The influence of economic development on the rise of populism
in Europe
University of Groningen Faculty of Economics and Business
Master’s Thesis
International Economics and Business
Student: Abe Dijkstra ID number: s2060191
Student email: a.dijkstra.30@student.rug.nl Date: 01-02-2019
ii Abstract
The gilets jaunes (yellow vest) protests in France against the government highlight the dissatisfaction that has long been on the rise in Europe, in the form of populism. In the last two decades, many developed democracies have seen a marked strengthening of populist parties. This study analyzes this trend through the votes for populist parties in 2000–2015, measuring on outcomes of 113 parliamentary elections in twenty-six European countries. This study develops an econometric model based on the theoretical economic insecurity perspective theory of Inglehart and Norris (2016). Globalization, automation and business cycles are discussed, economic phenomena influencing economic insecurity. This study finds a positive relationship between high national income inequality and the rising support for populism in European countries. Globalization, automation and to some extent business cycles all play their role in this. The polarization of the labor market, in which wage gains went disproportionately to those at the top and at the bottom of the income and skill distribution, is contributing to the support for populist political parties. Income inequality potentially plays a role in movements we see today, which share beliefs with the populist philosophy.
iii
C
ONTENTS
I. Introduction ... 1
II. Literature Review ... 4
1.1 Defining populism ... 4
1.1.1 Right Wing populism 6 1.1.2 Left Wing populism 7 1.2 Economic insecurity perspective ... 7
1.2.1 Globalization 9 1.2.2 Automation 12 1.2.3 Business cycles 14 1.2.4 Income inequality 15 III. Methodology ... 17 IV. Data ... 20 4.1 Descriptive statistics ... 22 4.2 Econometrical issues ... 25
V. Results and analyses ... 27
1
I.
I
NTRODUCTION
The ‘yellow vest’ movement is active in France and has shut down central Paris for ten weeks in a row since November ‘18.1If one have to pick one word to describe the dominant attitudes toward the political establishment in Europe – to describe attitudes toward politicians and the institutions of the European Union – that word would be ‘dissatisfaction’. The anti-government yellow vest movement started as a protest against pension cuts and fuel tax hikes and demanded democracy. It is organized without help from France’s powerful labor unions or official political parties. The protests are pro-nationalism and anti-globalism and are demanding ‘Frexit’ (the French exit from the European Union). They reject the propaganda of the mainstream media and the debt owed to the European Union (EU) and protest against austerity in general.2
Could this wave of dissatisfaction in European society have been predicted? Why is there such a rising tide of consensus for populist proposals since the start of this century? Is this a global shift in voters’ preferences or emotions? Is it related to economic development, and if so, through what channels?
2
What is driving the simultaneous shift towards populism in so many countries? Figure 1 shows that, particularly in the past decades, populist parties in Europe have seen an increase in voting shares in national elections. This study analyzes the growing populist trend across European countries by studying the political election results in Europe in the 21st century.
Figure 1: Support for populist parties over time among European countries with at least one populist party
Note: source data from Inglehart and Norris (2016).
Two arguments, brought forward by Inglehart and Norris (2016), have recently been used in academic literature to explain modern mass support for populist parties. The economic insecurity perspective focuses on rising levels of income insecurity and unhappiness among those ‘left behind’ from global markets and the cultural perspective emphasizes a generational backlash reacting against long-term shifts in progressive and liberal social values. This study examines the economic insecurity perspective while controlling for cultural factors. The main research question is:
3
The study focuses on the period 2000–2015, because data is available for that period regarding identification of political parties from Kessel (2015) and Inglehart and Norris (2016). They identified fifty populist political parties in twenty-six European countries that participated in several national elections. Populism gained interest and researchers (Arzheimer, 2012; Rooduijn, 2014; Inglehart and Norris, 2016) tried to analyze the rise of populism through characteristics of voters, whereas other researchers (Swank, 2003; Colantone and Stanig, 2017; Rodrik, 2017) have examined economic and trade motives. This brings us to the relevance of this study and the gap it aims to fill.
Previous research was performed at the individual voter-characteristic level, whereas this study examines macroeconomics at an international level. This study develops an econometric model based on the theoretical economic insecurity perspective by Inglehart and Norris (2016). From this point of view, the role of globalization, automation and business cycles is discussed in relationship with increased populist sentiment in Europe. To analyze the data and produce results the country-level fixed effect regression is used. This study finds that high income inequality is mainly influencing the rising support for populist political parties in Europe. Globalization, automation and to some extent business cycles all play their role in this.
4
II. L
ITERATURE
R
EVIEW
This literature review first defines and discusses populism in Europe. The economic insecurity perspective and economic developments influencing it are discussed, followed by an evaluation of various economic research done on populism, giving attention to the research methods and data used. Finally, this section lays out the main areas of interest of this study and hypotheses are formed.
1.1 DEFINING POPULISM
5
those values. Moreover, cosmopolitan ideas like emphasizing open borders and open societies, are combined with liberal values that challenge the authoritarian component of populism.
Figure 2: Heuristic model of party competition in western societies.
Note: source model Inglehart and Norris (2016).
After decades of pro-globalization sounds, another wave is going through Europe. European populists have either increased their numbers in parliament or won elections as heads of governments in several countries, including the Freedom Party (Austria), the People’s Party (Denmark), the National Front (France), the People’s Party (Switzerland), the Progress Party (Norway), the Party for Freedom (the Netherlands) and the Sweden Democrats (Sweden). Analyses of parties in Western Europe have often associated populism with the economic right wing (Mudde, 2016), but it is increasingly recognized that this fails to catch certain core features of populist parties around the world particularly in South and South-East Europe, where populist parties often favor economic left wing policies.
