the determinants of poverty dynamics in indonesia: evidence … · 2014. 5. 5. · model to examine...

This article was downloaded by: [Macquarie University]On: 08 June 2013, At: 04:38Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Bulletin of Indonesian Economic StudiesPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/cbie20

The determinants of poverty dynamicsin Indonesia: evidence from panel dataTeguh Dartanto a & Nurkholis aa University of Indonesia, JakartaPublished online: 21 Mar 2013.

To cite this article: Teguh Dartanto & Nurkholis (2013): The determinants of poverty dynamics inIndonesia: evidence from panel data, Bulletin of Indonesian Economic Studies, 49:1, 61-84

To link to this article: http://dx.doi.org/10.1080/00074918.2013.772939

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representationthat the contents will be complete or accurate or up to date. The accuracy of anyinstructions, formulae, and drug doses should be independently verified with primarysources. The publisher shall not be liable for any loss, actions, claims, proceedings,demand, or costs or damages whatsoever or howsoever caused arising directly orindirectly in connection with or arising out of the use of this material.

http://www.tandfonline.com/loi/cbie20

http://dx.doi.org/10.1080/00074918.2013.772939

http://www.tandfonline.com/page/terms-and-conditions

Bulletin of Indonesian Economic Studies, Vol. 49, No. 1, 2013: 61–84

ISSN 0007-4918 print/ISSN 1472-7234 online/13/010061-24 © 2013 Indonesia Project ANUhttp://dx.doi.org/10.1080/00074918.2013.772939

THE DETERMINANTS OF POVERTY DYNAMICS IN INDONESIA:

EVIDENCE FROM PANEL DATA

Teguh Dartanto* Nurkholis*University of Indonesia, Jakarta

We use the ‘spell’ approach to identifying poverty and apply an ordered logit model to examine the determinants of poverty dynamics in Indonesia, categorising households as poor, transient poor (–), transient poor (+) or non-poor. Observing the National Socio-Economic Survey (Susenas) balanced-panel data sets of 2005 and 2007, we found that 28% of poor households are classified as chronically poor (that is, remaining poor in two periods) while 7% of non-poor households are vulnerable to being transient poor (–). Our estimations confirmed that the determinants of pov-erty dynamics in Indonesia are educational attainment, the number of household members, physical assets, employment status, health shocks, the microcredit pro-gram, access to electricity, and changes in employment sector, employment status and the number of household members. We also found that households in Java–Bali are more vulnerable to negative shocks than those outside Java–Bali.

Keywords: poverty dynamics, transient poverty, vulnerability, shocks, government assistance

BACKGROUNDPoverty in Indonesia has been much researched, but most of this research – for example, Bidani and Ravallion (1993) – focuses on static poverty and analyses the proportion of the population falling below a given income threshold at a given time. Poverty is not, however, a purely static phenomenon (Muller 2002); house-holds currently not poor may later fall below the poverty line, owing to shocks

* We would like to thank the University of Indonesia and the Directorate General of High-er Education, Ministry of National Education, the Republic of Indonesia, for financing this research through the National Research Strategic Fund 2010 (DRPM/Hibah Strategis Na-sional/2010/I/4024). We would also like to thank Ms Lily Yunita and Mr Usman from the Institute for Economic and Social Research, University of Indonesia, for their assistance. We would like to thank Professor Mohamad Ikhsan (University of Indonesia) for providing Susenas panel data sets. Professor Shigeru Otsubo (Nagoya University) and his seminar participants provided valuable comments, as did Professor Hal Hill, and other partici-pants of the 2011 Singapore Economic Review Conference, and Assistant Professor Mark Rebuck. We would also like to thank the three anonymous referees for their constructive and valuable comments and suggestions, which helped improve the quality of this paper. Any remaining errors are our responsibility.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013

http://dx.doi.org/10.1080/00074918.2013.772939

62 Teguh Dartanto and Nurkholis

such as crop loss, job loss or death.1 Conversely, poor households may escape from poverty by gaining employment or a better job (Fields et al. 2003), increasing their level of education (Herrera 1999) or having access to improved infrastruc-ture (Sawada et al. 2008).

Under President Susilo Bambang Yudhoyono, the Indonesian government has changed its poverty alleviation policies from a macro, top-down approach to a community or household participatory approach. It has developed and implemented several policies to alleviate chronic poverty, including educational subsidies (Bantuan Operasional Sekolah), scholarships, conditional cash trans-fers, community empowerment programs (Program Nasional Pemberdayaan Masyarakat), credits for small and medium enterprises (SMEs) (microfinance) and infrastructure development projects (Program Pengembangan Kecamatan). The government has also provided social safety nets, including subsidised rice (Raskin), cash transfers (Bantuan Langsung Tunai) and health insurance targeted to the poor (Askeskin).2 These policies are intended to address transient poverty and protect the poor from external shocks. However, the effectiveness of govern-ment policies in alleviating poverty is questionable. It can be difficult to evaluate the impact of poverty alleviation policies in the short term, because many policies experience a delay between implementation and results. For instance, the impact of microcredit on SMEs often becomes apparent only after two or more years, requiring a longer period of observation. Further, it is generally acknowledged that the impact of investment in human capital (such as education and health) on household welfare cannot be investigated immediately.

As the poverty incidence can change over time, it is important to conduct a dynamic analysis to distinguish between chronic poor, transient poor and never poor; to discover which determinants differentiate among groups; and to evaluate the effectiveness of government policies in changing poverty status in Indonesia. The distinction between chronic and transient poverty is important not only for an accurate measurement of poverty but also for policy implication purposes; chronic and transient poverty call for different alleviation strategies. In a country or region where the poverty problem is characterised by the chronically poor, the appropriate strategy would be to redistribute assets, providing basic physical and human-capital infrastructure. If the predominant poverty problems relate to transient poverty, the strategy should be geared towards providing safety nets and coping mechanisms to reduce the vulnerability of the poor and to help them return to a non-poor situation (Hulme and Shepherd 2003; McCulloch and Calan-drino 2003).

This paper has three objectives. First, it aims to contribute to the literature of poverty studies. To date, there has been very little analysis of poverty dynamics in Indonesia – for example, studies that investigate the welfare movements of a

1 For example, Contreras et al. (2004) found that health problems were correlated with households falling into poverty in Chile. Dercon and Krishnan (2000) showed that risk contributes to poverty fluctuations in Ethiopia.2 Sparrow, Suryahadi and Widyanti (2012), using the National Socio-Economic Survey (Susenas) panels of 2005 and 2006, showed that Askeskin increases the use of outpatient health care among the poor. This policy may therefore protect households from falling into transient poverty because of health shocks.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013

The determinants of poverty dynamics in Indonesia: evidence from panel data 63

set of households over time. Second, it provides information to deepen under-standing of poverty in present-day Indonesia, particularly the factors that deter-mine households’ movements into and out of poverty and why some households remain poor. Third, as a pioneer paper, it deals with how economic shocks and risks, government assistance, and changes in socio-economic variables can change poverty status in Indonesia.3

This paper explains the concepts of poverty dynamics, and then describes the changes in household poverty status in Indonesia during 2005-07. It reviews the research methods of the ordered logit model and analyses the estimation results of determinants of poverty dynamics. It ends with some important findings and policy suggestions.

THEORETICAL FRAMEWORK

Concepts and measures of chronic and transient poverty, based on panel dataResearchers commonly identify and measure chronic and transient poverty (income- and consumption-based poverty) on the basis of panel data, using the ‘spell’ and ‘components’ approaches (Yaqub 2000). The spell approach focuses on the number or length of spells of poverty experienced by households, because the defining feature of both chronic and transient poverty is their extended duration (Hulme and Shepherd 2003). The components approach distinguishes the per-manent component of a household’s income or consumption from its transitory variations, classifying the chronically poor as those whose permanent component – commonly the intertemporal average of household income or consumption – is below the poverty line (McKay and Lawson 2003).

Under the spell approach, the term ‘chronic poor’ indicates that consumption expenditure or household income remained below the poverty line in all observa-tion periods. ‘Transient poor’ indicates that consumption expenditure or house-hold income was not always below the poverty line and was sometimes above it. ‘Non-poor’ (never poor) indicates that consumption expenditure or household income remained above the poverty line in all observation periods (Hulme, Moore and Shepherd 2001). Figure 1 shows a simple illustration of the spell approach.

Chronic poverty, then, can be described as the household condition of being poor over an extended period, whereas transient poverty (– or +) refers to a state of occasionally being poor or non-poor. This difference is typically based on longitudinal (or panel) data or a life-history survey, both of which observe the living conditions of the same individual or households at several points in time (McKay and Lawson 2002). The former provides information about individuals or households during one period or consecutive periods, while the latter captures the dynamic aspect of living conditions from a list of retrospective questions. A life history – for instance, the weight-for-height anthropometric measure – can fluctuate significantly in a short time. These fluctuations may reflect, for example,

3 Dercon and Shapiro (2007) contend that the impact of shocks and risks on poverty mo-bility has received relatively limited attention in the literature of poverty dynamics. Ana-lysing poverty dynamics can provide insights into the effects of socio-economic and anti-poverty policies on household poverty status. It can also help policy makers to identify effective ways to help households escape poverty.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


the period of the agricultural season or the effects of chronic disease. Over an extended observation period, an individual with a weight-for-height measure-ment below the standard could therefore be categorised as chronically poor, whereas an individual with a weight-for-height measurement occasionally equal to or below the standard could be categorised as transiently poor.

