00074910012331338903
Bulletin of Indonesian Economic Studies
ISSN: 0007-4918 (Print) 1472-7234 (Online) Journal homepage: http://www.tandfonline.com/loi/cbie20
Changes in Household Welfare, Poverty and
Inequality During the Crisis
Emmanuel Skoufias & Asep Suryahadi
To cite this article: Emmanuel Skoufias & Asep Suryahadi (2000) Changes in Household
Welfare, Poverty and Inequality During the Crisis, Bulletin of Indonesian Economic Studies,
36:2, 97-114, DOI: 10.1080/00074910012331338903
To link to this article: http://dx.doi.org/10.1080/00074910012331338903
Published online: 18 Aug 2006.
Submit your article to this journal
Article views: 134
View related articles
Citing articles: 9 View citing articles
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=cbie20
Download by: [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RA JA ALI HA JI
TANJUNGPINANG, KEPULAUAN RIAU]
Date: 19 January 2016, At: 22:05
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Bulletin of Indonesian Economic Studies
Vol 36 No 2, August 2000, pp. 97–114
CHANGES IN HOUSEHOLD WELFARE, POVERTY AND
INEQUALITY DURING THE CRISIS
Emmanuel Skoufias
International Food Policy Research Institute (IFPRI), Washington DC
Asep Suryahadi and Sudarno Sumarto*
Social Monitoring and Early Response Unit (SMERU), Jakarta
This study provides evidence about changes in the distribution of living
standards among Indonesian households during the economic crisis. It
uses consumption expenditure data from a panel of households that were
surveyed in May 1997, just before the onset of the crisis, and then again in
August 1998, about a year after the crisis began. A household-specific
deflator is used to make nominal consumption expenditures comparable
across this period. The results suggest that there was a considerable drop
in household welfare during the economic crisis. Average per capita
expenditures fell significantly, and at the same time inequality increased.
The poverty rate also appears to have doubled from the pre-crisis level.
However, transitions into and out of poverty before and after the crisis
reveal remarkable fluidity.
INTRODUCTION
In this article, we present evidence about changes in Indonesian
household living standards—measured by per capita real consumption
expenditures—and in the distribution of living standards across
households—measured by indices of inequality—during the economic
crisis. Our study has two distinguishing characteristics.
First, it is based on a set of households that were surveyed in May
1997, just before the onset of the crisis, and then 14 months later in August
1998, about a year after the crisis began. The use of ‘panel’ data (a
combination of cross-section and time-series data on the same households)
offers the opportunity to identify how the welfare of specific households
changed during the crisis.1
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
98
Skoufias, Suryahadi and Sumarto
The second characteristic relates to the price deflator we use to make
nominal consumption expenditures comparable across years. Given the
large shifts in the relative prices of food and non-food items during the
crisis, different price deflators result in very different estimates of the
magnitude and severity of its impact.2 We adopt a household-specific
deflator that is a weighted average of the food and non-food price indices.
The weights applied to food and non-food prices vary from household
to household and are calculated from an ‘Engel curve’, which predicts
each household’s food share in consumption expenditure, based on the
household’s (logarithms of) per capita consumption and family size. We
believe that such a deflator is more appropriate than the standard deflator
for evaluating the impact of the economic crisis, since it captures more
accurately the impact of higher food prices on poorer households.
DATA AND KEY VARIABLES
Data
The data we use are part of the 100 village survey conducted by
Indonesia’s central statistics agency, BPS, and funded by UNICEF.3 The
purpose of this survey is to monitor changes in health, education, nutrition
and socio-economic status in 100 villages purposively selected from 10
districts (kabupaten) in 8 provinces throughout Indonesia. In each village,
120 households were chosen—giving a total sample of 12,000
households—and information was collected for the household and all
family members about factors such as education, employment sector and
type of work. Although the sample is large in terms of number of
households, and represents a variety of areas across the country, it is
important to note that the selection of villages was not random. Hence,
readers should keep in mind that the findings of this study apply only to
this sample.
The data analysed here are from the 100 village survey rounds in
May 1997 and August 1998. Our analysis relies exclusively on a panel of
households that were interviewed in both rounds. In the August 1998
round, the sampling frame was changed from two enumeration areas of
60 households each to the original two plus a third enumeration area,
each with 40 households. Of the 120 households from the two
enumeration areas that were the same in both rounds, 80 were targeted
for re-interview. Unfortunately, in the second round the identifying codes
for households were changed. We therefore identified the households
within each enumeration area using the name of the household head,
and then cross-checked the matches using the demographic characteristics
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
99
of the households. This resulted in the matching of 8,141 (68%) of the
households across the two rounds, implying that, of the 120 households
in each village, an average of 82 were actually followed across rounds.4
Measuring Household Welfare
We use per capita consumption expenditures (PCE) as one indicator of
household living standard.5 Of course, consumption expenditure is only
one of many components of household welfare. Others include
employment, health conditions, and the ability to access and use basic
services such as water, sanitation, health care and education. We examine
the changes in household welfare by using household PCE in each year
to calculate a variety of poverty and inequality indices. As discussed in
detail in Deaton (1997) and Deaton and Zaidi (1999), the social welfare
function approach developed by Atkinson (1970) provides a useful
framework for interpreting statistical measures of poverty and inequality.
For example, if we were to describe social welfare in period t, W(t), as a
function of the PCE of all the households in the population in period t,
i.e.:
(
)
W (t) = W PCE1 (t), PCE2 (t),..., PCEK (t)
(1)
where K is the number of households in the population, then with a set
of relatively innocuous assumptions about the properties of the function
W,6 we may express social welfare in period t as:
(
)
W (t) = PCE(t) 1 − I (t)
(2)
i.e. as a function of the mean level of PCE in period t, denoted by the bar
over PCE(t), multiplied by one minus the level of inequality in the
distribution of PCE in period t (denoted by I(t)).7 Along similar lines, the
common indices of poverty (described in more detail below) can also
conceivably be regarded as particular forms of the social welfare function.
