See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267568043

Geography of Infrastructure Functionality at Schools in Nigeria:

Evidence From Spatial Data Analysis Across Local Government
Areas

Conference Paper in Papers in Applied Geography · October 2014

DOI: 10.1080/23754931.2015.1012443

CITATIONS READS

6 309

3 authors:

Shaky Sherpa Belinda Archibong

Columbia University Columbia University
7 PUBLICATIONS 172 CITATIONS 49 PUBLICATIONS 356 CITATIONS

SEE PROFILE SEE PROFILE

Vijay Modi
Columbia University
158 PUBLICATIONS 5,015 CITATIONS

SEE PROFILE

All content following this page was uploaded by Belinda Archibong on 31 July 2019.

The user has requested enhancement of the downloaded file.

This article was downloaded by: [Belinda Archibong]
On: 19 June 2015, At: 07:52
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Papers in Applied Geography

Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/rpag20

Geography of Infrastructure
Functionality at Schools in Nigeria:
Evidence From Spatial Data Analysis
Across Local Government Areas
a a a
Belinda Archibong , Vijay Modi & Shaky Sherpa
a
Columbia University
Published online: 11 Jun 2015.

Click for updates

To cite this article: Belinda Archibong, Vijay Modi & Shaky Sherpa (2015) Geography of Infrastructure
Functionality at Schools in Nigeria: Evidence From Spatial Data Analysis Across Local Government
Areas, Papers in Applied Geography, 1:2, 176-183, DOI: 10.1080/23754931.2015.1012443

To link to this article: http://dx.doi.org/10.1080/23754931.2015.1012443

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
 Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [Belinda Archibong] at 07:52 19 June 2015
 Geography of Infrastructure Functionality at Schools in Nigeria:
Evidence From Spatial Data Analysis Across Local Government Areas

Belinda Archibong, Vijay Modi, and Shaky Sherpa

Columbia University

Is functionality of electricity, sanitation and water infrastructure at schools unequally distributed geographically in Nigeria? Are
there significant disparities in infrastructure functionality between Northern and Southern geopolitical zones in the country as
has been posited in previous studies? In this study, we answer these questions with an examination of functionality at schools,
with metrics for functionality aggregated at the smallest administrative unit available in the country, the local government area
(LGA). We employ spatial statistical techniques to examine the spatial autocorrelation of power, sanitation and water (or
‘infrastructure’) non-functionality across 68,627 schools for 764 of 774 local government areas in Nigeria using a novel survey
dataset courtesy of Nigeria's Office of the Senior Special Assistant to the President on the Millennium Development Goals. We
find evidence for the existence of LGA clusters of infrastructure non-functionality, aligned along Nigeria’s six geopolitical zones.
The results also reveal a significant cluster of LGAs in the Northwest zone, the zone with the highest income-based poverty rate
(70%) in the country, outperforming LGAs in both other Northern and some Southern zones on all functionality indicators.
The results hold up to multiple testing correction, controlling the false discovery rate using the Benjamini-Hochberg method.
These results highlight the need for a spatially targeted policy approach, at finer spatial scales, to poverty reduction efforts
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

through infrastructure access expansion in the country. Keywords: clusters, geopolitical zone, infrastructure access, local
government area, Nigeria, spatial autocorrelation.

