Introduction

On the 25th of August 2016, Geert Wilders, the leader of the Dutch populist radical right party the PVV, announced its election manifesto. The text was only 1 page long and focused mainly on anti-immigrant policy proposals. Yet number three of the 11 bullet points stated: “direct democracy: introducing binding referendums, citizens get more power” (PVV 2016). The PVV is not the only populist party advocating referendums. In his 2007 landmark study of populist radical right parties in Europe, Mudde (2007, p. 151) highlights that referendums are one of the “key features” of a populist democracy (see also Taggart 2000, pp. 103–105; Mudde and Kaltwasser 2013, p. 207). Populist parties often claim to be the saviours of democracy, and direct democracy is one means to save the people from the corrupt elites. There are few studies on whether populist parties push for more referendums.Footnote 1 Yet the number of referendums organized in democratic countries is clearly on the rise (Altman 2011; Qvortrup 2014).

If there is a general lack of research regarding referendums and populism from the supply side (i.e. party), this lacuna is even more pronounced on the demand side (i.e. citizen). Recent research has found that citizens can have (a higher or lower degree of) populist attitudes (Hawkins et al. 2012; Akkerman et al. 2014, 2017; Spruyt et al. 2016). Indeed, citizens can to a higher or lower degree believe that society is ultimately separated into two groups, the good people and the corrupt elite, and they can believe that politics should be an expression of the general will (cf. Mudde 2004). Looking at this definition, one can expect that more populist citizens will be more in favour of referendums. Yet so far, no research has been conducted on whether populist citizens are actually in favour of referenda,Footnote 2 whether they participate more in referendums than their less populist fellow citizens, and whether they are more likely to vote ‘No’. It is precise this topic that interests us in this study. Specifically, we are interested in whether populist attitudes influence citizens’ referendum preferences and practices. The research question guiding this paper is the following: Are populist attitudes related to voters’ referendum preferences and practices?

Based on the Dutch 2016 National Referendum Survey (NRO 2016; Centerdata 2016b), we find that indeed populist attitudes correlate highly with referendum preferences. Regression analyses also show that populists are not more (or less) likely to turn out to vote, but are overwhelmingly more likely to vote ‘No’, even after we control for campaign topics, trust in government or party preference for a populist party. These results have important consequences for our understanding of the role of referendums in a democracy. If a government decides to organize a referendum, but there is a large group of citizens with a high degree of populist attitudes, our results suggest it may well be that they will lose the referendum regardless of the topic of the referendum, the campaign or even general trust in government.

In this paper, we first discuss populist attitudes and their relationship with referendums. Afterwards, we describe the dataset, the operationalisation and the type of analysis we carried out. The results are presented in the subsequent section. Lastly, we conclude the paper by rounding up the results and looking forward.

Theoretical framework

Defining populism

Any paper on populism must first define the concept. In this study, we define populism as a thin-centred ideology (Mudde 2004). Even though other perspectives exist, the thin-centred approach has become increasingly dominant (Akkerman et al. 2014, 2017). The advantage of the thin-centred approach is threefold (Akkerman et al. 2014, 2017). First, the thin-centred ideological approach is able to travel, scholars have successfully employed this definition of populism in Europe and Latin American (see Mudde and Kaltwasser 2013). Second, the thin-centred ideology approach allows scholars to measure populism (Akkerman et al. 2014). To date this has occurred in speeches (Hawkins 2009), in newspapers (Rooduijn 2013) and in party platforms (Rooduijn 2013; Rooduijn and Pauwels 2011). Third, and perhaps most important for this paper, the thin-centred ideology approach allows social scientists to measure populism among citizens (see Hawkins et al. 2012; Akkerman et al. 2014; Spruyt et al. 2016).Footnote 3

In this study, we define populism as a “thin-centered 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” (Mudde 2007, p. 23).

