Introduction
‘Whatever happened to the Italian Five Star Movement?’ (Bull, Reference Bull2022) was one of the conclusions of the Italian elections in 2022. This ‘valence populist’ party rode a wave of protest and distrust beyond associations with the left and right (Angelucci and Vittori, Reference Angelucci and Vittori2021), attracting voters disillusioned by traditional (mainly leftist) parties and those previously not interested in politics (Natale, Reference Natale2014). Yet within 18 months of its second consecutive general election (March 2018) as Italy's largest party, the 5-Star Movement (M5S) had dropped to polling 4th, halving its 30%+ vote-share, behind parties of both the right and the left. In those 18-months, the party participated in two governments, one with a party on the right and one with a party on the centre-left. While indeed incumbents have been shown to typically lose votes in subsequent elections, this penalty is usually around 2% (Rose and Mackie, Reference Rose, Mackie, Daadler and Mair1983), not the nearly 17% polling drop faced by M5S.
To what extent can this drop be associated with the phenomenon of perceived ideological changes and to what extent can the drop be associated with assessments of government performance? Once populist parties enter government, their own rhetoric changes (Venizelos Reference Venizelos2023) as do the levels of populism expressed by their constituents (Luo Reference Luo2024; Aytaç, Çarkoğlu & Elçi Reference Aytaç, Çarkoğlu and Elçi2021). Such reactions might lead voters to decrease their support for such parties. Alternatively, the pressures of incumbency, such as responsibility for the economy and the economic vote (Lewis-Beck and Stegmaier, Reference Lewis-Beck and Stegmaier2000) or being associated with policies that supporters dislike might lead voters to identify with non-government parties (Spoon & Klüver Reference Spoon and Klüver2019).
This paper joins other case studies in the field of examining intra-election changes in party preferences. Such changes have been attributed to income, education, and ideology in the Netherlands (Van Dermeer et al., Reference Van Dermeer2015), political sophistication and household-context in Switzerland (Kuhn, Reference Kuhn2009), and spousal partisan attachment and political interest in West Germany (Schmitt‐Beck et al., Reference Schmitt‐Beck, Weick and Christoph2006). These studies use multi-wave panel surveys to observe characteristics of any overall intention of changing party preferences. Here we seek something more specific: what individual attributes make one more or less likely to switch preferences away from a specific party – the M5S noted above.
Two possible avenues of intra-election volatility are either deserting (Van Dermeer et al., Reference Van Dermeer2015) a party by switching preferences to another or disengaging (Schmitt‐Beck et al., Reference Schmitt‐Beck, Weick and Christoph2006; Kuhn, Reference Kuhn2009) by switching between having a preference and not having a preference. We argue that given the populist nature of the M5S, different sources of discontent would lead down these pathways. Voters who previously viewed the party as beyond left-and-right or ideologically similar to themselves might return to being disengaged. Alternatively, voters who became dissatisfied with the party due to its performance in government would desert the party for others.
Using panel survey data of 326 respondents from March 2018 through September 2020, this manuscript identifies differences between those who remained loyal to the party, those who deserted the party, and those who became disengaged with all parties. Descriptive analysis will first call attention to differences between these groups and the full panel sample. Multivariate analysis will then identify to what extent dissatisfaction with incumbency and the party ideologically can predict changes in party preference. A discussion section then briefly compares the case of the M5S with other valence populist parties. Results of this analysis have implications for populist parties in power. Evidence suggests that once in power, voters view populists as they do any other party, and make their party choices based on evaluations of party leadership and governance. In a sense then, this ‘valence populist’ party becomes a political actor like any other, perhaps further burdened by no ideological basis on which to draw support.
Rise of valence populists
In reviewing the literature on populist parties, Zulianello (Reference Zulianello2020) noticed that the ideational approach of populism faced a lacuna. Classic definitions of populist parties are those that (a) emphasize a moral and Manichean contraposition between ‘pure people’ and ‘corrupt elite’ and (b) believe that politics should essentially secure the enactment of the will of the people (Mudde, Reference Mudde2004, p. 543). Given this ‘thin ideology,’ scholars have noted that this could be found on both the left – an inclusive version – and the right – an exclusive version (Mudde and Rovira Kaltwasser, Reference Mudde and Rovira Kaltwasser2013). Along these lines, studies of (European) voting behaviour have identified that voters for left populists have strong egalitarian and altruistic values and strongly support government redistribution (Rooduijn et al., Reference Rooduijn2017), while voters for right populist are ideologically right-leaning (Rooduijn, Reference Rooduijn2018), nationalistic, euroskpetic, perceive migrants as threatening, and view themselves as socially distant to Muslims (Lubbers and Coenders, Reference Lubbers and Coenders2017).
