Populism and Support for Protectionism: The Relevance of Opposition to Trade Openness for Leftist and Rightist Populist Voting in The Netherlands.

Leftist and rightist populist parties in Western Europe both oppose trade openness. Is support for economic protectionism also relevant for their electorates? We assess this in the Netherlands, where both types of populist parties have seats in parliament. Analyses of representative survey data (n = 1,296) demonstrate that support for protectionism drives voting for such parties, as do the well-established determinants of political distrust (both populist constituencies), economic egalitarianism (leftist populist constituency) and ethnocentrism (rightist populist constituency). Surprisingly, support for protectionism does not mediate the relationship between economic egalitarianism and voting for left-wing populists, or the link between political distrust and voting for either left-wing or right-wing populist parties. In contrast, support for protectionism partly mediates the association between ethnocentrism and voting for right-wing populists. We discuss the largely independent role of protectionism in populist voting in relation to the cultural cleavage in politics and electoral competition, and also provide suggestions for future research.

Populist politicians and parties on both sides of the Atlantic claim to ‘truly’ cater to ‘the needs of the common man’ (March, 2011; Rooduijn and Akkerman, 2015; Rooduijn et al., 2014; Schumacher and Rooduijn, 2013). One way in which they promise to do so is opposing free trade (also referred to as ‘support for protectionism’), which featured prominently during the 2016 presidential election in the United States, as it did in the prelude to the national parliamentary elections in various Western European countries in 2017. Donald Trump, for instance, stated, ‘I am all for free trade, but it’s got to be fair. When Ford moves their massive plants to Mexico, we get nothing. I want them to stay in Michigan’ (OnTheIssues, 2016). This is reminiscent of Ross Perot’s fierce opposition to the North American Free Trade Association (NAFTA) in the 1990s (Hawthorn, 1994). Perot, a presidential candidate, resisted NAFTA because it would lead to, as he put it, ‘a giant sucking sound going south’, which was a metaphor for the relocation of industrial production to Mexico and a loss of manufacturing jobs in the United States.

Similar stances can be found among populists in Europe. Leftist populist parties, such as the Socialistische Partij (SP; Socialist Party) in the Netherlands and die Linke (the Left) in Germany,[1] as well as right-wing populists like Front National (National Front) in France and Partij voor de Vrijheid (PVV; Party for Freedom) in the Netherlands, are fierce opponents of common markets, especially when it comes to their expansion. Such parties argued vociferously against the Bolkenstein Directive in 2006, for instance, by warning of the ‘unfair competition’ that workers would endure from proverbial ‘Polish plumbers’ entering North-West European labour markets. Furthermore, in the prelude to the parliamentary elections of 2017, they have fiercely opposed trade agreements such as the Transatlantic Trade and Investment Partnership (TTIP) (Charlemagne, 2015). This opposition to trade openness neatly fits into the more encompassing agendas of these parties, which aim to shield the ‘common man’ from the perceived vagaries of economic globalisation, agreements made in so-called ‘elitist institutions’ such as the European Union, and the alleged ‘violation’ of national cultures and sovereignty.

It is, however, not yet clear whether support for protectionism also characterises the populist electorate. Having clarity on this will deepen our understanding of the social bases of populism and uncover the potential electoral relevance of support for protectionism, especially when free trade treaties are politicised. We therefore aim to answer the following questions: Does support for protectionism drive voting for leftist and rightist populist parties? And, if so, how can this be explained? To do so, we compare the electorates of the populist left and populist right to those of non-populist parties. Focusing on both leftist and rightist populist constituencies is especially relevant, because these groups may have different reasons for opposing free trade due to their different attitudinal profiles.

As trade openness increases economic inequality due to the falling demand for lower skilled workers (Mayda and Rodrik, 2005), it may be that support for protectionism is part of the link between economic egalitarianism and support for left-wing populist parties. A right-wing objection to trade openness could also be in play. More specifically, since opposition to free trade is part of a more encompassing resistance to perceived infringements of national culture and may reflect a sense of national superiority (Mansfield and Mutz, 2009; Margalit, 2012; Mutz and Kim, in press; Van der Waal and De Koster, 2015), support for protectionism may be part of the link between ethnocentrism and nationalism on the one hand and support for right-wing populist parties on the other. Additionally, if trade associations and partnerships such as NAFTA and TTIP are considered to be projects of ‘corrupt political elites’ beyond the control of the common man (Kaina, 2008; Mudde, 2004), support for protectionism might be part of the well-established link between political distrust and support for both leftist and rightist populist parties (Hooghe and Oser, 2015; Hooghe et al., 2011; Kemmers et al., 2016).