Mudde’s definition of populism focuses on three key aspects of populism (Mudde 2004; see also Akkerman et al. 2014). First, defined as such populism is people centred. For populists ‘the people’, considered as one homogenous whole,Footnote 4 represent the focal point of democracy. Therefore, as noted above in Mudde’s definition, politics should be an expression of the voice of the people. Second, the people must be juxtaposed with the elite. It is the elite that threatens the people-centred notion of democracy. Third, the distinction between the people and the elite is framed as antagonistic (i.e. Manichean). Thus, the people are viewed as good and the elite as evil and corrupt (see Mudde 2004; Zaslove 2008; Akkerman et al. 2014; Rooduijn 2013; Hawkins 2009).

As mentioned earlier, one advantage of the thin-centred ideological approach to populism is that it is possible to measure populism. Since populism, from this perspective, is viewed as a set of ideas (ideology or worldview; see Mudde 2004; Hawkins 2009; Hawkins et al. 2012), we are able to measure populism among individuals (i.e. as an attitude). Building upon previous work on measuring populism at the individual level (Hawkins et al. 2012; Akkerman et al. 2014, 2017; Spruyt et al. 2016), we operationalize the thin-centred definition of populism, to capture the people centred, the anti-elite, and the antagonistic notion of populism (see below).

Populism and referendum preferences and practices

If we turn to the above definition of populism, it should come as no surprise that populists can be expected to favour direct forms of political representation, such as referendums. The reason for this is clear: referendums are a way to give power back to the people (Mudde 2007, p. 151). Referenda are a more direct means to represent the will of the people and can be used to challenge the corrupt elite. Referendums fit with each of the three key aspects of populism: they are people centred, reduce the power of the elite and are a means to keep the corrupt elite in check (at least to some extent) (Mudde 2007, p. 152).

So far, no empirical research has directly examined the relationship between populist attitudes and referendum preferences, but there are some indirect tests examining this relationship. These studies, for example, examine the relationship between voting for populist parties and referendum preferences (Pauwels 2014; Bowler et al. 2016). The first of these two studies indeed finds a link between referenda and populism: Pauwels’ study of populist parties in Europe (2014) notes that those who favour referenda are also more likely to vote for a populist party. At the same time, Bowler et al. (2016) find no such effect in their study of populist voters in Australia, Canada and New Zealand. This should not come as a surprise, as the electorate of populist parties is very diverse. While populist citizens are indeed more likely to vote for populist parties (Van Hauwaart and Van Kessel 2017), they are not the only ones who vote for them. Indeed, Akkerman et al. (2014, p. 343) found that populist parties not only attract the most populist citizens, but also elitists (who favour governments that lead rather than follow the people). Indeed, citizens can have different motivations for voting for populist parties and elitist citizens can e.g. be drawn to a populist radical right party’s authoritarian agenda of law and order. One can expect populist citizens to be very favourable of referendums, but elitist citizens can be expected to be highly unfavourable of referendums. Taken together both groups may cancel each other out resulting in an overall null effect.

It is therefore important to note that previous studies focus on voting for a populist party and not on populist attitudes per se. When one examines these attitudes directly, a stronger relationship can be expected.Footnote 5 Specifically, we expect that those with higher populist attitudes will be more supportive of referenda.

H1

The higher citizens score on populist attitudes, the more they will be in favour of referendums.

However, simply because a populist prefers more direct democracy, does not mean that this will automatically mean that populist citizens are more likely to turn out to vote or that they will vote differently than non-populists. Populist citizens often are less interested in politics, have lower levels of efficacy and they are often less satisfied with the workings of government (Spruyt et al. 2016). The reasons for this, it is often claimed, are based on the socio-economic characteristics of populists. They have lower levels of education and come from the lower socio-economic classes. In studies of elections, such characteristics are related to a lower voter turnout (Smets and Van Ham 2013). Immerzeel and Pickup (2015) indeed find that the presence of a populist party in a party system does not increase voter turnout in elections. They argue that supporters of populist parties are disengaged and they find that they are even less likely to vote (Immerzeel and Pickup (2015).

