Poor children living in rich households: A Blurred Picture ...€¦ · Poor children living in rich households: A Blurred Picture or Hidden Realities? Keetie Roelen [email protected]

1

Poor children living in rich households: A Blurred Picture or Hidden Realities?

Keetie Roelen

[email protected]

Institute of Development Studies (IDS), Brighton, UK

July 2016

DRAFT: do not cite without author’s permission

Abstract

An expanding evidence base suggests that child poverty is a diverse experience, challenging the

notion that monetary poverty measures can serve as a proxy for multidimensional poverty and vice

versa. This holds particularly true for children given their dependence on others for the provision of

their basic needs. Yet few studies have investigated explanations for differential child poverty

outcomes and their diverse experiences in a comprehensive manner. This study breaks new ground

by providing a unique mixed methods investigation of drivers of differential child poverty outcomes

in Burundi, Ethiopia and Vietnam. It considers the role of measurement error and factors in the

public and private spheres in explaining why some children experience monetary poverty but not

multidimensional poverty and vice versa. It does so by capitalising on secondary large-scale

quantitative panel data and combining this with purposively collected primary qualitative and

participatory data in all three countries.

This study finds that measurement error only provides a partial explanation for differential

outcomes in monetary and multidimensional poverty for individual children. Educational attainment,

occupation and marital status of the heads of household play a significant but highly context-specific

role in explaining diverse experiences. Parental awareness of and attitudes towards investments in

child wellbeing is found to be crucial for promoting children’s outcomes despite limited monetary

resources in all contexts. These factors may also play into a trade-off between household wealth and

child wellbeing with short-term gains in wealth being prioritised over long-term gains in child

development. Cross-contextual findings also indicate that the availability of infrastructure, services

and social protection policies present important enabling factors for or barriers to securing child

wellbeing in situations with limited or ample monetary resources. Finally, aspirations crucially inform

children’s own decisions towards improvements in short-term versus long-term outcomes.

This study exemplifies the need for comprehensive child poverty measurement and mixed methods

analysis of its context-specific underlying drivers and diverse experiences. Most crucially it highlights

the importance of a nuanced policy response that responds to differential outcomes and

experiences in aiming to reduce all forms of child poverty.

Acknowledgements

The author would like to acknowledge the invaluable support of Tsegazeab Kidanemariam Beyene

and Hayalu Miruts in Mekelle, Ethiopia; the Southern Institute of Social Studies in Ho Chi Minh City,

Vietnam; Francisco Cabrero Hernandez; Helen Karki Chettri and Kimberly Wied in the process of data

collection and analysis. This research was funded by ESRC grant ES-K001833-1.

mailto:[email protected]

2

Introduction

The dichotomy between monetary and multidimensional approaches is a crude but persistent

bifurcation within poverty measurement. It is grounded in normative, conceptual and empirical

notions about the extent to which monetary resources can be translated into non-monetary

outcomes (Alkire et al., 2015; Ruggeri Laderchi, Saith, & Stewart, 2003; Thorbecke, 2008).

Conceptually, monetary measures are predicated on the assumption that if individuals have a

certain degree of purchasing power they will be able to fulfil their basic needs (Thorbecke, 2008;

Tsui, 2002) while multidimensional approaches incorporate a broad base of attributes in their

measures, thereby aiming to direct reflect the many manifestations of poverty (Hulme, 2015). This

distinction is particularly pertinent for children given their inherent limited agency in translating

household resources into improved individual outcomes (Roelen & Sabates-Wheeler, 2012; Vijaya,

Lahoti, & Swaminathan, 2014). Monetary indicators are considered more likely to fluctuate in the

short term than non-monetary indicators (Clark & Hulme, 2005; Hulme & Shepherd, 2003),

suggesting that multidimensional poverty is more representative of a permanent situation (Ayala,

Jurado, & Pérez-Mayo, 2011) or structural condition of poverty (Battiston, Cruces, Lopez-Calva, Lugo,

& Santos, 2013).

Expanding empirical evidence from both developed and developing country contexts suggests that

outcomes based on monetary and multidimensional approaches are often loosely associated and

that one measure cannot serve as a proxy for another (Baulch & Masset, 2003; Bernard, Dercon,

Orkin, & Taffesse, 2014; Bradshaw & Finch, 2003; Gaihre, 2012; Klasen, 2000; Kumar, 2012; Levine,

2012; Nilsson, 2010; Perry, 2002; Ruggeri Laderchi et al., 2003; Sahn & Stifel, 2003; Santos, 2012;

Wagle, 2009). Evidence with respect to child poverty is less extensive but points towards similar

mismatch patterns (Notten, 2012; Roelen, Gassmann, & de Neubourg, 2012; Roelen & Notten, 2013)

and indicates that monetary and multidimensional child poverty are different phenomena (Roelen,

forthcoming).

Despite the expanding evidence base on differential poverty outcomes, little research has been

undertaken to explore drivers underlying this dissonance. While the approaches’ conceptual

underpinnings may offer theoretical explanations, we are not aware of comprehensive empirical

investigations into underlying drivers. Such an investigation would not only serve academic curiosity

but is also crucial for policy purposes as “results and the related analyses may reveal the need for

different policy responses depending on which form(s) of poverty different groups of people

experience” (De Neubourg et al 2004, p. 16). A case for prioritising children is easily made: children

have different basic needs than adults do and a denial of those needs has long-term adverse and

often irreversible consequences, both for children and society at large (Roelen & Sabates-Wheeler,

2012).

This article seeks to advance research and contribute to child poverty reduction efforts by exploring

potential explanations for differential child poverty outcomes in Ethiopia and Vietnam. More

specifically it considers the role of measurement error and explanatory factors in the private and

public spheres. It does so using a unique combination of data and methods by drawing on literature

from high, medium and low income country contexts and by capitalising on secondary quantitative

panel data that has been made available to third party users and complementing this with analysis

of purposively collected qualitative data in both countries.

3

This article commences with a short review of evidence regarding the association between monetary

and multidimensional child poverty outcomes and potential explanations for differential outcomes.

This is followed by a section describing data, methods and analytical strategy. Next, the article

provides a brief overview of child poverty overlap and mismatch in Ethiopia and Vietnam, followed

by a discussion of results based on the empirical investigation. The article concludes with an

overview of lessons learned and implications for policy and future research.

Association and mismatch of monetary and multidimensional child poverty

A review of available evidence reveals that outcomes based on monetary and multidimensional

measures are weakly correlated and identify different groups of children as being poor. In Indonesia,

for example, the majority of children living in calorie deficient households were found to be above

the poverty line (Hadiwidjaja, Paladines, & Wai-Poi, 2013). In Congo Brazzaville, indicators for

monetary poverty appeared to be a crude tool for identifying children at risk of deprivation in terms

of their physical environment (Notten, 2009). With respect to child poverty in Darfur, Trani and

Cannings (2013) conclude that income deprivation does not adequately reflect the reality and

complexity of child poverty, which is a finding considered of particular pertinence in emergency

contexts as time, resource and logistical challenges often lead to singular and over-simplified

interventions. Similar results can be observed in developed country contexts. A cross-country study

in four countries in the European Union finds that children in the EU living in monetary poor

households are not necessarily those suffering from deprivation in non-monetary dimensions, and

vice versa, (Roelen & Notten, 2013). Similar observations were made for children in Portugal (Bastos,

Fernandes, & Passos, 2004). And in their study on child poverty in the UK, Brewer et al (2009) point

towards a ‘hump-shaped’ profile; as household income rises, children’s levels of deprivation first rise

and then fall. In sum, monetary and non-monetary measures can firmly be considered to provide

different pictures of child poverty.

Drivers of differential child poverty outcomes

Evidence about drivers of incongruent outcomes is fragmented at best, and fails to provide a

coherent insight into what may underpin differential pictures of child poverty. A first explanation is

grounded in the conceptual notion that monetary resources can be translated into non-monetary

outcomes but that issues with reliability and validity of underlying measures lead to differential

outcomes (Bradshaw & Finch, 2003). Explanations predicated on the view that monetary and

multidimensional poverty are two distinct concepts can be divided into factors pertaining to the

private and public spheres, such as household characteristics, attitudes and awareness, public policy

and cultural norms and values. This section explores potential drivers of mismatch in more detail,

tapping into studies of child poverty and poverty at large in low, medium and high-income country

contexts.

Measurement error

Various studies have considered the role of measurement error in poverty mismatch, particularly in

developed country contexts. It is based on the premise that any poverty measurement is subject to

error and therefore does not represent an accurate reflection of reality (Bradshaw & Finch, 2003). It

follows that attempts to combine or contrast outcomes based on flawed measures will result in

4

compounded errors and inevitably lead to different groups being identified as being poor when

using different measures.

In seeking to explain differential outcomes for monetary and non-monetary indicators of child

poverty in the UK, Brewer, O'Dea, Paull, and Sibieta (2009) question the reliability of the income

measure with respect to its equivalence scale and indicator for disposable income as well as the

issue of underreporting of income. At the same time they consider measures of living standards

potentially practically or conceptually flawed. Similarly, Berthoud and Bryan (2011) suggest that

inconsistencies between income estimates and findings based on a deprivation index (including

items related to daily living, financial strain and durables) at the lower end of the income distribution

can be explained by under-reporting of incomes by those with very low incomes. In their cross-

country analysis of poverty in the EU, Ayala et al. (2011) also point towards measurement errors

resulting from misreporting of income with respect to income-based measures and the potential

impact this has on the significance of the relationship between measures of income-based and

multidimensional poverty.

Differential use of units of analysis may also lead to or compound measurement error. Monetary

measures are predicated on household-level aggregates of income, consumption or expenditures

and inferences about individual-level poverty are made based on assumptions about intra-

household distribution. Multidimensional measures generally aim to include more individual-level

indicators, often as a direct consequence of the criticism that monetary measures do not adequately

capture individuals’ living conditions. Indeed, Ayala et al. (2011) postulate that the focus on different

types of individual wellbeing components forms one argument in explaining the lack of a statistically

significant relationship between income poverty and multidimensional poverty. Yet the current most

widely used measure of multidimensional poverty - the Multidimensional Poverty Index (MPI) - is

exclusively based on household-level indicators (Alkire et al., 2015) and studies investigating poverty

using this method in conjunction with household-level monetary measures find considerable

mismatch (Santos, 2012; Tran, Alkire, & Klasen, 2015).

Notwithstanding the existence of measurement error, both empirical and conceptual reasons

suggest that it can fully explain mismatch patterns. Ayala et al. (2011), for example, do not consider

the presence of measurement error sufficient to refute the weak correlation between income-based

and multidimensional poverty as they observed only a small proportion of those exiting income-

based poverty also to fare better with respect to multidimensional aspects of poverty. Despite

considering measurement error as an explanation for different outcomes of child poverty in the UK,

Brewer et al (2009) challenge the role of such error for conceptual reasons, stating: “It should not be

surprising that income and the other measures of living standards often give differing impressions of

the relative position of a particular household as ‘disposable income’ and ‘material living standards’

are fundamentally different concepts, so households with low disposable incomes need not be the

same as those households with low material living standards, even if both were measured perfectly.”

