Unemployment as fertile ground for electoral support of the radical left: evidence from the european regions in the first two decades of the 21st century

ABSTACT

This paper aims to discover whether the rate of unemployment in the first two decades of the 21^st century influenced electoral support for the radical left parties (RLPs). The election results for 37 RLPs were analysed at the level of the NUTS 2 regions. The impact of unemployment was examined using four indicators – general, youth, long-term and male unemployment rates. The aim was achieved through a multilevel regression analysis conducted for parliamentary elections between 2002 and 2019. The results suggest that RLPs are more successful in regions with a higher unemployment rate (general, long-term, male) across the years under review. A 1% rise in the long-term unemployment rate led to an increase in support for RLPs of 0.2% to 1.9% in most NUTS 2 regions (179 out of 198). This effect is most pronounced in the regions where the centre-periphery cleavage associated with historically strong leftist tendencies can be observed, particularly in the case of German, Greek, Portuguese and Spanish regions. The main thrust of the paper is in the analysis of electoral support in parliamentary elections at the regional level, which is less represented in the thematic literature compared with the individual or national level.

KEYWORDS:

• Radical left
• unemployment
• electoral geography
• NUTS 2 regions

Introduction

The paper deals with the radical left parties (RLPs), which can be briefly defined as communist or socialist parties with membership in the Party of the European Left or the movement Now the People. A strengthening of the RLPs has occurred since the beginning of the new millennium. This has been achieved either through a rise in the number of countries in which they achieved at least a 1% election result – 16 countries before 2009, 18 after 2009 – or larger electoral gains, from 3% to 10% before 2009 to an average of 5% to 13% after 2009. These figures are also affected by significant differences, either across countries (relatively strong positions in the political systems of Cyprus, Czechia, Finland, France, Greece, Portugal and Spain) or across regions within a country, e.g. eastern and western Germany or north-east and central Spain.

Although these parties have strengthened because of the global financial crisis and the European debt crisis (Figure 1), predictions (March 2012; Vail and Bowyer 2012) that these parties would again be a major force have not materialised. This is evidenced by worsening election results in recent years at national level, e.g. Czechia and Spain. It is possible, therefore, that there could be a return to the results witnessed at the turn of the millennium as the RLPs have almost disappeared from the former post-socialist states that are now members of the European Union (EU). These two phenomena may have arisen because most of these parties failed to overcome their own long-standing problems, e.g. a historical stigma in eastern European countries, an unattractive strategic vision and ideological differences between factions (March 2012).

Figure 1. Electoral results of RLPs in parliamentary elections.

Left figure (2002–2008), right figure (2009–2019). Darker area means higher electoral support. Source: Own calculation based on data from the European Election Database (Norwegian Centre for Research Data 2020) and individual statistical offices; Software: Datawrapper GmbH.

The empirical literature agrees that the economic downturn and rising unemployment rates can be considered as major economic factors influencing electoral support for RLPs (Doležalová et al. 2017; Dorn et al. 2020; Hernández and Kriesi 2016; March and Rommerskirchen 2015; Proaño Acosta, Peña, and Saalfeld 2020; Visser et al. 2014). Therefore, this paper aims to discover whether the rate of unemployment influenced the electoral support of RLPs in the first two decades of the 21^st century. The literature describes the influence of unemployment at the individual (own personal situation) or national (state of the economy) level. However, Figure 1 shows that there are large regional differences in voter support for RLPs. Of course, there are articles dealing with the importance of regional characteristics (e.g. Caselli, Fracasso, and Traverso 2020; Elinder 2010; Georgiadou, Rori, and Roumanias 2018; Van Hamme, Vandermotten, and Lockhart 2018); nevertheless, only a few of these findings directly analyse the development of RLPs and, therefore, there is scope for supplementing the literature on the issue of the impact of regional unemployment rates on electoral support for RLPs. This relationship is monitored at the NUTS 2 regional level, which allows us to observe development of the RLPs in more than two hundred regions within 21 EU countries.

The main results suggest that RLPs are more successful in parliamentary elections in regions with higher rates of unemployment (general, long-term, male) across the period under review. This effect is most pronounced in the regions where we can observe the centre-periphery cleavage associated with historically strong leftist tendencies, particularly in the case of German, Greek, Portuguese and Spanish regions. On the other hand, the insignificant influence of youth unemployment rates can be explained by the fact that young voters prefer new topics such as ecology to traditional cleavages.

The following section will provide a definition of RLPs, a description of the relationship between unemployment and support for RLPs and the influence of regional characteristics. The third section introduces the proxies and a multilevel regression model. The fourth section presents the results of the regression analysis with an emphasis on regional differences. The Conclusions section summarises the individual findings and suggests opportunities for further research.

Literature review

Radical left-wing parties are located to the left of social democracy. The parties ‘reject the underlying socio-economic structure of contemporary capitalism and its values and practices. They advocate alternative economic and power structures involving a major redistribution of resources from the existing political elites’ and emphasise the ‘identification of economic inequity as the basis of existing political and social arrangements, and their espousal of collective economic and social rights as their principal agenda’ (March and Mudde 2005, 25). Within the left-wing political family, it is possible to distinguish four major subgroups (March 2012, 320): conservative communists, e.g. the Communist Party of Greece (KKE); reform communists, e.g. the Cypriot Progressive Party of Working People (AKEL); democratic socialists, e.g. the Finnish Left Alliance (VAS); and populist socialists, e.g. The Left Party (Die Linke). In the last decade, a new stream has emerged in this party grouping known as left-wing populism, e.g. We Can (Podemos) or the Coalition of the Radical Left (SYRIZA), which combines radical left-wing economic programmes with anti-elite rhetoric alleged to represent ‘ordinary people’. These parties are members of the Party of the European Left or the movement Now the People. In accordance with March (2012), the collective designation of radical left is used in this analysis for all these parties.

In line with the theoretical and empirical literature, this paper is based on the premise that the economic downturn and the associated rise in unemployment led to an increase in electoral support for RLPs as voters blamed the ruling party – typically the centre parties – for their problems, e.g. growing uncertainty about future incomes. Visser et al. (2014, 543) comment that ‘unemployed people are especially likely to participate in anti-capitalist movements because their relatively unfavourable socioeconomic situation, dissatisfaction with the way the capitalist system works and frustration might lead them to support radical social change’. Additionally, ‘jobless people possess fewer resources to make a decent living, which might intensify their support for income redistribution’. At the same time, the economic crisis is leading to a decline in wealth and a rise in income inequality, both of which exacerbate demand in society for a revamp of the capitalist system. This effect can be observed in many parliamentary elections (Doležalová et al. 2017; Dorn et al. 2020; Hernández and Kriesi 2016; March and Rommerskirchen 2015; Proaño Acosta, Peña, and Saalfeld 2020; Visser et al. 2014).

The effect was fully proven during the global financial crisis and the European debt crisis, when RLPs recorded significant increases in voter support. In addition to the adverse effects of unemployment, RLPs also attracted votes from Eurosceptics and from those resisting the imposition of austerity measures (Beaudonnet and Gomez 2017), especially in countries facing rescue programmes involving the European Central Bank, the International Monetary Fund and the European Commission (Hobolt and De Vries 2016). In western European countries, this rise in support was even more pronounced as the effects of the financial crash were more acute following several decades of dynamic economic development (Hernández and Kriesi 2016). The new left-wing populist parties (e.g. Podemos, SYRIZA), in particular, have succeeded in gaining the support of ideologically ambivalent and protest voters, making them more successful than the original radical leftist parties (Ramiro and Gomez 2017).

