The Cost and Gains of Remoteness:
European Integration and its Impact on
Dutch Border Municipalities
IE&B Master Thesis
Feb 2011
Author: R. Handgraaf (s1483781)
Supervisors: Prof. H. Garretsen, Prof. S. Brakman
Abstract
NEG theory provides a model that explains the geographical distribution of economic activity. A key implication is the different impact that trade cost changes have on border regions relative to non-border regions. This thesis tests this implication by focusing on Dutch municipalities and how they are affected by the impact of the establishment of the European Single Market Programme and the introduction of a European common currency. By applying an empirical strategy drawn from Redding and Sturm (2008), results provide an indication that border regions are differently affected by the Single Market Programme relative non-border regions, most notably those along the German border. In addition the impact of the Single Market Programme is found to be stronger for small municipalities than is the case for large municipalities. However, border regions maintain to have a disadvantage over centrally located municipalities.
- 2 -
Table of Contents
1. Introduction
3
2. Literature review
5
2.1. NEG Theory
5
2.2. Impact of Shocks
6
2.3. Redding and Sturm (2008)
8
2.4. Theoretical Model
10
2.5. EU Integration
12
2.5.1 Single Market Programme
2.5.2 European Monetary Union
2.6. Hypotheses
13
2.6.1 Hypothesis limitation
2.6.2 Labour Mobility
3. Research Strategy
15
4. Data Description
18
5. Empirics
20
5.1. Estimation Results – SMP
20
5.2. Estimation Results – EMU
23
5.3. Small vs. Large regions
26
- 3 -
1. Introduction
Economic activity is unevenly distributed across the world, continents and countries. Cities are clusters of bustling activity where ideas and products are developed and made, in contrast to rural areas. Economists have extensively developed models to explain this phenomenon. New Economic Geography Theory provides a model that explains the unequal distribution of economic activity based on economic factors1. Characteristic to the model are the agglomeration and dispersion forces that arise from a combination of increasing returns to scale, factor mobility and trade costs. It is highly valued for its relative simplicity and the endogeneity of demand, income and location in the model. An implication of the New Economic Geography Theory, henceforth named NEG, is the existence of multiple equilibria of spatial economic distribution and their sensitivity to external shocks; cities can arise and cities can fall apart. Recent research has focused on empirical testing of the model to proof its relevance in the field of economics. Among those publications is a stream of literature that focuses on the effects of external shocks on the spatial distribution of economic activity and whether they can lead to the establishment of a new equilibrium.
This paper builds on the study by Redding and Sturm (2008), which explores the effect of the post-war German division in to East and West Germany, and the resulting effects on West German cities. The authors find that the loss in market access results in a decline in population for especially border cities relative to other cities in the country. The shock, however, did not lead to a permanent new equilibrium, the reunification of both Germanys lead to a stream of labour moving back to the now again centrally located regions. The empirical research strategy is attractive due to its simple, yet, effective approach and requires only a small range of variables.
I use the same empirical specification in this study to test whether in the Netherlands border municipalities are differently affected by a dramatic decrease in trade cost than non-border municipalities. In my case, the dramatic cross border trade cost reductions stem from the formal completion of the Single Market Program in 1992 and the establishment of a common currency area in Europe in 1999. I hypothesize that border regions, that are more dependent on cross border trade compared to central regions, are differently affected by these events. Moreover, due to a relative higher dependence on neighboring regions, small regions will experience a larger impact by changes in market access than large regions. The expected effects of the shock are, however, more pronounced for the European Single Market than the European Monetary Union.
This paper adds to the existing literature a NEG analysis of the effects of economic spatial integration on the economic distribution in the Netherlands. The study is unique in its spatial focus on the municipal level with respect to the NEG studies that commonly focus on larger regions. Moreover, I draw from a self-constructed panel dataset containing the annual growth rates of 153 Dutch municipalities over a 48 year time period.
1
- 4 - This study finds that border regions have a systemic disadvantage over non-border regions because of their remoteness. However, there is evidence that the effect of European Integration affects border regions differently than non-border regions. However, these effects can be positive and negative, depending on the overall effect of the increased market access effect and the increased competition effect. In addition, the effects of the two shocks differ between the regions along the German border and the Belgium border. Moreover, the data suggests that trade cost reductions affect small cities more than large cities, in agreement with the hypothesis. In addition, I find that the border effect is more pronounced for regions located near the border than those actually at the border. Finally, by focusing on the agglomerations in the Netherlands I find that they have experienced significant higher population growth rates relative to the non-agglomerated regions as a result of both the European Single Market and the European Monetary Union.
- 5 -
2. Literature Review
2.1 NEG Theory
Central to the study by Redding and Sturm (2008) is a New Economic Geography model: a theory that aims to explain the uneven distribution of economic activity among regions by focusing on the role of increasing returns to scale, transport costs and factor mobility. These factors combined lead to agglomeration and dispersion forces that together determine the spatial distribution of economic activity. The influential paper Krugman (1991) introduced the interactions that arise when you combine the above mentioned factors. This model has been further developed theoretically in the years following the initial publication (Krugman and Venables 1995; Venables, 1996; Puga, 1999;). Furthermore, the theory has matured, evidenced by several comprehensive handbooks regarding NE (Fujita et al. 1999; Brakman et al. 2009). Finally, the NEG theory has been acknowledged its relevance in the field of economics when Paul Krugman won the 2008 Nobel Prize for his publications on economic geography and international trade models.
The assumptions of consumers’ love for variety, increasing returns to scale, transport costs and mobile production factors play a crucial role agents’ location decision (Redding, 2010), yet the NEG characteristic dynamics stem from the mobile labour assumption. Krugman and Venables (1995) find that instead of assuming a mobile labour force, an intermediate product market can be used as the source of agglomeration. Briefly, the basic model, as introduced in Krugman (1991), takes a space with a fixed and given distribution of consumers who supply labour. Businesses have a high incentive to concentrate close to the consumers to minimize transportation costs and benefit from returns to scale that will arise from serving larger markets. This force of agglomeration is termed the ‘home market effect’, one of the two agglomeration forces. The other force, named ‘the price index effect’, stems from the decreased prices that result from the increased competition among firms and attracts more consumers/labour to that region. On the other hand, an immobile sector such as agriculture lead to dispersion as high local demand increases its local wage/price. Helpman (1998) used the housing market as an immobile sector that leads to the dispersion of economic activity. As labour moves in, demand for housing increases, this deters further inflows of labour. The interaction of these forces leads to an uneven distribution in which a core and periphery pattern can emerge, as well as complete or no agglomeration; NEG theory is characterized by multiple equilibria (Redding, 2010).
