Does economic geography matter for Pakistan? A spatial exploratory analysis of income and education inequalities.
Ahmed, Sofia
1. INTRODUCTION
From the industrial revolution to the emergence of the so-called
knowledge economy, history has shown that economic development has taken
place unevenly across regions. A region's economy is a complex mix
of varying types of geographical locations comprising different kinds of
economic structures, infrastructure, and human capital. In this context
recent literature in regional sciences has highlighted how crucial it is
to analyse socio-economic phenomena in the light of spatial concepts
such as geography, neighbourhood, density, and distance [Krugman (1991);
Krugman and Venebles (1995); Quah (1996); Baldwin, et al. (:2003); van
Oort (2004); Kanbur and Venebles (2005); World Development Report
(2009)]. Keeping these recent developments in view, this paper
identifies, measures, and models the temporal relationship between
space, economic inequalities, human development, and growth for the case
of Pakistan. (1) Specifically, by using data at district level from 1998
and 2005, it utilises spatial exploratory techniques to determine the
effect of distance and contiguity among 98 of Pakistan's
administrative districts on their human capital characteristics and
inequalities. (2) This way it provides some of the first spatially
explicit results for clustering of socioeconomic characteristics across
Pakistani districts. (3)
Most of the existing research on Pakistan's socio-economy is
based on a provincial level, and it neglects the role of social
interactions the districts within the provinces. (4) This paper in
particular investigates whether spatial clustering of income and average
education levels can explain their distribution across Pakistani
districts. District level research has become even more important as
Pakistan has taken a major step towards fiscal decentralisation with the
enactment of the 18th Constitutional Amendment. Moreover the 7th
National Finance Commission Award has allowed the transfer of more funds
from the federation to the provinces which now have more authority over
the provision of health, educational and physical infrastructure
facilities. This fundamental shift towards the division of power between
the centre and the provinces bears significant implications for the
country's long term policy planning, management and implementation.
As education and other public and social services become the sole domain
of the provinces, there is a need for increased research at the district
level.
Furthermore, Pakistan is also characterised with spatial
disparities between its key socio-economic characteristics such as
education, health, physical infrastructure, etc. [Burki, et al (2010)].
While some districts have state of the art physical and human capital
infrastructure, others have made little or no progress at all. This
phenomenon is in line with the findings of the World Bank's World
Development Report (2009) that has demonstrated how and why the
clustering or concentration of people and production usually takes place
in particular favourable areas (coasts, cities, etc.) during the growth
process in any country. For the case of Pakistan, the most developed
districts are located in Northern and Central Punjab. It has been noted
that Pakistani districts with a population density of more than 600
persons per square km are characterised by industrial clusters, superior
education and health infrastructure and better sanitation facilities
that serve as attractive pull factors, e.g., Karachi, Lahore, Peshawar,
Charsadda, Gujranwala, Faisalabad, Sialkot, Mardan, Islamabad, Multan,
Swabi, Gujrat and Rawalpindi [Khan (2003)]. On the other hand, districts
with lowest population densities (or those having below 30 persons per
square km) are characterised by prevalence of various push factors such
as; absence of job opportunities due to lower education and health
facilities, poor agricultural endowments, barren or mountainous
topography, and lack of limited presence of industrial units [Khan
(2003)]. Moreover, the fact that the highly (and medium) concentrated
districts (except for Swat and Muzzaffargarh) are mostly clustered
around metropolitan cities of Karachi and Lahore [Burki, et aL (2010)]
demonstrates that a district's human and economic development is
being shared by its neighbouring districts, confirming that economic
geography matters for Pakistan.
In the light of the above mentioned issues, this study empirically
investigates the spatial clustering of economic inequalities, growth and
development across Pakistani districts by utilising ESDA techniques. The
paper is organised as follows: Section 2 describes the data; Sections 3
and 4 provide a detailed overview of the methodology utilised; Section 5
presents the empirical results; finally Section 6 discusses the policy
and methodological implications of the empirical results and concludes.
2. DATA
For district wise average earnings income and education levels,
this paper utilises micro data from the Pakistan Social and Living
Standards Measurement survey (PSLM) 2004-05. It is the only
socio-economic micro data that is representative at the provincial and
at the district level. Moreover, the sample size of the district level
data is also substantially larger than the provincial level data
contained in micro data surveys such as Household Income and Expenditure
Survey (HIES) of Pakistan and the Labour Force Survey (LFS) of Pakistan.
This has enabled researchers to draw socioeconomic information which is
representative at lower administrative levels as well. The survey for
2004-05 provides district level welfare indicators for a sample size of
about 76,500 households. It provides data on districts in all four
provinces of Pakistan namely; Punjab, Sindh, Khyber Pakhtunkhwa (KP),
and Balochistan. The federally administered tribal areas (FATA region)
along the Afghan border in the north-west and Azad Kashmir are not
included in the data.
To analyse the spatial differences in district wise primary,
secondary, and bachelor's education levels over time, this article
has utilised the district level data from the 1998 Population Census of
Pakistan. Since the data from PSLM (2004-05) is statistically comparable
with the Pakistan Census Data (1998) the two data sets together provide
a decent gap of 7 years to analyse the temporal changes in income and
development characteristics across Pakistan.
Finally, for investigating spatio-temporal differences in district
wise income, GDP growth rate, and human development levels, this paper
has taken its data from the National Human Development Report (2003) and
from Jamal and Khan (2008). Note that all income data from 2004-05 was
deflated using the Pakistani Consumer Price Index (CPI) of 1998.
3. METHODOLOGY
Due to the abundance in data collected at a provincial or a
rural/urban disaggregation, most socio-economic studies on Pakistan, are
a province based analysis. Pakistani provinces however have extreme
'within' diversity in terms of their economic structures,
development levels, cultures, language, natural resources and geography.
Hence regional policy making requires analysing socio-economic issues at
an even smaller geographical disaggregation. For this reason, the
spatial unit of analysis chosen for this study is the
'districts' of Pakistan. In terms of geographical
disaggregation Pakistan (excluding the Federally Administered Tribal
Area (FATA) region and Azad Kashmir) has 4 levels consisting of 4
provinces (Punjab, Sindh, Khyber Pakhtunkhwa (KP), and Balochistan), 107
districts, 377 sub-districts, and 45653 villages. A lower level unit of
analysis is not being used because of two main reasons. Firstly, data on
regional scales below the district level in Pakistan suffers from
reliability issues. The second issue is more technical. In order to give
information on 45,653 villages of Pakistan instead of 107 districts, the
project would need a matrix of distance with 45,653 x (45,653 + 1)/2 =
1,042,121,031 free elements to be evaluated, hence the utilisation of
district level data. Due to data constraints, this article analyses 98
out of 107 districts in Pakistan (see Table A1).
