Income, Deprivation, and Perceptions in Latin America and the Caribbean: New Evidence from the Gallup World Poll Leonardo Gasparini* Walter Sosa Escudero** Mariana Marchionni* Sergio Olivieri* * CEDLAS / National University of La Plata. Argentina ** Universidad de San Andres. Argentina 1
Motivation and Goals Deprivation has many dimensions: income, asset, subjective. Due to lack of information, there are no studies that analyze these dimensions for all LAC consistently with the same dataset. The Gallup Poll provided a valuable source. This paper Provides new evidence on the multiple dimensions of deprivation in LAC by exploiting the Gallup Poll. Validates its use by comparing results with a large sample of national household surveys. Descriptive and exploratory. 2
Outline of the paper 1. Introduction 2. Sources of information 3. Income in the Gallup Poll 4. Income deprivation 5. Objective non-monetary deprivation 6. Subjective deprivation 7. The dimensionality of deprivation 8. Deprivation and perceptions 9. Concluding remarks 3
2. Sources of information Gallup World Poll Mainly 2006 Poll because world coverage better coverage in the Caribbean lower rate of income non-response mean incomes more consistent with household surveys. National Household Surveys for 21 LAC countries most surveys for 2006 from SEDLAC (Cedlas, WB and Mecovi) 4
Number of children - hh surveys Comparison Gallup Household Surveys/Census 2.5 =0.64 2.0 1.5 1.0 0.5 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 Number of children - Gallup Positive correlations Statistically significant, sometimes economically low Usually due to problems in some (few) countries in the Gallup Poll 5
3. Income in the Gallup Poll Question on monthly household income too simple reported in brackets of local currency units placed at the end of the questionnaire asked to a randomly selected adult in the household We compute a household per capita income variable by (1) randomly assigning a value in the bracket (2) translating this value to US$ using country exchange rates adjusted for PPP, (3) adjusting by household size=children+estimated # of adults from household surveys. 6
Too much volatility in incomes between Gallup 2006 and 2007 Mean income (in PPP US$) increases 45% in the region! Correlation coefficient across countries=0.75 2006 Gallup figures match better NA figures 7
0 0.1.2.3.4.1.2.3.4 Density of log per capita income Non parametric estimates Latin America Original data Similar means Density of log p/c income Density of log p/c income -5 0 5 10 log pc income HHS Gallup -5 0 5 10 log pc income HHS Gallup Both distributions seem to match reasonably well in the case of Latin America 8
0 0.1.2.3.4.1.2.3.4 but not in the Caribbean Original data Similar means Density of log p/c income Density of log p/c income -2 0 2 4 6 8 log pc income -2 0 2 4 6 8 log pc income HHS Gallup HHS Gallup 9
Income - national surveys Incomes in Gallup and household surveys are correlated, although not strongly 400 300 =0.64 Honduras 200 100 Venezuela 0 0 100 200 300 Income - Gallup Only in Venezuela incomes in Gallup are higher than in the household survey. 10
Mean income (Gallup) Incomes in Gallup and national accounts are correlated, although not strongly. 400 =0.75 350 300 250 200 150 100 50 0 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 GDP per capita (NA) 11
Most countries are located in similar steps in the income ladder GDP (NA) Income (Gallup) Trinidad and Tobago Chile Argentina Uruguay Chile Trinidad and Tobago Costa Rica Costa Rica Mexico Brazil Uruguay Argentina Brazil Venezuela Panama Paraguay Dominican Republic Dominican Republic Venezuela Panama Peru El Salvador Paraguay Mexico El Salvador Ecuador Ecuador Peru Guatemala Nicaragua Nicaragua Bolivia Bolivia Guatemala Haiti Haiti _Spearman=0.85 12
4. Income deprivation Measure poverty with the US$ 1 and 2 lines Apply lines from WDI and SEDLAC. 13
Most countries are located in similar steps in the poverty ladder Gallup Surveys 1 Haiti Haiti 2 El Salvador Bolivia 3 Nicaragua Nicaragua 4 Bolivia El Salvador 5 Peru Guatemala 6 Paraguay Ecuador 7 Guatemala Venezuela 8 Ecuador Peru 9 Dominican R. Paraguay 10 Mexico Mexico 11 Panama Panama 12 Venezuela Brazil 13 Brazil Dominican R. 14 Uruguay Argentina 15 Costa Rica Costa Rica 16 Argentina Uruguay 17 Chile Chile _Spearman=0.90 14