I s functionality of electricity, sanitation, and water at

schools unequally distributed geographically in
Nigeria? Are there significant contiguous (or closely
Castro and Singer 2006). These results highlight the
benefits of detailed microspatial data for identifying
infrastructure functionality distribution in the coun-
located) clusters of schools with disparities in metrics try. They also suggest evidence of structural, geopo-
that define access to these infrastructure elements litically based disparities in functionality of public
between the northern and southern geopolitical zones infrastructure services at the school facility level in
in the country, as has been posited in previous studies Nigeria with potentially useful insights for planners
(Akinyosoye 2006; Kanbur and Venables 2007; and policymakers in the country.
Sowunmi et al. 2012)? We answer the aforementioned In the first known comprehensive analysis of infra-
questions with an examination of electricity, sanita- structure functionality at the LGA level, we use data
tion, and water functionality at schools, with metrics from a 2011–2012 survey conducted by OSSAP in an
for functionality aggregated at the smallest administra- effort spearheaded by the Nigerian government,
tive unit available in the country, the local government which received responses from point persons at
area (LGA). We employ spatial statistical techniques 68,627 schools (over 80 percent public or government
along with geographic information systems (GIS) to owned) in 764 LGAs on available functionality of
examine the spatial autocorrelation of electricity, sani- power, sanitation, and water. We then employed well-
tation, and water (henceforth referred to as infrastruc- known spatial statistical methods, including global
ture) functionality across 68,627 schools for 764 of Moran’s I and Getis–Ord Gi*(d) to analyze the spatial
774 LGAs in Nigeria. The underlying data were from autocorrelation of infrastructure functionality for each
a recent survey courtesy of the herculean efforts of the LGA and discover conclusively the answers to the fol-
Nigeria’s Office of the Senior Special Assistant to the lowing questions:
President on the MDGs (OSSAP; MDGs stand for
the Millennium Development Goals1). 1. How is infrastructure functionality spatially asso-
We find evidence for the existence of LGA clusters ciated at the local level in Nigeria? Is there spa-
of infrastructure nonfunctionality, aligned along tial autocorrelation of infrastructure functionality
Nigeria’s six geopolitical zones. Our results also at the LGA and zonal levels in Nigeria? Are
reveal a significant cluster of LGAs in the Northwest there LGA clusters of infrastructure nonfunc-
zone, the zone with the highest income-based poverty tionality (high values of infrastructure nonfunc-
rate (70 percent) in the country, outperforming tionality) in the country or clusters of
LGAs in both other northern and some southern infrastructure functionality (low values of infra-
zones on all functionality indicators. Our results hold structure nonfunctionality)?
up to multiple testing correction, controlling the false 2. If so, where are these clusters located? Do we see a
discovery rate using the Benjamini–Hochberg strict north–south zonal divide in the distribution
method (Benjamini and Hochberg 1995; Caldas de of functionality?

Papers in Applied Geography, 1(2) 2015, pages 176–183 © Applied Geography Conferences.
Published by Taylor & Francis Group, LLC.
 Infrastructure Functionality at Schools in Nigeria 177

Table 1 Summary statistics at national and geopolitical zonal levels

Number of Population Poverty Proportion Proportion Proportion

Zone LGAs in study density rate (%)a Proportion I D 0 power D 0 sanitation D 0 water D 0

National 764 1,028 61 0.41 0.78 0.55 0.68

North-Central 121 273 60 0.52 0.82 0.67 0.72
Northwest 184 912 70 0.37 0.86 0.54 0.60
Northeast 104 203 69 0.51 0.91 0.64 0.73
South-South 123 551 56 0.37 0.73 0.48 0.67
Southwest 137 3,017 50 0.32 0.62 0.48 0.59
Southeast 95 1,214 59 0.37 0.75 0.47 0.74

Note: Data are author estimates. LGA D local government area.

a
Poverty rate, measured as proportion of persons living on under US$1 per day (National Bureau of Statsitics 2010).

Nigeria’s northern geopolitical zones with average water, and sanitation. Important to note here is the fact
poverty rates2 of 66 percent, 11 percentage points that although electricity might not be a primary infra-
above the Southern average, are often identified as the structure objective at primary schools, the simple aggre-
underperformers of their regional neighbors on devel- gate LGA-wide measure appears to reasonably proxy
opment metrics, including access to public infrastruc- LGA-wide functionality of power in correlations against
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

