Claim for Independence
Determinants of Secessionism in the European Union
Abstract
Why do regions within a country want to secede, which are the factors that drive these proclaims? The regions face a trade-off between different forces. The “centripetal” forces, which are acting for the region to remain in the same country and the “centrifugal” forces, which are acting to secede the region from the country. Only the regions where the centrifugal forces overpass the centripetal forces are willing to pay the independence cost. In this thesis, I present the different factors (i.e. wealth, ethno-cultural factors and historical factors) that may make the electorate of a region support the claim for secession. By analyzing 128 elections in the lower chambers of the European Union countries using data on the regional GDPs, the historical languages and the independence history of the regions, the paper shows that these factors foster secessionist movements.
Keywords: Secessionism, wealth, language.
Master thesis.
MSc International Economics & Business 18 June, 2019
Lluc Areny Naves S3724255
l.areny.naves@student.rug.nl llucarenav94@gmail.com University of Groningen
Index
Introduction………1 Literature review………3 Hypothesis………..8 “Wealth” hypothesis 8 Ethno-cultural hypothesis 8Historical Independence hypothesis 9
Data and variables………11
Dependent variable “Vote share” 11
1
Introduction
After the dissolution of the Soviet Union (1991) and the Yugoslav wars (1991-2001), it seemed that the different regions of Europe had achieved a national state stability (Fukuyama, 1992). However, after almost two decades of solid recovery and growth driven by innovation, globalization. In recent years populations in various European regions have showed a desire to pursue their own identity by searching new ways of becoming independent of the national states they are currently part of. For instance, Kosovo became independent from Serbia in 2008, while the European Union did fail to recognize it as a new state because some of the member states refused to do so. This poses a dilemma: the European regions that form part of the European Union and want to secede will be recognize by the European Union and its national states.
After the independence of Kosovo the secessionist proclaims have increased in some regions like Scotland or Catalonia. Scotland held a referendum in 2014 asking to the people “Should Scotland be an independent country?”. Were the scots who voted answered by 55,3% “no vote”. Nevertheless, due to recent events related to Brexit, in which the UK is going to leave the European Union, the Scottish parliament may want to repeat the referendum.
Catalonia finds itself in a different situation. Since 2012 the various regional governments of Catalonia have attempted to achieve a referendum pact with the government of Spain. However, the different Spanish governments did not allow any pact.
Taking the cases of Scotland and Catalonia as an example, this paper addresses some of the critical reasons behind secessionist movements across the European Union.
The economic literature has suggested an economic explanation to secessionist sentiments. The richer or poorer regions want to secede from the central state because of a trade-off produced between the “centrifugal” and “centripetal” forces. The centripetal forces work for maintaining the country unite and the centrifugal work for separate the country. (Alesina & Spolaore, 2003)
In addition to, the economic explanation, cultural or historic differences between the regions and the central state can also play a role on the electorate. Creating a steeper difference on the preferences of the citizens in the different regions of a state (Desmet et al, 2011)
2 The paper is organized as it follows. The first part will consist on a literature review, which will present the main contribution of the economic theories about secessionism and its models. The second part presents the hypotheses. Thirdly, the Data and Variables chapter explain the inputs used in our regression model. The fourth part presents the specification and the tobit model that is used to fit the data. The fifth chapter will present the outcomes of the regressions. The sixth part contrasts the predictions of our model with the actual electon results. The seventh part consists of two different robustness checks of the results. Finally, the conclusion is presented.
Figure 1. Results of secessionist movements in Europe in the last elections.
3
Literature Review
The literature of secession and independence have different results and arguments when it’s applied to countries with non-democratic governments or countries with democratic governments. In this thesis, I will focus on the “democratic” governments because my study is focused on the European Union, assuming that the European Union member states are democratic states.
