Applying the small area estimation approach to multi-dimensional child poverty in Nigeria.


Poster Session

Poster no.16

Saleem O. Falowo (GIS Division, Monitoring and Evaluation Department, Ministry of Budget and National Planning)
Enrique Delamonica (Social Policy and Gender Equality, UNICEF Nigeria) (Speaker)

In this paper the Small Area Estimation approach (Rao, 2003) is used to estimate and analyse multi-dimensional Child Poverty for Local Government Areas (LGA) in Nigeria. There are 774 LGAs (the political-administrative level below the 36 States comprising the Nigerian Federation). This is the first ever estimate of Child Poverty with such detailed level of geographic granularity in Sub-Saharan Africa. These results are obtained using small area estimations combined with Geographic Information System (GIS) mapping. Besides the intrinsic value of applying the technique for scientific exploration and knowledge generation purposes, the results were used as inputs to decide geographical areas of focused interventions by UNICEF. The paper is divided in three sections.

First, the conceptual framework is presented. Using widely accepted and recognized international criteria, the human rights approach to estimate Child Poverty based on constitutive rights is described (Gordon et al, 2003; Minujin and Nandy, 2012). It consists of assessing simultaneously for each child deprivations in seven rights (education, health, information, nutrition, water, sanitation, and housing).

Second, the available data (i.e. the National Population Census data of 2006, and the National Demographic and Health Survey data of 2013) are used to estimate deprivation for each of these dimensions/rights. The study uses a two-stage small area model (Day and Harriot, 1979; Molina and Rao, 2010) to give estimates of deprivation and GIS for mapping the distribution of poverty dimensions across the geographical hierarchy (i.e. national, state and LGA). The 2013 Household survey data are combined to create small area estimates of the seven dimensions constituting Child Poverty. Adjustment of the weights attached to the survey cases (i.e. each person or household within the survey) are made in order to match as closely as possible the estimated small area counts of the indicators for each right to the ones calculated from census data. Then, in the second step, adjusted weights in the survey are used to calculate a value of each of the seven components of Child Poverty in each LGA. Thus, census (representative for all households) and survey data (only representative at the State level) are interpolated. However, this is not sufficient to estimate Child Poverty which is not an average nor a sum of the level of deprivation in each dimension. It is based on a composite holistic view of the deprivations suffered per child – information which cannot be interpolated using the standard tools. Thus, a technical innovation is introduced to smooth a series of interpolations between the state-level information on Child Poverty and the small area estimates of each of the dimensions. While this allows for a robust estimate of Child Poverty at the LGA level (as established by sensitivity analysis), there are still open questions about the most efficient way to calculate confidence intervals (fundamental for comparisons through time and across LGAs) of the estimates.

Third, the estimates of Child Poverty incidence in each of the 774 LGAs are translated into maps which graphically represent where the poor children are located in Nigeria. The maps show the distribution of the dimensions across the LGAs and their contribution to total Child Poverty. For instance, household and neighbourhood level problems such as sanitation and inadequate provision of water supplies, health, and shelter are most severe in the North-East of Nigeria. Also, their burdens fall primarily on the rural population. LGAs where there are the city centres have wider access to education and safe and clean water while showing lower access to shelter than rural ones. Spatial auto-correlation and Moran indices are used to explore the influence of geography and proximity in forming clusters of high Child Poverty LGAs. Two measures of intra-state inequality are also explored and analysed. One is the depth of poverty by LGA. The other one is the absolute gap between the best and worst performing LGA for each dimension of child poverty. Neither of these two estimates on inequality would have been possible without the small area estimates.

The paper ends with a discussion reflecting on the relevance of this geographic analysis to assess and evaluate national and donor-funded strategies and policies to reduce Child Poverty in Nigeria and monitoring the Sustainable Development Goals (SDGs). Using small area estimates has allowed for estimates of child poverty and inequality which are novel, expanding the frontier of knowledge about the situation of children and families in Nigeria. Moreover, the analysis presented here has had practical implications, beyond the purely intellectual contribution. It has been used for determining geographic areas of focus in the collaboration between UNICEF and the Government of Nigeria (Ministry of Budget and National Planning). Moreover, in order to deal with some challenges in applying small area estimates for measuring Child Poverty has resulted in some technical innovations and has raised some questions for further research.

Day, T. E. And Harriot, R. A. (1979), "Estimates of income for small Places: An Application of James -Stein Procedures to Census Data, "Journal of the America Statistical Association, 74, 269-277. Gordon, D., Nandy, S., Pantazis, C., et al. (2003), Child Poverty in the Developing World, The Policy Press, Bristol. Minujin, A. and Nandy, S. (2012) Global Child Poverty and Well-being: Measurement, concepts, policy and action, Bristol: Policy Press. Molina, I. And Rao, J. N. K (2010). Small Area Estimation of Poverty indicators. Canadian Journal of statistics. Rao, J. N. K. (2003). Small Area estimation. Wiley.
Keywords: multi-dimensional poverty; child poverty, SDGs, small area estimate; Moran index; spatial auto-correlation.