Construction of Key Variables
Per capita expenditures are defined as PCE(t) = C(t)/N(t), where C(t)
denotes deflated food and non-food consumption expenditures in period
t (see below for details on the deflators used) and N(t) denotes total family
size in period t.8 In various instances we also look at PCE for food and
non-food separately. Food expenditure is the sum of expenditures on
grains, meat, fish, eggs and milk, vegetables, beans and nuts, fruits,
seasonings, fats and oils, soft drinks, prepared food and other food items,
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
100
Skoufias, Suryahadi and Sumarto
and alcohol and tobacco.9 The reference period for expenditures on these
items was the week preceding the day of the interview. These weekly
expenditures were transformed into monthly expenditures by
multiplying by (30/7).
For non-food expenditures two measures were collected, each for a
different reference period—the previous month and the previous 12
months. To minimise recall errors, but at the risk of introducing exclusion
errors, we used the expenditures reported for the previous month. Nonfood expenditure is defined as the sum of expenditures on housing, health,
education, clothing and shoes, durable goods, taxes and insurance,
ceremonies, and other expenses.
It is important to take note of two qualifications about the data. First,
the surveys were conducted in different months of the calendar year (May
in 1997 and August in 1998), thus introducing the possibility that some
of the observed changes in consumption may be due to seasonality. This
is particularly true of items such as education and clothing expenditures,
which are influenced by the educational calendar. Second, although our
sample consists primarily of households in rural areas, there are some
households or villages that are classified as being in ‘urban’ areas (17.8%
of the sample). The reader is cautioned that villages in urban areas in our
sample are not part of large metropolitan agglomerations (such as
Jabotabek, Surabaya and Medan), but are villages that are close to the
district capitals. Such villages are classified administratively as kelurahan
instead of desa, and coded as urban areas in our sample.
Deflating Expenditures
The nominal consumption expenditures in the two rounds of the survey
need to be adjusted in order to allow meaningful comparisons about
household welfare across the two rounds. For example, because of the
large increases in the price of rice during the economic crisis, an
expenditure of Rp 10,000 on rice during August 1998 represents a much
smaller quantity of rice than the same expenditure in May 1997.
To control for the large price differences across the rounds we construct
a Laspeyres price index using the following steps. First, we construct a
deflator for food and non-food items using the mean shares of the food
and non-food items in the May 1997 survey as weights, and the price
indices published in the BPS monthly statistical bulletin Indikator Ekonomi
in May 1997 and in August 1998.10 We have not used region-specific
deflators for food or non-food items because the regional deflators
available in Indonesia are based explicitly on urban prices, so any crossregional comparisons should be made with caution.11
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
101
Second, we construct a household-specific deflator that is a weighted
average of the food and non-food price indices calculated above.
Specifically, if we denote by t the periods May 1997 and August 1998,
and the price deflators for food and non-food in period t by PF(t) and
PNF(t) respectively, then the price deflator for period t for household h,
Ph(t) can be expressed as:
(
)
P h (t) = W√Fh (97 )PF (t) + 1 − W√Fh (97 ) PNF (t)
(3)
The weights applied to food and non-food items vary from household
to household. The weight for each household was calculated from the
predicted value of the regression of household food share in May 1997,
W√Fh (97 ) on the logarithm of per capita consumption, ln( PCE(97 )) , and the
logarithm of household size.12 In this manner the influence of householdspecific unobserved components or tastes on the share of food is
eliminated.
As is the case for all Laspeyres price deflators, the share of food is
assumed to be constant. To the extent that the changes in relative prices
are such that the share of food also increases as a result of the crisis (as
indicated by the data), then the above deflator may be underestimating
the increases in prices. In an effort to check for this possibility, we also
constructed another deflator with variable weights for food based on the
coefficients from an Engel curve estimated separately for May 1997 and
for August 1998. However, the changes in the results obtained using the
deflators with fixed and varying food shares were very small, so we
choose to present only the results obtained using the deflator based on a
fixed food share.
WELFARE CHANGES DURING THE CRISIS
Distribution of Per Capita Expenditure
We begin with a graph that provides a quick visual impression of changes
in household consumption expenditures between May 1997 and August
1998. Figure 1 graphs the cumulative distribution functions (CDF) of
ln(PCE) in both periods. The figure shows that the 1998 CDF lies to the
left of the 1997 CDF with no ‘crossings’, implying that the 1998 CDF
‘stochastically dominates’ the 1997 CDF.13 This means that the poverty
rate will be higher in 1998 no matter what poverty line is chosen (Deaton
1997).14
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
102
Skoufias, Suryahadi and Sumarto
FIGURE 1 Cumulative Distribution Functions of ln(PCE)
Cumulative
distribution
1.0
0.8
1997
0.6
0.4
1998
0.2
0.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
ln(PCE)
Moreover, the shift to the left in the CDF between 1997 and 1998 was
not exactly parallel, with the lower part of the CDF shifting more to the
left than the upper part. For example, the respective implied falls in PCE
for the 1st, 5th, 10th and 20th percentiles are 41, 27, 22 and 20%, while for
the 80th, 90th, 95th and 99th percentiles they are 14, 12, 10 and 10%. This
means that the fall in consumption expenditures was greater for those at
the lower than at the upper end of the distribution, indicating a worsening
of the distribution and an increase in inequality.15
Poverty Rates
There are various methods of calculating a poverty line, and they can
produce widely differing results. Moreover, at any given point in time
the level of poverty reported is quite sensitive to the poverty line estimate
used. Equally reasonable poverty lines can produce very different poverty
rates for the same data. In this instance we are interested principally in
the changes in poverty over time. Hence, to match the official pre-crisis
poverty rate, we chose as the poverty line in 1997 the 11th percentile of
the distribution of ln(PCE) in the full sample of 12,000 households (not
just the matched sample).16 In other words, the poverty line was chosen
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
103
to produce an 11% poverty rate in the full sample. With this poverty line,
the poverty rate in our matched sample of 8,141 households in May 1997
turned out to be 12.4%. From that level we can calculate changes that are,
while not invariant, robust to the initial assumed level of poverty.