3.1. Spatial Economic Analysis and Spatial Effects
A fundamental concept in geography is that proximate locations
often share more similarities than locations far apart. This idea is
commonly referred to as the 'Tobler's first law of
geography' [Tobler (1970)]. Classical statistical inference such as
conventional regressions are inadequate for an in-depth spatial analysis
since they fail to take into account spatial effects and problems of
spatial data analysis such as spatial autocorrelation, identification of
spatial clusters and outliers, edge effects, modifiable areal unit
problem, and lack of spatial independence [Arbia, Benedetti, and Espa
(1996); Beck, Gleditsch, and Beardsley (2006); Franzese and Hays
(2007)]. (5) Moreover, as an uneven distribution of socio-economic
economic characteristics is shaping the economic geography of most
countries, spatial analysis also has increasing policy relevance [World
Development Report--WDR (2009)]. These reasons together necessitate the
use of spatial exploratory and explanatory methods that can explicitly
take spatial effects into account.
Spatial analysis investigates the presence (or absence) spatial
effects which can be divided into two main kinds: spatial dependence and
spatial heterogeneity. Spatial heterogeneity refers to the display of
instability in the behaviour of the relationships under study. This
implies that parameters and functional relationships vary across space
and are not homogenous throughout data sets. Spatial dependence on the
other hand, refers to the lack of independence between observations
often present in cross sectional data sets. It can be considered as a
functional relationship between what happens at one point in space and
what happens in another. If the Euclidean sense of space is extended to
include general space (consisting of policy space, inter-personal
distance, social networks etc.) it shows how spatial dependence is a
phenomenon with a wide range of application in social sciences. Two
factors can lead to it. First, measurement errors may exist for
observations in contiguous spatial units. The second reason can be the
use of inappropriate functional frameworks in the presence of different
spatial processes (such as diffusion, exchange and transfer, interaction
and dispersal) as a result of which what happens at one location is
partly determined by what happens elsewhere in the system under
analysis.
3.2. Quantifying Spatial Effects
Spatial dependence puts forward the need to determine which spatial
units in a system are related, how spatial dependence occurs between
them, and what kind of influence do they exercise on each other.
Formally these questions are answered by using the concepts of
neighbourhood expressed in terms of distance or contiguity.
Boundaries of spatial units can be used to determine contiguity or
adjacency which can be of several orders (e.g., first order contiguity
or more). Contiguity can be defined as linear contiguity (i.e., when
regions which share a border with the region of interest are immediately
on its left or right), rook contiguity (i.e. regions that share a common
side with the region of interest), bishop contiguity (i.e. regions share
a vertex with the region of interest), double rook contiguity (i.e. two
regions to the north, south, east, west of the region of interest), and
queen contiguity (i.e. when regions share a common side or a vertex with
the region of interest) [LeSage (1999)]. Other common conceptualisations
of spatial relationships include inverse distance, travel time, fixed
distance bands, and k-nearest neighbours.
The most popular way of representing a type of contiguity or
adjacency is the use of the binary contiguity [Cliff and Ord (1973,
1981)] expressed in a spatial weight matrix (W). In spatial econometrics
W provides the composition of the spatial relationships among different
points in space. The spatial weight matrix enables us to relate a
variable at one point in space to the observations for that variable in
other spatial units of the system. It is used as a variable while
modelling spatial effects contained in the data. Generally it is based
on using either distance or contiguity between spatial units. Consider
below a spatial weight matrix for three units:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Where [w.sub.12] or [w.sub.ij] may be the inverse distance between
two units i and j or it may be 0 and 1 if they share a border or a
vertex. The W matrix displays the properties of a spatial system and can
be used to gauge the prominence of a spatial unit within the system. The
usual expectation is that values at adjacent locations will be similar.
3.3. The Spatial Weight Matrix for Pakistan
The choice of the W matrix representation and its conceptualisation
has to be carefully based on theoretical reasoning and the historical
factors underlying the concept or phenomenon under study.
This paper has employed two W matrices for Pakistan. (6) The first
matrix is a simple binary contiguity W matrix (referred to as BC matrix
from now onwards) based on the concept of Queen Contiguity i.e. if a
district i shares a border or a vertex with another district j, they are
considered as neighbours, and [w.sub.i,j] takes the value 1 and 0
otherwise. This matrix is also zero along its diagonal implying that a
district cannot be a neighbour to itself. Hence it is a symmetric binary
matrix with a dimension of 98 x 98 (98 being the total number of the
districts being analysed). This matrix precisely tells us the influence
of geographically adjacent neighbours on each other. A simple binary
contiguity matrix is a standard starting point and its influence is
often compared with other types of W matrices.
The second W matrix developed for Pakistan is one based on inverse
average road distance from a district i to the nearest districtj which
has a 'large city' in it (referred to as ID matrix from now
onwards). Out of the 98 districts being studied there are only 14 that
come under the category of a district with a 'large size' city
as per the classification of the coding scheme for the PSLM survey.
These include Islamabad as the federal capital city; Lahore, Faisalabad,
Rawalpindi, Multan, Gujranwala, Sargodha, Sialkot, and Bahawalpur as
districts with a 'large size' city in Punjab; Karachi,
Hyderabad and Sukkur in Sindh, Peshawar in the Khyber Pakhtunkhwa and
Quetta in Balochistan. This matrix is a symmetric non-binary matrix,
again with a dimension of 98 x 98.
The reason for selecting road distance instead of train distance as
is normally done in most studies on regional analysis is that in
Pakistan, the road network is much better developed than the railway
network. As a result, Pakistan's transport system is primarily
dependent on road transport which makes up 90 percent of national
passenger traffic and 96 percent of freight movement every year
(Economic Survey of Pakistan, 2007-08). Inverse distance matrices have
more explanatory power as partitions of geographic space especially when
the phenomenon under study involves the exchange or transfer of
information and knowledge (in our case income and education). It
establishes a decay function that weighs the effect of events in
geographically proximate units more heavily than those in geographically
distant units. Since a country is not a plain piece of land, Euclidean
distance calculations or distance as 'the crow flies' make
little economic sense when we are trying to investigate the effect of
distance from districts with a large city on regional human development
characteristics. The effect of the density of country's
infrastructure network is an important influence for which reason road
distances have been utilised. For this reason this paper has utilised
the inverse of the average of the maximum and the minimum roads distance
between a district and its nearest district with a 'large
city'.
Finally both the matrices are row-standardised, which is a
recommended procedure whenever the distribution of the variables under
consideration is potentially biased due to errors in sampling design or
due to an imposed aggregation scheme.
