econstor
A Service of
zbw
Make Your Publications Visible.
Leibniz-Informationszentrum
Wirtschaft
Leibniz Information Centre
for Economics
Ye, Xinyue; Yue, Wenze
Working Paper
Comparative analysis of regional development:
Exploratory space-time data analysis and open
source implementation
Economics Discussion Papers, No. 2014-20
Provided in Cooperation with:
Kiel Institute for the World Economy (IfW)
Suggested Citation: Ye, Xinyue; Yue, Wenze (2014) : Comparative analysis of regional
development: Exploratory space-time data analysis and open source implementation,
Economics Discussion Papers, No. 2014-20
This Version is available at:
http://hdl.handle.net/10419/96486
Standard-Nutzungsbedingungen:
Terms of use:
Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen
Zwecken und zum Privatgebrauch gespeichert und kopiert werden.
Documents in EconStor may be saved and copied for your
personal and scholarly purposes.
Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle
Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich
machen, vertreiben oder anderweitig nutzen.
You are not to copy documents for public or commercial
purposes, to exhibit the documents publicly, to make them
publicly available on the internet, or to distribute or otherwise
use the documents in public.
Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen
(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten,
gelten abweichend von diesen Nutzungsbedingungen die in der dort
genannten Lizenz gewährten Nutzungsrechte.
http://creativecommons.org/licenses/by/3.0/
www.econstor.eu
If the documents have been made available under an Open
Content Licence (especially Creative Commons Licences), you
may exercise further usage rights as specified in the indicated
licence.
Discussion Paper
No. 2014-20 | April 30, 2014 | http://www.economics-ejournal.org/economics/discussionpapers/2014-20
Comparative Analysis of Regional Development:
Exploratory Space-Time Data Analysis and Open
Source Implementation
Xinyue Ye and Wenze Yue
Abstract
This paper aims to make contributions to comparative analysis of regional economic dynamics in
China and the US from the methodological perspective. More specifically, some recent advances in
exploratory space-time data analysis (ESTDA) will be suggested and implemented to conduct this
task in an open source environment. China and the US have been the subject of much discussion
about the patterns and trends of regional development, because of their importance in the world
economic system. Despite the rich and growing empirical literature on numerous case studies at
both national and sub-national scales, comparative space time analysis between and within these
two economic systems has just begun to catch attentions. Additionally, ESTDA and its open source
implementation can facilitate comparative studies of regional development.
JEL C21 C88
Keywords ETSDA; regional development; comparative analysis; open source; China; USA
Authors
Xinyue Ye, Kent State University, Kent, Ohio, USA
Wenze Yue, Zhejiang University, China, wzyue@zju.edu.cn
Citation Xinyue Ye and Wenze Yue (2014). Comparative Analysis of Regional Development: Exploratory Space-Time
Data Analysis and Open Source Implementation. Economics Discussion Papers, No 2014-20, Kiel Institute for the World
Economy. http://www.economics-ejournal.org/economics/discussionpapers/2014-20
Received April 4, 2014 Accepted as Economics Discussion Paper April 28, 2014 Published April 30, 2014
© Author(s) 2014. Licensed under the Creative Commons License - Attribution 3.0
2
1 Introduction
GIS and regional economics have been, and continue to be challenged by dealing with
temporal trend of geographic patterns and spatial dynamics of economic development
(Goodchild, 2008; Rey, 2009; Anselin, 2010). There is an imperative need for effective and
efficient methods to represent and examine the coupled space-time attributes of economic
phenomena in the comparative context (Janikas, 2007; Goodchild and Glennon, 2008). The
growth in international trade, the restructuring of economies under reform, and widening
regional inequality, have reemphasized the importance of geographical analysis. Theory,
methodology, and practice of temporal GIS has emerged as an active domain to address these
challenges while a large amount of efforts are still needed in integrating regional economics with
temporal GIS (Ye and Rey, 2011). As a multi-dimensional and multi-scale phenomenon,
regional development witnesses the role of geography and the awakening emphasis on space
among economists. Additionally, it is more interesting to what extent that geography matters,
which is also stimulated by the emergence of new economic geography (Krugman, 1999;
Fingleton, 2004; Anselin, 2010). This paper suggests some novel exploratory approaches to
compare spatial pattern and temporal trend of regional development between China and the US.
Accordingly, a reasonable number of research hypotheses can be developed. The crossfertilization between ESTDA (exploratory spatial data analysis) and regional studies is thus
identified and illustrated.
As Rey and Ye (2010) argue, “in order to develop a more spatially explicit growth theory
it is first necessary to develop operational measures that capture the spatial dynamics inherent in
regional datasets”. Hence, this research aims to provide new insights on comparative regional
studies by revealing and quantifying the changes and level of hidden variation across scales and
3
dimensions. More specifically, the paper reveals the comparative spatial dynamics that underlie
regional development by visualizing the similarity of time series across various intervals and
interactively investigating the growth trajectories of states and provinces using comparative
LISA (Local Indicators of Spatial Association) time path, in order to shed light on the debates on
mixed conclusions derived in many case studies. The above new advances are developed in an
open source environment, which offers a straightforward way of benefiting wider community. As
such, explanations of various regional development styles can be provided based on rigorous
analysis, and policy interventions are then proposed in light of the understanding of the spacetime data set.