However, voting in an election is not the same as voting in referendums. Webb (2013) examines the relationship between stealth democracy (which he too quickly links with populism) and the willingness to vote in a referendum. Webb finds that stealth democrats are more likely to vote in referendums than in elections. This is, however, once again, an indirect test, focusing on stealth democrats rather than on populists. It is important to emphasize that stealth democrats are different from populists. Stealth democrats are disengaged (apolitical) individuals who support “delegation, efficiency, and expert input in the decision-making process” (Lavezzolo and Ramiro 2018, p. 4; Webb 2013). To be sure, stealth democrats may share anti-establishment sentiments with populists. Yet they primarily demonstrate a desire to solve the most pertinent political problems using objective (non-political) solutions (experts). Populists, on the other hand, revert to the common sense and authenticity of the people (while stealth democrats only see referendums as a gun behind the door, a measure of last resort) (Hibbing and Theiss-Morse 2002, p. 2). Unlike populists, stealth democrats are satisfied with independent experts, or with businessmen, to solve political problems (Webb 2013) and thus have much in common with what Caramani (2017) refers to as technocrats, or what Mudde (2004) refers to as elitists.

It, therefore, remains to be seen whether one finds similar results (as the ones found studying stealth democrats) when testing the relationship between populist attitudes and referendum voting directly. Looking at our earlier theoretical discussion, our definition of populism and our first hypothesis, the expectation is that that populist citizens are more in favour of referendums, while it is also likely that they will be inclined to participate in referendums. From this line of reasoning we would expect populists to be more likely to vote in referendums.

H2

The higher citizens score on populist attitudes, the higher the probability they vote in referendums.

A final question is, do we expect populists to vote no or to vote yes in the referendum. More often than not we expect a ‘No’-vote, given that a referendum is a chance to vote No against ‘the’ elite.Footnote 6 Surely the vote choice in a referendum also hinges upon the topic, the cues a party gives its supporters and the campaign more in general (Reidy and Suiter 2015). However, given that populist attitudes are rooted in a worldview, a firm belief that the elites are bad, we anticipate that more often than not, giving your voice in this context is thus also about voting against the elite. Hence, we can expect that populist attitudes will still play an important role even when we control for all the aforementioned factors. This leads to our final hypotheses:

H3

The higher citizens score on populist attitudes, the more likely they cast a ‘No’-vote in referendums.

Methods

In this section, we begin by discussing the Dutch 2016 referendum as it provides the context within which the referendum survey took place. Afterwards, we discuss the data and detail how we operationalized our main independent variable: populist attitudes. We finish by detailing how we operationalized the other variables and indicate which type of models we ran. More detailed descriptives about the variables can be found in Appendix 1.

The Dutch 2016 referendum

After decades of discussions and failed attempts, the Dutch political parties agreed in 2014 on a referendum law that would allow citizens to collect signatures to force the government to organize a referendum on a bill or treaty that was recently approved by the parliament. Such a referendum would then, if a majority of the voters voted ‘No’ and if a turnout quorum of 30% of the electorate was met, force the government to reconsider the bill or treaty. On the first of July 2015, the new law became active. Three Euroskeptic organizations immediately started a signature collection effort to force the government to hold a referendum on the association agreement between the EU and Ukraine—the first bill or treaty that in some way dealt with the EU. They managed to collect enough signatures and on 6 April 2016, the referendum was held on the question “Are you in favour or against the law to approve the association agreement between the EU and Ukraine?” Voters were allowed to cast a vote in favour (the equivalent of a ‘yes’-vote) or against the law (the equivalent of a ‘No’-vote).

The debate during the campaign for the referendum mainly centred on four topics: (I) European integration (e.g. future accession of Ukraine to the EU); (II) the position of the Dutch government and whether it was trustworthy (III) the position of Russia and whether or not the agreement would help to protect the Netherlands and lastly (IV) whether or not the turnout quorum would be reached (I&O Research 2016; Jacobs, Forthcoming). Especially the latter was important for yes-leaning voters. In the end, 32.3% of the voters showed up. The ‘No’-camp won handsomely: 61.0% ‘No’ versus 38.2% ‘Yes’.