In table 1 we report the values of the Foster–Greer–Thorbecke (FGT)
poverty indices (Foster et al. 1984). This class of poverty measures is highly
regarded because it meets all the axioms desirable in consumption-based
poverty measures, and contains a parameter α that can be set to generate
results showing varying levels of sensitivity to income distribution among
the poor. Specifically, the FGT family of poverty measures is summarised
by the formula:
1
P(α ) =
N
q
z − ci
z
∑
i =1
α
(4)
where N is the number of households, ci is the per capita consumption
(or income) of the ith household, z is the poverty line, q is the number of
poor households, and α is the weight attached to the severity of household
poverty (or the distance from the poverty line). When α = 0, the FGT
measure collapses to the Headcount Index, or P(0), i.e. the proportion of
the population that is below the poverty line. This measure, while useful
for general poverty comparisons, is insensitive to differences in the depth
of poverty, in the sense that households far below the poverty line receive
the same weight as households just below the poverty line. Moreover, as
Deaton (1997) points out, it serves as an unsatisfactory indicator of
welfare, for it is possible for this measure to indicate a decline in headcount
poverty when some very poor households become even poorer and some
not so poor households’ expenditures increase sufficiently to push these
households above the poverty line.
This shortcoming is overcome by assigning higher values to the
parameter α. When α = 1, the FGT measure gives the Poverty Gap, or
P(1), a measure of the average depth of poverty, and indicates the average
TABLE 1 Foster–Greer–Thorbecke (FGT) Poverty Indicesa
1997
1998
Percentage changeb
P(0)
P(1)
P(2)
0.124
0.245
98
0.023
0.060
163
0.006
0.023
259
a
See text for explanation of P(0), P(1) and P(2).
b
Based on original unrounded numbers.
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
104
Skoufias, Suryahadi and Sumarto
money gap by which the consumption of the poor falls short of the poverty
line. When α = 2, the FGT index is called the Severity of Poverty index, or
P(2). This measure differs from P(1) in that it assigns relatively more
weight than P(1) to individuals whose expenditures are further away
from the poverty line and who are thus in more severe poverty.
Based on the poverty line in 1997, the poverty rate (Headcount Index)
doubled in our panel of households from 12.4% in May 1997 to 24.5% in
August 1998.17 Although this rate in 1998 is remarkably close to the rural
area poverty rate of 25.7% estimated by BPS (Sutanto 1999), the two rates
are not strictly comparable because they are derived by very different
methods.
The higher order poverty indices also increased, and by a factor higher
than the increase in the Headcount Index. For example, the Poverty Gap
index rose from 0.023 to 0.06, which means that the average poverty gap
increased from 2.3% to 6% of the poverty line. Meanwhile, the Poverty
Severity Index increased by a factor of almost 4, i.e. from 0.006 to 0.023.
Poverty Transitions
In table 2 we present a poverty transition matrix. We classify households
into one of four categories based on the relationship between their PCE
and the poverty line (PL): poor (PCE < PL); near poor, above the poverty
line but by less than 25% (PL ≤ PCE < 1.25*PL); near non-poor, more than
25% but less than 50% above poverty line (1.25*PL ≤ PCE < 1.5*PL); and
non-poor, 50% or more above poverty line (PCE≥ 1.5*PL). This allows us
to examine both how those in poverty in 1997 fared and who moved into
poverty in 1998. The first column shows the distribution of households
in 1997, indicating that 1,010 households were poor and 5,029 non-poor.
The first row shows the distribution of households across these categories
in 1998, with 1,997 poor households and 3,562 non-poor.
For each row, the columns show how households in that category in
1997 fared in 1998. For example, take the 988 households that were near
poor in 1997. In 1998 only 239 of these households (24.2%) were ‘on the
diagonal’ or in the same category of near poor. Among the rest, 309 (i.e.
140 + 169) or 31.3% of households had improved their economic status,
while 440 (44.5%) had fallen into poverty. Similarly by looking down the
columns one can see where the households in any category in 1998 were
in 1997. So, for instance, of the 3,562 households that were non-poor in
1998, 3,029 (85.0%) were also non-poor in 1997, while only 58 households
(1.6%) had come from being poor in 1997. Meanwhile, the bottom number
in each set gives the percentage of the total households. So 8.6% of the
population was poor in both periods, while 524 households (6.4% of 8,141)
were non-poor in 1997 but became poor in 1998.
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
105
In terms of percentages, only 10.4% of those non-poor in 1997 had
become poor in 1998. However, since the non-poor were 61.8% of the
1997 population, they are 26.2% of the poor in 1998. On the other hand,
even though 44.5% of the near poor in 1997 had become poor in 1998, as
only 12.1% of the 1997 population was near poor, only 22.0% of the poor
in 1998 came from the near poor category in 1997.
Although during the crisis many of the households that were
marginally poor before the crisis became impoverished, the transition
matrix reveals considerable fluidity. Approximately 31% of the poor in
TABLE 2 Poverty Transition Matrix
Poverty Status in 1998a
Total 1997
Poor
Near Poor
Total 1998
Row percentage
Column percentage
Total percentage
8,141
100.00
100.00
100.00
1,997
24.53
100.00
24.53
1,369
16.82
100.00
16.82
1,213
14.90
100.00
14.90
3,562
43.75
100.00
43.75
Poverty Status in 1997
Poor
Row percentage
Column percentage
Total percentage
1,010
100.00
12.41
12.41
697
69.01
34.90
8.56
177
17.52
12.93
2.17
78
7.72
6.43
0.96
58
5.74
1.63
0.71
Near poor
Row percentage
Column percentage
Total percentage
988
100.00
12.14
12.14
440
44.53
22.03
5.40
239
24.19
17.46
2.94
140
14.17
11.54
1.72
169
17.11
4.74
2.08
Near non-poor
Row percentage
Column percentage
Total percentage
1,114
100.00
13.68
13.68
336
30.16
16.83
4.13
282
25.31
20.60
3.46
190
17.06
15.66
2.33
306
27.47
8.59
3.76
Non-poor
Row percentage
Column percentage
Total percentage
5,029
100.00
61.77
61.77
524
10.42
26.24
6.44
671
13.34
49.01
8.24
805
16.01
66.36
9.89
3,029
60.23
85.04
37.21
a
Poor: PCE < PL (poverty line)
Near non-poor: 1.25*PL ≤ PCE
−∞
∫ g(x)dx
∀α .
−∞
14 This is true because if we draw any vertical line on the graph, it will always
cross the 1998 CDF at a higher distribution value than the 1997 CDF.
15 However, when the analysis is disaggregated into urban and rural areas, there
is some indication that in urban areas the falls in consumption were greater at
the upper end of the distribution. For more details on this, see the longer
version of this paper in Skoufias et al. (1999).
16 The alternative is to use the standard approach to estimating a poverty line.
However, this ‘standard’ approach is not free from questionable assumptions
about the composition of the food bundle, the reference population, or the
minimum level of calorie availability used to define the poverty level. For
more discussion of these issues, see Ravallion (1992) and Chesher (1998).