Previous research on poverty dynamicsMany studies have found that human capital, demographics, geographical loca-tion, physical assets and occupational status help determine poverty status. An increase in human capital, for example, indicated by educational attainment (years of schooling), decreases the probability of being chronically poor and improves a household’s ability to respond to transitory shocks (Adam and Jane 1995; Alis-jahbana and Yusuf 2003). Jalan and Ravallion (1998) contend that changes in demographics, such as increased household size, are related to chronic poverty. McCulloch and Calandrino (2003), in Sichuan, China, confirmed that chronic pov-erty is commonly found in rural (especially remote) areas, and Fields et al. (2003) found that households in urban areas have a higher probability of escaping from poverty. A lack of physical assets is also often associated with chronic poverty (Adam and Jane 1995), and occupational status can help determine household poverty status. Okidi and Kempaka (2002) found that self-employed farming households in Uganda are more likely to be chronically poor, and Kedir and McKay (2005) found that households in Ethiopia with a head working as a waged employee can escape poverty.

Chronic poor

Transient poor (+) Never poor

Y2

Z2

Y1Z10

Transient poor (–)

a Y1 and Y2 represent individual or household income or consumption in period 1 and period 2 respectively. Variables Z1 and Z2 represent the poverty lines in the same periods.

Source: Adapted from Grab and Grimm (2007).

FIGURE 1 The Distinction between Chronic Poor, Transient Poor (–), Transient Poor (+) and Never Poora

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


Grab and Grimm (2007), using the Indonesia Family Life Survey (IFLS) data set to measure poverty dynamics, compared chronic and transient poverty in two timespans. They found that absolute comparisons show a significant decline in chronic poverty from 1993–97 to 1997–2000. The decline in both chronic and tran-sient poverty was largely due to a substantial decline in poverty in rural Indone-sia. Fields et al. (2003), using the 1993 and 1997 IFLS panel data sets, found that the determinants of household income dynamics at those times were household location, age of the household head, employment status of the head, change in the gender of the head, change in employment status of the head, and change in the number of children. Alisjahbana and Yusuf (2003), using the IFLS data sets from 1993 and 1997, observed that of the 84.8 percentage points of non-poor in 1993, 11.6 percentage points had fallen into poverty by 1997. Of the 15.2 percentage points of poor in 1993, 7.8 percentage points had remained poor whereas the other 7.4 percentage points had escaped poverty.

OVERVIEW OF POVERTY DYNAMICS IN INDONESIA DURING 2005–07

Susenas and the economic condition We use data from the 2005 and 2007 National Socio-Economic Survey (Suse-nas), conducted by Indonesia’s Badan Pusat Statistik (BPS), the national statis-tics agency, to measure poverty dynamics in Indonesia.4 Susenas consists of two main data sets: core and module. The 2005 Susenas core data set recorded detailed characteristics of 278,352 households, from an estimated 59 million households nationally and covering various geographic regions of Indonesia. The 2005 Suse-nas module data set collected additional information on a subset (or 68,288 house-holds) of core households. It recorded detailed information on food and non-food consumption, as well as on household shocks and coping strategies.

BPS revisited around 10,600 households from the 2005 Susenas module sample in 2007. Merging the 2005 and 2007 Susenas panels and dropping observations that contained incomplete household information or that were outliers yielded a total of 8,726 households (balanced-panel data). The 2007 Susenas did not revisit those households that migrated, so the 8,726 revisited households are those that remained in one location during 2005–07.5

4 We intended to use a longer sequence of Susenas data sets (for instance, 2002–07), to cap-ture greater changes in poverty status. The 2002 and 2007 databases do not match, however, because Susenas modules collect information from different categories every three years. We also found many inconsistencies in the 2006 data, so we could not include them. Analysing poverty dynamics using panel data covering a short period (three years) may not reveal all of the long-run changes in poverty. Given the limitations of available data, however, analysing a short period of poverty dynamics using a Susenas data set that provides rich information about household socio-economic conditions and covers all provinces will nevertheless con-tribute to a deeper understanding of the recent situation of poverty. It will also provide useful insights into why some households remain poor and why others escape poverty. 5 In merging the 2005 and 2007 sample identifiers of Susenas core and module data sets, we found 9,491 balanced-panel samples. Around 1,120 samples were lost during the merg-er, which might be due to a split of provinces. We not only merged the sample identifiers but also included household information such as educational attainment, physical assets, shocks and the poverty line.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


In order to provide basic information about household welfare status, our analysis of poverty dynamics starts with a discussion of household expenditure, the poverty line and poverty incidence during 2005–07. In this period, household expenditure in Indonesia increased by an average of 30.4% (table 1). Households outside Java–Bali experienced an increase in expenditure of 39.4%, while house-holds in Java–Bali experienced an increase of only 23.6%. This large increase in household expenditure outside Java–Bali was not followed by a corresponding reduction in poverty in those areas, because the poverty line in those areas rose by 31.9% (largely owing to the rise of rice prices in 2005–06 (Lindblad and Thee 2007)). The national poverty incidence remained unchanged during 2005–07, but the poverty incidence outside Java–Bali decreased by around 0.5 of a percent-age point. Surprisingly, urban poverty also decreased by around 0.5 of a percent-age point, but rural poverty increased by almost 1.0 percentage point. Although both rural and urban households experienced a similar proportion of increase in expenditure, the rural poverty line increased by more than 25% while the urban poverty line increased by only 14%.

Poverty dynamics in Indonesia during 2005–07This paper uses the spell approach (as illustrated in figure 1), the poverty lines of 2005 and 2007, and the poverty measures of the Foster–Greer–Thorbecke (FGT) formula (Foster, Greer and Thorbecke 1984) to identify and measure poverty sta-tus in Indonesia.6 It analyses only the P0 (head-count index) of the FGT poverty measurement. As this paper uses a short period of panel data, it may be inappro-priate to refer to the chronic poor and never poor; both categories require at least five years of longitudinal data to provide a clear definition and analysis. Using expenditure-based poverty measures, we have categorised households into just four groups: poor, transient poor (–), transient poor (+) and non-poor. This adjust-ment does not reduce the significance of this analysis of poverty dynamics in Indonesia. The paper also applies three different poverty lines: the official poverty line, published by BPS; the lower poverty line (75% of the official poverty line); and the upper poverty line (125% of the official poverty line). These three differ-ent poverty lines help examine the sensitivity of poverty incidence to changes in the poverty line. Figure 2 shows Indonesian poverty dynamics during 2005–07 at a national level, using the official poverty line.

Table 2 shows that poverty in Indonesia is predominantly a rural phenome-non and very sensitive to changes in the poverty line: an increase of 25% in the poverty line causes an increase of more than 100% in the poverty rate. At the disaggregated level, 95% of urban poor households in 2005 were able to escape poverty during 2005–07, while 64% of rural poor households were able to do the same. Around 11% of rural non-poor households in 2005 subsequently fell into

6 The FGT class of poverty measures follows:

πα = 1n

z − yiz

⎛⎝⎜

⎞⎠⎟i=1

q∑

α

where π is the poverty index, n is the total population size, z is the poverty line, yi is the income of the ith individual (or household), q represents the number of individuals just below or on the poverty line, and α is a parameter for the FGT class.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


TABLE 1 Summary of Household Expenditure, the Poverty Line and Poverty Incidence (2005–07)

Household Expenditure Calculated Based on Balanced Panels of 2005 and 2007 (Rp/month/capita)

Region 2005 2007 Change (%)

Mean SDa Mean SD

National 288,579 260,391 376,175 330,679 30.4 Urban 401,305 348,171 521,161 409,812 29.9

Rural 208,434 119,911 273,093 205,269 31.0

Java–Bali 312,278 301,724 386,130 337,318 23.6 Outside Java–Bali 261,840 200,639 364,944 322,697 39.4

Official Poverty Line (Rp/month/capita)

Region 2005 2007 Change (%)

National 141,465 167,390 18.3 Urban 165,565 187,942 13.5

Rural 117,365 146,837 25.1

Java–Bali 145,569 169,031 16.1

Urban 170,153 192,974 13.4

Rural 120,985 145,088 19.9

Outside Java–Bali 135,768 179,015 31.9

Urban 156,456 197,909 26.5 Rural 115,080 160,121 39.1

Poverty Incidence, Calculated Based on the Total Sample of Susenas 2005 and 2007 (%)

Region 2005 2007 Change

National 16.6 16.6 0.0 Urban 13.0 12.5 –0.5

Rural 19.4 20.4 1.0

Java–Bali 15.8 16.0 0.2 Outside Java–Bali 18.0 17.5 –0.5

a SD = standard deviation.

Source: Authors’ calculations and several BPS publications.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


FIGURE 2 National Poverty Dynamics during 2005–07a

No. of HH 8,726

P-05 1,061 (12.2%)

NP-05 7,665 (87.8%)

P-07 Poor 292

(3.4%)

NP-07 Transient poor (+)

769 (8.8%)

P-07 Transient poor (–)

509 (5.8%)

NP-07 Non-poor

7,156 (82.0%)

a HH= households; P = poor; NP = non-poor. Figures in parentheses are the percentage of the total sample.

Source: Authors’ calculations.

poverty, compared with only 1% of urban non-poor households. Urban house-holds contributed more transient poor (+) and non-poor while rural households contributed more transient poor (–) and poor. Rural households rely mostly on agricultural activities for income, which are relatively unstable compared with industrial or service sectors in the urban area. Negative shocks such as crop loss, falling agricultural prices, or death and illness can therefore easily send rural households into poverty.