Foley (1967), Romer (1975) and Roberts (1977) initiated the economic theory of secession in the 1970s. Their theories focus on regional conflicts over fiscal policy that arise from different income contributions of the regions to the central state and the redistribution of those contributions between the regions by the central state. According to the theory, the role of the government is reduced to the provision of public and private goods determined through the voting behavior of the citizens of the state. In the central state, poor citizens favor a policy of high income tax rate while rich citizens favor a policy of low income tax rate. The equilibrium of this income tax rate is reached by using the theory of the median voter1, according to which the preferred tax of the majority will be imposed on the state. Naturally, the equilibrium of this tax rate will not coincide with the preferences of some of the individual regions. As a consequence, this will create a conflict between the central state and the regions who do not agree with the equilibrium tax proposed. The conflict can result in the separation of the region from the central state, because removing the institutional constrains of the union allows the institutions of the region to create new policies closer to the wishes of its voters. (Bolton & Roland, 1997)
The above argument was summarized by Alesina and Spolaore (2003). Who aim to determine which is the optimal size of countries. In the book, the force called “centrifugal force” refers to the force which accounts for the fact that large countries face more heterogeneous populations. In addition to the public goods argument, they found another force, which they define “centripetal”, this force can explain why the regions want to be united in one state. The centripetal force is explained by economies of scale, which benefit the region and the central state. In contrast the “centrifugal” forces are a handicap for the central state. Inevitably, these
1 The median voter theorem, which was develop by Anthony Downs in 1957, is an attempt to explain why parties
4 two forces create a tradeoff that takes place it in the regions that want to secede. (Alesina & Spolaore, 2003)
Alesina & Spolaore (1997) explain secessionism using a spatial differentiation model. The model induces some heterogeneity in the individual preferences of the public goods by assuming that the world consists in a line where the inhabitants are distributed uniformly. Countries correspond to a continuous subset of cuts of this line or segment. The political borders and country sizes are endogenously determined. In each country, a public good is provided in a fixed quantity and the location of this public good is determined by the majority of the votes. Applying the median voter theorem and with the assumption that individuals are uniformly distributed, the center of the country will be where we locate the public good, so individuals who are not living in the center of the country will face transportation cost. Although the location of the public good is the fairest, it will imply that the people who are living on the borders of the country will be indifferent with living in one country or another. Finally, secession is associated with political fragmentation, which reduces the distance between the inhabitants of the new country with the public good. The authors conclude that the model with democratic equilibrium that allows secessionism, leads to excessive fragmentation even though this is not the optimal distribution of the world’s welfare. (Alesina & Spolaore, 1997)
Bolton and Roland (1997) propose to model the tradeoff determining the secession of a region with a different idea. The idea puts emphasis on income inequalities across regions. The centripetal force on their model is total private income, which is higher if the countries are large, while the centrifugal force is the population heterogeneity. The model features two regions, where the income is distributed along a general2 continuous law. The aim of the
taxation is to redistribute the income. The taxation rate in these regions is decided through the majority of voting so the regions that decide to secede can modify the tax rate. Secession therefore improves on average the match between public policies and individual preference but reduces the private income. In summary, this approach considers not only individual heterogeneity through the income distribution but also regional disparities. It allows both richer and poorer regions to seek secession for different reasons, because the richer regions want to pay less tax to redistribute and the poorer regions want higher taxes to redistribute.
5 The models discussed so far ignore economic characteristics that also play a role in why regions want to secede. One characteristic is the role of globalization and trade openness during the past decades. Economic integration makes large countries less attractive because the size of the market of the different regions does not coincide only with the size of the state any longer. In an open world economy, the size of the regional markets is the whole world. Furthermore, the reduction of trade cost between countries favors secession, since the regional market does not coincide with the boundaries of the country. (Madies et al, 2018) In our paper we only use European Union countries, a supranational entity based on a single market between countries and a single currency. These factors, can favor the regions that want to secede believing that they will not lose their market potential when they secede from the country. (Schmitter, 1995: Van Houten, 2000)
In addition to the above arguments we also need to discuss strategic voting behavior in the economic theory of secession. The purpose of this stream of literature is to explain, the strategic delegation of the vote by the median voter to a party that does not have the same preferences as the median voter in the public redistribution of the goods. In this perspective, Gradstien (2004) views the secession option in a Coasian bargaining system3 between two provinces that are members of a federation. The author explains that the secession or union decision involves a strategic delegation problem. With the strategic delegation, the majority of each region tries to reduce it contribution in order to push the other region to pay a large share of the federal public good which is provided equally to the two regions. It turns out that the majorities of the regions choose the parties that have lower preference for the public good. These moves improve the regions bargaining power but leads to an inefficient allocation of the federal public good. The author suggests one way to limit the secessionist tendencies, this way is to allow a referendum in the region. Asking if the electorate wants to leave the country or not. In the referendum, the bargaining will not take place and we will know which is the real support for independence.
Finally, the models do not take into account that people are not distributed uniformly between the regions. This inspired some authors like Hetcher (2001) or Sorens (2005) to hypothesize that regions with larger populations should demand more sovereignty because it is easier to them to provide public goods. More population makes more efficient the provision and
3 Coase Theorem is an economic theory developed by Roland Coase. This theorem affirms that when there is a
6
Source: Own elaboration.
Figure 2.GDP ratio of the European regions compare to it’s population ratio.
expenditure on public goods. Nevertheless, the success of small countries like Iceland seems to contradict to this argument.
Figure 2 compares the variables GDPratio4 and Population ratio5 to refute the argument for European regions, that regions with bigger populations should be more successful than small ones. Also according to figure 2, the richer regions are not the ones with more population respect to its country so these regions can spend the same as other populated regions in public goods but with less people.
The models also ignore characteristics that are not economic which play a role like the ethnically or culturally characteristics of different regions inside a country. Sambanis and Milanovic (2014) state that the ethnic or other cultural differences between a region and the rest of the country can generate a conflict with the central government over where to provide the public goods exacerbated by if the region is politically excluded from the center. If individuals of one region identify more with the ethnic cultural group of the region rather than
4 The GDP ratio will be explain in the data chapter page 12
5 The population ratio is the population of one region divided by the population of the country distributed
uniformly between the regions of that country. (𝑅𝑒𝑔𝑖𝑜𝑛𝑝𝑜𝑝/(
𝐶𝑜𝑢𝑛𝑡𝑟𝑦𝑝𝑜𝑝
7 the nation, this can lead to reduce the investment of the government anticipating the secession, despite the fact that this will intensify the conflict.
