About the Author
Read an Excerpt
By Rebecca M. Blank
UNIVERSITY OF CALIFORNIA PRESSCopyright © 2011 The Regents of the University of California
All rights reserved.
A Broader Look at Changing Inequality
The next several chapters of this book provide a detailed comparison of the composition and distribution of income in the United States in 1979 and in 2007. I am interested in looking at the pretax income available to all nonelderly adults, which I refer to as the total-income distribution. Surprisingly, there is almost no research that takes an approach as comprehensive as that taken in this book, looking as it does at changes in the distribution of total income and its sources. There are a large number of papers that investigate changes in wage inequality over the past several decades, often focusing on either hourly or weekly wages. Very few researchers have looked at how changes in hours or weeks of work interact with wage inequality, however, or at how total-income inequality has changed in comparison to earnings inequality.
THE LITERATURE ON RECENT TRENDS IN U.S. INCOME INEQUALITY
The literature on changes in aggregate-income inequality is not extensive. All of these papers indicate that income inequality in the United States has been widening in recent decades. Most of this literature focuses either on married-couple families or on families with children, and ignores individuals who live alone or in other types of families. For instance, Juhn and Murphy (1997) look at changes in patterns of earnings within married-couple families, focusing particularly on the growing positive correlation between wives' and husbands' earnings. Pencavel (2006) updates and extends this work. Reed and Cancian (2001) simulate the effects of changes in husbands' and wives' earnings and capital income on the family income distribution. Gottschalk and Danziger (2005) look at male and female wage rates, earnings, and family income. Ahituv and Lerman (2007) are concerned with the effect of changes in family structure and work behavior on inequality. Other papers look at the role of family structure changes in explaining growing family income inequality.
One past paper that is more comprehensive is by Karoly (1996), who looks at changes in income among the entire U.S. population between 1973 and 1993. A more recent paper, by Heathcote, Perri and Violante (2010), looks at the historical relationship between labor-force behavior, wages, and total household income using a variety of different data sets. Their focus is on the macroeconomic implications of these relationships.
This book investigates how the distribution of income among all persons has been affected by changes in labor-force behavior, family structure, wages, and unearned income. In contrast to earlier papers, this book looks at all nonelderly adults, including those who live alone as well as those in families headed by married couples or single individuals. I investigate changes in the distribution of earnings and relate changes in earnings by gender to changes in the distribution of wages, taking account of simultaneous changes in the distribution of work hours. I also look at total income available to individuals and analyze the effects of changes in the distribution of earnings, governmental income, and other unearned income. In short, this quantitative analysis attempts to be more comprehensive, providing a better overall look at shifts in inequality among all of the components of earnings and income, and including a broader slice of the total population in the analysis.
THE DATA ON WHICH THIS ANALYSIS IS BASED
The remainder of this chapter describes the data on which all of the analysis in this book is based. Those who are most interested in the results and conclusions can skip to chapter 2, although I encourage even those readers to skim through the remainder of this chapter so that they may better understand the nature and the limitations of this analysis.
The primary years of comparison in this paper are 1979 and 2007. These years are comparable points in the business cycle and are both end years of economic expansion (the peak of the cycle); both were followed by a deep and extended recession. Rising wage inequality in the United States started in the mid-to late-1970s, so 1979 represents a time before people became concerned with increasing inequality. These two endpoints provide almost three decades of comparison.
The data analysis throughout this book is based on the Current Population Survey (CPS)—a large, random national sample of the population. For each primary year, I use the March data from the following year. These data include the Annual Social and Economic Supplement, which asks a variety of questions about individual and family income and work effort among all civilian adults during the previous year. Hence, my data is taken from the March 1980 and March 2008 CPS, providing information on 1979 and 2007, respectively.
I use data for all eighteen- to sixty-four-year-old adults in each year. I focus on nonelderly individuals because this group is typically considered the working-age population, and I am particularly interested in how changes in the distribution of wages over the past thirty years have affected changes in aggregate income. In part because nonelderly adults are attached to the work force, the policy options that one might consider to address inequality among this population are also very different from those that one would discuss if looking at income inequality among the elderly. Throughout the analysis, I use individuals as the key unit of observation—not families. Like the CPS, however, I assume that individuals share income with all other coresident persons who are related by marriage, adoption, or birth. Persons or families that live together but are not related to each other I assume to be separate income units.
