Anti-austerity voting in an era of economic crisis: Regional evidence from the 2015 referendum in Greece

Abstract

The austerity measures and structural reforms that Greece had to undertake since 2010 in exchange for financial aid divided the Greek political system into pro-austerity and anti-austerity camps. These divisions reached a climax with the July 2015 referendum. The paper attempts to assess the extent that to which the geographic patterns that emerged in voting were due to the differentiated economic regional impact of economic crisis. Using economic voting theory as a benchmark, and employing spatial econometric methods, the study contributes to a largely ignored topic, known as geographies of economic voting, providing new and valuable insights for an under-researched time period.

Keywords

2015 Greek referendum economic voting economic crisis regions spatial econometrics

Introduction

The most prominent ‘victim’ of the global financial crisis, erupted in late 2008, was Greece. During the period 2009–2015, Greece’s real gross domestic product (GDP) and disposable income have shrunk more than one-quarter and one-third, respectively. Its unemployment rate has nearly tripled, exceeding 26%, while the employment rate has reached 50%, a record low. Moreover, the at-risk-of-poverty or social exclusion rate has risen to alarming levels from 28.1 to 36.0% while several other well-being indicators related, inter alia, to health, education and social services have sharply deteriorated (Eurostat, 2016).

The effects of the crisis, however, go far beyond economic and social dimensions, generating important and far-reaching political consequences. The austerity measures and structural reforms that the country had to undertake, in exchange for financial aid through several bailout agreements (Memorandums of Understanding (MoUs)), with the European Commission, the European Central Bank (ECB) and the International Monetary Fund (IMF) (the ‘Troika’) divided the Greek party system and the electorate into pro-austerity and anti-austerity camps. At first, in the elections of 2012, the aggregate electoral share of the two dominant and pro-austerity parties (the centre-right New Democracy (ND) and the centre-left Panhellenic Socialist Movement (PASOK)) for almost four decades dropped more than 50% compared to the previous (pre-crisis) national election of 2009 (see Table 1). Three years later, in January 2015, the anti-austerity camp finally gained office, with the victory of Syriza (Coalition of the Radical Left), which formed a coalition government with the right-wing, nationalist-populist ANEL (Independent Greeks), founded in February 2012 by former ND members of parliament (MPs) who were against the MoUs. Soon after, in July 2015, the Greek people again voted against austerity in a referendum. The great electoral influence of the anti-austerity camp since 2012 and its dominance in 2015 dramatically changed the political landscape and signalled important societal and political divisions highlighting the need for further scientific research.

Table 1.

Greek election results, 2009–2015, and aggregated results of anti-austerity vote sharea (percentage of vote shares).

	2009	May 2012	June 2012	2014a	January 2015	July 2015 Referendum
Syriza	4.6	16.8	26.9	26.6	36.3
ND	33.5	18.9	29.7	22.7	27.8
GD	0.3	7.0	6.9	9.4	6.3
The River	–	–	–	6.6	6.1
KKE	7.5	8.5	4.5	6.1	5.5
ANEL	–	10.6	7.5	3.5	4.8
PASOK	43.9	13.2	12.3	8.0	4.7
KIDISO	–	–	–	–	2.5
DIMAR	–	6.1	6.3	1.2	0.5
Other (anti-austerity)	–	6.8	1.8	3.7	2.6
Other (pro-austerity)	–	9.4	3.2	0.9	1.0
Other	10.2	2.7	0.9	11.3	1.9
Total	100	100	100	100	100
Total anti-austerity	–	49.7	47.6	49.3	55.5	Nob 61.3
Other	–	50.3	52.4	50.7	44.5	Yesb 38.7
Turnout	71.0	65.1	62.5	59.3	63.9	62.2

ANEL: Independent Greeks; DIMAR: Democratic Left; GD: Golden Dawn; ND: New Democracy; KKE: Communist Party of Greece; KIDISO: Movement of Democrat Socialists; PASOK: Panhellenic Socialist Movement.

Note: Authors’ calculations. Grey shaded lines represent parties with anti-austerity stance in the referendum.

^aEuropean Elections.

^bTo the bailout agreement proposed by the ‘Troika’.

The paper aims to examine the determinants of the July 2015 referendum outcomes in Greece at the regional level, focusing on anti-austerity voting shares. Specifically, the study attempts to assess the extent to which the geographic pattern that emerged in the referendum was due to the differentiated regional impact of the economic crisis. For this reason, economic voting is selected as the main theoretical tool, in a period that economic issues are more salient and visible, playing a more critical role in voting behaviour (Kostadinova and Giurcanu, 2015). The analysis is conducted at the regional level because the role of regions, and space, have been found to be greatly influential to the understanding of political behaviours and processes, especially in voting patterns (Agnew, 1996; Darmofal, 2006; Lacombe et al., 2014). Additionally, the analysis focuses on the regional level because the ramifications of the crisis have proved anything but spatially uniform (Martin, 2010). Unevenness over space is not only one of the defining characteristics of capitalist development (Harvey, 1982) but is also a critical characteristic of periodic crises, which undoubtedly characterise the capitalist development (Hadjimichalis and Hudson, 2014). The spatial heterogeneity of the crisis has also been confirmed for Greece by several recent studies (Artelaris, 2017; Artelaris and Kandylis, 2014). Greece is a useful case to examine economic voting because the attribution of responsibility for economic policies is very clear as the power of local/regional governments is very weak; the unclear lines of responsibilities in a hierarchical political system make the economic voting being less pronounced (Hellwig, 2008).

The paper contributes in several ways. First, the paper contributes to an under-researched area and a largely ignored topic, especially by regional scientists and geographers; as Ragusa and Tarpey (2016) have put it, the so-called geographies of economic voting have received little attention and are far from settled. The results of the study highlight the need for a closer integration between political and economic science, and regional science and spatial econometrics. Second, given the scarcity of studies investigating the economic voting theory at the regional level, especially in times of crisis, this study fills an important gap in the international literature. Third, this study shows, for the first time to the best of our knowledge, the importance of two new variables in voting behaviour: regional poverty rate and bank deposits. Last, it casts light on the issue of voting in referendums, which has received growing attention and interest, especially in connection with the evolution of the European Union.

The political context in Greece during the economic crisis

The Greek crisis started in late 2009, triggered by the global financial crisis of 2008 and the subsequent wider economic recession, its inability to control its budget deficit as well as inherent problems related to the Economic and Monetary Union design. In May 2010, the PASOK government reached a €110bn bailout agreement with the ‘Troika’ to avoid a default, in exchange for a programme of a prolonged fiscal consolidation imposing harsh austerity measures and drastic structural reforms. From that moment on, the agreement became a central issue in the Greek political arena; all opposition parties, including the major opposition party, ND, voted against the agreement, except for a populist radical right party, LAOS (Popular Orthodox Rally), which voted for the agreement.

