A Measured Approach to Ending Poverty and Boosting Shared Prosperity Concepts, Data, and the Twin Goals Dean Jolliffe, Peter Lanjouw; Shaohua Chen, Aart Kraay, Christian Meyer, Mario Negre, Espen Prydz, Renos Vakis, and Kyla Wethli Policy Research Report Presentation Tokyo, December 8, 2014
Background: The World Bank Twin Goals Ending extreme poverty by 2030 (< 3% of global pop. below $1.25 a day) Boosting shared prosperity (Growth of incomes of bottom 40% of population in every country) PRR focuses on concepts, empirical evidence, strengths & weaknesses, and measurement related to the new twin goals 2
Overview of PRR structure Part I: Defining and assessing the goals > Ending Poverty by 2030 > Understanding Shared Prosperity Part II: The twin goals in a broader context > Alternative Notions of Poverty and Shared Prosperity > Challenges Posed by Uncertainty Part III: Data and measurement challenges > Tracking Poverty and Shared Prosperity, Nationally > Tracking Poverty and Shared Prosperity, Globally DEC Policy Research Talk
Measuring Global Poverty: The Basic Approach Builds on established practice of the past 25 years starting with 1990 WDR on Poverty 3 key ingredients Indicator of economic wellbeing Selection of poverty line Aggregation to single summary index Constructing a global database Assembling and cleaning country-level household survey data PovcalNet + Global Poverty Working Group DAC Development Debate 4
6 key takeaways from the report Assessing and decoding the goals Reaching the poverty goal is ambitious: Business as usual will not get us there by 2030 Shared prosperity is one articulation of inclusive growth with historical antecedents in the literature Synergies to twinning the goals Progress in boosting shared prosperity is critical to eliminate extreme poverty. Data challenges Progress is made at the country level Policies need to be informed by highquality household surveys. Research in data collection methods improves data quality and comparability. More frequent data is not the right call to action PRR calls for greater emphasis on data quality and data systems. 5
Global Poverty in 2011: Headcount and Number of Poor By Developing Region East Asia & Pacific 161 million South Asia 399 million 415 million Sub-Saharan Africa Middle East & North Africa 2 28 Europe & Central Asia 6 Latin America & the Caribbean Developing World 17.0% World 14.5% 1,010.7 million 1,010.7 million Headcount at $1.25 a day (2005 PPP), percent 6
3% by 2030 is far from assured: Business as usual will not get us there Global Poverty in 2030 at $1.25 per day (2005 PPP), assuming unchanged inequality Scenario Headcount (percent) Number of poor (million) Average income growth of 4% p.a. in each country 3 252 Each country sustains avg per capita growth during past 20 years 6.8 573 Each country sustains avg per capita growth during past 10 years 4.8 405.4 Each country sustains avg per capita growth during past 10 years (survey-based growth) 6.7 564.8 7
Forecasts of future growth and thus poverty reduction are highly uncertain Approach 1: Projections based on actual past average growth rates 25 Approach 2: Probabilistic scenarios based on random draws from past variation in growth rates between 2001 and 2011 25 20 Global Poverty Headcount Ratio (%) 15 10 5 3% target 1981-1991 1991-2001 2001-2011 Global Poverty Headcount Ratio (%) 20 15 10 5 3% target median 5 th to 95 th percentile 1 st to 99 th percentile Global headcount: 7.1% 0 0 3.8% Years 8
Sources of uncertainty about progress towards the twin goals Economic and financial crises, food price shocks Unrealistic to postulate stable growth for all countries Crises can affect the sustainability of programs that assist the poor Climate change and extreme weather patterns Effect on global poverty up to 2030 may be muted Key impact may be on sustainability of progress beyond 2030 State fragility, political, social, and armed conflict Up to 1/3 of the world s poor live in FCS Complicated by link with climate change Global disease risk (pandemics) Pandemics have generated episodes of profound disruption Globalization can hasten the spread of pathogens DAC Development Debate 9
What does past country experience suggest about the likely pace of poverty reduction in the future? In countries where poverty has ended poverty decline did not always slow DAC Development Debate 10
6 key takeaways from the report Assessing and decoding the goals Reaching the poverty goal is ambitious: Business as usual will not get us there by 2030 Shared prosperity is one articulation of inclusive growth with historical antecedents in the literature Synergies to twinning the goals Progress in boosting shared prosperity is critical to eliminate extreme poverty. Data challenges Progress is made at the country level Policies need to be informed by highquality household surveys. Research in data collection methods improves data quality and comparability. More frequent data is not the right call to action PRR calls for greater emphasis on data quality and data systems. 11
How do we define Economic Development? Historically: Growth in average income Usually defined in terms of GDP per capita. This is far from egalitarian (Chenery et al, 1974) Average income assigns greater weight to those in richer percentiles of the income distribution. Weight on Percentile j in Social Welfare Function PRR Shared Prosperity 0 0.2 0.4 0.6 0.8 121 Percentiles of income or consumption distribution
