integrating economic dualism and labor market ... - CiteSeerX

INTEGRATING ECONOMIC DUALISM A N D LABOR MARKET SEGMENTATION: The Effects of Race, Gender, and Structural Location on Earnings, 1974-2000 Lesley Williams Reid Georgia State University

Beth A. Rubin University of North Carolina at Charlotte Although the U.S. economy 01 the early twenty-first century is vastly different from the US. economy prior to the 1970s: the nature of these economic changes and their impact on US. workers is unclear. This article claims that despite contemporary economic shifts. differential labor and employer power continues to segment the economy, and workers' position in the labor market continues to predict their rewards. beyond the effects of gender, race, and human capital. Drawing on segmented labor market and dual economy research, we propose a four-category model of the structural factors that influence variance in work-relatcd rewards. We examine the distribution of jobs in each of four categories betwecn 1974 and 2000 and observe that losses and gains across categories are unevenly distributcd by race and gender. While white men have experienced the greatest dcclincs in employment and earnings, they have maintained their absolute advantage over women and nonwhites. In multivariate analyses, we fmd that the structural position of employmcnt continues to be a significant determinant of wages. Although women and racial minorities have expcrienced sizable increases in employment in primary labor market jobs in the core of the economy. both groups remain overrepresented in low-paying jobs. Moreover women, but not nonwhite men. consistently receive significantly fewer rewards for their labor in both low-paying and high-paying jobs. Our findings suggest that structural factors continue to influence earnings inequality, cspecially across race and gender lines.

It is virtually irrefutable that the U.S. economy has changed dramatically over the past thirty years, as the foundation of the country's economic infrastructure has shifted from a bureaucratic, industrial economy characterized by relative production and employment stability to a postbureaucratic, mixed economy characterized by flexibility Direct correspondence to Lesley Williams Reid, Department of Sociology, Georgia State University, University Plaza, Atlanta,

GA 30303-3083;c-mail: lcsleyreid(?pru.edu

The Sociological Quarterly, Volume 44, Number 3, pages 405432. Copyright 0 2003 by The Midwest Sociological Society. All rights reserved. Send requcsts for permission to reprint to: Rights and Permissions, University of California Press, Journals Division, 2000 Center St., Ste. 303, Berkeley, CA 94704-1223. ISSN: 0038-0253; online ISSN 1533-8525

406

THE SOCIOLOGICAL QUARTERLY Vol. 44/No. 3/2003

in its organizations, production techniques, and employment (Harvey 1989; Rubin 1995; 1996;Vallas 1999). It is similarly irrefutable that a primary result of this change has been the fundamental reorganization of the nature of work in the United States. The impact of these changes on the structure of US. wage inequality, however. is less clear. Some analysts of the new economy have focused on the increased need for and reliance on skilled labor, arguing that high performance work structures have improved working conditions for all workers and lessened racial and gender inequality as the premium on skill in the flexible economy has eroded historical patterns of discrimination (Bailey 1987; Appelbaum, Bailey, Berg, and Kalleberg 2001 ; Berg and Kalleberg 2001). In contrast, those who have focused on the “dark side” of the flexible economy (Harrison 1994; Rubin 1995) have suggested that while some jobs have been upgraded, a twopronged approach to obtaining flexibility has bifurcated working conditions and wages and has, in fact, decreased the overall quality of employment for many, especially for women and minorities. In efforts to untangle some of these debatcs, scholars have examined the wage polarization that has resulted from industrial and occupational shifts (Morris, Bernhardt, and Handcock 1994; Bernhardt, Morris, and Handcock 1995). Likewise, efforts to understand the stratification consequences of the new economy have turned to identlfying the operation of networks associated with labor market processes (Elliott 1999;2000; Petersen, Saporta, and Seidel2000), to reconceptualizing class (Wright 1985;2000; Sorensen 1996; 2000; Goldthorpe 2000), to debating the role of occupational sex segregation (Reskin 1993: England, Hermson, and Cotter 2000;Tam 2000), and to examining more mesolevel intraorganizational processes (Petersen 1992).What has fallen by the academic wayside is consideration of the continued role of segmented labor markets and dual economic sectors in shaping work outcomes. We do not contest Petersen’s (1992, p. 100) claim that the (not so) new structuralists’ position that economic and labor market structures are the primary determinants of inequality “may be too restrictive.” Disaggregated studies of stratification structures and processes, whether at the network or organizational levels, have illuminated important components of stratification. Still, we argue that understanding stratification processes at a more aggregate level may continue to be quite useful, that the point of stratification research is not to assert that either the micro, meso, or aggregate structures and processes are the primary determinants of stratification but rather to move back and forth among them in order to see how, as the overall economy and social world changes, these different structures and processes continue to, or fail to. act in “known“ ways. This is the motivation for our research. We expect that differential employer and labor positions continue to segment the economy and that workers’ position in the labor market continues to predict their rewards. We further expect that “good” jobs, located in the primary labor market and the core of the economy, and “bad” jobs, located in the secondary labor market and the periphery of thc economy, continue to be distributed unequally among black and white and male and female workers. In sum, we seek to assess whether labor market segmentation and dual economic sectors are still important contributors to structured inequality. Before we address these hypotheses empirically, we first provide a rationale for the continued theoretical utility of research on segmented labor markets and the dual economy. Second, we describe the conditions of the contemporary economy that make such research empirically compelling. We then turn to our quantitative investigation.

