Deindustrialization in the European Union:
An Analysis of Causes of Decreasing Shares in Value-added by the
Manufacturing Sector at the NUTS 2 Regional Level
University of Groningen
Faculty of Economics and Business
Master’s Thesis International Economics and Business
June 20, 2017
Student: Carmen Zürcher
Student ID number: s3183807
Student email: c.f.zurcher@student.rug.nl
Supervisor: Prof. Dr. Bart Los
Abstract
This study disentangles the causes behind deindustrialization in the form of falling shares of value-added
by the manufacturing sector in current prices in EU NUTS 2 regions between 2000 and 2007. For this
purpose, a Structural Decomposition Analysis is conducted based on the 2017 WIOD release of World
Input-Output Tables with regional breakdowns for the EU25 into 250 NUTS 2 regions. Findings indicate
that deindustrialization is mainly caused by a natural progression towards becoming a service economy
across almost all EU25 regions, but effects of increased international trade and specialization are
heterogeneous with some regional patterns emerging.
Contents
Introduction ... 2
Deindustrialization and its causes ... 3
Definition and terminology ... 3
Deindustrialization and structural transformation ... 3
Causes of Deindustrialization ... 8
Deindustrialization in Europe – Empirics ... 11
Methodology and Data ... 13
Identifying deindustrialization ... 13
Input-Output Analysis ... 14
World Input-Output Tables (WIOT) ... 15
Structural Decomposition Analysis... 16
Data ... 24
Findings and Discussion ... 26
Conclusions ... 31
Page | 2
Introduction
Deindustrialization – the decline of the manufacturing sector in an economy – has long since been a topic of interest in Europe and 'bringing back manufacturing jobs' has been at the heart of many a political debate. The sector is often seen as of great importance for economic growth and development, due to ascribed characteristics such as numerous backward and forward linkages, technological spillovers, easily tradable goods, creating generally well-paid jobs (Rowthorn & Coutts, 2004; Tregenna, 2009; 2011) and even for furthering democratization of a country (Rodrik, 2016). Consequently, deindustrialization – and reindustrialization – are on many government agendas (Bouhol & Fontagne, 2006; Tregenna, 2011). Numerous studies have found that many European Union member states started deindustrializing over past decades (Kollmeyer, 2009). This raises the question for policy-makers whether and how to respond to these changes. The situation has become more challenging over the past two decades as globalization and technological progress has led to an increasing international fragmentation of production, greatly changing way the manufacturing sector operates (Baldwin & Evenett, 2015). Moreover, regions within the European Union differ greatly in terms of their economic structure (Aumayr, 2007), which also raises the potential that they are affected differently by deindustrialization processes.
Identifying deindustrialization and its causes is no easy task. On the one hand, there is a conceptual debate as to what should actually constitute deindustrialization (see Tregenna, 2009;2011), on the other, it is difficult to disentangle the different economic mechanisms behind deindustrialization empirically (Kollmeyer, 2009). Researchers started using input-output analysis and spatial econometrics, among other techniques; each approach with its own set of challenges and issues. Nonetheless, identifying deindustrialization and its causes is an important task to enable successful industrial policy-making.
Based on the new World Input-Output Database (WIOD) 2017 release with breakdowns for the EU25 into 250 NUTS 2 regions (Thissen, Lankuizen & Los, 2017), this paper uses a Structural Decomposition Analysis to answer the following research question: What are the causes of deindustrialization in the form of declining shares of manufacturing in value added in current prices at the NUTS2 level between the years 2000 and 2007 for the European Union?
Page | 3
Deindustrialization and its causes
Definition and terminology
Deindustrialization relates to the „systematic disinvestment in a nation's core manufacturing industries“ (Bluestone & Harrison, 1982:6), such was it defined by Bluestone and Harrison in 1982 who were the first to use the term (Brady & Denniston, 2006). The 'disinvestment' can manifest in the sector's declining employment or output – in absolute terms or as a share in the overall economy (Tregenna, 2011; Rowthorn & Wells, 1987; Rodrik, 2016). In this paper, the term deindustrialization is used to refer to a decline in value added share of manufacturing in current prices (i.e. how much of a region’s total value added or GDP in current prices is contributed by its manufacturing sector) unless otherwise specified. The choice of share in value added over share in gross output as a measure of deindustrialization will be discussed in Section 2. Conversely, the term industrialization refers to an increase in the value added share. While the term re-industrialization is used in the context policy discussion, no distinction can be made in the analysis between industrialization and re-industrialization based on the dataset used. Any increases in the value added share of manufacturing found in the analysis are therefore referred to as industrialization. Having sorted out some of the necessary technicalities, what is deindustrialization conceptually?
Deindustrialization and structural transformation
Deindustrialization is part of what is known as structural change in an economy, in other words “the reallocation of economic activity across three broad sectors (agriculture, manufacturing, and services) that accompanies the process of modern economic growth” (Herrendorf, Rogerson & Valentinyi, 2013:4). As an economy develops and the nation’s income rises, the share of value added by agriculture declines in favor of manufacturing. With further development, the share of manufacturing then begins to drop again while services rise and agriculture declines further. This leads to an inverted U-shape for the value added share of manufacturing in an economy over time (Herrendorf, Rogerson & Valentinyi, 2013; Rodrik, 2016). Examining a historical time series of ten developed countries from 1800 to 2000 (Figure 1, left column), this process of structural transformation is evident. As the log of GDP per capita in 1990 international dollars (horizontal axis) increases, value added share in current prices (vertical axis) for agriculture falls, while the share of services increases and manufacturing exhibits an inverted U-shape. Deindustrialization refers to this fall in the value added share of manufacturing.
Page | 4
heterogeneity in the fall of the manufacturing value added shares in current prices within the European Union which is the focus of this paper.
Figure 1: Shares in total value added in current prices by sectors at different levels of Log of GDP per capita (1990 international $) for ten developed countries 1800-2000 (right column) and EU15 countries 1970-2007 (left column) (Source: Herrendorf, Rogerson & Valentinyi, 2013:8/12 and sources therein)
Page | 5
manufacturing in GDP in EU25 member countries (rather than just EU15 countries) for the year 2000 ranged from 9.9% in Cyprus to 32.7% in Ireland (Figure 2, Panel A). Apart from Ireland (32.7%), the largest shares are in the Czech Republic (26.8%), Finland (26.5%), Slovenia (25.8%), Slovakia (24.7%), Hungary (22.9%), and Germany (22.9%), – apart from Ireland, Germany and Finland, all at the time not yet EU members. Lower shares (apart from Cyprus) are found in Denmark (16.2%), France (16.0%), the Netherlands (15.6%), Latvia (13.7%), Luxembourg (11.3%), and Greece (11.1%). The change in manufacturing value added share in current prices between 2000 and 2007 for the EU25 (Figure 2, Panel B) ranges from -10.9% in Ireland to +0.8% in Germany (in percentage points, i.e. a -10.9% change for Ireland reduces its manufacturing value share from 32.7% to 21.8%), with 21 out of 25 EU member states experiencing deindustrialization. However, the aggregation at the country level conceals the extent of the heterogeneity, as the differences are far starker at the regional level (Figure 3).
Page | 6 Figure 2: Manufacturing value added share in GDP in current prices (% of GDP) in the year 2000 (Panel A) and absolute change in manufacturing value added share in GDP in current prices (percentage points) from 2000 to 2007 (Panel B) at the country level for the EU25 (author’s own calculations based on WIOD 2017 release data. For details on calculations, visualization and data, see Section 2.).