Yet Zulianello (Reference Zulianello2020) identified a third typeFootnote 1 of populist party that typically receive 15% vote share in the countries where present (Zulianello and Larsen, Reference Zulianello and Larsen2021), though could not be classified as right or left populist. These ‘valence’ populists (a) lack a positional character and, instead, (b) focus on valence issues (Stokes, Reference Stokes and Kavanagh1992) such as corruption and good governance (Zulianello, Reference Zulianello2020) offering up neither inclusionary nor exclusionary aspects. Further research identified that such parties do indeed take up deliberately blurry positions on both economic and socio-cultural dimensions of ideological competition, instead focusing heavily on anti-corruption appeals (Zulianello and Larsen, Reference Zulianello and Larsen2024).
M5S rise to power as a valence populist
The source of success of the Italian M5S has been its ability to occupy the space of radical protest against ‘old politics’ (Mény and Surel, Reference Mény and Surel2002). This protest included expressing discontent with élites, a furious criticism against the economic performance of the incumbents, and a broader socio-political systemic discontent. A significant number of its electoral base came from those previously unaligned (Natale, Reference Natale2014) with the parties that characterised Italian political space post-1994. These disparate aspects have been kept together thanks to a semi-charismatic leadership that has emphasized M5S's non-political and non-ideological bases (Passarelli and Tuorto, Reference Passarelli and Tuorto2018). The party has rejected being labelled as ‘left’ or ‘right,’ noting these concepts are outdated as they claim that parties of the left and right are indistinguishable (Vittori, Reference Vittori2020). Both the party and its voters had been seen as ‘multi-ideological’ or ‘post-ideological,’ for when taking positions on policies, some are more similar with left-wing parties (such as basic income, public services, and rights for new family models) while others are more similar with right-wing parties (such as immigration and EU integration) (Ceccarini and Bordignon, Reference Ceccarini and Bordignon2018; Passarelli and Tuorto, Reference Passarelli and Tuorto2018).
While the ‘five stars’ of the party's early platform focused on water, environment, mobility, development, and energy (Passarelli et al., Reference Passarelli, Tronconi, Tuorto, Corbetta and Gualmini2013), its electoral appeal stressed the relevance of short-term factors such as the lack of competence in existing candidates rather than a particular ideology, which fits well with Mudde's (Reference Mudde2007) classification of populist parties as having only a ‘thin ideology’ at best. Among the classifications, given its lack of clear left-right ideology and focus on valence issues like fighting corruption, increased transparency, and democratic reform, M5S has been classified as a valence populist party (Zulianello, Reference Zulianello2020).
The party rode a wave of anti-party sentiment in public opinion in the February 2013 elections to become the largest party by vote share (25.6%) and repeated this level of success in the March 2018 elections becoming the largest party by vote (32.7%) and seat share (Bergman et al., Reference Bergman, Passarelli and Serricchio2020). In 2013, its unique position opposed to EU-imposed austerity measures while taking moderate stances on immigration and tax-and-spend policies placed it on the outskirts of the political space. By 2018, however, the party had moderated its position on the EU and took a more negative stance towards immigration, along with most other Italian parties at the time in response to the 2015 ‘refugee crisis.’ Other parties had also shifted in the anti-EU direction given that the EU had imposed asylum and immigration policies on its members. This resulted in M5S being placed more towards the centre of political space and a potential coalition partner (Giannetti et al., Reference Giannetti, Umansky and Sened2024).
Subsequently, the party formed a populist coalition government (Conte I) with the Lega, a party more commonly associated with the populist radical right party grouping. When that government collapsed in 2019, the M5S then formed a government with the main centre-left Democratic Party (Conte II). It was around this time that regional elections began to demonstrate falling support of the M5S (Valbruzzi and Vignati, Reference Valbruzzi and Vignati2021), its parliamentarians began to exhibit reduced legislative activity (Bromo et al., Reference Bromo, Gambacciani and Improta2025), and the party leadership changed along with an organizational renewal (Crulli, Reference Crulli2023).
Ideological perception as a source of valence party decline
Having served in two governments, it would be difficult for a party to maintain its non-ideological character. Citizens perceive parties to be ideologically close to parties that they govern with at the national level (Fortunato and Stevenson, Reference Fortunato and Stevenson2013). In the situation above, there were two separate coalitions (before 2021) for which citizens might have altered their ideological perception of M5S. Furthermore, evidence suggests that voters ‘learn’ about party's policy shifts during the process of governing (Seeberg et al., Reference Seeberg, Slothuus and Stubager2017) or through the ministries it controls (Hjermitslev and Krauss, Reference Hjermitslev and Krauss2024). Perceptions can also form based on the welfare policies inherited or implemented (Adams et al., Reference Adams, Bernardi and Wlezien2020). A possible explanation then, for the reduced support for the M5S could then be challenges to its ideological character: having lost its ability to differentiate itself as beyond left-and-right politics (Vittori, Reference Vittori2020).
