An eroding centre : implications of globalisation and inequality for the existing political order

(1)

An Eroding Centre

Implications of Globalisation and Inequality for the

Existing Political Order

Master Thesis Political Science – International Relations Written and submitted by:

Simon Graf

11242884

Prins Hendrikkade 99-B-III 1011 AG Amsterdam sim.graf@hotmail.com

+31 6 50 57 29 56

Our Changing Global Economic Order: Growing Risks and Imbalances Second Semester 2016-2017

Supervisor: Prof. Dr. G.R.D. Underhill

Submission Date: 23-06-2017 Word count: 18,817

(2)

List of Tables

Table 1: SWIID 5.1: SE Comparison ... 19

Table 2: Analysed Elections ... 21

Table 3: Voting Behaviour of the Education Subsets ... 28

Table 4: H1.1 Model I ... 30

Table 5: H1.1 Model II – Excerpt ... 32

Table 6: H1.1 Model III – Excerpt ... 33

Table 7: H1.1 Model IV ... 34

Table 8: H1.1 Models V and VI ... 35

Table 9: H1.2 Models I and II – Excerpts ... 37

Table 10: H1.2 Models III and IV ... 38

Table 11: H2 Model I ... 39

Table 12: H1 Models I and II ... 40

Table 13: H1 Models III and IV ... 41

List of Figures

Figure 1: Theoretical Argument………...14

Figure 2: Correlations: Inequality – Globalisation………...24

Figure 3: Sample Distribution of Education Levels……….24

Figure 4: Sample Distribution of Age, Gender and Education Levels……….25

(4)

List of Abbreviations

AfD Alternative für Deutschland [Alternative for Germany]

AFTA ASEAN Free Trade Area

ASEAN Association of Southeast Asian Nations BIS Bank of International Settlement CHES Chapel Hill Expert Survey

CSES Comparative Study of Electoral Systems

EEA European Economic Area

FDI Foreign Direct Investment

FPÖ Freiheitliche Partei Österreichs [Freedom Party of Austria]

GAL Green, Alternative, Libertarian ILO International Labour Organization

LIS Luxemburg Income Study

MAD Mean Absolute Difference

MARPOR Manifesto Research on Political Representation

MFN Most Favoured Nation

NAFTA North American Free Trade Agreement

OECD Organisation for Economic Co-operation and Development

OLS Ordinary Least Squares

PPP Purchasing Power Parity

SD Standard Deviation

SES Socioeconomic Status

SWIID Standardized World Income Inequality Database

SYRIZA Synaspismos Rizospastikis Aristeras [Coalition of the Radical Left]

TAN Traditiona, Authoritarian, Nationalist

UP Unidos Podemos [United We Can]

(5)

Abstract

Many Western industrialised states have witnessed a remarkable rise of populist parties in recent years, especially since the outbreak of the 2008 global financial crisis. Interestingly, this phenomenon also occurred in countries that were comparatively little affected by it. Instead of the crisis itself, the author suspects that increasing inequality in developed states was one of the major drivers behind this development. In an attempt to shed some light on the consequences of changing economic inequality, the author analysed the voting behaviour of respondents in different skill cohorts in 52 elections among 20 OECD countries. Recent developments in the quality and quantity of cross-country comparable inequality data enabled the author to conduct a statistical analysis of the suspected links. While he did find support for changes in voters’ preferences on the economic and the new politics dimension, the question whether parties do adjust to these changing preferences could not definitively be answered. Furthermore and unfortunately, the data did not allow final conclusions with regard to developments in the overall political polarisation of polities. Nevertheless, the analysis shows that especially centre parties in developed countries need to pay attention to growing levels of inequality. It also indicates that if inequality keeps growing, social and political tensions could increase, both on the national as well as on the international level.

Introduction

The 2008 financial crisis has had devastating effects on the global economy. Although developing countries were affected too – mostly by spillover effects – the countries hardest hit by the crisis can be found among the more developed regions in the world (Ötker-Robe & Podpiera, 2013). Especially a number of countries in the European Union that was subsequently also hit by the 2011 sovereign debt or Euro crisis, faced massive economic downturns and came close to default. Greece and Spain saw unemployment rates of well over 20% over a period of 5 years (and the persisting in the former), with youth unemployment rates temporarily surpassing the 50% mark (Eurostat, 2017a; Eurostat, 2017b). In their working paper for The World Bank, Ötker-Robe and Podpiera (2013) note that economic downturns and crises especially and disproportionately negatively affect the poorest in societies, both via market mechanisms and the resulting pressure on employment and income, as well as via welfare political effects and the state’s decreasing ability to provide basic services to its citizens.

The aftermath of the crisis saw a number of political changes in the European political landscape. New parties – be they populist left, populist right, anti establishment, regional or protest parties –

(6)

challenged the (in recent history) traditionally stable (Western-)European continent’s established political order. What seems odd though is that the emergence and/or the remarkable increase of electoral success of these parties is not only limited to the especially crisis-torn European periphery such as the abovementioned Spanish (UP: congressional elections 2016 (21.2%)) and Greece (SYRIZA: national parliamentary election 2012II (26.9%), 2015I (36.3%)) polities, but also occurred in the relatively little affected core of the European Union such as Germany (AfD: 2016 regional elections in Baden-Württemberg (15.1%), Mecklenburg-Western Pomerania (20.8%), Saxony-Anhalt (24.3%)) or Austria (FPÖ: national council elections 2013 (20.5%)) as well as in Scandinavia (Sweden: Sverigedemokraterna: riksdag election 2014 (12.9%)) (Nordsieck, 2017). The question raised is whether the crisis was the primal reason for the rise of these challengers or just a trigger or a catalyst. If one presumes the latter, two potential alternative explanations appear on the scene: globalisation and inequality.

While globalisation and liberalised financial markets have not only fostered the velocity with which the crisis spread from the USA to Europe and the rest of the world, these phenomena also have special implications for the welfare state and income distribution in times of economic stability and expansion. Free flow of capital to markets in which it can be invested profitably and increasing market access for exporters offer unprecedented opportunities for companies and consumers worldwide. On the other hand, the low-skilled workforce in Western countries is facing increasing competition from low wage countries that have become highly competitive in international markets due to the depletion of trade barriers of all kinds and subsiding transportation costs. Whether the state manages to compensate the losers of globalisation will have a significant effect on the levels of inequality and, closely linked, on social stability in Western societies. Last year’s US elections and the British EU referendum showed the world the potential consequences of electoral success of politically peripheral entrepreneurs and, consequently, diplomats and interest groups of international trade and security relations might have to adapt and get used to new tones and logics in their field of work.