Integrating Economic Dualism and Labor Market Segmentation

407

SEGMENTED LABOR MARKET A N D DUAL E C O N O M Y RESEARCH

In the 1970s,researchers developed structural theories of wage determination that challenged the then-dominant status attainment models. One variant located processes of wage determination in industry structures, focusing specifically on the bifurcation of industries into core and peripheral sectors. This research identified the economic core as composed of monopolistic industries that are large in size and influence, diversified, capital and technology-intensive. and operating in national and international markets. The economic periphery, by contrast, is composed of small firms characterized by restricted markets, high risk, a small number of employees, and labor-intensive production techniques. Employment outcomes differ by sector. Higher wages, more generous benefits, career ladders, and greater work satisfaction characterize employment in the core, and the opposite conditions characterize employment in the economic periphery (Averitt 1968; Kalleberg, Wallace, and Althuaser 1981). A second structuralist theory of earnings inequality addressed the existence of two occupational labor markets, the primary labor market and the secondary labor market. The primary labor market consists of jobs that require skills, training, and education and that are characterized by high wages, stable employment, bureaucratic management, opportunities for advancement, and union protection. In contrast, secondary labor market jobs are constrained by neither structural limitations nor union demands. Jobs in this segment of the labor market generally require few skills, necessitate little job training, provide minimal job security, pay low wages, and provide limited mobility (Doeringer and Piore 1971; Gordon 1972; Osterman 1975). Though the structuralist approach to stratification improved considerably on purely individualistic approaches, research from both the dual economy and segmented labor market frameworks was infrequent by the 1990s.The decrease in structuralist research in the last decade resulted from internal weaknesses of the theoretical framework and external competition from new theoretical approaches to understanding income inequality. Internally, two weaknesses undermined dual economy and Segmented labor market research. First, economic dualism and labor market segmentation research was strongly criticized for the assumption of parallelism of sectors. The preponderance of rescarch on the structural determinants of work outcomes, either explicitly or implicitly, assumed a parallel structure between segmented labor markets and dual economic sectors, locating primary labor market occupations in the core of the economy and secondary labor market occupations in the periphery of the economy (Beck. Horan and Tolbert 1978; Hodson 1978; Edwards 1979;Bluestone and Harrison 1982;Raffalovich 1994; Gittleman and Howell 1995). A number o f theoretical problems are endemic to this approach. Equating the economic core with the primary labor market and the economic periphery with the secondary labor market is descriptive and atheoretical (Hodson and Kaufman 1982). Further, this orientation does not account for the ways that parallelism IS, or more likely is not, empirically supported (Parcel and Sickmcier 1988). In sum, the assumption of parallelism obscured the complexity of the organization of work (see also Grodsky and Pager 2001). Second, both dual economy and segmented labor market research faced continuing debate over the demarcation of sectors (Boston 1990). Most contentious in this debate were the allegations of tautological reasoning in research that attempted to empirically define either labor markets or economic sectors. In a lrrequently cited exchange, Hodson

