PROCEEDING

The 3rd International Conference on Economics, Business, and Accounting Studies (ICEBAST) 2017
“Social Cohesion, Public Policy Reformation, and Market Integration towards Inclusive Global Economy”
Faculty of Economics and Business - Universitas Jember, 24-25 November 2017

THE SPATIAL DISTRIBUTION OF INDONESIA’S MANUFACTURING INDUSTRIES:

AN EXPLORATORY SPATIAL DATA ANALYSIS
Edy Santoso* and Regina N. Wilantari
University of Jember, Jl. Kalimantan No. 37, Jember 68121, Indonesia
* Corresponding author. Tel.: +62-331-332150; fax: +62-331-332150 E-mail address: [email protected]

Abstract

The issues of spatial distribution of manufacturing industries are important to economic policy. In this paper we contribute to the empirical
literature on the spatial distribution of manufacturing sector by taking spatial dependence among region units explicitly into account. The
main objective of this paper is to analyze the space-time dynamics of spatial distribution of manufacturing industries in Indonesia. For the
methodology, Exploratory Spatial Data Analysis (ESDA) Global Moran’s I was used to determine the general level of spatial autocorrelation
in the data based on the province level of manufacturing industries establishments data for the year 2006 to 2015. The results show strong
evidences of global spatial autocorrelation in the distribution of manufacturing industries in Indonesia. The result shows that the distribution
of Indonesia’s manufacturing industries are unevenly distributed. The Moran’s I statistic analysis, indicates a strong positive spatial
autocorrelation in research area and discribes the phenomenon pattern expressed is clustered. A high concentration of manufacturing
industries is seen in the several provinces only, as well as in Java island. The moran’s I are tend to decrease along the periode of study. From
the LISA, we can ascertain that there was statistically significant high clustering. Only some provinces really show significant clustering and
these provinces are West Java, Banten, DKI Jakarta, Central Java, and East Java.
Keyword: Spatial Distribution, Manufacturing Industries, ESDA, Spatial Autocorrelation, LISA

INTRODUCTION
The main feature of geographically economic activity is concentration and unevenness. The concentration of
spatial economic activity indicates that industrialization is a selective process and occurs only in certain cases
when viewed in geographical terms. For example, in the United States, the majority of the manufacturing industry
is concentrated in a Northeast and Upper Midwest location known as manufacturing belt. Then the
manufacturing industry in the United States moved from the northern region to other country and certain areas
of the southern region have become very specialized in manufacturing and referred to as the new manufacturing
belt [1]. Similar industrial spatial concentrations are also found Axial Belt industrial estate in UK and Ruhr
manufacturing belts in Germany [2].

Source: Calculated from BPS, Industrial Survey

Figure 1. Distribution of Indonesia’s Manufacturing Industry, 2015
The same phenomena can also be found in Indonesia, industrial development and economic activities in
Indonesia for more than the last three decades is unevenly distributed. Indonesia's manufacturing industry has
tended to be spatially concentrated in Java since the 1970 [3, 4]. According to [5], the manufacturing industry
was spatially concentrated in Java since the 1970, although in the late 1980 it shifted to Sumatra and Bali.
Spatially concentrated industries, especially for large and medium manufacturing industries. This was due to a
series of deregulation and liberalization in mid-1980. In addition, this spatial concentrated was also caused by
the infrastructure and labor accumulated in some certain areas, especially Java [6]. Spatial concentrations also
can be found in most developing countries where population and industrial distribution are concentrated in large
cities such as Bangkok, New Delhi, Sao paulo and Jakarta which mark a spatial system based on capital and labor
accumulation in urban agglomeration [7].

