Capturing Nested Spheres of Poverty A Model for Multidimensional Poverty Analysis and Monitoring

1 1. Poverty is more than low income 1 2. Project context and methods 2 3. NESP – Multidimensional spheres of poverty 2 3.1. Nested Spheres of Poverty (NESP): A multidimensional


Poverty is more than low income
Over recent decades, poverty concepts have profoundly changed from the mere consideration of income or consumption, to definitions that include multiple dimensions of deprivation and wellbeing. 1Today, leading development organisations apply poverty definitions that comprise aspects like selfdetermined lifestyles, choice, assets, capabilities, social inclusion, inequality, human rights, entitlement, vulnerability, empowerment and subjective wellbeing.The new poverty concepts have found their way into the UN Human Development Report (UNDP 2005), the World Bank's World Development Report (2000/01; see also World Bank 2002) and into other, more qualitative poverty studies published by the World Bank. 2 While they are more sophisticated, these new concepts have been difficult to quantify. 3herefore, international agencies such as the World Bank and UNDP, but also national governments, still favour money-metric poverty lines like the famous $1 ($2) per day, 4 or unfilfilment of basic needs (see Box 1).
Although there are close correlations among economic growth, income, subjective wellbeing and non-material poverty aspects, 5 the gap between a multidimensional understanding of poverty and the extremely reductionist one-dimensional indicators is disturbing.The emphasis on quantitative poverty measurement based on economic or basic needs parameters Box 1: Who is poor?
At the global scale, the World Bank and the UN define extreme economic poverty as having an income of less than $1 per day in purchasing power parity (PPP).The Human Development Index (HDI) of UNDP measures three fields: longevity, knowledge and decent standard of living.Longevity is measured by the percentage of people who die before age 40; knowledge is measured by adult literacy combined with the gross enrolment ratio for primary, secondary and tertiary schools; and standard of living is measured by real GDP/capita.The Human Poverty Index (HPI1) uses the same fields, but measures standards of living in terms of access to safe water and healthcare, and by the percentage of underweight children younger than 5 years (UNDP 2005).
In Indonesia, various measures are used for assessing poverty and wellbeing, mainly assessing the satisfaction of basic needs.More detailed descriptions are given by Maksum  (2004) and Cahyat (2004).neglects many other dimensions of wellbeing.This could lead to the repackaging of old, simplistic poverty alleviation strategies that rely solely on macroeconomic growth, income generation, or infrastructural improvements such as building roads, schools and health posts.In addition, these figures often look far more precise than they really are due to the widespread difficulties in collecting reliable and accurate poverty data in many poor countries, and the need for extrapolation based on limited data. 6Hence, the challenge is to find a practical compromise between a comprehensive poverty concept and a model based on quantifiable poverty indicators.In addition, the concept has to be simple enough that it appeals to decision makers who need answers to the following questions: 7  Who are the poor? How poor are they? Where do they live? Why are they poor? What can be done? What are the changes over time?
Many local governments, which face new responsibilities for poverty alleviation 8 under recent decentralisation, have few good answers to these questions.They typically lack the resources and capacity to answer them, and they often lack appropriate poverty concepts and reliable data. 9 this paper, we suggest a multidimensional poverty concept, building on the capability approach of Amartya Sen (e.g.1993, 1997, 1999), the sustainable livelihood approach (SLA; e.g.Chambers and Conway 1991; Scoones 1998;  Baumann 2000; Solesbury and Daniels 2002), 10 and the World Bank's qualitative approaches mentioned above.Although our practical work focused on forest-dependent poor, our model is sufficiently general to be used in many other rural settings. 11 Using this model as a basis, we introduce a new quantitative tool to measure and monitor poverty at the household level, to provide some answers to the questions above.Examples from our work in Kutai Barat (Indonesia) are shown to illustrate the practical use of both the model and the monitoring tool.We conclude by showing how multidimensional poverty monitoring can lead to better planning and, thus, to more effective poverty alleviation.

