interregional trade and transport connectivity: an
TRANSCRIPT
1
Interregional Trade and Transport Connectivity: An Analysis of Spatial and Networks
Dependence by Using a Spatial Econometric Flow Model at NUTS3 level
Luisa Alamá-Sabater*
Laura Márquez-Ramos*
Celestino Suárez-Burguet*
José Miguel Navarro-Azorín** *Universitat Jaume I, Castellón, Spain
** Universidad Politécnica de Cartagena, Spain
Abstract
Recent research has analyzed whether existing transportation networks affect interregional trade in
goods in a spatial econometric model approach (LeSage and Polasek, 2008; Alamá-Sabater et al,
2012). Nonetheless, to our knowledge, previous research within this approach does not deal with
highly disaggregated regional trade data by disentangling the role of different types of transportation
networks. This paper aims to cover this lack in the existing literature.
In order to proxy for the transportation network structure, data for trade flows between Spanish
provinces (NUTS3)1 in the year 2007 are obtained, as well as data about the characteristics of logistics
platforms existing in each region. Previously, Alamá-Sabater et al (2012) provided evidence about the
role of the location of logistics platforms for satisfying existing demand for transport structure at
NUTS2 spatial disaggregation level, as spatial and transportation networks dependence play a
significant role on interregional trade flows. However, these authors only focused on the number and
area of logistics platforms existing in each region to proxy for the logistics facilities in contiguous
regions.
The present paper uses a gravity model of trade as a basis to explain trade flows between Spanish
intra-national regions in spatial approach terms, incorporating, in addition to the characteristics of each
region, information on transportation networks. Additionally, it extends the procedure followed by
LeSage and Polasek (2008) and Alamá-Sabater et al (2012) including weighting matrixes based on
the logistics performance by typology (i.e. sea ports) to analyze and compare the role of transportation
networks on trade of different sectors.
Keywords: Spanish regions, interregional trade, logistics platforms.
JEL classification: R12, R23, R48
Corresponding author: Luisa Alamá-Sabater, Department of Economics, Universitat Jaume I,
Campus del Riu Sec, Castellon (12071), Spain. E-mail: [email protected]
1 NUTS is a French acronym for Nomenclature of Territorial Units for Statistics used by Eurostat. In this nomenclature NUTS1 refers to European Community Regions and NUTS2 to Basic Administrative Units, with NUTS3 reflecting smaller spatial units most similar to counties in the USA.
2
1. Introduction
This paper uses a gravity model of trade (Bergstrand, 1985 and 1989; Deardorff, 1995) as a basis to
explain trade flows between intra-national regions in spatial approach terms, incorporating, in addition
to the characteristics of each region, information on transport connectivity.
In previous research for highly disaggregated interregional United States trade, Hillberry and Hummels
(2008) shown that goods produced at a particular distance are not purchased because there is no
local demand for them, and then most shipments occur only between proximate location pairs, as
shipments are extremely localized. Two main reasons are provided. First, consumers would buy a
good produced at a distance if price became competitive with local alternatives (Eaton and Kortum,
2002; Melitz, 2003). Second, Hillberry and Hummels (2008)’ results support that regions specialize in
the production of different goods and vary in the mix of inputs they absorb. Transport connectivity
might play an important role in both explanations.
Transport connectivity has recently been considered in gravity studies of trade. In this sense, we can
distinguish two ways of defining connectivity (Márquez-Ramos et al, 2011). On the one hand,
connectivity in a narrow sense is limited to the physical properties of the transport network. On the
other hand, connectivity in a broad sense includes those factors related to the features of the services
and cooperation of transport operators, which are essential for the efficiency and effectiveness of the
transportation network. Márquez-Ramos et al (2011), as well as other authors who have considered
transport connectivity, address the concept in a narrow sense (Limao and Venables, 2001; Sanchez et
al, 2003; Clark et al, 2004; Micco and Serebrisky, 2004; Wilson et al, 2004), finding that transport
connectivity increases trade flows between trading partners. Nonetheless, these studies do not
consider the existence of spatial dependence among regions that, following Alamá-Sabater et al
(2012), is introduced in this paper to analyze the effect of transport connectivity in the broad sense.
Using a spatial autoregressive model, Alamá-Sabater et al (2012) analyzed the presence of spatial
autocorrelation effects on Spanish interregional trade flows, which has also been analyzed in LeSage
and Llano (2006). There are also working papers of the same authors in the public domain, where
spatial and network autocorrelation effects are also tested for Spain using sector specific flows from
the same trade database that we use in the present paper. Nonetheless, in addition to the traditional
connectivity concept defined by the geographical criteria which Lesage and Llano (2006) used, Alamá-
Sabater et al (2012) used a broad transport connectivity concept by considering the presence of
logistics platforms. In order to do so, these authors extended the procedure followed by LeSage and
Polasek (2008) to consider the actual connectivity structure of the road/rail network in Austria at
NUTS3, including weighting matrices based on logistics performance in neighboring regions. One
advantage of using this framework is that it allows nearby regions to enter into the determination of
spatial lags, with the weight assigned increasing directly with neighbors’ logistics performance.
Therefore, previous research has shown that spatial correlation exists in heavily broken down
geographical data (LeSage and Llano, 2006; LeSage and Polasek, 2008; Alamá-Sabater et al, 2012).
Nonetheless, LeSage and Polasek (2008) only considered the land (road/rail) transportation routes
that pass through Austrian regions, but they did not distinguished among them, and did not consider
3
sectorial disaggregation. Furthermore, these authors pointed out that an issue that could be of great
importance is that of accessibility (or what we call transport connectivity) and stated that “this could be
quite different for rail versus road networks. For the case of commodity flows under examination here,
an important factor would be the relative amounts of rail versus road transportation of commodities. In
many parts of the USA, where an extensive road network exists and commodities are transported
primarily by road with few natural barriers such as mountains, rivers, or lakes, the unmodified
approach to forming the spatial weight structure set forth in LeSage and Pace (2008) should work well.
Our empirical illustration involves rail and truck commodity flows between 35 regions in Austria where
mountains and other natural barriers as well as more limited road networks place limitations on
access” (LeSage and Polasek, 2008; pp.230-231).
This paper analyzes interregional trade flows in Spain. Spain exported 73% of its total exports to the
European Union (EU) in 2009 and 67% of its total imports came from the EU in the same period.2
However, that country is far from the centre of economic activity in Europe, isolated on its peninsula
with Portugal from the rest of the continent. There are only two major road and rail passages to France
and the rest of Europe through the Pyrenees (Hendaye/Irun and Cerbere/Portbou). In fact, the
Mediterranean Corridor, which connects the Iberian Peninsula to the rest of Europe, has been pre-
identified in the EU core network in the field of transport (EC, 2011).
