Applied Geography 65 (2015) 1–12
Contents lists available at ScienceDirect
Applied Geography
journal homepage: www.elsevier.com/locate/apgeog
A spatio-temporal analysis of academic performance at the Basic
Education Certificate Examination in Ghana
David Ansonga,∗, Eric K. Ansongb, Abena O. Ampomahc, Stephen Afranied
aUniversity of North Carolina at Chapel Hill, 325 Pittsboro Street, Chapel Hill, NC 27599, United States
bUniversity of East London, United Kingdom
cDepartment of Social Work, University of Ghana, Legon, Ghana
dDepartment of Sociology, University of Ghana, Legon, Ghana
a r t i c l e i n f o a b s t r a c t
Article history: Over the last decade, Ghana has more than tripled investment in its basic education system. Conse-
Received 16 July 2015 quently, the country has made huge educational gains, primarily in providing universal access to basic
Received in revised form 6 October 2015 education. However, many stakeholders are worried that academic performance is lagging because of dis-
Accepted 6 October 2015
proportional attention to accessing basic education. Discussion of these concerns is hampered by ongoing
Available online 8 November 2015
disagreement about the true trajectory of academic performance at the basic education level and the
Keywords: widespread nature of students’ lagging academic performance. In part, this disagreement stems from the
Academic achievement failure of empirical studies to comprehensively examine trends in academic performance standards at the
Spatial modeling basic education level by concurrently considering a geographical and longitudinal perspective. Thus, this
Multilevel growth curve modeling study examines the spatio-temporal trends of academic performance at the junior high school level since
Educational inequalities 2009 by using multilevel growth curve modeling, spatial statistics, and district-level longitudinal data.
Ghana
Results reveal 3 statistically distinct trajectories of academic performance: erratic, accelerating, and de-
celerating changes. Results also show that rural–urban gaps explain 31% of the performance trajectories, a
trend which is expected to persist in the long term. In addition, we find extreme variations in academic
performance within rural areas. Given the varying trajectories and geographical variability in academic
performance, we suggest a localized approach to addressing challenges of low academic achievement at
the basic education level in Ghana.
© 2015 Elsevier Ltd. All rights reserved.
1 i
i
c
u B
( u
p n
t G
t p
G w
o u
i m
C l
u
h
0
. Introduction
Since the late 1980s, Ghana has substantially expanded its ed-
cational system and steadily increased investment in education
Akyeampong, Djangmah, Oduro, Seidu, & Hunt, 2007). Public ex-
enditure on education as a share of the country’s gross domes-
ic product (GDP) increased more than 600% from 1.8% in 1980
o 8.14% in 2011, which is well above the current average of 5%
DP for Sub-Saharan Africa (UNESCO, 2015). Ghana also embarked
n implementing major educational policies and programs, includ-
ng the free Compulsory Universal Basic Education (fCUBE) policy,
apitation Grant, and the School Feeding Program. The fCUBE pol-G
d
Abbreviations: JHS, Junior High School; SHS, Senior High School; BECE, Basic Ed-
cation Certificate Examination; fCUBE, Free Compulsory Universal Basic Education. s
∗ Corresponding author. i
E-mail addresses: ansong@email.unc.edu (D. Ansong), e.ansong@uel.ac.uk (
(E.K. Ansong), aoampomah@ug.edu.gh (A.O. Ampomah), afranie@ug.edu.gh C
(S. Afranie).
ttp://dx.doi.org/10.1016/j.apgeog.2015.10.003
143-6228/© 2015 Elsevier Ltd. All rights reserved.cy has focused primarily on increasing enrollment and improv-
ng physical infrastructure at the basic or primary school level that
omprises of the first 9 years of formal schooling (Nudzor, 2013).
ecause of the extent of attention and investment directed toward
niversal basic school education, the past two decades have wit-
essed substantial improvement in access to basic education in
hana (Darvas & Balwanz, 2014). These investments have sought to
rovide all Ghanaian children with basic education, putting Ghana
ell on track to achieving its goal of universal access to basic ed-
cation. This goal is also consistent with the international com-
unity’s cross-thematic development frameworks such as the Mil-
ennium Development Goals (MDGs) and Sustainable Development
oals (SDGs). MDG Goal 2 seeks to ensure that by 2015, all chil-
ren will be able to complete primary level schooling without re-
pect to their gender or the area or region in which they live. Sim-
larly, the SDGs push for inclusive and quality education for all
United Nations Economic Commission for Africa [UNECA], 2013).
learly, the improved access to basic education in Ghana is com-
2 D. Ansong et al. / Applied Geography 65 (2015) 1–12
m
f
d
W
f
w
f
(
o
i
c
p
e
t
b
2
e
t
i
v
e
s
2
o
t
u
p
s
r
2
d
d
p
a
e
t
A
g
a
(
a
i
i
i
a
G
t
g
s
s
s
h
p
t
i
s
d
m
s
v
mendable and deserving of the substantial investments the Ghana
government and donor agencies have made.
Despite the goal of the fCUBE policy to improve the access and
quality of basic education, improvements in learning outcomes at
this level have not kept pace with the remarkable improvement in
access (Darvas & Balwanz, 2014; Mettle-Nunoo & Hilditch, 2000).
Although stakeholders generally agree that academic performance
standards at the basic education level are low and of great concern
(Affum-Osei, Asante, & Forkuoh, 2014; Gyan, Mabefam, & Baffoe,
2014), they disagree on the true trajectory of learning outcomes.
This issue is clouded by inconsistent claims about whether perfor-
mance standards are falling, rising, or unchanged. The lack of clar-
ity on the true trajectory of primary level academic achievement is
due to, at least in part, most of the assertions being based on snip-
pets of results rather than comprehensive nationwide longitudinal
data.
Many in Ghana share the perception that academic achieve-
ment is falling (Degue, 2012; Okyerefo, Fiaveh, & Lamptey, 2011).
In particular, this perception becomes heightened each year with
the release of the annual Basic Education Certificate Examination
(BECE) results. BECE is the national standardized examination for
students who have completed junior high school (JHS) (i.e., Grade
9). The heightened perception of falling academic achievement is
in part because, for instance, 182,000 JHS candidates who sat for
the 2013 BECE did not earn a passing grade for at least one core
subject, and therefore, did not qualify to enter senior high school.
Importantly, some analysts have described this shared perception
of a downward trend in academic achievement as a national secu-
rity threat (Gyasiwaa, 2013).
Many of those who contend that academic achievement is
falling have also speculated that the expanded enrollment in the
basic education level has come at the cost of the quality of basic
education. As noted by Lewin and Akyeampong (2009), “… rapid
expansion in enrollments has degraded quality” (p.143). Similarly,
the United Nation’s 2014 MDG report asserted that Ghana’s ex-
panded access to education has steadily weakened the quality of
education, and linked the decline in educational quality and stu-
dent achievement to larger class sizes, the growing number of new
schools, and the government’s heavy reliance on poorly trained and
unqualified teachers (UNECA, 2014). However, significant increase
in educational investments does not necessarily have to lead to
a trade-off between enrollment (quantity) and learning outcomes
(quality). Indeed, data from the late 1980s to the early 2000s sup-
port the idea that Ghana’s massive investment in education sector
in the form of 8000 classroom blocks led to concurrent improve-
ments in enrollment and students’ learning outcomes. Why then is
there perception that Ghanaian students are on a downward slope
of academic achievement?
Simultaneously, many other stakeholders disagree with the as-
sertion that Ghanaian students’ academic performance is falling.
For instance, even the chief examiners’ report of the 2012 BECE
suggests that performance trends were mixed (West African Ex-
amination Council, n.d.). Other experts and educational scholars
contend that the perception of falling performance is false and,
not based on data. For instance, Francis Kodzo Amedahe, a pro-
fessor in educational measurement and statistics at the University
of Cape Coast, has contended that “… the issue of falling educa-
tional standards in Ghana is a perception rather than a reality”
(Ghana News Agency, 2014; para.1). Given conflicting viewpoints
on the academic performance trajectory of Ghana, there are impor-
tant empirical questions that warrant attention. The most pressing
unanswered questions include: (a) What are the factual trajectories
of academic performance standards at the basic education level?