The literature describes the influence of unemployment at individual and national level. However, there is also another level, the state of the regional economy (Pattie and Johnston 1998, 250). This is based on the premise that voting preferences in parliamentary elections depend on an assessment of the economic situation in a locality or region. It can be assumed, then, that people are more affected by changes in their regional surroundings than by national influences since economic development is regionally uneven (Stockemer 2017, 1540) and votes against the government parties would be more likely to be cast in regions that are in recession (Pattie and Johnston 1998, 250). This means that greater support for RLPs will be found in regions that have been more affected by the economic crisis or have suffered from structural problems.

The significant influence of regional characteristics on national elections can be seen in Germany (Dorn et al. 2020), Italy (Caselli, Fracasso, and Traverso 2020), Portugal (Veiga and Veiga 2010), Sweden (Elinder 2010), the United Kingdom (Jensen, Lacombe, and McIntyre 2013), western Europe (Van Hamme, Vandermotten, and Lockhart 2018) and the EU (Georgiadou, Rori, and Roumanias 2018). However, only a few of these findings directly analyse the development of RLPs and, therefore, there is an opening here for further research to complement the literature on the impact of regional rates of unemployment on voter support for RLPs.

Methodology

This section is divided into two parts. First, regional structure is described, RLPs are classified, and regression proxies are introduced. Second, a multilevel regression model is presented as the basic instrument of the paper.

Data

This paper analyses electoral support at regional level. Analysis of data at regional level has two advantages over a similar examination at national level (Stockemer 2017, 1540). First, the variation within countries at regional level is considered. For example, in Czechia, the average unemployment rate for the years under analysis ranged from 3% in Prague to 11% in the Moravian-Silesian region. Second, comparisons can be made between a greater number of areas. For example, instead of comparing election results across 21 states, the results may be compared among 205 regions. Compared with an analysis at individual level (surveys), ‘the election outcomes capture electorate’s revealed (not stated) political preferences’ (Dorn et al. 2020, 5).

Specifically, the NUTS 2 regional level was selected because socio-economic data according to a uniform methodology was not available for a more appropriate lower level (NUTS 3). The EU comprised 281 NUTS 2 regions in the 2016 NUTS classification. This paper does not analyse six states (67 regions) in which the RLPs have not obtained at least 1% of the votes in at least one parliamentary election (Bulgaria, Estonia, Latvia, Malta, Poland and the United Kingdom). In these six countries, it is not possible to analyse the electoral success of the RLPs since these parties have not participated in the electoral process (Malta), their voter support is in the low tenths of a percent (Estonia, Poland and the United Kingdom) or these parties have been part of coalitions (Bulgaria and Latvia). Additionally, Ireland (non-transferability of available economic data to NUTS 2 due to changes in the NUTS 2 classification during the period under review) and six island regions with specific election results (Guadeloupe, Martinique, Guyane, La Réunion, Mayotte and Åland) are omitted. In summary, therefore, this paper will concentrate on 205 NUTS 2 regions in 21 EU member states. The regions are listed in the Appendix (Table A1).

This paper deals with the RLPs that won at least 1% of the votes in at least one parliamentary election at national level (in accordance with Doležalová et al. 2017; Hix and Marsh 2007). The parties were selected based on the ParlGov database methodology (Döring and Manow 2020). Specifically, parties that are classified as ‘Communist/Socialist’ are analysed. This paper assumes the Europeanisation of party families, which means that ideological differences between these groupings are more significant than differences between parties within one group (Camia and Caramani 2012). The group of RLPs is relatively homogeneous ideologically compared with other party families (March and Rommerskirchen 2015, 4) and, simultaneously, it is distinguishable (Fagerholm 2017, 17). Thirty-seven RLPs are analysed. A list of individual political parties is shown in the Appendix (Table A2). Additionally, it should be noted that this paper does not distinguish between individual subgroups of the radical left, e.g. reform and conservative communists or democratic and populist socialists. Data on election results was drawn from two sources, the European Election Database (Norwegian Centre for Research Data 2020) and individual statistical offices. If the NUTS 2 region corresponded to a constituency (Austria, Belgium, Cyprus, Luxemburg, the Netherlands and Spain), then the election result was used, while in case of discrepancies, the calculation was performed by aggregating results from individual electoral districts within the region (the rest of the countries).

The impact of unemployment on electoral support is expressed by four indicators. The general (GeneralUnempl), long-term (LongUnempl) and male (MaleUnempl) unemployment rates are calculated for the age cohort from 20 to 64 years, while the youth unemployment rate (YoungUnempl) is shared between young people neither in employment nor education and training in the age cohort from 15 to 24 years. All four variables are based on the Eurostat methodology (2020d). A comparison of the impact of these four indicators should ensure the robustness of the results.

Control variables can be divided into two groups, regional and macroeconomic. The model contains 12 regional variables. The first five variables, voter turnout (Turnout), vote share of green (Green), social democratic (SocialDem), far-right (FarRight) and regionalist (Regionalist) parties can be described as electoral characteristics. Voter turnout is the most common control variable, assuming that lower turnout is more advantageous for these parties (Doležalová et al. 2017; Finseraas and Vernby 2014). The second and third variables are based on the premise that some voters alter their preferences only within the left spectrum, i.e. a larger share for the remaining two left-wing party families should have a negative effect on RLPs (March and Rommerskirchen 2015). The fourth variable (FarRight) is based on the assumption that the electorate of the far right and left often overlap, especially in the case of topics related to modernisation, globalisation and socio-economic problems (Fagerholm 2018; March and Rommerskirchen 2015). The classification of individual political parties, as in the case of RLPs, is carried out according to the ParlGov database methodology (Döring and Manow 2020) using the categories ‘Green/Ecologist’, ‘Social democracy’ and ‘Right-wing’. The last political variable (Regionalist) is based on the idea that in regions with strong regionalist or separatist parties, the RLPs (as well as other parties) have limited scope for electoral success, especially in economically developed regions (Mazzoleni and Mueller 2016) or in regions with strong centre-peripheral cleavage (Knutsen 2010). Regionalist parties are defined as ‘Ethnic and regional parties’ according to the Manifesto Project (Volkens et al. 2020). Data for all electoral proxies was drawn from the European Election Database (Norwegian Centre for Research Data 2020) and individual statistical offices. The procedure for adjusting the data was the same as for the dependent variable.

The following five variables represent regional economic and structural characteristics. The premise is that voters decide retrospectively, so these variables are delayed by one year. As with other research findings, regional GDP per capita (GDPpc) is considered as a basic indicator of economic development, with the assumption that better results will be achieved in regions with lower economic levels (Vail and Bowyer 2012). This indicator is measured by current market prices (Eurostat 2020d) and is expressed in the form of a logarithmic function. The other four regional variables, representing structural characteristics, are YoungPop (share of people aged 18 to 34 in the total population; Eurostat 2020d), ElderlyPop (share of people over 65 in the total population; Eurostat 2020d), Primary (share of people with less than primary, primary and lower secondary education in the age group 25 to 64; Eurostat 2020d) and Tertiary (share of people with a university degree in the age group 25 to 64; Eurostat 2020d). It is expected that electoral support decreases with increasing age (Bedock and Vasilopoulos 2015; Van der Brug, Hobolt, and De Vreese 2009) and increases with higher education (Bedock and Vasilopoulos 2015; Lubbers and Scheepers 2007; Visser et al. 2014).

In addition to the standard regional characteristics, two variables are used that represent significant long-term trends – structural changes in the manufacturing industry due to Chinese imports (ImportShockChina) and immigration flows (Immigration). Import shock is expressed as the share of the year-on-year change in Chinese imports of industrial products relative to the number of employees in industry. This value is then multiplied by the industrial employment rate in each region. The computation is based on Colantone and Stanig (2018, 940) with data from Eurostat (2020b, 2020d). The proxy represents the idea that globalisation combined with the disappearance of traditional industrial jobs leads to an increase in economic nationalism (Colantone and Stanig 2018), resulting in support for the RLPs (Fagerholm 2018; March and Rommerskirchen 2015; Vampa 2020). The variable Immigration is expressed as the share of foreign citizens in 2011 (Census 2011 round), with the other years being calculated by including a crude rate of net migration plus statistical adjustment (Georgiadou, Rori, and Roumanias 2018). This paper assumes a negative influence of immigration since Edo et al. (2019) on the example of France and Vampa (2020), in the case of Spain, found that a higher proportion of immigrants led to a reduction in regional support for far-left candidates.