In what follows, we elaborate on these three potential explanations of support for protectionism among populist constituencies. Subsequently, we will test our hypotheses by analysing survey data that are representative of the Dutch population in 2012. The Netherlands is a suitable case study, as it has had left- and right-wing populist parties in parliament simultaneously from 2006 onwards. It is thus ideal for mapping and explaining the protectionist views of the constituencies of both types of populist party in contemporary Europe. The Dutch populist parties are the SP on the left and Geert Wilders’ PVV on the right (Hakhverdian and Koop, 2007; Otjes and Louwerse, 2015; Schumacher and Rooduijn, 2013; Spierings and Zaslove, 2017).[2]

Support for Protectionism among Populist Constituencies: Three Explanations

Resistance to the Distributive Consequences of Trade Openness

A well-established pattern in the scholarly debate on the antecedents of protectionist views is that the less educated oppose free trade much more than their better educated counterparts (Hainmueller and Hiscox, 2006; Van der Waal and De Koster, 2015). The dominant explanation for this pattern is the so-called ‘factor endowment model’ (Bechtel et al., 2012; Mayda and Rodrik, 2005), which claims that the highly skilled in advanced economies have more to gain from free trade than their lower skilled counterparts. This is because the (relative) demand for low-skilled workers falls if international trade is conducted with countries with an abundance of these workers, while the demand for the highly skilled rises. As a result, the employment opportunities and wages of the latter increase, while those of the former decrease, resulting in growing economic inequality.

This is precisely why leftist parties resist the relentless distributive impact of unfettered markets. Their support for protectionism is part of a more encompassing economically egalitarian agenda that includes other policies aimed at mitigating the inequalities resulting from free trade, such as welfare arrangements and a tax system that redistributes income from the rich to the poor. As expected, this approach appeals to their electorate: support for traditional leftist parties that oppose economic inequality is primarily driven by support for their economically egalitarian agendas (Achterberg and Houtman, 2006; Houtman et al., 2008; Van der Waal et al., 2007). This pattern inspires the first potential explanation for protectionist views among populist constituencies: they may oppose the distributive consequences of trade openness.

This is most obvious for supporters of leftist populist parties, as they most strongly promote an economically egalitarian agenda and appeal to the most economically egalitarian constituency in the Netherlands (De Koster et al., 2013). Nevertheless, people who oppose trade openness because of an aversion to economic inequality might also be found among the electorate of rightist populist parties. Of course, these parties primarily appeal to their electorate on the basis of their opposition to immigration (De Koster et al., 2014; Ivarsflaten, 2008; Rydgren, 2008) and their nativist conception of welfare and distribution issues (Abts and van Kessel, 2015; De Koster et al., 2013; Eger and Valdez, 2015). Nonetheless, the level of economic egalitarianism among PVV voters is remarkably high for a right-wing party. While the electorate of the PVV is less egalitarian than that of the SP and the Partij van de Arbeid (PvdA; Labour Party), it is markedly more opposed to economic inequality than voters who prefer other right-of-centre parties (De Koster et al., 2013). Support for protectionism among the voters of right-wing populist parties may, therefore, also be explained by their economic egalitarianism.

Overall, the constituencies of populist parties may oppose trade openness because they resist the distributive consequences of free trade. This would mean that support for protectionism is part of the link between economic egalitarianism and support for populist parties. Technically, we expect that support for protectionism mediates a positive effect of economic egalitarianism on voting for populist parties (Hypothesis 1).

While what is set out above is especially to be expected for those supporting the leftist populist party under investigation, the SP, a cultural explanation might be more relevant when it comes to support for the rightist Dutch populist party, the PVV. This is elaborated on below.

Resistance to the Cultural Consequences of Trade Openness

Various studies cast doubt on the factor endowment model as an explanation for the protectionist views of the less educated (Hainmueller and Hiscox, 2006; Margalit, 2012; Van der Waal and De Koster, 2015), and suggest that alternative explanations are needed. One of these is that people may perceive trade openness to be a cultural-order instead of an economic-redistribution issue. As international trade entails a process that transcends national borders and cultural divides, a free flow of goods and services affects and alters the cultural and national order (Mudde, 2007; Van der Waal and De Koster, 2015). This may be a reason to oppose it, which is suggested in studies that demonstrate a substantial positive relationship between opposition to trade openness and ethnocentrism and nationalism (e.g. Edwards, 2006; Kaltenthaler et al., 2004; Mansfield and Mutz, 2009; Mayda and Rodrik, 2005; Mutz and Kim, in press; Van der Waal and De Koster, 2015; Wolfe and Mendelsohn, 2005).