(b) Are the academic performance trajectories generally consistent
across Ghana’s administrative districts, and are there salient spatial
variations in the direction and rate of change in academic perfor-ance?
This study aims to help fill this empirical gap in academic per-
ormance through systematic assessment of the spatio-temporal
ynamics of academic achievement in Ghana at the BECE level.
e examine 6 years’ of local and nationwide BECE data (collected
rom 2009 through 2014) to address fundamental questions on
hether (a) statistically significant trends exist in academic per-
ormance, (b) such trends are sloping upwards or downwards, and
c) whether these trends, if any, favor or disadvantage certain ge-
graphical areas of the country. Developing a better understand-
ng of the space-time trends of academic achievement in Ghana is
ritical because determination of the spatial trajectory of academic
erformance (and potential disparities) in the country will enable
ducation researchers, administrators, and policy makers to better
arget their energies to areas in need of attention, especially at the
asic education level.
. Rural–urban differences
Spatial inequality in Ghana is not a new phenomenon. How-
ver, concerns are growing among many scholars and practitioners
hat the extent of spatial inequality in education and other social
ndicators is widening despite the country’s economic growth, in-
estment in socioeconomic development, and falling poverty lev-
ls (Annim, Mariwah, & Sebu, 2012; Aryeetey, Owusu, & Men-
ah, 2009). Like many sub-Saharan African countries (Michaelowa,
004), significant rural–urban differences exist in Ghana in terms
f access to educational resources and the resultant outcomes of
hese resources. More often than not, the rural–urban gaps in ed-
cational outcomes favor urban areas because they have dispro-
ortionally more education resources (e.g., good classrooms and
chool furniture) and favorable living conditions such as good
oad and accommodation (Ansong, Ansong, Ampomah, & Adjabeng,
015; Kimosop, Otiso, & Ye, 2015; Senadza, 2012).
Despite the urban advantage, the caveat is that the rural–urban
ifferences may not always be a clear dichotomy. The educational
iscourses in Ghana often presume that the problem of spatial dis-
arities in educational outcomes is more acute in rural than urban
reas (Tsikata & Seini, 2004). However, in the absence of adequate
mpirical evidence from a contextualized and spatial perspective,
his presumption may be an over-simplification of the problem.
lthough a great deal is known about rural–urban inequalities re-
arding access to education and related resources, little is known
bout these inequalities in terms of specific educational outcomes
e.g., academic performance) given the scant nationwide empirical
ssessment of any such potential spatial disparities. Empirical stud-
es are yet to clarify the temporal nature of academic performance
n rural versus urban areas. One of the few studies to have exam-
ned the spatial dimension of learning outcomes used data from
random sample of 6000 junior high-school students from across
hana (Ansong & Chowa, 2013). The study found regional varia-
ions in educational performance when measured by math and En-
lish scores, but such variations favored predominantly rural areas
uch as the western and northern regions. However, because the
tudy focused on limited subject areas (i.e., only math and English
ubjects) and used cross-sectional data, its findings do not offer a
olistic overview of the country’s space-time trends in academic
erformance.
The absence of such a holistic overview is not unusual, and
he data used to shape conversations on growing spatial inequal-
ty are often incomprehensive, non-longitudinal, and inadequately
crutinized. To a large extent, public concerns about falling aca-
emic standards at the basic education level have been driven by
edia reports of the abysmal BECE performance of primary level
tudents. Although these reports are important in initiating con-
ersations about measures for addressing inequalities in learning
D. Ansong et al. / Applied Geography 65 (2015) 1–12 3
o d
r w
t a
B e
T g
d w
A i
6 d
a p
c c
s b
b c
s f
o t
o a
l
l g
t a
t d
e j
i a
c t
c w
b e
s a
i i
t t
b v
s
n 4
r
r 4
c
i
n f
s T
n w
a i
t s
B t
c t
a c
e i
e G
a
4
3
c
b
a m
G t
s (
& p
p m
c c
i i
t e
u t
u t
g w
O r
m
i g
r f
r r
t mutcomes, the external validity of such reports is often assumed
ather than verified. For instance, in 2012, media reports claimed
hat more than two-thirds of the 840 candidates who sat for the
ECE in the Sissala West District failed the exam (Farouq, 2012).
he Ghana News Agency (2010) also reported that not a single can-
idate from more than a dozen public and private schools in the
gona West Municipality scored high enough (i.e., aggregate scores
to 30) on the 2012 BECE examination to qualify for placement in
second-cycle institution (i.e., senior high, vocational, and techni-
al schools). While such media reports focus on select schools or
chool districts, they tend to skew public perception of the state of
asic education performance in Ghana. From a policy and research
tandpoint, it is critical to assess empirically whether the problem
f low academic performance of the BECE is ubiquitous in Ghana
r is confined to specific schools and regions.
Thus far, given what is known about educational outcomes by
ocation, enrollment, and educational progression, there is reason
o believe that substantial spatial variations exist in district-level
rajectories of academic performance. For instance, Akyeampong
t al. (2007) found that the education system favors students living
n urban districts. Specifically, Akyeampong and colleagues, found
hildren from urban areas were not only more likely than rural
hildren to pass the BECE and transition to senior high school,
ut that the rural–urban gap in school enrollment and progres-
ion was widening. Moreover, the rural–urban gap also manifests
n other ways that can affect educational outcomes. For example,
he disparity in the “educational climate” between rural and ur-
an areas was revealed by the 2010 Population and Housing Cen-
us, which showed that whereas 1 out of 7 urban residents had
ever attended school, that proportion was significantly higher in
ural areas at 1 out of 3 (Ghana Statistical Service, 2012). Further
ural–urban disparities exist in the quality of available school fa-
ilities and education offered. Thus, rural residents’ access to qual-
ty education is severely limited by poor infrastructure and the low
umbers of qualified teachers (Poku-Boansi & Amoako, 2014). Rural
chools often have inadequate classrooms, outdated or insufficient
umbers of textbooks, and poorly trained or untrained teachers—
ll of which exacerbate the relatively poor quality of education in
hese areas (Ansong et al., 2015; Atuahene & Owusu-Ansah, 2013).
ecause the Ghanaian society is socially, culturally, and economi-
ally heterogeneous (Avendal, 2011), it is reasonable to assume that
cademic performance trajectories vary across the country. How-
ver, compared to existing knowledge on access to education and
ducational resources in Ghana, little is known about spatial vari-
tions in academic performance.
. Inter-rural differences
Inter-rural variability (i.e., variations between rural areas) is
largely unexplored spatial dimension of rural deprivation in
hana. Historically, the rural–urban dichotomy has been empha-
ized in development theory and practice (Gough, Agergaard, Fold,
Moller-Jensen, 2009). However, even though rural Ghana is dis-
roportionately disadvantaged and has lower achievement out-
omes, the dynamics of rural deprivation and the associated learn-
ng outcomes are more intricate and nuanced than portrayed by
he traditional rural–urban divide literature. Further, the lack of
nderstanding of variability among rural areas has been perpet-
ated by the many research studies on rural Ghana that have ag-
regated all rural areas into a homogeneous category (Atuahene &
wusu-Ansah, 2013; Mettle-Nunoo & Hilditch, 2000).
The vast literature on rural challenges in Ghana notwithstand-
ng, questions remain about whether educational disadvantage in
ural areas is more acute in certain regions than others. Although
ural Ghana shares many qualities, it also varies considerably in the
ype and extent of deprivation experienced. For instance, most in-icators of human development (e.g., health, education, economic
ell-being) suggest that the three northern regions of Ghana have
more severe level of deprivation than either the middle or south-
rn regions (Otiso & Owusu, 2008; Seini, 2000). Given these re-
ional disparities and the structure of the educational system,
ith rural districts nested within regions, a particularly challeng-
ng question remains to be answered: Compared with the mid-
le and southern regions of Ghana, do distinct clusters of low-
erforming rural districts exist in the north? If so, how have these
lusters evolved and changed over time and space? While it may
e reasonable to infer that all rural areas are disadvantaged, espe-
ially those in the northern regions (Poku-Boansi & Amoako, 2014),
ew empirical analyses have been conducted on national space-
ime trends in academic performance. Thus, it is unclear whether,
cademic trajectories of the northern regions are unique or simi-
ar to those of the nation as a whole. To help fill this knowledge
ap, this paper uses multilevel growth curve modeling and spatial
nalysis to examine, (a) the 6-year (2009–2014) trajectory of aca-
emic achievement as measured by performance on the national
unior high school examinations; (b) whether any such academic
chievement trajectory is influenced by the rural/urban status of
he district, and if so, how such influence is manifested; and (c)
hether inequalities exist among rural districts in Ghana. The ben-
fit of combining multilevel growth curve modeling with spatial
nalysis in this study is the ability to measure change over time
n the BECE results at both the national and district levels and at
he same time geographically visualize the changes over time and
ariations across different areas of the country.