The literature discussed so far does not consider the concentration of ethnic cultural groups in a region of the state. (Hechter, 2001; Horowitz, 1991) This concentration can change the tradeoff values of the different forces giving more power to the centrifugal force relative to the centripetal force.
Giving these models and different assertions, I develop the hypotheses that could explain why different regions inside the European Union claim for independence from the national state. In other words, my research question is: which are the characteristics of European Union regions that want to proclaim the independence of his region.
8
Hypothesis
In this chapter I will define three hypotheses on why voters of a region decide to vote for a secessionism party:
“Wealth” Hypothesis
The theory claims that rich voters will seek different redistribution policies relative to the poor voters. In connection with the theory, Sambanis and Milancovic (2014) state that regions which are relatively rich will seek more autonomy or even secession because the trade-off is not as steep for them than for poorer regions. The authors also claim that when these rich regions are net transfer donors like Catalonia in Spain or Lombardy in Italy will intensify their claims. The Gradstien (2004) argument is related to this argument because these regions will use bargaining strategy to contribute less income to the central state. Soren (2004) adds that people vote for secessionism when they expect political and economic benefits from independence. The author persists in the argument by writing: “These territorial grievances are frequently based on perceptions that the region is a net “loser” from the existing political union”. These arguments, lead to hypothesis 1:
H1: Secessionism party votes should increase when the income share of the region is higher than the average of that nation.
Another way to interpret the hypothesis 1 is by Gourevitch (1979), who suggests that economically growing but politically weak regions develop secessionism because the center loses legitimacy.
Ethno-Cultural Hypothesis
9 To track this ethno-cultural difference we use the historical languages of the region, since one of the major political projects of the regional movements has often been the revival of the historical languages or dialects. (Laitin, 1989) In order to take into account the ethno cultural determinants, we define the hypothesis 2:
H2: If the large share of inhabitants in a region have a different language tied to a different culture or ethnic background than the one represented by the central state they will support secessionist parties.
Finally, Van Houten and Fearon (2002) proclaim that dormant regional languages provide the potential for a claim of “ethnic differences”. In line with the argument of Horowizt(1991) explained above, Sambanis and Milanovic (2014) point out that the main difference between an ethnic group and a nation inside another nation is that the people of the nation inside another nation is concentrated in a region, which they see as their homeland. Whereas, the people of an ethnic group, which is spread in different regions of the state, cannot have their ideal homeland.
Historical independence hypothesis
The last hypothesis is based on historical grievances and previous conflicts between regions and the central state. These previous conflicts can shape a sense of distance from the national identity making voters claim for nationalistic demands. Secessionist demands may be greater in regions that experienced independence during some periods in the past. They often developed their own institutions before being integrated in a larger entity through force or agreement. (Sorens, 2005)
In addition, majoritarian regional groups that had been excluded from taking part in central government ruling in the past may also want to secede. These groups have often been denied rights or economic opportunities by repressive states. Inevitably, these factors create a sentiment of illegitimacy and low levels of national identification, allowing regional groups to claim for their own territory. (Sambanis & Milanovic, 2014)
10 one because it is based on the Scottish history of independence, whereas the Welsh secessionism movement is based on the defense of the Welsh language. Given these arguments, hypothesis 3 can be formulated as follows:
H3: Regions with recent history of independence will support secessionism proclaims.
11
Data and Variables
As explained in the previous chapter, I choose the European Union NUTS 1 and NUTS 2 definitions depending on the country. Appendix 1 shows which countries are using the category NUTS 1 and which are using the category NUTS 2. While NUTS 1 was mostly used, some countries did not fit to this category, because the regions that have secessionism vote form part of the NUTS2 classification only. One of the countries that did not fit with the NUTS 1 is Spain, which is characterized by historical autonomies like the Basque country or Galicia. In the NUTS1 classification, these regions could not be separated from other regions that did not have secessionist parties or votes.
Most of the data for the regions was collected from Eurostat. However, there were countries like France, Netherlands and Poland for which the data was not available in the Eurostat database. For France and the Netherlands I collected the data from their national statistic institutes (France: INSEE (Institut national de la statistique et des études économiques)) and (Netherlands: CBS (Centraal Bureau voor de Statistiek)). For Poland, I could not obtain the data which was therefore excluded from the analysis.
Finally I choose the period between 2000 and 2017 for my analysis, because all of the data is available for this period. Moreover, the global is to track the recent secessionist movements. Dependent variable “Vote share”
The dependent variable in the analysis is the vote share of secessionist parties in every region. This vote share is obtained from the results of elections performed for the lower houses of the different countries that form part of the European Union. The advantage of using secessionist vote share is its comparability. For a well-defined set of countries like the democracies in the European Union, voting for secessionist parties is the method that civil society use to support a secessionist proclaims. The level of electoral support of the secessionism parties is used in both journalistic and academic work to track the popular support for independence. (Sorens, 2005)
The variable Vshare measures the vote share of the secessionist parties in every NUTS region. I constructed this variable from the constituency-level Elections Archive (CLEA), which has all votes for every constituency and every party within the period I study.