17 When we deflated nominal expenditures with the deflator that allows the
share of food to vary from year to year, the results changed little, yielding a
headcount poverty rate in 1998 of 25.6 % instead of 24.5%.
18 It is difficult to quantify this here, as the data do not include information on
types of crops cultivated. Furthermore, the prevalence of marginal farmers
who owned small plots of land makes the distinction of landowning farmers
and non-landowning agricultural workers less than clear cut. Still, another
study by Pritchett et al. (2000), who use the same data, finds that owning land
reduces vulnerability to poverty.
19 ‘Village-specific fixed effects’ are the effects of village characteristics that do
not change over time.
20 The coefficient of a dummy variable indicates distance from the reference
category. For more discussion of regression with dummy variables, see Berndt
(1991: ch. 5).
21 The Generalized Entropy index GE(α) is given by the expression:
GE(α ) =
1
1
α (1 − α ) n
n
∑ yi y
i =1
α
− 1 , α ≠ 0, α ≠ 1
Technically, GE(0) equals the standard deviation of ln(PCE), GE(1) is the Theil
index of inequality, and GE(2) is half the square of the coefficient of variation
of ln(PCE). The Atkinson index is given by the expression:
1
Y (ε )
A(ε ) = 1 − ede y , where Yede (ε ) =
n
n
∑ (yi )
i =1
1−ε
1
1−ε
, ε > 0, ε ≠ 1
For more details on these and other inequality indices, see Cowell (1995).
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
113
22 If we were to deflate nominal consumption expenditures with a common price
deflator such as the national consumer price index, the corresponding
inequality indices for each year would be identical to those obtained using
nominal consumption expenditures in each year. That is because inequality
indices are independent of the scale of the variable analysed. Since our price
deflator varies from household to household, our analysis of inequality is based
on the deflated PCE.
REFERENCES
Asra, A. (1999), ‘Urban–Rural Differences in Costs of Living and Their Impact on
Poverty Measures’, Bulletin of Indonesian Economic Studies 35 (3): 51–69.
Atkinson, A. (1970), ‘On the Measurement of Inequality’, Journal of Economic Theory
2: 244–63.
Berndt, E.R. (1991), The Practice of Econometrics: Classic and Contemporary, Addison
Wesley, Reading MA.
Chesher, A. (1998), Local Poverty Lines and Poverty Measures for Indonesia, Report
prepared for the World Bank, Department of Economics, University of Bristol,
Bristol (mimeo).
Cowell, F. (1995), Measuring Inequality, 2nd ed., LSE Handbooks in Economics Series,
Prentice Hall/Harvester WheatSheaf, London.
Deaton, A. (1997), The Analysis of Household Surveys: A Microeconometric Approach
to Development Policy, Johns Hopkins University Press, Baltimore MD.
Deaton, A., and S. Zaidi (1999), Guidelines for Constructing Consumption
Aggregates for Welfare Analysis, The World Bank, Washington DC (mimeo).
Foster, J., J. Greer and E. Thorbecke (1984), ‘A Class of Decomposable Poverty
Measures’, Econometrica 52: 761–6.
Frankenberg, E., D. Thomas and K. Beegle (1999), The Real Costs of Indonesia’s
Economic Crisis: Preliminary Findings from the Indonesian Family Life
Survey, July, RAND, Santa Monica CA (mimeo).
Levinsohn, J., S. Berry and J. Friedman (1999), ‘Impacts of the Indonesian Economic
Crisis: Price Changes and the Poor’, NBER Working Paper No. 7194, June,
National Bureau of Economic Research, Cambridge MA.
Molyneaux, J. (1999), Descriptive Statistics and Analysis of the First Two Rounds
of the 100 Village Survey, RAND, Santa Monica CA (mimeo).
Papanek, G.F., and B.S. Handoko (1999), The Impact on the Poor of Growth and
Crisis: Evidence from Real Wage Data, Paper presented at a Conference on
the Economic Issues Facing the New Government, LPEM–FEUI, August
18–19, Jakarta.
Poppele, J., S. Sumarto and L. Pritchett (1999), Social Impacts of the Indonesian Crisis:
New Data and Policy Implications, SMERU Report, February, Social Monitoring
and Early Response Unit, Jakarta.
Pritchett, L., A. Suryahadi and S. Sumarto (2000), ‘Quantifying Vulnerability to
Poverty: A Proposed Measure, with Application to Indonesia’, SMERU
Working Paper, January, Social Monitoring and Early Response Unit, Jakarta.
Ravallion, M. (1992), ‘Poverty Comparisons: A Guide to Concepts and Methods’,
LSMS Working Paper No. 88, The World Bank, Washington DC.
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
114
Skoufias, Suryahadi and Sumarto
Skoufias, E., A. Suryahadi and S. Sumarto (1999), ‘The Indonesian Crisis and Its
Impacts on Household Welfare, Poverty Transitions, and Inequality: Evidence
from Matched Households in the 100 Village Survey’, SMERU Working Paper,
September, Social Monitoring and Early Response Unit, Jakarta.
Sumarto, S., A. Wetterberg and L. Pritchett (1998), The Social Impact of the Crisis in
Indonesia: Results from a Nationwide Kecamatan Survey, SMERU Report,
December, Social Monitoring and Early Response Unit, Jakarta.
Suryahadi, A., and S. Sumarto (1999), ‘Update on the Impact of the Indonesian
Crisis on Consumption Expenditures and Poverty Incidence: Results from
the December 1998 Round of the 100 Village Survey’, SMERU Working Paper,
August, Social Monitoring and Early Response Unit, Jakarta.
Sutanto, A. (1999), The December 1998 Poverty in Indonesia: Some Findings and
Interpretations, Paper presented at the Round Table Discussion on the
Number of Poor People in Indonesia, July, National Development Planning
Agency (Bappenas), Jakarta.
Thomas, D., E. Frankenberg, K. Beegle and G. Teruel (1999), Household Budgets,
Household Composition and the Crisis in Indonesia: Evidence from
Longitudinal Household Survey Data, Paper prepared for the 1999 Population
Association of America Meetings, RAND, Santa Monica CA.