Project context and methods
The CIFOR-BMZ project 'Making local government more responsive to the poor: Developing indicators and tools to support sustainable livelihood development under decentralisation' worked with local governments in forested areas of Indonesia (Kutai Barat and Malinau) and Bolivia (Pando) from 2003 until 2006.It applied a participatory learning approach for improving the understanding of trends in local poverty and wellbeing, and for developing local monitoring and planning tools to strengthen the local governments' poverty alleviation efforts.
The methodology included community baseline surveys, focus group discussions, indepth anthropological case studies on local poverty concepts and socioeconomic change, and multistakeholder workshops (see Bullinger  2006; Haug in prep.).The methodology used for developing local poverty indicators and the monitoring system is further explained in Section 4.More details are given in Cahyat et  al. (2007).Data used for producing the charts and poverty maps of Section 4 were collected through standardised household interviews during the poverty monitoring of Kutai Barat.The monitoring survey covered all 223 villages of the district with a sample size between 100% (villages with 20 or fewer households) to a minimum of 33% (villages with more than 60 households; 20 households were interviewed in villages with between 21 and 60 households).The calculation of indices is explained in Section 3.

Nested Spheres of Poverty (NESP): A multidimensional model of poverty
Poverty is a time-dependent condition.It can be temporary (acute or short-term poverty) or persistent (chronic poverty).It can be a permanent threat for those living just above the poverty line and it can be a trap for those who cannot get out of it. 12Poverty is a lack of many things.It may be caused by insufficient income, or unsatisfied basic needs, such as health, education or housing.But poverty is also highly subjective and may be caused by feelings of deprivation, vulnerability, exclusion, shame, pain, and other forms of ill-being. 13In addition, poverty is a result of a lack of means, capabilities, freedom and options for a better future.
Both unsatisfied basic needs and means to address this deprivation explain why poverty is often a self-reinforcing problem.We propose to use these two conditions in a single concept of poverty.If there is no enabling environment for getting out of poverty, people get trapped in chronic poverty.Thus, poverty is not only 'having no fish' , it is also 'not knowing how to fish' , 'not knowing where to fish' , 'not having a rod and line' or 'lacking the right to fish' .Plus, in many cases, there are 'no fish' because a lake has been polluted or has dried up.Ultimately, however, it is the subjective feeling of 'being hungry because of not having eaten fish' that is the very essence of poverty.
In order to capture all these notions and attributes of poverty, we conceptualised our poverty model in a nested shape (Figure 1).Interacting with all four spheres are infrastructure and services provided by government institutions, the private sector, development projects or civil society organisations.
The dynamics of poverty are reflected by the different layers of NESP.SWB has a very time-specific nature.As discussed above, SWB often fluctuates due to many influences.The analysis of our field data 20 showed a moderate correlation of SWB with the combined core aspects.This correlation also underlines the general importance of basic needs, as well as the high priority that local people in our study site ascribed to these attributes.Hence, improvement of the core generally leads to improved SWB.By the same token, a poor core usually means low SWB.
On a longer time scale, both core and SWB are influenced by the context. 21For instance, knowledge increases as a result of improved education, health problems grow because of environmental pollution, SWB declines due to social conflict.
However, some of the context spheres may be means and ends at the same time.For instance, to have political freedom to participate in decision making may be important to improve core conditions, but it can also be regarded as an essential need and thus linked to subjective wellbeing. 22e dichotomy of core and context helps distinguish between the conditions of the poor and the quality of the enabling environment that directly affects future developments.In the case of poverty monitoring, each of the spheres in core and context requires different types of information and different responses from local governments.While the information about the core helps to measure impacts on individual living conditions and guides how to address shortcomings (for instance, through humanitarian aid), information about the context helps in the determination of the prospects for achieving a higher standard of wellbeing and can guide strategic support to local development processes.

Box 2. Linking nEsp to sLA
The five capitals that are included in the Sustainable Livelihood Approach (Chambers  and Conway 1991; Scoones 1998; Baumann  2000; Solesbury and Daniels 2002) are included in the NESP poverty model, but they are spread over both clusters.While some assets, such as knowledge and health (human capital) are found in the core, others, such as natural and social capital, are in the context of our model.Table 1 shows where the five SLA capitals appear in our NESP model.The NESP model, especially with its context spheres, offers clear links to the various government sectors. 23The examples shown in the next section illustrate how the poverty spheres can be related to the respective sectors.

