Changing final demand or changing production
structures: what was the main contributor to
changes in transport emissions as a result of
intra-European trade?
A structural decomposition analysis on road transport emissions generated by intra-European trade.
Master thesis International Economics & Business University of Groningen
Author: Roos de Bruijn, S2345641 Assessor: Prof. dr. B. Los Co-assessor: Prof. dr. M.P. Timmer
Word count:
Abstract
2
Table of Contents
1. Introduction ... 3
2. Trends in emissions, transport and trade ... 6
2.1 Transport emissions ... 6
2.2 Increased trade ... 7
2.3 Changed trade ... 9
2.4 Quantifying emissions from transport ... 10
3. Methodology ... 14
3.1 Input-output tables ... 14
3.2 Calculating Europe’s transport emissions ... 19
3.2.1. The distance matrix ... 22
3.2.2 Quantities traded ... 24
3.2.3 Intermediate inputs and final demand ... 25
4. Data ... 28
4.1 WIOD 2013 extension ... 28
4.2 ArcGIS distance matrix ... 28
4.3 Value-to-weight ratios ... 30
4.4 Data on CO2 emissions ... 31
5. Results ... 32
5.1 Total emissions ... 32
5.2 Decomposing the change in total emissions ... 32
6. Conclusions ... 37
References ... 39
Appendix ... 44
Appendix A – Data sources ... 44
3
1. Introduction
‘Transport is Europe’s biggest climate problem’ (EFTE, 2016), states the European
Federation for Transport & Environment in accordance with data from the European Environment Agency. It appears that for the first time, pollutions emitted by transport have grown while those of other sectors have decreased, which makes the transport sector the biggest emitter of greenhouse gases (GHG1) of the European economy (Transport & Environment, 2016).
Genty (2012)
collected data for the
World Input-Output
Database (WIOD) on carbon dioxide (CO2) emissions. Her results on the emissions from transport in Europe are presented in figure 1 on the right. From figure 1 it can be retrieved that indeed,
emissions from
transportation have known
an increasing trend over the past decades. Besides that, it is visible that especially the emissions from air transportation increased significantly in the last years, thereby completely cancelling out the decrease in emissions in inland and water transport. Still inland transport -comprising road and rail transport- accounts for most of the emissions in the European Union.
According to multiple scientists, increases in transport emissions are inevitable regarding our current trade behaviour (Vohringer et al., 2010, Martinez et al., 2014, van Veen-Groot & Nijkamp, 1998). As transport is a service that depends on final demand, an increase in goods traded between different locations also asks for these goods to be transported between those specific locations (van Veen-Groot & Nijkamp, 1998). Furthermore, the process of globalization that made an increase in trade and thus the growth of our economy possible went hand in hand with a significant change in the nature of trade. As innovations drove the costs of transportation down dramatically, distance is now a less constraining factor. This has led to the development of a more
4 complex form of trade that stretches out over many different stages of production and over a handful of different countries (Baldwin & Evenett, 2015). This is also often referred to as trade fragmentation, a process that slices up value chains in more separate parts and thus increases the locations of production. As a result, there is more trade in intermediate parts (Johnson & Noguera, 2012), which affects transport rates through an increase in the total quantity of goods traded.
This thesis examines how trade-related transport emissions changed for Europe by mapping the emissions from road transport as a result of intra-European trade between 2000 and 2010. The emissions from transport are quantified based on a method used by Cristea et al. (2013), who performed their calculations on bilateral trade data and the distances travelled in order for the goods to be traded. This is different from the quantification method employed by Genty (2012) in figure 1, as these numbers present total CO2 emissions from transport within Europe, which also includes emissions generated by transporting non-European trade when they are executed by for example a European transport service company, or trade with non-European countries. In this thesis, only trade flows between European countries transported per road are assessed to account for total transport emissions. The emissions quantified here thus only present a proportion of the total emissions as presented in figure 1. However, as the quantifications in this thesis are based on trade levels and there appears to be an inclining trend for both trade in final goods trade and trade in intermediates, it is expected that the emissions from transport calculated here will also increase.
5 this research. This study employs an inter-regional extension of the WIOD 2013 release prepared by Thissen, Lankhuizen and Los. This dataset differs from previous editions in that it includes trade data on EU NUTS2 regions. This offers the opportunity to assess not only inter but also intra-regional trade. Including regional trade data is a main strength of this analysis as normally trade data is only measured on a country level, whereas now also domestic trade is included. As national transport is a major component of European road freight (Eurostat, 2017), the inclusion of regions provides the opportunity to obtain a very accurate picture of road freight emissions within Europe as a result of its trade.
This paper will be structured as follows. Section two will reflect on existing literature, showing the severe increase in transport emissions and explaining its relationship with trade. Here, expectations will be formed on the outcomes of the structural decomposition analysis presented later in this thesis, to see whether the calculations match the academic information on these issues. Section three will then explain the logic behind input-output analysis and the calculations that were performed in order to obtain the results. As this research utilises multiple external data sources, section four will shine a light on these and in section five the results on the analysis will be discussed. Afterwards, section six will present some concluding remarks and limitations to this thesis, which give incentives for further research on this topic.
Surprisingly, it appears that total CO2 emissions from transport as a result of intra-European trade decreased between 2000 and 2010. Although changing production structures led total emissions to increase, this was completely cancelled out by a decrease caused by both lower emission intensities in 2010 than in 2000 and by a change of final demand. The final demand variable was responsible for the most severe drop in emissions, as it explains 62% of the change in total.
6
2. Trends in emissions, transport and trade
2.1 Transport emissions
In 2011, the European Commission initiated its first target to reduce greenhouse gas emissions (GHG) from transport (European Commission, 2016). The target has been adopted in the 2011 White paper ‘Roadmap to a Single European Transport Area’, containing concrete steps to build a coherent, competitive and green transport system for the European Union. The aim for a greener transport system induces a 60% decrease in CO2 emissions by 2050 and for vehicles to be less dependent on fossil fuels (European Commission, 2016).