- 6 - The most recent review of empirical NEG research advancements is Redding (2010). This paper also links the current progress to related literature in the fields of regional and urban economics. Moreover, it calls for a better interaction with the regional and urban economist to arrive at a better understanding of agglomeration forces. This cooperation between the two disciplines may lead to an “empirically-verified-all-encompassing theory of geographic distribution of economic activity at multiple levels of spatial aggregation”, which illustrates the high ambitions in the field of NEG.
A current trend in empirical research is the use of natural experiments. Contrary to natural sciences a controlled experiment is nearly impossible, especially for large macroeconomic phenomena; one cannot force a sudden change on a country for scientific purposes. Moreover, the complex interactions between the many parameters make it difficult to extract a meaningful causal relationship between two parameters. Researchers, therefore, often make use of natural experiments to find evidence of their hypothesized relationships between parameters. By focussing on one event in which an exogenous shock affects one parameter it is possible to single out its effect on another parameter in order to draw any conclusions regarding their relationship. Because of the exogenous nature of the shock, other parameters play no role in the shock and therefore do not distort the analysis.
The section below presents a selection of recent advances in empirical testing of NEG properties that used natural experiments as an approach in their aim to gain a better understanding of the NEG theory. The reader will find that each study touches upon one or more of the NEG implications set forward by Head and Meyer (2004).
2.2. Impacts of shocks
The interplay of agglomeration and dispersion forces can lead to either agglomeration or to more equal distributions of economic activity over space. These equilibria are either stable or unstable; a change in a parameter could lead either to no change, or to dramatic changes. As an effect, agglomerations might rise, fall apart and concentrate elsewhere, or economic activity might disperse. For a change of equilibrium to occur, a certain threshold or tipping point has to be passed to get from the initial to the new equilibrium. This requires a shock strong enough in magnitude to have an everlasting effect. A dramatic fall in transport costs, for example, might lead to a new consideration of businesses about their location preferences. They might for instance decide that at such levels of transport costs, it is more profitable to locate away from the city to benefit from lower wage rates and, nevertheless, willing to incur the transportation costs. The relocation of one firm might lead to further changes of parameters in the agglomeration, which leads to a further decrease in the number of firms. This process enhances itself and leads to the agglomeration to fall apart. However, a small change in transport costs might not affect the choice of businesses after all, therefore leaving the equilibrium unaffected. Shocks therefore can have a large impact on the spatial distribution of economic activity.
- 7 - has changed city-growth paths and potentially new equilibria. They found that despite the ‘spectacular destruction’, the relative city population size returned to the pre-war standards, suggesting that the external shock had only short term effects and no long-run consequences. Similar results were found by Miguel and Roland (2010). They investigated the effects of US bombing of Vietnam cities during the Vietnam War on later economic development. Their data shows that the massive destruction had no long-lasting impact on local economic infrastructure and development levels.
These research however do not confirm or reject the existence of multiple equilibria, it rather explores the stability of an initial equilibrium. Therefore, follow up papers Davis and Weinstein (2008) and Bosker, et al. (2008) extend the research by determining whether multiple equilibria are part of real economies, again using data on the allied bombing of Japan and Germany respectively. However, the former reports no evidence of the existence of new equilibria, the latter presents mixed results. However, the four papers together argue that location fundamentals –local geography- and economic interplays, similar to those put to use in Krugman (1991) –e.g. increasing returns and monopolistic competition- have an important role to play in city formation.
Redding, Sturm and Wolf (2010) uses another event related to the Second World War, namely the division and reunification of Germany in to East- and West-Germany. This division is an exogenous shock that significantly changed the relative geographical locations of German cities significantly. Originally central regions found themselves cut-off from nearby regions by the new drawn border and became located at the periphery of the larger region, either West- or East-Germany. Redding, Sturm and Wolf (2010) exploit this division to observe how this event changed the industry location, specifically that of West-German airports. They find that after the division of Germany the international airport of Berlin lost its dominant position to Frankfurt airport, due to its remote location from other economic centres in West-Germany. Frankfurt became, as a result of the division, centrally located and therefore a preferable destination for air traffic. This change proofed to be permanent as after the reunification Frankfurt maintained its leading position. This observation suggests that due to the division of Germany a new sustainable equilibrium was found.
Ahlfeldt, Redding, Sturm and Wolf (2010), a work in progress, takes the division of Germany a step further by zooming in on the division and reunification of Berlin after the Second World War and investigate how this affected the city’s economic structure and the dynamics of the agglomeration and dispersion forces. They find that the infamous Berlin Wall that divided the city in two separate parts had a significant effect on the city’s economic structures. The border that ran through the pre-war city centre left West Berlin without access to the commercial centre. This resulted in the development of a centrally located commercial centre in the new city West Berlin.
- 8 - This shows that in contrast to Redding, Sturm and Wolf (2010), the equilibrium is unstable and the division did not lead to a permanent shift to the temporary centre in West Berlin.
Other examples of external shocks are those of policy changes, e.g. trade liberalization. Hanson (1996, 1997) uses the US-Mexico trade liberalization as a shock that lead to the redistribution of the economic activity from the centre regions towards the US-border regions. Overman and Winters (2006) use the accession of the United Kingdom to the European Economic Community as a shock that would alter the economic distribution in the UK. Accession would lead to substantial increases in market access of the port-cities in the UK, but would increase the cross-border competition with other regions as well, meaning that some port cities will benefit more than others. The authors investigate the composition of manufacturing imports and exports at seaports and see if they changed with the new membership to the EU. In the data they find evidence that indeed the increase in market access has increased economic activity at ports with a relative better export market access, and economic activity decreased at ports that had relative bad export market access and suffered from the increase in import competition. Although the results differ across sectors, the authors find that overall that the changes in the spatial distribution of UK manufacturing follow the predictions of NEG theory.