4. EXPLORATORY SPATIAL DATA ANALYSIS
Exploratory spatial analysis aims to look for "associations
instead of trying to develop explanations" [Haining (2003), p.
358]. This article applies exploratory spatial data analysis (ESDA)
techniques to district wise data on income, education, growth and
development levels in order to detect the presence of spatial
dependence. ESDA describes and visualises spatial distributions,
"identifies spatial outliers, detects agglomerations and local
spatial autocorrelations, and highlights the types of spatial
heterogeneities" [van Oort (2004); Haining (1990); Bailey and
Gatrell (1995); Anselin (1988); Le Gallo and Ertur (2003)]. The
particular ESDA techniques employed in this study include the
computation of Moran's 1 and Geary's C spatial autocorrelation
statistics. They demonstrate the spatial association of data collected
from points in space and measures similarities and dissimilarities in
observations across space in the whole system [Anselin (1995)]. However
due to the presence of uneven spatial clustering, the Local Indicators
of Spatial Association which measure the contribution of individual
spatial units to the global Moran's I statistic have also been
utilised (Ibid). The results are illustrated using Moran scatter plots
that have been generated to demonstrate the spatial distribution of
district wage and education levels across Pakistan.
4.1. Measures of Spatial Autoeorrelation
(i) Global Spatial Autocorrelation
Spatial autocorrelation occurs when the spatial distribution of the
variable of interest exhibits a systematic pattern [Cliff and Ord
(1981)]. Positive (negative) spatial autocorrelation occurs when a
geographical area tends to be surrounded by neighbours with similar
(dissimilar) values of the variable of interest. As previously
mentioned, this paper utilises two measures Moran's I and
Geary's C statistics to detect the global sigatial autocorrelation
present in the data. (7) The Moran's I is the most widely used
measure for detecting and explaining spatial clustering not only because
of its interpretative simplicity but also because it can be decomposed
into a local statistic along with providing graphical evidence of the
presence of absence of spatial clustering.
It is defined as:
I = n/[s.sub.0] x [[summation].sup.n.sub.i]
[[summation].sup.n.sub.j][w.sub.i,j] ([y.sub.i] - [[bar.y]]) ([y.sub.j]
- [[bar.y]]] / [[summation].sup.n.sub.i][([y.sub.i] - [[bar.y]).sup.2]
(1)
Where [[gamma].sub.i] is the observation of variable in location i,
[bar.y]] is the mean of the observations across all locations, n is the
total number of geographical units or locations, [w.sub.ij] is one of
the elements of the weights matrix and it indicates the spatial
relationship between location i and location j.
[S.sub.0] is a scaling factor which is equal to the sum of all the
elements of the W matrix:
[S.sub.0] = [[summation].sup.n.sub.i][[summation].sup.n.sub.j]
[w.sub.ij] ... (2)
[S.sub.0] is equal to n for row standardised weights matrices
(which is the preferred way to implement the Moran's I statistic),
since each row then adds up to 1. The first term in Equation (1) then
becomes equal to 1 and the Moran's I simplifies to a ratio of
spatial cross products to variance.
Under the null hypothesis of no spatial autocorrelation, the
theoretical mean of Moran's I is given by:
E (I) = -1/(n-I) ... (3)
The expected value is thus negative and will tend to zero as the
sample size increases as it is only a function of n (the sample size).
Moran's I ranges from -1 (perfect spatial dispersion) to +1
(perfect spatial correlation) while a 0 value indicates a random spatial
pattern. If the Moran's I is larger than its expected value, then
the distribution of y will display positive spatial autocorrelation i.e.
the value of y at each location i tends to be similar to values of y at
spatially contiguous locations. However, if I is smaller than its
expected value, then the distribution of y will be characterised by
negative spatial autocorrelation, implying that the value of y at each
location i tends to be different from the value of y at spatially
contiguous locations. Inference is based on z-values computed as:
[Z.sub.I] = [I - E(I)]/[sd(I)] (4)
i.e. the expected value of I is subtracted from I and divided by
its standard deviation. The theoretical variance of Moran's I
depends on the assumptions made about the data and the nature of spatial
autocorrelation. This paper presents the results under the randomisation
assumption i.e. each value observed could have equally occurred at all
locations, (8) Under this assumption [z.sub.I] asymptotically follows a
normal distribution, so that its significance can be evaluated using a
standard normal table [Anselin (1992a)]. A positive (negative) and
significant [Z.sub.I] value for Moran's I accompanied by a low
(high)p-value indicates positive (negative) spatial autocorrelation. (9)
The second measure of spatial autocorrelation that has been
utilised is the Geary's C which is defined as:
C - (N-1)[[summation].sub.i]
[[summation].sub.j][w.sub.ij][([X.sub.i] - [X.sub.j]).sup.2]]/ 2W
[[summation].sub.i][([X.sub.i] - [bar.X].sup.2] (5)
where N is the number of spatial units (districts in our case); X
is the variable of interest; [w.sub.ij] represents the spatial weights
matrix, where W is the sum of all [w.sub.i,j]. The value of Geary's
C lies between 0 and 2. Under the null hypothesis of no global spatial
autocorrelation, the expected value of C is equal to 1. If C is larger
(smaller) than 1, it indicates positive (negative) spatial
autocorrelation. Geary's C is more sensitive to local spatial
autocorrelation than Moran's I. Inference is based on z-values,
computed by subtracting 1 from C and dividing the result by the standard
deviation of C:
[Z.sub.c] = [c-1]/[sd(c)] ... (6)
The standard deviation of C is computed under the assumption of
total randomness, implying that [Z.sub.c] is asymptotically distributed
as a standard normal variate [Anselin (1992a); Pissati (2001)].
Finally, the results of the Moran's I and Geary's C are
dependent on the specification of the weights matrix. Although
interpretations change depending on whether the matrix was based on the
use of physical distance or economic distance, a "pattern of
decreasing spatial autocorrelation with increasing orders of contiguity
(distance decay) is commonly witnessed in most spatial autoregressive
processes regardless of the matrix specification" [van Oort (2004),
p. 109].
(ii) Local Spatial Autocorrelation
Since the Moran's I and Geary's C are global statistics
based on simultaneous measurements from many locations, they only
provide broad spatial association measurements, ignore the location
specific details, and do not identify which local spatial clusters (or
hot spots) contribute the most to the global statistic. As a remedy,
local statistics commonly referred to as 'Local Indicators of
Spatial Association (LISA)'are used along with graphic
visualisation techniques of the spatial clustering such as a
Moran's Scatterplot [Fotheringham, et al. (2000); Haining (2003)].