Including transport emissions in international agreements has proven to be an exception. Due to the fact that international transportation often crosses multiple country borders, is executed by a third party that is from neither trading countries, or is transported through areas which are not bound by a specific country’s regulation –like with maritime or air transport- it is hard to put blame on the pollutants from transport (OECD, 2010). The inability to assign responsibility makes it hard to control and thus to regulate emissions from transport and is the reason that they are not included under international climate agreements as the Kyoto protocol (Cadarso et al., 2010), or the Copenhagen Accord (ITF, 2010). Both these treaties commit UN nations to mitigating greenhouse gas emissions within their areas of jurisdiction. As domestic transportation only makes up a modest 10 per cent of total transport (Martinez et al., 2014), most transport-related emissions are not accounted for within these treaties. Also in the EU Emission Trading System (ETS), the world’s first and biggest emission system to combat global warming, the transport sector is excluded (European Commission, 2017). This is worrisome since the International Transport Forum (ITF) found transportation to be responsible for 15% of all greenhouse gas emissions in 2005 (ITF, 2010), this was 23% of all worldwide emissions in that same year. Grether et al. (2010) even measured transport-related emissions to have increased from one-third to three-quarters of total trade-related emissions since 1990, thereby even surpassing the emissions from the production of the goods to be transported.
7 Road transport is the biggest transport sector within Europe. It executes 49% of all EU freight transport, not to speak of the transport of passengers (EEA, 2016b). After this, maritime transport was the second largest transport mode, whereas rail accounted for only 10% and intra-EU air transport only for 0.1% of the total transport activities. As apart from aviation, road transport emits the most CO2 per kilometre of all transport modes and is the most consulted, it emits the most CO2 from all transport modes within Europe, as also appears from figure 1.
Van Veen-Groot & Nijkamp (1999) state that innovations in transport technologies and more efficient planning can result in positive environmental effects for the transport sector, if these new inventions will substitute the more polluting alternatives already in use. Nederveen et al. (2003), also describe innovations in transport technologies as the key measure to achieve emission reduction. However, the implementation of new technologies can also lead to additional transport instead of substitution for the less environmental friendly options. Gwilliam & Geerlings (1994) argue this way, stating that increased efficiency in transport will reduce fuel costs and that this will eventually lead to extra transportation. Besides that, the current competitive business environment with severe competition and demanding customers, asks for speed and flexibility in deliveries (van Veen-Groot & Nijkamp, 1999), probably leading to a shift towards the faster, more polluting transport modes.
As in this thesis, only emissions generated by one transport mode are being analysed, there is not accounted for substitution by other transport modes over the years. Based on the assumption that innovations in transport technologies occur, I expect that the emission intensities, say the CO2 emissions per ton-kilometre, are expected to have decreased between 2000 and 2010. As a consequence, it is expected that from the structural decomposition analysis will follow that total transport emissions as a result of trade decreased due to the change in emission intensity.
2.2 Increased trade
Emissions from trade arise from two main sources, according to Cristea et al. (2013), namely from the production of the goods traded, and from the transportation of these goods between trading partners. According to van Veen-Groot & Nijkamp (1999), transport is a service that is driven by demand and thus directly depends on the size of trade flows. Furthermore, Lejour (2003) states that trade depends on the distance between demand and production and that this reflects the demand for transport.
8 of new transport modes drove the costs of transportation down significantly which was the main factor improving mobility. Baldwin et al. (2015) refer to this as the first unbundling of globalisation, where the industrial revolution brought new innovations -such as steam engines for ships and trains- with lower transport costs making it possible to separate production from consumption, thereby developing comparative advantages and thus bringing incentives to trade.
Trade and transport thus seem to be dependent on each other and move in cohesion. The OECD (2010) predicts that as a result of increased trade, demand for international transport will increase rapidly, continuing its growth path of pre-crisis years. However, Martinez et al. (2014) found in a research where they simulate trade and transport growth until 2050 that these two variables grow disproportionally. In their results, a growth in trade with factor 3.4 would result in an even bigger growth of global freight, namely with factor 4.3 measured from its 2010 level. They measured that in turn, global CO2 emissions would grow with a factor 3.9 in this scenario, indicating that part of the emissions were indeed evaded due to technological innovations, as mentioned in the previous paragraph. As the volumes traded will not deviate so much from the volumes transported, the disproportionate growth in transport can be explained by an increase in hauling distance, which is the distance between two trading partners. Martinez et al. (2014) simulated the average hauling distance to increase with 12% between 2010 and 2050, which inevitably increases transportation services required.
The economic crisis was an exceptional period in terms of international transport, which again emphasizes the interdependency between transport and trade. Timmer et al. (2016) studied world input output tables at the time of the crisis and call the period from 2008 to 2010 the ‘great trade collapse’. In accordance, the International Transport Forum (OECD/ITF, 2010) found that world transport volumes peaked in 2008 and decreased afterwards. For the EU, the crisis resulted in a decline of 15% in international transport and 9% in national transport comparing 2009 with 2008 levels (Eurostat, 2011). For road freight this decrease was 10%, thereby destroying six years of growth for this sector. Consequently, Nowakowska-Grunt, & Strzelczyk (2016) found a significant drop in transport emissions in 2009, which was undoubtedly the result of the crisis. They state that therefore, the market situation is the most important factor in affecting transport and its emissions. Another report of the ITF (2010) even states that the economic crisis led to the sharpest drop in transport-related emissions in the past 40 years.
9 notwithstanding a ‘kink’ in transport emissions around 2008, growth came quick again thereafter.
It became clear from this subchapter that trade and transportation move in the same direction considering their growth path. Even though the economic crisis might have had influence on trade and consequently on transport, taking into account the fast recovery of trade levels after the crisis, leads to believe that demand levels in 2010 were not lower than those of 2000. Furthermore, in their article Gurtu et al. (2017) found that inland transport was more suitable for regional trade and therefore not affected by globalization significantly. In the structural decomposition analysis conducted in this thesis, it is thus expected to find an increase in road transport emissions due to a change in the structure and level of trade in final goods between 2000 and 2010.
2.3 Changed trade
Referring back to the first unbundling discussed in the previous paragraph which significantly increased the volume of trade, the second unbundling of globalisation discussed by Baldwin & Evenett (2015) brought along a major change in the nature of trade. This second unbundling disassociated production and consumption even further by the innovation of information and communication technologies (ITC). These technologies drastically lowered communication costs, which complemented sourcing from a distance with the possibility to coordinate from a distance. This led to the formation of multinational companies that operate in multiple countries as to reap the benefits of different locational advantages. Buckley and Strange (2015) refer to this phenomenon as the Global Factory, as one company’s stages of production can be scattered all over the world. Many other terms have been used to describe this development, such as Global Value Chains (GVCs), international fragmentation of production, and the ‘slicing up’ of value chains. These all refer to the fact that production is executed in increasingly smaller steps, of which an increasing share is executed internationally. As a consequence, intermediate products have to be traded internationally as to move along the business cycle, leading to more intra-company and thus intra-industry trade. On a country level, this leads to vertical specialization, which means that a country imports intermediate inputs to produce goods it will later export (Hummels & Rapoport, 1998). As a consequence, trade in intermediate products increases significantly. The same study by Hummels & Rapoport (1998) confirms this growth as they studied input-output data of ten OECD countries and found that in the 21st century trade in intermediate inputs had increased with 20% with respect to the late 1960s. They also report that the increase in vertical specialization accounts for more than 25% of the increase of total trade. Johnson & Noguera (2012) can confirm the drastic growth of trade in intermediates, as their study on the GTAP 7 in 2004 measures this to account for two thirds of total international trade.