In addition, the country is mountainous with average altitude standing at 610 meter. These conditions
make access to the main ports and connecting to the European network via France particularly
important to connect interregional and international trade flows. Therefore, as is the case of Austria in
LeSage and Polasek (2008), the effect of transport connectivity on interregional Spanish trade flows
could be quite different depending on the type of transportation network considered, and then different
typologies of logistics platforms are distinguished.3 Methodologically speaking, we disentangle the role
of sea ports by introducing their presence in neighboring regions, as well as we consider their relative
importance in weighting matrices.4
LeSage and Llano (2006) and Alamá-Sabater et al (2012) already accounted for spatial dependence
by using Spanish regions in a gravity framework, involving origin-destination flows. Additionally,
Alamá-Sabater et al (2012) took the analysis of different sectors a step further by following a spatial
pattern in accordance with the structure of territory and the type of economic sector. Although both
LeSage and Llano (2006) and Alamá-Sabater et al (2012) focused on Spanish regions and their
results revealed a spatial pattern, they also revealed the limitations of the level of territorial breakdown
chosen; Autonomous Communities (NUTS2), which are a too large basic unit and too heterogeneous
to be treated as a whole. It is therefore necessary to reduce the spatial level and consider a smaller
basic unit area, as LeSage and Polasek (2008) did for the case of Austria. In this paper, we reduce the
geographical scale to provincial level (NUTS3), and we do not only provide evidence regarding the
2 Source: Instituto Nacional de Estadística, INE. 3 The RELOG project (“Red Logística Española”) has for the first time compiled comprehensive data on the Spanish network of logistics platforms. Professor Celestino Suárez-Burguet led this project. 4 It is important to note that a number of sea ports concentrate the main flows of merchandise traffics in Spain. In particular, the most important sea ports in terms of sea traffic (tonnes) are Bahía de Algeciras, Valencia, Barcelona and Bilbao.
4
convenience of introducing spatial dependence in gravity models of trade when analyzing the role of
transport connectivity on regional and sectorial trade flows, but we also obtain unbiased sectorial
elasticities by logistics platform type which capture the magnitude of the impact of different types and
modes of transport connectivity on interregional trade flows.
Focusing on logistics platforms and disentangling them by typology is of great interest as if
intermodality is an issue and logistics platforms mainly connect international and interregional-inter-
modal flows, the most important logistics platforms will be those located in the coast; additionally,
when different sector specific trade flows are considered, such as is the case of the “oil product”, the
logistics platforms of interest would be related to refineries, airports, chemical clusters, etc, which
could be located at the other side of a railway or a pipeline.
The rest of the paper is organized as follows. Section two describes the spatial econometric flow
model and highlights the main hypotheses to be tested. Section three outlines the data and variables
used in the study. The empirical analysis is performed in section four and finally, section five contains
the conclusions and policy implications.
2. The model and main hypotheses
The purpose of flow models is to explain variation in the magnitude of flows between each origin-
destination (O-D) pair. The model introduced by LeSage and Pace (2008) is based on the type of
spatial auto-regressive models appearing in equation (1):
y = ρ�W�y + ρWy +ρ�W�y+∝ +βX + β�X� + γD + ε (1)
As in gravity models (Bergstrand, 1985 and 1989; Deardorff, 1995), X’s matrix captures the
characteristics of origin and destination regions that could influence bilateral trade, as well as the
distance between the main city in origin-destination regions (D). Each variable produces an n2 by 1
vector with the associated parameters at origin i, βo, and destination j, βd. The dependent variable
represents an n by n square matrix of interregional flows from each of the n origin regions to each of
the n destination regions, where each of the n columns of the flow matrix represents a different
destination and the n rows represent origins. As in LeSage and Pace (2008), the model matrices are
defined as �� = �� ⊗ �, �� = � ⊗ �� and �� = �� ⋅ ��.
W matrix represents an n by n spatial weight matrix based on a neighbor’s criteria as geographical
first-order contiguity. Non-zero values for elements i, j denote that zone i is a neighbor to zone j, and
zero values denote that zones i, j are not neighbors. The elements on the diagonal are zero to prevent
an observation from being defined as a neighbor to itself.
The spatial lag vector ��y would be constructed by averaging flows from neighbors to the origin
region, and parameter ρ1 would capture the magnitude of the impact of this type of neighboring
observation on the dependent variable. The spatial lag vector ��� would be constructed by averaging
flows from neighbors to the destination region, and parameter ρ2 would measure the impact and
significance of flows from origin to all neighbors of the destination region. Finally, the third spatial lag in
the model ��� is constructed using an average of all neighbors to both the origin and destination
regions. Estimating parameters ρ1, ρ2 and ρ3 provides an inference of the relative importance of the
three types of spatial dependence between the origin and destination regions.
In order to test whether incorporating transport connectivity information into the spatial structure of the
model results in substantial differences in the estimates, a first
based on first-order contiguous neighboring
Ww. Second, the model was estimated based on
conjunction with the restriction that only first
spatial lags. Finally, we disentangle the role of sea ports by introducing
regions, as well as we consider their
The three spatial matrices used in the present study are represented in Figure 1. Matrix
based dependence) captures the spatial relationship between trade of regions
D) and B, matrix Wd (destination
neighboring B (E and F), and matrix
neighboring A (C and D) and regions
Figure 1: Trade flows taken into account (
In relation to transport connectivity,
arise from Figure 1. On the one hand, a low quality of transport networks in one region compared to its
neighbors could be an incentive for firms to locate their activities in a region with better transport
connectivity (diversion effect). On the other hand,
an origin region to a destination region would create similar flows to
effect). Alamá-Sabater et al (2012) showed that the creation effec
using interregional trade data at NUTS
particular province (NUTS3) could benefit from its
effect is higher than the diversion effect
level of disaggregation of geographical data, the greater
creation effect outweighs the diversion effect), as it is difficult
unit could produce many goods without the help of the surrounding areas, as well as
small economic unit would not benefit
markets if it could not reach without crossing them
two different effects related to the fact that
5 Relative importance of each Spanish then if a region has an important sea port and share a border, the matrix element is near 1; otherwise, if they border one another but the sea port moves a reelement is near zero and if they do not border one another the matrix element is zero.6 When considering logistics platforms, it seems plausible for the trade creation effect to be higher than the trade diversion effect the higher the level of disaggregation as, for example, Gerona (NUTS3) benefits from transport infrastructures in Barcelona and Zaragoza to reach the market in Madrid.
In order to test whether incorporating transport connectivity information into the spatial structure of the
model results in substantial differences in the estimates, a first variant of the model w
neighboring regions to construct the weighting matrices
estimated based on a matrix W which considers transport connectivity in
conjunction with the restriction that only first-order neighbors are included in the formation of the
we disentangle the role of sea ports by introducing their presence in neighboring
their relative importance in weighting matrices.5
es used in the present study are represented in Figure 1. Matrix
based dependence) captures the spatial relationship between trade of regions neighboring
destination-based dependence) reflects trade between A and
, and matrix Ww (O-D-based dependence) considers trade between
regions neighboring B (E and F), the three matrixes
Trade flows taken into account (Wo, Wd, Ww, respectively).
In relation to transport connectivity, Alamá-Sabater et al (2012) stated that two opposite effects might
arise from Figure 1. On the one hand, a low quality of transport networks in one region compared to its
could be an incentive for firms to locate their activities in a region with better transport
on effect). On the other hand, it seems plausible that forces leading to flows from
an origin region to a destination region would create similar flows to neighboring destination
Sabater et al (2012) showed that the creation effect overcomes the diversion effect by
nterregional trade data at NUTS2. Therefore, the first hypothesis to be tested is whether
could benefit from its neighbors’ transport networks, and then the creation
than the diversion effect (H1). The second hypothesis to be tested is that
geographical data, the greater we expect the positive effect
the diversion effect), as it is difficult to imagine that a small spatial economic
unit could produce many goods without the help of the surrounding areas, as well as
not benefit from the transport networks of surrounding areas to reach
d not reach without crossing them (H2).6 The third hypothesis tests
related to the fact that most shipments occur only between very proximate location
Spanish sea port is calculated by using a 0-1 standardization measure, then if a region has an important sea port and share a border, the matrix element is near 1; otherwise, if they border one another but the sea port moves a relative low quantity of sea traffic, the matrix element is near zero and if they do not border one another the matrix element is zero.