At least two elements make this referendum a suitable case to carry out a first analyses of the association between populist attitudes and a preference for referendums. To begin with, like most European referendums, the Dutch 2016 referendum dealt with an EU-related topic (cf. Qvortrup 2014). This makes our findings more comparable to other European referendums. Additionally, the European dimension has the added value that populists are often considered to be Euroskeptics (Mudde 2007, p. 159). Rather than relying on populist attitudes, in referendums that deal with a EU-related topic, voters are likely to use existing information shortcuts, such as their attitude towards the EU. However, if we still find an effect of populist attitudes here after controlling for EU attitudes, this would represent strong support for the expectation that populist attitudes influence referendum voting.

It is likely that populist attitudes play less of a role in countries with a lot of referendums as citizens in these countries have more experience with referendums and probably have readily available information shortcuts that they have used in the past. For instance, when a fifth referendum on immigration is held, a citizen is likely to have well-formed opinion on the topic. Generic attitudes such as populist attitudes are less likely to play a role under such circumstances. However, Switzerland is quite the exceptional case: no other country has so much experience with referendums. Most countries, especially European ones, have some experience with referendums, but not a lot (Qvortrup 2014). In that sense, the Netherlands is a more representative case than Switzerland.Footnote 7

The dataset

We use the 2016 National referendum survey (NRO 2016). The questionnaire was presented to 2888 randomly selected members of the nationally representative LISS-panel.Footnote 8 The response rate was 87% (2525 respondents). The questionnaire was open from 7 to 26 April.

The main independent variable: populist attitudes

For the measurement of the respondents populist attitude we use the scale of Akkerman et al. (2014). This populist attitude scale consists of six items referring to the elements of Mudde’s definition of populism: (1) the sovereignty of the people; (2) the distinction between the pure people and the elite and (3) the idea that an antagonistic relationship exists between the people and the elite (often referred as Manichean). In Box 1, we show the six items of the populist attitudes scale. Respondents rated their agreement with each statement on a Likert scale, ranging from 1 (I very much disagree) to 5 (I very much agree). Using principled component analysis, we calculated the respondents factor scores.Footnote 9

Box 1 The six-item ‘Populist Attitude Scale’
Full size table

Other (in)dependent, control variables and methods of analysis

Regarding our first hypothesis, we calculate a set of simple bivariate correlations between our populist attitudes variable and a set of variables relating to referendum preferences. Specifically, we examine bivariate correlations with the extent to which a respondent agreed with the following statements: (a) voters should be allowed to vote in a referendum on some important national topics; (b) the results of referendums should be binding; (c) my local government should allow referendums about local issues and (d) referendums are too expensive. Respondents were asked to rate their agreement with these statements on a Likert scale ranging from 1 (I very much disagree) to 5 (I very much agree). We expect a positive and significant correlation with the first three statements, but a negative and significant correlation with the last one.Footnote 10

