PACDEV 3/2010

33
Poverty and Income Inequality in Rural Brazil: An Analysis of the Recent Decline Steven Helfand, Rudi Rocha e Henrique Vinhais Forthcoming in Portuguese in Pesquisa e Planejamento Econômico as Pobreza e Desigualdade de Renda no Brasil Rural: Uma Análise da Queda Recente” http://www.ipea.gov.br/ PACDEV 3/2010

description

- PowerPoint PPT Presentation

Transcript of PACDEV 3/2010

Page 1: PACDEV 3/2010

Poverty and Income Inequality in Rural Brazil: An Analysis of the Recent Decline

Steven Helfand, Rudi Rocha e Henrique Vinhais

Forthcoming in Portuguese in Pesquisa e Planejamento Econômico as

“Pobreza e Desigualdade de Renda no Brasil Rural: Uma Análise da Queda Recente”

http://www.ipea.gov.br/

PACDEV 3/2010

Page 2: PACDEV 3/2010

I. Introduction: The Issues

• What happened to rural poverty in the past 15 years?

• “Recent decline” in inequality in Brazil (IPEA book, 2006).

– How does rural Brazil compare to Brazil?

• How have alternative factors contributed to the decline in rural poverty?

Page 3: PACDEV 3/2010

Poverty Incidence: 1992-2005(PNAD; poverty line = 1/2 min wage of 8/2000)

0.40

0.45

0.50

0.55

0.60

0.65

0.70

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

Rural

-15pp

Page 4: PACDEV 3/2010

Gini for Rural Brazil(w/out North)

0.500.510.520.530.540.550.560.570.580.590.60

19

91

19

92

19

93

19

94

19

95

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

Page 5: PACDEV 3/2010

I. Introduction: The Questions

• Why did rural poverty decline?Decompose changes in rural poverty into growth and inequality components for 1992-98 and 1998-05.

• Why did income grow? Analyze evolution of sources of income.

• Why did inequality decline?Decompose changes in inequality into changes in shares and changes in concentration of each income source.

• Provide estimates of the magnitude of contributions to the decline in rural poverty:

Earned income, Bolsa Familia, Social security

Page 6: PACDEV 3/2010

II. Methods and Data

1) Methods

a) Decomposition of changes in poverty

b) Decomposition of changes in Gini

2) Construction of variables

3) Limitations of PNAD and related issues

Page 7: PACDEV 3/2010

Methods

a) Decomposition of changes in poverty into income and inequality components (Datt and Ravallion, 1992)

Pt = a measure of poverty

z = the poverty line

t = mean income

Lt = a vector of parameters that describe the Lorenz curve

ttt LzPP ,/

Page 8: PACDEV 3/2010

• Poverty change due to change in mean income from t to t+n, holding inequality constant

+• Poverty change due to change in inequality from

t to t+n, holding mean income constant

+• Residual

Page 9: PACDEV 3/2010

Estimate a quadratic Lorenz curve:

L = cumulative % of income

p = cumulative % of population

a, b e c parameters to be estimated

ε = random error

)()1()()1( 2 LpcpbLLpaLL

Page 10: PACDEV 3/2010

And use the estimated parameters for the simulation

m = b2-4ª

n = 2be-4c

e = -(a+b+c+1)

r = (n2-4me2)1/2

21

2

0 2)2(2

1m

zb

zbrn

mP

Page 11: PACDEV 3/2010

Methods: Decomposition of change in Gini

Gini is a measure of overall income inequality (0 => 1)

Concentration coefficient is a measure of income inequality for each source of income (-1 => +1)

Gini = sum of concentration coefficients weighted by shares.

c = concentration coef. s = income share k = income source

k

kk scGini

Page 12: PACDEV 3/2010

)( kkk

kk csscGini

)( GcsscGini kkk

kk

Decomposition of change in Gini into

-Changes in shares (s) of income sources-Changes in concentration (c) of income sources

Page 13: PACDEV 3/2010

III. Results

1) Decomposition of evolution of poverty

Page 14: PACDEV 3/2010

Table 1. Changes in Income, Poverty, and Inequality: Brazil vs. Rural Brazil

1992-98 1998-05 1992-98 1998-05

Domicile income per capitaa 350.7 456.8 453.8 30% -0.7% 145.9 183.1 201.1 26% 9.8%

Gini indexa 0.580 0.596 0.565 2.7% -5.1% 0.541 0.549 0.504 1.4% -8.1%

Headcount povertyb 0.33 0.27 0.22 -0.058 -0.048 0.62 0.56 0.46 -0.060 -0.097

Source: Author's estimates based on PNAD Microdata.