10 trade in final products, this also has its implications for transportation. Cadarso et al. (2010) state that an increase in transport results from both an increase in trade volume and a change in the nature of trade flows becoming more fragmented. Furthermore, they argue that fragmented supply chains seem to separate consumers from the emissions generated by production and transportation, again pointing at the responsibility problem. Furthermore, complex value chains increase the requirements of transportation. As there is more pressure for on-time delivery, firms are seeking suppliers that meet fast deliveries, even if these are located further away (Rovers, 1999).
However, also trade in intermediate products suffered from the crisis. A study of Los et al. (2015) on Global Value Chains actually found a reverse in the trend of international fragmentation. From 1995 onward, the share of ‘foreign value added’ had been following an increasing path, meaning that an increasing part of the total value if a final product is generated abroad. Controversially, they found that in 2008, a big part of global value added became substituted by regional value added. In accordance, the ITF (2009) reports that increased protectionist measures are adopted in the wake of the crisis, making transport more regionally focused.
The trend of international fragmentation of production thus seams a recent one, however slowed down due to the crisis, obtaining a more regional focus. However, as the EU is a regional trading bloc, it is not expected that intra-European trade and as a consequence transport, will suffer from this significantly. Therefore it is expected that from the structural decomposition analysis as performed in this thesis it will follow that total transport emissions as a result of intra-European trade will increase as a consequence of a change in the European production structure between 2000 and 2010.
2.4 Quantifying emissions from transport
Earlier attempts in quantifying emissions from transport are scarce, especially for transport crossing national borders. The Intergovernmental Panel on Climate Change (IPCC) quantifies emissions for the members of the United Nations of which, as mentioned, only domestic emissions are included in assessment by the Kyoto Protocol (Cadarso et al., 2010). The IPCC uses a top-down accounting approach of emissions, in which they look at total emissions in a country and then assess which share was emitted by the transport sector by looking at transportation data for that same year (IPCC, 1997). The International Transport Forum (ITF) also employs this top-down approach in their multiple quantifications, combining data on fuel combustion with emissions per fuel type. However, these top-down approaches only give a very general assessment of total emissions in a region without the possibility to link them to a certain trade flow or industry.
11 very complex and ambitious task to perform this detailed analysis on a whole sector or country. Besides that, this time-consuming analysis also includes all non-transport emissions, which are irrelevant for this thesis.
Furthermore, Cadarso et al. (2010) and Cristea et al. (2013) quantified emissions from international transport using a bottom-up approach. Using this approach, a level of transportation follows from trade flows, which are then assigned to different transport modes. Emissions are then calculated based on the quantities traded, the transport mode used and the distance between two trading regions. Cristea et al. (2013) seek to quantify total emissions resulting from trade, including both production and transportation, and find that taking into account transportation presents a very different view on emissions per country than when only production-related emissions are measured. They point out that in quantifying emissions, trade has to be thought of in terms of transportation instead of in value terms and that the conversion of trade flows into quantities traded is a critical step. Eventually they build a dataset including emissions of 36000 trade flows, which totals match very well with the top-down approach employed by the ITF. The authors include all four main modes of transportation (rail, road, maritime and aviation), whereas they do not include transport emissions resulting from domestic trade.
Cadarso et al. (2010) perform a resembling analysis on transport emissions, hence focusing on emissions generated from offshoring as a means to quantify the emissions resulting from increased fragmentation. This is executed for the Spanish economy by analysing input-output tables and thereby focusing on imports. By applying the ‘Broad Consumer Principle’, assigning emission responsibility from transport to the country of consumption, a very detailed view on emissions of Spanish imports is illustrated. Although they make a distinction between imports of final goods and imports of intermediate products used for further production, they only account for the emissions generated by transportation of first-order intermediate inputs. Accordingly, there is thus not accounted for the emissions generated by transportation further down the value chain, not presenting a complete view on transport emissions generated for Spanish consumption.
12 total amount of transports necessary to move one trade flow using a certain transport mode.
Lastly, Steenhof et al. (2006) apply a decomposition analysis on the Canadese freight transportation sector to see what factors have led to an increase in the volume of GHG emissions within Canada. In their analysis they distinguish the following causes: an increase in trade, switching to another transport mode, different type of fuel consumed and efficiency changes. They found that, despite cleaner transportation for all modes, transport emissions went up due to more economic activity and a shift towards heavier trucks.
As in this thesis the emissions from transporting traded goods are analysed, a method is used that resembles the approaches of Cristea et al. (2013), Cadarso et al. (2010) and Gurtu et al. (2017). As I am only interested in transport emissions resulting from trade and not that of production, conducting a Life Cycle Analysis would be too broad and therefore too time-consuming for the scope of this thesis. Instead, I follow the other three papers in analysing bilateral trade data as I am using input-output tables on trade within and between European regions. The emissions generated from moving a certain trade flow are then calculated by combining the trade data with the distances between trading partners. As in Cristea et al. (2013), trade flows will be handled as quantities instead of in value terms, which means the input-output tables are converted using a value-to-weight ratio. Afterwards, in order to be conclusive on the different variable’s contributions in explaining changes in European transport emissions resulting from trade, a structural decomposition analysis was used as in Steenhof et al. (2006). Another difference is that all of the authors mentioned include multiple modes of transport in their analysis, whereas I only take into account the emissions from road transport, thus drawing a more focused illustration on emissions from transport. However, I do include domestic trade and thus transport as the WIOD 2013 extension also holds data on regions, something that the other researchers neglect and definitely improves the accuracy of this work again.