As voting behaviour is influenced not only by regional characteristics but also by political and macroeconomic conditions, four variables are added at the state level. The political arrangement is represented by Threshold (minimum percentage of votes required to enter the lower house of parliament; various sources) and PolarVote (voter polarisation index; the ParlGov database methodology; Döring and Manow 2020). It is assumed that RLPs should be more successful in a political system in which the threshold for entering parliament is lower and, simultaneously, the party system is more polarised on the left-right scale (March and Rommerskirchen 2015). Two variables are employed in the case of economic conditions, Growth (GDP per capita growth; Eurostat 2020c) and GINI (Gini coefficient of equivalised disposable income; Eurostat 2020a). Both variables capture the state of the economy one year before an election, with the suggestion that RLPs are stronger in times of economic hardship (economic recession and/or higher income inequality in the economy).

Regression strategy

Due to the structure of panel data at the regional level within the European countries, mixed-effects regression analysis is an essential instrument for fulfilling the aim of this paper. The hierarchical model has three levels (election years, NUTS 2 regions and countries). While maintaining the hierarchical structure, the influence of only regional characteristics is first tested before variables at the state level are included to ensure the robustness of the results. Simultaneously, all regression equations assume that the effect of regional unemployment varies across regions. In conclusion, the employed regression method has three econometric characteristics: the regression estimates are obtained by maximum likelihood; the unstructured covariance matrix in which random-effect variances and covariances are distinctly estimated and all models contain robust standard errors due to the occurrence of heteroscedasticity (cluster option at state level).

Parliamentary elections are analysed for the whole period under review (2002–2019) as well as the sub-period covering the periods before and after the beginning of the financial crisis (2002–2008; 2009–2019). Finally, all panel data regression equations contain annual dummy variables that are associated with election years (Table A3 in the Appendix). The above description can be summarised by the following regression equations:

(1) $Y_{tjk}={\beta }_0+{\beta }_{1}X{a}_{tjk}+{\beta }_{2}X{b}_{t-1jk}+{\beta }_{3}Unemp{l}_{t-1jk}+{\mu }_{0jk}+{\mu }_{1jk}Unemp{l}_{t-1jk}+{\omega }_t+{\in }_{tjk}$(1)

(2) $Y_{tjk}={\beta }_0+{\beta }_{1}X{a}_{tjk}+{\beta }_{2}X{b}_{t-1jk}+{\beta }_{3}Z{a}_{tk}+{\beta }_{4}Z{b}_{t-1k}+{\beta }_{5}Unemp{l}_{t-1jk}+{\mu }_{0jk}+{\mu }_{1jk}Unemp{l}_{t-1jk}+{\omega }_t+{\in }_{tjk}$(2)

where $t$ indexes year, is the region and $k$ denotes the country; $Y_{tjk}$ is the electoral results of RLPs in a NUTS 2 region; is political control proxies at regional level (NUTS 2; Turnout; Green; SocialDem; FarRight; Regionalist); $X{b}_{t-1jk}$ is socio-economic control proxies at regional level (NUTS 2; log GDPpc; YoungPop; ElderlyPop; Primary; Tertiary; ImportShockChina; Immigration); $Z{a}_{tk}$ is political proxies at state level (Threshold; PolarVote); is socio-economic proxies at state level (Growth; GINI); $Unemp{l}_{t-1jk}$ represents regional unemployment rates (GeneralUnempl, YoungUnempl, LongUnempl, MaleUnempl) that are tested separately; ${\mu }_{0jk}$ is random intercepts; ${\mu }_{1jk}$ is random slopes; ${\omega }_t$ is dummies for election years and ${\in }_{tjk}$ represents an unobserved error term.

Results

This section analyses the influence of regional factors on the election results of RLPs. First, the results of multilevel regression analysis for the parliamentary elections with graphic addition of regional differences are presented and then the regression outputs are described for the period before and after the beginning of the financial crisis.

Whole period

Table 1 shows the outputs of multilevel regression analysis. Models 1 to 4 contain only regional characteristics, while models 5 to 8 are supplemented by the influence of four indicators at the state level.

Table 1. Influence of regional unemployment rates on electoral results of RLPs in parliamentary elections.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
General unemployment rate	0.39*** (2.64)				0.33** (2.41)
Youth unemployment rate		−0.12 (−0.76)				−0.02 (−0.15)
Long-term unemployment rate			0.65*** (3.13)				0.58*** (3.53)
Male unemployment rate				0.37*** (2.67)				0.33** (2.49)
Turnout	−0.12** (−2.28)	−0.12* (−1.94)	−0.14** (−2.45)	−0.12** (−2.36)	−0.13*** (−3.27)	−0.13** (−2.52)	−0.14*** (−3.27)	−0.14*** (−3.76)
Share of green parties	−0.52*** (−3.46)	−0.59*** (−3.35)	−0.46*** (−3.27)	−0.52*** (−3.4)	−0.57*** (−4.4)	−0.61*** (−4.27)	−0.53*** (−4.32)	−0.59*** (−4.61)
Share of social democratic parties	−0.25*** (−3.18)	−0.39*** (−3.95)	−0.27*** (−3.36)	−0.25*** (−3.03)	−0.18*** (−2.78)	−0.29*** (−3.34)	−0.19*** (−3.05)	−0.19*** (−2.71)
Share of far-right parties	−0.09* (−1.71)	−0.14** (−2.39)	−0.04 (−0.71)	−0.09 (−1.53)	−0.23*** (−4.37)	−0.25*** (−5.23)	−0.2*** (−3.48)	−0.23*** (−4.09)
Share of regionalist parties	−0.1*** (−2.62)	−0.16*** (−3.8)	−0.09** (−2.39)	−0.1** (−2.39)	−0.08** (−2.26)	−0.11*** (−3.93)	−0.07** (−2.07)	−0.08** (−2.15)
Log GDP per capita	−3.71* (−1.77)	−6.58*** (−3.31)	−3.32* (−1.79)	−3.81* (−1.72)	−3.52** (−2.36)	−7.08*** (−3.88)	−3.0** (−2.21)	−3.22** (−2.15)
Share of young population	0.43 (1.11)	0.35 (0.97)	0.61* (1.7)	0.46 (1.18)	0.19 (0.67)	0.19 (0.67)	0.24 (0.9)	0.21 (0.74)
Share of elderly population	0.42 (1.4)	0.23 (0.89)	0.39 (1.44)	0.44 (1.53)	0.19 (0.72)	0.1 (0.43)	0.15 (0.64)	0.22 (0.9)
Share of population with primary education	−0.1* (−1.65)	−0.13* (−1.79)	−0.11** (−2.05)	−0.09 (−1.46)	−0.25*** (−3.96)	−0.29*** (−3.4)	−0.22*** (−3.69)	−0.22*** (−3.82)
Share of population with tertiary education	0.002 (0.03)	0.04 (0.54)	−0.04 (−0.56)	0.01 (0.12)	−0.04 (−0.5)	0.02 (0.29)	−0.05 (−0.69)	−0.03 (−0.41)
Import shock from China	0.13** (2.02)	0.14** (2.3)	0.13** (2.11)	0.13* (1.86)	0.13* (1.83)	0.15** (2.02)	0.13* (1.8)	0.12* (1.7)
Share of immigrants	0.1*** (2.74)	0.09*** (3.31)	0.11*** (2.71)	0.1** (2.53)	0.07** (2.5)	0.06** (2.35)	0.07** (2.48)	0.07** (2.32)
Threshold (state level)					−1.41** (−1.96)	−1.49 (−1.51)	−1.1 (−1.37)	−1.22* (−1.66)
Voter polarisation index (state level)					53.91*** (6.29)	53.01*** (5.94)	52.84*** (6.4)	57.43*** (6.85)
GDP per capita growth (state level)					−0.06 (−0.45)	−0.26* (−1.74)	−0.24* (−1.73)	−0.04 (−0.27)
GINI (state level)					0.13 (0.52)	−0.07 (−0.28)	0.13 (0.51)	0.16 (0.6)
Variance	0.26	0.01	0.59	0.23	0.15	0.01	0.31	0.15
ICC Country	0.61	0.71	0.7	0.67	0.61	0.77	0.67	0.65
ICC NUTS 2|Country	0.87	0.83	0.87	0.86	0.88	0.87	0.87	0.87
Countries	20	20	20	20	20	20	20	20
NUTS 2	198	197	198	198	198	197	198	198
Observations	889	865	866	867	889	865	866	867
AIC	4819.15	4851.58	4661.92	4735.87	4611.39	4691.64	4456.73	4510.24
BIC	4991.6	5023.04	4833.42	4907.41	4802.99	4882.14	4647.29	4700.84

