T HE USE OF INDIVIDUAL FIRM DATABASES TO RESPOND TO THE LIMITS OF SPATIALLY AGGREGATED
DATABASES : T HE CASE OF THE ESTIMATION OF REGIONAL EXPORTS
Moritz Lennert
Promoteur : Jean-Michel Decroly
Travail en vue de l'obtention du titre de Docteur en Sciences Département Géosciences, Environnement et Société
Faculté des Sciences
Université Libre de Bruxelles
Curiously enough, the dolphins had long known of the impending demolition of Earth, and had made many attempts to alert mankind to the danger. But most of their communications were misinterpreted as amusing
attempts to punch footballs, or whistle for tidbits, so they eventually gave up and left the Earth by their own means - shortly before the Vogons
arrived. The last ever dolphin message was misinterpreted as a surprisingly sophisticated attempt to do a double-backwards somersault through a hoop, whilst whistling the ‘Star-Spangled Banner’. But, in fact,
the message was this:
“So long, and thanks for all the fish”
Douglas Adams, The Hitch Hiker’s Guide to the Galaxy
Table of Contents
A. WHAT'S IT ALL ABOUT ? THE INTERPRETATIVE
GUIDE TO THE LENNERT CHRONICLES...9
A.1. There's something missing: Current theoretical and practical contexts of regional development...13
A.1.1. Introduction...13
A.1.2. Show me what you have: The supply-side turn in regional policy...14
A.1.3. Choices: Space-blind or place-based...15
A.1.4. Blind on one eye: the neglected demand-side...17
A.1.5. Conclusions...19
A.2. Far out: The importance of geographical distance in contemporary geography... 21
A.2.1. Introduction...21
A.2.2. The gravity model in economic modeling of international trade : a success story...21
A.2.3. The gravity model in geography : distrust of physical explanations and laws ...22
A.2.4. The demise of geographic distance in economic geography...23
A.2.5. Travelling to other worlds : the importance of distance in related fields....25
A.2.6. A short note on the measure of distance...28
A.2.7. Conclusions...29
A.3. Big is the new black: hints of a future world of social physics and big data in economic geography ?...30
A.3.1. Introduction...30
A.3.2. How big is Big ?...31
A.3.3. A new paradigm for science ?...33
A.3.4. Is big data science really theory-free ?...34
A.3.5. Where to go from here ?...36
B. ORIGINAL PAPERS LEADING TO THE SPECIFIC
QUESTIONS DISCUSSED IN THIS THESIS...38
B.1. The Territorial Futures of Europe: ‘Trends’, ’Competition’ or
’Cohesion’ (Paper 1)...41
B.1.1. Abstract...41
B.1.2. Introduction...42
B.1.3. Context and methodology...43
B.1.4. Hypotheses...44
B.1.5. Trend Scenarios...48
Storyline...48
The final image...50
B.1.6. Cohesion-Oriented Scenarios...52
Storyline...52
The final image...53
B.1.7. Competitivenes-Oriented Scenario...55
Storyline...55
The final image...56
B.1.8. Roll-back scenario...58
Introduction...58
Final image...58
Storyline...59
B.1.9. Comparing scenarios...60
B.1.10. Conclusion...62
B.2. ESPON 3.4.2 Territorial impacts of EU economic policies and location of economic activities - Executive Summary (Paper 2)...63
B.2.1. Introduction...63
B.2.2. Always changing and still the same: the economic geography of Europe...63
GDP growth...63
From regional production to regional disposable wealth...66
Sectoral structures of Europe's regions...66
B.2.3. Regions embedded in nation-states...72
B.2.4. Towards complementary indicators of regional economic regulation...73
B.2.5. But why ?: Elements of explanation of the state and evolution of the localisation of activities...74
It's the economy, stupid!: The shifting economic, technological and political context of regional development...74
Now you're talkin': Shifting theoretical perspectives on regional growth and competitiveness...74
Spinning the globe: factors of localisation of firms...75
B.2.6. Oh no, don't press that button! : How macro-economic policies (might)
affect regional development...78
B.2.7. Small is beautiful: Regional Policies...82
B.2.8. So what is regional competitiveness? : a concluding concept discussion. . .83
B.2.9. Policy recommendations...85
At regional (micro-economic) level...85
At the general macro-economic level...87
B.3. Of bacteria and men: some reflections on modeling for regional policy (Paper 3)...91
B.3.1. The virtual petri dish...91
B.3.2. Identifying and feeding the bacteria...91
B.3.3. The difficulties of using the petri dish...92
B.3.4. From the petri dish to the future: the ESPON scenarios...93
B.3.5. Coming to the rescue of models...94
C. DEVELOPEMENT OF GIS METHODS...96
C.1. The use of exhaustive micro-data firm databases for economic geography : the issues of geocoding and usability in the case of the Amadeus database (Paper 4)...99
C.1.1. Abstract...99
C.1.2. Introduction...99
C.1.3. Micro-data in economic geography...100
The modifiable areal unit problem...100
New insights through micro-data...101
C.1.4. Data...102
C.1.5. Public Resources for Geocoding...103
The role of the INSPIRE directive...104
Overview of different forms of data dissemination...104
National idiosyncrasies in formats and contents of data...106
C.1.6. Geocoding...108
Introduction...108
Integrated web services...108
Address parsing...109
Techniques used for geocoding...110
C.1.7. Usability...112
Exhaustivity...112
The use of headquarters instead of establishments...115
Conclusions on the usability of the Amadeus database...117
C.1.8. Some maps as results...118
C.1.9. Conclusions...126
C.2. Building applications with FOSS4G bricks: two examples of the use of GRASS GIS modules as a high-level “language” for the analyses of continuous space data in economic geography (Paper 5)...127
C.2.1. Abstract...127
C.2.2. Introduction...127
C.2.3. The tools...129
A modular GIS for efficient raster processing and projection handling...129
The glue: Python...130
The GRASS Python scripting library...130
The accelerator: Simple parallel processing using Python’s multiprocessing module. 130 C.2.4. Example 1: Elaboration of new indicators for flexible-scale estimation of exports...131
Continuous space analysis in economic geography...131
Estimation of regional exports...132
Huff-like indicator...132
The modified Marcon-Puech indicator...133
Translating these equations into GRASS GIS “language”...134
Huff-like indicator...135
The modified Marcon-Puech indicator...135
Parallelization of the processes...136
C.2.5. Example 2: Calculating neighborhood matrices of raster objects...136
C.2.6. Conclusion...138
C.2.7. Annexe...139
D. BUILDING THE MODEL...140
D.1. Approaching regional openness through measures of specialization and spatial market shares : experimentation with micro-data on enterprises (Paper 6)...144
D.1.1. Abstract...144
D.1.2. Introduction...144
D.1.3. Huff-like model...144
D.1.4. M function-based model...146
D.1.5. Data and tools...147
D.1.6. Results...147
D.1.7. Conclusions...148
D.2. From point to point: estimating regional exports from the bottom-
up by applying a gravity model on individual firms (Paper 7)...150
D.2.1. Abstract...150
D.2.2. Acknowledgements...150
D.2.3. Introduction...150
D.2.4. Literature review...151