To test our second hypothesis, we carry out a logistic regression to estimate the effect of populist attitudes on turning out to vote (1) or not (0). As a set of controls, we include variables covering three clusters of control variables that are typically included in the (election) voter turnout literature.Footnote 11 This literature stresses the impact of socio-economic status, social integration, political attitudes and socio-demographics (Bühlmann and Freitag 2006; Jacobs and Spierings 2010; Smets and Van Ham 2013). (I) Regarding socio-economic status, we use education (6 categories) and income (net income in 12 categories). (II) We operationalize social integration—an individual’s integration in primary (family) and secondary groups (friends, neighbours) (Bühlmann and Freitag 2006, p. 17)—by using two control variables indicating (1) whether a respondent lives alone (0) or not (1) and (2) the degree to which (s)he lives in a rural environment (5 categories). (III) Political attitudes should be understood as evaluations of the political system (Bühlmann and Freitag 2006, p. 17). We include variables measuring trust in the national government (0–10) and trust in the EU (0–10), Russia (0–10) and Ukraine (0–10), as all four were referred to during the campaign. Lastly, regarding socio-demographics, we include age (7 categories) and gender (female–male). We also include a dummy variable for voting PVV and SP. Our prime objective is to test the extent to which populist attitudes predict referendum vote turnout. The focus is thus on political attitudes (i.e. the demand side). However, it is also possible that party choice, i.e. the fact that parties such as the PVV and the SP both advocate referenda, influences individual preferences. However, the direction of this relationship is unclear: do voters who favour referendums vote for populist parties because these also favour referendums or is it the other way around (cf. Rooduijn et al. 2016)? By controlling for party choice we cannot determine causality. However, if the populist attitude variable remains significant even while controlling for party choice, it is possible to assess whether populist attitudes have an effect independent of party choice.Footnote 12

To test our third hypothesis, we again carry out a logistic regression to estimate the effect of populist attitudes on voting against the elites in the referendum. In the case of the Dutch 2016 referendum, this is the No-option on the referendum-ballot (i.e. against the association agreement between the EU and Ukraine). The literature on voting behavior in referendums distinguishes between three clusters of explanatory variables (Reidy and Suiter 2015, pp. 137–138): (1) political attitudes related to (the topic of) the referendum, (2) campaign information and (3) socio-demographic factors. From the beginning, the referendum was framed as a referendum on European integration, the trustworthiness of Ukraine (and the degree to which it would be able to comply to the terms of the agreement), about the performance of the Dutch government and the influence of Russia in the region (I&O Research 2016). (I) These political attitudes are once again operationalized by including the aforementioned four trust variables (trust in government, trust in the EU, Ukraine and Russia), all measured on a scale ranging from 0 (no trust at all) to 10 (fully trust). As an extra control, we also added two dummies measuring party preference for one of the populist parties (PVV and SP). The reasons for doing this have been explained above. (II) We measure exposure to campaign information by looking at how frequently the respondent read, heard or saw information about the referendum in the past few weeks (5 ascending categories). (III) Regarding socio-demographic variables, we include education (6 categories), the degree to which she lives in a rural environment (5 categories),Footnote 13 age (7 categories) and gender (female–male). The descriptives of all the aforementioned variables can be found in Appendix 1.

Results

In this section, we first present an overview of the degree of populism among our respondents. Subsequently, we analyse to what extent populist attitudes are correlated with general referendum preferences. We then move to the more explanatory part of our analysis. We examine to what extent populist attitudes have an effect on the decision to turn out to vote and finish this section with a similar analysis about the decision to vote ‘No’ in the 2016 referendum.

Descriptives: populist attitudes

We start by providing a brief overview of the distribution of populist attitudes in our sample. Are there many populists or just a few? To begin with, Table 1 shows the results of the measurement of populist attitude per item of the scale.

Table 1 Descriptives of populist attitudes.
Full size table

While the factor scores we use in the regression models do more justice to the weight of individual items, they are less easy to interpret. For the description of the populist attitude variable here we therefore use the simple sum of the scores of the populism items divided by 6, as this is more straightforward to interpret. The scores of our respondents are normally distributed with a mode of 3, a mean of 3.49, the median at 3.34. This slight discrepancy between the median and the mean is due to the high frequency on the right extreme value (= 5) on the populist attitude scale. Almost 25% of the respondents scores 4 or higher on the populist attitude scale. In short, there is quite a substantial group of highly populist citizens in the Netherlands.

Are populist people more likely to support referendum provisions?

In this section, we examine to what extent a higher degree of populism among citizens is correlated with referendum preferences. Here, we show descriptive analysis because the patterns they lay bare are so strong that the tables and figures speak for themselves. The full multivariate regression analyses are provided in Appendix 2. In each of these four regressions, populist attitudes had a significant effect in the expected direction (p < 0.001).