Panel A
Page | 7 Figure 3: Manufacturing value added share in GDP in current prices (% of GDP) in the year 2000 (Panel A) and absolute change in manufacturing value added share in GDP in current prices (percentage points) from 2000 to 2007 (Panel B) at the NUTS2 level for the EU25 (author’s own calculations based on WIOD 2017 release data. For details on calculations, visualization and data, see Section 2.) (Larger versions of the maps are available in Appendix B).
Panel A
Panel B
Page | 8
These differences between EU regions then raise the question, what led some regions to deindustrialize while others industrialized between 2000 and 2007? It is not clear yet what the exact mechanisms are behind these different deindustrialization experiences across Europe (Nickell, Redding & Swaffield, 2008) and it is the endeavor of this paper to shed more light on the different economic forces behind deindustrialization and their interplay which have led to this heterogeneity. To this end, it is important to analyze not only why some regions experienced a fall in manufacturing value added shares, but also why others experienced an increase or only a moderate decline in the shares.
The remainder of this section discusses the mechanism behind causes of deindustrialization identified from the literature, followed by a review of existing empirical work on the causes of deindustrialization at the regional level in Europe. Section 2 then discusses methodology and data, while Section 3 presents and discusses findings of the study. Section 5 concludes with limitations, policy implications and suggestions for future research.
Causes of Deindustrialization
Over time, deindustrialization has attracted the attention of numerous disciplines, from economists, to sociologists to geographers, and there has been much debate surrounding the causes of deindustrialization and especially their respective weight (Alderson, 1999; Rowthorn & Coutts, 2004; Brady & Denniston, 2006). Reviewing the literature indiscriminately of which field it originated from, three broad main causes of deindustrialization – in terms of employment or output – seem to have been established: 1. higher productivity in the manufacturing sector, 2. increased consumer spending on services, and 3. trade liberalization and globalization (Kollmeyer, 2009). Rowthorn and Coutts (2004) also put forth another cause which they termed ‘specialization’ and which may potentially have gained traction over recent years in light of globalization. The mechanisms behind structural change or transformation – and consequently deindustrialization – are often explored using multi-sector models (see Herrendorf, Rogerson & Valentinyi (2013) for an overview). While discussing such models in detail would go beyond the scope of this paper, they are drawn upon to illuminate the economic mechanisms behind the identified causes of deindustrialization, and later also for insights in the interpretation of findings. Since the EU regions analyzed in this paper are open to trade within and outside of the EU, the focus is placed on multi-sector models of open economies, such as Rodrik (2016) and Matsuyama (2009).
Page | 9
service sector (Rowthorn & Coutts, 2004; Bouhol & Fontagne, 2006; Rodrik, 2016). For the mathematical explanation of why higher productivity growth in the manufacturing sector leads to employment, but not necessarily output deindustrialization, see the two-sector model of Rodrik (2016). This does not mean, however, that productivity does not play a role in output deindustrialization. While it is not included as a standalone cause, it is discussed in the context of the service economy argument in the next paragraph, international trade, as well as other causes.
A natural progression towards a service economy
de-Page | 10
pendent on “demand, supply and relative price effects”. This progression towards a service economy is often named as the main cause of deindustrialization (Bouhol & Fontagne, 2006; Rowthorn & Coutts, 2004).
International trade
Globalization has revolutionized the manufacturing sector. With what are known as globalization’s first and second unbundling (transportation and ICT revolutions respectively), trade costs were reduced dramatically. Not only did it become possible to, for example, produce manufactures in China and profitably sell them in Germany, but ICT has enabled firms to outsource individual production stages and processes (Baldwin & Evenett, 2015). This has increased competition for both intermediate inputs, as well as final products and led to a fragmentation of production (Palma, 2005; Timmer, Los, Stehrer & de Vries, 2013). In the case of the EU, the European Single Market also needs to be considered. Herrendorf, Rogerson and Valentinyi (2013) attribute the heterogeneity in manufacturing value added shares in GDP (current prices) across the European Union at least partially to the existence of the European Single Market. However, whether a country or region will deindustrialize as a result of increased international trade depends. If productivity is high enough in a region and it becomes an exporter of manufactures, the value added share of manufacturing increases. It may also lead regions to specialize in higher value added activities in manufacturing, while lower value added activities are replaced by imports. If a region however, cannot increase its productivity quickly enough to outpace the fall of relative prices (which fall due to increased competition), the country/region deindustrializes (Rodrik, 2016; Rowthorn & Coutts, 2004; see also Iacovone, Rauch & Winters, 2013). The weight of international trade as a source of deindustrialization is disputed, but has gained more attention since the 1980s (Kollmeyer, 2009).
Changing production structures or specialization
Page | 11 Other causes
Other causes which bear mentioning, are policy decisions which may lead to deindustrialization and the Dutch disease effect. In the case of the Dutch disease effect, if a country or region develops a strong trade surplus in primary resources (e.g. discovering new natural resources) or services (e.g. tourism), the demand for its currency rises and the currency appreciates. Its manufacturing exports become more expensive and imports cheaper, leading to deindustrialization in the country or region if productivity is not sufficiently improved to counter the effect (Bogliaccini, 2013; Palma, 2005; Rodrik, 2016; Stijns, 2003).
Overall, deindustrialization is not necessarily the result of a single cause, but rather depends on how different causes interact with each other (Palma, 2005). In a large economy which imports many of the manufacturing products it consumes, an increase in productivity of its firms would not decrease relative prices. Increased demand for services in an economy might not lead to deindustrialization if firms are able to sell their manufactures on the world markets instead (Rodrik, 2016). It is this interplay of the different causes which makes it difficult to identify the drivers behind deindustrialization (Kollmeyer, 2009). What has been done to empirically study causes of deindustrialization at the regional level before?
Deindustrialization in Europe – Empirics
While causes of deindustrialization – mainly in terms of employment – have been studied relatively widely at the country level in Europe (e.g. Bouhol & Fontagne, 2006; Kollmeyer, 2009; Tregenna, 2009; 2011), only few researchers have analyzed the phenomenon at the regional level in Europe (Stojcic & Aralica, 2015). Though there are numerous papers which have analyzed structural change as a whole for regions within single EU countries (e.g. Krieger-Boden, Morgenroth & Petrakos, 2008). There are also a few studies which have taken yet a different approach, opting instead to analyze at a micro-level (e.g. Bathelt & Kappes, 2008 who study a merger between two chemical companies as a cause of regional deindustrialization). The only comparable study to this paper at the regional level across parts of the EU appears to be a 2015 working paper by Stojcic and Aralica, which is discussed in more detail in the following.
Page | 12
and an overall labor productivity effect at the country and regional level. Relating these effects back to the causes discussed in the previous section, the labor intensity effect refers to the productivity differential between manufacturing and service sectors, the value added share effect covers output deindustrialization relevant causes, and the overall labor productivity effect is a residual. Results are presented as national averages and show that the employment deindustrialization is predominantly rooted in increased labor productivity, while sector share effects had positive effects in almost all countries. To further investigate the causes behind the employment deindustrialization in the different regions, a spatial Durbin panel model is then used to examine the influence of foreign direct investment, the share of high tech manufacturing, urbanization and localization economies, corruption, as well as a range of other variables. By investigating these variables also as spatial lag regressors, Stojcic and Aralica (2015) were able to determine the influence of inter-regional effects based on a region’s location. What they observe is that while experiences were often heterogeneous across the regions of individual countries, some regional patterns still emerged. Regions bordering Western Europe tended to industrialize rather than deindustrialize, while regions to the East mostly deindustrialized in terms of employment. Regions moving towards a focus on high tech manufacturing also experienced industrialization. Whether a region de- or industrialized in terms of output was largely dependent on the competitiveness of its manufacturing sector according to Stojcic and Aralica (2015). They also found a number of inter-regional effects, noting for example that a greater focus on services in one region leads to an increase of manufacturing in others and an increasing manufacturing employment share also increases shares in other regions. Their findings lead them to conclude that because of the regional heterogeneity, EU industrial policy-making should be based on the regional level, rather than national or supranational levels.