Table 2 also shows poverty dynamics at the disaggregated, regional level of Java–Bali and outside Java–Bali.7 In Indonesia, it is generally observed that there are two types of regional segregation: Java–Bali versus outside Java–Bali, and Western Indonesia versus Eastern Indonesia. Western Indonesia comprises Suma-tra, Java, Bali and Kalimantan, while Eastern Indonesia consists of Sulawesi, Nusa Tenggara, Maluku and Papua. Java and Bali have significantly larger populations and more developed economic activities and infrastructure than the other islands. Manufacturing activities and service sectors dominate the economy of Java–Bali; agricultural and mining activities dominate outside Java–Bali. Suryadarma et al. (2006), using the 2003 Village Potential Survey (Podes) and the 2002 and 2004 Susenas panel data sets, showed that households in Java–Bali had better access to basic services, such as education and health, than households outside Java–Bali.

7 According to BPS, the 2005 and 2007 Susenas panel data sets should be presented at the national, rural and urban levels but not at the provincial level. However, there is still the possibility and validity of analysing at the regional level both Java–Bali and outside Java–Bali, because the 2005 and 2007 Susenas balanced panel had been distributed propor-tionally between Java–Bali (4,626 households) and outside Java–Bali (4,100 households). A regional analysis would follow Suryadarma et al. (2006), who used the 2002 and 2004 Susenas panel data sets to analyse the level of access to basic services at regional level.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


TABLE 2 Overview of Poverty Status in 2005 and 2007a (number of households)

Condition in 2007

Lower Poverty Line Official Poverty Line Upper Poverty Line

Total Poor Non- poor



Condition in 2005 Urban Poor 74 2 72 281 13 268 690 171 519 Non-poor 3,552 2 3,550 3,345 32 3,313 2,936 220 2,716 Rural Poor 209 35 174 780 279 501 1,627 832 795 Non-poor 4,891 153 4,738 4,320 477 3,843 3,473 783 2,690

Java–Bali Poor 108 16 92 475 143 332 1,088 472 616 Non-poor 4,518 16 4,502 4,151 243 3,908 3,538 513 3,025 Outside Java–Bali Poor 175 21 154 586 149 437 1,229 531 698 Non-poor 3,925 139 3,786 3,514 266 3,248 2,871 490 2,381

National Poor 283 37 246 1,061 292 769 2,317 1,003 1,314 Non-poor 8,443 155 8,288 7,665 509 7,156 6,409 1,003 5,406

Total 8,726 192 8,534 8,726 801 7,925 8,726 2,006 6,720

a The official poverty line is that published by BPS. The lower poverty line is 75% of the BPS line, and the upper poverty line is 125% of the BPS line.

Source: Authors’ calculations based on Susenas data.

Almost 20% of villages outside Java–Bali had no primary school, compared with less than 1% of villages in Java–Bali.

The regional segregation between Java–Bali and outside Java–Bali might influ-ence household poverty characteristics, owing to differences in economic struc-ture and infrastructure availability. At the disaggregated, regional level, we found that 70% of poor households in Java–Bali in 2005 were able to escape poverty during 2005–07 and that 75% of poor households outside Java–Bali were able to do the same (table 2). Around 6% of non-poor households in Java–Bali in 2005 fell into poverty, compared with 8% of those outside Java–Bali. Further, around 30% of poor households in Java–Bali and around 25% of poor households outside Java–Bali remained poor in two periods of observation.

Non-poor households outside Java–Bali seemed more vulnerable to falling into poverty than those in Java–Bali, while poor households outside Java–Bali tended to move more easily out of poverty than those in Java–Bali. The economic dependence of areas outside Java–Bali on agricultural and mining commodities may explain this vulnerability: fluctuations in the prices of these commodities often lead to fluctuations in household income and expenditure.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


RESEARCH METHODOLOGY

Model specification As discussed above, the spell approach categorises households in Indonesia as poor, transient poor (–), transient poor (+) or non-poor. We contend that house-hold poverty status has an order in which one status might be more favourable than others. Non-poor is the most preferred condition; poor the least preferred. The order of transient poor (–) and transient poor (+) is in between poor and non-poor. This paper assumes that the improvement condition of transient poor (+) is more favourable than the degradation condition of transient poor (–).

We used an ordered logit model to examine the determinants that can change household poverty status and enable the poor to escape from poverty. Such a model is useful for understanding the relative effect of different household char-acteristics on poverty status, but it is less useful for distinguishing between pov-erty categories. Independent variables (predictors) in this model are essentially divided into two groups: the 2005 initial variables and the 2005–07 change vari-ables.

The initial variables represent household conditions and positions that may change household poverty status in the future. For instance, poor agricultural households with a small area of land in the initial year might become continu-ously poor later: the land might not produce above a subsistence level, and the household might not have enough resources to invest in modern agricultural technology or to buy good seed for the next production. Uninsured households that experience health shocks in the initial year might become poor in the future – their members might be unable to work, or they might have to allocate all of their resources to medical treatments. Households forced to sell land for medical treat-ments might later become impoverished; and non-poor households in the initial year might become poor households in the next, because of changes in variables such as marital or job status.

Independent variables included in the model consider the data availability in the 2005 and 2007 Susenas, as well as variables used in research by Jalan and Ravallion (1998), Herrera (1999), Okidi and Kempaka (2002), Alisjahbana and Yusuf (2003), Bigsten et al. (2003), Fields et al. (2003), Haddad and Ahmed (2003), McCulloch and Calandrino (2003), McKay and Lawson (2003), Contreras et al. (2004), Kedir and McKay (2005), Woolard and Klasen (2005), and Widyanti et al. (2009). The ordered logit model (equation (1)) is as follows:

yi = HHCi0β + SECOi

0x + ShockGovi0ϕ + ΔVARi

05−07φ + ei

(1)

where

• yi = a household’s poverty status: 0 = poor, 1 = transient poor (–), 2 = transient poor (+), 3 = non-poor;

• HHCi0 = a vector of family characteristics in 2005, including marital status, age,

educational attainment, number of household members, dummy variables for location and island;

• SECOi0 = a vector of socio-economic characteristics in 2005, including dummy

variables for employment sector and employment status, land ownership (in hectares), size of house (in square metres), access to electricity for lighting, and

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


a dummy variable for households with a family member working as a migrant overseas;

• ShockGovi0 = a vector of shocks, risks and policy variables received by a

household in 2005;8 • ΔVARi

05−07 = a vector of changes in variables during 2005–07, including change in marital status, number of household members, employment sector, employment status, access to electricity for lighting, and microcredit;

• e is an error term; and• i is the household identifier (i = 1,…, 8,726).

Appendix table 1 describes the variables and their expected signs.

Ordered response modelEquation (1) is an ordered response model (probit or logit) with four outcomes, y = 0,1,2,3{ }. An ordered logit model for y (conditional on explanatory variables x) can be derived from a latent variable model. Assume that a latent variable, y∗, is determined by:

y∗ = xβ + e, e x Normal 0,1( ) (2)

where β is a Kx1 coefficient vector and where vector x does not contain a constant (for a detailed explanation of the ordered response model, see Wooldridge 2010). The parameters of the model can be estimated by using maximum likelihood estimation. The signs of the estimated coefficients from the ordered probit (logit) models have the exact meaning with the result of ordinary least square (OLS) estimations. A negative sign determines whether the choice probabilities shift to lower categories when the independent variable increases. The partial effect of estimated coefficients, however, cannot be interpreted directly as the result of OLS estimation. In most cases, we are interested in the response probabilities or partial effects, P y = j x( ), of the ordered probit model (see Wooldridge 2010).

The ordered logit model (equation (1)) uses three sample groups: Java–Bali, outside Java–Bali and national (entire sample). Although our analysis of poverty dynamics focuses on the national group, dividing the sample helps show the con-sistency and robustness of estimation results. It also checks whether there are sig-nificant differences in poverty characteristics between Java–Bali and the rest of the country.

DESCRIPTIVE DATA ANALYSISHouseholds in Indonesia can be divided into four groups, based on their pov-erty experience in 2005–07 (table 3): poor (292 households), transient poor (–) (509 households), transient poor (+) (769 households) and non-poor (7,156 households).

8 Negative shocks and risks include economic risks and health shocks. Positive shocks in-clude improvements to public facilities surrounding living areas, more new jobs, and micro-credit. Economic risks include crop loss, job loss, falling crop prices and increased production costs. This vector also includes interaction variables between savings and socio-economic shocks, and the policy variables of Raskin and Askeskin. These variables are intended to ex-amine the effectiveness of saving and government policies in coping with negative shocks.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


We observed that members of the poor group are largely uneducated or have a low level of educational attainment; live in a rural area, rely on income from agri-cultural production and the informal sector; and either own a small area of land or are landless. Unlike the other groups, the poor group is excluded from access-ing modern utilities (around 39% of households in the poor group do not have electricity) and financial services (none of the households in the poor group have received microcredit from either the government or other sources).

The demographic characteristics and socio-economic variables of the tran-sient poor (–) group were slightly better than those of the poor group – including higher educational attainment, better access to electricity, and ownership of larger land areas – and fewer transient poor (–) households experienced economic risks and health shocks. We found that the major variable change faced by the transient poor (–) group during 2005–07 was an increase of one household member moving from formal to informal employment (14%).

Compared with the transient poor (–) group, the transient poor (+) group has a higher level of educational attainment; lives in an urban area; has better access to electricity in 2007 than it did in 2005; has a low percentage of members work-ing in the agricultural sector; and has a low percentage of households experienc-ing economic and health shocks, as well as sufficient savings to cope with such shocks. The greatest differences in the change variables of the transient poor (+) group and the poor and transient poor (–) groups is a decrease in household size by almost one member, a higher proportion of households receiving microcredit, a higher proportion of households gaining access to electricity, and a lower pro-portion of households moving from formal to informal employment.

