Poverty, Vulnerability and Social Protection in a Period of Crisis: The Case of Indonesia

World Development Vol. 30, No. 7, pp. 1211–1231, 2002 Ó 2002 Published by Elsevier Science Ltd. Printed in Great Britain 0305-750X/02/$ - see front matter

www.elsevier.com/locate/worlddev

PII: S0305-750X(02)00028-1

Poverty, Vulnerability and Social Protection in a Period of Crisis: The Case of Indonesia SHAFIQ DHANANI UN Building, Jakarta, Indonesia and IYANATUL ISLAM * Griffith University, Brisbane, Australia Summary. — Before the financial crisis of mid-1997, estimates of consumption poverty in Indonesia were based on rather modest poverty line thresholds when seen in relation to estimates of capability poverty. The reasons behind this discrepancy are identified and alternative estimates of consumption poverty for the pre-crisis period proposed. During the crisis, the behavior of consumption poverty can be described as transient in nature and is relevant in understanding the notion of vulnerability, that is, the risk that individuals and households can experience temporary episodes of poverty. Vulnerability could have worsened, however, in the absence of government intervention, entailing macroeconomic stabilization measures and social protection initiatives. Based on this experience, a fiscally sustainable social safety net, that is able to reinforce household coping mechanisms and social capital, is recommended as part of the country’s medium-term strategy to combat poverty. Ó 2002 Published by Elsevier Science Ltd. Key words — East Asian crisis, Indonesia, poverty, vulnerability, social protection

1. INTRODUCTION When the financial crisis hit Indonesia in mid-1997, few observers expected it to turn into a full-blown social crisis. Over the course of 1998, a preoccupation with tracking exchange rate fluctuations gave way to concerns about the human dimensions of the Indonesian crisis. Unfortunately, the paucity of data, coupled with different methods for estimating poverty incidence, meant that such concerns ended up in considerable controversy. Views varied from those who regarded a crisis-induced, severe increase in aggregate poverty as a highly plausible scenario, to those who argued that the rise in nationwide poverty was moderate. 1 It is against such a background that the paper attempts to revisit the salient issues in understanding the behavior of poverty during the Indonesian crisis. The paucity of data that characterized much of 1998 has now been compensated by a plethora of information on the social consequences of the Indonesian crisis.

The paper seeks to develop four themes. 2 The first one is to demonstrate that ‘‘. . .official statistics. . . prior to . . .mid-1997. . . underestimated the magnitude and intensity of the poverty that still existed throughout the country’’ (Breman, 2000, p. 8). Such an assessment is reinforced when judged in relation to what Sen (1999, chapter 4) has called ‘‘poverty as capability deprivation.’’ The latter pertains to nonincome dimensions of poverty and focuses on unmet basic needs in health, housing and education. 3 Poverty, seen as the deprivation of basic capabilities, was––and continues to be––significant in Indonesia and is inadequately captured by the available consumptionbased indicators of deprivation.

* The

authors are grateful for the comments and suggestions of two anonymous referees on an earlier version of this paper. Final revision accepted: March 6, 2002.

1211

1212

WORLD DEVELOPMENT

Second, consumption-based poverty incidence can change rapidly in a period of economic crisis such as that which engulfed Indonesia in the late 1990s. 4 This is particularly the case when the crisis is accompanied by rapidly rising food prices. It is thus necessary to distinguish between the transient and long-term behavior of consumption-based poverty. The former is useful in capturing the notion of ‘‘vulnerability,’’ that is the risk that even those who are not poor may experience transient episodes of poverty triggered by a systemic economic shock. The paper uses this construct––and annual data from the core questionnaire of the national socio-economic survey Susenas (the statistical authorities and researchers have generally used the tri-annual consumption module of this survey for calculating consumption poverty incidence)––to argue that the main reason for the rapid rise in poverty incidence as measured by the headcount ratio method is the large number of people living at or in the neighborhood of the poverty line. Small changes in the poverty line can produce relatively large changes in the estimates of people living below this poverty line. The elasticity of the number of people below the poverty line with respect to changes in the poverty line is shown to be quite high. Third, the paper argues that one should distinguish between overall poverty and the depth or severity of poverty, a distinction rarely undertaken before the crisis. Poverty severity increased substantially during the crisis. This paper seeks to measure the severity of poverty in two ways. It estimates the number of poor falling below 80% of the poverty line and below the ‘‘food poverty line,’’ and assesses how they have changed between the pre-crisis period and 1999. It then reports the poverty severity index––also known as P2 ––created by Foster, Greer, and Thorbecke (1984). Fourth, the paper undertakes a preliminary assessment of the measures put in place by the government to combat the worst effects of transient poverty. The paper suggests that in the absence of the twin measures of macroeconomic interventions to tame inflation and stabilize the exchange rate on one hand, and the direct microeconomic interventions of targeted subsidized rice and scholarships for the poor, transient poverty would almost certainly have become more persistent. The final section draws conclusions and highlights lessons learned.

2. CAPABILITY VS. CONSUMPTION POVERTY: A PRE-CRISIS PERSPECTIVE (a) Estimating capability poverty in Indonesia As a prelude to understanding the rather low thresholds that were used to estimate consumption poverty in Indonesia in the pre-crisis period, it is useful to discuss the evidence on capability poverty. Underlying capability poverty can be primarily revealed by unmet basic needs in the housing and health conditions of households, and the educational characteristics of the population. Before the crisis, about 40% of Indonesian households lived in a house with an earthen or wooden floor, and 30% did not have access to improved drinking water (Table 1). About 50% of the households used shared or public toilet facilities and nearly 70% disposed of their faeces in rivers, ponds, lakes or open spaces. About 25% of all households did not have access to electricity for lighting, and relied on kerosene and other fuels instead. As for health conditions, nearly 40% of all persons falling ill never sought medical help in a hospital, clinic or primary health care center, relying on traditional healers and self-treatment instead. Nearly 50% of all households did not have a doctor or midwife present during childbirth. More than a third of all children under the age of five were underweight (see Table 3). Turning now to education, nearly 40% of the population aged 10 and above had either not gone to school or had not completed primary education (Table 2). A further 30% had completed primary school, so 70% of the population had an educational attainment of primary school or less. The remaining 30% had gone on to junior secondary school, however only half of these had continued to senior secondary school. As for literacy, over 10% of the population aged 10 and above was illiterate, this proportion being twice as high for females as for males (15% and 7%). Finally school attendance rates were 95%, 78% and 49% for schoolage children in primary, junior secondary and senior secondary school respectively before the crisis. Rural areas, which accounted for two-thirds of the total population, experienced considerably poorer housing, health and education conditions than urban areas. Some 50% of rural households lived in a house without a solid floor, while 30% of households used mainly kerosene for lighting. The share of rural

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

1213

Table 1. Selected housing and health conditions, 1993–2000 (% of total households) Urban Pre-crisis

Earthen or wooden floor Without improved drinking waterb Shared/public toilets Defecate in rivers/ ponds/open air No electricity for lighting Traditional/ self-treatmentc Traditional healer/family and other birth attendante

Urban þ Rural

Rural Crisis

Pre-crisis

Crisis

Pre-crisis

Crisis

93

96

97

98

99

00

93

96

97

98

99

00

93

96

97

98

99

00a

18

16

14

13

13

12

61

55

51

49

48

46

47

41

38

35

34

32

18

12

10

10

10

11

50

42

38

36

36

36

40

31

28

26

26

25

44

35

33

32

31

31

72

60

56

55

55

55

63

51

48

46

46

45

46

42

40

41

37

37

89

85

83

82

80

79

75

69

67

67

63

61

10

5

4

3

3

2

61

41

34

29

25

23

45

28

23

20

16

14

d

42

45

39

38

39

–

40

44

37

38

38

–

74

64

60

60

53

56

60

50

46

48

40

37

37

41

34

37

38

–

29

21

19

26

18

18

Source: Welfare Statistics, Annual National Socio-economic Survey Susenas (Tables 6.3, 6.7, 6.9, 6.11, 6.5, 2.7B), February, various years, CBS. Welfare Indicators (Tables 2.3 and 2.9 for self-treatment for 1998 and 1999), CBS. a Preliminary figures for 2000. b Not using drinking water from pipe, pump, bottle, protected well or protected spring. c Not using medical services of hospital, doctor, health center, clinic or paramedic personnel. d ‘–’ not available. e Last birth of under-fives after 1998.

households using open spaces for disposing of their faeces was nearly 85%, while that using traditional or family birth attendants was as high as 60%. Nearly half of the rural population had less than primary school education and over 80% with primary education or less. The proportion of illiterate women was more than twice as high in rural areas, 19% compared with 8% in urban areas. So far, individual housing, health and education characteristics have been described without any attempt to aggregate some or all of these measures into an overall index of poverty or family welfare. The human poverty index (HPI) developed by the United Nations Development Program (UNDP, 1997) attempts to provide such an aggregate measure of capability poverty by combining four indicators, namely life expectancy, adult literacy rate, access to improved drinking water and proportion of underweight children below the age of five. The HPI for Indonesia can be estimated at about 25% before the crisis (Table 3), down from about 35% in 1990. It is clear that, while the level of HPI was quite high in pre-crisis Indonesia, there was a

rapid decline over time, entailing significant achievements in each of the four components of the HPI throughout 1990–96. This partly reflects the effect of rapid and sustained growth under the Soeharto regime, enabling the average Indonesian the purchasing power to acquire basic services. It also reflects the effect of direct government intervention in the provision of basic services, notwithstanding the waste and corruption that characterized the regime. (b) Official poverty series as an inadequate indicator of capability poverty How do the above measures of capability poverty compare with the more conventional headcount measure of consumption poverty produced by the Central Bureau of Statistics (CBS) before the crisis? CBS produces poverty estimates every three years, based on the consumption module of the Susenas household survey, which is fielded only once every three years. Its poverty incidence for 1996 was 11% nationally (10% in urban areas and 12% in rural areas), or two and half times less than the human poverty index for that year (Figure 1).