When considering logistics platforms, it seems plausible for the trade creation effect to be higher than sion effect the higher the level of disaggregation as, for example, Gerona (NUTS3)
benefits from transport infrastructures in Barcelona and Zaragoza to reach the market in Madrid.
5
In order to test whether incorporating transport connectivity information into the spatial structure of the
he model was estimated
matrices Wo, Wd, and
which considers transport connectivity in
are included in the formation of the
their presence in neighboring
es used in the present study are represented in Figure 1. Matrix Wo (origin-
neighboring A (C and
reflects trade between A and regions
considers trade between regions
are n2*n2.
two opposite effects might
arise from Figure 1. On the one hand, a low quality of transport networks in one region compared to its
could be an incentive for firms to locate their activities in a region with better transport
it seems plausible that forces leading to flows from
destinations (creation
t overcomes the diversion effect by
the first hypothesis to be tested is whether a
, and then the creation
The second hypothesis to be tested is that the higher the
the positive effect to be (the
to imagine that a small spatial economic
unit could produce many goods without the help of the surrounding areas, as well as the fact that a
surrounding areas to reach
The third hypothesis tests the co-existence of
most shipments occur only between very proximate location
1 standardization measure, then if a region has an important sea port and share a border, the matrix element is near 1; otherwise,
lative low quantity of sea traffic, the matrix element is near zero and if they do not border one another the matrix element is zero.
When considering logistics platforms, it seems plausible for the trade creation effect to be higher than sion effect the higher the level of disaggregation as, for example, Gerona (NUTS3)
benefits from transport infrastructures in Barcelona and Zaragoza to reach the market in Madrid.
6
pairs. First, goods are more likely to be exported to a particular importing region when price becomes
competitive with local alternatives, as a good transportation connection network to surrounding regions
increases neighbors’ exports (origin-dependence). Second, a good transportation connection network
to surrounding regions might also be highly beneficial to satisfy the existing local industrial demand
and then to increase neighbors’ imports (destination-based dependence). The hypothesis stated by
Hillberry and Hummels (2008) with highly disaggregated trade data that variation in regional industrial
structure should help explain the pattern of bilateral shipments would be supported if destination-
dependence is found to be higher than origin-based dependence. Additionally, Spanish interregional
traders might find it easier to cooperate in destination than in origin, as the industrial structure forces
importers to cooperate in logistics to benefit in terms of trade costs to a higher extent than exporters
(H3). Hypothesis four states that first-order contiguity dependence may include a series of factors such
as cultural proximity, a shared history, and a perception of closeness and information costs rather than
acting exclusively as a proxy for transport connectivity. By estimating the two variants of equation (1),
the constructed from first-order contiguity relationships (first-order contiguity) and the model based on
transport connectivity considerations (connectivity), the effect of transport connectivity is isolated.
Therefore, coefficients ρ�, ρ and ρ�in the connectivity model (by different types of logistics platforms)
should be lower in magnitude than those obtained in the first-order contiguity model (H4), as the
connectivity model is a more accurate way to estimate the role of transport connectivity, understood in
the broad sense. Finally, we disentangle the role of sea ports on trade of different sectors, as if
intermodality is an issue and logistics platforms mainly connect international and interregional-inter-
modal flows, neighboring to logistics platforms located in the coast will play a significant role on
interregional trade flows (H5).
3. Data and variables
We generate a dataset with total commodity flows transported between 47 Spanish regions
(provinces)7 during the year 2007.8 As we are considering the interregional trade in the mainland and
the effect of trade with bordering regions, the Canary Islands and the Balearic Islands, Ceuta and
Melilla are not taken into account.9 The regions were based on the NUTS3 and the interregional trade
flow matrices were supplied by C-Interreg.10 We used 16 origin-destination matrices; one with total
trade flows in tonnes, while the others correspond to 15 branches of activity.11 We focus on extending
gravity equations and then consider a number of the characteristics of the origin and destination
regions. In order to construct the matrices Xo (origin) and Xd (destination), and following LeSage and
Polasek (2008), we used the log of the area, the log of population, the log of GDP per capita and the
7 See Figure A.1, in Appendix. 8 We have worked with 47 regions, so the weighting matrix is 2209 rows and 2209 columns (47x47). 9 Note that neither “ship-land” nor “plane-land” connections between the peninsula and the islands play a significant role to connect international and interregional-inter-modal flows, as there is not an important commodity trade flow between inner regions and the islands. Islands-peninsula logistics platform connections are used mainly for passengers transportation. 10 We are using “unrestricted” trade flows supplied by C-intereg project. This dataset is explained in Llano et al (2008), where a difference is made between “restricted” and “unrestricted” flows. 11 See Table A.1 in the Appendix.
7
log of unemployment in each region12 as explanatory variables. A vector of (logged) distances (km)
between the capitals of each O-D region was also included as an explanatory variable, along with an
intercept vector. We would expect area, population and GDP per capita to display a positive sign,
leading to higher levels of commodity flows (weights) in both the origin and destination regions. The
coefficient of unemployment is expected to present an ambiguous sign as this variable might be
reflecting sector-specific characteristics such as the degree of resources intensity and technological
innovation achievement, whereas the expected coefficient estimate on distance is ambiguous. First,
distance might be negative, indicating a decrease in commodity flows with distance and, second,
distance variable might be positive if Spanish provinces trade more, for example, with the largest
economic centers, which are not necessarily the nearest ones when a much disaggregated territorial
level is taken into account.
In order to explain the model and the dependent variable in equation (1), we generate an n2 by 1
vector by stacking the columns of the matrix. If we consider a model with 4 regions, the flow matrix
would be represented as in Table 1. Columns show the dyad label (4 origin regions x 4 destination
regions = 16), identifier (ID) of the origin region and ID of the destination region, y denotes the
dependent variable (exports) and X’s the explanatory variables (area, population, GDP per capita and
employment, together with geographical distance). Only four regions (Seville, Zaragoza, Barcelona
and Madrid) are considered in Table 1 for simplicity.
Table 1: Data organization
Dyad label
Region origin ID origin
Region destination
ID destination
Origin explanation variables
Destination explanation variables
Distances
Y X1 X2 X3 X1 X2 X3
1 Seville 1 Seville 1 y11 a11 a12 a13 b11 b12 b13 d11
2 Zaragoza 2 Seville 1 y21 a21 a22 a23 b11 b12 b13 d21
3 Barcelona 3 Seville 1 y31 a31 a32 a33 b11 b12 b13 d31
4 Madrid 4 Seville 1 y41 a41 a42 a43 b11 b12 b13 d41
5 Seville 1 Zaragoza 2 y12 a11 a12 a13 b21 b22 b23 d12
6 Zaragoza 2 Zaragoza 2 y22 a21 a22 a23 b21 b22 b23 d22
7 Barcelona 3 Zaragoza 2 y32 a31 a32 a33 b21 b22 b23 d32
8 Madrid 4 Zaragoza 2 y42 a41 a42 a43 b21 b22 b23 d42
9 Seville 1 Barcelona 3 y13 a11 a12 a13 b31 b32 b33 d13
10 Zaragoza 2 Barcelona 3 y23 a21 a22 a23 b31 b32 b33 d23
11 Barcelona 3 Barcelona 3 y33 a31 a32 a33 b31 b32 b33 d33
12 Madrid 4 Barcelona 3 y43 a41 a42 a43 b31 b32 b33 d43
13 Seville 1 Madrid 4 y14 a11 a12 a13 b41 b42 b43 d14
14 Zaragoza 2 Madrid 4 y24 a21 a22 a23 b41 b42 b43 d24
15 Barcelona 3 Madrid 4 y34 a31 a32 a33 b41 b42 b43 d34
16 Madrid 4 Madrid 4 y44 a41 a42 a43 b41 b42 b43 d44
12 The Spanish Statistical Institute (INE) is the data source of explanatory variables. Regarding the dependent variables, dates refer to 2007.