Supply-side focus in EU regional policy...151
On the estimation of intra-national trade...152
Conclusion...154
D.2.5. A point-based bottom-up gravity model...155
The gravity equation...157
Determining the input data...157
Stoppage criterion...160
Transferability and possible improvements of the model depending on data availability ...160
Practical implementation of the model...161
D.2.6. Concrete application of the model on Belgian firm data...161
Firm data...161
Input-output tables...163
Resolution of analysis, distance measure and spatial aggregation level...163
Validation...163
Geography of trade...166
D.2.7. Discussion...173
D.2.8. Conclusions and perspectives...173
D.3. Who's the customer ? How the supply-side turn of regional policy neglects the spatial production-consumption patterns of regional economic activity (Paper 8)...175
D.3.1. Abstract...175
D.3.2. Introduction...175
D.3.3. Literature review...176
All roads lead back to economic base theory...176
All roads ?...178
Whose demand for whose production ?...178
Difficulties of estimation...179
D.3.4. A new model for estimating regional export share...181
D.3.5. Who consumes the regional economic production ?...182
D.3.6. Exploring the impact of trade on economic and social performances...196
Economic performance...196
Social performance...197
D.3.7. Conclusions...199
E. CONCLUSIONS: ACKNOWLEDGING THE LIMITS AND
IMAGINING THE FUTURE...201
E.1. New Data and New Tools...202
E.2. The Gravity Model for Estimating Regional Exports...204
E.3. Regional Export Markets and their Relevance for Policy...206
F. COMBINED REFERENCES OF ALL PAPERS...208
G. ANNEXES... 230
G.1. List of Publications by the Author...231
G.1.1. Contributions to collective works...231
G.1.2. Peer-reviewed journal articles...231
G.1.3. Papers published in national and international conferences or symposium proceedings...234
G.1.4. Research reports, reading notes, book reviews, letters to the editor, working papers...235
G.2. The Author's Software Developments in GRASS GIS...236
A. WHAT'S IT ALL ABOUT ? THE
INTERPRETATIVE GUIDE TO THE LENNERT CHRONICLES
This PhD thesis is a combination of, on the one hand, previous work realized in the course of many years of research in different projects which I coordinated and which covers fairly largely scoped topics and, on the other hand, more recent work realized during my years as teaching assistant, which allowed me to do more classical individual research focused on a very narrow, specific topic. As such, it is thus as much an invitation to recognize my previous experience as well as actual “doctoral” work as proof of my capacity to do scientific research.
There is a clear line of continuity across all the papers and reports presented here as all deal, in one way or another, with issues related to regional development and economic geography which has been a topic for me ever since my first master’s thesis almost twenty years ago (Vandermotten and Moritz Lennert, 1999). Box Erreur : source de la référence non trouvée provides an overview of the papers that constitute this PhD thesis.
As presented in chapter B, the subject of the more narrow, focused work in the recent papers was mostly born out of the experience in the previous larger-scope projects in the ESPON program where a recurring theme of discussion was the lack of actual knowledge about many of the economic interactions and their spatial patterns at fine-grained resolution, but also the sometimes quite one-sided official view on regional development and the policies necessary for it. The two papers I chose to represent this reflection and to show my personal trajectory on that road are papers 1 and 2, which are the results of two different projects I coordinated:
ESPON 3.2 on spatial scenarios and ESPON 3.4.2 on the location of economic activities and territorial impacts of economic policies. These papers lay the groundwork motivating the later work. I present some further, updated, reflections on these issues in the first section of this chapter, where I show that the question of the geography of the distribution of regional production has clear political relevance as most current regional development policies are supply-side policies with demand considered as global and infinite at the scale of the individual firm or region and thus not relevant to regional policy. In continuation of this debate, the more recent work uses very fine-grained data in an attempt to contribute to an answer to the following scientific question : How much of their local economic production regions export to elsewhere and where do they export to ?
It is quite surprising to see how little we actually know about this issue. Data is either
unavailable, or it is not collected and treated in a way to allow its scientific study and
analysis. Analyses in the literature are generally either small case studies based on data
available for a few firms, on limited proxy data such as shipment of transport statistics or on
very simplistic models. Much of the data used is also aggregated at scales that make any
attempt at differentiating short-distance sales from longer distance interactions futile, leading to contestable results.
Most data used in the ESPON projects and in much of quantitative economic geography in Europe, as most data collected at Eurostat, is also at best at NUTS 2 level, which often does not allow differentiations that would be necessary to understand phenomena that rule economic activities in peoples’ daily lives. This data issue was the second motivation behind this PhD. In addition, based on the experience in the ESPON projects where we tried to work with individual firm data, and as the PhD was supposed to be in applied geomatics, the idea was to combine specialized techniques and new spatial data sources based on individual point-based data in order to see how they could improve our comprehension of local interactions of economic actors.
Three recent papers, regrouped in chapter D represent this attempt. The first (paper 6) covers early experiments using individual firm data which I did not continue at such, but which remain an interesting path of inquiry. These experiments formed the basis for the second paper (paper 7) which presents a model which uses national input-output data as constraint to a bottom-up gravity model based on individual firm data in order to estimate the share of regional production that is exported to outside each region. Finally, the third (paper 8) uses the results of that model to analyze the spatial patterns of consumption of different regions' production, as well as very tentatively evaluating the impact of exports on the economic and social performance of regions. The approaches presented in these three papers are very far from incontestable. They rely on a series of simplistic assumptions and on very limited information available in the data. May they be seen as some of the many shots in the (almost) dark of human science in the hope that they arrive at least somewhere near the dartboard.