Eighteen months later, the Greek Prime Minister (PM) Papandreou asked for a vote of confidence and proposed a referendum in order for a second bailout agreement to be approved. Although his government got the vote of confidence, extreme international pressure for the formation of a national unity government,¹ which would secure the implementation of the new austerity package and avoid a referendum, led to Papandreou’s resignation. Subsequently, PASOK, ND and LAOS formed a coalition government, with Papademos, a technocrat and former president of the Greek central bank and former vice-president of the ECB, as PM. Obviously, the participation of ND in the government marked a change in the party’s stance, which was now in favour of austerity measures. The new government negotiated the second MoU, which was passed by the Greek parliament in February 2012.

In May 2012 a national election was held, in which no party got more than 19% of the votes, in a country where the two first parties usually got around 80% of the vote. ND was first with 19%, Syriza was surprisingly second with 17% and PASOK, which in the previous election of 2009 had got 44%, was third with 13%. Νo government was formed from the May election, so the ‘earthquake’ was completed with the election of June 2012, because in this new election, a total of seven parties entered parliament: ND and Syriza, which increased their vote shares by around 10% each; PASOK; the reformist left DIMAR (Democratic Left), founded in 2010 after a split from Syriza, ANEL, KKE (Communist Party of Greece) and the neo-Nazi Golden Dawn (GD) (see Table 1). Finally, a pro-austerity coalition government was formed comprising ND, PASOK and DIMAR, with Samaras, the president of ND, as PM.

The 2012 elections marked a peak of fluidity, dealignment and fragmentation for the Greek party system, with high levels of volatility (Teperoglou and Tsatsanis, 2014; Vasilopoulou and Halikiopoulou, 2013). They also marked a process of structural transformation of the Greek party system, from classic two-partyism with a single peak of the distribution of voters along the main ideological dimension, towards a bimodal system with two peaks, each near the two extremes of the dimension (Dinas and Rori, 2013). Moreover, the twin elections of 2012 signalled the consolidation of the pro-austerity/anti-austerity division, which also incorporated stances in favour or in criticism of Greece’s membership in the EU (Vasilopoulou and Halikiopoulou, 2013). The pro/anti-austerity division, having as its main content the favourable or critical stances towards the EU, was also the dominant division amongst Greek political elites (Tsirbas and Sotiropoulos, 2015).

The second MoU (and loan agreement) was due to expire at the end of December 2014 but the ND-PASOK government (DIMAR had left the coalition in 2013) failed to meet the Troika’s requirements and pass the evaluation procedure in order for the last instalment to be sent. Hence, the government asked for and secured a two-month extension of the agreement and immediately initiated an early presidential election process. The failure of the Greek Parliament to endorse the coalition government’s candidate² for president led to early parliamentary elections in January 2015 (for a detailed account see Tsirbas, 2016). Syriza won the election with 36.3%, ending more than four decades of dominance of PASOK and ND, and, from this perspective, representing a fundamental change in Greece’s political landscape. At the same time, PASOK, the political party that had played the most significant role since 1981, recorded the lowest electoral share in its history, with less than 5%. In essence, the Greek electorate promoted a different response to the crisis, after five years of the implementation of austerity policies by several consecutive governments, and expressed its will for a total change of the governing paradigm. A Syriza–ANEL coalition government was immediately formed, with Tsipras, the president of Syriza as PM. The seemingly unorthodox government coalition between a radical left-wing and a nationalist populist right-wing party can be explained in terms of new dimensions of conflict that had emerged after the outburst of the economic crisis, which are defined by stances towards the European Union (see Tsatsanis et al., 2014; Tsirbas and Sotiropoulos, 2015; Vasilopoulou and Halikiopoulou, 2013). In terms of these issue dimensions, Syriza and ANEL both stood towards the anti-EU and the anti-bailout end. The new government started prolonged and tough negotiations with Greece’s lenders for a new agreement. Six months later, in late June 2015, a proposal was temporarily made to Greece, entailing a new austerity measures package in exchange for further economic assistance. Greece, at that time, was in urgent need of funding and was heading towards a default, once again.

Then, in the early hours of 27 June, the Greek PM, Syriza’s leader Alexis Tsipras, decided to proclaim a referendum to be held only eight days later, on 5 July, in order for the Greek people to accept or reject the proposed package. The parties that supported the ‘NO’ vote were Syriza, ANEL and GD, as well as several small parties from the Left not represented in parliament. Although there were various meanings attached to the referendum’s dilemma by the ‘no’ camp, the framing of the referendum dilemma was that it was the zenith of democracy and a ‘NO’ vote would mean the end of austerity by providing extra negotiational power to the government. The parties that called for a ‘yes’ vote were ND, PASOK and To Potami (The River), a newly formed liberal, pro-European party, comprising former MPs, mainly from DIMAR, under the leadership of a famous journalist, as well as some minor neoliberal, non-parliamentary parties. They framed the referendum as a vote in or out of the Eurozone, accusing the Syriza–ANEL government of populism and of jeopardising Greece’s position in Europe. The Communists argued that the referendum was about a false dilemma and called for invalid ballots.

The result was a landslide victory for ‘no’ to the austerity package with 61.3%. ‘No’ won in every one of the 56 electoral constituencies but with large variation throughout the country, as can be seen in Figure 1. In essence, the outcome was largely a confirmation of the previously described divisions identified in the January 2015 election and of Syriza’s dominance, as well as the dominance of anti-austerity sentiments amongst the Greek electorate (Tsatsanis and Teperoglou, 2016). However, one week later, ‘no’ became ‘yes’ in the sense that the referendum result did not much help the Greek PM in ending austerity. A third bailout agreement and harsh austerity package were agreed upon with Greece’s lenders on 12 July 2015. In the subsequent snap election of September 2015, Syriza won once again and formed a coalition government with ANEL again.

Figure 1.

Regional differences of ‘NO’ vote in the 2015 referendum. Note: Authors’ calculations.

Theoretical and empirical overview

The search for determinants of voting behaviour amidst a serious economic crisis leads to economic voting as a main theoretical framework. There are some good reasons to believe that in times of crisis economic issues are more salient and visible, playing a more critical role in voting behaviour (Kostadinova and Giurcanu, 2015). For instance, in these eras, economic issues tend to receive more attention from voters and greater coverage by mass media.³

In the economic voting field, an impact of the economy on voters’ choices is assumed. The central idea behind economic voting is the responsibility hypothesis, namely the fact that electorates tend to hold governments responsible for the economic situation in a given country (Alesina and Rosenthal, 1995: 4; Lewis-Beck and Paldam, 2000; Nannestad and Paldam, 1994). The responsibility hypothesis is linked to the ‘grievance hypothesis’, namely the fact that electorates respond more to negative developments than to positive ones (Lewis-Beck and Paldam, 2000) or, as Kahneman and Tversky (1979) put it, people attach greater importance to losses than to gains. Moreover, people tend to make decisions on relative gains and losses rather than on absolute levels of utility, as classical political economic models would assume (Wayland, 1998). In other words, people are more prone to evaluate changes than absolute values.