6 1.1.1 Right Wing populism
To date (January 2019), politicians of the right wing populist movements have taken part in the governments of Austria (the Freedom Party of Austria), Hungary (Fidesz) and Sweden (the Sweden Democrats). Figure 3 shows an overview of all countries in the sample and the corresponding total share of populist votes. Before the Sweden Democrats gained seats in government, the party rose from negligible size (1.4%) in 2002 to Sweden’s third-largest party in 2014 (12.9%), and remained third in 2018 with 17.5% of the total votes. These new political entrants share some broad features. Most of them stress traditional values, law and order, and glorify the past. Their programs are typically nationalistic and push a populist anti-immigration and anti-establishment message (Mudde, 2007). In continental Europe, populist movements stress protection from immigrants, often linking them with Islamic terrorism. They also stress the problems of Chinese imports (Guiso et al. 2017). In Figure 2, the ‘economic right’ favor free markets and private enterprise, a more moderate role for the state, deregulation, and low taxation.
Figure 3: Support for populist parties among European countries over time
Note: source data political parties from Kessel(2015) and Inglehart and Norris (2016), election results from Nordsieck (2018). 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 P o p u lis t v o te sh ar e (%) Country Year
Support for populist parties among European countries over time
7 1.1.2 Left Wing populism
To date (January 2019), politicians of the left wing populist movements have gained support in Spain (Podemos) and have gained seats in the governments of Greece (Syriza) and Italy (Five Star Movement). Before Syriza gained seats in government of Greece, the party rose from negligible size (3.3%) in 2004 to the second-largest party in 2012 (16.8%), and in 2015 it became the largest party, with 36.3% share of the total votes. In southern Europe left wing populist movements call for a guaranteed minimum income and are in opposition to European central banking and the European illusion of fiscal discipline (Guiso et al. 2017).
The parties direct their anger towards their country’s establishment or the EU headquarters in Brussels (March and Keith, 2016). Left wing populism opposes capitalism and thereby rejects the presence of large multinational corporations and central banks. The main implication according to Rodrik (2017) is an increasing sentiment of anti-globalization. The horizontal axis depicted in Figure 2 locates the ‘economic left,’ favoring state management of the economy, economic redistribution through progressive taxation, and strong welfare states and public services.
Inglehart and Norris (2016) argue that the presence of populist parties in the political arena is heavily linked to economic insecurity.
1.2 ECONOMIC INSECURITY PERSPECTIVE
8
can be explained as a retro reaction to progressive value change by once-ruling masses of the population (Inglehart and Norris, 2016).
Inglehart and Norris (2016) started with the main theoretical point of view that rising economic insecurity and social disadvantage in society creates increasing support for populist parties. In particular, changes in society with regard to economic welfare and transformation of the workforce are seen to push people towards populist parties. However they found the most consistent evidence in support of the cultural perspective. In contrast, Guiso et al. (2017) concluded that populism does not exclusively has a cultural cause, but rather an economic insecurity cause with an important detectable cultural channel. This suggests that economic shocks activate populism both through voting and avoidance. Economic shocks shift beliefs and attitudes, which have traditionally been classified as cultural traits. It is crucial to distinguish economic shocks from cultural shocks. According to Guiso et al. (2017), when economic insecurity intensifies people demand short-term economic protection. At the same time, populist parties find their own space within the political landscape, with an agenda based on the creation of the contradictory ‘the people versus the elite’. This leads to the promise of short-term economic protection, as the long-term is criticized as an elite interest.
anti-9
elite rhetoric, whereas economic insecurity is the key driver of populist demand. This study will look specifically at the economic insecurity perspective, the demand side of populism and therefore the underlying grievances. To control for alternative explanations this study includes cultural variables.
Through what economic channels is the economic insecurity perspective influenced? The literature review discusses three main themes from various perspectives: the role of globalization, automation and business cycles. From here, the support for populism is discussed. Table 1 shows an overview.
Table 1: Overview of economic developments influencing economic insecurity perspective
Globalization Automation Business cycles
International trade and FDI: Inequality
Standards of living Offshoring and migrants:
Unemployment
Job polarization: Inequality unemployment Labor saving technology:
unemployment
Economic shocks:
Standards of living Unemployment Inequality
Note: FDI = foreign direct investment
1.2.1 Globalization
10
the long run, open economies fare better in aggregate than closed ones, and that relatively open policies contribute to long-term development (Winters, 2002). However, Colantone and Stanig (2017) argue that globalization might not be sustainable in the long run in the absence of appropriate redistribution policies aimed at compensating the people who are ‘left behind’, or the ‘have-nots’ of global markets: those segments of society that bear most of the adjustment costs of international trade. The cost of adjustment is often largely borne by the individual and depends critically on them finding another job. Similarly, capital in the import-competing sector which is highly specific (e.g. specific machinery) may be lost completely without provision for compensation, while some types of generic capital (e.g. buildings) may find uses in other sectors.
11
well as abroad, for relatively highly-skilled workers. This shapes international trade’s uneven impact. The increasing demand for highly-skilled workers at home can eventually lead to more unemployment in two different ways. If we assume that workers are not perfectly mobile, problems with mobility impede workers from moving across industries, firms and local labor markets. Regions hit by this uneven impact, with jobs outsourced to low-wage countries, could suffer high unemployment rates. The sentiment of the left behind from global markets is interpreted and promoted—especially by nationalist and right wing populist parties, whose policy proposals tend to bundle support for domestic free market policies with strong protectionist viewpoints (Colantone and Stanig, 2017).
12
businesses from Europe to elsewhere in the world, mainly Asia, there is higher demand for imported goods. The imported goods and services as a percentage of gross domestic product (GDP) is used as a proxy for international trade to measure the effect of international trade on support for populist parties. Based on the literature, the next hypothesis is:
Hypothesis 1: support for populist parties in Europe is high when imports of goods and services relative to GDP is high
1.2.2 Automation
13
warehouse workers. Changes in technology alter the types of jobs available and what those jobs pay. Autor (2014) comes with a solution and suggests that it is not middle-class workers that are doomed by automation and technology, but instead that human capital investment must be at the heart of any long-term strategy for producing skills that are complemented by rather than substituted by technological change.