408

THE SOCIOLOGICAL QUARTERLY Vol. 44I“o. 3/2003

and Kaufman (1981) contended that Tolbert, Horan, and Beck (1980) used circular reasoning in their attempt to define economic sectors. According to Hodson and Kaufman’s (1981) critique, Tolbert, Horan, and Beck (1980) defined their economic sectors with the same variables their analysis was attempting to explain. Horan, Tolbert, and Beck (1981) responded that their demarcation of sectors was based on aggregate-level data while their empirical analyses examined individual-level data. Despite the rigor of Horan, Tolbert, and Beck’s (1981) response to Hodson and Kaufman (1981), the tautology critique continues to be levied against segmented labor market and dual economy research. More important, perhaps, is the theoretical competition to segmentation approaches. In the past ten years, research has increasingly focused on less highly aggregated explanations of structured inequality. A key critique of status attainment research and hence contribution of the segmentation approach is that position determines rewards independent of individual performance. Yet subsequent research has challenged this claim. Petersen’s (1992) analysis of establishment-level data, for example, demonstrated the failure of structuralist explanation% at the industry or labor market level, to account for wage variation within establishments. He demonstrated that within-establishment variable pay-systems differences created opportunities for variable performance that generated unequal rewards. Similarly, research at lower levels of aggregation have addressed issues of race and gender inequality to which neither segmented labor market or dual economy research adequately attended (Boston 1990).*Reskin (Reskin 1988; Reskin and Roos 1990) and others (Bielby and Baron 1986:Tomaskovic-Devey 1993) have demonstrated that much of the observed wage difference between men and women was accounted for by occupational and, more importantly, job-level sex segregation. Likewise, studies increasingly focused on networks as sources of inequality. Scholars demonstrated that the size, type, and location of one’s networks are important contributors to structured inequality (Granovetter 1974; Munch, McPherson, and Smith-Lovin 1997; Elliott 1999;Fernandez, Castilla, and Moore 2000; Petersen et al. 2000). While it is certainly possible to examine industries and occupations as networked, that has not been the practice, and the network analyses within stratification have focused on networks of individuals. We do not dispute the importance of these approaches and, moreover, recognize some of the limitations of segmentation approaches that led to their decline. In light of the apt criticisms of the segmented labor market and dual economy approaches, why do we suggest that they remain useful theoretical tools? Although not without flaws, structural theories of work outcomes continue to aid in understanding contemporary earnings inequality for three reasons. First, growing wage inequality is a persistent characteristic of the contemporary economy. Recent multilevel studies of race and gender stratification suggest that industrial and occupational structures may still be relevant to understanding wage dispersion (Grodsky and Pager 2001; McCall 2001). The segmented labor market/dual economy approach may inform our understanding of the persistence of aggregate-level structural factors that influence wage inequality. Second, debates about the new economy have renewed research emphasis on the role of human capital in determining work outcomes, including wages (Cappelli 1997). Though human capital may continue to predict wages in the new economy, labor market and industry effects also may remain important.

Integrating Economic Dualism and labor Market Segmentation

409

Finally, race and gender remain significant sources of inequality in flexible employment (McCall2001). New economy analysts have shown that one of the ways by which employers create more flexible uses of labor is to use a strategy of “static” flexibility in which they use labor as needed. Research indicates that women and minorities are most likely to occupy those positions in which workers have the most insecure, nonstandard employment contracts and are least likely to occupy positions in which they develop new skills and opportunities for mobility (Smith 1993; Cornfield, Campbell, and McCammon 2001). Research on the “casualization” of the labor market implies a growth in the size of the secondary labor market. Inasmuch as this is the case, it certainly points to the continued utility of thinking in segmentation termc The segmented labor market/ dual economy framework allows for the integration of race, gender, and class effects in studies of the determinants of work outcomes (Cotter, Hermsen, and Vanneman 1999). With this in mind, we turn now to a discussion of how earnings inequality has varied by race and gender over the time period of interest to this research,

CONTEMPORARY TRENDS IN EARNINGS INEQUALITY After thirty years of unprecedented expansion, US. wages began to deteriorate in the 1980s (Mishel, Bernstein, and Schmitt 1999).During that decade. average wages grew less than 1 percent per year. During the 1990s,wages remained virtually stagnant (Mishel et al. 1999). Despite recent increases, average wages remain lower today than they were in 1970.The era of stagnant U.S. wages has coincided with an era of increasing wage inequality. While wage inequality during the 1980s was the result of an increasing distance between the wages of high- and low-income workers, wage inequality in the 1990s has been the result of increasing distance between high- and middle-income workers (Freeman 1997;Mishel et al. 1999). Trends in wage inequality are more complex when broken down by race and gender. Available Current Population Survey data indicate divergent trends in earnings for men and women, and for blacks and whites.2 Figure 1 indicates that men. both black and white, have experienced an overall decline in real earnings since 1979, with a recent upsurge in the late 1 9 9 0 White ~ ~ men, however, retain their substantial absolute earnings advantage over black men and over both black and white women. Although white men have maintained their earnings advantage, the decline in black male earnings since 1979, combined with contemporaneous increases in white female earnings, has resulted in a reordering of the earnings hierarchy by race and gender. In 1979, black men earned substantially more than both black and white women. As early as 1982, a shift in the direction of these trends is evident (Figure 1). As white women experienced increases in average earnings, the earnings of black men declined. As a result, by 2000, black men earned somewhat less, on average, than white women. Black female earnings remained essentially stable in 1979-1996, with increases after 1996.While not sharing the same earnings gains of white women, black women avoided the early 1990s decline in earnings of their male counterparts. In their disaggregated analysis of these trends, Bernhardt, Morris, and Handcock ( 1 995) show that the shrinking gap between men’s and women’s wages has less to do with absolute gains on the part of women and more to do with polarization in white men’s wages during these years. The decline in real wages for men is even more acute after taking into account changes in the economic returns to education. All male workers, with the exception of