The use of trade liberalization as external shock is, however, ‘potentially problematic’ (Redding and Sturm, 2008). First the effect of trade liberalization on the market access of regions is small and usually takes place over a period of time, rather than a sudden shock. Secondly, trade policy might not be an exogenous source of change, as it is often drafted with the economic characteristics of the regions in mind.2
Central to my paper is the publication by Redding and Sturm (2008) who use the division of post-war Germany in to East- and West-Germany as an external shock, which they assume will affect the distribution of economic activity across cities in West-Germany. This natural experiment makes an interesting case as the shock is large in magnitude, exogenous in nature, does not affect institutional quality, keeps natural endowments unchanged and the reunification allows for testing the reversed causality. Moreover, the empirical strategy they designed is firmly grounded in the NEG theory and can easily be employed in further studies. In the following section I will provide a more elaborate discussion of this paper and introduce their theoretical model on which my empirical research is based.
2.3 Redding and Sturm (2008)
As mentioned before, the division after the Second World War and reunification in 1990 of Germany into East- and West-Germany provides an interesting case to explore the existence of multiple equilibria and the shock sensitivity of the distribution of economic activity. Redding and Sturm (2008) carefully analyzes the effect of the division on the spatial distribution of economic activity across cities in West-Germany. The intuition is similar to Redding, Sturm and Wolf (2010): the division leads to a dramatic change in the strategic location of cities that previously were centrally located and became border cities. These new border cities experienced a sudden decline in trading partners as a large group of them became unreachable as a result of the new border. On the other hand, cities more distant from the new border are less affected by the change due to this distance. This distinction creates two groups of cities
2
- 9 - in West-Germany, namely border cities and non-border cities. The authors argue that the loss in trading partners of border cities leads to a new spatial distribution of economic activity as labour and businesses will reconsider their optimal location given the new circumstances.
Based on their model, derived from Helpman (1998), the authors develop a research strategy in which a mobile labour force plays a crucial role. Labour will allocate itself across cities searching for the highest return to their labour, considering local price level, until at equilibrium real wages are equalized for all cities and the need for labour to change location is eliminated. In such a model the population size can be used as a proxy for economic activity as people live where they work. In light of the NEG theory this means that after the division of Germany, the West-German border cities are expected to experience a decline in population size relative to the non-border cities in West-Germany as a result of their geographical disadvantage of being a border city.
However, the authors primarily focus on the effect of a decrease in market access of the regions. From the theoretical framework a second effect arises with the decrease in market access, namely the price competition effect. With the decrease access to the markets comes the decrease in competition among the regions. This decrease in competition will positively affect regions that previously had to compete with other markets. With the decrease in competition their industries could profit more from the absence of their rivals. This could attract labour to the region after the event. This effect is overlooked in the Redding and Sturm (2008) paper.
The authors constructed a database that contains city-level population size and other local parameters for a broad time range. From this data the authors find that there is a disproportional population growth difference between border and non-border regions. They observe that the border regions experience significant lower population growth rates as a result of the division of Germany. Moreover, the data indicates that the reversed is true for the reunification. The argument, however, is not strong as it implies that at time of the reunification the economy had reached its post-division equilibrium, which the authors find plausible. These results suggest that such an external shock can lead to a new distribution of economic activity and a new equilibrium can be reached. However, contrary to the findings of Redding, Sturm and Wolf (2010) the equilibrium is unstable since it does not hold when the shock is undone, as is the case with the reunification of Germany in 1990.
In addition, Redding and Sturm (2008) find that the negative population growth rates are more pronounced in small cities at the West-German border regions compared to the larger cities along that border. Larger cities have their own market size to support its consumers to a large extend and are therefore able to specialize in exporting their products, whereas small cities are less able to do so and depend more on the cities in their proximity.
- 10 -
2.4 Theoretical Model
In this section I will briefly discuss the model on which this thesis builds and is similar to the one used in Redding and Sturm (2008). The model is an adaption of the multiregional version constructed by Helpman (1998), which in turn incorporates parts of the Krugman papers (Krugman, 1991). It incorporates the NEG characteristics introduced above, such as the agglomeration and dispersion forces that determine the distribution of economic activity/population across regions. For a complete discussion of the model I refer to Helpman (1998) and Redding and Sturm (2008), but I will briefly mention the key points below.
The model takes a fixed number of regions that are each endowed with their own fixed stock of a non-tradable good housing. A population of L mobile consumers is distributed among these regions. The income of consumers is spent with a fixed share μ on tradable goods and 1- μ on housing. The consumption of tradable goods consists of purchases of a wide range of differentiated varieties. Firms pay fixed costs and wages to produce these product varieties using only labour as production factor. And in order to supply other regions firms incur trade costs.
A crucial assumption in this model is that of mobile labour. In the initial models labour was assumed to be a fixed production factor which leads to differences in wages across regions. Firms would locate near large markets and thus local demand for labour would increase. Labour is immobile and cannot move in order to serve the increased demand, consequently local wages increase and at equilibrium wages would differ across regions. Assuming a mobile labour force ensures that labour would locate at regions that offer the highest wage adjusted for local price levels. This inflow of labour would lower wages and at equilibrium the real wages equalize. Equation (1) provides the formal formulation in which the labour mobility ensures equal real wages for all regions:
𝜔𝑖 ≡
𝑤𝑖
𝑃𝑖𝑀 𝜇 𝑃𝑖𝐻 1−𝜇 = 𝜔 for each 𝑖 region. (1) 𝜔 is the real wage; w is the nominal wage; PM is the price index for tradable goods and services; PH is the price index for non-tradable goods and services; and 𝜇 is the share of income spend on tradable goods and 1- 𝜇 the share on non-tradable goods.
Instead of wages, one can look at the population sizes of a region, which will differ across regions when the labour mobility condition is in place. Moreover, the attractiveness of a region is not only determined by its price level and wage, but also by its proximity to other markets, hence the inclusion of market access is necessary. By replacing the price levels and wages in the previous equation by Market Access variables and the population size of a city, one can capture the same intuition in the following equation:
𝐿𝑖 = χ FMAi μ σ 1−μ 𝐶𝑀𝐴 𝑖 𝜇 1−𝜇 𝜎 −1 𝐻 𝑖, (2) where 𝐹𝑀𝐴𝑖 ≡ 𝑤𝑗𝐿𝑗 𝑃𝑗𝑀 𝜎−1 𝑇𝑖𝑗 1−𝜎 𝑗 , (3) 𝐶𝑀𝐴𝑖 ≡ 𝑛𝑗 𝑝𝑗𝑇𝑗𝑖 1−𝜎 𝑗 (4)
- 11 - good, and σ the constant elasticity of substitution between varieties.3 Equation (2) shows how the equilibrium population size of a region is linked to real wage𝜔, two types of market access and the available fixed stock of non-tradable good H. The formula differentiates between two types of market access, namely firms market access (FMA) and consumer market access (CMA), equation (3) and (4) respectively. FMA, explains the highest nominal wage a firm can offer and is determined by its trade costs with other markets, their economic size (measured in terms of population size) and price levels. CMA, equation (3), summarizes the access to tradable goods from other regions by consumers. Their consumption of 𝑛 varieties from other markets depends on the trade costs and their cost-price, which depend on the local prices (𝑝). Trade costs (T) is defined as a function of the distance between the two regions. This rather simplistic approach tells us that the strength of impacts of other regions is decreasing in distance4. However, the trade costs should not be confused with transportation costs, which are part of trade costs. Trade costs include other costs as well, administrative, regulatory and many more. These are however assumed to be increasing with distance. (T) is therefore of negative influence on the market access.