I will use the term "family unit" to refer to the group with whom an individual shares income. I use this, rather than the more common term "family," because a family in CPS terminology includes at least two related individuals. Since I include single individuals who rely solely on their own income in my analysis, a "family unit" can refer either to a group of related individuals or to a single individual.
One of the most difficult problems in the CPS for distributional analysis is the use of "top-coding" for various income variables. Not only do I not have information on actual income among those at the top of the distribution, but the top-codes change from year to year, so the extent of missing information varies across years. Furthermore, because top-coding occurs by income source, even individuals whose total income is not at the top of the income distribution may be top-coded on one particular income variable. Burkhauser et al. (2008) show that failure to account for these top-coding issues greatly affects estimates of trends in income inequality.
Fortunately, Larrimore et al. (2008) have recently completed an extensive analysis of the impact of top-coding in the CPS. They were able to access internal Census data, and they provide information on the mean values for top-coded data by income source for the Current Population Surveys fielded in each year from 1976 through 2007. For wages and self-employment income, they provide these mean values by gender, race, and full-time/ part-time status. This allows me to adjust for top-coding in wages and self-employment income, and in the income families receive from Social Security, other government assistance programs, Unemployment and Workers' Compensation, interest, dividends, pensions, and a host of other income sources.
One of the problems in the data is the presence of negative income, typically because of negative earnings or negative net rental income. Self-employed individuals may report negative earned income when their business is doing poorly, for instance. I set all of these negative earnings or rental-income reports to zero.
A further problem is that some individuals are in family units that report zero income for the entire year. I have looked closely at these individuals and am persuaded that many of them are actually in this situation. They may be supported by someone outside their family unit or may be living on savings for the year. In 1979, this group represented 0.45 percent of the sample, but that figure increased to 1.74 percent of the sample for 2007. About two-thirds of these zero-income persons are in single-person family units in both years. In both years, about 80 percent say they are taking care of either home or family (perhaps taking care of a relative they do not live with), claim they are ill or disabled (although they report no disability income), claim they are going to school, or report themselves unable to find work. They have disproportionately low levels of education. I have repeated the analysis in chapters 2 and 3, both including and excluding the zero-income family units. Excluding them makes no difference to any of the data trends that I discuss. They are included in all of the results reported in this book.
It is worth emphasizing that all of the analysis in this book is based on cash income as reported to the government. Some types of income tend to be underreported in government surveys. For instance, there appears to be underreporting of cash-assistance programs among lower-income persons. But the most significant underreporting probably occurs among higher-income individuals, particularly those who are self-employed or engaged in illegal activities. My guess is that the inclusion of unreported or illicit income would result in an even wider distribution of income than these data indicate. But we know little about the trends in unreported income, and it is hard to say whether taking them into account would accentuate or offset the widening income inequality of the past thirty years.
Since this analysis is based on cash income, these data take no account of taxes. As I note in chapter 4, taxes are generally progressive, and the after-tax distribution of income is less unequal than the before-tax distribution of income. These data also do not include noncash benefits from employers (such as pension payments, sick days, and health-insurance subsidies) or from the government (such as assistance for housing, food, and child-care costs). In chapter 4, I return to these concerns and indicate how the inclusion of tax and noncash benefits might affect my conclusions.
ADJUSTING FOR DIFFERENCES IN FAMILY SIZE
I look at the distribution of income and income components at the individual level. This assumes that the key question is how economic well-being is distributed among persons. When discussing earnings and its components, I can look at individual earnings among workers. But when I turn to a discussion of total income and its components (earnings, government income, and other income), I must allocate the income within a family unit to each individual. One option is make a simple per person income calculation, dividing total income by family size. This is not a satisfying approach, however. Individuals live in different-size family units, and there are economies of scale within these units.
For this reason, most economists make an adjustment for economies of scale when calculating per person income. Such an adjustment assumes that the additional income needed to keep family well-being at the same level will decline as more and more persons are added to the family. Said in another way, to keep a family unit equally well off as family size increases, income has to increase more when a second person joins the family than it does when a third person enters; it has to increase more when a third person enters than it does when a fourth person enters; and so on.