410


750

FIGURE 1. MEDIAN WEEKLY EARNINGS, 1979-2000 (2000 DOLLARS)

those holding college degrees, experienced a loss of real wages between 1973 and 1999 (Mishel et al. 2001). The decline in wagcs of the least-cducated men is lhe most pronounced of all changes in the returns to education. Between 1979 and 1999, wages for men without a high school diploma dropped by approximately 28 percent.The wages of male high school graduates and men with some college experienced similar, if less dramatic, declines over the same period, falling 15 percent and 7 percent respectively (Mishel el al. 2001). While some argue that economic changes favor women (O’Neill and Polachek 1993; Wellington 1993; Furchtgott-Roth and Stolba 1996; Blau and Kahn 1997; Yamagata, Yeh, Stewman, and Dodge 1997), evidence suggests that the restructured economy continues to create an unequal and gendered division of labor. While white women have experienced income gains over recent years, they still earn 75 percent of what white men earn on a weekly basis (Figure 1). Further, black women face a further earnings deficit, making 65 percent of white male earnings and 87 percent of black male earnings. While black workers have historically been overrepresented in secondary labor market jobs (Edwards, Reich, and Gordon 1975; Lang and Dickens 1988), black women have been especially overrepresented. This trend, paired with current efforts to roll back affirmative action hiring, suggests that black female workers will continue to face the greatest disadvantage in the new economy. What do these trends reveal about work rewards in the new economy? It appears that, despite very recent gains, male workers earn less today than they did twenty years ago. Even taking into account the bidding up of wages for more-educated workers, only the wages of the most-educated workers have increased over this time period (Mishel et al. 2001). This trend is consistent with our argument.That is, as male employment declines in primary labor market jobs in the core of the economy, men, especially less-educated men, may earn less. Women, by contrast. have made some inroads in an economy that,

lntegrating Economic Dualism and Labor Market Segmentation

411

regardless of industry, relies less on brawn and more on human-capital and soft skills (Moss and Tilly 2001).

LINKING STRUCTURALTHEORIES AND CONTEMPORARY TRENDS IN EARNINGS INEQUALITY Segmented labor market and dual economy theories originated out of attempts to explain persistent structural inequality, despite increases in human capital. It would seem, then, that structural theories of wage determination, including segmented labor marketidual economy theories, would be similarly useful for understanding some of the contradictions that characterize the new economy (Rubin 1995; 1996;Vallas 1999; Smith 2001). Most germane here are the arguments that identify workplace restructuring as a source both of deskilling and enskilling of workers across the economy (Zuboff 1994; Cappelli 1997). While an overview of this debate is beyond the scope of this study, several points are particularly relevant. First, the contradictory skill trends in which strategies of dynamic flexibility upgrade skill and valorize human capital, and strategies of static flexibility downgrade skill and rely on various cheap, nonstandard work contracts, are relevant both for services and manufacturing: Second, while many analysts focus their attention on the increased importance of service industries to claim that human capital is an increasingly important predictor of stratification outcomes, as Appelbaum and her colleagues (2001) indicate, manufacturing and goods production continue to comprise a major component of the new economy’s health. Thus, in examining economic transformation and changes in the economic reward structure, we argue that it is vitally important to consider thc overall economy not just the service sector. Here too, we think a reconsideration of segmentation theories is valuable. Beyond the contradictory effects of human capital across indurtries and occupations, stratification researchers have demonstrated that human capital arguments alone are deficient as explanations of wage inequalities between white and black and male and female workers (Jacobs and Steinberg 1990; Tomaskovic-Devey 1993; McCall 2001). Evidence indicates that both race and gender explain work outcomes beyond the effects of human capital at all levels of the occupational hierarchy; hence race and gender may further segment the labor force within industries and occupations (Reskin and Padavic 1994; Mishel el al. 1999). Petersen and Morgan (lY95) have shown that within selected occupations, the wage-gap betwecn men and women disappears entirely once the combination of occupation and establishment is controlled for; however, their analysis focused on blue-collar clerical and white-collar administrative and professional occupations. Likewise, Tam (1997) argues that observed gender inequalities result from differences in the distribution of specialized human capital by sex (though see the fairly convincing rebuttal by England et al. 2000). Examining the shifting mix of occupational positions and economic sectors into which workers are slotted in today’s new economy can contribute to our understanding of these persistent inequalities. The aim of our research is to examine the outcomes of the structural location of workers in terms of wages and the temporal shifts in both structural location and subsequent outcomes. In sum, we seek to establish that, independent of human capital, race, and gender. labor market and economic segmentation continue to determine the ways in which workers are rewarded for their labor.