12 Social Union (CSU) in Bavaria or Democratic Union of Catalonia (UDC) in Catalonia are organized to defend the regional interest and traditions, but they often do this by partaking in the central government and negotiate with them new deals for the regions. These movements typically do not support the independence of the region. (Sorens, 2004)
The dependent variable includes all the parties that Jason Sorens (2005) classifies as ethier moderate6 or radical7 secessionist parties. However, this classification was not comprehensive enough because in ten years the political parties can change their names and objectives. I therefore also used another source, i.e. the European Free Alliance (EFA). This association is composed of several political parties that seek the independence of their region or want their region to be more autonomous from the central state. On their webpage EFA provides the postulates of the different parties. By examining these postulates, it is possible to discern which parties favor secessionism movements.
With these two sources discuss above, I constructed the list of secessionist parties, which is reproduced in appendix 2. Some of the secessionist parties that are participants in the EFA cannot be included in the list, because the regions they represent are too small to be even included in the NUTS 2 classification. One example is the Moravian movement in Czech Republic, which advocates for the self-determination of Moravia and Silesia.
With the list of secessionist parties and the data provided by CLEA I constructed the Vshare, which, as explained before, measures the vote share of the secessionism parties in every parliamentary election in the lower national chamber which take place in the NUTS region. Independent variables
With the data described in the introduction of the chapter, I could create a variety of different independent variables to test the hypotheses outlined above. The variables, which I use to test the first hypothesis, are GDPratio and GDPshare. GDPratio is the relative regional Gross Domestic Product (GDP) per capita relative to the nation-wide mean per capita GDP. In addition, the GDPshare is the GDP share of the region in country’s national GDP.
The maximums for GDPratio is 2,4 Bucharest region in Romania, while for GDPshare is 1 in countries with only one region like the Czech Republic. The minimums for GDPratio is 0.53
6 Moderate parties pursues a more ambiguous strategy emphasizing the word self-determination more than
independence.(e.g PNV, BNG, …)
7 Radical parties are the ones, which pursue a more direct strategy willing to proclaim the independence of the
13 in the French Ultramar territory while for the GDPshare is 0.0012 in Ciudad de Melilla, Spain.
The variable, which I use to test the ethno-cultural hypothesis, is the Langdif. This variable is based on the work of linguistics, who classify languages into families and subfamilies and yet smaller categories according to their structural components. (Fearon and Laitin, 2000)
An example of this classification can be two languages of the British Islands: English, which is ranked as an Indo-European language, Germanic, Western, North Sea, English, while Irish Gaelic is an Indo-European Language, Celtic, Insular, Goiledic. As we can see in the example, these two languages diverge in the second family or group.
To code these differences Fearon and Van Houten (2002) use the classification data of Grimes (1996) on basis of which they constructed the variable LANGFAM, which is defined as the number of the first level at which the historical language of the region and the language spoken commonly in all the country diverge. When the central language and the regional language are the same, they assign the value 10. Finally, with the variable LANGFAM they construct the variable LANGDIF, which is the variable that I use in this study. The variable ranges between 1 to 0,1. The 1 is for the regional languages, which proceed from entirely different families than the central languages (e.g. Spanish and Basque), while the 0,1 is for regions which have the same historical languages of the center.
Finally, for testing the historical independence hypothesis I will use a variable called Hind. This variable is a dummy between 1 and 0. The 1 is coded when the region had a period of independence as a country after the 1648, and 0 if the region didn’t have any period of independence. The year 1648 is chosen because this is when the different treaties of Westphalia were signed and is often defined as the end of the era of the feudal system and the beginning of the national states. (Sorens, 2005)
Control variables
14 Autonomy is a dummy variable in which the 1 accounts for the regions that are defined as a federal state like Germany or a devolved8 state like Spain and the 0 is for central states like
France. This variable controls for the possibility that secessionist parties might receive a greater support in federal or devolved states than in central states.
IDC is a variable measuring whether the regions represents an island as part of the European Union, form part of a national state. The variable assumes a value of 1 if the island is part of a bigger state whereas it is equal to 0 if the island is an independent state. I use this variable because the islands normally are the most distant regions from the capital.
Finally, the last dummy variable is Capital, which is equal to 1 if the region is the capital of the state and 0 if it is not. The capital variable controls for a region, which represent or includes the capital of a country, which normally is the richest region of the country, for example London in the United Kingdom or Île de France in France.
Table 1, depicts all the variables that we are going to use in the model and shows their means, standard deviation, maximum and minimum.
8 Devolved state are the states, which delegate some powers from the central state to a government in a
subnational level. It is a form of administrative decentralization. Regions with devolved powers have the power on legislate in that powers. (e.g. Spain, Italy)
Table 1. Descriptive statistics.
15 Finally, it is important to control whether the explanatory variables present multicollineaity problems. The table 2 shows that most of the variables do not have linear correlations between them. The only variables that have correlation issues are Capital and Autonomy variables. The two of them are control variables; Capital is moderately correlated with the GDPratio and strongly correlated with the GDPshare. This might happen because Capital regions are usually the richest region of the country and they agglomerate a huge part of the national economic activity, as we can see in figure 3. However, those regions do not have secessionist movements because the center of power of the country is there. With respect to Autonomy, the control variable is negatively correlated with the GDPshare.
Source: Own elaboration.