Differences in poverty between Gallup and household surveys: where do they come from? Do they arise mainly from differences in the distribution of observable income determinants (age, education, children, etc.)? Microsimulation analysis for 2007 Poverty Poverty Characteristics Constant Parameters Residual Total Diff. Gallup Hh. Surveys Effect Effect Effect Effect Chile 20.7 4.0 16.7 0.1 4.2 9.3 3.1 El Salvador 58.3 31.5 26.8 0.5-5.4 32.9-1.1 Peru 54.6 22.0 32.6-8.9 8.2 33.5-0.3 Uruguay 23.6 7.9 15.7-3.7 20.9-8.6 7.0 Venezuela 24.6 25.8-1.2-14.9 23.3-12.8 3.1 The characteristic effect is NOT the main driving force of the poverty differences between the two sources. 15
Even when incomes are poorly measured, the Gallup survey will be helpful if we can reasonably identify the income poor and the non poor. Comparison of poverty profiles Argentina Household surveys Gallup Quintile 1 Rest Quintile 1 Rest Share males 44.2 46.5 39.4 52.2 Family size 5.5 3.7 5.0 4.2 Children (<15) 2.2 0.8 2.7 1.8 Water 96.2 99.4 95.6 94.9 Employed 42.9 57.1 35.0 53.6 Chile Household surveys Gallup Quintile 1 Rest Quintile 1 Rest Poverty profiles are roughly similar for most variables in most countries. Share males 46.4 48.6 42.0 48.4 Family size 5.2 4.2 4.8 4.1 Children (<15) 1.7 0.9 1.9 1.2 Water 90.4 96.4 95.0 99.8 Employed 32.8 54.7 35.2 48.8 16
0 0 0 lat/sas.1.2.3.4.5 lat/weu.1.2.3.4.5 lat/noa.1.2.3.4 0 0 0 lat/car.1.2.3.4 lat/eap.1.2.3.4 lat/eca.1.2.3.4 The income distribution in Latin America is clearly different from the distributions in other regions The Caribbean East Asia and Pacific Eastern Europe and Central Asia Density of log p/c income Density of log p/c income Density of log p/c income 4 6 8 10 12 log pc income 4 6 8 10 12 log pc income 4 6 8 10 12 log pc income lat car lat eap lat eca South Asia Western Europe North America Density of log p/c income Density of log p/c income Density of log p/c income 4 6 8 10 12 log pc income 4 6 8 10 12 log pc income 4 6 8 10 12 log pc income lat sas lat weu lat noa 17
Western Europe North America Eastern Europe The Caribbean Eastern Asia & Pacific South Asia Latin America Is Latin America the most unequal region in the world? 50 45 40 35 30 Yes, if regional inequality is average of national inequalities 18
Western Europe North America Eastern Europe Latin America South Asia The Caribbean Eastern Asia & Pacific Is Latin America the most unequal region in the world? 60 55 50 45 40 35 30 No, if regional inequality ignores countries 19
5. Non-monetary deprivation Household consumption of some services and durable goods. Gallup 2006 (i) basic services: water, electricity, sanitation (ii) phone (fixed and cellular), TV, computer and Internet Gallup 2007: adds automobile, cable TV, washing machine, freezer and DVD player. 20
Services and durables in Gallup and household surveys Linear correlation coefficients between both information sources is positive and significant, but not large: 1. 0.64 for water 2. 0.75 for electricity 3. 0.66 for telephone 4. 0.63 for PC 21
Index of non-monetary multidimensional deprivation Key steps 1. Define the set of goods services and durables 2. Define a structure of weights factor analysis 3. Define a threshold line to match the income deprivation headcount ratio The approach identifies relative deprivation in terms of an index based on the consumption and access to some durable goods and services available in the Gallup survey. 22
Puerto Rico Chile Jamaica Uruguay Venezuela Argentina Costa Rica Ecuador Colombia Trinidad & T. Brazil Bolivia Peru Panama Dominican R. Mexico Haiti El Salvador Guatemala Cuba Honduras Paraguay Nicaragua Multidimensional deprivation in LAC 45 40 35 30 25 20 23
North America Western Europe Eastern Europe & Central Asia Latin America & The Caribbean Eastern Asia & Pacific South Asia Sub-Saharan Africa Multidimensional deprivation in the world 80 70 60 50 40 30 20 10 0 24
6. Subjective deprivation Recent literature has emphasized measures targeted directly at self perceived notions of well being. Gallup has several relevant questions: wp16: feel about life (ladder) wp30: satisfied or dissatisfied with standard of living wp40: not enough money to buy food 25
In the 0-10 ladder of subjective welfare, LAC average is 5.88. High income OECD is 7.10 and Sub-Saharan countries average is 4.24 Overall Satisfaction is 67% (OECD: 83%, Sub- Sahara: 39%). Satisfacion with food purchases is 67% (OECD: 91%, Sub-Sahara: 46% 26
In terms of general satisfaction LAC performs like a country in the upper-middle income group. In terms of food satisfaction LAC compares to a lowmiddle income country. Individuals perceive an improvement in their lives, and are optimistic for the future. 27