Lastly, this thesis goes deeper than the study by Cadarso et al. (2010), where they also look at trade in terms of intermediate inputs and final goods but only account for the first-order transport emissions, whereas in this thesis the trade flows from all upstream suppliers are included in order to calculate trade-related transport emissions. Also, I improve Steenhof et al. (2006) in that, other from analysing the effect of trade in transport emissions, I also make a distinction in the type of trade that was most relevant for changing the emissions.
14
3. Methodology
In this section I will explain the methodology used in order to quantify the emissions resulting from transporting intra-European trade and to perform the structural decomposition analysis in this thesis. The aim of this decomposition is to figure out what was the main contributor to the increase in transport-related emissions within Europe: a change in final demand or a change in the production structure? First, I will give a thorough explanation of the structure of input-output tables and how they will be used for these calculations, as the WIOD by Thissen, Lankhuizen and Los will be the main data source of this thesis.
Details on the data sources consulted will be provided in section 4: Data.
3.1 Input-output tables
Input Output tables tell something about production structures within an economy. They are made up as a matrix consisting of different parts, showing which part of production is executed in certain countries and industries. A simplified illustration of an input-output table is presented below.
Intermediate inputs Z Final demand FD Total output x Countries (c) c1 c2 c1 c2 Industries (n) a b a b c1 a 𝑧!!!! 𝑧!"!! 𝑧!!!" 𝑧!"!" 𝑓𝑑!!! 𝑓𝑑!!" 𝑥!! b 𝑧!"!! 𝑧!"!! 𝑧!"!" 𝑧!!!" 𝑓𝑑!!! 𝑓𝑑 ! !" 𝑥 !! c2 a 𝑧!"!" 𝑧!"!! 𝑧!!!! 𝑧!"!! 𝑓𝑑!!" 𝑓𝑑!!! 𝑥!! b 𝑧!"!" 𝑧 !!!" 𝑧!"!! 𝑧!!!! 𝑓𝑑!!" 𝑓𝑑!!! 𝑥!! Primary
inputs w’ Value added … … … … Total output
x’ … … … …
Table 1: Illustration of an input-output table.2
The table is build up as a large matrix consisting of several separate parts: the Z matrix, presenting trade in intermediate inputs, the primary inputs vector w’, consisting of value added, a matrix representing final demand FD and two vectors presenting total output
x/x’. In this thesis, the main focus will be on the Z and FD matrix.
2 In matrix notation, capital letters indicate that the variable is shaped as a matrix. Lower case letters
15 A matrix shape is used in order for the table to represent production structures of different countries (c) and bilateral trade flows between them, represented by c1 and c2 in the example above. The production in a country can be divided into n industries (here: a and b). The industries in the columns are the producing or ‘demanding’ country’s industries, and those in the rows the ‘delivering’ country’s industries.
For the Z matrix, this means the following. An element in this matrix shows the deliveries of goods from a certain country-industry to another. The Z matrix has an identical number of rows and columns to assure that the country-industry in row six is the same as the country-industry of column six. Let’s take the element in the second row and the first column of the Z matrix:
𝑧
!"!!, as an example.
Here, the element z represents the unity of the cell, namely the intermediate inputs. The superscripts indicate between which countries the intermediate inputs are traded (1-1) and the subscripts indicate the industries between which this takes place (b-a). The element𝑧
!"!! thus represents the quantity of goods that industry b in country 1 delivers to industry a in the same country (1).For the final demand matrix FD, the columns indicate the demanding countries and the rows the delivering industries per country. A product is a product of final demand when it has left the production process and is being sold for different use. Originally, in input-output tables final demand is divided into different segments, rather referred to as final demand categories (k). These categories could be for example household consumption, government consumption or investments. In this thesis, the categories of final demand are not further used in the analysis of transport emissions and therefore these will not be taken into consideration. The value of typical final demand element 𝑓𝑑!!! shows how much country 1 demands products that are produced by industry a in
country 1.
Finally, the vectors x and x’ present total output3. This is the total of all produced intermediate inputs and goods sold as final products for a certain country’s industry. The first element 𝑥!! here thus shows the total production of goods of industry a in country 1,
either used as intermediate inputs for other country-industries or sold to meet final demand. The formula summarizing the total output for industry a in country 1 in input-output tables is thus:
𝑥!! = 𝑧
!!+ 𝑓𝑑!! (1)
Where the summation signs show that all intermediate inputs produced by this country-industry, independent from the country and industry they are delivered to, are included. The same holds for the sales of final goods.
3 As vectors, indicated with a bold small case letter (x), originally are column vectors, the prime in x’
16 In most input-output tables also primary inputs are taken into account, these can be found in the bottom of the table, typically consisting of value added. These are presented in a row vector, indicating the distinct inputs per country-industry. As primary inputs are not relevant for this research, they are not further specified in table 1 and in the rest of this thesis.
So, as to conclude, the input-output tables present the production processes of different industries in different countries and how these trade with each other. As the demanding country-industries are depicted above the columns of the matrix, the numbers in a typical column in matrix Z show all the inputs necessary for total production of this specific industry in this country. In that way, a row thus represents all the intermediate inputs a certain country’s industry delivers to the industries in other countries. In the same fashion, a column in the final demand matrix FD displays the total demand for final goods coming from this country, whereas the rows show from which country’s industries this is demanded.
The WIOT used for this thesis
The WIOD 2013 extension, in preparation by Thissen, Lankhuizen & Los, that is used for this thesis is a multiregional input-output table presenting world trade data for the years 2000 until 2010. It presents not only trade data on countries but for some of them also on different regions within these countries. County-regions are included for all EU-25 countries. For Bulgaria, Romania, Cyprus and a handful of non-European countries including ‘Rest of the World’, no separate regions are specified in the database. These countries are specified as one region. As the amount of regions per country differs as they are based on the NUTS2 classification of the European Union, the dimensions per country in the Z matrix also differ. Therefore, I will look at Z only in terms of regions (r), thereby neglecting the existence of the countries these regions are located in. For every region trade data is divided amongst fourteen industries, referred to as n. The total dimensions of the Z matrix in Thissen, Lankhuizen & Los (in preparation) is thus nr x nr. As the total amount of regions is 266 and there are fourteen industries per region, the dimensions of the Z matrix are thus 14x266 by 14x266, or 3724x3724.
17 However, in this thesis there will not just be worked with the Z matrix as mentioned above, as modifications will have to be applied in order to perform the structural decomposition analysis in a later stage.
