Trends in Income Inequality, Volatility, and Mobility Risk

Trends in Income Inequality, Volatility, and Mobility Risk Via Intertemporal Variability Decompositions Austin Nichols Abstract A unified measure of inequality, volatility, and mobility risk is developed from well-known decompositions of a generalized entropy inequality measure. Decomposition by population subgroup is applied to panel data measuring income across people over time, where subgroups are taken to be individual people, so “within-group” inequality is a measure of the variability of individual income over time and “between-group” inequality is a measure of inequality in long-run mean incomes. The variability of individual income over time is further decomposed into “volatility” and “mobility risk” using individual-specific trends in income. I apply the decompositions to several decades of U.S. data and find every component (inequality, volatility, and mobility risk) increasing over time, and a large impact of taxes. I further find large swings in the progressivity of income growth after taxes that are not observed in pretax income, consistent with the known tax regimes in recent U.S. history. Introduction There has been a renewed interest in recent years in income inequality, but also economic mobility (both moving up and moving down), and income volatility, the year-to-year variations in income that families may or may not be able to smooth over. This is not just about the business cycle: the percentage of Americans worried about being laid off nearly quadrupled from 12 to 46 percent1 from 1982 to 1998, as the economy improved dramatically. Over the same time frame, workers reported their subjective probability of job loss fell from close to 20 to less than 10 percent.2 To address these concerns, researchers, journalists, and politicians have joined the fray, seeking to measure and explain income inequality, mobility and volatility. These related phenomena have different implications; as Senator Schumer said, “If you’re holding a job but your share of the pie is getting smaller, that’s a different set of policy needs than if you keep losing your job” (quoted by Leonhardt 2007). To date, however, there has been no unified approach to measuring these phenomena. I define an aggregate measure of income risk as half the squared coefficient of variation (or the general entropy measure with parameter 2, denoted GE2) of incomes measured over both people and time. The aggregate measure can be decomposed into an inequality component measuring dispersion in mean incomes, a volatility component measuring the average dispersion of fluctuations about person-specific trends, and a mobility component measuring the dispersion of person-specific trends. I then apply this decomposition to Panel Study of Income Dynamics (PSID) data from the United States to characterize trends in inequality, volatility, and mobility over the last several decades. I also examine the regressivity of income growth in these data. Background It is of little use to discuss inequality without some mention of changing incomes over time. They may be changing due to short-lived transitory shocks, or more permanent changes, but either kind of change induces greater volatility in the income stream and greater relative3 mobility. Some view these changes as mitigating inequality (frequently citing Schumpeter 1955 or Friedman 1962), but if these changes reflect income risk, they lower well-being, holding constant the mean level of income. This paper inclines toward the latter viewpoint, characterizing observed changes in income as reflecting dispersion, and I define an aggregate measure of income risk that can be decomposed into a long-run inequality component, a volatility component, and a mobility component. Of course, greater absolute mobility, especially growth in real incomes, may change our interpretation of these other features. A doubling of real incomes may make us less Original version June 3, 2008; this draft November 4, 2008. The author acknowledges helpful comments from Gregory Acs and participants at the IRISS 10th Anniversary Workshop in Differdange, Luxembourg, October 24-25, 2008, including Stephen Jenkins, Markus Jäntti, Martin Biewen, and Philippe Van Kerm. 1 Nichols and Zimmerman (2008), quoting poll results from the International Survey Research Corporation in Chicago; see also Greenspan (1997). 2 Leonhardt (2007), quoting Gallup poll results. 3 I am calling growth in incomes relative to average growth “relative mobility” so both volatility and mobility may result in reranking of individuals within the income distribution in any one period or over time. 2 worried about increasing trends in volatility combined with constant relative mobility and increasing inequality, though the change in measured (scale-invariant) inequality due to a doubling of incomes is nil. I will briefly discuss income growth toward the end, but for most of the text, I will restrict my attention