Inequality in 3D: Income, Consumption, and Wealth David Johnson Jonathan Fisher Tim Smeeding Jeff Thompson WID.world conference Dec 14-15, 2017 Thanks to Russell Sage Foundation and Washington Center for Equitable Growth
Various Ginis for income show increasing inequality 0.90 0.80 0.70 0.60 0.50 0.40 SCF Household Income (Wolff) Money Income (Census) After tax and Transfer Income (CBO) 0.30 0.20 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012
Wealth inequality is higher and has increased more than income in recent years 0.90 0.80 SCF Household Wealth (Wolff) 0.70 0.60 0.50 0.40 SCF Household Income (Wolff) Money Income (Census) After tax and Transfer Income (CBO) 0.30 0.20 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012
Consumption inequality is lower and has increased at rates similar to income 0.90 0.80 SCF Household Wealth (Wolff) 0.70 0.60 0.50 0.40 0.30 0.20 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 SCF Household Income (Wolff) Money Income (Census) After tax and Transfer Income (CBO) Consumption (CE - FJS) Consumption (Attanasio/Pistaferri)
Why three dimensions? Report by the Commission on the Measurement of Economic Performance and Social Progress (Stiglitz et al., 2009) wrote the most pertinent measures of the distribution of material living standards are probably based on jointly considering the income, consumption, and wealth position of households or individuals. Income, consumption, and wealth are not perfectly correlated, and there are life-cycle patterns in all three Need to account for budget constraint cc tt = yy tt - aa tt+1 + aa tt (1 + rr tt )
The need for using one data set The conclusion we draw is that one should be very cautious when combining data on inequality in wages and earnings from the CPS or PSID, and data on inequality in net worth from the SCF. - Heathcote, Perri, Violante (2010)
Data landscape in the United States Survey of Consumer Finances (SCF) Dual-Frame Sample National Area Probability Sample List Sample High wealth households Triennial: 1989-2016 Unit of observation is the primary economic unit Income, wealth, and some consumption Consumption for food, housing, and vehicles Consumer Expenditure (CE) Survey National Area Probability Sample Annual: 1980-2017 Used for weights for Consumer Price Index Unit of observation is the consumer unit Income, consumption, and some wealth Wealth includes owned home, vehicles, and some assets Panel Study of Income Dynamics (PSID) Nationally representative data beginning in 1968 Biennial survey since 1997 Unit of observation is the family Income every wave Wealth and consumption every wave since 1999
SCF captures more income at the top Ratio of SCF to CE income
Wealth and Income in SCF closely match In the beginning national aggregates SCF Wealth and Financial Accounts SCF Income and National Accounts
Definitions of income, consumption, and wealth Disposable Income (similar to Luxembourg Income Study + CG) Money income: income from employment, investment, government cash transfers, and interhousehold transfers of money Plus in-kind transfers Plus realized capital gains Less net taxes (using NBER TAXSIM) Consumption Total spending on food, housing, nondurables, transportation, other durables, education, health, and child care. Imputed service flow for homeowners. Imputed service flow from vehicles. Imputed rent for those living in subsidized housing Wealth (similar to Luxembourg Wealth Study) Assets including stocks, bonds, mutual funds, home-equity, residential real estate, and business assets Less all debt including mortgage, credit cards, student debt, and business debt
Imputing consumption to the Survey of Consumer Finances Skinner (1987) and Fisher and Johnson (2006) ln(total Consumption) = α 0 + α 1 *food home +α 2 *food away + X γ + ν Blundell, Pistaferri, and Preston (2008) ln(food at home) = M μ + β*ln(total Consumption) + e
Imputing unreported consumption Follow Skinner (1987) and Fisher and Johnson (2006) with some modifications: Minimize the prediction error by only imputing unreported consumption Use predictive mean matching across the CE and SCF Impute ratio of reported to total separate regressions by year impute five times Separate regression by spending-to-income groups Over the past year, would you say that your spending: - exceeded your income - that it was about the same as your income - that you spent less than your income?
