TALL POPPIES IN THE LAND OF THE FAIR GO: WHY HAS AUSTRALIAN INEQUALITY RISEN AND DOES IT MATTER?

Transcription

1 TALL POPPIES IN THE LAND OF THE FAIR GO: WHY HAS AUSTRALIAN INEQUALITY RISEN AND DOES IT MATTER? ACE 2012 Andrew Leigh MP (+many co-authors)

2 The Wealth Ladder 2 Imagine a ladder in which each rung represents $1 million of household wealth. 50% of all households are closer to the ground than they are to the first rung. The typical Australian household is halfway to the first rung. Someone in the top 10 percent is at least 1½ rungs up. A household in the top 1 percent is at least 5 rungs up. Gina Reinhardt is 8 kilometres off the ground.

3 But Should Economists Care? 3 Robert Lucas: of all the tendencies that are harmful to sound economics, the most seductive, and in my opinion, the most poisonous, is to focus on questions of distribution. Marty Feldstein: there is nothing wrong with an increase in well-being of the wealthy or with an increase in inequality that results [solely] from a rise in high incomes. Isn t this just the politics of envy?

4 Roadmap 4 Top incomes methodology What do we know about inequality (especially at the top)? What drives inequality? Why should economists care? Conclusions & ideas for future research

5 5 Methodology

6 Top Incomes Using Tax Data 6 The use of tax data to study inequality goes back to Kuznets in the 1950s Piketty in 1990s revisits the data using external control totals for population and income. Advantages: annual data back to early-c20th But also some risks.

7 7

8 Tax Evasion? 8 TCMP indicates no major changes in the US from 1970s to 2000s (and in Sweden from 1930s to 1990s) Across countries, random audit studies give similar results. In the WVS, respondents are asked whether cheating on taxes is ever justifiable. 1= never justifiable, 10= always justifiable. Across 13 developed countries, the mean is between 2.1 and 2.7 for all except two nations (Japan is 1.5, France is 3.1). Across income types, underreporting is lowest for wage income, which has grown as a share of top incomes.

9 Other Methodological Issues 9 Banded income Part-year incomes Tax year not always calendar year Adult population control total: 15+, 18+, 20+ Income unit: couples or singles Personal income control total Income definition: taxable and total Capital gains: excluded from what follows (where possible)

10 Relation to Other Ineq. Measures 10 Limitations of top incomes for looking at the whole distribution: - They miss the bottom 90% (only weakly satisfy the Pigou-Dalton transfer principle) - Generally based on pre-tax incomes - Not adjusted for household size. But they track the Gini pretty well anyway.

11 Gini.3 Change in Gini Figure 5: Top Income Shares and LIS Gini Coefficients Inequality levels (circa 2000) Inequality changes (c1980 to c2000) GBR USA USA SWE AUS ESP IRL GBR CAN ESP DEU AUS CAN CHE FRA DEU SWE CHE FRA IRL NLD NLD Top 10% Share Change in Top 10% Share Notes: 1. New Zealand is not shown, since it is not in the LIS. 2. Gini Coefficients are based on equivalised post-tax income. 3. Top income shares are unequivalised and based on pre-tax income. 11

12 12 Trends in Inequality

13 From , noone makes William Rubinstein s all-time rich list In 1960s, the rich feel under some pressure to be accepted by ordinary working Australians rather than the other way round (Craig McGregor) Since 1980, top 1% have gotten 13% of income gains. Two >$1M political donations in recent years 13

14 14

15 15

16 Fig 3: Income Share of Richest 1% in Anglo-Saxon Countries Tax Year Australia Ireland UK Canada New Zealand US 16

17 Fig 4: Income Share of Richest 1% in Non Anglo-Saxon Countries Tax Year France Japan Spain Switzerland Germany Netherlands Sweden 17

18 Wealth Inequality 18 Wealth is slower-moving than income, since it depends on savings and intergenerational transfers. In the US, top wealth shares have not risen as fast as top income shares over the past generation. Wealth also throws up larger issues of comparability.

19 19

20 20 What Drives Inequality?

21 Drivers of Inequality 21 Yes Superstar labor markets (English-speaking) Top tax rates on labor & capital Wars destruction of capital stock, tax rates Possibly Left/right government Yet to be properly tested SBTC, firm size, trade, immigration, inflation, demographic structure

22 Figure 8: Share of the Top 1% (solid line, left axis) & 1-Top Marginal Tax Rate (dashed line, right axis) Australia Canada New Zealand UK US

23 Drivers - Taxes 23 Empirical strategy: Atkinson & Leigh use data for five Anglo-Saxon countries since 1920, estimating median rate paid by top 1% and top tax rate. Two channels: immediate incentive effects and long-term accumulation effects. We use both, in an IV strategy. Result: - A 10 percentage point cut in the median MTR paid by the top 1% raises their income share by percent. - Since the mean of the top 1% share is around 10%, this equates to a percentage point increase.

24 Drivers - Taxes 24 This suggests that taxes can explain 1/3 rd to ½ of the rise in top income shares in Anglo countries over the period Our result also allows us to estimate Laffer-type effects, ie. what is the MTR that maximizes tax revenue from the top 1%? We calculate this at 63-83% (higher than the top tax rate in any Anglo country today).

