Poverty in Nigeria is particularly worrisome because of the country’s available natural and human resources. The World Bank report [1] established that Nigeria with about 170 million population falls among countries with extreme poverty whose over 70% population live on $1.25 (N200) or even less per day. National Bureau of Statistics (NBS) put the poverty rate in Nigeria at 54.4% and 69.0% in 2005 and 2010 respectively, while [2] and [3] reported 70.2% and 70% respectively. United Nations Habitat [4] put the poverty rate at 76%. With quantum of poor unemployed youth, poverty has been linked with the rising number of youth engaging in social vices such as robbery, 419 activities (including internet fraud), political thugs, campus cultism, drug and human trafficking, militancy and Boko haram insurgence in the northeast which has claimed many lives.

Over the years, a number of Poverty Reduction Strategy (PRS) has been initiated in Nigeria. This includes the recently designed National Economic Empowerment Development Strategies (NEEDS). Also, a special Federal Government institution to alleviate poverty in the country; the National Poverty Eradication Programmes (NAPEP) was created. These previously initiated PRS in the country appear only to have addressed the various manifestations of poverty, such as unemployment, lack of access to credit and functional rural and urban infrastructures, and gender inequality among others. While these PRS were well intentioned, none had any significant, lasting, or sustainable positive effects on the people they were planned for ( [3] , [5] ). This may be attributed among others to the non-consideration of the heterogeneous nature of poverty and spatial contiguity of geographical units in the design of PRSPs. Virtually, all existing poverty studies ( [6] - [16] ) treated geographic unit, such as a local government area, a state, ora geographical zone, as an independent isolated entity rather than as an entity surrounded by other geographic units with which it interacts. In a village or urban community, many of the households may have similar sources of income, and all households affected by the same agro-climatic and geographic conditions. Households may also have other circumstances in common including road conditions, availability of public facilities such as health, water supply and education. Hence, it is reasonable to suppose that households living in the same area tend to act in similar ways and to influence one another ( [17] ). The poverty level of a community is not only determined by factors highlighted above but also by conditions in the neighbouring communities. For instance, if the poverty level of a specific geographical entity is low, the neighbouring communities will be affected by the spillover effects of low poverty. The same is also true if an area is prosperous, the spillover effect of the prosperity will lead to a reduction in poverty in the neighbouring areas. According to [18] , if this bias (spatial dependence) is ignored, econometric results may be incorrect and produce policy recommendations that are counterproductive. We use spatial data analysis methods in this study to explore the spatial contiguity of senatorial districts based on poverty incidence, and spatial econometric methods to incorporate the spatial bias formally into the econometric models.

The focus of this study is on senatorial districts. In addition to legislative duties, senators also have tasks with the economic development of their respective senatorial district through lobby for sitting capital projects and judicious use of monthly constituency allowance meant for execution of projects that will enhance the well-being of the people. A thorough knowledge of the poverty situation and incidences at the senatorial levels will provide sufficient information to state governors to know how to intervene effectively. Furthermore, each senatorial district, on average is made up of 7 local government areas which also imply that knowledge of poverty situation in the senatorial district can guide local government chairmen and councilors on how to tackle poverty in their domains.

There is scarce information on spatial decomposition and spillover of poverty across the Senatorial Districts (SD) in Nigeria. The objective of the study is to determine the similarity (high-high, low-low) or dissimilarity in poverty incidence across the neighbouring senatorial districts. This will enable a senator to know whether his SD’s performance is better or worse than the performance of neighbouring SDs in terms of poverty reduction. The study also identified the determinants of cold (low-low) and hot (high-high) spots of poverty. The paper is divided into four sections; namely introduction, theoretical framework, methodology, results and discussion and conclusion.

The concepts and assumptions of spatial analysis measure geographic variables that exhibit properties of spatial dependency (the tendency of the same variables measured in locations in close proximity to be related). While traditional statistical techniques have treated this feature as nuisance, spatial statistics considers them explicitly. In the past, models that explicitly incorporated “space” (or geography) were primarily found in specialized fields such as regional science, urban and real estate economics and economic geography (examples are reviews in [19] - [22] ). However, spatial econometric methods have increasingly been applied in a wide range of empirical investigations in more traditional fields of economics as well, including, among others, studies in agricultural and environmental economics ( [17] [18] [23] - [30] ).