ISSN: 0007-4918 (Print) 1472-7234 (Online) Journal homepage: http://www.tandfonline.com/loi/cbie20
Changes in Household Welfare, Poverty and
Inequality During the Crisis
Emmanuel Skoufias & Asep Suryahadi
To cite this article: Emmanuel Skoufias & Asep Suryahadi (2000) Changes in Household
Welfare, Poverty and Inequality During the Crisis, Bulletin of Indonesian Economic Studies,
36:2, 97-114, DOI: 10.1080/00074910012331338903
To link to this article: http://dx.doi.org/10.1080/00074910012331338903
Published online: 18 Aug 2006.
Submit your article to this journal
Article views: 134
View related articles
Citing articles: 9 View citing articles
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=cbie20
Download by: [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RA JA ALI HA JI
TANJUNGPINANG, KEPULAUAN RIAU]
Date: 19 January 2016, At: 22:05
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Bulletin of Indonesian Economic Studies
Vol 36 No 2, August 2000, pp. 97–114
CHANGES IN HOUSEHOLD WELFARE, POVERTY AND
INEQUALITY DURING THE CRISIS
Emmanuel Skoufias
International Food Policy Research Institute (IFPRI), Washington DC
Asep Suryahadi and Sudarno Sumarto*
Social Monitoring and Early Response Unit (SMERU), Jakarta
This study provides evidence about changes in the distribution of living
standards among Indonesian households during the economic crisis. It
uses consumption expenditure data from a panel of households that were
surveyed in May 1997, just before the onset of the crisis, and then again in
August 1998, about a year after the crisis began. A household-specific
deflator is used to make nominal consumption expenditures comparable
across this period. The results suggest that there was a considerable drop
in household welfare during the economic crisis. Average per capita
expenditures fell significantly, and at the same time inequality increased.
The poverty rate also appears to have doubled from the pre-crisis level.
However, transitions into and out of poverty before and after the crisis
reveal remarkable fluidity.
INTRODUCTION
In this article, we present evidence about changes in Indonesian
household living standards—measured by per capita real consumption
expenditures—and in the distribution of living standards across
households—measured by indices of inequality—during the economic
crisis. Our study has two distinguishing characteristics.
First, it is based on a set of households that were surveyed in May
1997, just before the onset of the crisis, and then 14 months later in August
1998, about a year after the crisis began. The use of ‘panel’ data (a
combination of cross-section and time-series data on the same households)
offers the opportunity to identify how the welfare of specific households
changed during the crisis.1
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
98
Skoufias, Suryahadi and Sumarto
The second characteristic relates to the price deflator we use to make
nominal consumption expenditures comparable across years. Given the
large shifts in the relative prices of food and non-food items during the
crisis, different price deflators result in very different estimates of the
magnitude and severity of its impact.2 We adopt a household-specific
deflator that is a weighted average of the food and non-food price indices.
The weights applied to food and non-food prices vary from household
to household and are calculated from an ‘Engel curve’, which predicts
each household’s food share in consumption expenditure, based on the
household’s (logarithms of) per capita consumption and family size. We
believe that such a deflator is more appropriate than the standard deflator
for evaluating the impact of the economic crisis, since it captures more
accurately the impact of higher food prices on poorer households.
DATA AND KEY VARIABLES
Data
The data we use are part of the 100 village survey conducted by
Indonesia’s central statistics agency, BPS, and funded by UNICEF.3 The
purpose of this survey is to monitor changes in health, education, nutrition
and socio-economic status in 100 villages purposively selected from 10
districts (kabupaten) in 8 provinces throughout Indonesia. In each village,
120 households were chosen—giving a total sample of 12,000
households—and information was collected for the household and all
family members about factors such as education, employment sector and
type of work. Although the sample is large in terms of number of
households, and represents a variety of areas across the country, it is
important to note that the selection of villages was not random. Hence,
readers should keep in mind that the findings of this study apply only to
this sample.
The data analysed here are from the 100 village survey rounds in
May 1997 and August 1998. Our analysis relies exclusively on a panel of
households that were interviewed in both rounds. In the August 1998
round, the sampling frame was changed from two enumeration areas of
60 households each to the original two plus a third enumeration area,
each with 40 households. Of the 120 households from the two
enumeration areas that were the same in both rounds, 80 were targeted
for re-interview. Unfortunately, in the second round the identifying codes
for households were changed. We therefore identified the households
within each enumeration area using the name of the household head,
and then cross-checked the matches using the demographic characteristics
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
99
of the households. This resulted in the matching of 8,141 (68%) of the
households across the two rounds, implying that, of the 120 households
in each village, an average of 82 were actually followed across rounds.4
Measuring Household Welfare
We use per capita consumption expenditures (PCE) as one indicator of
household living standard.5 Of course, consumption expenditure is only
one of many components of household welfare. Others include
employment, health conditions, and the ability to access and use basic
services such as water, sanitation, health care and education. We examine
the changes in household welfare by using household PCE in each year
to calculate a variety of poverty and inequality indices. As discussed in
detail in Deaton (1997) and Deaton and Zaidi (1999), the social welfare
function approach developed by Atkinson (1970) provides a useful
framework for interpreting statistical measures of poverty and inequality.
For example, if we were to describe social welfare in period t, W(t), as a
function of the PCE of all the households in the population in period t,
i.e.:
(
)
W (t) = W PCE1 (t), PCE2 (t),..., PCEK (t)
(1)
where K is the number of households in the population, then with a set
of relatively innocuous assumptions about the properties of the function
W,6 we may express social welfare in period t as:
(
)
W (t) = PCE(t) 1 − I (t)
(2)
i.e. as a function of the mean level of PCE in period t, denoted by the bar
over PCE(t), multiplied by one minus the level of inequality in the
distribution of PCE in period t (denoted by I(t)).7 Along similar lines, the
common indices of poverty (described in more detail below) can also
conceivably be regarded as particular forms of the social welfare function.
Construction of Key Variables
Per capita expenditures are defined as PCE(t) = C(t)/N(t), where C(t)
denotes deflated food and non-food consumption expenditures in period
t (see below for details on the deflators used) and N(t) denotes total family
size in period t.8 In various instances we also look at PCE for food and
non-food separately. Food expenditure is the sum of expenditures on
grains, meat, fish, eggs and milk, vegetables, beans and nuts, fruits,
seasonings, fats and oils, soft drinks, prepared food and other food items,
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
100
Skoufias, Suryahadi and Sumarto
and alcohol and tobacco.9 The reference period for expenditures on these
items was the week preceding the day of the interview. These weekly
expenditures were transformed into monthly expenditures by
multiplying by (30/7).