As can be seen, I approached this question from a theoretical side, but mostly from a very practical, methodological side. Scientific research today is more and more centered on data, and as such the nature of the data, the technical options how to treat it and the epistemological questions around the role of data and methodologies have significantly gained in importance. Paper 3 is an early paper in line with these more methodological issues as it is a reflection note I wrote following a seminar at DG Regio on the integration of economic modeling into the regional policy processes at the European Commission, based on my experiences with the different modeling efforts in the ESPON 3.2 project. The debate between this earlier, critical approach to modeling, and my later extensive use of a spatial model is part of the larger epistemological debate I will come back to in the last section of this introductory chapter.
Beyond the specific question of modeling, however, I consider other practical aspects, such
as data collection and treatment and the practical issues linked to the development of
relevant algorithms, just as important. Unfortunately, these aspects, which I would estimate
to represent at least 80% of the daily work of modern researchers, are rarely recognized and
papers are often not deemed worth publication if they do not provide substantive conceptual debate, with basic methodological and data issues often relegated to annexes or completely left out. As I do not agree with this tendency, I deliberately decided to include several practical methodology papers in this PhD, notably in chapter C. Paper 4 deals with the question of the availability and usability of individual firm-level data, including the issue of geocoding such databases. For the actual modeling exercise I ended up using a different data set than the one discussed in this paper, but this takes nothing away from the conclusions drawn therein. Paper 5 is a discussion on how existing GIS packages, especially their raster-based routines, can be very useful tools for economic geography once it moves towards the use of large, point-based database analyzed in real space (as opposed to the theoretical spaces used in many econometric models, notably in the so-called “New Economic Geography”). All the modeling in the recent papers was done using such a GIS packages, GRASS GIS (Neteler et al., 2012).
Paper 5 also represents another issue that I consider of prime importance for research, and more largely for modern society: the need for open science, including in its practical manifestations of accessible open source code and open data. The paper was presented at the peer-reviewed academic track of the 2016 FOSS4G conference which gathers actors of free and open source software in geomatics. As discussed in the conclusions of this paper, and in the last section of this chapter, open source approaches are an indispensable part of science.
This is also reflected in the work I’ve done parallel to this PhD integrating state-of-the-art algorithms in remote sensing into GRASS GIS in order to provide open source alternatives to the black-box proprietary solutions. This work has lead to a series of co-authored journal and conference papers independent of those presented here (Beaumont et al., 2017;
Georganos et al., 2017; Tais Grippa et al., 2017; Taïs Grippa et al., 2017; Vanhuysse et al., 2017).
In the next sections of this first chapter, I will provide some deeper discussions of some of the aspects discussed in the different papers. I begin by giving a more up to date vision of current regional development policies and research, arguing that its heavy supply-side bias misses some of the most important aspects of regional economies. I then go on to discuss a fundamental issue which lies at the basis of the gravity model used, but which has been subject of intense discussion in the regional science community: the role of distance in the interactions between economic actors. I argue that this role is important and thus justifies the use the gravity approach. Finally, I examine some issues and debates around the use of “big data”, notably a possible shift in scientific research paradigms linked to the move to a more data-based science, including in economic geography. I end with global conclusions discussing the limits of the work presented, and perspectives for future research endeavors.
A final note of disclosure: in line with postmodern lines of thought, I agree that the position
of the researcher is definitely not neutral, whatever the cautions taken in the application of
methodologies. My research clearly comes out of a line of critical thought about economic
development based on a moral imperative of social and spatial justice (Harvey, 1973; Smith,
1984) and on a radical reading of socio-economic relations, grounded in Marx’s analysis of
capitalism and his more modern disciples (Harvey, 1989, 1984). While this positioning
underlies all of the present papers, it is particularly visible in in this introductory chapter, but
also in the last paper with its discussion of the importance of external trade for regional
economies. Earlier papers such as the one on the spatial scenarios, but also the ESPON
project 3.4.2 report, also reflect thus vision. It is important to note, however, that this
situatedness does not preclude a positivist approach to science and that the results of the
methods presented are hypotheses that can and should be tested and (in)validated by others.
A.1. T HERE ' S SOMETHING MISSING : C URRENT
THEORETICAL AND PRACTICAL CONTEXTS OF REGIONAL DEVELOPMENT
A.1.1. Introduction
Theories and practices of regional development have evolved significantly in the last decades. Papers 1 and 2 relate to the debates a decade ago, when it was framed essentially around the opposition competitiveness vs cohesion, but where the notion of “place-based”
approaches to regional policies started to gain ground.
Excellent overviews about the general historical evolutions exist (Barca et al., 2012; Capello, 2009; Cochrane and Poot, 2014; Dawkins, 2003; Pike et al., 2017; Torre and Wallet, 2016), so I will not repeat them here. In brief, the general movement of both theory and practice of regional economic development, in line with general economic development, has been from demand-side approaches in a Keynesian tradition to supply-side approaches in line with the neo-liberal theories which came to the forefront since the 1980s. At the same time, micro- economic and micro-behavioral theories have gained in importance, supplementing, or sometimes even replacing, macro-economic theories. In regional development, this is combined with the debate about place-based policies versus space blind (or, for some authors, people-based) policies (Barca, 2009; Barca et al., 2012).
In addition, on a more normative, or even moral, side, policies have shifted from the idea of convergence through redistribution, i.e. a collective responsibility for the well-being of the individual (be it a person or a region) to a notion of convergence through development, i.e.
the individualization of that responsibility, generally linked to the notion of “untapped resources” (Barca et al., 2012; Capello, 2009) that, if activated, would allow the individual some form of self-sustaining development (Pike et al., 2017).
All this has led to a situation where policies for individual regions are tailored to the
specificities of these regions, often, however, without taking into account the role these
regions play within larger contexts, and the role these larger contexts play for the
development of the regions. As Pike et al put it: “a focus only on the endogenous
characteristics of places as determinants of local growth draws attention away from other
critically important and co-constitutive exogenous factors that shape local and regional
development including the restructuring of international divisions of labour, national political
economies and macro-economic shifts” (Pike et al., 2017, p. 48). This lack of embeddedness,
as well as the prime focus on supply-side elements, without asking the question of who
consumes the products produced, and where they are consumed, thus limits regional development policies and calls for more analysis of such questions, such as those presented in my recent work (notably papers 7 and 8). In the following, I will discuss these issues more in detail.
