OCCUPATIONS AND THE STRUCTURE OF WAGE INEQUALITY IN THE UNITED STATES, 1980s-2000s Word Count: 11,942

Occupations have long been regarded as central to the stratification systems of industrial countries, but have played little role in empirical attempts to explain the well-documented increases in wage inequality that occurred in the United States in the 1980s and 1990s. We address this deficiency by assessing occupation-level effects on wage inequality using data from the Current Population Survey for 1983-2008. We model the mean and variance of wages for each occupation, controlling for education and demographic factors at the individual level, in order to test three competing explanations for the increase in wage inequality: the growth of between-occupation polarization, changes in education and labor force composition, and “residual” inequality unaccounted for by occupations and demographic characteristics. After correcting for a problem with imputed data that biased recent results by Kim and Sakamoto (2008), we find that between-occupation changes explain 66% of the increase in wage inequality from 1992-2008, although 23% of this is due to the switch to the 2000 occupation codes in 2003. Sensitivity analysis reveals that 18% of the increase in inequality from 1983-2002 results from changes in just three occupations: managers “not elsewhere classified,” secretaries, and computer systems analysts. Occupational Inequality 1 Occupations and the Structure of Wage Inequality in the United States, 1980s-2000s Sociologists have long regarded occupations as the key indicators of workers ́ positions in the societal division of labor. Their widespread use of measures of occupational prestige, status, or average earnings to represent the structure of inequality assumes that occupations are the basic components of stratification systems in industrial nations such as the United States (e.g., Parkin 1971). Many occupations constitute distinct labor markets within which supply and demand for particular types of labor intersect, and so occupational differences help explain how inequality is generated (Stolzenberg 1975, Spilerman 1977; Bielby and Kalleberg 1981). The centrality of occupations to sociologists’ theoretical conceptions of the stratification system suggests that changes in the occupational structure—i.e. in the relative size and wages of different occupations—would play a prominent role in sociological attempts to explain the well-documented increase in wage inequality in the United States over the past 25 years. Indeed, popular accounts of the increase in inequality often assume that there has been a growing polarization of occupations, illustrated by changes such as the loss of high-wage blue-collar occupations through de-industrialization, the increase in poorly paid service sector jobs, and the rapid growth in wages of managerial and professional occupations. Surprisingly, though, systematic sociological research on the role of occupations in generating patterns of wage inequality is conspicuously scarce (DiPrete 2007; Morris and Western 1999; Myles 2003). The relatively few sociological studies that have explicitly examined how occupations generate economic inequality appear to confirm the assumption that differences between occupations are primarily responsible for the growth of wage inequality (e.g., Massey and Hirst 1998; Wright and Dwyer 2003; Weeden et. al. 2007). Moreover, some economists (e.g., Autor, Katz, and Kearney 2006) also argue that increases in inequality since 1990 are due to the polarization between high and low wage occupations. Recently, however, this view of the centrality of the occupational structure in explaining changes in inequality has been challenged by several sociologists. Kim and Sakamoto (2008) use data from the Current Population Survey (CPS) to examine trends in betweenand within-occupation inequality from 1983-2002. They find that the withinoccupational component of inequality increased sharply over their study period, while the between-occupation Occupational Inequality 2 component remained relatively constant (see also Raffalovich 1993). Kim and Sakamoto (2008: 153) conclude “that the traditional sociological focus on occupations may be inadequate as the primary explanation of rising wage inequality” and suggest that occupations are declining in significance and becoming less influential in the New Economy. Their conclusion that the role of occupations in explaining wage inequality is declining has serious implications for sociologists’ assumptions about the central role played by occupations in the process of stratification. In this paper, we reevaluate the role of occupations in explaining the increase in wage inequality by using data from the Current Population Surveys from 1983-2008 to estimate models that simultaneously test individual and occupation-level effects on changes in the mean and variance of wages. This allows us to analyze the impact of occupational structure on trends in inequality controlling for changes in the returns to education and the demographic composition of the labor market. In addition, we present results at both the aggregateand detailed-level of occupations. In so doing, we are able to estimate the contribution of specific occupations to the increase in inequality over this time period, adjusting for changes in individual-level variables. Finally, we correct for the Census Bureau’s treatment of missing wage data, which imputes missing wages based on non-missing data from aggregate occupational categories rather than detailed occupations, thereby mechanically increasing within-occupation inequality. We show that much of the trend in within-occupation inequality reported by Kim and Sakamoto (2008) is an artifact of this incorrect approach to handling missing data. EXPLAINING THE INCREASE IN WAGE INEQUALITY IN THE U.S. Debates on trends in wage inequality typically identify three main sources. First, sociologists (but usually not economists) generally assume that wage differences between occupations are important for generating wage inequality. Second, economists—and many sociologists—argue that the growth of wage inequality reflects primarily changes in the composition of the labor force with regard to human capital characteristics such as education, experience, and skill otherwise measured. Finally, a growing literature in economics (as well as by sociologists such as Kim and Sakamoto 2008) has maintained that there is “residual inequality” in