ture services (Madu 2006; Foster and Brice~ no- the Defense Meteorological Satellite Program and
Garmendia 2010; National Bureau of Statistics 2010; Operational Linescan System (DMSP-OLS) night-
Sowunmi et al. 2012; Foster and Pushak 2011). When lights data from the 2011–2012 period of study with a
infrastructure functionality, a component of access, is significant correlation of ¡0.67.
examined at the small-scale LGA level, which lists We then run simple Pearson correlation tests to
among its duties the management of sanitation and examine the aspatial bivariate relationship between
water supply, we view unexpected, relative to the North- our infrastructure indicators. Initial visualization,
west zone’s status as the area with the worst reported descriptive statistics, and correlation test results are
poverty rate in the country (at 70 percent), LGAs that presented in the next section. We then assess overall
consistently outperform other northern and some patterns of spatial association in the LGA infrastruc-
southern LGAs on all infrastructure functionality indi- ture functionality indicators with the global Moran’s I
cators in ways not fully explained by population density statistic. Finally, we identify local patterns of spatial
alone (see Table 1). We also find certain Southeast autocorrelation in the infrastructure functionality
LGAs that outperform southern neighbors on the sani- indicators with the Getis–Ord Gi*(d) statistic.
tation functionality variable, an unexpected finding
given the Southeast zone’s status as the zone with both
the highest poverty rate (59 percent) among its southern Data, Descriptive Statistics, and Initial Visualization
neighbors and the worst water functionality at schools of In an effort spearheaded by the Nigerian government,
zones in our sample. These LGAs, outliers for their researchers from the OSSAP in collaboration with the
zones, could serve as blueprints for the creation of poli- Sustainable Engineering Lab at Columbia University
cies facilitating infrastructure functionality expansion conducted extensive, comprehensive surveys of
and improvement in the country. Our results also point schools at LGAs, obtaining responses to questions
to evidence for geopolitically based inequality of infra- concerning power, water, and sanitation functionality,
structure functionality in Nigeria. In the subsequent among other indicators. The surveys were collected
sections, we present our approach, testing our methods from principals at 68,627 schools across 764 of 774
on the novel survey data set, with the accompanying LGAs in Nigeria (with the last ten LGAs dropped due
results presented. A brief discussion of our results is to unreliable data). Over 80 percent of the schools
then proffered. were public schools. For power functionality, respond-
ents were asked true–false questions about both avail-
ability and functionality. An aggregate power score of 0
Methods or 1 was assigned to a school depending on if the
respondent answered false or true to the question of
To examine the spatial association of infrastructure whether the respondent had available functional power
functionality at the LGA level in Nigeria, we employ from the national grid, functional power from a genera-
aspatial and spatial statistical techniques along with GIS tor, or functional power from a solar system. Similarly,
to analyze spatial autocorrelation of infrastructure func- an aggregate sanitation score of 0 or 1 was assigned to a
tionality across schools in the country. We do this in school depending on if the respondent answered false
three steps. First, we create a simple, aggregated mea- or true to the question of whether the respondent had
sure for each LGA of the proportion of schools in each functional improved sanitation in the form of a func-
LGA reporting zero functionality (nonfunctionality) to tional flush or improved pour flush toilet, or a func-
a particular infrastructure metric, namely electricity, tional improved ventilated latrine or pit latrine with a
 178 Archibong et al.

Table 2 Descriptive statistics: Pearson’s product–moment correlation coefficient

Variable Proportion power D 0 Proportion sanitation D 0 Proportion water D 0

Proportion power D 0 1.00 0.62* 0.57*

Proportion sanitation D 0 0.62* 1.00 0.47*
Proportion water D 0 0.57* 0.47* 1.00

Note: Data are author’s estimates.