First of all, let me introduce matrix A. Within this A matrix, the rows hold the amount of intermediate inputs that are necessary for one unit of output for the column industries. Calculation of the A matrix is performed by dividing the elements of Z (z) by the corresponding element of total output x for that country’s industry. This means the following calculation is performed:
𝐀 = 𝐙 ∗ 𝐱!! (2)
Where the hat indicates that the diagonal matrix of vector x was taken. As division of matrix Z by output vector x is not possible following the rules of matrix calculation4, this calculation is performed by multiplying Z with the inverse of the diagonalized output vector, which leads to the same answer.
Now, the matrix A holds input shares instead of total inputs as in matrix Z. A typical row here shows the shares of that region-industry’s intermediate inputs in the production of one unit of output for the column region-industry. Also, in a typical column of this matrix can be found that many different intermediate inputs are necessary to produce one unit of output.
However, matrix A holds the intermediate inputs necessary to produce one unit of output per column country-industry. When interested in the inputs necessary for the actual demand in the economy, the A matrix has to be multiplied with the level of final demand. For this calculation, I first converted the final demand matrix FD into a final demand vector f. This was done by summing over its columns in order to arrive at one single column presenting the total amount of goods demanded per industry. This is done since for calculating the amount of inputs required, only the level of final demand matters, not the regions this demand is coming from. This new final demand vector f shows how many final goods are demanded from each industry in each region, thus having a dimension of 14x266 by 1, or: 3724x1. Now, the multiplication with A will become Af, of which the product represents the direct inputs necessary to produce total final demand. The region-industries for which this gives an outcome, are the producers of
4
18 inputs for final demand. These are also referred to as the first tier suppliers of this economy’s supply chain.
However, as one can imagine, also these direct inputs are not made out of thin air, but again require inputs. This set of indirect inputs (inputs for the direct inputs) are calculated by multiplying the direct inputs with A another time: A(Af). The producers of these inputs are then the suppliers of the first tier suppliers, also referred to as second tier suppliers. When these second tier suppliers also demand inputs from third tier suppliers, these are calculated with A(A2f), the inputs for these products as A(A3f) and so forth. The
total output of this industry than becomes:
x = f + Af + A(Af) + A(A2f) + A(A3f) + … + A(Anf) (3)
Another way of writing this, still including all direct and indirect inputs (and thus all tier suppliers) necessary for the production of final demand is the following:
x = [(I –A)-1f] (4)
In which the part (I – A)-1 is also referred to as the Leontief inverse5. The use of this inverse is used as a shortcut to assure all trade in direct and indirect inputs needed for the production of the total demand for final goods is included, which is necessary to include all trade flows where transport emissions are generated. The calculation of this inverse can be performed over matrix A in Matlab and produces matrix M with the same dimensions. Equation four can thus also be rewritten as: x = Mf and holds the production in all country-industries in an economy in order to produce the level of final demand held in f. However, as this only presents the production in different regions to generate a certain level of final goods, a multiplication with matrix A is necessary to also include the trade between the different regions’ industries that this production entails. But the multiplication of matrix A with the vector holding the outcomes of Mf, would result again in a vector, presenting only the total trade and production to acquire a certain level of final demand. In order to specify the amounts of goods traded within specific trade flows between different region-industries as necessary to calculate transport flows later,
A is multiplied with a diagonalized version of Mf.
The calculation that now is performed is 𝐀 ∗ 𝐌𝐟 and results in a matrix presenting all intermediate inputs produced and traded in a certain economy in a certain year to meet a corresponding level of final demand. Recalling from the structure of input-output tables, this perfectly describes matrix Z and thus 𝐀 ∗ 𝐌𝐟 = 𝐙. Presenting this difficult way of calculating Z is to show that this matrix holds all direct and indirect inputs necessary to produce the total level of final demand in an economy. The notation of Z as
5 The Leontief inverse is named after its inventor: Wassily Leontief. The construction and rationale of the
19 𝐀 ∗ 𝐌𝐟 will be used in the structural decomposition analysis (SDA) as modifications will be made to this equation as to account for changes in its separate parts, which will simulate the different situations of the SDA.
3.2 Calculating Europe’s transport emissions
3.2.1. The main model
As was discussed in the literature review of this thesis, trade of goods implies transportation of these goods, and this transport generates emissions. From the last section, two situations emerge where transportation and thus transport-emissions arise. These are the trade in direct and indirect inputs necessary for the production of final demand, and the trade in final goods themselves. In this thesis, both are included in calculating transport emissions to measure both their contributions to the total change in transport emissions. Based on the attempts of Cadarso et al. (2010), Cristea et al. (2013) and Gurtu et al. (2017) the following variables were included to construct a method quantifying transport emissions: the quantities traded, an emission coefficient presenting the CO2 emissions generated when moving a ton of goods one kilometre using a certain mode of transport, and the distances to be travelled when trading.
Hence, in order to calculate total emissions from road transport (e) in a certain year (t), the following formula was used:
𝐞! = [e!∗ sum 𝐐𝐀!𝐌!𝐟!°𝐃𝐙, 2 ] + [e!∗ sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 ] (5)
Where e! presents the emission coefficient holding the CO2 emissions per ton/kilometre for road transport in a certain year, 𝐃𝐙 and 𝐃𝐅𝐃 are distance matrices for
the intermediate input matrix and final demand matrix respectively, and matrices 𝐐𝐀!𝐌!𝐟!
and 𝐐𝐅𝐃! present the quantities traded per road between all European region-industries in either intermediate inputs or final goods. As only trade flows within Europe are included for calculation of European transport emissions, the quantity matrices only present the quantities traded per road between European regions. As a result, 𝐐𝐀!𝐌!𝐟! presents trade data on fourteen industries (n) per region (r), for 252 regions in total (rEU). Its dimensions
are thus nrEU x nrEU, or 3528x3528. Also for 𝐐𝐅𝐃!, the non-European regions are
excluded for the multiplication of transport emissions. As in the final demand matrix FD, industries are included in its rows, but in its columns only one vector of final demand is included per region, its dimensions become nrEU x rEU, or 3528x252.
The notation sum(…, 2) stands for summing over a matrix’ columns. As the intermediate and final demand matrices have different dimensions in terms of columns, but not in rows, summing over their columns gives two column vectors with the same amount of rows, namely dimensions nrEU x 1, and makes it possible to add them to
20 coefficient. As the total emissions from road transport e are presented in a column vector, this vector has to be summed in order to arrive at a scalar presenting one numeric value.