1214

WORLD DEVELOPMENT Table 2. Educational attainment, literacy levels and school attendance, 1993–2000 (% of total population aged 10 and above)a Urban Pre-crisis 93

Educational attainment Less than primary 30 school Primary school 28 Primary or less 58

Urban þ Rural

Rural Crisis

Pre-crisis

Crisis

Pre-crisis

Crisis

96

97

98

99

00

93

96

97

98

99

00

93

96

97

98

99

00

25

23

23

22

23

54

49

45

45

44

43

45

40

37

37

35

35

28 53

28 51

28 51

27 50

27 50

32 86

35 84

36 81

36 81

36 80

36 79

31 76

32 72

33 70

33 70

33 68

32 67

Junior school Senior school and above

17 25

16 31

18 31

18 31

19 31

19 31

8 6

9 7

11 8

11 8

11 9

12 9

11 13

13 15

14 16

14 16

15 17

15 18

% Illiterate Male Female

7 4 11

6 3 9

6 3 8

5 3 8

5 3 8

6 3 9

18 12 24

16 11 22

14 9 19

14 9 19

13 9 18

14 9 18

14 9 19

13 8 17

11 7 15

11 7 14

10 6 14

10 6 14

Still attending school a Aged 7–12 (primary) Aged 13–15 (jun. school) Aged 16–18 (sen. school)

(%) 96 97 84 87

98 88

98 89

96 88

98 88

91 61

93 69

94 72

94 71

94 74

94 73

93 69

94 76

95 78

95 77

95 79

96 79

63

67

68

69

65

30

34

36

36

38

36

43

48

49

49

51

49

66

Source: Welfare Statistics, Annual National Socio-economic Survey Susenas (Tables 3.2 and 3.3), February, Various years, Central Bureau of Statistics. Premliminary data for 2000. a Still attending school: special tabulations of core questionnaire of Susenas before 2000, Table c p. 99 for 2000 (preliminary).

Table 3. Human poverty index, 1990–2000 (% of population)a Indicator

1990

1993

1995

1996

1997

1998

1999

2000

P1 Not expected to survive to age 40 P2 Adult illiteracy rate aged 15þb P3 Average P31 No improved drinking water c P33 Underweight children under 5d Human poverty index (HPI)h

25.7 18.5 46.7 (47.6) 45.8e 34.7

(21.7) (16.0) 41.6 39.7 43.4f 30.6

(19.4) 14.5 35.5 34.9 36.1 26.4

18.3 (13.8) 33.6 31.3 (35.9) 25.0

(17.2) (13.2) 31.7 27.8 (35.6) 23.6

(16.2) (12.6) 30.8 26.2 35.4 22.8

(15.3) (12.0) 32.3 25.7 38.9g 23.5

(14.4) (11.4) 32.1 25.4 (38.9) 23.2

Source: P1 : UNDP/CBS/Bappenas (2001). P2 : Population census 1990 and intercensal population survey 1995, CBS. P31 : Welfare Statistics, National Socioeconomic survey Susenas (Tables 3.3 and 6.7), February, various years, CBS. Preliminary data for 2000. P33 . Welfare Indicators, various issues (Table 2.3 for 1999), based on tri-annual Susenas health module CBS. a Figures in brackets are authors’ trend estimates. b Illiteracy rate for adults aged 15þ (census data) differs from Table 2 (population aged 10þ, annual Susenas survey data). c Percentage of households without access to water from pipe, pump, bottle, protected well or protected spring. d Moderately underweight (23.0% and 21.2% in Feb. 1995 and Dec. 1998) and severely underweight (13.1% and 16.6% in Feb. 1995 and Dec. 1998). e 1989. f 1992. g December 1998. h HPI calculated using the following formula: HPI ¼ ½ðP13 þ P23 þ P33 Þ=31=3 where P3 ¼ 1=2ðP31 þ P32 Þ. See UNDP (2001, p. 241). The HPI estimates above for 1999 is higher than in the UNDP Human Development Report (23.1 vs. 21.3) due mainly to the lower values for P1 and P33 in the latter (13 and 34).

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

1215

Figure 1. Capability and consumption poverty: pre-crisis perspective. Source: CBS: Statistical Yearbook of Indonesia 2000 (Table 12.1.A, old poverty line standards). Human poverty index: Table 3. Authors’ estimates: Table 4 for 1996. Note: Authors’ estimates for 1976–93: based on 1996 urban and rural poverty lines adjusted for inflation consumer price index (see text).

Over 1976–99, CBS introduced several changes in its calculation method. The key changes entailed: (i) adjustments for food consumed outside the home in 1987, based on special food surveys; (ii) moving from a caloriecost method for setting the food poverty line to the pricing of a food bundle method in 1993; (iii) producing national estimates of poverty incidence by adding up provincial estimates in 1996, the latter obtained using province-specific food and nonfood bundles; and (iv) the introduction of new poverty line standards in 1999. The CBS methodology is presented in detail in CBS/UNDP (1999), CBS (2000a) and CBS (2000b). 5 Some of these changes have wide ramifications for interpreting poverty statistics in Indonesia. The adoption of higher poverty line standards in 1999 resulted in a major upward revision of poverty incidence for 1996 from 11% to 18% nationally. Consequently, the new 1996 estimate is no longer comparable to poverty estimates for previous years shown in Figure 1, nevertheless it remained 7% below the human poverty index for that year. Furthermore, the use by the statistical authorities in 1996 of an approach that computes national poverty incidence as an aggregation of provincial poverty estimates complicates comparability over time and introduces contentious issues of regional differences in prices and

consumption patterns in Indonesia. Should one use a poverty line that can be applied across all regions––and rural and urban areas within regions––or should one highlight local representation and thus use region-specific poverty lines? These questions were raised in World Bank (1993), Bidani and Ravallion (1993), and Ravallion and Bidani (1994). They have been recently revisited––but not necessarily resolved––by Sutanto, Irawan, and Said (1999), Sutanto and Irawan (2000) and Pradhan, Suryahadi, Sumarto, and Pritchett (2000). 6 CBS uses province-specific food bundles (varying in food types and quantities as well as prices), while the World Bank prefers using a national reference food bundle based on the national consumption pattern of the poorest 15% of the population (varying in price across provinces) to provide a consistent poverty profile over space and time (World Bank, 1993, p. 103). A variation of the latter method is the use of an ‘‘iterative’’ method to determine both the reference population and its typical nonfood budget share (Suryahadi, Sumarto, Suharso, & Pritchett, 2000, p. 13; Pradhan et al., 2000), while CBS computes the cost of province-specific nonfood bundles in the same way as it does the food bundle (CBS, 2000c). 7 Chesher (1998) has shown that moving to region-specific poverty lines compounds the extremes in the spatial distribution of poverty,

1216

WORLD DEVELOPMENT

making the estimates higher for ‘‘high-poverty’’ provinces and lower for ‘‘low-poverty’’ provinces. One could argue that, despite these modifications, the CBS method yields ranking of provinces from the least to the most poor that are quite consistent with the iterative method, with a Spearman rank correlation coefficient of 0.92 (Pradhan et al., 2000, p. 14). This conclusion, on the other hand, is less valid if urban and rural estimates are considered separately. Certainly, in rural areas, the iterative method yields rankings that are different from the CBS approach. While the above discussion of regional poverty is by no means exhaustive, it provides a flavor of the many issues still surrounding regional poverty calculations in Indonesia. In what ways do the debate on the use of regionspecific as opposed to a national poverty line in dealing with the spatial distribution of poverty in Indonesia affect the findings of this paper? There are two issues that are germane to the substance of the current discussion. First, a key argument of this paper that the CBS method underestimated national poverty in pre-crisis Indonesia still stands and stems from the way CBS estimated rural poverty vis- a-vis urban poverty. This is a core methodological issue that is noted by World Bank (1993) and that is revisited at some length at a subsequent juncture. Second, an examination of the spatial distribution of poverty is useful in supplementing the current state of knowledge on the effects of the 1997 financial crisis in Indonesia. For example, a well-known hypothesis is that Java was particularly badly affected vis- a-vis the provinces outside Java. 8 There is some evidence that some of the Javanese provinces were indeed badly hit. Consider one conspicuous case. The official CBS figures, published in CBS (2000c), the Statistical Yearbook of Indonesia 2000 (Tables 12.5–12.7), show that one of the highest absolute increases in poverty in the country took place in the populous, highly urbanized and industrialized province of West Java. This is consistent with the view that the crisis-induced increase in poverty in Indonesia reflected the collapse of the construction sector in West Java and the capital city (Jakarta) and the decline in other nonagricultural employment opportunities. 9 Clearly, the issue of regional variations in poverty is significant is supplementing one’s understanding of the social consequences of the Indonesian crisis, and ought to be the subject of further research.