The economic insecurity perspective gives us strong reason to believe that unemployment is likely to motivate populist reactions. However, some authors found that unemployment has a negative relationship with voting for populist parties. In a study by Swank (2003) focusing on right wing populist parties, did not find a significant relation between unemployment and support for right wing populist parties. Knigge (1998) argues that when unemployment becomes a real problem in the economy, people tend to rely on mainstream political parties and therefore there is no growth related to populist parties. Following the economic insecurity perspective, logic predicts that mass support for populism should be observed to be concentrated among economically-marginalized sectors who are the main left-behinds from global markets, technological advances, and knowledge societies (Inglehart and Norris, 2016). Thus populist votes should be strongest among unskilled workers, the unemployed, and those lacking college degrees. In regard to economic insecurity perspectives in times of crisis Algan et al. (2017) found a correlation between an increase in unemployment and a decline in trust in national and European political institutions. Inglehart and Norris (2016) came to the same conclusion when it comes to unemployment, as the support for non-mainstream parties increased during the Great Recession in Europe.
14
perspective and populist support could also be predicted by subjective feelings of economic insecurity. Based on the literature the following hypothesis is formed:
Hypothesis 2: support for populist parties in Europe is high when the unemployment share of the working-age population is high
1.2.3 Business cycles
Algan et al. (2017) argue that when a country has to deal with a crisis of systemic economic insecurity resulting in decreasing levels of GDP, the traditional established parties (whether left-leaning and relying on government-based policies, or right-left-leaning and relying on markets) find it hard to address this issue, with the consequence that their voters lose faith in them. Financial- and business- cycles are creating ‘booms’ and ‘busts’. A bust often results in a financial and economic crisis. Algan et al. (2017) argue that the crisis-driven economic insecurity is a substantial driver of populism and political distrust. History teaches us that in the ‘busts’, the middle class often takes the biggest hits and suffers most employment and financial wise. In the late 20th century, Dornbusch and Edwards (1991) stated that populist policymakers and the population at large were deeply dissatisfied with the economy’s performance and felt that the situation should and could be improved. This could be an initial condition for growing support for populism.
15
poorer economic conditions of the Euro-area countries— is correlated with the populist vote. The study by Swank (2003), using a Tobit’s maximum likelihood estimator, showed an insignificant association between slower economic growth and support for right wing populist parties. Their study focused on right wing populist voting shares in elections in 16 western European countries between 1981 and 1998. According to Dornbusch and Edwards (1991), very moderate growth, stagnation, or outright depression increases the support for populist parties. These parties are likely to gain support when mainstream parties and national institutions fail to manage shocks to their economies. Institutional constraints increase frustration among voters and result in a turn towards populist parties; according to Mudde (2016), this is especially true for Eurozone countries. Based on more recent literature, the next hypothesis is as follows:
Hypothesis 3: support for populist parties in Europe is high when GDP per capita is low
1.2.4 Income inequality
16
distribution, but not to those in the middle (Autor, 2014). There is overwhelming evidence of powerful trends toward greater income and wealth inequality in Western society based on the rise of the knowledge economy, technological automation, financialization of the economy, the collapse of manufacturing industry, and global flows of labor, goods and people—especially the inflow of migrants and refugees (Inglehart and Norris, 2016).
Rodrik (2017) theoretically argued that the main implication of the rising support for populism in the U.S. is the increasing economic inequality. From the literature review, the following hypothesis is formed:
Hypothesis 4: support for populist parties in Europe is high when national income inequality is high
17
III. M
ETHODOLOGY
This section explains the empirical strategy used in this research. The main element is the model specification, which sets out the variables used. The data sources are discussed in the next section.
This study analyzes the outcomes of 113 parliamentary elections in twenty-six European countries4 in 2000–2015 to analyze whether the hypotheses are supported. The time window was chosen due to the data availability regarding identification of political parties that participated in several elections researched by Kessel (2015) or Inglehart and Norris (2016). Table 2 (Appendix I) gives an overview of the political parties in the sample.
Swank (2003) argues that data from parliamentary elections are relatively consistent in structure and political importance, both across nations and time. The following model specification tests what effect the macroeconomic variables have on the share of votes for populist parties during parliamentary elections:
pop (i, t ) = αi + β1 gdpc (i, t -1) + β2 imp (i , t -1) + β3 unemp(i, t -1) + β4 gini (i, t -1)
+ β5 imi (i, t -1) + β6 edu (i, t -1) + β7 age (i, t -1) + ε (i, t) (1)
Equation (1) is estimated using panel data, in which the behavior of variables are observed across time. Estimate (1) is calculated with the fixed effects model, for which αi is the country-specific
18
is the import of goods and services as share relative to gross domestic product. unemp is the unemployment share of the working age population. gini is the Gini coefficient of equivalised disposable income and has a value between 0 and 1, where 1 expresses maximal income inequality and 0 maximum income equality. ε is the error term.
To test for these hypotheses, this study includes some cultural control variables: education, median age of the population and immigration stock. edu is the percentage share of the population between 15 and 64 years who have less than primary, primary and lower secondary education. imi is the share of total immigrants who were welcomed in a country that year relative to the population living in country i. Immigration as a stock or flow is used several times in political research (Lubbers et al. 2002; Algan et al. 2017 & Guiso et al. 2017) as a control variable. Consequence of globalization are increasing migration flows around the world, particularly in Europe open border policies result in higher migration flows of people (Rodrik, 2017). Through the migration flow in Europe, mainly from outside Europe, logically the share of migrants becomes larger.