Research indicates that PVV voters are primarily characterised by high levels of ethnocentrism (De Koster et al., 2014; Ivarsflaten, 2008; Rydgren, 2008), and the above suggests that this could explain their support for protectionism. Although less obvious, it could also explain the opposition to trade openness among SP voters, even if the protectionism at the left-wing populist party level is informed by other reasons. This is in line with a recent study which indicates that, due to the absence of parties that explicitly combine economic egalitarianism with resistance to immigration in Western Europe, the part of the electorate with this specific attitudinal profile opts for leftist populist parties at the ballot box (Lefkofridi et al., 2014). These voters comprise substantial parts of the electorate, especially in the Netherlands, where they often vote for the SP (Lefkofridi et al., 2014: 71–72). In general, SP voters are certainly not as ethnocentric as PVV voters (Akkerman et al., 2014; De Koster et al., 2014), but ethnocentrism might nevertheless be relevant when it comes to explaining why their opposition to trade openness is higher than among supporters of non-populist parties. A similar argument applies to the potential role of nationalism. At the party level, nationalism is more prominently associated with rightist rather than leftist populism (Immerzeel et al., 2016), but this does not, by definition, imply that notions of national superiority can only be found among the former’s constituency. This may, therefore, explain the support for protectionism of both leftist and rightist populist constituencies. Again, the latter case is, of course, most likely.

In short, support for protectionism among supporters of leftist and, especially, rightist populist parties may be due to resistance to the cultural consequences of trade openness. This would mean that support for protectionism is part of the link between ethnocentrism and nationalism on the one hand and support for populist parties on the other. Technically, we expect that support for protectionism mediates positive effects of ethnocentrism and nationalism on voting for populist parties (Hypotheses 2a and 2b).

Distrust of Political Institutions and Politicians

The two explanations of protectionist views among populist constituencies outlined above align with two types of concern that are often found among the public at large: economic and cultural. As such, they are related to recognisable ideological positions: trade openness might be opposed by those who are ‘economically leftist’ or by those who are ‘culturally rightist’. These two explanations are most likely to account for the electorates of left-wing and right-wing populist parties, respectively. The final explanation, in contrast, is not related to such clear, overarching, ideological and party positions. Instead, it revolves around what is considered to be the defining characteristic of both contemporary and past populism in both its leftist and rightist guise: the claim that there is an unwarranted and unacceptable power differential between ‘the people’ and ‘the elite’ (Akkerman et al., 2014; Canovan, 2005; Mudde, 2004, 2007; Rooduijn, 2014; Schumacher and Rooduijn, 2013; Spruyt et al., 2016; Taggart, 2000, 2004).

The opposition between ‘the pure people’ and ‘the corrupt elite’ (Mudde, 2004: 543) may drive the protectionist views of populist constituencies: they could perceive trade openness as a project of untrustworthy political elites. Trade openness is all but a natural state of affairs. On the contrary: enabling free trade across borders often calls for long and difficult negotiations between the politicians, diplomats and policy-makers of the countries involved, and this includes the usual bickering, trickery and vague engagements. Consider, for instance, the political processes behind the NAFTA, the Bolkenstein Directive and the TTIP. Trade openness is, in short, an intrinsically political phenomenon, and the trade agreements upon which it is based result from opaque political processes.

Support for protectionism among populist constituencies might, therefore, not so much be driven by their economic or cultural concerns; instead, it could be inspired by their well-established ‘strong anti-institutional impulse’ (Kriesi, 2014: 363, compare Canovan, 1999). More specifically, it may be motivated by their distrust of political institutions like the government and parliament, as well as those that populate them, namely politicians (Hooghe et al., 2011; Kaina, 2008; Van der Meer, 2010). This political distrust is characteristic of the constituencies of both leftist and rightist populist parties (Hooghe et al., 2011; Hooghe and Oser, 2015; Kemmers et al., 2016). As a result, support for protectionism among supporters of these parties might be high because they conceive free international trade to be a project of untrustworthy political elites. This would mean that support for protectionism is part of the link between political distrust and support for populist parties. Technically, we expect that support for protectionism mediates a positive effect of political distrust on voting for populist parties (Hypothesis 3).

Data and Operationalisation

The data for this study (Achterberg et al., 2012) were collected in 2012 by the research institute CentERdata, which is affiliated with Tilburg University. It maintains a panel aged 16 and older that is representative of the Dutch population. Panel members complete questionnaires online, with respondents who do not have Internet access provided with the necessary equipment. A total of 1707 individuals were invited to participate in the study, with 1302 completing the questionnaire (response rate: 76.3%). Those who finished the survey in less than 10 minutes, which is the minimum time reasonably required to provide valid answers, were removed, leaving a dataset of 1296 respondents.

The dependent variable support for protectionism is measured by the following item:

Some people have suggested that the Dutch government should limit imports in order to protect jobs. Others say that such limits would raise consumer prices and would hurt Dutch exports. Do you favour or oppose import limitations?

The response categories ranged from (1) strongly favour to (5) strongly oppose, and were reversed to measure support for protectionism.[3] This item was inspired by the one that is used most often in the relevant literature (most notably in studies based on the widely used American National Election Studies, ANES; e.g. Hainmueller and Hiscox, 2006; Scheve and Slaughter, 2001), and has been utilised before in the Dutch context (Van der Waal and De Koster, 2015). It should be noted that the item is formulated with reference to the economic consequences of free trade. This suggests that our analyses will provide: a lenient test for the explanation for the protectionist views of populist constituencies which focuses on economic egalitarianism (Hypothesis 1), and strict tests for the two non-economic explanations (Hypotheses 2 and 3).