8
Alamá-Sabater et al (2012) constructed weighting matrices by using a geographical criterion
(contiguity-based model) and introducing the presence of logistics platforms (transport connectivity
model), i.e. the regions adjacent to A (origin) or B (destination) that also have logistics platforms. In
order to proxy the quality and level of logistics factors between O-D regions, they calculated a
connectivity index as a simple average of two dimension indices, these being the number and size of
logistics platforms.
Figure A.2 in Appendix presents an example to illustrate the accuracy difference regarding transport
connectivity on interregional trade flows between both the first-order contiguity and the connectivity
model. In the example, Zaragoza has eight neighbors (Huesca, Lleida, Tarragona, Teruel,
Guadalajara, Soria, La Rioja and Navarra) and, whereas in the contiguity model the eight regions
present equal weights (the geographical criteria are the same for all regions), in the connectivity model
the imposed filter weights the eight regions with the connectivity index defined above, and then three
regions present the highest weightings (Tarragona, Guadalajara and Navarra).
Following Alamá-Sabater et al (2012), Figures 2 and 3 show the number of logistics platforms and the
logistics surface area by Spanish province, respectively. Madrid, Barcelona, Zaragoza and Cadiz,
have the largest surface area of logistics platforms, mainly due to the presence of very large logistics
platforms in these regions (such as the Zaragoza Logistics Centre in the region of Aragon, the Madrid
Barajas centre in the Madrid region and the Port of Algeciras in Andalusia), which increase the
average size of platforms. Provinces such as Valence, in the Valencian Community, and a number of
provinces in Andalusia-Extremadura (Seville, Malaga, Granada and Badajoz13) also present a large
number of logistics platforms. In contrast, provinces in Extremadura, Castile La Mancha and Castile
and Leon show a real shortage of square meters devoted to logistics activities. The Balearic and
Canary Islands are also home to only a small number of large platforms linked to their ports. This
transport connectivity picture is in line with the international or supra-regional intermodal nodes
identified in PEIT (2005), which are located in the area of Madrid, the area of Barcelona/Catalonia, the
area of the Basque Country, and Valencia, Zaragoza, Algeciras and Seville, as well as with the main
national combined traffic corridors that are located on the Mediterranean Axis, the Central Corridor
(Asturias-Madrid, Basque Country-Madrid and from here to Andalusia) and the Ebro Axis. Traffic levels
are also significant in the Madrid-Levante Corridor.
Figure 2: Number of logistics platforms (by Spanish region in 2007).
13 The Madrid-Badajoz-Portugal axis is also a corridor of great importance.
9
Source: The RELOG project.
Figure 3: Logistics surface area (by Spanish region in 2007) as a percentage of total logistics surface
area in Spain.
Source: The RELOG project.
19
0
NUMBER OF LOGISTIC PLATFORMS
17.48
0
2007
LOGISTIC SURFACE AREA
10
In order to introduce logistics characteristics by modifying weighting matrixes in a spatial
autoregressive specification to proxy for transport connectivity, Alamá-Sabater et al (2012) calculated
a connectivity index as a simple average of the number and size of logistics platforms. Scores of every
dimension were derived as an index relative to the maximum and minimum achieved by both origin
and destination regions, based on the assumption that logistics play a comparable role in O-D. The
performance of the connectivity index took a value between 0 and 1 calculated according to equation
(2):
)minmax(
)min(
valueobservedvalueobserved
valueobservedvalueactualCI
−
−= (2)
According to this index, if regions i, j have good logistics infrastructure and share a border, the matrix
element is near 1; otherwise, if they border one another but the logistics infrastructure is poor, the
matrix element is near zero and if they do not border one another the matrix element is zero.
Nonetheless, a number of shortcomings should be mentioned related to the definition of what is
considered as “logistics platform”.
One possible definition of logistics activities is the process of planning, implementing, and controlling
the efficient, cost effective flow and storage of raw materials, in process inventory, finished goods and
related information flows from point of origin to point of consumption for the purpose of conforming to
customer requirements.14 According to this definition, under the RELOG project, logistics platforms are
locations where goods can be stored, transshipped between different means of transport and where
their journeys are organized. The facilities considered logistics platforms are: dry ports, logistics
platforms, logistics zones, centers for commodity exchanges, inter-modal centers, logistics centers,
transport centers, ports, loading terminals, centers for boarding located in airports and merchandize
terminals. Then, a number of questions arise. First, whether the “logistics platform” term includes “free
ports” or “free zones”, located in the ports. These areas are mainly connected to the activity of
international customs and are built mainly for promoting the international trade (not the interregional
one). Therefore, if the size and number of “logistics platform” include these areas, the situation of the
coastal regions, with large “maritime ports” and intense international trade in/out- flows, will be biased
compared to the inner regions. Second, whether the “logistics platform” term includes “stocking-
facilities” and then, whether they are owned by the firms producing the commodities (self-stocking-
facilities) or by “logistics operators”. And third, whether the “logistics platforms” term includes “oil-
chemical-gas” tanks strategically located around big airports, ports and refineries. Depending on the
answer to these questions, the effects captured by the model will be different, as well as the
interpretation of the results obtained in sectorial regressions.
Therefore, in the empirical analysis, we firstly run regressions by using highly disaggregated regional
trade data at NUTS3 with the same methodology as used in Alamá-Sabater et al (2012). Thus is, our
first variant of the model includes only first-order contiguity (contiguity-based model), whereas our
second variant of the model (transport connectivity model) reflects the logistics performance in
Spanish provinces by using surface and size of logistics platforms. Finally, logistics platforms by
typology are disentangled for a better understanding of the role of spatial and transportation networks
14 US Council of Supply Chain Management. Webpage: www.cscmp.org
11
dependence on interregional trade flows. Table A.2 in Appendix lists logistics platforms included
separately under every issue.
4. Empirical analysis
4.1. Descriptive analysis
First of all, we present a map of Spain showing regions containing the total trade flows, as export-trade
(Figure 4) and as import-trade (Figure 5). The areas where the most important trade flows are
concentrated are identified with darker colors (darker red colors reflect higher levels of flows, while
lighter red colors indicate lower flow levels). These maps represent total trade flows, so the analysis
should be carried out from a general point of view. According to our data, the Spanish regions with the
greatest outward and inward intensity are Barcelona, Madrid, Seville and Valencia.15
Figure 4: Spanish regions (NUTS3) by export intensity.16
Figure 5: Spanish regions (NUTS3) by import intensity.