Note: (a) Domicile income per capita in R$ of 9/2005; changes are percentage changes. (b) Changes are percentage points. North excluded.

Change1992 1998 2005

Change

Brazil Rural Brazil

1992 1998 2005

Rural BR vs BR:1992-98: Similar: income dominates.1998-05: Distinct: income grows, Gini falls more, pov. falls more.

Page 15: PACDEV 3/2010

Table 3. Decomposition of Changes in Rural Poverty into Growth and Inequality Components

Estimated change in headcount poverty (percentage points)

Counterfactual 1: Change in income with

inequality constant

Counterfactual 2: Change in inequality with income constant

Residual

1992-1998 -6.97 -9.25 2.50 -0.22

Share of change attributable to 133% -36% 3%component

1998-2005 -9.32 -3.99 -5.17 -0.16

Share of change attributable to component

43% 55% 2%

But in robustness check to alternative definition of “rural” for 1998-05, the magnitudes are reversed.

Page 16: PACDEV 3/2010

III. Results

2) Why did income grow?

Page 17: PACDEV 3/2010

Table 4. Changes in Domicile Income per Capita in Rural Areas by Income Source

Change % Change Change % Change

Domicile income per capita 145.9 183.1 201.1 37.2 26% 17.97 10%

Labor income Mean 118.7 144.3 144.1 25.6 22% -0.28 0% Share of total 0.81 0.79 0.72 -0.03 -3% -0.07 -9% Contribution to change in total income (%) 69% -2%

Social security income Mean 23.2 33.3 45.7 10.1 43% 12.47 37% Share of total 0.16 0.18 0.23 0.02 14% 0.05 25% Contribution to change in total income (%) 27% 69%

Rents, remittances, gifts Mean 1.29 3.71 2.60 2.42 187% -1.12 -30% Share of total 0.01 0.02 0.01 0.01 128% -0.01 -36% Contribution to change in total income (%) 6% -6%

Other income Mean 2.68 1.81 8.71 -0.9 -32% 6.90 382% Share of total 0.02 0.01 0.04 -0.01 -46% 0.03 338% Contribution to change in total income (%) -2% 38%

Source: Author's estimates based on PNAD Microdata.

Notes: Domicile income per capita and components in R$ of 9/2005. North excluded.

1998-2005

1992 1998 2005

1992-1998

Page 18: PACDEV 3/2010

Income growth for rural Brazil:

Earned income fell as a share of total income from 81% to 72%

Social security grew steadily, and in 2005 represented 23% of rural income

-Changes in 1988 Constitution

-Women, retirement age, benefits

-Increasing real value of min. wage

Bolsa Familia grew significantly in recent years

Page 19: PACDEV 3/2010

Table 5. Percentage of People Living in Domiciles that Received Income Included in the Category "Other Income": Rural Areas by Deciles of the Income Distribution

Percentage of people in decile with

"other income"

Mean value per capita of

"other income"

Percentage of people in decile with

"other income"

Mean value per capita of

"other income"

Percentage of people in decile with

"other income"

Mean value per capita of

"other income"

Deciles

1 2.7% 3.1 3.1% 15.2 66.1% 11.4

2 2.0% 5.4 3.9% 16.9 62.2% 13.5

3 4.2% 4.8 4.5% 22.4 59.7% 14.6

4 4.0% 12.8 3.8% 24.1 48.4% 15.9

5 5.3% 5.1 3.0% 28.4 54.0% 15.8

6 5.7% 13.6 2.1% 20.3 38.1% 19.9

7 8.4% 8.1 2.6% 26.6 30.4% 22.7

8 10.8% 11.8 3.7% 30.1 26.9% 30.8

9 14.7% 21.0 5.1% 31.6 14.6% 50.8

10 21.7% 92.1 9.3% 112.9 11.8% 137.2

Source: Author's estimates based on PNAD Microdata.

Notes: Values are in R$ of 9/2005; means are based on domiciles that received "other income." North excluded.

Bottom 50% 3.6% 21.4 58.1% 14.2

Top 50% 4.6% 44.3 24.4% 52.3

1992 1998 2005

Bolsa Familia

Coverage:

Leakage?