Structural factors

Research within childhood and child development traditions has long recognised that multiple risk

factors play a role in determining children’s outcomes. Parental health and education, child-parent

relationships and neighbourhood conditions, for example, have all been found to influence various

5

aspects of child wellbeing in different ways (Ciula & Skinner, 2015). Studies of intergenerational

transmissions of poverty distinguish between factors operating at different levels, such as individual-

and community-level factors (Engle, 2012), household level and extra-household level factors (Bird,

2007) or private and public transmissions (Harper, Marcus, & Moore, 2003). For the purposes of the

analysis in this paper, we explore factors in the ‘private’ and ‘public’ spheres in explaining overlap or

mismatch of child poverty outcomes, while recognising that in practice these factors are often so

interlinked that they do not form distinct processes per se (Engle, 2012; Harper et al., 2003). The

issue of aspirations is considered separately as it sits on the interface between the private and public

sphere.

Private sphere

As a result of widespread availability of household surveys, the correlation between individual and

household characteristics and poverty has been widely studied (Dercon, 2012). It is also our first

point of call for investigating drivers of differential child poverty findings, and examples of studies

considering the association between characteristics and various outcomes abound. Halleröd,

Rothstein, Daoud, and Nandy (2013) point towards the importance of educational attainment of

household heads in explaining lower levels of deprivation in a range of different non-monetary

dimensions of child wellbeing in low and middle-income countries. This positive effect can be

attributed to the interaction between education and productivity, but might also contribute to

practices that are beneficial to child wellbeing. A study on child care and nutrition in Ghana found

that the level of mother’s education played an important factor in childcare, including optimal

feeding practices (Ruel et al. 2001). In Zimbabwe, children living in a household with a disabled adult

were found to be more vulnerable to disease or chronic health issues regardless of household asset

ownership. The lack of a significant relationship between assets and health issues is attributed to the

fact that disabled adults lack the ability to seek medical care for their children regardless of

household wealth (Crea et al., 2013).

Beyond simple observable characteristics, parental awareness and attitudes appear a vital factor

with studies from especially high but increasingly also low-income country contexts establishing

linkages between parental engagement and children’s outcomes. Such engagement can reinforce or

override the monetary situation within a household. While parental attitudes and behaviours are

often correlated with household income (Goodman and Gregg 2010), it can also work to counteract

income effects. A systematic review of studies on parenting and child maltreatment in low-income

contexts indicates that parental awareness and parenting practices can reduce violence against

children and promote safe and nurturing environments (Knerr et al 2011). A global review shows

that parents’ roles in making connection, controlling behaviour, respecting individuality, modelling

appropriate behaviour and offering protection are crucial in preventing health risk behaviours and

improving health outcomes for adolescents (WHO 2007).

A related issue that is crucial for explaining different outcomes between monetary and

multidimensional child poverty refers to the potential trade-off between child wellbeing and

household wealth with respect to children’s role in work and supporting household (re)production.

Economic models of children’s time use assume that households make decisions about children’s

time allocation so that it maximises household utility, thereby balancing short-term income against

returns to investments in children’s long-term development (Orkin, 2012). It follows that if

6

households derive greater utility from short-term gains in income, children’s immediate wellbeing

may be compromised, either directly through children engaging in productive activities or indirectly

through children substituting for adult contributions to unpaid care work. The extent to which short-

term gains are prioritised over long-term gains will interact with parents’ as well as children’s

attitudes and aspirations (discussed in greater detail below).

It has to be noted that children’s engagement in work and domestic chores has also been found to

equip them with essential skills and can foster self-esteem (Woodhead, 2004), thereby challenging

binary notions about child work as being either good or bad. This holds particularly true for

adolescents as a strict dichotomy between childhood and adulthood “obscures our understanding of

what happens in the lives of children” (Bourdillon, 2006). Hence, the extent to which children’s

contributions to household wealth - either directly or indirectly – constitute a differential poverty

outcome should be considered with caution.

Public sphere

The public sphere pertains to the wider enabling environment that can have implications for either

monetary or multidimensional child poverty or both. The access and quality of services is a first

important aspect that may explain differential poverty outcomes. Bhutan’s low level of development

with weak infrastructure, incipient markets and poor access to services was found to be a crucial

factor in explaining why a large group of households experienced multidimensionally poverty but

was not considered monetary poor (Santos, 2012). The existence of user fees and lack of health

insurance however may compound the inability to access services (Halleröd et al., 2013), particularly

for children living in income-poor households and thereby leading to overlap of both types of

poverty. This ties into the issue of wider socioeconomic structures; a functioning labour market and

availability of employment opportunities are crucial for supporting children’s outcomes, both from a

monetary and non-monetary perspective (Harper et al., 2003).

Although the role of social relations and belief systems are less widely studied as they are less

amenable to being captured in quantitative methods (Harper et al., 2003), they play a crucial role in

determining children’s outcomes. At the same time, some of the greatest risks to child wellbeing

may originate from traditional cultures that are strongly grounded in patriarchal values and promote

practices such as early marriage and child labour (Boyden 2012). Broader conceptions about

parental sacrifice for children’s development, children’s contributions to household production and

welfare policies acting as a safety net for the most vulnerable will also play an important role in

outcomes for children regardless of their families’ financial status.

Aspirations

An explanatory factor that straddles the private and public spheres is that of aspirations. Aspirations

operate at the individual level but is highly influenced by external factors. They form an important

part of ‘mental models’ that influence decision-making processes and can override or bind

rationality (Bernard, Dercon, Orkin, & Tafesse, 2014). They are both consequence and cause of living

in deprivation but it can be argued that it is the ‘aspirations gap’ - the difference between someone’s

actual and desired standard of living - is what affects behaviour (Ray, 2003). This gap is bounded

though; if it is too small, there will be no desire to change current living conditions but if it is too

large, it will fail to provide positive encouragement (ibid). This non-linear relationship is

7

corroborated by findings from Andra Pradesh in India regarding the effect of mother’s aspirations on

children’s educational outcomes, indicating that medium aspirations had a larger effect than low or

high aspirations (Serneels & Dercon, 2014). Recent experimental research in rural Ethiopia suggests

that aspirations have a positive impact on long-term investments, which include investments in child

development through school enrolment and increased spending on education (Bernard, Dercon,

Orkin, & Tafesse, 2014). Further empirical evidence about the role of aspirations in determining

poverty and child wellbeing is limited (Gorard, Huat See, & Davies, 2012), particularly in reference to

the role of adults’ versus children’s aspirations.

Data and methods

The investigation of explanations underpinning differential child poverty outcomes takes a unique

mixed methods approach combining secondary quantitative and primary qualitative data. Mixed

methods approaches are widely acknowledged to offer breadth and specificity that quantitative and

qualitative measures in isolation fail to achieve (Shaffer, 2013). Approaches can vary in their degree

of integration (Carvalho & White, 1997) ranging from the use of participatory methods as a

complement to quantitative data for incorporating issues that are often overlooked or ignored

(Camfield, Crivello, & Woodhead, 2009) to a tightly integrated and iterative study aiming to duly

acknowledge and unpick poverty’s complexities (Roelen & Camfield, 2015). This study seeks to find a

middle ground by combining secondary quantitative panel data with primary qualitative data in an

iterative process.

Data

Sources of secondary quantitative data included in this study are the Ethiopian Rural Household

Survey (ERHS) waves from 1999, 2004 and 2009i and the Vietnam Household Living Standards Survey

(VHLSS) waves from 2004, 2006 and 2008ii.

The ERHS is a panel survey data set focusing on rural livelihoods with rounds in 1994, 1995, 1997,

1999, 2004 and 2009. Despite its relatively small size - it included 15 villages and a sample of 1,477

households in the first full round in 1994, it is representative of the main agricultural systems in

Ethiopia. Sample attrition between 1994 and 2009 is low, with a loss of only 16.1 percent (or 1.1

percent per year) and most of the attrition occurs in the early years of the study; attrition between

2004 and 2009 is less than 0.6 percent per year (Dercon, Hoddinott, & Woldehanna, 2012; Dercon &

Porter, 2011). This study uses data from the last three waves.

The VHLSS is a nationally representative data set and is based on the former Vietnam Living

Standards Survey (VLSS), which was conducted in 1993 and 1998. The VHLSS and has since been

undertaken every second year since 2002 by the Government Statistical Office (GSO), following the

World Bank’s Living Standards Measurement Survey (LSMS) methodology. Survey samples from

2002 to 2010 were drawn from a master sample, which is a random sample of the 1999 Population

Census enumeration areas and includes a rolling sample. It provides micro-data at the level of both

the household and its individual members on a range of issues related to children’s well-being and

poverty as well as social protection. Previous studies using the VHLSS data did not find attrition bias

(Baulch & Masset, 2003) and assumed an unbiased sample (Günther & Klasen, 2009).

Sample sizes per cross-sectional wave and for the full panel data sets are presented in Table 1.

8

Table 1 Sample statistics - quantitative data

1999 2004 2009 panel

Ethiopia (ERHS) 5054 3709 4937 1497

2004 2006 2008 panel

Vietnam (VHLSS) 12154 10696 9960 1068

Qualitative data collection took place in four sites in each country from August to December 2013.

Site selection was informed by analysis of secondary data, including quantitative data and other

reports, and pragmatic considerations. In Ethiopia, qualitative fieldwork took place in the northern

region of Tigray in Harresaw and Limat kushets, Harresaw tabia in Atsbi woreda and Kaslen and

Wela-Alabur kushets, Geblen tabia in Subhasaesie woreda. Tigray region was selected given its

relatively high poverty figures; research sites were chosen to mirror those included in the ERHS data

set. In Vietnam, qualitative data collection was undertaken in southern Mekong River Delta region in

Xã Mỹ Hòa and Xã Long Hậu communes in Dong Thap province and Xã Mỹ Hòa and Thị trấn Óc Eo in

An Giang province. These sites were selected as analysis of survey data indicated mismatch of

poverty outcomes was most prominent in these four sites. These sites were selected following

analysis of VHLSS data finding that mismatch of poverty outcomes was most outspoken in Mekong

River Delta region and the selected four sites within that region. Sample sizes per country are

presented in Table 2.

Table 2 Sample statistics - qualitative data

adults children Total

Ethiopia 88 61 159

Vietnam 145 78 223

Qualitative fieldwork engaged both adults and children and consisted of focus group discussions, key

informant interviews, household case studies and both individual- and group-based participatory

exercises. They aimed to elicit views and experiences regarding manifestations and causes of child

poverty. Given the technical nature of and negative connotation with the terms monetary poverty

and multidimensional poverty, questions for adults and children were framed around the positive

concepts of household wealth and child wellbeing as applicable in local languages. Adults and

children were asked about manifestations of child wellbeing and household wealth, the extent to

which they overlapped or not and explanations for differential outcomes. Community members in all

four sites formulated criteria for household wealth and child wellbeing and subsequently discussed

households’ situations with respect to these criteria.