Three common consumption or monetary measures would suggest a much higher poverty incidence of the population than the 11% officially sanctioned estimate for pre-crisis Indonesia, or even the revised 18% estimate for 1996. First, the food budget share of the average household was nearly 60% in 1996, this proportion reaching nearly 70% for the bottom 50% of population. Another way of looking at Engel’s law is to note that nearly 70% of households spent more than 65% of their expenditure on food, the proportion ranging from 18% in urban areas to 92% in rural areas. Second, the mean and median per capita expenditure amounted to just $0.90 and $0.70 per day in that year (calculated using the prevailing exchange rate) according to the Susenas survey. Finally, the mean earnings of wage employees were <$90 per month, or around $0.70 per capita per day for a family of four members according to the National Labor Force survey Sakernas. As for ownership of basic consumer goods, about a third to a half of households did not own many everyday durable goods, particularly in rural areas. The latest 1995 intercensal population survey data indicate that just over a half of the households surveyed possessed a kitchen stove, this proportion falling to a third in rural areas. About a third of all households did not own a radio or cassette player, and more than half did not own a television set, this share declining to <30% in rural areas. As for means of transport, only half of the households owned a bicycle, and 30% and 10% owned a motorcycle in respectively urban and rural areas. Car ownership was limited to 9% of urban households and just 2% of rural households, or 4% nationally. In sum, the relatively high food budget shares, the low earnings and expenditures levels of the average Indonesian household (in relation to the used $1 per capita per day norm commonly used in international comparisons of living standards), and the relatively restricted ownership of basic durable goods, would suggest a higher incidence of consumption poverty than officially admitted. The low pre-crisis CBS estimate of consumption poverty in Indonesia for 1996 and in previous years is due to two main factors: a relatively low poverty line in rural areas, and an underestimated consumption of nonfood items. First, the official food poverty line in rural areas was just 78% of its urban equivalent, a gap far in excess of expected urban-rural price differentials (Table 4). This is because the CBS

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

1217

Table 4. Pre-crisis alternative estimates of consumption poverty incidence, 1996 (head count ratio or HCR method) Poverty line (Rupiah/capita/month) Food (2,100 Kcal) % of Urban

Total (Food þ Nonfood) % of Urban

Percent (food share)

Percentage poor

Poor people People (million)

Percent share

CBS, old poverty lines Urban 29,681 Rural 23,197 a 25,596 Urban þ Rural

100 78

38,246 27,413 31,421

100 72

78 85 81

9.7 12.3 11.3

7.2 15.3 22.5

32 68 100

CBS, new poverty lines Urban 30,455 Rural 23,844 Urban þ Rurala

100 78

42,032 31,366 35,312

100 75

72 76 80

13.6 19.9 17.7

9.6 24.9 34.5

28 72 100

Authors’ Estimatesb Urban 30,455 Rural 27,105 28,312 Urban þ Rurala

100 89

48,341 39,283 43,607

100 81

63 69 67

20.9 39.0 32.5

14.7 48.8 63.5

23 77 100

Source: CBS (2000c): Tables 2.2 and Table 3.1, Statistical Yearbook 2000, CBS. a Urban þ rural poverty lines calculated. b Author’s estimates: Total poverty line ¼ CBS food poverty line scaled up by nonfood share in consumption. Urban food poverty line ¼ CBS new urban food poverty line. Rural food poverty line ¼ urban food poverty line  89% to adjust for cheaper rural prices.

method, while using a standard bundle of food for both urban and rural households, allowed them to consume these commodities in different quantities, thus implicitly allowing urban consumers to purchase more expensive food items. The alternative use of an identical food basket for both urban and rural areas, in terms of both type and quantity as proposed by the World Bank (1993) would significantly increase the rural food poverty line. A later section will show that poverty estimates are quite sensitive to small changes in the poverty line, suggesting that a higher rural food poverty line would substantially raise the estimate of rural poverty. Second, the value of nonfood items in the total official poverty line amounted to <20% of total household expenditure until 1996 (22% in urban and 15% in rural areas). This is quite low relative to the expenditure pattern of households in the neighborhood and below the poverty line observed in Susenas consumption surveys, who spent around a third of their total consumption expenditure on nonfood items. The new poverty line standards used for the revised 1996 poverty estimates only partially corrected the extent of nonfood consumption (28% in urban and 24% in rural areas), while they did not correct for the low rural food poverty line relative to the urban one. As such, the CBS method continues to yield relatively low total poverty lines, and correspondingly

lower poverty estimates, particularly in rural areas. 10 (c) Trends in pre-crisis absolute poverty: authors’ estimates Consumption-based measures of poverty ought to reflect the situation not only of the many households whose most basic needs in terms of housing, health and education remain unfulfilled, but also of the many households who spend more than two-thirds of their expenditures on food because of their low earnings in employment or self-employment, amounting to <$1 per family member per day (at the prevailing exchange rate). The original CBS consumption-based measures, which estimated the overall poverty incidence level of Indonesia before the crisis at 11%, does not meet these two criteria, nor does the revised CBS estimate for 1996. Table 4 presents an alternative estimate of poverty incidence for 1996 which addresses some of the weaknesses of the CBS methodology, and which is more closely aligned with the indicators of capability poverty discussed above. The starting point for this method is the CBS national food poverty line. But instead of using a nonfood consumption bundle, it arrives at the total poverty line by using the Orshansky scaling-up factor to allow households at the

1218

WORLD DEVELOPMENT

poverty line to spend a third of their total expenditure on nonfood consumption (the 1996 Susenas data show that urban households in the neighborhood of the poverty line spent 37% of their total consumption on nonfood items, while the corresponding figure for rural households was 31%). This is essentially the same scaling-up method as the one used by the World Bank in its country report for 1984 (see Rao, 1983). A second key modification attempts to correct for the substantially lower rural food poverty line relative to its urban counterpart. In the CBS calculations for 1996, this difference was almost 20%, while the difference in the price of basic food commodities between urban and rural areas was of the order of 11% (World Bank, 1993; Ikhsan, 1999; CBS/UNDP, 1999; Asra, 1999). The rural food poverty line was obtained by simply deflating the cost of the standard CBS urban food bundle, considered the national standard bundle of food in terms of types of food and corresponding quantities consumed, by the above price difference between urban and rural areas. Due to the sensitivity of poverty incidence estimates to the level of the poverty line selected, this second adjustment is enough to raise the estimate of rural poverty substantially. Applying these two modifications to the official CBS method, national poverty incidence in 1996 can be re-estimated at 32.5% (20.9% in urban areas and 39.0% in rural areas). While this alternative national estimate is around three times higher than the original CBS estimate of 11%, it is nevertheless closer to indicators of unmet basic needs described earlier. In addition, unlike the CBS estimates, the urban-rural poverty incidence gap is significantly higher, at 19% points compared with just two in the original CBS estimates (6% points in the revised estimate). The higher differential obtained by the authors appears credible because it is consistent with the rapid rural–urban migration taking place before the crisis. Finally, it would be useful to track changes in absolute poverty over time, i.e., to measure poverty incidence in terms of a poverty line built from a constant bundle of food and nonfood items. This is because, in estimating poverty incidence over the years, and especially in adopting the new poverty line standards for 1996, CBS has used the alternative concept of relative poverty. This concept allows basic needs to change over time to reflect higher per capita expenditures, and the decision on the

part of households to consume more and better food, clothing, housing, education and other goods and services. There is a large literature on the concept of absolute and relative poverty that argues for the need to keep track of both measures. In order to produce estimates of absolute poverty over time, the basket of needs constituting the poverty line should be kept fixed, particularly during a period of economic crisis characterized by rapid changes in relative prices. For the pre-crisis period 1976–96, an approximate way of keeping the basket of needs constant is to simply adjust the 1996 poverty lines in Table 4 for the overall rate of inflation in previous years. While the implicit assumption of fixed relative prices in this method, both between food items and between food and nonfood items, is admittedly a simplification of reality over such a long period of time, Figure 1 nevertheless indicates that absolute poverty, by the consumption standards of 1996, was around 70% in the late 1970s, and was halved to about 32% in the mid-1990s. 3. CRISIS-INDUCED TRANSIENT POVERTY, 1996–99 (a) Revisiting the poverty debate As noted, the initial phase of the discourse on the impact of the crisis on poverty was mired in controversy, although a fair degree of consensus has now emerged. With the benefit of hindsight, it is now clear that two key issues were either missing or insufficiently emphasized in the early phase of the Indonesian poverty debate. First, there was a general lack of appreciation of the need to highlight the phenomenon of transient poverty. This is particularly important in the case of the Indonesian crisis, which was characterized by a deep recession as well as an inflation shock. It is the latter, as will be argued, that holds the key to understanding transient poverty in Indonesia––that is, people moving in and out of poverty in a relatively short period of time. Second, the early phase of the poverty debate in the context of the Indonesian crisis paid insufficient attention to the need to distinguish between the overall incidence of poverty and the severity of poverty. As Sen (1976) showed more than two decades ago, the two may not behave in an identical fashion. The overall incidence of poverty is best captured by the