Please note that for 𝐐, the letters in the subscripts indicate of which year the vectors or matrices were used for the calculation of the quantity matrix. For 𝐐𝐀!𝐌!𝐟! this
thus means that matrix A, matrix M and final demand vector f were used for its calculations. However, as Q is a matrix, we used the diagonalized outcome of Mf and thus not its vector. The letters here are thus not to indicate calculations, but to indicate the year we used for each value.
The distance matrices present the distances in kilometres between all European regions included in the dataset. As the dimensions of the Z matrix and that of FD matrix differ, two separate distances matrices are produced to make sure the right trade flows are multiplied with the corresponding distances they have to travel in order to arrive at the region traded with.
As in this thesis, I will be looking at the change in total emissions between 2000 and 2010 for Europe, the formulas in order to calculate emissions in both 2000 and 2010 are presented below.
𝐞! = e! ∗ [sum 𝐐𝐀!𝐌!𝐟!°𝐃𝐙, 2 + sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 ] (6) 𝐞! = e!∗ [sum 𝐐𝐀!𝐌!!!°𝐃𝐙, 2 + sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 ] (7)
Where the subscripts t=0 and t=1 represent a variables’ values for either the base year situation t=0 (2000) or the end situation t=1 (2010). Distances between regions are assumed not to change during the time period and therefore the distance matrices D are assumed to be constant variables.
The change in total emissions over the period studied is then calculated by:
∆𝐞 = 𝐞!− 𝐞! (8)
or:
∆𝐞 = {e!∗ [sum 𝐐𝐀!𝐌!!!°𝐃𝐙, 2 ] + sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 ]} −
21 the contributions of different variables in this growth, SDA seemed like the best analysis to perform to investigate this. The question asked in the decomposition analysis thus becomes: which part of changes in total transport emissions as a result of intra-European trade can be attributed to changes in the emission coefficient, which part to changes in the European production structure and which to changes in European final demand? The inclusion of the emission coefficient in the SDA is to see whether less-polluting alternatives were created and put to use between 2000 and 2010. When this is the case and emissions per ton-km decreased, while total emissions increased over time, the effects of increased trade levels on emissions is even more severe than presented at first sight. Inclusion of this variable in the analysis is thus a crucial step in discovering the true effects of both changes in final demand and changes in production on total emissions.
For the structural decomposition analysis, I use a method for additive decomposition with multiple variables by Dietzenbacher & Los (1998). Their method decomposes the change in the dependent variable by using two polar equations and eventually taking the mean of their outcomes. This comes from the idea that decomposition can be executed in two ways, and that there is no specific reason to prefer one above the other. Therefore, both are performed and as thereafter the mean of both calculations is used, one will arrive at the most accurate and objective answer.
The first option for decomposing equation 9 is:
𝐞!− 𝐞! = e!− e! ∗ sum 𝐐𝐀!𝐌!!!°𝐃𝐙, 2 + sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 + {e!∗ [sum 𝐐𝐀!𝐌!!! − 𝐐𝐀!𝐌!𝐟! °𝐃𝐙, 2 +
{e!∗ sum (𝐐𝐀!𝐌!!!− 𝐐𝐀!𝐌!𝐟!)°𝐃𝐙, 2 + sum 𝐐𝐅𝐃!− 𝐐𝐅𝐃! °𝐃𝐅𝐃, 2 ]} (10)
Where there is accounted for the change in emission coefficients e!− e! , the
change in production structures (𝐐𝐀!𝐌!𝐟! − 𝐐𝐀!𝐌!𝐟!) and the change in final demand 𝐐𝐀!𝐌!!! − 𝐐𝐀!𝐌!𝐟! and 𝐐𝐅𝐃! − 𝐐𝐅𝐃! . As final demand is also used in the calculations of the different Z matrices (namely in the final demand vector), also the change in the final demand vectors is taken into account when looking at the changes in total final demand as to also include the effects of its change on production structures.
The alternative way of decomposing equation 9 is:
𝐞!− 𝐞! = e! − e! ∗ sum 𝐐𝐀!𝐌!𝐟!°𝐃𝐙, 2 + sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 + {e!∗ sum 𝐐𝐀!𝐌!!!− 𝐐𝐀!𝐌!𝐟! °𝐃𝐙, 2 + {e!∗ sum (𝐐𝐀!𝐌!!! − 𝐐𝐀!𝐌!!!)°𝐃𝐙, 2
+sum 𝐐𝐅𝐃! − 𝐐𝐅𝐃! °𝐃𝐅𝐃, 2 ]} (11)
22 executed, but ‘approached’ from a different direction6. It can be seen that for a variable interacting with another variables in the initial equation 9, its change is being valued against both a base-year and an end-year situation of the variable of interaction, of which one is adopted in the first polar and the other in the second. In calculating the contributions of the changes in different variables on the change of total emissions, per variable the average should be taken of the part measuring its change in the two polar equations.
This means that the effect of changes in the emission coefficient on the change in total emissions can be expressed as:
∆e = 0.5 e!− e! ∗ sum 𝐐𝐀!𝐌!!!°𝐃𝐙, 2 + sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 +
0.5 e!− e! ∗ sum 𝐐𝐀!𝐌!𝐟!°𝐃𝐙, 2 + sum 𝐐𝐅𝐃!°𝐃𝐅𝐃, 2 (12)
Where the average is taken of the effect of changes in e on the production structure and final demand in 2010 and 2000 respectively. For production structures and final demand this results in the following two calculations:
∆𝐐𝐀..= 0.5 ∗ {e!∗ sum 𝐐𝐀!𝐌!!! − 𝐐𝐀!𝐌!𝐟! °𝐃𝐙, 2 } + 0.5 ∗ {e!∗ sum 𝐐𝐀!𝐌!!! − 𝐐𝐀!𝐌!𝐟! °𝐃𝐙, 2 ]} (13) ∆𝐐𝐅𝐃 = 0.5{e!∗ [sum 𝐐𝐀!𝐌!!!− 𝐐𝐀!𝐌!𝐟! °𝐃𝐙, 2 + sum 𝐐𝐅𝐃𝟏 − 𝐐𝐅𝐃𝟎 °𝐃𝐅𝐃, 2 ]} +
0.5{e!∗ [sum 𝐐𝐀!𝐌!!!− 𝐐𝐀!𝐌!! !)°𝐃𝐙, 2 + sum 𝐐𝐅𝐃! − 𝐐𝐅𝐃! °𝐃𝐅𝐃, 2 ]} (14) The total change in emissions and the results of the decomposition analysis will be presented in section 5: Results. Below, the computations and dimensions of distance and quantity matrices used in the main model will be further explained.