*p < .001.

slab. If respondents responded true to any one of these nonfunctionality indicators compared to lower non-
improved sanitation options, they were assigned a sani- functionality patches in the southern zones of the
tation score of 1. If they responded false to all of the country. The initial visualization gives us an idea of
aforementioned options, then they were assigned a san- the trends to expect in terms of the distribution of
itation score of 0. Lastly, for water functionality, which infrastructure functionality across LGAs in the coun-
was potable water functionality, we assigned an aggre- try. However, it tells nothing about which LGAs are
gate potable water functionality score of 0 or 1 to a in the extremes of the functionality distribution; that
school depending on if respondents answered false or is, areas of intense infrastructure nonfunctionality,
true to the question of whether they had available func- with functionality scores significantly below the global
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

tional potable water in the form of functional piped mean or areas of significant infrastructure functional-
water or functional borehole or tube well water. ity, with functionality scores above the global mean
Next, an aggregated infrastructure score between 0 for each variable. It is also difficult to identify which
and 3 was created for each school, which was a simple LGAs outperform or underperform significantly on
sum of the three functionality scores for power, water, all indicators based solely on the initial visualization.
and sanitation. Finally, for each LGA, the proportions
of schools (as a fraction of the total number of schools) Results
with overall infrastructure and individual power, water,
and sanitation scores of 0 were calculated and used as
our metrics of overall infrastructure nonfunctionality Global Association of Infrastructure Functionality
and power, water, and sanitation nonfunctionality. Rates Across LGAs
Tables 1 and 2 show the summary statistics for the total
The results for the global Moran’s I test for overall
number of schools sampled, population density, and
spatial autocorrelation in our 764 LGA sample are
infrastructure functionality aggregated at the national
given using Equation 1:
and geopolitical zonal levels. Table 2 provides correla-
tion coefficients on the relationship between each of XX
the infrastructure functionality measures. n w .x ¡ x/.xj ¡ x/
j ij i
ID XX X
i
Infrastructure functionality rates at schools are low- (1)
wij .x ¡ x/2
i i
est for power nationally, with 78 percent of schools in i j
LGAs reporting no available, functional power. Water
and sanitation functionality rates are the second and wherewij is a contiguity weight matrix that equals 1 if
third lowest of the three functionality indicators, locations i and j are neighbors and 0 otherwise; xi is
respectively, with respondents reporting nonfunction- the variable of interest, in our case each of the infra-
ality rates of 68 percent and 55 percent nationally. structure functionality rates; n is the number of obser-
Note that almost all northern LGAs report nonfunc- vations, equal to 764 LGAs in our study; and x is the
tionality figures higher than the national average, with global mean for infrastructure functionality variable x
the Northwest zone managing to remain below the in the sample.
national mean on the sanitation and water nonfunc- The results are presented in Table 3. We settle on
tionality indicators. the k D 8 nearest neighbors weight matrix for the rest
We also see a significant correlation between infra- of this study, based on our knowledge of the study
structure functionality indicators, with the strongest area, as it allows for every entity to have an adjacent
positive linear association between the power and sani- neighbor (not the case in the Queen’s matrix) and
tation functionality indicators, the next strongest asso- does not overemphasize smaller LGAs due to the
ciation between power and water, and the weakest, but irregularity of polygon sizes in the Nigeria region (as
still significantly positive, association between sanita- is the case with the 80 km conceptualization, where
tion and water functionality indicators. the minimum distance for every entity to have a neigh-
Initial visualization of these functionality indicators bor is about 72 km). To assess the significance of the
is presented in Figure 1 and was done in ArcGIS’s observed I statistic, the I statistic is compared to the
ArcMap 10.2 (Environmental Sciences Research expected value of I in the absence of spatial autocorre-
Institute [ESRI], Redlands, CA). Note the broad lation, E(I)D ¡1/(n ¡ 1), which tends to 0 as n gets
swaths of northern LGAs with high infrastructure larger; a larger I (i.e., I > E(I)) reflects positive spatial
 Infrastructure Functionality at Schools in Nigeria 179
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

Figure 1 Initial visualization of infrastructure nonfunctionality variables by local government area (LGA) with (A) overall
infrastructure, (B) power, (C) sanitation, and (D) water.

