Long term trends in fair and unfair inequality in the United States

ABSTRACT

This article analyses the microeconomic sources of wage inequality in the United States from 1967–2012. Decomposing inequality into factors categorized by degree of personal responsibility, education explains over twice as much of inequality today as 45 years ago. However, neither hours worked nor education, industry, marital status, or geographical location can explain the rise in income inequality. In fact, ‘unfair’ inequality (income disparity derived from non-responsibility factors) has risen faster than ‘fair’ inequality (income disparity derived from responsibility factors), regardless of the set of variables chosen as fair sources of inequality. We further examine income inequalities within gender and racial groups, finding substantial heterogeneity. Overall, using micro data to understand the sources of inequality and how these changes over time can provide better information for policymakers motivated to combat rising inequality.

KEYWORDS:

• Income inequality
• Gini coefficient
• unfair inequality

JEL CLASSIFICATION:

• D31
• D63
• J31

Acknowledgement

We thank the Universidad EAFIT for providing excellent research support in this project. We are grateful to Susana Berrio Montoya and Alejandra Montoya Agudelo for their outstanding research assistance. All remaining errors are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ See, for example, Autor, Katz, and Kearney (2006) and Dynan and Ravina (2007) for potential utility consequences of inequality.

² A closely related concept studies the inequality of opportunity; see Pistolesi (2009), Lefranc, Pistolesi, and Trannoy (2009), Aaberge et al. (2011), and Roemer (2000, 2002). See Bourguignon, Ferreira, and Menéndez (2007) for a study of the inequality of opportunity in Brazil and Ramos and Van de Gaer (2015) or Pignataro (2012) for a general summary. Hopkins and Kornienko (2010) construct a theoretical framework to study the effects from both inequality of endowments and outcomes.

³ Note that, a priori, the literature has produced no strict preference for analysing the Gini index over alternative measures of inequality, such as the Theil index. Both indices carry particular advantages and disadvantages. Nevertheless, the Gini index remains the most prominent measure for income inequality in popular and public policy debates – an aspect potentially owed to its intuitive derivation and graphical representation from the Lorenz curve. Another attractive feature of employing the Gini index relates to the notion that comparing Ginis across groups with different sample sizes remains less of a concern than for other indices. We thank an anonymous referee for a respective comment on different inequality indices.

⁴ Specifically, they distinguish between circumstances and effort – concepts that relate closely to non-responsibility and responsibility factors in our narrative. Another distinction comes from their usage of the Michigan Panel Study of Income Dynamics, which allows them to include parental characteristics and place of birth (among other attributes), whereas our sample is derived from the CPS. Note that our sample is representative for the entire U.S. with more than 50,000 observations per year, whereas Pistolesi (2009) focus on Michigan with 1,561 and 2,246 observations in 1968 and 2001, respectively.

⁵ While tax policies and other transfers have changed substantially over the last several decades, and naturally affect after-tax income inequality, our analysis is limited to pre-tax income at the individual level. This is done due to the impossibility of determining the tax filing status among married individuals, the desire to accurately reflect marriage and children to be choice variables within our framework, and the difficulty of aggregating individual-level attributes such as age and education to the household level. Furthermore, the widely used NBER Taxsim programme, which calculates tax burdens for families, will not allow us to estimate after-tax income for a substantial part of our sample period. For a discussion regarding the choice between alternative income variables, see Armour, Burkhauser, and Larrimore (2013).

⁶ In 1994, major changes were made in the way that the March CPS was collected, and changes were also made in internal censoring points (see Burkhauser, Feng, and Jenkins 2009).

⁷ We collapse the industry categories provided by the CPS to the following broader groups: (1) Agriculture, Forestry, and Fishing; (2) Mining; (3) Construction; (4) Manufacturing – Durable Goods; (5) Manufacturing – Nondurable; (6) Transportation; (7) Telecommunications; (8) Utilities and Sanitary Services; (9) Wholesale Trade; (10) Retail Trade; (11) Finance, Insurance, and Real Estate; (12) Business and Repair Services; (13) Personal Services Goods; (14) Entertainment and Recreation Services; (15) Professional and Related Services; (16) Public Administration; (17) Other.

⁸ The number of racial categories in the CPS has increased over time. Prior to 1987, categories were limited to black, white, and other. In 1988, American Indian and Asian categories were introduced. Respondents could report more than one race beginning in 2003, drastically increasing the number of potential categories. Similarly, information on Hispanic origin was not collected prior to 1971. In a balance between maintaining comparability over time and capturing the important rise in the Hispanic demographic over time, we collapse race and ethnic categories into four: black, white (non-Hispanic), Hispanic (non-black), and other. Removing the Hispanic variable from our analysis in 1971 and years shortly after had no effect on our estimated Gini coefficients (adjusted or otherwise), suggesting the lack of data on Hispanic origin prior to 1971 is not a serious concern.

⁹ Almås, Havnes, and Mogstad (2012) provide details on the respective code in Stata.

¹⁰ Recent papers have documented a wage premium for married men with children (e.g., de Linde Leonard and Stanley 2015). One explanation argues that men may become more stable and committed workers.

Additional information

Funding

We thank the Universidad EAFIT for providing excellent research support in this project.