منابع رقابت نابرابری در درآمد: مقایسه مولفه های واریانس

Using data from the National Longitudinal Survey of Youth, this study partitions competing sources of earnings variation into four components: a component for individual-level factors, a component for transitory factors, a component for occupational factors, and a component for geographic factors. From these variance components intraclass correlation coefficients are calculated and compared within and across seven ascriptive statuses: non-Hispanic white, non-Hispanic black, Latino, women, men, foreign-born and U.S. born statuses. Among those of age 25 to middle adulthood, the results indicate that the transitory component is the largest source of variation in earnings, followed by individual-level factors. Occupational factors and geographic factors combined account for nearly the amount of earnings variation as individual-level factors. The omission of these socio-structural factors inflates the size of the individual-level variance component by over 35 percent. Counter to expectations, the variance profiles are remarkably similar across ascriptive statuses. Indeed, the variance profiles for whites and blacks are nearly identical. Modest differences in intraclass correlation coefficients are found between men and women. Relative to men, less of the variation in earnings for women is attributed to individual-level factors, and slightly more of the variation in earnings for women is attributed to occupational factors and transitory factors. This research draws attention to universal theories of earnings inequality.

This research studies several major and often competing sources of earnings inequality. In line with past research, there are four societal components that are deemed essential for the creation of earnings inequality (in general) and for the creation of ascriptive earnings inequality (in particular). First, individual-level factors are taken into account to capture the effect of parental socioeconomic resources, childhood socialization, and educational quality on the acquisition of skills and attributes that are later rewarded in the labor market. Second, a transient component is taken into account to capture all the seemingly “random” events that affect peoples’ earnings on a yearly basis over the course of a working career. Thirdly and fourthly, are two socio-structural factors (the occupational structure and the geography of opportunity) that affect the distribution of earnings beyond individual and transitory factors. The occupational structure is generally understood to affect earnings inequality because society rewards some occupations more handsomely than others. Moreover, particularistic theories of ascritpive inequality often implicate occupational segregation in the devaluation of pay for disadvantaged groups. For similar reasons, geographic factors affect earnings inequality. At minimum, differences in the cost of living and geographic differences in socioeconomic opportunity contribute to earnings variation. Particularistic theories of inequality suggest that geographic factors are especially relevant for disadvantaged groups because differential access to socioeconomic opportunities often correspond closely with relatively benign factors like historical settlement patterns, and more malicious factors, like residential segregation and group threat processes. This study advances prior work on individual sources of earnings inequality by simultaneously incorporating these two socio-structural factors into a variance decomposition of earnings attainment. Data from the National Longitudinal Survey of Youth are used to partition the variation in earnings attainment into these four components – an individual-level, a residual/transitory level, an occupational-level, and a geographic-level – then intraclass correlation coefficients are calculated and compared within and across ascriptive statuses. In general, the findings indicate that the transitory component is the greatest source of earnings variation. The explanatory power of individual-level factors is secondary to transitory factors. Occupational factors and geographic factors (added together) account for nearly the amount of earnings variation as do individual-level factors. Importantly, however, this research finds that the omission of occupational factors and geographic factors inflates the size of the individual-level variance component by at least 35 percent and, in some cases, over 50 percent depending on the ascriptive status of the group. This finding demonstrates how previous research is liable to overstate the importance of individual-level characteristics in explaining earnings inequality. In addition to this general pattern, a comparative analysis across ascriptive statuses is also intriguing. Foremost, the results demonstrate that the sources of within group earnings inequality do not differ greatly between ascriptive statuses. Indeed, the variance profiles for blacks, whites (and to a lesser extent Latinos) are nearly identical to each other. This finding is quite unanticipated, and maybe even perplexing given the extensive research on racial and ethnic inequality that went into developing the main theoretical premise of the article. The theoretical premise of this research is that socio-structural factors are more likely to affect earnings for disadvantaged groups than advantaged groups because of such factors like occupational and residential segregation. Accordingly, I hypothesized that socio-structural factors would account for more of the variation in earnings for disadvantaged groups than for advantaged groups. For the most part, however, this premise is not substantiated by the data. Although puzzling, this finding is also encouraging. It suggests that the U.S. stratification system is not structured in such a fundamentally different manner for minority groups as to render their sources of earnings inequality incomparable with that of whites. In other words, the sources that generate earnings