. Methods
.1. Data and analyses
The present study uses six waves of data (2009–2014) drawn
rom the BECE results of junior high school students in Ghana.
hese longitudinal data are aggregated at the school district level
ith varying sample sizes: 142 districts in 2009, 138 in 2010, 154
n 2011, 156 in 2012, 169 in 2013 and 166 in 2014. These sample
izes vary by year because of the creation of new districts during
he study period and non-availability of BECE data from a few dis-
ricts in 2010. Using two rigorous approaches— multilevel growth
urve modeling and spatial statistics analysis–the data were exam-
ned to determine the spatio-temporal curves of BECE pass rates in
hana.
.1.1. Multilevel growth curve modeling
The use of repeated measures data make multilevel growth
urve models an invaluable statistical tool in educational research
ecause this technique estimates changes in student outcomes
ore accurately by taking into account the hierarchical nature of
he data (i.e., nesting of repeated measures within individual units)
Hox & Stoel, 2005). Moreover, as compared with traditional ap-
roaches for analysis of longitudinal data, multilevel growth curve
odeling offers the advantage of being able to easily and effi-
iently handle unbalanced data, such as the varying sample sizes
n our data (Luke, 2004). We used multilevel growth curve mod-
ling to model the process of change in BECE pass rates over
ime, and to assess rural and urban differences in the initial sta-
us and growth rates. The time variable is the academic year (Year),
hich indicates the measurement occasion for each of the six BECE
esults starting from 2009. The variable academic achievement is
easured as the percentage of candidates who earned an aggre-
ated score between 6 and 30, which is considered a passing score
or admission to a senior high school (Mereku, 2012). To test for
ural–urban variation in BECE pass rates, we included a binary
easure of whether a district is predominantly rural or urban.
4 D. Ansong et al. / Applied Geography 65 (2015) 1–12
fi
a
a
fi
c
i
5
a
(
t
a
A
b
r
d
d
i
fi
4
u
c
a
e
fi
f
i
u
p
b
j
H
g
H
t
r
B
o
l
8
i
m
c
5
Fig. 1 presents a plot of the mean BECE pass rate over 6 con-
secutive years, starting from 2009. The figure depicts a nonlinear
growth curve; the percentage of candidates passing the BECE de-
creased from 2009 to 2010 and then increased between 2010 and
2012 before decreasing again between 2012 and 2014. Because of
these nonlinear trends in BECE results over time as shown in Fig. 1,
we modeled quadratic (Year2) and cubic slopes (Years3) (Shek &
Ma, 2011) to account for potential nonlinear individual growth tra-
jectories in the way in which districts in Ghana perform at the
BECE over time.
Based on the recommended analytical strategy for longitudinal
data analysis (Singer & Willett, 2003) we tested five nested mod-
els, starting with the most restrictive and progressing to the least
restrictive model. Model 1, the unconditional means model, exam-
ined possible differences across all districts in the mean percentage
of candidates who passed the exam, regardless of the exam year.
This model also assessed average district level variations in the
BECE pass rate and the amount of variance accounted for by the
inter-district differences as well as differences between years. The
intra-class correlation (ICC) was .49 [i.e., 208.06/(208.06 + 215.87)
see Table 1], suggesting that inter-district differences explained
49% of the variability in the BECE pass rate. The ICC is greater than
the recommended cutoff of .25, which further supports the need
to use individual growth curve modeling (Kreft, 1996).
Next, we modeled an unconditional linear growth curve model
(Model 2) to examine the existence of any significant differences
in the districts’ trajectory over time. Because the time variable was
statistically significant in Model 2 (i.e., the baseline model) we pro-
ceeded to examine two higher-order change trajectories. First, we
added a quadratic term (Year2) to test for higher-order change tra-
jectories over time. The resulting model (Model 3) was needed to
examine whether the individual rate of change accelerated or de-
celerated over time. In the next test for higher-order change trajec-
tory (Model 4) we added a cubic term (Year3) to capture the rate
of acceleration or deceleration over a long period of time. Model
5, the last in the five-step modeling process, tested the potential
moderation role of a district’s status as either predominantly ru-
ral or urban. In other words, this model tested whether the change
trajectories were statistically the same for rural and urban districts.
Because we did not expect the rural–urban designation of districts
to change significantly over the observation period, we examined
the rural–urban status as a time-invariant phenomenon. The model
is estimated as follows:
Yij = γ 00 + γ 01(Rural)j + γ 10(Year)ij + γ 11(Rural)j(Year)ij
+ γ 20(Year2)ij + γ 21(Rural)j(Year2)ij + γ 30(Year3)ij +
γ 331(Rural)j(Year )ij + u1j(Year)ij + u0j + rij
where Yij is the BECE pass rate of year i for district j, γ00 is the
adjusted average BECE rate across all districts, γ 01 to γ 31 are thed
t
t
r
t
t
t
a
t
t
t
a
m
t
Fig. 1. Mean BECE pass rate profiles of districts from 2009 to 2014. cxed effects, u1j and u0j are the two random effects (or error terms
t Level 2) of district j on BECE pass rate and rij is the error term
t Level 1.
To determine which model best fits the data, we compared the
t indices (−2 log likelihood [−2LL] and the Akaike information
riterion [AIC]) of the two models with the lowest AIC values, that
s, the cubic growth (Model 4) and predictor effect models (Model
) (Kuha, 2004; Shek & Ma, 2011). A pairwise comparison indicated
statistically significant difference between the two models: χ2
4) = 7434.42–7412.63 = 21.79, p < .001. The AIC difference be-
ween the lowest AIC model (predictor effect model AIC = 7436.63)
nd all other models (unconditional linear growth curve, Model 2
IC = 7651.19; quadratic growth curve Model 3 AIC = 7493.28; cu-
ic growth curve Model 4 AIC = 7450.42) were greater than the
ecommended cutoff of 10 (Burnham & Anderson, 2002). This AIC
ifference means that among all the nested models, only the pre-
ictor effect model (Model 5) explained any substantial variation
n the data. Thus, Model 5 (predictor effect) was retained as the
nal and best fitting model.
.1.2. Spatial pattern analysis
Spatial pattern analysis was the second analytical approach we
sed to examine the data. This method is a nontraditional statisti-
al tool for mapping and quantifying statistically significant trends
nd clusters in geographically referenced data. Specifically, we gen-
rated thematic maps and conducted grouping analyses. The bene-
ts of the spatial analyses include the ability to (a) validate results
rom the multilevel growth curve models, (b) map and visualize
dentifiable growth trajectories, and (c) highlight inter-rural, inter-
rban, and rural–urban variations.
First, we used the Grouping Analysis tool, a relatively new ex-
loratory method in ArcGIS, to identify distinct groups of districts
ased on the trajectories of the BECE pass rate. To avoid pre-
udging the number of distinct trajectories, we used the Calinski–
arabasz pseudo F-statistic to determine the maximum number of
eographically diverse trajectories that exist in the data (Calinski &
arabasz, 1974). Lastly, we used six consecutive choropleth maps
o reveal the spatio-temporal variations that exist within Ghana’s
ural areas based on the percentage of candidates who passed the
ECE from 2009 to 2014. To make the map easy to interpret with-
ut excessive loss of detail, we used graded colors with a five-
evel equal interval classification scheme—that is, 20%, 40%, 60%,
0%, and 100%—for each wave of data. This visualization scheme
s based on the recommended four to six classes used in most
aps (Yang, 2005). We also computed and mapped out the annual
hanges in BECE pass rates across rural areas.