This paper is organized as follows. The next section presents a literature review of
ESTDA and open source spatial analysis, along with the competing interpretation of regional
economic dynamics between and within China and the US. This is followed by an illustration of
some recent development of ESTDA in an open source environment, using the GDP per capita
data in China and the US over the 1978–2008 period. This paper closes with some final points.
The purpose of this paper is to promote and realize an open source implementation to conduct
comparative exploratory space-time analysis of regional economic data, which is viewed as an
initial step towards the goal of exploring and sharing development opportunities between and
within China and the US.
2 Literature Review
China’s mounting importance in the world economic system since the launch of reforms
in the late 1970s has long been the subject of intense academic debates, in terms of its
spectacular growth and poverty reduction. After overtaking Japan as the world’s secondary
4
largest economy, China is now challenging the hegemonic status of the US. As Wei (2006)
comments, “the rise of China and its challenge to US hegemony is a hot topic in the media and
popular magazines such as Foreign Policy”. Such phenomenal development has been a hallmark
of China's growth, and has been gaining attentions and recognition (Fan and Sun, 2008). The
speed and unevenness of economic restructuring and development in favor of specific regions, is
truly unprecedented and merits a careful examination and assessment both theoretically and
methodologically. Scholars have been intensely debating the impacts of the reform on regional
convergence and inequality (Weng 1998; Wei 1999; Li and Haynes, 2010). Research has
revealed that patterns of regional inequality differ with geographical scales, thus creating a
multiscalar typology of regional inequality in China (Wei and Ye, 2009). Multi-scale and multimechanism approach has been adopted, with a synthesis of global, state and local forces in
understanding the process of regional development (Yu and Wei 2003; Ye & Wei, 2005; Li and
Wei, 2011).
Many studies on US regional income growth patterns have been conducted from the
perspective of either geographical dynamics or time series (Fan and Casetti, 1994; Rey and
Montouri, 1999; Tomljanovich and Vogelsang, 2002; Rey 2004a, 2004b; Sommeiller, 2007). Fan
and Casetti (1994) argue that the inverted-U hypothesis cannot explain the rich details of
regional dynamics and conventional approach of measuring systemic inequality conflates the
large regional variations. Rey and Montouri (1999) examine US regional convergence from a
spatial econometric perspective, finding that the level of global spatial autocorrelation temporally
associates with regional income dispersion. Tomljanovich and Vogelsang (2002) detect strong
evidence for regional convergence based on time series methods. Rey (2004a, 2004b) develops a
series of spatially explicit methods of data analysis to address the geographical and distributional
5
dimensions of regional convergence and inequality. However, the findings are mixed,
ambiguous, and sometimes conflicting due to different research design and analytical strategies.
Sommeiller (2007) states that “the neoclassical economists support convergence whereas the
geographers marshal evidence for divergence, at least in the case of the United States at the subnational level”.
Not surprisingly, growing inequality across various spatial scales threatens the social
harmony and thus erodes the political basis for economic growth in both China and the US. This
concern is exemplified by Levy and Chowdhury (1995)’s comment that “large income and
wealth differences between countries and regions generated acts of aggression which inflicted
considerable human suffering, loss of resources and knowledge, destruction of civilizations and
environmental damage.” Meanwhile, a number of fascinating debates on the trajectories and
mechanisms of regional development are reflected in numerous empirical studies of specific
regions in these two countries. Despite this rich and growing list of empirical literature,
comparative analysis of regional development between and within China and the US remains
largely unexplored. It is fascinating to detect a daunting list of differences and similarities
between and within multiple regional systems. Researchers and policy makers can gain better
understanding of different economic development mechanisms and policy implementation
schemes based on comparative analysis (Ye and Rey, 2011).
The region serves as a useful unit of observation to test competing economic growth
theories and regional economic hypothesis (Quigley, 2001). Abundant evidence has exhibited
that spatial variation of regional development fails to be predicted or explained (Rey and
Montouri, 1999; Anselin, 2010). Though rich conceptual frameworks have highlighted the
spatial dynamics and unevenness of income distribution processes, many implementations of
6
growth theories in estimable econometric specifications do not appropriately treat dynamic
spatial effects in the data (Rey and Janikas, 2005). A key issue is that the interaction between
regional economics and space time analysis has not been as strong as perhaps it could be.
Fundamentally, the most crucial step is to systematically understand the data before testing
hypotheses. Researchers have to get acquainted with the data before formulating meaningful
research questions or carrying out action plan.
