Vous êtes sur la page 1sur 50

REGIONAL WELFARE

DISPARITIES IN THE
PHILIPPINES:
Household Endowments
or Returns?
Emmanuel Skoufias
Katja Vinha

The World Bank Group
August 20, 2014
Intro: Objectives
Document disparities in well-being across space in the
Philippines
How does the standard of living vary within and between regions in the
Philippines?
Across provinces?

Investigate the determinants of the observed differences
in the standards of living
Are differences between any two regions due
to differences in returns to characteristics
or difference in the characteristics themselves?

Provide general diagnostic framework
Useful for many other outcomes (e.g. reducing regional disparities in
child stunting, wasting and other MDGs)
2
Intro: Objectives
Inform the policy debate surrounding regional and
province level inequalities in the Philippines.
Provide empirical evidence on the sources of regional
welfare differences
To what extent is the current allocation of fiscal transfers
and LGU expenditures related to needs?
How does access to better infrastructure (paved roads,
clean water, sanitation, and electricity) affect the
(marginal welfare) returns of household endowments (e.g.
the level of education and the demographic composition
of households)?





3
Intro: Methodological Framework
Welfare measure: household consumption
expenditures per capita (PCE), adjusted for
cost of living differences across regions in the
Philippines.
The official province-specific poverty lines in 2009 are used to make
these adjustments for spatial cost of living differences.

Determinants of welfare (model):
a set of household portable/mobile characteristics such as age,
family composition, and the level of education of the household head
and his/her spouse; and
a set of corresponding parameters that summarize the marginal
welfare effects or returns of these characteristics associated with
living and working in any given geographic location.





4
Welfare measure
Welfare measure: Per Capita Consumption
adjusted by spatial cost of living differences



1. Where
,

= nominal per capita consumption of


household i in province p
2. PL

= 2009 poverty line in province p
3. PL
NCR
= 2009 poverty line in NCR
4. CPI in year t in region p
5
Welfare model
The log of the welfare measure is
modeled as a linear function of
portable/mobile household endowments



X(i) = main household characteristics:
whether the household is headed by a female
the marital status of the head (married, single, or separated/divorced),
the age of the household head and its square,
the composition of the household (number of infants under one, number of children aged 1 to
6, number of children aged 7 to 14, number of young adults aged 15 to 24, number of adults
aged 25 to 59, and number of adults aged 60 and older), and
the educational attainment of the household head (no education or some elementary school;
elementary school completed; high school completed; college completed.)


6
Methodological Framework
Question addressed :
Whether welfare and poverty differences
across spacee.g. between urban and rural
areasare explained primarily by the
differences in characteristics of the people
living in these areas or by the differences in
returns that they obtain from those
characteristics in their region of residence.





7
Some Equations

1
1

1
2

1
85

= +
1

1
1

1
2

1
85
+
2

2
1

2
2

2
85
+

85

(1)

+
1

1
+
2

2
+

+
1

,
2

;
1
+
2

,
2

;
2
+

,
2


1
+

,
2


2

8
2. Welfare
Disparities Between
and Within Regions:
2009
9
Welfare Disparities within &
between
Question 1: How large are the disparities in
welfare within regions of the Philippines in
2009?
Welfare differences: urban vs rural areas within each
region

Question 2: How large are the disparities in
welfare between/across regions of the
Philippines in 2009?
Welfare differences: urban vs NCR
Welfare differences: rural vs rural CALABARZON

10
Figure 1: Within regions
11
Figure 2: Between regions, whole
12
Figure 3: Between regions, urban
13
Figure 4: Between regions, rural
14
-
.
4
-
.
2
0
.
2
P
e
r
c
e
n
t
a
g
e

d
i
f
f
e
r
e
n
c
e
A
R
M
M
C
e
n
t
r
a
l

V
i
s
a
y
a
s
Z
a
m
b
o
a
n
g
a

P
e
n
i
n
s
u
l
a
N
o
r
t
h
e
r
n

M
i
n
d
a
n
a
o
B
i
c
o
l
C
A
R
A
G
A
D
a
v
a
o
M
I
M
A
R
O
P
A
E
a
s
t
e
r
n

V
i
s
a
y
a
s
S
O
C
C
S
K
S
A
R
G
E
N
C
A
R
W
e
s
t
e
r
n

V
i
s
a
y
a
s
I
l
o
c
o
s
C
e
n
t
r
a
l

L
u
z
o
n
C
a
g
a
y
a
n

V
a
l
l
e
y
Source: Author estimates based on 2009 FIES
Rural households in region vs. rural households in CALABARZON
Welfare differences between regions, 2009
Figure 5
Figure 5: By occupations
15
Table 4:
Importance of endowment and returns at the national level, 2009
Differential Endowment Returns
Predicted Welfare: Urban area
10.622***