. Results
The mean academic performance profiles for rural and urban
istricts and the national average are displayed in Fig. 1. Overall,
he downward direction of the trend for both rural and urban dis-
ricts is similar to the national trend. The fairly parallel profile for
ural and urban districts suggests no interaction between district
ype and time. In other words, rural districts did not catch up with
he urban districts during the 6-year observation period nor did
he urban ones fall to the level of the rural areas.
Fig. 2 builds on Fig. 1 by illustrating the rural–urban gaps
nd their statistical significance at the national level from 2009
o 2014. The extent of this gap remained statistically significant at
he .01 and .001 significance levels over the six-year period. Al-
hough the rural–urban gap narrowed marginally after 2009, the
verage performance of urban districts in subsequent years re-
ained significantly higher than rural districts. As Fig. 2 shows,
he largest differences between the rural and urban districts oc-
urred in 2009 (15.07%, t = 4.42, p < .001) followed by 2014
D. Ansong et al. / Applied Geography 65 (2015) 1–12 5
Table 1
Results of multilevel growth curve models.
Model 2
Model 1 Unconditional linear growth Model 3 Model 4 Model 5
Unconditional mean model curve model Quadratic growth curve model Cubic growth curve model Predictor effect model
b(SE) b(SE) b(SE) b(SE) b(SE)
Fixed effects
Intercept (γ ) 47.76(1.21)∗∗∗00 63.26(1.51)∗∗∗ 44.73(2.02)∗∗∗ 62.58(3.29)∗∗∗ 80.56(6.97)∗∗∗
Year (γ 10) −4.24(.26)∗∗∗ 9.35(1.03)∗∗∗ −12.93(3.42)∗∗∗ −21.04(7.20)∗∗
Year2 (γ ∗∗∗20) −1.92(.13) 5.38(1.08)∗∗∗ 7.52(2.26)∗∗
Year3 (γ 30) −.69(.10)∗∗∗ −.86(.21)∗∗∗
Rural (γ 01) −23.15(7.88)∗∗
Year× Rural (γ 11) 10.37(8.18)
Year2 × Rural (γ 21) −2.73(2.57)
Year3 × Rural (γ 31) .22(.24)
Random effects
Residual (rij) 215.87(11.11)
∗∗∗ 143.65(8.49)∗∗∗ 106.59(6.34)∗∗∗ 99.10(5.89)∗∗∗ 98.47(5.86)∗∗∗
Intercept (u ) 208.06(27.29)∗∗∗0j 228.85(43.25)∗∗∗ 264.62(42.97)∗∗∗ 273.27(43.03)∗∗∗ 244.33(39.73)∗∗∗
Year (u ) 2.31(1.48) 6.38(1.69)∗∗∗1j 6.67(1.66)∗∗∗ 6.80(1.67)∗∗∗
Model Fit
−2LL 7904.53 7639.19 7479.28 7434.42 7412.63
AIC 7910.53 7651.19 7493.28 7450.42 7436.63
AICc 7910.56 7651.28 7493.40 7450.57 7436.97
BIC 7925.02 7680.17 7527.09 7489.05 7494.59
Note. −2LL = −2 log likelihood; Akaike information criterion; AICc = corrected Akaike information criterion; BIC= Bayesian information criterion.
∗∗∗p < .001, ∗∗p < .01, ∗p < .05.
( i
2 a
t
t
5
c
e
s
t
B
(
T
T
t
e 5
q
p
c t
t C
c v
2 T
d
t
l
c
F
b
12.70%, t = 3.88, p < .001). The smallest difference occurred in
013 (11.11%, t = 2.99, p < .01).
.1. Multilevel growth curve modeling results
Results of the final model presented in Table 1 show that
he intercept (b = 80.56, p < .001) and linear growth parameter
b = −21.04, p < .001) were both highly statistically significant.
his suggests that the initial status and the linear growth rate of
he BECE pass rate changed over time. This finding means that the
stimated initial average BECE pass rate decreased after 2009. The
uadratic effect was positive and statistically significant (b = 7.52,
< .01) but the cubic effect was negative and statistically signifi-
ant (b = −.85, p < .001). These results, which are consistent with
he mean profiles in Fig. 1, show that the pass rate initially de-
reased in 2010, increased in 2011, and then decreased again in
013 and 2014. However, the rate of deceleration gradually slowed
own over time.
The rural–urban status of the districts was a significant predic-
or of the initial BECE pass rate (b = −23.15, p < .01) but not the
inear (b = −10.37, p = .21), quadratic (b = −2.73, p = .29), or
ubic (b = −23.15, p = .37) changes in the BECE pass rate. Thatig. 2. Illustration of statistically significant annual differences in BECE pass rate F
etween rural and urban districts from 2009 to 2014. 2s, rural and urban districts had a statistically similar quadratic
nd cubic rate of change in the BECE results. The rural–urban sta-
us of districts explained 31% [(143.65–98.47)/143.65 = .3145] of
he inter-district variations in the BECE pass rate. The negative
orrelation between the intercept and the linear growth param-
ter (b = −17.54, p < .01) means that districts with high BECE
cores had a slower linear decrease, whereas districts with low
ECE scores had a faster decrease in their linear growth over time.
hus, the situation of the low scoring districts got worse over time.
.2. Spatial grouping analysis results
The parallel box plot in Fig. 3 illustrates the optimal dis-
inct trajectories in the BECE results from 2009 to 2014. The
alinski–Harabasz pseudo F-statistic for the grouping analysis re-
ealed three geographically-diverse trajectories (pseudo F = 74.87).
hese three trajectories are analogous to accelerating, decelerat-ig. 3. A parallel box plot summarizing the global trajectories in BECE results from
009 to 2014.
6 D. Ansong et al. / Applied Geography 65 (2015) 1–12
t
5
t
t
b
u
P
a
u
r
b
i
o
c
d
t
5
r
h
B
I
l
u
f
t
t
o
r
p
d
s
i
t
a
t
t
6
a
t
b
t
t
y
t
L
f
c
s
a
C
o
S
h
B
T
2
ing, and erratic changes. The first statistically distinct trajectory, il-
lustrated with a star–green line, reflects districts whose BECE re-
sults are neither improving nor declining consistently. Although
this group had relatively steady results over the years, their group
mean hovered below the global mean throughout the 6-year ob-
servation period. The second statistically distinct trajectory is il-
lustrated with a triangular point red line. The line shows districts
whose BECE pass rates have decreased consistently since 2009. Al-
though still above the global mean, this group’s mean fell below
the global upper quartile for the first time in 2014. The last statis-
tically significant trajectory is illustrated with a circular point blue
line that shows districts whose results increased steadily during
the 6-year period.
Throughout the study period, the BECE pass rate scores for this
group (i.e., districts with steady increases over time) were greater
than the global mean and global upper quartile. The parallel box
plot also shows that the number of outliers increased over time;
meaning that the number of districts with exceptionally good re-
sults increased over the 6 years.
Fig. 4 presents the localized results of the grouping analy-
sis, and shows the three statistically significant unique trajectories
identified using the ArcGIS Grouping Analysis tool. The green dis-
tricts have relatively stable trajectories. One noticeable result from
the map is that majority of districts in the northern regions are
neither improving nor declining in their BECE performance. The
red districts experienced steady decreases in BECE pass rates over
the study period. A disproportionate number of districts with de-
clining trajectories were located in the southern regions. The blue
districts experienced steady improvement in the BECE pass rates
over the 6-year study period, and were located in the southern re-
gions, especially in southwest Ghana. The unshaded white areas
indicate districts with growth trajectories that did not fit any sta-
tistically distinguishable pattern for growth trajectories.
Overall, Fig. 4 shows that most districts in northern and eastern
Ghana are neither improving nor declining compared to those in
southwest Ghana. The southern region is the most diverse because
its districts have a noticeable variety of trajectories. Specifically, the
Ashanti Region in the south is the most promising region in terms
of the number of districts with an accelerating growth trajectory
or improving performance. In contrast, the Western Region is the
worst performing region given its high number of districts with a
declining BECE performance trajectory.
5.3. Inter-rural spatial variability
Fig. 5 shows the spatial distribution of Ghanaian rural districts
by academic performance from 2009 to 2014. Each panel, particu-
larly before 2014, illustrates the dynamic nature of the BECE pass
rates, which suggests some rural areas outperform others. The spa-
tial distribution shows a wide but shrinking inter-rural variation in
the BECE pass rate over time. In other words, the differences that
exist between rural districts have narrowed over time. However,
the results show that the primary reason for the narrowing gap be-
tween rural districts is due largely to the fact that many districts
that used to perform well on the BECE have experienced gradual
decline in their BECE results. That is, the darker shades have pro-
gressively gotten lighter, which suggests worsening trends in rural
areas.