The variable vote choice consists of three categories and is based on responses to a question concerning which party respondents would vote for if there were elections for the national parliament tomorrow. We coded those preferring the SP as a vote for the populist left and those opting for the PVV and the splinter parties Democratisch Politiek Keerpunt (DPK; Democratic Political Turning Point) and Trots op Nederland (TON; Proud of the Netherlands) as a vote for the populist right. Finally, the category vote for non-populist parties comprises respondents favouring centrist, small Christian and new-leftist parties. The three main centre parties are as follows: Partij van de Arbeid (PvdA, Labour Party), Volkspartij voor Vrijheid en Democratie (VVD; Liberal Conservatives) and Christen-Democratisch Appèl (CDA; Christian Democratic Appeal). The smaller Christian parties are as follows: ChristenUnie (CU; ChristianUnion) and Staatskundig Gereformeerde Partij (SGP; Reformed Political Party), while the new-leftist parties are Democraten 66 (D66; Liberal Democrats) and GroenLinks (GL; Green Left). The variable vote for non-populist parties serves as the category of reference in our main analyses.[4] A more narrowly defined version consisting of respondents who voted for the three main centre parties (PvdA, VVD and CDA) is used as the reference category in our first robustness check. This check also entails a comparison of those who vote for the SP with people who prefer the left-wing centre party, the PvdA, and the people who vote for the PVV with those opting for the right-wing centre party, the VVD.

Voting intentions for the splinter parties Partij voor de Dieren (PvdD; Party for the Animals, n = 12) and 50Plus (50+; Party for the Elderly, n = 13) were coded as missing because they are hard to classify in one of the three categories described above. Respondents who indicated they would not vote (n = 30), are ‘not allowed to vote’ (n = 16), ‘don’t know yet’ (n = 201) or ‘won’t say’ (n = 14) were also coded as missing.

Economic egalitarianism was measured by four items previously used in the Dutch context (e.g. Achterberg and Houtman, 2009; De Koster et al., 2013):

The state should raise social benefits.

There is no longer any real poverty in the Netherlands.

Large income differences are unfair because everyone is essentially equal.

Companies should be obliged to allow their employees to share the profits.

Each item had five response categories ranging from (1) strongly disagree to (5) strongly agree. A factor analysis of these items yielded a first factor explaining 49% of the variance, with factor loadings ranging from 0.59 to 0.76. The sum scale calculated on the basis of these four items for respondents with at least a valid answer to three of them proved to be moderately reliable (Cronbach’s α = 0.64). Because of this, the second robustness check performed consists of four analyses that each uses one separate item for economic egalitarianism instead of the scale.

Ethnocentrism was measured with six items derived from Eisinga and Scheepers (1989) that indicate negative prejudice towards out-groups. The sum scale constructed from the following six items has recently been used in other Dutch research (De Koster et al., 2014; Van Bohemen et al., 2012):

Foreigners carry all kinds of dirty smells around.

With Moroccans you never know for certain whether or not they are going to be aggressive.

Most people from Surinam work quite slowly.

Most Turks are rather self-indulgent at work.

Foreigners living in the Netherlands should adapt to Dutch uses and customs.

The Netherlands should have never let foreign guest workers in.

The response categories ranged from (1) strongly disagree to (5) strongly agree. A factor analysis of these items revealed a first factor explaining 62% of the variance, with factor loadings ranging from 0.62 to 0.87. We calculated the sum score for the respondents with a valid answer to at least four of these items (Cronbach’s α = 0.87).

Nationalism was measured by three items borrowed from the nationalism measure included in the ISSP National Identity Modules in 1995, 2003 and 2013 (compare Davidov, 2008):

I would rather be a citizen of the Netherlands than of any other country in the world.

The world would be a better place if people from other countries were more like the Dutch.

People should support the Netherlands even if it is in the wrong.

Their response categories ranged from (1) strongly disagree to (5) strongly agree. The first factor revealed by a factor analysis of these items explained 59% of the variance. The factor loadings ranged from 0.69 to 0.81. We calculated the sum score for respondents with a valid answer to at least two of these items (Cronbach’s α = 0.65). As the reliability is moderate, a third robustness check is applied, with three analyses each utilising one separate nationalism item instead of the scale.

We utilised three items previously used in the Netherlands to measure political distrust (Achterberg and Mascini, 2013). The first two asked to what extent the respondents trust (1) the institution of politics and (2) politicians. These items, with five answer categories ranging from (1) absolutely no trust to (5) certainly a lot of trust, were reverse coded. The third item asked to what extent the respondents agreed with the statement ‘Politicians seldom speak the truth’, with answer categories ranging from (1) strongly disagree to (5) strongly agree. The first factor in a factor analysis of these three items explained 80% of the variance, with factor loadings ranging from 0.83 to 0.93. These added up to a reliable scale (Cronbach’s α = 0.87) created for the respondents with valid answers on all three items.

