Slide 1

Household Dynamics,
Chronic Poverty and
Social Protection in Indonesia
Wenefrida Widyanti†, Asep Suryahadi,
Sudarno Sumarto and Athia Yumna
The SMERU Research Institute
www.smeru.or.id
Outline
1.
2.
3.
4.
5.
6.
7.
8.
9.
Introduction
Data
Poverty and Chronic Poverty in Indonesia
Household Composition Change and Chronic Poverty
Household Dynamics as a Protection Instrument
Economic Viability and Chronic Poverty
Household Dynamics and the Concept of Chronic Poverty
Household Dynamics and Social Protection
Conclusion
2
Introduction




A typical household usually consists of several individuals with different
characteristics, including their economic capacities, which in the end
determines the economic capacity of the household as a unit.
A change in household composition will affect the economic capacity and
economic condition of a household and most likely entail simultaneously
both positive and negative effects on a household’s economic capacity and
economic condition.
The direction of causation can also go in the opposite direction. A change in
the economic condition of a household can induce the household to change
its household composition.
The existence of relationships between household composition and
household’s economic capacity and condition indicates that household
composition may play an important role in explaining why some households
fall into chronic poverty (i.e. severe and persistent poverty).
3
Data

Indonesia Family Life Survey (IFLS) Data from RAND, a longitudinal
household survey with a sample which is representative of about 83
percent of the Indonesian population (13 provinces in Indonesia):




IFLS1 was conducted in 1993/94 by RAND in collaboration with the
Demographic Institute of the University of Indonesia (LDUI)  7,224
households (over 22,000 individuals) were interviewed.
IFLS2 was subsequently conducted in 1997 by RAND in collaboration with
UCLA and LDUI  94% of IFLS1 households were relocated and reinterviewed + 878 split-off households (over 33,000 individuals) were
interviewed.
IFLS3 was fielded in 2000, conducted by RAND in collaboration with the
Centre for Population and Policy Studies, Gadjah Mada University (PSKKUGM)  10,400 households (around 39,000 individuals) were interviewed.
A complete panel of 6,403 households were interviewed in IFLS1, IFLS2 and
IFLS3.
4
Poverty and Chronic Poverty in Indonesia (1)
Table 1. Poverty Indicators of Panel Data Households (%)
Indicator
Poverty headcount (P0)
Poverty gap (P1)
Poverty severity (P2)
1993
23.05
6.79
2.92
1997
14.56
3.87
1.56
2000
15.02
3.70
1.37
Number of observations (N)
6,403
6,403
6,403
Source: Authors’ calculation using IFLS data

There was clear improvement in the household welfare between 1993 and
1997; however, due to the advent of an economic crisis starting in the
second half of 1997, there was stagnation in household welfare between
1997 and 2000.
5
Poverty and Chronic Poverty in Indonesia (2)
Table 2. Poverty Dynamics of Panel Data Households
Poverty Pattern
Always poor
Twice poor
Once poor
Never poor
Number of observations (N)
1993
1997
2000
Incidence (%)
Poor
Poor
Poor
Not poor
Poor
Not poor
Not poor
Not poor
Poor
Poor
Not poor
Poor
Not poor
Poor
Not poor
Not poor
Poor
Not poor
Poor
Poor
Not poor
Not poor
Poor
Not poor
4.23
4.33
3.56
9.89
2.00
10.93
4.00
20.16
5.23
65.72
6,403
Source: Authors’ calculation using IFLS data



Always poor + twice poor  chronic poor (14.1%)
Once poor  vulnerable (20.2%)
Never poor  non-poor (65.7%)
6
Household Composition Change and Chronic Poverty (1)
Table 3. Household Distribution by Poverty Groups across the Existence of Household
Composition Change (%)
Existence of Household Composition
Change
Chronic
Poor
Vulnerable
Non-poor
N
No change in composition
Experienced change in composition
15.00
13.66
19.10
20.71
65.90
65.63
2,173
4,230
Total
14.12
20.16
65,72
6,403
Source: Authors’ calculation using IFLS data

The distributions by poverty groups of both households that
experienced household composition change and those that did not
are similar to each other as well as to the total distribution 
Household composition change is not a major cause of the chronic
poverty phenomenon in Indonesia
7
Household Composition Change and Chronic Poverty (2)
Table 4. Household Distribution by Poverty Groups across the Types of Household Composition
Change (%)
Type of Composition Change
Death of breadwinner
Death of other household member
Birth of a child
Divorce or separation
Additional working adult
Additional non-working adult
Others
Total
Chronic
Poor
0.00
15.00
11.81
21.43
14.34
13.92
5.05
13.66
Vulnerable
Non-poor
N
33.33
15.00
15.28
14.29
20.58
21.30
22.22
66.67
70.00
72.92
64.29
65.08
64.78
72.73
12
20
288
14
1,074
2,723
99
20.71
65.63
4,230
Source: Authors’ calculation using IFLS data

The distributions by poverty groups of the households which experienced
household composition change by the type of the composition change that
occurred are similar to the total distribution (except divorce or separation but based
on a small number of observations)  There is no evidence that certain types of
household composition change cause a higher probability for households to be
chronically poor
8
Household Dynamics as a Protection Instrument
Table 5. The Proportions of Households Having a Bad State in Previous Period among
Those which Experienced Household Composition Change (%)
Bad State in Previous Period

