Volume 32, Issue 2: Paper 2

European Left-Wing Populism and Unemployment

Lauren Garrison, Furman University

 Though populism is not a new phenomenon, the United States and Europe have seen a dramatic rise in populism over the past decade. This populism has been right-wing as in the case of Donald Trump in the US, the United Kingdom Independence Party (UKIP) in the UK, and Marine Le Pen in France, and less commonly, though not infrequently, left-wing as in the case of SYRIZA in Greece and Podemos in Spain. The graph below displays the striking rise in the vote share of non-mainstream parties in general since the 1990s across 31 European countries, with the increase in far-right populist and far-left populist parties primarily driving this upward trend. Although this rise of populism has attracted the attention of researchers from many fields, discussion of populism is confounded by the lack of a single definition of populism. Regardless of the specific definition employed, populists generally pursue short-sighted policies without regard to long-term consequences and many, though not all, demonstrate authoritarian tendencies and weaken democratic institutions.[1]

 

This rapid rise of populism and the threat which it presents to modern liberal democracies caught some observers off guard. In a 2018 theoretical paper, however, Dani Rodrik asserts that populism has not come out of nowhere and proposes a theory that right-wing populism is primarily driven by cultural cleavages, such as immigration, while left-wing populism is primarily driven by economic inequalities, such as those caused by trade and global financial systems.[2] These cultural cleavages are especially difficult to measure. In the case of immigration, for example, an analysis of actual levels of immigration may yield different results than an analysis that employs perceptions of level of immigration, since politicians’ rhetoric can cause perceptions to not match reality. Because of the difficulties in measuring cultural cleavages, I focus on the economic cleavages prong of Rodrik’s theory. In this paper, I partially test Rodrik’s theory by examining the relationship between unemployment, an economic factor, and the vote share for left-wing populist parties. Specifically, I investigate whether there is a causal relationship between increases in unemployment and rising support in the form of a higher vote share for left-wing populist parties.

In short, my analysis displays that there is a positive relationship between unemployment and left-wing populist vote share, which lends empirical support to Rodrik’s theory about the differing causes of right-wing and left-wing populism. Furthermore, my analysis suggests that there is little to no impact of unemployment on support for right-wing populists. In Part I of this paper, I will provide a more comprehensive review of the existing literature regarding the causes of voter support for populists. In Part II, I will describe in further detail the models and methods that I used to estimate these effects and in Part III, I will describe the data that I used. In Part IV, I will present and explain my results. Finally, I will discuss the limitations of this study, areas for future research, and policy implications in Part V.

II. Theory and Existing Literature

As explained above, in this paper, I empirically test a theory by Dani Rodrik, a respected macroeconomist whose research focuses on globalization and political economy, about the differing causes of left-wing vs right-wing populism. Rodrik goes so far as to label left-wing populism “economic populism” and right-wing populism “cultural populism.” Because a populist must claim to represent the ordinary people against a corrupt elite, the type of populism a populist pursues prescribes who the enemy must be. In the case of cultural populism, the enemy is often foreigners or minorities. In the case of left-wing populism, the enemy is the wealthy and large corporations.[3] Specifically, Rodrik argues that the cultural shocks which fuel right-wing populism are made salient primarily through issues of immigration and asylum seekers. In terms of left-wing populism, the focus of this paper, Rodrik identifies trade, finance, and foreign investment as the primary avenues through which economic factors drive left-wing populism.[4]

Other researchers have pointed out that a clear delineation between economic and cultural factors is likely not realistic. In a comprehensive literature review of the economic work regarding populism, Guriev and Papaioannou stress that cultural factors are by definition constant over time so it is unclear how cultural shocks alone would lead to a sudden rise in populism. Thus, they overview two theories that consider the joint effects of economic and cultural factors. “Culture plus economics” contends that both economic and cultural factors considered together explain a rise in populism whereas “culture times economics” treats cultural and economic factors as interactive, so that economic shocks trigger dissatisfaction with the status quo.[5] Although I will not explore these additive and interactive effects of culture on economic factors in this paper, it is important to remember that economic shocks do not occur in a vacuum.