The non-poor group is generally more educated than other groups; has fewer household members; lives in urban areas; has a larger proportion of households connected to electricity; has experienced fewer economic and health shocks; and has sufficient savings to cope with such shocks. The daily activities of non-poor households are disrupted by health shocks only 3.7 days per month, around half the time experienced by poor households. Furthermore, members of households in the non-poor group tend to work in the formal and non-agricultural sectors, so their income is less volatile and less likely to depend on government assistance.

Table 4 shows that significant differences exist between households in Java–Bali and those outside Java–Bali. The latter, for example, have more family mem-bers, live mostly in rural areas and have larger areas of agricultural land. Fewer households outside Java–Bali have electricity for lighting. Furthermore, house-holds outside Java–Bali experienced more economic risks and health shocks than households in Java–Bali. The daily activities of households outside Java–Bali are disturbed by health shocks half a day more than those of households in Java–Bali. Households outside Java–Bali are more vulnerable to being transient poor, both (–) and (+), than households in Java–Bali.

THE DETERMINANTS OF POVERTY DYNAMICS IN INDONESIAThis paper uses three models – Java–Bali, outside Java–Bali and national – based on household location, which were estimated using maximum likelihood estima-tion, with robust standard errors. Tables 5 and 6 show the estimation results of the ordered logit model. The signs of coefficients in the three models are almost

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


TABLE 3 Descriptive Data on Poverty Statusa

Poor Transient Poor (–)

Transient Poor (+)

Non-poor

Mean SD Mean SD Mean SD Mean SD

Demographic variables in 2005Marital status of HH head (1 = married; 0 = other) 0.88 0.33 0.85 0.35 0.87 0.34 0.85 0.36 Age of HH head (years) 47.43 14.28 46.17 14.90 47.43 14.23 45.53 13.71 Educational attainment of HH head (years) 4.74 3.15 5.10 3.37 5.65 3.19 6.91 4.38 Number of HH members 4.72 1.79 4.06 1.74 4.88 1.77 3.85 1.60

Island dummy (1 = Java–Bali; 0 = other) 0.49 0.50 0.48 0.50 0.43 0.50 0.55 0.50 Location dummy (1 = urban; 0 = rural) 0.04 0.21 0.06 0.24 0.35 0.48 0.46 0.50

Socio-economic variables in 2005Employment sector of HH head (1 = agriculture; 0 = other)

0.80 0.40 0.72 0.45 0.64 0.48 0.45 0.50

Employment status of HH head (1 = formal; 0 = other)

0.16 0.36 0.18 0.38 0.17 0.38 0.30 0.46

Land ownership (ha) 0.64 0.79 0.86 1.19 0.74 1.26 0.52 1.59

House size (m2) 59.77 50.19 58.17 27.92 56.67 55.95 70.32 65.37

HH with a migrant-worker (TKI) member (1 = has TKI member; 0 = other)

0.04 0.19 0.04 0.20 0.04 0.19 0.04 0.21

Access to electricity for lighting (1= no; 0 = yes)

0.39 0.49 0.27 0.44 0.27 0.44 0.10 0.30

Shocks/risks and policy variables in 2005ECSHRS (1= experience; 0 = no experience) 0.28 0.45 0.26 0.44 0.23 0.42 0.16 0.37 Used Raskin for ECSHRS (1 = yes; 0 = no) 0.02 0.14 0.02 0.12 0.03 0.16 0.01 0.08 Daily activities disrupted by health problems for all family members (days/month)

6.36 11.20 4.45 8.61 4.85 8.70 3.73 7.80

Health insurance (1 = Askeskin; 0 = other) 0.04 0.19 0.03 0.16 0.02 0.15 0.01 0.10 Saving as a coping strategy for ECSHRS (1 = having savings; 0 = no savings)

0.01 0.08 0.01 0.08 0.02 0.14 0.03 0.16

Microcredit (1 = yes; 0 = no) 0.00 0.00 0.03 0.16 0.02 0.12 0.03 0.18 Source of microcredit (1 = government; 0 = other)

0.00 0.00 0.01 0.09 0.01 0.07 0.01 0.10

Family member gaining employment (1 = yes; 0 = other)

0.06 0.24 0.05 0.21 0.10 0.30 0.08 0.27

Improved public facilities nearby (1 = yes; 0 = no)

0.13 0.34 0.09 0.29 0.08 0.27 0.10 0.29

Change variables during 2005–07Change in number of HH members –0.07 1.27 0.64 1.50 –0.59 1.67 0.07 1.53 Change in marital status of HH head (1 = divorced; 0 = other)

0.05 0.23 0.05 0.21 0.06 0.24 0.06 0.23

Change in employment sector of HH head (1 = agricultural to non-agricultural; 0 = other)

0.11 0.32 0.11 0.31 0.13 0.34 0.14 0.35

Change in employment status of HH head (1 = formal to informal; 0 = other)

0.11 0.32 0.14 0.34 0.08 0.27 0.12 0.32

Change in access to electricity for lighting (1 = access in 2007 but not in 2005; 0 = other)

0.11 0.31 0.08 0.27 0.13 0.34 0.04 0.21

Change in access to credit (1 = credit in 2007 but not in 2005; 0 = other)

0.03 0.16 0.04 0.19 0.05 0.22 0.07 0.26

Number of observations 292 509 769 7,156

a SD = standard deviation; HH = household; TKI = Tenaga Kerja Indonesia (Indonesian migrant worker); ECSHRS = economic shocks and risks; Raskin = rice for the poor; Askeskin = health insurance for the poor.


Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


TABLE 4 Descriptive Data Used in the Ordered Logit Modela

Java–Bali Outside Java–Bali

National

Mean SD Mean SD Mean SD

Demographic variables in 2005Marital status of HH head (1 = married; 0 = other) 0.85 0.36 0.85 0.35 0.85 0.36 Age of HH head (years) 46.73 14.03 44.75 13.59 45.80 13.86 Educational attainment of HH head (years) 6.51 4.27 6.74 4.22 6.62 4.24 Number of HH members 3.79 1.54 4.21 1.76 3.98 1.66 Island dummy (1 = Java–Bali; 0 = other) 0.53 0.50 Location dummy (1 = urban; 0 = rural) 0.51 0.50 0.31 0.46 0.42 0.49

Socio-economic variables in 2005Employment sector of HH head (1 = agriculture; 0 = other) 0.41 0.49 0.58 0.49 0.49 0.50 Employment status of HH head (1 = formal; 0 = other) 0.30 0.46 0.26 0.44 0.28 0.45 Land ownership (ha) 0.23 1.09 0.94 1.83 0.56 1.53 House size (m2) 73.38 62.55 62.04 62.37 68.05 62.72 HH with a migrant-worker (TKI) member (1 = has TKI member; 0 = other)

0.04 0.20 0.05 0.21 0.04 0.20

Access to electricity for lighting (1= no; 0 = yes) 0.03 0.16 0.26 0.44 0.13 0.34

Shocks/risks and policy variables in 2005ECSHRS (1= experience; 0 = no experience) 0.16 0.37 0.19 0.39 0.17 0.38 Used Raskin for ECSHRS (1 = yes; 0 = no) 0.01 0.08 0.01 0.12 0.01 0.10 Daily activities disrupted by health problems for all family members (days/month)

3.74 7.67 4.21 8.53 3.96 8.09

Health insurance (1 = Askeskin; 0 = other) 0.01 0.10 0.02 0.12 0.01 0.11 Saving as a coping strategy for ECSHRS (1 = having savings; 0 = no savings)

0.03 0.16 0.02 0.14 0.02 0.15

Microcredit (1 = yes; 0 = no) 0.05 0.21 0.01 0.10 0.03 0.17 Source of microcredit (1 = government; 0 = other) 0.02 0.13 0.00 0.04 0.01 0.10 Family member gaining employment (1 = yes; 0 = other) 0.08 0.27 0.08 0.26 0.08 0.27 Improved public facilities nearby (1 = yes; 0 = no) 0.10 0.30 0.09 0.28 0.10 0.29

Change variables during 2005–07Change in number of HH members 0.07 1.42 0.01 1.69 0.04 1.55 Change in marital status of HH head (1 = divorced; 0 = other) 0.05 0.22 0.06 0.24 0.06 0.23 Change in employment sector of HH head(1 = agricultural to non-agricultural; 0 = other)

0.14 0.34 0.14 0.34 0.14 0.34

Change in employment status of HH head(1 = formal to informal; 0 = other)

0.12 0.32 0.12 0.32 0.12 0.32


0.02 0.13 0.10 0.30 0.06 0.23


0.08 0.27 0.05 0.22 0.07 0.25

Poverty statusPoor 143 149 292Transient poor (–) 243 266 509Transient poor (+) 332 437 769Non-poor 3,908 3,248 7,156

Number of observations 4,626 4,100 8,726

a SD = standard deviation; HH = household; TKI = Tenaga Kerja Indonesia (Indonesian migrant worker); ECSHRS = economic shocks and risks; Raskin = rice for the poor; Askeskin = health insurance for the poor.


Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


the same, except in the following variables: age of household head (outside Java–Bali), economic shocks and risks (outside Java–Bali), source of microcredit (out-side Java–Bali) and change in marital status (Java–Bali). All three models show that the Wald chi-square statistics of the ordered logit model are statistically sig-nificant, indicating that at least one of the covariates or independent variables affects household poverty status. Generally, the ordered logit models of poverty dynamics are consistent and robust.