We also include controls for various background characteristics that are known to be related to economic concerns, cultural concerns, political distrust and voting behaviour. These include gender (male = 0; female = 1), age in years, non-native (parents born in the Netherlands = 0; parents not born in the Netherlands = 1), education (the minimum number of years formally needed to attain the highest level of education achieved by the respondent, ranging from 8 = only primary education to 18 = university degree), net monthly household income (four categories: 1. €1,150 or less; 2. €1,151–€1,800; 3. €1,801–€2,600; 4. €2,601 or more), labour-market position (dummies: 1. not in labour market (ref.); 2. employed; 3. partially employed 4. unemployed), religious denomination (dummies: 1. no religious denomination (ref.); 2. Protestant; 3. Catholic; 4. other religious denomination); and attendance at religious services (dummies: 1. No attendance (ref.); 2. occasional attendance, (ranging from once a month to less than once a year); 3. frequent attendance (once a week or more often)). Table 1 reports the descriptive statistics.

Table 1.

Descriptive Statistics.

	Mean	SD	Range	n
Voting behaviour
Vote for non-populist parties (ref.)
Vote for populist left	0.17		0–1	980
Vote for populist right	0.06		0–1	980
Protectionism	2.32	0.87	1–5	1126
Economic egalitarianism	3.24	0.70	1–5	1259
Ethnocentrism	2.57	0.78	1–5	1213
Nationalism	3.58	0.71	1–5	1244
Political distrust	3.24	0.90	1–5	1243
Gender (female)	0.44		0–1	1296
Age	56.63	15.24	16–90	1296
Non-native	0.08		0–1	1274
Education	14.60	2.73	8–18	1295
Net household income	3.11	1.00	1–4	1294
Labour-market position
Not in labour market (ref.)
Employed	0.47		0–1	1271
Partially employed	0.02		0–1	1271
Unemployed	0.06		0–1	1271
Religious denomination
No religious denomination (ref.)
Protestant	0.21		0–1	1296
Catholic	0.28		0–1	1296
Other religious denomination	0.06		0–1	1296
Attendance at religious services
No attendance at religious services (ref.)
Occasional attendance at religious services	0.32		0–1	1280
Frequent attendance at religious services	0.13		0–1	1280

Results

Main Analyses

Prior to mapping and explaining the support for protectionism among Dutch populist constituencies, we explored whether the measures introduced above relate to support for protectionism. Table 2 indicates that they do, although some do so more convincingly than others. The relationships between support for protectionism and economic egalitarianism, ethnocentrism, nationalism and political distrust are in the expected direction: the more economically egalitarian, ethnocentric, nationalist and politically distrusting that people are, the higher their level of opposition to trade openness. Nevertheless, the strength of these relationships is modest, except for the association between ethnocentrism and support for protectionism. This suggests that trade openness is primarily a cultural concern among the Dutch electorate.

Table 2.

Zero-order Correlations between Potential Drivers of Voting for Populist Parties and Protectionism.

	r	n
Economic egalitarianism	0.10***	1116
Ethnocentrism	0.27***	1077
Nationalism	0.06*	1108
Political distrust	0.13***	1104

p < 0.001; **p < 0.01; *p < 0.05.

To provide insight into the variations in attitudinal profiles across the constituencies of left-wing populist, right-wing populist, and non-populist parties, Table 3 reports the mean scores on the potential drivers of populist voting. The first row provides an answer to the descriptive part of our research question, showing that the constituencies of populist parties are indeed characterised by support for protectionism. The remaining rows in Table 3 provide an initial indication of the potential explanations for this pattern. Unsurprisingly, economic egalitarianism can particularly be found among the leftist populist constituency, while its rightist counterpart stands out because of its high level of ethnocentrism. This suggests that economic concerns may explain the protectionist views of the former constituency, and cultural concerns the protectionist opinions of the latter. Somewhat surprisingly, populist constituencies are not more nationalistic than their non-populist counterparts; the leftist populist constituency is even less nationalistic. This means that nationalism is an unlikely candidate as an explanation for the greater support for protectionism among populist constituencies. Finally, both of the populist electorates report far more political distrust than the non-populist constituencies, indicating that distrust is a potential explanation for their support for protectionism.

Table 3.

Mean Scores of Potential Drivers of Voting for Populist Parties.

	Non-populist party constituency	Left-wing populist party constituency	Right-wing populist party constituency	n
Protectionism	2.18	2.51***	2.72***	889
Economic egalitarianism	3.11	3.67***	3.18	968
Ethnocentrism	2.45	2.52	3.44***	938
Nationalism	3.66	3.46**	3.49	963
Political distrust	2.97	3.54***	4.08***	961

p < 0.001; **p < 0.01; *p < 0.05 (left-wing and right-wing populist constituencies compared to non-populist party constituency; Bonferroni test for multiple comparisons).