412


ANALYSIS

Data and Variables This study uses data from the National Opinion Research Center’s General Social Survey (Davis and Smith 2000) for the years 1974,1980,1988,1996, and 2000.5We strategically chose to analyze NORC’s General Social Survey in lieu of other equally effective, but more often used data sets, such as the U.S. Census Bureau’s Current Population Survey. Between the 1970 census and the 1980 census, the Census Bureau made significant changes in both its industry and occupation classifications6 The differences between these two categorizations make longitudinal comparisons cumbersome. In attempting to address this problem, the Census Bureau recommended comparisons of 1970 and 1980 classifications using the same sample as the most effective method of comparing 1970 and 1980 census occupation and industry classifications. In the 1988 administration of the General Social Survey, NORC asked respondents to indicate their job title, the tasks they perform at work, and the type of company for which they work. Survey administrators then coded industry and occupation for each respondent using, first, 1970 criteria and, second, 3980 criteria. Using these data allows us to confirm the validity of our industry and occupation categorizations by comparing respondents’ classifications based on 1970 and 1980 Census Bureau definitions of occupations and industries. The respondents selected for analysis include all those who reported working fulltime or part-time, not working due to illness, strike, or vacation, being laid-off, or unemployed and looking for work. We eliminated those respondents who reported being retired, in school, keeping house, or not participating in the paid labor force for some other reason. In addition, we eliminated all respondents who did not answer questions addressing either occupation or industry of employment. These limitations yielded sample sizes of 638 respondents in 1974,719 respondents in 1980,804 respondents in 1988, 1,690 respondents in 1996,and 1,558 respondents in 2000.’

Dependent Variable Earnings were measured by asking respondents their annual wages from the occupation they specified as their primary source of income.x Since earnings data are reported categorically in the GSS, we use the midpoint of the 1974 response categories to estimate each respondent’s earnings in 1974,1980,1988,1996, and 2000.9In order to compare the effects of economic dualism and labor market segmentation on earnings over time we use a standardized measure of earnings. We computed real earnings by multiplying 1996, 1988,1980,and 1974 earnings each by the degree of inflation between 1974 and 2000 (Xgges 1988).We recoded these adjusted earnings into the categories used in the 1974 GSS.

Independent Variables The primary relationship of interest in this research is the intersection of economic segments and labor markets. We constructed a four-category model measuring the interaction between one’s occupation and the industry in which one works.1° Attempting to categorize occupations into labor market segments and industries into economic segments entails significant methodological problems. To ground our categories sufficiently in theory, we would havc liked to have developed a data set at both the industry and


413

occupation levels of analysis. Since it is not the goal here to develop a new method of categorizing occupations or industries, we used two previously developed, data-driven categorizations to create our industry-occupation classifications.We recognize the problems entailed in using a static classification scheme to study historical change." While far from ideal, such a methodology nonetheless allows for exploring changes in economic structures, such as labor markets and sectors of the economy (Xggcs 1988). We use Tolbert, Horan, and Beck's (1980) classification of economic sectors. Tolbert and his colleagues (1980) categorized industries as being in the core or periphery of the economy on the basis of empirical measures of market concentration, monopolistic characteristics, and labor force size. While a definitive categorization method for labor market segments has not developed out of the labor market segmentation research, Boston's (1 990) categorbation method strikes us as the most empirically rigorous Boston uses degree of necessary skills or required training to categorize occupations as part of either the primary or secondary labor market.I2 We next created a variable that coded the industry-occupation association as coreprimary, core-secondary, periphery-primary, or periphery-secondary. We tested the validity of our categorizations across the census occupational coding changes by using the 1988 responses to industry and occupation based on 1970 census definitions and 1980 census definitions. We cross-tabulated the categorization of individuals into cells based upon the 1970 census definition and the 1980 census definition of both their occupation and their employing industry. We found that the coding of a respondent's job into an industry-occupation combination using 1970 criteria was not significantly different from the coding of that same respondent's job using 1980 criteria. Our intent in cross-classifying occupation and industry of employment is to explore changes in the wage effects of employment in the primary labor market versus the secondary labor market and employment in the core of the economy versus the periphery of the economy. In our analyses, we examine the relative effects on earnings of employment in one of four cells: the core-primary. core-secondary, periphery-primary, and periphery-secondary. Core-primary employment, for example, would be employment in a primary labor market occupation in an industry in the core of the economy. We anticipate that employment in similar occupations may have different wage effects in different economic sectors. Likewise, we anticipate that employment in similar industries may have different wage effects in different labor markets. For example, drawing from the sample cases listed in Table 1 , although a financial manager in the securities, commodity brokerage, and investment industry and a manager in an automotive repair shop both hold primary labor market jobs, we would anticipate their wages to differ greatly not just due to differences in skills required lor both occupations but also due to the structural position of each industry in the US. economy. Structure alone does not explain earnings. Research demonstrates that racial and ethnic minorities are disadvantaged in terms of rewards for work.The labor market disadvantage of being female is another consistent finding in previous research. We include, therefore, the combined race and sex classification of each respondent (Table 2 provides descriptive statistics of the variables in our analyses). Due to the very small samples of non-black ethnic groups in early administrations of the GSS, we were constrained to a dichotomous variable that measures race and ethnicity as either white or n ~ n w h i t e . 'The ~ resulting race-ethnicity-gender cross-classifications are white male, white female, nonwhite male, and nonwhite female.