(0.006)

Predicted Welfare: Rural area
10.078***

(0.005)

Difference
0.544***

(0.008)

Welfare difference attributable to:

Demographics

0.055*** -0.223***

(0.004) (0.073)
Education

0.192*** 0.003

(0.004) (0.011)
Constant

0.516***

(0.074)
Total

0.248*** 0.296***
(0.006) (0.006)
Observations 38,400
Source: Author calculations based on FIES 2009.
Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

16
Figure 6
17
3. Decomposing
Welfare Disparities
Between and Within
Regions: 2009
18
Decomposing Welfare Differences
Question: What is the primary
explaining factor (correlate) of welfare
disparities across/between regions in
the Philippines in 2009?
Is it differences in the returns to education, or family
demographic composition
OR is it differences in the characteristics themselves?


19
Decomposing Regional Welfare
Disparities
If differences in characteristics:
Migration of households across space is able to equalize returns across regions, but
concentration of individuals with poor endowments (low education, low skilled occupations) in
the poorer regions.

Policy Implication: Primary focus on targeted (not universal) human capital
development programs ( e.g. CCTs such as 4P etc)

If differences in returns:
How large are the differences in returns (e.g. to education) across regions?
Limits of migration in equalizing the returns to characteristics across regions
Markets that are segmented across regions

Policy implication: Primary focus on increasing the returns to household
endowments in backward regions (e.g. territorial development programs,
investments in infrastructure, transportation, connectivity)
20
Figure 7a: Between Rural 2003
21
Figure 7b: Between Rural 2009
22
Figure 8a: Between Urban 2003
23
Figure 8b: Between Urban 2009
24
Figure 9a: Within, or U vs. R 2003
25
Figure 9b: Within, or U vs. R 2009
26
Regional Welfare Disparities:
Summary of Findings
BETWEEN REGIONS: Differences in the returns are the
primary explanation of the welfare differences between
regions.
Policies aimed at increasing the returns to household mobile
characteristics have the potential of reducing welfare inequalities
between regions

WITHIN REGIONS (R vs. U) : Differences in household
endowments/characteristics are main factors behind the
lower average welfare in rural areas relative to that in urban
areas in the same region
human capital development programs (e.g. CCTs such as 4P etc)
likely to be very effective in decreasing welfare inequalities with in
regions (i.e. ,U and R areas within regions)
27
Regional Welfare Disparities:
Summary of Findings
These findings consistent with the findings in
other EAP countries (Thailand and Indonesia)
Consistent with the presence of agglomeration economies
Limits of migration in equalizing the returns to characteristics
across regions (migration seems to work better within regions)
Market segmentation

But in sharp contrast with results in Latin
American countries where differences in the
characteristics are the main correlates of
welfare differences between regions


28
4. Welfare
Differences
Between/Among
Provinces in 2009
29
Figure 11
30
Decomposing Welfare Differences Across
Provinces
We investigate the correlation
between the actual welfare ratio and
two simulated welfare profiles:
(i) Returns (Geographic profile): the simulated welfare
ratio holding characteristics at the province level constant (and
allowing returns to differ across provinces)
(ii) Characteristics (Concentration profile): the simulated
welfare ratio holding returns at the province level constant
(and allowing characteristics to differ across provinces)

31
STEP 1: We estimate mean welfare ratio profile across the 85 provinces by estimating a separate
regression for each province, using ordinary least squares, i.e.,

where = 1, ,85 (B.1)


STEP 2: We derive the mean simulated welfare ratio for each province by holding constant
household characteristics at the national average, denoted by

, and allowing returns to vary


by province, as estimated in the step 1 above, (returns profile) i.e.,



=

where = 1, ,85 (B.2)


STEP 3: We derive the mean simulated welfare ratio for each district by holding constant
returns as the national average and allowing characteristics to vary by province, (endowments
profile)



