Transcript China`s (Uneven) Progress Against Poverty

China’s (Uneven) Progress
Against Poverty
Martin Ravallion and Shaohua Chen
Development Research Group, World Bank
1
Questions
• How much progress has China made against absolute
poverty?
• When and where was the greatest progress made?
• What happened to inequality? A poverty-inequality
trade off?
• What were the proximate causes of uneven progress
over time and across provinces? What role was
played by public policies?
• What lessons does China’s past success against
poverty hold for China in the future and for the rest of
the developing world?
2
Data
Five findings
Five lessons
3
Data
4
Distributional data for China
• Newly constructed poverty lines
– Old lines seen as out of date: “too low” + no allowance for
geographic COL differences
– New lines: 850 Yuan per year for rural areas and 1200 Yuan for
urban areas, both in 2002 prices; also province-specific lines
• Newly assembled distributional data
– much of which has not previously been analyzed
– Rural Household Surveys (from 1980) and Urban Household
Surveys (1981) of National Bureau of Statistics
– Early surveys small for 30% of provinces, but no sign of bias
– Time series of tabulated distributions (micro data not available)
• Incomplete data at provincial level
– though we can still provide estimates of the trends.
5
New poverty lines
• Region-specific food bundles for urban and rural
areas, valued at median unit values by province.
• Food bundles based on the actual consumption of
those between the 15th and 25th percentile nationally.
• These bundles are then scaled to reach 2100 calories
per person per day, with 75% of the calories from
foodgrains.
• Allowances for non-food consumption are based on
the nonfood spending of households in a
neighborhood of the point at which total spending
equaled the food poverty line in each province (and
separately for urban and rural areas).
6
Deflators over time
• Urban and rural CPI
• Urban inflation rate higher than rural, esp., in the
1990s (higher costs of previously subsidized goods)
240
CPI (100 in 1990)
Urban
200
Rural
160
120
80
40
1980
1985
1990
1995
2000
7
Rising urban-rural COL differential
Urban-rural cost-of-living differential (%)
44
40
36
32
28
24
20
16
1980
1985
1990
1995
2000
8
Corrections for 1990 change in
valuation method in RHS
• 1990 change in valuation methods for imputing
income from consumption of own-farm output
• Distributions by both methods for 1990 are used to
correct the data for the late 1980s
Y (new) / Y (old )  1.19272  0.20915 p  0.23457 p 2  0.12562 p 3  ˆ
( 5421.5)
( 111.8)
( 54.5)
( 44.9 )
R2=0.99959
9
Corrections for 1990 change in
valuation method in RHS
• 1990 change in valuation methods for imputing
income from consumption of own-farm output
• Distributions by both methods for 1990 are used to
correct the data for the late 1980s
Alternative
estimates
for 1990
Old method
New method:
1. Actual
2. Estimated using
our correction model
Mean income
(Yuan per
person)
629.70
Gini
index
(%)
31.53
Headcount
index (%)
686.30
29.87
29.93
688.05
30.05
29.86
37.63
10
Poverty measures
• Headcount index (H): % living in households with
income per person below the poverty line.
• Poverty gap index (PG): mean distance below the
poverty line as a proportion of the poverty line
• Squared poverty gap index (SPG): poverty gaps are
weighted by the gaps themselves, so as to reflect
inequality amongst the poor (Foster et al., 1984).
• Parameterized Lorenz Curves
– alternative functional forms (Beta+general elliptical)
– checks for theoretical consistency and accuracy
11
Inequality measures
• Relative Gini index based on sum of income
differences normalized by the mean for that
distribution
• Absolute Gini index based on sum of income
differences normalized by a fixed mean
12
Persistent data problems
• Sample frame based on registration system =>
underestimation of urban poverty
• Survey compliance problems, esp., urban areas
• Single price indices, independent of level of income
13
Five findings
14
1. Huge overall progress against poverty, but
uneven progress
2. Rising inequality, though more so in some
periods and places
3. The pattern of growth matters to both poverty
and inequality in China
4. No sign of an aggregate growth-equity trade
off
5. Poverty would have fallen much faster
without rising inequality
15
Finding 1: Huge overall progress
against poverty, but uneven progress
• In the 20 year period after 1981, the proportion living
below our new poverty lines fell from 53% to 8%.