25 Table 1: Top Income Shares and Partisanship Since 1960 Country ΔTop 10% Share ΔTop 1% Share Total years Rightwing govt Left-wing govt Diff (R-L) P-value on diff Rightwing govt Left-wing govt Diff (R-L) P-value on diff Rightwing govt Left-wing govt Australia Canada France Germany Ireland Japan Netherlands New Zealand Spain Sweden Switzerland United Kingdom United States Mean Source: Leigh (2009) 25

26 26 Why Should We Care?

27 Consequences of Inequality 27 Yes No In last 40 years, top incomes are good for growth (consistent with Forbes 2000 using DS) Health and savings are unaffected Maybe Campaign contributions, industrial innovation, political polarization, residential segregation, trust, happiness, social mobility

28 Table 3: Top Incomes and Growth Since 1960 Dependent variable: Average annual per capita growth (5 year periods) [1] [2] [3] [4] [5] [6] OLS OLS RE RE FE FE Panel A: Top 10% Share Income share of richest 10% 0.073** *** *** 0.112*** [0.026] [0.032] [0.021] [0.025] [0.016] [0.025] Log(GDP per capita) *** *** *** *** [0.264] [0.418] [0.566] [0.743] [2.917] [3.017] Investment and Schooling Controls? No Yes No Yes No Yes Country Fixed Effects? No No No No Yes Yes Year Fixed Effects? No No Yes Yes Yes Yes R Countries Observations Source: Andrews, Jencks & Leigh (2011) 28

29 9 Figure 3: Trends from 1970 to 2003 for the Income of the Top Decile and the Bottom Nine Deciles in the United States: Observed Means versus Predicted Means if the Top Decile s Share Had Stayed at Its 1970 Level Observed mean for top decile Log of mean income Estimated mean for top decile if inequality stayed at 1970 level Estimated mean for bottom nine deciles if inequality stayed at 1970 level Observed mean for bottom nine deciles Simulations assume that the ratio of personal income to GDP is 0.83 (the mean value for the US over this period). 29

30 Source: Leigh & Jencks (2007) 30

31 Table 1: Inequality and Savings, Dependent variable is the gross national savings rate [1] [2] Lag: T-1 Measure: Top 10% Top 1% Panel A: Without country and year fixed effects Inequality * ** [0.167] [0.244] Observations R-squared Panel B: With fixed effects, without interest rate controls Inequality [0.177] [0.257] Observations R-squared Panel C: With fixed effects and interest rate controls Inequality [0.178] [0.261] Observations R-squared Source: Leigh & Posso (2009) 31

32 Other Impacts of Inequality 32 Social mobility (data are poor, but lags are long)

33 Other Impacts of Inequality 33 In maximising utility, many people exhibit a preference for equality. Ultimatum game: offers <25% are often rejected This sense of egalitarianism reflects what Lester Thurow termed the income distribution as a public good Practical implications: positional goods (prestigious jobs, university places, home in a good suburb)

34 34 Future Research

35 Future Research 35 New countries: Southern Europe Who are the rich? Gender, age, occupation, education. Mobility over the lifecycle and across generations? Sub-national top incomes estimates (especially US states) Testing additional causes and consequences.

36 Contact Details 36 Web: Top income papers available on website: - Inequality in Indonesia: What Can We Learn from Top Incomes? (with Pierre van der Eng) (2009), Journal of Public Economics - Top Incomes in W. Salverda, B. Nolan, and T. Smeeding (eds) (2009) The Oxford Handbook of Economic Inequality - Top Incomes and National Savings (with Alberto Posso) (2009) Review of Income and Wealth - Top Incomes in New Zealand : Understanding the Effects of Marginal Tax Rates, Migration Threat, and the Macroeconomy (with A.B. Atkinson) (2008) Review of Income and Wealth, 54(2): How Closely do Top Income Shares Track Other Measures of Inequality? (2007) Economic Journal, 117(524): F619 F633 - Inequality and Mortality: Long-Run Evidence from a Panel of Countries (with Christopher Jencks) (2007), Journal of Health Economics, 26(1): The Distribution of Top Incomes in Australia (with A.B. Atkinson) (2007) Economic Record, 83(262): Do Rising Top Incomes Lift All Boats? (with Dan Andrews and Christopher Jencks) (2011) B.E. Journal of Economic Analysis & Policy 11(1):Article6 - The Distribution of Top Incomes in Five Anglo-Saxon Countries over the Twentieth Century (with A.B. Atkinson)

37 Coda: Housing Wealth Inequality 37 Since most Australian wealth is in housing stock, we can ask is appreciation faster in affluent suburbs? Data: House & unit sales by postcode for Using 1993 data, assign a percentile to all postcodes. For each percentile, estimate average annual growth in prices from (dropping 1993 avoids regression to the mean). I ll show , , then full sample.

38 House price changes by initial postcode (pre-2000) Initial percentile of prices in Year T0 Data is only. Each point represents the beta coefficient from a regression of log house and unit prices on the year. 38

39 House price changes by initial postcode (2000 onwards) Initial percentile of prices in Year T0 Data is only. Each point represents the beta coefficient from a regression of log house and unit prices on the year. 39

40 House price changes by initial postcode Initial percentile of prices in Year T0 Data is for most states. Each point represents the beta coefficient from a regression of log house and unit prices on the year. 40

41 Put another way Gini coefficient of all house sales Year Houses Units Sample is , covering only ACT, NSW, Qld, SA & WA 41