Fig. 6 shows the annual changes in rural area performance from
2009 to 2014. Panels a through e show that some rural districts ex-
perienced yearly increases in their pass rates (blue shades) while
others experienced declines (red shades). The five panels show
clear patterns or clusters, thereby confirming variability within ru-
ral areas. The maps also confirm results from Fig. 5 that the nar-
rowing inter-rural gap is not because low-performing districts have
improved, but rather well-performing districts have declined inheir performance on the BECE.
.4. Inter-urban spatial variability
Trends in the inter-rural comparison is similar to trends in
he inter-urban comparison. As panels a through f in Fig. 7 show,
he spatial distribution of academic performance in Ghanaian ur-
an districts varies. The spatial distribution shows moderate inter-
rban variations in the BECE pass rate during the study period.
anels e and f in Fig. 7 show that the inter-urban differences have
lso narrowed over time, but there is still a fair amount of inter-
rban variation compared to the inter-rural variation.
Fig. 8 shows the pattern of annual changes in the BECE pass
ate in Ghana’s urban areas. Panels a through c show that the ur-
an areas of Ghana experienced a mix of improvement and decline
n their BECE results from 2009 through 2012. However, the extent
f inter-urban variation shrunk from to 2012 to 2014 largely be-
ause more urban districts experienced decline in their BECE result
uring the last two years of the study period. This trend is similar
o the trends observed among rural districts.
.5. Rural–urban spatial variability
There are rural–urban similarities and differences in BECE pass
ates. Results from Fig. 9 show that both rural and urban districts
ave representation of the three statistically distinct trajectories of
ECE pass rate in Ghana: improving, declining, and stable trends.
n addition, both rural (31%) and urban (32%) areas have equiva-
ent proportions of districts whose trajectories are not statistically
nclassified. On the other hand, there are many rural–urban dif-
erences in Ghana’s BECE pass rate. First, most of the rural dis-
ricts are located in the northern and middle sectors of the coun-
ry, whereas nearly all the urban districts are located in the south
f the country. Second, as evident in panels a and b of Fig. 9, ru-
al districts are largely neither improving nor declining (58%) com-
ared to urban districts (41%). Third, a higher proportion of urban
istricts showed consistent signs of improvement (24%) during the
tudy period compared to a handful of rural districts (5%). Another
mportant rural–urban difference in performance at the BECE is
hat the declining districts, though few, are somewhat spread out
mong rural districts in the south, middle, northern Ghana, while
hose of the urban districts are clustered in the south, particularly
he southwestern part.
. Discussion
The goal of this study was to model changes in the percent-
ge of candidates who passed the BECE over a 6-year period and
o determine whether the pattern of change over time differed
ased on the urban or rural status of Ghana’s administrative dis-
ricts. At the national level, the BECE performance trends suggest
hat the initial results followed a fluctuating pattern until the last 3
ears when they declined consistently. The recent trend at the na-
ional level, lends credence to concerns expressed by Degue (2012),
ewin and Akyeampong (2009), and others on the falling BECE per-
ormance and academic standards in Ghana. Our results are also
onsistent with Ghana’s poor performance at several international
chooling rankings. For instance, in a 2015 global report on math
nd science education published by the Organization for Economic
o-operation and Development (OECD), Ghana placed last behind
ther African countries such as Tunisia, Morocco, Botswana, and
outh Africa (Darko, 2015). Ghana has also consistently lagged be-
ind other peer African countries, including South Africa, Morocco,
otswana, and Egypt in the 2003, 2007, and 2011 editions of the
rends in International Math and Science Study (TMISS) (Abukari,
010; Etsey et al., 2009).
D. Ansong et al. / Applied Geography 65 (2015) 1–12 7
Fig. 4. Results of grouping analysis showing spatial variability in the growth trajectories in BECE results from 2009 to 2014.
m
i s
e i
p c
s i
g T
i f
t s
t r
d p
p G
C
m m
w i
l t
i c
n c
t eThe clear implication of our findings and those of other studies
s that the Ghanaian government and other education stakehold-
rs should worry about the trajectory of the country’s academic
erformance. The results of our spatial grouping analysis demon-
trate that of the 120 districts statistically classified into distinct
roups, a significant majority (66%) experienced neither consistent
mproving nor declining BECE performance. Moreover, these dis-
ricts consistently performed below the national average during
he study period. In other words, this group, which dominates the
istricts in most of the regions in Ghana (Upper East = 90%, Up-
er West = 78%, Northern = 60%, Volta = 70%, Eastern = 56%, and
entral = 79%), has not experienced meaningful, steady improve-
ent in BECE performance since 2009. This finding confirms the
idely held view that academic standards at the basic education
evel in Ghana remain low (Gyan et al., 2014). Moreover, this find-
ng is alarming because the districts in the stable group represent
early two-thirds of districts in Ghana and have a large number of
he country’s junior high school students. While internal migrationay be a predictive factor for the observed trends, emigration of
tudents from low-performing districts may not possibly be a dom-
nant reason why low-performing areas continue to do poorly be-
ause internal migration between rural and urban areas in Ghana
s not highly skewed in one direction (Anarfi, Kwankye, Ababio, &
iemoko, 2003; Castaldo, Deshingkar, & McKay, 2012). In fact, data
rom the fifth round of the Ghana Living Standards Survey (GLSS 5)
how that majority of internal migrants move largely towards ru-
al areas (Ghana Statistical Service, 2008). Certainly, this persistent
oor performance at the BECE points to persistent weaknesses in
hana’s efforts to improve its basic education system.
In the face of such poor performance at the BECE, reassess-
ent of the current education programs is warranted. The BECE
tself may need to be evaluated on whether it is currently the op-
imal assessment tool for academic performance at the JHS level. A
ross-country assessment between Ghana and other West African
ountries such as Gambia, Liberia, and Sierra Leone that use the
quivalent of the BECE as an assessment tool would be useful to
8 D. Ansong et al. / Applied Geography 65 (2015) 1–12
Fig. 5. Variability in percentage of candidates in deprived areas that earned a pass grade at the BECE from 2009 to 2014.
Fig. 6. Annual changes in BECE pass rates in deprived areas from 2009 to 2014.
Fig. 7. Variability in percentage of candidates in non-deprived areas that earned a pass grade at the BECE from 2009 to 2014.
Fig. 8. Annual changes in BECE pass rates in non-deprived areas from 2009 to 2014.
1
m
o
2
t
a
&
e
determine whether the problem of low performance at the JHS
level is largely peculiar to Ghana or is a problem with the assess-
ment format. Education stakeholders and administrators of exam-
inations in Anglophone countries of West Africa already recognize
that external examination should not be the only measure of learn-
ing outcomes at the JHS level; thus, they reformed the BECE nearly
a decade after its inception to take into account teachers’ contin-
uous assessment of students’ in-class performance (Akyeampong,997). The ratio of the external exam to the continuous assess-
ent, which started from 60:40, is now 70:30, and there are rec-
mmendations to change to a 50:50 ratio (Dery & Addy-Lamptey,
010). There is similar ongoing debate in several African coun-
ries about the appropriate weighting system for the continuous
ssessment and external examination (Kapambwe, 2010; Kellaghan
Greaney, 2003; Pudaruth et al., 2013). It is important that the
nsuing decisions to replace or augment the one-shot external ex-
D. Ansong et al. / Applied Geography 65 (2015) 1–12 9
Fig. 9. Rural–urban spatial variability in the BECE results based on grouping analysis (a) Rural (b) Urban.
a h
n h
u (
a e
t e
2
t e
o g
i d
n t
d t
i m
( t
v t
m
m i
s T
t a
g c
c fi
I n
d h
t f
a h
a m
s
t N
r e
y (
t t
m c
t t
F B
a
t m
O Gminations with more classroom and school-based assessments do
ot overlook the fact that the inter-rural, inter-urban, and rural–
rban disparities in teaching and infrastructural (i.e., classroom
nd school) conditions could have implications on performance at
he BECE across different geographical areas.