414

TABLE 1.

INDUSTRY-OCCUPATION CLASSIFICATIONS

Dual Economy Core Segmented Primary Labor Market

Secondary

Periphery

Core-Primary Sector Sample cases Chemical engineer in the plastics. synthetics, and resins industry Financial manager in the securities, commodity brokerage, and investment industry

Periphery-Primary Sector Sample cases Manager in an automotive repair shop Accountant in the apparel and accessory retail trade industry

Core-Secondary Sector Sample cases Ticket agent in the air transportation industry Machine operator in the glasdglass products industry

Periphery-Secondary Sector Sample cases Sales worker in a mobile home dealership Dispatcher in the taxicab service industry

Recent literature indicates that one's work outcomes arc influenced by region of employment (Colclough and Tolbert 1992). Shifts in capitalist investment lead to regionally disparate work outcomes, as industries abandon regions with entrenched workingclass organizations in favor of regions with unorganized, lower paid workers (Storper and Walker 1989; Grant and Wallace 1994). We therefore include a measure assessing geographic residence of the respondent defined as Northeast, South, Midwest, or West.14We use the West as a reference category to compare the effects of regional differences in wages between 1974 and 2000. In order to examine how economic and labor market structures affect work outcomes independent of human capital, we include two measures of human capital as controls. First, we include each respondent's level of education defined as the number of years of education completed by the respondent. Second, we include the work experience of each respondent, measured as the respondent's age less years of education less six (Smith 1997).15We recognize, however, that this second measure is a better indicator of experience for men than for women, who spend less time in the labor force on average. Finally, to offset the effects of part-time work on wages, we also include the number of hours worked per week as a further control.

Hypotheses

In exploring the determinants of earnings, we expect that three categories of variables will have significant effects: human capital, race and sex, and structural sector of employment. Research indicates unequivocally that human capital determines a significant share of earnings. We therefore expect education and experience to increase earnings. Contrary

415


TABLE 2.

DESCRIPTIVE STATISTICS

Mean (Std. Dev.) or Percentage

Earnings (2000 dollars) Education Experience Hours worked per week White male White female Nonwhite male Nonwhite female Northeast South Midwest West

1974

1980

1988

1996

2000

34,247.44 (24.475.88) 12.44 (3.04) 24.42 (14.71) 39.84 (12.38) 54.2 33.7 6.2 5.8 23.8 31.0 29.3 15.9

33,859.77 (25,111.89) 12.75 (3.00) 20.29 (14.56) 41.01 (14.44) 52.5 38.7 3.6 5.3 19.5 35.0 27.8 17.7

30,291.31 (22,946.91) 13.22 (2.85) 20.00 (13.59) 41.33 (13.85) 42.7 40.4 7.8 9.1 18.0 35.0 27.3 19.7

31,794.37 (22,445.53) 13.88 (2.67) 20.26 ( I 2.50) 42.35 (14.17) 42.2 39.0 8.2 10.5 19.5 34.0 23.7 22.8

33,337.65 (22,522.83) 13.76 (2.68) 20.93 (12.65) 41.92 (13.45) 40.4 38.6 9.0 12.0 20.6 3s. I 23.8 20.5