Page | 13
Decomposition Analysis. This type of analysis not only enables the identification of different causes, but to some extent also disentangles the channels through which they operate. This study therefore contributes to the existing body of research not only through the analysis of deindustrialization at the regional level, but also by being one of the few studies which examines multiple possible causes at the same time (Kollmeyer, 2009).
Methodology and Data
The section introduces the methods and data used in this paper in more detail, starting with the calculation of the key variable of interest and the motivation behind its choice.
Identifying deindustrialization
As suggested by Tregenna (2011) and Rodrik (2016), deindustrialization – or industrialization – is measured in terms of the change in the share of value added by manufacturing in the total value added or
gross domestic product (GDP) of a region σr which is calculated as
, where tm = the total manufacturing value added in million € (calculated as the sum of p, the value added by industries classified as manufacturing M in region r at time t) and tv = GDP of region r at time t (calculated as the sum of value added q for all industry S in an economy). In Section 1, the change in a region’s value-added share of manufacturing was then computed as both the absolute change between two periods, i.e.
- , and as a ratio, i.e. / . However, for purposes of the analysis which follows, change in the
manufacturing value added shares is used as a ratio. Both methods of calculation are included in the overall results table in Appendix A.
Page | 14
the manufacturing sector, value added from contributions such as cotton from the agricultural sector and retail services are attributed to the other two sectors (Herrendorf, Rogerson & Valentinyi, 2013). It is also a more suitable measure because of the previously discussed fragmentation of production, limiting the value added contributions to what was generated by the manufacturing sector in one country or region (for example, value added in the partial assembly of a manufacture in China is not counted towards the manufacturing value added in Germany where final assembly takes place). For a detailed discussion of this topic, see Baldwin and Evenett (2015) and Timmer, Los, Stehrer and de Vries (2013).
Another point to consider is the question of constant versus current prices. As the dataset used in this analysis is not available in constant prices, manufacturing value added shares are analyzed in current prices. While Bouhol and Fontagne (2006) voice reservations regarding the use of value added shares in current prices rather than constant prices, the two measures have been found to exhibit similar patterns of development over time (Kuznets, 1966; Herrendorf, Rogerson & Valentinyi, 2013). Rodrik (2016) did find that shares in current prices peaked at lower levels of development than shares in constant prices and Bouhol and Fontagne (2006) further noted that shares in current prices also closely follow manufacturing shares in employment. This issue is raised again in Section 4 on limitations.
With the change in manufacturing value added share in GDP in current prices defined as the key variable of interest, input-output analysis can be used to disentangle the different causes behind the change.
Input-Output Analysis
Input-output (I/O) analysis refers to an “analytical framework developed by Professor Wassily Leontief in the late 1930s” (Miller & Blair, 2009:1) and is conducted on the basis of Input-Output Tables which track linkages between industries by giving information on “the amount and type of intermediate inputs needed in the production of one unit of output, [and] […] the gross output in all stages of production that is needed to produce one unit of final demand” (Timmer, Los, Stehrer & de Vries, 2013:620). This study uses an interregional extension of such tables (called World Input-Output Tables), which track not only linkages across industries of a single economy, but also across countries and regions. For a better understanding of the methodology used in this paper, a brief introduction to the set-up of the regional World Input-Output Tables as well as key I/O identities is included, starting with information on the notations used.
Notation
Page | 15
World Input-Output Tables (WIOT)
WIOTs are constructed based on National Accounts, Supply and Use Tables and international trade data (Dietzenbacher, Los, Stehrer, Timmer & de Vries, 2013). Figure 4 provides a stylized version of a regional WIOT with two countries and the rest of the world (ROW). Both countries have two regions with two industries each and two industries for the rest of the world. Below a very brief introduction is added with descriptions of the table and explanations of key I/O identities based on Miller and Blair (2009) and Los, McCann, Springford and Thissen (2017). Details on the construction of WIOTs can be found in Dietzenbacher, Los, Stehrer, Timmer and de Vries (2013), with additional details for the regional WIOTs in the forthcoming Thissen, Lankhuizen and Los (2017).
Figure 4: Stylized World Input-Output Table (Source: Los, McCann, Springford & Thissen, 2017).
An I/O table traces transactions across sectors and the final consumption of goods for a set period of time (a year in the case of the WIOD dataset). These transactions are in monetary terms rather than physical quantities (i.e. €5 million of steel, rather than 5 tons of steel) (Miller & Blair, 2009). What are the different components of a WIOT as stylized in Figure 4?
N = number of industries in a region or country R = number of regions / countries
Z = a NR x NR matrix on the value of intermediate input sales and use, zij gives the value of industry i sales
to industry j, which industry j uses in the production of its gross output xj;
F = a NR x R matrix which tracks the value of products delivered to final users, i.e. which are not used as
Page | 16
divided into the four categories final consumption by households and non-profit organizations, final consumption by government, net capital formation, and inventory adjustment (Los, McCann, Springford & Thissen, 2017). However, since the individual demand categories are not relevant to the research of this paper, demand categories and demands by different countries are summed into total final demand of industries f;
w’ = value added or the value of primary inputs (e.g. labor and capital) which an industry uses in producing
its gross output;
x = a NR x 1 vector of the values of gross output produced by the different sectors. Due to the double-entry
bookkeeping property, gross output can be calculated as the value intermediate inputs plus value added, as well as value of intermediate inputs plus the value of final demand.
Based on these elements, three key identities can be calculated which are needed for the analysis in this paper:
A = a NR x NR matrix with elements calculated as zij/xj = aij, aij gives the value of intermediate input from
industry i which industry j uses per one euro of gross output, also called a direct input coefficient;
L = the Leontief inverse – named after Economist Wassily Leontief – is a NR x NR matrix constructed as L
= (I – A)-1, where I is an NR x NR identity matrix with ones on the diagonal and zeros elsewhere (see Miller
and Blair (2009) for the derivation), element lij gives the value of intermediate inputs from industry i which
industry j uses in the production of one euro of final demand;
v = a NR x 1 vector of value added coefficients with elements calculated as wj/xj = vj, element vj gives the
value added by industry j per euro of gross output xj.
Based on these variables, gross output can then be calculated as x = Lf and the value added, which is of interest to this paper, is calculated through the value added coefficients: v’x = v’Lf. Overall, these variables deliver the inputs for many different I/O analyses (for a detailed introduction, see Miller and Blair (2009)).
Structural Decomposition Analysis
Page | 17
shown here to explain the basic properties of SDA. As an example, the previously introduced equation x =
Lf can be used, where changes in gross output x are a result of changes in the two determinants L and f (for
a detailed, step-by-step introduction to SDA, see Miller & Blair, 2009). The basic decomposition then takes the following form, with indices indicating different time periods:
∆ ∆ ∆ 1
The first term in this equation results in the change in x which would have occurred if only the Leontief inverse had changed and final demand had remained constant, the second term gives the result of the opposite case, i.e. if only final demand had changed (Dietzenbacher & Los, 2000). The same applies to the multiplicative approach, though results are relative changes rather than absolute. However, Equation 1 is not unique, meaning the decomposition equation could also take the form of Equation 2 in the case of x =
Lf.