For non-food expenditures two measures were collected, each for a
different reference period—the previous month and the previous 12
months. To minimise recall errors, but at the risk of introducing exclusion
errors, we used the expenditures reported for the previous month. Nonfood expenditure is defined as the sum of expenditures on housing, health,
education, clothing and shoes, durable goods, taxes and insurance,
ceremonies, and other expenses.
It is important to take note of two qualifications about the data. First,
the surveys were conducted in different months of the calendar year (May
in 1997 and August in 1998), thus introducing the possibility that some
of the observed changes in consumption may be due to seasonality. This
is particularly true of items such as education and clothing expenditures,
which are influenced by the educational calendar. Second, although our
sample consists primarily of households in rural areas, there are some
households or villages that are classified as being in ‘urban’ areas (17.8%
of the sample). The reader is cautioned that villages in urban areas in our
sample are not part of large metropolitan agglomerations (such as
Jabotabek, Surabaya and Medan), but are villages that are close to the
district capitals. Such villages are classified administratively as kelurahan
instead of desa, and coded as urban areas in our sample.
Deflating Expenditures
The nominal consumption expenditures in the two rounds of the survey
need to be adjusted in order to allow meaningful comparisons about
household welfare across the two rounds. For example, because of the
large increases in the price of rice during the economic crisis, an
expenditure of Rp 10,000 on rice during August 1998 represents a much
smaller quantity of rice than the same expenditure in May 1997.
To control for the large price differences across the rounds we construct
a Laspeyres price index using the following steps. First, we construct a
deflator for food and non-food items using the mean shares of the food
and non-food items in the May 1997 survey as weights, and the price
indices published in the BPS monthly statistical bulletin Indikator Ekonomi
in May 1997 and in August 1998.10 We have not used region-specific
deflators for food or non-food items because the regional deflators
available in Indonesia are based explicitly on urban prices, so any crossregional comparisons should be made with caution.11
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
101
Second, we construct a household-specific deflator that is a weighted
average of the food and non-food price indices calculated above.
Specifically, if we denote by t the periods May 1997 and August 1998,
and the price deflators for food and non-food in period t by PF(t) and
PNF(t) respectively, then the price deflator for period t for household h,
Ph(t) can be expressed as:
(
)
P h (t) = W√Fh (97 )PF (t) + 1 − W√Fh (97 ) PNF (t)
(3)
The weights applied to food and non-food items vary from household
to household. The weight for each household was calculated from the
predicted value of the regression of household food share in May 1997,
W√Fh (97 ) on the logarithm of per capita consumption, ln( PCE(97 )) , and the
logarithm of household size.12 In this manner the influence of householdspecific unobserved components or tastes on the share of food is
eliminated.
As is the case for all Laspeyres price deflators, the share of food is
assumed to be constant. To the extent that the changes in relative prices
are such that the share of food also increases as a result of the crisis (as
indicated by the data), then the above deflator may be underestimating
the increases in prices. In an effort to check for this possibility, we also
constructed another deflator with variable weights for food based on the
coefficients from an Engel curve estimated separately for May 1997 and
for August 1998. However, the changes in the results obtained using the
deflators with fixed and varying food shares were very small, so we
choose to present only the results obtained using the deflator based on a
fixed food share.
WELFARE CHANGES DURING THE CRISIS
Distribution of Per Capita Expenditure
We begin with a graph that provides a quick visual impression of changes
in household consumption expenditures between May 1997 and August
1998. Figure 1 graphs the cumulative distribution functions (CDF) of
ln(PCE) in both periods. The figure shows that the 1998 CDF lies to the
left of the 1997 CDF with no ‘crossings’, implying that the 1998 CDF
‘stochastically dominates’ the 1997 CDF.13 This means that the poverty
rate will be higher in 1998 no matter what poverty line is chosen (Deaton
1997).14
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
102
Skoufias, Suryahadi and Sumarto
FIGURE 1 Cumulative Distribution Functions of ln(PCE)
Cumulative
distribution
1.0
0.8
1997
0.6
0.4
1998
0.2
0.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
ln(PCE)
Moreover, the shift to the left in the CDF between 1997 and 1998 was
not exactly parallel, with the lower part of the CDF shifting more to the
left than the upper part. For example, the respective implied falls in PCE
for the 1st, 5th, 10th and 20th percentiles are 41, 27, 22 and 20%, while for
the 80th, 90th, 95th and 99th percentiles they are 14, 12, 10 and 10%. This
means that the fall in consumption expenditures was greater for those at
the lower than at the upper end of the distribution, indicating a worsening
of the distribution and an increase in inequality.15
Poverty Rates
There are various methods of calculating a poverty line, and they can
produce widely differing results. Moreover, at any given point in time
the level of poverty reported is quite sensitive to the poverty line estimate
used. Equally reasonable poverty lines can produce very different poverty
rates for the same data. In this instance we are interested principally in
the changes in poverty over time. Hence, to match the official pre-crisis
poverty rate, we chose as the poverty line in 1997 the 11th percentile of
the distribution of ln(PCE) in the full sample of 12,000 households (not
just the matched sample).16 In other words, the poverty line was chosen
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
103
to produce an 11% poverty rate in the full sample. With this poverty line,
the poverty rate in our matched sample of 8,141 households in May 1997
turned out to be 12.4%. From that level we can calculate changes that are,
while not invariant, robust to the initial assumed level of poverty.
In table 1 we report the values of the Foster–Greer–Thorbecke (FGT)
poverty indices (Foster et al. 1984). This class of poverty measures is highly
regarded because it meets all the axioms desirable in consumption-based
poverty measures, and contains a parameter α that can be set to generate
results showing varying levels of sensitivity to income distribution among
the poor. Specifically, the FGT family of poverty measures is summarised
by the formula:
1
P(α ) =
N
q
z − ci
z
∑
i =1
α
(4)
where N is the number of households, ci is the per capita consumption
(or income) of the ith household, z is the poverty line, q is the number of
poor households, and α is the weight attached to the severity of household
poverty (or the distance from the poverty line). When α = 0, the FGT
measure collapses to the Headcount Index, or P(0), i.e. the proportion of
the population that is below the poverty line. This measure, while useful
for general poverty comparisons, is insensitive to differences in the depth
of poverty, in the sense that households far below the poverty line receive
the same weight as households just below the poverty line. Moreover, as
Deaton (1997) points out, it serves as an unsatisfactory indicator of
welfare, for it is possible for this measure to indicate a decline in headcount
poverty when some very poor households become even poorer and some
not so poor households’ expenditures increase sufficiently to push these
households above the poverty line.