Analysis of qualitative data involved a process of reading and re-reading, followed by a

categorisation and coding of responses. The standardised coding scheme was grounded in Maslow’s

hierarchy of needs theory (Maslow, 1954), Bronfenbrenner’s ecological model of human

development (Bronfenbrenner, 1979) and Minkkinen’s structural model of child wellbeing

(Minkkinen, 2013) with codes reflecting recurrent themes in both countries, ensuring consistency of

analysis.

9

Analytical strategy

A multitude of methods is used to investigate explanations for differential poverty outcomes,

including parametric and non-parametric analysis of quantitative data and analysis of interview,

group discussion exercise and case study data. An overview of the analytical strategy is provided in

Table 3.

Table 3 Analytical strategy

Analytical categories Method

Measurement error

Descriptive statistics

Community wealth and wellbeing ranking

Focus group discussions

Structural factors

Private sphere Multinomial regression analysis

Community wealth and wellbeing ranking

Focus group discussions

Case studies

Public sphere

Aspirations

The use of descriptive statistics includes simple poverty profiles as well as transition matrices

considering movements between poverty groups across waves. Multinomial regression models

estimate associations between individual, household and community level factors and child poverty

outcomes. Analysis is undertaken for each wave based on the panel sample, allowing for the

inclusion of previous time periods to control for poverty status in previous periods.

The dependent variable refers to ‘poverty group status’ with children belonging to either one of four

groups: (1) poverty overlap: children that are both monetary poor and multidimensionally poor (AB);

(2) positive mismatch: children that are monetary poor but are not multidimensionally poor (B); (3)

negative mismatch: children that are multidimensionally poor but are not monetary poor (A); and (4)

no poverty overlap: children that are not multidimensionally poor and are not monetary poor (C).

Independent variables at individual level include gender and age of the child. Household factors

include gender, age, marital status, educational attainment and occupational status of the

household head, household size and location. The Ethiopia models include an indicator for the

presence of household members in bad health or being immobile while the Vietnam models include

indicators for proportions of children in the household. Community indicators for Ethiopia include

distance to town, availability of electricity, piped water, schools and government hospital (see also

Dercon et al 2012, Dercon and Porter 2011). Community indicators in Vietnam include living in an

area where a disaster happened in 2004, living in an area with limited opportunities and living in an

area with an ECD centre. As the availability of community indicators in Ethiopia and in Vietnam are

limited to rural areas (Baulch 2011), we estimate an overall model and rural modeliii.

Given the volume of data incorporated in this mixed methods analysis, the results section will only

include selected findings but will otherwise refer to overall results rather than report detailed

outcomes. Readers are referred to the respective annexes for detailed information.

10

Mismatch monetary and multidimensional child poverty in Ethiopia and Vietnam

Before preceding to analyse explanations of different poverty outcomes, we report outcomes for

monetary and multidimensional child poverty in Ethiopia and Vietnam and their degree of mismatch

as investigated in earlier work (Roelen mimeo). Measures of monetary child poverty are based on

real per capita consumption in Ethiopia and real per capita expenditures in Vietnam while measures

of multidimensional child poverty include country-specific sets of indicators and employ the

‘counting approach’ for aggregation (Atkinson, 2003), mirroring methodologies as applied by OPHI’s

Multidimensional Poverty Index (MPI) (Alkire et al., 2015) and UNICEF’s Multiple Overlapping

Deprivation Analysis (de Neubourg, de Milliano, & Plavgo, 2014)iv.

Findings indicate that substantial groups of children are either multidimensionally poor (negative

mismatch) or only monetary poor (positive mismatch). Proportions of poverty mismatch are largest

in Ethiopia, with limited correlation between monetary and non-monetary indicators. Despite

greater correlation between monetary and non-monetary outcomes in Vietnam, children living in

multidimensional poverty are not necessarily monetary poor and vice versa. Sensitivity analysis

shows that these levels of mismatch persist across the income distribution. An overview of poverty

group proportions is presented in Table 4.

Table 4 Poverty overlap and mismatch in Ethiopia and Vietnam

Ethiopia

N

(# children)

monetary poor and

multidimensionally

poor (%)

multidimensionally

poor but not

monetary poor (%)

monetary poor but not

multidimensionally

poor (%)

non-

poor

(%)

Total

(%)

1999 2,893 19.2 23.9 25.6 31.3 100

2004 2,726 24.7 25.4 26.8 23.1 100

2009 3,230 13.5 23.2 26.0 37.3 100

Vietnam

N

(# children)

monetary poor and

multidimensionally

poor (%)

multidimensionally

poor but not

monetary poor (%)

monetary poor but not

multidimensionally

poor (%)

non-

poor

(%)

Total

(%)

2004 12,154 22 16 16 45 100

2006 10,696 15 13 16 56 100

2008 9,960 14 13 13 60 100

Source: Roelen, mimeo

Analysis of poverty dynamics points to many transitions between poverty groups over time in both

Ethiopia and Vietnam with large proportions of children changing poverty group from one period to

the next, including moves out of poverty but also falls into poverty. It should be noted that while the

empirical investigation in this article does include longitudinal analysis, the analysis focuses on

explanations for poverty group membership in a given wave as opposed to transitions between

poverty groups over time.

11

Explaining poverty mismatch

This section explores explanations of child poverty overlap and mismatch based on the combined

analysis of quantitative and qualitative data, focusing on measurement error, factors in the private

and public spheres, and aspirations.

Measurement error

In assessing the role of measurement error, we firstly consider the extent to which mismatch is

sensitive to the poverty lines used for both poverty measures. A clustering of mismatch around the

poverty lines would suggest that differential outcomes are a result of the ambiguous establishment

of such lines rather than of monetary and multidimensional poverty being distinct experiences.

Analysis of multidimensional child poverty across the income distribution in both Ethiopia and

Vietnam do not provide evidence for such clustering (see Table 5).

Table 5 Multidimensional poverty across income distribution (as denoted by consumption and expenditures)

Ethiopia Vietnam

1999 2004 2009 2004 2006 2008

multidimensionally

poor (%)

multidimensionally

poor (%)

multidimensionally

poor (%)

multidimensionally

poor (%)

multidimensionally

poor (%)

multidimensionally

poor (%)

deciles real per capita consumption/ expenditures

1 43.6 51.2 34.9 78.5 65.6 66.3

2 41.6 46.8 32.9 58.6 53.6 49.4

3 40.7 45.2 33.0 52.6 42.8 39.7

4 45.1 49.1 35.4 46.3 37.0 29.1

5 45.1 47.9 39.9 41.4 33.1 27.1

6 46.3 56.9 36.0 37.1 29.1 25.1

7 46.9 47.6 40.8 32.7 23.4 18.6

8 39.5 48.9 38.2 30.1 19.9 14.3

9 40.9 59.7 33.2 16.9 13.7 10.1

10 40.7 49.2 43.8 7.5 5.3 6.7

Source: Authors’ own calculations based on EHRS 1999, 2004, 2009 and VHLSS 2004, 2006 and 2008

In Ethiopia, multidimensional poverty rates fluctuate across deciles with multidimensional poverty

decreasing across the first three deciles but then increasing with greater per capita consumption,

suggesting the occurrence of positive mismatch (children being monetary but not

multidimensionally poor) at the bottom of the distribution and negative mismatch (children being

multidimensionally but not monetary poor) at the top of the distribution. In Vietnam,

multidimensional poverty rates decline as per capita expenditures increase. Nevertheless, one in 12

children in the highest decile also experience multidimensional poverty.

Qualitative findings shed further light on the extent to which measurement error may form an

explanation. We consider the extent to which indicators for household wealth and child wellbeing as

identified by community members in community wealth and wellbeing exercises match indicators

employed for the quantitative analysis. In Ethiopia, indicators defined by community members did

not directly mirror those available in the quantitative data, particularly with respect to household

wealth. The availability of livestock, land and labour was identified over access to monetary

12

resources. Indicators for child wellbeing were more similar with a strong focus on going to school

and working in or outside the home. In Vietnam, the identification of criteria for household wealth

and child wellbeing by community members was more strongly in line with quantitative indicators

available, including income and employment to denote household wealth and education, sanitation

and shelter for child wellbeingv.

While the discrepancies between indicators identified in qualitative exercises and those used in the

quantitative analysis may lead to measurement error, it does not suffice to explain poverty

mismatch. Community members across all sites in both countries indicated household wealth and

material child wellbeing to be distinct concepts with their own criteria. All communities also

identified households with inconsistent situations regarding household wealth and child wellbeing.

While there is no doubt that measurement of both monetary poverty and multidimensional poverty

is subject to error, both quantitative and qualitative findings suggest that child poverty mismatch

cannot be fully attributed to such error.

Structural factors – private sphere

Individual and household characteristics

The role of individual household characteristics was investigated using multinomial regression

models applied to three waves of data in both countries (see Tables A1 to A6 in the Annex). A range

of factors at the individual and household level were found to be important in determining poverty

overlap or mismatch but that the role they play is highly context-specific.

In Ethiopia, estimates indicate that living in a larger household increases the likelihood to

experiencing poverty overlap and positive mismatch but decreases the odds of negative mismatch.

Greater household size may lessen the need for children to withdraw from school or work many

hours in household production, which is an important component of the measure of

multidimensional poverty in Ethiopia. Education of the household head is also important, and this

importance intensified overtime. In 2009, living with a household head without any education

considerably increased the odds of poverty overlap and negative mismatch. In 2004 and 2009, living

with a household head who had completed primary education or more considerably increased

chances of negative mismatch, suggesting that while education may lead to improved economic

outcomes it does not necessarily go hand in hand with greater child wellbeing in terms of school

attendance and work in and outside of the house. Children’s individual characteristics of age and

gender were not significant.

In Vietnam, estimates suggest that living with a household head that is single, divorced or separated

increases the odds for belonging to any poverty group (poverty overlap, positive mismatch or

negative mismatch). Living with a household head that is widowed, however, decreases the odds for

poverty overlap and positive mismatch. Education of the household head appears to play an

important role in terms of poverty overlap: living with a household head having no education

increases the odds for experiencing poverty overlap while living a household head having secondary

education reduces those odds. Living with an unemployed household head is associated with higher

odds to experiencing poverty overlap and negative mismatch while living with a household head that

is a skilled professional is associated with lower odds for belonging to those poverty groups.

13

Ethnicity is strongly associated with poverty overlap; being of ethnic minority increases the odds for

being simultaneously poor across all years and regression models. Similarly, living in North West and

North Central Coast regions is associated with higher odds to being simultaneously poor. Living in

Mekong River Delta, however, is associated with higher odds for poverty overlap and negative

mismatch but decreases the odds for positive mismatch. Again, children’s individual characteristics

were not found to be significant factors in any of the multinomial models.

Trade-off household wealth and child wellbeing

Findings in both countries provide evidence for households’ delicate balancing act between

household wealth and child wellbeing in explaining both positive and negative poverty mismatch.

In Ethiopia, descriptive statistics show that children in higher income quintiles engage in a greater

number of hours worked in household production (see Figure 1).