( 62%+ in 1980.)
• Half of the decline in poverty came in 1981-84.
• However, there were many setbacks for the poor.
– Poverty rose in the late 1980s and stalled in early 1990s,
– recovered pace in the mid-1990s,
– but stalled again in the late 1990s.
16
Headcount index, 1981-2001
60
Headcount index (%)
50
Upper line
40
30
20
Lower
line
10
0
1980
1985
1990
1995
2000
17
Headcount index for “$1/day”, 1981-2001
70
60
China
50
40
Developing world
less China
30
20
10
1980
East Asia
less China
1985
1990
1995
2000
18
Effect on headcount index of our correction for
the change in valuation methods
Headcount index for rural areas
80
70
Upper
line
60
50
Old valuation
method
40
30
20
10
Lower
line
0
1980
1985
1990
1995
2000
19
Trend rates of change in rural headcount index
(upper line; by province; %/year; 1983-2001)
log X it   iX   iX t  itX
7
6
5
4
3
2
1
0
-30
-20
-10
0
20
Trend rates of change in rural headcount index
(upper line; by province; %/year; 1983-2001)
log X it   iX   iX t  itX
7
6
5
4
3
Fujian,
Jiangsu
2
1
Beijing
Guangdong
0
-30
-20
-10
0
21
Finding 2: Rising inequality
But not continuously and more so in some
periods and some provinces
• Relative inequality is higher in rural than urban areas
– in marked contrast to most developing countries.
• Though steeper increase in urban inequality.
• Relative inequality between urban and rural areas has
not shown a rising trend once one allows for the
higher rate of increase in the urban cost-of-living.
• Absolute inequality has increased appreciably
– between and within both urban and rural areas,
– and absolute inequality is higher in urban areas.
22
Relative inequality between urban and rural areas
Ratio of urban to rural mean income
2.8
W ithout adjustment for
urban-rural COL differential
2.4
2.0
1.6
W ith adjustment for COL
1.2
1980
1985
1990
1995
2000
23
Absolute inequality between urban and rural areas
2.4
Difference between urban and rural mean
(divided by 1990 national mean)
2.0
1.6
W ithout COL adjustment
1.2
0.8
W ith COL adjustment
0.4
0.0
1980
1985
1990
1995
2000
24
Relative inequality in rural and urban areas
and nationally
40
Gini index (%)
National
35
30
Rural
25
Urban
20
15
10
1980
1985
1990
1995
2000
25
Absolute inequality in rural and urban areas
and nationally
Absolute Gini index (relative to 1990 mean)
140
Urban
120
100
National
80
60
Rural
40
20
0
1980
1985
1990
1995
2000
26
Effect on Gini index and mean of our correction
for the change in valuation methods
Mean income in rural areas
(Yuan/person/year; 1980 prices)
36
Gini index
(right axis)
700
Old valuation method
(broken lines)
600
32
500
28
400
Mean
(left axis)
300
24
Gini index of income inequality (%)
40
200
100
1980
1985
1990
1995
2000
27
Finding 3: The pattern of growth matters
• Economic growth was clearly a key proximate cause
of poverty reduction
 ln Pt   0   1 ln Yt   t
• Growth elasticity of poverty reduction
= – 3.2 (t= – 8.7) (using survey means)
– 2.6 (t= – 2.2) (using GDP per capita)
28
The sectoral pattern of growth matters
• The gains to the poor from aggregate economic
growth depended on its sectoral composition.
• Decomposition of change in poverty:
r
r
r
u
u
u
u
r
u
u
P01  P81  [ n01
( P01
 P81
)  n01
( P01
 P81
)]  [( P81
 P81
)( n01
 n81
)]
Within-sector effect
Population shift effect
– Within-sector effect is the change in poverty measures over
time weighted by final year population shares
– Population shift effect measures the partial contribution of
urbanization over time, weighted by the initial urban-rural
difference in poverty measures. (Kuznets process of migration.)