Populism research generally considers the issue from either the demand side, what motivates voters to want populist candidates and parties, or the supply side, what causes populist parties and candidates to exist.[6] In this paper, I focus on the demand side of populism. Although Rodrik does not specifically discuss unemployment, I have chosen it as my variable of interest because unemployment is often a manifestation of the shocks which Rodrik identifies which is especially personal and salient to voters. Events like the China trade shock, for instance, have been shown to increase unemployment due to global forces including trade, finance, and foreign investment.[7] Rodrik argues that globalization, and events caused by it like the 2008 financial crisis, drove wedges in society that were economic in nature, such as those between workers of different skill levels, rural and urban areas, and elites and ordinary people.[8]

Several studies have focused on the rising support of populism in the context of Europe specifically. Guiso et al found a significant effect of Eurozone membership on support for populists. When a country is a member of the Eurozone, some fiscal and monetary policy tools are removed from the domestic level, meaning that national policymakers may have a harder time responding to financial crises.[9] Dijkstra et al examined the impact of several demographic and economic factors on the vote share for parties opposed to European integration, which is strongly correlated with populism and anti-elite rhetoric, and found that economic and industrial decline coupled with rising unemployment were primarily responsible for driving a higher vote share for Eurosceptic parties.[10]

Autor, Dorn, and Hanson analyzed the impact of trade shocks, specifically the increase in Chinese imports in the US from 1990-2007 as a result of China joining the World Trade Organization (WTO), and found that trade shocks can affect unemployment. Specifically, higher import volumes led to higher unemployment rates.[11] Thus, unemployment is not independent of these economic shocks and cleavages which Rodrik describes. In a follow-up study, Autor, Dorn, and Hanson found electoral consequences of trade shocks. In particular, the authors found evidence of shifting political ideology related to increased trade-exposure in local labor markets, generally finding an increase in polarization in either direction on the political spectrum.[12]

This study is most similar to a study by Algan et al which examined the impact of unemployment on voting for nonmainstream, in particular populist, parties in Europe during the Great Recession. Although the authors find a statistically significant positive relationship between populist vote share and unemployment, this study does not differentiate between left populists and right populists.[13] Thus, my study adds to the existing literature by examining the differential impact of unemployment on left-wing versus right-wing populists.

III. Econometric Models

I estimated the effect of unemployment on far-left populist vote share for 31 European countries from 2000-2019 using Ordinary Least Squares (OLS) regressions. I ran multiple linear regressions of the effect of unemployment on far-left populist vote share which controlled for age, education, European Union (EU) membership, Eurozone membership, employment in blue collar jobs, inflation, voter turnout, and real GDP per capita growth, as shown by the model below. Most regressions include region fixed effects and some version of time fixed effects.

(1)
Leftpopulist_voteshare = β0 + β1 * Unemployment + β2 * Age + β3 * Education + β4 * EU + β4 * Euro + β5 * Blue Collar + β6 * Inflation + β7 * Turnout + β8 * rGDP_percap_growth + u

Left populist vote share, the outcome variable, is the summed total of the vote share of every far-left populist party in each election. The outcome variable is given as the vote share for all left-wing populist parties summed within each election because the type of party matters much more in Rodrik’s theory than do the individual parties themselves. Most elections have no left populist parties, and of those that do have left populist parties, there is usually only one left populist party in each election. However, several elections have two populist parties, and two elections (Spain in 2016 and November 2019) had three left populist parties. Unemployment is the variable of interest, and it represents the percentage of the labor force that does not have a job and is actively seeking one.

In order to establish a causal relationship between unemployment and left-wing populist vote share, variables that might be correlated with unemployment and also might directly affect left-wing populist vote share must be added to the model. If these variables were omitted from the model, the effect of unemployment on left-wing populist vote share would be biased (based on a comparison with simple linear regressions) because the coefficient on unemployment would be capturing the effect of these omitted variables. Thus, the other variables in the model are control variables intended to reduce the bias that could occur from omitting these variables. First, two demographic factors are included as controls: age and education. I split age into three categories, ages 0-24, ages 25-64, and ages 65 and older, because voting behavior is often influenced by age. In the case of populists, previous studies suggest that older voters may be more likely to vote for populists.[14] Similarly, I included education because education level is known to have an impact on voting behavior. I also split education into a three category variable which represents the percentage of the country’s population whose highest level of education falls into each category: less than high school, only high school, and more than high school. Theory predicts that the lower the level of educational attainment, the more likely someone is to vote for a populist.[15]

EU and Eurozone membership are binary variables that are equal to 1 if a country is a member in that year and are included in the model because populist parties tend to be Eurosceptic, and EU and Eurozone regulations often restrict domestic power in handling economic and financial matters, which populists may seize as an opportunity to paint Brussels bureaucrats as the corrupt elite against whom they fight.[16] Turnout is the percentage of eligible voters who participate in each election. Theory predicts that voters who are economically disadvantaged and should vote for left-wing populists according to Rodrik’s theory are less likely to turn out to vote.[17] Thus, when turnout is higher, more economically disadvantaged voters may show up to vote, resulting in a higher vote share for left-wing populist parties.