Beyond the poverty line: Classes of functions that do not distinguish between poor and non-poor Weight on Percentile j in Social Welfare Function Atkinson 1 αα ww yy pp = yy (pp) Higher αα higher inequality aversion αα = 22 αα = 00 (mean income) αα = 11 Weight on Percentile j in Social Welfare Function Bottom 40 percent ww yy pp = II pp<0.4 yy (pp) Sen real national income w y p = 1 gg y p 0 0.2 0.4 0.6 0.8 1 Percentiles of income or consumption distribution 0 0.2 0.4 0.6 0.8 1 Percentiles of income or consumption distribution Weights normalized to sum to one; drawn for hypothetical lognormal income distribution (mean $2000, Gini.30). 13
Growth and social welfare: Decomposing growth in social welfare into (a) change in average incomes, and (b) change in inequality Atkinson (1) Sen index Bottom 40% Poverty headcount DAC Development Debate 14
Shared Prosperity: towards inclusive growth Boosting the per capita income or consumption growth of the poorest 40 percent in a given country Simplicity Closely linked to average income per capita Focus 40% cut-off is arbitrary but with some precedence Robert S. McNamara: 1972 Annual meetings the poorest 40 percent of the citizenry is of immense urgency since their condition is in fact far worse than national averages suggest. [ ] PRR Shared Prosperity DECRG Departmental Review Meeting 15
Comparing the bottom 40% to the national average This provides a means to assess changes in inequality of outcomes even though shared prosperity goal is not in and of itself an inequality goal. South Africa Uganda Annual growth (1993 = 100) 90 110 130 150 170 190 Bottom 40 mean Population mean Annual growth (1989 = 100) 90 110 130 150 170 190 Bottom 40 mean Population mean 1980 1990 2000 2010 Year 1980 1990 2000 2010 Year DEC Policy Research Report 16
Shared Prosperity: Country-level indicator Mean bottom 40% income distribution United States Average US household in bottom 40% would be in the richest 10% in Brazil. Mean bottom 40% Mean bottom 40% income distribution Brazil consumption distribution India Average Brazilian household in bottom 40% would be at about the 90 th percentile in India. DEC Policy Research Report Source: Lakner and Milanovic (2013) 17
6 key takeaways from the report Assessing and decoding the goals Reaching the poverty goal is ambitious: Business as usual will not get us there by 2030 Shared prosperity is one articulation of inclusive growth with historical antecedents in the literature Synergies to twinning the goals Progress in boosting shared prosperity is critical to eliminate extreme poverty. Data challenges Progress is made at the country level Policies need to be informed by highquality household surveys. Research in data collection methods improves data quality and comparability. More frequent data is not the right call to action PRR calls for greater emphasis on data quality and data systems. 18
Twinning the goals: Three scenarios for global poverty under differential shared prosperity 30 Historic 2002-2011 Simulations 2012-2030 25 Global poverty rate (<$1.25/day) 20 15 10 5 3% goal for 2030 0 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 Source: Lakner, C., Negre, M., & Prydz, E. B. (2014). Twinning the goals: how can promoting shared prosperity help to reduce global poverty?. World Bank Policy Research Working Paper, (7106). 19
6 key takeaways from the report Assessing and decoding the goals Reaching the poverty goal is ambitious: Business as usual will not get us there by 2030 Shared prosperity is one articulation of inclusive growth with historical antecedents in the literature Synergies to twinning the goals Progress in boosting shared prosperity is critical to eliminate extreme poverty. Data challenges Research important both for policy & for improving data collection methods. The importance of country level policy and poverty profiles. Progress is made at the country level Policies need to be informed by highquality household surveys. More frequent data is not the right call to action PRR calls for greater emphasis on data quality and data systems. 20
Spatial price differences within countries matter for sub-national poverty profiles Non-metro poverty less metro poverty in the US In the US, official poverty estimates do not account for cost of living differences Official poverty estimates Adjusted poverty estimates In official estimates, poverty in non-metropolitan areas is higher than in metropolitan areas. Once adjusted for cost-ofliving differences, poverty in non-metro areas is 15% lower than in metro areas (Jolliffe, 2004) DAC Development Debate 21
Research in data collection methods Questionnaire changes => changes in cons., poverty, & shared prosperity Diary: HH, infreq. Diary: HH, freq. Recall: Usual, 12 month Recall: Collapse, 7 day Recall: Subset, 7 day Recall: Long, 7 day Recall: Long, 14 day change from benchmark personal diary -0.136*** -0.173*** -0.207*** -0.283*** -0.071* -0.039-0.161*** Household consumption surveys vary widely (over time & countries) Beegle et al. (2012) provide experimental evidence on the effect Exact same instrument except increase recall period => 12% drop in avg consumption => 8 point (%) increase in poverty Same recall period, but long list collapsed to comprehensive groups => 24% drop in avg consumption => 32% drop in shared prosperity Research can inform questionnaire design & provide bridges across otherwise noncomparable data Source: Adapted from Beegle et al. (2012). 22