SCF has higher reported and imputed consumption than CE at the higher ventiles Reported consumption Imputed consumption
Inequality in 1D In the beginning
Own Shares for top 5% of Wealth, Income, and Consumption 70% Wealth 60% 50% Saez-Zucman 40% 30% Piketty-Saez Income 20% 10% Consumption 0% 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
Own Shares for top 5% of Wealth, Income, and Consumption 70% Wealth 60% 50% 40% Income 30% 20% Disposable Income Consumption 10% 0% 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
SCF has a higher Consumption Gini than CE (1989-2016)
Mainly due to the high income oversample in the SCF (1989-2016)
Inequality in 2D In the beginning
2-D inequality: Percent of households in top 5% of two measures
Recall 1D: Own Shares for top 5% of Wealth, Income, and Consumption 70% Wealth 60% 50% 40% 30% 20% Income Consumption 10% 0% 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
2D: Shares of Wealth, Income and Consumption for top 5% of Wealth 70% 60% Wealth 50% 40% 30% Income 20% 10% Consumption 0% 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
2D: Wealth Shares (own and cross) for top 5% of Wealth, Income, and Consumption 70% 60% Own share: Wealth 50% 40% Cross share: Consumption Cross share: Income 30% 20% 10% 0% 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
Top 5% have a larger increase in Income In the beginning for those in top 5% of Wealth and Consumption 500,000 450,000 400,000 350,000 300,000 250,000 Top 5% of Wealth & Consumption 205% 187% 215% 200,000 150,000 Top 5% of Income Top 5% of Wealth 100,000 50,000 Median Income 122% 0 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
Inequality in 3D In the beginning
Recall 2D: Wealth Shares (own and cross) for top 5% of Income and Consumption 70% 60% 50% 40% Cross share: Income Own share: Wealth Cross share: Consumption 30% 20% 10% 0% 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
3D: Wealth Shares (own and cross) for top 5% of Income/Wealth and Income/Consumption 70% 60% 50% 40% 30% Cross share: Income Own share: Wealth Cross share: Consumption Cross share: Income and Wealth Cross share: Income and Consumption 20% 10% 0% 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
2D and 3D Wealth Shares increase more than 1D own Wealth Shares 1.60 1.50 1.40 1.30 Cross share: I&C Cross share: I&W Cross share: Cons Cross share: Income 1.20 1.10 Own share: Wealth 1.00 0.90 0.80 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
2D and 3D Income Shares increase more than 1D own Income Shares 1.60 1.50 Cross share: W&C 1.40 1.30 1.20 1.10 1.00 Cross share: Y&C Own share: Income Cross share: Cons Cross share: Wealth 0.90 0.80 1989 1992 1995 1998 2001 2004 2007 2010 2013
Same pattern occurs for top quintile 1.30 1.25 Cross share: Consumption 1.20 1.15 1.10 1.05 1.00 Cross share: Income Own share: Wealth 0.95 0.90 0.85 0.80 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016
PSID shows similar increases in 2D (using quintiles) 1.6 1.5 1.4 Cross share: Income 1.3 1.2 1.1 Cross share: Consumption 1 0.9 Own share: Wealth 0.8 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Mainly because the Percent of Households in Top 5% of all three measures increases
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: all quintiles
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: all quintiles
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: Q1
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: Q2
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: Q3
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: Q4
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: Q5
In the beginning Inequality in 3D: Transition matrix for income and consumption quintiles by wealth quintile: all quintiles
Determining 2D and 3D measures using entire distribution Using pairwise Shorrocks measures shows correlations do not rise over time Using Gini correlation shows falling correlations over time
Next Steps Create a summary measure for inequality in 3D Integrate distributions with National accounts Examine the off-diagonal people those who aren t at the top and bottom for all three Fully examine the life-cycle relationships and mobility Compare to PSID and impute consumption and wealth back to 1968 Use PSID to measure intra- and inter-generational mobility Eurostat-OECD Expert Group on Measuring the Joint Distribution of Household Income, Consumption and Wealth at Micro Level