∆ ∆ ∆ 2
For n determinants, n! equivalent decomposition equations exist which result in a range of different outcomes for the determinants and the different results would be averaged (Dietzenbacher & Los, 1998) (or the geometric mean taken for multiplicative). While this may be feasible for small numbers of determinants, calculations quickly balloon and six determinants would already have 720 equivalent decomposition equations. However, using the average of the two polar decomposition equations gives very similar results to the average across all decomposition forms (for additive SDA) (Dietzenbacher & Los, 1998). This polar decomposition approach is also used in this study. However, for expositional simplicity, only one polar equation will be shown for partial steps, with both polar equations being shown for the final SDA form.
Applying what was discussed in this short introduction to I/O analysis and SDA to the key variable of interest in this study, the change in manufacturing value added share in GDP at current prices, the variable can be decomposed into changes in value added coefficients, the Leontief inverse and final demand (Equation 3). Based on these determinants and basic arithmetic properties of fractions, the following basic first polar decomposition can be set up:
er = a NR x 1 summation vector with 1 for industries of EU25 region r and zeros elsewhere;
m = manufacturing value-added coefficients of all EU25 regions and non-EU countries, a NR x 1 vector
Page | 18 3 4 ′ ′ ′ ′ ′ ′ ′ ′ ′ ′ ′ ′
In this basic set-up of the decomposition (Equation 4) where indices indicate the period used, the first factor explains how much of the change in the manufacturing value added share is attributed to the effect of changes in the value added coefficients of a region, the second covers effects of changes in the Leontief inverse, and the final factor reflects the effect of changes in final demand. By adding a regional summation vector er, changes are decomposed for each region individually. It is important here to note that while
changes in the value added coefficients and the Leontief inverse are not entirely independent from each other (Dietzenbacher & Los, 2000), though this has no bearing on the interpretation of the factors as will be discussed at a later point. However, this level of decomposition does not yet disentangle causes of deindustrialization (or industrialization) sufficiently. While the first factor of changes in value added coefficients captures deindustrialization due to increased specialization, it is not yet possible to distinguish international trade from service economy effects. A further step of decomposition of the Leontief inverse and final demand factors is therefore necessary to resolve these issues. To this end, a distinction is made between effects of changes in the aggregate production structure or final demand levels and effects of changes in where intermediate inputs and final products are bought from, i.e. international trade. For this purpose, the trade share matrix approach of Dietzenbacher, Hoen and Los (2000) is adopted. To illustrate the approach, a simple 4 x 4 A matrix of two regions with two industries each can be used:
in million €, current prices Region A Region B
Manufacturing Non-Manuf Manufacturing Non-Manuf
Region A Manufacturing 0.5 0.3 0.4 0.2
Non-Manuf 0.2 0.6 0.1 0.1
Region B Manufacturing 0.3 0.1 0.4 0.3
Non-Manuf 0 0 0.3 0.4
Table 1: Example A matrix for two regions with two industries
Page | 19 AAggregate inputs = 0.8 0.4 0.8 0.5 0.2 0.6 0.4 0.5 0.8 0.4 0.8 0.5 0.2 0.6 0.4 0.5 , ATrade shares = 0.625 0.75 0.5 0.4 1.0 1.0 0.25 0.2 0.375 0.25 0.5 0.6 0 0 0.75 0.8
The element AAggregate inputs
11 states that the manufacturing industry of Region A uses €0.8 million of
intermediate inputs from manufacturing industries to produce €1 million of gross output (€0.5 million + €0.3 million). Of those €0.8 million in manufacturing inputs, 37.5% are delivered by the manufacturing industry
of Region B (according to the element ATrade shares31 (€0.3 million/€0.8million)). In other words, the A matrix
is equal to the aggregate intermediate input matrix multiplied by the trade share matrix. The same concept can be applied to final demand (a more formal explanation of the matrices and their construction follows). Applying this approach to the decomposition for this paper gives:
A* = a NR x NR matrix which consists of R identical N x NR matrices with “aggregate intermediate inputs
per unit of gross output by industry” by region (Dietzenbacher, Hoen & Los, 2000:428) stacked,
∀ : ∗ ∑ ;
TA = a NR x NR matrix “of intermediate trade coefficients, representing the shares of each [region] in
aggregate inputs, by input by industry by [region]” (Dietzenbacher, Hoen & Los, 2000:428),
/ ∗ , it follows that ∑ 1;
F* = a NR x R matrix which consists of R identical N x R matrices “of final demand for product i by [region
c]” (Dietzenbacher, Hoen & Los, 2000:429) stacked, ∀ : ∗ ∑ ;
TF = a NR x R matrix “of final demand trade coefficients, representing the shares of [region] r in aggregate
final demand for product i in [region c]” (Dietzenbacher, Hoen & Los, 2000:429), / ∗ , it
follows that ∑ 1;
The Leontief inverse and final demand can then be written as ∗∘ and
∗∘ in the decomposition, e denotes a NR x 1 summation vector and ∘ is the Hadamard product
operator for elementwise multiplication of matrix and vector elements. The first polar decomposition equation then reads
Page | 20
Factor (5.2) in Equation 5 now captures changes in the production structure (A*) (i.e. technological change) which is one of the causes of deindustrialization of interest in this paper. However, the separation between international trade and service economy effects is not sufficiently clear yet in the remaining three factors, since, for example, a decrease in a region’s value added share of manufacturing from effects of changes in TA may be the result of that region’s manufacturing sector losing to international competition
(attributed to the international trade cause) or the region’s non-manufacturing sectors gaining (attributed to the service economy cause). Therefore the two effects need to be disentangled, which can be achieved through the additive property of matrices (see Miller & Blair, 2009). The approach can again be illustrated
based on the example in Table 1. Using the previously constructed input trade share matrix ATrade shares from
Table 1, separating manufacturing sectors (G) and non-manufacturing sectors (H) gives the following matrices: = 0.625 0.75 0.5 0.4 0 0 0 0 0.375 0.25 0.5 0.6 0 0 0 0 , = 0 0 0 0 1.0 1.0 0.25 0.2 0 0 0 0 0 0 0.75 0.8
ATrade shares can then be written as ATrade shares = ( + . Separating manufacturing and
non-manufacturing sectors in the variables TA, F* and TF of Equation X with the method above, then results in
TA = ( + , F* = ( ∗+ ∗ , and TF = ( + . Due to the points raised by Herrendorf, Rogerson
and Valentinyi (2013) regarding the European Single Market and the possibility of avoiding deindustrialization by selling manufactures on the world markets (Rodrik, 2016), one last further
decomposition is made for manufacturing sector variables , ∗, and to identify whether effects of
changes are rooted in intra- (E) and extra-EU25 regions/countries (O) based on Miller and Blair (2009).