3.2.2. The distance matrix
In order to account for the distances travelled when transporting a certain flow of goods traded, a distance matrix D was used. This matrix was hand-made, especially for
6
23 this thesis. Acquiring this matrix was done using ArcGIS, more on this can be found in section 4: Data. The matrix D contains the distances over road in kilometres between all European regions in the WIOD 2013 extension, which are the NUTS2 regions for the EU25 countries, plus Bulgaria, Romania and Cyprus (these are included as just one country without regions). As for the Z matrix, I will only use the total amount of regions, not specifying these per country. Countries without specified regions will thus be presented as one region. The regions specified can also be found in table 1 appendix A.
The amount of regions for Europe is 252 in total, also referred to as rEU. This
means that the initial distance matrix D, presenting the distances between all European regions has a dimension of rEUxrEU, or 252x252.
Below one can find a simplified illustration of the distance matrix that was constructed. A typical element in this matrix 𝑑!!!!, represents the distance travelled per
road between region one and region three in kilometres. Here, the red cells indicate the distance from a region to itself. This may sound odd, but what is used here is the intra-regional distance 𝑑!"#. Recall from the red cells in the illustration of the input output
tables that also intra-regional trade occurs. As this trade also requires transportation, including intra-regional distances presents a more complete view on total emissions in the end. How I calculated intra-regional distances can be found in section 4.
Distance matrix D r1 r2 r3 r4 r5 r6 r7 r8 r1 𝑑!"# 0 𝑑!!!! 𝑑!!!! 𝑑!!!! 𝑑!!!! 𝑑!!!! 𝑑!!!! r2 0 𝑑!"# 0 0 0 0 0 0 r3 𝑑… 0 𝑑!"# 𝑑… 𝑑… 𝑑… 𝑑… 𝑑… r4 𝑑… 0 𝑑… 𝑑!"# 𝑑… 𝑑… 𝑑… 𝑑… r5 𝑑… 0 𝑑… 𝑑… 𝑑!"# 𝑑… 𝑑… 𝑑… r6 𝑑… 0 𝑑… 𝑑… 𝑑… 𝑑!"# 𝑑… 𝑑… r7 𝑑… 0 𝑑… 𝑑… 𝑑… 𝑑… 𝑑!"# 𝑑… r8 𝑑… 0 𝑑… 𝑑… 𝑑… 𝑑… 𝑑… 𝑑!"#
Table 3: Illustration of the distance matrix.
As can be seen in table 3 above, both the rows and the columns of region 2 are filled with zeros. This is to indicate that region 2 is an overseas region that is not connected to the road network and thus no distance can be calculated from this region to any other region in the network. For all overseas regions, distances are thus set to zero as to exclude transportation to this region in the calculation of road transport emissions from trade within Europe. However, as transportation within these overseas regions is assumed to happen at least partly through road transportation, internal distances of these regions are included in calculating total transport emissions.
24 Since the distance matrix has to be multiplied with the quantity matrices later, presenting trade data for all industries within a region, the distance matrix has to be modified to match the dimensions of the quantity matrices. Since the location of industries within a region is non-specified, the same distance can be used for all fourteen industries of a region to indicate its distance to one other region’s industries.
As a consequence, the distance matrix compatible with quantity matrix 𝐐𝐀!𝐌!𝐟!, has to hold the same dimensions, namely nrEU x nrEU. As the initial dimensions of the
distance matrix were rEU x rEU and the inter-regional distances are identical for each
industry within a specific region, the original matrix can be ‘stretched out’ in order to produce new distance matrix DZ. This is executed by using every element in the original
matrix D, for a matrix consisting of 14x14 cells in the new matrix DZ, stretching the
original distance matrix with dimensions 252x252 to a new distance matrix which dimensions are 3528x3528.
For the other distance matrix, suitable for multiplication with matrix 𝐐𝐅𝐃!, the dimensions should become nrEU x rEU, as only one column of final demand is included
per region since the different categories of final demand are being neglected. The new distance matrix here thus should hold the same interregional distance on each fourteen rows, but a different one for all its columns, presenting 3528x252 elements of distance in the new distance matrix DFD.
3.2.3 Quantities traded
The quantities of products traded in the European economyin any specific year are represented by the matrices 𝐐!.. and 𝐐𝐅𝐃, holding quantities traded of intermediate inputs and final demand respectively. Quantity matrices were formed by multiplying the WIOT data, presenting trade in million euros, with a vector vw, which holds the ‘value-to-weight ratios’ for all of the fourteen industries per region included in the database. This value to weight ratio transforms the trade matrices values into quantities traded, since transport and thus its emissions depend on weights and not on value (Cristea et al. 2013). A modification of the value-to-weight ratios was made to convert the trade values in euros into quantities in tons. Furthermore, since not all traded goods are moved through road transportation, a weight has to be assigned for this mode to indicate its share of total transport facilitating intra-European trade. As was mentioned in section 2.1, the European Environment Agency found that within Europe, 49% of all freight was transported per road in 2013. No data was available on the share of road transport in the years 2000 or 2010, so I divided the trade matrices by two and thereby assume that half of the total amount of traded goods within Europe is transported per road. The quantity matrices are produced in the following fashion:
𝐐𝐀… = 0.5 ∗ 𝐯𝐰 ∗ 𝐓𝐀… (15)
25 Where the vector with the value to weight ratios is first diagonalized before multiplication with the trade matrices in order to create a matrix Q eventually, which has dimensions of 3528x3528 for production and 3528x1008 for final demand. As the quantity matrices Q are a modification of trade matrices T, quantity matrix QA1M1f0 uses
trade matrix TA1M1f0 etcetera. These trade matrices will be further explained under section
3.2.4.
As discussed, vw is a vector representing the value-to-weight ratios per industry for each region. As the same fourteen industries repeat itself for every region in the trade matrix, the vw vector presents the same ratio every 15th row to make sure every identical industry is assigned the same weight7. As vw is a vector presenting information for fourteen industries for all 252 European regions, it has a dimension of 3528x1.
3.2.4 Intermediate inputs and final demand
This section discusses the creation of the initial matrices used for calculation of the quantities traded. These are based upon the methods mentioned under section 3.1 and are referred to as the trade matrices in the previous section. The initial calculations present the different situations sketched under the structural decomposition analysis. As can be found in equations 6 until 14, these are:
A0M0f0; showing the total amount and structure of direct and indirect inputs required (Z matrix) in the economy using the production structure of the year 2000 measured against the final demand level of 2000 in order to express total intermediate production in the year 2000.