Table 6 shows the partial effects (dy/dx) of changes in the probability of house-holds categorised as poor, transient poor (–), transient poor (+) and non-poor, responding to change in independent variables (predictors). The partial effects (the predicted probability of household poverty status) were evaluated at means of independent variables y = j x( ).Demographic variablesAll three models in Table 5 confirm that educational attainment, location and the number of household members are the most important demographic deter-minants of household poverty status. The variables of marital status and age of the household head are both statistically significant to poverty status in model 3 (national level); but marital status is not significant in model 1 (Java–Bali), nor is age in model 2 (outside Java–Bali). Married households outside Java–Bali have a higher probability of being non-poor; most of the households outside Java–Bali are working in the labour-intensive agricultural sectors, so a married household has more workers than a single household and therefore has the potential to pro-duce greater output and generate greater income.

Table 6 shows that an increase in the number of household members decreases the probability of being non-poor, while increasing the probability of being poor, transient poor (–) and transient poor (+). This finding is similar to those of Her-rera (1999), Haddad and Ahmed (2003), and Woolard and Klasen (2005). Given a fixed income, an increase in the number of members forces households to reduce per-person consumption to support the additional members. A better education increases the probability of being non-poor, because a higher level of education provides greater opportunities for a better job and, subsequently, a higher income. These findings confirmed the conclusions of other studies, such as Bigsten et al. (2003), and Widyanti et al. (2009).

The location dummy variable reveals that those living in urban areas have a higher probability of being non-poor. This finding confirms findings in studies of other countries, such those of Fields et al. (2003), and Kedir and McKay (2005). Urban areas, where most industries and economic activities are located, provide more job opportunities in the formal and the informal sectors.

Socio-economic variablesAs many studies have found (Dercon and Krishnan 2000; Okidi and Kempaka 2002), households with a head working in the agricultural sector have a high probability of being poor, owing to low productivity and wage rates. This prob-ability increases by 1.3% in Java–Bali, 1.1% outside Java–Bali and 1.4% nationally, as table 6 shows. Furthermore, households with a head working in the formal sec-tor – that is, working for an agency, office or company for a fixed salary, either in cash or in goods – have a higher probability of being non-poor. Those working in

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


formal sectors increase their probability of being non-poor by 4.6% in Java–Bali, 6.8% outside Java–Bali and 5.8% nationally. The formal sector in Indonesia guar-antees a stable income and pays higher wage rates than the informal sector, which complements the finding of Kedir and McKay (2005) that waged employees in rural Ethiopia have a higher probability of escaping from poverty.

Owing to the lack of job opportunities in Indonesia, individuals who cannot find jobs in the formal sector or start a business (as an entrepreneur) are forced to work either in the domestic, informal sector, for a low wage, or outside Indonesia, as migrant workers overseas. Most migrant workers also work in the informal sector, as domestic helpers, but they are paid a higher wage. This paper con-firms that households with a family member working outside Indonesia tend to be non-poor. Their remittances may take the form of family transfers to support basic needs, or entrepreneur capital transfers to support their families starting a business. Hall (2007) showed that remittances play an important role in poverty dynamics in Latin America. Notably, the coefficient for this variable is insignifi-cant for the outside Java–Bali sample.

Land ownership as an indicator of physical assets significantly affects house-hold poverty status. Table 6 shows that a one-hectare increase in land size would increase the probability of being non-poor by 1.7% in Java–Bali, 1.3% outside Java–Bali and 1.7% nationally. Landless and small-landholder households tend to be chronically poor, because their productive assets are inadequate for increas-ing their income. Land reforms aimed at increasing access to land as a productive asset by poor households could alleviate chronic poverty. This finding is similar to those of Haddad and Ahmed (2003), and Woolard and Klasen (2005). House size as an indicator of physical assets can also determine a household’s poverty status: a larger house is associated with a lower probability of being non-poor. Both findings may possibly also imply that the certification of agricultural land and house ownership may help alleviate poverty: certification would legalise land and house ownership, which could then be used as collateral for gaining productive credit from formal institutions.

Other socio-economic variables, such as access to modern electricity utilities, significantly increase the probability of escaping poverty. The unit cost of light-ing with electricity is cheaper per kilowatt-hour than lighting with candles or an oil lamp (Foster and Tre 2003). Households could possibly reduce their energy expenditure and potentially re-allocate savings to income-generating activities or, if children are part of the household, to education. This could ultimately help lift households out of poverty. Table 4 shows that households in Java–Bali have better access to electricity than households outside Java–Bali, owing more to the avail-ability of the electricity grid than to a household’s ability to pay the connection fee (LPEM FEUI, PSE-KP UGM and PSP-IPB 2004a). Increasing household access to electricity could be considered as a poverty alleviation policy, especially for households outside Java–Bali.

Shocks, risks and government assistanceHouseholds in Java-Bali are more vulnerable to negative shocks than those out-side Java-Bali; the latter are relatively resilient. Even so, households outside Java–Bali experienced more negative shocks than households in Java–Bali (table 4), although our estimation results showed that the coefficients of economic risks

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


TABLE 5 Estimation Results of Ordered Logit Modela

Java–Bali Outside Java–Bali

National

Coeff. Robust Std Error



Demographic variables in 2005Marital status of HH head (1 = married; 0 = other) 0.198 0.145 0.295 0.134** 0.239 0.097***Age of HH head (years) –0.007 0.004* 0.004 0.004 –0.002 0.003***Educational attainment of HH head (years) 0.079 0.012*** 0.052 0.011*** 0.068 0.008***Number of HH members –0.431 0.032*** –0.421 0.028*** –0.402 0.021***Island dummy (1 = Java–Bali; 0 = other) –0.410 0.073***Location dummy (1 = urban; 0 = rural) 1.283 0.105*** 0.291 0.115** 0.868 0.079***


–0.822 0.109*** –0.540 0.113*** –0.720 0.077***


0.544 0.161*** 0.544 0.161*** 0.544 0.113***

Land ownership (ha) 0.182 0.091** 0.095 0.033*** 0.149 0.032***House size (m2) 0.006 0.002*** 0.007 0.003** 0.006 0.002***HH with a migrant-worker (TKI) member (1 = has TKI member; 0 = other)

0.716 0.247*** 0.097 0.219 0.337 0.159**

Access to electricity for lighting (1= no; 0 = yes) –1.984 0.290*** –1.033 0.124*** –0.916 0.108***

Shocks/risks and policy variables in 2005ECSHRS (1= experience; 0 = no experience) –0.377 0.111*** 0.005 0.114 –0.173 0.079**Used Raskin for ECSHRS (1 = yes; 0 = no) –0.241 0.378 –0.204 0.282 –0.107 0.229 Daily activities disrupted by health problems for all family members (days/month)

–0.010 0.005** –0.007 0.005 –0.007 0.004*

Health insurance (1 = Askeskin; 0 = other) –1.164 0.280*** –0.337 0.307 –0.646 0.212***

Saving as a coping strategy for ECSHRS (1 = having savings; 0 = no savings)

0.558 0.309* 0.653 0.368* 0.596 0.243***

Microcredit (1 = yes; 0 = no) 0.920 0.382** 0.118 0.400 0.639 0.278**Source of microcredit (1 = government; 0 = other) –0.254 0.608 0.475 1.049 0.085 0.492 Family member gaining employment (1 = yes; 0 = other)

0.364 0.173** 0.062 0.156 0.219 0.115*

Improved public facilities nearby (1 = yes; 0 = no) –0.318 0.136** 0.601 0.178*** 0.092 0.108

Change variables during 2005–07Change in number of HH members –0.152 0.031*** –0.184 0.026*** –0.160 0.020***Change in marital status of HH head (1 = divorced; 0 = other)

0.048 0.218 –0.342 0.176** –0.190 0.135


0.528 0.148*** 0.240 0.129* 0.393 0.096***


–0.265 0.213 –0.675 0.194*** –0.500 0.141***


1.318 0.356*** 0.151 0.137 0.150 0.128


0.431 0.179** 0.826 0.237*** 0.531 0.138***

/cut0 –4.510 0.289*** –4.614 0.275*** –4.631 0.200***/cut1 –3.327 0.288*** –3.430 0.270*** –3.465 0.197***/cut2 –2.496 0.282*** –2.460 0.265*** –2.576 0.193***

Number of observations 4,626 4,100 8,726Log pseudolikelihood –2,345.27 –2,629.68 –5,055.63 Wald Chi-squared 708.78 561.21 1,102.26 Pseudo R-squared 0.1462 0.1105 0.1170

a ***, ** and * indicate significance at the 1%, 5% and 10% levels respectively; HH = household; TKI = Tenaga Kerja Indonesia (Indonesian migrant worker) ECSHRS = economic shocks and risks; Raskin = rice for the poor; Askeskin = health insurance for the poor. Table 4 reveals the average change in the number of household members.


Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


TABL

E 6

Est

imat

ions

of P

arti

al E

ffect

(dy/

dx) (

Pro

babi

lity

in %

)a

Java

–Bal

iO

utsi

de Ja

va–B

ali

Nat

iona

l

Poor

TP–

TP+

NP

Poor

TP–

TP+

NP

Poor

TP–

TP+

NP

Dem

ogra

ph

ic v

aria

ble

s in

200

5M

arita

l sta

tus

of H

H h

ead

(1 =

mar

ried

; 0 =

oth

er)

–0.3

1 –0

.63

–0.9

6 1.

90

–0.7

1 –1

.37

–2.2

5 4.