Using the input trade share matrix of the manufacturing sectors based on Table 1 to again illustrate
this step and assuming that Region A is part of the EU25 while Region B is not, the following two matrices are constructed: = 0.625 0.75 0 0 0 0 0 0 0.375 0.25 0 0 0 0 0 0 , , = 0 0 0.5 0.4 0 0 0 0 0 0 0.5 0.6 0 0 0 0 ,
In a decomposition, now provides information on what change in a variable can be
attributed to effects of changes in the trade shares of manufacturing sectors in industries within the EU25
and on what change can be attributed to changes in industries outside of the EU25, with =
Page | 21
yields = ( + ), ∗ = ( ∗ + ∗), and = ( + ). Implementing the separation of
manufacturing and non-manufacturing sectors and EU25 versus non-EU25 effects for manufacturing in the decomposition of the change in manufacturing value added shares, results in the following breakdowns of the variables TA, F* and TF to be added to the decomposition (a summary and interpretations for all
decomposition factors are included at the end):
= a NR x NR matrix of trade shares of manufacturing sectors in intermediate inputs used by industries in EU25 regions (non-EU25 country industry columns and non-manufacturing industry rows set to zero);
= a NR x NR matrix of trade shares of manufacturing sectors in intermediate inputs used by industries in non-EU25 countries (EU25 region industry columns and non-manufacturing industry rows set to zero);
= a NR x NR matrix of trade shares of non-manufacturing sectors in intermediate inputs used by
industries of all countries and regions (manufacturing industry rows set to zero);
∗
= a NR x R matrix of final demand levels for products produced by manufacturing sectors by EU25 regions (non-EU25 country columns and non-manufacturing industry rows set to zero);
∗
= a NR x R matrix of final demand levels for products produced by manufacturing sectors by non-EU25 countries (non-EU25 region columns and non-manufacturing industry rows set to zero);
∗
= a NR x R matrix of final demand levels for products produced by non-manufacturing sectors by all
countries and regions (manufacturing industry rows set to zero).
= a NR x R matrix of trade shares of manufacturing sectors in final demand of manufacturing products by EU25 regions (non-EU25 country columns and non-manufacturing industry rows set to zero);
= a NR x R matrix of trade shares of manufacturing sectors in final demand of manufacturing products
by non-EU25 countries (EU25 region columns and non-manufacturing industry rows set to zero);
= a NR x R matrix of trade shares of non-manufacturing sectors in final demand of non-manufacturing
products by all countries and regions (manufacturing industry rows set to zero)
This yields TA = (( + )+ ), F* = (( ∗+ ∗)+ ∗) and TF = (( + )+ ).
Extending the SDA equation of Dietzenbacher, Hoen and Los (2000) (Equation 5), the first polar equation of the final decomposition form is then written as
Page | 22 (6.4) = ∗∗∘ ∘ ∗∘ ∗∘ (6.5) = ∗∗∘ ∘ ∗∘ ∗∘ (6.6) = ∗ ∗ ∗ ∘ ∗ ∗ ∗ ∘ ∗ ∗ ∗ ∘ ∗ ∗ ∗ ∘ (6.7) = ∗ ∗ ∗ ∘ ∗ ∗ ∗ ∘ ∗ ∗ ∗ ∘ ∗ ∗ ∗ ∘ (6.8) = ∗ ∗ ∘ ∗ ∗∘ ∗ ∗ ∘ ∗ ∗ ∘ (6.9) = ∗∗∘ ∘ ∗∘ ∗∘ (6.10) = ∗∗∘ ∘ ∗∘ ∗∘ (6.11) = ∗∗∘ ∘ ∗∘ ∗∘
And the second polar equation:
Page | 23 (7.10) = ∗∗∘ ∘ ∗∘ ∗∘ (7.11) = ∗∗∘∘ ∗∗∘∘
Page | 24
case for factors 5, 6, 7, and 8). In summary:
Cause Attributed factors
Service economy 5, 6, 7, 8, 11
International trade 3, 4, 9, 10
Changing production structures 1, 2
Other 5, 6, 7, 8, 11
Calculations are performed in MATLAB 2016b and MS Office Excel 2016, the MATLAB code is included in Appendix C.
Data
This study uses the ‘beta-version’ (Los, personal communication, April 26, 2017) of the 2017 World Input-Output Database (WIOD) release by the Groningen Growth and Development Center (GGDC) (Thissen, Lankhuizen & Los, 2017). The release covers 250 NUTS 2 regions of the EU25 (Figure 5) and 15 countries, plus the rest of the world. Today’s EU members Bulgaria and Romania, are also included in the dataset, but do not include the disaggregation to the regional level. Croatia is not included separately in the dataset, but is instead part of the rest of the world estimates. Other countries included are Australia, Brazil, Canada, China, India, Indonesia, Japan, Mexico, Russia, South Korea, Taiwan, Turkey, and the United States.
Page | 25
The European Union’s Nomenclature of Territorial Units for Statistics 2 or NUTS 2 regions represent the second of three levels of regional disaggregation in the EU and were drawn to be comparable in terms of population size (between 800’000 to 3 million people), though they may differ widely with regards to other characteristics such as economic size (Eurostat, 2011; EC No 1059/2003). The WIOD 2017 release follows the 2003 NUTS 2 classifications of the European Union and the OECD (compare Regulation on the Establishment of the NUTS EC No 1059/2003 and OECD (2003)). Visualizations of data and results in this
study are created in ArcGIS using shapefiles released by the European Commission (2008).1
The dataset itself is available as a time series for the years 2000 to 2010 and consists of WIOTs in million euros in current prices which cover 14 industries for each of the 266 countries and regions listed above. The industries included are Agriculture; Mining, Quarrying and Energy Supply; Food, Beverages and Tobacco; Textiles and Leather etc.; Coke, Refined Petroleum, Nuclear Fuel, and Chemicals, etc.; Electrical and Optical Equipment and Transport Equipment; Other manufacturing; Construction; Distribution; Hotels and Restaurant; Transport, Storage and Communication; Financial Intermediation; Real Estate Renting and Business Activities; Non-Market Service. Two key limitations of the dataset also need to be mentioned. As a result of limited data availability, lower quality data and a lack of harmonization among countries, technologies and import use are assumed to be homogenous across industries in a country or region. Data on trade in services is less accurate than that of other sectors and is also influenced by business practices such as transfer pricing (Timmer, Dietzenbacher, Los, Stehrer & de Vries, 2015). Both of these issues will affect the outcomes of this study by understating the heterogeneity between industries and potentially overstating the impact of trade in services. Additionally, the conversion into euros also implies that intermediate input and final demand level changes may be affected by exchange rate fluctuations (Los, McCann, Springford & Thissen, 2017). The time period studied in this paper is limited to the change between 2000 and 2007 to avoid effects of the financial crisis of 2008-2009. Further details on the construction of the dataset are available in the forthcoming Thissen, Lankhuizen and Los (2017). Additional details on limitations of the dataset can be found in Timmer, Dietzenbacher, Los, Stehrer and de Vries (2015).
Sectors are classified as manufacturing based on the overlaps between the WIOD 2017 release sector labels and the EU’s Statistical Classification of economic Activities in the European Community (NACE) (Eurostat, 2002 Revision 1.1) manufacturing sector classifications, which also served as the base in the construction of the GGDC’s WIOTs (Dietzenbacher, Los, Stehrer, Timmer & De Vries, 2013). Consequently, manufacturing is comprised of a bundle of 5 of the 14 sectors included in the regional WIOT for 2000-2007: Food, Beverages and Tobacco; Textiles and Leather etc.; Coke, Refined Petroleum, Nuclear Fuel, and
Page | 26
Chemicals, etc.; Electrical and Optical Equipment and Transport Equipment; Other manufacturing.
As in Dietzenbacher, Hoen and Los (2000), the construction of trade share matrices in aggregate intermediate inputs TA required changes, since not every industry uses inputs from all industries in its
production which results in a division by zero in the calculation of trade shares ( / ∗ ) for
the industries. If an aggregate intermediate input element for an industry is zero in 2000 and 2007, i.e. if an industry did not use a type of intermediate input in 2000 and does still not use it in 2007, the undefined trade share results can be assigned the value zero in both years to enable further computations without affecting results. However, if the aggregate intermediate input element is zero for only one of the years, i.e. an industry started or stopped using a type of intermediate input between 2000 and 2007, the undefined trade share has to be assigned a nonnegative value for the decomposition. Following Dietzenbacher, Hoen and Los (2000), the corresponding trade share from the other year is chosen as the replacement value. This choice affects results, since all effects of changes stemming from an industry starting or stopping the use of a type of input are assigned to production structure changes and none to trade share changes.