29
Decomposition of the change in poverty
Migration to urban areas helped, but the bulk of the
reduction in poverty came from within rural areas
Poverty measures
(% point change 1981-2001)
Within rural
Within urban
Population shift
Total change
H
-32.53
PG
-10.39
SPG
-4.51
(72.5)
(74.0)
(75.0)
-2.08
-0.32
-0.09
(4.6)
(2.3)
(1.5)
-10.27
-3.32
-1.42
(22.9)
(23.7)
(23.6)
-44.87
-14.04
-6.01
Note: % of total in parentheses.
Note: Quite rapid urbanization despite restrictions on migration
• Urban share of 19% in 1980; rose to 39% in 2002
30
Regression decomposition for mean income
growth
r r
u u


n


n
• Mean income: t t t
t t
• Growth rate:
 ln t  str  ln tr  stu  ln tu  [ str  stu ( ntr / ntu )] ln ntr
sti  nti  ti /  t
• Test equation:
 ln Pt  0   r str  ln tr   u stu  ln tu

n
( str

r
u nt
st . u
nt
)  ln ntr   t
• Null hypothesis:
H0: i   for i=r,u,n
31
 ln Pt  0   r str  ln tr   u stu  ln tu
r
n
  n ( str  stu . tu )  ln ntr   t
nt
Constant
Growth rate of mean
rural income (shareweighted) ( r )
Growth rate of mean
urban income (shareweighted) ( u )
Population shift
effect ( n )
R2
D-W
Headcount
index
0.033
(0.808)
-2.563
(-8.432)
Poverty gap
index
0.040
(0.690)
-3.341
(-7.768)
Squared poverty
gap index
0.039
(0.510)
-3.722
(-6.637)
0.092
(0.201)
0.519
(0.797)
0.744
(0.877)
0.735
(0.159)
0.823
2.671
2.189
(0.335)
0.796
2.653
3.941
(0.462)
0.739
2.661
32
 ln Pt  0   r str  ln tr   u stu  ln tu
r
n
  n ( str  stu . tu )  ln ntr   t
nt
Constant
Growth rate of mean
rural income (shareweighted) ( r )
Growth rate of mean
urban income (shareweighted) ( u )
Population shift
effect ( n )
R2
D-W
Headcount
index
0.033
(0.808)
-2.563
(-8.432)
Poverty gap
index
0.040
(0.690)
-3.341
(-7.768)
Squared poverty
gap index
0.039
(0.510)
-3.722
(-6.637)
0.092
(0.201)
0.519
(0.797)
0.744
(0.877)
0.735
(0.159)
0.823
2.671
2.189
(0.335)
0.796
2.653
3.941
(0.462)
0.739
2.661
33
Decomposing GDP growth
• Standard classification of its origins, namely
– “primary” (mainly agriculture),
– “secondary” (manufacturing and construction) and
– “tertiary” (services and trade).
• The primary sector’s share fell from 30% in 1980 to
15% in 2001, though not montonically.
• Almost all of this decline was made up for by an
increase in the tertiary-sector share.
34
Shares of GDP by sector
60
Share of GDP
Secondary
50
40
Tertiary
30
20
Primary
10
1980
1985
1990
1995
2000
35
Regression decomposition for sectoral
decomposition
• Test equation:
n
 ln Pt   0    i sit  ln Yit   t
i 1
• Null hypothesis:
H0: i   for i = 1,..n
36
n
 ln Pt   0    i sit  ln Yit   t
i 1
Headcount index (log difference)
Constant
0.116
0.163
(1.059)
(1.656)
Growth rate of
-2.595
GDP per capita
(-2.162)
-8.067
Primary ( 1 )
(-3.969)
-1.751
Secondary ( 2 )
(-1.214)
-3.082
Tertiary ( 3 )
(-1.239)
Secondary+
Tertiary
R2
0.207
0.431
D-W
1.553
1.725
1   2
-6.317
(-3.231)
1.331
2 3
(0.405)
0.155
(1.761)
-7.852
(-4.092)
-2.245
(-2.199)
0.423
1.768
-5.607
(-3.14)
37
n
 ln Pt   0    i sit  ln Yit   t
i 1
Headcount index (log difference)
Constant
0.116
0.163
(1.059)
(1.656)
Growth rate of
-2.595
GDP per capita
(-2.162)
-8.067
Primary ( 1 )
(-3.969)
-1.751
Secondary ( 2 )
(-1.214)
-3.082
Tertiary ( 3 )
(-1.239)
Secondary+
Tertiary
R2
0.207
0.431
D-W
1.553
1.725
1   2
-6.317
(-3.231)
1.331
2 3
(0.405)
0.155
(1.761)
-7.852
(-4.092)
-2.245
(-2.199)
0.423
1.768
-5.607
(-3.14)
38
Primary sector was the main engine of
poverty reduction
• Growth in the primary sector (primarily agriculture)
did more to reduce poverty than either the secondary
or tertiary sectors.