*, ** and *** denote significance at 10%, 5% and 1% level, respectively; t-statistics are reported in parentheses; Variance means variance of unemployment rates at regional level; ICC Country means interclass correlation at country level; ICC NUTS 2|Country means interclass correlation at region-within-country level; AIC means the Akaike information criterion; BIC means the Bayesian information criterion; constant and time (election-year) dummies are not reported.

Regarding political control variables, the electoral success of RLPs is negatively affected by higher turnout (Doležalová et al. 2017; Finseraas and Vernby 2014) and higher electoral support for green, social-democratic and far-right parties (March and Rommerskirchen 2015). Thus, competition in the left spectrum (green and social democracy) and importance of societal issues (far right) can be observed both at regional (Vampa 2020) and national (March and Rommerskirchen 2015) levels. At the same time the results confirm that the RLPs have limited scope for electoral success in regions with strong regionalist parties (Knutsen 2010; Mazzoleni and Mueller 2016).

In the case of socio-economic regional characteristics, RLPs achieve a higher election result in regions with a lower economic level (Vail and Bowyer 2012), in regions with a lower proportion of people with primary education (Visser et al. 2014), in regions with a significant loss of jobs in industry due to pressures from globalisation (proxy import shock from China; Fagerholm 2018; March and Rommerskirchen 2015; Vampa 2020) and in regions with a higher proportion of immigrants. In the case of immigration, the results are different from most empirical findings (e.g. Bedock and Vasilopoulos 2015; Edo et al. 2019; Vampa 2020). An explanation is offered in that there is competition in the left spectrum, which means that voters will move from the centre-left to the RLPs if they have strong anti-immigration attitudes (Santana and Rama 2018).

The remaining three regional variables (YoungPop, ElderlyPop, Tertiary) can be considered statistically insignificant. Looking at variables at the state level, RLPs are more successful in a political system with strong polarisation (March and Rommerskirchen 2015), while the level of income inequality is likely to have no effect (Rooduijn and Burgoon 2018).

This paper focuses on the impact of unemployment rates. The results of the regression analysis suggest that the unemployment rate is a significant factor (Doležalová et al. 2017; March and Rommerskirchen 2015; Proaño Acosta, Peña, and Saalfeld 2020). Thus, RLPs achieve better election results in regions with a higher general rate of unemployment or with a higher share of unemployed men or the long-term unemployed. Of these three indicators, the long-term unemployment rate has the highest impact on RLPs (see regression coefficients and Akaike and Bayesian information criterion values). The closer relationship between the long-term unemployment rate and the electoral support for RLPs in European regions is analysed in more detail in the following paragraphs (Figures 2 and 3), while the insignificant impact of the young unemployed is explained in the following section.

Figure 2. Influence of long-term unemployment rate on electoral results of RLPs in NUTS 2 regions within European countries.

The number of observations is reported in parentheses; white dot represents the median; thick black bar represents the interquartile range. Source: Own calculation based on data from the seventh regression equation in Table 1.

Figure 3. List of regions with the highest positive and negative impact of long-term unemployment rate.

Own calculation based on data from the seventh regression equation in Table 1.

Figures 2 and 3 present the average effect of a 1% increase in the long-term unemployment rate on electoral support for RLPs in the individual regions. The calculation of the values in both graphs is based on data from the seventh regression equation (Table 1) and is the sum of the random slope coefficient and beta coefficient for each election event.

Figure 2 shows the impact of the regional long-term unemployment rate within European countries. The values demonstrate that in most NUTS 2 regions (179 out of 198), a 1% increase in the long-term unemployment rate will lead to a rise in support for RLPs of 0.2% to 1.9%, with a median value of 0.59%. The highest median impact can be observed in Greece (1.18%) and in Czechia (0.92%), while the largest regional differences are typical for France, Greece, Italy, Portugal and Spain. In the case of the general unemployment rate, there would be an increase in electoral support of 0.2% to 1.3% in 146 out of 198 NUTS 2 regions. Also, there are five states, Cyprus, France, Italy, Romania and Spain, in which the regional unemployment rate can have a negative effect. For these reasons, Figure 3 analyses the regions in which the long-term unemployment rate has the highest positive and negative impacts.

Figure 3 is divided into two parts: the left-hand side includes the ten regions with the highest average positive effect of the long-term unemployment rate, while the right-hand side contains the nine regions in which the average impact of the long-term unemployment rate is negative.

Within the left-hand group there are six Greek regions (Central Greece, Crete, Epirus, Ionian Islands, South Aegean and Western Greece), two Spanish (Basque Country and Navarra), one German (Thuringia) and one Portuguese (Alentejo). That the unemployment rate has the highest impact on election results for the RLPs in Greek regions is not surprising given the effects of the global financial crisis, which has led to a significant increase in electoral support for SYRIZA (Bedock and Vasilopoulos 2015; Hernández and Kriesi 2016; Hobolt and De Vries 2016). Simultaneously, Greece has long been one of the countries in which RLPs have held a strong position (Van Hamme, Vandermotten, and Lockhart 2018). In more detail, all thirteen Greek regions are located in the top thirty regions with the highest influence from the long-term unemployment rate. This could significantly affect the results of regression analysis. For this reason, an additional regression analysis, in which the Greek regions were omitted, was performed as a robustness check (see Table A4). The results do not differ significantly from the main findings in Table 1; hence, Greek regions do not significantly distort the achieved results. The remaining four regions can be characterised as regions with the centre-periphery cleavage associated with strong leftist tendencies, the German Thuringia (Dorn et al. 2020), the Portuguese Alentejo (Vandermotten and Lockhart 2000) and the Spanish Basque Country and Navarra (Vampa 2020).

The right-hand part of the graph includes the ten regions in which the growth of unemployment led to worse results for RLPs. In Cyprus, a 1% increase in the rate of long-term unemployment can result in a drop in electoral support for RLPs of up to 0.9%. The finding is related to the specific situation in Cyprus, where the Progressive Party of Working People (AKEL) ruled at a time of low unemployment in the first decade of the 21^st century, while the financial crisis between 2012 and 2013 resulted in a significant rise in unemployment (general from 7% to 16% and long-term from 1% to 8%), which was reflected in a decline in voter preferences in the 2016 parliamentary elections (Katsourides 2016). For the remaining regions, the negative but also minimal impact may be because there are regions in which RLPs have had lower electoral support for a long time, namely peripheral parts of France (Corsica), Italy (Calabria, Campania and Sicily) and Spain (Ceuta). In the other three regions, this negative impact can be explained by the fact that these regions are the bases for the Spanish Socialist Workers Party (Castilla–La Mancha and Extremadura) and the Hungarian minority in the Romanian region, Centru (Haydukiewicz 2011).