Findings and Discussion
As predicted by the literature, the single largest contributor to deindustrialization in the EU25 overall are effects from increases in final demand for non-manufactures, with an average decrease of 14.5% in the value added shares of manufacturing. International trade, value added coefficient and technology changes seem to have had more heterogeneous effects, which will be discussed in more detail (Table 2). Figure 6 includes again the map of relative changes in manufacturing value added shares between 2000 and 2007, followed by visualizations of the decomposition results for each factor.
Description Number of regions with relative change < 1.00 out of 250 regions
Geometric mean across all regions
σr1/σr0 Relative Change Manuf Value Added 195 0.891
Factor 1 Value Added Coefficients 193 0.968
Factor 2 Technological Change 194 0.974
Factor 3 Input Trade Shares Manuf EU25 109 1.008
Factor 4 Input Trade Shares Manuf non-EU25 134 0.999
Factor 5 Input Trade Shares Non-manufactures 137 0.993
Factor 6 Final Demand Manufactures EU25 0 1.072
Factor 7 Final Demand Manufactures non-EU25 15 1.034
Factor 8 Final Demand Non-Manufactures 244 0.855
Factor 9 Final Demand Trade Shares Manuf EU25 139 1.001 Factor 10 Final Demand Trade Shares Manuf
non-EU25 147 0.997
Page | 27 Figure 6: Mapped results of the decomposition analysis displayed in order of their respective terms in the decomposition equations 6 and 7. From top left: Relative change in manufacturing value added 2000-2007; factor 1 change in value added coefficients; factor 2 technological change; factor 3 input trade share changes for manufactures within the EU25; factor 4 input trade share changes of manufactures non-EU25; factor 5 changes in input trade shares of non-manufactures; factor 6 changes in final demand levels of manufactures EU25; factor 7 changes in final demand levels of manufactures non-EU25; factor 8 changes in final demand levels of non-manufactures; factor 9 changes in final demand trade shares of manufactures EU25; factor 10 changes in final demand trade shares of manufactures non-EU; factor 11 changes in final demand trade shares of non-manufactures (author’s own calculations based on WIOD 2017 release data. A complete results table, as well as larger versions of the maps are available in Appendices A and B respectively).
1 2
3 4 5
6 7 8
Page | 28
As in the paper of Stojcic and Aralica (2015), comparing effects of the different determinants on manufacturing value added shares across EU25 regions based on the maps in Figure 6 reveals a number of regional patterns which will be the focus of the discussion.
Contrasting experiences of East German and Polish regions
Region Total Fac-tor 1 tor 2 Fac- tor 3 Fac- Fac-tor 4 tor 5 Fac- tor 6 Fac- Fac-tor 7 tor 8 Fac- tor 9 Fac- tor 10 Fac- tor 11
Fac-Poland Lodzkie 0.974 0.974 0.961 0.973 0.996 1.012 1.075 1.073 0.870 0.985 0.961 1.002 Mazowieckie 0.975 0.975 0.987 0.967 1.000 0.986 1.077 1.066 0.870 0.981 0.968 0.992 Malopolskie 0.981 0.981 0.984 0.962 1.000 0.997 1.075 1.057 0.868 0.986 0.970 0.994 Slaskie Lubelskie 0.931 0.986 0.990 0.967 0.965 0.988 1.064 1.121 0.948 0.962 0.955 0.996 Podkarpackie 0.826 0.939 0.990 0.934 0.982 0.985 1.068 1.077 0.940 0.947 0.968 0.991 Swietokrzyskie 0.909 0.941 0.990 0.956 0.979 1.002 1.064 1.099 0.949 0.979 0.963 0.996 Podlaskie 0.844 0.936 0.991 0.917 0.971 1.015 1.096 1.092 0.910 0.955 0.961 1.009 Wielkopolskie 0.928 0.977 1.003 0.948 0.980 0.994 1.140 1.056 0.910 0.924 0.971 1.045 Zachodniopomorskie 0.916 0.938 0.990 0.965 0.977 0.989 1.081 1.107 0.955 0.972 0.964 0.989 Lubuskie 0.930 0.973 0.990 0.943 0.980 1.024 1.071 1.094 0.943 0.961 0.962 0.997 Dolnoslaskie 0.905 0.935 0.974 0.949 0.978 1.021 1.064 1.107 0.948 0.975 0.960 1.002 Opolskie KujawskoPomorskie 1.048 0.969 1.040 0.968 0.974 1.045 1.044 1.094 0.962 0.979 0.969 1.014 WarminskoMazur-skie 0.895 1.002 0.960 0.921 0.993 1.024 1.063 1.105 0.892 0.978 0.969 1.000 Pomorskie 0.911 1.058 0.966 0.920 0.996 0.965 1.046 1.104 0.925 0.976 0.969 0.999 East Germany Brandenburg Nordost 1.157 0.983 1.001 1.069 1.009 0.971 1.007 1.018 1.044 1.041 1.017 0.991 Brandenburg Sud-west 1.118 0.968 1.000 1.061 1.014 0.995 1.040 1.019 0.975 1.032 1.019 0.995 Chemnitz 1.140 0.980 1.006 1.056 1.012 0.981 1.038 1.016 0.993 1.033 1.017 1.001 Dresden 1.162 0.980 1.003 1.063 1.017 0.982 1.051 1.019 0.979 1.048 1.021 0.993 Leipzig 1.224 0.994 1.001 1.081 1.020 0.990 1.048 1.020 0.991 1.054 1.024 0.986 Dessau 1.244 1.014 0.993 1.119 1.020 0.956 1.015 1.015 1.015 1.111 1.027 0.949 Halle 1.092 0.958 0.985 1.019 1.004 1.016 1.020 1.017 1.035 1.011 1.010 1.015 Magdeburg 1.413 1.034 0.993 1.147 1.026 0.970 1.025 1.018 1.018 1.124 1.031 0.979 Table 3: Decomposition results for Polish and East German regions. The column ‘Total’ contains the ratio of manufacturing value added shares in 2007 to 2000.
Page | 29
based on Rodrik (2016). A similar observation is made for the Baltic states (Estonia, Lativa and Lithuania) which would have experienced steep declines in manufacturing value added shares based on increases in final demands of non-manufacturing sectors, but were able to counter it through increased trade shares in intermediate inputs and final demand inside and outside the EU25. However, the changes in value coefficients (factor 1), suggest that Polish regions and Latvia may be specializing in lower value added activities while Estonia and Lithuania are focusing on higher value added activities of manufacturing. East German regions also experienced increases in manufacturing value added shares due to trade share factors, especially the regions Dessau and Magdeburg which gained 11.9% and 14.7% from intra-EU25 intermediate input trade share changes (factor 3) and 11.1% and 12.4% from intra-EU25 final demand trade share changes (factor 9). At the same time, East German regions did not experience the decrease in manufacturing value added share linked to the increased final demand levels for non-manufactures (factor 8) as Polish and Baltic regions. Overall, most Polish, East German and Baltic regions seem to have experienced industrialization or more moderate deindustrialization due to their increased competitiveness in EU25 and world markets for both intermediate inputs and final demand. Nordic regions made the opposite experience.