NESP indicators and indices
The NESP model offers a comprehensive basis for multidimensional poverty and wellbeing assessments.In order to convert it into a locally meaningful and specific concept, however, one needs to represent the model's spheres with a set of local poverty indicators.These indicators should comply with the following minimum set of 'SMART' criteria: 24 Simple means that an indicator is easy to understand and practical to use.
Measurable means that the indicator can be reasonably quantified and assessed.It also means that the indicator can be measured by locally available means (e.g.no expensive scientific methodology is needed).
Adapted means that the indicator is location specific, i.e. it should be relevant in its sociocultural and natural-geographic context.With its broad basis, the model might be considered to be an ill-being/wellbeing model rather than a poverty model.However, we understand the two terms in a reciprocal way along a gradient from ill-being (in our definition: poverty) to high wellbeing (i.e.prosperity), and we use both concepts in an interchangeable manner.Although this definition is not conventional, it is useful when trying to accommodate different national concepts and helpful when assessing and analysing various dimensions of poverty.Furthermore, 'poverty' often has a negative connotation of passivity, incompetence or backwardness, and the use of the term can be offensive or demeaning.The term 'wellbeing' allows discussion of poverty in more positive terms.Hence, 'poverty' should be read as 'lack of wellbeing' and 'wellbeing' as 'reduced poverty'.In our practical work, we started with a comprehensive perspective and later refined this angle according to local perceptions of poverty and wellbeing, and the practical demand of the local government.This approach was welcomed by the local government as the various dimensions of our poverty model could easily be linked to the respective government sectors.
Robust means that the indicator value ideally does not depend on who the assessor is or when the assessment is conducted (unless seasonality is a factor that needs to be captured).Robustness makes an indicator credible and acceptable to policy makers.
Timely means that the indicator changes on the same time scale as the poverty aspects.This facilitates adequate policy responses to the monitoring findings.E.g. if an indicator lags too far behind, impacts cannot be linked to policy action.
In our example from Indonesia, we used a set of indicators for each of the model's spheres, and each indicator received two or three possible values.For instance, in the case we assigned three conditions to a particular indicator, score The indicators of each sphere can be aggregated into normalised indices, one for each of the nine spheres, with a value range between 0 and 1.The resulting nine NESP indices can be analysed and presented separately.However, if desired, they can also be combined into composite indices for the core and the context. 27cally collected NESP data reflect the local understanding of poverty and are, therefore, relative measures of poverty.However, this does not exclude comparison across sites, as long as one is aware that relative poverty rates are being compared.In areas with similar living conditions, such comparisons seem reasonable. 28e following section illustrates how NESP indicators and indices are created and used for local poverty monitoring in an example from Indonesia.

Local adaptation of NESP
The government of Kutai Barat district was not satisfied with the national poverty assessments, as they were perceived as unreliable and biased towards conditions on Java.So, they asked for help in developing a locally specific poverty monitoring system to allow the districts to better target poverty alleviation.
First, we developed the NESP model.We filled the general model with locally specific content based on an understanding of poverty that made sense to all project partners (local government, NGOs, communities, CIFOR researchers).Sixty focus group discussions, various workshops and in-depth community studies were used to compile long-lists of possible indicators that described the nine spheres in a locally meaningful manner. 29After two rounds of field trials in 40 villages a final short-list was prepared 30 and used to measure poverty in all 21 subdistricts and 223 villages of the district.
The poverty indicators used in the 2006 Kutai Barat poverty monitoring survey are compiled in Table 2.In addition, the survey collected information on household structure, the use of forest products, income sources and on the perceptions of selected local government programmes for analytical purposes.

NESP-based visualisation of data
The indicators listed in the right column of Table 2 can be aggregated into nine normalised indices corresponding with the nine spheres in the left column. 31In order to allow quick comparisons of different villages or subdistricts, we applied a simple colour code (Figure 3).  4.
The NESP model allows easy comparison of different units (households, villages, subdistricts, etc.) at a glance.If a more quantitative comparison is needed, the real index scores can be compared in bar diagrams (e.g. Figure 5). 33The two ways of representing poverty data have different strengths and weaknesses.While the NESP model (Figure 4) gives a quick overview of the overall poverty situation of a village (or household or subdistrict, etc.), including critical sectors and possible trade-offs, the bar diagrams (Figure 5) provide a more differentiated picture that also allows comparing indices of the same colour code in a more quantitative way.This is especially helpful for index values which are close to the boundary between two colours and might therefore disguise significant differences between two villages.
Both versions instantly show which sectors are in a critical condition.In the example of Figure 4, Village A lacks education and healthcare and has problems in the economic sphere, Village B lacks education, while Village C clearly has environmental problems, and Village D suffers from inadequate infrastructure and government services.All these red spheres are signs of alert for the respective government agencies which then need to follow up with a more in-depth analysis of the underlying causes.
Another way of illustrating poverty monitoring results is to use data lists with colour codes (as in Table 3).
Poverty data can also be organised in thematic maps (see Figures 7 and 8).