to the optimists of the equalizing effects of the new economy, however, we do not expect the influence of education and experience to have increased since 1974 and thus expect the effects of race, sex, and sector to persist net of human capital. Again unequivocally, women and racial minorities have experienced greater benefits of employment in recent years than in the early 1970s. Despite these improvement%we expect these ascribed statuses to continue to affect wages regardless o f the new economy’s valorization of skill. White women and racialiethnic minorities, both male and female, may have experienced grcalcr rewards to their employment, but those rewards remain substantially lower than the rewards accorded to white men. In part, we expect the employment disadvantages experienced by women and raciaUethnic minorities to be related to their disproportionate employment in the less lucrative sectors of thc occupational and industrial hierarchy. We hypothesize that “good” jobs, located in the primary labor market and the core of the economy, and “bad” jobs, located in the secondary labor market and the periphery of the economy, continue to be distributed unequally among nonwhite and white and male and female workers. Independent of the effects of human capital and raceigender on earning%we hypothesize that differential employer and labor positions will continue to segment the economy and that workers’ position in the labor market will continue to predict their rewards. We anticipate that the greatest economic rewards to employment will still be garnered by workers in the primary labor market employed in core industries. We anticipate, however, that levels of employment will have shifted across sectors, moving away from core-primary and core-secondary sectors and toward periphery-primary and

416


periphery-secondary sectors. Despite these shifts, we do not anticipate that the effects of employment in each of these sectors on earnings will have lessened since 1974.Earnings will continue to be greatest in the core-primary sector and least in the periphery-secondary sector, with periphery-primary and core-secondary earnings between, in that order. Methods

We begin our analyses by examining the shifts in the distribution of jobs across occupational and industrial sectors between 1974 and 2000. We examine both changes in the relative shares of employment and average earnings by sector. To explore how these changes have benefited, or failed to benefit, different categories of workers, we examine changes in the distribution of employment and earnings by raceiethnicity and gender. Then, to explore the relative determinants of earnings (including human capital, race and gender, and structural sector of employment), we use OLS regression to estimate five separate models, one model for each year.16 To compare coefficients in the various models, we conduct z-tests for the equality of coefficients across time periods.17 Beyond the effects of race and gender controlling for structural sector of employment, we also want to know the effects of race and gender within sectors. Hence, we conclude our analyses by repeating our regression models separately for each structural sector: coreprimary, core-secondary, periphery-primary, and periphery-secondary. RESULTS Distributional Shifts in Industry-Occupation Cells

We expected that continued deindustrialization paired with the outsourcing of nonessential tasks by core industries would lead to a relative decline in employment in the core-primary and core-secondary sectors. The data are consistent with this assertion. The data in Table 3 indicate that employment in the core-primary and core-secondary sectors declined by 1 percent and 30 percent, respectively. Core-primary jobs, representing skilled labor in core industries, are those jobs that are most likely to bring the greatest rewards to workers. Core-secondary jobs would likely be jobs that would bring the greatest rewards to less-skilled and less-educated workers. While the occupations represented by these jobs require less skill, the industries in which these jobs are located are more likely to be unionized and more likely to have a broad benefits-base for employees. Therefore, losses in these sectors would not be advantageous for workcrs As the percentage of workers in the core-primary and core-secondary sectors decreased, the percentage of workers employed in the periphery-primary and peripherysecondary sectors increased. Between 1974 and 2000, the percentage of workers in periphery-primary jobs increased by 19 percent, while the percentage of workers in periphery-secondary jobs incrcased by 3 percent. The core-primary and peripherysecondary sectors, which have experienced the least change in relative employment, have experienced the greatest change in earnings. Earnings in the core-primary sector declined by 7 percent while earnings in thc periphery-secondary cell declined by 12 percent. In the periphery-primary cell, earnings remained essentially stable, while in the core-secondary cell, earnings increased by 10 percent. Only the core-secondary cell saw