The Moran scatterplot is derived from the global Moran I statistic.
Recall that the Moran's I formula when we use a row standardised
matrix can be written as:
I = [[summation].sup.n.sub.i]([y.sub.i] - [bar.y])
([[summation].sup.n.sub.i] [w.sub.i,j] ([y.sub.j] - [bar.y])) (7)
This is similar to the formula for a coefficient of the linear
regression b, with the exception of ([[summation].sup.n.sub.i]
[w.sub.i,j] ([y.sub.j] - [bar.y])), which is the so-called spatial lag
of the location i.
Therefore I is formally equivalent to the regression coefficient in
a regression of a location's spatial lag (Wz) on the location
itself. This interpretation is used by the Moran's scatterplot,
enabling us to visualise the Moran's I in a scatterplot of Wz
versus z, where z = [Y.sub.i] - [bar.y])/([y.sub.i]). Moran's I is
then the slope of the regression line contained in the scatterplot. A
lack of fit in this scatterplot indicates local spatial associations
(local pockets/non-stationarity). This scatterplot is centred on 0 and
is divided in four quadrants that represent different types of spatial
associations.
5. EMPIRICAL RESULTS
5.1. Spatial Autocorrelation Estimates for District-wise Earnings
Income Inequality Levels
Our first empirical estimation involves calculating measures of
spatial dependence for district income inequality (measured as Gini
coefficient of average district earnings income) in the year 2004-05.
Table 1 provides the results of Moran's I statistic and
Geary's C statistic for district income inequality levels using the
two weight matrices. In both the cases, the null hypothesis of no
spatial dependence of income inequality between districts is rejected at
the significance level of 1 percent as the measures demonstrate a weakly
positive spatial autocorrelation amongst district inequality levels
(0.21 under BC matrix specification and 0.25 under ID matrix
specification). The results for Geary's C statistic have been
reported in Table A2a in the Appendix. This implies that income
inequality in one district is not strongly spatially associated with
income inequality in its neighbouring districts in the case of Pakistan.
5.2. Local Spatial Association between District-wise Income
Inequality Levels
The Moran scatterplot provides a more disaggregated view of the
nature of the global autocorrelation. It not only provides us
information on the presence of clusters in the data but also on the
outliers contained in it (see Figure 1). This scatterplot is divided
into four quadrants, each of which represents a different type of
spatial association. The upper right quadrant (High-High zone)
represents spatial clustering of a district with a high level of the
variable under study (income inequality in our case) around neighbours
that also have high values of income inequality as demonstrated by the
high values of both, the Z-score and the Wz (the spatial lag). The upper
left quadrant (Low z - High Wz zone) represents spatial clustering of a
district with a low level of income inequality with neighbouring
districts that have a high income inequality levels. The lower left
quadrant (Low z - Low Wz zone) represents spatial clustering of a
district with a low income level around neighbours that also have low
incomes. The lower right quadrant (High z - Low Wz zone) represents
spatial clustering of a high income inequality district with neighbours
that have low income inequality levels.
Figure 1 illustrates the results obtained in Col I of Table 1 via a
Moran scatterplot for Gini coefficient of district per capita incomes
using the binary contiguity weights matrix. It shows a positive global
Moran's I (z-score = 2.98), which is represented by the slope of
the black line. Due to the weakly positive spatial autocorrelation, we
are unable to detect any substantial clusters of high (or low)
inequality districts in particular for the year 2005. Similarly, Figure
A8 (see Appendix) also shows a Moran scatterplot for Gini coefficient of
district per capita incomes, however it has utilised an inverse distance
weights matrix instead. The overall spatial autocorrelation is although
statistically significant, it still remains weak.
[FIGURE 1 OMITTED]
5.3. Spatial Association between District-wise Education Levels
The role of human capital in generating growth is important since
the distribution of income is mainly driven by the distribution of human
capital within a country [Golmm and Ravikuman (1992); Saint-Paul and
Verdier (1993); Galor and Tsiddon (1997)]. Hence the operation of human
capital externalities and knowledge spillovers plays an important role
in generating regional dependencies and disparities. It has been
demonstrated that regions located in an economic periphery experience
lower returns to skill attainment and hence have reduced incentives for
human capital investments and agglomerations. However spatial
externalities do not spread without limits [Darlauf and Quah (1999)] as
a result of which closely related economies or regions tend to have
similar kinds of human capital externalities and technology levels as
compared to the more distant ones [see Quah (1996); Mion (2004)]. This
section investigates the spatial disparities in education levels across
Pakistan, the extent to which neighbouring districts share similar
levels of education, and examines whether district human development
level inequalities are spatially associated.
In order to do so, this paper uses the average district wise
education attainment level (which is measured as the average number of
schooling years completed in a district) as a proxy for human capital.
It is expected that neighbours of districts with high education
attainment should also have high educational awareness and hence similar
if not equal attainment levels. Again the Moran's I global and
local indices along with a Moran scatterplot and Geary's C
statistic have been utilised.
Our results indicate that there exists a greater possibility of
knowledge spillovers between districts that share a border, as compared
to when they do not (see Table 2). The global Moran's I for average
district education level (measured as the average education attainment
of a district's citizens) is positive and statistically significant
when neighbourhood is defined in terms of contiguity, however it is
negative and statistically insignificant when neighbourhood is defined
in terms of proximity. These results imply that for a Pakistani
district, sharing a border with a district whose individuals have a high
(low) education level, 'may' result in rising (lowering) its
own education levels.
The positive pattern for spatial autocorrelation for average
district education levels demonstrated by the BC matrix shows more
clusters with low education levels (in the case of Balochistan) and high
education levels (in the case of Punjab) as compared to outliers.
Districts in northern Punjab emerge in the High-High quadrant and
confirm our assumption about high human capital districts being located
close to each other (Figures 2 and A5). Similar empirical findings have
also been put forward in a recent study on agglomeration patterns of
industries across Pakistani districts in a study by Burki and Khan
(2010).
[FIGURE 2 OMITTED]
The neighbouring districts of Karachi and Thatta emerge as the most
significant outliers when we analyse the local Moran's I values
using the BC and the ID matrices. While Karachi fails into the High-Low
zone, Thatta falls in the Low-High zone. However, the fact that being a
neighbour with Karachi (a district with one of the highest average
education levels in Pakistan) does not translate in Thatta having
improved human capital characteristics is not very surprising. Regional
science and regional economics literature has demonstrated that the
economic influence and knowledge spillover effects of coastal cities
(such as Karachi) are quite different from the pattern of spillovers
generated by landlocked regions [Glaeser, et al. (1992); Henderson
(2003)]. The overall spatial pattern of autocorrelation is quite
diffused when we use the ID matrix for analysis (see Figure A5). However
under both the neighbourhood structures Rawalpindi, Abbottabad, Chakwal
and Jhelum emerge as a statistically significant cluster of districts
with high average education attainment levels.