19
Table 3: Expected signs of the coefficients in relation to the voting share on populist parties
Hypothese Variable Expected sign of coefficient
I gdpc -
II imp +
III unemp +
IV gini +
Note: hypotheses are tested on the share of votes on populist parties. gdpc; average real gross domestic product per capita, imp: import share relative to gross domestic product, unemp: unemployment share of the working age population, gini: gini coefficient.
First, a pooled OLS regression is used. In the pooled regression, all 113 observations are pooled together, neglecting the cross section and time series nature of the data. In Guiso et al. (2017) the Tobit maximum likelihood approach was used because the sample had number of elections with no votes for populist parties. This study has controlled for that and differences between OLS and Tobit maximum likelihood were minimal. When using the OLS estimator, the regression might be biased. The fixed effects model is a solution to this bias, at least when the bias is constant over time (Chatterjee and Hadi, 2015). When using fixed effects, one assumes that something within the individual intercept αi biases the predictor or outcome variables (Chatterjee and Hadi, 2015). This study performed a ‘Hausman’ test and found an insignificant chi-square.5 The latter implies that the null hypothesis was not rejected as a result that the random effect model is appropriate. Intuitively, the fixed effects model is more appropriate. For example, beliefs and attitudes around economic insecurity and populism are captured by country-level fixed effects.
20
IV. D
ATA
This section explains which data sources are used, followed by an overview of the descriptive statistics and finally the econometrical issues are discussed.
The first issue is the identification of political parties. Mudde (2007) has been influential in the literature, suggesting that populist philosophy is a loose set of ideas that share three core features: anti-establishment, authoritarianism and nativism. Kessel’s (2015) definition is more or less the same and uses the following definition of populism to identify populist parties (i) portray ‘the people’ as virtuous and essentially homogeneous; (ii) advocate popular sovereignty, as opposed to elitist rule; and (iii) define themselves against the political establishment, which is alleged to act against the interest of ‘the people’. As will be discussed in the next chapter, one of the difficulties of using the populist concept as a tool of classification is that populism is not always a stable core attribute of certain political parties. There are a substantial number of parties that can be seen as “borderline cases of populism” (Kessel, 2015 p.33). Populist rhetoric can in theory be voiced by all parties, and political actors may also modify their degree of populism over time. In the end, however, Kessel (2015) identified parties that can, at least for a certain period of their existence, be considered as genuine cases of populism following this definition.
21
political parties within each country. Thirteen selected indicators examined where experts rated the position of European parties on a range of populist items, such as support for traditional values, liberal lifestyles, and multiculturalism, as well as their economic stance towards market deregulation, state management of the economy and preferences for either tax cuts or public services. Parties are coded as populist in this dataset if they are labeled as such in the studies by Kessel (2015) or Inglehart and Norris (2016), that identified political parties that participated in several national elections. Calculations for the total share of populist votes in each election year t for each country i in the sample are done by the author.
Methodologically, it is important to note here that pop measures the total percentage of votes for all populist parties (left wing and right wing) in country i at year t. For each country i in election year t, three ideologies are possible: left wing populist, right wing populist, and not populist. Hence, if there are two right-wing and two left-wing populist parties in country i, all their support is added up. All countries are included in which Kessel (2015) and Inglehart and Norris (2016) identified one or more populist parties. There have been years in which there was no populist party participating in elections in these countries. However, these election years are included in the sample to prevent a selection bias.
The election results were taken from Nordsieck (2018), a reference guide to the parliamentary elections and governments in the European countries since 1945. More than 700 parties are listed. The guide includes basic data of these parties (founding years, political orientations and websites) and a chronological summary of their history (origins, name changes, mergers and splits).
22
Bank.7 Average real gross domestic product per capita is measured in constant 2011 international dollars (in purchasing power parity). The second explanatory variable is the share of imports of goods and services relative to domestic GDP. The data on this variable were collected from the World Bank. The third explanatory variable, unemployment, is the level of unemployment as a percentage of the total domestic workforce population. The data comes from the World Bank. The fourth and final explanatory variable is the Gini coefficient in the year before the election; the data were collected from Eurostat.8 The Gini coefficient data set has some missing observations, mainly from Croatia, Latvia, Lithuania and Iceland.9
The control variable education is the percentage share of the population with less than primary, primary and lower secondary education (levels 0-2 on the ISCED level of education)10 and is taken from Eurostat. The median age data is taken from Eurostat. Cultural insecurity could also be produced by the amount of immigrants entering a country due to immigration. Unfortunately, there are no data on immigration flows by country of origin and region of destination. To capture the fear of ‘harming the culture’ due to the arrival of immigrants, the stock of immigrants entered the country in that year (to measure an inflow) as a percentage relative to the population is used. Calculations are done by the author. The data of the population and immigration used for these calculations comes from Eurostat. Immigration data has some missing observations, mainly from Bulgaria, France and Romania.
4.1DESCRIPTIVE STATISTICS
23
2010 elections in Hungary. The FIDESZ-MPSZ party received about half of the votes, and there were also smaller populist parties that received a significant percentage of the vote. There are two more elections in which pop was higher than 50%: Bulgaria in 2009 and Italy in 2015. Some countries in which pop is low are Iceland, Luxembourg and the United Kingdom.
The number of elections per country differs and is taken from the years 2000-2015. France and Luxembourg had three elections and Greece six. This leads to an unbalanced panel. Most of the countries had four or five elections in the sample period. The data for the independent variables is taken from one year before the year of national elections and thus taken from 1999-2014.