It turns out that citizens who have higher levels of populist attitudes indeed favour referendums more than citizens with lower levels of populist attitudes. Indeed, the more a citizen exhibits populist attitudes the more (s)he is in favour of holding referendums on national and local topics and the more (s)he favours binding rather than non-binding referendums—the strongest correlation of the four variables in Table 2. The latter does not come as a surprise: populist citizens are anti-elite, so measures that curb the freedom of elites to ignore referendum outcomes are consistent with this anti-elitism (cf. supra). Most interestingly, it turns out that populist attitudes are also (negatively) correlated with the idea that referendums are too expensive. Comparatively speaking, populists are more willing to spend money on organizing referendums.

Table 2 Correlation matrix: populist attitudes and referendum preferences.
Full size table

To explore these findings in more detail, we also provide a set of bar charts depicting the mean of each of the four variables (Figs. 1, 2, 3, 4). For this specific analysis—and presentational purposes—we divided the populist attitudes scale in its four quartiles.Footnote 14 In line with Table 2, the results are very clear: in all four figures, the mean scores increase/decrease linearly and the confidence intervals rarely overlap between two consecutive quartiles. For instance, while the 25% least populist respondents on average do not want national referendums (mean: 2.63, scale: 1–5), the 25% most populist respondents are clear proponents of national referendums (mean: 4.21). The results for the fourth variable (referendums are too expensive) show the flattest slope, but even here relationship appears to be linear and it remains substantial: the 25% least populist respondents on average clearly agree more with the statement (mean: 3.85) than do the 25% most populist respondents who in essence believe the price of referendums is about right (mean: 3.32). A robustness check using multivariate regressions shows the same (see Appendix 2): populist attitudes have a strong and significant effect on referendum preferences.

Fig. 1
figure 1

National referendum preference per populism quartile

Full size image
Fig. 2
figure 2

Referendums should be binding preference per populism quartile

Full size image
Fig. 3
figure 3

Source NRO (2016)

Local referendum preference per populism quartile. Note 95% confidence intervals; Scales run from 1 to 5, with 3 being the cut-off point.

Full size image
Fig. 4
figure 4

Referendums too expensive per populism quartile

Full size image

In sum, our first hypothesis stating that those with higher levels of populist attitudes are more in favour of referendums seems to be corroborated.Footnote 15

Are populists more likely to turn out to vote?

So far, we have analysed the association between referendum preferences and the levels of populist attitudes of citizens. What about the process of voting in a referendum? Are populist citizens more likely to turn out to vote in referendums?

At first sight, citizens with higher levels of populist attitudes seem to vote more. For instance, of the quartile respondents that is the least populist 57.9% said they voted, while 61.9% of the quartile most populist individuals said they voted (see Appendix 3).Footnote 16 However, it appears that these differences are artefacts of the impact of other variables. Our multivariate logistic regression analysis (Table 3) shows that after controlling for the classic voter turnout control variables, populist attitudes do not have a significant effect: citizens with higher levels of populist attitudes are not more likely to turn out to vote. It needs to be added that neither are they less likely to turn out to vote. The unstandardized coefficient is positive, but not significant. In a way this is by itself already remarkable. Most of the literature sees populist citizens as ‘losers of globalization’, dissatisfied with politics (cf. supra), typically the type of citizens who are more likely to stay at home. As noted, other studies noted that voters for populist radical right parties were less likely to turn out to vote in elections (Immerzeel and Pickup 2015). We find that this does not hold for referendums: Table 3 shows they are just as likely to vote as citizens with lower levels of populist attitudes.Footnote 17,Footnote 18

Table 3 Logistic regression analysis turning out to vote
Full size table

Which of the other variables matter most? In the methods section, we highlighted that little is known about what drives voter turnout in referendums (but see: Schuck and De Vreese 2009). To accommodate this, we used the clusters of factors that determine voter turnout decisions in elections. In line with Schuck and De Vreese, we find that age has a positive effect on voter turnout decisions: young voters were more likely to stay at home. A variable that almost shows a significant effect (p < 0.1) is education: higher educated citizens were more likely to vote.Footnote 19 Similarly, political attitudes and vote intention for the populist radical right PVV seem to matter as well. Thus, turning out to vote for the referendum can be in part explained by the supply-side mobilization of the party.