=

where = 1, ,85 (B.3)



where

represents the sample mean characteristics of households by province and the


parameters are calculated as population-weighted means:

=1

(B.4)

=1

32
Figure 12
33
Decomposing Welfare Differences Across
Districts
In 2009 the correlation between the actual
welfare profile and the returns profile is
stronger than the correlation between the
actual welfare profile and the characteristics
profile
reconfirms earlier results at regional level

Policy question comes down to what type of
policies can increase returns to mobile hh
characteristics (examined in more detail in part 7)
34

Q: How do province level returns to
education vary across regions?
Table 5: Differences in marginal returns to welfare across regions -
Educational (NCR omitted) reference category: no or some
elementary school, 2009

Highest degree completed by Head of
Household
REGION
Elementary
High
school
College

Ilocos
0.131*** 0.098 0.010
(0.046) (0.059) (0.083)
Cagayan Valley
0.058 0.117 0.132
(0.055) (0.071) (0.099)
Central Luzon
0.016 -0.004 -0.112*
(0.037) (0.047) (0.066)
Bicol
0.076* 0.113** 0.177**
(0.044) (0.056) (0.079)
Western Visayas
0.071* 0.136** 0.121
(0.041) (0.052) (0.073)
Central Visayas
0.162*** 0.335*** 0.140*
(0.041) (0.053) (0.073)
Eastern Visayas
0.095* 0.229*** 0.249***
(0.048) (0.062) (0.086)
Zamboanga Peninsula
0.191*** 0.286*** 0.340***
(0.054) (0.069) (0.095)
Northern Mindanao
0.137*** 0.269*** 0.280***
(0.049) (0.063) (0.087)
Davao
0.179*** 0.221*** 0.125
(0.049) (0.062) (0.087)
SOCCSKSARGEN
0.132** 0.179*** 0.133
(0.050) (0.064) (0.090)
CAR
0.059 0.099 -0.042
(0.073) (0.093) (0.130)
ARMM
-0.040 -0.162** -0.486***
(0.054) (0.070) (0.097)
CARAGA
0.108* 0.139* 0.238**
(0.060) (0.077) (0.107)
CALABARZON
0.044 0.057 -0.003
36

Q: How do province level returns to
household mobile endowments
correlate with province characteristics
(roads, connectivity, health facilities
etc)?
Table 6: Correlation of marginal welfare effects and provincial level variables, 2006