33

–0.4

9 –0

.95

–1.5

0 2.

94

Age

of H

H h

ead

(yea

rs)

0.01

0.

02

0.03

–0

.06

–0.0

1 –0

.02

–0.0

3 0.

05

0.00

0.

01

0.01

–0

.02

Educ

atio

nal a

ttai

nmen

t of H

H h

ead

(yea

rs)

–0.1

2 –0

.24

–0.3

7 0.

72

–0.1

1 –0

.22

–0.3

7 0.

70

–0.1

3 –0

.25

–0.4

0 0.

78

Num

ber

of H

H m

embe

rs0.

63

1.29

2.

00

–3.9

2 0.

91

1.79

3.

05

–5.7

5 0.

76

1.49

2.

39

–4.6

4 Is

land

dum

my

(1 =

Java

–Bal

i; 0

= ot

her)

0.77

1.

51

2.42

–4

.70

Loca

tion

dum

my

(1 =

urb

an; 0

= r

ural

)–2

.00

–4.0

3 –6

.04

12.0

7 –0

.60

–1.1

9 –2

.05

3.84

–1

.57

–3.0

7 –4

.94

9.58

S

ocio

–eco

nom

ic v

aria

ble

s in

200

5Em

ploy

men

t sec

tor

of H

H h

ead

(1 =

agr

icul

ture

; 0 =

oth

er)

1.32

2.

67

4.02

–8

.01

1.14

2.

23

3.82

–7

.19

1.40

2.

71

4.30

–8

.40

Empl

oym

ent s

tatu

s of

HH

hea

d (1

= fo

rmal

; 0 =

oth

er)

–0.7

2 –1

.49

–2.3

4 4.

55

–1.0

5 –2

.09

–3.6

7 6.

82

–0.9

3 –1

.83

–3.0

1 5.

77

Land

ow

ners

hip

(ha)

–0.2

6 –0

.54

–0.8

4 1.

65

–0.2

1 –0

.41

–0.6

9 1.

30

–0.2

8 –0

.55

–0.8

9 1.

72

Hou

se s

ize

(m2 )

–0.0

1 –0

.02

–0.0

3 0.

06

–0.0

1 –0

.03

–0.0

5 0.

09

–0.0

1 –0

.02

–0.0

4 0.

07

HH

with

a m

igra

nt-w

orke

r (T

KI)

mem

ber

(1 =

has

TK

I mem

ber;

0 =

oth

er)

–0.7

7 –1

.63

–2.6

3 5.

03

–0.2

0 –0

.40

–0.6

9 1.

29

–0.5

5 –1

.10

–1.8

3 3.

48

Acc

ess

to e

lect

rici

ty fo

r lig

htin

g (1

= no

; 0 =

yes

)7.

98

12.8

4 13

.23

–34.

05

2.95

5.

44

8.21

–16

.60

2.46

4.

50

6.42

–13

.37

Sh

ock

s/ri

sks

and

pol

icy

vari

able

s in

200

5EC

SHR

S (1

= ex

peri

ence

; 0 =

no

expe

rien

ce)

0.63

1.

27

1.91

–3

.80

–0.0

1 –0

.02

–0.0

4 0.

07

0.35

0.

67

1.07

–2

.09

Use

d R

aski

n fo

r EC

SHR

S (1

= y

es; 0

= n

o)

0.39

0.

80

1.21

–2

.41

0.49

0.

94

1.56

–2

.99

0.21

0.

41

0.66

–1

.28

Dai

ly a

ctiv

ities

dis

rupt

ed b

y he

alth

pro

blem

s fo

r al

l fam

ily m

embe

rs (d

ays/

mon

th)

0.01

0.

03

0.05

–0

.09

0.01

0.

03

0.05

–0

.09

0.01

0.

03

0.04

–0

.08

Hea

lth in

sura

nce

(1 =

Ask

eski

n; 0

= o

ther

)3.

07

5.73

7.

44 –

16.2

3 0.

86

1.64

2.

64

–5.1

3 1.

68

3.11

4.

52

–9.3

0 Sa

ving

as

a co

ping

str

ateg

y fo

r EC

SHR

S (1

= h

avin

g sa

ving

s; 0

= n

o sa

ving

s)

–0.6

4 –1

.33

–2.1

5 4.

12

–1.0

6 –2

.16

–3.9

6 7.

18

–0.8

7 –1

.74

–2.9

6 5.

57

Mic

rocr

edit

(1 =

yes

; 0 =

no)

–0.9

2 –1

.94

–3.1

7 6.

03

–0.2

4 –0

.48

–0.8

3 1.

56

–0.9

2 –1

.84

–3.1

4 5.

90

Sour

ce o

f mic

rocr

edit

(1 =

gov

ernm

ent;

0 =

othe

r)0.

42

0.85

1.

28

–2.5

5 –0

.83

–1.6

7 –3

.02

5.52

–0

.15

–0.3

0 –0

.49

0.95

Fa

mily

mem

ber

gain

ing

empl

oym

ent (

1 =

yes;

0 =

oth

er)

–0.4

6 –0

.96

–1.5

1 2.

93

–0.1

3 –0

.26

–0.4

4 0.

83

–0.3

8 –0

.75

–1.2

3 2.

36

Impr

oved

pub

lic fa

cilit

ies

near

by (1

= y

es; 0

= n

o)0.

52

1.06

1.

60

–3.1

9 –1

.04

–2.1

0 –3

.80

6.93

–0

.17

–0.3

3 –0

.53

1.03

C

han

ge v

aria

ble

s d

uri

ng

2005

–07

Cha

nge

in n

umbe

r of

HH

mem

bers

0.22

0.

46

0.70

–1

.38

0.40

0.

78

1.34

–2

.52

0.30

0.

59

0.95

–1

.85

Cha

nge

in m

arita

l sta

tus

of H

H h

ead

(1 =

div

orce

d; 0

= o

ther

)–0

.07

–0.1

4 –0

.22

0.43

0.

86

1.64

2.

66

–5.1

6 0.

39

0.75

1.

19

–2.3

3 C

hang

e in

em

ploy

men

t sec

tor

of H

H h

ead

(1 =

agr

icul

tura

l to

non–

agri

cultu

ral;

0 =

othe

r)–0

.65

–1.3

5 –2

.14

4.13

–0

.48

–0.9

5 –1

.66

3.10

–0

.65

–1.2

9 –2

.14

4.08

Cha

nge

in e

mpl

oym

ent s

tatu

s of

HH

hea

d (1

= fo

rmal

to in

form

al; 0

= o

ther

)0.

43

0.87

1.

31

–2.6

1 1.

90

3.55

5.

44 –

10.9

0 1.

15

2.18

3.

31

–6.6

3 C

hang

e in

acc

ess

to e

lect

rici

ty fo

r lig

htin

g (1

= a

cces

s in

200

7 bu

t not

in 2

005;

0 =

oth

er)

–1.1

0 –2

.34

–3.8

8 7.

32

–0.3

1 –0

.61

–1.0

6 1.

98

–0.2

7 –0

.53

–0.8

6 1.

65

Cha

nge

in a

cces

s to

cre

dit (

1 =

cred

it in

200

7 bu

t not

in 2

005;

0 =

oth

er)

–0.5

3 –1

.11

–1.7

6 3.

40

–1.2

8 –2

.61

–4.8

4 8.

74

–0.8

1 –1

.62

–2.7

3 5.

17

Pro

bab

ilit

y y=jx

()

1.50

3.

20

5.40

89

.90

2.20

4.

70

9.50

83

.60

1.90

4.

00

7.40

86

.70

a dy

/dx

= di

scre

te c

hang

e of

dum

my

vari

able

fro

m 0

to

1; T

P– =

tra

nsie

nt p

oor

(–);

TP+

= tr

ansi

ent

poor

(+)

; NP

= no

n-po

or; H

H =

hou

seho

ld; T

KI

= Te

naga

Ker

ja I

ndon

esia

(I

ndon

esia

n m

igra

nt w

orke

r) E

CSH

RS

= ec

onom

ic s

hock

s an

d ri

sks;

Ras

kin

= ri

ce fo

r th

e po

or; A

skes

kin

= he

alth

insu

ranc

e fo

r th

e po

or. T

able

4 r

evea

ls th

e av

erag

e ch

ange

in

the

num

ber

of h

ouse

hold

mem

bers

.So

urce

: Aut

hors

’ cal

cula

tions

bas

ed o

n Su

sena

s da

ta.D

ownl

oade

d by

[M

acqu

arie

Uni

vers

ity]

at 0

4:38

08

June

201

3


and health shocks affecting household poverty status outside Java–Bali are sta-tistically insignificant. This might be due to these households working largely in agriculture and owning larger areas of land, and therefore being able to reduce agricultural risks such as crop loss and falling prices by diversifying agricultural cultivations.

Households in Java–Bali experiencing economic risks due to crop loss, job loss and falling prices have a tendency to be poor and transient poor (–). Health shocks – represented by a number of daily activities disrupted by health problems – can change poverty status, and those households experiencing such shocks tend to be poor. This finding is consistent with that of Contreras et al. (2004), in Chile. Our three models confirmed that non-poor households experiencing either economic or health shocks but with sufficient savings should maintain their poverty status. Model 3 in Table 6 shows that having savings will decrease the probability of a household being poor and transient poor (–) by 0.9% and 1.7% respectively.