To summarize this paragraph, hypothesis 2 stating that the higher citizens score on populist attitudes, the higher the probability they vote in referendums is not supported by our analysis.

Are populist people more likely to vote ‘No’?

As mentioned in the previous section, a fair amount of populist citizens voted in the Dutch 2016 referendum—not more or less than their less populist counterparts. This begs the question: do citizens with higher levels of populist attitudes vote differently than those with lower levels of populist attitudes? Are they more likely to vote ‘No’? And if so, is this a genuine effect or rather an artefact of other factors such as the party they sympathize with?

At first glance, populists clearly seem more likely to vote ‘No’: merely 18.2% of the least populist quartile said they voted ‘No’, while a whopping 89.5% of the most populist quartile said so (see Appendix 3). This does not come as a surprise: populists are against the elites, so they are more likely to vote against them as well (cf. supra). A more interesting question is to what extent populist attitudes still matter after controlling for other factors such as political attitudes, campaign exposure and socio-demographic factors. We find that populist attitudes indeed are significantly related to voting against the establishment after controlling for these and other variables (see Table 4).Footnote 20,Footnote 21

Table 4 Logistic regression analysis voting ‘No’ in the referendum
Full size table

Regarding the control variables, most of them perform as expected: the main argument of the campaign seems to have landed with the electorate and perform as expected: voters who trusted the EU and Ukraine were far less likely to vote ‘No’. Furthermore, people who intended to vote for the two main populist parties, the socialist SP and the populist radical right PVV, were more likely to vote ‘No’, which is in line with the position of these parties during the campaign. What is most striking through is that populist attitudes still have a significant effect, even after controlling for these party preferences. Thus, voting ‘No’ and scoring higher on the populist scale increases the likelihood independent of the supply-side message of the two populist parties. This is another indication of the explanatory power of populist attitudes beyond mere party preference and it suggests that it is important to test claims about populist attitudes and direct democracy directly rather than in an indirect way by looking only at voters of populist parties. Lastly, both the higher educated and the elderly were less likely to vote ‘No’.

Turning to our hypothesis 3, stating that the higher citizens score on populist attitudes, the more likely they cast an anti-elite vote in referendums, the evidence indicates that this is indeed the case: even when controlling for political attitudes, campaign information and classic socio-demographic factors, populist attitudes have a significant and substantial positive effect on the likelihood to choose the anti-elite vote option in a referendum. Whether this holds for all referenda, is another question. As noted in the theory, theme and context can make a difference. Thus, a referendum proposed by a populist party on a different topic may yield a different result.

Conclusion and discussion

Summary

While populist parties are typically in favour of referendums, we know very little about to what extent populist citizens also favour referenda. The research question of this paper is:

Are populist attitudes related to voters’ referendum preferences and practices?

We answered this question in three steps. First, it seems that citizens with a higher degree of populist attitudes are more likely to favour referendums: they support local and national referendums, they are supportive of the idea that they should be binding and they are less likely to find referendums too expensive. Second, while such citizens can be expected to vote less in general elections, they are not more or less likely to vote in referendums. Third, the more populist a citizen is, the more likely (s)he is to vote ‘No’, regardless of their trust in government, regardless of the topic of the referendum, regardless of socio-demographic factors and regardless of campaign factors. Whether this also holds for other sorts of referenda, further research is necessary.

These findings are in line with the limited research on populism and referenda, while they also provide new insights. As Mudde argues (2007), populists have a people centred, direct and unmediated understanding of democracy, one that is favourable to referenda. Thus, we should also expect that populist citizens should favour referenda. This is indeed what we find.

