Indian income inequality dynamics, 1922-2014: From British Raj to - PowerPoint PPT Presentation

Indian income inequality dynamics, 1922-2014: From British Raj to Billionaire Raj? Lucas Chancel PARIS SCHOOL OF ECONOMICS & IDDRI Thomas Piketty EHESS & PARIS SCHOOL OF ECONOMICS World InequalityLab | December 15, 2017

This presentation 1. 1. Introduction Top 1 % income share in India : 1922 - 2014 25 2. Me 2. Methodology & & da data Combination of historical and latest tax data, household surveys 20 and national accounts in a systematic manner. % Total income 3. Re 3. Results 15 Top 1% national income share in 2014 back to its historical high (22%) in benchmark scenario. Since 1980, top 0.1% captured 10 more total growth than bottom 50% (12% vs. 11%). 4. Di 4. Discussion 5 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 Results consistent with economic policy shifts over last decades. Year Need for release of tax tabulations of 2000s. Per adult pretax national income. Systematic combination of tax, survey and national accounts data. Benchmark scenario displayed (A0B1C1D1). Source: Authors' computations using tax and survey data and national accounts. 2

Introduction 3

India stopped publishing tax data when it entered the digital age Important transformations of the Indian economy since 2000s (pursuit of § deregulation/privatization initiated in the mid-1980s). Little available data to assess the distributional impacts of growth. Some evidence of growing § inequality: Banerjee and Piketty (2005) show decreasing inequality 1940-1980 followed by an • increase, but series stop due to lack of data. NSSO consumption data suggests consumption inequality increased since 1980s, but • misreporting at the top, and no income inequality. Anand & Thampi (2017) document a sharp rise on wealth inequality since 1990. • We seek to reconstruct (cautiously and critically) income inequality series from bottom to § top. 4

Methodology & data sources 5

Public pressure led the Indian government to release recent tax data, used to update the Banerjee-Piketty series Figure 4 - Evolution of the proportion of income-tax taxpayers in India Tax da Ta data Number of taxpayers in India, 1922-2014 Tax data available from 1922-1923 to 1998- 8 99. In 2016, the government released data Share of total adult population ( % ) for recent period (2011-12 to 2013-14). 6 NB: strong increase in number of Indian tax 4 filers over recent decades, in line with evolution in France & USA during interwar period (10-15%) or post WWII (>50%) 2 (Piketty, 2001; Piketty and Saez, 2003). 0 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 Year Data from Indian Income Tax Department and WID.world population estimates. Source: Authors' computations using data from Indian Income Tax Departement and UN population data. 6

We use NSSO and IHDS surveys to reconstruct bottom incomes since 1951 Figure 2 - Top 20% consumption share from NSSO surveys NSSO co NS consumption da data Top 20% share in total consumption in India, 1951-2011 1951-2011 data, obtained directly 46 from NSSO or indirectly via the World Bank India Database (Ozler et al. 44 1996). We use Universal Reference % Total consumption Period (longest time span). 42 IHDS/IC IH ICPSR in incom ome an and co consumption 40 survey su 2005 and 2011-12 surveys include 38 income and consumption: used to 1950 1960 1970 1980 1990 2000 2010 infer income from consumption in Year Data from United Nations WIDER World Income Inequality Database and World Bank India Database. NSSO. Source: Authors’ computations using data from United Nations WIDER Income Inequality Database and World Bank India Database (based upon NSSO surveys) 7

The gap between National accounts growth and household survey growth is huge Figure 5 - Cumulated growth rates according to NAS and NSSO National ac Na accou ounts da data Total real growth rate in India, 1983-2011 Well documented NSSO / NA growth 500 mismatch (see for e.g. Deaton and Kozel, 2005). 400 Over 1983-2011: 200% growth in HH Total growth (%) consumption in NSSO, 300% in NA 300 and 480% income growth in NA. 200 à We explain a fraction of the gap 100 with top incomes, but not all of it. 0 National income (Nat. Accounts) Household consumption (Nat. Accounts) Household consumption (NSSO) Source: Authors' computations using national accounts and NSSO data. 8

Our approche in 4 steps: estimating tax and survey incomes incomes: Method similar to Banerjee Piketty St Step 1 - Es Estimate fi fiscal in § (2005) except we use Generalized Pareto Interpolation (Blanchet et al. 2016) --> more precise estimates, relaxing strict pareto assumption at the top. incomes : We observe survey consumption St Step 2 - Es Estimate su surv rvey in § distribution over time, as well as consumption-income ratios for each percentile in IHDS data. We use it to infer income from NSSO for each percentile group. For income groups with reported income < consumption, 3 alternative strategies followed. 9

Our approche in 4 steps: combining tax and survey data Step 3 - In St Inter erpolate fi fiscal in income fo for mi missing ye years: We use 2005 IHDS to compute Tax vs. survey income levels: convex profile (illustration) Tax vs. survey income levels in India : 2011 1500 percentile level growth rates between 1999 and 2005 and 2005 and 2011. 1000 Thousand INR St Step 4 - Co Combine ta tax an and su survey da data: We assume that surveys are reliable from p=0 to 500 p 1 and tax data reliable from p 2 to the top of the distribution. Assume different possible values for p 1 . p 2 given by the number of tax 0 filers. p 2 p 1 94 96 98 100 Percentile Between p 1 and p 2 : different possible Tax data Survey data profiles (linear, concave, convex) with little impact on results. 10

In total, we define 54 alternative strategies Savings profile Post-2000 survey Surveys Junction profile among the poorest data based on… representativeup betweensurvey to p1=… and tax data AverageA1 and A2 p1=80 Convex junction IHDS [A0] [C1] [D1] [B1] x 54 54 alternative x x Large negative p1=90 Linear junction = savings [A1] [C2] [D2] scenarios sc NSSO [B2] No negativesavings p1=95 Concave junction [A2] [C3] [D3] 11

Results 12

Sharp rise in top share post 1980s robust across all scenarios Top 1 % income share in India : 1922 - 2014 30 25 % Total income 20 15 10 5 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 Year Per adult pretax national income. Systematic combination of tax, survey and national accounts data. All scenarios displayed. Thick red line corresponds to the benchmark scenario (A0B1C1D1). Source: Authors' computations using tax and survey data and national accounts. 13

Decrease in bottom share robust across all scenarios Appendix 15 - Bottom 50% income shares across 54 scenarios Bottom 50 % income share in India : 1951 - 2014 30 25 % Total income 20 15 10 1940 1960 1980 2000 2020 Year Per adult pretax national income. All scenarios displayed. Thick red line corresponds to the benchmark (A0B1C1D1). 14

The top 10% and the middle 40% inverted their relative positions Top 10 % vs. Middle 40 % income shares in India 1951-2014 55 50 % Total income 45 40 35 30 1950 1960 1970 1980 1990 2000 2010 Year topshare_A0B1C1D1 middle40_A0B1C1D1 Per adult pretax national income. Systematic combination of tax, survey and national accounts data. Benchmark scenario displayed (A0B1C1D1). Source: Authors' computations using tax and survey data and national accounts. 15