality in U.S. Labor Markets." Demography.

Mishel, Lawrence and Jared Bernstein. 1994. The State of Working America 1994-1995. Armonk, NY: M. E. Sharpe. . 1996. "Technology and the Wage Struc- ture: Has Technology's Impact Accelerated Since the 1970s?" Paper presented at the NBER Labor Studies Workshop, July, Cam- bridge, MA.

Mishel, Lawrence, Jared Bernstein, and John Schmitt. 1997. The State of Working America 1996-1997. Armonk, NY: M. E. Sharpe.

Morris, Martina, Annette D. Bernhardt, and Mark

S. Handcock. 1994. "Economic Inequality: New Methods for New Trends." American So- ciological Review 59:205-19.

Murnane, Richard and Frank Levy. 1996. Teaching the New Basic Skills. New York: Free Press.

Murphy, Kevin and Finis Welch. 1993. "Indus- trial Change and the Rising Importance of Skill." Pp. 101-32 in Uneven Tides: Rising In- equality in America, edited by S. Danziger and

P. Gottschalk. New York: Russell Sage Foun- dation.

Nelson, Joel and Jon Lorence. 1988. "Metropoli- tan Earnings Inequality and Service Sector Employment." Social Forces 67:492-5 11.

Nielsen, Fran~ois and Arthur S. Alderson. 1997. "Income Inequality in U.S. Counties, 1970 to 1990." American Sociological Review 62: 12-


Peck, Jamie. 1989. "Reconceptualizing the Local Labour Market: Space, Segmentation and the State." Progress in Human Geography 13:42

61. . 1996. Work-Place: The Social Regula- tion of Labor Markets. New York: Guilford.

Reskin, Barbara and Patricia Roos. 1990. Job Queues, Gender Queues: Explaining Women's Inroads into Male Occupations. Philadelphia, PA: Temple University Press.

Riche, Richard, Daniel Hecker, and John Burgan. 1983. "High Technology Today and Tomor- row: A Small Slice of the Employment Pie." Monthly Labor Review 16(11):50-59.

Rodrik, Dani. 1997. Has Globalization Gone Too Far? Washington, DC: Institute for Interna- tional Economics. Sassen, Saskia. 1991. The Global City. Princeton, NJ: Princeton University Press. . 1995. "Immigration and Local Labor Markets." Pp. 87-127 in The Economic Soci- ology of Immigration, edited by A. Portes. New York: Russell Sage.

Schoepfle, Gregory. 1982. "Imports and Domes- tic Employment: Identifying Affected Indus- tries." Monthly Labor Review 105(8): 13-26.

Smith, Shelley. 1991. "Sources of Earnings In- equality in the Black and White Female Labor Forces." Sociological Quarterly 32: 117-38.

Snipp, C. Matthew and Leonard E. Bloomquist. 1989. "Sociology and Labor Market Structure: A Selective Overview." Research in Rural So- ciology and Development 4: 1-27.

Spalter-Roth, Roberta and Heidi Hartmann. 1995. "Contingent Work: Its Consequences for Economic Well-Being, Gender Division of La- bor, and the Welfare State." Washington, DC: Institute for Women's Policy Research.

Strober, Myra and Carolyn Arnold. 1987. "Inte- grated Circuits/Segregated Labor: Women in Computer Related Occupations and High-Tech Industries." Pp. 136-82 in Computer Chips and Paper Clips, vol. 2, edited by H. Hartmann. Washington, DC: National Academy Press.

Teixeira, Ruy. 2000. "America's Forgotten Ma- jority." Retrieved January 16, 2000 (http:/f

7424.asp). Tilly, Chris. 1996. HalfA Job: Bad and Good Part-Time Jobs irz a Changirzg Labor Market.

Philadelphia, PA: Temple University Press.

Topel, Robert. 1994. "Regional Labor Markets and the Determinants of Wage Inequality." AEA Papers and Proceedings 84(2): 17-22.

U.S. Department of Commerce. 1994a. 1987 Economic Census CD-ROM. Vol. 1, Report Se- ries 1E [MRDF]. Washington, DC: The Bu- reau of the Census [producer and distributor]. . 1994b. Regional Economic Information System CD-ROM, 1969-1992 [MRDF]. Wash- ington, DC: Bureau of Economic Analysis [producer and distributor].

Waldfogel, Jane. 1997. "The Effect of Children on Women's Wages." American Sociological

Revievt, 62:209-17.

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

Weinberg, Daniel. 1996. Current Population Re- ports P60-191. Washington, DC: U.S. Bureau of the Census.

Williamson, Jeffrey. 1996. "Globalization and Inequality Then and Now: The Late 191h and Late 20fh Centuries Compared." Working Pa- per No. 5491. National Bureau of Economic Research, Cambridge, MA.

Wilson, William Julius. 1987. The Truly Disad- vantaged. Chicago, IL: University of Chicago Press. . 1997. "The New Social Inequality and Affirmative Opportunity". Pp. 57-77 in The New Majority: Tovt,ard a Popular Progressive Politics, edited by S. Greenberg and T. Skoc- pol. New Haven, CT: Yale University Press.

Xie, Yu and Emily Hannum. 1996. "Regional Variation in Earnings Inequality in Reform Era Urban China." American Journal of Sociology 101:950-92.

  • Recommend Us