Figure 1 Livestock ownership and family work across consumption

deciles for children aged 10-15 in rural Ethiopia

810

12

14

16

ave

rage

ho

urs

fam

ily w

ork

(p

er

we

ek)

23

45

6

ave

rage

liv

esto

ck o

wn

ers

hip

(T

LU

)

0 2 4 6 8 10

real per capita consumption (deciles)

livestock (TLU) hours family work (per week)

Source: Roelen (2015)

This finding is corroborated by qualitative data in which adults and children indicate that children in

wealthier households work are usually more involved in herding livestock, contributing to family

production or doing domestic chores. This may go at the expense of studying at home or going to

school: “Sometimes children in rich households are obliged to work in farm activities rather than

going to school” [female caregiver, Geblen, Ethiopia].

A gendered effect appears at work with qualitative data suggesting that children living in male-

headed households more likely to work and experience negative mismatch. Cockburn and Dostie

(2007) find similar results, suggesting that female heads might give greater priority to schooling or

that there are fewer possibilities for children to engage in productive work in female-headed

households. A gender effect is also at play on behalf of children: girls were more likely to undertake

domestic chores and boys to work on the family farm and herding livestock.

14

Qualitative data from Vietnam does not provide strong evidence for direct contributions of children

to productive activities but does point to the existence of substitution effects. Many parents in the

communities included in qualitative fieldwork were observed to be working far away from home,

often leaving their children to care for elderly and disabled household members. While Cuong and

Linh (2013) find effects of parental migration on children’s time use to be negligible, qualitative

findings in this study suggest that time allocation was considerably impacted through reductions in

time spent on studying or leisure: “I stopped my study 2 years ago at grade 5. I help my sister to take

of her children at home” [girl child, An Giang, Vietnam].

Awareness and attitudes

Perspectives from adults and children in the qualitative data indicate that awareness and attitudes

of parents are an important factor in explaining poverty mismatch; they can secure child wellbeing

even when the household has few monetary resources but also contribute to poor child wellbeing

despite the availability of monetary resources.

Adults’ and children’s responses across the board highlight that parents have an important influence

on children’s outcomes beyond the availability of monetary resources. Some respondents suggested

that wealth and personal attention to children may be inversely related, with households

experiencing monetary poverty placing greater emphasis on children’s education and future

opportunities as well as mitigating the effects of limited economic resources: “We are poor but we

try to let our children study properly because we do not want our children to feel disadvantaged

compared to other children” [female caregiver, Dong Thap, Vietnam].

Qualitative findings indicate that general awareness and attitudes regarding child wellbeing have

greatly improved in recent years, particularly in Ethiopia. Parents and social workers indicate how

government campaigns and extension services has instilled the importance of education,

immunisation, pre- and antenatal care and family planning, as indicated by a woman from Limeat:

“People’s general attitudes towards raising and caring for children have significantly changed over

time. For example, most mothers follow up pre and anti-natal care, follow vaccinations, most parents

send their kids to school on time, reduced underage marriages and love and attention for children

increased” [woman, Limeat, Ethiopia]. These findings were corroborated in reference to the balance

between schooling and work with respondents attaching great value to education and prioritised

school over work as education is considered crucial for securing future livelihoods.

Notwithstanding these positive effects, researcher observations and discussions with children

revealed that such expressed perspectives were not necessarily in line with reality as children’s

education was discontinued or interrupted when reaching secondary school. This discrepancy

appears subject to a gender effect with girls’ education receiving less priority than boys. A gender

effect also extends to the household head; children that were identified in the qualitative data as

experiencing good wellbeing despite living in a poor household were more likely to be part a female-

headed household, while children experiencing poor wellbeing despite living in a relatively affluent

household were more likely to be part of a male-headed household.

15

Structural factors – public sphere

Access to services

An important component of the public sphere in explaining poverty overlap and mismatch is the

access to services. In Ethiopia, school attendance rates in the quantitative sample, for example, rose

from 35 percent in 1999 to 65 percent in 2009 and qualitative findings suggest that the availability of

primary education played an important role in this increase. More generally, qualitative findings

indicate that availability of services such access to schools, health posts and safe drinking water is

important in driving positive mismatch - i.e. ensuring children’s wellbeing even if children live in

monetary poor households. By the same token, the absence of such infrastructure can lead to

negative mismatch - multidimensional child poverty even if a child is living in a household with

greater wealth, as illustrated by a social worker from Harresaw: “The wellbeing situation of children

in this community has generally improved over time because infrastructure like health posts, and

primary education are established near to our community. Nevertheless, there are still some critical

problems affecting children like long distance to get to school above grade 4 and lack of potable

water” [social worker, Harresaw, Harresaw].

In Vietnam, regression estimates and qualitative discussions do not point towards a significant role

for services and infrastructure in influencing poverty status in explaining poverty mismatch, largely

due to widespread availability of services and therefore little variance in the data (Baulch & Dat,

2011). Qualitative findings strongly indicate that government social protection programmes play a

positive role in securing children’s needs despite household poverty and therefore driving poverty

mismatch. The most frequently mentioned policy was the ‘poverty certificate’ or ‘poverty book’

policy, which applies to monetary poor households and gives access to support such as tuition fee

waivers, health insurance and commune support: “My child saw other children having poor

household certificate and he asked me why we did not have one. People with such a certificate

receive a great amount of support whereas we don’t receive any” [female caregiver, Dong Thap,

Vietnam].

At the same time, experiences with government involvement were not altogether positive and

qualitative findings suggest that it could also lead to negative mismatch or poverty overlap.

Respondents pointed to the importance of having legal documentation for accessing services

regardless of income status and how the access of such documentation can lead to negative

mismatch: “I have never gone to school because my family lives in a rental house that means we are

temporary residents, so I cannot have legal documents, like birth certificate for school application"

[child, An Giang, Vietnam]. A number of respondents also indicated how they were no experiencing

poverty overlap due to having been moved from the area that securing their main livelihoods: “My

family is in poverty. The hamlet’s officers have had us move here and now we are in a difficult

condition and there is no foundation to work anymore” [female adult, Dong Thap, Vietnam].

Socioeconomic context

Wider socioeconomic conditions were found to be an important driver for explaining poverty

overlap and mismatch. In Vietnam, the absence of stable jobs was considered an important barrier

to securing a stable situation for children, both in terms of income and other areas of wellbeing. It

creates a difficult reality for parents having to work long hours away from home, sometimes leaving

16

children in the care of others with potential adverse effects on child wellbeing: “Household poverty

means that we do not have stable job, which results in unstable income” [female adult, Dong Thap,

Vietnam]. In Ethiopia, lack of economic opportunities beyond agricultural activities was mentioned

as posing barriers to both adults and children in their attempts to improve monetary and non-

monetary outcomes.

Cultural norms and values

Although the role of cultural norms and practices was not explicitly incorporated in fieldwork scripts,

their role inevitable emerged in discussions around what constitutes child wellbeing and what

contributes to child wellbeing. In both countries, looking clean and well-clothed was deemed

important by both adults and children for gaining respect from family and community members.

Although the purchase of clothing and soap was considered to be linked to the availability of

monetary resources, hygienic practices were linked to how strongly this aspect was valued by adult

caregivers.

In Vietnam, living up to societal norms and standards was deemed particularly important. Adult

respondents referred to the importance of obeying parents and teachers, of studying hard and not

being lazy and dressing appropriately. Various respondents pointed towards a direct mismatch

between the emphasis on this component of child wellbeing and availability of monetary resources

with wealthier parents being unable to spend adequate time with children to instil those values: “A

well-off family can have a lot money for children but if parents just pay attention to their business

and have less time to take care of their children, those children surely do not feel happy and in many

cases, those children will be easily deprived” [teacher, An Giang, Vietnam].

Another recurrent element in Vietnam referred to children’s responsibilities towards caring for

elderly and disabled adults in the households, particularly when parents work in areas far from

home: “Parents advise me that I should not go out too much and help my paternal grandparents”

[child, An Giang, Vietnam]. While children appeared to take pride in care responsibilities, there were

also signs that they undermined the opportunity to take part in school, study or leisure activities, as

discussed above.

In Ethiopia, findings indicate that engaging in domestic chores or working on the household farm is a

positive attribute for children: “I don’t send my children to work for other households but I believe

children should work at home in household production” [Male caregiver, Harresaw, Ethiopia] While

the role of work in child wellbeing has to be considered with caution (as discussed above), children’s

responses in this study suggest that the balance often tips in such a way that child wellbeing may

undermined: “I can say my wellbeing is good and bad. It is good because I am in school.

My wellbeing is bad because I am working at home when I return from school” [girl child, Harresaw,

Ethiopia]. This corroborates other research in rural Ethiopia on patterns of children’s work (Abebe,

2007).

Aspirations

An explanatory factor that straddles the private and public spheres is that of aspirations. Children’s

aspirations are important in their own considerations of what constitutes child well-being and how

to achieve it. Qualitative findings in Ethiopia in particular indicate that lack of local economic

17

opportunities and role models within the community shape children’s aspirations and ambitions for

the future, inspiring choices regarding school and work and playing into differential outcomes with

respect to income and non-income based child poverty.

In Ethiopia, children’s aspirations were also considered an important factor in determining whether

a child was going to school or not regardless of monetary resources. Seeking low-skilled work in

Saudi-Arabia was frequently mentioned as a more desirable opportunity than continuing education

in pursuit of a skilled job in the local area. Limited economic opportunities and lack of role models

appeared to feed into these aspirations. In other cases children’s aspirations and parents’ attitudes

appeared in conflict with each other: “If I pass the national examination, I want to continue my

education in the town of Atsbi. I want to be an engineer in order construct roads to my community in

particular and my country in general. But my father wants me to join the Dera high school in order to

support him” [girl child, Harresaw, Ethiopia].

Conclusion

While the empirical evidence on differential outcomes of monetary and multidimensional child

poverty is steadily expanding, few studies have considered underlying drivers. This article aimed to

provide a unique and rich empirical contribution by exploring drivers that differential child poverty

outcomes in Ethiopia and Vietnam. It applies a mixed methods approach using secondary repeated

cross-sectional and longitudinal data and primary qualitative data of adults’ and children perceptions

that allows for a strong level of breadth of specificity.

Findings suggest that measurement error only provides a partial explanation for differential findings

but cannot account for the full extent of mismatch. Structural factors at individual, household,

community and government level are strong in securing child wellbeing despite monetary poverty or

vice versa. Educational attainment, occupation and marital status of heads of household can explain

differential poverty outcomes for children but the strength and direction of the association is highly

context-specific. Parental awareness of and attitudes towards the raising and educating children

plays an important role in translating high or low levels of household resources in improved child

wellbeing. Qualitative findings in Ethiopia point towards a gender dimension with ‘positive’

mismatch (low household resources but good child wellbeing) occurring more frequently in female-

headed households and ‘negative’ mismatch (higher household resources but low child wellbeing)

being more prevalent among male-headed households. Greater emphasis on short-term gains in

household production over long-term investments in child development may lead to a trade-off

between household wealth and chid wellbeing, particularly in terms of the balance between school,

work and leisure. Findings in Vietnam point to the important role of government services and stable

local employment opportunities in preventing both monetary and multidimensional poverty for

children. Lack of such local economic opportunities as well as lack of role models in Ethiopia were

found to shape children’s aspirations and their own choices with respect to the balance between

monetary and non-monetary outcomes.

