Bulletin of Indonesian Economic Studies Vol 36 No 2, August 2000, pp.

95112

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 and transitions into and out
of quartiles of consumption expenditure before and after the crisis reveal
remarkable fluidity.

INTRODUCTION

In this article, we present evidence about changes in Indonesian

household living standardsmeasured by per capita real consumption
expendituresand in the distribution of living standards across
householdsmeasured by indices of inequalityduring the economic
crisis. Our study has two distinguishing characteristics.
First, it is based on a set of households that were surveyed in May
Is the 1997, just before the onset of the crisis, and then 14 months later in August
explanation 1998, about a year after the crisis began. The use of panel data (a
of panel combination of cross-section and time-series data) offers the opportunity
data OK? to identify how the welfare of specific households changed during the
crisis.1
96 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 households food share in consumption expenditure, based on the
households (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
Indonesias Central Statistics Agency (BPS, Badan Pusat Statistik) 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 householdsand 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
 Changes in Household Welfare, Poverty and Inequality during the Crisis 97

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.:

Please ( )
W (t) = W PCE1 (t), PCE2 (t),..., PCEK (t) (1)
check
where K is the number of households in the population, then with a set
equations
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,
98 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 periodthe 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. Non-
food 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 cross
regional comparisons should be made with caution.11
Changes in Household Welfare, Poverty and Inequality during the Crisis 99

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) = WFh (97 )PF (t) + 1 WFh (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,
WFh (97 ) on the logarithm of per capita consumption, ln( PCE(97 )) , and the
logarithm of household size.12 In this manner the influence of household-
specific 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
100 Skoufias, Suryahadi and Sumarto

FIGURE 1 Cumulative Distribution Functions of ln(PCE)

Cumulative
distribution
1.0

0.8

0.6 1997

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
 Changes in Household Welfare, Poverty and Inequality during the Crisis 101

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 FosterGreerThorbecke (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 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:
q
1 z ci
P( ) =
N z
(4)
i =1

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 is the weight attached to the severity of household
poverty (or the distance from the poverty line). When = 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 . When = 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 FosterGreerThorbecke (FGT) Poverty Indicesa

P(0) P(1) P(2)

Please check 1997 0.124 0.023 0.006

% change for 1998 0.245 0.060 0.023
P(1) and P(2). Percentage change 98 163 259
(161 and
283?) a
See text for explanation of P(0), P(1) and P(2).
 102 Skoufias, Suryahadi and Sumarto

money gap by which the consumption of the poor falls short of the poverty
line. When = 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
Slight are not strictly comparable because they are derived by very different
wording methods. What are comparable are the poverty rates of the pre-crisis 1997
change survey and the survey conducted in 1998 during the crisis.
here. 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.
Changes in Household Welfare, Poverty and Inequality during the Crisis 103

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 Near Non-Poor

Non-Poor

Total 1998 8,141 1,997 1,369 1,213 3,562

Row percentage 100.00 24.53 16.82 14.90 43.75
Column percentage 100.00 100.00 100.00 100.00 100.00
Total percentage 100.00 24.53 16.82 14.90 43.75
Poverty Status in 1997
Poor 1,010 697 177 78 58
Row percentage 100.00 69.01 17.52 7.72 5.74
Column percentage 12.41 34.90 12.93 6.43 1.63
Total percentage 12.41 8.56 2.17 0.96 0.71
Near poor 988 440 239 140 169
Row percentage 100.00 44.53 24.19 14.17 17.11
Column percentage 12.14 22.03 17.46 11.54 4.74
Total percentage 12.14 5.40 2.94 1.72 2.08
Near non-poor 1,114 336 282 190 306
Row percentage 100.00 30.16 25.31 17.06 27.47
Column percentage 13.68 16.83 20.60 15.66 8.59
Total percentage 13.68 4.13 3.46 2.33 3.76
Non-poor 5,029 524 671 805 3,029
Row percentage 100.00 10.42 13.34 16.01 60.23
Column percentage 61.77 26.24 49.01 66.36 85.04
Total percentage 61.77 6.44 8.24 9.89 37.21

a
Poor: PCE < PL (poverty line) Near poor: PL PCE < 1.25*PL
Near non-poor: 1.25*PL PCE <1.5*PL Non-poor: PCE 1.5*PL
104 Skoufias, Suryahadi and Sumarto