“Getting acquainted with data” is the task of exploratory data analysis (EDA) (Andrienko
and Andrienko, 2006; Anselin, et al. 2006). The debates on regional convergence and inequality
dynamics have been informed by, and to some extent inspired by, a parallel development of new
space-time methods which has quantified the magnitude of regional dynamics (Miller and
Wentz, 2003; Getis et al., 2004). The space-time perspective has become increasingly relevant to
the understanding of economic development and novel methods are needed to truly integrate
space and time (Ye and Carroll, 2011). A number of EDA techniques have been applied to the
dynamics of regional income distributions. Indeed, EDA is to analyze data for the purpose of
interactively formulating hypotheses instead of testing hypotheses (Andrienko and Andrienko,
2006; Guo and Mennis, 2009). Moreover, EDA can reveal complex space-time economic
phenomenon not identified otherwise (Le Gallo and Ertur, 2003). Rey and Ye (2010) argue that
space-time methods can be generated by either incorporating spatial dependence into the time
series or extending static spatial associations to a temporal context.
During the past several decades, a number of efforts have been witnessed on the
development and implementation of spatial statistical analysis packages, which continues to be
an active area of research. Familiar reviews of these issues are represented in, among others,
Anselin and Getis (1992); Getis et al. (2004); Rey and Anselin (2006); Anselin (2010). This is
7
exemplified by the growing software list with a quite considerable collection which is
maintained by Center for Spatially Integrated Social Science (CSISS). The history of open
source movement is much younger, but its impact on GIS world is impressive (Rey, 2009). Open
source, therefore, represents a paradigm shift in geospatial research that has facilitated
collaboration across disciplines. According to Lewis (2011), about 300 Open Source / Free GIS
related software projects are created after 15 years of open source movement. As Rey (2009)
comments, “a tenet of the free software (open source) movement is that because source code is
fundamental to the development of the field of computer science, having freely available source
code is a necessity for the innovation and progress of the field”. This user-led innovation in the
open source movement is also the philosophy of conducting research in regional economic
geography. The development of open source GIS packages has been boosted. However, open
source projects in the areas of advanced spatial analysis are very few (Rey and Anselin, 2007;
Rey, 2009).
3 Comparative Analysis
Exploratory data analysis has a significant appeal for comparative analysis because it can
help generates a comprehensive list of novel hypothesis. During the course of analysis, as
something interesting to researchers is detected in the data, new research questions arise, causing
specific regions or relationships to be scrutinized in more detail. In the rest of this section, some
methods are highlighted drawing on examples from per capita GDP distributions of China and
the United States from 1978 to 2008. The recent interest on comparative spatial economic
dynamics centers on two main dimensions of regional evolutions. The first concerns how
similar/dissimilar time series distribute over space. The second dimension relates to the temporal
8
trend of spatial pattern. Rey and Ye (2010) suggests spatial network/spider graph and LISA time
path to address these two concerns. Some initial designs of these geo-visual analytical methods
are available in the package STARS (Space-Time Analysis of Regional Systems) (Rey & Janikas
2006). As an open source toolbox designed for the analysis of space-time data, STARS is
featured with an array of dynamically linked graphical views. Written entirely in a scripting
language Python, STARS can be easily expanded to support more flexible and specific types of
analyses. As such, the current research extends these two general notions into a more interactive
and comprehensive context.
In the following sections, comparative space-time analysis of regional systems will be
conducted using case studies of China and the US. The two data sets are relative per capita gross
domestic product (GDP) over the 1978-2008 period at the province and state (contiguous United
States) levels. These two data sets are comparable because the states (contiguous United States)
and provinces (China) are self-contained and well-functioning units which form the theoretical
structure for spatial interaction models in spatial economy (Fan and Casetti, 1994; Fan and Sun,
2008).
3.1. Comparative Network and Spider
The network approach visualizes the covariance matrix of regional growth on a single
map. Figure 1 is generated using the package STARS. The temporal correlation matrix is a
matrix of correlation coefficients between the time series of per capita GDP of each state
(province). Through integrating a spatial component, the time series can be represented
geographically using the network approach. More specifically, this pairwise approach provides a
measure of the strength of the correlation between two sets of incomes in the spatial setting.
9
Links connecting the centroid of each region (polygon) are defined by spatial weights matrix,
which decide whether any pair of observations are neighbor. In Figure 1, first order contiguity is
employed as the default option, but other types of spatial weight matrix can also be used to
define the neighboring spatial units of a focal observation. In other words, researchers have the
flexibility to define which observations are neighbors to a focal observation based on the
conceptual framework.
In the current design of STARS, the color of the links on the map are conditioned on the
strength of the temporal covariance between each pair of neighboring regions, with the red
segment indicating an below-average correlation (weak temporal linkage) while the blue one
suggesting an above-average correlation (strong temporal linkage). Two neighboring provinces
connected by blue links are considered to be similar in the temporal dynamics since the
correlation of their income time series is above the national average. This closeness can be
interpreted that the two involved economies share some similar economic development
mechanisms, or interaction might exist between them. If a region has strong temporal linkages
with all of its spatial neighbors, there might be strong space-time economic integration for the
focal economy.