Having discovered these initial, somewhat crude, insights, we conducted more elaborate analyses to test our hypotheses on the mediating role of support for protectionism in the links between economic egalitarianism, ethnocentrism, nationalism and political distrust on the one hand and populist voting on the other. Table 4 reports three multinomial regression analyses. The first model is in line with the main findings presented in Table 3. If controlled for various potential confounders, economic egalitarianism drives voting for the populist left, ethnocentrism voting for the populist right and political distrust voting for both left-wing and right-wing populist parties. Nationalism is not significantly (populist left) or negatively (populist right) related to a preference for a populist party, which means that Hypothesis 2b is rejected. Model 2 demonstrates that support for protectionism drives voting for populist parties, also when various potential confounders are included in the analysis.

Table 4.

Explaining Support for Protectionism among Populist Constituencies.

	Model 1		Model 2		Model 3
	Populist left	Populist right	Populist left	Populist right	Populist left	Populist right
Economic egalitarianism	1.30*** (0.19)	0.11 (0.25)			1.32*** (0.19)	0.16 (0.25)
Ethnocentrism	0.15 (0.16)	1.76*** (0.27)			0.05 (0.16)	1.67*** (0.28)
Nationalism	−0.23 (0.16)	−0.39 (0.23)			−0.27 (0.16)	−0.50* (0.24)
Political distrust	0.63*** (0.14)	1.31*** (0.23)			0.64*** (0.14)	1.40*** (0.24)
Protectionism			0.41*** (0.12)	0.83*** (0.16)	0.48*** (0.14)	0.75*** (0.21)
Gender (female)	0.09 (0.23)	−0.43 (0.38)	−0.08 (0.21)	−0.85* (0.33)	−0.08 (0.23)	−0.63 (0.40)
Age	−0.02* (0.01)	−0.02 (0.02)	−0.01 (0.01)	0.00 (0.01)	−0.02* (0.01)	−0.02 (0.02)
Non-native	0.09 (0.45)	0.49 (0.62)	0.00 (0.40)	0.46 (0.52)	0.10 (0.45)	0.57 (0.63)
Education	−0.08 (0.04)	0.07 (0.07)	−0.11** (0.04)	−0.05 (0.06)	−0.06 (0.05)	0.12 (0.08)
Household income	−0.23* (0.12)	−0.24 (0.19)	−0.3*** (0.11)	−0.26 (0.15)	−0.20 (0.12)	−0.13 (0.20)
Not in labour market	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Employed	0.71* (0.30)	0.03 (0.48)	0.46 (0.28)	−0.09 (0.42)	0.65* (0.30)	−0.09 (0.49)
Partially employed	0.15 (0.78)	0.66 (1.21)	−0.15 (0.73)	−0.48 (1.14)	−0.13 (0.80)	0.37 (1.21)
Unemployed	0.65 (0.50)	0.77 (0.62)	0.47 (0.46)	1.10* (0.52)	0.73 (0.50)	0.83 (0.63)
No religion	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Protestant	−0.32 (0.35)	−0.47 (0.55)	−0.36 (0.33)	−0.26 (0.48)	−0.37 (0.36)	−0.57 (0.57)
Catholic	0.02 (0.30)	−0.22 (0.45)	−0.03 (0.27)	−0.05 (0.41)	−0.00 (0.30)	−0.19 (0.45)
Other religion	−1.18 (0.70)	−1.93 (1.26)	−0.46 (0.65)	−0.71 (1.08)	−1.12 (0.71)	−2.06 (1.42)
No attendance at religious services	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
Occasional attendance at religious services	0.04 (0.27)	−0.52 (0.45)	−0.16 (0.25)	−0.67 (0.40)	0.05 (0.28)	−0.50 (0.46)
Frequent attendance at religious services	−1.91** (0.65)	−0.79 (0.73)	−2.22*** (0.63)	−1.41* (0.68)	−1.91** (0.66)	−0.77 (0.74)
Constant	−4.53** (1.42)	−9.87*** (2.43)	0.90 (0.93)	−2.35 (1.37)	−5.80*** (1.50)	−12.43*** (2.65)
Pseudo R2	0.28		0.11		0.30

Multinomial regression analyses. Reference category: vote for non–populist parties. Entries are log odds, standard errors in parentheses; n = 832.

p < 0.001; **p < 0.01; *p < 0.05.