Table 2. Correlation of the independent variables
GDP share in 2017 of the regions inside its country. Source: Eurostat (2019). Gross domestic product (GDP) at current market prices by NUTS 2 regions (nama_10r_2gdp). Own elaboration
16
Methodology
This chapter presents the empirical strategy that was used in this study. Model
This study analyzes the outcomes of 128 parliamentary elections in the lower national chambers covering 127 European Union regions in the period going from 2000 to 2017. By choosing the parliamentary elections, I follow the argument of Swank and Betz (2003) that the data from the parliamentary elections in the lower chambers is relatively consistent in structure and political importance, across states and time. The following model will test the effects of three different variables on the vote share for the secessionist parties (Vshare) during the parliamentary elections.
(1) 𝑉𝑠ℎ𝑎𝑟𝑒𝑖𝑡 = 𝛼 + 𝛽1𝐺𝐷𝑃𝑟𝑎𝑡𝑖𝑜𝑖𝑡+ 𝛽2𝐺𝐷𝑃𝑠ℎ𝑎𝑟𝑒𝑖𝑡 + 𝛽3𝐿𝑎𝑛𝑔𝑑𝑖𝑓𝑖 + 𝛽4𝐻𝑖𝑛𝑑𝑖 + ∑ 𝛽𝑖𝐶𝑖 + 𝜀
The dependent variable, 𝑉𝑠ℎ𝑎𝑟𝑒𝑖𝑡, is the vote share of the secessionist parties every election year. 𝐺𝐷𝑃𝑟𝑎𝑡𝑖𝑜𝑖𝑡 is the ratio of the GDP per capita of the region to the GDP per capita of the
country, 𝐺𝐷𝑃𝑠ℎ𝑎𝑟𝑒𝑖𝑡 is the share of GDP of the region in each country´s GDP. 𝐿𝑎𝑛𝑔𝑑𝑖𝑓𝑖 is
difference of language families between the central language and the historical regional language. 𝐻𝑖𝑛𝑑𝑖 is a dummy variable which accounts if the region had a period of independence after 1648. Finally, ∑𝛽𝑖𝐶𝑖 is the set of control variables explained in the
17 Estimation
Our focal dependent variable, Vshare, is a limited dependent variable with 0 in more than 75% per cent of the cases. In figure 4 we can see that in 602 results of elections of the sample most of them are 0, only 123 are higher than 0. Only 123 election results have secessionist vote.
In such cases the, ordinary least square (OLS) estimator model would be biased and inconsistent. An alternative to OLS is Tobit, which estimates the betas with the maximum likelihood procedure. Tobit recognizes that the dependent variable have two sorts of data, the limit observations (y=0) and the non-limit observations (y>0). With this type of data, Tobit extends the Probit procedure for estimating a non-linear probability function and produces consistent and unbiased estimates for equation (1). (Hill, Griffiths & Lim, 2011) The maximum likelihood regression procedure has also been used before in some studies for the dependent variable vote share.(Jackman & Volpert, 1996)
The standard Tobit model looks like this:
Source: Own elaboration.
18 Where the 𝑦𝑖∗ is the latent variable,
𝑦𝑙∗ = 𝛽𝑥
𝑖 + 𝑢𝑖, 𝑢𝑖~𝑁(0, 𝜎2)
Censoring at value 𝑦𝑙 (left censoring) is used in this study,
𝑦
𝑖{
𝑦
𝑦
𝑖∗𝑖𝑓 𝑦
𝑖∗> 𝑦
𝑙 𝑙𝑖𝑓 𝑦
𝑖∗≤ 𝑦
𝑙However, the Tobit model assumes normal distribution of the residuals. To achieve the normal distribution, the dependent variable Vshare had to be converted Vshare=ln (1+Vshare). To achieve a normal distribution the secessionist vote shares, which have less than two percent of the population support, had to be dropped out of the sample.
In addition, to test the normality assumption. The Jarque Bera test9 is runned for the latent variable. The next output shows the result and the p-value of the Jacque Bera test applied to 𝑦𝑖∗.
The test refuse the null hypothesis on five per cent significance. Meaning that 𝑦𝑖∗ is normally
distributed.
9 The Jarque Bera test is named after Carlos Jarque and Anil k. Bera, is a goodness of fit test analyzing if the
19
Results
Table 3 presents the signs of the results that are expected from the different hypotheses and variables.
Table 4 reports the results of five model specifications. The first one includes only the main explanatory variables, whereas in the second to the fifth all the variables included. The third and the fifth also include robust standard errors to control for potential heteroscedasticity of the data. The fourth and the fifth model exclude the Vlaams Gewest in Belgium since the scatterplot showed it as a potential outlier, see appendix 4. Finally, the reported sigma of the table is the standard error of the regression, which is comparable to the root mean square error obtained in the OLS regressions.
The five models are used to show the effect of including the control variables, which made the estimations stronger, while also including the robust errors to account for the heteroscedasticity problem.
Table 4 shows in the first column that without control variables the coefficients relative to the hypotheses two and three are positive and significant. However, the first hypothesis presents an opposite sign: while the sign of the GDPratio is positive and significant, the sign of the GDPshare is negative and significant. Intuitively, this is explained by the absence of the control variable capital. The capital region of the countries normally is where the major economic activity of a country is located. Furthermore, the people who are living there are normally not supporting a secessionist movement because the political power of the country is concentrated there.