These developments do not only challenge actors in the international arena. With extreme parties occupying increasing numbers of seats in national parliaments, decision-making processes become more and more difficult. In addition to this and despite the fact that some researchers point out the advantages of a better and broader legislative representation of the electorate, the systematic exclusion of those pole parties and their voters from governmental responsibility does deteriorate the quality of democratic political systems (e.g. Morlino, 2004). This thesis addresses the issue of polarisation of national parliaments. The main research question of this work will be: Does globalisation induced economic inequality increase polarisation of national parliaments? The

(7)

author’s main argument is that an increase of inequality leads the disadvantaged voters to either support classic left parties in order to support more welfare state to compensate for their losses or parties of the so-called new right or new left, which are often anti-globalisation and protectionist and would therefore fight one of the causes for growing inequality. What is innovative about this approach is that inequality measures the relation between the incomes of different cohorts. The author is therefore not primarily interested in changes of the economic situation of the different social cohorts per se, but in the change of the ratio.

First, we will take a look at some literature on the subject as well as on the broader field of this thesis’ theoretical framework, mainly consisting of research from economic voting and election and from populism studies. This thesis then proceeds by introducing the theoretical argument as well as the hypotheses before discussing operationalisation and the employed data. A brief chapter of descriptive statistics will precede the statistical tests of the hypotheses, before a reflection of the analysis and a conclusion will round out this thesis.

Literature Review

There are a number of good reasons why researchers should be interested in societal and political polarisation. One of the most mentioned reasons in the literature is its undermining effect on democratic regimes. Dalton (2008) for example names the Weimar Republic, the Austrian democracy in the 1930s as well as the French Fourth Republic and the Chilean democracy in the 1970s as some of the victims of too highly fractionalised and polarised party systems. Remmer (1991) noted that the strength of this effect is significantly mediated by the type of party system and that it generally is stronger in more fragmentised systems. Of course, a system collapse is the extreme case. However, Jones (2001) provides evidence that, even in a two party system, polarisation is hampering democratic processes, by showing that it leads to increased probabilities of government gridlocks in the USA. In their study on the Swiss welfare state development, Häusermann, Mach, and Papadopoulos (2004) also point out the negative effect of polarisation on the ability and willingness to compromise of representatives of employers and unions, and the political right and left. Especially in a consensus democratic political system like Switzerland, where the so-called social partners occupy a crucial role in the political decision-making process, this is a serious problem and drives the country away from the ideal type of a consensus democracy (Vatter, 2008).

Despite those serious issues, there are also positive aspects to a highly polarised party system. Dalton (2008) argues that as a result of political polarisation, voters are presented with clearer and more

(8)

distinct party choices. There is also a more intense partisan competition, which in turn stimulates the public discourse and participation and therefore positively affects representation. Bornschier (2015) too points out the positive aspect of a better, more congruent parliamentary representation. According to him, the example of Switzerland showed that the potential for a certain vote share of the new right had been there since the 1970. Because the class cleavage remained the main driver of political distinction, this new line of conflict failed to surface for years. With the arrival of the new right party family in the mainstream political arena of Europe and the politicisation of the cultural dimension, the electorate is arguably better represented.

Apart from its effects, we are naturally interested in the causes of polarisation. What factors are held responsible for increases in political polarisation in the literature? In 1991, Remmer linked the erosive and delegitimising effects of political polarisation to the various economic crises of the 1980s. With regards to welfare policy in Switzerland, Häusermann, Mach, and Papadopoulous (2004) affiliate the increasingly polarised positions to a number of economic factors such as budget deficits and recessions, as well as pressure on spending levels and taxation originating in increasing global competition. Mughan (2007) sees a high explanatory power for the rise of far right parties on the micro-level, in personal economic factors, such as job insecurity. The driving force behind this development is the deindustrialisation of advanced economies and the offshoring of high-paying second sector jobs due to globalised markets. The far right often occupies protectionist economic policies and opposes globalisation in more general terms, also including migration and cultural aspects. Bornschier (2015) claims that the polarisation on the new politics dimension in connection with the emergence of the new right is simply a late reaction to the emergence of the new left in the 1960s and 1970s. After the fading importance of religious (in Switzerland between catholic and protestant groups and cantons) and class cleavages, the cultural conflict lines emerged from their latency. Kriesi et al. (2008) further point out that the emergence of new right parties can trigger a self-reinforcing pull mechanism and eventually lead to a more fractionalised party system if centre-right parties move towards the pole. This is very plausible, since according to Adams, Clark, Ezrow and Glasgow (2006), mainstream parties do react to shifting voter preferences and the success of other parties by adjusting their party positions accordingly. Adams and Merrill (1999) further find, that party positions are generally more extreme than the ones of their voters (also in Leimgruber, Hengartner, & Leemann, 2010).

(9)

Theoretical Framework

The research question touches several subfields of social and political science. Most prominently represented is literature from globalisation studies as well as studies on elections and populism. While the latter will explain the supply side of the theoretical argument, one needs the former to determine which countries and therein, which population groups or worker segments are most negatively affected by the mechanisms and effects of globalisation. This thesis will then focus on those cohorts that are specifically exposed to labour market risks and analyse their voting behaviour. The line of argument, the hypotheses, as well as most other aspects of this thesis will all be looked at from the viewpoint of economic voting theory.

Economic Voting

The fact that personal and public economic situations drive electoral processes is of course not surprising at all. The literature on economic voting is huge and in its broadness only of marginal importance to this thesis. What is important is that macroeconomic factors do drive electoral outcomes, although not all variables have the same effects on the different incumbent parties.

One example of this is presented by Remmer (1991). She analysed 21 elections in Latin America during the debt crises and found that macroeconomic as well as a voter’s personal economic situation have an effect on the incumbent party’s vote share. She further concludes that economic downturns might trigger political polarisation, however, without necessarily evoking political extremism or even endangering democracy as a whole. Another important contribution in the field came from Powell and Whitten (1993). They analysed over 100 national elections, and, with their sample being extended by another 40 in a replication and extension of their original work (Whitten & Palmer, 1999) they found different effects of economic indicators on the success of different parties. Against the backdrop of the traditional emphasis of the respective party families, it hardly comes as a surprise that high inflation figures turn out to be especially harmful to centre-right parties, while more leftist governments suffered disproportional losses due to high unemployment. What mediates the effects, a conclusion also found by Remmer (1991), is the configuration of the respective political systems and especially the clarity of political and democratic accountability.