(Figure 1 about here)
However, the above visualization masks two important issues. First, positive correlation
and negative correlation are mixed. Second, the spatial distribution of correlation coefficients are
oversimplified by the classification based on the cutting point of the average. Limiting attention
to either magnitude or sign of correlation coefficients may result in a misguided or partial
understanding of regional dynamics. As Rey and Janikas (2006) state, “explore patterns through
various interfaces and the views are dynamically integrated with one another, giving rise to the
10
second meaning of dynamic spatial data analysis (the first meaning is incorporating time to
spatial data analysis).” The details of temporal linkages for each possible region-pair can be
further interactively visualized.
It is not surprising that similar time series tend to geographically cluster, as shown in
Figure 1. However, it is more interesting to see how the spatial pattern of links changes
according to the varying size and direction of correlation coefficients. Figures 2, 3, and 4 thus
visualize how neighboring economies relate to each other over time in these two regards.
Ignoring the rich details of these relationships leads to overlooking many possible interactions
and dependence among regions. Figures 2, 3, and 4 extend the previous work of STARS by
distinguishing the direction and size of correlation coefficients in an interactive manner, which in
turn generate different perspectives of looking at the same time series data. In other words, it is
valuable to consider all possible spatial perspectives before formulating research questions.
(Figure 2 about here)
In Figures 2, 3, and 4, solid blue segments indicate positive temporal linkages while
dotted red segments signify negative temporal linkages. More specifically, this extended network
graph identifies both similar and dissimilar economic growth trends quantitatively (the absolute
value of the correlation coefficient) and qualitatively (the sign of the value). As such, various
levels of correlation can be visualized, and this will more distinctly identify cross-sectional
relationships. The relationships between a focal economy and any of its neighboring economies
are divided into multiple groups based on user’s input. For instance, the left graphs on Figure 2
display the distribution of direction of correlation since the threshold of absolute correlation
value is set to be zero. The China case shows that most of the negative correlation links exist
between coastal provinces and their immediate inland provinces, while all the neighboring inland
11
provinces and coastal provinces have positive correlation coefficients among themselves. This
further justifies the widening and consistent inequality between coastal and inland China (Wei
and Ye, 2009). At the same time, it also vividly shows the spatial footprint of coast-oriented
regional development policy, because the coastal provinces are sharply separated from inland
provinces. When only the absolute correlation values above 0.5 are allowed to be drawn (the top
right graph in Figure 2), most of the disappearing links are located between central provinces and
their immediate neighboring coastal and other inland/western provinces. This disconnection,
instead of dissimilarity, also proves the widening coast-inland inequality. The US case
demonstrates an obvious northwest-southeast corridor of negative links (the bottom left graph in
Figure 2), while the threshold value of 0.5 further identify several regional economic groups,
such as economically-integrated Northeastern US and diverse Western US. Florida, a coastal
state largely based on tourism and elderly residents, is isolated from its neighboring states, part
of a large cluster of poverty trap.
Figure 3 and Figure 4 list the snapshots of the maps based on inputs of 0.2, 0.4, 0.6, and
0.8. It should be noted that any value can be selected by the sliding bar, and the corresponding
network map can be immediately drawn. Figure 3 and Figure 4 show that when the absolute
value of correlation coefficient increases, majority of segments left on the map are solid blue in
both countries. Larger values in correlation mainly indicate similarity of growth prevail among
neighboring province or states. However, it is worth noticing that only two strong negative links
(<= -.0.8) exist in China, and both of them connect to Guandong, a province on the South China
Sea coast. Meanwhile, no strong negative links are identified in the US.
Two robust phenomena survive the changing correlation value thresholds. First, there is a
‘rectangular-size’ cluster of strong segments in southwestern China, where a poor trap has
12
formed over years, while at the same time period there is a liner-shape of wealthy provinces with
similar time series displayed along the coast. Second, there are two persistent regional clusters in
the US: the Northeast-Mid Atlantic cluster of high income states and deep Southeast cluster of
low income states.
(Figure 3 about here)
(Figure 4 about here)
A spider graph reflects the correlation of incomes between a province/state and any of the
rest provinces (states) (Figure 5) (Rey and Janikas, 2006). This graph relaxes the contiguity
constraints in Figure 1, and visualizes the time series correlation between a specific region and
any of the rest of the regional system. Hence, the spider graph reveals the possible economic
integration of each focal region with their respective national system. As such, this method aims
to identify the specific regions with which they share common dynamics, as reflected in aboveaverage standardized pairwise correlations in economic dynamics. These are indicated by edges
connecting each focal region to the dynamically similar regions. This is illustrated in the spider
graphs of two regions by STARS: Guangdong in China and Florida in the US. Guangdong is in
sync with coastal provinces while all of Florida’s similar sets of linked states do not locate in any
neighboring spatial units (Figure 5). However, this map also masks the details as Figure 1 does.
In other words, this macro division can frequently mask a great deal of turbulence at finer scales.