This shortcoming is overcome by assigning higher values to the
parameter α. When α = 1, the FGT measure gives the Poverty Gap, or
P(1), a measure of the average depth of poverty, and indicates the average
TABLE 1 Foster–Greer–Thorbecke (FGT) Poverty Indicesa
1997
1998
Percentage changeb
P(0)
P(1)
P(2)
0.124
0.245
98
0.023
0.060
163
0.006
0.023
259
a
See text for explanation of P(0), P(1) and P(2).
b
Based on original unrounded numbers.
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
104
Skoufias, Suryahadi and Sumarto
money gap by which the consumption of the poor falls short of the poverty
line. When α = 2, the FGT index is called the Severity of Poverty index, or
P(2). This measure differs from P(1) in that it assigns relatively more
weight than P(1) to individuals whose expenditures are further away
from the poverty line and who are thus in more severe poverty.
Based on the poverty line in 1997, the poverty rate (Headcount Index)
doubled in our panel of households from 12.4% in May 1997 to 24.5% in
August 1998.17 Although this rate in 1998 is remarkably close to the rural
area poverty rate of 25.7% estimated by BPS (Sutanto 1999), the two rates
are not strictly comparable because they are derived by very different
methods.
The higher order poverty indices also increased, and by a factor higher
than the increase in the Headcount Index. For example, the Poverty Gap
index rose from 0.023 to 0.06, which means that the average poverty gap
increased from 2.3% to 6% of the poverty line. Meanwhile, the Poverty
Severity Index increased by a factor of almost 4, i.e. from 0.006 to 0.023.
Poverty Transitions
In table 2 we present a poverty transition matrix. We classify households
into one of four categories based on the relationship between their PCE
and the poverty line (PL): poor (PCE < PL); near poor, above the poverty
line but by less than 25% (PL ≤ PCE < 1.25*PL); near non-poor, more than
25% but less than 50% above poverty line (1.25*PL ≤ PCE < 1.5*PL); and
non-poor, 50% or more above poverty line (PCE≥ 1.5*PL). This allows us
to examine both how those in poverty in 1997 fared and who moved into
poverty in 1998. The first column shows the distribution of households
in 1997, indicating that 1,010 households were poor and 5,029 non-poor.
The first row shows the distribution of households across these categories
in 1998, with 1,997 poor households and 3,562 non-poor.
For each row, the columns show how households in that category in
1997 fared in 1998. For example, take the 988 households that were near
poor in 1997. In 1998 only 239 of these households (24.2%) were ‘on the
diagonal’ or in the same category of near poor. Among the rest, 309 (i.e.
140 + 169) or 31.3% of households had improved their economic status,
while 440 (44.5%) had fallen into poverty. Similarly by looking down the
columns one can see where the households in any category in 1998 were
in 1997. So, for instance, of the 3,562 households that were non-poor in
1998, 3,029 (85.0%) were also non-poor in 1997, while only 58 households
(1.6%) had come from being poor in 1997. Meanwhile, the bottom number
in each set gives the percentage of the total households. So 8.6% of the
population was poor in both periods, while 524 households (6.4% of 8,141)
were non-poor in 1997 but became poor in 1998.
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
105
In terms of percentages, only 10.4% of those non-poor in 1997 had
become poor in 1998. However, since the non-poor were 61.8% of the
1997 population, they are 26.2% of the poor in 1998. On the other hand,
even though 44.5% of the near poor in 1997 had become poor in 1998, as
only 12.1% of the 1997 population was near poor, only 22.0% of the poor
in 1998 came from the near poor category in 1997.
Although during the crisis many of the households that were
marginally poor before the crisis became impoverished, the transition
matrix reveals considerable fluidity. Approximately 31% of the poor in
TABLE 2 Poverty Transition Matrix
Poverty Status in 1998a
Total 1997
Poor
Near Poor
Total 1998
Row percentage
Column percentage
Total percentage
8,141
100.00
100.00
100.00
1,997
24.53
100.00
24.53
1,369
16.82
100.00
16.82
1,213
14.90
100.00
14.90
3,562
43.75
100.00
43.75
Poverty Status in 1997
Poor
Row percentage
Column percentage
Total percentage
1,010
100.00
12.41
12.41
697
69.01
34.90
8.56
177
17.52
12.93
2.17
78
7.72
6.43
0.96
58
5.74
1.63
0.71
Near poor
Row percentage
Column percentage
Total percentage
988
100.00
12.14
12.14
440
44.53
22.03
5.40
239
24.19
17.46
2.94
140
14.17
11.54
1.72
169
17.11
4.74
2.08
Near non-poor
Row percentage
Column percentage
Total percentage
1,114
100.00
13.68
13.68
336
30.16
16.83
4.13
282
25.31
20.60
3.46
190
17.06
15.66
2.33
306
27.47
8.59
3.76
Non-poor
Row percentage
Column percentage
Total percentage
5,029
100.00
61.77
61.77
524
10.42
26.24
6.44
671
13.34
49.01
8.24
805
16.01
66.36
9.89
3,029
60.23
85.04
37.21
a
Poor: PCE < PL (poverty line)
Near non-poor: 1.25*PL ≤ PCE
−∞
∫ g(x)dx
∀α .
−∞
14 This is true because if we draw any vertical line on the graph, it will always
cross the 1998 CDF at a higher distribution value than the 1997 CDF.
15 However, when the analysis is disaggregated into urban and rural areas, there
is some indication that in urban areas the falls in consumption were greater at
the upper end of the distribution. For more details on this, see the longer
version of this paper in Skoufias et al. (1999).
16 The alternative is to use the standard approach to estimating a poverty line.
However, this ‘standard’ approach is not free from questionable assumptions
about the composition of the food bundle, the reference population, or the
minimum level of calorie availability used to define the poverty level. For
more discussion of these issues, see Ravallion (1992) and Chesher (1998).