A third and more recent example of economic voting literature, this time focussing on the European Union, is presented by Bosco and Verney (2012). Of the 12 elections they studied – national, regional, and municipal – only two did not result in major losses for the incumbent party, following the outbreak of the European debt crisis. Albeit not offering any evidence for the direct influence of economic factors (rather than insufficient crisis management, bad governance etc.), their case study shows the enormous impact the euro crisis had on European politics.

(10)

This list of economic voting literature can go on and on, however, for this thesis the important conclusion to be drawn from it is that most scholars do find evidence that economic indicators influence election outcomes (Hibbs, Rivers, & Vasilatos, 1982; Schneider, 1984; Nannestad & Paldam, 1997; Narud & Valen, 2001). This claim is even reinforced and supplemented by the fiscal voting literature, focussing on adjustments in incumbent governments’ fiscal policies to artificially adjust business cycles in their constituencies prior to elections (Alesina, Roubini, & Cohen, 1997; Mulas-Granados, 2004; Eslava, 2011). This also suggests that voters do in fact react to short-term changes of economic variables prior to elections.

Globalisation and Inequality

Globalisation did not start after the Second World War. Depending on the focus of studies and the definition of the term in use, one will find first signs of globalisation as early as 5000 years ago. However, if one follows the suggestions of O’Rourke and Williamson (2000) and defines globalisation as “the integration of international commodity markets” (p. 3), one must also come to the conclusion that this form of globalisation has occurred only since the early 19th century with the massive technological advantage and the consequential decrease of transportation costs to a level that would finally allow foreign trade to affect domestic prices and resource allocation. Although the precise definition of the concept globalisation is not central for the course of this thesis and it will not be measured but only used in this theoretical discussion, it should be emphasised that the focus will be mainly on the economic aspects of the term, i.e., on international economic integration. This term was described by Rodrik (2000) as an integration of “markets for goods, services and factors of production” (p. 178) or by O’Rourke (2001) as “declining barriers to trade, migration, capital flows, foreign direct investment (FDI), and technological transfers” (p. 1).

The current third wave of globalisation, according to the compelling arguments of Chase-Dunn, Kawano, and Brewer (2000), has been one of unprecedentedly deep integration, not only because of diminishing transportation costs but also and especially because of the information revolution and the emergence of highly technologised global financial markets. On the output side, this development was not only limited to the economic, but also included cultural aspects (Castells, 1993; Kellner, 2002). In 2004, the global export to GDP ratio was over 25%, having doubled since the 1970s and clearly outpaced growth rates. Other indicators can be found in international finance. For example foreign exchange markets turnovers increased more than fourfold between 1995 and 2016 (BIS, 2017). Projects such as the NAFTA, the EEA or the AFTA and of course the success of the WTO – according the WTO World Trade Report 2011, over 50% of global trade was subject to

(11)

MFN rates of zero – clearly show the importance of international trade as a driver of growth and development for states.

The major beneficiaries have been developing countries, whose growth rates outpaced the ones of the developed world in the 1990s by the factor three (OECD, ILO, World Bank, & WTO, 2010). Due to the huge populations in countries such as China or India, it is therefore not surprising that overall global inequality (not to confuse with international inequality, which measures inequality between country mean incomes) as measured by the Gini coefficient PPP has decreased by eight percentage points since the 1970s (Melchior, Telle, & Wiig, 2000; Milanovic, 2007). Although this gives reason for a bright outlook in terms of global poverty rates, these numbers shall not veil the problematic effects that globalisation processes can have on inequality. While globalisation processes are also often criticised for their environmental consequences or their arguably neo-colonial results, the protection of domestic jobs is one of the main arguments for supporting protectionist policies in the West. What this implies is that globalisation not only has a converging between-country effect, but that there are also within-country or national effects that can be much more diverse. Because this thesis focuses on effects on national elections, the most important concept of inequality will be the within-country or national one.

Country Effects

The effects of globalisation processes on income and income distribution are a little more controversial and less clear on the national level. While general trade theory expects an aggregate increase of income, a number of scholars put even these assumptions into question. Without neglecting gains from trade in general, Antràs, de Gortari, and Itskhoki (2016) question the suggested amount of these gains. They argue that because simple Kaldor-Hicks efficient improvements are generally not accepted in societies, one also has to take into account the effects of progressive redistributional policies that intend to move society closer to Pareto efficiency. This will cause the optimum level of production input of the most favoured persons to drop and therefore annihilate up to one third of the original gains of trade. They further stress the different societal impact of an increase in wealth depending on which cohorts gain most. This is not only the case from a normative point of view but also from a democracy theoretical one. Using a very similar concept to the one defined by Rodrik (2000) described above, Dollar claims the absence of a correlation between globalisation and within-country inequality in his 2001 article for the World Bank. His main arguments and findings in this matter are borrowed from Dollar and Kraay (2002). In their article, the two researchers agree with the generally accepted view of decreasing global inequality (as outlined in the previous section), but did not find any evidence for a within-country inequality

(12)

fuelling effect of globalisation. From a trade theoretical perspective, this result on a global level is not surprising, however, the fact that in their analysis this held true among different regions, might certainly be. Different factor endowments and redistributional systems as well as different levels of development would be considered to have an influence on whether inequality in- or decreases with the opening to global trade and capital. The researchers themselves acknowledge the difficulty of cross-country comparisons with inequality data as well as a couple of other issues that could possibly skew their results. In the revised and updated 2005 version of Dollar’s article, he further acknowledges, based on new data from the Luxemburg Income Study, that inequality levels in developed countries could in fact be rising, while a decrease in the developing world is observable. This is also the credo of a majority of scholars in the field. Gozgor and Ranjar (2015) find a positive effect of globalisation on within-country inequality as well as redistribution in their sample of 140 countries over a 42-year period. After splitting their sample into two groups of developing and developed economies, they further find the expected bigger impact in advanced economies. The findings of Dreher and Gaston (2008), who analyse different aspects of globalisation, point in the same direction. They too found a positive impact of globalisation on income and wage inequality. After splitting their sample into an OECD and a non-OECD group, effects in the latter became largely insignificant (depending on the model in use) and stayed significant and positive in the OECD group. Their findings also “provide an empirical ‘justification’ for Bhagwati’s (1999) observation that it is the developed countries, rather than the developing countries, that oppose greater integration” (Dreher & Gaston, 2008, p. 531). Bergh and Nilsson (2010) as well as various other researchers (e.g. Williamson, 1997; Reuveny & Li, 2003) come to similar conclusions, again finding significant positive effects that largely stem from the developed rich countries in the sample. What does this imply with regards to the research question? Based on this rich ground of empirical evidence, we can assume that globalisation processes increase income (and wage) inequality mainly in rich countries. This is an important factor contributing to the building of the theoretical argument of this thesis. Although it does definitely not account for the total change of inequality levels, it does account for parts of it. This already allows parties to politicise this issue and oppose globalisation as a measure of fighting inequality. It points towards a potential division of countries that would be important to make. The next section uses some implications of basic trade theory to shed some light on this matter and draw conclusions for the case selection.