1997 had moved out of poverty in 1998, although mainly to the near
poor category (17.5%). Also, 44.5% of the near poor in 1997 had become
poor in 1998, but there were 17% that had managed to become non-poor.
More surprisingly, almost 17% of the poor households in 1998 were near
non-poor and more than a quarter (26.2%) were non-poor in 1997. These
are the households that in 1997 had expenditures that were more than 25
and 50% above the poverty line respectively. Only 35% of the poor in
1998 are those who were also poor in 1997. This implies that reaching
the poor in 1998 will be difficult, as many families who otherwise would
not have been at all poor have suffered large reversals in fortune during
the crisis and become poor.
Thus the first impressions created by the shift to the left of the CDF of
ln(PCE) in figure 1 miss a large part of the story. That is, while those
classified as poor in 1998 were poorer than those classified as poor in
1997, it is not simply that poor households got poorer. Many of the
households classified as poor in 1998 were entering newly into poverty,
in many instances replacing previously poor households that had moved
out of poverty. A possible explanation for this is that in rural areas the
crisis is more likely to have had a negative effect on those without land
or with little land and relying primarily on wage income, as the data on
wages suggest very large falls in real wages (Papanek and Handoko 1999).
In contrast, the real incomes of producers of some food and export crops
in rural areas were likely to remain unchanged, if not to increase, as they
benefited from relative price shifts favouring these commodities.18 In
urban areas, evidence from other studiesparticularly the Indonesian
Family Life Survey known as IFLS2, and particularly in provinces on
Java (Frankenberg et al. 1999)suggests the shock has affected the
relatively well-off.

Household Correlates of Transitions

What household characteristics are associated with the patterns of change
in per capita consumption observed during the crisis? To examine the
household covariates of the change in ln(PCE) between the two rounds,
we have estimated a number of exploratory regressions, reported in table
3. Column (A) in this table contains the estimates obtained from regressing
the change in ln(PCE) on a set of household and head of household
characteristics in 1997, such as family size, age of the head, education
level of the head, sector of employment and type of work of the head,
detailed age and gender composition of the household, and some
variables characterising the geographic location of the household.
Meanwhile, columns (B) and (C) control for village-specific fixed effects.19
In column (C), the dependent variable is the difference in nominal ln(PCE),
 Changes in Household Welfare, Poverty and Inequality during the Crisis 105

TABLE 3 Correlates of the Change in ln(PCE)

(dependent variable: ln(PCE98)ln(PCE97))

No Fixed Effects Village-Specific Fixed Effects

Variablea (A)b (B)b (C)b
Coeff. t-value Coeff. t-value Coeff. t-value

Log family size in 1997 0.113 3.24 0.134 4.24 0.138 4.39
Age of household head in 1997 0.003 1.25 0.001 0.41 0.001 0.36
Age squared 0.004 1.49 0.000 0.03 0.000 0.07
Heads education level in 1997
Primary 0.045 0.93 0.038 0.87 0.042 0.95
Junior high school 0.015 0.31 0.021 0.50 0.024 0.57
Senior high school 0.011 0.23 0.019 0.44 0.021 0.50
Vocational senior high school 0.064 1.34 0.068 1.56 0.069 1.60
Tertiary 0.055 1.13 0.038 0.88 0.037 0.86
Heads employment sector in 1997
Mining 0.058 0.99 0.061 1.16 0.060 1.15
Manufacturing 0.086 3.33 0.051 2.15 0.051 2.14
Utilities 0.140 1.21 0.107 1.02 0.105 1.01
Construction 0.051 1.73 0.021 0.76 0.021 0.76
Trade 0.042 2.21 0.023 1.32 0.024 1.39
Transport & communication 0.063 2.30 0.040 1.61 0.041 1.65
Financing 0.041 0.42 0.138 1.56 0.138 1.58
Services 0.062 2.39 0.046 1.94 0.046 1.96
Heads type of work
Self-employed with family help 0.042 3.57 0.024 2.13 0.024 2.13
Employer 0.068 1.07 0.055 0.96 0.054 0.95
Government employee 0.021 0.60 0.025 0.78 0.024 0.76
SOE employeec 0.017 0.23 0.012 0.18 0.011 0.16
Private employee 0.012 0.61 0.019 1.09 0.019 1.10
Unpaid family worker 0.090 1.75 0.054 1.16 0.053 1.1
Gender & age composition
dummies Yes Yes Yes
Geographic dummies Yes No No
District dummies Yes No No
Constant 0.242 3.27 0.445 6.90 0.310 4.85
Is it clear why Number of observations 8,140 8,140 8,140
athe F statistic F(46, 8,093) 18.33 4.92 5.2
has these R2 0.094 0.2788 0.2795
numbers in
a
brackets? Excluded dummy variables (reference category): no schooling for education;
agriculture for employment sector; self-employed without any family help for
type of work.
b
See text for explanation of (A), (B) and (C).
c
SOE = state-owned enterprise.
106 Skoufias, Suryahadi and Sumarto

whereas in columns (A) and (B) it is the difference in real ln(PCE). The
purpose of the specification in (C) is to allow inflation rates to vary across
villages. The results of the three specifications do not differ greatly.
Briefly, the results provide a weak indication that households with a
tertiary educated head experienced greater falls in consumption.
However, the negative coefficient is not statistically significant at
conventional significance levels. Households with heads working in the
manufacturing, transport and communication, and services sectors in
1997 seem to have experienced larger falls in consumption than those
whose heads were employed in agriculture (the reference employment
sector).20 In contrast, households with a head who is self-employed and
working with help from paid or unpaid family workers seem to have
experienced significantly higher consumption growth than those of the
reference category (table 3).