In conclusion it should be noted that explanations for differential poverty outcomes are not mutually

exclusive and that factors fungible. Perceptions about the appropriate balance between the creation

of household wealth and child wellbeing, for example, are influenced by children’s and adults’

attitudes and aspirations in light of the situation of the household as a whole and grounded in wider

18

structures of service provision, socioeconomic contexts and cultural norms and practices. In other

words, while explanations can be explored and tested empirically - as undertaken in this paper - the

weight that is attached to each of these explanations will inevitably be informed by the researcher’s

conceptual, ideological, disciplinary and methodological underpinnings.

References

Abebe, T. (2007). Changing Livelihoods, Changing Childhoods: Patterns of Children's Work in Rural

Southern Ethiopia. Children's Geographies, 5(1-2), 77-93. doi: 10.1080/14733280601108205

Alkire, S., Foster, J., Seth, S., Santos, M. E., Roche, J. M., & Ballon, P. (2015). Multidimensional

Poverty Measurement and Analysis. Oxford: Oxford University Press.

Atkinson, A. B. (2003). Multidimensional deprivation: contrasting social welfare and counting

approaches. Journal of Economic Inequality, 1, 51-65.

Ayala, L., Jurado, A., & Pérez-Mayo, J. (2011). Income Poverty and Multidimensional Deprivation:

Lessons from Cross-Regional Analysis. Review of Income and Wealth, 57(1), 40-60. doi:

10.1111/j.1475-4991.2010.00393.x

Bastos, A., Fernandes, G. L., & Passos, J. (2004). Child income poverty and child deprivation: an essay

on measurement. International Journal of Social Economics, 31(11/12), 1050-1060. doi:

doi:10.1108/03068290410561168

Battiston, D., Cruces, G., Lopez-Calva, L., Lugo, M., & Santos, M. (2013). Income and Beyond:

Multidimensional Poverty in Six Latin American Countries. Social Indicators Research, 112(2),

291-314. doi: 10.1007/s11205-013-0249-3

Baulch, B., & Dat, H. V. (2011). Poverty dynamics in Vietnam, 2002 to 2006. In B. Baulch (Ed.), Why

Poverty Persists. Poverty dynamics in Asia and Africa (pp. 219-254). Cheltenham: Edward

Elgar Publishing.

Baulch, B., & Masset, E. (2003). Do Monetary and Nonmonetary Indicators Tell the Same Story About

Chronic Poverty? A Study of Vietnam in the 1990s. World Development, 31(3), 441-453.

Bernard, T., Dercon, S., Orkin, K., & Tafesse, A. S. (2014). The Future in Mind: Aspirations and

Forward-Looking Behaviour in Rural Ethiopia. CSAE Working Paper, WPS/2014-16.

Bernard, T., Dercon, S., Orkin, K., & Taffesse, A. S. (2014). The Future in Mind: Aspirations and

Forward-Looking Behaviour in Rural Ethiopia. In C. f. t. S. o. A. Economies (Ed.), CSAE

Working Paper (Vol. WPS/2014-16). Oxford: University of Oxford.

Berthoud, R., & Bryan, M. (2011). Income, Deprivation and Poverty: A Longitudinal Analysis. Journal

of Social Policy, 40(01), 135-156. doi: doi:10.1017/S0047279410000504

Bird, K. (2007). The intergenerational transmission of poverty: An overview. CPRC Working Paper, 99.

Bourdillon, M. (2006). Children and Work: A Review of Current Literature and Debates. Development

& Change, 37(6), 1201-1226. doi: 10.1111/j.1467-7660.2006.00519.x.

Boyden, Jo (2012). The rise of modernity and competing ideologies of childhood in the 21st Century.

Presentation at the Department of Childhood Studies, Rutgers University, 02 October 2012.

Bradshaw, J., & Finch, N. (2003). Overlaps in Dimensions of Poverty. Journal of Social Policy, 32, 513-

525.

Brewer, M., O'Dea, C., Paull, G., & Sibieta, L. (2009). The living standards of families with children

reporting low incomes. DWP Research Reports 577. London: Department for Work and

Pensions.

Bronfenbrenner, U. (1979). The Ecology of Human Development: Experiments by Nature and Design.

Cambridge: Harvard University Press.

19

Camfield, L., Crivello, G., & Woodhead, M. (2009). Wellbeing Research in Developing Countries:

Reviewing the Role of Qualitative Methods. Social Indicators Research, 90, 5-31.

Carvalho, S., & White, H. (1997). Combining the Quantitative and Qualitative Approaches to Poverty

Measurement and Analysis. The Practice and Potential. World Bank Technical Paper, 366.

Ciula, R., & Skinner, C. (2015). Income and Beyond: Taking the Measure of Child Deprivation in the

United States. Child Indicators Research, 8(3), 491-515. doi: 10.1007/s12187-014-9246-6

Clark, D., & Hulme, D. (2005). Towards A Unified Framework for Understanding the Depth, Breadth

and Duration of Poverty. Paper presented at the The Many Dimensions of Poverty, Brasilia,

Brazil, 29-31 August 2005.

Cockburn, J., & Dostie, B. (2007). Child Work and Schooling: The Role of Household Asset Profiles and

Poverty in Rural Ethiopia. Journal of African Economies, 16(4), 519-563. doi:

10.1093/jae/ejl045

Crea, T. M., Lombe, M., Robertson, L. A., Dumba, L., Mushati, P., Makoni, J. C., . . . Gregson, S. (2013).

Asset ownership among households caring for orphans and vulnerable children in rural

Zimbabwe: the influence of ownership on children's health and social vulnerabilities. AIDS

Care, 25(1), 126-132. doi: 10.1080/09540121.2012.687812

Cuong, N. V., & Linh, V. H. (2013). Should Parents Work Away from or Close to Home? The Effect of

Temporary Parental Absence on Child Poverty and Children’s Time Use in Vietnam. Young

Lives Working Paper, 104.

de Neubourg, C., de Milliano, M., & Plavgo, I. (2014). Lost (in) Dimensions: Consolidating progress in

multidimensional poverty research Office of Research Working Paper No. 2014-04. Florence:

UNICEF Office of Research.

Dercon, S. (2012). Understanding Child Poverty in Developing Countries: Measurement and Analysis.

In J. Boyden & M. Bourdillon (Eds.), Childhood Poverty. Multidisciplinary Approaches (pp. 52-

74). Basingstoke: Palgrave Macmillan.

Dercon, S., Hoddinott, J., & Woldehanna, T. (2012). Growth and Chronic Poverty: Evidence from

Rural Communities in Ethiopia. The Journal of Development Studies, 48(2), 238-253. doi:

10.1080/00220388.2011.625410

Dercon, S., & Porter, C. (2011). A poor life? Chronic poverty and downward mobility in rural Ethiopia,

1994 to 2004. In B. Baulch (Ed.), Why Poverty Persists. Poverty dynamics in Asia and Africa.

(pp. 65-91). Cheltenham: Edward Elgar Publishing.

Engle, P. (2012). Poverty and Developmental Potential. In J. Boyden & M. Bourdillon (Eds.),

Childhood Poverty. Multidisciplinary Perspectives (pp. 129-147). Basingstoke: Palgrave

Macmillan.

Gaihre, R. H. (2012). Comparison between Multidimensional Poverty Index and Monetary Poverty for

Nepal Paper presented at the Dynamic Comparison between Multidimensional Poverty and

Monetary Poverty, Oxford.

Goodmann, A. & Gregg, P. (2010). Poorer children’s educational attainment: how important

are attitudes and behaviour? Joseph Rowntree Foundation.

Gorard, S., Huat See, B., & Davies, P. (2012). The impact of attitudes and aspirations on educational

attainment and participation. York: Joseph Rowntree Foundation.

Günther, I., & Klasen, S. (2009). Measuring Chronic Non-Income Poverty. In T. Addison, D. Hulme &

R. Kanbur (Eds.), Poverty Dynamics: Interdisciplinary Perspectives. New York: Oxford

University Press.

20

Hadiwidjaja, G., Paladines, C., & Wai-Poi, M. (2013). The Many Dimensions of Child Poverty in

Indonesia; Patterns,Differences and Associations. Jakarta: World Bank.

Halleröd, B., Rothstein, B., Daoud, A., & Nandy, S. (2013). Bad Governance and Poor Children: A

Comparative Analysis of Government Efficiency and Severe Child Deprivation in 68 Low- and

Middle-income Countries. World Development, 48(0), 19-31. doi:

http://dx.doi.org/10.1016/j.worlddev.2013.03.007

Harper, C., Marcus, R., & Moore, K. (2003). Enduring Poverty and the Conditions of Childhood:

Lifecourse and Intergenerational Poverty Transmissions. World Development, 31(3), 535-

554.

Hulme, D. (2015). Global Poverty. Global governance and poor people in the post-2015 era.

Abingdon: Routledge.

Hulme, D., & Shepherd, A. (2003). Conceptualizing Chronic Poverty. World Development, 31(3), 403-

423.

Klasen, S. (2000). Measuring Poverty and Deprivation in South Africa. Review of Income and Wealth,

46(1), 33-58.

Knerr, W., Gardner, F. & Cluver, L. (2011). Parenting and the prevention of child maltreatment in

low- and middle-income countries. Sexual Violence Research Initiative. Oxford: University of

Oxford.

Kumar, R. (2012). Comparing Multidimensional Poverty and Consumption Poverty based on Primary

Survey in India Paper presented at the Dynamic Comparison between Multidimensional

Poverty and Monetary Poverty, Oxford.

Levine, S. (2012). Comparing Multidimensional and Monetary Poverty in Uganda Paper presented at

the Dynamic Comparison between Multidimensional Poverty and Monetary Poverty, Oxford.

Maslow, A. (1954). Motivation and personality. New York: NY: Harper.

Minkkinen, J. (2013). The Structural Model of Child Well-Being. Child Indicators Research, 6(3), 547-

558. doi: 10.1007/s12187-013-9178-6

Nilsson, T. (2010). Health, Wealth and Wisdom: Exploring Multidimensional Inequality in a

Developing Country. Social Indicators Research, 95(2), 299-323. doi: 10.1007/s11205-009-

9461-6

Notten, G. (2009). Is Monetary Poverty a Suitable Proxy for Deprivation in the Physical Environment?

Children, Youth and Environments, 19(2), 20-35.

Notten, G. (2012). A multidimensional poverty profile of child poverty in Congo Brazzavill. In A.

Minujin & S. Nandy (Eds.), Global child poverty and well-being. Bristol: The Policy Press.