131 ISBN: 978-602-5617-05-8

 Edy Santoso and Regina N. Wilantari

Source: Calculated from BPS, Industrial Survey

Figure 2. Labor Growth of Manufacturing Industry by Province (000 people), 2011-2014
Figure 2 shows that an interregional disparities associated with manufacturing industry for the growth of
manufacturing labor for the year 2011-2014. The largest labor growth of manufacturing industry is dominated
by the provinces that became the center of manufacturing industry in Indonesia, mostly found in Java. These
provinces are Banten Province, East Java, Central Java, West Java, and DKI Jakarta. For the province outer of Java
Island, there are North Sumatra Province and Riau Islands Province. Hence, it can be concluded that the growth
of labor of manufacturing industry in Indonesia is linear with spatial concentration of manufacturing industry.
This means that high labor growth occurs in provinces that do have a high concentration of manufacturing
industry activity (see figure 1).
The concentration of spatial economic activity, especially in the manufacturing industry, has become an
interesting phenomenon to be analyzed. For the manufacturing industry, spatial concentrations are determined
by wage costs, transportation costs and market access and the externalities of spatial concentration associated
with saving localization and saving urbanization. Uneven distribution of resources leads to disparities in spatial
industrial concentrations between regions. The inequality of this resource is reflected in the concentration of
economic activity occurring in certain areas only. A region where the concentration of economic activity occurs
benefits (a so-called agglomeration economies). As a result the industry will move and be spatially concentrated
in areas that can benefit from the geographic proximity of economic activity. Instead the industry will leave areas
that are no longer able to provide benefits from the geographic proximity of economic activity.
The concentration of spatial economic activity can be understood by the location theory approach. The purpose
of the location theory is to determine the reasons and explain why certain factors are important for one industry
but not for other industries [8]. It also involves the principle of substitution, in which the industry chooses a
location from some alternative location, which in economic theory, is similar to the problem of how to replace
labor production factors with capital or land and vice versa. Many factors are considered in relation to the
determination of the appropriate location for the industry. These factors vary depending on the peculiarities of
an industry [9]. Generally speaking location theory is broadly divided into three theoretical schools [10], which
include: Neo-classical theory, New trade theory, and New Economic Geography (NEG). Neo-Classical theory is
characterized by perfect competition, homogenous products and non-increasing returns to scale [10]. This
theory indicates that the location of an industry is determined exogenously, by the so called first nature (term
used by [11] and [12], which indicates a given spatial distribution of natural endowments, climate, technology,
and factors of production. Conversely, in the NEG theory, location is determined endogenously, by second nature
where spatial distribution of economic activities is independent from natural advantage. The concept of second
nature is characterized by mobility of production factors and/or firms, pecuniary and technical externalities, and
input-output linkages [10].
The present study focuses on spatial exploratory analysis of Indonesia’s manufacturing industries. Spatial
exploratory analysis of industry location will aid in capturing important intangible aspects concerning spatial
dependence and heterogeneity, which was not acknowledged by most of the previous studies. This spatial
analysis can also identify some of manfacturing industries that may clustered or randomly distributed. Analyzing
these spatial concepts and location factors will aid state and local economic development agencies in designing
strategies to better retain and attract manufacturing industries and their clusters, which in turn will boost their
economy and provide employment opportunities for local residents. The primary objective of this paper is to
evaluate the presence and extent of manufacturing industries spatial correlation for province level data. This
objective is achieved by utilizing Global Moran’s I Statistic and Local Indicators of Spatial Association (LISA)
statistics.

132
The Spatial Distribution of Indonesia’s ..........

METHODS
[13] noted that traditionally, each of economies is assumed as independent unit, and has ignored the possibility
of space interactions across regions. [14] stated that everything is related to everything else, but near things are
more related than distant things. Since, data obtained from points in space are generally associated with spatial
dependence and heterogeneity [15, 16]. Spatial dependence indicates that observations at a particular location
depend on observations at other locations. This spatial dependence in some instances is also expressed as spatial
autocorrelation [15]. Conversely, spatial heterogeneity refers to variations in spatial structure in the form of non-
constant error variance or model estimates [15]. Many economists have been interested in the use of spatial
econometrics in a regional economic studies. One of the advantages of this method is its ablity to capture the
spatial effect or spatial relationship in a geographical economies. Spatial effects are important in explaining
spatial distribution of economic activities, however it has been largerly ignored in the traditional economic
literature that pool data for large samples of countries or regions. This study analyse an exploratory spatial data
associated with spatial distribution of Indonesia’s manufacturing industries, which examines whether a
phenomenon pattern expressed is clustered, dispersed, or random. Using manufacturing industries
establishments for periode 2007-2014. To achieve the objectives, we were used Global Moran’s I Statistic and
Local Indicators of Spatial Association (LISA) statistics. The brief explanation for each method as follows:
1. Global Moran’s I
The spatial association of data collected from points in space is tested using a Global Moran’s I, which measures
similarities and dissimilarities in observations across space [15]. Moran’s I is a test for spatial autocorrelation,
which examines whether a phenomenon pattern expressed is clustered, dispersed, or random (Stieve, 2012). For
the number of establishments y, Moran’s I is:

   i j i
 n   ( y   )( y j   ) 
I  
  w    i ( yi   )2  .................................................................... (1)
 i j ij   