5.4. The Dynamics of Spatial Association between District-wise
Earnings Income Inequality and Education Levels
This section analyses the temporal change in the spatial
distribution of district wise real per capita GDP growth rate, district
wise per capita incomes, and district human development levels between
1998 and 2005. It also examines the spatial association between district
wise primary, secondary, and bachelors education levels in 1998.
Figures A3a, A3b, A3c, and A3d in the Appendix each demonstrates a
Moran scatterplot which provides a disaggregated picture of the nature
of spatial autocorrelation for district per capita income in 1998 and
2005, using the BC and ID matrix respectively. The spatial lag (Wz) in
this situation is a weighted average of the incomes of a district's
neighbouring districts. The scatter plots in both the years (using both
the matrices) demonstrate that the overall pattern of spatial dependence
between district income levels has remained positive and statistically
significant. However, the overall value of the global Moran's I
statistic has reduced from being 0.81 to 0.38 between 1998 and 2005 when
the results are reported using the BC matrix. Similarly, the value of
global Moran's I statistic has reduced from being 0.91 to 0.51
between 1998 and 2005 under the results produced using the ID matrix.
Furthermore a spatial analysis of the growth rate between 1998 and
2005, also indicates a positive and a statistically significant spatial
autocorrelation pattern when neighbourhood is defined in terms of
contiguity but a statistically insignificant pattern when neighbourhood
is defined in terms of proximity as measured by the ID matrix (see Table
3). This implies that districts with a high (low) real GDP growth rate
may be spatially associated with their contiguous neighbouring districts
which also have high (low) real GDP growth rates.
Moreover, since our macro-data from 1998 provides district wise
statistics on individual education attainment levels (measured as the
percentage of individuals having completed an education level), it has
allowed us to analyse whether education levels in neighbouring districts
are spatially associated or how the distance from large neighbouring
cities (or provincial capitals) affects the incentives to obtain
education in a district. Table 4 demonstrates that whether neighbourhood
is measured in terms of geographic proximity (using ID matrix) or in
terms of geographic contiguity (using BC matrix), there exists a
positive and highly significant spatial autocorrelation for levels of
education below high-school (i.e. primary, matric i.e., grade 10, and
inter i.e., grade 12). However, for higher levels (Bachelors and above),
geographic contiguity to a district with a high percentage of graduates
could be more influential than the distance from the provincial capital
or the nearest large city.
Finally, although spatial association between district development
levels (as measured by the Human Development Index (HDI) calculated by
the UNDP in NHDR, 2003) has reduced between 1998 and 2005 from 0.40 to
0.311, it still remains positive and significant (see Table 5). These
results for Pakistani districts again confirm the findings of the new
economic geography literature that a region's development levels,
depend on the development levels prevailing in its neighbouring regions.
6. CONCLUSIONS
This paper has performed an exploratory analysis of socio-economic
disparities across Pakistan for the first time and has provided useful
insights for the conduct of economic regional policy in Pakistan. It has
investigated the spatial distribution of income inequality, income,
education, growth and development levels for 98 districts between 1998
and 2005. The overall finding that emerges from this article is that the
distribution of district wise income inequality, income, education
attainment, growth, and development levels, exhibits a significant
tendency to cluster in space (i.e. the presence of spatial
autocorrelation is confirmed), thereby highlighting the importance of
understanding economic geography in the context of Pakistan.
Specifically the following main findings emerge from this article.
First, the province of Punjab contains the largest cluster of high per
capita income districts in both 1998 and 2005. Second, district wise
income inequality levels demonstrate weak spatial" association.
Moreover district education levels reveal high spatial association, and
districts with a high (low) real GDP growth rate have been spatially
associated with contiguous neighbouring districts which also have high
(low) real GDP growth rates between 1998 and 2005. Third, there exists
positive spatial dependence for education levels below bachelors (i.e.,
primary, matric i.e., grade 10, and inter i.e., grade 12). However,
for higher levels (Bachelors and above), geographic contiguity to a
district with a high percentage of graduates, is more influential than
the distance from the provincial capital or the nearest large city. This
result is corroborated by the findings from Burki and Khan (2010) which
confirms that districts located away from urban centres are also the
ones with lowest education levels in Pakistan. Our empirical analysis
also reveals that except for Lahore, none of the other 3 provincial
capitals of Pakistan (Karachi, Peshawar, Quetta) have high knowledge
spillovers. While this finding is not surprising for Karachi, since
coastal cities have different spillover mechanisms as compared to
landlocked cities, it indicates that infrastructure and cluster
development can facilitate increased knowledge spillovers at least from
the centres of economic activity in Pakistan if not from all large city
districts. Finally, spatial association of district wise Human
Development Indicators confirms that a district's development
levels may depend on the development levels prevailing in its
neighbouring districts in Pakistan.
The methodological implication of the above mentioned results is
that studies which utilise Ordinary Least Squares to investigate intra-
Pakistan socio-economic issues could possibly be producing inaccurate
statistical inferences. By assuming spatial-independence, they may
produce estimates that are biased and overestimated, since our results
show that observations for socio-economic district characteristics do
tend to cluster in Pakistan. The main policy implication that emerges
from our results is that growth and development policies need to focus
on infrastructure and cluster development that can cater to large
segments of the population. This is particularly because the spatial
pattern of income inequality, district incomes, education levels, and
development levels shows how development in Pakistan is concentrated in
Punjab (in particular Northern Punjab especially in terms of human
development indicators).
The presence of possible spatial spillovers as demonstrated in this
paper also implies that cluster development can play an extremely
important role in generating knowledge externalities, domestic commerce,
and employment creation by bringing work and knowledge to people instead
of them travelling to it. Pakistan already has many pseudo-clusters that
have developed over time. Examples include the IT cluster
'Karachi', textile and leather cluster 'Faisalabad',
automotive manufacturing cluster 'Port Qasim', furniture
cluster 'Gujranwala', light engineering cluster
'Gujrat', sports and surgical cluster 'Sialkot',
heavy industries cluster 'Wah' and even light weapons
manufacturing cluster 'Landikotal'. An emphasis on regional
and industrial regeneration policies can play a crucial role in reducing
spatial disparities and enhancing the regional advantages of these
districts [Pakistan (2011)]. Finally, this paper has highlighted the
importance of additional research on Pakistan that takes into account
spatial effects. Since it has only considered spatial changes in
socio-economic phenomena in 8 years between 1998 and 2005, an immediate
possibility could be to extend this spatio-temporal analysis may include
extending it over a longer period of time. Another possibility may
involve a spatial econometric analysis of the effect of a
district's inequality, income and education levels on its growth.