The average real gross domestic product per capita (gdpc) is highest in Luxembourg ($90.700) and lowest in Bulgaria ($9400). gdpc mean in the sample period is $34.100. Imports of goods and services relative to gross domestic product (imp) is highest in Luxembourg (158%), the Slovak Republic (92%), Hungary (82%), Belgium (81%) and the Czech Republic (79%). The lowest imp is France (25%). Low values are not restricted to a particular area in the Europe, but have more to do with the size of the countries’ economy and their demographics. The unemployment share of the working age population (unemp) is largest in Greece (24.9%), followed by Spain (24.1%), and is lowest in Norway (2.3%). The unemp mean in the sample period is 8.9%, with the highest unemp in 2012–2014. Western European countries tend to perform better on this variable. The income inequality measured by gini is highest in Latvia (.39), Romania (.38) and Bulgaria (.37) and lowest
Table 4: Descriptive statistics.
24
in Norway (.22). This is not restricted to a particular area in Europe. Gini has a lot of missing observations (17). Total immigrants welcomed each year in a country as share relative to the population (imi) is highest in Luxembourg (4.1%) and lowest in Bulgaria (.0017%) and Romania (.06%). The percentage of the population which has less than primary, primary and lower secondary education decreases over time in all countries. Population who have less than primary, primary and lower secondary education was highest (.58) in Spain in 2000 and Italy (.56) in 2001. The average age in in the sample is 39.6 years and the range varies between 33.5 and 45.4 years. Median age is typically lower for Eastern European countries.
Table 5 provides the simple bivariate correlation between all variables. The independent variables do not show high correlation with the dependent variable pop. The highest coefficient (.76) is the positive correlation between gdpc and imi. The relative high correlation can be explained due to when gdpc and standards of living are increasing, the country is more attractive to immigrants. Nevertheless, when gdpc is high, assuming no concentration of the wealth, more jobs can be created, the degree of unemployment is decreasing and countries welcome more immigrants. However, it is too early to draw conclusions regarding this correlation, because it is the bivariate relationship between variables. Therefore, the regression results are discussed in the next section. Before introducing the results of the models, econometrical issues are discussed.
Table 5: Correlation matrix of coefficients of pooled OLS regression
25
4.2ECONOMETRICAL ISSUES
There were several econometrical issues. First, this study tested for the multicollinearity problem. Multicollinearity makes it hard to distinguish between the individual effects of the explanatory variable on the dependent variable. One way to measure multicollinearity is by the variance inflation factor (VIF), which assesses how much the variance of an estimated regression coefficient increases if the predictors are correlated. According to Chatterjee and Hadi (2015), a VIF of less than 10 indicates no signs of multicollinearity. The results of the bivariate correlation matrix in Table 5 already suggested this.
Second, one issue when constructing an econometric model is that the regression is influenced by omitted variable bias when important variables are missing. Using control variables for culture in this study was helpful, but far from sufficient to control for culture, resulting in omitted variables. For example, factors like trust, happiness, (social) media, and attitudes towards specific problems were not included, either because it would have led to too many variables in the regression or because the data was not available on a country level.
Third, a common source of endogeneity is reverse causality. This could occur when taking data of the dependent and independent variables from the same year, which can influence each other. For example, populist votes may influence economic performance and economic performance may influence populist votes in the same year. This study minimizes endogeneity by using a one-year lag for the independent variables.
26
27
V. R
ESULTS AND ANALYSES
In this section, the results of the general model specifications are discussed to show whether the hypotheses are supported. Followed by the results of the robustness checks.
5.1 Regression results
Table 6 shows the results of the pooled OLS regression. The main variables of interest are gdpc, imp, unemp and gini. In column (I) gdpc, imp and unemp are significant at the 0.01 and 0.05 level. The coefficient of gdpc was as expected negative, meaning that lower level of gdpc correlates with higher level of support for populist parties. Because of the high bivariate correlation, either imi or gdpc is excluded from the regression in columns I and II. imi and gdpc are both significant at the 0.01 level, whereas imp in both scenarios is significant at the 0.05 level.
Table 6: Pooled OLS regression results
(I) (II) (III) (IV) (V) (VI)
Variables pop pop pop pop pop pop
gdpc t-1 -0.449*** -0.204** -0.334** (0.160) (0.096) (0.157) imp t-1 0.189** 0.195** 0.141** 0.180** 0.176** 0.195** (0.079) (0.078) (0.070) (0.073) (0.080) (0.078) unemp t-1 -0.783** -0.542 -0.426 -0.542 (0.335) (0.327) (0.303) (0.327) gini t-1 -0.382 -0.162 -0.604 -0.162 (0.522) (0.535) (0.507) (0.535) edu t-1 0.149 0.153 -0.100 0.018 0.112 0.153 (0.224) (0.219) (0.163) (0.183) (0.225) (0.219) age t-1 0.031*** 0.022*** 0.020*** 0.017** 0.029*** 0.022*** (0.009) (0.008) (0.006) (0.007) (0.008) (0.008) imi t-1 -8.253*** (2.940) Constant -0.851** -0.686* -0.571** -0.509* -0.787** -0.686* (0.358) (0.353) (0.280) (0.299) (0.344) (0.353) Observation s 96 94 111 111 96 96 R-squared 0.234 0.203 0.156 0.168 0.202 0.203
28
When including gini in the regression in column (I) (II) (V) and (VI), observations for the regression drop by 20% because of the national elections that are now not included due to the missing observations of gini. However, the coefficients of gdpc and imp hold at the 0.05 significance level and the coefficient is larger, meaning that the effect is larger for the number of votes for populist parties. The control variable age is significant at the 0.01 or 0.05 level in all columns, whereas the control variable edu is insignificant in all columns.
When both gini and imi are included in the regression, the observations drop from 113 to 94; gini alone is responsible for 17 missing observations. Due to the small number of observations, the regression results are sensitive and therefore interpretations of the results have to be made carefully. The number of countries do not drop in the sample because every country has at least one national election with all data of the variables in equation (1). However, this makes the weight of countries in the sample who do not have missing variables of gini and imi higher in the regression when these variables are included.