Economic voting also manifests itself at the regional level for a great many of reasons. For instance, voters can form views of their regional contexts because of the direct (e.g. general impression of the strengths and weaknesses of the regional economies) and indirect observation and experience (views and opinions of others, conversations, local media) (Pattie et al., 2015). Although the role of ‘intermediate’ geographies on economic voting has been long recognised (Kramer, 1983), the literature is scarce leading to a ‘spatial gap’ that can be explained by, at least, three factors (Ragusa and Tarpey, 2016): (1) the field has been dominated by political scientists and economists, (2) the economic voting literature has focused largely on ‘valence’ theories which have minimised other questions and (3) this kind of analysis requires data at sub-national level, usually more difficult to be obtained.

In empirical terms, at the national level, unemployment rates and GDP changes have been recognised as the most critical macroeconomic determinants of voting behaviour (see Lewis-Beck and Paldam, 2000). Yet, the empirical findings are not unequivocally characterised by instability, both in terms of time and geography. In relation to the punishment argument, evidence from 55 countries during the period 1960–2012 shows that economic crises affect negatively the electoral performance of incumbents (Vasselai and Mignozzetti, 2012). Furthermore, all governing parties in Southern Europe have suffered great losses and have lost power after the crisis (Bosco and Verney, 2012). According to a study of 30 European countries up to October 2014 (Hernandez and Kriesi, 2016), the increase in economic misery is associated with the worsening performance for incumbents. This decline is even greater in hard-hit countries and in their second post-crisis election.

In the Greek case, besides the above-presented electoral analysis, research has also shown that when incumbents perform poorly in the economy they have little chance of being re-elected (Nezi, 2012). In an analysis of Eurobarometer data for the period 1985–1999, it has been concluded for Greece that both sociotropic and egocentric perceptions about the economic conditions were significant determinants of voting (Freire and Costa Lobo, 2005). Also, it has been argued that the January 2015 election is a case of retrospective and economic voting, through which the punishment of the parties that had handled the economic crisis was completed (Tsirbas, 2016).

On the contrary, the empirical research at the regional level is rather scarce and it has focused mainly on the USA and the UK (Leib and Quinton, 2011: 12). For the USA, the empirical evidence lends support to the retrospective hypothesis, using income growth and/or unemployment rate as main indicators (e.g. Cho and Gimpel, 2009; Kim et al., 2003; Lacombe and Shaughnessy, 2007; Wright, 2012), while similar findings have also been reported for the UK (e.g. Gibson, 1992; Jensen et al., 2013; Johnston et al., 2007; Pattie et al., 2015). The few studies analysing economic voting in other countries have resulted in mixed findings: some have fully confirmed the hypothesis (e.g. Auberger and Aubois, 2005, for French legislative elections), some have only partially confirmed the hypothesis (e.g. Elinder, 2010, for Swedish general elections) and others have not confirmed it at all (e.g. Fernández-Durán et al., 2004, for Mexico). Moreover, at the level of regional and local elections, the potential explanatory power of a more specialised type of economic voting, namely second-order economic voting, has been highlighted (Bosch, 2016; Jastramskis, 2014).

Since there have been some important referendums in recent years – like the referendum for Scottish independence, the referendum on the membership of the UK in the EU, the Italian constitutional referendum and the Greek referendum on the bailout agreement – the importance of referendums is on the rise. It is also worth noting that referendums present some special characteristics that differentiate them from parliamentary elections and must be taken into account in any thorough analysis (for an extensive review see Mendez et al., 2014).

First, the dichotomous questions of referendums, by simplifying complex issues, tend to divide the electorate and to spark fierce competition and intense rhetoric between opposing sides. Second, although political parties usually have an openly expressed stance for the referendum’s dilemma (see Table 1) and actively try to influence the electorate, the latter can bypass them and deal with the referendum dilemma with a more holistic way, especially if the referendum has polity-changing implications. In other words, referendums are often seen as questions about the status quo of a political system as a whole (Hobolt, 2006), i.e. quite differently than parliamentary elections. It has been argued that there is a ‘status quo bias’ in voters’ decision making regarding referendums: voters opt for the ‘devil they know’ (Le Duc’s law) (Le Duc, 2003; Hobolt, 2015). However, in the Greek referendum case Le Duc’s law did not hold.

Study area, data and variables

Study area description

The main administrative–territorial structure in Greece includes 13 regions corresponding to NUTS 2 level of the Eurostat, 51 regions (prefectures) corresponding to NUTS 3 level⁴ and 325 municipalities that can be subdivided into municipal or local communities (LAU 1 level). The empirical analysis is conducted at NUTS 3 level, the finest spatial scale for this study, for several reasons. First, this scale reflects a politically relevant geographical division because NUTS 3 regions are also electoral regions⁵ (constituencies). Second, these regions exhibit much less heterogeneity than would be observed in the larger NUTS 2 regions both in terms of development level and regional economic structures (Petrakos and Artelaris, 2008) and in terms of historical and political characteristics. This homogeneity, partly explained by a historical point of view because the prefectures have been long the country’s main administrative unit, can alleviate the potential problem of ecological fallacy (see below). Third, this spatial scale is the lower spatial aggregation level to obtain reliable economic variables; economic data at the municipal level are scarce and less reliable in Greece.

It is worth noting that although there is a certain ‘individualistic primacy’ (Loney and Nagelkerke, 2014) amongst social researchers, meaning that associations at the individual level are considered to be more valid, mainly because of the ecological fallacy, the importance of aggregate-level analysis has been long recognised (Wellhofer, 1991). Ecological fallacy refers to the possibility of making false inferences using aggregate (or macro level) data to explain individual (or micro level) behaviour. However, there are several strong arguments for the use of this type of data in this study.

First and foremost, there is a lack of individual-level data at the regional level, not only in Greece but also in other countries. Second, the direction of associations discovered at the aggregate level is usually the same with that of associations discovered at the individual level, although they would probably differ in size (Loney and Nagelkerke, 2014). Third, there are some variables that have meaning only at the aggregate level, like living in a crisis-hit region or not. Fourth, an aggregate analysis using objective measures of economic conditions and real voting data avoids problems related to measurement error and non-response in survey data as well as other problems raised by subjective perceptions and survey responses (Elinder, 2010). Finally, aggregate variables are instrumental, in the sense that they are correlated with the individual exposure and can only affect the outcome through the individual exposure (Loney and Nagelkerke, 2014). For example, although the average regional change in an economic indicator might not accurately reflect the degree in which each individual was exposed to it, it largely refers to the same people over time. At the same time, the possibility of having people with reverse effects (i.e. people who became richer in the crisis) exists, but there is good reason to believe their number would be small and, thus, unable to affect the direction and magnitude of results. Moreover, the above-presented literature suggests that voters are more sociotropic than egotropic, that is they are more sensitive to how the majority of others are doing. In any case, these types of data do not preclude detailed analyses of elections as several studies have suggested (e.g. O’Loughlin et al., 1994).