where wij indicates elements of the spatial weight matrix W (distance/contiguity weight matrix) between two
points ( i and j ),  the mean of all y observations, and i, j = 1,…,n. In general, a Moran's Index value near +1.0
indicates clustering while an index value near -1.0 indicates dispersion. A positive and significant value for
Moran’s I indicate positive spatial correlation, showing that provinces have a high or low number of
establishments similar to their neighbouring province. Conversely, a negative and significant value for Moran’s I
indicates negative spatial correlation, showing that provinces have high or low number of establishments unlike
neighbouring provinces [17]. The study calculate’s Moran’s I for the number of manufacturing industries across
33 province in Indonesia for the year 2006-2015, by employing GeoDa, a spatial data analysis software.
2. Local Indicators of Spatial Association (LISA)
In the case of uneven spatial clustering, global spatial indicators such as Moran’s I are found to be less useful.
This resulted in a new general class of local spatial indicators such as Local Indicators of Spatial Association (LISA,
also known as Local Moran) statistics developed by [15] are designed to test individual sites for membership in
clusters, which measures the contribution of individual provinces to the global Moran’s I statistic [15]. Local
Moran’s I is useful for showing places where significant spatial autocorrelation exists and designed to test
individual sites for membership in clusters. The LISA statistic is calculated for the ith province as:

I i  zi  wij z j .................................................................... (2)

j

where wij indicates elements of the spatial weight matrix W (distance/contiguity weight matrix) between two

points ( i and j ), z i and z j indicates the standardized number of establishments for province i and j,
respectively. The sum of LISA ( I
i
i
) for all observations is proportional to global Moran’s I, implying that LISA

statistic can be interpreted as indicators of local spatial clusters and as diagnostics for local instability (spatial
outliers) [15].
The LISA cluster map indicates the locations with a significant Local Moran statistic classified by type of spatial
correlation: (a) high-high association (HH), a province with many manufacturing industries has neighboring
province with many manufacturing industries; (b) low-low association (LL), a province with few manufacturing
industries has neighboring province with few manufacturing industries; (c) low-high association (LH), a province
with few manufacturing industries has neighboring province with many manufacturing industries; and (d) high-
low association (HL), a province with many manufacturing industries has neighboring province with few

133
 Edy Santoso and Regina N. Wilantari

manufacturing industries. The HH and LL locations suggests clustering of similar values (positive spatial
correlation), whereas the HL and LH locations indicate spatial outliers (negative spatial correlation) [15].

FINDINGS AND ARGUMENT

Uneven distribution of resources can lead to disparities in the economic development between regions. The
inequality of this resource is reflected on concentration of economic activities occurring on certain region. A
region with a concentration of manufacturing industry would be growing faster than a region with no a
concentration of manufacturing industry. The figure 1 illustrates that the spatial distribution of manufacturing
industries in Indonesia for the year 2015 is unevenly distributed, which is some provinces becoming central to
this distribution of industries. Provinces with high concentrations of industries are dominated by provinces in
Java Island such as Banten, West Java, DKI Jakarta, Central Java, and East Java. Some provinces outside of Java
island also have a high concentration of industries, these provinces are North Sumatra and Riau Islands. Hence,
it can be concluded that the distribution of manufacturing industry in Indonesia is unequally distributed and still
concentrated in Java Island.
Spatial Exploratory Analysis
The spatial distribution of manufacturing industries for the year 2015 is presented in figure 3. The figure
illustrates that the distribution of Indonesia’s manufacturing industries is unevenly distributed for all 33
provinces. Mostly, the manufacturing industries are located in Province of Banten, West Java, DKI Jakarta, Central
Java, East Java, and North Sumatra, where most of these provinces are located in Java Island. As a result, a high
concentration of manufacturing industries is seen in the Java Island. This implies that the Indonesia’s
manufacturing industries shows a spatial pattern and it is not independently distributed over space.

Source: Calculated from BPS, Industrial Survey

Figure 3. The Spatial Distribution of Indonesia’s Manufacturing Industries, 2015

To prove that the Indonesia’s manufacturing industries are not independently distributed over space. we could
use the Global Moran’s I. This index analyze spatial dependence (autocorrelation) of manufacturing industries
establishments acrooss regions (provinces).
Global Moran’s I
The spatial association of manufacturing industries is tested using a global Moran’s I, which measures similarities
and dissimilarities in manufacturing industries across neighbouring counties [19]. A positive and significant
value for Moran’s I indicate positive spatial correlation, showing that province with a high or low numbers of
manufacturing industries are similar to their neighbouring province. Conversely, a negative and significant value
for Moran’s I indicates negative spatial correlation, showing that province with a high or low number of
manufacturing industries are unlike their neighbouring province. We calculate Moran’s I for the number of
manufacturing industries establishments across all contiguous Indonesia’s provinces for the year 2015, by
employing GeoDa, spatial data analysis software. The Moran’s I statistic is equal to 0.581859, indicating a strong
positive spatial relationship (autocorrelation) in my research area and shows that the phenomenon pattern
expressed is clustered. But, it is significant?. Use pseudo p-value in the randomization for reporting p-value, we
get ρ = 0.002, so it’s definitely significant.