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

headcount ratio, but the latter is unable to distinguish between those who are hovering just under the poverty line and those who are located well below it. It is now customary in the poverty measurement literature to capture the severity of poverty by estimating whether inequality among the poor has worsened. If the latter occurs, then this implies that the very poor have become worse off vis- a-vis the marginally poor. This section reports a poverty severity index compiled by CBS, but adds an intuitive dimension to the findings by assessing changes in the number of the poor below 80% and around 65% of the poverty line, the latter commonly referred to as the food poverty line. (b) The evidence Estimates prepared by four different sources confirm that headcount poverty increased substantially in the first year of the crisis. It now appears that this increase was transient in nature. The reasons underpinning this trend and the policy implications that follow from this are highlighted at a subsequent stage. First, data compiled by CBS using revised poverty lines indicate a rapid rise in poverty incidence during the crisis, from 18% to 23% between February

1219

1996 and February 1999, peaking at 37% in September 1998 (double the February 1996 rate) when food prices were at their highest level, before declining to the 1996 level by August 1999 11 (Figure 2). Second, the analysis of matched households in the periodic ‘‘100 Village Survey’’ funded by UNICEF and conducted by CBS found that poverty incidence in rural areas doubled from 12% to 24% from May 1997 to August 1998 (Skoufias, Suryahadi, & Sumarto, 1999). Third, a careful assessment of the evolution of poverty during 1996–99, comparing different methods and aiming at reaching a consensus estimate, estimated that poverty incidence doubled between August 1997 and February 1998 and, at its peak in August–December 1998, was nearly three times higher (264%) than during its lowest point in August 1997, following the surge in the price of rice. Poverty incidence declined thereafter at the start of the stabilization of general inflation (Suryahadi et al., 2000, p. 22). Finally, estimates prepared by the authors, using poverty lines that are different from CBS, indicate that poverty incidence increased from an estimated 29% in February 1997 to 44% in September 1998. Food prices began to decline thereafter, leading to a rapid decline in poverty

Figure 2. Poverty incidence during crisis, 1996–2000 (% population). Source: CBS: Statistical Yearbook of Indonesia 2000 (Table 12.1.B, new poverty line standards). CBS/UNDP (1999, p. 58) for September 1998. Human poverty index: Table 3. Authors’ estimates for February 1996: Table 4. Note: Authors’ estimates for February 1997–February 2000 using method given in Table 4, and based on following data: February 1998, 1999 and 2000, based on expenditure distribution of core module of annual Susenas survey. December 1998 and August 1999, based on consumption profile of ad hoc Mini-Susenas survey.

1220

WORLD DEVELOPMENT

to 34% in February 2000. Nevertheless, the headcount poverty incidence remained some 5% higher than before the crisis. A feature of the evidence worth emphasizing is the contrast between the stability of capability poverty and the volatility of consumption poverty during the crisis, clearly depicted in Figure 2. Transient poverty increased in both urban and rural areas according to both CBS and the authors, however this increase was more marked in urban areas, where poverty incidence doubled from 16% to 33%, while rural poverty incidence rose from 38% to 55% between February 1997 and February 1999. Both the magnitude of change in overall poverty and its severity in urban areas are consistent with the significant increase in the share of household incomes devoted to purchasing food. The household food budget share rose from 50% to 56% in urban areas, and 67% to 73% in rural areas. In fact, the percentage of the population spending more than 65% of its total expenditure on food more than doubled from 18% to 39% in urban areas, and rose from 92% to 96% in rural areas in this period, according to the annual national Susenas survey. The economic crisis not only increased the number of people falling below the poverty line substantially, but also increased extreme poverty. The number of people falling below around 65% of the total poverty line, or below the food poverty line, and the number of people falling below 80% of the total poverty line both increased faster than the overall number of poor people between February 1996 and February 1999. These increased by around 50% compared with 37% for the population below the authors’ poverty lines (Table 5). While extreme poverty rose more rapidly in urban areas,

due to the low number of poor before the crisis, the number of people below the food poverty line rose by 5.5 million in rural areas, compared with 3.5 million people in urban areas. The CBS poverty lines yield similar results for the population falling below the food poverty line. A second way of illustrating the severity of poverty is to use a measure called P2 developed by Foster et al. (1984). CBS estimates that P2 increased rapidly during the crisis. It rose by 31% from 0.71 to 0.93 in urban areas, and by 23% from 0.96 to 1.18 in rural areas between February 1996 and February 1999 (Figure 3). In both cases, the index rose sharply in the intervening period (December 1998). More recent data for August 1999 indicate a decline in the urban severity index back to pre-crisis level, but the index remained above the pre-crisis level for rural areas. Suryahadi et al. (2000, p. 25), using the old CBS poverty line standards, found that the P2 index increased even more rapidly than the above suggests, by 202% in urban areas and 84% in rural areas, indicating a deepening of poverty incidence in Indonesia, especially in urban areas. The rise in extreme poverty during the crisis is consistent with deteriorating health and nutritional standards of the very poor reported by other studies. According to one such study, the prevalence of micronutrient deficiencies and wasting increased markedly in rural Central Java between 1995–96 and early 1999, while the prevalence of wasting among children was very high in early 1999 in the urban slums of Jakarta, Surabaya and Makassar, a situation usually only detected in emergency or disaster situations, and indicative of severe food shortages (Helen Keller International, 1999). The prevalence of anemia and night-blindness

Table 5. Population below selected poverty lines, 1996–99 (millions of people, headcount ratio method) CBS new poverty lines Feb. 96

Feb. 99

Authors’ poverty lines

Change

Feb. 96

Million

%

Feb. 99

Change Million

%

Below 65% of poverty line Urban Rural

12.29 2.38 9.91

19.35 5.95 13.40

7.06 3.57 3.49

57 150 35

18.44 2.38 16.06

27.49 5.87 21.62

9.05 3.49 5.56

49 147 35

Below 80% of poverty line Urban Rural

16.52 4.25 12.27

20.15 7.00 13.15

3.63 2.75 0.88

22 65 7

32.94 7.24 25.71

50.51 14.16 35.94

17.16 6.92 10.23

52 96 40

Below total poverty line Urban Rural

35.19 9.66 25.53

48.43 15.48 32.95

13.24 5.82 7.42

38 60 29

63.51 14.75 48.76

86.77 25.12 61.65

23.26 10.37 12.89

37 70 26

Source: Calculated from sources given in Table 4.

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

1221

Figure 3. Trends in poverty severity index (P2), 1996–99. Source: CBS (2000b) Table 2.4, p. 18. Note: P2 is in essence the square of the efficient of variation of expenditure distribution below the poverty line (see Foster et al., 1984).

among children and mothers in both rural Central Java and the city slums also increased during the crisis, and continued to do so in the first half of 1999. An analysis of weight-for-age data contained in Susenas survey indicates that, while there was little change in children below the age of five over 1995–98, the prevalence of underweight children aged 6–17 months in both urban and rural areas, and of 6–23 months in rural areas increased in this period (Jahari, Sandjaja, Soekirman, Herman, & Jalal, 1999). The same study notes an increase in the cases of severe malnutrition reported to the crisis center of the Ministry of Health, nutrition clinics and in the media (including 24,000 cases of kwashiorkor and marasmus, diseases not reported since the early 1980s), and an increased prevalence of underweight children under five during 1997–98 in a study conducted by the University of Indonesia. (c) Explaining the evidence The major thrust of the evidence is that the worst is over and that the incidence of nationwide poverty is apparently moving back toward pre-crisis levels. Thus, a social recovery seems to be in progress. Is it supported by alternative evidence? Furthermore, how can the volatile behavior of poverty during the crisis be ex-

plained? Figure 4 provides a framework for explaining the responses to the crisis. The key feature of the evidence presented above is the volatility of poverty statistics. Perhaps the best way to understand this phenomenon is to highlight the following points. The Indonesian crisis was characterized not just by a deep recession, but also by an ‘‘inflation shock’’ in 1998 of over 100% per annum. Nevertheless, the inflation rate fell equally sharply. For the first nine months of 1999, Indonesia experienced virtually zero inflation. The behavior of the inflation rate is reflected in the behavior of the poverty line. The poverty line for urban and rural areas rose sharply between 1996 and December 1998, achieved a plateau in the early part of 1999 and then declined by August 1999. When the impact of inflation on the poverty line is combined with the fact that a significant component of the population is clustered around the poverty line, even moderate movements in the poverty line can trigger significant changes in the incidence of poverty. Thus, in the short run, inflation is the key determinant of rapid changes in observed poverty. Hence, the hypothesis of transient poverty. One could argue that the fall in prices during 1998–99 was the natural outcome of an economy in recession, and that the fall in consumer demand and the rise in excess capacity put downward pressure on prices. But the inflation shock did not

1222

WORLD DEVELOPMENT

Figure 4. Response to the crisis.

subside by accident, nor was the protection of the poor from the ravages of inflation an unplanned phenomenon. In both cases government intervention, combining anti-inflation strategies with social protection measures, played an important role in mitigating the incidence of transient poverty. This point will be taken up in greater detail at a subsequent stage. The volatility in poverty estimates in Indonesia is due to the well-known observation that the degree of inequality in household consumption expenditures reported in the Susenas survey is rather low, and that a large number of households live in the neighborhood of the poverty line (Booth, 1997, p. 6). In rural areas in particular, the shape of the frequency distribution curve of the population is relatively narrow, with a small standard deviation about the mean. With so many people clustered around the poverty line, a small change in it can produce relatively large changes in the number of people living below this poverty line using the headcount measure. This is further illustrated by the vastly different poverty incidence estimated using the one-dollar and the two-dollar poverty lines in the World Development Report 2000/01. While only 15% were below the one-dollar line, this share rose to 66% below the two-dollar line (World Bank, various years, pp. 280–281). This also means that many more people are vulnerable to episodes of

poverty than those below the poverty line. One estimate puts this at 30–50% of all Indonesian households, assuming a headcount poverty rate of 20% (Pritchett, Suryahadi, & Sumarto, 2001, p. 25). The sensitivity of the headcount measure of poverty to small changes in the poverty line can be quantified by calculating the elasticity of people below the poverty line with respect to changes in the CBS poverty line as follows: qpl ¼

% Change in poor people % Change in poverty line

where qpl is the elasticity of poverty with respect to the poverty line. For February 1999, the elasticity with respect to the CBS poverty line can be calculated as 3.1. In other words, a 1% change in the CBS poverty line leads to a 3.1% change in the number of people living below this poverty line. When poverty incidence is estimated at around 24% of the total population by CBS, this corresponds to a change in the percentage of people living below the poverty line of around 24%  0:031 ¼ 0:74%. Other things being equal, a 10% increase in the cost of living of the people living in the neighborhood of the poverty line will result in nearly 7.5% more people being counted as newly poor by the headcount measure.