• Starting in 1981, if the same aggregate growth rate
had been balanced across sectors then it would have
taken 10 years to bring the national poverty rate down
to 8%, rather than 20 years.
• But could a more equitable growth process have
allowed the same rate of growth?
39
Province level
• Complete series of mean income from 1980
• But less complete distributional data; 11-12 years
• Marked differences in initial conditions; Gini index
around mid-1980s varied from 18% to 33%.
• OLS estimates of province specific trends:
log X it   iX   iX t  itX
40
Trend in rural headcount index (% per year)
Provinces with higher growth rates in rural
mean income saw faster poverty reduction
5
Beijing
Shanghai
Tianjin
0
-5
-10
-15
-20
-25
-30
0
1
2
3
4
5
6
7
Trend in mean income (% per year)
41
Trend in rural headcount index (% per year)
Provinces with higher growth rates in rural
mean income saw faster poverty reduction
5
Beijing
Shanghai
Reliability?
H<2%
Tianjin
0
-5
-10
-15
-20
-25
-30
0
1
2
3
4
5
6
7
Trend in mean income (% per year)
42
Trend in rural headcount index (% per year)
Provinces with higher growth rates in rural
mean income saw faster poverty reduction
5
Beijing
Shanghai
Tianjin
0
Elasticity = -2.4 (t = -4.3)
-5
(dropping Beijing, Shanghai, Tianjin)
-10
-15
-20
-25
-30
0
1
2
3
4
5
6
7
Trend in mean income (% per year)
43
Wide variation in growth elasticities of
poverty reduction
• 95% CI for the impact of a 3% growth rate on H is
(0%, 9%)
• Dropping Beijing, Shanghai and Tianjin the 95% CI
for 3% growth rate is (4%, 10%)
• Growth elasticity calculated as ratio of trend in H to
trend in mean varies from –6.6 ro 1.0 (mean=-2.3)
• Geographic composition of growth mattered to
aggregate rate of poverty reduction….
44
Trend rate of growth in mean rural income (%/year)
Growth did not occur where it would
have most impact on poverty
7
6
5
4
3
Henan
2
1
0
-.8
-.7
-.6
-.5
-.4
-.3
-.2
-.1
.0
.1
Share weighted total elasticity of the headcount index to growth
45
Inequality and the pattern of growth
• The composition of growth also mattered to the
evolution of aggregate inequality.
• Agricultural growth was inequality decreasing.