15 Note that the maps on intensity of trade flows are measured in tonnes, therefore, they do not control for the value/volume relation of the flows. As a consequence, a number of regions could appear to be as very important trading regions, but they are not. 16 These maps are constructed by setting flows within regions to zero to emphasize interregional flows.
1.5e+08
720621
Total trade (tonnes) 2007
SPANISH PROVINCES (NUTS 3) BY EXPORT INTENSITY (OUTFLOWS)
12
Finally, Figure 6 shows the map of the connectivity index from equation (2) in the case of NUTS3.17
Examining the maps in Figures 4 and 5 in conjunction with that of the logistics network in Figure 6,
there appear to be more flows in origin and destination regions in the provinces where logistics
networks are more extensive than in provinces with less developed logistics networks.
In the Spanish case, a clear differentiation can be made between provinces in terms of logistics
performance and then, this descriptive test emphasizes the need of explicitly incorporating such prior
information into the spatial and networks dependence structure of a spatial econometric flow model
when analyzing trade flows, as it might result in substantial differences in the estimates and
inferences.
Figure 6: Spanish regions (NUTS3) - Connectivity index
17 The regions containing the highest logistics performance index values are dark red.
1.5e+08
4.5e+06
Total trade (tonnes) 2007
SPANISH PROVINCES (NUTS 3) BY IMPORT INTENSITY (INFLOWS)
13
4.2. Main results
In order to analyze the spatial dependence of interregional Spanish trade flows, we estimate equation
(1) by maximizing the log-likelihood function concentrated with respect to the parameters ρ1, ρ2 and
ρ3, and the parameters βi.
Our first variant of the model includes only first-order contiguity, whereas our second variant of the
model reflects the logistics performance in Spanish regions discussed in Section 3, as we employ a
matrix W which considers logistics performance in conjunction with the restriction that only first-order
neighbors are included in the formation of the spatial lags. This results in a direct relationship between
increased numbers of the nearest neighbors and the performance of the logistics segments that go on
to form the spatial lag variables. Full results are only presented for the transport connectivity model for
simplicity.18
Different columns in Table 2 present the results obtained when estimating equation (1) for total trade
and different activity branches (R1-R15). Column (1) shows that area, population and income display
the expected positive sign and are significant. The bigger surface, population and income are in a
region the higher trade flows. Unemployment is found to be not significant, whereas distance is
positive signed and significant.19 The positive sign found in the variables of area, population and
income, in conjunction with the positive sign found in distance variable might be pointing towards the
idea that provinces trade more with the largest economic centers, which are not necessarily the
nearest ones when a much disaggregated territorial level is taken into account. The negative sign for
distance variable found in sectors R4 (Textile and Clothing), R5 (Leather and Footwear Industry), R9
(Manufactures of Rubber and Plastic Products) and R12 (Manufactures of Machinery and Mechanical
18 The full results for the first-order contiguity model are available upon request from the authors. 19 Similar results are obtained for the first-order contiguity model.
Deviation from the mean[+]Deviation from the mean[-]
CONNECTIVITY INDEX
14
Equipment) seems to indicate a higher importance of interregional land (road/rail) transportation costs
in these sectors, as the higher the geographical distance, the lower trade, and as a consequence they
might be tending to locate nearer to the most important economic and geographical Spanish centers.
The variable of occupation/unemployment is found to be positive and significant in sectors R1
(Agriculture, Forestry and Fishing), R2 (Mining and Quarrying), R11 (Basic Metals and Fabricated
Metal Products), R14 (Manufactures of Transport Equipment) and R15 (Diverse Industries). This result
might be reflecting that these industries are intensive in labor, as a higher number of workers engaged
in these industries (higher occupation/lower unemployment), the higher production and exports.
With regard to the sectorial parameters which capture the magnitude of the impact of transport
connectivity on Spanish interregional trade flows (ρ1, ρ2 and ρ3), we find a positive and significant
effect of transport connectivity, understood in its broad sense, on total trade flows. Therefore, a
particular region benefits from its neighbors’ transportation networks, as the creation effect is higher
than the diversion effect and the spatial lags for the origin and destination (associated with parameters
ρ1 and ρ2) average of neighboring regions on the logistics network are positively associated with the
level of commodity flows (H1).
15
Table 2: Estimates from the transport connectivity spatial model Total
trade R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15
Origin Area 0.73***
(4.16)
1.18***
(5.70)
0.32
(1.52)
0.42*
(1.91)
0.09
(0.94)
0.23***
(2.58)
0.42**
(2.33)
0.26
(1.55)
0.66***
(3.32)
0.33**
(2.12)
0.62***
(2.94)
0.84***
(4.18)
0.17
(1.46)
0.47***
(2.76)
0.26
(1.50)
0.57***
(3.83)
Origin
Population
1.25***
(11.34)
1.12***
(9.16)
1.05***
(8.40)
0.98***
(7.37)
0.49***
(8.59)
0.52***
(10.13)
0.87***
(8.14)
1.32***
(11.94)
1.25***
(10.27)
1.17***
(12.76)
1.39***
(10.68)
1.59***
(12.47)
0.21***
(3.27)
1.15***
(11.00)
0.93***
(8.80)
1.01***
(11.47)
Origin GDPpc 2.42***
(3.93)
4.20***
(5.72)
2.32***
(3.13)
2.02***
(2.62)
0.26
(0.73)
0.23
(0.70)
1.34**
(2.12)
2.39***
(4.01)
2.55***
(3.67)
1.35**
(2.45)
2.81***
(3.77)
4.97***
(6.91)
0.39
(0.95)
1.82***
(3.03)
2.16***
(3.50)
2.17***
(4.14)
Origin
Occupation
0.04
(0.74)
0.13**
(2.21)
0.11*
(1.81)
0.01
(0.21)
-0.02
(-0.61)
-0.03
(-1.12)
-0.007
(-0.13)
-0.02
(-0.45)
0.06
(0.97)
-0.05
(-1.19)
0.06
(0.99)
0.26***
(4.24)
0.02
(0.48)
0.03
(0.57)
0.09*
(1.68)
0.10**
(2.38)
Destination
Area
0.62***
(3.52)
1.22***
(5.75)
0.28
(1.32)
0.94***
(4.36)
0.11
(1.10)
0.12
(1.33)
0.44**
(2.40)
0.21
(1.22)
0.65***
(3.22)
0.31**
(2.00)
0.74***
(3.50)
0.40*
(1.93)
0.17
(1.50)
0.71***
(4.22)
0.22
(1.24)
0.32**
(2.16)
Destination
Population
0.76***
(7.33)
0.69***