Page 20: PACDEV 3/2010

III. Results

3) Why did inequality decline?

Page 21: PACDEV 3/2010

Table 6. Components of Domicile Income per Capita and Their Contribution to Rural Income Inequality

1992-1998 1998-2005

Gini index 0.541 0.549 0.504 1.4% -8.1%

Labor income

Concentration indexa 0.532 0.543 0.507 2.1% -6.6%

Share of component in domicile income per capita (%)b 81% 79% 72% -0.03 -0.07

Contribution to Gini index (%)b 80% 78% 72%

Social security income

Concentration indexa 0.560 0.564 0.571 0.5% 1.2%

Share of component in domicile income per capita (%)b 16% 18% 23% 0.02 0.05

Contribution to Gini index (%)b 16% 19% 26%

Rents, remittances, gifts

Concentration indexa 0.501 0.642 0.592 28.2% -7.8%

Share of component in domicile income per capita (%)b 1% 2% 1% 0.01 -0.01

Contribution to Gini index (%)b 1% 2% 2%

Other income

Concentration indexa 0.816 0.565 0.083 -30.8% -85.3%

Share of component in domicile income per capita (%)b 2% 1% 4% -0.01 0.03

Contribution to Gini index (%)b 3% 1% 1%

Source: Author's estimates based on PNAD Microdata.

Notes: (a) percentage change; (b) percentage points. North excluded.

1992 1998 2005Change

Page 22: PACDEV 3/2010

Table 7. Decomposition of Changes in the Rural Gini into Changes in Concentration and Shares

Contribution to increase in Gini

Contribution to decrease in Gini

Labor income 0.009 0.000 120% -0.027 0.000 60%

Social security income 0.001 0.000 12% 0.001 0.002 -7%

Rents, remittances, gifts 0.002 0.000 31% -0.001 -0.001 3%

Other income -0.004 -0.001 -63% -0.013 -0.007 44%

Total 0.008 0.000 -0.039 -0.005

Change in Gini

Source: Author's estimates based on PNAD Microdata.

Note: North excluded.

0.008 -0.045

1992-1998 1998-2005

kk CS kk S)GC( kk CS kk S)GC(

88% de-concentration ; 12% change in shares

Page 23: PACDEV 3/2010

What have we learned about 1998-05?

Rural poverty reduction (9.7pp)-Income growth 43%, ineq. decline 55%-Brazil: -5%, 105%

Income growth per capita (9.8%)-Earned income flat-Social security explains 69% ( 37%); BF 38% ( 382%)-Brazil: income -1%; EI -5% ; SS 18%; BF 96%

Inequality decline-60% earned income-44% BF-Social security has increased inequality in both periods-Brazil: BF only accounts for 25% of decline

Page 24: PACDEV 3/2010

Final issues:

• Robustness to change in defn. of rural?• Results robust, and contrast w/ Brazil even greater

• Regional differences matter a lot

Page 25: PACDEV 3/2010

Conclusions

• The “persistence of rural poverty” is not an accurate characterization:

-Poverty still unacceptably high (46% in 2005)-But poverty fell by 15pp 1992-05-Poverty and ineq. fell faster in rural areas

• Rural poverty reduction did not occur uniformly across regions.

Page 26: PACDEV 3/2010

Conclusions

• BF explained around 40% of poverty reduction in recent period

- Through income and inequality channels

- Expansion of coverage

• Social security explained about 25%! - Through income channel only

- Expansion of coverage

- Increase in real benefits (tied to min. wage)

• Labor income explained about 33%

• In exclusively rural areas, social security and Bolsa Familia have similar magnitudes:

BF: 37% SS: 27%

Page 27: PACDEV 3/2010

Conclusions

• Expansion of social security in 1990s and BF in 2000s are unlikely to be repeated in the future. – These programs combined to explain all of income growth

1998-05– Alternative engines of income growth must be found– Earned income still represents over 70% of income