Before and after the beginning of the global financial crisis

The second part of the regression analysis compares the effects of unemployment rates in the periods before and after the beginning of the global financial crisis. The results are presented in Tables 2 and A5. Table 2 presents the outputs for the control variables examined at regional and national level, while the models with only regional variables are given in Table A5 in the Appendix. In both periods, the electoral success of RLPs was negatively affected by competition from the green parties and positively by the unemployment rates (general, long-term and male). Thus, these findings confirm previous conclusions in this paper.

Table 2. Influence of regional unemployment rates on electoral results of RLPs in parliamentary elections before and after the beginning of the global financial crisis.

	2002–2008				2009–2019
	(5)	(6)	(7)	(8)	(5)	(6)	(7)	(8)
General unemployment rate	0.67*** (5.3)				0.53*** (3.29)
Youth unemployment rate		0.33*** (4.44)				0.18 (1.19)
Long-term unemployment rate			0.83*** (3.1)				0.63*** (2.79)
Male unemployment rate				0.65*** (6.0)				0.47*** (3.07)
Turnout	−0.01 (−0.12)	0.01 (0.19)	−0.04 (−0.95)	−0.03 (−0.55)	−0.13** (−2.54)	−0.16*** (−2.63)	−0.15*** (−3.0)	−0.15*** (−2.84)
Share of green parties	−0.37** (−2.43)	−0.64*** (−3.76)	−0.21* (−1.71)	−0.49*** (−3.02)	−0.6*** (−3.98)	−0.65*** (−4.49)	−0.57*** (−3.72)	−0.62*** (−4.17)
Share of social democratic parties	0.01 (0.19)	0.05 (0.97)	−0.04 (−1.33)	0.02 (0.46)	−0.21*** (−3.43)	−0.28*** (−3.86)	−0.22*** (−3.3)	−0.23*** (−3.68)
Share of far-right parties	−0.11 (−1.37)	−0.09 (−0.99)	0.04 (0.77)	−0.11 (−1.34)	−0.16** (−2.2)	−0.2*** (−3.07)	−0.17** (−2.31)	−0.17** (−2.2)
Share of regionalist parties	−0.01 (−0.4)	0.02 (0.58)	−0.01 (−0.65)	−0.01 (−0.69)	−0.09** (−2.02)	−0.12*** (−2.74)	−0.1** (−2.18)	−0.1** (−2.12)
Log GDP per capita	4.04 (1.44)	−0.98 (−0.25)	0.55 (0.23)	3.34 (1.33)	−1.24 (−0.83)	−4.31*** (−2.72)	−1.98 (−1.4)	−1.66 (−1.12)
Share of young population	−0.03 (−0.07)	0.01 (0.03)	0.07 (0.21)	−0.06 (−0.17)	−0.36 (−1.0)	−0.31 (−0.83)	−0.29 (−0.77)	−0.33 (−0.89)
Share of elderly population	0.16 (0.61)	0.17 (0.46)	0.13 (0.76)	0.11 (0.5)	−0.11 (−0.36)	−0.03 (−0.11)	−0.14 (−0.47)	−0.08 (−0.28)
Share of population with primary education	−0.08 (−1.1)	−0.21* (−1.73)	−0.08 (−1.48)	−0.09 (−1.21)	−0.28*** (−4.18)	−0.33*** (−3.72)	−0.26*** (−3.52)	−0.25*** (−3.81)
Share of population with tertiary education	0.02 (0.21)	0.24 (1.34)	0.001 (0.00)	0.04 (0.4)	−0.004 (−0.06)	0.07 (0.81)	0.01 (0.1)	0.01 (0.16)
Import shock from China	1.39* (1.86)	0.36 (1.47)	0.78* (1.68)	0.99 (1.54)	0.13*** (2.71)	0.15*** (2.7)	0.13** (2.42)	0.12** (2.47)
Share of immigrants	−0.001 (−0.01)	−0.07 (−0.81)	0.02 (0.29)	0.005 (0.08)	0.08* (1.76)	0.09* (1.7)	0.08* (1.66)	0.08* (1.73)
Threshold (state level)	2.7** (2.21)	−0.69 (−0.74)	1.71* (1.85)	1.84** (2.24)	−0.49 (−0.53)	−0.38 (−0.33)	−0.48 (−0.48)	−0.36 (−0.39)
Voter polarisation index (state level)	33.65** (2.23)	59.22*** (3.31)	6.316 (0.81)	43.82*** (2.63)	58.19*** (5.26)	63.06*** (4.51)	59.89*** (5.2)	61.56*** (5.5)
GDP per capita growth (state level)	−1.61 (−1.58)	0.27 (0.20)	−0.85 (−1.07)	−1.11 (−1.01)	0.18 (0.98)	0.19 (1.05)	0.02 (0.11)	0.22 (1.19)
GINI (state level)	1.43** (2.1)	0.83 (1.16)	1.55*** (3.47)	1.34** (2.06)	1.31*** (3.14)	1.73*** (3.51)	1.24*** (2.68)	1.38*** (3.2)
Variance	0.16	0.01	0.24	0.14	0.11	0.002	0.24	0.11
ICC Country	0.96	0.75	0.98	0.93	0.66	0.74	0.69	0.79
ICC NUTS 2|Country	0.99	0.83	0.98	0.98	0.9	0.88	0.89	0.89
Countries	16	16	16	16	19	19	19	19
NUTS 2	175	176	174	174	192	191	192	192
Observations	342	325	325	329	547	540	541	538
AIC	1665.71	1694.87	1521.84	1609.62	2904.79	2955.63	2862.19	2863.17
BIC	1776.92	1804.6	1631.58	1719.71	3046.84	3097.25	3003.87	3004.67

*, ** and *** denote significance at 10%, 5% and 1% level, respectively; t-statistics are reported in parentheses; Variance means variance of unemployment rates at regional level; ICC Country means interclass correlation at country level; ICC NUTS 2|Country means interclass correlation at region-within-country level; AIC means the Akaike information criterion; BIC means the Bayesian information criterion; constant and time (election-year) dummies are not reported.

The positive impact of youth unemployment or positive influence of a larger proportion of the young population within the regions surveyed have not been confirmed. These findings differ from the current literature (Bedock and Vasilopoulos 2015; Van der Brug, Hobolt, and De Vreese 2009). The statistically insignificant influence of the young population can be explained by the fact that young voters prefer new topics such as ecology to traditional cleavages (Walczak, Van der Brug, and De Vries 2012).

If comparing the period before and after the beginning of the financial crisis, it can be seen that the crisis has amplified the influence of several factors that determined the electoral success of RLPs, such as more polarised political systems (Hernández and Kriesi 2016), rising income inequality (Winkler 2019), the increasing negative effects of globalisation in typical European industries (March and Rommerskirchen 2015; Vampa 2020) and greater competition, not only with other parties of the left but also regionalist (Bosco and Verney 2012) and, under certain circumstances, far-right parties (March and Rommerskirchen 2015).

Figure 4 shows the ten regions in which the long-term unemployment rate had the highest average positive impact on the electoral success of the RLPs from 2002 to 2008 (left side) and from 2009 to 2019 (right side). The calculation of the values is based on data from the seventh regression equation for each period in Table 2 and is the sum of the random slope coefficient and beta coefficient for each election event. The presented values are the average for each region. The values demonstrate that in all the presented regions, a 1% increase in the long-term unemployment rate will lead to a rise in support for RLPs of 1.1% to 2.6%.