Inequality
In order to examine the impact of the crisis on the distribution of living
standards across households we have also calculated the values of a
variety of inequality indices, such as the Generalised Entropy class of
indices (denoted by GE()), the Gini index, and the Atkinson index
(denoted by A()). The GE() and A() indices offer the advantage of
being more sensitive to differences in different parts of the expenditure
distribution, depending on the value of the sensitivity parameters and
. For example, the larger is, the more sensitive GE() is to consumption
differences at the top of the distribution; and the more negative is, the
more sensitive GE() is to differences at the bottom of the distribution.
Along similar lines, in Atkinsons index of inequality, the larger (known
as the inequality aversion parameter) is, the more sensitive A() is to
income differences at the bottom of the distribution.21
From the viewpoint of the social welfare function framework, the
Generalised Entropy and Atkinson indices of inequality are quite similar
to the Severity of Poverty index P(2) when greater emphasis is placed on
the lower end of the distribution (e.g. for = 1 and = 2). In this case,
the implicit social marginal utility is higher for households with
expenditures lower than the mean. Along the same lines, the social
marginal utility implied by P(2) is higher for households whose
expenditures are farther below the poverty line. The main difference is
that the inequality indices assign a welfare weight to all households in
the sample, whereas P(2) assigns a zero welfare weight to all households
above the poverty line.
Changes in Household Welfare, Poverty and Inequality during the Crisis 107

Another advantage offered by the GE() and A() inequality indices

is that both are additively decomposable into within-group and between-
group inequality. This allows us to examine whether consumption
inequality changed differently within and across (or between) urban areas
and rural areas or within and between districts. Such a decomposition is
not possible for the Gini index, although this is the measure more
commonly used as an index of inequality.22
In table 4 we report, for each of the two rounds of the survey, the
values of the GE() index for selected values of (1, 0, 1 and 2); the
values of the Gini coefficients of inequality; and the values of the Atkinson
index for = 0.5, 1 and 2. As can be seen from the table, inequality in real
expenditures increases substantially according to the GE(1) (23%) and
A(2) (17%) measures. Both of these measures are more sensitive to
consumption differences at the bottom of the distribution. The Gini
coefficient, meanwhile, increases by only 7%. Inequality using the GE(2)
measure, however, falls by 7%.
These findings appear to reflect two phenomena during the crisis.
First, inequality in rural areas increased, as the nominal wages of rural
wage earners did not keep pace with price inflation of their consumption
basket. While real wages fell, some rural net producers benefited from

TABLE 4 Generalised Entropy, Gini and Atkinson Indices of Inequality

GE(1) GE(0) GE(1) GE(2) Gini A(0.5) A(1) A(2)

1997 0.140 0.133 0.152 0.245 0.283 0.068 0.125 0.219

1998 0.173 0.154 0.166 0.228 0.304 0.076 0.143 0.257
% change 23 16 9 7 7 13 15 17
By district
Within-group
1997 0.119 0.112 0.131 0.223 0.057 0.104 0.179
1998 0.141 0.123 0.134 0.195 0.061 0.111 0.192
% change 18 9 3 13 6 7 7
Between-group
1997 0.021 0.021 0.021 0.022 0.011 0.023 0.049
1998 0.032 0.031 0.032 0.033 0.017 0.036 0.080
% change 50 50 50 53 52 54 62
108 Skoufias, Suryahadi and Sumarto

the depreciation and relative price shift. This is consistent with the
findings of Sutanto (1999), which show increasing P(2) indices, potentially
indicating rising inequality among the poor. Second, even though this
sample does not capture the major metropolitan areas, it does provide
an indication of some large falls in the expenditure of the rich in urbanised
areas. Thus, inequality as measured by indices sensitive to the upper
taile.g. GE(2)appears to be falling.
The increase in inequality in the sample was accompanied by an
increase in both within-group and between-group inequality. What is
more interesting, however, is that the growth in inequality between
districts was proportionately higher than the growth in inequality within
districts. Thus, inequalities in mean consumption across districts that
were present before the crisis were reinforced as a result of it.
We have also recalculated all of the inequality indices reported in
table 4 using nominal instead of real consumption expenditures. We still
found that inequality was higher in 1998 than in 1997, but the proportional
increase from the 1997 level was smaller. Though suggestive, these results
indicate that using nominal consumption expenditures or using deflators
that vary only across regions but not across households is likely to
underestimate the changes in inequality in Indonesia during the crisis.