Nordic regions (Denmark, Sweden and Finland)
Region Total Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Factor 7 Factor 8 Factor 9 Factor 10 Factor 11
Denmark
Hovedstadsreg 0.974 0.974 0.961 0.973 0.996 1.012 1.075 1.073 0.870 0.985 0.961 1.002 Ost for
Sto-rebelt 0.975 0.975 0.987 0.967 1.000 0.986 1.077 1.066 0.870 0.981 0.968 0.992 West for
Page | 30 Aland 0.949 0.867 1.017 1.030 0.986 0.893 1.034 1.147 0.950 1.007 0.964 1.085 Table 4: Decomposition results for regions of Nordic countries (Denmark, Sweden and Finland). The column ‘Total’ contains the ratio of manufacturing value added shares in 2007 to 2000.
As with East German regions, Nordic regions saw a smaller decline in manufacturing value added shares from increased final demand of non-manufactures than most other EU25 regions (Table 4). However, if there had been no changes otherwise, increasing final demand levels for manufactures would have led manufacturing value added shares in the Nordic regions to increase between 3.4% - 14.0% based on within the EU25 markets (factor 6) and 5.6% - 14.7% based on outside the EU25 (factor 7). But decreasing trade shares for manufactures in both intermediate inputs and final demand inside and outside the EU25 (factors 3, 4, 9 and 10) seem to have negated these effects. Ita Suomi of Finland did experience an increase in the manufacturing value added share, but this seems to be the result of its declining competitiveness in trade of non-manufactures (factors 5 and 11). The regions of Greece are another interesting case, given the large heterogeneity in relative changes discussed in Section 1.
Greek regions
Region Total Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Factor 7 Factor 8 Factor 9 Factor 10 Factor 11
Greece Anatoliki Makedonia Thraki 0.651 0.909 0.964 0.835 0.990 0.984 1.193 1.016 0.810 0.970 0.999 0.961 Kentriki Makedonia 0.484 0.957 0.950 0.780 0.992 1.002 1.212 1.013 0.728 0.770 0.988 1.010 Dytiki Make-donia 0.848 0.977 0.904 0.941 0.992 0.915 1.136 1.033 0.880 1.146 0.997 0.954 Thessalia 0.737 0.992 0.940 0.904 0.994 1.011 1.223 1.015 0.809 0.872 0.997 0.996 Ipeiros 0.835 1.059 0.917 0.899 0.993 1.019 1.216 1.020 0.787 0.941 0.992 1.039 Ionia Nisia 0.878 1.001 0.935 0.863 1.025 0.925 1.221 1.041 0.945 1.110 1.018 0.843 Dytiki Ellada 0.770 1.013 0.919 0.917 0.996 0.972 1.216 1.019 0.810 0.951 1.003 0.974 Sterea Ellada 0.661 0.971 0.937 0.846 0.993 1.016 1.175 1.008 0.838 0.831 0.991 1.043 Peloponnisos 0.668 0.966 0.941 0.822 0.990 1.026 1.177 1.010 0.812 0.905 0.992 1.015 Attiki 2.875 1.265 0.947 1.633 1.043 0.973 1.229 1.046 0.762 1.451 1.024 0.995 Voreio Aigaio 4.609 1.353 0.961 1.339 1.065 0.987 1.333 1.057 0.787 2.036 1.097 1.018 Notio Aigaio 0.692 0.921 0.915 0.837 1.010 0.943 1.239 1.029 0.847 0.984 1.009 0.961 Kriti 0.628 0.925 0.946 0.846 0.999 0.984 1.276 1.033 0.808 0.839 0.983 0.982 Table 5: Decomposition results for Greek regions. The column ‘Total’ contains the ratio of manufacturing value added shares in 2007 to 2000.
Page | 31
in value added coefficients (factor 1), and large increases in trade shares for manufactures as intermediate inputs and final demand within the EU25 (factors 3 and 9). However, the majority of Greek regions seems to have deindustrialized more than average from effects of trade share changes of manufactures within the EU25, especially for intermediate inputs (compare factor 3 and Table 2).
Noteworthy is also the case of the two regions of Ireland, whose deindustrialization appears to be almost entirely rooted in the increased demand levels for non-manufactures (with a factor 8 of 0.738 and 0.750). French regions mainly deindustrialized as a result of increases in final demand of non-manufactures and changes in value added coefficients which may be linked to the outsourcing of services described by Rowthorn and Coutts (2004) (see Appendix A).
Conclusions
Using Structural Decomposition Analysis to decompose changes in value added shares of manufacturing in current prices across the EU25 regions reveals different experiences with the causes at work. As was expected based on previous research, the largest contribution to deindustrialization across EU25 regions stems from increased final demand for products from non-manufacturing services – most likely driven by services, though non-manufactures sectors were not further decomposed in the empiric analysis. Findings have also highlighted the key importance of international trade within the EU25 and on the world markets as a source of de- or industrialization. Thus, adding to Herrendorf, Rogerson and Valentinyi (2013) that the heterogeneity appears to not only stem from trade in the European Single Market, but also from trade on world markets in general. However, the findings of this paper do not seem to support the notion that industrial policy making capabilities should be built at the regional level as suggested by Stojcic and Aralica (2015). While there is heterogeneity in the deindustrialization or industrialization experiences for different regions from a quantitative perspective, regions within a country seem to be affected by the same causes through the same channels (as for example was the case for Greek and Nordic regions).
Page | 32
Page | 33
References
Alderson, A. S. (1999). Explaining deindustrialization: globalization, failure, or success?. American
Sociological Review, 701-721.
Aumayr, C. M. (2007). European Region Types in EU-25. The European Journal of Comparative
Economics, 4(2), 109.
Baldwin, R. E., & Evenett, S. J. (2015). Value Creation and Trade in 21st Century Manufacturing. Journal
of Regional Science, 55(1), 31-50.
Bathelt, H., & Kappes, K. (2008). Regional Deindustrialization and Re-bundling: Evidence from the Merger of the Former German Hoechst and French Rhône-Poulenc Groups. European Planning Studies,
16(10), 1329-1352.
Bell, D. (1976, May). The coming of the post-industrial society. In The Educational Forum (Vol. 40, No. 4, pp. 574-579). Taylor & Francis Group.
Bluestone, B., & Harrison, B. (1982). The Deindustrialization of America: Plant Closings, Community
Abandonment, and the Dismantling of Basic Industry. Basic Books.
Bogliaccini, J. A. (2013). Trade Liberalization, Deindustrialization, and Inequality: Evidence from Middle-income Latin American Countries. Latin American Research Review, 79-105.
Boulhol, H., & Fontagné, L. (2006). Deindustrialisation and the fear of relocations in the industry. CEPII Working Paper (No 2006-07, March).
Brady, D., & Denniston, R. (2006). Economic Globalization, Industrialization and Deindustrialization in Affluent Democracies. Social Forces, 297-329.
Dietzenbacher, E., Hoen, A. R., & Los, B. (2000). Labor Productivity in Western Europe 1975–1985: An Intercountry, Interindustry Analysis. Journal of Regional Science, 40(3), 425-452.
Dietzenbacher, E., & Los, B. (1998). Structural Decomposition Techniques: Sense and Sensitivity.
Econom-ic Systems Research, 10(4), 307-324.
Dietzenbacher, E., & Los, B. (2000). Structural Decomposition Analyses with Dependent Determinants.
Page | 34
Dietzenbacher, E., Los, B., Stehrer, R., Timmer, M., & De Vries, G. (2013). The Construction of World In-put–Output Tables in the WIOD Project. Economic Systems Research, 25(1), 71-98.
European Commission (2008). Nomenclature of Territorial Units for Statistics (NUTS) 2006 – Statistical Units – Data set [shapefile]. Available at http://ec.europa.eu/eurostat/web/gisco/geodata/reference-data/administrative-units-statistical-units/nuts.