Figure 4. List of 10 regions with the highest impact of general unemployment rate before and after the beginning of the global financial crisis.

Left figure (2002–2008), right figure (2009–2019). Source: Own calculation based on data from the seventh regression equation for each period in Table 2.

In the period before the financial crisis, the strongest influence of the unemployment rate on voter support was in the regions where the RLPs had historically achieved above-average election results (e.g. French Auvergne, Italian Tuscany and Portuguese Lisbon metropolitan area; Van Hamme, Vandermotten, and Lockhart 2018) while, simultaneously, the other regions could be considered as peripheral parts of the states in question, namely Alentejo (Portugal), Brandenburg, Mecklenburg-Western Pomerania and Thuringia (Germany), Ionian Islands (Greece), North and East Finland (Finland) and Upper Norrland (Sweden).

Regarding developments after 2008, despite a partial change in the regional order, there are still regions in which the centre-periphery cleavage associated with strong leftist tendencies can be observed. These regions include the Czech Northwest (Pink 2012), German Brandenburg, Bremen, Mecklenburg-Western Pomerania, Saxony-Anhalt and Thuringia regions (Dorn et al. 2020), the Greek regions of Crete and Ionian Islands (Van Hamme, Vandermotten, and Lockhart 2018) and the Spanish Basque Country and Navarra (Vampa 2020).

Conclusions

This paper aimed to discover whether unemployment rates influenced the electoral support of RLPs in the first two decades of the 21^st century. The election results of RLPs that received at least 1% of the votes in at least one parliamentary election at the national level were analysed at the NUTS 2 region level through multilevel regression analysis. The categorisation of these parties was based on the methodology of the ParlGov database (Döring and Manow 2020). Four indicators of unemployment were worked on – general, youth, long-term and male unemployment rates. Eighty-six parliamentary elections from 2002 to 2019 were analysed.

The basic findings are in line with the empirical literature (Doležalová et al. 2017; March and Rommerskirchen 2015; Proaño Acosta, Peña, and Saalfeld 2020), and the results, therefore, suggest that RLPs are more successful in parliamentary elections in regions with higher rates of unemployment (general, long-term, male) across the period under review. A 1% increase in the long-term unemployment rate would lead to an increase in support for RLPs of 0.2% to 1.9% in most NUTS 2 regions (179 out of 198). The highest influence was particularly identified in Germany (Brandenburg, Mecklenburg-Western Pomerania and Thuringia), Greece (Crete, Ionian Islands and South Aegean), Portugal (Alentejo) and Spain (Basque Country and Navarra). The centre-periphery cleavage associated with historically strong leftist tendencies is typical for all these regions (Van Hamme, Vandermotten, and Lockhart 2018). In the case of the general unemployment rate, there would be an increase in electoral support of 0.2% to 1.3% in 146 out of 198 NUTS 2 regions. In contrast, a rise in the rate of unemployment in Cyprus could have a negative effect on the electoral success of RLPs. This can be explained by the fact that the Progressive Party of Working People (AKEL) was in government during the recession, resulting in a significant decline in party support at the next election (Katsourides 2016). It should be noted also that the insignificant influence of youth unemployment rates (or share of young population in general) can be explained by the fact that young voters prefer new topics such as ecology to traditional cleavages (Walczak, Van der Brug, and De Vries 2012).

Focusing on the political control variables, the impact of turnout is negative or ambiguous, which is in accordance with the literature, while simultaneously there was competition from green, social democratic, far-right and regionalist parties throughout the period, especially during the global financial crisis. Turning to variables at the state level, RLPs are more successful in political systems that have a strong polarisation (March and Rommerskirchen 2015).

Regarding socio-economic regional characteristics, the RLPs achieve better election results in regions with a lower economic level (Vail and Bowyer 2012), in regions with a significant loss of jobs in industry due to the pressures of globalisation (proxy representing import shock from China; Fagerholm 2018; March and Rommerskirchen 2015; Vampa 2020) and in regions with higher proportion of immigrants (Santana and Rama 2018). The positive impact of the latter two factors was amplified during the global financial crisis.

The main thrust of this paper is in the analysis of electoral support in parliamentary elections at the regional level, which is less represented in the thematic literature compared to the individual or national level. The findings suggest that regional rates of unemployment impact the electoral success of RLPs, but the influence is, first, conditioned by historical developments and, second, strengthened by the global financial crisis and the European debt crisis. This paper deals with the NUTS 2 regional level, which offers much interesting data. However, the NUTS 3 level would be more appropriate because the analysis performed would contain several times more observations and these regions are more naturally formed in most countries compared with the artificially created NUTS 2 regions that are mainly used for statistical comparisons. Within the NUTS 3 regions, it is only possible to work with electoral characteristics, while socio-economic data is available either partially (the OECD Regional Database) or not at all (Eurostat), according to a uniform methodology. An analysis of the RLPs that gained at least 1% of the votes in at least one parliamentary election is another benefit of the article, as all relevant parties that participated in the electoral process during the period under review are included.

As a possible extension of the research, this paper proposes three options. First, distinction should be made between the results for the traditional RLPs and for the left-wing populist parties. Second, a comparison of the results for RLPs with the results of the green, social democratic, far-right and regionalist parties should be offered. Third, an analysis at the level of NUTS 3 regions is proposed.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Appendix

Table A1 List of NUTS 2 regions.

Country	NUTS 2 regions
Austria	Burgenland (AT), Niederösterreich, Wien, Kärnten, Steiermark, Oberösterreich, Salzburg, Tirol, Vorarlberg
Belgium	Région de Bruxelles-Capitale/Brussels, Hoofdstedelijk Gewest, Prov. Antwerpen, Prov. Limburg (BE), Prov. Oost-Vlaanderen, Prov. Vlaams-Brabant, Prov. West-Vlaanderen, Prov. Brabant Wallon, Prov. Hainaut, Prov. Liège, Prov. Luxembourg (BE), Prov. Namur
Croatia	Jadranska Hrvatska, Kontinentalna Hrvatska
Czechia	Praha, Strední Cechy, Jihozápad, Severozápad, Severovýchod, Jihovýchod, Strední Morava, Moravskoslezsko
Cyprus	Kypros
Denmark	Hovedstaden, Sjælland, Syddanmark, Midtjylland, Nordjylland
Finland	Länsi-Suomi, Helsinki-Uusimaa, Etelä-Suomi, Pohjois- ja Itä-Suomi
France	Île de France, Centre – Val de Loire, Bourgogne, Franche-Comté, Basse-Normandie, Haute-Normandie, Nord-Pas-de-Calais, Picardie, Alsace, Champagne-Ardenne, Lorraine, Pays-de-la-Loire, Bretagne, Aquitaine, Limousin, Poitou-Charentes, Languedoc-Roussillon, Midi-Pyrénées, Auvergne, Rhône-Alpes, Provence-Alpes-Côte d’Azur, Corse
Germany	Stuttgart, Karlsruhe, Freiburg, Tübingen, Oberbayern, Niederbayern, Oberpfalz, Oberfranken, Mittelfranken, Unterfranken, Schwaben, Berlin, Brandenburg, Bremen, Hamburg, Darmstadt, Gießen, Kassel, Mecklenburg-Vorpommern, Braunschweig, Hannover, Lüneburg, Weser-Ems, Düsseldorf, Köln, Münster, Detmold, Arnsberg, Koblenz, Trier, Rheinhessen-Pfalz, Saarland, Dresden, Chemnitz, Leipzig, Sachsen-Anhalt, Schleswig-Holstein, Thüringen
Greece	Anatoliki Makedonia, Thraki, Kentriki Makedonia, Dytiki Makedonia, Ipeiros, Thessalia, Ionia Nisia, Dytiki Ellada, Sterea Ellada, Peloponnisos, Attiki, Voreio Aigaio, Notio Aigaio, Kriti
Hungary	Budapest, Pest, Közép-Dunántúl, Nyugat-Dunántúl, Dél-Dunántúl, Észak-Magyarország, Észak-Alföld, Dél-Alföld
Italy	Piemonte, Valle d’Aosta/Vallée d’Aoste, Liguria, Lombardia, Provincia Autonoma di Bolzano/Bozen, Provincia Autonoma di Trento, Veneto, Friuli-Venezia Giulia, Emilia-Romagna, Toscana, Umbria, Marche, Lazio, Abruzzo, Molise, Campania, Puglia, Basilicata, Calabria, Sicilia, Sardegna
Lithuania	Sostines regionas, Vidurio ir vakaru Lietuvos regionas
Luxembourg	Luxembourg
Netherlands	Groningen, Friesland (NL), Drenthe, Overijssel, Gelderland, Flevoland, Utrecht, Noord-Holland, Zuid-Holland, Zeeland, Noord-Brabant, Limburg (NL)
Portugal	Norte, Algarve, Centro (PT), Área Metropolitana de Lisboa, Alentejo, Região Autónoma dos Açores (PT), Região Autónoma da Madeira (PT)
Romania	Nord-Vest, Centru, Nord-Est, Sud-Est, Sud – Muntenia, Bucuresti – Ilfov, Sud-Vest Oltenia, Vest
Slovakia	Bratislavský kraj, Západné Slovensko, Stredné Slovensko, Východné Slovensko
Slovenia	Vzhodna Slovenija, Zahodna Slovenija
Spain	Galicia, Principado de Asturias, Cantabria, País Vasco, Comunidad Foral de Navarra, La Rioja, Aragón, Comunidad de Madrid, Castilla y León, Castilla-la Mancha, Extremadura, Cataluña, Comunidad Valenciana, Illes Balears, Andalucía, Región de Murcia, Ciudad Autónoma de Ceuta (ES), Ciudad Autónoma de Melilla (ES), Canarias (ES)
Sweden	Stockholm, Östra Mellansverige, Småland med öarna, Sydsverige, Västsverige, Norra Mellansverige, Mellersta Norrland, Övre Norrland