A.1.2. Show me what you have: The supply-side turn in regional policy
To begin with, current regional policy is without doubt clearly oriented towards the supply- side. Any aspects of redistribution - the classic, Keynesian means of enhancing territorial and social cohesion by spatially distributing demand - is completely absent from any of these documents. Be it the EU regional policy currently focused on “smart specialization” (DG Regional and Urban Policy, 2017; European Commission, 2017, 2015, 2014, 2010) or the recommendations of the OCDE (OECD, 2016, 2012) at supra-national scale, or policies at regional scale such as, in Belgium, the Walloon Plan Marshall and the Brussels international development plan (Ministre-Président de la Région de Bruxelles-Capitale, 2007; Walloon Region, 2015).
Implicit in most of these approaches is the notion that regions have to be competitive in outside markets, with several reasons advanced on why this is important. Storper and many of the evolutionist approaches see this competitiveness as source of monopoly rents that ensure firms with a high share of value added in the value chain (Boschma and Martin, 2010;
Storper, 2008). Such monopoly rents can also provide firms in these regions with the means to gain financial control over firms in other regions, thus gaining a role of command and control within the international economic circuits. Others, notably in the heterogeneity literature, but also in the reflections on milieu and in institutional economic geography, highlight effects of export activities on regional productivity (Amin and Thrift, 1995; Bernard et al., 2012; Camagni, 2005; Greenaway and Kneller, 2007; Melitz and Redding, 2015). It is important to note, however, that the more a region is active on outside markets the more its export sectors are also dependent on the general evolution of these markets, thus determining which sectors are growing or shrinking, often independently of the regional structures.
None of the above-mentioned policy documents ever raise the question of the demand needed
to realize the value of the regional production, however. Be it internal demand within the
region, or external demand. In general, supply-side factors are seen as those that hamper
regional growth, notably in lagging regions. The Commission notes that “The
macroeconomic framework has a significant impact on regional economic growth”, but does
not follow through on that insight, except to call for less restrictive macro-economic regulations in order to give individual regions more flexibility to develop their own supply- side policies (DG Regional and Urban Policy, 2017). There is some admission that the current economic crisis has been “particularly tough for those regions relying more on domestic demand”, but no single policy recommendation concerning means to stimulate this demand can be found. On the contrary, the main element brought forward concerning these lagging regions are elements of cost factors, notably rising labor costs, which are deemed responsible for the reduction of export shares, but no reflection can be found concerning other aspects of demand evolution during the crisis. Similarly, when the OECD presents scenarios of how improvements of certain supply-side factors could influence regional growth, there is no mention whatsoever of the fact that this growth can obviously only happen if there is sufficient demand for the regions' products, and no reflection on the fact that this logic only works if all other (competing) regions stagnate in terms of their supply-side factors during that time (OECD, 2012). Demand is considered as infinite from the point of view of any given region, because demand is considered as being global, without analyzing into any depth the actual share of economic production that really is in contact with global demand.
A.1.3. Choices: Space-blind or place-based
These options are not limited to policies, however. Research in regional development has clearly accompanied and fostered the political institutions' approach. One debate particularly present in Europe is the one about space-blind versus place-based policies, represented respectively by two reports for the European Commission (Barca, 2009; Sapir, 2004). The space-blind approach called for by Sapir (but also by the World Bank (2009)) is in line with neoclassical approaches in which “space plays no part in determining the development path of a local economy” (Capello, 2009), even in the so-called “New Economic Geography (Krugman, 1991) where space “is stylized into points devoid of any territorial dimension”
(Capello, 2009).
In contrast, the place-based approach called for by Barca and relayed today by the new
“smart specialization” model (McCann and Ortega-Argilés, 2015; Morgan, 2015) is grounded in a community of schools of thought in economic geography
1which include those dealing with agglomeration (Becattini and Società Italiana degli Economisti, 1989; Benko and Lipietz, 1992; Marshall, 1890; Piore and Sabel, 2000), innovation and innovative “milieu” or context (Aydalot and Groupe de Recherche Européen sur les Milieux Innovateurs, 1986;
Camagni, 2005; Maillat, 1995; Storper, 2008), evolutionist approaches (Boschma and Frenken, 2005; Boschma and Martin, 2010; Grabher, 2009; Martin and Sunley, 2006), clusters (Porter, 2003), institutional economic geography (Amin and Thrift, 1995; Martin, 2003) and many other forms of what Capello calls theories around “diversified-relational space” (Capello, 2009). All these territorialized theories were generally born from a critical vision of economic growth models that were deemed too simple. They have introduced many
1 From here on, when I refer to economic geography I mean economic geography « proper » as opposed to the « new economic geography » of economists. I will always use the additional « new » to designate the latter.
tangible and intangible factors that explain why some regions have more economic resources than others or are more successful in exploiting them. They have clearly fundamentally enriched our vision of regional development and have lead to a turn in regional policies away from one-size-fits-all approaches.
Two debates that arise out of the different views of two schools are interesting to highlight.
One is a consequence of the focus on intangible elements in economic geography. This focus has created the image that many regions have important economic resources, but that they have just not been able to use them efficiently. Policy's role, therefore becomes to unleash these potentials, all on the supply-side, which will automatically lead to economic growth.
One can find many references to the idea of such “untapped resources”, both in the scientific and the political literature (Barca et al., 2012; Capello and Kroll, 2016). This echos many of the “activation” schemes applied to the unemployed, based on the notion that unemployed are unemployed because they do not supply the right skills, not because there is not enough demand. The same logic seems to enter more and more into regional development policy and theory where “underdevelopment traps that limit and inhibit the growth potential of regions or perpetuate social exclusion are the result of a failure of local elites to act”, regions need a
“sense of community” and “the success of place-based policies requires that local and regional actors are galvanized to play a constructive and leading role in the policy” (Barca et al., 2012). If this is not the case, fault for the lack of development in these regions seems to be solely their own, without any of these authors providing definitions of regions, nor clear indications of what this “community” is that is so desperately needed. Barca et al go even further by declaring that “if convergence is to be promoted, this is to be done by development rather than by redistribution”, thus clearly rejecting any spatial redistribution of wealth. There is, thus a sense that regions are autonomous, well-defined entities that are responsible for their own well-being, with no notion of multi-scalar factors, inter-dependencies and exogenous influences that could influence regional development. By trying to avoid the pitfalls of over-simplified, neoclassic, one-size-fits-all, and space agnostic approaches, theory and practice seems to now swing the pendulum to the other extreme.