Figure 6 extends the new design in Figure 2 to the case of spider graph. When the
threshold values change from 0.2 to 0.8 using 0.2 as the step length, Guandong’s positive links
display a higher degree of stability with six coastal provinces staying through 0.6 (Hainan leaves
the list when 0.8 applies). This is complicated by the radically different levels of negative links
over time (22, 19, 14, and 10). At the same time, a casual inspection suggests that Florida has
13
relatively changeable lists of both positive and negative links. When the threshold is set to be
0.8, only one positive link (California) and two negative links (Nebraska and Kentucky) remain.
Hence, researchers can further identify which provinces or states might have very different
economic development paths from the rest economic units, as well as the extent to which the
difference is.
(Figure 5 about here)
(Figure 6 about here)
(Figure 7 about here)
3.2 Comparative LISA Time-Path
LISA (local indicator of spatial association) is an indicator to examine local spatial
dependence (Anselin, 1995). LISA Time Path extends this static spatial statistics to a temporal
setting by plotting the pair-wise movement of a given variable of the focal unit (X coordinate)
and its spatial lag (Y coordinate) over time (Rey et al., 2005). At a given time, each region can
be identified with these X and Y coordinates. Hence, each region has a directional path
connecting all the coordinates by temporal order. Figure 8 illustrates the co-movement of a
region’s GDP per capita and its spatial lag over time in STARS, which can be used to identify
levels of stability of a region with a given structural process (the x-axis refers to a
province/state’s per capita GDP relative to the national average, and the y-axis refers to its
spatial lag). A variety of geometric properties can be summarized for each region’s LISA time
path, since individual aspects of the same contemporaneous process can be dissected by interval
gaps. When viewed in a comparative context, the geometry of the paths (the trajectory of LISA
of specific economies) can illuminate aspects of various regional growth processes.
(Figure 8 about here)
14
(Figure 9 about here)
(Figure 10 about here)
As Figure 8 illustrates, Guandong (top) and Florida (bottom) have been witnessing
sharply different trends of relationships in their local linkages. However, this visualization can be
further improved if any selected regions can be visualized on one map (under the same scale)
with the arrow of path indicating the direction. Figure 9 thus compares Guandong and Florida
under the same scale. Some interesting phenomena are detected. First, the distinct performance
of the two coastal regions reemphasizes the importance of location factors in the economic
landscape. Compared to Guandong, Florida has a very torturous path in a much smaller activity
space on the Figure, while it depicts a relatively sharper decline and rise on the neighbor side.
Over 21 years, the range on the X axis is 0.747 for Guandong (0.042 for Florida) and the range
on the Y axis is 0.082 for Guandong (0.005 for Florida). The national status of GDP per capita of
Guandong’s neighbors is much lower than Florida’s, and both of them are below the national
average in China and the US. Guandong’s average neighbor increases from 0.585 in 1978 to
0.667 in 2008, while Florida’s moves slightly from 0.828 to 0.883 during the same time period.
The striking feature is Guandong’s rocketing economic status. Guandong used to be lower than
that of Florida in the US, but it dramatically increased from below the national average to one of
the richest provinces (comparing the X axis). Guandong moves from 0.785 in 1978 to 1.532 in
2008, while Florida drops from 0.907 to 0.865 during the same time period. In other words,
Guandong makes a dramatic transition from 21.5 percent below the national mean to 53.2
percent above the national average. Second, Guandong’s LISA Time Path has contained most of
the big jumps in status move-up and move-down, which take place when the segments lean
towards the X axis. This phenomenon means the role of Guandong is decisive in its local regime.
15
However, Florida demonstrates an opposite trend, which indicates that Florida is overwhelmed
by its neighboring states in development.
Figure 10 further demonstrates how this new method can be used to compare the
evolution of local economic regime in each country, by contrasting the LISA time paths of
various regions at the same scale. Guandong and Fujian have been identified to have strong
positive links in the spider figure, and their LISA time paths are compared on the top graph of
Figure 10. Since Florida and California have strongly-correlated time series, their LISA time
paths are also compared. Guandong grows much faster than Fujian (Guandong increases from
0.785 to 1.532, while Fujian is from 0.581 to 1.056), while the latter’s neighbors benefit more
than the former’s (Guandong’s Y value move from 0.585 to 0.667, and Fujian’s increases from
0.692 to 1.192), which further validates the persistent poverty of Guandong’s neighbors at this
finer scale. A U shape (California) and an inverted U shape (Florida) with many short-lived
phenomena of status change are identified on the bottom graph of Figure 10. California (Florida)
and its average neighbor change from 1.127 (0.907) and 1.058 (0.828) to 1.105 (0.865) and 0.971
(0.833) respectively. The network graph, in conjunction with the spider figure in the previous
section highlight that California and Florida are largely isolated from their neighboring states due
to the low correlation coefficients, which can be partly explained by the existence of many path
segments parallel to either X or Y axis. These segments dramatically decrease the relevance of
the two states to their neighbors in the statistical results. Hence, the comparison of LISA TimePaths can reveal important insights as to regional income disparities. Additionally, a closer
inspection reveals that China has much more dispersed spatial dynamics in Figure 9 and Figure
10, which indicate the possible existence and practice of a variety of economic development
models, as well as a large and persistent inequality in China due to both economic success and
16
problematic development (Fan and Sun, 2008; Wei and Ye, 2009). The peaks and troughs shown
in Figure 9 and Figure 10 also imply that the economic reform policy starting from 1978 are vital
to explain changes in regional growth and inequality.