Table A2 List of RLPs.

Country	Radical left parties
Austria	Communist Party of Austria (Kommunistische Partei Österreichs)
Belgium	Workers’ Party of Belgium (Partij van de Arbeid – Parti du Travail de Belgique)
Croatia	Croatian Labourists – Labour Party (Hrvatski laburisti – Stranka rada)
Czechia	Communist Party of Bohemia and Moravia (Komunistická strana Čech a Moravy)
Cyprus	Progressive Party of Working People (Anorthotikó Kómma Ergazómenou Laoú)
Denmark	Red-Green Alliance (Enhedslisten – De Rød-Grønne)
Finland	Left Alliance (Vasemmistoliitto – Vänsterförbundet)
France	French Communist Party (Parti communiste française) Left Front (Front de gauche) Other far-left (Extrême gauche) Revolutionary Communist League (Ligue communiste révolutionnaire) Unbowed France (La France Insoumise) Workers’ Struggle (Lutte ouvrière)
Germany	The Left Party (Die Linke)
Greece	Coalition of the Radical Left (Synaspismós Rizospastikís Aristerás) Communist Party of Greece (Kommounistikó Kómma Elládas) European Realistic Disobedience Front (MeRA25) Front of the Greek Anticapitalist Left (Antikapitalistiki Aristeri Synergasia gia tin Anatropi) Popular Unity/Course of Freedom (Laïkí Enótita/Plefsi Eleftherias)
Hungary	Hungarian Workers’ Party (Magyar Munkáspárt)
Italy	Communist Refoundation Party (Partito della Rifondazione Comunista) Left (Sinistra) Party of the Italian Communists (Partito dei Comunisti Italiani)
Lithuania	Socialist People’s Front (Socialistinis liaudies frontas)
Luxembourg	Communist Party of Luxembourg (Kommunistesch Partei Lëtzebuerg) The Left (Déi Lénk)
Netherlands	Socialist Party (Socialistiese Partij)
Portugal	Bloc of the Left (Bloco de Esquerda) Communist Party of the Portuguese Workers/Re-Organised Movement of Party of the Proletariat (Comunista dos Trabalhadores Portugueses/Movimento Reorganizativo do Partido do Proletariado) Unified Democratic Coalition (Coligação Democrático Unitária)
Romania	People’s Party – Dan Diaconescu (Partidul Poporului – Dan Diaconescu)
Slovakia	Communist Party of Slovakia (Komunistická strana Slovenska)
Slovenia	The Left (Levica) United Left (Združena levica)
Spain	Communist Party/United Left (Partido Communista/Izquierda Unida) United We Can/We Can (Unidas Podemos/Podemos)
Sweden	Left Party (Vänsterpartiet)

Table A3 List of parliamentary elections.

Country	Parliamentary elections
Austria	2002, 2006, 2008, 2013, 2017, 2019
Belgium	2007, 2010, 2014, 2019
Croatia	2011
Czechia	2002, 2006, 2010, 2013, 2017
Cyprus	2006, 2011, 2016
Denmark	2005, 2007, 2011, 2015, 2019
Finland	2003, 2007, 2011, 2015, 2019
France	2002, 2007, 2012, 2017
Germany	2002, 2005, 2009, 2013, 2017
Greece	2004, 2007, 2009, 2012 (May), 2015 (January), 2019
Hungary	2002
Italy	2006, 2008, 2013, 2018
Lithuania	2008, 2012
Luxembourg	2004, 2009, 2013, 2018
Netherlands	2002, 2003, 2006, 2010, 2012, 2017
Portugal	2002, 2005, 2009, 2011, 2015, 2019
Romania	2012
Slovakia	2002, 2006, 2010, 2012, 2016
Slovenia	2014, 2018
Spain	2004, 2008, 2011, 2015, 2016, 2019 (April)
Sweden	2002, 2006, 2010, 2014, 2018