Data and variables

Most studies in the economic voting literature use the following three objective economic indicators in their empirical analysis: unemployment rate, GDP (or income) growth and inflation rate (see, e.g. Bouvet and King, 2016; Burnett and Lacombe, 2012; Elinder, 2010; Kim et al., 2003; Nezi, 2012). These measures are very salient mainly because they are frequently cited by parties, receive great coverage by media as well as provide a good picture of the economic situation and environment.

In this study, however, focusing on the regional level, two important modifications have been made to this common practice for several reasons. First, in the absence of household income data at NUTS 3 level, the growth rate of per capita bank deposits during the period of crisis is used in the regression analysis, instead of per capita GDP. Although GDP is the most widely used measure of economic performance, there are several issues concerning its reliability and its ability to measure economic well-being (see, for instance Van den Bergh, 2007), especially in periods of economic crises (OECD, 2013). Moreover, GDP measures what and where is produced but it does not measure the regional income because a part of the income is consumed in other or foreign areas. In Greece, this problem is very important because both of Greece’s largest regions, Attiki and Thessaloniki, ‘export’ a large part of their economic activity to their neighbouring-satellite regions and many employees are commuting to work on a daily basis (Petrakos and Artelaris, 2008). The considerable and frequent changes and revisions of the regional GDP from the Hellenic Statistical Authority (ELSTAT) also highlight the inappropriateness of this indicator for this study (Petrakos and Psycharis, 2016).

As a result, bank deposits seem to be a more useful variable for this study, mainly because they are a more direct measure of income capacity, security and well-being. Moreover, in Greece, throughout the crisis period, there has been a recurring discourse about the possibility of a bail-in of depositors (deposits’ ‘haircut’) to support the banks financially. In addition to this, in neighbouring Cyprus, a country of the same language and nationality and very close historical and social ties with Greece, bank deposits were indeed cut in 2013 because of the crisis, a development that received much media attention in Greece. Thus, bank deposits, besides a measure of well-being and income security, have also become a central and salient issue in the crisis-related public debate. Not surprisingly, a negative association is expected for the growth rate of bank deposits per capita and ‘NO’ vote.

The second main distinction from previous studies is the use of regional income poverty as a critical determinant of voting outcomes. Although relative poverty (or income inequality) has been a matter of great academic and political concern for decades, the interest has increased not only as a result of the recent Great Recession but also because poverty seems to be a relatively persistent phenomenon and a threat to economic prosperity and social cohesion of the EU. However, despite the great interest and the growing media coverage at the international level, it has not attracted interest from voting behaviour literature (Bouvet and King, 2016). Especially for Greece, the economic crisis has brought poverty to the forefront of public debate and the media has intensified its coverage.

In terms of theory, Alesina et al. (2004) have shown why inequality (which is often associated with high relative poverty rates) negatively affects individual utility even after controlling for individual income and several other personal characteristics. In terms of empirical analysis, Bouvet and King (2016), using data on national parliamentary election outcomes in 32 the Organisation for Economic Co-operation and Development countries, have identified income inequality as a potential determinant of voting behaviour, suggesting that the media attention, given to the ‘Occupy Wall Street’ movement during the Great Recession and Thomas Piketty’s book (2014), Capital in the Twenty-First Century, have placed inequality at the centre of public debate. The importance of this variable for voting has been confirmed by several national polls as well (Bouvet and King, 2016).

In this study, the ‘at-risk-of-poverty’ rate, the most widely recognised indicator of income poverty, is used in the econometric analysis. This indicator is defined as the percentage of persons in households whose equivalised disposable income is less than 60% of the national median. Using estimations from recent studies employing microsimulation models (a combination of sample survey and population census information) to calculate regional poverty solves the common problem of paucity of data at low aggregation levels, such as prefectures (Artelaris and Kandylis, 2014; Copus et al., 2015). It is logical to assume that a positive association is expected for poverty rates and anti-austerity voting shares.

Finally, following previous studies, the level of regional unemployment also reflects the economic conditions for the last year before the referendum. For this variable, it is logical to expect that a higher regional unemployment rate is related to a higher anti-austerity vote, not only for punishment reasons but also because the anti-austerity camp, and especially left-wing parties, traditionally focuses on the reduction of unemployment.

Economic characteristics, however, are not the only ones that can affect voting. As a result, to avoid model misspecification, the model should also include control variables. The most influential variable of this type is typically the outcome of the previous national elections; this variable can capture a large number of factors, including inter alia, history, cultural and religious features, ideological predispositions and socio-demographic factors, facilitating the neutralisation of possible contextual effects. This ‘strong’ variable is expected to have the most significant impact on voting. Thus, the inclusion of additional socio-demographic or political controls is not appropriate because these influences have already manifested themselves in the previous elections (Bouvet and King, 2016; Martinsson, 2013). As a result, the average percentage of regional voting shares in the previous three national elections is included in the regression model. This variable is operationalised by adding together the vote shares of right-of-centre and right-wing parties on the one hand, and left-of-centre and left-wing parties on the other hand, thus attaining the political ‘identity’ of each region in the main ideological axis. Naturally, this ‘regional political identity’ variable encompasses all political trends, as well as the special characteristics and historical ‘memory’, for each region.

The data used in the empirical analysis are retrieved from several sources. Table 2 presents a description of the variables used and especially their unit of measurement, data sources and summary statistics.

Table 2.

Variables and summary statistics.

Variable	Unit of measurement	Source	Min	Max	Mean	Median	St. Dev	Exp. Sign
Anti-memorandum vote/‘No’ vote	% of total votes	Ministry of Interior	51.2	73.8	61.4	60.9	5.1	NA
Bank deposits per capita	% change (2010–2014)	Bank of Greece	−26.0	−11.5	−18.7	−19.1	3.4	−
Unemployment rate	2014 level	Labour Force Survey (ELSTAT)	12.8	38.8	25	25.4	5.6	+
Poverty	60% of the median income, 2011	Micro-simulation estimations^a	21.6	31.0	26.8	26.7	1.7	+
Regions’ political identity	Average vote share of left and left-of-centre parties in 2004, 2007 and 2009 national elections	Ministry of Interior	41.0	65.7	52.1	51.9	5.0	+

^aSee Artelaris and Kandylis (2014) and Copus et al. (2015).