1997
2000
Poverty:
- Poor in previous period
- Not poor in previous period
21.99
78.01
14.59
85.41
N
4,230
4,230
Unemployment:
- Head unemployed in previous period
- Head employed in previous period
N
Source: Authors’ calculation using IFLS data
15.26
84.74
4,155
20,52
79,48
4,006
The proportion of bad state households among those that experienced a change in
household composition is similar to the total households  There is no evidence
that households change their composition to cope with poverty and unemployment.
9
Economic Viability and Chronic Poverty
Table 8. Results of Ordered Probit of the Effects of Household Composition on the Probability
to be Chronic Poor or Vulnerable
Independent
Variable
Chronic Poor
Coefficient
Std. Error

Vulnerable
Coefficient
Std. Error
Household composition:
Husband-wife with
children households
Single male/father with
and without children
households
Single female without
children households
Single mother with
children households
Other household
compositions
Household
characteristics:
Number of household
members
Dependency ratio
Proportion of male in a
household
Proportion of adult in a
household
Proportion of working
household members
Proportion of household
members with secondary
education or higher
0.01592
0.02367
0.01416
0.08498
0.06077
0.05223
*
0.02562
0.00477
-0.21484
**
0.00591
0.03290
0.00901
-0.11640
**
0.01100
0.02190
0.02580
0.06705
*
0.03200
0.04668
**
0.01784
0.02383
**
0.00188
0.02035
**
0.00174
-0.00003
0.00004
-0.00002
0.00003
-0.00008
0.00019
-0.00007
0.00016
0.03525
0.02767
0.03011
0.02368
0.02319
*
0.01204
0.01981
*
0.01028
-0.61423
**
0.02719
-0.52458
**
0.03156
Note: The independent variables used in the model are based on 1993 data.
** Significant at 1%
*
Significant at 5%


Single female without children
households have the lowest
probability to be either
chronically poor or vulnerable,
while single father households
have the highest probability to
be vulnerable.
The larger the number of
household members, the higher
the probability a household to be
chronically poor or vulnerable.
Higher proportion of household
members with senior secondary
or higher education reduces the
probability of a household to be
either chronic poor or
vulnerable.
10
Household Dynamics and the Concept of Chronic Poverty
Table 9. Poverty Headcount Rates for Various Household Groups in the Data (%)
Population Group
Poverty Headcount (%)
1993
1997
2000
N
First round households:
- First round households in the
complete panel
- First round households visited in the
second round but not visited in the
third round
- First round households not visited in
the second round but visited in the
third round
- First round households not visited in
the second and third rounds
23.05
14.56
15.02
6,403
14.93
5.97
–
201
12.07
–
10.34
232
10.00
–
–
300
21.92
(N=7,136)
14.29
(N=6,604)
14.86
(N=6,635)
7,136
- New households in the second
round visited in the third round
- New households in the second
round not visited in the third round
–
8.94
11.91
705
–
13.39
–
224
- Total second round households
–
10.01
(N=929)
11.91
(N=705)
929
–
–
9.30
2,818
21.92
(N=7,136)
13.77
(N=7,533)
13.11
(N=10,158)
10,883
- Total first round households
Second round households:
Third Round Households:
- New households in the third round
All Households in the Data

The complete panel
households are poorer than
the total sample in each
round:


The households dropped
out of sample are less
poor
The new households
added into sample are
less poor
 The use of household as the
unit of analysis for poverty
may undermine the
conceptualisation &
measurement of chronic
poverty.
Source: Authors’ calculation using IFLS data
11
Household Dynamics and Social Protection
Table 12. Results of probit analysis of household participation in government social
protection programmes in 2000 (%)
Basic needs assistance
Other assistances
Coefficient
Coefficient
Independent variable
Std. Error
Std. Error
Poverty status:
Chronically poor
0.33698
0.22719
0.12199
0.42655
Vulnerable
0.29597
0.10773
-0.02686
0.20418
Poor in 1993
0.03295
0.10902
-0.13284
0.20336
Poor in 1997
-0.06977
0.10596
-0.08762
0.20563
Poor in 2000
0.12706
0.10269
-0.09073
0.19605
Change in household
composition:
Change in household
composition 1993-1997
Change in household
composition 1997-2000
0.07934
0.04978
0.07779
0.08664
0.01345
0.04076
-0.00663
0.07303
**
Household characteristics:


Poverty status in general does not increase the probability of receiving assistance,
except for the vulnerable in receiving basic need assistance.
Change in household composition does not increase the probability of receiving
assistance either.
12
Conclusion (1)




Household composition change is not a major cause of the
chronic poverty phenomenon in Indonesia.
There is no evidence that certain types of household
composition change cause a higher probability for households
to be chronically poor.
There is no evidence that households change their composition
to cope with poverty as well as unemployment.
Husband-wife households have the highest probability to be
non-poor, while single mother households have a higher
probability to be non-poor than single father households.
13
Conclusion (2)




The larger the number of household members, the higher the
probability a household to be chronically poor or vulnerable.
A higher proportion of household members with senior
secondary or higher education reduces the probability of a
household to be either chronic poor or vulnerable.
Because of household dynamics, the use of household as the
unit of analysis for poverty could undermine the
conceptualisation & measurement of chronic poverty.
Poverty status and change in household composition in general
do not increase the probability of a household to receive an
assistance.
14
15