Trade Theory

The question of who gains and who loses from trade has occupied the minds of economists and researchers since the 18th century, when Adam Smith published his one of his most highly regarded

(13)

work “The Wealth of Nations”. An alternative approach, using comparative rather than absolute advantage, but still regarding labour as the only factor of production, was presented by David Ricardo in the early 1800s. This theory of comparative advantages of production was the basis of many international trade theories. One of the classic models of the early 20th century trade theory was developed by the Swedish economists Eli Heckscher and Bertil Ohlin, which includes multiple factors of production. Following this, two different models emerged: Factor-endowment models, based on the Stolper-Samuelson theorem (Stolper & Samuelson, 1941), which assume inter-industrially mobile factors of production, whereas specific-factors or Ricardo-Viner models assume immobility (Hiscox, 2001). To illustrate the difference, imagine a simple two-country two-factor two-commodity model. The two countries start trading and, because of comparative advantages, the developed country, where capital is the abundant factor, specialises in the production of the capital-intensive good. If capital and labour can move across sectors, both move to the capital-capital-intensive sector. The result is a drop in wages and an increase in capital returns. If the factors of production cannot move, then the capital and labour employed in the capital-intensive sector gain.

Some authors in the literature generally disagree with those mechanisms. Wood (1997) as well as Topalova (2010) focus on the trade liberalisation processes in India, Mexico or Latin America of the 1970s, 1980s, and 1990s. The conclusions they reach are inconsistent with the expectations of both the Stolper-Samuelson theorem and the specific-factors model, namely that inequality in many of these countries in fact increased despite labour clearly being the abundant factor. A potential explanation for their findings that would still be compatible with the prevailing theories comes from Goldberg and Pavcnik (2007). They promote the Stolper-Samuelson theorem and argue that the theory is not applicable to countries such as Mexico, Morocco, and Colombia. These countries, according to the argument of the scholars, belong to a group of middle-income developing countries that have a comparative disadvantage in low-skilled labour intensive industries (Hanson & Harrison, 1999; Robertson, 2000; Currie & Harrison, 1997; Attanasio, Goldberg, & Pavcnik 2004; all in Goldberg & Pavcnik 2007).

A number of scholars find evidence for a hybrid theory, a so-called partially-mobile-factors model. Brawley (1997) analyses the lines of conflict behind the support for and opposition to protectionist policies in Imperial Germany and finds that they are largely based on the level of factor mobility. Beaulieu (2002) follows a similar approach to analyse the societal cleavages responsible for the positions on the 1999 Canadian free trade agreement with the USA. Hiscox (2001) analyses the historical formation of the most influential lobbying organisation in six Western developed countries and again finds that the level of factor mobility at a period clearly shaped the conflict line along

(14)

which an organisation emerged – class or industry based. More interesting contributions come from Artuç and McLaren (2015) and Mayda and Rodrik (2005).

Even though the literature is quite divided on this topic, most scholars agree with the general mechanisms behind them. All approaches would therefore generally see the abundant factor in a country as the main winner, since most of that factor is employed in the abundant-factor-intensive industry anyway. Whether the second profiteer is the rest of the factor or the other factor in the same sector is of less importance. It further supports the conclusions of the previous section, at least among a potential sample of developed countries.

Electoral and Populism Studies

After the sections above have shed some light on the expected effects of globalisation processes on different countries and social groups, we shall now turn our attention to possible consequences with regards to policy and party preferences of the affected voters. For this reason, this section will analyse some literature from the broad field of electoral and populism studies in order to develop a theoretical framework that allows for the establishment of proper and testable hypotheses in the next section. Logic would suggest that due to economic disadvantages, a voter might either decide to fight the symptoms, enlarge the welfare state and increase social services, or to fight the causes, which, in the case of the abovementioned countries, include globalisation.

The Left

Do people in fact vote for more welfare state when they elect left parties into government? Social policies of a state can absorb some of the inequality shocks that stem from globalisation but also from a variety of other sources (Smeeding, 2002). Not basing his argument directly on inequality but on trade openness, Rodrik (1998) analyses government sizes in a sample of over 100 countries, building on the findings of Cameron (1978), who argued that the best predictor for an OECD country’s tax revenue is its economic openness and found a very high correlation between the two measures. Rodrik’s explanation for his findings is that the state actively tries to reduce the risks that result from a country’s exposure to an internationally volatile environment and trade. But what about left parties more specifically?

Evidence for the so-called compensation hypothesis is presented by Walter (2010). Motivated by the mixed findings of macro-level research, Walter investigates the micro-foundations of the hypothesis in her article, which suggests that economic insecurity leads to an increase in welfare state preferences among workers. Taking into out sectoral as well as factorial considerations, she finds that higher exposure to global trade leads to higher economic insecurity, especially among

(15)

low-skilled workers. This translates into favouring more welfare state expansion and votes for the Social Democrats. Albeit using the slightly different concept of economic insecurity, Walter shows that such negative effects of trade can lead to a shift to the left in a national polity. It is noteworthy that she chose Switzerland, a small and very open economy that heavily relies on exporting and free capital flows (Quinn, 1997, as cited in Walter, 2010). Furthermore, it is important to note that her economic risk indicator consists of low income and high unemployment risk, the former of course also being partially included in inequality measures, deployed by this thesis. Rehm, Hacker, and Schlesinger (2012) conduct research in the same direction. They specifically analyse the two aspects of the economic risk indicator separately and conclude that both low wages as well as high employment insecurity lead to higher support for welfare state policies, such as unemployment insurance. They further find that the intensity of this policy preference depends on the societal distribution of the two aforementioned attributes and is higher if they fall on the same group.

An interesting insight into the psychological aspects of these employment and economic insecurities is provided by Anderson and Pontussen (2007). Their findings show that more skilled workers tend to feel more secure not only about their current position but also about the chance to find another job in case they would slide into unemployment among their sample of 15 OECD countries. This might also indicate that factor mobility is higher among skilled labour in skill and capital abundant economies. Furthermore, the researchers dismember the economic insecurity indicator and show the positive effects that different welfare policies have on the different aspects of insecurity. Again, this shows that state interventions indeed offer some cure for the economically disadvantaged, the globalisation losers. Another article on the subject, however, with more nuanced findings, was authored by Emmenegger, Marx and Schraff (2015). Not only taking into account the implications of the field of political economy and the compensation hypothesis but also other research strands such as political alienation literature and protest voting, the scholars draw a more complex picture. With the inclusion of these fields, the researchers propose a tripartite approach. They argue that economic disadvantage either leads to the rational expression of the redistributional policy preference and hence, in line with the findings of the aforementioned articles, to a left mainstream party vote (i.e., the Social Democrats); to a protest vote, either on the extreme left or the extreme right; or to political alienation and consequently to an abstention from voting. While they did not find support for the proposed lower voter turnout because of economic disadvantages for their case, which was the Netherlands, they did not systematically exclude the existence of such a mechanism in other polities, especially in two-party systems. More important, however, are their positive results regarding both left party votes as well as protest party preferences.