While the presence of spatial clustering of income and education in
Pakistan (as demonstrated in this paper) could support the use of a
spatial lag model to capture the spillover of inequality between
districts, missing data on district incomes or omitted variables could
also necessitate the use of a spatial error model (which reflects
spatial autocorrelation in measurement errors) in analysing the effect
of inequality on district income levels.
APPENDIX
[FIGURE A3a OMITTED]
[FIGURE A3b OMITTED]
[FIGURE A3c OMITTED]
[FIGURE A3d OMITTED]
[FIGURE A4 OMITTED]
[FIGURE 6a OMITTED]
[FIGURE 6b OMITTED]
[FIGURE A7a OMITTED]
[FIGURE A7b OMITTED]
[FIGURE A8 OMITTED]
Table A1
List of Districts
PUNJAB
1 Rawalpindi
2 Jhelum
3 Chakwal
4 Attock
5 Gujranwala
6 Mandi Bahauddin
7 Hafizabad
8 Gujrat
9 Sialkot
10 Narowal
11 Lahore
12 Kasur
13 Sheikuhupura
14 Okara
15 Faisalabad
16 Jhang
17 TT Singh
18 Sargodha
19 Khushab
20 Mianwali
21 Bhakkar
22 Multan
23 Khanewal
24 Lodhran
25 Vehari
26 Sahiwal
27 Pakpattan
28 Bahawalpur
29 Bahawalnagar
30 R. Y. Khan
31 D.G. Khan
32 Muzaffar grah
33 Layyah
34 Rajanpur
SINDH
35 Hyderabad
36 Dadu
37 Badin
38 Thatta
39 Mirpur Khas
40 Sanghar
41 Tharparkar
42 Sukkur
43 Ghotki
44 Khair put
45 Nawab shah
46 Larkana
47 Jaccobabad
48 Shikarpur
49 Nowshero Feroz
50 Karachi
KPK
51 Peshawar
52 Charsadda
53 Nowshera
54 Kohat
55 Kark
56 Hangu
57 Mardan
58 Sawabi
59 Abbottabad
60 Haripur
61 Mansehara
62 Batagram
63 Kohistan
64 Swat
65 Lower Dir
66 Upper Dir
67 Chitral
68 Malakand Agency
69 Shangla
70 Bannu
71 Lakki Marwat
72 D.I. Khan
73 Tank
74 Bunir
BALOCHISTAN
75 Quetta
76 Sibi
77 Nasirabad
78 Kalat
79 Pishin
80 Qilla Abd
81 Bolan
82 Pangjur
83 Barkhan
84 Chagai
85 Jaffarabad
86 Jhal Magsi
87 Mastung
88 Awaran
89 Gwadar
90 Turbat
91 Kharan
92 Ziarat
93 Khuzdar
94 Killa Saif
95 Lasbella
96 Loralai
97 Musa Khel
98 Zhob
Table A2a
Global Autocorrelation Results for Earnings Income
Inequality--Geary's C (2005)
Weight Matrix I II
i [not equal to] j [w.sub.i,j] = 0 or 1 [w.sub.i,j] = 1/[d.sub.i,j]
I = j [w.sub.i,i] = 0
Geary's C 0.824 1.458
E(C) 1.000 1.000
Sd(C) 0.082 0.324
Z -2.138 1.413
p-value 0.033 0.158
Source: Author's Calculations.
Table A2b
Global Autocorrelation Results for District Per Capita Earnings
Income--BC Matrix
Weight Matrix 1998 2005
i [not equal to] j [w.sub.i,j] = 0 or 1 [w.sub.i,j] = 0 or 1
i = j [w.sub.i,i] = 0
Moran's I 0.818 0.380
E(I) -0.010 -0.010
Sd(I) 0.103 0.101
Z 8.048 3.856
p-value 0.000 0.000
Source: Author's Calculations.
Table A5
Global Autocorrelation Results for Education Attainment--Geary's C
(2005)
Weight Matrix I II
i [not equal to] j [w.sub.i,j] = 0 or 1 [w.sub.i,j] = 1/[d.sub.i,j]
i = j [w.sub.i,i] = 0
Geary's C 0.584 1.092
E(C) 1.000 1.000
Sd(C) 0.080 0.275
Z -5.230 0.336
p-value 0.000 0.737
Source: Author's Calculations.
REFERENCES
Anselin, L. (1988a) Lagrange Multiplier Test Diagnostics for
Spatial Dependence and Spatial Heterogeniety. Geographical Analysis 20,
1-23.
Anselin, L. (1988b) Spatial Econometrics: Methods and Models.
Dordrecht: Kluwer Academic Press.
Anselin, L. (1992a) Space and Applied Econometrics. Special Issue.
Regional Science and Urban Economics 22.
Anselin, L. (1992b) SpaceStat, A Software Programme for the
Analysis of Spatial Data. Santa Barbara, CA: National Center for
Geographic Information and Analysis, University of California.
Anselin, L. (1995a) SpaceStat. A Software Programme for the
Analysis of Spatial Data (Version 1.80). Morgantown: Regional Research
Institute, West Virginia University.
Anselin, L. (1995b) Local Indicators of Spatial Association--LISA.
Geographical Analysis 27, 93-115.
Anselin, L. (1996) The Moran Scatterplot as an ESDA Tool to Assess
Local Instability in Spatial Association. In M. Fisher, H. J. Scholten,
and D. Unwin (eds.) Spatial Analytical Perspectives on GIS. London:
Taylor and Francis.
Arbia, G., R. Benedetti, and G. Espa (1996) Effects of the MAUP on
Image Classification. Geographical Systems 3, 123-41.
Arif, G. M., N. Iqbal and S. Farooq (2010) The 2010 Flood and
Poverty in Pakistan: A Preliminary District Level Analysis. Conference
Paper for The Environments of the Poor. New Dehli.
Bailey, T. C. and A. C. Gatrell (1995) Interactive Spatial Data
Analysis. Addison Wesley Longman.
Baldwin, R., R. Forslid, P. Martin, G. Ottaviano, and F.
Robert-Nicoud (2003a) Economic Geography and Public Policy. Princeton
University Press.
Balisacan, A. M. and F. Nobuhiko (2004) Changes in Spatial Income
Inequality in the Philippines: An Exploratory Analysis. World Institute
for Development Economic Research (UNU-WIDER). (Working Papers UNU-WIDER
Research Paper).