Although in the long run real wages are the same across regions, the distribution of labour and their economic output and consumption is unequally distributed across the regions. The agglomeration forces are traceable in the equation; FMA includes the home market effect in which large markets, or those near large markets attracts labour inflows; CMA incorporates the price index effect, the downward pressure on prices in large markets or those close to large markets. Both agglomeration forces impact the region’s population size positively. The fixed stock of non-tradable, however, reduces the incentive to locate near large regions. Due to its fixed and immobile nature, the price of this good increase as demand for it increases. As part of the aggregate price level, this pushes the price level up.
Using equation (2) we can see the implications that change in market access has on the population size of regions. On the one hand, a reduction in the trade costs will increase the market access of a region and attract more economic activity, which is represented by an increase in population size. On the other hand an increase in market access due to reductions in trade costs can lead to an increase in competition. Regions that are formerly protected by their remote location will have to compete with other regions as their consumers and producers gain access to new markets with other products and prices. Regions that lose the competition will experience an outflow of economic activity to the other regions. These two effects lie at the heart of this thesis and it is its goal to find evidence that the mechanisms discussed here are observed in the real world.
3
The model takes the Dixit-Stiglitz approach in differentiating between varieties, which implies a constant elasticity of substitution between varieties: σ > 1.
4
- 12 -
2.5 EU integration
2.5.1 Single Market Programme
The creation of the European Single Market in 1992 is the outcome of a long process of economic integration. Since the end of the Second World War Europe is characterized by economic and political cooperation to reach economic growth and political stability. Through the accession of more member states5, the European Union has become the largest market in the world (2009 data) with a population size of over 500 million and an average GDP per capita of almost 30.000 dollar (IMF).
The European Single Market Programme (SMP) has lead to dramatic decreases in trade barriers through the free mobility of goods, labour and capital. Next to free trade, SMP lead to better integrated markets, increased innovation and competition, all in favour of the European consumer. It is understood that such trade reductions had dramatic effect on the economic infrastructure and spatial distribution of economic activity.
Redding and Sturm (2008) argue that trade liberalization is unlikely to be a complete independent exogenous shock, as policy is likely to be drafted related to local economic issues. Trade liberalization is more of a political process than an economic one, and the European Union is, like any centre of power, suspected to be under some influence of lobby groups. However, the level of analysis in this thesis is at the municipal level, whereas the drafting of the SMP is an international event. It is very unlikely that Dutch municipalities have been able to exert any influence on the Single Market Programme. It is possible that at a national level some policies are drafted in the interest of the Netherlands, this will, however, affect all municipalities in the Netherlands and is unlikely to have different impact on border regions versus non-border regions.
A second point of critique raised by Redding and Sturm (2008) is the fact that trade liberalization is a process rather than a shock. The SMP, for example, was gradually implemented in the years before the deadline of 1992. However, the consequence with respect to this thesis is the reduced impact of the ‘shock’ on the population size of the municipalities. It is possible that labour, anticipating the change, already decided to relocate. By focussing on the year 1992 as the break-year, I will not catch the labour movements prior to the event. This notion makes it less likely that I will find evidence that there has been a change in population sizes of border regions relative to non-border regions after the shock. On the other hand, positive significant proof of the hypothesis will therefore be a strong supportive finding.
2.5.2 European Monetary Union
The latest step in economic integration of European states has been the introduction of the Euro as a common currency and an active independent European Central Bank in control. A single currency would improve the ease of transactions, reduced currency risks, increase market transparency and through a credible stable price policy reduce uncertainty. The euro was introduced in 1999 and coins and notes went in to circulation in January 2002. Next to an independent central bank, the Maastricht Treaty of 1992 formulated criteria governments have to fulfil regarding their, among others, sovereign debt, budget deficits and inflation policy.
5
- 13 - Although the euro is to considered to have had large consequences on the business environment in Europe and the world, its effect on trade patterns between the Netherlands, Belgium and Germany are limited. The introduction of the euro effectively lead to a fixed exchange rate regime between the member states that eliminated any currency risks. The Netherlands and Germany had already been engaged in a rigid exchange rate regime for many years. The same holds for the exchange rate between the Dutch Guilder and the Belgium Franc. These currencies have hardly changed in the years preceding the euro. The only significant effect of the common currency was the reduction in transaction costs, as there was no need to exchange currencies. Therefore it is unlikely that the establishment of the Euro currency area had any effect on the market access of regions. Moreover, the distance to other countries with which the Netherlands had a floating exchange rate regime is too big to have any discernible effects on the market access of border regions. Therefore it is doubtful to find a difference in population growth rates of border regions compared to non-border regions after the event.
2.6 Hypotheses
From the previous sections we have learned that an external shock that increases market access of a region can have a dramatic impact on the distribution of economic activity. The impact of such shock, however, reduces over distance. Therefore, I argue that the effects of a shock that reduces cross border trade barriers is stronger for regions close to the border relative to non-border regions due to their geographical proximity to the shock. In this paper I will test this notion by using the formal establishment of the European Single Market in 1992 and the introduction of the European Monetary Union in 1999 as external shocks that reduced the trade costs between countries significantly. However, given the discussion in the previous section on the European Monetary Union and the Single Market Programme, it is not surprising that I expect each event to impact the Dutch regions with a different magnitude. These expectations lead to the following testable hypotheses:
Hypothesis 1a: The reduction in trade costs across national borders due to the establishment of the European Single Market increases market access between regions and will affect border regions stronger than non-border regions.
Hypothesis 1b: The introduction of the Monetary Union did not have a different impact on border regions compared to non-border regions.