417

real increases in earnings, but these increases benefited a much smaller pool of workers than they would have in 1974 because of the 30 percent drop in core-secondary employment. Some changes in employment across sectors affected all raceigender groups. All four groups lost shares of relative employment in the core-secondary cell. This finding is not surprising. Outsourcing of nonessential tasks instituted by core industries since 1974 lessened low-skill workers’ employment in core industries. Instead, these same workers would likely work in peripheral industries under contract to core industry firms. Lessskilled workers are still necessary in the new economy, but their employment is less likely to be in core industries and, hence, they are less likely to garner the wages and benefits more typical of the core than the periphery of the economy. Although coresecondary employment declined between 1974 and 2000, average wages for remaining white core-secondary workers increased, while average wages for nonwhite coresecondary workers decreased. This pattern suggests that the jobs lost in the core-secondary sector were the less-skilled and hence the less lucrative ones. Relative employment in the periphery-primary sector increased for all race and gender groups, although this increase was quite slight for white women. Like the shared decline in relative employment in the core-secondary sector, the shared growth in relative employment in the periphery-primary sector was not accompanied by a shared growth in earnings. White men and nonwhite women employed in the peripheryprimary sector lost income, 11 percent and 3 percent respectively; both white women and nonwhite men increased their income by over 30 percent. Looking at changes in employment and earnings separately for race and gender groups, white men appear to have experienced the greatest negative changes between 1974 and 2000. They experienced sub5tantial declines in core sector employment, with an 11percent drop in primary labor market employment and a 20 percent drop in secondary labor market employment. They experienced parallel increases in employment in the periphery of the economy, with over 20 percent gains in both primary and secondary labor market employment. These shifts in relative employment are not necessarily negative, but white male relative gains in periphery-primary and periphery-secondary employment have been paired with losses in average earnings in these sectors. White men experienced an 11 percent decline in periphery-primary sector earnings and a 28 percent decline in periphery-secondary sector earnings. By contrast, white men experienced only a 4 percent decline in core-primary sector earnings and actually gained earnings in the core-secondary sector. The relative losses of white men eroded, but did not eliminate, their absolute advantages over nonwhite men and all women. Even as of 2000, white men employed in the least rewarding sector, the periphery-secondary, earned on average over $7,000 more than any other raceisex group. And in the most rewarding sector, the core-primary, white men earned $14,000 more on average than other groups as of 2000. Despite their relative disadvantage in comparison to white men, white women experienced positive gains in both employment and earnings between 1974 and 2000. White women saw substantial gains in their relative share of employment in the most lucrative core-primary sector and notable losses in their relative employment in the less lucrative core-secondary and periphery-secondary sectors. Their share of employment in the periphery-primary sector remained stable. Across all sectors, white women gained earnings between 1974 and 2000. These gains ranged from 21 percent in the core-primary sector to 38 percent in the periphery-secondary sector. Although changes in hours

White female 1974 1980 1988 1996 2000 Percent change 1974-2000

Percent change 1974-2000

1996 2000

White male 1974 1980 1988

Total 1974 1980 1988 1996 2000 Pcrcent change 1974-2000

Year

34.3

21 .o

37.2

20.4 16.6 16.7 14.5 12.1 10.9

4.0

27,289 (20,364) 29,918 (19,901) 32,280 (20,804) 35,264 (21,386) 33,029 (19,963)

-

19.6 19.8 18.6 16.6 15.6

19.9 25.8 22.4 24.0 27.3

10.7

50,165 (24,108) 52,040 (26,853) 46,195 (24,405) 48,092 (24,679) 48,151 (24,197)

30.0

-7.4

-0.9

43.8 44.2 37.0 36.6 39.1

20.3 17.9 17.1 15.4 14.2

44,925 (25,084) 44,480 (27,273) 40,952 (23,521) 42,060 (23,585) 41,611 (23,421)

Mean Earnings Percent $ (SD) Employment

28.9

19,472 (9,671) 19,856 (12,584) 27.539 (18,543) 23,431 (15,790) 25,100 (14,042)

15.8

32,386 (16.876) 40,585 (20,813) 36,840 (20.067) 31,186 (16,Y39) 37,507 (16.214)

10.3

28,667 (1 6,Y11) 31,671 (20,198) 31.749 (19,305) 27,736 (16,470) 31,621 (16,328)

Mean Earnings $ (SD)

Core-Secondary

0.0

36.9 31.4 36.3 40.1 37.0

24.9

22.9 23.6 24.6 30.1 28.6

19.3

27.5 27.3 29.2 33.0 32.8

Percent Employment

3 1.9

23,808 (18,594) 24,962 (16,567) 25,260 (18,591) 27,535 (19,691) 31.397 (20,693)

- 10.7

47,957 (27,294) 39,910 (26,364) 35,115 (24,712) 41 ,099 (24,865) 42,808 (24,823)

0.4

35,187 (26,145) 32,285 (23,149) 29,710 (22,571) 33,335 (23,141) 35,326 (23,155)


Periphery-Primary

COMPOSITION OF INDUSTRY-OCCUPATION SECTORS

31.8 34.0 28.7 29.3 32.1

Percent Employment

Core-Primary

TABLE 3.

-6.8

26.6 26.1 26.7 23.9 24.8

21.1

13.7 12.4 19.7 16.7 16.6

3.0

20.3 20.8 24.9 22.3 20.9

Percent Employment

37.6

11,927 (8.967) 14,486 (9.157) 12,663(13,991) 14,410 (9,274) 16,413 (14,355)

-28.3

33,346 (23,250) 29,921 (27,490) 26,629 (23,156) 21,484 (16,387) 23,901 (17.964)

-11.8

21,148 (19,133) 19,694 (10.109) 18,087 (18,827) 17,686 (13,613) 18,661 (15,389)


Periphery-Secondary

8 W

x

?

Z

P P

.

2

2