wages after occupations and labor force characteristics are taken into account. Occupational Inequality 3 1. Between-Occupation Inequality Sociologists often argue that “occupational structure”—i.e., the relative size and average wages of different occupations—is central to explanations of inequality (e.g., Blau and Duncan 1967; Treiman 1977; Featherman and Hauser 1978; Bielby and Kalleberg 1981). Recent descriptions such as “dissagregate structuration” (e.g., Grusky and Sorenson 1998, see also Weeden et al. 2007) contend that members of detailed occupational categories are fairly homogenous in terms of their life chances, and so economic inequalities should lie primarily between occupations, not within them. The validity of the assumption that occupations are integral to explanations of economic inequality depends on the degree to which wages differ between, relative to within, occupations. Theoretically, between-occupation wage differences—and changes in average occupational wages—result from several sources. First, occupations vary in their skill, i.e., the degree of complexity of occupational activities and the amount of training time required to perform them adequately. The skill level of an occupation may be reflected in its average education level as well as by occupation-specific skills not captured by broad educational categories. Between-occupation wage differences may also be represented by differences in the returns to education and other labor force characteristics such as gender, which also vary by occupation. In our empirical analysis, we will test for occupation effects above and beyond the returns to labor force composition by estimating models that include individual-level controls for education and other wage-related individual characteristics. Second, occupations also differ in the extent to which they are able to adopt and maintain institutional mechanisms of social closure such as licensing, educational credentialing, voluntary certification, association representation and unionization (Weeden 2002). These institutional mechanisms of social closure increase earnings by restricting the supply of workers, increasing demand and channeling it to the occupation, and signaling the quality of service. Weeden’s analysis found that occupational earnings were affected by occupational differences in both technical complexity and measures of social closure. Recently, several papers have found evidence of between-occupation effects on wage inequality, suggesting that the occupational structure has become increasingly polarized: both highand low-paying occupations have grown in size, especially in the 1990s. Massey and Hirst (1998) found empirical support for an “hourglass” pattern of growth in both highand low-paying occupations in the period 1969-89 for men (but not women); this represents a Occupational Inequality 4 change from the “escalator” pattern of the 1949-1969 period in which people worked in progressively more highly paid occupations. Wright and Dwyer (2003) examined the growth and decline of jobs of varying quality during the economic expansions in the U.S. since the 1960s and concluded that there has been a growth of jobs in the top and bottom quintiles in the 1990s, and a decline in the middle quintile. Autor, Katz and Kearney (2006) showed that the share of total hours worked by members of both high-wage and low-wage occupations increased during the 1990s, while the employment shares of middle-wage occupations declined (see also Goos and Manning [2007] for a similar analysis for the United Kingdom). Finally, Weeden et.al. (2007) found that much of the growth in inequality over this time period is due to differences across “big classes” (aggregate groups of occupations), especially for men. At the same time, they recognize that the absence of controls for education in their analysis may confound their interpretation of between-occupation changes in inequality with changes in the returns to education, an issue to which we now turn. 2. Human Capital and Labor Force Composition A different interpretation of the growth in wage inequality is that it results from changes in the returns to and/or the levels of education and skill in the labor force, not from changes in occupational structure. It is possible that the well-documented increase in the returns to education, as exemplified by the widening college/high-school wage gap (e.g., Katz and Murphy 1999; Autor 2004), could explain much of the apparent polarization of the labor force into good and bad jobs. There is considerable support for the notion that the rise in inequality is due to a general increase in the returns to education and skill. The most prominent explanation of rising inequality in the economics literature during the 1990s was the theory of “skill-biased technical change” (SBTC), which argues that recent changes in inequality are driven by fundamental technological advances such as the growing use of computers that are “biased” in favor of 1 Wright and Dwyer (2003) note that trends in income inequality are not necessarily related to job growth across job quality quintiles: e.g., increasing income inequality could be associated with flat growth since (1) the income spread between the best and worst jobs could be increasing even if they are growing at the same rate, and (2) increasing earnings inequality can occur within job types (p. 303). Occupational Inequality 5 high-skilled workers (Katz and Murphy 1992; Levy and Murname 1992; Acemoglu 2002; see Lemieux 2008 for a recent review). An increasing demand for skilled workers driven by technological changes would help to explain the deterioration of wages for less-skilled workers during the 1980s and the growth in demand for workers with high levels of education (e.g., Levy and Murnane 1992). An alternative human-capital based explanation for the increase in inequality is that it is not due to technological changes per se but changes in the composition of the labor force. It is well known—and well grounded in human capital theory (see Mincer 1974)—that more educated and experienced workers tend to have higher variances in wages. As a result, an increase in the average educational level and experience (i.e. age) of the labor force will tend to increase wage inequality (Lemieux 2006). With respect to trends in occupation-level inequality, this composition-based perspective implies the same empirical prediction as SBTC: observed measures of human capital, such as education and experience, should attenuate or explain away trends in between occupational inequality. Given the possibility that the returns to skill or changes in the composition of the labor force could account for the impact of changes in occupational structure on inequality, we will incorporate the human capital model directly into our empirical estimates of occupational inequality. In the methods section below, we present individuallevel models of the mean and variance of different occupations, controlling for education and potential labor market experience as well as other key individual characteristics that might explain the growth of inequality, such as gender, race, and union membership. 