The second debate relates to an almost moral question of how important spatial diversity and the maintenance of diversified populated spaces actually are. In the ESPON 3.2 scenarios, notably the normative scenario, we pleaded for the idea that policy makers and inhabitants would probably have to accept the fact that some regions will lose population, going as far as imagining that “some are used as biodiversity reserves and areas for soft tourism, with only very few people actually living in them” (Lennert and Robert, 2007, Vol 2). Behind this idea was a certain compromise between the desire inherent to most regional planners to allow regional populations and structures to maintain their existence, and the force of agglomeration tendencies which cause very high levels of out-migration in some, notably Eastern European regions. The above territorially-based approaches, and notably the policies promoted by the EU, follow the first postulate that all regions in Europe should develop, and that all inhabitants have the right to adequate living standards where they are.
The neoclassic, space-blind conception, on the other hand, advocates perfect factor mobility
in order to increase efficiency, and thus higher levels of growth which could then lead to
higher levels of well-being for individuals. Hence the notion that these space-blind approaches are people-based, instead of place-based. In its 2009 report (World Bank, 2009), the World Bank goes as far as moving from the observation of agglomeration to a moral
“imperative of agglomeration” (Van Hamme et al., 2010), based on the idea that since agglomeration drives development, policies should encourage such agglomeration in large cities, in the hope that the wealth created through the process can then serve as basis for further development of the country. While this approach is highly criticized, notably for its negligence of negative externalities of agglomeration (Barca et al., 2012; Dijkstra, 2013;
Dijkstra et al., 2013), the possible existence of such negative effects does not seem sufficient to rule out the fact that possibly agglomeration is the path that will lead to more wealth, thus leaving the moral question difficult to decide. The debate raises or reinforces questions of scales of action and scales of definition of goals of policies. Without a clear understanding of the interactions between scales and between regions, be they neighboring or further apart, it is difficult to actually relate this moral debate to the necessary evidence.
A.1.4. Blind on one eye: the neglected demand-side
The previous sections have shown that current public policy is clearly oriented towards supply-side policies, that these policies are grounded in a notion that regions have untapped resources explored by the current economic geography literature and that the policy consensus (or imperative) within the EU seems to be that all regions should develop (generally limited to the sense of increasing GDP/capita) to avoid excessive agglomeration, based on an activation of their resources. Demand is seen as global and infinite at the scale of the region, and is thus a negligible factor, except possibly for extreme crises situations as the economic recession that began in 2008, but even then the common responses are supply-side policies in order to enhance productivity.
Very few authors in the above-mentioned currents actually mention demand, or only when explicitly discussing demand-driven theories and models in history. Cochrane and Poot hint at a reason for this when they write “demand-led growth is not a regional development policy as such but rather a national or international project within which regional strategies can be articulated” (Cochrane and Poot, 2014). This means that such policies cannot be lead as pure regional development policies, but are multi-scalar. In the EU, however, regional policy is a very distinct competency of the EU institutions while macro-economic regulatory competencies are only partly transferred from the nation-states to the EU, notably through the so-called “Stability and Growth Pact” (European Parliament and the Council, 2011). One can, therefore, hypothesize that in terms of institutional logic, regional policy has to remain regional, especially since this allows the European Commission to interact directly with regions without always having to compromise with the national level.
So how does demand play in the regional economic development processes ? As we show in
the ESPON project on territorial impacts of EU policies, regions are clearly embedded in
national development trajectories, with two thirds of variance between NUTS 2 regions' GDP
growth explained by variance between countries (see paper 2). In line with other authors
(Duménil and Lévy, 1993; Harvey, 1989; Husson, 1996; Savage, 2005; Stockhammer, 2011), we argue that it is the general reduction of the share of productivity gains going to wages that has lead to a crisis of under-consumption, i.e. insufficient demand as a result of neoliberal reforms, and that regional economic development is embedded in this process.
When exploring demand-driven models of regional growth, however, it is interesting to see that in many models, demand is modeled as a single variable of external income, thus representing one single external market (Cochrane and Poot, 2014). This is also the fundamental idea of the entire literature on the economic base model (Krikelas, 1992), which has been severely criticized, notably the different strands of territorialized, supply-side theories cited above, but still dominates a non-negligible part of practical regional development approaches. Cochrane and Poot claim that “the greatest source of regional fluctuation in autonomous demand is likely to be regional exports, which are - for given competitiveness - fully driven by incomes in other regions and countries” (Cochrane and Poot, 2014). It is worth highlighting that they acknowledge that other regions in the same country, and thus the same system of economic regulation (Aglietta and Fernbach, 2000;
Boyer, 1990), also play a role in regional growth. However, the literature clearly lacks a more spatially differentiated approach to demand, mostly because of the absence of relevant data.
An exception is the survey work by Peter Cabus on the spatial inter-firm sales relationships in Flanders (Cabus and Vanhaverbeke, 2006), which shows that a large part of interactions are within the Flemish region, and between one tenth and one third of output is sold within a radius of 15km.
At the same time, starting with Kaldor, a series of cumulative causation models of endogenous growth include notions of local demand-driven growth, by postulating that a combination of exogenous demand and local demand (represented by output of the preceding year) drives productivity growth, and thus economic growth, notably, as Torre and Wallet note, through “reciprocal spillovers between production and consumption activities” (Dixon and Thirlwall, 1975; Kaldor, 1970; Krugman, 1991; Torre and Wallet, 2016). Internal demand thus seems a factor that cannot just be ignored when thinking about regional development.
Obviously, the notion of internal is highly dependent on the size of regions and it is thus necessary to develop tools that allow avoiding modifiable areal unit problems.
