Factors behind rising U.S. wealth inequality: a decomposition and policy implications
0 likes69 views
We offer a novel decomposition of wealth inequality, to quantify the three principal factors of changes in wealth inequality: saving behavior, structural changes in the age distribution of population, and the generational exchange. Our findings indicate that the rise in U.S. wealth inequality since mid-1970s is largely driven by between-cohort disparities, a dimension often underestimated. Whereas saving behavior effectively mitigates inequality, demographic dynamics exert an increasingly amplifying effect. Quantitatively, rising longevity magnifies wealth disparities across cohorts. In sum, rising wealth inequality does not signify jeopardizing equity, with important implications for public policy. We also discuss implications of our research for structural models of wealth inequality.
![Factors behind rising U.S. wealth inequality
A decomposition and policy implications
Marcin Lewandowski [FAME|GRAPE & University of Warsaw]
Krzysztof Makarski [FAME|GRAPE & Warsaw School of Economics]
Jan Svejnar [Columbia University & CERGE-EI]
Joanna Tyrowicz [FAME|GRAPE, University of Warsaw & IZA ]
ECINEQ
Washington, July 2025
1 / 33](https://image.slidesharecdn.com/presentation-250712190410-5403d97d/85/Factors-behind-rising-U-S-wealth-inequality-a-decomposition-and-policy-implications-1-320.jpg)
![Identifying the role of specific birth-cohorts
Question: Do cohorts with ↑ LE65 contribute ↑ to overall inequality?
Two tricks:
• Deaton and Paxson (1994) decomposition
• Recentered Influence Functions: Firpo et al. (2009) Rios-Avila (2020)
RIF{wealthi , GE()} = βc birth cohortc + βaagea + βy yeary + i
GE() = E[RIF{wealthi , GE()}]
βc – unconditional partial effect of cohort on distributional statistics (GE)
27 / 33](https://image.slidesharecdn.com/presentation-250712190410-5403d97d/85/Factors-behind-rising-U-S-wealth-inequality-a-decomposition-and-policy-implications-27-320.jpg)
Factors behind rising U.S. wealth inequality: a decomposition and policy implications
- 1. Factors behind rising U.S. wealth inequality A decomposition and policy implications Marcin Lewandowski [FAME|GRAPE & University of Warsaw] Krzysztof Makarski [FAME|GRAPE & Warsaw School of Economics] Jan Svejnar [Columbia University & CERGE-EI] Joanna Tyrowicz [FAME|GRAPE, University of Warsaw & IZA ] ECINEQ Washington, July 2025 1 / 33
- 2. Motivation 1: The importance of wealth inequality • Wealth inequality in the United States has been rising since mid-1970s • Much attention from both policy makers and academics • Barack Obama calling it “the defining challenge of our times” • According to Pew Research Center, 61% Americans share this sentiment Can the story of wealth inequality be told without demography? 2 / 33
- 3. Motivation 2: The relevance of demographics Life expectancy past 65 increased substantially Up 5+ years: 14.1 years (1950) → 19.5 years (2015), more than 30% Theory: longevity affects wealth distribution 1. BEHAVIORAL: ↑ incentives for old-age saving ↑ widening gap between young and around-retirement 2. STRUCTURAL: ↑ shifts in cohorts shares (baby boomers) Empirical question: do demographic changes matter quantitatively for wealth inequality? 3 / 33
- 4. Our contribution: • New decomposition of (changes in) wealth inequality applied to the U.S. data (1949-2016): 1. Adjustment in saving behavior of surviving cohorts 2. Changes in population structure due to shifts in shares of surviving cohorts 3. Impact of generational exchange i.e. inequality contributed by born vs deceased cohorts • Main findings: 1. Magnitudes of the three factors large 2. Demography (orange+green) trumps saving behavior 3. Saving behavior ⇓ inequality 4. Cohorts with the largest longevity gains contribute the most to wealth inequality 4 / 33