autocorrelation (spatial cluster) and a smaller I (I < E that there is positive spatial autocorrelation in the dis-
(I)) reflects negative spatial autocorrelation (spatial dis- tribution of infrastructure functionality across LGAs
similarity; Yu and Wei 2008). Significance of the in Nigeria.
global Moran’s I statistic is assessed by a test of a null
hypothesis of spatial randomness, rejection of which
indicates a spatial pattern to the data. Significance is
then tested by comparison to a reference distribution. Local Association: Clustering of Infrastructure
The issue of the reference distribution against which Functionality Rates Along Geopolitical Zone
to test significance is much discussed in the literature Local spatial association is examined using the Getis–
(Anselin 1995; Bivand, Pebesma, and G omez-Rubio Ord Gi*(d) statistic, which is a distance-based metric
2008; Yu and Wei 2008) as the exact distribution of that measures the proportion of a variable located
the statistic is often computationally restrictive. Two within a specific radius of a point, respective to the
reputable methods in the literature are the normal total sum of the variable in the study region (Paez and
approximation (Anselin 1995) and the saddlepoint Scott 2005). In other words, it measures “the overall
approximation method (Tiefelsdorf 2002). The results concentration of all pairs xi, xj such that i and j are
from both are equivalent and significant. This implies within d of each other” (Getis and Ord 1992) as

Table 3 Global Moran’s I statistic for spatial autocorrelation in infrastructure nonfunctionality indicators using different
spatial weights and different methods

Global Moran’s I: Spatial weight ID0 Pp0 Wp0 Sp0

* * *
Queen 0.457 0.481 0.377 0.467*
Nearest neighbors (8) 0.430* 0.466* 0.366* 0.427*
80 km 0.345* 0.360* 0.261* 0.325*
Global Moran’s I: Method (w D knn(8))
Normal approximation 0.430* 0.466* 0.366* 0.427*

*All significant at p < .001.

 180 Archibong et al.

depicted in Equation 2 (all variables defined as in When multiple tests are performed on a sample, the
Equation 1): probability that some effects will be retained as sig-
nificant solely by chance rises as the number of tests
Xn grows and must be controlled. The Type I error rate
w .d/xj
j D 1 ij
Gi
ðd Þ D X (2) or the probability of rejecting one or more null
n
x hypotheses when they are, in actuality, true grows
jD1 j
with the number of tests, leading to false discoveries
in our results. To control for these false discoveries,
Testing for the Gi*(d) is straightforward because the we apply the Benjamini–Hochberg (BH) procedure,
Gi*(d) values are expressed as standard normal variates regarded as the most powerful in its class of multiple
Z[Gi*(d)]). Under the null hypothesis of spatial ran- comparison procedures (MCPs) for controlling the
domness, Z[Gi*(d)] are asymptotically normally dis- proportion of Type I errors, to adjust the p values
tributed, N(0, 1) as n ! 1 (Getis and Ord 1992; from our original Gi*(d) results (Benjamini and
Caldas de Castro and Singer 2006). Significance is Hochberg 1995; Caldas de Castro and Singer 2006;
determined by an examination of these z scores, with Williams, Jones, and Tukey 1999). Given a signifi-
large positive z scores and a p value < .05 indicating cance level of a D .05, the BH procedure controls
clustering of high values and large negative z scores the FDR < a where:
with a p value < .05 indicating clustering of low values
within distance d. It has a straightforward interpreta-  
F
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

tion with clustering of high (positive) values regarded FDR D E jR > 0 P.R > 0/
as high clusters or hot spots and indicating clusters of R
extreme values above the global mean and clustering
of low (negative) values regarded as low clusters or FDR is the false discovery rate, F is the number of false
cold spots and indicating clusters of extreme values positives (false discoveries) or Type I errors, and R is
below the global mean. the set of rejected null hypotheses or the sum of false
As a final note on the local Gi*(d) statistic results, positives (F) and true positives (T). The BH adjusted
as local spatial statistics depend on tests of spatial Gi*(d) results are presented here.
association for each location in the sample, the prob- An examination of local association in our 764 LGA
lem of the effect of multiple comparisons on the sig- sample provides some very interesting results depicted
nificance levels of the tests has been raised by a in Figure 2. In summary, although the Northwest
number of authors (Anselin 1995; Williams, Jones, zone is stated as the most income poor in the country,
and Tukey 1999; Caldas de Castro and Singer 2006). it actually registers the second highest infrastructure