The notion identified above of an abstract, global and thus more or less infinite demand thus has to be qualified into a combination of differentiated demand sources, from within the region, from its neighbors and from within its national context, from levels of larger regulation systems and free trade zones such as the EU, and global demand. Regional economic development can, therefore, not be thought as independent of the macro-economic evolutions within which it takes place. This includes questions of wage distribution (Lavoie and Stockhammer, 2013), but also more generally distribution of resources within the national society. As Pike et al note, “increasing inequalities also have wider economic impacts dragging upon further growth, limiting the expansion of demand and consumption by groups experiencing stagnating or declining relative incomes” (Pike et al., 2017).
Going back to the quote from Cochrane and Poot above, it is important to note that the
authors speak about “fluctuation” in demand, not total demand. This highlights another issue
generally not made explicit in the entire literature about supply-side approaches aiming at competitiveness of regional production on external markets: most production within a region is actually not within this realm of competition, but caters to the local population. While the economic base literature has caused a long tradition of considering these sectors as unimportant, or just derived from the resources created by those sectors that are traded with the outside, others actually highlight that these sectors provide the large bulk of employment in most regions and offer different, but plausible development paths for regions. Some authors have developed the notion of the “foundational economy, which employs 40% of the workforce and is both private and public, is the sector of the economy that provides goods and services taken for granted by all members of the population and is therefore territorially distributed” (Bentham et al., 2013). Speaking of this literature, Pike et al note that “such approaches though require supportive political and economic frameworks, including multilevel systems of government and governance and fiscal redistribution”. Notions that seem to go against the current trend of place-based, self-sustaining forms of development.
But the proponents of the “foundational economy” also note that “direct measurement of the foundational economy is impossible because official statistics were not designed to do this”
and thus the classification of what is considered as being part of this economy “involves judgement, its boundaries are necessarily imprecise, and will need further specification”
(Bentham et al., 2013). Analyzing the spatial patterns of production and consumption of outputs of regional firms might be one path to shedding some light on the contours of this part of the economy that might offer a different solution to the issues of regional development, notably, but not only, in lagging regions (Servillo et al., 2016).
A.1.5. Conclusions
The discussions of the policies and science of regional development in the light of current debates has shown that there are still a series of areas where we do not have enough knowledge in order to make informed decisions. One is the level and intensity of multi-scalar factors, inter-dependencies and exogenous influences that influence regional development, allowing us to move away from a too simplistic vision based on the sole endogenous resources of a region. The second open field are the questions of scales of action and scales of definition of goals of policies, which are difficult to answer without adequate empirical knowledge of which factors play out within the region. Third, we need a spatially more differentiated analysis of the demand-base of regional economies, instead of just assuming one single external market, as different markets might be more or less influenced by specific macro-economic evolutions that have to be taken into account when trying to understand regional development. Finally, a more thorough analysis, including a clearer identification of these large parts of the economy that are sheltered from foreign competition, but that fulfill a fundamental role in the well-being of a region's population will allow formulating more differentiated policies.
On a more theoretical level, a synthesis of the seemingly opposed macro- and micro-
foundational approaches, as formulated by Capello seems a desirable perspective: “still
required is the further step forward which would produce an approach combining the economic laws and mechanisms which explain growth, on the one hand, with the territorial features that spring from the intrinsic relationality present at local level on the other. Such an approach would represent the maximum of cross-fertilization among location theory, development theory, and growth macroeconomics; a synthesis which would bring out the territorial micro-foundations of macroeconomic growth models.” (Capello, 2009)
An important methodological question linked to the issue of scales of analysis is that of
whether the distance between economic actors and between economic production and
economic consumption plays an important role in understanding interactions. The next
section provides an overview over the debate on this subject.
A.2. F AR OUT : T HE IMPORTANCE OF GEOGRAPHICAL DISTANCE IN CONTEMPORARY GEOGRAPHY
A.2.1. Introduction
The use of a gravity model in papers 7 and 8, as well as the other methods presented in paper 6, raise a series of issues in connection to the debate in contemporary economic geography. The gravity model as such has known criticism, both for specific aspects of the model itself and for its seemingly « physicist » vision of human geography, with the notion of
« laws » of geography leading to intense discussions. At the same time, the basic notion that lies underneath the logic of the model, i.e. that geographical distance is an important factor in explaining, or at least predicting, geographic patterns and behaviors, is also subject to intense scrutiny and critique, especially among economic geographers.
A.2.2. The gravity model in economic modeling of international trade : a success story
The gravity model has a history of over a century, but was introduced more forcefully in the 1950s in line with the quantitative turn of social sciences (Bergstrand and Egger, 2013), including its development in geography, generally under the name « spatial interaction models », with Wilson’s formulations probably the most well-known for their formalization and anchoring in theoretical background (Wilson, 1971). In economics, Tinbergen is credited to be the first to have introduced the « benchmark » gravity equation of international trade (Tinbergen, 1962).
Interestingly, whereas the postmodern turn in geography has led to rejection of the gravity
model by many economic geographers (although it is still widely used in transport and
migration studies), economists have adopted it more and more for the reason that it is one of
the most empirically successful models of spatial interaction (Anderson, 2011; Bergstrand
and Egger, 2013; Head and Mayer, 2014). Today, while subject to continuing criticism and
evolution (Anderson and van Wincoop, 2004), it is used extensively in modeling of
international trade, FDI, but also migration. Spatial friction, an important component of
spatial interaction, is also one of the fundamental elements of the so-called « new economic
geography » in the form of transport costs (Krugman, 1991).
A.2.3. The gravity model in geography : distrust of physical explanations and laws
In economic geography, the gravity model has met with increased criticism linked to different reasons : a critique of the absence of explanatory foundations to the gravity model, general, postmodern distrust of mechanistic, « social physics » (Stewart, 1948) models combined with the rejection of the notion of laws in human geography and, as we will see in the next section, the growing perception of the concept of geographical distance as too limited for explaining spatial patterns in the economy.
Social physics in itself can be defined as a science “applying principles and methods of physics to subject matter traditionally investigated by other disciplines such as economics, psychology, sociology, political science, and geography” (Barnes and Wilson, 2014). As one of its proponents, Stauffer, notes: “It is the law of big numbers which allows the application of statistical physics methods” (Stauffer, 2013). General criticisms of this approach are based on the idea that “the call to quantify and to represent social-spatial activity enacts a specific visibility that potentially renders invisible other phenomena”, such as real-life human beings, but also power relations and inequalities (Barnes and Wilson, 2014).