Besides the need to reassess the BECE weighting format, addi-
ional government actions may be needed to address the problem
f low performance at the BECE and the rural–urban differences. It
s not as though government and stakeholders have completely ig-
ored the problem in the past decade. In fact, government expen-
iture on basic education has tripled in the past decade, yet learn-
ng outcomes at the junior high school level have not improved
Darvas & Balwanz, 2014). Perhaps, the government’s education in-
estment choices have targeted the access-challenges to the detri-
ent of academic quality concerns. Notably, the introduction of
ajor educational initiatives in Ghana such as the Free Compul-
ory Universal Basic Education (FCUBE) policy in 1996, the Capi-
ation Grant Scheme in 2004, and the Ghana School Feeding Pro-
ram in 2005, and free school uniforms in 2009 have had signifi-
ant impact on universal access to basic education (CREATE, 2008).
t is however, not clear the extent to which these programs have
irectly affected the lagging performance standards. It is essential
hat government, policy makers, and stakeholders focus adequate
ttention on performance standards, similar to the efforts directed
t investments in access basic education.
Another key finding of this study is the varying space-time pat-
erns of academic performance at the BECE in the 2009–2014 pe-
iod. In contrast to the national trends, in-depth district-level anal-
sis shows more nuanced variations in BECE results than the global
rends suggest. Significant geographical variations exist in perfor-
ance trends, which is likely due to spatial disparities in access
o pedagogical resources and infrastructure (Ansong et al., 2015).
or some districts, mostly in the central, eastern, Volta, northern,
nd upper regions of Ghana, the proportion of candidates passing
he BECE remained relatively stable over the 6-year study period.
ther districts, most of which are in the Ashanti (southern) region,ave consistently experienced annual increases. The Ashanti region
ouses Ghana’s second largest city and a network of urban centers
Otiso & Owusu, 2008), hence the region is well-endowed with
ducation resources, such as a high proportion of trained teach-
rs, which is critical for learning outcomes (Ministry of Education,
013). In contrast, another group of districts mostly in the west-
rn (i.e., Nzema East, Aowin-Suaman, and Bia) and Brong Ahafo re-
ions (i.e., Tano South, Pru, and Jaman South) experienced steady
ecline over the 6-year study period. The only glimmer of hope for
he academic performance of these downward-trending districts is
hat they have not yet fallen below the national average. This opti-
istic assessment gives local authorities a window of opportunity
o thoroughly explore localized, targeted interventions to reverse
heir decline in BECE performance.
We also found that some districts, the majority of which are
n southern Ghana, did not fit any of the three distinct profiles.
hese districts are not in the blue, red, or green zones of Figs. 4
nd 9, which means they have not consistently experienced de-
lines, improvements, nor stability in their BECE performance. This
nding does not mean that the statistically unclassified districts do
ot have academic performance challenges. Rather, those districts
ave unpredictable trajectories and this uncertainty can hamper ef-
ective educational planning. The fact that those districts do not
ave an upward trajectory (i.e., do not show consistent improve-
ent) require attention because their BECE performance does not
eem to be commensurate with increased education investment.
oticeably, most of the statistically unclassified districts in south-
rn Ghana (i.e., 76% in Fig. 4) are adjacent to blue-shaded districts
i.e., districts with ideal growth trajectories). Because of data limi-
ations, our study is not able to offer insights into the reasons for
ontrast between contiguous districts. It will be important for fu-
ure studies to investigate the reasons for the sharp contrasts in
ECE performance between contiguous districts.
In the normal scheme of things, variability in academic perfor-
ance should not necessarily be a cause for concern. However, in
hana’s case, stakeholders should be concerned given the conspic-
10 D. Ansong et al. / Applied Geography 65 (2015) 1–12
t
e
N
(
i
t
i
c
f
c
u
t
E
6
i
t
f
w
p
o
f
t
t
o
t
p
c
i
s
s
7
v
e
p
d
f
c
t
t
f
t
i
2
a
g
c
t
t
t
i
o
z
t
a
l
R
A
uous decline of some districts on the BECE. Moreover, even more
worrisome is the long-range forecast for Ghana’s academic perfor-
mance. Specifically, the multilevel growth curve modeling used in
this study revealed higher-order polynomial trends (i.e., quadratic
and cubic slopes) that suggest the low-performing districts, partic-
ularly the red-shaded districts, are unlikely to improve their per-
formance at the BECE in the near future. In other words, with-
out more effective targeted programs and interventions, the worse-
performing school districts will continue to get worse, further
widening the educational inequality gap, which has a huge impact
on individual welfare. Government, policy makers, and stakehold-
ers cannot afford to ignore this potential for further decline in per-
formance at the BECE.
The potential for further decline can be reversed if serious lo-
calized measures are undertaken to improve performance stan-
dards in the low-performing districts. Evidence from a growing
body of empirical studies and program reports from Ghana and
other African countries suggest that decentralized measures that
devolves power to schools and school districts fosters commu-
nity and parental involve which are associated with improved aca-
demic outcomes (Dowd, 2001; Hyde, Kadzamira, Sichinga, Chib-
wana, & Ridker, 1997; Odonkor, 2000). Results from our study sug-
gest that poor academic performance at the BECE is most likely a
local issue because multiple trajectories exist within each region
or even within the rural clusters shown in Fig. 6. The localized
nature of the academic performance problem is precisely the rea-
son why southwestern Ghana not only has the highest number of
districts with positive trends (blue-shaded districts) but also has
a disproportionally large share of poor-performing districts (red-
shaded districts). There are promising districts within rural clus-
ters, as shown in Fig. 6, because some rural districts have experi-
enced improvements in BECE pass rates from time-to-time. It will
be important for future studies to examine and glean lessons from
the unique characteristics and success stories of districts such as
Kintampo South and Kassena Nankana West that are deprived ar-
eas that have managed to experience a steady upward trajectory in
the proportion of candidates passing the BECE.
Another finding of this study is that the noticeable spatio-
temporal patterns in academic performance at the BECE generally
follow the lines of rural/urban divide, which is in line with findings
from a similar study in Kenya (Kimosop et al., 2015) and Uganda
(Kasirye, 2009). Results of our multilevel growth curve model-
ing revealed that rural districts experienced disproportionate BECE
performance declines whereas urban districts largely experienced
increased fortunes. As shown in Fig. 2, whereas the rural–urban
gap narrowed marginally after 2009, a substantial gap still remains
partly due to the well documented rural–urban gap in socioe-
conomic development and educational resources (e.g., textbooks,
trained teachers, physical infrastructure, and school supervision)
that is in favor of urban areas (see Otiso & Owusu, 2008; Poku-
Boansi & Amoako, 2014). Worryingly, forecasts from our higher-
order polynomial trends suggest that if nothing is done, the rural–
urban BECE achievement gap will continue to widen in the future.
Therefore, the critical question is whether stakeholders can inter-
vene in ways that will enable rural areas to narrow the gap with
urban districts. Although daunting, such improvement in rural aca-
demic performance is possible given the right intervention.
The practical way forward should perhaps begin with the ad-
mission that a blanket global view of academic standards in Ghana
masks the needs of certain areas that require additional localized
interventions beyond the general efforts of government. This ob-
servation has important implications for the approaches that pol-
icy makers could use to address local challenges in BECE perfor-
mance. Spatially targeted approaches to addressing the education
challenges in Ghana are needed. Evidence suggests that interven-
tions that target specific marginalized communities produce bet-er outcomes. One such intervention is the Education for Empow-
rment program, which is a partnership between IBIS, a Danish
on-governmental organization, and the Ghana Education Services
GES). The program, which seeks to bridge the educational inequal-
ty gap between rural and urban Ghana has yielded positive educa-
ional outcomes (Darvas & Balwanz, 2014). Other spatially-oriented
nterventions that have the potential to improve educational out-
omes in deprived communities include the GES’s supplemental
unding to basics schools in deprived districts; the government’s
ash transfer program to support poor households’ health and ed-
cational expenditure; and incentives such as salary increments for
eachers in deprived districts (Darvas & Balwanz, 2014; Ministry of
ducation, 2012).