Figure 2 GI*(d) Benjamini–Hochberg FDR results: Distribution of significant low and high clusters by infrastructure
nonfunctionality variable for (A) overall I D0, (B) power D 0, (C) sanitation D 0, and (D) water D 0.
Note: LGA D local government area; FDR = false discovery rate.
 Infrastructure Functionality at Schools in Nigeria 181

Table 4 GI*(d) Benjamini–Hochberg FDR corrected results: Intensity of clustering by zone, (percentage of total local gov-
ernment areas in each zone in high clusters for all (I D 0) and each infrastructure nonfunctionality variable) along with pov-
erty rate and average infrastructure nonfunctionality rates for each zone

Geopolitical zone Poverty rate High Gi*_all High Gi*_power High Gi*_sanitation High Gi*_water

North-Central 0.60 0.45 0.01 0.33 0.24

Northeast 0.69 0.37 0.19 0.19 0.13
Northwest 0.70 0.04 0.01 0.07 0.00
South-South 0.56 0.04 0.00 0.05 0.05
Southeast 0.59 0.07 0.00 0.03 0.25
Southwest 0.50 0.01 0.00 0.01 0.00
Geopolitical zone Poverty rate Proportion I D 0 Proportion power D 0 Proportion sanitation D 0 Proportion water D 0
North-Central 0.60 0.52 0.82 0.67 0.72
Northeast 0.69 0.51 0.91 0.64 0.73
Northwest 0.70 0.37 0.86 0.54 0.60
South-South 0.56 0.37 0.73 0.48 0.67
Southeast 0.59 0.37 0.75 0.47 0.74
Southwest 0.50 0.32 0.62 0.48 0.59
Geopolitical zone Poverty rate Low Gi*_all Low Gi*_power Low Gi*_sanitation Low Gi*_water
North-Central 0.60 0.01 0.02 0.00 0.03
Northeast 0.69 0.01 0.00 0.00 0.05
Northwest 0.70 0.22 0.04 0.16 0.17
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

South-South 0.56 0.27 0.13 0.31 0.07

Southeast 0.59 0.15 0.03 0.39 0.00
Southwest 0.50 0.27 0.29 0.23 0.21

functionality rates in the country (tied with the South- poverty and infrastructure nonfunctionality to be
east and South-South zones at .37 for I D 0) as shown strongly positively correlated as discussed earlier.
in Table 4. In terms of intensity of clustering, this In contrast, the North-Central zone appears to be
zone registers the second greatest proportion of LGAs the worst off both in terms of overall infrastructure
in the low cluster for all infrastructure nonfunctional- functionality and intensity of clustering of infra-
ity (I D 0) and the second least proportion of LGAs in structure nonfunctionality in the region. Again this
the high cluster for I D 0 as shown in Table 4, located contrasts with its position as the zone with only the
primarily in Kano, Katsina, and Jigawa states. Again, third highest income-based poverty rate in the
this is an unexpected result, if we expect income country, as shown in Table 4, and the zone

Figure 3 Six geopolitical zones in Nigeria with thirty-six states labeled and 774 local government areas in faint outline.
 182 Archibong et al.