Second, as mentioned in the discussion of the Redding and Sturm (2008) paper, the impact of an external shock on a city or region could be vary according to the economic size of a city or region. The argument is that small cities are more dependent on other markets than their own. Large cities have their own markets to support their income and wealth and therefore shocks will tend to have less of an impact on their economic fundamentals. This theoretical idea can be formalized in the following testable hypothesis:
Hypothesis 2: An external shock that increases a regions market access will have a larger effect on economically small regions relative to economically large regions.
- 14 - trade costs between the Netherlands and its two neighbours significantly. In the next sections I will set out the empirical research strategy, introduce the data used and present the testing results with respect to the hypotheses. First, however, there are limitations with respect to the hypothesis that need to be discussed.
2.6.1 Hypothesis Limitations
The formulation of hypothesis 1a poses a problem for empirical testing. As discussed earlier, the effect of a reduction in trade costs has two effects, on the one hand the increase in market access that attracts labour, and on the other hand the increase in competition that could lead to an outflow of labour. An observed outflow of labour suggests the dominance of the price competition effect, and an inflow of labour suggests the dominance of the increased market access. These two effects combined could also lead, however, to an aggregate outcome in which there is no apparent change in population size. The possible finding that there is no change in population sizes after the ‘shock’ does therefore not reject the hypothesis: it could mean both effects eliminate each other. This leads to a problem in analyzing the data and interpreting the results.
2.6.2 Labour mobility
As discussed in the previous section, labour mobility ensures that real wages are similar across all regions and is the crucial factor that ensures equilibrium in the spatial distribution of economic activity in the NEG model I apply. However, one could question the validity of this assumption. Within the Netherlands an estimated 10 percent of the population moves to another location, of which only a third moves away from the region they lived (Feijten and Visser, 2005). This group of cross-region movers consists mostly of students or graduates; once a job is found, households are very reluctant to move. The rather immobility of labour in the Netherlands gives another possible bias in the data towards a rejection of hypothesis 1a and 2. If labour is indeed not mobile, then it is unlikely to find a dramatic flow of labour towards or from border regions despite increases in local demand for labour.
- 15 -
3. Research Strategy
The aim of this paper is to find proof of the suggested effect of an increase in market access on the attractiveness of a region. And if so, whether the economic size of a region is relevant to the impact of an increase of market access on that region. These two hypotheses will be the prime focus of the following sections. In this section I will introduce the research strategy and the empirical testing specifications I will employ to find supportive evidence.
I test these two hypotheses by using a simple so-called difference-in-differences methodology following the Redding and Sturm paper (2008). This methodology measures the difference between a test group and control group before and after the effect of a treatment or shock. In this case I compare the population growth rates of Dutch border municipalities (treatment group) with the population growth rates of Dutch non-border municipalities (control group). By comparing the growth rates before and after the event at 1992 in which the European Single Market formally was completed the hypothesis can be accepted or rejected. The empirical specification I will use reads as follows6:
𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑏𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝑏𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (5) In which citypop represents the annual population growth of region i at time t, Border is a dummy variable that takes the value of 1 if region i represents the treatment group of border regions and otherwise 0, SMP is a dummy variable that takes the value of 1 if at time t the Single Market Programme has been introduced in 1993 and otherwise the value of 0. 𝛿 are time dummy variables which control for the growth trends of that particular year. Finally, e is the error term.
The specification is not suspect to unobserved region fixed effects, for they are differenced out when I use population growth rates. The first coefficient, 𝛽1, captures the systemic difference in population growth rates between the border and non-border regions. The coefficient of interest to the thesis is 𝛽2, which captures the effect of the formal completion of the European Single Market on the relative population growth rates of border and non-border regions. A significant positive coefficient would support the hypothesis that border regions in the Netherlands after the completion of the European Single Market did experience an increase in population size relative to non-border regions.
In addition to the European Single Market, I will also use the introduction of the Euro in 1999 as an event that has reduced trade costs significantly. The intuition is similar to the previous case, as well as the empirical method. Testing the effect of the introduction of the EMU is identical to the baseline regression. Simply by replacing variable SMP in equation (5) by EMU, which is defined as a dummy variable that takes the value 1 if at time t the EMU was formally in place in 1999, and 0 if not:
𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑏𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝑏𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (6)
The third specification distinguishes between the Dutch-German border and the Dutch-Belgium border. There are reasons to believe that the trade costs between the Netherlands and Belgium are different
6
- 16 - from those between the Netherlands and Germany. First there is the Benelux, a pre-European customs and economic union between Belgium, the Netherlands and Luxemburg. As early as in 1960 the first agreements were made regarding the cross border mobility of labour, products and capital, economic integration and international trade coordination. Moreover, the common language along the Dutch-Belgium border reduces trade costs as communication barriers are relatively low. Finally, cultural differences are less pronounced between the Netherlands and Belgium compared to the Germans. These factors validate the possible difference in impact of European integration on the regions along the Belgium and German border. Formally:
𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑏𝑒𝑙𝑏𝑜𝑟𝑖+ 𝛽2 𝑏𝑒𝑙𝑏𝑜𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛽3𝑔𝑒𝑟𝑏𝑜𝑟𝑖+ 𝛽4 𝑔𝑒𝑟𝑏𝑜𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (7) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑏𝑒𝑙𝑏𝑜𝑟𝑖+ 𝛽2 𝑏𝑒𝑙𝑏𝑜𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛽3𝑔𝑒𝑟𝑏𝑜𝑟𝑖+ 𝛽4 𝑔𝑒𝑟𝑏𝑜𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (8)
Furthermore, the fourth specification focuses only on the border regions of the Limburg province. This province is geographically located near large foreign markets and is expected to be most influenced by the European integration process. Limburg is in the close proximity of the Ruhr-gebiet, the largest industrial agglomeration in Germany, as well as close to Brussels a large economic centre in Belgium. Moreover, Limburg is at a relative distant location from the Dutch agglomeration named Randstad in the west of the country, therefore more dependent on the cross-border markets. Therefore I will perform the specification as defined in (5) using only a dataset comprising of Limburg regions as border regions relative to the non-border set of Dutch regions:
𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑙𝑖𝑚𝑏𝑜𝑟𝑖+ 𝛽2 𝑙𝑖𝑚𝑏𝑜𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (9) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑙𝑖𝑚𝑏𝑜𝑟𝑖+ 𝛽2 𝑙𝑖𝑚𝑏𝑜𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (10)
Fifth, the Limburg regions are located at the German border and/or the Belgium border; therefore I will differentiate again between these two borders to find a possible divergence in population growth rates between the two borders. I will test these using the following specification:
𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑙𝑖𝑚𝑏𝑒𝑙𝑖+ 𝛽2 𝑙𝑖𝑚𝑏𝑒𝑙𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛽3𝑙𝑖𝑚𝑔𝑒𝑟𝑖+ 𝛽4 𝑙𝑖𝑚𝑔𝑒𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (11) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑙𝑖𝑚𝑏𝑒𝑙𝑖+ 𝛽2 𝑙𝑖𝑚𝑏𝑒𝑙𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛽3𝑙𝑖𝑚𝑔𝑒𝑟𝑖+ 𝛽4 𝑙𝑖𝑚𝑔𝑒𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (12)
- 17 - dependent variable which becomes the change in population share against the total population (cityprop). The specification reads as follows:
𝑐𝑖𝑡𝑦𝑝𝑟𝑜𝑝𝑖𝑡 = 𝛽1𝑏𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝑏𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (13) 𝑐𝑖𝑡𝑦𝑝𝑟𝑜𝑝𝑖𝑡 = 𝛽1𝑏𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝑏𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (14)
Furthermore, the seventh specification formulates the test to examine the second hypotheses regarding the different impact of the shocks on small and large cities. To test this I will have to run a test for each sub group. By comparing the 𝛽2 coefficients I will able to draw conclusions. The specifications are defined in the following expressions:
𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝐿𝐵𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝐿𝐵𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (15) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝐿𝐵𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝐿𝐵𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (16) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑆𝐵𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝑆𝐵𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (17) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑆𝐵𝑜𝑟𝑑𝑒𝑟𝑖+ 𝛽2 𝑆𝐵𝑜𝑟𝑑𝑒𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (18)
In addition, the effect of the shock on the different city sizes may differ across border type. Therefore, similar to previous specifications, I will differentiate between the German border regions and the Belgium border regions. The specifications read as follows:
𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝐿𝐵𝑒𝑙𝑖+ 𝛽2 𝐿𝐵𝑒𝑙𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛽3𝐿𝐺𝑒𝑟𝑖+ 𝛽4 𝐿𝐺𝑒𝑟𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (19) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝐿𝐵𝑒𝑙𝑖+ 𝛽2 𝐿𝐵𝑒𝑙𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛽3𝐿𝐺𝑒𝑟𝑖+ 𝛽4 𝐿𝐺𝑒𝑟𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (20) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑆𝐵𝑒𝑙𝑖+ 𝛽2 𝑆𝐵𝑒𝑙𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛽3𝑆𝐺𝑒𝑟𝑖+ 𝛽4 𝑆𝐺𝐸𝑅𝑖∗ 𝑆𝑀𝑃𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (21) 𝑐𝑖𝑡𝑦𝑝𝑜𝑝𝑖𝑡 = 𝛽1𝑆𝐵𝑒𝑙𝑖+ 𝛽2 𝑆𝐵𝑒𝑙𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛽3𝑆𝐺𝑒𝑟𝑖+ 𝛽4 𝑆𝐺𝐸𝑅𝑖∗ 𝐸𝑀𝑈𝑡 + 𝛿𝑡+ 𝑒𝑖𝑡 (22)
These base specifications will be extended in the final section of the empirical section. The approach will be similar; the main difference will lie in the definition of a border region, and the empirical specification will therefore not differ from the specifications introduced above.
- 18 -
4. Data
The regressions introduced above are tested by using a structured panel data set consisting of population growth rates calculated from a dataset on 153 municipalities in the Netherlands. This data is taken from the Central Bureau of Statistics Netherlands (CBS), which keeps a detailed record of the population sizes of all Dutch municipalities. I obtained annual population size data for the period of 1960 to 2008. These 153 municipalities together cover two thirds of the national population size and have a fair distribution of 51 border and 102 non-border regions7.
There are three factors that affect the population size of a municipality. First there is the birth/mortality rate that leads to an increase or decrease in the population. Second, there is the mobility of labour that determines growth or decline of the population size; migration of households affects population sizes. Thirdly, the restructuring of municipalities has lead to new drawn borders, changing the composition of the population size. Some cities/villages became part of the neighbouring municipality, other municipalities merged completely. The restructuring of municipalities was part of national policy to create more efficient and less costly local government. In total, the number of municipalities was reduced from 1036 in 1960 to 430 in 2010. The restructuring of the municipalities contributes to most large population size variation in the dataset.
The latter cause of variation in population size is purely of an administrative nature and does not reflect labour migration. Therefore, I had to adjust the data by hand, to arrive at a dataset that would reflect the labour movements only. All municipalities in the dataset have been subject to these adjustments. The original dataset includes a list with all the municipal restructuring policies for each municipality and the number of people affected. I adjusted the dataset using the following two rules:
Rule I: A reduction (increase) in population of a municipality as a result of a restructuring policy at a certain time before 1985 will lead to a subtraction (addition) of that amount in the years prior to the event.
Rule II: A reduction (increase) in population of a municipality as a result of a restructuring policy at a certain time after 1985 will lead to an addition (subtraction) of the same amount in the years after the event.
For example, in 1966 Amsterdam received 2953 citizens from the neighbouring municipality Landsmeer as a result of a restructuring of the local borders. Following rule I, my procedure to remove this unwanted variation is by adding that population figure to the Amsterdam population in the years prior to the actual event. Through this method the sudden increase in the population size of Amsterdam at 1966 does not appear in the data; in fact I redefined the border of 1960 to the 1966 situation. In case of a reduction of population I subtracted the amount of people lost from the population size in the years prior to the event. On the other hand, in case of such adjustments at later years, after 1985, I would do the reverse to maintain the actual population sizes as original as possible. For example, in the case of Emmen, that received 8041 citizens from Schoonebeek through a merger in 1998, I subtracted that
7
- 19 - figure from the population in the years after the event. That way the 48 years of observations for Emmen are maintained as original as possible, except for the last ten years. In case of complete mergers without any adjustments with other municipalities I considered the merged municipalities as one for the whole period.
Through this method I arrive at an improved dataset in which changes of population sizes are due to natural growth rates and migration. This procedure of adding and subtracting, however, affects the absolute population numbers in a way that they do not represent the actual population size of the municipality at specific times. This leads to flawed results when I use these new figures to calculate the yearly population growth rates. Moreover, this also affects the municipalities’ population share of the total population, see equation (11) and (12).