3. Residual Inequality A third explanation argues that the rise in wage inequality is due neither to changes in occupational structure nor to observed human capital variables but to unexplained, “residual” inequality. In the recent literature in economics on human capital and rising inequality considerable attention has focused on whether the increase can be explained by observed characteristics of the labor force as opposed to residual inequality (Juhn, Murphy, and Pierce 1993; Lemieux 2005, Acemoglu 2002; Autor, Katz, and Kearney 2006; Card and DiNardo 2002; Katz and Autor 1999). In addition, the concept of residual inequality in human capital models has a direct parallel with the measure of within-occupation inequality used by Kim and Sakamoto (2008) in the debate on occupational inequality, i.e., the component of inequality not explained by occupational structure, or at least by a particular set of occupational codes. Occupational Inequality 6 The examination of trends in residual inequality has attracted attention because it is useful in helping to adjudicate between SBTC and alternative theories (Juhn, Murphy, and Pierce 1993; Acemoglu 2002; Card and DiNardo 2002; Lemieux 2006). SBTC predicts that inequality will increase both withinand between-educational levels. Within-education inequality increases because of the greater demand for skills, some of which are unobserved to the researcher and vary among workers with the same level of education. Workers who are best able to adapt to the new technologies earn a wage premium, leaving less adaptable workers with the same level of education behind (Lemieux 2006; Juhn, Murphy and Pierce 1993; Acemoglu 2002). Overall, the rise of residual inequality in the 1970s and 1980s is consistent with the prediction of SBTC: Katz and Murphy (1999) find that residual inequality explained around 60% of the increase in inequality between 1963 and 1995, and Juhn, Murphy, and Pierce (1993) and Acemoglu (2002) find significant increases in residual inequality since the late 1970s after controlling for individuallevel predictors of wages such as education, experience, and demographic categories. In a prominent sociological account of increasing inequality, Bernhardt et al. (2001) note that more than half of the increase in inequality for men has been within groups of workers with the same age, education, and experience. However, the SBTC explanation has recently been challenged by a revisionist argument that points out two shortcomings based on trends in residual inequality during the 1990s. First, the evidence suggests that residual inequality grew sharply during the 1980s but then slowed, or ceased, during the 1990s (Card and DiNardo 2002). Second, Lemieux (2006) shows that after adjusting for changes in the proportion of workers in different education and experience categories, the trend in residual inequality after 1990 disappears, suggesting that the recent increase in inequality is the result of changes in the demographic composition of the labor force rather than an underlying process of skill-biased technological change. The debate between the SBTC and revisionist interpretations of inequality trends is relevant for our discussion of occupational inequality for two reasons. First, as we mentioned above, there is a conceptual parallel between residual inequality in human capital models and within-occupation inequality. Consequently, the results found by Card and DiNardo (2002) and Lemieux (2006) indicating little or no increase in residual inequality during the 1990s raise questions about the trend in residual within-occupational inequality over the same time period reported by Kim and Sakamoto (2008). Occupational Inequality 7 Second, a careful consideration of the role of occupational structure is missing from this literature in economics (exceptions are Autor, Katz and Kearney 2006 and Goos and Manning 2007). Given the focus on observed and residual components of inequality, this is a nontrivial omission. For example, one of the reasons why the variance of wages increases with education (e.g., Lemieux 2006) is that specialized training encourages workers to move into heterogeneous occupations. As a result, changes in occupational structure may explain an increase in within-education “residual” inequality or average wages. Weeden et.al. (2007) suggest that a growing demand for high-skill occupations may explain some of the increase in residual inequality in human capital models. On the other hand, the reverse might also be true: an increase in the returns to higher education might explain trends in betweenoccupation inequality, but in this case it is not the occupational structure per se that is generating inequality but a change in the fundamental relationship between education and earnings. Summary: Three Competing Explanations Based on the preceding discussion, we propose to test the following three competing explanations of the recent rise in wage inequality: 1) Changes in occupational structure: This is the “between” occupation argument that predicts that changes in the average wage and size of occupations will explain the increase in inequality. 2) Education and labor force changes: Changes in the returns to human capital and the demographic composition of the labor force will explain the increase in inequality. 3) Residual inequality, net of controls for occupation and observed human capital variables. This combines the explanations based on trends in residual inequality in human capital models with Kim and Sakamoto’s (2008) argument about trends in within-occupation inequality. Our models for the mean and variance of occupational wages control for education and basic demographics at the individual level, allowing us to test these three competing hypotheses simultaneously. Occupational Inequality