Figure 4 GI*(d) Benjamini–Hochberg corrected results: Intensity of clustering by zone and each infrastructure nonfunc-
tionality variable along with poverty rate for each zone.
Note: LGA D local government area.

containing the country’s federal capital, Abuja. For patterns and speak to structural inequality of infra-
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

intensity of clustering, even when the effect of the structure functionality aligned along geopolitical
presence of the federal capital Abuja is accounted zones in Nigeria. This is the first known study investi-
for, we see significant high clusters of infrastructure gating the spatial distribution of infrastructure func-
nonfunctionality across the North-Central zone tionality at schools at the LGA level in Nigeria and
concentrated in Plateau, Nasarawa (a state that the first, to our knowledge, to apply a BH procedure
shares a border with the Federal Capital Territory to correct the false discovery rate to investigate local
(FCT) where Abuja is located, as shown in Figure 3) spatial association of infrastructure functionality at the
and Benue states. Northeast LGAs, with the highest LGA level in Nigeria. The results presented here have
proportion of power nonfunctionality (0.91 for important policy implications and the lauded efforts of
power D 0), also score highly on the intensity of OSSAP’s microfacility and LGA-level survey efforts
clustering of power nonfunctionality, with enable a shift away from broad, macro, purely income-
19 percent of all Northeast LGAs located in the based poverty assessments toward a more holistic
high cluster for power, and located primarily in approach to poverty reduction through infrastructure
Borno, Yobe, and Taraba states. access and functionality expansion that considers the
The Southeastern LGAs appear to underperform role of geography in shaping differential functionality
significantly relative to their zonal neighbors for water of public infrastructure in the country. They also serve
functionality in the country (0.74 for water D 0), also to partly debunk the narrative of pervasive northern
scoring highly on the intensity of clustering of water underperformance in the country, by identifying clus-
nonfunctionality concentrated in Enugu and Imo ters of LGAs in the Northwest zone that outperform
states. There also appears to be an outlier cluster (for southern and other regional neighbors on our overall,
its zone) of high sanitation and water nonfunctionality power, sanitation, and water infrastructure functional-
LGAs in Bayelsa state in the South-South zone of the ity indicators, which might serve as a blueprint for
country. LGAs in lagging northern and some southern zones.
Overall, the Northeast and North-Central LGAs The results also reveal a similar cluster of Southeast
appear to dominate the high clusters across most infra- LGAs that outperform their regional neighbors on the
structure indicators, with the Northwest LGAs stand- sanitation functionality score, partly upsetting the
ing out as the outperformers of all LGAs in their trend of Southeastern underperformance among their
zonal neighbors and Southeast LGAs outperforming southern neighbors. The detection of nonfunctionality
on sanitation, but lagging significantly on water func- hot spots or LGAs with critically high rates of infra-
tionality. The results are summarized in Figures 2 and structure nonfunctionality and cold spots can be a use-
4 and Table 4. ful starting point for planning and policy decision
making in determining which areas have the most
pressing infrastructure needs overall and disaggre-
Discussion and Further Research gated by variable: power, water, and sanitation in the
country.
Recognition of spatial autocorrelation is important for
adjustment of regression models for further research Acknowledgements
investigating why these patterns occur due to the vio-
lation of the independence assumption (Anselin 1995). Special thanks go to Nigeria’s Office of the Senior
The results presented here are an important first step Special Assistant to the President (OSSAP), Prabhas
toward investigating the drivers of these spatial Pokharel, Chris Natali, Salah Chafik, Zaiming Yao,
 Infrastructure Functionality at Schools in Nigeria 183