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

Support for the view that a tenuous social recovery is in progress comes from real wage data. Nominal wage increases were unable to keep pace with the rising cost of food and other essential commodities between mid-1997 and early 1999 following the collapse of the rupiah. In addition, the employment situation worsened. While nominal earnings increased by <20% in the first year of the crisis, inflation of the order of 100% in the same period eroded the purchasing power of the nominal wage to just 60% of its pre-crisis value. The manufacturing and construction sectors, but also virtually all other nonagricultural sectors, shed many jobs following the beginning of the economic crisis in mid-1997. This in turn had a multiplier effect on informal sector employment and incomes. The crisis was further compounded by the drought affecting mainly Eastern Indonesia in the second half of 1997 and also some parts of Java and Sumatra, as well as forest fires raging in Sumatra, Kalimantan and Eastern Indonesia in the same period, both leading to loss of agricultural incomes (ILO/UNDP, 1998; ILO, 1999). Nominal wages continued to grow by 20% per annum in 1999 and 2000, while inflation was brought under control, rising by just 5–10% per annum. Real wages were nevertheless still only about 80% of their pre-crisis level by August 2000 (Dhanani & Islam, 2001, p. 34). (d) Implementation of social safety net programs An anti-inflation strategy pursued within a macroeconomic framework cannot on its own fully protect the poor from the ravages of inflation. The point is that not all socioeconomic groups face a uniform inflation rate or suffer to the same extent from a surge in prices. As is well known, food prices rose faster than the overall inflation rate throughout 1998. More importantly, there was a substantial rise in the price of rice by 180% while nonfood items rose by 80% between February 1996 and February 1999, according to the Susenas survey data (World Bank, 2000b). Given that the poor––both in rural and urban areas––are net buyers of food, the wedge between the food and nonfood inflation rate goes some way towards explaining why the poor bore the brunt of the crisis. Net sellers of food in the rural economy and producers of export-oriented cash crops, who are likely to be located above the poverty line, would have benefited from rising food

1223

prices and the currency devaluation at the expense of more vulnerable groups such as landless rural workers, whose real wages collapsed. There is now evidence that those in extreme poverty actually faced a higher inflation rate vis- a-vis others. One study, on the basis of regular Susenas data, estimated that the bottom 10% of households actually experienced a higher inflation rate than the top 10% of households during the crisis period, particularly in urban areas (Levinsohn, Berry, & Friedman, 1999). This suggests that a two-track policy––one focusing on aggregate price stability and the other on subsidizing the price of key goods and services consumed by the poor––may be more effective in mitigating the effects of the inflation shock on the poor. This in turn leads one to a discussion of social protection policy, or social safety net (SSN) as it is widely known in Indonesia. The government was slow in responding to the crisis, and did not launch a SSN program until a year after the start of the crisis, prompted by food riots and surging food prices in the first half of 1998. Nevertheless, several programs were launched in mid-1998, including a large rice subsidy program, a relatively large scholarship and block grant program, free medical, family planning and childbirth services for very poor households, a nutritional program for pregnant women and babies, and block grants to local communities for labor-intensive public works. The rice subsidy program and scholarships for primary and senior secondary school students were entirely government-funded, while the World Bank, the Asian Development Bank, the UN agencies, and several bilateral donors contributed to other programs. The rice subsidy program, known by its Indonesian acronym OPK, aimed at providing up to 10 million food-insecure households with 10–20 kg of rice per month at substantially subsidized prices. The scholarship program consisted of providing scholarships to some four million school children (6% of primary school children, 17% of junior secondary school students and 10% of senior secondary school students). The above SSN programs have come under criticism for their undercoverage, i.e., a large number of poor people were not reached, as well as their leakage, i.e., a large proportion of the benefits had gone to the nonpoor, due to their inadequate design and implementation, and waste and corruption. One such evaluation, conducted by the social monitoring and

1224

WORLD DEVELOPMENT

early response unit (SMERU), a donor-funded project, and based on the analysis of the February 1999 nationwide Susenas survey SSN module, concluded that the SSN had near random targeting, resulting in almost proportionate distribution of benefits between poor and nonpoor households (Sumarto, Suryahadi, & Widyanti, 2001, p. 20). Also, except for the rice subsidy program, which covered 40% of all households and 53% of poor households, and the nutrition program, which covered 16% of people, the SSN program coverage reached just 5–8% of the total people and 5–10% of the poor (Sumarto et al., 2001, p. 54). Finally, this study, as well as a previous SMERU study based on the ‘‘100 Village Survey,’’ found that program effectiveness varied across programs and regions from location to location and from program to program (Suryahadi, Suharso, & Sumarto, 1999; Sumarto et al., 2001). 12 The implication one could draw from the above findings is that the impact of SSN programs on poverty alleviation was rather limited, despite their size and national coverage. A corollary conclusion is that the government was unable or unwilling to ensure program effectiveness, and that future SSN programs may encounter the same fate, unless drastic measures are taken to avoid past mistakes and shortcomings. On the other hand, a number of field studies and evaluations have reached a different assessment, namely that the rice subsidy and scholarship programs were relatively effective in cushioning the impact of the crisis on the poor. Several such studies were conducted by SMERU and at least one evaluation used the same nationwide Susenas survey SSN module, making it difficult to ignore them. A complete evaluation of SSN programs is outside the present scope, which is limited to assessing the effectiveness of the large income transfer programs such as rice subsidy and scholarships in alleviating transient consumption poverty. The preservation of critical social services, particularly health and education, and sustaining local economic activity through regional block grants and small-scale credits, are therefore not discussed here, nor is the relatively minor labor-intensive public works program. The rice subsidy program, according to administrative data, reached 44.2 million people in 1998 and peaked at approximately 50 million people in early 1999, almost equivalent to the entire population recorded as poor by CBS in December 1998. An assessment of the program,

based partly on independent fieldwork commissioned by the Ministry of Food and Horticulture and conducted by universities and NGOs (Rachman, Pujihastuti, Sabannah, & Sukandar, 1999), arrived at the following conclusions. The program was highly cost-effective; in its absence, the poor would have suffered an 11% income reduction and the very poor a 22% decline in income; the poor would have reduced their calorie consumption by about 8% and protein consumption by about 15%; the program had made an important contribution to price stability; and, since the introduction of the program, there had been a pronounced absence of food riots. The study concluded that the rice subsidy program should be part of Indonesia’s long-term tool-kit of social protection measures (Tabor & Sawit, 1999). A December 1998 field study of 21 urban areas and 19 rural areas in five provinces conducted by SMERU concluded that . . .OPK program is reaching needy people, but not all needy people received OPK,’’ and that ‘‘. . . no information was uncovered concerning wastage, re-sale, corruption or malfeasance’’ (Sri Kusumastuti et al., 1999). A later SMERU survey of around 450 households in Northern Java found that ‘‘There is no doubt that the OPK rice program has assisted the poorer members of the community in hamlets where 10 kg of rice was provided three or four times’’ (Hardjono, 1999, p. 28). While both these studies have pointed out that coverage has been insufficient or inadequate due to remoteness and presumably limited resources, they do agree that whatever OPK rice was available had gone to those who needed it, lessening their poverty burden. As for the scholarship program, the Northern Java survey cited above observes that . . .the children who were granted scholarships very largely come from genuinely poor households; unlike most government programs, the scholarship program received almost universal praise from respondents, with only two complaining that their child was passed over in favor of better-off neighbors. (It also notes that) The scholarship program. . . displays a positive bias towards poor households and at the same time has given very tangible assistance to a relatively large number of beneficiaries (Hardjono, 1999, pp. 31–32).