Opposing trade openness is a more substantial driver of voting for a rightist populist party than for a leftist one. Model 3 is relevant when it comes to testing our hypotheses on the mediating role of protectionism in the link between economic egalitarianism, ethnocentrism, nationalism and political distrust, on the one hand, and support for populist parties on the other. The most salient finding of Model 3 is that support for protectionism is largely an independent driver of voting for a populist party. The positive effects of economic egalitarianism, ethnocentrism and political distrust in Model 1 are barely affected by including support for protectionism. Meanwhile, the effects of support for protectionism remain stable vis-à-vis those in Model 2. Nevertheless, Model 3 in Table 4 is not enough when it comes to establishing the precise mediating role of support for protectionism. As a result of the rescaling that occurs in each separate model in logistic regression analyses, coefficients cannot simply be compared between models (Mood, 2010). Consequently, we applied the KHB method (Breen et al., 2013), which was specifically designed to correct for this characteristic of logistic regression analyses, and also tests to what extent the effects of independent variables of interest on a dependent variable are mediated by another variable. The KHB analyses demonstrated that only one of the significant drivers of populist voting is significantly mediated by support for protectionism: 10% of the effect of ethnocentrism on right-wing populist voting is mediated by support for protectionism (p < 0.01). Put differently, the opposition to trade openness among the PVV constituency is part of the well-established link between their ethnocentrism and vote choice, and this corroborates Hypothesis 2a. In contrast, support for protectionism proved to be part of neither the well-established link between economic egalitarianism and support for a leftist populist party (which means that Hypothesis 1 is rejected), nor the well-established link between political distrust and support for both leftist and rightist populist parties (in contrast to Hypothesis 3).

We will now report the results of our robustness checks before discussing these findings in the concluding section.

Robustness Checks

The first robustness check tested the hypotheses by using different operationalisations of the reference category for voting behaviour. We started by comparing leftist and rightist populist constituencies to a smaller set of ‘non-populist parties’, that is, the main centre parties (PvdA, CDA, and VVD). This led to conclusions being reached with respect to our hypotheses that are similar to those reached in the main analyses. Taking the effects of all the other variables into account, support for protectionism proved to be an additional driver of voting for both of these parties (p < 0.01 and p < 0.01). Again, only Hypothesis 2b was corroborated: part of the effect of ethnocentrism on voting for the PVV (12%) was significantly mediated by support for protectionism (p < 0.01). We also conducted two types of binary logistic regression analysis on an even more restricted sample: the SP versus the PvdA and the PVV versus the VVD. Again, only Hypothesis 2b was corroborated: in the analyses comparing voting for the SP with voting for the PvdA, support for protectionism was not a significant mediator, but this variable did significantly mediate 12% of the effect of ethnocentrism in terms of preferring the PVV over the VVD (p < 0.05).

The second robustness check consisted of four analyses: one for each separate item of economic egalitarianism, instead of the moderately reliable scale. In all four cases, economic egalitarianism drove voting for the SP (p < 0.001), but support for protectionism did not significantly mediate these effects. Just like the main analyses, this indicates that Hypothesis 1 has to be rejected.

The third robustness check entailed three analyses: one for each separate item of nationalism instead of the moderately reliable scale. This produced the same conclusion as the main analyses: nationalism is not or even negatively related to voting for a populist party, and its effect is not mediated by opposition to trade openness. In accordance with the main analyses, this indicates that Hypothesis 2b is not corroborated.

Conclusion and Discussion

We studied whether support for protectionism drives voting for leftist and rightist populist parties, and, if so, how this can be explained. We did this by analysing survey data concerning the Netherlands, a Western European country that has had both a left-wing and a right-wing populist party in parliament for a substantial period of time.

We found that those who vote for the populist left and populist right do indeed report higher levels of opposition to trade openness than supporters of non-populist parties. Support for protectionism drives their vote, even if various other determinants of voting are taken into account. To answer the explanatory part of our question, it is relevant that left-wing and right-wing populist voters might oppose trade openness for different reasons. We have explored three explanations, revolving around: (1) economic concerns, because free trade increases economic inequality; (2) cultural concerns, because open borders affect the cultural and national order; and (3) political distrust, because free trade treaties result from opaque political processes.

At the level of political parties, it seems safe to assume that, in line with their ideological profile, the support for protectionism of leftist populist parties is part of their aversion to economic inequality (Otjes and Louwerse, 2013), while opposition to free trade among their rightist counterparts is in line with their ethnocentric and nationalist cultural agenda (Otjes and Louwerse, 2013; Vossen, 2011). In addition, the anti-establishment agenda that the two types of populist party have in common (Otjes and Louwerse, 2015) may fuel their support for protectionism if free-trade treaties are viewed as projects of political elites. Our analyses, which focus on the level of voters, demonstrate that the explanation for the protectionist views of populist constituencies differs from what one might expect on the basis of the ideological profile of the parties they vote for at the ballot box.