In a pooled regression, all observations are pooled together, neglecting the cross section and time series nature of the data. Therefore, when using the OLS estimator, the regression might be biased. The country-level fixed effects model is a solution to this bias, as beliefs and attitudes around economic insecurity and populism are captured by country-level fixed effects.
29
primary and lower secondary education and vice versa. This seems counterintuitive. However, a higher median age means more elderly people. In the past, the education system consisted of
Table 6: Correlation matrix of coefficients of fixed effect model
Variables (1) (2) (3) (4) (5) (6) (1) gdpc 1.000 (2) imp -0.230 1.000 (3) unemp 0.604 -0.082 1.000 (4) gini -0.034 -0.008 -0.106 1.000 (5) imi -0.215 -0.175 0.111 -0.086 1.000 (6) edu 0.150 0.063 0.181 -0.274 -0.010 1.000 (7) age -0.381 -0.147 -0.356 -0.287 0.064 0.588 Note: list of abbreviations in Appendix II.
30
columns. Moreover, the sign is contrary to what this study expected. The negative sign suggests that an increase in the unemployment rate leads to a decrease in the votes for populist parties.
Table 7: Country-level fixed effects regression results
(I) (II) (III) (IV) (V)
Variables pop pop pop pop pop
gdpc t-1 0.381 0.627 0.776 (0.620) (0.468) (0.479) imp t-1 0.120 0.123 0.152 0.163 0.235 (0.153) (0.149) (0.144) (0.153) (0.147) unemp t-1 -0.226 -0.364 -0.229 (0.387) (0.293) (0.303) gini t-1 1.343** 1.362** 1.338** 1.552** 1.622*** (0.595) (0.575) (0.579) (0.593) (0.601) edu t-1 -0.147 -0.562** -0.755*** (0.363) (0.262) (0.269) age t-1 0.025* 0.025*** 0.033*** (0.013) (0.009) (0.009) imi t-1 -1.120 -1.304 -0.445 -1.793 -0.269 (3.005) (2.905) (2.895) (3.001) (3.038) Constant -1.345** -1.488*** -1.564*** -0.463* -0.170 (0.540) (0.293) (0.301) (0.246) (0.216) Observations 94 94 94 94 94 R-squared 0.436 0.446 0.444 0.400 0.379 Number of countries 26 26 26 26 26
Note: standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Columns I, II, III, IV and V show results of the country-level fixed effects regression. List of abbreviations in Appendix II.
This outcome is unexpected, however, the results of unemp are not significant, and therefore conclusions cannot be drawn from those results. Table 7 columns I, II and IV show the coefficient of gdpc, which has the unexpected positive sign. When country-level fixed effects are included, gdpc is no longer significant. Therefore the results do not support the third hypothesis.
31
for pop was relatively low, influence the results and make the results sensitive for interpretation. The weight of the other countries without missing observations of gini is therefore higher in the sample.
Control variables age (columns II and III) and edu (columns IV and V) were separately tested and show a significance level between 0.05 and 0.10. When controlling for both (column I), age shows a strong significant relationship with pop. edu becomes insignificant due to the strong bivariate relationship between the two control variables with age as the strongest predictor.
5.2 Robustness checks
32
Table 8: Country-level fixed effects regression results with two year lag independent variables
(I) (II) (III) (IV) (V)
Variables pop pop pop pop pop
gdpc t-2 -0.252 -0.253 -0.237 (0.674) (0.512) (0.517) imp t-2 0.050 0.079 0.049 0.071 0.052 (0.230) (0.221) (0.212) (0.224) (0.212) unemp t-2 0.144 0.235 0.265 (0.402) (0.319) (0.319) gini t-2 1.572** 1.249** 1.382** 1.722*** 1.734*** (0.690) (0.617) (0.642) (0.626) (0.616) edu t-2 -0.320 -0.630** -0.487* (0.421) (0.294) (0.273) age t-2 0.011 0.020** 0.015* (0.014) (0.009) (0.008) imi t-2 2.539 0.254 2.054 -0.929 2.147 (3.200) (2.889) (3.014) (3.137) (3.367) Constant -0.543 -0.939*** -0.873*** -0.074 -0.217 (0.547) (0.290) (0.281) (0.248) (0.229) Observations 94 94 94 94 94 R-squared 0.262 0.252 0.253 0.253 0.253 Number of countries 26 26 26 26 26
33
VI. C
ONCLUSION
Over the past two decades, populist sentiments have increased, a trend that has manifested itself across continents and electoral systems (Mudde, 2016). This trend can be observed globally, but particularly in Europe. Politicians of populist movements have taken part in the governments of Austria (the Freedom Party of Austria), Hungary (Fidesz), Sweden (the Sweden Democrats), Greece (Syriza) and Italy (Five Star Movement).
Signs of dissatisfaction in society are visible in the form of the non-political yellow vest movement that has attracted global attention. This movement’s demands to the government share some broad values with the populist political parties and the populist definition of Mudde (2007). Moreover, as Judis (2017) argued, there is no set of features that exclusively defines which movements, parties, and people are populist. It is not an ideology, but a way of thinking about politics and society.
This study analyzed the trend of rising populism using the economic insecurity perspective of Inglehart and Norris (2016). What distinguishes this research from others (e.g., Swank, 2003; Funke et al., 2016; Inglehart and Norris, 2016; Guiso et al., 2017) is that it analyzed the consequence of developments from a macroeconomic perspective. The effects of globalization, automation and business cycles on economic insecurity are discussed in relationship with the votes for populist parties in national elections.