Orkin, K. (2012). Are Work and Schooling Complementary? In J. Boyden & M. Bourdillon (Eds.),

Childhood Poverty. Multidisciplinary Approaches. (pp. 298-313). Basingstoke: Palgrave

Macmillan.

Perry, B. (2002). The Mismatch between Income Measures and Direct Outcome Measures of

Poverty. Social Policy Journal of New Zealand, 19, 101-127.

Ray, D. (2003). Aspirations, Poverty and Economic Change. In N. Y. U. a. I. d. A. E. (CSIC) (Ed.). New

York: New York University and Instituto de Analisis Economico (CSIC).

Roelen, K., & Camfield, L. (Eds.). (2015). Mixed Methods Research in Poverty and Vulnerability.

Sharing Ideas and Learning Lessons. Basingstoke: Palgrave Macmillan.

Roelen, K., Gassmann, F., & de Neubourg, C. (2012). False positives or hidden dimensions: What can

monetary and multidimensional measurement tell us about child poverty in Vietnam?


21

International Journal of Social Welfare, 21(4), 393-407. doi: 10.1111/j.1468-

2397.2011.00836.x

Roelen, K., & Notten, G. (2013). The Breadth of Child Poverty in Europe: An Investigation into

Overlap of Deprivations. Poverty & Public Policy, 5(4), 319-335. doi: 10.1002/pop4.53

Roelen, K., & Sabates-Wheeler, R. (2012). A child-sensitive approach to social protection: serving

practical and strategic needs. Journal of Poverty and Social Justice, 20(3), 291-306. doi:

10.1332/175982712X657118.

Ruel, M., Armar-Klemesu, M. & Arimond, M. (2001). A Multiple-Method Approach to Studying

Childcare in Urban Environment: The Case of Accra, Ghana. FCND Discussion Paper No. 116.

Washington DC: IFPRI.

Ruggeri Laderchi, C., Saith, R., & Stewart, F. (2003). Does it Matter that we do not Agree on the

Definition of Poverty? A Comparison of Four Approaches. Oxford Development Studies,

31(3), 243-274.

Sahn, D., & Stifel, D. (2003). Exploring Alternative Measures of Welfare in the Absence of

Expenditure Data. Review of Income and Wealth, 49(4), 463-489.

Santos, M. E. (2012). Tracking Poverty Reduction in Bhutan: Income deprivation alongside

deprivation in other sources of happiness. Paper presented at the Dynamic Comparison

between Multidimensional Poverty and Monetary Poverty, Oxford.

Serneels, P., & Dercon, S. (2014). Aspirations, Poverty and Education. Evidence from India. Young

Lives Working Paper, 125.

Shaffer, P. (2013). Ten Years of “Q-Squared”: Implications for Understanding and Explaining Poverty.

World Development, 45, 269-285. doi: http://dx.doi.org/10.1016/j.worlddev.2012.12.008

Thorbecke, E. (2008). Multidimensional Poverty: Conceptual and Measurement Issues. In N. Kakwani

& J. Silber (Eds.), The Many Dimensions of Poverty. New York: Palgrave Macmillan.

Tran, V. Q., Alkire, S., & Klasen, S. (2015). Static and Dynamic Disparities between Monetary and

Multidimensional Poverty Measurement: Evidence from Vietnam. OPHI Working Paper No.

97.

Trani, J.-F., & Cannings, T. I. (2013). Child Poverty in an Emergency and Conflict Context: A

Multidimensional Profile and an Identification of the Poorest Children in Western Darfur.

World Development, 48(0), 48-70. doi: http://dx.doi.org/10.1016/j.worlddev.2013.03.005

Tsui, K.-y. (2002). Multidimensional Poverty Indices. Social Choice and Welfare, 19, 69-93.

Vijaya, R. M., Lahoti, R., & Swaminathan, H. (2014). Moving from the Household to the Individual:

Multidimensional Poverty Analysis. World Development, 59(0), 70-81. doi:


Wagle, U. (2009). Capability Deprivation and Income Poverty in the United States, 1994 and 2004:

Measurement Outcomes and Demographic Profiles. Social Indicators Research, 94(3), 509-

533.

WHO. (2007). Helping parents in developing countries improve adolescents’ health. Geneva: WHO.

Woodhead, M. (2004). Psychosocial impacts of child work: a framework for research, monitoring and

intervention. The International Journal of Children’s Rights, 12(4), 321-377.




22

Table A1 Multinomial regression Ethiopia 1999

1999

Multinomial model Multinomial model rural (with inclusion of community factors)

AB A B AB A B

b/se b/se b/se b/se b/se b/se

Child is female 1.237 0.665 1.030 1.072 0.707 0.924

(0.334) (0.188) (0.265) (0.328) (0.223) (0.262)

Age of child 1.000 1.013 0.975 0.955 0.991 0.951

(0.059) (0.066) (0.055) (0.069) (0.076) (0.063)

Household is female 2.154 0.351 1.225 2.192 0.359 0.771

(1.463) (0.286) (0.957) (1.720) (0.319) (0.729)

Age of household head 1.021 1.007 1.008 1.024 1.010 1.009

(0.012) (0.012) (0.012) (0.014) (0.014) (0.014)

Household size 1.015 0.872 1.085 1.112 0.933 1.140

(0.072) (0.065) (0.075) (0.093) (0.080) (0.090)

Household head is single 1.106 0.513 0.855 0.880 0.561 0.742

(0.373) (0.176) (0.282) (0.360) (0.232) (0.293)

Household head is single/divorced/widowed 0.870 0.874 0.795 0.470 0.799 0.520

(0.487) (0.661) (0.420) (0.316) (0.636) (0.314)

Household head is polygamous 0.697 1.118 0.719 0.447* 0.926 0.465*

(0.238) (0.379) (0.235) (0.176) (0.358) (0.173)

23

Household head has primary education or more 0.399* 0.294* 0.401* 0.317* 0.270 0.342*

(0.165) (0.164) (0.151) (0.162) (0.182) (0.150)

Household head does domestic work 0.334 1.655 0.304 0.347 1.235 0.512

(0.251) (1.419) (0.281) (0.296) (1.136) (0.549)

Household head does manual work 0.000 0.000 0.577 0.000 0.000 0.014*

(0.000) (0.000) (0.840) (0.000) (0.000) (0.023)

Household head does non-manual work 0.094* 0.583 0.071* 0.008*** 0.359 0.007***

(0.103) (0.522) (0.077) (0.011) (0.334) (0.009)

Household head is not in labour force 0.000 0.621 0.000 0.591 1.97e+15 1.416

(0.000) (0.769) (0.000) (5.80e+07) (1.08e+23) (1.30e+08)

Number of members in bad health/immobile 1.108 1.067 0.971 1.004 0.951 0.904

(0.099) (0.106) (0.085) (0.103) (0.105) (0.091)

Child lives in Tigray 6.540*** 3.648** 2.902 96.144*** 6.715 34.206**

(3.600) (1.803) (1.726) (103.480) (6.899) (37.419)

Child lives in Oromya 1.495 0.508 1.068 29.843*** 1.211 14.770***

(0.649) (0.196) (0.430) (25.052) (0.734) (11.705)

Child lives in SNNPR 8.206*** 0.500 7.270*** 0.510 0.152* 0.804

(3.312) (0.210) (2.639) (0.368) (0.114) (0.526)

Distance to town in kilometres

0.595*** 0.824 0.672***

(0.058) (0.083) (0.062)

24

Community has electricity

243.441*** 8.119 60.855**

(352.421) (11.835) (85.066)

Community has piped water

0.000*** 0.000 0.003**

(0.001) (0.000) (0.007)

Community has primary school

0.070*** 0.397 0.048***

(0.050) (0.244) (0.033)

Community has junior school

26.957* 6.369 9.000

(37.006) (9.528) (11.937)

Community has high school

22.031 5.36e+09*** 9.484

(38.205) (9.37e+09) (13.959)

Community has government hospital

8.974** 2.484 5.630**

(6.095) (1.459) (3.720)

Number of observations 511.000

468.000

P-value 0.000

0.000

Pseudo R-Square 0.139

0.215

BIC 1535.214 . . 1455.502

Note: omitted categories are: child is male; household head is male, household head is married; household head has less than primary education; household head is farmer

or does family work; child lives in Tigray.


2004

25


AB A B AB A B


Child is female 1.166 0.940 0.785 1.092 0.803 0.809

(0.210) (0.169) (0.135) (0.230) (0.169) (0.160)

Age of child 1.033 1.031 1.022 1.110** 1.091* 1.059

(0.033) (0.033) (0.031) (0.042) (0.041) (0.037)


(0.496) (0.407) (0.346) (0.518) (0.296) (0.399)


(0.009) (0.009) (0.009) (0.011) (0.011) (0.011)

Household size 0.958 1.062 0.932 0.952 1.076 0.963

(0.038) (0.043) (0.037) (0.048) (0.054) (0.045)


(0.353) (0.527) (0.297) (0.474) (0.764) (0.474)

Household head is single/divorced/widowed 4.661** 1.181 2.414 2.408 0.783 1.419

(2.430) (0.744) (1.326) (1.647) (0.640) (0.946)

Household head is polygamous 1.311 1.516 1.134 1.323 1.449 1.059

(0.300) (0.371) (0.247) (0.356) (0.413) (0.271)

Household head has primary education or more 0.716 1.783* 0.950 0.621 1.727 1.140

(0.190) (0.461) (0.230) (0.207) (0.532) (0.321)

26

Household head does domestic work 3.136* 0.688 5.339*** 4.889* 0.813 10.622***

(1.538) (0.392) (2.676) (3.093) (0.637) (6.867)

Household head does manual work 0.000 2.90e+10*** 6.07e+09 0.000 4.27e+10*** 1.50e+10

(1502.476) (3.65e+10) . (3171.484) (5.63e+10) .

Household head does non-manual work 0.397* 0.153** 0.286** 0.497 0.155** 0.193**

(0.181) (0.099) (0.131) (0.255) (0.105) (0.108)

Household head is not in labour force 0.140 0.344 1.785 0.353 0.000 2.253

(0.162) (0.305) (1.209) (0.461) (0.000) (2.070)

Number of members in bad health/immobile 0.977 0.905* 1.093 1.146* 0.980 1.219**

(0.047) (0.042) (0.051) (0.078) (0.064) (0.077)

Child lives in Tigray 3.808*** 1.820 2.307* 1.442 0.948 4.013

(1.403) (0.632) (0.855) (1.206) (0.797) (3.172)

Child lives in Oromya 3.616*** 0.589* 2.199** 0.572 0.083*** 0.749

(0.992) (0.158) (0.581) (0.271) (0.048) (0.371)

Child lives in SNNPR 4.875*** 0.888 5.028*** 3.249 0.138 7.873**

(1.229) (0.217) (1.156) (2.880) (0.139) (6.192)

Child was monetary poor in 1999 2.225*** 1.066 2.029*** 2.159** 0.888 2.535***

(0.425) (0.206) (0.367) (0.525) (0.221) (0.572)


1.136*** 1.082* 1.108**

(0.041) (0.039) (0.039)

27


0.247 0.192 4.233

(0.382) (0.332) (5.389)


1.035 2.278 0.300*

(0.645) (1.690) (0.164)


3.875 0.633 17.442***

(3.535) (0.540) (11.622)


0.405 0.941 0.622

(0.483) (1.160) (0.548)

Community has high school

1.898 5.681 3.046

(1.469) (5.642) (2.137)


0.426 0.311* 0.420

(0.291) (0.149) (0.204)


968.000

P-value 0.000

0.000


0.182

BIC 3285.307 . . 2744.320


or does family work; child lives in Tigray. Only monetary poverty status in previous waves is taken into account; the inclusion of multidimensional poverty status did not

improve the fit of the model.