(16)

The (New) Right and Populist Parties

The last example in the previous section showed that the left is not the only party family to profit from economic grievances and labour market disadvantages. New challengers or anti-parties have emerged on the left and the right side of the political spectrum and are now challenging the old established mainstream parties, suffering from a growing conflict between their responsiveness and responsibility functions (Mair, 2009). Kriesi and Pappas (2015) attribute the emergence of many of the European new radical parties to the economic crisis that left many Europeans punish the incumbent parties and turn to new alternatives. They also emphasize that it is often precisely the group of globalisation losers that are targeted as potential new clients of these emerging parties (also Oesch, 2008). Hobolt and Tilley (2016) argue in a similar direction. According to them, disappointment over the mainstream parties’ politics, whether in office or opposition, is likely to be the highest among economically disadvantaged voters. These dynamics would have left the vote shares of mainstream right parties dropping by six and mainstream left parties by three percentage points if elections were held in 2014, alone compared to the last elections. Since the outbreak of the crisis in 2008, mainstream parties have lost 12 percentage points of their votes, mostly to right- and left-wing alternatives.

It is however not only the 2008 global financial crisis and the Eurozone crisis that lead to this dramatic transformation. Kriesi et al. (2008) found that the globalisation processes themselves had already begun transforming the political landscape in Western Europe by creating a new integration-demarcation cleavage, dividing the electorate into globalisation winners and losers. This demand side of the formula then meets the supply side, the far right, according to Golder (2016), the fastest growing party family in Europe. Golder further presents a good overview of the different movements within the far right party family, which is far from homogenous. There are certain tendencies, though, that can be found in most of them. Euroscepticism literature offers some helpful insights here. Paul Taggart (1998) for example analysed various European polities and party systems. While mainstream parties largely supported European integration, resistance to it was found almost exclusively in the peripheries of the political system. Although mainstream left parties might oppose some capitalist and free trade aspects and conservative parties might not agree with all environmental policies, the coherence with their ideological backgrounds and the positive aspects of the remaining policy areas mostly lead to supporting positions. But on the far left and the far right of the spectrum, the opposition to integration is by no means exclusively built on their ideological views but also chosen strategically (Taggart, 1998; Marks & Wilson, 2000; Kopecky & Mudde, 2002). Another contribution is made by Kuhn, van Elsas, Hakhverdian, and van der Brug (2016), who argue that higher levels of inequality lead to increases in Euroscepticism. Of course, this is only

(17)

applicable to European countries, nevertheless it shows that blaming foreign powers and processes on the international arena for domestic problems is not farfetched. The results of unstoppable developments, such as technological advance and automation, that certainly have negative effects on economic security of low-skilled workers in developed economies are likely to be framed as a problem of free trade and globalisation, since those processes can be stopped, at least according to the rhetoric of the new right. Although the main drivers behind electoral success of right wing populist parties are most likely to be found in cultural grievances (Oesch, 2008; Kriesi et al., 2008), protectionist policies are quite popular among protest and populist parties as well.

Argument and Hypotheses

The literature review presented above has a number of implications for expected impacts on democratic polities around the globe. In the following chapter, predicted outcomes and causal mechanisms will be presented to merge the various literature and research strands into one coherent theoretical argument, from which the hypothesis for the subsequent analysis will be deduced.

Based on the literature introduced in the earlier chapters, it is expected that in developed economies, in which capital and high-skilled labour are the abundant factors of production, while, depending on the factor mobility, it is either (low-skilled) labour throughout all sectors or all factors of production deployed in low-skilled labour intensive industry sectors that face the most severe negative consequences (Brawley, 1997; Hiscox, 2001; Beaulieu, 2002). Either way, one can expect that most globalisation losers in these countries will be found in the low skill labour segments.

Instead of using economic insecurity as main predictor, this thesis advances using income inequality. There are two reasons why this approach is chosen. First and foremost, the author expects a shortcoming in the argumentation line of the compensation hypothesis. The argument goes that scholars mainly focused on the side of globalisation losers and therefore largely discussed the topic from a perspective of absolute scales. A large number of articles in the field of psychology focussing on equity theory have shown that comparative scales, i.e., unequal returns, must be taken into account as well (Adams, 1965; Mowday, 1991; Sweeny, 1990). This even holds true for some of our most closely related species, most famously presented by Brosnan and de Waal (2003) in their experiment on unequal returns, using capuchin monkeys. According to this approach, the dissatisfaction of the ‘globalisation losers’ does not per se increase due to economic hardship. Rather than that, it is their comparative economic loss that increases dissatisfaction, e.g. income stagnation in times of substantial economic prosperity. Secondly, it has been argued that inequality poses a significant threat to democracy and might have given rise to populist parties, movements and

(18)

political entrepreneurs (Remmer, 1991; Jones, 2001; Dalton, 2008). Despite the frequency of this claim, there is an agitating gap in the previous literature with regards to quantitative analyses. This might partially stem from the, until very recently, low availability of cross-country comparable income inequality data, as will be outlined in more detail bellow. Reliable national inequality data is scarce as well and would more likely allow for case study based research. This thesis will contribute

to filling this gap. Figure 1 graphically summarises the literature analysed above. Based on this, the author deducts his hypotheses. Hypothesis 1 (H1) builds the overall expectations of the proposed theoretical argument, visualised in Figure 1, and can be regarded as the main hypothesis of this thesis:

H1: In developed democratic economies, growing income inequality increases the level of parliamentary polarisation.