.1. Limitations and strengths
Two data-related limitations of this study are worth mention-
ng. First, data on newly created districts were not available, and
herefore, the present study could not examine the academic per-
ormance of these districts. Moreover, this study modeled only six
aves of BECE academic performance data, starting from 2009. Ex-
anding the data to include BECE pass rates prior to 2009 might
ffer further insights into the long-term trends in academic per-
ormance. However, earlier waves of data were not available at the
ime the present study was conducted. Despite these limitations,
his study has the potential to advance the existing body of work
n academic performance in Ghana in two important ways. First,
he ability to model six waves of BECE data is an important im-
rovement over the prior use of cross-sectional data to inform dis-
ussions about academic performance trends. Second, the analyt-
cal approach of combining growth curve modeling with spatial
tatistics is a superior approach that offers rich insights into the
patial and temporal dimension of academic performance.
. Conclusion
Ghana has outpaced most developing countries in education in-
estments and expenditure, and has made huge gains in access to
ducation, but quality concerns remain. While the results of the
resent study validate the concerns of some stakeholders that aca-
emic standards are generally low, the trajectory of academic per-
ormance shown herein is more nuanced, with some districts de-
lining, improving, or reaming stable. An important lesson from
his study is that there are diverse geographical trajectories, and
herefore, a localized approach to addressing the academic per-
ormance challenges at the junior high school level in Ghana is
he prudent approach. As discussed earlier, the patterns of change
n the percentage of candidates passing the BECE from 2009 to
014 differs not only between the rural and urban districts but
lso within rural districts. This study cautions that barring any tar-
eted intervention, rural–urban gap in performance at the BECE
ould widen. Government, researchers, education stakeholder and
he media should also discuss the problem from a spatial perspec-
ive because some districts need additional attention and support
o reverse their downward trend at the BECE. If social inequality
n Ghana is to be addressed, then spatial disparities in the quality
f education must be addressed in order to offer all young citi-
ens better and equitable educational opportunities. Such a situa-
ion would ultimately go a long way in enhancing the well-being
nd quality of life of the citizenry as well as ensuring sustainable
ivelihoods.
eferences
bukari, Z. (2010). Risk and protective factors associated with academic achievement
among Ghanaian youth (PhD. Dissertation). University of Denver.
D. Ansong et al. / Applied Geography 65 (2015) 1–12 11
A G
A
H
A
H
A
K
A K
A
K
K
A
K
A K
A L
A L
M
B
C M
C M
M
C
D
M
D
N
D
O
D
O
D O
E P
F P
G
S
G S
S
G
G S
G T
G
U
ffum-Osei, E., Asante, E. A., & Forkuoh, K. S. (2014). Perceived stress and academic
performance of senior high school students in western region, Ghana. European
Journal of Business and Social Sciences, 2(11), 88–101.
kyeampong, A. K. (1997). Continuous assessment in post-secondary teacher training
in Ghana: A case study evaluation (BL) (Order No. U091298) Available from Pro-
Quest Dissertations & Theses Global. (301516848) http://search.proquest.com/
docview/301516848?accountid=14244 Accessed 29.09.15.
kyeampong, K., Djangmah, J., Oduro, A., Seidu, A., & Hunt, F. (2007). Access to basic
education in Ghana: The evidence and the issues Available online at http://sro.
sussex.ac.uk/1872/1/Ghana_CAR.pdf Accessed 15.07.15.
narfi, J., Kwankye, S., Ababio, O., & Tiemoko, R. (2003). Migration from and
to Ghana: A background paper Development Research Centre on Migra-
tion, Globalization and Poverty, Working Paper C4. Brighton: University of
Sussex.
nnim, S. B., Mariwah, S., & Sebu, J. (2012). Spatial inequality and household poverty
in Ghana. Economic Systems, 36(4), 487–505.
nsong, D., Ansong, E. K., Ampomah, A. O., & Adjabeng, B. K. (2015). Factors con-
tributing to spatial inequality in academic achievement in Ghana: analysis of
district-level factors using geographically weighted regression. Applied Geogra-
phy, 62, 136–146. http://dx.doi.org/10.1016/j.apgeog.2015.04.017.
nsong, D., & Chowa, G. A. N. (2013). Visual demonstration of academic performance
and parental involvement in Ghana (YouthSave research brief 13–17). St. Louis, MO:
Washington University, Center for Social Development Available online at http:
//csd.wustl.edu/Publications/Documents/RB13-17.pdf Accessed 22.07.15.
ryeetey, E., Owusu, G., & Mensah, E. J. (2009). An analysis of poverty and regional
inequalities in Ghana Working Paper 27. New Delhi, India: Global Development
Network.
tuahene, F., & Owusu-Ansah, A. (2013). A descriptive assessment of higher educa-
tion access, participation, equity, and disparity in Ghana. SAGE Open, 3(3), 1–6.
http://dx.doi.org/10.1177/2158244013497725.
vendal, C. (2011). Social work in Ghana: engaging traditional actors in professional
practices. Journal of Comparative Social Work, 2011(2), 1–19. Available online at
http://www.soch.lu.se/en/christel-avendal/publication/4195224 .
urnham, K. P., & Anderson, D. R. (2002). Model selection and multi-model inference:
a practical information-theoretic approach (2nd ed.). Secaucus, NJ, USA: Springer.
alinski, T., & Harabasz, J. (1974). A dendrite method for cluster analysis. Communi-
cations in Statistics, 3, 1–27. http://dx.doi.org/10.1080/03610927408827101.
astaldo, A., Deshingkar, P., & McKay, A. (2012). Internal migration, remittances and
poverty: Evidence from Ghana and India Working Paper 7. University of Sussex,
Migrating out of Poverty Research Programme Consortium.
REATE (2008). Educational access in Ghana. Country policy brief Retrieved from
http://www.create-rpc.org/pdf_documents/Ghana_CAR_policy_brief.pdf .
arko, S. (2015). Asia tops biggest global school rankings Available online at http://
www.bbc.com/news/business-32608772 (13.05.2015).
arvas, P., & Balwanz, D. (2014). Basic education beyond the millennium development
goals in Ghana: How equity in service delivery affects educational and learning
outcomes (World bank document No. 82726). Washington, DC: World Bank.
egue, J. (2012, September 13). Reversing the decline in academic perfor-
mance at the BECE level in Volta region: The role of stakeholders. Modern
Ghana News Available online at http://www.modernghana.com/news/417607/
1/reversing-the-decline-in-academic-performance-at-t.html Accessed 15.07.15.
ery, R. G., & Addy-Lamptey, W. (2010). Effects of classroom assessment scores on
the final scores used in grading students at senior high school in Ghana. Pa-
per presented at the 36th International Association for Educational Assessment
(IAEA) Annual Conference held in Bangkok Thailand, from Sunday 22 to Friday
27 August 2010 Available online at http://selectscore.com/fullpaper/127.pdf
(24.09.15).
owd, A. J. (2001). The impact of community support for education on standard 2
mathematics learning in Malawi. Paper presented at the 2001 Annual Conference
of the Comparative and International Education Society, Washington, DC.
tsey, K., Smith, T. M., Gyamera, E., Koka, J., de Boer, J., Havi, E., et al. (2009). Ba-
sic education in Ghana: Progress and problems. U.S. Washington DC: Agency for
International Development.
arouq, M. U. (2012, September 12). Ghana: 591 students fail BECE in Sissala West.