- 6. Two dimensions of wealth inequality Total inequality = Inequality within cohorts + Inequality between cohorts 1. Within: same birth cohort 2. Between: different birth cohorts U.S. wealth inequality has been increasingly driven by between-cohort wealth inequality 6 / 33
- 7. U.S. wealth inequality has been increasingly driven by between-cohort wealth inequality 7 / 33
- 8. U.S. wealth inequality has been increasingly driven by between-cohort wealth inequality 8 / 33
- 9. Key concepts: Generalized entropy Generalized entropy GE() = GEbetween + GEwithin = 1 ( − 1) C X c=1 h nc ac a − 1 i + C X c=1 h nc ac a GEc () i • c – cohort • – sensitivity • nc – share of particular c in the population • ac a – average c wealth relative to global average • GEc () – wealth inequality inside each c 9 / 33
- 10. Key concepts: measuring the change in wealth inequality A complete decomposition of ∆ in wealth inequality: generalized entropy ∆GE =∆GEbetween + ∆GEwithin ∆GEbetween = ∆nc + ∆ ac a | {z } surviving cohorts +generational exchange ∆GEwithin = ∆nc + ∆ ac a + ∆GE()c | {z } surviving cohorts +generational exchange 10 / 33
- 11. Decomposing wealth inequality We formulate three propositions, for the part played by 1. exchange of generations 2. population structure 3. saving behavior in the change of wealth inequality. 11 / 33
- 12. Decomposing wealth inequality: the exchange of generations (ins outs) Proposition 1: Generational exchange The ∆ in wealth inequality implied by the exchange of generations between periods t1 and t2 can be decomposed: • Component 1: a between-cohort effect of the generational exchange (recently born cohorts, indexed by b; recently deceased indexed by d) 1 2 − B X b=1 nt2,b · at2,b at2 − D X d=1 nt1,d · at1,d at1 # • Component 2: a within-cohort effect of the generational exchange B X b=1 nt2,b · GEw ()t2,b · at2,b at2 − D X d=1 nt1,d · GEw ()t1,c · at1,d at1 12 / 33
- 13. Decomposing wealth inequality: the change of population structure of survivors Proposition 2: Population shares. The ∆ in wealth inequality implied by a change in the population structure between periods t1 and t2 can be decomposed into: • Component 3: a between-cohort effect change in population shares of surviving cohorts 1 2 − C X c=1 (nt2,c − nt1,c ) · at2,c at2 • Component 4: a within-cohort effect of the change in population shares of surviving cohorts C X c=1 (nt2,c − nt1,c ) · GEw ()(t1,c) · at2,c at2 13 / 33
- 14. Decomposing wealth inequality: saving behavior of survivors Proposition 3: Saving behavior The ∆ in wealth inequality implied by a change in individuals’ saving behavior between periods t1 and t2 can be decomposed into: • Component 5: a between-cohort change in cohort wealth (saving behavior) of surviving cohorts 1 2 − C X c=1 n(t1,c) · at2,c at2 − at1,c at1 • Component 6: a within-cohort change in cohort wealth (saving behavior) of surviving cohorts C X c=1 n(t1,c) · GEw ()(t1,c) · at2,c at2 − at1,c) at1 • Component 7: a change in within-cohort GE index (saving behavior) of the surviving cohorts C X c=1 n(t1,c) · GEw ()(t2,c) − GEw ()(t1,c) · at2,c at2 14 / 33
- 15. Implications for modeling wealth inequality The three propositions yield the following components: Saving behavior: • Component 7: Within-cohort driven by GE()c Xfor infinitely lived agents • Component 6: Within-cohort driven by shifts in average cohort wealth requires OLG • Component 5: Between-cohort driven by shifts in average cohort wealth requires OLG Population shares: • Component 4: Within-cohort driven by shifts in population structure requires OLG+ • Component 3: Between-cohort driven by shifts in population structure requires OLG+ Exchange of generations • Component 2: Within-cohort driven by exchange of generations requires OLG+ • Component 1: Between-cohort driven by exchange of generations requires OLG+ 15 / 33