The middle section of Figure 1 shows the two proposed causal mechanisms. The first one suggests that the loser of globalisation processes might increase their preferences for welfare state intervention. This basically picks up the propositions of the compensation hypothesis discussed above. However, the primary reason for this is not economic insecurity per se, but the comparative disadvantage they suffer due to economic integration into global markets. The author suspects that dissatisfaction with the personal economic situation is particularly strong among this group of individuals if the overall macroeconomic numbers are strong and in glaring contradiction to the individual circumstances. This causal mechanism could be called the symptom-oriented curing preference. Based on the literature above it can be stated that these preferences are best represented

Globalisation fuelled Inequality

Globalisation Economic _Inequality

Policy Preference: Welfare State Policy Preference: Protectionism Vote Share: Left Parties Vote Share: Far Right Parliamentary Polarisation Independent Variable Causal Mechanisms Dependent Variable Vote Share: Far Left

(19)

among parties of the social democratic and socialist party families on the economically left side of the spectrum. It is those families that historically represented the workers’ side of Lipset and Rokkan’s (1967) capital-labour cleavage.

The second mechanism works with a cause-oriented path. Instead of favouring policies that treat the effects of globalisation – in this case increasing inequality to the disadvantage of low-skilled workers in developed economies – they could also turn to globalisation and free-trade opposing parties on the far right and the far left. Although it is often argued that far right parties tend to mobilise more on the new politics or cultural axis, protectionist policies, either in the form of opposition to immigration, but also as trade barriers and subsidies for low-skill labour intensive industries (agriculture, mining, etc.), are not uncommon among the programmes of far right parties either (Oesch, 2008). And as argued above, protest parties also position themselves strategically. On the far left, strategic and ideological opposition to trade integration overlap. From these lines of argumentation, we then deduce the second hypothesis (H1.1):

H1.1: Higher levels of income inequality increase the probability to vote for far right, left, or far left parties among the low-skilled workforce.

This mechanism might even be reinforced by the findings of Kedar (2005). She argues that voters often cast their vote for parties that represent more extreme views and policy preferences than the voters themselves, or as Adams and Merrill (1999) described it: “One of the most discussed findings from the literature on political representation is that political parties and candidates typically present policy positions that are similar to, but more extreme than, the positions of their party supporters” (p.765). This compensational mechanism is especially well-marked in representative polities, in which power-sharing takes place to a bigger extent than is the case in majoritarian systems. Duverger’s Law further states that majoritarian systems tend to lead to two party polities, while proportional representation erases the necessity to maximise a party’s vote share (Duverger, 1954; Riker, 1982). Therefore, we would expect this effect to be stronger with a higher number of parties in a polity.

An argument that goes hand in hand with the claim of Kedar is strategic party positioning and adjustment. According to the spatial model of elections, parties do change their positions in reaction to changes in voters’ policy preferences and other parties’ successes (Adams, 2012). It is not surprising to expect mainstream parties closer to the centre to adapt their preferences if peripheral parties gain vote shares, such as suggested above. Kriesi et al. (2008) find support for this claim in six European polities, however with certain differences with regard to shifts in the cultural or

(20)

economic dimensions. If parties on the right win, the losing mainstream party adjusts towards the right pole as well. The same should apply on the left side. Adams et al. (2006) also find evidence that mainstream parties do adjust their positions as reaction to shifting preferences of the electorate. Another supporting piece of literature was written by Pontussen and Rueda (2010). Although depending on the level of low-income voter mobilisation, left parties are found to shift leftwards with growing inequality, in light of shifting voter preferences towards more welfare state intervention. At the mean voter turnout level of their sample, albeit not statistically significant on the 5% level throughout all models, the coefficient already points towards more leftist party positions with growing inequality, an effect whose intensity and statistical significance grow with increasing voter turnout. The author therefore suggests the third hypothesis:

H1.2: Higher levels of income inequality lead mainstream parties to strategically adjust their party positions towards the poles before elections take place.

Albeit there are numerous logically stringent and empirically supported arguments substantiating the hypotheses, there are two possibly confounding factors that have to be mentioned. First of all, one might argue that because the overall effects of globalisation are positive, so is the number of people that are better off. Workforce situated in the third or fourth income quintile, improving their income situation and thus shifting their political preferences away from the welfare state and interventionist left, towards centrist and liberal parties. The aggregate effect on the vote shares of left parties (and therefore also on the incentive for strategic adjustment) could then turn out to be zero or even negative. In contrast to this, it is argued here that voters are more sensitive to losses than to gains. Support for this claim comes from Nannestad and Paldam (1997). Using a time-series data on Denmark between 1985-1992, they show that negative economic developments translate into electoral punishment much more drastically than is the case for positive change.

The second possible constraint stems from a possible lack of mobilisation potential that could coincide with economically disadvantageous personal situations. Political alienation is most spread among the poorest and lowest educated in society (for an overview see Lijphart, 1997). The socioeconomic status (SES) of an individual is generally regarded as a good indicator for the probability of his or her participation in the political process, i.e., voting (Lazarsfeld, Berelson, & Gaudet, 1949; Kleinhenz, 1995). The SES comprises factors such as income, occupational status and education. Higher inequality would at least drive down the incomes in the already worse educated segments of society, which possibly also results in a higher level of structural unemployment.

(21)

Consequently, it could be argued that the main drivers behind this thesis’ theoretical argument, the economically disadvantaged, could slide into the politically alienated segments of society, rather than changing their preferences and subsequently also expressing this change in the voting booth. On the other hand, the author believes that while the SES might be a good indicator for meso-levels of political participation, he argues that changes in income or occupational status of an individual have a different result in the short term. Rather than adapting to the behaviour of the new SES group, these changes in the personal economic situation might in the short term be a potential mobilisation factor. The author therefore cautiously formulate hypothesis 2:

H2: Higher levels of income inequality increase the voter turnout among the low-skilled workforce in the short term.

After having deducted these four hypotheses, graphically summarised in Figure 1, the next section discusses the data used to test them.

Operationalisation and Data

The choice of methods is one of the central elements of every academic analysis and depends mainly on the research question as well as on the availability of data, time and resources. The approach chosen for this thesis is a quantitative one. Although quantitative and statistical analysis are often still regarded as being superior, especially in economically shaped fields of social science, this is a misperception. Rather than that, numbers often tend to give a false feeling of truth and objectivity, something they only provide if they are backed and supported by a well-founded theoretical framework and solid data measurement. Especially the latter is very often problematic. Availability of qualitatively sufficiently high and large datasets is very often a major obstacle in social sciences. Furthermore, cross-country comparability can also not always be guaranteed, be it because of non-transparent and different methods, concepts and definitions used by national statistical agencies or even worse, deliberate deception with wrong information, as in the case of Greece’s reports of its household figures to the European Union (European Commission, 2010). Not only were Greek methods insufficient for European standards, the reported numbers were also intentionally adjusted to assure the acceptance of Greece in the Eurozone.