6 key takeaways from the report Assessing and decoding the goals Reaching the poverty goal is ambitious: Business as usual will not get us there by 2030 Shared prosperity is one articulation of inclusive growth with historical antecedents in the literature Synergies to twinning the goals Progress in boosting shared prosperity is critical to eliminate extreme poverty. Data challenges Research in data collection methods improves data quality and comparability. Progress is made at the country level Policies need to be informed by highquality household surveys. More frequent data is not the right call to action PRR calls for greater emphasis on data quality and data systems. 23
Poverty and Prosperity Policies, made at the country level, need highquality data Vietnam poverty maps Pockets of poverty & dynamics of Poverty Profiles require research & country context & data 1999 2009 24
6 key takeaways from the report Assessing and decoding the goals Reaching the poverty goal is ambitious: Business as usual will not get us there by 2030 Shared prosperity is one articulation of inclusive growth with historical antecedents in the literature Synergies to twinning the goals Progress in boosting shared prosperity is critical to eliminate extreme poverty. Data challenges Progress is made at the country level Policies need to be informed by highquality household surveys. Research in data collection methods improves data quality and comparability. More frequent data is not the right call to action PRR calls for greater emphasis on data quality and data systems. 25
Complementary data needed to estimate poverty and shared prosperity Household surveys are a necessary input to measuring global poverty and shared prosperity, but they are not sufficient. Several complementary data sources are also needed Population (census) data Purchasing power parity (PPP) indices Inflation and national accounts growth to estimate total number of the poor (as product of poverty rate and population) population frame for survey samples Make poverty line comparable across countries Inflation data to keep measures of wellbeing in real terms NA data to line up surveys into reference years 26
Error in population counts => error in poverty estimates Bangladesh, 2005 to 2015 UN WPP 2008 Rev. / WDI 2011 and earlier Poverty rate in 2010 43% 71m poor 65m poor 64m poor UN WPP 2012 Rev. / WDI 2014 and later UN WPP 2010 Rev. / WDI 2012 and 2013 US National Research Council: 4.8% average absolute error in UN/WB 5-year projections => 50 million mis-identified UN World Population Prospects estimates serve as inputs to WDI and for poverty estimates. One example, Bangladesh: Census in 2011 UN WPP pre-censal estimates significantly higher than post-censal => 6-7 million fewer poor. Good population counts relies on good data systems (vital statistics systems, construction/housing, fertility, mortality, etc.) 27
New PPPs can have substantial implications for overall level & regional profile of global poverty Regional poverty headcount 1993 poverty headcount based on three PPP Indices 60% 50% 40% 30% 20% 10% 0% 1.35 bn total 1.3 bn total 1.8 bn total 50% 51% 47% 43% 42% 39% 26% 25% 24% 15% 10% 57% 1985 ICP 1993 ICP 2005 ICP $1.01 a day $1.08 a day $1.25 a day 2005 ICP PPP: the developing world is poorer than we thought Release of 2011 ICP brings same challenges Our view: additional research will be necessary before international poverty rates can be estimated using the ICP PPPs (International Comparison Program, 2014). East Asia and the Pacific South Asia Latin America and the Caribbean Sub-Saharan Africa Source: Deaton (2010) 28
Projection errors in poverty estimates due to divergence between national accounts and surveys: Let s pretend we didn t have India s 2009/2010 survey Income distribution (density estimate) Poverty headcount, percent (at $1.25 a day, 2005 PPP) 04/05 actual 04/05 scaled to 09/10 using survey means 41.6 actual 04/05 scaled to 09/10 using survey means 04/05 scaled to 09/10 using national accounts 04/05 scaled to 09/10 using national accounts 222 m people 09/10 actual Daily consumption per capita (2005 PPP) Critical to better understand divergence between NA and surveys Sources: Povcal, India NSS, WDI, Ravallion (2008) 29
Concluding observations and looking forward There is a merit to twinning the goals While shared prosperity indicator is not in and of itself an inequality measure, it does open avenues to broadening the discussion to include inequality of outcomes Introduction of shared prosperity reasserts need to focus on types of growth Quality of data needs as much attention as frequency Data systems architectures at the country level are needed not only to support credible measurement of twin goals, but also for effective national development policy DAC Development Debate 30
Thank you www.worldbank.org/ameasuredapproach A Measured Approach to Ending Poverty and Boosting Shared Prosperity Concepts, Data, and the Twin Goals World Bank Policy Research Report