46
Inequality and GDP growth by origin
Constant
Growth rate of
GDP per capita
Primary ( 1 )
1
-0.072
(0.429)
0.012
(0.544)
Secondary ( 2 )
Tertiary ( 3 )
R2
D-W
1   2
2 3
0.018
2.112
2
0.038
(1.278)
3
0.038
(3.598)
-1.798
(2.244)
0.170
(0.432)
-0.218
(-0.272)
0.326
2.112
-1.968
(2.263)
0.388
(0.381)
-1.755
(2.819)
0.316
2.202
Note: The dependent variable is the first difference over time in the log of the Gini
47
Inequality and GDP growth by origin
Constant
Growth rate of
GDP per capita
Primary ( 1 )
1
-0.072
(0.429)
0.012
(0.544)
Secondary ( 2 )
Tertiary ( 3 )
R2
D-W
1   2
2 3
0.018
2.112
2
0.038
(1.278)
3
0.038
(3.598)
-1.798
(2.244)
0.170
(0.432)
-0.218
(-0.272)
0.326
2.112
-1.968
(2.263)
0.388
(0.381)
-1.755
(2.819)
0.316
2.202
Note: The dependent variable is the first difference over time in the log of the Gini
48
Inequality and growth in mean
urban and rural incomes
Constant
Growth rate in mean
rural income
Growth rate in mean
rural income lagged
Growth rate in mean
urban income
R2
D-W
Rural
0.013
(0.880)
-0.476
(-3.206)
0.510
(4.322)
0.075
(0.830)
0.491
Urban
0.006
(0.386)
-1.430
(-5.808)
1.014
(4.635)
0.687
(3.305)
0.690
1.741
Rural economic growth reduced inequality within both
urban and rural areas, as well as between them 49
Inequality and growth in mean
urban and rural incomes
Constant
Growth rate in mean
rural income
Growth rate in mean
rural income lagged
Growth rate in mean
urban income
R2
D-W
Rural
0.013
(0.880)
-0.476
(-3.206)
0.510
(4.322)
0.075
(0.830)
0.491
Urban
0.006
(0.386)
-1.430
(-5.808)
1.014
(4.635)
0.687
(3.305)
0.690
1.741
Rural economic growth reduced inequality within both
urban and rural areas, as well as between them 50
Finding 4: No sign of an aggregate
growth-equity trade off
• The strong positive correlation over time between
China’s GDP per capita and inequality is driven by
common time trends.
• Near zero correlation between changes in (log) Gini
and growth rate.
• The periods of more rapid growth did not bring more
rapid increases in inequality. Indeed,…
51
The periods of falling inequality had highest
growth in mean household income
Annualized log difference (%/year)
Inequality
1.
2.
3.
4.
1981-85
1986-94
1995-98
1999-2001
Falling
Rising
Falling
Rising
Gini
index
-1.12
2.81
-0.81
2.71
Mean
household
income
8.87
3.10
5.35
4.47
GDP per
capita
8.80
7.99
7.75
6.61
52
The periods of falling inequality had highest
growth in mean household income
Annualized log difference (%/year)
Inequality
1.
2.
3.
4.
1981-85
1986-94
1995-98
1999-2001
Falling
Rising
Falling
Rising
Gini
index
-1.12
2.81
-0.81
2.71
Mean
household
income
8.87
3.10
5.35
4.47
GDP per
capita
8.80
7.99
7.75
6.61
53
Provinces with higher growth did not have
steeper rises in inequality
log X it   iX   iX t  itX
Trend in rural Gini index (% per year)
3.2
2.8
2.4
2.0
1.6
1.2
0.8
0.4
r = -0.18
0.0
-0.4
0
1
2
3
4
5
6
7
Trend growth rate in mean rural income (% per year)
54
Double handicap in more unequal
provinces
More unequal provinces faced two handicaps in
rural poverty reduction:
1.
High inequality provinces had a lower growth
elasticity of poverty reduction:
 iH /  iY  ( 5.935 0.0136 y80R i )(1  G83R i )  1.365  iG  ˆt
( 4.487 )
( 2.560 )
( 2.392 )
R2=0.386; n=29
2.
High inequality provinces had lower growth:
signs of “inefficient inequality” both within rural areas,
and between urban and rural areas =>
55
Regressions for provincial trends in poverty
and mean incomes
Initial conditions (mean and distribution) + location
R
 iH   67.877  0.141Y80i  0.463 G83
i  6.797 URi
( 6.239)
( 8.090)
( 3.313)
( 3.201)
 9.291 COASTi  25.012 GDONGi  ˆt
( 5.292)
( 15.160)
R2=0.827
R
 iY  14.143 0.007 Y80i  0.149 G83
i  1.632 URi
( 3.759 )
( 1.294 )
( 2.526 )
( 2.682 )
 0.507 COASTi  1.290 GDONGi  ˆt
( 0.913)
(1.875 )
R2=0.423
56
Initially poorer and less unequal provinces
had higher rates of poverty reduction
• Large effects; going from the lowest initial inequality to
the highest inequality cuts 7% points off the annual rate of
poverty reduction.