Beck, N., K. Gleditsch, and K. Beardsley (2006) Space is More than
Geography: Using Spatial Econometrics in the Study of Political Economy.
International Studies Quarterly 50, 27-44.
Burki, A. A. and M. A. Khan (2010) Spatial Inequality and
Geographic Concentration of Manufacturing Industries in Pakistan. The
Pakistan Development Review.
Burki, A., A. K. Munir, M. Khan, U. Khan, A. Faheem, A. Khalid, and
S. T. Hussain (2010) Industrial Policy, Its Spatial Aspects and Cluster
Development in Pakistan. Analysis Report for the Industrial Policy 2010,
Lahore University of Management Sciences.
Celebioglu, F. and S. Dall'erba (2010) Spatial Disparities
across the Regions of Turkey: An Exploratory Spatial Data Analysis. The
Annals of Regional Science 45:2, 379-400.
Cliff, A. D. and J. K. Ord (1981)Spatial Processes. London: Pion.
Darlauf, S. N. and D. T. Quah (1999) The New Empirics of Economic
Growth. In J. B. Taylor and M. Woodford (eds.) Handbook of
Macroeconomics Vol. IA, Chap 4. North-Holland, Amsterdam.
Dominicis, L., G. Arbia, and H. L. F. de Groot (2007) The Spatial
Distribution of Economic Activities in Italy. Tinbergen Institute.
(Tinbergen Institute Discussion Papers 07-094/3).
Fotheringham, A., C. Bunsden, and M. Charlton (2000) Quantitative
Geography. London: Sage.
Franzese, R. and J. Hays (2007b) Spatial Econometric Models of
Cross-Sectional Interdependence in Political Science Panel and
Time-Series-Cross-Section Data. Political Analysis 15:2, 140-64.
Galor, O. and D. Tsiddon (1997) Technological Progress, Mobility,
and Economic Growth. American Economic Review 87, 363-382.
Glaeser, E., H. Kallal, J. Sheinkman, and A. Shliefer (1992) Growth
in Cities. Journal of Political Economy 100, 1126-1152.
Glomm, Gerhard and B. Ravikumar (1992) Public Versus Private
Investment in Human capital: Endogenous Growth and Income Inequality.
Journal of Political Economy 100:4, 818-834.
Haining, R. (2003) Spatial Data Analysis: Theory and Practice.
Cambridge: Cambridge University Press.
Henderson, J. V. (2003) Marshall's Scale Economies. Journal of
Urban Economics 53:1, 1-28.
Jamal, H. and A. J. Khan (2003) The Changing Profile of Regional
Inequality. The Pakistan Development Review 42:2, 113-123.
Jamal, H. and A. J. Khan (2005) The Knowledge Divide: Education
Inequality in Pakistan. The Lahore Journal of Economics 10:1, 83-104.
Jamal, H. and A. J. Khan (2008) Education Status of Districts: An
Exploration of Inter-Temporal Changes. (Research Report 71).
Jamal, H. and A. J. Khan (2008) Trends in Regional Human
Development Indices. (Research Report 73).
Jamal, H., A. J. Khan., I. A. Toor, and N. Amir (2003) Mapping the
Spatial Deprivation of Pakistan. The Pakistan Development Review 42:2,
91-1 I.
Kanbur, R. and T. Venables (2005) Introduction: Spatial Inequality
and Development. Journal of Economic Geography 5:1.
Khan, R. I. A. (2003) Spatial Distribution of Population with
Special Refernce to 1998 Population Census. In A. R. Kemal, M. Irfan,
and N. Mahmood (eds.). Population of Pakistan: An Analysis of 1998
Population and Housing Census. Islamabad: Pakistan Institute of
Development Economics.
Krugman, P. (1991) Geography and Trade. Cambridge: MIT Press.
Krugman, P. and A. J. Venables (1995) Globalisation and Inequality
of Nations. Quarterly Journal of Economics 11, 857-880.
Le Gallo, G. and C. Ertur (2003) Exploratory Spatial Data Analysis
of the Distribution of Regional Per Capita GDP in Europe, 1980-1995.
Journal of Economics 82:2, 201.
LeSage, J. P. (1999) The Theory and Practice of Spatial
Econometrics. Department of Economics, University of Toledo.
(Unpublished).
Mion, Giordano (2004) Essays in Spatial Economics. Open Access
Publications from Universite catholique de Louvain.
Naqvi, S. A. A. (2007) A Look at Spatial Inequality in
Pakistan--Case Study of District Sargodha. Centre for Global,
International and Regional Studies. Mapping Global
Inequalities--Conference Papers. (Paper mgi-9).
Pakistan, Government of (2011) Framework for Economic Growth
Pakistan. Islamabad: Planning Commission of Pakistan.
Pose, A. R. and V. Tselios (2007) Mapping The European Regional
Educational Distribution: Educational Attainment And Inequality.
Instituto Madrileno de Estudios Avanzados (IMDEA) Ciencias Sociales.
(Working Papers, 2007-18).
Quah, D. (1996) Regional Convergence Clusters in Europe. European
Economic Review 35, 951-958.
Rey, S. J. and B. D. Montouri (1999) US Regional Income
Convergence: A Spatial Econometric Perspective. Regional Studies 33:2,
143-156.
Saint-Paul, G. and T. Verdier (1993) Education, Democracy and
Growth. Journal of Development Economics 42:2, 399-407.
Siddiqui, R. (2008) Income, Public Social Services, and Capability
Development: A Cross-district Analysis of Pakistan. Pakistan Institute
of Development Economics, Islamabad. (PIDE Working Papers 2008:43).
Sokal, R. R., N. L. Oden, and B. A. Thomson (1998) Local Spatial
Autocorrelation in a Biological Model. Geographical Analysis 30,
331-354.
Tobler, W. (1970) A Computer Movie Simulating Urban Growth in the
Detroit Region. Economic Geography 46:2, 234-240.
United National Development Programme (2003) Pakistan National
Human Development Report. Karachi: Oxford University Press.
Comments
This paper analyses the spatial clustering of income, income
inequality, education, human development and growth by employing spatial
exploratory data analysis technique. The author needs to be commended
for taking forward the research agenda for sub-national development
issues and we really need this type of research for policy making
purposes at the local/regional levels.
The data employed in the paper comes from the district
representative PSLM 2004-05 CWIQ survey data set which gives information
only on cash earnings of the employed. The author has used this variable
for calculating income and income inequality at district level which is
technically incorrect, as this does not represent all the income flows
coming to a household.