I adjust for these differences in economies of scale by dividing family income by the square root of family size. This is the equation:
Yi = Yf / [square root of Nf] (1–1)
where Yi is the per person income allocated to person i; Yf is the total income available in the family unit where person i resides; and Nf is the number of persons in this family unit. If this is a single individual who lives alone, his or her income constitutes the total income in the family unit. If this is an individual who lives with others, he or she will be allocated a share of income from the family unit—a share that decreases as family-unit size rises, but that does so at a slower rate than it would if I did not account for economies of scale.
Table 1 shows how this adjustment affects per person income, showing how per person income changes as family size grows. Column 1 shows total family income, column 2 shows family size, column 3 indicates the result of a simple per person income calculation, and column 4 shows per person income adjusted for economies of scale. As column 1 indicates, I assume that family income equals $100,000. With only one person in the family unit, per person income equals $100,000 regardless of how it is calculated. When a second person enters the family unit, a simple per person income calculation would give each person $50,000. But the family size adjustment assumes there are economies of scale when two people live together and share rent, food, and other purchases. This means that adjusted per person income for this two-person household is $70,711 each. This is the equivalent amount of income each person would have to have to be as well off if they lived alone as they are living together in a family unit with $100,000 in income.
The remaining columns show how this calculation changes as family size rises. With four people in the household, each person has the equivalent of $50,000 in per person income in my calculation, which is the amount they would need to live alone at the same level of well-being as they achieve living together as a four-person family.
This adjusted per person income calculation allows me to directly compare the economic well-being of individuals no matter what their living arrangements might be. Hence, rather than comparing total family income levels, I will compare adjusted per person income levels and will look at how the distribution of adjusted per person income is changing over time. A single individual whose total income equals $50,000 is assumed to be as well off as an individual living in a family of four whose total income is $100,000. Both of these individuals have $50,000 in adjusted per person income.
Throughout this book I look at three different measures of inequality in the distribution of income and its components. First, I report the Gini coefficient, a measure of statistical dispersion that rises as income inequality rises. In a situation of perfect equality, where all incomes are equal, the Gini coefficient equals zero. At extreme inequality (where one person has all the income and everyone else has zero), the Gini coefficient equals one. Hence, as the Gini coefficient rises, income inequality increases. Gini coefficients for the distribution of family income in most advanced industrialized nations have historically been between 0.22 and 0.40.
Second, I show the coefficient of variation (CV), which is the standard deviation of the data divided by its mean. The CV provides a measure of how dispersed the data are around their mean. Like the Gini coefficient, it equals zero when all incomes are equal and rises with growing inequality. Highly unequal incomes can result in CVs that are well above one, however. For any significant changes in inequality, one would expect the Gini and the CV to move together; by looking at both of them I can confirm that the trends in the data do not vary with measurement choice.
Third, I report comparisons of the data at several different points in the distribution, typically looking at the 90/10 ratio, which compares the data at the ninetieth percentile in its distribution with the data at the tenth percentile. Growing inequality means a rising 90/10 ratio. I will also look at the 90/50 ratio and the 50/10 ratio (the fiftieth point in the distribution is the median), in order to see if rising inequality is concentrated at the top or the bottom of the distribution. As noted earlier, changes in inequality driven by rises in top incomes may be viewed differently from changes in inequality driven by declines in bottom incomes. If inequality is changing only slightly or only at certain points in the distribution, these measures might provide conflicting information. If inequality is shifting across the entire distribution, all of these measures should indicate rising inequality.
Excerpted from Changing Inequality by Rebecca M. Blank. Copyright © 2011 The Regents of the University of California. Excerpted by permission of UNIVERSITY OF CALIFORNIA PRESS.
All rights reserved. No part of this excerpt may be reproduced or reprinted without permission in writing from the publisher.
Excerpts are provided by Dial-A-Book Inc. solely for the personal use of visitors to this web site.
Table of ContentsList of Illustrations
Part I. Changes in Income and Earnings
1. A Broader Look at Changing Inequality
2. Changing Inequality in Annual Earnings and Its Components
3. Changing Inequality in Total Income and Its Components
4. Understanding These Changes
Part II. Can Inequality be Reduced?
5. How Economic Shocks Change Income Distribution
6. Ways to Reduce Inequality (and Their Limits)
7. Changing Inequality in the United States Today
Appendix 1. Details of the Chapter 2 Simulation and Appendix Figures
Appendix 2. Income Components by Decile
Appendix 3. Details of the Chapter 4 Simulations
Appendix 4. Details of the Chapter 6 Simulations
What People are Saying About This
"Blank offers insight on a topic of much current debate."Choice