Owing to limited data availability in the Susenas panel data sets, we have referred to only four types of government assistance in this paper: Raskin, Aske-skin, microcredit and improved public facilities. The interaction variable of Raskin and economic shocks and risks does not statistically affect household poverty sta-tus. The probability of households being poor did decrease, however, when the government distributed cheap rice to households experiencing economic shocks and risks in Java–Bali. This paper confirms Sumarto, Suryahadi and Widyanti’s (2005) findings that the subsidised rice program appears to reduce the risk of pov-erty.

An unexpected result is that the effectiveness of both policies (Raskin and Aske-skin) in protecting the poor is not confirmed. This might be due to incorrect target-ing of the programs and/or an uneven distribution of this government assistance, as indicated in tables 3 and 4. At the disaggregated, regional level, the proportions of government assistance (Raskin and Askeskin) received by households expe-riencing health shocks and economic shocks and risks are also relatively small. The government should not only focus on providing assistance based on poverty status but also pay attention to the shocks and events experienced by households.

In contrast, microcredit functions well as a poverty alleviation program – par-ticularly in Java–Bali, where 5% of households have access to it (table 4). The positive coefficient of microcredit in all three models indicates that households receiving credit programs tend to be non-poor. Increasing access to either micro-credit or financial institutions, particularly outside Java–Bali, might reduce pov-erty significantly.

The positive shock of gaining employment can alleviate household poverty and is identical to increased income or expenditure – that is, both can lift a house-hold out of poverty. If a household member can find a job, the probability of their being poor in Java–Bali and in Indonesia as a whole will decrease. This confirms the findings of Fields et al. (2003) that gaining employment would lift a household out of poverty in Indonesia. In addition, the improvement of public facilities, such as the development of bridges and roads, helps alleviate poverty – particularly outside Java–Bali, because these regions often face infrastructure bottlenecks. The probability of households being non-poor outside Java–Bali increases by 6.9% with the development of public facilities in this area (table 6). In contrast, it is surprising in that infrastructure developments in Java–Bali do not reduce poverty

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


status significantly. This is most likely because Java–Bali is a well-developed region that already has reasonably well-developed infrastructure, while new con-structions (such as toll roads) sometimes lead to land acquisitions or evictions. Another example, renovations of traditional markets into modern markets, occa-sionally marginalises previous traders, because of their inability to afford new building prices.

Changes in household indicators during 2005–07This section discusses the effect on poverty status of changes in the demographic, socio-economic and government assistance variables during 2005–07. An increase in household size by one family member decreases the probability of a household being non-poor by 1.9% nationally (table 6). An increase of one family member is associated with falling into poverty, because a given amount of resources needs to be redistributed to support the new member. Large households are less able to save resources or allocate them to other productive activities to help them escape poverty. Changes in marital status due to divorce are also increasing the probabil-ity of households being poor and transient poor (–) in areas outside Java–Bali. A divorce results in the loss of a productive family member, which might reduce a household’s economic power. This is consistent with Woolard and Klasen’s (2005) finding that female-headed households tend to fall into poverty in South Africa.

Changes in employment status from the agricultural to the non-agricultural sector increase the probability of households being non-poor. Non-agricultural sectors are likely to pay higher and more stable wages, so households are able to increase and smooth their consumption levels. Table 6 suggests that house-holds whose members are able to find a job in the non-agricultural sector increase their probability of being non-poor. To alleviate poverty further, the Indonesian government should consider policy options to shift the economic basis from agri-culture to non-agriculture, or consider changing traditional agriculture into agri-culture-based industries.

In contrast, a change in employment status from the formal to the informal sector can send a previously non-poor household into poverty. Households with members being laid off or finding new jobs (as either an employee or a self-employed person) in the informal sector have a higher probability of being either poor or transient poor (–), reducing their probability of being non-poor (table 6).

Model 1 in Table 6 confirms the importance of infrastructure development and microcredit in reducing poverty in Indonesia. Increasing access to electricity, for example, can substantially enhance the productivity of households and house-hold-based microenterprises. Appliances such as pumps, sewing machines and power tools can generate income, while information and communication technol-ogies can make market information more accessible and enable social and politi-cal participation (LPEM FEUI, PSE-KP UGM and PSP-IPB 2004a, 2004b). Table 6 also shows that households receiving microcredit during 2005–07 increased their probability of being non-poor, possibly by starting small businesses and creating jobs. Increasing access to microcredit, particularly outside Java–Bali, where finan-cial institutions are not yet well developed, could help reduce poverty.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


CONCLUDING REMARKSObserving the 2005 and 2007 Susenas panel data sets and using the spell approach to determine household poverty status, we found that 28% of poor households in Indonesia could be considered chronically poor (that is, remaining poor in two periods) and 7% of non-poor households are vulnerable to being transient poor (–). Poverty in Indonesia is a rural phenomenon, sensitive to changes in the poverty line: an increase of 25% in the poverty line causes an increase of more than 100% in the poverty rate. Rural households are more vulnerable than urban households to falling into poverty. Around 11% of rural non-poor households fell into poverty during 2005–07, compared with only 1% of urban non-poor house-holds. At regional level, 30% of poor households in Java–Bali and 25% of poor households outside Java–Bali are categorised as poor (either chronically poor or remaining poor in two periods). Areas outside Java–Bali contributed more tran-sient poor.

Using an ordered logit model, we found that the determinants of poverty dynamics in Indonesia are educational attainment; the number of household mem-bers; physical assets (land and house ownership); employment sector; employ-ment status; access to modern electricity utilities and microcredit; and changes in the number of household members, employment sector and employment status. Our estimation of the partial effects of change in independent variables confirmed that an increase of one family member decreases the probability of a household being non-poor by 1.9% nationally. Households that received microcredit during 2005–07 increased their probability of being non-poor by 3.4% in Java–Bali, 8.7% outside Java–Bali and 5.2% nationally.

We also found that households in Java–Bali are more vulnerable to negative shocks than those outside Java–Bali, which are relatively resilient. No consistent statistical evidence from our three models supports the hypothesis that govern-ment policies such as Raskin and Askeskin are effective in helping households cope with such shocks. Microcredit programs, however, have helped to allevi-ate poverty. Although there is no consistent statistical evidence of these policies changing poverty status, the government should improve household targeting and consider negative shocks and events when distributing assistance.

Our estimation results confirmed that poverty alleviation policies should not be generalised to all regions, because of the differences in poverty characteris-tics between Java–Bali and areas outside Java–Bali. For example, the government could consider providing more safety nets such as Raskin and Askeskin to help households in regions vulnerable to negative shocks. Other policy suggestions targeting rural households include promoting family planning; redistributing land and certifying both land and house ownership; increasing access to elec-tricity, to improve the productivity of households and household-based micro-enterprises; increasing access to microcredit; and providing technical assistance for those starting and running a business.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013


REFERENCESAdam, R.H. and Jane, J. H. (1995) ‘Sources of income inequality and poverty in rural Paki-

stan’, International Food Policy Research Institute Research Report 102, International Food Policy Research Institute, Washington DC.

Alisjahbana, A. and Yusuf, Arief A. (2003) ‘Poverty dynamics in Indonesia: panel data evi-dence’, Working Paper in Economics and Development Studies No. 200303, Padjadja-ran University, Bandung.

Bidani, B. and Ravallion, M. (1993) ‘A new regional poverty profile for Indonesia’, Bulletin of Indonesian Economic Studies 29 (3): 37–68.

Bigsten, A., Kebede, B., Shimeles, A. and Taddesse, M. (2003) ‘Growth and poverty reduc-tion in Ethiopia: evidence from household panel surveys’, World Development 31 (1): 87–106.

Contreras, D., Cooper, R., Herman, J. and Nielson, C. (2004) ‘Dinámica de la pobreza y movilidad social: Chile 1996–2001’, Departamento de Economía, University of Chile, Santiago, available at <http://www.bcn.cl/carpeta_temas_profundidad/temas_pro-fundidad.2007-11-13.0283659577/documentos_pdf.2007-11-13.0655127334/archivos_pdf.2007-11-13.4643314898/archivo1/baja_archivo>.

Dartanto, T. and Nurkholis (2011) ‘Finding out the determinants of poverty dynamics in Indonesia: evidence from panel data’, MPRA Paper No. 41185, available at <http://mpra.ub.uni-muenchen.de/41185/1/MPRA_paper_41185.pdf>.

Dercon, S. and Krishnan, P. (2000) ‘Vulnerability, seasonality, and poverty in Ethiopia’, Jour-nal of Development Studies 36 (6): 25–53.

Dercon, S. and Shapiro, J.S. (2007) ‘Moving on, staying behind, getting lost: lessons on poverty mobility from longitudinal data’, in Moving Out of Poverty: Cross-disciplinary Perspectives on Mobility, eds D. Narayan and P. Petesch, World Bank and Palgrave Mac-millan, New York: 77–126.

Fields, G., Cichello, P., Freije, S., Menendez, M. and Newhouse, D. (2003) ‘Household income dynamics: a four-country story’, Journal of Development Studies 40 (2): 30–54.

Foster, J.E., Greer, J. and Thorbecke, E. (1984) ‘A class of decomposable poverty measures’, Econometrica 52: 761–76.

Foster, V. and Tre, J-P. (2003) ‘Measuring the impact of energy interventions on the poor: an illustration from Guatemala’, in Infrastructure for Poor People: Public Policy for Private Provision, eds P.J. Brook and T.C. Irwin, World Bank, Washington DC: 125–77.

Grab, J. and Grimm, M. (2007) ‘Robust multiperiod poverty comparisons’, Ibero-America Institute for Economic Research Working Paper No. 160, University of Göttingen, Ger-many, available at <http://www2.vwl.wiso.uni-goettingen.de/ibero/papers/DB160.pdf>.