CONCLUDING REMARKS

The findings of this study suggest that there was a considerable drop in
the welfare of Indonesian households during the first year of the economic
crisis. Average per capita expenditures declined significantly, and at the
same time inequality increased. The poverty rate appears to have doubled
in the first year of the crisis. However, transitions into and out of poverty
reveal remarkable fluidity. Many households have indeed suffered from
falling incomes and entered into poverty, but some appear to have
experienced improvements in their welfare. This implies that reaching
the poor after the crisis will be difficult, as many families who otherwise
would not have been poor at all have suffered large reversals in fortune
during the crisis and have entered into poverty.
Changes in Household Welfare, Poverty and Inequality during the Crisis 109

NOTES

* We would like to thank Lant Pritchett and two referees for valuable comments
and suggestions, Yusuf Suharso for research assistance, and BPS and UNICEF
for providing access to the data. The remaining errors and weaknesses,
however, are solely ours.
1 For other studies of social changes during the Indonesian crisis, see
Frankenberg et al. (1999); Levinsohn et al. (1999); Papanek and Handoko (1999);
Poppele et al. (1999); and Sumarto et al. (1998).
2 This is shown by Suryahadi and Sumarto (1999) and Thomas et al. (1999).
Asra (1999) also shows that headcount poverty estimates in Indonesia are
sensitive to the choice of inflation rates and cost of living differences between
urban and rural areas.
3 See Suryahadi and Sumarto (1999) for a more detailed description of the 100
village survey, and Molyneaux (1999) for an analysis of its first two rounds of
data.
4 This figure slightly exceeds the target number of re-interviews, a result
probably due to variations in the rules actually applied in the field.
5 Consumption expenditures here include those on items purchased at the
market as well as imputed values of own production consumed by the
household, and gifts. For further details, see Skoufias et al. (1999).
6 These assumptions include W being non-decreasing in each of its arguments,
symmetric and quasi-concave, and homogeneous of degree one.
7 Note that this formulation allows for the possibility that social welfare may
increase from one period to another even if PCE falls, as long as inequality in
the distribution of PCE falls by more than mean PCE.
8 For cross-household comparisons it is more appropriate to use C/N, where
is a parameter that represents economies of scale at the household level
(e.g. = 1 implies no economies of scale). For our present purposes of
comparison over time, the use of the special case of = 1 is not overly limiting.
9 In our calculation of total food expenditures we include alcohol and tobacco
for consistency with BPS practice.
10 Beginning in April 1998, the base year used to calculate price indices in BPS
publications was changed from April 1988 May 1989 to 1996. As a result,
month-specific values of the price indices in the first and second rounds of
the survey are calculated using different base years. Therefore, before
constructing the deflator for non-food items we had first to convert the base
of the May 1997 food and non-food price indices to August 1996.
11 Alternative price indices include the general price indices for 44 cities, or the
category-specific prices indices for the same 44 cities (BPS, Indikator Ekonomi).
All measures suffer from the disadvantage that price indices based on the
prices of food items or groups of food items in cities may be quite different
from indices based on prices prevailing in rural areas.
Please 110 Skoufias, Suryahadi and Sumarto
check
equations
12 In other words, we estimated a semi log-linear Engel curve for food.
13 A distribution f stochastically dominates another distribution g if

f ( x)dx > g(x)dx .

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() is given by the expression:

1 n
yi y
1
GE( ) = 1 , 0, 1
(1 ) n
i =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
1
Y ( ) 1 n
1
A( ) = 1 ede y

, where Yede ( ) =
n
(yi )

, > 0, 1
i =1

For more details on these and other inequality indices, see Cowell (1995).
 Changes in Household Welfare, Poverty and Inequality during the Crisis 111

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), UrbanRural Differences in Costs of Living and Their Impact on

Poverty Measures, Bulletin of Indonesian Economic Studies 35 (3): 5169.
Atkinson, A. (1970), On the Measurement of Inequality, Journal of Economic Theory
2: 24463.
Berndt, Ernst 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: 7616.
Frankenberg, E., D. Thomas and K. Beegle (1999), The Real Costs of Indonesias
In IFLS Economic Crisis: Preliminary Findings from the Indonesian Family Life
reference, Survey, July, RAND, Santa Monica CA (mimeo).
changed Levinsohn, J., S. Berry and J. Friedman (1999), Impacts of the Indonesian Economic
Indonesia to Crisis: Price Changes and the Poor, NBER Working Paper No. 7194, June,
Indonesian. National Bureau of Economic Research, Cambridge MA.
OK? 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, LPEMFEUI, August
1819, 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.
112 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.