Besides this general critique of social physics (for further elements of this discussion also see the section on big data), an early line of critique of the gravity model was based on the apparent lack of its theoretical grounding as well as elements of misspecification (Barnes, 1991; Olsson, 1970, 1970; Sheppard, 1978), leading some to outrightly reject the model because « a rationalist justification of the gravity model based upon logical impeccability has yet to be realized » (Barnes, 1991). Others have found potential weakness of the model in specific difficulties such as the MAUP issue related to the use of spatially aggregated data in gravity equations (Fotheringham, 1984; Openshaw, 1977; Sheppard, 1984; Webber, 1980).
Besides the obvious problem of using mean distances between aggregation unit instead of distances between actual origin and destination sites, another main issue regarding aggregation is the difficulty to estimate intra-unit distances: “The intrazonal cost problem is more difficult since there has never been much agreement about how to estimate realistic costs for intrazonal travel”. He goes on to say that “if no satisfactory theory can be developed to handle the zone-design problem, and quite independent of whatever else it may be, an optimal zoning approach may be the only practical solution” (Openshaw, 1977). With the work presented here I claim that instead of trying to find optimal zoning solutions, using individual, precisely located data solves the issue even better. Webber, highlighting “the problem that empirical estimates of parameters seem to depend on the zonal system employed”, also seemingly supports the idea of using individual data by saying: “individuals behave, zones do not” (Webber, 1980).
Today, while specification issues are still being discussed among economists, the existence of
at least partially satisfying theoretical groundings of the gravity model as used for
international trade seem widely accepted: “Theoretical foundations have been developed over
the last thirty years that provide a rigorous, logical, economic rationale for its econometric
use, although such foundations are not yet complete” (Bergstrand and Egger, 2013).
When modeling in search for causal explanations, theoretical foundations are obviously important. Economists have added such foundations in later years. In the beginning, the gravity model was however considered very useful, even without theoretical justifications, as it allowed useful predictions, as even recognized by geographers: « The [...] gravity model may well provide powerful short-term predictions but [...] its explanatory value is dubious .»
(Olsson, 1970). The reason Olsson advances for this power of prediction is that “the specification errors are masked and statistically counteracted by existing autocorrelations for which distance can serve as an effective proxy variable” (Olsson, 1970).
This explanation is particularly interesting as Getis has shown that it is possible to formally link the notion of spatial autocorrelation with that of spatial interaction (Getis, 1991). This means that there is more than just random autocorrelations at work as Olsson seems to suggest, but that the foundation of the gravity model, spatial interaction, is actually linked with one of the fundamental elements of analytical geography, phrased by Tobler into his
“first law of geography”: “Everything is related to everything else, but near things are more related than distant things” (Tobler, 1970). This idea has lead to a whole series of advances in spatial statistics, heavily used both in geography and in economics (Anselin, 1995; Getis and Ord, 1992).
Although to many geographers and spatial analysts Tobler's law seems like an obvious given, it has also been the subject of debate (see notably Barnes, 2004; Goodchild, 2004; Miller, 2004; Phillips, 2004; Smith, 2004; Sui, 2004; Tobler, 2004), spurned by the distrust by modern geographers of the notion of laws in human geographers: “Any suggestion that there are geographic laws excites strong feelings in geography, a discipline whose incomplete transformation into a nomothetic science in the 1960s has been further complicated by more recent imports of antifoundationalism, post-structuralism, and the critical insights of science studies” (Smith, 2004). This distrust of mechanistic explanations of human behavior (as those found, for example, in Asimov's fictitious “psychohistory (Asimov, 1951)), and of spatial patterns resulting from this human behavior has led not only to the rejection of notions such as spatial autocorrelation as having no meaning if not filled with relational content (Phillips, 2004), but also to the rejection of geographical distance and its corollary, proximity, as valid factors of explanation in geography: “The degree of relatedness between two things is best understood as a function of properties other than their spatial proximity (Smith, 2004).”
A.2.4. The demise of geographic distance in economic geography
« Economists […] have rediscovered agglomeration and geographical distance as a
fundamental driver of the location of economic activity. Geographers […] seem to have
become liberated from the yoke of analysing interaction in geographically-constrained spaces
and have put their sights into exploring relationships which need not be limited by physical
contiguity » (Rodriguez-Pose, 2011). This evolution in economic geography can be
summarized under the general terms of « relational turn » (Boggs and Rantisi, 2003). Capello makes the distinction between “diversified-relational” space of geographers and regional scientists, as opposed to the “diversified-stylized” space used by the “new economic geography” (Capello, 2009; see also Corpataux and Crevoisier, 2007).
This redefinition of space can be linked to a host of recent theories in economic geography (see list of schools in the chapter on place-based vs space-blind approaches), all of which have in common that they focus on the importance of different dimensions of relationships between economic actors inside and outside a given region. They have led to a more complex definitions of proximity, with geographical proximity only one in many - and certainly not the most important - (Rutten, 2017; Simandan, 2016; Smirnov, 2016). In a simplified manner, one can say that these schools defend a somewhat Schumpeterian view of regional development in that they explore which types of relational and institutional patterns enable regions to engender innovations that allow these regions to compete on the globalized market.
These patterns are often multi-scale, and based on “social spaces [that are seen as not]
necessarily connected to any specific geographical place” (Rutten, 2017). Or, as Rodriguez- Pose writes: « For geographers, the dominance of large metropoli derives from the relational proximity among economic agents [...]. These relations create their own multifarious space within their interactive processes (Garretsen and Martin, 2010, pp. 141–42), a space completely detached from geometric space. » (Rodriguez-Pose, 2011).
Many economic geographers, however, do not share this rejections of geographical space as prime explanation of spatial economic patterns. Crevoisier, for example, calls space a “prime, irreducible reality that irrevocably places each individual in a specific economic structure that is both constraining and liberating”, leading to the idea that proximity is “a backcloth of inter- relations that structures the economy and actors' behaviour” (Crevoisier, 1996). Healy and Morgan denounce “an over-exaggerated sense of what can be accomplished at a distance”
which goes counter to the idea that “distanciated mechanisms do not offer the same scope for reciprocity, trust, understanding and serendipity that is afforded by sustained [face-to-face]
contact” (Healy and Morgan, 2012).