34
populist political parties in twenty-six European countries that participated in elections in 2000– 2015. The results for the parliamentary elections were taken from Nordsieck (2018). A country-level fixed effect regression was used to answer the main research question.
This study has several main results. One hypothesis is supported and another is partly supported. The findings show that increasing economic inequality has a positive relationship with the rising support for populism. This is the same conclusion about Europe that Rodrik (2017) found for the US. Piketty’s (2014) research has brought renewed attention to rising levels of wealth and income inequality. There is overwhelming evidence of powerful trends toward greater income and wealth inequality in Western society based on the rise of the knowledge economy, technological automation, and the collapse of the middle class (Inglehart and Norris, 2016). This study found a positive relationship between income inequality and increasing populist sentiment. These findings are robust when taking a two year lag for the independent variables.
35
to low-wage countries causes more competition for jobs and companies in Europe, because efficiency-seeking companies opt for lower production costs in foreign countries. When country-level fixed effects are used, the findings do not support the hypothesis.
There are some limitations to this study. First, there are some measurement errors in this study. Change over time in the party identification is not accounted for in the sample and some important cultural variables in the model are missing. Another measurement error is the definition of populist parties and how the party identification is influenced by some degree of subjectivity, therefore the results are debatable. Second limitation is the number of observations. This is low due to the missing observations of immigration and income inequality. The results of a regressions with small number of observations are sensitive and therefore the conclusions must be taken carefully. The time period for this study and number of countries in the sample are a cause for the small number of observations. Another limitation is concerning the cross-national level of this study. Macroeconomic developments examine indicators on national and international level, therefore the findings from the study are very general.
right-36
37
R
EFERENCES
Acemoglu, D., Egorov, G., Sonin, K. (2013). A political theory of populism. The Quarterly Journal of Economics, 128(2), 771-805.
Algan, Y., Guriev, S., Papaioannou, E., Passari, E. (2017). The european trust crisis and the rise of populism. Brookings Papers on Economic Activity, 2017(2), 309-400.
Ark, B., O'Mahoney, M., Timmer, M. P. (2008). The productivity gap between Europe and the United States: trends and causes. Journal of economic perspectives, 22(1), 25-44. Brynjolfsson, E., McAfee, A. (2014). The second machine age : Work, progress, and prosperity
in a time of brilliant technologies (First ed.). New York, NY: W.W. Norton & Company. Chatterjee, S., Hadi, A. Regression analysis by example (Fifth edition, Wiley series in
probability and statistics). Hoboken: Wiley; 2015.
Colantone, I., Stanig, P. (2018). The trade origins of economic nationalism: Import competition and voting behavior in western europe. American Journal of Political Science, 62(4), 936-953.
Deaton, A. (2012). The financial crisis and the well-being of Americans. Oxford Economic Papers, 64(1), 1-26.
Dornbusch, R., Edwards, S. The macroeconomics of populism in latin america (A national bureau of economic research conference report). Chicago: University of Chicago Press; 1991.
38
Frey, C. B., Osborne, M. A. (2017). The future of employment: how susceptible are jobs to computerisation?. Technological forecasting and social change, 114, 254-280. Funke, M., M. Schularick, and C. Trebesch (2016), “Going to the extremes: Politics after
financial crises, 1870-2014”, European Economic Review 88, 227– 260.
Grusec, J., & Hastings, P. Handbook of socialization : Theory and research (second ed.) New York: Guilford Press; 2015.
Guiso, L., Herrera, H., Morelli, M., Sonno, T. (2017). Populism: Demand and Supply. Center for Economic Policy Research Discussion Paper, 11871. 1-63.
Inglehart, R., & Norris, P. (2016). Trump, Brexit, and the rise of populism: Economic have-nots and cultural backlash.HKS Working Paper No. RWP16-026.
Judis, J. The populist explosion : How the great recession transformed american and european politics(Columbia global reports). New York: Columbia Global Reports; 2016
Kessel, van S. Populist parties in europe agents of discontent? Houndmills, Basingstoke, Hampshire: Palgrave Macmillan; 2015.
Knigge, P (1998). The ecological correlates of right-wing extremism in Western Europe, European Journal of Political Research 34, 249-279.
Lewis-Beck, M. S., Nadeau, R. (2011). Economic voting theory: Testing new dimensions. Electoral Studies, 30(2), 288-294.
Los, B., McCann, P., Springford, J., Thissen, M. (2017). The mismatch between local voting and the local economic consequences of Brexit. Regional Studies, 51(5), 786-799.
39
March, L., Keith, D. (2016). Europe's radical left from marginality to the mainstream? (second ed.) London: Rowman & Littlefield International; 2016.
McClure, P. K. (2018). “You’re Fired,” Says the Robot: The Rise of Automation in the Workplace, Technophobes, and Fears of Unemployment. Social Science Computer Review, 36(2), 139-156.
Mudde, C (2004), “The Populist Zeitgeist”, Government and Opposition 39(4), 541– 63.
Mudde, C., Kaltwasser, C. R. Populism in europe and the americas threat or corrective for democracy? Cambridge: Cambridge University Press; 2012.
Mudde, C. (2013). Three decades of populist radical right parties in Western Europe: So what? European Journal of Political Research, 52(1), 1-19.
Mudde, C. (Ed.). (2016). The populist radical right : A reader (Routledge studies in extremism and democracy). London: Routledge, Taylor & Francis Group.
Nordsieck, W. Parties and elections in Europe: parliamentary elections and governments since 1945, european parliament elections history and political orientation of parties.
Düsseldorf (Germany): BOD publishers; 2018.
Piketty, T. Capital in the twenty-first century. Cambridge Massachusetts: Belknap Press of Harvard University Press; 2014.
Rodrik, D. (2018). Populism and the economics of globalization. Journal of International Business Policy, 1(1-2), 12-33.