- 16. Take this method to the data 16 / 33
- 17. Data source Survey of Consumer Finances (SCF+) data for 1950-2020 by Kuhn et al (2020) • Assets - Financial assets (including defined-contribution retirement plans), real estate, cars • Liabilities - Personal debt and housing debt • Our measure - Net wealth Demographic characteristics match Current Population Survey and U.S. Census data 17 / 33
- 18. Choose the (sensitivity) parameter: match Gini 18 / 33
- 19. But this parameter does well also for other measures: top10 19 / 33
- 20. Results 20 / 33
- 21. Results 1: Decomposition with net impact 21 / 33
- 22. Results 2: Decomposition with all seven components 22 / 33
- 23. Results 2 : Decomposition with all channels 1. Pattern consistent over time: cohort-level effects are HUGE 2. Pattern consistent over time: Within generational exchange0 (the solid bar) 3. Pattern consistent over time: Between generational exchange0 (the striped bar) 4. Pattern changing over time: Within GE change shifts the sign, though its magnitude is modest 5. Pattern changing over time: the magnitude of Between population structure ⇑ 6. Pattern changing over time: the magnitude of Between: generational exchange ⇓ 23 / 33
- 24. This is a consistent pattern, only magnitudes change (driving net changes) 24 / 33
- 25. The full decomposition • Overall cohort-level effects are HUGE • Within:generational exchange0, but outweighed by Between:generational exchange0 • Within:GE change shifts the sign and its magnitude is increasing 25 / 33
- 26. This is a consistent pattern, only magnitudes change (driving net changes) 26 / 33
- 27. Identifying the role of specific birth-cohorts Question: Do cohorts with ↑ LE65 contribute ↑ to overall inequality? Two tricks: • Deaton and Paxson (1994) decomposition • Recentered Influence Functions: Firpo et al. (2009) Rios-Avila (2020) RIF{wealthi , GE()} = βc birth cohortc + βaagea + βy yeary + i GE() = E[RIF{wealthi , GE()}] βc – unconditional partial effect of cohort on distributional statistics (GE) 27 / 33
- 28. Greater life expectancy =⇒ greater contribution to wealth inequality 1880 1900 1895 1885 1890 1905 1910 1915 1925 1930 1935 1940 1945 1950 1955 1960 1965 1970 −1 −.8 −.6 −.4 −.2 0 .2 .4 .6 .8 1 1.2 GE: Cohort effects relative to cohort born 1920−24 (adjusting for age and year effects) −2 0 2 4 6 Change in life expectancy at age=65, relative to cohort born in 1920−1924 GE: CI (left axis) GE: point estimate of efect size (left axis) 28 / 33
- 30. Demographics matters • Population structure is a key force ⇑ wealth inequality (on the net) • Generational exchange used to ⇓ wealth inequality (on the net) • Saving behavior ⇓ wealth inequality (on the net) • Demographic factors saving behavior • Demographics implies shifts much larger than their net effect • Not per-se evidence of growing inequity Overall: • Demographic factors should not be missed in the policy debates • Economic modeling cannot ignore demographics 30 / 33
- 31. Thank you for your attention! Questions or suggestions? w: grape.org.pl | Working Paper #100 t: grape org f: grape.org e: [email protected] 31 / 33
- 32. Bibliography Deaton, A. S. and Paxson, C.: 1994, Saving, growth, and aging in taiwan, Studies in the Economics of Aging, University of Chicago Press, pp. 331–362. Firpo, S., Fortin, N. M. and Lemieux, T.: 2009, Unconditional quantile regressions, Econometrica 77(3), 953–973. Rios-Avila, F.: 2020, Recentered influence functions (rifs) in stata: Rif regression and rif decomposition, The Stata Journal 20(1), 51–94. 32 / 33