First, in line with studies scrutinising the relatively high support for protectionism among the less well educated (Hainmueller and Hiscox, 2006; Van der Waal and De Koster, 2015), economic egalitarianism proves to be only weakly related to opposition to free trade, and we found that support for protectionism is not part of the well-established link between economic egalitarianism and a preference for a leftist populist party. Even in times of economic crisis (our survey was conducted in 2012), protectionist tendencies among the populist electorate cannot be understood as reflecting economic concerns. This is especially remarkable if one considers that the standard indicator for support for protectionism that we used has strong economistic overtones. This finding adds to the expanding literature that demonstrates that various anti-globalisation opinions, ranging from anti-immigrant attitudes (Hainmueller and Hopkins, 2014; Van der Waal and Houtman, 2011) to welfare chauvinism (Van der Waal et al., 2010) and support for protectionism (Margalit, 2012; Van der Waal and De Koster, 2015), cannot be understood by taking into account the economic interests of the economically weak and their concomitant preference for economic redistribution. Our findings are therefore relevant for the expectation that the threat of ‘“neo-liberal” globalization [will reinvigorate] the “politics of security” advocated by left-wing populist parties’ (Grande, 2008: 340–341). This may well be true at the level of political parties, but the above indicates that it is unlikely that higher levels of international free trade increase the salience of economic redistribution issues among voters.

Second, in accordance with studies demonstrating a substantial relationship between support for protectionism and cultural concerns such as nationalism and ethnocentrism (e.g. Edwards, 2006; Kaltenthaler et al., 2004; Mansfield and Mutz, 2009; Mayda and Rodrik, 2005; Van der Waal and De Koster, 2015; Wolfe and Mendelsohn, 2005), support for protectionism among rightist populist electorates seems to be part of their cultural concerns, albeit moderately so: part of the relationship between ethnocentrism and a vote for a right-wing populist party is mediated by support for protectionism. This indicates that the cultural cleavage between the supporters of right-wing populist parties and new-leftist ones is more encompassing than commonly assumed. This not only concerns issues related to immigrants, ethnic minorities, and law and order, but also includes views on trade openness. Instead of reviving the economic redistribution agenda of leftist populist parties, the politicisation of trade openness is therefore more likely to broaden the scope of the cultural appeal of the populist right. Put differently, instead of reinvigorating the left-right dimension, the politicisation of trade openness will probably increase the salience of the GAL-TAN dimension (Hooghe et al., 2002) during elections.

Third, political distrust is higher among supporters of both types of populist party than among the electorate of non-populist parties (Hooghe et al., 2011; Hooghe and Oser, 2015; Kemmers et al., 2016), but support for protectionism is not part of this link between political distrust and voting behaviour. This indicates that resistance to ‘elite political projects’ does not explain the higher level of support for protectionism among populist constituencies.

Our most remarkable finding is that support for protectionism largely serves as an independent driver of voting for a populist party, that is, it is a determinant of voting in addition to both a wide range of socio-demographic factors and the already established attitudinal drivers of populist voting behaviour: economic egalitarianism, ethnocentrism and political distrust. Put differently, in addition to the economic and cultural concerns (Achterberg and Houtman, 2006; Van der Waal et al., 2007) and the anti-establishment attitudes (Hooghe and Oser, 2015; Kemmers et al., 2016) that pit populist constituencies against non-populist ones, opposition to trade openness matters when it comes to voting for a populist or non-populist party at the ballot box, at least in the present-day Netherlands. Our finding that views on free trade currently play a largely independent electoral role suggests that support for protectionism among the electorate is politicised and shaped by cues from populist party elites. As such, their emphasis on opposition to trade openness is an additional aspect of the way in which their agenda-setting weighs on the political field (compare Daenekindt et al., 2017; Minkenberg, 2001).

This finding has implications for electoral competition. The Dutch leftist populist party is in competition for the electorate with its right-wing counterpart on more grounds than previously assumed. Along with their shared political distrust (Hooghe et al., 2011; Hooghe and Oser, 2015; Kemmers et al., 2016) and Euroscepticism (compare De Vries and Edwards, 2009; Halikiopoulou et al., 2012; Van Elsas and van der Brug, 2014), they also have opposition to trade openness in common. During elections in which anti-establishment sentiments, the role of the EU and trade issues are salient, the SP and PVV are therefore likely to compete for the same segment of the electorate. This resonates with research on electoral volatility that reports ‘a continuous exchange from SP to PVV just below the 15 per cent rate’ until 2010 (Van der Meer et al., 2012).

Finally, our study inspires two questions for future research. First, more research is needed to assess whether agenda setting and active mobilisation by populist party elites is indeed an explanation for the link between support for protectionism and populist voting behaviour uncovered here. Second, a salient question concerns how our findings travel beyond the Dutch case, especially because the Dutch population at large is one of the most pro free trade in the world (Mayda and Rodrik, 2005). Support for protectionism might, therefore, be an even more prominent fault line among the electorate in other countries. Note, however, that a low average level of support for protectionism at the national level does not rule out the possibility that this issue is highly contested and has significant political consequences. Indeed, our study shows that notwithstanding the high number of pro free-trade citizens in the Netherlands, support for protectionism clearly divides the voters for leftist and rightist populist parties from those who vote for non-populist parties.