Three other economic factors are included as controls. First, blue collar represents the percentage of the workers of each country that works in a sector that is traditionally considered blue collar: construction, agriculture, and industry/manufacturing. Unemployment has been especially pronounced in former industrial areas and theory suggests that voters that feel left behind from industrial decline are more likely to vote for populist parties.[18] Inflation is the rate at which the price level is rising compared to the last year. Outside the simultaneous relationship with unemployment, I expect high inflation to increase support for left-wing populists, as inflation could be considered an economic factor as per Rodrik’s theory. Finally, real GDP per capita growth is the percentage by which real GDP grew or declined in each country compared to the previous year. This variable is included as proxy for income, which is often included as control variable in studying individual voting behavior. In the case of Eurosceptic parties, which are strongly correlated with populist parties, voters with lower income are more likely to vote for anti-EU parties.[19]

For fixed effects, each country is classified into one of four regions: northern Europe (Denmark, Finland, Iceland, Ireland, Norway, Sweden, and the UK), southern Europe (Cyprus, Greece, Italy, Malta, Portugal, and Spain), central Europe (Austria, Belgium, France, Germany, Luxembourg, the Netherlands, and Switzerland), and eastern Europe (Bulgaria, Croatia, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, and Slovenia).[20] For time fixed effects, some specifications include year fixed effects, whereas some specifications include controls for time based on the timing of financial crises and periods of high unemployment. Because most of the regressions only include about 130 observations due to the limited frequency of elections during my study period, including every year as a control variable presents a degree of freedom issue resulting in a lower adjusted R². To remedy this, I have created four time category variables based on the variation in unemployment related to financial crises. Pre-crisis includes 2000-2007, Great Recession includes 2008-2009, Sovereign Debt Crisis includes 2010-2013, and post-crisis includes 2014-2019.

The level of analysis for each control variable, with the exception of voter turnout, is country and year, while the level of analysis for the outcome variable and voter turnout is each individual election, which generally equates to country and year except for a few instances in which countries had more than one election per year. It is important to remember that we do not observe every country in every year since elections are not held annually. The model below includes subscripts with c representing country, t representing year, and i representing election.

(2)

Leftpopulist_voteshare_cti = β0_cti + β1 * Unemployment_ct +β2 * Age_ct + β3 * Education_ct + β4 * EU_ct + β4 * Euro_ct + β5 * Blue Collar_ct + β6 * Inflation_ct + β7 * Turnout_cti + β8 * rGDP_percap_growth_ct + u_cti

 

IV. Data

Given that the definition of populism varies so much, choosing an outcome variable to measure populism is not straightforward. For this paper, I use data from the PopuList, a project initiated by The Guardian, a credible center-left British newspaper, which classifies political parties for 31 European countries as populist, far-right, far-left, and Eurosceptic. Using this data, parties can be further classified as both populist and far-left or both populist and far-right, which is how I identify parties for the outcome variable in this paper. The data has been reviewed by over 80 academics and spans the time period of 1989-2019. The PopuList employs Cas Mudde’s definition of populism, which he describes as “an ideology that considers society to be ultimately separated into two homogeneous and antagonistic groups, ‘the pure people’ versus ‘the corrupt elite’, and which argues that politics should be an expression of the volonté générale (general will) of the people.”[21] The variables on populist, far-right, far-left, and Eurosceptic are binary and therefore do not account for the severity of the ideology or whether a party becomes more populist or ideologically extreme over time.

The PopuList also uses Cas Mudde’s definition for far-right parties: “parties that are nativist (which is an ideology that holds that states should be inhabited exclusively by members of the native group and that nonnative elements are fundamentally threatening to the homogenous nation-state) and authoritarian (which is the belief in a strictly ordered society, in which infringements of authority are to be punished severely).”[22] The PopuList uses Luke March’s definition of far-left parties: “parties that reject the underlying socio-economic structure of contemporary capitalism and advocate alternative economic and power structures. They see economic inequality as the basis of existing political and social arrangements and call for a major redistribution of resources from existing political elites.”[23] These definitions are consistent with Rodrik’s theory that right-wing populism is generated by cultural factors while left-wing populism is a response to economic factors.

I merged the PopuList with data from ParlGov, a political science data infrastructure which contains information about political parties and elections for most Organization for Economic Cooperation and Development (OECD) countries, so that information about whether a party is populist is linked with information about that party’s performance in each election. I collected the data for most of the independent variables (unemployment, age, education, blue collar employment, inflation, and real GDP per capita growth) from Eurostat, the agency of the European Commission which collects and compiles national statistics from EU Member States and several other European countries. I coded the variables for EU and Eurozone membership using information from the European Union country profiles.[24] Voter turnout data comes from the Voter Turnout Database provided by the International Institute for Democracy and Electoral Assistance (IDEA). Although the PopuList can be applied to elections beginning in 1989, I chose to narrow the time period to 2000-2019 for this study to limit the frequency of missing values in Eurostat data in the 1990s, particularly for Eastern European countries.