Spatial econometric methods

Political data are, in essence, spatial data because the political attitudes, processes and events occur at specific geographic locations (Darmofal, 2006). Given the spatial nature of the data, it is reasonable to expect the possibility of the existence of spatial autocorrelation or dependence in the data (Jensen et al., 2013; O’Loughlin et al., 1994). Theoretically, spatial autocorrelation is based on the First Law of Geography, suggested by Tobler in 1970, which states that near things are more related than things that are more distant. Several theories in political science suggest that spatially proximate entities are more likely to behave similarly than spatially distant entities, predicting positive spatial autocorrelation, that is the spatial clustering of similar behaviours among neighbouring observations (Darmofal, 2006).

Although reasonable and convincing from a theoretical point of view, spatial dependence is not included in most electoral studies, while, when included, it is not often adequately addressed (Lacombe et al., 2014). The presence of spatial dependence may lead to serious bias and/or inefficiency in the estimates of the coefficients (Anselin and Bera, 1998). In the voting literature, several studies have found clear evidence of spatial dependence, rejecting the ordinary least squares (OLS) (a-spatial) model in favour of the spatial models (see for the US Burnett and Lacombe, 2012; Cho and Gimpel, 2009; Kim et al., 2003; Lacombe and Shaughnessy, 2007; O’Loughlin et al., 1994. See for the UK Cutts and Webber, 2010; Jensen et al., 2013. See for Mexico Fernández-Durán et al., 2004. See for Spain Tapiador and Mezo, 2009).

Spatial dependence, in a regression model, can be taken into account via a spatial (N × N) weight matrix W. This matrix typically is a formal expression of spatial adjacency between regions and specifies the degree of interdependence between any two observations. Several forms of spatial weight matrices have been suggested in the literature (see Anselin and Bera, 1998). Typically, the specification of a weight matrix is based on three criteria: simple contiguity, decreasing functions of distance and k-nearest neighbours. In this study, the most frequently used weights matrix, the contiguity matrix, is used where regions sharing a common border are considered neighbours.⁶

Two main types of spatial regression models have been extensively used in the literature: spatial autoregressive (or lag) model (SAR) and the spatial error model (SEM) (Anselin, 2010). In the former model (SAR), spatial autocorrelation pertains to the dependent variable, namely voting. This model can be expressed in matrix notation as

y = ρ W y + X β + ε

(1)

ε \sim N(0, σ^{2} I)

where y is a vector (n×1) of observations on the dependent variable, ρ is the coefficient of the spatial lag term Wy measuring how neighbouring observations affect the dependent variable (a spatial autoregressive coefficient), W defines the (n×n) weight matrix, X is the matrix (n×k) of observations on the independent variables, β is the vector of parameters to be estimated and ε is a vector of error terms.

In the latter model (SEM), the spatial autocorrelation is considered in the error term implying that the factors not included in the model are correlated over space. Formally, this model can be expressed as follows

y = X β + ε with ε = λ W ε + μ

(2)

where λ is a coefficient indicating the extent of spatial correlation between the residuals, and the rest of the notations are the same as before. If appropriate, these models are estimated using maximum likelihood or method of moments techniques (Anselin and Bera, 1998).

More recently, a few scholars have suggested that the restriction to SAR and SEM might be questionable (see, e.g. LeSage and Pace, 2009). As a result, alternative spatial models should be considered to draw correct inferences. Each of these spatial models has a different motivation containing, typically, more than one spatial interaction effect. Examples of such models include, among others, the spatial Durbin model (SDM), which contains both a spatially lagged-dependent variable and spatially lagged explanatory variables, and the Kelejian–Prucha, or SAC, model that makes use of both a spatially lagged-dependent variable and a spatially autocorrelated error term.

Empirical results and discussion

Based on the above discussion, the effect of regional economic conditions on electoral outcomes is tested by the following cross-sectional OLS regression model

N O = a + β_{1} P_E l e c t + β_{2} D e p_G r + β_{3} U n e m p l + β_{4} A R O P + ε

(3)

where NO is the regional percentage of the anti-austerity voting share in the referendum, P_Elect is the left-leaning regional vote shares in the previous national elections, Dep_Gr is the regional growth rate of per capita bank deposits, Unempl is the regional unemployment rate and AROP is the regional ‘at-risk-of-poverty’ rate.

The results of seven different models are reported in Table 3. Model 1 is the basic a-spatial regression OLS model; Models 2, 3 and 4 include the spatial variables (lag and error) to capture spatial dependence while model 5 include the standardised effects of our best specification to assess the relative impacts of the variables. In order to control for national effects, equation (3) is augmented (model 6), following Casi and Resmini (2014), to include four dummy variables, one for every explanatory variable, which measure each region’s value with the national average (equals 1 if the value is above the mean of the country and 0 otherwise). For the sake of comparison, the last column presents the basic model⁷ using, however, as the dependent variable the (aggregate) regional voting shares of the national elections of January 2015 for the political parties that were against the memorandum in the referendum (model 7). It is worth noting that several other specifications and variables are used (such as age, the number of government employees), but the specifications presented here provide the best fit of the data and the most stable coefficient estimates implying no significant bias.

Table 3.

The impact of regional economic conditions on voting shares in Greece, OLS, ML and GMM estimations.

	Model 1: OLS Referendum	Model 2: SAR referendum	Model 3: SEM referendum	Model 4: SEM (GMM-HET) referendum	Model 5: SEM (standardised) referendum	Model 6: SEM with national dummies referendum	Model 7: OLS January 2015 National Elections
Constant	−0.16435(0.159)	−0.16606(0.135)	−0.16048(0.124)	−0.16022*(0.077)	−0.01718(0.873)	−0.254137*(0.058)	−0.24507(0.100)
Regions political identity	0.00798***(0.000)	0.00799***(0.000)	0.00816 ***(0.000)	0.00818***(0.000)	0.81529***(0.000)	0.00882***(0.000)	0.00808***(0.000)
Bank deposits growth	−0.53146***(0.000)	−0.53410***(0.000)	−0.49703***(0.000)	−0.4936***(0.000)	−0.33157***(0.000)	−0.49951**(0.001)	−0.49432***(0.008)
Unemployment	0.00146*(0.076)	0.00145*(0.068)	0.00129*(0.074)	0.00127***(0.009)	0.14080*(0.074)	0.00298***(0.003)	0.00148(0.156)
Poverty	0.84025***(0.005)	0.83943***(0.002)	0.82642***(0.001)	0.8244***(0.002)	0.27550***(0.001)	0.86047**(0.016)	0.76827**(0.040)
Regions political identity (national dummy)	–	–	–	–	–	−0.01126(0.469)	–
Bank deposits growth (national dummy)	–	–	–	–	–	−0.00399(0.767)	–
Unemployment(National Dummy)	–	–	–	–	–	0.02876**(0.015)	–
Poverty(national dummy)	–	–	–	–	–	−0.00145(0.910)	–
Spatial coefficient (ρ/λ)	–	0.00165(0.936)	0.32069**(0.042)	0.4046**(0.023)	–	0.33609**(0.031)	–
Coefficient of determination (R²)a	0.62	0.65	0.68	0.65	0.68	0.71	0.50
Akaike information criteria (AIC)	−203.6	−201.6	−206.1	–	97.02	−204.3	−179.0
Schwarz criteria (SC)	−193.9	−190.0	−196.4	–	106.7	−186.9	−169.4
Koenker–Bassett	6.69(0.153)	–	–	–	–	–	4.18(0.381)
Breusch–Pagan	8.52*(0.074)	8.33*(0.080)	8.94*(0.062)	–	5.95(0.113)	2.35(0.124)	4.81(0.307)
White test	16.60(0.277)	–	–	–	–	–	16.74(0.270)
Jarque–Bera (normality)	0.92(0.629)	–	–	–	–	–	2.22(0.329)

GMM: Generalised Method of Moments; HET: Heteroscedasticity; ML: maximum likelihood; OLS: ordinary least squares; SAR: spatial lag model; SEM: spatial error model.