Unlike in the developed countries, there are few literatures on the application of spatial econometrics to poverty studies in sub-Saharan Africa. It is generally new in Nigeria. According to [26] , the landscape of poverty is a result of many forces acting independently and in interaction with other social and structural forces to produce a set of opportunities and constraints. These are manifested in the economic realities of wealth and poverty. Poverty is a highly heterogeneous phenomenon showing a wide spatial variability. A large difference in the standard of living of the populations in different geographical locations is common. Spatial heterogeneity between areas can be introduced in a model for a variety of reasons, including differences in agroclimatic conditions, geographic conditions (particularly access to main urban centres and markets), the presence of natural resources (particularly water for irrigation), other non-physical conditions (especially, historical and ethnic) and facets of public policy ( [31] ).

In a study on poverty and inequality in Vietnam: spatial patterns and geographic determinants, [27] findings revealed that 10 percent point increase in the poverty rate in a district results in 8 percent increase in the poverty rate in a neighbouring district. In a similar study on spatial approach to social and political forces as a determinant of poverty in US, [18] indicate that a 10 percentage point increase in the poverty rate in a county results in a 2% increase in the poverty rate in a neighbouring county. This is strong evidence that spillover effects exist between counties with respect to poverty. Neighbourhood effects on poverty as a result of similarities in socioeconomic and environmental factor are well documented in studies ( [32] [33] ) carried out in Kenya and Uganda respectively. This finding is corroborated by [24] . They reasoned that poverty of a neighbourhood is tied to the fortunes of neighbouring areas: there are geographic spillovers in poverty reduction. Reducing poverty in particular neighbourhoods affects the poverty of neighbouring tracts. A study on the topography of poverty in US, [26] findings showed that 51.9% of the total counties belong to similar spatial concentration (low-low and high-high), whereas only 7.8% were categorized as being spatial outliers (high-low and low-high).

Writers in the sociological, political science, and regional science literatures point out that certain community attributes are empirical correlates of successful communities ( [34] - [40] ). These and other studies suggest that many factors influence the level of community and economic development of a place [41] . For example, view poverty as a condition of the local social structure. Numerous studies have found a positive association between economic development and social capital [18] - [42] . investigated the independent effects of social capital on family poverty rate; their result revealed that counties/local government areas rich in social capital have lower family poverty rates.

According to [43] , the intrinsic value of education in raising individual capabilities and freedoms and consequently contributing to higher incomes cannot be over-emphasised. It has been severally documented (example is [44] ) that labour is the main asset of the poor. One of the important ways of enhancing and preserving this asset is through education. Indeed, the [45] showed that investment in education and other forms of human capital particularly health is an important element of a poverty reduction strategy.

The study covered the 109 Senatorial Districts (SD) of Nigeria. Each Senatorial district is composed of Federal constituencies (360 Federal constituencies) while Federal constituency is made up of Local Government Areas. Figure 1 shows the senatorial districts map of Nigeria. The study utilized secondary data obtained from 2006 Core Welfare Indicator Questionnaire Survey (NLSS) and 2003/2004 National Living Standard Survey data by National Bureau of Statistics (NBS). The only NBS data disaggregated into senatorial districts. Other sources were the Nigerian Institute of Meteorological Services (NIMET) for agro-climatic and environmental data and Food and Agricultural Organisation for the fertility soil map of Nigeria.

Although the 2006 Core Welfare Indicator Survey data were collected two years after the National Living Standard Survey (NLSS), the 2004-2006 periods was one of relatively slow growth and low inflation in Nigeria, so it was reasonable to assume that there was relatively insignificant change. Similar studies carried out with lagged data are:

1) [17] in Ecuador combined Census of population and households conducted in 1990 and the World Bank’s Living Standard Measurement Surveys (LSMS) conducted in 1995.

2) [32] in Uganda utilized two household data sets collected by the Uganda Bureau of Statistics (UBOS):

census data for 1991 and sample survey data from 1992 (IHS) to derive welfare estimates and maps.