 

Table 1: Summary Statistics

	Observations	Mean	Std. Dev.	Minimum	Maximum
Left Populist Vote Share	168	2.88%	6.84%	0%	36.34%
Unemployment	165	8.72%	4.93%	2.3%	24.9%
Percent Aged 0-24	167	28.79%	3.22%	23.1%	38.1%
Percent Aged 25-64	167	54.49%	1.82%	50.5%	58.5%
Percent Less than High School	166	29.51%	12.88%	12.1%	78.1%
Percent High School Only	166	46.46%	12.71%	13.4%	72.2%
EU Member	162	0.80	0.40	0	1
Uses Euro	162	0.51	0.50	0	1
Percent Blue Collar	167	31.08%	10.17%	16.5%	71.8%
Inflation	159	2.14%	2.33%	-1.65%	16.27%
Voter Turnout	168	68.59%	13.31%	37.79%	95.7%
rGDP_Per Capita Growth	168	2.24%	3.14%	-7.6%	13%

 

The mean for left populist vote share is relatively low at 2.88% which suggests that left populist parties are not frequently experiencing giant electoral victories and entering into government. Depending on the specific structure and context of each country’s system, a vote share of 2.88% could be enough to win a seat and have some influence over government. After winning a single seat in 2015, for instance, UKIP was able to play a role in staging the Brexit referendum.[25] The maximum of 36.34% is relatively large in a parliamentary system with more than two parties, meaning that left-wing populists could have significant influence in the government. This maximum occurred in Greece in 2015 and, although mainstream parties had a combined plurality of 41%, left-wing populists, notably SYRIZA, exercised considerable influence in the government at the time. It is important to consider that 125 of the 168 elections (74%) in this study period had no left populist parties at all, compared to only 37% which had no right-wing populists. The average summed left-wing populist vote share from 2000-2019 is shown for each country in the map below.

The mean unemployment rate of 8.72% is consistent with the fact that European unemployment tends to be naturally higher than US unemployment. There is much variation in unemployment across time and across countries, which has an effect in examining the relationship between unemployment and left-wing populist vote share. Interestingly, the maximum unemployment rate of 24.9% occurred in Greece in 2015, the same year and country in which the maximum left populist vote share occurred. The average unemployment rate from 2000-2019 is shown for each country in the map below.

 

V. Results

As Table 2 displays, the effect of unemployment on the vote share for far-left populist parties is significant at the one percent level in all but one specification. Column 1 represents a model with year fixed effects, which has a lower adjusted R² than does column 2, which uses the four category time variables instead of years to address the degrees of freedom problem associated with using year fixed effects. Columns 3-5 introduce lagged variables for the variable of interest, unemployment, and for several of the non-demographic factor control variables, inflation and real GDP per capita growth. These variables are lagged because I suspected that voters might be reacting to economic conditions from the previous year, especially in elections that occur earlier in the calendar year. I also introduce clustered standard errors by country in regressions 3-4.

Regressions 4-5 include a lagged left populist vote share variable, which is the total vote share of all far-left populist parties in each country’s previous election. I include this variable to account for the fact that voters may be reacting to a party that was successful in the last election if left-wing populist parties are gaining support and influence in the government over time. The adjusted R² increases dramatically when lagged left-wing populist vote share is added to the model. Regression 5 differs from regression 4 in that it is a country fixed effects model employed to control for unobservable factors, like culture and institutional structure, that are vary across countries but are relatively constant within each country over time. Regression 5 is my preferred specification because the country fixed effects are more specific than region fixed effects and the r-squared and adjusted r-squared remain high despite the increased number of variables in the regression. Overall, the positive relationship between unemployment and far-left populist vote share is robust to several specifications and is consistent with Rodrik’s theory that economic factors increase support for left-wing populist parties.