Table A4 Influence of regional unemployment rates on electoral results of RLPs in parliamentary elections (without Greek regions).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
General unemployment rate	0.27* (1.82)				0.21 (1.57)
Youth unemployment rate		−0.09 (−0.47)				0.06 (0.4)
Long-term unemployment rate			0.49* (1.9)				0.42** (2.27)
Male unemployment rate				0.29** (2.1)				0.23* (1.83)
Turnout	−0.09 (−1.53)	−0.06 (−1.18)	−0.1 (−1.58)	−0.09 (−1.53)	−0.07 (−1.58)	−0.04 (−0.84)	−0.09* (−1.81)	−0.08* (−1.86)
Share of green parties	−0.4*** (−2.97)	−0.44*** (−3.21)	−0.38*** (−2.84)	−0.41*** (−3.04)	−0.52*** (−4.25)	−0.51*** (−4.19)	−0.5*** (−4.26)	−0.53*** (−4.49)
Share of social democratic parties	−0.13** (−2.23)	−0.14** (−2.2)	−0.11** (−2.33)	−0.12** (−2.18)	−0.08* (−1.75)	−0.08* (−1.81)	−0.09** (−2.11)	−0.08* (−1.67)
Share of far-right parties	−0.07 (−1.07)	−0.14** (−2.14)	−0.03 (−0.42)	−0.06 (−0.86)	−0.23*** (−3.55)	−0.25*** (−3.91)	−0.19*** (−2.71)	−0.22*** (−3.37)
Share of regionalist parties	−0.07** (−2.03)	−0.13*** (−4.32)	−0.05* (−1.83)	−0.06* (−1.74)	−0.07* (−1.76)	−0.1*** (−2.84)	−0.06* (−1.66)	−0.06 (−1.64)
Log GDP per capita	−3.57 (−1.51)	−2.74 (−1.48)	−3.29 (−1.56)	−3.41 (−1.36)	−2.73* (−1.65)	−2.49 (−1.45)	−2.38 (−1.52)	−2.39 (−1.43)
Share of young population	0.25 (0.49)	−0.14 (−0.33)	0.46 (0.96)	0.29 (0.56)	−0.03 (−0.08)	−0.21 (−0.62)	0.11 (0.3)	−0.01 (−0.02)
Share of elderly population	0.36 (0.93)	−0.07 (−0.22)	0.34 (1.0)	0.37 (1.01)	0.058 (0.18)	−0.17 (−0.66)	0.07 (0.23)	0.091 (0.3)
Share of population with primary education	−0.07 (−1.23)	−0.07 (−1.02)	−0.09* (−1.72)	−0.06 (−1.09)	−0.25*** (−4.51)	−0.29*** (−4.23)	−0.23*** (−4.26)	−0.22*** (−4.4)
Share of population with tertiary education	0.02 (0.25)	0.01 (0.12)	−0.03 (−0.37)	0.03 (0.38)	−0.03 (−0.43)	−0.02 (−0.21)	−0.05 (−0.73)	−0.02 (−0.24)
Import shock from China	0.11 (1.63)	0.09 (1.54)	0.11* (1.76)	0.11 (1.55)	0.1 (1.5)	0.09 (1.52)	0.1 (1.57)	0.1 (1.47)
Share of immigrants	0.12*** (2.99)	0.09*** (3.28)	0.12*** (2.9)	0.12*** (3.02)	0.08*** (3.1)	0.07*** (2.91)	0.08*** (3.17)	0.08*** (3.19)
Threshold (state level)					−1.22* (−1.67)	−1.34 (−1.5)	−1.07 (−1.34)	−1.06 (−1.35)
Voter polarisation index (state level)					49.48*** (5.57)	48.95*** (6.45)	46.99*** (5.35)	52.06*** (6.09)
GDP per capita growth (state level)					0.11 (0.46)	0.12 (0.43)	0.03 (0.11)	0.12 (0.48)
GINI (state level)					0.32 (1.57)	0.18 (0.82)	0.33 (1.4)	0.33 (1.57)
Variance	0.18	0.08	0.61	0.13	0.04	0.02	0.19	0.03
ICC Country	0.74	0.73	0.82	0.78	0.67	0.71	0.7	0.68
ICC NUTS 2|Country	0.88	0.76	0.9	0.88	0.86	0.85	0.87	0.86
Countries	19	19	19	19	19	19	19	19
NUTS 2	185	184	185	185	185	184	185	185
Observations	811	787	788	791	811	787	788	791
AIC	4285.79	4187.19	4107.03	4196.18	4060.94	3974.29	3912.87	3953.07
BIC	4454.93	4355.24	4275.14	4364.42	4248.87	4161.02	4099.65	4140.0

*, ** and *** denote significance at 10%, 5% and 1% level, respectively; t-statistics are reported in parentheses; Variance means variance of unemployment rates at regional level; ICC Country means interclass correlation at country level; ICC NUTS 2|Country means interclass correlation at region-within-country level; AIC means the Akaike information criterion; BIC means the Bayesian information criterion; constant and time (election-year) dummies are not reported.

Table A5 Influence of regional characteristics on electoral results of RLPs before and after the beginning of the global financial crisis.

	2002–2008				2009–2019
	(1)	(2)	(3)	(4)	(1)	(2)	(3)	(4)
General unemployment rate	0.7*** (4.66)				0.74*** (3.72)
Youth unemployment rate		0.2 (1.52)				0.14 (0.9)
Long-term unemployment rate			0.77*** (2.94)				1.06*** (4.04)
Male unemployment rate				0.66*** (5.53)				0.71*** (3.58)
Turnout	−0.05 (−0.68)	−0.1 (−1.09)	−0.08 (−1.31)	−0.04 (−0.66)	−0.03 (−0.53)	−0.08 (−1.36)	−0.04 (−0.73)	−0.01 (−0.17)
Share of green parties	−0.42** (−2.25)	−0.59*** (−3.2)	−0.19 (−1.61)	−0.5** (−2.55)	−0.58*** (−3.38)	−0.63*** (−3.13)	−0.51*** (−2.88)	−0.61*** (−3.34)
Share of social democratic parties	−0.02 (−0.55)	0.04 (0.78)	−0.04 (−1.09)	−0.01 (−0.15)	−0.26*** (−3.15)	−0.38*** (−3.91)	−0.27*** (−3.06)	−0.28*** (−3.25)
Share of far-right parties	−0.05 (−0.64)	0.04 (0.41)	0.06 (1.55)	−0.03 (−0.4)	−0.08 (−0.97)	−0.13* (−1.74)	−0.04 (−0.59)	−0.0 (−1.03)
Share of regionalist parties	−0.004 (−0.31)	0.03 (0.74)	−0.004 (−0.24)	−0.01 (−0.7)	−0.1** (−2.02)	−0.15*** (−2.61)	−0.1** (−2.13)	−0.11** (−2.11)
Log GDP per capita	2.29 (0.81)	−2.89 (−0.90)	−0.72 (−0.28)	1.9 (0.75)	−0.83 (−0.45)	−5.75*** (−4.22)	−1.09 (−0.61)	−1.31 (−0.77)
Share of young population	−0.35 (−1.01)	0.1 (0.25)	−0.12 (−0.47)	−0.26 (−0.83)	0.024 (0.06)	0.18 (0.41)	0.19 (0.44)	0.03 (0.07)
Share of elderly population	0.09 (0.4)	0.28 (0.9)	0.11 (0.68)	0.07 (0.35)	0.09 (0.26)	0.23 (0.68)	0.13 (0.41)	0.11 (0.31)
Share of population with primary education	−0.07 (−1.04)	−0.04 (−0.61)	−0.08* (−1.7)	−0.07 (−0.96)	−0.19** (−1.96)	−0.17* (−1.68)	−0.19** (−2.06)	−0.15 (−1.64)
Share of population with tertiary education	0.09 (1.1)	0.37** (2.09)	0.02 (0.28)	0.11 (1.23)	−0.04 (−0.46)	0.07 (0.85)	−0.06 (−0.77)	−0.03 (−0.33)
Import shock from China	0.58 (1.0)	0.53 (1.51)	0.16 (0.55)	0.52 (1.05)	0.11** (2.02)	0.14** (2.29)	0.11** (2.07)	0.09* (1.69)
Share of immigrants	0.08 (1.0)	−0.04 (−0.43)	0.12 (1.36)	0.06 (0.72)	0.12** (2.18)	0.13** (2.2)	0.12** (2.15)	0.12** (2.04)
Variance	0.15	0.03	0.26	0.13	0.16	0.01	0.29	0.17
ICC Country	0.9	0.87	0.96	0.89	0.76	0.69	0.81	0.75
ICC NUTS 2|Country	0.96	0.88	0.97	0.95	0.87	0.79	0.88	0.86
Countries	16	16	16	16	19	19	19	19
NUTS 2	175	176	174	174	192	191	192	192
Observations	342	325	325	329	547	540	541	538
AIC	1712.07	1735.11	1552.94	1651.07	3075.28	3144.82	3028.33	3041.05
BIC	1807.94	1829.71	1647.54	1745.97	3200.1	3269.28	3152.84	3165.4

*, ** and *** denote significance at 10%, 5% and 1% level, respectively; t-statistics are reported in parentheses; Variance means variance of unemployment rates at regional level; ICC Country means interclass correlation at country level; ICC NUTS 2|Country means interclass correlation at region-within-country level; AIC means the Akaike information criterion; BIC means the Bayesian information criterion; constant and time (election-year) dummies are not reported.