Marginal welfare effect of
IMMOBILE PROVINCIAL
VARIABLES
Female
head
Child (<1
year)
Child
(1-6
years)
Child
(7-14
years)
Y Adult
(15-25
years)
Adult
(25-59
years)
Elderly
(60+
years)
HOH w/
elementary
HOH w/
high
school
HOH
w/college
Expenditure gini, 2009
0.021 -0.387* 0.026 -0.074 0.145** 0.206* 0.169 1.267*** 2.691*** 3.545***
(0.308) (0.222) (0.083) (0.061) (0.071) (0.106) (0.165) (0.187) (0.231) (0.336)
Share of population urban, 2009
-0.165 0.137* 0.023 0.008 0.023 0.007 0.011 -0.064 0.092 -0.098
(0.102) (0.073) (0.028) (0.020) (0.023) (0.035) (0.055) (0.062) (0.076) (0.111)
Population density, 2010
-0.013*** -0.006* -0.001 -0.001 -0.001 -0.000 -0.001 0.003 -0.002 0.008
(0.004) (0.003) (0.001) (0.001) (0.001) (0.002) (0.002) (0.003) (0.003) (0.005)
% of household heads with at
least a high school diploma, 2009
0.838*** -0.262* -0.041 -0.066 -0.021 0.007 -0.022 0.226* -0.045 0.028
(0.213) (0.154) (0.058) (0.042) (0.049) (0.073) (0.115) (0.129) (0.160) (0.233)
% of barangays with an
elementary school, 2010
-0.097 -0.150 -0.007 -0.030 0.027 -0.016 -0.009 0.250*** 0.191* 0.354**
(0.138) (0.099) (0.037) (0.027) (0.032) (0.047) (0.074) (0.083) (0.103) (0.150)
% of barangays with a health
center within 5 km, 2010
0.141 -0.039 -0.071 0.007 -0.119** -0.007 -0.155 0.059 -0.021 -0.011
(0.232) (0.167) (0.063) (0.046) (0.054) (0.080) (0.125) (0.141) (0.174) (0.254)
% of barangays with access to the
highway, 2010
-0.108 0.122 -0.053 0.021 -0.015 -0.033 0.004 -0.086 -0.021 -0.056
(0.174) (0.125) (0.047) (0.034) (0.040) (0.060) (0.093) (0.105) (0.130) (0.189)
% of barangays with cell phone
reception, 2010
0.095 0.132 0.031 -0.116* 0.072 -0.079 0.123 -0.346* -0.197 0.320
(0.317) (0.228) (0.085) (0.062) (0.073) (0.109) (0.170) (0.192) (0.237) (0.345)
% of barangays w/o
manufacturing within 2km, 2007
-0.090 0.076 0.018 0.019 0.042 -0.055 -0.057 0.096 0.170 0.020
(0.141) (0.101) (0.038) (0.028) (0.032) (0.048) (0.076) (0.085) (0.105) (0.154)
% of barangays w/o commercial
establishments within 2km, 2007
0.216 -0.101 0.005 -0.010 -0.048 0.064 0.080 -0.057 -0.089 -0.020
(0.142) (0.102) (0.038) (0.028) (0.033) (0.049) (0.076) (0.086) (0.106) (0.155)
Dynasty, 2009
-0.009 -0.019 -0.000 0.003 -0.005 0.018 0.012 -0.008 -0.046 -0.046
(0.057) (0.041) (0.015) (0.011) (0.013) (0.020) (0.030) (0.034) (0.043) (0.062)
Constant
-0.221 0.047 -0.086 0.007 -0.078 0.041 -0.042 -0.245 -0.483** -0.670**
(0.254) (0.183) (0.068) (0.050) (0.058) (0.087) (0.136) (0.154) (0.190) (0.277)
Observations 79 79 79 79 79 79 79 79 79 79
R-squared 0.366 0.165 0.200 0.344 0.277 0.108 0.085 0.530 0.740 0.724
38
Province Endowment & Returns
Surprisingly few variables summarizing the immobile characteristics
of the province that are significantly correlated with the estimated
marginal welfare effects of household portable endowments.

There is a statistically significant and sizable correlation between
expenditure inequality and returns to educationthe greater the
expenditure inequality in the province, the larger the returns to all
levels of educational attainment.
In fact, a one standard deviation increase from the mean expenditure inequality is associated with education
coefficients that are 0.07 (40%), 0.15 (30%) and 0.2 (18%) larger for elementary, high school and college
attainment, respectively.

Provinces where a larger share of the barangays had an elementary
school in the barangay also had higher returns to all levels of
education.
5. District needs vs.
Fiscal Transfers
(Revenues) of LGUs
40
District needs vs. Fiscal
Transfers
Is there is a needs-based allocation
of transfers to provincial governments?
Central government transfers allocation based
on need expect a significantly negative
correlation between the level of per capita
transfers and needs



41
Figure 14
42

the allocation of fiscal transfers is
based on needs.
Or poorer regions receive higher
revenues (IRA) pc

6. District needs vs.
Fiscal Expenditures at
LGU level
44
District needs and LGU Fiscal
Expenditures
Do LGUs in poorer provinces have
higher fiscal expenditures per capita?
expect a significantly negative correlation
between the level of per capita fiscal
expenditures and needs/poverty



45
Figures 15 & 17
46

LGU govt expenditures unrelated to
needs (poverty level)
In fact LGUs in richer provinces have higher
expenditures on education pc than LGUs in
the poorer provinces likely to contribute to
greater regional disparities in welfare


Take away messages-1
The welfare differences among
regions/provinces in the Philippines are
primarily due to differences in the returns to
household mobile endowments
Policies aimed at increasing returns
CCT programs (e.g. 4P) part of a greater package needed for
equalizing opportunities for welfare
Increases in HK need be accompanied by investments that increase
returns to HK (connections across space) in SR to MR




48
Take away messages-2
The design of the fiscal transfer system is
consistent with promoting the equality of
opportunities for welfare across provinces

However, improvements in the capacity of
local government units appear necessary to
ensure that the opportunities for welfare are
promoted among the poorer provinces.
spending pattern by LGUs in the different provinces
appears to be unrelated to needs



49
Thank you
50

Vous aimerez peut-être aussi