Source: Own elaboration.
20 The models presented in column two and column three, include all the variables. In these two models, all the explanatory variables are positive and highly significant. Nevertheless, between the control variables there are differences. The capital variable is never significant in the models. The variable becomes insignificant because of multicollinearity problems with the GDPshare, see table 2 in the Data and Variables chapter. The use of robust standard errors does not seem to affect the estimations.
The models of the column four and five include all the variables, presenting a different outcome. After controlling for the outlier (Belgium), mostly all the variables are positive and highly significant (excluding Capital) but GDPshare becomes less significant, it is only significant at 10% level.
Regarding the different hypotheses exposed in this chapter, the results confirm the hypotheses. The table 4 confirms the wealth hypothesis: the GDPratio and the GDPshare are significant and positive in the preferred models including all variables. However, when Belgium is taken out, it is less clear that the regions with secessionist vote have a greater
(1) Explanatory variables. (2) All variables. (3) All variables control for heteroscedasticity. (4) All variables without outlier. (5) All variables without outlier control by heteroscedasticity. Source: Own elaboration.
21 GDPshare than the rest of the regions of the country. These results seem in line with the studies of Sorens (2005), Van Houten and Fearson (2002) which report similar estimates. In conclusion, the relatively rich regions of a country, with the exception of the capitals, present a greater support for secessionist parties.
The ethno-cultural hypothesis is also confirmed by Table 4. The table illustrates that the presence of a historical regional language positively affects the support for secessionist parties. Van Houten and Fearon (2002) also found that the presence of secessionist parties is mostly related to historical regional languages. Sambanis and Milancovic (2014) found similar results, showing how the increase of regional distinctiveness is likely to increase the sovereignty demands of a region.
Finally, the model in Table 4 gives positive and significant results for the historical hypothesis. As in the study of Sorens (2005) the Hind variable confirms that regions with an independent history show more support for secessionist parties.
22
Predictions
Using the estimates of models (2) and (4), this section is going to show the predictions of those models and the outcomes with the last elections of four different regions known for its claims of sovereignty. These regions are Scotland, Catalonia, Euskal Herria and Vlaams Gewest.
The table 5 presents the predictions of the models (2), (4) and the actual outcome of the different elections. As we can see the differences between the models are small, however we use model (4) to show that the outlier Vlaams Gewest is not biasing our results.
The table shows that the predictions of our models are somewhat underestimating the actual vote share of secessionist parties during the elections. The most interesting cases are Catalonia and Scotland, which have an underestimation of 20% and 19%.
However, we can explain this under performance with two arguments. Gradstien (2004) sees the secessionist parties as bargaining actors for the region. In connection with the results of table 5, the electorate of these regions may try to reduce its financial contribution to the central state by voting more than is expected for secessionist parties.
The second argument is that maybe these four regions have other characteristic for what my models doesn’t account. For example, is the role of historical fiscal deficits against the central state (Catalonia) or natural resources (Scotland). (Sambanis & Milancovic, 2014)
Table 5. Results and predictions
23 Brandon (2015) is focus on the Catalan secessionist movement. Brandon (2015) found that they support the independence because they feel exploited by the Spanish fiscal policy. The huge bias of the prediction may be explained by the historical fiscal deficits.
Sambanis and Milancovic (2014) find out that the presence of natural wealth like oil or gas reserves in a region is correlated with higher demands of sovereignty. The underperformance of the model in the Scottish electoral results may be explained by Scottish reserves of oil and gas in the North Sea.
24
Robustness
In this chapter, I will compare the Tobit model with the OLS model. I will also compare the Tobit estimations with the Probit estimations.
Table 6 shows the results of the different regression OLS and Tobit. It shows that the OLS estimators are highly significant and with the sign expected from the different hypothesis. However, the values of the coefficients are very small.
The appearance of zeros can make the OLS to underestimate the data (William H. Greene, 1981). Consequently, the Tobit model is a better estimation process than the OLS in this particular case. The Tobit coefficients are highly significant and much higher than the ones estimated by the OLS model.
Source: Own elaboration.
25 Secondly, this section is going to test if the estimates of Tobit are consistent with the estimates of Probit, since the models are similar in many ways. Greene (2003) found out that the Probit estimates 𝛽𝑝 should be consistent with the Tobit estimates divided by the standard error of Tobit 𝛽𝑡/𝜎𝑡. This result will hold if the Tobit model is correctly employed.
In Table 7, the two models are compared, dividing the Tobit estimates by the standard error (Sigma). The table also shows that Probit omits the variable capital because of multicolinearity problems. As Table 2 in the Data and Variables chapter shows, Capital has a strong correlation with GDPshare. However, Table 7 also confirms that the Tobit model fits the dataset well, because the estimates of column one and three are similar.
Table 7. Probit, Tobit and 𝜷𝒕/𝝈𝒕.