To have a long period of observations in my dataset, I have not included some municipalities that were founded in the late 1970’s and grew at rapid rates. Almere, for example, was founded in 1976 in the then newly developed island and province Flevoland, and is not included.
- 20 -
5. Empirics
In total 18 regressions were performed, each focusing on a specific aspect of the base regression as introduced above. The results prove to be ambiguous with respect to the hypothesized values of the coefficients. The output was generated using Eviews, version 7. Table 1 below presents the estimation output with respect to the creation of the Single European Market and the results will be discussed below. The second subsection will discuss the estimation results of the specifications regarding the European Monetary Union. The third subsection presents the results regarding the second hypotheses introduced in section 2. The then following section present additional specifications that provide additional insights regarding the effect of the two key events on the distribution of economic activity in the Netherlands.
5.1 Estimation Results – Single Market Program
The first column in Table 1 presents the base regression as introduced in equation (5). The results show that the border dummy variable has a significant negative coefficient, which suggests that border regions have a systemic half percent point lower growth rate compared to the non-border control regions prior to the creation of the European Single Market. The key coefficient, interaction term β2, is
positive, suggesting that the introduction of the European Single Market did indeed lead to higher population growth rates in border regions. However, this coefficient is not found to be different from zero and therefore does not provide the evidence to support the hypothesis. In addition, the low R2 (15%) tells that the model is not strong in explaining the variation in growth rates among Dutch municipalities. However, it is not the aim of this paper to provide a model that explains population growth; I merely focus on the shock effect of the creation of European Single Market on city populations. On the other hand, a high r-square would imply that the geographical location of a city does affect the population growth rates of Dutch municipalities to a large extent.
The second regression, column (2), differentiates between municipalities located at the German border and at the Belgium border against the non-border regions. The results illustrate the two opposing effects that trade liberalization can have on border regions. First, it is worth noting that being located at either border has an overall significant negative impact on the population growth rate of the municipality, similar to the first case. However, the effect of the European Single Market on population growth of border regions relative to non-border regions differs across the two borders. On the one hand, border regions along the Belgium border have experienced negative growth rates after the creation of the European Single Market relative to non-border regions, as is reflected by the negative interaction term. Β2 tells that Belgium border regions experienced almost 0.2% point lower population
growth rates relative to the control group. Moreover, this negative relationship is found to be significantly different from zero with 90% confidence. On the other hand, border regions along the German border, relative to the control group, did experience significant positive population growth in the years after the SMP was installed. The model estimates with high certainty that these border regions had 0.3% point higher growth rates.
- 21 - that my hypothesis holds true. Moreover, the opposite signs for each border illustrate that somehow the competition effect has been stronger for Belgium borders, and have benefited the regions along the German border. Apparently, border regions along the Belgium border have lost in the price competition against other regions. On the other hand, regions along the border regions have experienced an inflow of labour through the increase in market access and did not lose their competitiveness. This result seems to be in contrast with the previous test in which we found no significant result. However, it seems fairly possibly that the negative and the positive coefficient of the current test singled out each other and lead to an insignificant result when testing for all border regions. Therefore, the finding of an insignificant result does not directly mean that the dynamics described in the NEG model are not observed in real world, it could mean that both forces balance each other out and result in an insignificant change in population size.
The reason, however, why there are such strong differences between the two types of border regions is hard to pinpoint. The historical close ties, large cultural similarities and already high economic integration between the Netherlands and Belgium could partly explain the diverging effects of further economic integration relative to the German border regions where such close ties are absent.
Table 1/Regression Results/SMP
- 22 - Column (3) presents the output of the estimation of regions of the Dutch province of Limburg against the non-Limburg regions. Similar to the base regression, a region in Limburg has lower growth rates compared to the non-Limburg regions. The model estimates that Limburg regions have a 0.34 percent points lower growth rates relative to other regions in the Netherlands. In addition, there is no evidence that the introduction of the European single market has had a different impact on the population growth rates of Limburg’s regions than on other regions, despite its strategic location near cross-border economic centres in Germany and Belgium. This specification therefore does not provide evidence to support the hypothesized relationship, as is evidenced by the insignificant interaction term.
However, similar to the previous tests, there is a potential difference in the effect of the SMP on regions along the Belgium border and the German border. Therefore, in column (4) I differentiate between the two border types. However, the results are similar to the findings in (3), namely no significant effect on the population size of border region relative to non-border regions after the event.
Moreover, in the fifth specification, column (5), I employ a different estimation strategy in which the explanatory variable is changed in the percentage change of the region’s population share of the whole population. This is simply calculated by first dividing the population size of municipality (ί) at year (t) by the national total population size. Second, I calculate the yearly percentage change in the population share. Thus an increase in the population share would mean that the region has had an increase in population size with respect to all other regions in the Netherlands. This way I capture not only the share of populations in the sample, but include those municipalities that are not represented in the sample. A positive border effect would mean that border regions have increased their share of the total population, meaning they had higher population growths relative to non-border regions. The intuition with respect to the hypothesized relationship does not change: A positive significant interaction effect supports the hypothesis.
However, despite the different estimation strategy I find no different results compared to the base regression. Border regions experienced a half percent point decline in their share of the total population, which suggests they lost households against the non-border regions. Moreover, the creation of the European Single Market did not affect the population share of the border regions as the interaction affect is not different from zero.
- 23 - The above results provide mixed support for the hypothesized relationships. Initially I find little evidence that suggest that border regions have benefited relatively more from non-border regions from better access to foreign markets. The distinction between the Belgium and German border, however, shows that regions at different borders are differently affected by the creation of the European Single Market. In the next section I will focus on the second ‘shock’ I investigate, namely the introduction of the Euro and the European Monetary Union in 1999. The approach is identical to the one used previously.
5.2 Estimation Results – European Monetary Union
Table 2 presents the outcome of the introduction of the Euro in 1999 and its effect on the population growth rates of border regions. As outlined in section 2, the introduction of the Euro had far reaching consequences trade patterns within the European Union. Column (6) presents the base regression on border regions against non-border regions. The estimation shows that border regions experienced a significant 0.5 percent point lower population growth rates compared to non-border regions prior to the event. The introduction of the euro, however, did not lead to different growth rates between border and non-border regions. The small negative, yet insignificant, interaction effect shows that the introduction of the Euro, despite its potential impact on economies and businesses, has no impact on the population size of border regions compared to non-border regions.