Brett Gleitsmann, Carson Farmer, and the members National Economic Empowerment, Development Strategy
of the Columbia University Sustainable Engineering (Nigeria), & Nigeria. National Planning Commission.
Lab for the data and discussion of methods used in 2005. Nigeria, non-functionality reduction strategy paper:
this article. National Economic Empowerment and Development
Strategy No. 5–433, International Monetary Fund,
Washington, DC.
Notes Paez, A., and D. M. Scott. 2005. Spatial statistics for urban
1 analysis: A review of techniques with examples. GeoJournal,
The MDGs from can be found at http://nmis.mdgs. 61(1): 53–67.
gov.ng/. Sowunmi, F. A., V. O. Akinyosoye, V. O. Okoruwa, and B. T.
2
Poverty rates are measured as the percentage of per- Omonona. 2012. The Landscape of Poverty in Nigeria: A
sons living on under US$1 per day (National Bureau Spatial Analysis Using Senatorial Districts-level Data.
of Statistics 2010). American Journal of Economics, 2(5): 61–74.
Tiefelsdorf, M. 2002. The saddlepoint approximation of
Moran’s I’s and local Moran’s Ii’s reference distributions
References
and their numerical evaluation. Geographical Analysis, 34(3):
Akinyosoye, V. O. 2006. Government and agriculture in 187–206.
Nigeria: Analysis of policies, programme and administration. Williams, V. S., L. V. Jones, and J. W. Tukey. 1999.
Anselin, L. 1995. Local indicators of spatial association— Controlling error in multiple comparisons, with examples
LISA. Geographical Analysis, 27(2): 93–115. from state-to-state differences in educational achievement.
Journal of Educational and Behavioral Statistics, 24(1): 42–69.
Downloaded by [Belinda Archibong] at 07:52 19 June 2015

Benjamini, Y., and Y. Hochberg. 1995. Controlling the false

discovery rate: A practical and powerful approach to Yu, D., and Y. D. Wei. 2008. Spatial data analysis of regional
multiple testing. Journal of the Royal Statistical Society, Series development in Greater Beijing, China, in a GIS
B (Methodological), 57: 289–300. environment*. Papers in Regional Science, 87(1): 97–117.
Bivand, R. S., E. J. Pebesma, and V. G omez-Rubio. 2008.
Applied spatial data analysis with R (Vol. 747248717). New
York: Springer. BELINDA ARCHIBONG is a doctoral candidate in
Caldas de Castro, M., and B. H. Singer. 2006. Controlling the School of International and Public Affairs, Colum-
the false discovery rate: A new application to account for
multiple and dependent tests in local statistics of spatial
bia University, 500 West 120th Street, New York, NY
association. Geographical Analysis, 38(2): 180–208. 10027. E-mail: [email protected]. Her research
Foster, V., and C. Brice~ no-Garmendia. 2010. Africa’s interests include political economy, economic and
infrastructure. In A time for transformation (1–27). political geography, development and environmental
Paris: Agence Française de Developpement and the World economics, and energy policy.
Bank. http://webcache.googleusercontent.com/search. (last
accessed November 20, 2014). VIJAY MODI is Professor in the Department of
Foster, V., and N. Pushak. 2011. Nigeria’s infrastructure: A Mechanical Engineering, Columbia University, 500
continental perspective. Washington, DC: World Bank.
West 120th Street, New York, NY 10027. E-mail:
Getis, A., and J. K. Ord. 1992. The analysis of spatial
association by use of distance statistics. Geographical [email protected]. His research interests include
Analysis, 24(3): 189–206. energy infrastructure, CO2 sequestration, fuel cells,
Kanbur, R., and A. J. Venables. 2007. Spatial disparities and distributed sensing/control of flow, and heat transfer.
economic development. Chapter 9: 204–15.
Madu, I. A. 2006. Spatial inequality in Nigeria: The SHAKY SHERPA is GIS Research Analyst at the
imperative of geographic perspectives in the development Earth Institute, Columbia University, 500 West 120th
process. Journal of Social and Economic Development, 8(2): 105. Street, New York, NY 10027. E-mail:
The Landscape of Poverty in Nigeria: A Spatial Analysis [email protected]. His research interests include
Using Senatorial Districts-level Data
mapping and modeling approaches for geospatial
National Bureau of Statistics. 2010. National Bureau of
Statistics home. http://www.nigerianstat.gov.ng/. (last planning in areas of energy, infrastructure, health,
accessed November 20, 2014). environment and community planning.

View publication stats