26
Conclusion
Since the 2014 Scottish referendum and the 2017 Catalonian referendum, the European media have been focus on these regions, for its political situation, due to concerns about their secessionist demands and what that would imply for the integrity of the European Union. This thesis attempts to answer which factors are the greatest determinants of secessionist support across Europe. The literature about the optimal size of the states (Alesina & Spolaore, 2003) and the model of Bolton and Roland (1997) provide an economic explanation. These models predict that when economies of scale are relatively unimportant and the cost of population heterogeneity is high the region will secede. Nevertheless, the desire of sovereignty cannot be express only by a narrow economic calculus. (Madiès et al, 2018) These models should include more factors like the ones Sambanis & Milancovic (2014) and Sorens (2005) include. Using the work of these authors, the thesis tests the different hypothesis and finds out that they are all valid. As Van Houten and Fearson (2002) wrote in his study “These results suggest that ethnic and economic explanations of regional autonomy movements are complementary”.
Although it seems an obvious argument, most of the economic models do not take into account the cultural or ethnic differences. This occurs because it is difficult to assess mathematically what are the cultural or ethnic differences between two regions. However, these factors do play a role. Reynaerts and Vanschoonbeek (2016) report in their article that the newly independent countries lower its GDP per capita by 20%, thus, showing how independence, when obtained, results in an economic recession. While the economic prospects are not positive, secessionist movements gather a great support at the elections. The conclusion must be that the cultural, ethnic or historic differences lower the “centripetal” forces of the trade-off.
27
Datasets
Centraal Bureau voor de Statistiek (2019). Regional accounts; key figures 1995-2015 [data file]. CBS [Producer and distributor].Retrieved from http://statline.cbs.nl
Eurostat (2019). Gross domestic product (GDP) at current market prices by NUTS 2 regions (nama_10r_2gdp) [data file]. Eurostat [Producer and distributor].Retrieved from http://www ec.europa.eu
Eurostat (2019).Population on 1 January by age, sex and NUTS 2 region (demo_r_d2jan) [data file]. Eurostat [Producer and distributor].Retrieved from http://www ec.europa.eu Institut national de la statistique et des études économiques (2019). Produits Intérieurs Bruts Régionaux (PIBR) 1990-2015 [data file]. Insee [Producer and distributor].Retrieved from http://www.insee.fr
Kollman, K., Hicken, A., Caramani, D., Backer, D., & Lublin, D. (2018). Constituency-level elections archive [data file and codebook]. Ann Arbor, MI: Center for Political Studies, University of Michigan [producer and distributor]. Retrieved from
http://www.electiondataarchive.org.
Sorens, J. (2017). Nationalistvote.xls [data file]. Jason Sorens, Dartmouth College [producer and distributor] Retrieved from http://www.dartmouth.edu
References
Alesina A, & Spolaore, E. (2003). The Size of Nations, The MIT Press
Alesina, A, & Spolaore, E. (1997). On the number and size of nations. Quarterly Journal of Economics, 112(4), 1027–1056.
Bolton, P, & Roland, G. (1997). The breakup of nations: a political economy analysis. Quarterly Journal of Economics, 112(4), 1057–1090.
Boylan, B. (2015). In pursuit of independence: the political economy of Catalonia’s
secessionist movement. Journal of the association for the study of ethnicity and nationalism, 21(4), 761-785.
Desmet, K, Breton, M, Ortuño-Ortín, I, Weber, S. (2011). The stability and breakup of nations: a quantitative analysis. Journal of Economic Growth, 16(3), 183–213.
Fearon, J, & Laitin, D. (1999). Weak states, rough terrain, and large-scale ethnic violence since 1945. Paper prepared for delivery at the 1999 Annual Meetings of the American Political Science Association (APSA), 2-5 September 1999, Atlanta, GA.
Fearon, J, & Laitin, D. (2000). Ordinary Language and External Validity:
SpecifyingConcepts in the Study of Ethnicity. Presented at APSA Annual Meeting, Washington DC
28 Foley, D. (1967). Resource Allocation and the Public Sector, Yale Economic Essays, VII, 45–98.
Fukuyama, F. (1992). The End of History and The Last Man, Free Press
Gradstein, M. (2004). Political bargaining in a federation: Buchanan meets Coase. European Economic Review, 48(5), 983–999.
Grimes, B. (1996). Ethnologue: Languages of the World (13th ed). Dallas, Summer Language Institute
Hechter, M. (2001). Containing nationalism. Oxford, UK: Oxford University Press.
Hill R.C., Griffiths W.E., Lim G.C. (2011): Principle of Econometrics. 6th Eds John Wiley & Sons, Inc., New Jersey.
Horowitz, D. L. (1985). Ethnic Groups in Conflict. Berkeley, University of California Press Horowitz, D. L. (1991). Self-determination: Politics, philosophy, and law. In I. Shapiro & W. Kymlicka (Eds.), Ethnicity and group rights: NOMOS XXXIX. New York, New York University Press, 421-461.
Jackman, R, & Volpert, K. (1996). Conditions Favoring Parties of the Extreme Right in Western Europe. British Journal of Political Science, 26(4), 501-521.
Keating, M. (2001). Plurinational Democracy. Stateless Nations in a Post-Sovereignty Era, Oxford, Oxford University Press.
Madiès, T, Rota-Grasiozi, G, Tranchant, J, Trepier, C. (2018). The Economics of Secession: A Review of Legal, Theoretical, and Empirical Aspects. Swiss Journal of Economics and Statistics 154:19.