An evaluation of the February 1999 SSN module of the Susenas survey found that there was a strong bias toward the poorest groups: 63% of all scholarships went to the poorest two quintiles, while 18% went to the highest two

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

quintiles, the latter share falling to 11% in the case of primary school (Jones & Hagul, 2001, p. 224). These findings led the study to conclude that ‘‘. . .(the scholarship and block grant program) is. . . one of the few programs in Indonesia that have demonstrably had a positive social equity impact, bringing benefits to poorer students and poorer schools’’ (Jones & Hagul, 2001, p. 228). Finally, the 1998 and 1999 Susenas survey data indicate that school attendance rates at all three levels were maintained during the crisis (Table 2 above). This suggests that this program may have prevented large numbers of poor students from dropping out of school. While other social safety net programs such as the labor-intensive program and the provision of free medical assistance through the health card program may have had limited success (Hardjono, 1999), the direct income transfer due to the subsidized rice program and the scholarship program may have made a difference of up to 15% in the total income of a poor household. A family able to purchase 20 kg of subsidized rice per month at Rp. 1,000 per kg, compared with the prevailing market price of Rp. 2,500 per kg in 1998, could save Rp. 30,000 per month. Add to this the value of a primary school scholarship of Rp. 20,000 per month, and a family of four members living in the neighborhood of the CBS (old) rural poverty line of Rp. 76,000 per capita or Rp. 304,000 per household per month in early 1999, could have increased its income by 16% if registered in both these programs. The subsidized rice program alone could have added around 10% to the income of a poor rural household (see also a comparable estimate of 9% by Tabor & Sawit (1999)). Depending on their participation in either or both programs, this is equivalent to preventing 7–12% of households from falling below the poverty line, using the poverty elasticity estimate with respect to the poverty line calculated earlier. How can the former view that the rice subsidy and scholarship programs did not deliver their benefits to the poor be reconciled with the opposing view, using the same data sets in some cases, that they were relatively successful at reaching their intended beneficiaries? In an attempt to resolve this question, the separate issues of coverage and targeting should be examined in turn. First, all the above evaluations agree that, while the coverage of the rice subsidy program was adequate, this was not the case for the scholarship program, nor for other

1225

SSN programs for that matter. In the absence of donor funds for scholarships at the primary and senior secondary school level, the government’s own resources only targeted 6% of primary school children (and in grades 4, 5 and 6 only), and 10% of senior secondary school students (World Bank, 1999, p. 5). Turning now to targeting, the rice subsidy program, being an entirely government-funded program, did not benefit from independent donor-funded monitoring, unlike the education and health programs. Lack of sufficient data, as well as weaknesses in the existing data, preclude a definite assessment of the effectiveness of its targeting. The SSN module of the Susenas survey of early February 1999 was designed to fill this gap, however, it has produced more questions than answers. This survey was in any case fielded too early in the program cycle, barely five to six months into implementation of what were ambitious nationwide programs, and probably without adequate pre-testing. The timing of the survey, and ambiguities in the questionnaires on scholarships, may have led to potentially large under-reporting (Jones & Hagul, 2001, p. 223) as well as misreporting, such as the almost equivalent proportion of primary school grades 1 to 3 children claiming to be recipients (Sumarto et al., 2001, p. 27), a finding which has yet to be reported elsewhere. Questionnaire design may also have affected the response to the rice subsidy program, though in this case, the number of households allegedly receiving rice from the OPK program may have been overestimated, at 20 million households compared to 10 million households in administrative records (Sumarto & Suryahadi, 2001, p. 20). In addition to the above reliability problems affecting the Susenas 1999 SSN module, one must add the issue of data interpretation. Two studies, using this same data set, arrive at quite different estimates of beneficiaries, one finding that 7.1% and 0.9% of primary school scholarship recipients came from respectively the poorest and richest quintile (Jones & Hagul, 2001, p. 223), while corresponding figures for the second study are 5.8% and 2.0% (Sumarto et al., 2001, p. 54). In both cases, the total proportion of recipients was 4.0%, with very different implications for targeting effectiveness, a discrepancy that neither study addresses. Moreover, the SMERU interpretation was based on its designation of the first quintile only of the sample households as poor. In the absence of any other way of identifying the

1226

WORLD DEVELOPMENT

location of poor households, the OPK and scholarship programs targeted the households classified as ‘‘pre-prosperous’’ and ‘‘prosperous stage I’’ of the National Family Planning Board. Before the crisis, these two categories accounted for nearly 40% of all households (16% and 22% in each), a share which rose to nearly 50% (23% and 26%) of all households at the peak of the crisis in mid-1998 and early 1999 (Dhanani & Islam, 2000, p. 26), a ratio not very different from that shown in Figure 2 above. Were the SMERU results to be re-calculated using the lowest two quintiles as poor, as would appear to be more representative of the absolute and transient poor population during the crisis, their estimates of program leakages would be considerably reduced, as indeed demonstrated by Jones and Hagul (2001). Two final comments apply to the findings of Suryahadi et al. (1999) and Sumarto et al. (2001). The first of these relied on a supplementary questionnaire to the periodic ‘‘100 Village Survey’’. This questionnaire did not specifically identify the OPK program by name at a time when many institutions such as the army, political parties, NGOs, religious organizations, companies and wealthy individuals provided food parcels to the needy. The second study, while noting that the low program coverage in many districts affected poor and nonpoor households alike, does not give due weight to the constraints of rice availability and distribution problems as opposed to deficient targeting. Furthermore, in a situation of general rice shortage in the peak of the crisis, nonpoor households may have had little choice but to purchase rice from the rice subsidy program delivered in their neighborhood. The above discussion of the impact of the rice subsidy and scholarship programs suffers from lack of adequate and sufficiently good quality data, which should be remedied as soon as possible, as well as the dearth of independent and comprehensive monitoring and evaluation, which SMERU has commendably sought to address to some extent. Nevertheless, the evidence gathered so far from large surveys as well as field studies leads one to reach a preliminary assessment that the rice subsidy and scholarship programs performed reasonably well under the circumstances, given the speed at which they were forced into implementation, and numerous constraints in the field in a large and diverse country such as Indonesia.

To illustrate the appropriateness of the above conclusion, it is worth highlighting a recent SMERU report that observed: The OPK program is an example of a crisis initiative that worked reasonably well. . . Pressure at the local level for a ‘‘fairer’’ distribution of the rice was overwhelming, since the ‘‘almost poor’’ or the ‘‘newly poor’’ families had no official entitlement to the subsidized rice. . . Hence it is possible that some of what was recorded as going to those who are ‘‘noneligible’’ is not really mis-targeting, but is a justifiable correction to the official eligibility criteria (Sumarto & Suryahadi, 2001, p. 17).

The evidence and discussion of the monetary impact of the SSN programs presented suggest that some social protection measures taken by the government to combat the crisis may have been relatively effective. The discussion of the efficacy of government intervention in protecting the poor will be incomplete unless the relative significance of other explanatory variables in understanding the transient nature of the crisis-induced increase in poverty is considered. It is possible to argue that the reason why the social impact of the crisis was muted, and apparently shortlived, is rooted in the coping mechanisms of households. Certainly, both the ‘‘old’’ poor (those who were poor pre-crisis) and the ‘‘new’’ poor (those who descended into poverty as a result of the crisis) have tried to cope with the pressures of a recession-cum-inflation shock in 1998 in a number of ways. Thus they sold assets, reduced consumption of micronutrientrich food, cut down on ‘‘nonessential expenditure,’’ sought refuge in the agricultural sector and the informal sector and migrated overseas. Agricultural employment rose from 41% to 45% between 1997 and 1998 (ILO, 1999, p. 36); nonwage or informal employment increased from 65% to 68% (ILO, 1999, p. 37); legal migration increased from 235,000 to 412,000 between 1997–98 and 1998–99 (ILO, 1999, p. 347), Figures which are likely to be considerably higher if one allows for illegal migration. These adjustments also enable one to understand why there was a relatively moderate rise in open unemployment from 4.7% to 5.4% during 1997–98 (ILO, 1999, p. 26). They also imply that the labor market adjustment due to the crisis-induced recession was not through open unemployment but through a drop in real wages. Ultimately, the coping mechanisms reflected a combination of individual determination to survive and the resilience of social

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS

capital, as assistance flowed to the victims of the crisis through the informal network of friends and family. At least one study claims that the average value of assistance received from informal sources was significantly higher than the average value of such assistance from formal sources (Frankenberg, Beegle, Sikoki, & Thomas, 1998). Clearly consumption smoothing at the household level in a context of well-developed social capital is an important part of the story in understanding the dynamics of the social impact of the Indonesian crisis. It is, however, difficult to believe that such processes alone can explain why poverty has stabilized and that the worst may be over, in the absence of an antiinflation strategy and OPK-type interventions as well as the scholarship program. Had inflation continued unabated at the rate seen in 1998, and had social safety net interventions been absent, household-level coping mechanisms and the informal network of support forged together by friends and family in the crisis period would have been overwhelmed. Ominous scenarios of an uncontrollable social crisis would have become a stark reality. Hence, the need to combine macroeconomic stability with a long-term, fiscally sustainable social protection policy which was able to reinforce existing social capital. 13 4. CONCLUSION Several themes emerge from this paper. First, a distinction should be made between capability poverty and poverty based on current consumption. The former focuses on such nonincome dimensions as education and health. Measured along such dimensions, around 25% of Indonesians were unable to meet basic needs even before the crisis. This is considerably higher than the 11% incidence of consumption-based poverty that gained wide currency in discussing the outcomes of the Soeharto regime. The publicity that such a statistic generated conveyed the optimistic implication that the incumbent government managed to ‘‘solve’’ poverty as a generic problem and that future strategies ought to focus on fighting pockets of poverty––among isolated and remote communities, the handicapped, the old and the infirm etc. The failure to recognize the significant incidence of capability poverty thus belittled the challenges that remained. The current government ought to