Another example for applying NESP: Poverty maps
Poverty maps are a powerful tool to visualise poverty patterns.They show where poverty hotspots are and which poverty spheres are critical in which area.This helps answering the question 'Where are the poor?'However, the patterns revealed by poverty maps do not automatically provide answers to the problem 'Why are they poor' , but only show correlations between different aspects of poverty. 34onetheless, correlations are useful as they generally make a good starting point to look for causal links (see Section 6).
In order to demonstrate the illustrative power of poverty maps, we present a few examples from the poverty and wellbeing monitoring survey in Kutai Barat. 35gure 6 shows an overview map of the study area on the island of Kalimantan (Borneo).
The examples in Figures 7 and 8 are based on the NESP approach applied in the poverty monitoring survey of Kutai Barat, February-March 2006.

Using NESP for more effective poverty alleviation
NESP as a multidimensional local poverty monitoring system provides comprehensive and relevant information important for district and subdistrict planning.The core and context information can help planning agencies to:  Alert the local government on poverty hotspots  Alert responsible government sectors  Identify needs for addressing acute poverty (basic needs)  Anticipate future impoverishment caused by an unfavourable context  Identify strategic entry points to reduce chronic poverty  Identify strategic entry points to strengthen the enabling environment (context)  Identify priority areas for regionally more balanced development  Identify which poverty alleviation measures worked and which did not  Track changes of poverty data over time.
Through these actions, the local government can get answers to our initial questions.

Who are the poor?
Depending on the survey resolution, we can identify poor households, poor villages, and poor subdistricts, or geographic regions with high poverty. 36Poverty lists and maps help local governments to identify poverty hotspots and allocate their aid on the basis of clear demand.Additional information on household structure shows whether poverty is especially related to ethnicity or certain clusters of households (e.g.those with only one adult, with disabled family members).

How poor are they?
The use of poverty indices allows quantification of the nine poverty spheres, which helps in allocating government support and aid.Analysing core and context separately can shed some light on the differentiation of acute

Where do they live?
The spatial information shown in poverty maps (Figures 7 and 8) helps the local government to identify hotspot areas.These areas can be poor for many different reasons.However, their identification is the first step to addressing the problem.Overlaying spatial information, such as infrastructure maps and poverty maps, can help to identify patterns of poverty.

Why are they poor?
Many poverty causes, such as natural hazards, fluctuating world market prices, or national political and economic crises, are beyond the local government's control.The analysis of multidimensional data sets and poverty maps can generate hypotheses and ideas on poverty causes.However, the visualisation of poverty data is no substitute for in-depth analysis.Therefore, any index that shows a critical value is only a sign of alert that must trigger some serious discussion or more detailed studies about the underlying causes.A basic causal analysis that aims to improve planning can be conducted at the village level (see Figure 9) with additional input from technical agencies, researchers or civil society organisations.

What can be done?
NESP flags critical conditions and helps identify priority areas and sectors.This can make development planning far more effective.In the case of Kutai Barat, the monitoring approach will be linked to the existing annual planning system (as shown in Figure 9). 37e monitoring results will be distributed to subdistricts and villages, where the findings are checked for plausibility by comparing rankings of the NESP spheres at village level.Critical spheres then become priorities for the annual village planning sessions.For instance, if the health condition is critical in Village A and the villagers agree on this fact, it becomes a top priority for planning the development activities of next year.As the monitoring system does not explain why health is critical, the village assembly conducts a basic causal analysis and elaborates suitable measures which are then proposed to the subdistrict level.Here the proposals are collected from all villages and discussed by the subdistrict government and related technical agencies.At the subdistrict planning session-where the villages are also represented-an annual development plan is prepared and submitted to the district government.In addition, information can be requested from other government agencies, or from researchers and civil society organisations familiar with the area.If these steps are conducted properly, a revised poverty alleviation strategy should reflect the spatial and sectoral priorities that emerge from monitoring.Such a strategy would need to address basic needs, as well as contextual constraints and opportunities in order to facilitate self-driven poverty alleviation.