Studies on vote choice, on the other hand, are mixed. Pauwels shows that those who support populist radical right parties are more likely to support referenda, while Bowler et al. (2016) show that this is not the case. Given that our dependent variable is populists attitudes, we contend that ours is a more pure test. The results, however, beg the question: what are the implications for the link between referenda and party choice? For one we know that, not only populists vote for populist parties. Elitists may vote for a populist party, or voters may vote for a populist party since they agree with the party’s anti-immigrant ideology. Thus, voters for populist parties may also oppose referenda. However, it is also possible that the three parties studied by Bowler et al. (2016) are less populist (people centred) than is assumed by the authors. If this is the case, this would explain why in their study voters were less supportive of referenda. In any case, this points to two important points. First, it highlights the strength of our test, i.e. testing populist attitudes and not vote choice. And second, it points to the need for further research and more cases.

Implications for theory and society

The findings from this paper have important implications. For populism researchers, they suggest that while referendums are often neglected in the populism literature, they may be important once we move away from populist parties, to populist attitudes among citizens. Populist parties may not implement demands for referendums once they are in government (Jacobs 2011). However, this is not the case for populist citizens. If they are given the chance to vote in a referendum, it appears that those with higher populist attitudes are not less likely to vote than not populist. This can have real consequences for referendum results. For referendum researchers, our research suggests not only classic political attitudes, such as trust in government may matter, but that populist attitudes should be included in the standard battery of variables used to explain (especially) referendum outcomes. For governments organizing a referendum the results are not hopeful: regardless of the government’s popularity or campaign effort, the more populist citizens there are the more likely there will be a ‘No’ vote. In countries where a large portion of the voters is populist, it may imply that governments will find it difficult to win a referendum. For democratic reformers the results are also sobering. In some countries, democratic reformers have suggested more referendums are a means to reduce citizen dissatisfaction with the way democracy works (Bowler and Donovan 2013). Whether citizens indeed become less dissatisfied is something we did not examine. However, populist citizens’ higher likelihood to vote ‘No’ makes an overall ‘No’-result more likely. This creates a difficult situation for the government, i.e. it may force governments to change course and in the process this may simply further feed into dissatisfaction. The more referendums are held, the more the government’s flexibility is tested. This becomes all the more pressing, as populists are not the only actors who favour referendums. Postmaterialists also demand referenda, pacing the mainstream parties even more in a difficult situation (Dalton et al. 2001).

Future research

These implications also suggest that more research is needed. First of all, if populist attitudes matter so much in referendums we need more insight into which factors determine the degree of populist attitudes of individual citizens. Are these attitudes stable? Can they be reversed? And which factors influence them in general?

Second, our results need to be replicated in other referendum cases and in other countries. Most countries, especially European ones, have some experience with referendums, but not a lot (Qvortrup 2014). In that sense, the Netherlands is a more representative case than say Switzerland, but clearly replication of our research in other countries is needed. Furthermore, populist attitudes may play a smaller role in referendums that deal with tangible, salient topics that touch the everyday lives of citizens. It could also be that they play a different role in political systems where citizens have more opportunities to influence the decision-making process. However, it must be said that this would imply that populist attitudes probably play an even larger role in less open political systems. After all, the Netherlands is a relatively open political system where smaller parties have a high chance to gain representation and citizens have a fair amount of opportunities to influence the decision-making process. Lastly, it could be that populist voters become frustrated when governments ignore ‘No’-outcomes. After all this was only the second time a referendum was held in the Netherlands. Hence, depending on the reaction to these referendums, populist citizens may become more inclined to stay at home, especially when referendums become a frequently used feature of the politics of that given country.

Our results cannot be definitive, but they do suggest that whatever their role, it is clear that populist attitudes deserve a more prominent role in referendum research and should be included in the standard toolkit of referendum researchers.