Generating large-N data is of course also extremely time-consuming and expensive, even more so when one needs to cover multiple nations. In this case though. as will be described bellow, data is available in a sufficient quality and comparability that allows for a statistical analysis, a method, that

(22)

is in this case superior, given the advanced theoretical discussion in the matter. Election studies are generally speaking a very complex field of research and there is an almost inexhaustible number of variables to potentially play a role on the micro and macro level. In the realm of economic voting, the processing of large-N data makes sense and potentially uncovers cross-border valid mechanisms that influence election results independent from the polity. Before moving on to the modelling and analysis section, one needs to clearly define the used concepts and means of measurement. This section will explain the conceptualisation and introduce the sources of the data used for this thesis. At this point it is important to note that due to limited resources and for the sake of cross-country comparability, compromises are necessary. We will begin this section with the explanatory variable, moving to the control variables and the dependent variable.

Inequality

There is a big variety of inequality concepts in use in today’s research. One could start with a couple of very simple empirical concepts, such as the range of income, deviation, or the coefficient of variation. Of course, these simple descriptions tell us little about the distribution of income in a society. The two most used concepts in social sciences are probably the Theil index and the Gini coefficient. While the former is more sensitive to changes around the extremes, the Gini coefficient better captures changes around the mode (Milanovic, 2007). Furthermore, Gini is naturally bound between zero (perfect equality) and one (maximum inequality, i.e., all income concentrated in one person) and Theil has no natural upper bound and is only limited by log N (Allison, 1978). Although both might capture interesting insights and could be valuable to this thesis, it is data availability, which finally decides to the benefit of Gini coefficient data. An example of insufficient data availability is Smeeding’s (2002) article on globalisation and inequality. His main problem for achieving reliable and precise predictions and trendlines for income inequality was the lack of cross-country comparable data.

Although there are various ways to calculate the Gini (Ceriani & Verme, 2012), what is generally done when calculating it, is comparing all available pairs of incomes in a society. Although the Gini coefficient is widely used as a measure for income inequality, for example by the Eurostat, the OECD, and the World Bank, most of these datasets contain only a very limited amount of country-year observations. The OECD data for example, covers only the country-years between 2003 and 2014, and even in this rather short range, the observations are not complete (e.g., data points for Norway: 2004, 2008-2013 (OECD, 2016)). The same situation applies to the Luxemburg Income Study (LIS), which, in terms of comparability, has the highest standard in the field (Solt, 2016).

(23)

This thesis uses Frederik Solt’s (2016) Standardized World Income Inequality Database (SWIID). Solt addressed the issue of data availability by constructing his data points in the SWIID. Although there have been attempts before (most notably by Milanovic, 2014) they either lack observations or comparability due to the usage of different data sources or working with different concepts and definitions. Solt avoids this by estimating his data. To do so, he first collected source data, which he categorised systematically into 11 categories, according to the respective definitions used. He added these categories to the two existing LIS categories, market and net Ginis, describing pre and post tax and transfer settings. With this existing set, he was then able to calculate the ratios between the categories and estimate the missing data. To account for the uncertainty of his estimations, Solt generated each variable 1,000 times using a Monte Carlo simulations and applying a moving-average algorithm to each simulation. From the 71 new country-year observations published by the LIS since 2008, only five showed substantive and significant differences to the data points generated for the SWIID. Arguing that the respective countries were exclusively found in the developing world in countries for which no data existed prior to the update, Solt manages to credibly explain those differences and convincingly argue that the quality of his most recent versions has increased significantly. Summarising his 1,000 estimations for each observation produces cross-country comparable net and market Gini coefficients including standard errors for 174 countries and over 4,600 country-years. The distinction between net and market inequality measures is of course central to this thesis, since governmental social policies can and do absorb some of the inequality shocks that partially originate in globalisation (Smeeding, 2002).

By using change in net Ginis instead of market Ginis, the statistical models applied here already take into account existing redistributive policies and therefore country differences. Furthermore, following the argument elucidated above, we would not expect frustration based on low wages but based on a low income. If an economy manages to distribute the gains from trade equally among the

members of society, social fragmentation and political polarisation would not result, according to the theoretical argument. The data for the country year observations in the final sample of this thesis is of comparatively high quality, as is shown in Table 1. Standard errors for net Gini coefficients of over 1 are exclusively to be found in three observations for Portugal, for market Ginis over 2 only in

Comparison of Standard Errors (SE)

Max. Gini Net SE Mean Gini Net SE Median Gini Net SE Max. Gini Market SE Mean Gini Market SE Median Gini Market SE SWIID 5.1 Total 8.0156 1.5230 1.3894 21.93363 2.21830 2.02845 SWIID 5.1 Sample 1.3959 0.4841 0.4075 2.88279 0.89949 0.70305

(24)

Portugal as well as Germany 2013 (with the Ginis presented as percentages). Generally speaking, one can be confident that data is of sufficient quality to conduct the proposed research.

Micro-Level Data

Micro-level data can be accessed via the Comparative Studies of Electoral Systems (CSES) project. The CSES is a collaboration between various election studies from all over the globe in an effort to create cross-country comparable post-election datasets (CSES, 2017). The result are four modules, each having produced datasets covering 29 to 50 elections and comprising between 50,000 and 80,000 observations on the individual level. The project started in the 1990s and coverage started in 1996. The survey-based data sets cover a broad variety of indicators and variables, from gender, age and marital status to voting behaviour in the most recent election, party sympathies to union

membership, educational level and income quintile of the respondents.

The data also includes categorisation of the competing parties on a left-right scale, based on participating expert interviews. However, the explanatory notes on this variable are rather poor. Even David Howell (personal communication, 21 May, 2017), Director of Studies at CSES, noted that on this matter, the data is far from ideal because CSES does not actually know who the experts behind the judgements are and on what basis the judgements were made. Furthermore, one must assume that this dimension is mainly based on economic policy preferences of the parties included and, although economic mechanisms are the main focus of this thesis, the party categorisation must take more aspects into account. While the left side is rather less problematic, especially the parties of the new right are not consequentially economically liberal or opposing state intervention. While this might have been the case during the rise of the far right in the 1980s and 1990s, de Lange (2007) shows that the far right – as a general trend since one can not talk about a coherent party family here – has positioned itself more centrally on the economic dimension since the beginning of the new century. Last but not least, there is also an issue of practicability. Unfortunately the variables are coded so inconsistently that a one-on-one analysis is impossible and recoding of each country year observation would be necessary in any case. The qualitatively better and more feasible solution is therefore to rely on the Chapel Hill Expert Survey (CHES, 2016), which covers party classifications in a number of European states as well as their election successes between 1999 and 2014 (Polk et al., 2017). The parties are classified both on a classic left-right dimension as well as on a new politics dimension, which ranges from a green, alternative and libertarian pole (GAL) to a traditional, authoritarian and nationalist (TAN) one (Hooghe, Marks, & Wilson, 2002; Kitschelt, 1995). Additionally, a general left-right dimension combines the two. These classifications are also amended by a large number of party positions towards specific policies. The disadvantage of using this data is the limited

(25)

geographic coverage (i.e., mainly European Union member states) and its temporal limitations. More arguments for this decisions will be presented in the section on polarisation.