What are the changes over time?
The poverty monitoring system we suggest above allows governments to track changes of poverty over time.NESP is not a static concept.In particular, the context spheres can be extremely dynamic.Capturing changes of core poverty and the enabling environment requires regular repetition of monitoring surveys.Depending on available resources, annual to biannual cycles will provide decision makers with sufficiently updated information.Some figures and lessons learned from our case study in Kutai Barat are listed in Box 5.

Village Planning Village Planning Technical Agencies Technical Agencies
Future Development Future Development Figure 9. Intended monitoring and planning cycle in Kutai Barat

Conclusion
In summary, the NESP model is a practical tool for measuring and monitoring poverty or wellbeing in its many dimensions.The model is dynamic and reaches well beyond basic needs and consumption expenditures.It covers qualitative aspects (such as subjective wellbeing), core spheres of poverty (like health, minimal wealth and knowledge and its related dimensions), as well as contextual spheres (including natural, economic, social, and political environment, and infrastructure and services).
Each of the spheres can be assessed by using a set of indicators which can be combined into sectoral indices (see Figure 7 and Table 3).If so desired, the indicators can also be converted into composite indices of the core and the context (see Figure 7), or into an overall aggregate.Teachers were used as village assessors.The relative ranking of villages within selected subdistricts matched the judgment of long-term experts and local informants far better than any other official poverty report.Several quality checks are built into the system.For detailed information see Cahyat et al. (2007).
Time and costs can be significantly reduced if the approach only needs to be locally adapted (e.g. in other rural areas of Indonesia).
Government support is crucial.As a monitoring and planning tool, the NESP approach has been warmly welcomed by district planners.The monitoring system was strongly supported by the local government between 2003 and 2005, covering the full costs of implementation.However, in early 2006 the government changed and the institutionalisation of the approach got temporarily halted.
In the meantime, the regional planning agency (Bappeda) together with GTZ and CIFOR prepared an action plan for the future application and institutionalisation of the approach (see Figure 9).
Other organisations also showed interest in applying the approach in other districts of Indonesia.
Experience from its use in our Indonesian sites demonstrates that the model can be easily adapted to local conditions.In contrast to national poverty indicators, the NESP indicators reflect local peculiarities, such as cultural preferences of lifestyles, seasonality or local modes of subsistence.However, the approach can also be scaled up to larger areas because of the design of the indicator scores.However, attention has to be given to the temptation of mistaking the approach for an 'automatic poverty alleviation programme' where decision makers only sits in their chairs, look at colourful maps and take decisions to address the poverty problems in their regency.Like any other monitoring system, NESP can only indicate and hint where and of what kind the problems are.With its multisphere setup, the model visualises how complex poverty is and that solutions also mean trade-offs.Understanding the underlying causes of poverty and finding balanced ways for poverty alleviation remains the creative task of concerned people.

Endnotes
1 For comprehensive summaries of this trend see Kanbur and Squire (1999) or Angelsen and Wunder  (2003). 2 E.g. the 'Voices of the Poor' study (Narayan et al.  2000a, b; Narayan and Petesch 2002).3 See Sumner (2004, p. 14), Kanbur and Squire (1999,  pp.4-5), Angelsen and Wunder (2003, pp.4-7, 10-11).4 This poverty line was introduced in the 1990 Word Development Report of the World Bank.For a concise summary of the main points of critique, see Kanbur and  Squire (1999), Reddy and Pogge (2005)   5 See Angelsen and Wunder (2003, p. 10).6  This argument is also one of the main points of critique brought forward by Reddy and Pogge (2005)  against the World Bank's poverty reports.While admitting some methodological shortcomings, Ravallion (undated) emphasises the international consensus on the $1/$2 poverty lines.7 Some of these questions are also asked by Reddy  and Pogge (2005, p. 4).8 Poverty alleviation is understood here in the sense of Angelsen and Wunder (2003, p. 2) as encompassing poverty reduction as well as poverty prevention.9 E.g. data from the Indonesian Central Statistics Agency (BPS) contradicted the figures collected in parallel by the Indonesian Family Planning Coordination Agency (BKKBN), which left the local governments in deep confusion about the poverty situation in their districts.On the other hand, composite indices like the HDI and one-dimensional economic measures such as the $1 poverty line may allow cross-country comparisons, but they do not provide a solid basis for day-to-day poverty alleviation efforts especially by local decision makers.