Haddad, L. and Ahmed, A. (2003) ‘Chronic and transitory poverty: evidence from Egypt, 1997–99’, World Development 31 (1): 71–85.

Hall, A.L. (2007) ‘Moving away from poverty: migrant remittances, livelihoods, and devel-opment’, in Moving Out of Poverty: Cross-disciplinary Perspectives on Mobility, eds D. Narayan and P. Petesch, World Bank and Palgrave Macmillan, New York: 307–32.

Herrera, J. (1999), ‘Ajuste económico, desigualdad, y movilidad’, in Pobreza y economía social: análisis de una encuesta ENNIV-1997, eds R. Webb and M. Ventocilla, Institute Cuanto, UNICEF and USAID, Lima: 101–42.

Hulme, D., Moore, K. and Shepherd, A. (2001) ‘Chronic poverty: meanings and analytical frameworks’, Chronic Poverty Research Centre Working Paper No. 2, University of Manchester, Manchester, available at <http://www.chronicpoverty.org/uploads/pub-lication_files/WP02_Hulme_et_al.pdf>.

Hulme, D. and A. Shepard (2003) ‘Conceptualizing chronic poverty’, World Development 31 (3): 403–24.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013

http://www.bcn.cl/carpeta_temas_profundidad/temas_profundidad.2007-11-13.0283659577/documentos_pdf.2007-11-13.0655127334/archivos_pdf.2007-11-13.4643314898/archivo1/baja_archivo



http://mpra.ub.uni-muenchen.de/41185/1/MPRA_paper_41185.pdf

http://mpra.ub.uni-muenchen.de/41185/1/MPRA_paper_41185.pdf

http://www2.vwl.wiso.uni-goettingen.de/ibero/papers/DB160.pdf

http://www2.vwl.wiso.uni-goettingen.de/ibero/papers/DB160.pdf

http://www.chronicpoverty.org/uploads/publication_files/WP02_Hulme_et_al.pdf

http://www.chronicpoverty.org/uploads/publication_files/WP02_Hulme_et_al.pdf


Jalan, J. and Ravallion, M. (1998) ‘Determinants of transient and chronic poverty: evidence from Rural China’, World Bank Policy Research Working Paper No. 1936, Washington DC.

Kedir, A.M. and McKay, A. (2005) ‘Chronic poverty in urban Ethiopia: panel data evidence’, International Planning Studies 10 (1): 49–67.

Lindblad, J.T and Thee, K.W. (2007) ‘Survey of recent developments’, Bulletin of Indonesian Economic Studies 43 (1): 7–33.

LPEM FEUI, PSE-KP UGM and PSP-IPB (2004a) Macroeconomic Policy and Poverty Reduc-tion in Indonesia: A Summary of Mid Term Reports, LPEM-FEUI, University of Indonesia, Jakarta.

LPEM FEUI, PSE-KP UGM and PSP-IPB (2004b) Infrastructure and Poverty Alleviation, LPEM-FEUI, University of Indonesia, Jakarta.

McCulloch, N. and Baulch, B. (2000) ‘Simulating the impact of policy upon chronic and transitory poverty in rural Pakistan’, Journal of Development Studies 36 (6): 100–130.

McCulloch, N. and Calandrino, M. (2003) ‘Vulnerability and chronic poverty in rural Sichuan’, World Development 31 (3): 611–28.

McKay, A. and Lawson, D. (2002) ‘Chronic poverty: a review of current quantitative evi-dence’, CPRC Working Paper No. 15, University of Manchester, available at <http://www.chronicpoverty.org/uploads/publication_files/WP15_Mckay_Lawson.pdf>.

McKay, A. and Lawson, D. (2003) ‘Assessing the extent and nature of chronic poverty in low income countries: issues and evidence’, World Development 31 (3): 425–39.

Muller, G.P. (2002) ‘Explaining poverty: on the structural constraints of income mobility’, Social Indicator Research 59 (3): 301–19.

Okidi, John A. and Kempaka, G. (2002) ‘An overview of chronic poverty and develop-ment policy in Uganda’, CPRC Working Paper No. 11, University of Manchester, Manchester, available at <http://www.dfid.gov.uk/r4d/PDF/Outputs/ChronicPov-erty_RC/11OkidiMugambe.pdf>.

Sawada, Y., Shoji, M., Sugawara, S. and Shinkai, N. (2008) ‘The role of infrastructure in mitigating poverty dynamics: the case of an irrigation project in Sri Lanka’, JBICI Dis-cussion Paper No. 16, JBIC Institute, Tokyo, available at <http://www.jbic.go.jp/en/research/report/discussion-paper/pdf/dp16_e.pdf>.

Sparrow, R., Suryahadi, A. and Widyanti, W. (2012) ‘Social health insurance for the poor: targeting and impact of Indonesia’s Askeskin programme’, in Social Science & Medicine, forthcoming, available at <http://dx.doi.org/10.1016/j.socscimed.2012.09.043>.

Sumarto, S., Suryahadi, A. and Widyanti, W. (2005) ‘Assessing the impact of Indonesian social safety net programmes on household welfare and poverty dynamics’, The Euro-pean Journal of Development Research 17 (1): 155–77.

Suryadarma, D., Widyanti, W., Suryahadi, A. and Sumarto, S. (2006) ‘From access to income: regional and ethnic inequality in Indonesia’, SMERU Working Paper, SMERU Institute, Jakarta, available at <http://www.eaber.or/node/22547>.

Widyanti, W., Suryahadi, A., Sumarto, S. and Yumna, A. (2009) ‘The relationship between chronic poverty and household dynamics: evidence from Indonesia’, SMERU Working Paper, SMERU Institute, Jakarta.

Woolard, I. and Klasen, S. (2005) ‘Determinants of income mobility and household poverty dynamics in South Africa’, Journal of Development Studies 41 (5): 865–97.Wooldridge, Jeffrey M. (2010) Econometric Analysis of Cross Section and Panel Data: Second

Edition, MIT Press, Cambridge, US. Yaqub, S. (2000) Poverty Dynamics in Developing Countries, Institute of Development Studies, Sussex, UK.

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013

http://www.chronicpoverty.org/uploads/publication_files/WP15_Mckay_Lawson.pdf

http://www.chronicpoverty.org/uploads/publication_files/WP15_Mckay_Lawson.pdf

http://www.dfid.gov.uk/r4d/PDF/Outputs/ChronicPoverty_RC/11OkidiMugambe.pdf

http://www.dfid.gov.uk/r4d/PDF/Outputs/ChronicPoverty_RC/11OkidiMugambe.pdf

http://www.jbic.go.jp/en/research/report/discussion-paper/pdf/dp16_e.pdf

http://www.jbic.go.jp/en/research/report/discussion-paper/pdf/dp16_e.pdf

http://dx.doi.org/10.1016/j.socscimed.2012.09.043

http://www.eaber.or/node/22547


APPENDIX TABLE 1 Description of Independent Variables and Expected Signsa

Demographic variables in 2005Marital status of HH head (1 = married; 0 = other) Marital status of household head +Age of HH head (years) Age of household head (in years) +Educational attainment of HH head (years) Years of schooling completed by household head +Number of HH members Number of household members –Island dummy (1 = Java–Bali; 0 = other) Household living either in Java–Bali or in Sumatra,

Kalimantan, Sulawesi, Papua or Nusa Tenggara–

Location dummy (1 = urban; 0 = rural) Whether household is living in an urban or rural area

+


Household head working in agricultural production

–


Household head working for a fixed salary (cash or goods)

+

Land ownership (ha) Land owned by household +House size (m2) Size of house owned by household +HH with a migrant-worker (TKI) member (1 = has TKI member; 0 = other)

Household member an Indonesian migrant worker (Tenaga Kerja Indonesia)

+

Access to electricity for lighting (1= no; 0 = yes)

Household does not use electricity for lighting +

Shocks/risks and policy variables in 2005ECSHRS (1= experience; 0 = no experience) Household experiencing economic shocks or risks –Used Raskin for ECSHRS (1 = yes; 0 = no) Household experiencing economic shocks or risks

and receiving Raskin+

Daily activities disrupted by health problems for all family members (days/month)

Days per month disrupted by health shocks –

Health insurance (1 = Askeskin; 0 = other) Household experiencing health shocks and receiving Askeskin

+

Saving as a coping strategy for ECSHRS (1 = having savings; 0 = no savings)

Household able to use savings to offset economic shocks or risks

+

Microcredit (1 = yes; 0 = no) Household gaining microcredit +Source of microcredit (1 = government; 0 = other)

Microcredit obtained via government programs +

Family member gaining employment (1 = yes; 0 = other)

Household member gaining employment +

Improved public facilities nearby (1 = yes; 0 = no)

Household benefiting from improved public facili-ties nearby

+

Change variables during 2005–07Change in number of HH members Change in the number of household members –Change in marital status of HH head (1 = divorced; 0 = other)

Household becomes single-parent household –


Household head moves from agricultural to non-agricultural sector

+


Household head moves from formal to informal employment

–

Change in access to electricity for lighting(1 = access in 2007 but not in 2005; 0 = other)

Household had access to electricity for lighting in 2007 but not in 2005

+


Household received microcredit in 2007 but not in 2005

+

a Using a correlation matrix, we confirmed that there are no concerns about multicolinearity. See Dartanto and Nurkholis (2011).

Dow

nloa

ded

by [

Mac

quar

ie U

nive

rsity

] at

04:

38 0

8 Ju

ne 2

013

the determinants of poverty dynamics in indonesia: evidence … · 2014. 5. 5. · model to examine...

Documents