134
The Spatial Distribution of Indonesia’s ..........

Figure 4. Global Moran’s I Scatter Plots of Indonesia’s Manufacturing Industries, 2015

How about the patern of the Moran’s I for Indonesia’s manufacturing industri?. Figure 5 shows the global Moran’s
I statistic for spatial distribution of manufacturing industries in Indonesia for the year 2006 to 2015 using the
spatial weight matrix. The values of global Moran’s I statistic show a downward trend from 2006 to 2015 and
significant positive spatial autocorrelation. The results suggest that the locations of province with high or low
the number of manufacturing industries are clustered, and the spatial pattern has a developing trend of decrease
spatial autocorrelation.

0.800000
0.700000
0.600000
Global Moran's I

0.500000
0.400000
0.300000
0.200000
0.100000
0.000000
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Year
Source: Calculated from BPS, Industrial Survey
Figure 5. The chart of Moran’s I statistics for Indonesia’s Manufacturing Industries, 2006-2015
The Moran’s I is a global statistic. Hence, we know there is clustering going on in the research area, but we do not
know exactly where clustering exist. To investigate further, we would do a Local Indicators of Spatial Association
(LISA). This test not only tests for regional clustering, it can also show the presence of significant spatial clusters
or outlier.
Local Indicators of Spatial Association (LISA)
Moran’s I gives us a global statistic, and a single result to assess the spatial association pattern for the entire
study area. Therefore, it cannot identify the regions contributing more to the results of global spatial
autocorrelation, and detect the hot spot or a typical localization. In order to solve these problems, the Moran
scatter plot and Local Indicators of Spatial Association (LISA, also known as Local Moran) are adopted. LISA
statistics was developed by [19], which measures the contribution of individual province to the global Moran’s I
statistic. Local Moran’s I useful for showing places where significant spatial autocorrelation exists.
The figure 3 shows a high concentration of manufacturing industries is seen in the several provinces, as well as
in java island. This implies that, the Indonesia’s manufacturing industries exhibits a spatial pattern and it is not
independently distributed over space.

135
 Edy Santoso and Regina N. Wilantari

Figure 6. Local Moran’s I scatter plot for Indonesia’s Manufacturing Industries, 2006-2015
The HH and LL locations suggests clustering of similar values (positive spatial correlation), whereas the HL and
LH locations indicate spatial outliers (negative spatial correlation) [19]. A positive and high autocorrelation is
found in some provinces really show significant clustering. Figure 6 indicates locations with a significant Local
Moran statistic for the year 2006 to 2015. It illustrates that no significant changes for HH location for along
research period. Provinces of West Java, Banten, DKI Jakarta, Central Java, and East Java still have the most
significant manufacturing industrial clusters among all Indonesia Provinces. So, it can be concluded that along
the periode of study, the provinces which is have significant manufacturing industrial clusters are the same, these
provinces are West Java, Banten, DKI Jakarta, Central Java, and East Java. although, the moran’s I tend to decrease
along the year 2006-2015.

Source: Calculated from BPS, Industrial Survey

Figure 7. Local Indicator of Spatial Association (LISA) Cluster Map for Manufacturing Industries, 2006

136
The Spatial Distribution of Indonesia’s ..........

The LISA Cluster Map shows that how the attribute industries clusters. The red color show the tracts where high
rate cluster with high rates, and blue shows where low rates cluster with low rates. Figure 7 and figure 8, we
could see there is a cluster of high industries cluster and cluster of lower industries. For the year of 2006 and
2015, A high cluster (HH) industries occurs in five provinces, these provinces are West Java, Banten, DKI Jakarta,
Central Java, and East Java. For outer java, there is no high cluster (HH) industries for any province.