Table 2: Results for Left Populists

	Far-Left Populist Vote Share
	1	2	3	4	5
Unemployment	0.624*** (0.183)	0.599*** (0.181)
Lagged Unemployment			0.631* (0.321)	0.305*** (0.111)	0.692*** (0.241)
Lagged Left Populist Vote				0.804*** (0.077)	0.431** (0.188)
% Aged 0-24	-0.101 (0.230)	-0.113 (0.224)	-0.157 (0.407)	-0.010 (0.157)	-1.7628 (1.067)
% Aged 25-64	0.269 (0.392)	0.220 (0.367)	0.232 (0.505)	0.023 (0.198)	-0.642 (0.921)
% Less than High School	-0.017 (0.110)	-0.136 (0.094)	-0.098 (0.159)	-0.067 (0.727)	-0.080 (0.204)
% High School Only	-0.026 (0.141)	0.031 (0.106)	0.058 (0.193)	0.004 (0.074)	0.032 (0.127)
EU Member	-1.017 (1.285)	-0.546 (1.025)	-0.815 (1.261)	-0.653 (0.601)	-4.368** (2.083)
Uses Euro	3.864* (2.059)	3.797** (1.892)	4.082 (2.920)	1.594 (1.305)	-0.983 (2.179)
% Blue Collar	-0.115 (0.102)	-0.132 (0.100)	-0.127 (0.183)	-0.054 (0.072)	0.338 (0.383)
Inflation	0.058 (0.182)	0.185 (0.180)
Lagged Inflation			0.297** (0.139)	0.066 (0.073)	-0.027 (0.170)
Voter Turnout	-0.082* (0.049)	-0.082** (0.041)	-0.083 (0.159)	-0.047 (0.029)	-0.041 (0.160)
RGDP Per Capita Growth	-0.092 (0.221)	-0.009 (0.196)
Lagged RGDP Per Capita Growth			0.040 (0.164)	-0.024 (0.148)	0.038 (0.184)
Pre-Crisis		-0.289 (1.467)	-0.238 (2.283)	0.995 (1.054)	1.694 (2.367)
Great Recession		0.937 (3.046)	1.438 (2.684)	2.247 (1.994)	4.528 (3.931)
Sovereign Debt Crisis		-2.624* (1.461)	-1.741 (1.591)	-0.984 (0.850)	0.121 (1.682)
Constant	-0.050 (26.009)	0.647 (22.933)	-1.420 (37.361)	4.011 (16.771)	77.707 (69.178)
Year FE	X
Region FE	X	X	X	X
Clustered SE			X	X	X
Country FE Model					X
R2	0.4135	0.3933	0.3920	0.7739	0.8540
Adj R2	0.2531	0.3152	0.3094	0.7372	0.7784
n	150	150	143	130	130

*** indicates significant at 1% level, ** indicates significant at 5% level,
* indicates significant at 10% level. Robust standard errors are given in parentheses.

 

Table 3: Left Populists vs Right Populists

	Left Populists	Right Populists
	1	2
Lagged Unemployment	0.692*** (0.241)	-0.325 (0.171)
Lagged Left Populist Vote	0.431** (0.188)
Lagged Right Populist Vote		0.151 (0.094)
% Aged 0-24	-1.7628 (1.067)	-0.307 (1.185)
% Aged 25-64	-0.642 (0.921)	-0.534 (1.006)
% Less than High School	-0.080 (0.204)	0.557 (0.319)
% High School Only	0.032 (0.127)	0.818 (0.276)
EU Member	-4.368** (2.083)	-5.566 (2.058)
Uses Euro	-0.983 (2.179)	-4.274 (2.865)
% Blue Collar	0.338 (0.383)	-0.049 (0.421)
Lagged Inflation	-0.027 (0.170)	0.287 (0.228)
Voter Turnout	-0.041 (0.160)	0.082 (0.103)
Lagged RGDP Per Capita Growth	0.038 (0.184)	-0.036 (0.162)
Pre-Crisis	1.694 (2.367)	-10.115 (4.360)
Great Recession	4.528 (3.931)	-6.492 (3.750)
Sovereign Debt Crisis	0.121 (1.682)	-5.332 (2.196)
Constant	77.707 (69.178)	3.588 (80.357)
Region FE	X	X
Clustered SE	X	X
R2	0.8540	0.9337
Adj R2	0.7784	0.8993
n	130	130

*** indicates significant at 1% level, ** indicates significant at 5% level,

* indicates significant at 10% level. Robust standard errors are given in parentheses.

To further test Rodrik’s theory about the differing causes of right-wing versus left-wing populism, Table 3 compares the preferred specification for left-wing populists – column 4 from Table 2 – to the same model with right-wing populist vote share as the outcome variable. The average summed right-wing populist vote share from 2000-2019 is shown for each country in the map below. The effect of unemployment on right-wing populist vote share is negative but statistically insignificant, suggesting that unemployment has little to no effect on support for right-wing populist parties. Taken together, the results of these regressions lend even more support to Rodrik’s theory that economic factors increase support for left-wing populists but not for right-wing populists.