(5.95)
0.66***
(5.49)
1.02***
(7.94)
0.47***
(8.25)
0.35***
(7.12)
0.43***
(4.31)
1.22***
(11.23)
1.16***
(9.32)
1.06***
(10.88)
0.95***
(7.20)
1.18***
(9.24)
0.37***
(5.78)
1.11***
(10.87)
0.96***
(8.98)
0.85***
(9.17)
Destination
GDPpc
1.76***
(2.90)
3.98***
(5.49)
1.66**
(2.22)
2.39***
(3.20)
0.13
(0.38)
0.23
(0.72)
0.47
(0.75)
2.78***
(4.64)
2.18***
(3.07)
1.49***
(2.71)
1.65**
(2.24)
4.30***
(5.98)
1.32***
(3.20)
3.34***
(5.62
2.87***
(4.62)
-0.002
(-0.005)
Destination
Occupation
0.01
(0.20)
0.20***
(3.35)
0.06
(0.99)
0.04
(0.69)
-0.05*
(-1.67)
-0.004
(-0.16)
-0.07
(-1.40)
-0.05
(-1.01)
-0.02
(-0.31)
-0.15***
(-2.90)
-0.01
(-0.12)
0.19***
(3.22)
0.06*
(1.68)
0.13***
(2.62)
0.07
(1.32)
-0.15***
(-3.06)
Distance 0.88***
(10.52)
0.60***
(5.78)
0.64***
(5.53)
0.63***
(5.94)
-0.13***
(-2.94)
-0.08**
(-2.03)
1.13
(1.46)
-0.03
(-0.39)
0.22**
(2.36)
-0.17**
(-2.44)
0.56***
(5.03)
0.19*
(1.96)
-0.17***
(-2.99)
-0.12
(-1.35)
0.009
(0.11)
-0.08
(-1.14)
Constant term -30.07***
(-8.12)
-21.65***
(-5.17)
-18.90***
(-4.36)
-25.99***
(-5.88)
-11.51***
(-5.49)
-12.13***
(-6.40)
-18.33***
(-4.88)
-16.38***
(-4.76)
-27.14***
(-6-64)
-21.09***
(-6.67)
-30.80***
(-7.00)
-16.54***
(-3.92)
-3.52
(-1.49)
-21.15***
(-6.16)
-10:86***
(-2.94)
-23.07***
(-7.43)
ρ1. Connectivity 0.22***
(5.75)
0.31***
(8.42)
0.26***
(7.17)
0.49***
(14.51)
-0.02
(-0.54)
-0.02
(-0.73)
0.24***
(6.41)
0.23***
(6.91)
0.18***
(4.87)
-0.08**
(-2.30)
0.22***
(5.69)
0.09**
(2.45)
0.08**
(2.47)
0.16***
(4.68)
0.27***
(8.46)
-0.07*
(-1.89)
ρ2. Connectivity 0.41***
(11.45)
0.41***
(12.13)
0.30***
(8.61)
0.42***
(11.56)
0.10***
(3.39)
0.12***
(4.16)
0.26***
(7.96)
0.22***
(6.62)
0.36***
(10.98)
0.22***
(6.70)
0.51***
(15.33)
0.41***
(12.64)
0.05
(1.35)
0.18***
(5.60)
0.31***
(9.98)
0.37***
(12.25)
ρ3. Connectivity 0.46***
(8.61)
0.33***
(6.91)
0.54***
(11.19)
-0.03
(-0.71)
0.05
(1.23)
0.18***
(4.30)
0.33***
(6.73)
-0.06
(-1.43)
0.19***
(4.05)
0.12***
(2.65)
0.23***
(4.88)
0.22***
(4.96)
0.25***
(4.72)
0.23***
(5.27)
-0.04
(-1.00)
0.16***
(3.54)
ρ1. First-order
contiguity
0.27***
(6.62)
0.32***
(8.07)
0.28***
(6.73)
0.51***
(14.30)
-0.01
(-0.17)
-0.05
(-1.19)
0.26***
(6.44)
0.30***
(7.92)
0.22***
(5.44)
-0.05
(-1.24)
0.23***
(5.52)
0.12***
(2.82)
0.04
(1.01)
0.19***
(4.69)
0.36***
(9.74)
-0.06
(-1.55)
16
ρ2. First-order
contiguity
0.42***
(11.91)
0.46***
(12.63)
0.32***
(7.94)
0.48***
(12.82)
0.14***
(3.52)
0.22***
(5.71)
0.31***
(8.06)
0.31***
(8.15)
0.41***
(11.48)
0.32***
(8.66)
0.61***
(17.30)
0.47***
(13.86)
0.11***
(2.71)
0.28***
(7.09)
0.39***
(10.85)
0.50***
(14.75)
ρ3. First-order
contiguity
0.53***
(9.14)
0.37***
(7.02)
0.69***
(11.92)
0.04
(0.887)
0.12*
(1.83)
0.36***
(5.04)
0.46***
(7.53)
-0.05
(-1.03)
0.22***
(4.14)
0.22***
(3.86)
0.31***
(5.62)
0.32***
(6.05)
0.44***
(5.85)
0.36***
(6.27))
-0.05
(-0.85)
0.31***
(5.02)
Notes: ***, **, * indicate significance at 1%, 5% and 10%, respectively. Z-statistics are given in brackets.
17
Overall, O-D-based dependence, i.e. that dependence considering trade between regions neighboring
origin and regions neighboring destination, is found to be of greater importance than origin-based and
destination-based dependence. If we compare these results with those obtained in Alamá-Sabater et
al (2012), they are in line with the expectation that the higher the level of disaggregation of
geographical data, the greater the positive effect of transport connectivity on interregional trade flows.
Therefore, we are able to provide empirical evidence supporting H2.
Furthermore, when different sectors are distinguished, three different patterns emerge. First, those
sectors for which origin-based dependence is the most important (R3: Food Industry and R7: Paper,
printing and Graphic Arts), where an origin region with a good transportation connection network to
surrounding regions benefits the most in terms of exports. Second, we find sectors for which
destination-based dependence is the most important (R1, R4, R8, R9, R10, R11, R14 and R15),
where a destination region with a good transportation connection network to surrounding regions
benefits the most in terms of trade. This result is in line with the hypothesis stated by Hillberry and
Hummels (2008) that variation in regional industrial structure should help explain the pattern of
bilateral shipments. Therefore, a good transportation connection network to surrounding regions is
highly beneficial to satisfy the existing local industrial demand (H3). Finally, we also find those sectors
for which O-D dependence is the most important (R2, R5, R6, R12 and R13). As previous research
which revealed a spatial pattern when analyzing interregional trade flows in Spain (Alamá-Sabater et
al, 2012), there is a consistent pattern of parameter ρ2 being of greater importance than ρ1,
suggesting that neighbors to the destination region in the analyzed logistics model represent the most
important determinant of higher levels of industrial commodity flows between O-D pairs. These results
differ of those obtained by LeSage and Polasek (2008) for the case of Austria, who find that ρ1 is
always larger than ρ2, pointing towards more importance assigned to the spatial lag involving
neighbors to the origin, relative to the destination region.
We also provide evidence supporting H4, as coefficients ρ1, ρ2 and ρ3 are always of higher
magnitude in the case of the first-order contiguity model than in the case of the transport connectivity
model. Nonetheless, the sign and significance of ρ1, ρ2 and ρ3 are similar in both models, excluding
the case of R12 (Manufacture of machinery and mechanical equipment), for which origin dependence
seems to be of higher importance than destination dependence in the transport connectivity model,
whereas the opposite conclusion holds for the first-order contiguity model.
With regards to H5, Table 3 and Table 4 present the results obtained for agriculture and industrial
sectors, respectively, when introducing only sea ports in the connectivity model.
The results obtained in Table 3 show that the spatial lags are positive and significant, and that both
origin and destination dependence seem to be of similar magnitude in agriculture. Therefore, spatial
and transportation networks play a significant role on agriculture interregional trade flows when only
sea ports are considered as logistics platforms.
Table 3: Estimates from the transport connectivity spatial model (only sea ports are considered).