The relative levels and pace of change of a region can thus be investigated and compared
at the individual scale. That is, a particular region’s economic status fluctuates, or moves
up/down relative to the national average. At the local scale, a focal economy might have a
different velocity of development rate from its neighboring ones over time. This comparison
provides important insights to the finer-scale aspects of stability and distinct directional
movement within various dynamic regional income distributions, because the convergence
hypothesis is concerned with these distributions over time.
5 Summary
Spatial turn in many socioeconomic theories has been noted in a vast field, encompassing
both social and physical phenomena (Krugman, 1999; Goodchild, 2008; Sui, 2011). This
intellectual and technological change has yielded important insights on physical sciences, social
sciences and the humanities, with an explosion of interest across disciplines (Peuquet, 2002;
Batty, 2005; Sui, 2011). Regional variations in economic development are highly topical subjects
for intellectual enquiry and have long been the focus of policy initiatives. The empirical
observations in the current research lend support to the notion that the multi-scale and multidimension methods can expose some hidden space-time patterns and trends that otherwise would
be very difficult to detect in regional economic studies.
Bode and Rey (2006) call for “further research on integrating space into formal
theoretical models of growth and convergence as well as on developing the next generation of
17
analytical methods needed to implement those models” as “the preconditions for reliable policy
recommendations, one of the primary goals of economic research”. The fast growth in spatial
economic analysis is increasingly seen as attributable to the availability of panel regional
economic datasets. By contrast, regional geographers have been slower to adopt and implement
new spatially explicit methods of data analysis due to the lack of extensible software packages,
which becomes a major impediment to promote spatial thinking in regional geography studies.
Meanwhile, spatial economic analysis is increasingly being supported by the emergence
of advanced analytical methods in space-time data analysis and data visualization. The
interactive spatial data analysis has motivated, if not directly provoked, new queries on spatial
economic theories. Therefore, the current research implements the new methodological advances
in an open source environment for exploring data that has both temporal and spatial dimensions,
which lend support to the notion that space and time cannot be meaningfully separated. This
paper thus demonstrates an example to interface academic spatial analysis with the open source
revolution, which is among the burgeoning efforts seeking the cross-fertilization between the two
fast-growing communities (Rey, 2009). The research provides an extension of some current
functions in the open source package STARS with an eye towards promoting comparative space
time analysis of regional systems. As Rey (2009) suggests, “increased adoption of open source
practices in spatial analysis can enhance the development of the next generation of tools and the
wider practice of scientific research and education”. As such, this paper helps to speed that
adoption by investigating the role of space and scale in the comparative analysis of regional
income inequality in China and the US over the 1978-2008 period. The concept of exploratory
space-time data analysis is strongly associated with visualization because graphical presentation
enables the analyst to open-mindedly explore the structure of the data set and gain some new
18
insights, as illustrated by the comparative network/spider figures and LISA time paths.
Shneiderman (1996) argues that exploratory data analysis can be generalized as a three-step
process: “overview first, zoom and filter and then details-on-demand”. More importantly, it is
worth noticing that this process should be iterative, and the methods implemented in the current
research addressed the challenge. To explain the observed patterns and trends, a follow-up
research is needed on collecting determinants of economic growth.
This open source work procedure can facilitate the interdisciplinary research due to “the
collaborative norms involving positive spillover effects in building a community of scholars”
(Rey, 2009). The package of STARS is entirely open source, which can promote collaboration
among researchers who want to improve current functions or add extensions to address specific
research questions in regional studies.
Based on the strength of scientific visualization techniques, this paper stresses the need to
study the space-time dimension underlying regional economic data sets. This work provides
insights as to the possible spatial pattern of the selected correlation coefficients over time,
temporal dynamics of associated regions, and economic growth of a specific region. Finally, a
new interactive tool is suggested and demonstrated as providing an explanatory framework for
space-time economic data. On this basis, the sincere hope here is that this dialogue between
regional economics and ESTDA will embrace the real world challenges of economic inequality
issues.
ESDA is not new to economic convergence and inequality studies, whereas extending
ESDA into a dynamic context, no doubt, would give better support for the scientific investigation
and management of regional economic data sets, including its description, representation,
analysis, visualization, and simulation. Additionally, comparative space-time analysis enables
19
access to a much wider thinking which addresses the role of space at different stages of
economic growth and thus identifies the research gaps and opportunities for more in-depth study.