The Ghanaian Chronicle Available from All Africa Global Media website http://
allafrica.com/stories/201209130757.html Accessed 15.07.15.
hana News Agency (2010, September 23). Ten schools in Agona West and East
score zero in BECE Available online at http://www.ghananewsagency.org/
education/ten-schools-in-agona-west-and-east-score-zero-in-bece-20696 Ac-
cessed 15.07.15.
hana News Agency (2014, April 13). Falling educational standards in Ghana is a
perception—Prof Amedahe Available online at http://www.ghananewsagency.
org/education/falling-educational-standards-in-ghana-is-a-perception-prof-
amedahe-73531 Accessed 15.07.15.
hana Statistical Service (2008). Ghana living standards survey report of the firth
round (GLSS 5). Accra: Ghana Statistical Service.
hana Statistical Service (2012). 2010 population and housing census: Summary report
of final results. Accra: Ghana Statistical Service.
ough, K. V., Agergaard, J., Fold, N., & Moller-Jensen, L. (2009). Conceptualizing
and performing comparison of rural-urban dynamics. In J. Agergaard, N. Fold,
& K. Gough (Eds.), Rural-urban dynamics: Livelihoods, mobility and markets in
African and Asian frontiers (pp. 9–22). Florence, KY: Routledge.
yan, C., Mabefam, M. G., & Baffoe, M. (2014). Push out or drop out? taking a crit-
ical look at the poor performance and drop-out of students of the JSS/JHS pro-
gramme in Ghana. Academic Journal of Interdisciplinary Studies, 3(1), 409–420.
http://dx.doi.org/10.5901/ajis.2014.v3n1p409.yasiwaa, A. (2013, October 3). 182,000 failed BECE candidates a national
security threat: A security analyst Available online at http://www
.myjoyonline.com/news/2013/October-3rd/182000-failed-bece-candidates-a-
national-security-threat-security-analyst.php Accessed 15.07.15.
ox, J., & Stoel, R. D. (2005). Multilevel and SEM approaches to growth curve mod-
eling. In B. S. Everitt, & D. C. Howell (Series Eds.), Encyclopedia of statistics in
behavioral science: Vol. 3 (pp. 1296–1305). Chichester, England: John Wiley &
Sons, Ltd.
yde, K., Kadzamira, E., Sichinga, J., Chibwana, M., & Ridker, R. (1997). Village based
schools in Mangochi, Malawi: An evaluation. (IPR working paper series: Determi-
nants of educational achievement and attainment in Africa). Washington, DC: In-
stitute for Policy Reform.
apambwe, W. M. (2010). The implementation of school based continuous assess-
ment (CA) in Zambia. Educational Research and Reviews, 5(3), 99–107.
asirye, I. (2009). Determinants of learning achievement in Uganda. Economic pol-
icy research Centre, Uganda Available online at http://www.csae.ox.ac.uk/
conferences/2009-EdiA/papers/325-Kasirye.pdf Accessed 24.09.15.
ellaghan, T., & Greaney, V. (2003). Monitoring performance: Assessment and exami-
nations in Africa. Paris: ADEA.
imosop, P. K., Otiso, K. M., & Ye, X. (2015). Spatial and gender inequality in the
Kenya certificate of primary education examination results. Applied Geography,
62, 44–61. http://dx.doi.org/10.1016/j.apgeog.2015.04.006.
reft, I. G. (1996). Are multilevel techniques necessary? An overview including simula-
tion studies. Los Angeles: California State University [Unpublished manuscript].
uha, J. (2004). AIC and BIC comparisons of assumptions and performance.
Sociological Methods Research, 33(2), 188–229. http://dx.doi.org/10.1177/
0049124103262065.
ewin, K. M., & Akyeampong, K. (2009). Education in sub-Saharan Africa: research-
ing access, transitions and equity. Comparative Education, 45(2), 143–150. http:
//dx.doi.org/10.1080/03050060902920492.
uke, D. A. (2004). Multilevel modeling. Thousand Oaks, CA: Sage http://dx.doi.org/
10.4135/9781412985147.
ereku, K. (2012). Should BECE mark the end of JHS education? Available from
the Ghana Schoolsnet website: http://www.ghanaschoolsnet.com/forum/topics/
should-bece-mark-the-end-of-jhs-education accessed 15.07.15.
ettle-Nunoo, R. J., & Hilditch, L. (2000). Donor participation in the education sector
in Ghana. Accra, Ghana: ActionAid.
ichaelowa, M. (2004). Quality and equity of learning outcomes in Francophone
Africa, Monograph. Montreal: UNESCO Institute for Statistics.
inistry of Education (2012). Education strategic plan 2010 to 2020: Vol. 1. Policies,
Strategies, Delivery, Finance Available online at http://www.globalpartnership
.org/content/government-ghana-education-strategic-plan-2010-2020-volume-1-
policies-strategies-delivery Accessed 24.09.15.
inistry of Education (2013). Education sector performance report, 2013. Republic of
Ghana. Ministry of Education Available online at http://www.moe.gov.gh/assets/
media/docs/FinalEducationSectorReport-2013.pdf Accessed 9.09.15.
udzor, H. P. (2013). Unearthing the discursive shift in the ‘fCUBE’ policy im-
plementation in Ghana: using critical discourse analysis. International Journal
of Research & Method in Education, 36(2), 179–201. http://dx.doi.org/10.1080/
1743727X.2012.696242.
donkor, M. (2000). Final evaluation of care international’s school and community ori-
ented education project. Frontier Analysis Consulting Associates.
kyerefo, M. P. K., Fiaveh, D. Y., & Lamptey, S. N. L. (2011). Factors
prompting pupils’ academic performance in privately owned junior high
schools in Accra, Ghana. International Journal of Sociology and Anthropol-
ogy, 3(8), 280–289. Available from http://www.academicjournals.org/article/
article1379512112_Okyerefo%20et%20al.pdf .
tiso, K. M., & Owusu, G. (2008). Comparative urbanization in Ghana and
Kenya in time and space. GeoJournal, 71, 143–157. http://dx.doi.org/10.1007/
s10708-008-9152-x.
oku-Boansi, M., & Amoako, C. (2014). Dimensions of spatial inequalities in Ghana-
ian cities. Journal of Geography and Regional Planning, 8(5), 131–142. http://dx.
doi.org/10.5897/JGRP2014.0477.
udaruth, S., Moloo, R., Chiniah, A., Sungkur, R., Nagowah, L., & Kishnah, S. (2013).
The impact of continuous assessments on the final marks of computer science mod-
ules at the University of Mauritius. Proceedings of Global Engineering, Science and
Technology Conference 3–4 October 2013, Bay View Hotel, Singapore.
eini, W. (2000, September 11). The deprived communities in Ghana: Rural and urban
Available from the All Africa website http://allafrica.com/stories/200009110127.
html Accessed 15.07.15.
enadza, B. (2012). Education inequality in Ghana: gender and spatial dimen-
sions. Journal of Economic Studies, 39(6), 724–739. http://dx.doi.org/10.1108/
01443581211274647.
hek, D. T. L., & Ma, C. M. S. (2011). Longitudinal data analyses using linear mixed
models in SPSS: concepts, procedures and illustrations. The Scientific World Jour-
nal, 11, 42–76. http://dx.doi.org/10.1100/tsw.2011.2.
inger, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis. New York, NY:
Oxford Press http://dx.doi.org/10.1093/acprof:oso/9780195152968.001.0001.
sikata, D., & Seini, W. (2004). Identities, inequalities and conflicts in Ghana (CRISE
working paper 5). Oxford, England: Centre for Research on Inequality, Human
Security and Ethnicity, Department of International Development, University of
Oxford Available from http://www3.qeh.ox.ac.uk/pdf/crisewps/workingpaper5.
pdf.
NESCO Institute for Statistics (2015). Public expenditure on education as % of GDP.
UNESCO Institute for Statistics Data Center Available online at http://data.un.
org/Data.aspx?d=UNESCO&f=series3AXGDP_FSGOV Accessed 15.07.15.
12 D. Ansong et al. / Applied Geography 65 (2015) 1–12
W
Y
United Nations Economic Commission for Africa (2013). Africa regional consul-
tative meeting on the sustainable development goals (Outcome Document).
Addis Ababa, Ethiopia: Economic Commission for Africa Available from
https://sustainabledevelopment.un.org/content/documents/2878Africa-SDGs%
20Outcome%20Doc%20ENG.pdf Accessed 06.16.15.
United Nations Economic Commission for Africa, African Union, African Devel-
opment Bank, & United Nations Development Programme (2014). MDG re-
port 2014: Assessing progress in Africa toward the millennium development
goals. Addis Ababa, Ethiopia: Economic Commission for Africa Available online
at http://www.afdb.org/fileadmin/uploads/afdb/Documents/Publications/MDG_
Report_2014_11_2014.pdf Accessed 15 July1205.est African Examination Council. (n.d.). April 2012 Basic Education Certificate Ex-
amination: General resumé of the chief examiners’ reports on the standard of the
papers. Available from http://www.ghanawaec.org/Exams/examiner_BECE.aspx.
ang, T. C. (2005). Choropleth map GIS Resource Document 05–72 (GIS_RD_05-72).