1227

redirect the debate on fighting poverty in Indonesia by focusing on capability poverty rather than consumption-based indicators of deprivation. At least, CBS ought to be encouraged to eschew its focus on consumptionbased indicators of poverty by aligning them with indicators of capability poverty. The second major theme that emerges is the crucial need to distinguish between transient poverty and its long-term behavior. Consumption-based indicators of poverty are extremely sensitive to variations in prices. In a highinflation environment, which characterized the Indonesian crisis, this can lead to a good deal of volatility in the poverty line. This, combined with the fact that a significant component of the Indonesian population is clustered around the poverty line, can lead to large––but transient––shifts in poverty incidence when measured by consumption indicators. Capability poverty, on the other hand, is an underlying structural––and even chronic––phenomenon. It responds gradually to long-term growth and government interventions to provide the community with broad-based access to basic services. Not surprisingly, capability poverty behaved in a much more stable manner even during the crisis than measures of deprivation based on current consumption. The third key message is the necessity of distinguishing between overall poverty and the severity of poverty in fully appreciating the social consequences of the Indonesian crisis. The popular headcount ratio merely looks at the overall numbers in poverty but is unable to distinguish between the marginally poor and the very poor. Using a number of indicators, the paper was able to demonstrate that the incidence of extreme poverty rose faster than the incidence of overall poverty. The primary policy message of this paper is that the reversal of the sharp increase in poverty did not happen by accident. Government interventions played an important role, even though they were introduced on a national scale almost a year after the start of the crisis, perhaps because the government underestimated the impact of the crisis and food price hikes on the poor and near-poor. An anti-inflation strategy combined with exchange rate stability managed to bring the ‘‘inflation shock’’ of 1998 under control. Indeed, for the first nine months of 1999, Indonesia experienced deflation. Given the basic premise that consumption-based indicators of poverty are highly sensitive to inflation, the control of inflation

1228

WORLD DEVELOPMENT

and the subsequent onset of deflation partly helps to explain why the crisis-induced swelling in the ranks of the poor turned out to be transient. Finally, the paper argues that an anti-inflation strategy within a macroeconomic framework cannot adequately explain why the social consequences of the Indonesian crisis turned out to be less severe than initially anticipated. Since food intake dominates the consumption bundle of the poor, the government sensibly sought to offer subsidized rice to poor households. This became a key component of the government’s social protection policy. The evaluations reviewed in this study suggest that,

on balance, it was effective in providing some protection to the poor during the crisis. Another possible success story is the scholarship program, which sought to protect the human capital investments of the poor by seeking to stabilize school enrollment rates. When converted into cash equivalent, the rice subsidy program and the scholarship program represented a significant share of a poor household’s income. The paper was inspired by these success stories to suggest that a fiscally sustainable social protection policy that is able to reinforce household coping mechanisms and social capital should become part of Indonesia’s mediumterm strategy for battling poverty.

NOTES 1. A thorough review of the debate can be found in Booth (1999a,b). See also UNSFIR (1999). 2. This paper is a revised and updated version of a working paper prepared for UNSFIR, Jakarta (see Dhanani & Islam, 2000). 3. The recent poverty assessment literature has increasingly focused its attention on ‘‘participatory poverty assessment’’ (PPA) techniques. The approach tries to delineate the nonincome dimensions of poverty by drawing on discussions with poor men and women and other stakeholders. A major volume has recently been compiled by Narayan et al. (1999), which draws on 78 PPA reports ranging across 47 countries, including Indonesia. The authors are struck by the ‘‘. . .commonality of the human experience of poverty across countries’’ (Narayan et al., 1999, p. 6) and highlight five dimensions of poverty. These include: lack of access to food; lack of access to basic infrastructure, rural roads, transportation and water; such psychological dimensions as powerlessness, voicelessness, dependency, shame and humiliation; a thirst for literacy; and the need to manage assets (physical, human, social and environmental), rather than merely income, as a way of coping with the vulnerability of poverty. See Narayan et al. (1999, p. 7). 4. Kanbur and Squire (1999, p. 7) observe: ‘‘the distinction between transient and chronic poverty has emerged as an important issue in the context of the East Asian crisis.’’ 5. See Dhanani (1994, pp. 21–34) for a detailed discussion of changes in CBS methodology up to 1993.

6. See also Islam and Khan (1986) who develop a statistical profile of poverty and inequality in Indonesia using 1976 province-level data. 7. The World Bank uses the following scaling method (World Bank, 1993, Annex 1.2, p. 111): z ¼ zf ð2  aÞ where z is the total poverty line, zf is the food poverty line and a is the food share. This is different from the usual Orshansky method of scaling up the food poverty line as follows: z ¼ zf =a. For example, if zf ¼ 100 and a ¼ 0:65, the World Bank method will produce z ¼ zf ð2  0:65Þ ¼ 135, while the Orshansky method will produce z ¼ zf ð0:65Þ ¼ 154, a total poverty line which is 14% higher than that of the World Bank, and which would correspondingly produce significantly higher estimates of poverty incidence. For details of the Orshansky method see Booth (1993) and World Bank (1993, p. 104). 8. See UNSFIR (1999) for a discussion of the regional dimensions of Indonesian poverty during the crisis. 9. The authors are grateful to an anonymous referee for drawing attention to this point. 10. It should be noted that CBS was not the only agency to underestimate consumption poverty in Indonesia. The World Bank, in applying its ‘‘global method’’ for estimating poverty to pre-crisis Indonesia, reported a figure of 11.8% in 1995 and 7.7% in 1996 (see World Development Reports 1998/99 and 1999/2000, p. 236). This was based on its widely advocated ‘‘one dollar a day poverty line’’ derived from the purchasing power parity (PPP) method. The first estimate appears pretty

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS close to CBS’s estimate of 11.3% in 1996. In a report released in 2000, the Indonesia office of the World Bank drew attention to the use of ‘‘penurious’’ and ‘‘narrow’’ poverty definitions in pre-crisis Indonesia (World Bank, 2000a, Executive Summary). For a discussion of the appropriateness of the PPP method or the international poverty line approach as applied to Indonesia, see Dhanani and Islam (2000). See also an ADB study by David, Asra, and Castro (1999, p. 14) who found that the World Bank methodology underestimated poverty incidence relative to national poverty lines, sometimes by quite high margins, in eight out of eleven Asian countries reviewed. The study noted that ‘‘. . .in its present formulation, the international poverty line approach has limited applicability in quick monitoring of the impact of a crisis or a program on a country or a group of countries.’’ 11. The September 1998 figure is based on the nominal expenditure distribution of the Mini-Susenas survey

1229

December 1998. It has not been adjusted for possible substitution effects due to relative prices and higher imputed consumption of own products caused by higher food prices (CBS/UNDP, 1999, p. 58). Neither has it been adjusted for possible changes in nominal incomes and expenditures in the intervening period. The net effect of these two changes therefore remains unknown. 12. See also Cameron (2001) who also uses the ‘‘100 Villages Survey’’ to study the impact of the Indonesian financial crisis on children and, in this context, draws attention to the role of the SSN programs. 13. This also seems to be the key message of the growing literature on the role of social protection policies in developing countries. (See, for example, Morduch, 1999; Subbarao et al., 1998.) There is also an extensive literature on social capital and its role in economic development. (See, for example, Evans, 1996; Collier, 1998; Woolcock, 1998.)

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. Bidani, B., & Ravallion, M. (1993). A regional poverty profile for Indonesia. Bulletin of Indonesian Economic Studies, 29, 37–68. Booth, A. (1993). Counting the poor in Indonesia. Bulletin of Indonesian Economic Studies, 29(1). Booth, A. (1997). Poverty in Indonesia. South Asia Multidisciplinary Advisory Team Working Paper. New Delhi: International Labor Organization. Booth, A. (1999a). The impact of the crisis on poverty and equity. In H. W. Arndt, & H. Hill (Eds.), Southeast Asia’s Economic Crisis: Origins, Lessons and the Way Forward. Singapore: Institute of Southeast Asian Studies. Booth, A. (1999b). Survey of recent developments. Bulletin of Indonesian Economic Studies, 35(3), 3– 38. Breman, J. (2000). The impact of the Asian crisis on work and welfare in village Java. Dies Natalis 2000 Address, delivered on October 12 on the occasion of the 48th Anniversary of the Institute of Social Studies. The Hague: Institute of Social Studies. Cameron, L. (2001). The impact of the Indonesian financial crisis on children: an analysis using 100 villages survey. Bulletin of Indonesian Economic Studies, 37(10), 43–64. CBS (2000a). Perkembangan Tingkat Kemiskinan dan Beberapa Dimensi Sosial-Ekonominya 1996–1999 (Evolution of Poverty Level and Socio-economic Dimensions 1996–1999). Publication series Minis Susenas 1999, Book 2, Catalogue no. 4706. Jakarta: Central Bureau of Statistics. CBS (2000b). Penyempurnahaan Metodologi Penghitungan Penduduk Miskin dan Profil Kemiskinan 1999