Eurostat (2002). Statistical Classification of Economic Activities in the European Community, Rev. 1.1
(NACE Rev. 1.1). Available from
http://ec.europa.eu/eurostat/ramon/nomenclatures/index.cfm?TargetUrl=LST_NOM_DTL&StrNom= NACE_1_1&StrLanguageCode=EN&IntPcKey=&StrLayoutCode=EN.
Eurostat (2011). Regions in the European Union: Nomenclature of Territorial Units for Statistics NUTS
2010/EU-27. LU: Eurostat Methodologies & Working papers. Available at http://ec.europa.eu/eurostat/documents/3859598/5916917/KS-RA-11-011-EN.PDF.
Herrendorf, B., Rogerson, R., & Valentinyi, Á. (2013). Growth and Structural Transformation (No. w18996). National Bureau of Economic Research.
Iacovone, L., Rauch, F., & Winters, L. A. (2013). Trade as an Engine of Creative Destruction: Mexican Ex-perience with Chinese Competition. Journal of International Economics, 89(2), 379-392.
Kollmeyer, C. (2009). Explaining Deindustrialization: How Affluence, Productivity Growth, and Globalization Diminish Manufacturing Employment. American Journal of Sociology, 114(6), 1644-1674.
Krieger-Boden, C., Morgenroth, E., & Petrakos, G. (Eds.). (2008). The Impact of European Integration on
Regional Structural Change and Cohesion. New York, NY: Routledge.
Kucera, D., & Milberg, W. (2003). Deindustrialization and Changes in Manufacturing trade: Factor Content Calculations for 1978–1995. Review of World Economics, 139(4), 601-624.
Los, B., McCann, P., Springford, J., & Thissen, M. (2017). The Mismatch Between Local Voting and the Local Economic Consequences of Brexit. Regional Studies, 51(5), 786-799.
Page | 35
Miller, R. E., & Blair, P. D. (2009). Input-Output Analysis: Foundations and Extensions. UK: Cambridge University Press.
Nickell, S., Redding, S., & Swaffield, J. (2008). The Uneven Pace of Deindustrialisation in the OECD. The
World Economy, 31(9), 1154-1184.
OECD (2003). Territorial Grids of OECD Member Countries. Available at http://www.oecd.org/cfe/regional-policy/42392313.pdf.
Palma, G. (2005). Four Sources of “De-Industrialisation” and a New Concept of the “Dutch Disease”. In J. A. Ocampo (ed.), Beyond Reforms: Structural Dynamics and Macroeconomic Vulnerability. New York: Stanford University Press and World Bank.
Rodrik, D. (2016). Premature Deindustrialization. Journal of Economic Growth, 21(1), 1-33.
Rowthorn, R., & Coutts, K. (2004). De-industrialisation and the Balance of Payments in Advanced Econo-mies. Cambridge Journal of Economics, 28(5), 767-790.
Rowthorn, B., & Wells, J. R. (1987). De-Industrialization Foreign Trade. Cambridge, UK: Cambridge University Press.
Saeger, S. S. (1997). Globalization and Deindustrialization: Myth and Reality in the OECD. Review of
World Economics, 133(4), 579-608.
Stijns, J. P. (2003). An Empirical Test of the Dutch Disease Hypothesis Using a Gravity Model of Trade. Available at SSRN: https://ssrn.com/abstract=403041.
Stojcic, N., & Aralica, Z. (2015). Regional Patterns of Deindustrialization and Prospects for
Reindustriali-zation in South and Central East European Countries (No. 118). The Vienna Institute for
Interna-tional Economic Studies, wiiw.
Thissen, M., Lankhuizen, M., & Los, B. (Forthcoming 2017). Construction of a Time Series of Fine-grained
Detailed NUTS2 Regional Input-Output Tables for the EU Embedded in a Global System of Country Tables (Mimeo). The Hague: PBL Netherlands Environmental Assessment Agency.
Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R., & Vries, G. J. (2015). An illustrated user guide to the world input–output database: the case of global automotive production. Review of International
Page | 36
Timmer, M. P., Los, B., Stehrer, R., & De Vries, G. J. (2013). Fragmentation, Incomes and Jobs: An Analy-sis of European Competitiveness. Economic policy, 28(76), 613-661.
Tregenna, F. (2009). Characterising Deindustrialisation: An Analysis of Changes in Manufacturing Employment and Output Internationally. Cambridge Journal of Economics, 33(3), 433-466.
Tregenna, F. (2011). Manufacturing Productivity, Deindustrialization, and Reindustrialization (No. 2011, 57). Working paper//World Institute for Development Economics Research.
Appendix B
Larger Versions of
Maps (in order of
% Appendix C - MATLAB Code
% Deindustrialization in Europe - Carmen Zurcher
% Code may still contain snippets from various stages of debugging % Basic Calculations
% Load data
mat0 = cell2mat(struct2cell(load('Complete_00.mat'))); mat7 = cell2mat(struct2cell(load('Complete_07.mat')));
% calculate total manufacturing, total output, manufacturing share in % output, and share in value added per region
TM0 = CalcTM(mat0); Q0 = CalcQ(mat0); MShare0 = TM0./Q0; MVal0 = CalcMval(mat0); TM7 = CalcTM(mat7); Q7 = CalcQ(mat7); MShare7 = TM7./Q7; MVal7 = CalcMval(mat7);
[MValadd0, QVal0] = CalcMval3(mat0); [MValadd7, QVal7] = CalcMval2(mat7);
%xlswrite('X:\My Desktop\Thesis\wetransfer\MShare.xls', MVal0', 9, 'C2:C267');
% xlswrite('X:\My Desktop\Thesis\wetransfer\MShare.xls', QVal7', 8, 'I2:I250');
% xlswrite('X:\My Desktop\Thesis\wetransfer\MShare.xls', QVal0', 8, 'C2:C250');
%
% calculate change value added between 2007 and 2000
DeltaValAll = CalcDeltaValAll(MVal0, MVal7); DeltaSigma = MVal7 ./ MVal0;
%xlswrite('X:\My Desktop\Thesis\wetransfer\MShare.xls', DeltaValAll', 8, 'C2:C250');
%xlswrite('X:\My Desktop\Thesis\wetransfer\SDA.xlsx', DeltaValAll', 1, 'A2:A250');
%xlswrite('X:\My Desktop\Thesis\wetransfer\MShare.xls', DeltaSigma', 9, 'D2:D267');
Published with MATLAB® R2016b
function ValShare = CalcMval(vari) index = 1; index2 = 1; QVal = zeros(0, 266); MVal = zeros(0, 266); EmpComp = vari(3725, 1:3724); OtherVal = vari(3726, 1:3724); output = EmpComp + OtherVal; for i = 1:266 QV = sum(output(index:index+13)); QVal(i) = QV; index = index +14; end for i = 1:266 MV = sum(output(index2+2:index2+6)); MVal(i) = MV; index2 = index2 +14; end ValShare = MVal./QVal; end
Published with MATLAB® R2016b
function [MValadd7, QVal7] = CalcMval2(vari7) index = 1; index2 = 1; QVal7 = zeros(0, 249); MValadd7 = zeros(0, 249); EmpComp = vari7(3725, 1:3486); OtherVal = vari7(3726, 1:3486); output = EmpComp + OtherVal; for i = 1:249 QV = sum(output(index:index+13)); QVal7(i) = QV; index = index +14; end for i = 1:249 MV = sum(output(index2+2:index2+6)); MValadd7(i) = MV; index2 = index2 +14; end end
Published with MATLAB® R2016b