Agriculture sectors
ρ1 ρ2 ρ3
R1 0.15*** 0.17*** -0.03
18
(5.92) (7.06) (-1.18) R2 0.16*** 0.18*** 0.01 (6.67) (7.36) (0.44) R3 0.17*** 0.15*** -0.16*** (7.19) (6.59) (-6.39)
Notes: ***, **, * indicate significance at 1%, 5% and 10%, respectively. Z-statistics are given in
brackets. The dependent variable is measured in Tons.
Table 4 shows that ρ1, ρ2 and ρ3 are positive and significant for the industrial exports of most sectors.
Additionally, there is a consistent pattern of parameter ρ2 being more times positive and significant
than ρ1 in a number of sectors, suggesting that neighbors of the destination region in the transport
connectivity model represent the most important determinant of higher levels of industrial commodity
flows between O-D pairs. Overall, sector-by-sector results suggest that a particular region could
benefit from its neighbors’ sea ports, and then intermodality is an issue as logistics networks connect
international and interregional trade flows.
Table 4: Estimates from the transport connectivity spatial model (only sea ports are considered).
Industrial sectors
Notes: ***, **, * indicate significance at 1%, 5% and 10%, respectively. Z-statistics are given in brackets. The dependent variable is measured in Tons.
5. Conclusions and policy implications
This paper analyses the role of transport connectivity on interregional trade flows using a spatial
approach. We find evidence of the importance of transport connectivity, understood in a broad sense,
ρ1 ρ2 ρ3
R4 0.001 0.06** 0.06** (0.08) (2.52) (2.47) R5 -0.06*** 0.09*** 0.08** (-2.59) (4.2) (2.37) R6 0.10*** 0.13*** 0.10*** (3.89) (5.47) (3.54) R7 0.13*** 0.11*** -0.10*** (5.86) (5.31) (-4.61) R8 0.10*** 0.18*** -0.01 (4.41) (7.69) (-0.49) R9 -0.05** 0.11*** 0.04* (-1.96) (5.35) (1.78) R10 0.13*** 0.21*** -0.07*** (5.15) (8.49) (-2.84) R11 0.03 0.18*** -0.02 (1.25) (8.32) (-1.05) R12 -0.004 0.02 0.08*** (-0.157) (0.82) (2.61) R13 0.06** 0.09*** -0.07*** (2.49) (3.94) (-2.98) R14 0.12*** 0.16*** -0.09*** (5.21) (7.53) (-4.07) R15 -0.03 0.16*** 0.06** (-1.33) (7.04) (2.53)
19
on trade. Additionally, it is confirmed that the gravity equation replays the determinants of interregional
trade with a large degree of significance in terms of the use of economic and geographical variables
(income, population, area, and distance).
First, we provide evidence that forces leading to flows from an origin region to a destination region
would create similar flows to neighboring destinations and then, a particular region benefits from its
neighbors’ transport networks (H1). Second, we find that the higher the level of territorial breakdown
the higher the positive effect of logistics networks on interregional exports, as a smaller territorial unit
depend to a greater extent on their neighbors’ transportation networks (H2). Third, destination-based
dependence seems to be of higher importance than origin-based dependence in the case of Spanish
interregional trade flows. Opposite to other countries for which transport networks might play a more
important role to increase price competitiveness for exporters (such as Austria), obtained results
support that Spanish provinces mostly specialize in the production of different goods and transport
networks contribute to increase imports related to that industries in destination neighboring regions, as
it is easier to cooperate as importers of commodities than as exporters (H3). By estimating the two
variants of equation (1), we obtain that the connectivity model is a more accurate way to estimate the
role of transport connectivity than the first-order contiguity model (H4). Finally, we disentangle the role
of sea ports on trade of different sectors, and we find that neighboring to logistics platforms located in
the coast plays a significant role on interregional and international trade flows; Spanish sea ports
connect intranational traders with international destinations (H5).
References
• Alamá-Sabater, L. Márquez-Ramos, L. and Suárez-Burguet, C. (2012), “Trade and transport
connectivity: A spatial approach,” Applied Economics, forthcoming.
• Bergstrand, J. H. (1985): “The gravity equation in international trade: Some microeconomic
foundations and empirical evidence”, The Review of Economics and Statistics 67(3), 474-481.
• Bergstrand, J. H. (1989): “The generalized gravity equation, monopolistic competition, and the
factor-proportions theory in international trade”, The Review of Economics and Statistics 71(1),
143-153.
• Clark, X., D. Dollar, and A. Micco (2004): “Port efficiency, maritime transport costs, and bilateral
trade”. Journal of Development Economics 75(2), 417–450.
• Deardorff, A. V. (1995): “Determinants of bilateral trade: Does gravity work in a Neo-classical
word?” NBER Working Paper 5377.
• Eaton, J. and Kortum, S. (2002): “Technology, geography, and trade”. Econometrica 70, 1741–
1779.
• European Commission (EC) (2011) “Proposal for a regulation of the European parliament and of
the council establishing the connecting Europe facility”, COM (2011) 665/3, Brussels.
• Hillberry, R. and Hummels D. (2008): “Trade responses to geographic frictions: A decomposition
using micro-data”. European Economic Review 52, 527–550.
• LeSage J.P. and Pace R.K. (2004): Introduction to Spatial and Spatiotemporal in Spatial and
Spatiotemporal Econometrics. Published by James P. LeSage and R. Kelley Pace. Vol. 18. Oxford:
Elsevier Ltd.
20
• LeSage J.P. and Pace R.K. (2008): "Spatial econometric modeling of origin-destination flows."
Journal of Regional Science 5, 941-967.
• LeSage, J. P. and Llano, C. (2006): “A Spatial Interaction Model With Spatially Structured Origin
and Destination Effects”, SSRN: http://ssrn.com/abstract=924603
• LeSage, J. P., and Polasek W. (2008): "Incorporating transportation network structure in spatial
econometric models of commodity flows." Spatial Economic Analysis 3 (2), 225-245.
• Limao, N, and Venables, AJ (2001): “Infrastructure, Geographical Disadvantage and Transport
Costs.” World Bank Economic Review 15, 451-479.
• Llano C.; Esteban, A., Pérez, J., Pulido, A. (2008): “Opening The Interregional Trade “Black Box”:
The C-Intereg Database For The Spanish Economy (1995-2005)”. International Regional Science
Review. Doi:10.1177/0160017610370701.
• Márquez-Ramos, L., Martínez-Zarzoso, I., Pérez-García, E. and Wilmsmeier, G. (2011): “Special
Issue on Latin-American Research” Maritime Networks, Services Structure and Maritime Trade.
Networks and Spatial Economics 11(3), 555-576.
• Melitz, M.J. (2003): “The impact of trade on intra-industry reallocations and aggregate industry
productivity”. Econometrica 71, 1695–1725.
• Micco, A. and Serebrisky, T. (2004): “Infrastructure, competition regimes, and air transport costs:
Cross-country evidence.” Policy Research Working Paper Series 3355, The World Bank.
• PEIT (2005), Plan Estratégico de Infraestructuras y Transporte - Strategic Infrastructures and
Transport Plan (2005-2020), Centro de Publicaciones Secretaría General Técnica Ministerio de
Fomento, Madrid. http://www.fomento.gob.es/MFOM/LANG_CASTELLANO/_ESPECIALES/PEIT/
• Sanchez, R.J., J. Hoffmann, A. Micco, G.V. Pizzolitto, M. Sgut and Wilmsmeier, G. (2003): “Port
Efficiency and International Trade: Port Efficiency as a Determinant of Maritime Transport Costs”.