40
Swank, D. (2003). Globalization, the welfare state and right-wing populism in western europe. Socio-Economic Review, 1(2), 215-245.
41
A
PPENDIX
I
Table 2: Identification of populist parties in the sample
Country Party
AT FPO
AT Alliance for the Future of Austria
AT Team Stronach BE Vlaams Blok BE FRONT NATIONAL BE List Dedecker BG NDSV BG Coalition Ataka
BG Law, Order and Justice (Red, Zakonnost, Spravedlivost)
BG Citizens for European Development of Bulgaria (GERB)
BG VMRO-BND Bulgarian National Movement
BG NFSB National Front for the Salvation of Bulgaria
BG HSS Croatian Peasants Party
CH Swiss People’s Party
CH Swiss Democrats
CH Lega dei Ticinesi
CH Geneva Citizen’s Movement
CZ ANO
CZ Public Affairs (Veci Verejne)
CZ Usvit
DE Die Linke (The Left)
DE NPD National Democratic Party
DE AfD Alternative for Germany
DK Dansk Folkeparti
ES Podemos
FI True Finns
FR FN (Front National)
FR MPF Popular Republican Movement
GB British National Party
GB UK Independence Party
GB NF National Front
GR SYRIZA
GR ANEL
GR XA Golden Dawn
GR LAOS Popular Orthodox Rally
GR ND New Democracy
HR HSP-AS
HR HSS Croatian Peasants Party
42
HR HSP Croatian Party of Rights
HR HDZ Croatian Democratic Union
HU FYD-HDF Fed.of Young Democrats&Hungarian Dem.Forum
HU Justice and Life Party (MIEP)
HU Movement for a Better Hungary
HU FIDESZ-MPSZ
IS Citizen’s Movement (BF)
IT Forza Italia
IT Lega Nord
IT Movimento Cinque Stelle
IT Il Popolo della Liberta (PdL)
IT Fdl Brothers of Italy
LT Labour Party (DP)
LT Party ”Order and Justice” (TT)
LT DK The Way of Courage
LU Alternative Democratic Reform Party
LV For Fatherland and Freedom/ LNNK
LV All for Latvia
LV NA National Alliance
NL List Pim Fortuyn
NL Liveable Netherlands
NL Geert Wilders’ Freedom Party (PVV)
NL SGP Political Reformed Party
NO Progress Party (FrP)
NO Democrats
PL Samoobrona Rzeczypospolitej Polskiej
PL Prawo i Sprawiedliwosc
PL SP United Poland
PL KNP Congress of the New Right
RO Greater Romania Party
RO People’s Party
SE Sweden Democrats
SI Slovene National Party (SNS)
SI SDS Slovenian Democratic Party
SI SDS Slovenian Democratic Party
SK HZDS Movement for a Democratic Slovakia
SK SMER
SK KDH Christian Democratic Movement
SK Slovak National Party (SNS)
SK Ordinary People and Independent Personalities (OLaNO)
43
A
PPENDIX
II
Table 4: Abbreviations and descriptions of the variables
Abbreviation Description
pop Share of votes for populist political parties
gdpc Average real gross domestic product per capita
imp Import share relative to gross domestic product
unemp Unemployment share of the working age population
gini Gini coefficient, income inequality measurement
imi Immigration stock relative to population
edu Percentage of population between 15 and 64 years who have less than primary, primary and lower secondary education
44 Notes
1 * French central bank slashes GDP forecast on back of yellow vests protests. Financial times. Paris: D. Keohane; 10-12-18 [11-12-18] Available on: https://www.ft.com/content/d2b8956a-fc5b-11e8-aebf-99e208d3e521
* France's 'Yellow Vest' Protests Keep Pressure on Macron as They March Into 11th Weekend. Time. Paris: A. Charlton; 26-01-2019 [27-01-2019] Available on:
http://time.com/5513858/yellow-vest-protestors-france-macron/
2 Why are france yellow jackets so angry? Politico Paris J. Lichfield. 141218 [141218] -https://www.politico.eu/article/why-are-france-yellow-jackets-so-angry
3 1980s-1990s Washington consensus: ‘the Market’. A reaction to failing import-substitution industrialization strategies (ISI). It was implemented voluntarily or by ‘conditionalities’ attached to International Monetary Fund (IMF) and Worldbank (WB) support; and in the World Trade Organization (WTO) rules. Key points of the consensus: the liberalization of trade; import and exports; foreign direct investment (FDI); international capital flows; privatization of state-owned enterprises.
4 Countries in the sample: Austria, Belgium, Bulgaria, Croatia, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Italy, Latvia, Lithuania, Luxembourg, Netherlands, Norway, Poland, Romania, Slovak Republic, Slovenia, Spain, Sweden,
Switzerland, United Kingdom. 5 Prob > chi2 = 0,978
6 Bakker, Edwards, Hooghe, Jolly, Marks, Polk, Rovny, Steenbergen and Vachudova. 2015. "2014 Chapel Hill Expert Survey." Version 2015.1. Available on https://www.chesdata.eu. Chapel Hill, NC: University of North Carolina, Chapel Hill.
7 The Development Data Group of the World Bank Group, coordinates data work and maintains a number of macro, financial and sector databases. Much of the data comes from the statistical systems of member countries. Data is available on their website: https://data.worldbank.org/ 8 Eurostat is the statistical office of the European Union situated in Luxembourg. Providing the European Union governments, businesses, the education sector, journalists and the public with statistics at European level that enable comparisons between countries and regions. Data is available on their website: https://ec.europa.eu/eurostat/data/database
9 Database of the World Bank andThe United Nations University World Institute for Development Economics Research give no solution for this problem
10The International Standard Classification of Education (ISCED) is a statistical framework for organizing information on education maintained by the United Nations Educational, Scientific and Cultural Organization (UNESCO).