Note: ***, ** and * denote a statistical significance level of 1, 5 and 10%, respectively.

^aFor spatial models: pseudo-R².

For each basic model, the estimated coefficients, the level of statistical significance (p values), the adjusted $R^{2}$ figures, the Akaike (AIC) and Schwarz (SC) information criteria to assess the improvement in fit during the modelling stages as well as diagnostic tests for heteroscedasticity and normality are reported. As can be seen from Table 3, all estimated parameters of the basic model (model 1) have the expected signs and are highly significant by the conventional criteria. In other words, regional economic conditions seem to have played a major role in the referendum outcomes. The model fit is very satisfactory explaining more than 60% of the variation in the ‘NO’ vote. However, this model is largely improved when an SEM is estimated; both information criteria (i.e. AIC and SC) indicate the superiority of SEM over any other model employed,⁸ while the coefficient λ for the spatial error variable is positive and highly significant, implying a strong correlation between observations located nearby. It is worth mentioning that the results show no significant problems of heteroscedasticity or non-normality; diagnostic tests for heteroscedasticity (i.e. White test and Koenker–Bassett test) do not reject homoscedasticity, whereas the Jarque–Bera test is also not significant, suggesting that errors are normally distributed.

However, since the Breusch–Pagan test did reject homoscedasticity at only 10% level, the SEM is estimated using the Generalised Method of Moments (GMM). Kelejian and Prucha (2010) have developed an estimator for the autoregressive parameter of a spatially autoregressive disturbance process that allows for heteroscedastic innovations (see also Artelaris and Petrakos, 2016). A variable bandwidth based on the four nearest neighbours is used in this study and the results are robust to all alternative specifications of kernel function employed, i.e. Epanechnikov, Triangular and Gaussian. Although the use of GMM does not affect the previous results in terms of the magnitude of coefficients, the effect on significance is profound; all parameter estimates, and especially unemployment, appear stronger in terms of statistical significance. Interestingly, the results of the SEM model are also identical both for model 6, controlling for national effects, and for model 7, focusing on the national elections of January 2015.

In support of the above hypotheses, the empirical results indicate the association between regional electoral outcomes and economic conditions. More specifically, the grievance hypothesis is confirmed, because a greater regional unemployment and poverty rate are both significantly positively associated with greater anti-austerity voting shares. In this respect, ‘classic’ economic voting assumptions are verified. Moreover, the growth rate of per capita bank deposits, a measure of income capacity, security and well-being, is found to be negatively associated with the anti-austerity vote. Interestingly, standardised coefficients (model 5) reveal that all variables have a high impact on voting share but, as expected, political identity has a bigger impact than the rest. On the other hand, unemployment plays a less important role compared to the other economic variables (i.e. bank deposits growth and poverty) but not negligible. Last but not least, the results also show evidence of spatial autocorrelation in residuals from OLS regression models, highlighting a strong correlation between observations located nearby. SEM, found to be the most appropriate model in this study, has been selected as the most appropriate specification in several electoral studies because certain areas might vote in a certain manner due to local customs or historical reasons (Lacombe and Shaughnessy, 2007). The latter point is also connected to the fact that regional political identity, measured by its electoral history, is also significant, as expected.

Conclusion

The economic crisis has affected Greece more than any other country and has shaped political competition from 2010 onwards. The structural reforms and harsh austerity measures that Greece had to adopt, to get financial aid from the EU and IMF as well as to avoid default, became a central issue of Greek politics, forming two opposite camps: one pro-austerity and pro-EU, and one anti-austerity and critical of the EU. The latter camp finally gained office in January 2015 and received a popular mandate in a referendum in July 2015 to end austerity, without achieving this goal.

Using recent economic data and focusing on the Greek referendum, this study attempted to assess the extent to which the geographic pattern that emerged in the referendum was due to the differentiated regional impact of the economic crisis. The main result is that the extent of the exposure to the economic crisis of each region seems to greatly affect the voting behaviour, in a period of crisis where economic issues are more important and salient. In other words, the impact of economic conditions, their changes during the crisis and the way these conditions and changes are perceived by the electorate at the regional level seem to be of great importance for voting behaviour.

In terms of specific empirical findings, regional voting shares in the referendum, besides its expected positive correlation with the degree of left and left-of-centre historical influence in each region, are also affected by the unemployment rate, the poverty rate and the degree of losses in bank deposits. Since the two latter variables have been totally overlooked by previous similar studies, the results of the analysis suggest they can serve as adequate predictors of voting for future relevant studies. Moreover, the results confirm the importance of the space in electoral behaviour employing, in contrast to most of the previous studies, spatial regression models that normally yield more accurate estimates. In other words, if, according to Edward Tufte (1978: 65), ‘when you think economics, think elections and when you think elections, think economics’, then the present study shows that when you think elections, especially in times of crisis, you should indeed think economics but you should also think geography.

In terms of theory, the findings of this study argue in favour of the grievance hypothesis, implying that negative changes matter more. Intriguingly, this might bridge together economic voting theory with Kahneman and Tversky’s (1979) prospect theory, according to which people tend to prefer the ‘lottery’ of avoiding any cost at all, embracing, at the same time, the possibility of a great loss. Given the described below dichotomous character of referendums, as well as the fierce public debate which preceded the Greek referendum, it can be argued that it was a case of ‘lottery’ behaviour, since the ‘no’ camp advocated the end of austerity (the avoidance of any cost at all) and the ‘yes’ camp highlighted the possibility of Greece exiting the EU in case ‘no’ won (the great loss). Furthermore, it has to be noted that the central characteristic of referendums as total questions on the status quo might have facilitated the translation of economic grievance to a ‘NO’ vote in the Greek case. However, providing our model with convergent validity, the results are analogous for the national elections of January 2015 as well, demonstrating similar associations between the aforementioned variables and the voting shares of the political parties that were against the memorandum in the referendum.