Maritime Economics & Logistics, 5, 199–218.
• Wilson, J. S., C. L. Mann, and T. Otsuki (2004): “Assessing the Potential Benefit of Trade
Facilitation: A Global Perspective”. Working Paper 3224, The World Bank.
21
Appendix
Figure A.1. Provinces in Spain (NUTS3).
Figure A.2. First-order contiguity versus transport connectivity model.
Table A.1: Activity branches. R1- Agriculture, forestry and fishing R2- Mining and quarrying R3- Food Industry R4- Textile and clothing R5- Leather and Footwear Industry R6- Manufacture of wood and cork R7- Paper, printing and graphic arts R8- Chemical Industry R9- Manufacture of rubber and plastic products R10- Industry, non-metallic mineral products R11- Basic metals and fabricated metal products R12- Manufacture of machinery and mechanical equipment R13- Electrical equipment, electronic and optical R14- Manufacture of transport equipment R15- Diverse industries Source: Spanish Statistical Institute, INE, Spain (2010). www.ine.es
22
Table A.2. Spanish logistics platforms, by typology.
Air Cargo Province (NUTS3) Autonomous Region (NUTS2)
TC de Palma de Mallorca Palma de Mallorca Islas Baleares
TC de Sevilla Sevilla Andalucía
TC de Zaragoza Zaragoza Aragón
TC de Gran Canaria Las Palmas Islas Canarias
TC de Málaga Málaga Andalucía
TC de Bilbao Vizcaya País Vasco
Puerto de Sta. C. De Tenerife Santa Cruz de Tenerife Islas Canarias
TC de Fuerteventura Las Palmas Islas Canarias
TC de Tenerife sur Santa Cruz de Tenerife Islas Canarias
TC de Santander Santander Cantabria
TC de Santiago Santiago Galicia
TC de Asturias Asturias Principado de Asturias
TC de Tenerife norte Santa Cruz de Tenerife Islas Canarias
TC de Reus Tarragona Cataluña
TC de Girona Gerona Cataluña
TC de Vigo Pontevedra Galicia
TC de A Coruña La Coruña Galicia
TC de San Sebastián Guipúzcoa País Vasco
Centros de Carga Aérea de Madrid-Barajas Madrid Comunidad de Madrid
Centros de Carga Aérea Barcelona-El Prat Barcelona Cataluña
Centros de Carga Aérea Valencia Valencia Comunidad Valenciana
Carga terrestre intermodal (dry ports) Province (NUTS3) Autonomous Region (NUTS2)
Puerto Seco de Castilla-León Palencia Castilla La Mancha
Puerto Seco de Antequera Málaga Andalucía
Puerto Seco de Toral de los Vados (El Bierzo) León Castilla La Mancha
Puerto Seco La Robla León Castilla La Mancha
Puerto Seco Venta de Baños Ventasur Palencia Castilla La Mancha
Puerto Seco de Coslada Madrid Comunidad de Madrid
Puerto Seco Santander-Ebro Zaragoza Aragón
Puerto Seco TMZ. Terminal Marítima de Zaragoza Zaragoza Aragón
Puerto Seco de Azuqueca de Henares Guadalajara Castilla La Mancha
Zona de actividades logísticas (ZAL) Province (NUTS3) Autonomous Region (NUTS2)
ZAL Puerto Real (Cádiz) Cádiz Andalucía
ZAL Bahía de Algeciras Cádiz Andalucía
CILSA-ZAL del Puerto de Barcelona Barcelona Cataluña
Zaldesa Salamanca Castilla La Mancha
ZALIA Zona de Actividades Logísticas de Asturias Asturias Principado de Asturias
ZAL del Puerto de Tarragona Tarragona Cataluña
ZAL del Puerto de Vigo Pontevedra Galicia
ZAL del Puerto de Valencia Valencia Comunidad Valenciana
ZAL del Puerto de Cartagena Murcia Región de Murcia
ZAL del Puerto de Alicante Alicante Comunidad Valenciana
ZAL Sevilla Sevilla Andalucía
ZAL del Puerto de Motril Granada Andalucía
23
Parque Gran Europa ZAL Azuqueca A-2 Guadalajara Castilla La Mancha
Train load Province (NUTS3) Autonomous Region (NUTS2)
TM Madrid Abroñigal/Sta.Catalina Madrid Comunidad de Madrid
TM Fuencarral Madrid Comunidad de Madrid
TM Zaragoza PLAZA Zaragoza Aragón
TM Barcelona-Can Tunis Barcelona Cataluña
TM Vicálvaro Madrid Comunidad de Madrid
TM Los Prados Málaga Andalucía
TM Venta de Baños Palencia Castilla La Mancha
TM Miranda Ebro Burgos Castilla y León
TM Noain Navarra Navarra
TM Pla de Vilanoveta Lérida Cataluña
TM La Llagosta Barcelona Cataluña
TM Irún Guipúzcoa País Vasco
TM León León Castilla La Mancha
TM Albacete Albacete Castilla La Mancha
TM Cosmos León Castilla La Mancha
TM Valencia Fuente San Luís Valencia Comunidad Valenciana
TM Murcia Murcia Región de Murcia
Complejo Valladolid Valladolid Castilla La Mancha
TM Huelva Huelva Andalucía
TM Alcázar de San Juan Ciudad Real Castilla La Mancha
TM Salamanca Salamanca Castilla La Mancha
TM Portbou Gerona Cataluña
TM Júndiz Álava País Vasco
TM Ourense Ourense Galicia
TM La Nava de Puertollano Ciudad Real Castilla La Mancha
TM Girona Gerona Cataluña
TM Sevilla Majarabique Sevilla Andalucía
TM Sevilla La Negrilla Sevilla Andalucía
TM Torrelavega Cantabria Cantabria
TM Montcada-Bifurcacio Barcelona Cataluña
Sea Ports Province (NUTS3) Autonomous Region (NUTS2)
A Coruña La Coruña Galicia
Alicante Alicante Comunidad Valenciana
Almería Almería Andalucía
Avilés Asturias Principado de Asturias
Bahía de Algeciras Cádiz Andalucía
Bahía de Cádiz Cádiz Andalucía
Baleares Palma de Mallorca Baleares
Barcelona Barcelona Cataluña
Bilbao Vizcaya País Vasco
Cartagena Murcia Murcia
Castellón Castellón Comunidad Valenciana
Ceuta Ceuta Ceuta
Ferrol-San Cibrao La Coruña Galicia
24
Gijón Asturias Principado de Asturias
Huelva Huelva Andalucía
Las Palmas Las Palmas Islas Canarias
Málaga Málaga Andalucía
Marín y Ría de Pontevedra Pontevedra Galicia
Melilla Melilla Melilla
Motril Granada Andalucía
Pasajes Guipúzcoa País Vasco
Santa Cruz de Tenerife Santa Cruz de Tenerife Islas Canarias
Santander Cantabria Cantabria
Sevilla Sevilla Andalucía
Tarragona Tarragona Cataluña
Valencia Valencia Comunidad Valenciana
Vigo Pontevedra Galicia
Vilagarcía Pontevedra Galicia Source: The RELOG project.