Reynaerts, J. and J. Vanschoonbeek, 2016. The economics of state fragmentationassessing the economic impact of secession. MPRA Paper 69681.
Roberts, K. (1977). Voting over Income Tax Schedules, Journal of Public Economics, VIII, 329–340.
Romer, T. (1975). IndividualWelfare, Majority Voting, and the Properties of a Linear Income Tax. Journal of Public Economics, IV, 163–185.
Sambanis, N, & Milanovic, B. (2014). Explaining regional autonomy differences in decentralized countries. Comparative Political Studies, 47(13), 1830–1855.
Schmitter, P. (1995). If the Nation-state Were to Wither Away in Europe: What Might Replace It? Oslo: Arena.
Sorens, J. (2004). Globalization, secessionism, and autonomy. Electoral Studies 23, 727-752. Sorens, J. (2005). The cross-sectional determinants of secessionism in advanced democracies. Comparative Political Studies, 38(3), 304–326.
29 Van Houten, P. (2000). Regional Assertiveness in Western Europe: Political Constraints and the Role of Party Competition. Dissertation, University of Chicago
Appendix
Appendix 1 List of regional classification in European countries.
Country
Region
Classification
Austria Nuts 1 Belgium Nuts 1 Bulgaria Nuts 1 Czechia Nuts 1 Croatia Nuts 1 Cyprus Nuts 1 Denmark Nuts 1 Estonia Nuts 1 Finland Nuts 1 France Nuts 2 Germany Nuts 1 Greece Nuts 1 Hungary Nuts 1 Ireland Nuts 1 Italy Nuts 2 Latvia Nuts 1 Lithuania Nuts 1 Luxembourg Nuts 1 Malta Nuts 1 Netherlands Nuts 1 Portugal Nuts 1 Romania Nuts 2 Spain Nuts 2 Slovenia Nuts 1 Sweden Nuts 1 United Kingdom Nuts 130 Appendix 2 List of secessionist parties by country.
Country
Secessionist Political Parties
Belgium Direct (Vlaams Gewest)
New Flemish Alliance (Vlaams Gewest) Libertair (Vlaams Gewest)
Vlaams Belang (Vlaams Gewest) Wallon Rally (Wallonia)
Finland Future of Aland (Äland) France A Manca Naziunale (Corsica)
Corsica Naziune (Corsica) Corsica Libera (Corsica) Corsica Viva (Corsica) Liga de la Savouè (Savoy)
Parti pour l’Organisation d’une Bretagne Libre (Bretagne) Partit Occitan (Occitania)
Union Democratique Bretonne (Bretagne) UPC (Corsica)
Rinnovu Naziunale (Corsica) Germany Bayern Partei (Bayern)
Italy Independenza Veneta (Veneto/Venezia Guilia) Lega Nord “Secessionist party until 2016”
(Liguria/Lombardia/Piemonte/Trentino/ Venezia Guilia/Veneto) Lega Lombarda (Lombardia)
Lega Sard (Sardegna)
Liga Fronte Veneto (Veneto/Venezia Guilia) Liga Veneta (Veneto/Venezia Guilia)
Liga Veneta Republica (Veneto/Venezia Guilia) Mesa Sardos Liberos (Sardegna)
Noi Siciliani—FNS (Sicilia)
Partidu Indipendentistu Sardu (Sardegna) Partito Sardo d’Azione (Sardegna)
Progetto nordest (Veneto/Venezia Guilia/Lombardia) Veneto stato (Veneto/Venezia Guilia)
Sardigna Natzione (Sardegna)
Portugal Partido Democratico do Atlantico (Azores) Romania Erdélyi Magyar Néppárt (Transilvania)
Spain Amaiur (Euskal Herria/ Navarra) BNPG/BNG/Bloque PSG (Galicia) Bloque Nacionalista Galego (Galicia)
CNC-FREPIC-ANAWAC/CCI “Canarias por la Independencia” (Islas Canarias)
31 Convergència i Unió (Catalunya)
Esquerra Republicana de Catalunya (Catalunya/ Illes Balears/ Comunitat Valenciana)
Estat Catala (Catalunya/ Illes Balears/ Comunitat Valenciana) Estau Aragonés (Aragón)
Euskadiko Ezkerra (Euskal Herria/ Navarra)
Euskal Herria Bildu / Reunir Euskal Herria (Euskal Herria/ Navarra) Eusko Alkartasuna (Euskal Herria/ Navarra)
Frente Popular Galego (Galicia) Més Mallorca (Illes Balears) Més Menorca (Illes Balears) Més Formentera (Illes Balears) Nación Andaluza (Andalucia)
Partido Nacional Canario (Islas Canarias)
Partido Nacionalista Vasco – Euzko Alderdi Jeltzalea (Euskal Herria/ Navarra)
Partit Demòcrata Europeu Català (Catalunya) Partit Socialista per Menorca (Illes Balears) Puyalon de Cuchas(Aragón)
United Kingdom
Plaid Cymru/Party of Wales (Wales) Scottish National Party (Scotland)
32 Appendix 3Searching for outliers on the GDPratio and the GDPshare.
Figure Appendix 1. Scatter plot of the latent variable y* vis-à-vis GDPratio.