inequality among blacks and Latinos are similar to the sources that generate earnings inequality among whites. Therefore, a better understanding of the general causes of inequality in the population at large will be to society's betterment. The implication of this research is that it lends empirical support to Tilly's (1998:16) keen observation: Observers often ground explanations for each form of inequality separately in perennial but peculiar forces. Each one seems sui generis, constituting its own mode of existence. If sexism springs from age-old patriarchy, racism from the heritage of slavery, denigration of non-citizens from xenophobic state traditions, however, it is hard to see why the mechanisms of inclusion and exclusion in all these cases have such striking resemblances. They must have more common causal properties than particularistic accounts suggest. It is important to note that Tilly's insight applies to categorical inequalities, in general, and does not preclude, whatsoever, the fact of an earnings gap between whites, blacks and Latinos (which is seen in this analysis and amply reported elsewhere). This research indicates, however, that by the time individuals reach the age of 25, the role of individual-level factors, transitory factors, and socio-structural factors in the creation of earnings inequality appears to be quite similar among racial and ethnic groups. This knowledge should caution researchers from isolating unique factors that seem only to pertain to certain groups, and instead, direct them to consider broader mechanisms that cause inequality both within and between groups. This is especially true in light of recent trends toward increasing inequality within race and ethnic groups ( Leicht, 2008:241–242). As previously mentioned, the consequences of ascriptive inequality can be equally harmful regardless of whether it is driven by particularistic and/or universalistic forces. Addressing the issue, however, depends on our knowledge of which force predominates. Apart from the comparison of minority groups, the comparison between women and men is more in line with expectations. Only among women do the findings support the proposition that those faced with structural barriers – such as occupational segregation – are more likely to have their earnings determined by socio-structural factors when compared to advantaged groups that are relatively free from similar market and institutional constraints/influences. The results here demonstrate that earnings attainment for women has more to do with the occupational structure and more to with transitory factors, but less to do with individual-level skills and attributes than for men. This research is not without several potential limitations. First, there is a chance that an omitted grouping factor would account for additional individual and transitory variation and possibly alter the variance profiles between ascriptive statuses. For example, one theoretically important grouping factor that is omitted from the analysis is the family household unit. The NLSY79 does contain siblings in its sampling frame, so it is possible to include housing unit as another variance component. However, in a supplemental analysis the addition of a variance component for sibling household does not reduce any of the residual variation. In fact, the household unit does not contribute any information beyond the individual-level component. The inclusion of a random intercept for household unit simply decomposes the individual variance into its respective within and between household variances. For example, among black respondents, the household variance is estimated to be .0357 and the individual component is reduced to .0768; adding these components together (.0768 + .0357 = .112) reproduces the initial amount of individual variance for blacks illustrated in Table 1. Therefore, the omission of household units does not change the conclusions drawn from the variance profiles presented in this research. Another grouping factor that is omitted are peoples’ neighborhoods. However, experimental research has shown that neighborhoods have little impact on adult labor market outcomes (e.g., The Moving to Opportunity Demonstration; Orr et al., 2003). More reasonably, the effect of neighborhoods on earnings is likely to operate indirectly through childhood development and socialization, which in this analysis is captured by the individual-level component. Although this research has not exhausted alternative grouping factors, according to the breath of the existing literature, it is difficult to think of another grouping factor that is as essential as the ones included in this analysis. Given the novelty of this research approach there are ample avenues for future inquiry. One avenue for future research should consider alternative socioeconomic outcomes. Although earnings attainment is perhaps the most common outcome, it by no means represents all of the many different facets of social inequality. Alternative outcomes could include occupational prestige, workplace responsibility/authority, and cumulative work experience. Also, future research should utilize alternative longitudinal datasets to see if the results presented here are reproducible. Two potential alternative data sets are the Panel Study of Income Dynamics and the 1997 National Longitudinal Survey of Youth. Research using these data would provide profiles for other birth cohorts that may or may not conform to the results presented here. Such an analysis would help determine if earlier immigrant cohorts (like those in this study) experience the sources of earnings inequality differently from more recent immigrant cohorts. Analyses of a similar type are also warranted for other countries, as well. Determining how individual-level, transitory level, and socio-structural level factors affect socioeconomic attainment in different countries for different groups of people would potentially provide great insight into general processes of social stratification.