While open to the idea that different forms of proximity do play an important role, several authors reject the idea that this means that geography does not matter anymore: “The 'Death of Distance' argument that dominated much of the early literature on the Internet and cyberspace (e.g., Caincross 1997) is simplistic because it assumes that communication has only a substitution relationship with transportation (i.e., more virtual interaction implies less physical movement). In fact, empirical evidence suggests that the opposite is the case: the rise of telecommunication demand has been paralleled by a corresponding increase in travel demand at all geographic scales (Couclelis, 2000)” (Miller, 2004). And “organizational proximity may be a partial substitute for geographical proximity, especially for people who are already part of a community of practice, but partial is the operative word (Blanc and Sierra, 1999)” (Morgan, 2004).
It is important to note that this apparent demise of distance in economic geography is based
on a focus of research on factors of differences in economic growth (or more abstractly
economic competitiveness). This means that what these authors try to explain is not the bulk
of functioning economies, but rather the marginal differences that explain relative differences in wealth and income. As such these approaches also focus on micro-behaviour elements, sometimes forgetting the macro (Capello, 2009). In a certain way they can also be considered as somewhat elitist as they completely neglect the foundational economy (Bentham et al., 2013; Bowman et al., 2014) that determines everyday life of most of us and that is not in competition with the outside world. This everyday life is obviously physically and spatially bounded by many practical aspects and so most economic interaction takes place within these bounds. Finally, such spatial bounds are also put in place by the political regulation systems that determine most of the rules and practicalities of economic activities (Benko and Lipietz, 1992; Boyer, 1990; Morgan, 2004). By neglecting these bounds, modern economic geography sometimes seems to have lost sight of the general situation of the forest the trees of which it is counting.
The debate is still ongoing within the discipline and contrary to what the above selection of quotes may indicate, many authors do not have a simplistic black-and-white vision of the role of geographical space. Much of the strong rejection of the role of geographical space and proximity/distance has been a reaction to forms of “spatial fetishism”, i.e. the hiding of underlying social relationships behind a too strong focus on spatial structures (Castells, 1972;
Harvey, 1973; Lefebvre, 1974). Just as Marx denounced commodity fetishism in the Capital (Marx, 1872), the idea here is that spatial relations are just expressions of social relations and should thus not be taken at simple face value, nor as the natural way of things. As Lefebvre puts it “The fetishism of an abstract economics is being transformed into the fetishism of an abstract economic space” (Lefebvre, 1974). Combating such spatial fetishism remains an important social project as it means combating the maintenance of structures of exploitation posing as immutable spatial structures.
However, as I will show in the next section, many other disciplines have advanced on the question of the importance of proximity in a more empirical way and highlight that distance is definitely not dead. While it is necessary to remain critical of these approaches, it is also important to remain open to the idea that for many people geographical space is a relevant dimension in which they organize and structure their lives and which serves as an important reference in decision-making and relationships.
A.2.5. Travelling to other worlds : the importance of distance in related fields
One field of study where distance serves as an important element, at least in the form of spatial friction generally through transport costs, is that of trade. An important result in the study of international trade through the gravity model has been that contrary to the generalized intuition of space-time compression through globalization and new communication technologies (Cairncross, 1997; Harvey, 1989; Leamer and Storper, 2001;
Tranos and Nijkamp, 2013), the importance of distance has actually increased (Carrère and
Schiff, 2005; Disdier and Head, 2008), including for tradable services (Kimura and Lee,
2006). While some authors attempt to explain this “distance-puzzle” as resulting from misspecification (Mr. David T. Coe et al., 2002; Yotov, 2012) others advance elements such as increasing regional integration, resulting in a higher increase of trade between geographically closer countries, or the proportionally higher decrease of so-called « dwell » (distance independent) costs than that of distance dependent costs, leading to a larger importance of the geographical distance in the total cost of trade (Carrère and Schiff, 2005).
Most of the literature on this is limited to international trade. Nevertheless, some patterns can be transferred to domestic trade as well. Anderson and van Wincoop show that “both international trade costs and local distribution costs are very large and together dominate the marginal cost of production” (Anderson and van Wincoop, 2004). And Baldwin and Lopez- Gonzalez, based on WIOD input-output table analysis highlight the fact that for many countries intermediate products are still sourced mainly from intranational sources: “imported intermediates share of total world manufacturing is only 16 per cent; for the production of all goods and services, it is just 8 per cent. In other words, the world is not flat and distance has not died – especially when it comes to international production networks. At the level of aggregation available today, most nations are largely self-sufficient in terms of intermediate inputs” (Baldwin and Lopez-Gonzalez, 2015).
An interesting issue that illustrates the ambiguous knowledge about the role distance plays is the debate about the so-called “home bias”
2. McCallum launched the debate with his estimation in 1995 that inter-provincial trade in Canada was supposedly much higher than expected when compared to equal-distance trade with US-American states (McCallum, 1995). This provoked many similar studies that all reproduced some sort of border effect, even though most of the time this effect was not as high as that estimated by McCallum. This effect was interpreted as showing trade barriers linked to the presence of a border. Recently studies, however, have put this into question. Whereas, again, some have identified specification errors (Anderson and van Wincoop, 2003; Hillberry and Hummels, 2003), others have focused on the measurement of distance showing that more precise measurement of distance reduces the border effect (Head and Mayer, 2009; Hillberry and Hummels, 2003).
Another line of research identified border effects even within countries, i.e. between intra- country divisions such as US States (Wolf, 2000). This obviously raised the question of the origins of such border effects in a situation where these borders are generally not considered as trade barriers, not even informal ones. Hillberry and Hummels provide an interesting take in that debate: Using very precisely located individual trade flow data, they find that
“‘borders’ between 5-digit zip codes represent a sizable barrier to trade. We consider these zip-code effects the reductio ad absurdum of the home bias literature. While one can imagine many barriers to trade that operate at national borders, it is harder to conceive of what barriers plausibly operate at state borders, and harder still to imagine those associated with 5- digit zip codes. Our results suggest that ‘‘home bias,’’ at least in state borders, is an artifact of geographic aggregation. Since shipments drop off extraordinarily rapidly over very short distances, attempts to measure border effects on larger geographical groupings are nearly certain to ascribe the non-linear effects of distance to ‘‘home bias’’ dummy variables
2 This section on the debate on the home bias comes from the first version of paper 7. In the revised version of this paper presented below, this section is much shorter, so I reproduce the longer version, here.