However, similar as to the SMP case the result can be insignificant due to the two effects that work opposite directions. When I differentiate between the two borders the estimation output, presented in column (7), I do find an interesting result in contrast to the hypothesis. Similar to the second regression (2) in table 1, Belgium and German border regions had lower population growth rates compared to the control group. The introduction of the Euro, however, did have an effect on the Belgium border regions, but left the German border regions unaffected. The German border regions did see a relative increase in their population growth rates after the introduction of the Euro, yet with less certainty than academic standards allow for.
Belgium border regions, on the other hand, experienced a 0.27 percent point lower population growth as a result of the introduction of the Euro compared to the non-border regions. This result is in contrast with expectations. Apparently, the introduction of the Euro has been large enough to create a difference in population growth rates between border regions and non-border regions along the Belgium border. Despite the argument that the Euro did not affect the trade patterns between the Netherlands and Belgium due to the already fixed exchange rate regime, trade patterns are affected. The negative coefficient shows that border regions along the Belgium border experienced a population outflow, which indicates they did not manage to compete with other regions as their market access increased as well.
- 24 - Column (7) presents the output of the estimation in which the Limburg province regions are the treatment group and are tested against the regions in the control group. The estimation shows that the data does not provide any evidence that supports the hypothesized relationship. Limburg regions have had lower population growth rates relative to those regions in the control group: an estimated 0.4 percent point. This divergence in growth rates was not altered by the introduction of the Euro. The estimated effect is found to be not different from zero, and hence provides support for my hypothesis that the impact of the euro is less profound as the establishment of the SMP was. The expectation that the proximity of Limburg regions to German and Belgium markets would provide them an advantage over other regions proved to be wrong.
In column (8) I make the distinction between the two border types for the Limburg regions as I did earlier for the SMP case. In this regression I find no evidence that the euro had a significant different impact on either of the two border types. This finding is consistent with the previous finding of (7), the
Table 2/Estimation Results/EMU
- 25 - Table 3/Summary results/
*,**,*** indicate significance levels of 0.1, 0.05, 0.01 respectively; (0) means coefficient is not different from zero.
introduction of the euro had no impact on either of the two types of border regions in the province of Limburg.
Finally, in the last specification, column (9), I use the proportional share of the total population to measure any relative changes in the population share as a result of the introduction of the euro. From the results I have to conclude that the introduction of the Euro did not lead to a change in the population share of border regions with respect to the non-border regions. Similar to the specification in column (5), the use of a regions change in the share of the national population has not provided any new insights.
Summarizing, table 3 provides a brief overview regarding the estimation results. It shows that there are mixed findings with respect to the hypotheses. First, the introduction of the European Single Market appeared to have had a significant different effect on border regions vis-à-vis non-border regions. This difference appears when I differentiate between the regions along the Belgium and the German border. In addition, the tests regarding the introduction of the euro resulted in the interesting observation that border regions along the Belgium border again lost their economic appeal as they experienced an increase in competition due to trade cost reductions. The other tests, however, indicated no different impact on border regions relative to non-border regions.
Specification
Coefficient
Support Hypotheses
SMP-border 0.094 No, (0)
- 26 -
5.3 Small vs. Large cities
Redding and Sturm (2008) find that the effect of reductions in trade costs affects small border cities more than large border cities. The intuition is that small cities are more dependent on other markets whereas large cities have the support of their own economic mass that could sustain their economic size. This means that trade cost reductions have a proportionate larger impact on small border cities than on large border cities. Hypothesis II I introduced in section 2 formulizes this argument. In this section I will provide evidence that indeed small regions compared to large regions are stronger affected by changes in market access through, in this case, trade cost reductions.
The empirical approach is similar to that employed in the previous sections. I performed the same base regression and tested for both the 1991 SMP event, as well as the 1999 EMU shock. I divided the dataset in two sub-sets, one including all Dutch municipalities below the median population size at 1986, and one with all those above the median. This division obviously leads to smaller subsamples, but also to a different distribution of border and non-border regions. In the sample of small regions the group of border regions is better represented than in the group of large regions. The small region sample consists out of 76 regions of which exactly half is a border region. The large region sample has 77 municipalities of which 13 are along the border. This unequal division will affect the quality of the regressions performed as border regions might not be sufficiently represented in the large region sample. I ran estimations for each subset for each event, plus additional regressions in which I make a distinction between municipalities along the Belgium border and those along the German border. The eight resulting regressions are presented below in table 4.
Table 4/ Estimation results/Small vs. large cities
- 27 - Column (11) and (12) present the base regression for both sub sets in the case of the European Single Market. In both cases the border regions have experienced negative growth rates, as I found in all previous cases. The interaction term, however, proves to be more interesting. Namely, I find that small border cities have indeed experienced significant positive population growth rates with respect to small non-border cities. For large cities this is not the case, and we observe no change in the population growth rates. This result provides strong evidence that small border cities are stronger affected by the SMP trade cost reductions relative to the large border cities, which is in line with hypothesis 2.
Column (13) and (14) provide additional evidence that supports the hypothesis. The distinction between the Belgium and German border regions shows again that the effect of the trade reductions due to the Single European Market affects each group of border regions differently. Similar to previous findings, both groups of small and large border regions experienced prior to the shock lower population growth rates relative to non-border regions. However, the smaller regions tend to have structural lower population growth rates compared to the large regions.
For the subset of small regions I find evidence that along the German border, the small border regions experienced a highly significant increase in population growth rates relative to small non-border regions. This result supports the first hypothesis that border regions are stronger affected by the reduction in trade costs than non-border regions. For the subset of large regions I find for both the German and Belgium border no significant change in population growth rates as a result of the Single European Market. Therefore, I can conclude that the reduction in trade costs that stem from the Single European Market is more pronounced for small regions compared to large regions. This result, however, only holds for those regions along the German border, which is in line with the empirical results in the previous sections. Moreover, the fact that the sample set of large regions only includes 13 border regions, which are also, divided over two border groups, must be taken in to account when one read these results.
These results suggest that the creation of the European Single Market has a larger impact on smaller border regions than on large border regions. The second set of results in Table 4 present the case of further European economic integration, namely the introduction of the Euro and the European Monetary Union. The results, however, provide no evidence that trade cost reductions, which result from a single currency area, have any effect on border regions. Moreover, although the small border regions have had higher positive population growth rates compared to large border regions after the introduction of the EMU, in both sets this interaction term is found to be insignificant.