The importance of this finding is even greater taken by responsible political parties’ point of view. On the one hand, governing and office-seeking parties should realise that the economic policies they pursue matter both for their own success and for a political system’s stability. On the other hand, it has to be realised, from a policy-making point of view, that economic policies do not have the same outcome in every region and that regional differences are accountable for differences in voting behaviour.

In this respect, future studies could also investigate the geographies of other referendums. Also, as economic voting and prospect theory share some common notions, the latter could be operationalised and tested in future studies, providing a theoretical foundation for a better understanding of this issue. Moreover, subsequent investigations need to examine the role and the importance of the two new variables proposed in this study, namely regional poverty rate and bank deposits. Last but not least, studies might analyse the voting pattern at different spatial levels as well as their interaction. In any case, it can be asserted that future studies must take advantage of the integration between political and economic science, and regional science and spatial econometrics.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Panagiotis Artelaris is a lecturer at the Department of Geography (Harokopio University) and research associate at the Department of Planning and Regional Development (University of Thessaly). His research interests are in the areas of Spatial Economics, Economic Growth and Decline, Spatial Inequalities, and Spatial Econometrics. He has been involved in several research projects (FP6, FP7, ESPON, HORIZON) and has published several papers in international academic journals including, among others, GEOFORUM, International Regional Science Review, GeoJournal, Local Economy, Region et Developpement, Journal of Economic Studies, Growth and Change, Eastern Journal of European Studies and Italian Journal of Regional Studies.

Yannis Tsirbas is a lecturer at the Department of Political Science and public administration of the University of Athens. He holds a PhD in Political Communication from Panteion University and a MA in Political Behavior from the University of Essex. His academic interests include electoral research, political elites and political communication, areas in which he has published a monograph, several contributions to international collective volumes and several articles in scientific journals like Historical Social Reasearch, South European Society and Politics and Greek Political Science Review.

References

Agnew

(1996) Mapping politics: how context counts in electoral geography. Political Geography 15: 129–146.

Alesina

and Rosenthal

(1995) Partisan Politics, Divided Government and the Economy. Cambridge: Cambridge University Press.

Alesina

Di Tella

and MacCulloch

(2004) Inequality and happiness: Are Europeans and American’s different? Journal of Public Economics 88: 2009–2042.

Anselin

(2010) Thirty years of spatial econometrics. Papers in Regional Science 89(1): 3–25.

Anselin

and Bera

(1998) Spatial dependence in linear regression models with an introduction to spatial econometrics. In: Ullah

and Giles

(eds) Handbook of Applied Economic Statistics. New York: Marcel Dekker, pp.237–289.

Artelaris

(2017) Geographies of crisis in Greece: A social well-being approach. Geoforum 84: 59–69.

Artelaris

and Kandylis

(2014) Mapping poverty at regional level in Greece. Region et Developpement 39: 131–147.

Artelaris

and Petrakos

(2016) Intraregional spatial inequalities and regional income level in the EU: Beyond the inverted-U hypothesis. International Regional Science Review 39(3): 291–317.

Auberger

and Aubois

(2005) The influence of local and national economic conditions on French legislative elections. Public Choice 125: 363–383.

10.

Bosch

(2016) Types of economic voting in regional elections. Journal of Elections, Public Opinion and Parties 26(1): 115–134.

11.

Bosco

and Verney

(2012) Electoral epidemic: The political cost of economic crisis in Southern Europe, 2010–11. South European Society and Politics 17(2): 129–154.

12.

Bouvet

and King

(2016) Income inequality and election outcomes in OECD countries: New evidence following the Great Recession of 2008–2009. Electoral Studies 41: 70–79.

13.

Burnett

and Lacombe

(2012) Accounting for spatial autocorrelation in the 2004 presidential popular vote: A reassessment of the evidence. The Review of Regional Studies 42: 75–89.

14.

Casi L and Resmini L (2014) Spatial complexity and interactions in the FDI attractiveness of regions. Papers in Regional Science 93(S1): S51–S78.

15.

Cho

and Gimpel

(2009) Presidential voting and the local variability of economic hardship. The Forum 7(1): 1–21.

16.

Copus

Melo

Kaup

et al . (2015) Regional poverty mapping in Europe – Challenges, advances, benefits and limitations. Local Economy 30(7): 742–764.

17.

Cutts

and Webber

(2010) Voting patterns, party spending and relative location in England and Wales. Regional Studies 44(6): 735–760.

18.

Darmofal

(2006) Spatial Econometrics and Political Science. Working Paper. Columbia: Department of Political Science, University of South Carolina.

19.

Dinas

and Rori

(2013) The 2012 Greek parliamentary elections: Fear and loathing in the polls. West European Politics 36(1): 270–282.

20.

Elinder

(2010) Local economies and general elections: The influence of municipal and regional economic conditions on voting in Sweden 1985–2002. European Journal of Political Economy 26: 279–292.

21.

Eurostat (2016) Smarter, Greener, More Inclusive? Indicators to Support the Europe 2020 Strategy. Luxembourg: Eurostat Statistical Books.

22.

Fernández-Durán

Poire

and Rojas-Nandayapa

(2004) Spatial and temporal effects in Mexican direct elections for the chamber of deputies. Political Geography 23: 529–548.

23.

Freire

and Costa Lobo

(2005) Economics, ideology and vote: Southern Europe, 1985–2000. European Journal of Political Research 44: 493–518.

24.

Gibson

(1992) The effects of unemployment on voting in British elections: A new specification of a political-economic model of constituency voting. Environment and Planning C: Government and Policy 10: 451–465.

25.

Hadjimichalis

and Hudson

(2014) Contemporary crisis across Europe and the crisis of regional development theories. Regional Studies 48(1): 208–218.

26.

Harvey

(1982) The Limits to Capital. Oxford: Basil Blackwell.

27.

Hellwig

(2008) Globalization, policy constraints and vote choice. Journal of Politics 63(4): 1128–1141.

28.

Hernandez

and Kriesi

(2016) The electoral consequences of the financial and economic crisis in Europe. European Journal of Political Research 55: 203–224.

29.

Hobolt

(2006) How parties affect vote choice in European integration referendums. Party Politics 12(5): 623–647.

30.

Hobolt

(2015) What lessons can Britain learn from other EU referendums? Available at: http://bit.ly/1RyxOaz (accessed 15 November 2016)

31.

Jastramskis

(2014) Learning the economic vote at local elections: Case of Lithuania, 1995–2011. Baltic Journal of Political Science 3: 100–112. DOI: 10.15388/BJPS.2014.3.4868.

32.

Jensen

Lacombe

and McIntyre

(2013) A Bayesian spatial econometric analysis of the 2010 UK General Election. Papers in Regional Science 92(3): 651–667.

33.

Johnston

Jones

Propper

et al . (2007) Region, local context, and voting at the 1997 general election in England. American Journal of Political Science 51(3): 640–654.