This paper explores the potential for the new methods to function in regional studies,
specifically, in the comparison of income distribution dynamics between different economic
systems. In other words, the current work is mainly from an exploratory perspective, which can
motivate regional and economic geographers to design a series of regional economic analysis
questions and formulate new hypotheses from theoretical and policy perspectives. This spacetime work provides an important contribution to the current economic convergence and
inequality literature, which lacks in comparative space-time studies. Although this comparative
study arose in the analysis of income distribution dynamics, it broadly aims to analyze the role of
geography and location in socioeconomic phenomena. Hence, it can also be applied to a wide set
of socioeconomic processes with geo-referenced data measured over areal units at multiple time
periods. In addition, the methods are built in open source environments and thus easily extensible
and customizable.
References
Andrienko, N. and Andrienko, G. (2006). Exploratory Analysis of Spatial and Temporal Data: A
Systematic Approach. Springer, Berlin.
Anselin, L. (1995). Local indicators of spatial association-LISA. Geographical Analysis, 27:93–
115.
Anselin L (2010) Thirty years of spatial econometrics. Pap Reg Sci 89(1):3–25
Anselin, L. and Getis, A. (1992). Spatial statistical analysis and geographic information systems.
The Annals of Regional Science, 26:19–33.
Anselin, L., Syabri, I., and Kho, Y. (2006). GeoDa: An introduction to spatial data analysis.
Geographical Analysis, 38:5–22.
Batty, M. (2005) Approaches to modeling in GIS: spatial representation and temporal dynamics.
20
In GIS, spatial analysis, and modeling, ed. D.J. Maguire, M. Batty, and M.F. Goodchild, 4161. Redlands, CA: ESRI Press.
Bode, E. and Rey, S. J. (2006). The spatial dimension of economic growth and convergence.
Papers in Regional Science, 85(2):171–176.
Fan, C. and Casetti, E. (1994). The spatial and temporal dynamics of US regional income
inequality, 1950-1989. Annals of Regional Science, 28:177–196.
Fan, C.C. and M.J. Sun. 2008. Regional Inequality in China, 1978–2006. Eurasian
Geography and Economics 48, 1–20.
Fingleton, B. (2004). Theoretical economic geography and spatial econometrics: Bridging the
gap between theory and reality. In Getis, A., M´ur, J., and Zoeller, H., editors, Spatial
Econometrics and Spatial Statistics, pages 8–27. Palgrave, Hampshire.
Getis, A., M´ur, J., and Zoeller, H., editors (2004). Spatial Econometrics and Spatial Statistics.
Palgrave, Hampshire.
Goodchild, M. F. (2008). Geographic Information Science: the grand challenges. In Wilson, J.
and Fotheringham, A., editors, The Hand-book of Geographic Information Science, pages 596–
608. Malden, MA: Blackwell.
Goodchild, M. F. and A. Glennon (2008) Representation and computation of geographic
dynamics. In K.S. Hornsby and M. Yuan, editors, Understanding Dynamics of Geographic
Domains. Boca Raton: CRC Press, pp. 13–30.
Guo, D. and J. Mennis (2009). Spatial data mining and geographic knowledge discovery-An
introduction, Computers, Environment and Urban Systems, 33(6) pp. 403-408.
Janikas, M. (2007). Comparative regional income dynamics: Clustering, scale, and
geocomputation. PhD thesis, University of California, Santa Barbara and San Diego State
University.
Krugman, P. (1999). The role of geography in development. Interna-tional Regional Science
Review, 22(2):142–161.
Le Gallo, J. and Ertur, C. (2003). Exploratory spatial data analysis of the distribution of regional
Per Capita GDP in Europe, 1980–1995. Papers in Regional Science, 82:175–201.
Levy, A. and Chowdhury, K. (1995). A geographical decomposition of intercountry income
inequality. Comparative Economic Studies, 37:1–17.
Miller HJ, Wentz EA (2003) Representation and spatial analysis in geographic information
systems. Ann Assoc Am Geogr 93:574–594.
21
Lewis B (2011) Open source GIS. http://opensourcegis.org/
Li, H. and Haynes, Kingsley E. (2010) Economic Structure and Regional Disparity in China:
Beyond the Kuznets Transition. GMU School of Public Policy Research Paper No. 2010-16.
Available at SSRN: http://ssrn.com/abstract=1601051
Li, Y.R. and Y.H.D. Wei. (2010) The spatial-temporal hierarchy of regional inequality of China.
Applied Geography 30: 303–316.
Peuquet, D. J. (2002). Representations of Space and Time. Guilford, New York.
Quigley, J. M. (2001). The renaissance in regional research. The Annals of. Regional Science,
35(2):167–178.
Rey, S. J. (2009) Show me the code: Spatial analysis and open source. Journal of Geographical
Systems 11: 191–207
Rey, S. J. (2004). Spatial dependence in the evolution of regional in-come distributions. In Getis,
A., M´ur, J., and Zoeller, H., editors, Spatial Econometrics and Spatial Statistics, pages 194–213.
Palgrave, Hampshire.
Rey, S. and Anselin, L. (2006). Recent advances in software for spatial analysis in the social
sciences. Geographical Analysis, 38:1–4.