• The Northeast– Home to 2/3 of rural poor; poverty fell least here– Earned income declined 1995-06– Requires special attention in terms of research and

policies

Page 28: PACDEV 3/2010

The End

Page 29: PACDEV 3/2010

Additional tables from paper

Page 30: PACDEV 3/2010

Table 2. Share of Total Income by Deciles

1992 1998 2005 1992-98 1998-05 1992 1998 2005 1992-98 1998-05

1 0.94 1.26 1.27 0.32 0.01 0.77 0.84 1.00 0.07 0.16

2 2.94 3.47 4.00 0.21 0.52 2.51 2.51 2.97 -0.07 0.30

3 5.89 6.46 7.24 0.04 0.25 5.15 5.03 6.41 -0.12 0.92

4 9.9 10.39 11.9 -0.11 0.76 8.6 8.6 9.5 0.07 -0.47

5 15.2 15.6 17.4 -0.04 0.26 13.3 13.1 14.8 -0.11 0.75

6 21.9 21.85 24.8 -0.47 1.13 19.4 18.3 20.5 -1.01 0.61

7 30.6 30.04 33.8 -0.54 0.78 27.5 25.9 28.4 -0.51 0.27

8 42.4 42.28 45.5 0.52 -0.53 38.1 36.2 39.3 -0.20 0.60

9 58.9 56.13 61.4 -2.73 2.04 54.3 52.5 55.1 0.03 -0.54

10 100 100 100 2.80 -5.22 100 100 100 1.85 -2.60

Source: Author's estimates based on PNAD Microdata.

Note: North excluded.

Rural Brazil Brazil

DecilesDecile share in cumulative distribution Gain/loss for each decile Decile share in cumulative distribution Gain/loss for each decile

Page 31: PACDEV 3/2010

Table 8. Robustness of Results to Alternative Definitions of "Rural" Areas: "Official Rural" vs. "Exclusively Rural"

1992-98 1998-05 1992-98 1998-05

(A) Basic indicators from Table 1 (changes in % and p.p.)

Domicile income per capitaa 26% 9.8% 20% 15.4%

Gini indexa 1.4% -8.1% -0.9% -6.1%

Headcount povertyb -0.06 -0.097 -0.05 -0.106

(B) Counterfactual contribution to the decline in poverty (Table 3)

Income growth 133% 43% 116% 60%

Change in inequality -36% 55% -20% 37%

Residual 3% 2% 4% 2%

(C) Change in mean income by component (Table 4)

Labor income 22% 0% 15% 6%

Social security income 43% 37% 41% 41%

Rents, remittances, gifts 187% -30% 195% -31%

Other income -32% 382% -33% 368%

(D) Contribution to change in Gini by component (Table 7)

Labor income 120% 60% 67% 42%

Social security income 12% -7% -21% -7%

Rents, remittances, gifts 31% 3% -58% 6%

Other income -63% 44% 111% 59%

Source: Author's estimates based on PNAD Microdata.

Notes: (a) percentage change; (b) percentage points. North excluded.

Exclusively rural areas are based on census tract codes. They exclude rural areas classified as "urban extensions" or "rural towns."

Exclusively rural areasOfficial definition

of rural areas

Page 32: PACDEV 3/2010

Figure 1. Income, Poverty, and Inequality in Rural Brazil

155

160

165

170

175

180

185

190

195

200

205

210

0.40

0.42

0.44

0.46

0.48

0.50

0.52

0.54

0.56

0.58

0.60

1995 1996 1997 1998 1999 2001 2002 2003 2004 2005

Poverty Gini Income

Income in R$, 9/2006

Source: Helfand and Del Grossi (2008).Note: Includes Tocantins, excludes the other states in the Northern Region.

Page 33: PACDEV 3/2010

Appendix 1

a coefficient on (P2-L) 0.925 0.747 0.891

b coefficient on L(P-1) -0.026 0.070 -0.340

c coefficient on (P-L) 0.103 0.187 0.140

R2 1.000 1.000 1.000

μ mean income 145.88 183.13 201.10

z poverty line 117.19 117.19 117.19

e -(a+b+c+1) -2.002 -2.004 -1.692

m b2-4a -3.701 -2.982 -3.450

n 2be-4c -0.307 -1.030 0.589

r (n2-4me2)1/2 7.710 6.997 6.313

s1 (r-n)/(2m) -1.083 -1.346 -0.830

s2 -(r+n)/(2m) 1.000 1.001 1.000

L(H) -0.5(bH+e+(mH2+nH+e2)0.5) 0.235 0.185 0.147

H FGT0 0.620 0.550 0.457

PG FGT1 0.327 0.261 0.204

SPG FGT2 0.215 0.156 0.119

Source: Author's estimates based on PNAD Microdata.

Note: North excluded.

Variable 1992 1998 2005