2009

28


AB A B AB A B


Child is female 0.877 0.914 0.692* 0.941 0.971 0.746

(0.180) (0.147) (0.121) (0.203) (0.172) (0.134)

Age of child 1.038 1.020 0.977 1.046 1.030 0.972

(0.037) (0.029) (0.030) (0.039) (0.033) (0.030)


(1.127) (0.923) (1.451) (0.868) (0.474) (1.393)

Age of household head 0.982 1.000 1.004 0.978* 0.991 0.999

(0.010) (0.008) (0.009) (0.011) (0.009) (0.009)

Household size 1.155** 0.937 1.273*** 1.222*** 0.952 1.317***

(0.062) (0.040) (0.058) (0.071) (0.045) (0.062)


(0.247) (0.342) (0.259) (0.320) (0.536) (0.277)

Household head is single/divorced/widowed 1.501 1.148 1.417 1.621 0.609 1.521

(0.884) (0.508) (0.784) (1.019) (0.305) (0.896)

Household head is polygamous 3.819*** 2.893*** 2.313*** 4.393*** 2.725*** 2.678***

(1.068) (0.687) (0.560) (1.312) (0.733) (0.687)

Household head has primary education or more 1.089 1.510* 1.229 1.020 1.538 1.156

(0.316) (0.312) (0.281) (0.318) (0.353) (0.276)

29

Household head does domestic work 0.853 0.272** 0.933 2.409 0.505 1.706

(0.556) (0.136) (0.545) (1.733) (0.286) (1.093)

Household head does manual work 1.741 1.124 1.032 2.364 1.528 1.066

(1.606) (0.716) (1.039) (2.284) (1.142) (1.101)

Household head does non-manual work 0.405 1.253 0.505 0.358 1.243 0.466

(0.250) (0.499) (0.218) (0.224) (0.529) (0.203)

Household head is not in labour force 1.108 1.176 0.447 1.668 1.666 0.627

(0.730) (0.588) (0.313) (1.217) (0.954) (0.459)

Number of members in bad health/immobile 1.161 1.008 0.922 1.156 1.101 0.873

(0.162) (0.110) (0.121) (0.165) (0.132) (0.122)

Child lives in Tigray 9.177*** 0.492 15.633*** 169.314*** 1.275 13.172**

(4.242) (0.254) (6.464) (252.085) (0.866) (10.417)

Child lives in Oromya 0.536 1.120 0.324** 46.788*** 1.735 0.548

(0.199) (0.226) (0.115) (53.946) (0.677) (0.384)

Child lives in SNNPR 2.570*** 0.521** 3.744*** 2013.148*** 0.387 11.330**

(0.732) (0.123) (0.898) (3470.081) (0.207) (9.614)

Child was monetary poor in 1999 1.652* 0.998 2.669*** 1.174 1.242 1.959**

(0.358) (0.174) (0.497) (0.286) (0.245) (0.403)

Child was monetary poor in 2004 2.356*** 1.165 1.966*** 2.157** 0.873 2.018***

(0.510) (0.208) (0.359) (0.513) (0.181) (0.398)

30


1.354*** 0.971 1.076

(0.092) (0.030) (0.046)


0.142*** 0.743 0.607

(0.074) (0.222) (0.210)


62.932*** 0.550* 3.090*

(52.054) (0.167) (1.438)


0.282** 0.133*** 0.959

(0.125) (0.056) (0.313)


0.202 0.296** 1.063

(0.204) (0.117) (0.645)


16.972** 2.458** 2.637

(16.014) (0.690) (1.484)


1113.000

P-value 0.000

0.000


0.216

BIC 3053.348 . . 2920.843


or does family work; child lives in Tigray. Only monetary poverty status in previous waves is taken into account; the inclusion of multidimensional poverty status did not

improve the fit of the model.

31

Table A4 Multinomial regression Vietnam 2004

2004

Multinomial model

Multinomial model rural

(with inclusion of community factors)

AB A B AB A B


Child is female 1.068 0.941 0.942 1.026 0.918 0.830

(0.234) (0.194) (0.183) (0.246) (0.215) (0.174)

Age of child 0.975 1.004 0.956 0.965 0.988 0.959

(0.034) (0.035) (0.030) (0.038) (0.039) (0.032)


(0.205) (0.260) (0.473) (0.386) (0.266) (0.707)


(0.014) (0.013) (0.013) (0.015) (0.015) (0.014)

Household head is single 30.959** 0.000 3.573 33.710** 0.000 3.716

(36.096) (0.000) (4.415) (44.819) (0.000) (5.196)

Household head is widowed 1.530 1.944 0.464 1.498 1.824 0.378

(0.927) (1.024) (0.252) (1.045) (1.192) (0.230)

Household head is divorced 22.851* 16.148* 2.505 1.68e+10*** 2.26e+10 1.51e+09***

(31.546) (21.585) (3.464) (2.30e+10) . (2.30e+09)

Household head is separated 4.200 0.000 2.454 7.419 0.000 3.217

32

(5.544) (0.000) (2.746) (10.117) (0.000) (3.619)

Household head has no education 1.969* 1.116 1.599 2.066* 1.128 1.573

(0.608) (0.359) (0.528) (0.717) (0.418) (0.575)

Household head has secondary education 0.210*** 0.806 0.491** 0.254*** 0.976 0.646

(0.061) (0.209) (0.115) (0.081) (0.285) (0.167)

Household head has post secondary

education 0.238* 0.366* 0.328* 0.353 0.652 0.586

(0.145) (0.174) (0.147) (0.243) (0.353) (0.292)

Household head is unemployed 2.321 6.039*** 0.645 2.424 10.376*** 0.789

(1.225) (2.707) (0.367) (1.641) (6.468) (0.538)

Household head is government staff 0.039** 1.046 0.000 0.026** 0.474 0.000

(0.042) (0.720) (0.000) (0.036) (0.450) (0.000)

Household head is a skilled professional 0.379** 1.185 0.751 0.660 1.165 0.917

(0.142) (0.304) (0.188) (0.257) (0.337) (0.248)

Child has other ethnicity 16.964*** 2.129 3.945*** 18.727*** 1.887 4.622***

(6.181) (0.918) (1.556) (7.645) (0.951) (1.963)

Children aged 5-11 present in household 0.612 0.595 1.140 0.562 0.539 1.181

(0.305) (0.254) (0.523) (0.320) (0.281) (0.612)

Children aged >11 present in household 0.949 0.474 1.440 0.791 0.467 1.282

(0.512) (0.229) (0.718) (0.488) (0.272) (0.715)

33

<25% of household members are children 0.146 0.496 0.133 0.026* 0.247 0.093*

(0.157) (0.311) (0.149) (0.041) (0.188) (0.111)

40-49% of household members are

children 2.551* 0.839 1.885 1.761 0.543 1.511

(1.050) (0.337) (0.719) (0.801) (0.256) (0.632)

>49% of household members are children 1.708 1.071 1.577 1.355 0.850 1.468

(0.637) (0.343) (0.538) (0.549) (0.301) (0.541)

Child lives in a rural area 3.026** 3.658*** 4.364***

(1.096) (1.147) (1.457)

Child lives in Red River Delta 0.381 0.801 1.075 0.328 0.650 1.068

(0.204) (0.347) (0.368) (0.190) (0.300) (0.401)

Child lives in North East 1.351 1.400 1.359 0.893 0.655 1.112

(0.668) (0.690) (0.554) (0.473) (0.363) (0.495)

Child lives in North West 15.038** 0.976 2.201 33.317* 2.622 7.948

(13.280) (1.266) (2.068) (53.991) (4.536) (13.326)

Child lives in North Central Coast 2.778* 0.671 1.744 2.583 0.416 1.788

(1.233) (0.344) (0.625) (1.259) (0.237) (0.717)

Child lives in Central Highlands 1.266 1.823 0.264* 1.171 1.525 0.300*

(0.643) (0.909) (0.148) (0.634) (0.800) (0.175)

34

Child lives in South East 0.314* 1.152 0.156*** 0.240** 0.920 0.198***

(0.152) (0.502) (0.073) (0.129) (0.435) (0.097)

Child lives in Mekong River Delta 1.695 4.403*** 0.384* 2.206 4.099** 0.377

(0.755) (1.819) (0.172) (1.057) (1.829) (0.193)

Child lives in area where a disaster

happened in 2004

0.780 0.759 1.132

(0.214) (0.203) (0.261)

Child lives in area where there are non-

farm employment opportunities

0.600 0.614 0.993

(0.158) (0.157) (0.236)

Number of observations 1068.000 858.000

P-value 0.000

0.000


0.287

BIC 2571.035 . . 2231.498

Note: Reference values are: child is male; household head is male; household head is married; household head has primary education; household head is unskilled worker;

child is of Kinh/Chinese ethnicity; children aged <5 are present in the household; % of children in household is 25-39%; child lives in South Central Coast; child is non-poor

in 2004; child is non-poor in 2006

35


2006

Multinomial model



AB A B AB A B


Child is female 1.561 1.657* 0.994 1.630 1.727* 1.064

(0.425) (0.370) (0.237) (0.480) (0.433) (0.279)

Age of child 0.931 0.944 0.918* 0.954 0.953 0.928

(0.042) (0.037) (0.037) (0.047) (0.042) (0.041)


(0.551) (0.328) (0.661) (0.972) (0.653) (1.173)


(0.017) (0.014) (0.015) (0.019) (0.017) (0.018)

Household head is single 1.06e+11*** 1.27e+10*** 4.00e+09 1.40e+11*** 5.41e+09*** 4.30e+09

(1.37e+11) (1.77e+10) . (1.98e+11) (9.25e+09) .