Gönner et al.
comprising decent housing condition, appropriate clothing, some basic equipment, such as TVs, a bicycle or motorbike, etc.The definition of wealth depends on local standards.Knowledge includes both formal education and informal or traditional knowledge.
18 Income was also frequently mentioned, but mainly as a tool for improving health, wealth and knowledge.19 The segregation into these four spheres is somewhat arbitrary.However, it was the most practical way to aggregate the poverty attributes compiled in our project workshops in terms of redundancy and comprehensiveness.20 Spearman's rank correlation: r = 0.528, based on 10 431 household interviews covering 223 villages in Kutai Barat (Indonesia).21 However, this is not a unidirectional causality; e.g.improved knowledge can influence the context through different resource use, new conflict resolution mechanisms, etc.
23 Although there is a wide discussion about crosssectoral governance approaches, practice lags far behind.Weak institutional coordination, lack of communication about and competition over budgets, resources and power are frequent constraints, and sectoral setups generally prevail.24 This SMART system used here differs from the more conventional one, where S stands for 'specific', M for 'measurable', A for 'achievable and attributable', R for 'relevant and realistic' and T for 'time-bound, timely, trackable and targeted' (e.g.GEF undated).
25 Statistical findings from our field tests suggest that three to five indicators per poverty sphere index are enough to describe the nine poverty spheres.We tested a long-list of preliminary indicators and conducted correlation tests for different subsets to identify a representative group of indicators for each poverty sphere.
! All of these questions refer to conditions over the last 12 months, and apply only to your household or village.Please provide only one answer for each question.
! Household members include only those living together in one house or those being supported by the household.

Health and Nutrition
Var 12 Have there been any shortages of food for more than 1 month during the past 12 months?

Figure 2 .
Figure 2. SMART criteria for poverty indicators

Figure 5 .Figure 4 .
Figure5.Bar diagrams of poverty sphere scores for the same 4 villages as in Figure4SWB subjective wellbeing; H health, W wealth, K knowledge, N natural sphere, E economic sphere, S social sphere, P political sphere, I & S infrastructure and services.

Figure 6 .Figure 7 .Figure 7 Figure 8 .Figure 8
Figure 6.Overview map of Kutai Barat and Malinau(Andrianto 2006) your household have access to clean drinking water (not necessarily from PDAM)the event of sickness, do members of your household always receive modern medical treatment from a doctor, nurse, midwife, or traditional care from a shaman or healer?sick during the last 12 monthsMaterial WealthVar 15 (PLEASE ASSESS FOR YOURSELF, DO NOT ASK) What is the quality of the respondent's PLEASE ASSESS FOR YOURSELF, DO NOT ASK) Does the household own a satellite dish or a is the highest level of education among the adult members of your household (including the household head)SLTA) or higher or passed Packet C Var 19 Are there any children aged between 7 and 16 years old in your household attending school (children funded by your household)there any household members with additional off-farm qualifications (e.g.healing, making handicrafts, carpentry, driving)?

Table 1 .
Location of SLA capitals in NESP model The colours depend on the composition of the respective indices.32

Table 2 .
Aspects of the three wellbeing clusters: SWB, core and context

Table 3 .
Data list with colour code (village names have been changed) Notes: Colour codes as per Figure

Box 5. somE figurEs And LEssons LEArnEd from kutAi BArAt
below $1 a day and 2.7 billion lived on less than $2 a day.See http://web.worldbank.org.
and Sumner  (2007).The global poverty lines are set at $1 per day (extreme economic poverty) and $2 per day (more precisely $1.08 and $2.15 in 1993 PPP).It has been estimated that in 2001, some 1.1 billion people had consumption levels