17 When we deflated nominal expenditures with the deflator that allows the
share of food to vary from year to year, the results changed little, yielding a
headcount poverty rate in 1998 of 25.6 % instead of 24.5%.
18 It is difficult to quantify this here, as the data do not include information on
types of crops cultivated. Furthermore, the prevalence of marginal farmers
who owned small plots of land makes the distinction of landowning farmers
and non-landowning agricultural workers less than clear cut. Still, another
study by Pritchett et al. (2000), who use the same data, finds that owning land
reduces vulnerability to poverty.
19 ‘Village-specific fixed effects’ are the effects of village characteristics that do
not change over time.
20 The coefficient of a dummy variable indicates distance from the reference
category. For more discussion of regression with dummy variables, see Berndt
(1991: ch. 5).
21 The Generalized Entropy index GE(α) is given by the expression:
GE(α ) =
1
1
α (1 − α ) n
n
∑ yi y
i =1
α
− 1 , α ≠ 0, α ≠ 1
Technically, GE(0) equals the standard deviation of ln(PCE), GE(1) is the Theil
index of inequality, and GE(2) is half the square of the coefficient of variation
of ln(PCE). The Atkinson index is given by the expression:
1
Y (ε )
A(ε ) = 1 − ede y , where Yede (ε ) =
n
n
∑ (yi )
i =1
1−ε
1
1−ε
, ε > 0, ε ≠ 1
For more details on these and other inequality indices, see Cowell (1995).
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
Changes in Household Welfare, Poverty and Inequality during the Crisis
113
22 If we were to deflate nominal consumption expenditures with a common price
deflator such as the national consumer price index, the corresponding
inequality indices for each year would be identical to those obtained using
nominal consumption expenditures in each year. That is because inequality
indices are independent of the scale of the variable analysed. Since our price
deflator varies from household to household, our analysis of inequality is based
on the deflated PCE.
REFERENCES
Asra, A. (1999), ‘Urban–Rural Differences in Costs of Living and Their Impact on
Poverty Measures’, Bulletin of Indonesian Economic Studies 35 (3): 51–69.
Atkinson, A. (1970), ‘On the Measurement of Inequality’, Journal of Economic Theory
2: 244–63.
Berndt, E.R. (1991), The Practice of Econometrics: Classic and Contemporary, Addison
Wesley, Reading MA.
Chesher, A. (1998), Local Poverty Lines and Poverty Measures for Indonesia, Report
prepared for the World Bank, Department of Economics, University of Bristol,
Bristol (mimeo).
Cowell, F. (1995), Measuring Inequality, 2nd ed., LSE Handbooks in Economics Series,
Prentice Hall/Harvester WheatSheaf, London.
Deaton, A. (1997), The Analysis of Household Surveys: A Microeconometric Approach
to Development Policy, Johns Hopkins University Press, Baltimore MD.
Deaton, A., and S. Zaidi (1999), Guidelines for Constructing Consumption
Aggregates for Welfare Analysis, The World Bank, Washington DC (mimeo).
Foster, J., J. Greer and E. Thorbecke (1984), ‘A Class of Decomposable Poverty
Measures’, Econometrica 52: 761–6.
Frankenberg, E., D. Thomas and K. Beegle (1999), The Real Costs of Indonesia’s
Economic Crisis: Preliminary Findings from the Indonesian Family Life
Survey, July, RAND, Santa Monica CA (mimeo).
Levinsohn, J., S. Berry and J. Friedman (1999), ‘Impacts of the Indonesian Economic
Crisis: Price Changes and the Poor’, NBER Working Paper No. 7194, June,
National Bureau of Economic Research, Cambridge MA.
Molyneaux, J. (1999), Descriptive Statistics and Analysis of the First Two Rounds
of the 100 Village Survey, RAND, Santa Monica CA (mimeo).
Papanek, G.F., and B.S. Handoko (1999), The Impact on the Poor of Growth and
Crisis: Evidence from Real Wage Data, Paper presented at a Conference on
the Economic Issues Facing the New Government, LPEM–FEUI, August
18–19, Jakarta.
Poppele, J., S. Sumarto and L. Pritchett (1999), Social Impacts of the Indonesian Crisis:
New Data and Policy Implications, SMERU Report, February, Social Monitoring
and Early Response Unit, Jakarta.
Pritchett, L., A. Suryahadi and S. Sumarto (2000), ‘Quantifying Vulnerability to
Poverty: A Proposed Measure, with Application to Indonesia’, SMERU
Working Paper, January, Social Monitoring and Early Response Unit, Jakarta.
Ravallion, M. (1992), ‘Poverty Comparisons: A Guide to Concepts and Methods’,
LSMS Working Paper No. 88, The World Bank, Washington DC.
y [Universitas Maritim Raja Ali Haji], [UNIVERSITAS MARITIM RAJA ALI HAJI TANJUNGPINANG, KEPULAUAN RIAU] at 22:05 19
114
Skoufias, Suryahadi and Sumarto
Skoufias, E., A. Suryahadi and S. Sumarto (1999), ‘The Indonesian Crisis and Its
Impacts on Household Welfare, Poverty Transitions, and Inequality: Evidence
from Matched Households in the 100 Village Survey’, SMERU Working Paper,
September, Social Monitoring and Early Response Unit, Jakarta.
Sumarto, S., A. Wetterberg and L. Pritchett (1998), The Social Impact of the Crisis in
Indonesia: Results from a Nationwide Kecamatan Survey, SMERU Report,
December, Social Monitoring and Early Response Unit, Jakarta.
Suryahadi, A., and S. Sumarto (1999), ‘Update on the Impact of the Indonesian
Crisis on Consumption Expenditures and Poverty Incidence: Results from
the December 1998 Round of the 100 Village Survey’, SMERU Working Paper,
August, Social Monitoring and Early Response Unit, Jakarta.
Sutanto, A. (1999), The December 1998 Poverty in Indonesia: Some Findings and
Interpretations, Paper presented at the Round Table Discussion on the
Number of Poor People in Indonesia, July, National Development Planning
Agency (Bappenas), Jakarta.
Thomas, D., E. Frankenberg, K. Beegle and G. Teruel (1999), Household Budgets,
Household Composition and the Crisis in Indonesia: Evidence from
Longitudinal Household Survey Data, Paper prepared for the 1999 Population
Association of America Meetings, RAND, Santa Monica CA.