34.

Kahneman

and Tversky

(1979) Prospect theory: An analysis of decision under risk. Econometrica 47(2): 263–291.

35.

Kelejian

and Prucha

(2010) Specification and estimation of spatial autoregressive models with autoregressive and heteroscedastic disturbances. Journal of Econometrics 157: 153–167.

36.

Kim

Elliott

and Wang

(2003) A spatial analysis of county-level outcomes in US presidential elections: 1988–2000. Electoral Studies 22: 741–761.

37.

Kostadinova

and Giurcanu

(2015) Political blackmail, institutional infighting, and electoral politics: The fate of governing parties during the great recession (2008–12) in Romania and Bulgaria. Politics & Policy 43(6): 789–821.

38.

Kramer

(1983) The ecological fallacy revisited: Aggregate-versus individual-level findings on economics and elections, and sociotropic voting. American Political Science Review 77(1): 92–111.

39.

Lacombe

Holloway

and Shaughnessy

(2014) Bayesian estimation of the spatial Durbin error model with an application to voter turnout in the 2004 presidential election. International Regional Science Review 37(3): 298–327.

40.

Lacombe

and Shaughnessy

(2007) Accounting for spatial error correlation in the 2004 presidential popular vote. Public Finance Review 35(4): 480–499.

41.

Le Duc

(2003) The Politics of Direct Democracy: Referendums in Global Perspective. Toronto: Broadview.

42.

Leib

and Quinton

(2011) On the shores of the moribund backwater?: Trends in electoral geography research since 1990. In: Warf

and Leib

(eds) Revitalizing_Electoral_Geography. Surrey, United Kingdom: Ashgate, pp.9–27.

43.

LeSage

and Pace

(2009) Introduction to Spatial Econometrics. Boca Raton, FL: Taylor and Francis.

44.

Lewis-Beck

and Paldam

(2000) Economic voting: An introduction. Electoral Studies 19: 113–121.

45.

Loney

and Nagelkerke

(2014) The individualistic fallacy, ecological studies, and instrumental variables: A causal interpretation. Emerging Themes in Epidemiology 11: 18.

46.

Martin

(2010) The local geographies of the financial crisis: From the housing bubble to economic recession and beyond. Journal of Economic Geography 11: 87–618.

47.

Martinsson

(2013) Economic voting in Sweden 1985–2010. Electoral Studies 32: 470–475.

48.

Mendez

and Triga

(2014) Referendums and the European Union. A Comparative Inquiry. Cambridge: Cambridge University Press.

49.

Nannestad

and Paldam

(1994) The VP-function: A survey of the literature on vote and popularity functions after 25 years. Public Policy 79: 213–245.

50.

Nezi

(2012) Economic voting under the economic crisis: Evidence from Greece. Electoral Studies 31: 498–505.

51.

OECD (2013) Measuring Well-Being and Progress. Paris: OECD Statistics Directorate. Available at: http://www.oecd.org/measuringprogress (accessed on 15 February 2017).

52.

O’Loughlin

Flint

and Anselin

(1994) The geography of the Nazi vote: Context, confession, and class in the Reichstag election of 1930. Annals of the Association of American Geographers 84: 351–380.

53.

Pattie

Johnston

Schipper

et al . (2015) Are regions important in British elections? Valence politics and local economic contexts at the 2010 general election. Regional Studies 49(9): 1561–1574.

54.

Perrons

(2012) Regional performance and inequality: Linking economic and social development through a capabilities approach. Cambridge Journal of Regions, Economy and Society 5: 15–29.

55.

Petrakos

and Artelaris

(2008) Regional inequalities in Greece. In: Coccossis Η and Psycharis Y (eds) Regional Analysis and Policy: The Greek Experience, pp. 121–139. Heidelberg: Springer/Physica Verlag HD.

56.

Petrakos

and Psycharis

(2016) Regional Development in Greece. Athens: Kritiki (in Greek).

57.

Piketty

(2014) Capital in the Twenty-First Century. Cambridge, MA: Belknap.

58.

Ragusa

and Tarpey

(2016) The geographies of economic voting in presidential and congressional elections. Political Science Quarterly 131: 101–132.

59.

Tapiador

and Mezo

(2009) Vote evolution in Spain, 1977–2007: A spatial analysis at the municipal scale. Political Geography 28: 319–328.

60.

Teperoglou

and Tsatsanis

(2014) Dealignment, de-legitimation and the implosion of the two-party system in Greece: The earthquake election of 6 May 2012. Journal of Elections, Public Opinion and Parties 24(2): 222–242.

61.

Tobler

(1970) A computer movie simulating urban growth in the Detroit region. Economic Geography 46: 234–240.

62.

Tsatsanis

Freire

and Tsirbas

(2014) The impact of the economic crisis on the ideological space in Portugal and Greece: A comparison of elites and voters. South European Society and Politics 19(4): 519–540.

63.

Tsatsanis E and Teperoglou E (2016) Realignment under stress: the july 2015 referendum and the september parliamentary election in greece. South European Society and Politics 21 (4): 427–450.

64.

Tsirbas

(2016) The January 2015 parliamentary election in Greece: Government change, partial punishment and hesitant stabilisation. South European Society and Politics. 21(4): 407–426. DOI: 10.1080/13608746.2015.1088428

65.

Tsirbas

and Sotiropoulos

(2015) What do Greek political elites think about Europe and the crisis? An exploratory analysis. L.S.E. Hellenic Observatory Papers on Greece and Southeast Europe. GreeSE paper No. 93. Available at: http://www.lse.ac.uk/europeanInstitute/research/hellenicObservatory/CMS%20pdf/Publications/GreeSE/GreeSE-No93.pdf.

66.

Tufte

(1978) Political Control of the Economy. Princeton, NJ: Princeton University Press.

67.

van den Bergh

(2007) Abolishing GDP. Tinbergen Institute Discussion, 019/3.

68.

Van den Bergh J. (2007). Abolishing GDP. TI 019/3. Amsterdam: Tinbergen Institute Discussion Paper.

69.

Vasilopoulou

and Halikiopoulou

(2013) In the shadow of Grexit: The Greek election of 17 June 2012. South European Society and Politics 18(4): 523–542.

70.

Vasselai

and Mignozzetti

(2012) The impacts of international economic crises on the succession of incumbents in 55 countries (1960–2012). In: Paper presented at the 2012 annual meeting of APSA. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2108034 (accessed 9 March 2018).

71.

Wayland

(1998) The political fate of market reform in Latin America, Africa, and Eastern Europe’. International Studies Quarterly 42: 645–674.

72.

Wellhofer S (1991) Confounding sources of variance in the macro-analysis of electoral data. European Journal of Political Research 19(4): 425–39.

73.

Wright

(2012) Unemployment and the democratic electoral advantage. American Political Science Review 106(4): 685–702.