Skills tasks, and class- an integrated class based approach to understanding recent trends in economic inequality in the USA

Abstract

This paper examines the large body of literature regarding the evolution of income inequality in the USA over the last several decades, and interprets it in an integrated class framework. Namely, the canonical supply–demand model for skilled labor, the skill biased technological change/routine biased technological change literature, and Piketty’s “super-manager” hypothesis are examined and integrated into Erik Olin Wright’s class framework. In doing so, this paper clarifies the causal links between income and skills, tasks, class, and income, as well as the flows of people from skills to jobs/occupations and class by mapping these connections on a class framework. By breaking down the complementary factors involved in growing economic inequality described in the literature, this paper (1) lays the foundation for separately analyzing the evolution of the causal links and flows described in the literature, (2) identifies elements in the literature that are sometimes conflated, and (3) endogenizes the institutional factors that shape the causal mechanism. These connections are supported with calculations of wage trends from 1980 to 2010 using May/ORG data-sets, and the income variance for the period is decomposed by skill, tasks, and class.

Appendix: Decomposition methodology

Following (Uni 2008), log variance of income is calculated with the following equation:

$${\text{LV}} = \frac{1}{n}\sum\limits_{i = 1}^{n} {(ly_{i} - \mu )^{2} } ,$$

where ${\text{LV}}$ is log variance, $n$ is the number of samples,$\mu$ is the mean of log income, and $ly_{i}$ is log-income of the ith sample. Log variance is decomposed into within group variance and between group variance with m groups using the following:

$${\text{LV}} = \sum\limits_{i = 1}^{m} {s_{i} LV_{i} } + \sum\limits_{i = 1}^{m} {s_{i} (\mu_{i} - \mu )^{2} } ,$$

where $s_{i}$ is the share of the ith group, $s_{i}$ is ($\frac{{n_{i} }}{n}$), the proportion of the sample size of the ith group to the size of all samples, and $\mu_{i}$ is the mean of log income in the ith group. The first right term is the within-group inequality and the second between-group inequality. The intertemporal change in log variance ($\Delta LV$) is decomposed into contributions within groups, between groups, and two structural affects as follows:

$$\Delta {\text{LV}} = \sum\limits_{i = 1}^{m} {s_{i} \Delta {\text{LV}}_{i} } + \sum\limits_{i = 1}^{m} {s_{i} \Delta (\mu_{i} - \mu )^{2} } + \sum\limits_{i = 1}^{m} {\Delta s_{i} {\text{LV}}_{i} } + \sum\limits_{i = 1}^{m} {\Delta s_{i} (\mu_{i} - \mu )^{2} } ,$$

where the first right term is within group contributions to variance, the second term the between group contributions to variance, and the third and fourth terms the contribution from changes in the compositions of groups.