Wealth inequality and the Great Recession Evidence from

Wealth inequality and the Great Recession: Evidence from Sweden Jacob Lundberg and Daniel Waldenström Uppsala University Presentation at Mi. SOC/ISER, September 16, 2014 1

The economy in Sweden 1999 -2012 U and Y Disposable income Gini 2

This paper Two main objectives: • First, estimate of wealth distribution in Sweden since 2008. – Wealth tax repealed in 2007, Property tax reformed in 2008 – No household wealth data exist after 2007 in Sweden! • Second, study role of Great Recession for wealth distribution – Shocks to output, growth, unemployment and stock prices – Were the rich able to escape the crisis or did they take the strongest blow? – Did crisis experiences differ among the rich, across occupations etc? Paper structure • Estimating private assets and liabilities • Estimate wealth distribution 1999 -2012 3

Recent research on other countries • United States – Wolff (SCF, 2012): "Asset price meltdown for the middle class" • Plummeting house prices main explanation – Saez and Zucman (2014): "Wealth inequality is making a comeback" • New estimates using capitalized capital income • Germany: – Grabka and Westermeier (GSOEP, 2014): Small changes in both wealth levels and household wealth inequality 4

Data • Database: LINDA – 3, 35% nationally representative sample (300, 000 individuals) – Panel – Household members added (in total about 800, 000 obs) • Wealth data sources – Wealth Register 1999 -2007 – Other registers 2008 - (Income and Tax Register, Property Tax Register, Land survey register) • Additional information – Firm-specific dividends (SIX financial market information) – Financial Accounts (macro aggregate on household financial balances) 5

Results Part I: New estimates of individual wealth 2008‒ 2012 6

Estimating individuals' wealth in 2008 -2012 • We use different approaches – – Observed actual stocks (non-financial assets, debts) Observed dividend income and payout/share Regression-based "Investment income method" Chain with price changes assuming no transactions • Calibrate model using "true" outcomes during 1999 -2007 • Problem: too little noise in imputed variables and is added noise independent of the variables – Rubin (1976), Pagan (1984), Dardanoni et al (2011) – We abstract from this issue so far • Compute inequality estimates on individuals (so far) 7

Estimation of real estate/property • 8

Average housing value: estimated vs actual 1200000 1000000 SEK 800000 600000 400000 200000 0 1997 1999 2001 2003 Average housing value in Wealth Register 2005 2007 2009 2011 2013 Our estimated average houseing value 9

Real estate estimations vs actual Mean Gini 10

Financial asset estimations, 2008 • 11

Bank deposit estimations vs actual Mean Gini 12

Estimation of listed shares • 13

Value of shares in 2007: Estimation / "Truth" Capitalization method (Dividend / rate of return) Our direct estimation 14

Estimation of other financial assets • Mutual funds – Problematic since dividends typically reinvested – We use value of 2007 and calculate forward using value indices – Assumes no net savings • Bonds and fixed-interest securities – Not fully covered. Premium lottery bonds not observed. – Use interest earnings to estimate value • Condominiums – Simple approach: Extrapolate individual's 2007 ownership using municipality price indices. Assuming no transactions! • Solvable using new Apartment Register at Statistics Sweden (Census 2011) 15

Estimation of liabilities • Two types of loans: State student loans (15%), Mortgages (85%) – We observe student loans • Mortgages are observed in Inteckningar/Pantbrev – Land Survey Register (Lantmäteriet) – By year, property and individual (incl. debt share) • Differs from capitalization method (Debt = interest payments/interest rate) 16

Liabilities projection vs actual Mean Gini 17

LIabilities in 2007: Estimation / "Truth" Capitalization method (Interest payment / rate of return) Our direct estimation from mortgage contracts 18

Results Part II: How did the wealth of Swedes fare during the Great Recession? 19

Results 1: How well do our estimations perform? • Average total wealth: Actual vs Estimated 800000 700000 600000 500000 Estimated net wealth Net wealth 400000 300000 200000 100000 0 2005 2010 20

Results 1: How well do our estimations perform? • Histogram of the ratio of predicted to true net wealth in 2007, using only data for the years up to 2004 for the prediction 21

Results 2: Wealth levels across the distribution • Average wealth across deciles. No strong crisis impact! 6000000 5000000 4000000 3000000 2000000 1000000 0 2002 2004 2006 2008 2010 2012 2014 -1000000 22

Results 2: Wealth inequality • Gini coefficient of net wealth according to Wealth Register (– 2007) and our imputations (2007–) – A level break in 2007, but no strong trend in 2008 -2012 0. 75 0. 73 0. 71 0. 69 0. 67 0. 65 1997 1999 2001 2003 2005 2007 2009 2011 2013 23

Results 2: Wealth inequality • Figure: Top wealth shares (Top 10%, Top 1%, Top 0. 1%) according to the Wealth Register (– 2007) and our imputations (2007–) – No strong trend after crisis (similar to Gini) 0. 7 0. 6 Top 10% 0. 5 0. 4 0. 3 0. 2 Top 1% 0. 1 Top 0. 1% 0 1997 1999 2001 2003 2005 2007 2009 2011 2013 24

Results 3: Did net wealth trends of financial sector employees differ? • Figure: Average wealth of financial sector employees vs all others – A drop in 2008, but after that continued increase in difference 2000000 1800000 1600000 1400000 Financial sector employees 1200000 1000000 800000 600000 All others 400000 200000 0 2002 2004 2006 2008 2010 2012 25

Results 3: Wealth trajectories across age and sex Age Sex 26

To be done. . . • Remake analysis using household • Improve measurement further – Non-listed equity • Half of all privately owned business equity (Waldenström, 2014) • Match estimated firm values from Firm register FRIDA – Consumer durables • Car Register (estimate car values) • Finer checks on the impact of the crisis – Impacts across asset classes – Decomposition of Gini across assets and time – Look at distribution of capital incomes 27

Concluding remarks • We investigate how the wealth of Swedish households developed before, during and after the Great Recession • The 2007 wealth tax repeal implies a lack of official wealth statitics – We estimate household wealth using income and property tax registers • Preliminary findings: Seemingly small effects of the crisis overall • Wealth inequality (Gini, top wealth shares) exhibit no breaks in levels or trends • Financial sector employees seem to experience a short-run hit, but long-run trend unbroken 28