Source: Calculated from BPS, Industrial Survey

Figure 8. Local Indicator of Spatial Association (LISA) Cluster Map for Manufacturing Industries, 2015

CONCLUSION
This study analysis an exploratory spatial data associated with spatial distribution of Indonesia’s manufacturing
industries, which examines whether a phenomenon pattern expressed is clustered, dispersed, or random. In this
paper we contribute to the empirical literature on the spatial distribution of manufacturing sector by taking
spatial dependence among region units explicitly into account. We apply an approach that considers
simultaneously geographical concentration and spatial dependence to characterize the spatial distribution of
manufacturing industries. The result from the Moran’s I statistic analysis, indicating a strong positive spatial
relationship (autocorrelation) in research area and shows that the phenomenon pattern for the distribution of
manufacturing industries expressed is clustered, and a high concentration of manufacturing industries based is
seen in the several provinces, as well as in java island. Report at least the moran’s I are tend to decrease along
the periode of study. From the LISA, we can ascertain that there was statistically significant high clustering in
Indonesia. Only some provinces really show significant industrial clustering and these provinces are West Java,
Banten, DKI Jakarta, Central Java, and East Java.
The findings are not only strengthened the regional economic theory especially the economic geography theory,
but also important to develop a policy recommendation due to industrial development planning. As a policy
implication, the analyse showed that spatial dependence across regions was matter and in other side will be
beneficial for the economic growth of the regions. It implied that all regions need to cooperate due to economic
integration.
For future research, it might be interesting to develop model take into account some variables which influence
to spatial distribution of manufacturing industries. Hence, the model can be able to explain the spatial
distribution of manufacturing industries process.

REFERENCES
[1] J. H. Thomas and J.J. Stevens. (2003) Spatial Distribution of Economics Activities in North America. Handbook of
Regional and Urban Economic: Cities and Geography. Vol. 4. North Holand. 2003.
[2] Roger Hayter. (2000). The Dynamics of Industrial Location: The Factory, The Firm and The Production System. John
Wiley and Sons Ltd. West Sussex. England, 2000.
[3] Iwan Jaya Aziz, Iwan. (1994). Ilmu Ekonomi Regional dan Beberapa Aplikasinya di Indonesia (Regional Economics
and Its Some Applications in Indonesia). Lembaga Penerbit Fakultas Ekonomi Universitas Indonesia, Jakarta ,1994.
[4] Hall Hill. (1997). Indonesia’s Industrial Transformation. Singapore: Institute of Southeast Asian Studies, 1997.
[5] M. Kuncoro. 2012. Ekonomika Aglomerasi: Dinamika dan Dimensi Spasial Kluster Industri di Indonesia. UPP STIM
YKPN, Yogyakarta, 2012.
[6] Wahyuddin. (2004) Industri dan Orientasi Ekspor Dinamika dan Analisis Spasial. Universitas Muhamadiyah
Surakarta, 2004.

137
 Edy Santoso and Regina N. Wilantari

[7] M. Kuncoro. (2002), Analisis Spasial dan Regional: Studi Aglomerasi dan Kluster Industri.. UPP AMP YKPN. Yogyakarta,
2002.
[8] Melvin L. Greenhut. (1982). Plant Location in Theory and in Practise, the Economics of Space. The Theory of Least-
Cost Location, Chapter one. Greenwood Press, Publishers, 1982.
[9] P. R. Sambidi. (2007). Spatial Econometric Analysis of Agglomeration Economies Associated With The Geographical
Distribution of The U.S. Biotech Industry. A Dissertation. The Graduate Faculty of the Louisiana State University and
Agricultural and Mechanical College, 2007.
[10] M. Brülhart. 1998. Economic Geography, Industry Location, and Trade: The Evidence. World Economy 21, 1998,
pp.775 - 802.
[11] William Cronon. (1991). Nature’s Metropolis: Chicago and the Great West. New York: Norton,1991.
[12] P. Krugman. (1993). First nature, second nature, and metropolitan location. Journal of Regional Science 33, 1993, pp.
129-14.
[13] Sergio Rey and B. Montouri. 1999. US Regional Convergence: A Spatial Econometric Perspective. Regional Studies 33,
1999, pp. 148-156.
[14] Tobler. (1979). Cellular Geography, Philosophy in Geography. Reidel, Dordrecht, 1979, pp. 379-386.
[15] Luc Anselin. (1988). Spatial Econometrics Methods and Models. Kluwer Academics Publishers, Dordrecht, 1988.
[16] J. P. LeSage. (1999). The Theory and Practice of Spatial Econometrics. Unpublished, Dept. of Econ., University of
Toledo, 1999.
[17] Andrada Pacheco L and Timothy J. Tyrrell. (2002). Testing Spatial Patterns and Growth Spillover Effects in Clusters of
Cities. Journal of Geographical Systems 4, 2002, pp. 275-285.
[18] L. Anselin. (1995). Local indicators of spatial association – LISA. Geographical Analysis,27. 1995, pp. 93-115.

138