Satisfacion past and present Top Venezuela, Costa Rica, Puerto Rico Bottom Haiti, Peru, DR Satisfaction, future Brazil, Venezuela Paraguay, Haiti Food / Housing Chile, Argentina Haiti, Honduras, Venezuela Basic needs: The top is occupied by countries in the middle of the distribution of overall satisfaction (Argentina, Chile). Venezuela is at the top in terms of general satisfaction and at the bottom of basic needs. Haiti is at the bottom of all dimensions. 28
7. The Dimensionality of deprivation Extreme case I: a single underlying notion of welfare, all questions related to welfare are seen as proxies of this same notion. Extreme case II: welfare is a truly multidimensional concept that cannot be appropriately captured by any single notion. All questions add something new. Correlations are significant but far from high: Subjective Non-Monetary Income Subjective 1 Non-Monetary 0.486 1 Income 0.433 0.428 1 29
A Factor Analytic Approach Input: 12 welfare variables used in sections 5 to 7. Output: optimal number of orthogonal factors best representing welfare Results The welfare space can be appropriately represented by three underlying orthogonal notions: income, subjective welfare (an average of different measures), basic needs (water, electricity). Income-based assessments of welfare status are still relevant, even when subjective notions are available. The relevance of the two other factors suggests that welfare is a truly multidimensional phenomenon that cannot be fully captured by income. 30
The adequacy of poverty lines Implicit poverty line (IPL): income needed to achieve a certain level of satisfaction. Let p=g(y), p=probability of being satisfied, y=income Suppose that p* separates the satisfied from the nonsatisfied: IPL: a level of income y p such that y p = G -1 (p*) (the level of income that best separates the satisfied from the non-satisfied) G can be estimated non-parametrically 31
The IPL for food satisfaction is US$ 37 when p*=0.5 US$ 163.1 when p*= 0.659 (the unconditional proportion of satisfied individuals). The widely-used US$1-a-day line is equivalent to a monthly income of US$ 32.7. The US$1-a-day threshold would be a reasonable poverty line to measure and analyze food deprivation. 32
7. Deprivation profiles Do deprivation profiles change as we change the definition of poverty? 33
Deprivation profiles by age 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 [16,25] [26,40] [41,64] [65+] Income Subjective Income poverty is decreasing in age, while subjective poverty is increasing Important result for the debate on the measurement of old age poverty, and on targeting of social policies. 34
Difference in family size (poor non poor) 1.2 1.0 0.8 0.6 0.4 0.2 0.0 Income poverty Subjective poverty Differences in family size are much smaller when considering subjective poverty. Targeting schemes based on # children may imply significant biases when other dimensions of deprivation are considered. 35
LAT BOL VEN GTM URY PAN COL MEX NIC HND CHL CRI BRA ARG SLV PER PRY ECU Satisfaction with efforts to deal with the poor 60% 50% 40% 30% 20% 10% 0% Poor Non poor 36
Top rates of approval in Bolivia, Cuba and Venezuela Real? Propaganda? The income poor are more satisfied with social policy than the non-poor. Are governments really doing good things for the poor? Are the non-poor better informed on the weakness and failures of the social protection system? Results change when considering the subjective definition of deprivation: the poor are now less satisfied with public policies. Is it just pessimism? 37
8. Concluding remarks Income poverty analysis with Gallup data confirms results with household surveys, allow for comparable world comparisons and reveals new information for some countries (eg. Cuba). Welfare can be appropriately summarized by three dimensions: Income, an average of the subjective welfare measures, variables associated to basic needs. Welfare is a truly multidimensional phenomenon that cannot be fully captured by income. Income is an essential component of welfare. 38
The Gallup poll provides a unique opportunity to carry out a truly international analysis of social issues, in a multidimensional context However, some inconsistencies with household surveys, Census and National Accounts. Cross-country correlations are positive and significant, but sometimes too low. Still believe that the Gallup Poll is a very valuable source for international comparisons, and that future improvements in the quality of the survey in some countries and in some questions could turn it into an essential source for international research. 39