3) [18] in the United States of America utilized family poverty rate from the 2000 Census while the explanatory variables were measured in 1990.

The study provided answers to the following research questions, using senatorial district as the reference geographical location:

1) Is the poverty incidence of a senatorial district significantly influenced by poverty incidence of neighbouring senatorial district(s)?

2) What are the factors influencing poverty levels in senatorial districts with similar spatial patterns of poverty?

The detail of the variables (dependent and independent) used in the analysis, the a priori expectations and the literatures supporting the a priori expectations are presented in Table 2.

The soil classification per senatorial district data were obtained from the Food and Agricultural Organization (FAO) soil map of Nigeria based on natural fertility and traditional agricultural practices. FAO classified Nigerian soil into very low productivity, no productivity, high productivity, medium productivity and low productivity. For this study, a dummy variable was used to capture the soil classification. The soil of very low productivity and no productivity were classified as bad soil (bad soil = 0) while soils of high productivity, medium productivity and low productivity were classified as good soil (good soil = 1). Good soil (soils of high productivity, medium productivity and low productivity) was expected to bring about reduction in poverty rate ( [32] [33] ).

Data were analyzed using descriptive and spatial regression analyses. A diagnostic ordinary least squares regression analysis was carried out by expressing the prevalence of poverty per SD as a function of selected spatial variables. This was to ascertain the presence of spatial dependence. The OLS regression model was estimated as:

where:

y_i is a vector average poverty rates;

X_i is a matrix of independent variables,

β_i is a vector of coefficients,

e is a vector of random errors.

The geo-referenced independent variables data were grouped under demographic, agro-ecological and environmental, infrastructural, sociopolitical and economic characteristics (see Table 2). The data under each group- ing were based on senatorial district.

Through the diagnostic, the presence of spatial autocorrelation is revealed as well as the cause (spatial lag or spatial error) for necessary correction. Either of the models below corrects the defect (the presence of spatial dependence):

1) Spatial-error model:

Note: Nigerian Institute of Meteorological Services (NIMET), Central Bank of Nigeria (CBN), National Bureau of Statistics (NBS), Core Welfare Indicator Survey Data (CWIS), National Living Standard Survey Data (NLSS) & Food and Agriculture Organization. The detailed information on soil classification is given below.

2) Spatial-lag model:

where:

y is an n × 1 vector of dependent variable (average poverty rate),

x_i an n × k matrix of covariates (independent variables),

b_i is the regression coefficient for the independent variables,

e is a zero-mean error term,

W_(l) and W_(e) are n × n spatial lag and error weights matrices, respectively.

{r, l} the associated scalar spatial parameters (measures the extent of spillover).

Following [18] [26] - [29] [63] , we calculated local Moran’s I and obtain Local Indicator of Spatial Association (LISA) map (this identifies the hotspots and coldspots of poverty).

The study examined the determinant of hotspots and coldspots of poverty in Nigeria using spatial econometrics technique. The study revealed that the individual geographic units that made up the country were not independent and isolated entities; but individual entity was surrounded by other geographical units which interacted significantly with one another. The result affirmed that spillover of poverty from one senatorial district to the other was present in significant proportion. The study also identified the senatorial districts with low-low and high-high poverty incidence. People employed in agriculture and household size have positive influence on poverty incidence while annual rainfall, literate adult and households’ access to basic infrastructure have negative influence on poverty incidence in coldspots and hotspots of poverty

Based on the findings of this study, it is recommended that possible spillover of poverty from neighbouring geographical area should be incorporated while designing poverty reduction programmes. The need for increased productivity of farmers through adoption and availability of modern farm inputs rather than increase in the number of farmers is recommended by the study. This will not only bring about genuine contribution of agriculture to country’s GDP but also the increase. General increase in the accessibility of populace to basic infrastructure (safe water, public electricity and health) and education by government, most especially in the high- high senatorial districts is important in order to achieve a sustainable poverty reduction in Nigeria.

Fatai Abiola Sowunmi, (2016) Spatial Analysis of Hotspots and Coldspots of Poverty in Nigeria. Journal of Geographic Information System,08,301-315. doi: 10.4236/jgis.2016.82026

Source: the results of data analyses (2010).