Household head is widowed 0.179* 0.742 0.179** 0.261 0.424 0.119**

(0.146) (0.453) (0.119) (0.247) (0.350) (0.096)

Household head is divorced 1.378 0.000 0.000 2.090 0.000 0.000

(1.927) (0.000) (0.000) (3.311) (0.000) (0.000)

Household head is separated 0.000 0.000 17.984 0.000 0.000 35.530

(0.000) (0.000) (30.467) (0.000) (0.000) (83.556)

36

Household head has no education 0.967 1.242 1.048 0.666 1.141 0.574

(0.356) (0.399) (0.365) (0.275) (0.417) (0.227)

Household head has secondary education 0.481* 0.989 0.624 0.487 0.705 0.664

(0.177) (0.290) (0.183) (0.192) (0.232) (0.218)

Household head has post secondary

education 0.237 0.429 0.000 0.152 0.168* 0.000

(0.289) (0.250) (0.000) (0.200) (0.132) (0.000)

Household head is unemployed 8.319** 4.404** 1.692 22.945*** 11.038*** 4.113

(5.846) (2.159) (1.054) (19.600) (7.405) (3.439)

Household head is government staff 2.495 0.000 0.680 1.622 0.000 0.378

(3.522) (0.000) (0.894) (2.469) (0.000) (0.525)

Household head is a skilled professional 0.312* 1.117 0.645 0.281* 1.507 0.489*

(0.168) (0.311) (0.209) (0.172) (0.468) (0.178)

Child has other ethnicity 24.195*** 4.434*** 3.912** 47.880*** 4.586** 5.505***

(10.942) (1.923) (1.728) (25.852) (2.350) (2.706)


(1.129) (0.295) (0.468) (1.129) (0.204) (0.559)


(1.204) (0.353) (0.431) (0.992) (0.317) (0.436)

37

<25% of household members are children 0.482 0.907 0.919 0.427 0.686 0.931

(0.500) (0.565) (0.706) (0.483) (0.545) (0.822)

40-49% of household members are

children 2.579 1.079 2.002 2.548 0.997 1.392

(1.355) (0.454) (0.930) (1.445) (0.496) (0.741)

>49% of household members are children 1.718 0.886 1.545 2.045 1.051 1.814

(0.768) (0.294) (0.622) (0.978) (0.393) (0.791)

Child lives in a rural area 2.141 1.525 1.127

(1.199) (0.507) (0.441)


(1.532) (0.377) (0.354) (2.172) (0.548) (0.466)


(0.825) (0.368) (0.309) (0.876) (0.521) (0.311)

Child lives in North West 3.332 1.851 1.248 4.520 1.960 1.345

(3.019) (1.603) (1.020) (4.879) (2.096) (1.346)

Child lives in North Central Coast 5.717* 1.545 2.320 8.446** 1.707 2.137

(4.019) (0.834) (1.040) (6.550) (1.077) (1.090)

Child lives in Central Highlands 7.770** 1.952 4.430** 7.956* 1.411 3.831*

(5.839) (1.168) (2.492) (6.477) (0.992) (2.337)

Child lives in South East 6.011* 1.881 1.445 3.459 2.508 1.757

38

(4.425) (0.979) (0.743) (2.848) (1.478) (0.997)

Child lives in Mekong River Delta 7.010** 4.076** 0.307 11.770** 5.758** 0.070*

(4.850) (1.929) (0.190) (8.859) (3.141) (0.080)

Child is multidimensionally + monetary 368.042*** 10.195*** 29.214*** 277.592*** 12.370*** 26.417***

poor in 2004 (304.626) (3.781) (11.537) (230.377) (5.104) (11.703)

Child is only multidimensionally poor in 28.920*** 4.663*** 0.922 19.140*** 4.477*** 0.817

2004 (24.216) (1.334) (0.543) (16.067) (1.448) (0.506)

Child is monetary poor in 2004 83.886*** 2.912** 24.466*** 66.823*** 2.629* 20.378***

(70.795) (1.105) (8.473) (55.829) (1.100) (7.516)

Child lives in area where a disaster happened in 2004

1.703 1.265 2.399**

(0.573) (0.353) (0.727)

Child lives in area where there are non-farm employment opportunities

1.027 0.708 0.681

(0.362) (0.209) (0.198)

Child lives in area with ECD centre

0.784 1.513 1.066

(0.252) (0.414) (0.306)


P-value 0.000

0.000


0.417

BIC 2169.349 . . 1955.165


child is of Kinh/Chinese ethnicity; children aged <5 are present in the household; % of children in household is 25-39%; child lives in South Central Coast; child is non-poor in

39

2004; child is non-poor in 2006


2008

Multinomial model



AB A B AB A B


Child is female 1.116 1.112 1.607* 1.185 1.150 1.609

(0.319) (0.272) (0.388) (0.377) (0.320) (0.424)

Age of child 1.028 1.153** 0.994 1.043 1.214*** 1.009

(0.048) (0.053) (0.041) (0.055) (0.066) (0.046)


(0.694) (0.394) (0.587) (0.777) (0.495) (0.696)

Age of household head 0.970 0.976 1.005 0.949* 0.980 1.003

(0.019) (0.016) (0.016) (0.020) (0.018) (0.018)

Household head is single 5.866 0.000 0.000 58.435* 0.000 0.000

(8.889) (0.000) (0.000) (110.464) (0.000) (0.000)

Household head is widowed 0.180 0.432 0.674 0.260 0.399 0.666

(0.170) (0.284) (0.443) (0.299) (0.336) (0.485)

Household head is divorced 135.182 0.000 0.000 6.10e+17 6.757 0.493

(388.833) (0.000) (0.000) (3.60e+25) (8.37e+08) (6.45e+07)

40

Household head is separated 14.201 9.462* 2.824 84.979* 13.918* 1.656

(21.590) (9.531) (3.407) (149.166) (17.190) (2.412)

Household head has no education 2.430* 1.739 2.135* 3.358** 1.962 2.797**

(0.877) (0.607) (0.758) (1.383) (0.773) (1.088)

Household head has secondary education 0.230*** 1.054 0.851 0.274** 1.257 0.977

(0.100) (0.337) (0.252) (0.127) (0.448) (0.310)

Household head has post secondary education 0.000 1.273 0.433 0.000 1.135 0.633

(0.000) (0.673) (0.299) (0.000) (0.697) (0.452)

Household head is unemployed 2.990 5.687*** 0.426 8.664** 12.540*** 0.211

(2.015) (2.830) (0.259) (7.011) (8.019) (0.171)

Household head is government staff 37.156** 0.000 0.743 68.094* 0.000 0.710

(50.794) (0.000) (0.804) (113.679) (0.000) (0.874)

Household head is a skilled professional 0.372 0.430* 0.373* 0.301* 0.451* 0.554

(0.198) (0.154) (0.147) (0.182) (0.176) (0.226)

Child has other ethnicity 4.454*** 2.892** 2.687* 6.727*** 3.237** 2.234

(1.929) (1.121) (1.080) (3.392) (1.460) (0.987)


(1.359) (1.076) (0.687) (3.752) (1.222) (0.418)


(2.589) (0.771) (0.829) (6.905) (0.594) (0.461)

41

<25% of household members are children 0.684 1.630 1.663 0.801 1.933 1.610

(0.543) (0.849) (0.887) (0.689) (1.125) (0.946)

40-49% of household members are children 0.965 1.759 1.568 1.036 2.250 2.190

(0.516) (0.779) (0.702) (0.611) (1.157) (1.061)

>49% of household members are children 1.130 1.928 2.215* 1.266 2.421* 2.566*

(0.479) (0.698) (0.797) (0.604) (1.013) (1.011)

Child lives in a rural area 1.628 1.862 1.345

(0.863) (0.680) (0.573)


(1.078) (0.743) (0.590) (0.338) (0.542) (0.313)


(1.955) (1.650) (0.416) (2.672) (1.217) (0.232)

Child lives in North West 6.173* 0.569 0.965 9.337* 0.614 0.504

(5.459) (0.623) (0.685) (9.126) (0.714) (0.395)

Child lives in North Central Coast 7.052* 2.245 1.942 10.364** 1.657 0.835

(5.398) (1.628) (0.878) (8.809) (1.289) (0.419)

Child lives in Central Highlands 4.412 4.323* 1.324 6.488* 4.050 0.661

(3.612) (3.176) (0.752) (5.861) (3.135) (0.417)

Child lives in South East 3.122 3.302 0.670 2.214 2.753 0.726

42

(2.515) (2.274) (0.367) (2.040) (1.991) (0.423)

Child lives in Mekong River Delta 8.897** 7.364** 0.236* 28.161*** 8.663** 0.374

(6.922) (4.929) (0.174) (25.196) (5.976) (0.289)

Child is multidimensionally + monetary poor in 2004 20.097*** 3.891*** 6.808*** 16.488*** 3.088* 6.998***

(13.112) (1.583) (2.906) (11.918) (1.438) (3.329)

Child is only multidimensionally poor in 2004 4.496* 2.722** 1.801 2.552 1.973 1.232

(3.115) (0.886) (0.978) (2.029) (0.731) (0.765)

Child is only monetary poor in 2004 7.325** 0.886 6.150*** 7.392** 0.655 6.555***

(4.979) (0.440) (2.306) (5.515) (0.360) (2.675)

Child is multidimensionally + monetary poor in 2006 15.627*** 3.640** 9.380*** 15.395*** 4.018** 11.306***

(7.970) (1.708) (4.155) (8.896) (2.095) (5.467)

Child is only multidimensionally poor in 2006 2.987* 4.911*** 1.413 2.770 5.677*** 1.283

(1.543) (1.522) (0.667) (1.637) (2.003) (0.664)

Child is only monetary poor in 2006 2.554 0.971 3.885*** 2.964 0.913 4.682***

(1.265) (0.480) (1.264) (1.677) (0.518) (1.697)

Child lives in area where a disaster happened in 2004

2.843** 1.461 0.849

(1.051) (0.456) (0.241)

Child lives in area where there are non-farm employment

opportunities

1.261 1.181 0.548*

(0.471) (0.375) (0.157)

43

Child lives in area with ECD centre

0.791 0.915 2.168**

(0.273) (0.267) (0.647)


P-value 0.000

0.000


0.454

BIC 2089.016 . . 1879.155 . .


child is of Kinh/Chinese ethnicity; children aged <5 are present in the household; % of children in household is 25-39%; child lives in South Central Coast; child is non-poor in

2004; child is non-poor in 2006

i These data have been made available by the Economics Department, Addis Ababa University, the Centre for the Study of African Economies, University of Oxford and the

International Food Policy Research Institute. Funding for data collection was provided by the Economic and Social Research Council (ESRC), the Swedish International

Development Agency (SIDA) and the United States Agency for International Development (USAID); the preparation of the public release version of these data was

supported, in part, by the World Bank. AAU, CSAE, IFPRI, ESRC, SIDA, USAID and the World Bank are not responsible for any errors in these data or for their use or

interpretation.

ii Data has been made available by the Government Statistical Office (GSO) in Hanoi, Vietnam with support from UNICEF Vietnam. iii The inclusion of community indicators in the rural model has little explanatory power due to lack of variation; primary schools are available in all areas and the inclusion

of road accessible to auto and secondary school does not improve fit of the model. iv A more elaborate discussion of the measures for monetary and multidimensional child poverty and empirical findings can be found in Roelen (mimeo). v A more elaborate discussion of the comparison between quantitative and qualitative indicators used for reflecting multidimensional child poverty and child wellbeing

respectively can be found in Roelen (forthcoming).

Poor children living in rich households: A Blurred Picture ...€¦ · Poor children living in rich households: A Blurred Picture or Hidden Realities? Keetie Roelen [email protected]

Documents