The total household income which is the combination of income flows
coming from nearly 12 different sources, is only captured in the
national level HIES data set and is not available at the district level
data set used in this study.
Similarly, where the author mentions income inequality she is
actually referring to cash earnings inequalities across the districts.
As the author has carried out a very original piece of research, I
would suggest that the author should tailor her conclusions and policy
implications accordingly as well as mention the limitations of her
analysis using cash earnings income as opposed to income only.
A minor comment of an editorial nature is that author has used the
word chapter repeatedly throughout the paper which needs to be replaced
by 'paper/study'.
Lubna Shahnaz
Planning Commission, Islamabad.
(1) Economic inequalities refer to education, earnings income
inequalities in particular.
(2) Examples of studies similar to this paper include: Rey and
Montouri (1999) on convergence across USA, Balisacan and Fuwa (2004) for
income inequality in Philipines, Dall'erba (2004) analyses
productivity convergence across Spanish regions over time, Dominicis,
Arbia and de Groot (2005) analyses spatial distribution of economic
activities in Italy, Pose and Tselios (2007) investigates education and
income inequalities in the European Union, and Celebioglu and
Dall'erba (2009) analyses spatial disparities in growth and
development in Turkey.
(3) The only other exception includes Burki, et al. (2010) that has
explicitly considered spatial dependencies in its analysis. However it
has analysed 56 districts.
(4) Exceptions include Jamal and Khan (2003, 2003a), Jamal and Khan
(2008, 2008a), Naqvi (2007), Arif, et al. (2010), Siddique (2008) and a
few others. Except for Jamal and Khan (2003, 2003a), Jamal and Khan
(2007, 2007a), most of them only study selected districts/villages from
the same province e.g. Naqvi (2007) only analyses the districts/villages
of Punjab.
(5) Modifiable Areal Unit Problem: When attributes of a spatially
homogenous phenomenon (e.g., people) are aggregated into districts, the
resulting values (e.g., totals, rates and ratios) are influenced by the
choice of the district boundaries just as much as by the underlying
spatial patterns of the phenomenon.
(6) Usually two or more weights matrices are utilised in spatial
exploratory and econometric studies as a robustness measure. It is way
of demonstrating whether strength of spatial effects are robust to
changing definitions of neighbourhood.
(7) Another well-known measure of spatial autocorrelation is Getis
and Ord's G statistic [see Anselin (1995a), p. 22-23].
(8) The other two assumptions include the assumption of normal
distribution of the variables in question (normality assumption) or a
randomisation approach using a reference distribution for I that is
generated empirically (permutation assumption). For details and formulas
of the randomisation assumption, [see Sokal, et al. 1998)].
(9) Negative spatial autocorrelation reflects lack of clustering,
more than even the case of a random pattern. The checkerboard pattern is
an example of perfect negative spatial autocorrelation.
Sofia Ahmad <
[email protected] > is Research
Economist, Pakistan Institute of Development Economics, Islamabad.
Author's Note: I would like to thank Dr Jannette Walde
(University of Innsbruck Austria), Dr Maria Sassi (University of Pavia,
Italy), Dr Alejandro Canadas (Mount St Marys University, USA), Dr
Giuseppe Arbia (University G. D'Annunzio of Chieti, Spatial
Econometric Association), and Dr Richard Pomfret (University of
Adelaide) for their comments on an earlier version of this paper. I
would also like to thank Khydija Waked and Muhammad Qadeer at the
Planners Resource Centre Pakistan, for providing me with the shape
files. Finally, I would like to acknowledge, the data management staff
at the Pakistan Institute of Development Economics (PIDE) Islamabad,
Federal Bureau of Statistics, Islamabad, and Dr Amir Jahan Khan and Dr
Haroon Jamal (Sustainable Policy Development Centre, SPDC Karachi) for
their generous data support. This is a preliminary version of this paper
and comments are welcome.
Table 1
Global Autocorrelation Results for Income Inequality--Moran's I (2005)
Weight Matrix I II
I [not equal to] j [w.sub.I,j] = 0 or 1 [w.sub.ij] = 1/[d.sub.I,j]
I = j [w.sub.i,i] = 0
Moran's I 0.211 0.257
E(I) -0.010 -0.010
Sd(I) 0.074 0.103
Z 2.985 2.601
p-value 0.003 0.009
Table 2
Global Autocorrelation Results for Education Attainment--Moran's I
(2005)
Weight Matrix I II
I [not equal to] j [w.sub.i,j] = 0 or 1 [w.sub.i,j] = 1/[d.sub.i,j]
I = j [w.sub.i,i] = 0
Moran's I 0.395 -0.003
E(I) -0.010 -0.01
Sd(I) 0.075 0.103
Z 5.440 0.072
p-value 0.000 0.943
Table 3
Spatial Autocorrelation of per capita GDP Growth Rate between
1998-2005
GDP Growth Rate
(1998-2005)
BC Matrix ID Matrix
Moran's I 0.430 0.140
E(I) -0.010 -0.010
Sd(I) 0.071 0.099
Z 6.204 1.524
P-value 0.000 0.128
Source: Author's own calculations.
Table 4
Spatial Autocorrelation for Education Levels (1998)
Primary Education
BC ID
Moran's I 0.494 0.559
E(I) -0.010 -0.010
Sd(I) 0.075 0.103
Z 6.745 5.501
P-value 0.000 0.000
Geary's C 0.497 0.983
E(c) 1.000 1.000
Sd(c) 0.079 0.244
Z -6.401 -0.069
P-value 0.000 0.945
Matric
BC ID
Moran's I 0.391 0.247
E(I) -0.010 -0.010
Sd(I) 0.074 0.102
Z 5.443 2.523
P-value 0.000 0.012
Geary's C 0.610 0.703
E(c) 1.000 1.000
Sd(c) 0.085 0.379
Z -4.573 -0.783
P-value 0.000 0.434
Higher Education-Bachelors
BC ID
Moran's I 0.327 -0.014
E(I) -0.010 -0.010
Sd(I) 0.074 0.102
Z 4.582 -0.038
P-value 0.000 0.969
Geary's C 0.610 1.643
E(c) 1.000 1.000
Sd(c) 0.086 0.392
Z -4.538 4.193
P-value 0.000 0.000
Source: Authors own calculations. BC: Binary Contiguity Matrix, ID:
Inverse Distance Matrix.
Table 5
HDI Spatial Autocorrelation Using the Binary Contiguity Matrix
District Human
Development
Index (HDI)
1998 2005
Moran's I 0.405 0.311
Standard Deviation (I) 0.075 0.074
Z-value 5.573 4.341
P-value 0.000 0.000
Source: Author's calculations using data from NHDR (2003).