With these limitations, the dataset is reduced from an original 250,000 observations to 83,000 from a total of 20 countries. This is unfortunate but necessary if one aims to work with high quality data only. Table 2 gives an oversight over the included elections and the distributions of the observations.

Macro-Level Data

Basic macroeconomic data was accessed mainly via The World Bank (2016). The World Bank collects a great variety of data from different sources, such as national statistical bureaus or the International Labour Organisation (ILO). What distinguishes The World Bank from other sources such as the EU or the OECD is their bigger range of available data. The earliest country observation in the OECD unemployment data for the Netherlands can for example only be found in 2000, which would at least exclude the 1998 election (at least because of the potential inclusion of a lag or a moving average). In addition to unemployment and youth unemployment, the constructed data frame

# Country Elections Observations

1 Austria 2008 2013 2131 2 Belgium 1999 2003 6037 3 Czech Republic 2002 2006 2010 2013 6038 4 Denmark 1998 2001 2007 5272 5 Estonia 2011 966 6 Finland 2003 2007 2011 3702 7 France 2002 2007 2900 8 Germany 1998 2002 2005 2009 2013 10218 9 Greece 2009 2012 2038 10 Hungary 2002 768 11 Ireland 2002 2007 2011 5359 12 Italy 2006 1139 13 Netherlands 1998 2002 2006 2010 7774 14 Poland 2001 2005 2007 2011 7782 15 Portugal 2002 2005 2009 6431 16 Slovak Republic 2010 1178 17 Slovenia 2004 2008 2006 18 Spain 1996 2000 2004 2008 3930 19 Sweden 1998 2002 2006 3702 20 United Kingdom 1997 2005 3672

(26)

includes GDP growth rates, inflation rates as well as the change of these variables versus the last year.

However, since one might also be interested in some indicators for globalisation, the data also contains three variables, indicating economic, political and social globalisation, developed by the Swiss Federal Institute of Technology Zurich (Dreher, 2006; Dreher, Gaston, & Martens, 2008; KOF, 2017). The economic globalisation indicator is built on actual flow data of trade, FDI, cross-border income and portfolio investments as well as on various economic restrictions, while social globalisation is measured by data on personal contacts and information flows.

Lastly, the data frame includes some effective numbers of parties in both the electoral and the legislative phase to control for effects connected to the median voter theorem and an effectively very low number of potential choices (Gallagher & Mitchell, 2008). Furthermore, the inclusion of this variable takes the finding of Sartori (1976) and Kedar (2005) into account. Sartori showed that a high number of parties in a polity tends to be an indicator for centrifugal forces that drives parties away from the centres and can lead to higher polarisation. Kedar broke this down to the micro level. She finds, that instead of platform-oriented, people do in fact vote outcome-oriented. This means that the electorate considers the institutional settings of their polities, which are responsible for the post-electoral bargaining processes. In a majoritarian system (with a low number of parties) for example, the winning party implements their political programme without many constraints, whereas in proportional systems, compromises must be found within the winning coalition. The voter anticipates this and in the latter case, votes more extreme than his or her actual preferences. The number of effective parties is therefore an indicator for a greater range of options (which, according to Dalton (2008) already leads to higher polarisation among the voters), but also controls for strategic voting in order to reach the preferred outcome.

Polarisation

The operationalisation of polarisation of political ideology is quite tricky and there is no consensus on this matter in the academic literature. Brühlmann and Freitag (2006) for example use absolute distances of competing parties in order to analyse its effect on voter turnout. This approach is of course not applicable here since this thesis aims to measure factual polarisation in parliaments. Remmer (1991) simply measures the support for small and extreme parties close to the poles. But again, this does not show what happens in between, i.e., if the mainstream and more centrist parties adjust, either by moving closer to the poles or by clearly cementing their positions as moderate parties. There are more possible ways of operationalisation and a very interesting overview of the varieties of concepts, their advantages and shortcomings is summarised and presented by Schmitt

(27)

(2016). As he notes, there are not only different approaches to the operationalisation but also to the question of how to evaluate a party’s ideological position. He names three common strategies: survey based, manifesto based and expert based. Dalton (2008) for example proposes the use of voters’ perceived party positions from the CSES data (survey based approach). Again, this does not suit the purpose of this thesis because the voter does not objectively assess the positions but also takes the salience of certain topics into account. Leimgruber et al. (2010) alter this survey based approach by including answers from candidates of political parties and thus construct two dimensions based on both candidates and voters. The most known approach of a manifest based approach is probably The Manifesto Project (MARPOR). The author chose the third approach by using data of the Chapel Hill Expert Survey (CHES). Compared to survey based data, this approach is more objective because it uses standardised criteria and is based on experts who take into account more data than what is generally visible to the public. The main advantage compared to a manifesto based approach lies within its use of practical data. How a party behaves in parliament or government might significantly differ to what it writes in its manifesto.

Based on the deliberations of Schmitt (2016), the author decides to use a polarisation concept that measures the weighted distance of the weighted mean of a polity’s party positions, i.e., its ideological centre. For this, a variance based as well as a standard deviation based indicator is calculated for each election and each dimension (general left-right position, economic left-right position, and new politics GAL-TAN position). The author further excluded all elections for which less than 90% of the total votes and/or 90% of the total seats are captured by the CHES.

Using only the elections included in the CSES data reduces the data frame to only 48 elections, which clearly fails to reach the proposal of Tabachnick and Fidell (2007), according to which a regression must include not less than 50 + 8m observations, where m indicates the number of predictors. For this reason, all available data points from the CHES are included in a separate data frame to test H1. This enables us to create a panel data frame including 87 data points. Even though this is still not ideal, the original number of observations is almost doubled.

Descriptive Statistical Analysis

Before conducting the statistical analysis, one should first analyse the final dataset at hand. As a very first step, we take a look on the assumption developed in the theoretical section and albeit not central to the causal mechanism, still of great importance to the overall question: Does globalisation lead to higher inequality in states included in this sample? The very short answer is that the data used here cannot really confirm this. While there is an overall positive correlation between the change in the level of globalisation and the change in the level of (market) inequality, the decomposition of the

An eroding centre : implications of globalisation and inequality for the existing political order