• Initial distribution matters independently of growth; both
inequality measures remain significant (though with smaller
coefficients) when one adds the trend growth rate to the
regression for trend poverty reduction
57
Finding 5: Poverty would have fallen
much faster without rising inequality
• Lack of aggregate growth-equity trade-off implies that:
– Growth has more impact on poverty
– Rising inequality puts a brake on poverty reduction
• If not for the rise in inequality within rural areas, the
national poverty rate in 2001 would have been 1.5%
rather than 8%.
• In most provinces, rapidly rising rural inequality meant
far lower poverty reduction than one would have
expected given the growth.
– An exception was Guangdong, which achieved rapid rural
poverty reduction by combining growth with stationary
inequality. Why?
• Nor did higher inequality permit higher growth
58
Steeper increases in inequality did not
mean faster poverty reduction
Trend in headcount index (% per year)
5
Beijing
0
-5
-10
-15
-20
-25
Guangdong
-30
-0.4 0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
Trend in inequality (% per year)
59
Actual poverty incidence in 2001 and simulated
level without the rise in inequality
Actual headcount index in 2001 (%)
28
24
20
16
12
8
4
0
0
4
8
12
16
Simulated headcount index in 2001
using 1981 Lorenz curve (%)
60
Five lessons
61
Lesson 1: Low-lying fruit of agrarian reform
• Great Leap Forward and the Cultural Revolution left a
legacy of pervasive and severe rural poverty by the late
1970s.
• Yet much of the rural population that had been forced
into collective farming (with weak incentives for work)
could still remember how to farm individually.
• Undoing these failed policies called for decollectivizing agriculture and shifting the responsibility
for farming to households.
• This brought huge gains to the country’s (and the
world’s) poorest. Possibly half of the total decline in
poverty in China 1981-2001 was due to this reform.
• But it was a one-time reform.
62
Lesson 2: Agricultural growth is good for
poor people
• Important lesson for other developing countries.
• Though here too are unusual historical circumstances:
– the relatively equitable land allocation that could be achieved at
the time of breaking up the collectives.
• With fairly equal access to land (at least for the present)
and relatively few distortions to incentives, achieving
higher agricultural growth in China will require
– sound investments in research and development,
– and in rural infrastructure.
• Evidence that targeted poor-area development programs
can help in this setting.
63
Lesson 3: Some forms of public spending
and taxation matter more than others
• Taxation: Don’t tax poor farmers to subsidize urban
consumers! Higher procurement prices reduced poverty.
 ln H t   0.082 1.257  ln PPt 1  1.249 2 ln CPI t 1  ˆt
( 3.058 )
( 3.688 )
( 2.492 )
• These are distributional effects in large part:
 ln H t  0.060  1.040  ln PPt 1  0.882 2 ln CPI t 1  2.335  ln Yt
( 3.791)
( 8.049 )
( 4.651)
( 9.843 )
• This too is an unusual country circumstance
– a procurement system that taxed farmers by setting quotas and
fixing procurement prices below market levels.
• This was a powerful anti-poverty lever in the short-term.
• Public spending: Local – but not central – public spending
reduced poverty, but not inequality.
64
Lesson 4: Less clear on economy-wide
policies (macro stability and free trade)
• Support for the view that macroeconomic stability (esp.,
avoiding inflationary shocks) has been good for poverty
reduction:
 ln H t   0.082 1.257  ln PPt 1  1.249 2 ln CPI t 1  ˆt
( 3.058 )
( 3.688 )
( 2.492 )
• But the score card for trade reform is blank!
– Neither the trade reforms nor the trade expansions coincided
with the times of falling poverty.
– Zero correlation between changes in trade volume (TV) and
changes in poverty. Nor with lagged TV up to two years.
– Also holds with controls (inflation, proc. price, mean Y).
– Endogeneity of trade? Yes, but bias probably goes against the
view that trade reform was poverty reducing in short-term.
65
Lesson 5: Inequality is now an issue
for China
• High inequality in many provinces will inhibit future
prospects for both growth and poverty reduction.
• Aggregate growth is increasingly coming from sources
that bring limited gains to the poorest.
• Inequality is continuing to rise
Elasticity of
H to Gini
and poverty is becoming much
0.0
more responsive to rising inequality. 1981
2001
3.7
• Perceptions of what “poverty” means are also changing,
which can hardly be surprising in an economy that can
quadruple its mean income in 20 years.
66