 

Table 4: Left Populists vs Right Populists Subsample^†

	Left Populists	Right Populists
	1	2
Lagged Unemployment	1.268* (0.639)	-0.106 (0.415)
Lagged Left Populist Vote	-0.043 (0.266)
Lagged Right Populist Vote		0.055 (0.107)
% Aged 0-24	-2.226 (4.064)	-0.610 (2.007)
% Aged 25-64	-0.058 (4.984)	-0.104 (1.468)
% Less than High School	-2.038 (0.)	0.328 (0.402)
% High School Only	0.241 (1.502)	0.900*** (0.307)
EU Member	Omitted	-9.035** (4.237)
Uses Euro	3.986 (6.166)	-9.343** (4.222)
% Blue Collar	2.710 (2.260)	-0.352 (0.885)
Lagged Inflation	-1.567 (3.078)	0.396 (0.339)
Voter Turnout	-0.265 (0.356)	0.251* (0.133)
Lagged RGDP Per Capita Growth	0.120 (0.261)	0.033 (0.312)
Pre-Crisis	11.654 (19.399)	-9.520 (6.082)
Great Recession	15.056 (21.621)	-2.689 (5.843)
Sovereign Debt Crisis	7.087 (15.424)	-5.586* (3.227)
Constant	54.415 (305.883)	-4.922 (117.160)
Country FE	X	X
Clustered SE	X	X
R2	0.8740	0.9361
Adj R2	0.6510	0.8886
n	37	90

*** indicates significant at 1% level, ** indicates significant at 5% level,
* indicates significant at 10% level. Robust standard errors are given in parentheses.
† (Restricted to elections in which left populist vote share and right populist vote share, respectively, did not equal 0)

 

As a robustness check, I ran the same regressions for left-wing and right-wing populists from Table 3 but restricted the data to elections in which the left-wing populist vote share and the right-wing populist vote share, respectively, do not equal 0. These results are only a robustness check because the number of observations, particularly in the left-wing populist regression, is so low. Even so, the magnitude of the coefficient on unemployment for left-wing populists increases significantly: it increases from 0.692 percentage points to 1.268 percentage points. This coefficient significant at the ten percent level while the coefficient on unemployment in the right-wing populist regression is not significant. The implication of this subsample robustness check is that if there were more elections with left-wing populists in the study period (with much of my limitation being that elections in general do not occur very frequently which inherently limits my number of observations), the magnitude of unemployment may be even stronger than the results in my preferred specification suggest.

Finally, I ran a model which examines changes in unemployment rather than levels of unemployment. The change in unemployment is measured by the change in percentage points in the unemployment from the previous year. Similarly, the outcome variable in this model measures the change in percentage points in the summed left-populist vote share from the previous election. The results of this regression are shown in table 5 below. Importantly, the variable of interest, unemployment, is no longer significant when analyzed through changes rather than levels as in the regressions above.

When compared to tables 1-4, the results from table 5 suggest that changes in unemployment do not matter so much as levels of unemployment.[1] In other words, temporarily high unemployment rates do not appear to be driving vote shares for left populist parties in the same way that persistently high levels of unemployment do. This finding could have important policy implications. If policymakers want to prevent increased left-wing populism, they need not be as concerned with preventing temporarily high unemployment rates. Rather, they should focus their efforts on addressing persistently high rates of unemployment.

 

Table 5: Analysis Using Changes

	Left Populists
Change in Unemployment	-0.013 (0.035)
Lagged Left Populist Vote	-0.033* (0.018)
% Aged 0-24	0.065 (0.070)
% Aged 25-64	-0.022 (0.054)
% Less than High School	0.008 (0.016)
% High School Only	0.006 (0.011)
EU Member	-0.580* (0.325)
Uses Euro	-0.122 (0.147)
% Blue Collar	-0.012 (0.030)
Lagged Inflation	-0.009 (0.017)
Voter Turnout	-0.025** (0.011)
Lagged RGDP Per Capita Growth	-0.001 (0.010)
Pre-Crisis	0.059 (0.193)
Great Recession	0.173 (0.238)
Sovereign Debt Crisis	-0.035 (0.117)
Constant	5.284 (4.921)
Country FE	X
Clustered SE	X
R2	0.4094
Adj R2	0.1000
n	129

*** indicates significant at 1% level, ** indicates significant at 5% level,

* indicates significant at 10% level. Robust standard errors are given in parentheses.