Rey, S. and Anselin, L. (2007). PySAL: A Python library for spatial analytical functions. The
Review of Regional Studies, 37:5–27.
Rey, S. J. and Janikas, M. V. (2005). Regional convergence, inequality, and space. Journal of
Economic Geography, 5:155–176.
Rey, S. J. and Janikas, M. V. (2006). STARS: Space-time analysis of regional systems.
Geographical Analysis, 38:67–86.
Rey, S., Janikas, M., and Smirnov, O. (2005). Exploratory geovisualization of spatial dynamics.
Geocomputation 2005 Proceedings (CD-ROM).
Rey, S. J. and Montouri, B. D. (1999). U.S. regional income conver-gence: A spatial
econometric perspective. Regional Studies, 33:143– 156.
Rey, S. J. and Ye, X. (2010). Comparative spatial dynamics of regional systems. In P´aez, A.,
Gallo, J. L., Buliung, R., and Dall’Erba, S., editors, Progress in Spatial Analysis: Methods and
Applications, pages 441–463. Springer.
Shneiderman, B. (1996). The eyes have it: A task by data type taxonomy for information
visualizations. In Burnett, M. and Citrin, W., editors, Proceedings of the 1996 IEEE Symposium
on Visual Languages, page 336343. IEEE Computer Society Press, Piscataway.
22
Sommeiller, E. (2007). Regional income inequality in the United States, 1913–
2003. PhD thesis, University of Delaware.
Tomljanovich, M. and Vogelsang, T. J. (2002). Are U.S. regions converging?
Using new econometric methods to examine old issues. Empirical Economics,
27(1):49.
Wei, Y.H.D. (1999) Regional Inequality in China. Progress in Human Geography 23(1): 49-59.
Wei. Y.H.D. (2006) Geographers and Globalization: The Future of Regional Geography.
Environment and Planning A 38(8): 1395-1400.
Wei, Y. H. D. and Ye, X. (2004). Regional inequality in China: A case study of Zhejiang
Province. Tijdschrift voor Economische en Sociale Geografie (Journal of Economic and Social
Geography), 95:44–60.
Weng, Q. (1998), Local Impacts of the Post-Mao Development Strategy. International Journal
of Urban and Regional Research 22, pp. 425–442.
Ye, X. and Carroll, M. (2011) Exploratory Space-Time Analysis of Local Economic
Development. Applied Geography, 31: 1049-1058.
Ye, X. and Rey, S.J. (2011) A Framework for Exploratory Space-Time Analysis of Economic
Data. Annals of Regional Science DOI 10.1007/s00168-011-0470-4
Ye, X. and Wei, Y. D. (2005). Geospatial analysis of regional devel-opment in China: The case
of Zhejiang Province and the Wenzhou Model. Eurasian Geography and Economics, 46:342–
361.
Yu, D.L. and Y.H.D. Wei. 2003. Analyzing Regional Inequality in Post-Mao China in a GIS
Environment. Eurasian Geography and Economics 44: 514-534.
23
Figures:
Figure 1 Spatial Correlation Network in China and the US Generated by STARS
Note: cnpcr indicates per capita income relative to the national mean in China; uspcr indicates
per capita income relative to the national mean in the USA. These abbreviations apply to all
other figures.
Figure 2 New Design of Spatial Correlation Network in China and the US
Figure 3 Snapshots of Spatial Correlation Network in China
Figure 4 Snapshots of Spatial Correlation Network in the US
Figure 5 Spatial Spider of Guandong Province, China and Florida, the US Generated by STARS
Figure 6 Snapshots of New Design of Spatial Correlation Network in China
Figure 7 Snapshots of New Design of Spatial Correlation Network in the US
Figure 8 LISA Time Path of Guandong Province, China and Florida, the US Generated by
STARS
Figure 9 LISA Time Path of Guandong Province, China and Florida, the US in One Map
Figure 10 LISA Time Path of Guandong and Fujian; Florida and California
24
Figure 1 Spatial Correlation Network in China and the US Generated by STARS
25
Figure 2 New Design of Spatial Correlation Network in China and the US
26
27
Figure 3 Snapshots of Spatial Correlation Network in China
28
29
Figure 4 Snapshots of Spatial Correlation Network in the US
30
Figure 5 Spatial Spider of Guandong Province, China and Florida, the US Generated by STARS
31
32
Figure 6 Snapshots of New Design of Spatial Correlation Network in China
33
34
Figure 7 Snapshots of New Design of Spatial Correlation Network in the US
35
Figure 8 LISA Time Path of Guandong Province, China and Florida, the US Generated by
STARS
36
Figure 9 LISA Time Path of Guandong Province, China and Florida, the US in One Map
37
Figure 10 LISA Time Path of Guandong and Fujian; Florida and California
Please note:
You are most sincerely encouraged to participate in the open assessment of this
discussion paper. You can do so by either recommending the paper or by posting your
comments.
Please go to:
http://www.economics-ejournal.org/economics/discussionpapers/2014-20
The Editor
© Author(s) 2014. Licensed under the Creative Commons Attribution 3.0.