(Improving Methodology for Calculating Poverty Level and Poverty Profile 1999). Catalogue no. 2321. Jakarta: Central Bureau of Statistics. CBS, (2000c). Statistical Yearbook of Indonesia, Jakarta: CBS. CBS/UNDP (1999). Crisis, poverty and human development in Indonesia, 1998. Jakarta: Central Bureau of Statistics and United Nations Development Program. Chesher, A. (1998). Local poverty lines and poverty measures for Indonesia, Report Prepared for the World Bank, Department of Economics, University of Bristol, UK as cited in Pradhan et al. (2000). Collier, P. (1998). Social Capital and Poverty. Social Capital Initiative Working Paper No. 4. Washington, DC: World Bank. David, I. P., Asra, A., & Castro, M. (1999). Poverty incidence in the Asian and Pacific Region: data situation and measurement issues. Mimeo. Manila: Asian Development Bank. Dhanani, S., & Islam, I. (2000). Poverty, inequality and social protection: lessons from the Indonesian crisis. UNDP/United Nations Support Facility for Indonesian Recovery (UNSFIR) Working Paper 00/01. Jakarta: United Nations Development Program. Dhanani, S., & Islam, I. (2001). Indonesian wage structure and trends, 1976–2000. Background paper prepared for the Infocus Socio-Economic Security Program (ILO/SES). Geneva: International Labor Organization. Dhanani, S. (1994). A review of alternative national and regional poverty measures and trends in Indonesia. Interim technical report no. 22, Regional Manpower Planning and Training Project. Jakarta: National Development Planning Board.

1230

WORLD DEVELOPMENT

Evans, P. (1996). Government action, social capital and development: renewing the evidence on synergy. World Development, 26(6), 1119–1132. Foster, J., Greer, J., & Thorbecke, E. (1984). A class of decomposable poverty measures. Econometrica, 52, 761–766. Frankenberg, E. K., Beegle, B., Sikoki, & Thomas, D. (1998). Health, Family Planning and Well-being in Indonesia during an economic crisis, Final Report, December, The Policy Project. Hardjono, J. (1999). The micro data picture: results of a SMERU social impact survey in the PurwakartaCirebon corridor. Field report, Social Monitoring and Early Response Unit (SMERU). Jakarta: SMERU. Helen Keller International (1999). Indonesia’s crisis: a comparison of its impact on nutrition and health in urban and rural populations. Jakarta: Helen Keller International. Ikhsan, M. (1999). Some notes on poverty line calculation in Indonesia. Paper presented at Seminar on ‘‘Poverty, Inequality and the Indonesian Crisis: Issues in Methodology’, 28 April. Jakarta: United Nations Support Facility for Indonesian Recovery (UNSFIR). ILO (1999). Indonesia: strategies for employment-led recovery and reconstruction. Geneva: International Labor Organization. ILO/UNDP (1998). Employment challenges of the Indonesian economic crisis. Jakarta: International Labor Organization. Islam, I., & Khan, H. (1986). Spatial patterns of poverty in Indonesia: a taxonomic approach. Bulletin of Indonesian Economic Studies, 22(2). Jahari, A. B., Sandjaja, I. J., Soekirman, Herman, S., & Jalal, F. (1999). Nutrition status of under fives in Indonesia during the period 1989 to 1999. Paper presented at National Workshop on Food and Nutrition, November 23–25, Jakarta: Indonesian Academy of Sciences (LIPI). Jones, G. W., & Hagul, P. (2001). Schooling in Indonesia: crisis-related and longer-term issues. Bulletin of Indonesian Economic Studies, 37(2), 207– 221. Kanbur, R., & Squire, L. (1999). The evolution of thinking about poverty: exploring the interactions. Mimeo. Washington, DC: World Bank. Levinsohn, J., Berry, S., & Friedman, J. (1999). Impacts of the Indonesian crisis: price changes and the poor. Working Paper No. 7194. Cambridge, MA: National Bureau of Economic Research. Morduch, J. (1999). Between the market and the state: can informal insurance patch the market? Stiglitz Summer Research Workshop on Poverty, 6–8 July. Washington DC: World Bank. Narayan, I. et al. (1999). Can anyone hear us? Voices from 47 Countries, vol. 1, Poverty Group, PREM. Washington, DC: World Bank. Pradhan, M., Suryahadi, A., Sumarto, S., & Pritchett, L. (2000). Measurements of poverty in Indonesia, 1996, 1999, and beyond. Social Monitoring and Early Response Unit (SMERU), Jakarta. World Bank

Working Paper 2437. Washington, DC: World Bank. Pritchett, L., Suryahadi, A., & Sumarto, S. (2001). Quantifying vulnerability to poverty: a proposed measure, applied to Indonesia. Social Monitoring and Early Response Unit (SMERU), Jakarta. World Bank Working Paper 2437. Washington, DC: World Bank. Rachman, L., Pujihastuti, U., Sabannah, S., & Sukandar, D. (1999). Deskripsi Hasil Evaluasi Pelaksanaan Operasi Pasar Khusus Beras. In S. R. Tabor, & M. H. Sawit (Eds.), The OPK Program: Economy-wide Impacts Prepared for the State Ministry of Food and Horticulture. Jakarta: Economic Managedment Services International. Rao, V. V. B. (1983). Poverty in Indonesia, 1970–1980: trends, associated characteristics and research issues. Report No. DC-248. Washington, DC: World Bank. Ravallion, M., & Bidani, B. (1994). How robust is a poverty profile? The World Bank Economic Review, 8(1), 75–102. Sen, A. K. (1976). Poverty: an ordinal approach to measurement. Econometrica, 44(2), 219–231. Sen, A. K. (1999). Development as Freedom. New York: Alfred Knof. Skoufias, E., Suryahadi, A., & Sumarto, S. (1999). The Indonesian Crisis and Its Impacts on Household Welfare, Poverty Transitions and Inequality: Evidence From Matched Households in 100 Village Survey. Research working paper, Social Monitoring and Early Response Unit (SMERU). Jakarta: SMERU. Sri Kusumastuti, R. et al. (1999). Implementation of Special Market Operation (OPK) Program: Results of a SMERU Rapid Field Appraisal Mission in Five Provinces. Field report, Social Monitoring and Early Response Unit (SMERU). Jakarta: SMERU. Subbarao, K. et al. (1998). Safety net programs and poverty reduction: lessons from cross-country experience. Directions in Development Series. Washington, DC: The World Bank. Sumarto, S., & Suryahadi, A. (2001). Principles and approaches to targeting: with reference to the Indonesian social safety net programs. Research working paper, Social Monitoring and Early Response Unit (SMERU). Jakarta: SMERU. Sumarto, S., Suryahadi, A., & Widyanti, W. (2001). Designs and implementation of the Indonesian social safety net programs: evidence from the JPS module of the 1999 Susenas. Research working paper, Social Monitoring and Early Response Unit (SMERU). Jakarta: SMERU. Suryahadi, A., Suharso, Y., & Sumarto, S. (1999). Coverage and targeting in the social safety net programs: evidence from 100 village survey. Research working paper, Social Monitoring and Early Response Unit (SMERU). Jakarta: SMERU. Suryahadi, A., Sumarto, S., Suharso, Y., & Pritchett, L. (2000). The evolution of poverty during the crisis in Indonesia, 1996 to 1999 (using full Susenas sample). Research working paper, Social Monitoring and Early Response Unit (SMERU). World Bank Research Working Paper 2435. Jakarta: SMERU.

POVERTY, VULNERABILITY AND SOCIAL PROTECTION IN A PERIOD OF CRISIS Sutanto, A., Irawan, P. B., & Said, A. (1999). Poverty measurement: problems and development. Paper presented at the International Conference on Methodologies of Poverty Calculation in Indonesia, 30 November. Jakarta: Central Bureau of Statistics and World Bank. Sutanto, A., & Irawan, P. B. (2000). Regional dimensions of poverty: some findings on the nature of poverty. Paper presented at the International Conference on Poverty Measurement in Indonesia, 16 May. Jakarta: Central Bureau of Statistics and World Bank. Tabor, S.R., Sawit, M.H. (1999). The OPK Program: economy-wide impacts. Prepared for the State Ministry for Food and Horticulture. Jakarta: Economic Management Services International. UNDP (1997). Human development report 1997. Oxford University Press: New York and Oxford for United Nations Development Program. UNDP (2001). Human development report 2001. Oxford University Press: New York and Oxford for the United Nations Development Program. UNSFIR (1999). The social implications of the Indonesian economic crisis: perceptions and policy. Discus-

1231

sion Paper No. 1, project of United Nations Development Program. Jakarta: United Nations Support Facility for Indonesian Recovery (UNSFIR). Woolcock, M. (1998). Social capital and economic development: toward a theoretical synthesis and policy framework. Theory and Society, 27(2), 151– 208. World Bank (1993). Indonesia: Public Expenditures, Prices and the Poor. Report no. 11293-IND, Indonesia resident mission and country department III (East Asian and Pacific). Jakarta: World Bank. World Bank (1999). Back to-school Indonesia 1998/99: what worked well? what didn’t. A report of the scholarships and Grants Program, July 1999. Jakarta: World Bank. World Bank (2000a). Poverty reduction in Indonesia: constructing a new strategy, Indonesia Country Office. Jakarta: World Bank. World Bank (2000b). Indonesia: a macroeconomic update. Jakarta: World Bank. World Bank (various years). World development report. Washington, DC: World Bank.