VI. Conclusion and Limitations

Overall, my results suggest that higher unemployment is positively related to far-left populist vote share. The effect is statistically significant at the 1% level but small: in my preferred specification an increase of one percentage point in the unemployment rate leads to a 0.692 percentage point increase in the vote share for left populists. The effect of unemployment on the vote share for right populists is negative, but statistically insignificant. When taken together, these results lend support to Rodrik’s theory about the differing causes of left-wing and right-wing populism. Although a 0.692 percentage point increase is highly unlikely to sway an election, these results suggest that economic factors do increase support for left-wing populists while having little or no effect on right-wing populists.

This study has several limitations. First, research on populism is inherently limited by the lack of a clear definition, though the consensus-based approach taken by the PopuList is likely to remedy some of this concern. Furthermore, although the authoritarian nature of populists is much of what presents concerns from a policy standpoint (e.g., through the weakening of democratic institutions), Cass Mudde’s definition does not require that a party be authoritarian to be considered a populist, so my data does not necessarily capture that component of populism. In addition, the variables for populism, far-left, and far-right are binary and therefore do not account for parties with changing severities of ideology over time. Finally, this study is limited in that the analysis was performed on a country level rather than a regional or electoral district level, which could allow for more precise analysis.

Since the period covered by this study ends before 2020, future research could consider how the COVID-19 pandemic has affected voters’ demand for populists in relation to unemployment. In addition, future research could explore the other side of Rodrik’s theory: cultural factors, and whether they differentially impact support for right-wing versus left-wing populists. Future research on either the economic or cultural side of Rodrik’s theory could also employ the additive versus interactive effects of economics and culture. Future research could also consider using two stage models to examine the supply and demand for left populists together.

In terms of policy implications, there is more work to be done in the realm of understanding what factors are driving demand before concrete policy can be suggested. To aid in understanding and research, more candidates and parties, particularly those on a local level, could be identified in a manner similar to the PopuList to allow an analysis at a more precise level, and to see if relationships between populist vote share and variables such as unemployment hold at multiple levels of government. Regarding populism in general, Rodrik asserts that policymakers must take on the challenge of rebalancing the gains of globalization to curb the excesses of an open world economy.[26] Certainly, this call to action is relevant in the case of unemployment, as global trade and financial shocks differentially affect different areas and countries.

 

VII. References

Autor, David H., David Dorn, and Gordon H. Hanson. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.” American Economic Review 103, no. 6 (2013). DOI: 10.1257/aer.103.6.2121, 2121-2168.

Autor, David, David Dorn, Gordon Hanson, and Kaveh Majlesi. “Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure.” American Economic Review 110, no. 10 (2020). DOI: 10.1257/aer.20170011, 3139-3183.

Algan, Yann, Sergei Guriev, Elias Papaioannou, and Evgenia Passari. “The European Trust Crisis and the Rise of Populism.” Brookings Papers on Economic Activity (2017) https://www.jstor.org/stable/90019460. 309-382.

Bergman, Tarbjörn, Hanna Bäck, and Johan Hellström. Coalition Governance in Western Europe. Oxford: Oxford University Press, 2021.

Dijkstra, Lewis, Hugo Poelman, and Andrés Rodríguez-Pose. “The Geography of EU Discontent.” Regional Studies 54, no. 6 (2020). https://doi.org/10.1080/00343404.2019.1654603. 737-753.

Döring, Holger, Constantin Huber, and Philip Manow. Parliaments and Governments Database (ParlGov). 2022. https://www.parlgov.org/data-info/.

European Union. Country Profiles (website). https://european-union.europa.eu/principles-countries-history/country-profiles_en.

Eurostat. Various Datasets. Available from Eurostat website: https://ec.europa.eu/eurostat/web/main/data.

Guiso, Luigi, Helios Herrera, Massimo Morelli, and Tommaso Sonno. “Global Crises and Populism: The Role of Eurozone Institutions.” Economic Policy 34, no. 97 (2019). https://doi.org/10.1093/epolic/eiy018, 95-139.

Guiso, L., H. Herrera, M. Morelli, and T. Sonno. “Populism: Demand and Supply.” CEPR Discussion Paper Series (2017). https://www.tcd.ie/Economics/assets/pdf/Seminars/20172018/H.%20Herrera.pdf.

Guriev, Sergei and Elias Papaiaonnou. “The Political Economy of Populism.” Journal of Economic Literature 60, no. 3 (2022). https://doi.org/10.1257/jel.20201595. 753-832.

March, Luke. Radical Left Parties in Europe. New York: Routledge, 2011.

Mudde, Cass. Populist Radical Right Parties in Europe. Cambridge: Cambridge University Press, 2007.

Mudde, Cass. “The Populist Zeitgeist.” Government and Opposition 39, no. 4 (2004). https://doi.org/10.1111/j.1477-7053.2004.00135.x. 541-563.