A1M1f1; showing the total amount and structure of direct and indirect inputs required (Z matrix) in the economy using the production structure of the year 2010 measured against the final demand level of 2010 in order to express total intermediate production in the year 2010.
A0M0f1; showing the total amount and structure of direct and indirect inputs required (Z matrix) in the economy using the production structure of the year 2000 measured against the final demand level of 2010. In order to see the effect of changing demand levels on European production.
A1M1f0; showing the total amount and structure of direct and indirect inputs required (Z matrix) in the economy using the production structure of the year 2010
7It might seem not too realistic to use the same weight to value ratio for every industry –indifferent of its
26 measured against the final demand level of 2000. In order to see the effect of only a change in the production structure, holding demand levels constant.
FD0; showing the total amount and structure of final goods demanded in the economy in 2000.
FD1; showing the total amount and structure of final goods demanded in the economy in 2010.
For both the Z and FD matrices, there is both accounted for a change in the amount of products demanded as for changes in their structures. This means that there is also accounted for a change in emissions as a result of a shift in sourcing locations.
In the calculation of the matrices presented above, both European and non-European trade was included, as non-non-European demand or intermediate production might also generate trade within Europe, when the supply chains of countries are linked. This means that for the calculation of the Z matrices the rows and columns presenting non-European countries were maintained when using the A and M matrices. Also for final demand, the elements presenting non-European countries were maintained.
However, as for the calculation of road transport emissions resulting from trade within Europe, trade and thus transportation to non-European countries should not be included and therefore modifications have to be performed on the matrices presented above: I named these the trade matrices which only present trade within Europe.
Therefore, for the Z matrices, only the 14x252 rows and columns presenting trade for European regions need to be included. Adjusting the matrices in order to only present rows and columns presenting trade between European countries leaves a new matrix of nrEU x nrEU or 3528x3528 cells. These present the trade matrices used for further
calculations: TA0M0f0, TA1M1f1, TA0M0f1 and TA1M1f0, presenting the total production within
Europe necessary in order to fulfil total final demand.
For the FD matrices, this means that its dimensions should be cropped from nr x r to nrEU x rEU (3528x252), again by cutting out final demand levels of non-European
countries in order to avoid including them in calculating the transport emissions from intra-European trade. This results in trade matrices TFD0 and TFD1, presenting only trade
coming from demand from European countries in final goods produced by European countries.
28
4. Data
As a significant amount of external data sources are consulted for this thesis, this section will deliberately describe their structure and adaptations that were made in order to use them for this research.
4.1 WIOD 2013 extension
The main data source for this thesis is the World input-output database. For this thesis an extension of the 2013 database was used, produced by Thissen, Lankhuizen and Los (in preparation). This extension contains trade data on 41 countries including a ‘Rest of the World’, of which 27 European countries. In this dataset, the EU25 countries are further divided into regions based on the NUTS2 classification8. However, a few differences exist between the NUTS2 classification by the European Union and the regions included in the input-output tables. For example no NUTS2 division has been used for Cyprus, Romania and Bulgaria, and for Denmark some regions have been merged. A precise list of regions in the WIOD data can be found in the appendix. Inclusions of country-regions brings novelty and precision to this thesis since it is estimated that domestic freight represents around 10% of total trade related freight globally and 30% of total trade related CO2 emissions (Martinez et al 2014). Including also domestic trade between regions thus provides a more complete picture of total road freight emissions for Europe. Furthermore, the WIOD database splits total trade into fourteen industries9, based on the ISIC Rev 4 classification.
4.2 ArcGIS distance matrix
The distance matrix, which structure has been discussed in the methodology section, is a self-produced dataset, which was build with the help of ArcMap. This is a program of ArcGIS, where GIS stands for Geographical Information System. Within the program a world topographic map was loaded and different layers were added to this, namely: a layer presenting a road network for Europe, a layer dividing Europe into its NUTS2 regions and a layer indicating Europe’s most important cities.
Hereafter, the function of the OD cost matric was consulted, which ‘finds and measures the least-cost paths along the network from multiple origins to multiple destinations’ (ArcGIS, 2017). I selected the layer with European cities to become the Origins and Destinations for the calculation of the OD Cost matrix. As not for all regions
8
The NUTS2 classification divides Europe into 252 regions. More info on NUTS2 regions can be found in Eurostat (2011).
9 These industries include: Agriculture; Mining, quarrying and energy supply; Food, beverages and
29 in the WIOD dataset a city was included or for some there were multiple big cities within the region, I modified the origins and destinations until for every European region in the dataset (which are 252), its capital city was indicated as both an origin and a destination for the OD cost matrix. The capitals used for every region can be found in table 1 in appendix A: ‘Data for distance matrix calculation’. As for some regions it was not clear which was the capital city, the cities were selected that were positioned at a central location and were among the biggest cities within that region.
Solving the OD cost matrix presents a table with the shortest paths for all combinations of origins and destinations in the dataset. Unfortunately, ArcMap only calculates these in travel times, and not in distances. After exporting the OD Cost matrix from ArcMap, this thus has to be divided by the average speed driven on the European roads in order to arrive at the distance in kilometres between two locations. Assumed here is that road transport is executed by trucks, which often have a maximum speed of 90 km/h, and since trucks often perform transportation until point of delivery it is assumed they are also driving in built-up areas part of their driving time. The average speed that was chosen was 60 km/h for a complete transport route. Since this is equal to driving 60 kilometres in 60 minutes, it means the trucks drive a kilometre per minute and thus the matrix with travel times perfectly represents the distance in kilometres and thus no modifications have to be made.
When importing the distance table into Matlab, it was transformed into a matrix whereafter all rows and columns which represent distances to overseas regions were set to zero. These regions can be found again in the appendix where they are indicated with an asterix.
Hereafter, intra-regional distances were inserted in the matrix. For the calculation of intra-regional distances, Head and Mayer’s (2000) formula was used. This formula uses a region’s total surface in square kilometres as to calculate the average distance travelled in order to reach a location within this region’s area. This average distance is based upon the average distances between buyers and sellers. As in this thesis trade data is used, distance between buyers and sellers are a perfect measurement for the analysis. Furthermore, the unit of measurement is the size of an area in km2. The formula reads: 𝑑!"# = 0.67 area/π.