Rodríguez-Pose, Andrés. “The Revenge of the Places that Don’t Matter (and what to do about it).” Cambridge Journal of Regions, Economy and Society 11 (2018). doi:10.1093/cjres/rsx024. 189-209.

Rodrik, Dani. “Is Populism Necessarily Bad Economics?” American Economic Association 108 (2018). https://doi.org/10.1257/pandp.20181122. 196-199.

Rodrik, Dani. “Populism and the Economics of Globalization.” Journal of International Business Policy (2018). https://doi.org/10.1057/s42214-018-0001-4. 12-33.

Rooduijn, M., S. Van Kessel, C. Froio, A. Pirro, S. De Lange, D. Halikiopoulou, P. Lewis, C. Mudde, and P. Taggart. The PopuList: An Overview of Populist, Far Right, Far Left and Eurosceptic Parties in Europe. 2019. www.popu-list.org.

[1] These findings still hold when the level of left populist vote share is used as the outcome variable.

VIII. Endnotes

[1] Dani Rodrik, “Is Populism Necessarily Bad Economics?” American Economic Association 108 (2018), https://doi.org/10.1257/pandp.20181122, 196.

[2] Dani Rodrik, “Populism and the Economics of Globalization,” Journal of International Business Policy (2018), https://doi.org/10.1057/s42214-018-0001-4, 2.

[3] Ibid., 13.

[4] Ibid., 2.

[5] Sergei Guriev and Elias Papaiaonnou, “The Political Economy of Populism,” Journal of Economic Literature 60, no. 3 (2022), https://doi.org/10.1257/jel.20201595, 792.

[6] Luigi Guiso et al., “Populism: Demand and Supply,” CEPR Discussion Paper Series (2017), https://www.tcd.ie/Economics/assets/pdf/Seminars/20172018/H.%20Herrera.pdf, 1.

[7] Rodrik, “Populism and the Economics of Globalization,” 5.

[8] Ibid., 12.

[9] Luigi Guiso et al., “Global Crises and Populism: The Role of Eurozone Institutions,” Economic Policy 34, no. 97 (2019), https://doi.org/10.1093/epolic/eiy018, 98.

[10] Lewis Dijkstra, Hugo Poelman, and Andrés Rodríguez-Pose, “The Geography of EU Discontent,” Regional Studies 54, no. 6 (2020), https://doi.org/10.1080/00343404.2019.1654603, 737.

[11] David H. Autor, David Dorn, and Gordon H. Hanson, “The China Syndrome: Local Labor Market Effects of Import Competition in the United States,” American Economic Review 103, no. 6 (2013), DOI: 10.1257/aer.103.6.2121, 2121.

[12] David Autor, David Dorn, Gordon Hanson, and Kaveh Majlesi, “Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure,” American Economic Review 110, no. 10 (2020), DOI: 10.1257/aer.20170011, 3139.

[13] Yann Algan et al., “The European Trust Crisis and the Rise of Populism,” Brookings Papers on Economic Activity (2017), https://www.jstor.org/stable/90019460, 330.

[14]Lewis Dijkstra, Hugo Poelman, and Andrés Rodríguez-Pose, “The Geography of EU Discontent,” 743.

[15] Ibid.

[16] Guiso et al., “Global Crises and Populism: The Role of Eurozone Institutions,” 109.

[17] Rodrik, “Populism and the Economics of Globalization,” 12.

[18] Andrés Rodríguez-Pose, “The Revenge of the Places that Don’t Matter (and what to do about it),” Cambridge Journal of Regions, Economy and Society 11 (2018), doi:10.1093/cjres/rsx024, 196.

[19] Lewis Dijkstra, Hugo Poelman, and Andrés Rodríguez-Pose, “The Geography of EU Discontent,” 743.

[20] Algan et al., “The European Trust Crisis and the Rise of Populism,” 320.

[21] Cas Mudde, “The Populist Zeitgeist,” Government and Opposition 39, no. 4 (2004), https://doi.org/10.1111/j.1477-7053.2004.00135.x, 543.

[22] Cass Mudde, Populist Radical Right Parties in Europe, (Cambridge: Cambridge University Press, 2007).

[23] Luke March, Radical Left Parties in Europe (New York: Routledge, 2011), 8.

[24] European Union, Country Profiles (website), https://european-union.europa.eu/principles-countries-history/country-profiles_en.

[25] Torbjörn Bergman, Hanna Bäck, and Johan Hellström, Coalition Governance in Western Europe (Oxford: Oxford University Press, 2021) 6.

[26] Rodrik, “Populism and the Economics of Globalization,” 16.