H1a: M5S voters who supported the party because of its unique beyond left-right dimension, but then associated the party with the left-right dimension are more likely to disengage from the M5S
Spatial theories of voting suggest that voters choose parties that are ideologically close to them (Downs, Reference Downs1957). Being neither more left nor right led had led voters to vote for M5S in the 2013 election (Passarelli and Tuorto, Reference Passarelli and Tuorto2018), which suggests one possibility for the party's appeal was its centrist nature. That said, should ideologically minded voters have shifted perceptions given the possibilities noted above, one cannot discount the fact that such ideological shifts might move a target party away from a voter's ideological positions. Should a party move too far away, then voters would be more likely to identify with a party that is more ideologically closer (Ferland and Dassonneville, Reference Ferland and Dassonneville2021). Another option present in research suggests that individuals who lack ideological congruence with parties are less likely to turnout (Lefkofridi et al., Reference Lefkofridi, Giger and Gallego2014). In this case of party preferences, they might not have an attachment to any party.
H1b: M5S voters who perceive that the distance between them and the M5S has increased are more likely to disengage from the M5S
Penalties of incumbency and party decline
Instead of relying upon specific cues of a party's ideology or raison d'être, especially in a context far from an election, a voter might simply rely upon a heuristic of asking herself if she is better off since the previous election in determining a level of support for an incumbent government (Key, Reference Key1966). The Italian case was an early one in which researchers identified that economic performance was an important factor in voting behaviour (Bellucci, Reference Bellucci1984), specifically in that poor economic performance favoured the opposition (Lewis-Beck and Bellucci, Reference Lewis-Beck and Bellucci1982). One's confidence in the governing elite could thus be affected by one's subjective economic assessments of the current and future state of the economy (Lewis-Beck and Stegmaier, Reference Lewis-Beck and Stegmaier2000). Italians have held governments accountable at the time of election for subjective economic perceptions (Bellucci, Reference Bellucci2012) and previous evidence suggests that Italian voters have also held the government economically accountable between elections at the time of referenda (Leininger, Reference Leininger2019; Bergman, Reference Bergman2020). In the 2013 and 2018 elections, regions with declining incomes for low-income workers saw the highest levels of support for the M5S (Bloise et al., Reference Bloise, Chironi and Pianta2021) - then in opposition. In the 2013 election, economic blame attribution had a significant impact on vote choice (Bellucci, Reference Bellucci2014). As the XVIII parliament (2018–2022) was the first in which the M5S participated in cabinet, they might now be affected by this economic judgement on their incumbency.
H2a: M5S voters with more negative economic perceptions are more likely to desert the M5S
A key driver of both positive and negative voting behaviour is the accountability that voters place in governments as it relates their performance and handling of the nations’ problems. From 1994 onward, policy performance has been a key driver of Italian voting behaviour (Bellucci, Reference Bellucci2007). Displeasure with previous governments was one of the primary drivers of M5S support in the first place (Passarelli and Tuorto, Reference Passarelli and Tuorto2018). With increased voter volatility and the rise of new parties, voter attachment to parties is attitudinal, in that is it influenced more by valence characteristics, like good government performance, as opposed to social cleavages (Bellucci, Reference Bellucci2012). Thus, a party seen as being competent (even above ideological reasons) has an independent effect on vote choice (Bellucci, Reference Bellucci2006). As the XVIII parliament (2018–2022) was the first in which the M5S participated in cabinet, they might now be affected by this competence or favourability towards their incumbent governance.
H2b: M5S voters that become dissatisfied with M5S governance are more likely to desert the M5S
Leadership evaluations have been considered a ‘hard test’ of voting behaviour (Mauerer et al., Reference Mauerer, Thurner and Debus2015). Some consider them to be the most important non-policy related concern in theories of party competition (Adams et al., Reference Adams, Merrill and Grofman2005). Empirically, they have been found to be key to party attachment across Europe (Garzia, Reference Garzia2013), especially in Italy (Garzia, Reference Garzia2014; Bellucci et al., Reference Bellucci, Garzia and Lewis-Beck2015) during the Second Republic (Bellucci, Reference Bellucci2007) up to and including the 2022 elections (Puleo et al., Reference Puleo, Carteny and Piccolino2025). They determine not only which pre-electoral coalition a voter selects, but also the specific party within that coalition (Bergman, Reference Bergman2021). Just as citizens become aware of government actions and their subsequent approval via television and media (Bellucci and De Angelis, Reference Bellucci and De Angelis2013), so too do they become exposed to party leaders through various media channels (Barisione, Reference Barisione2014); a favourable portrayal can lead to feelings of support. With the increased reliance of internet usage (over television), M5S leader support in particular has an advantage, which then translates into affirmative voting behaviour (Garzia, Reference Garzia2017). Should the evaluation of the party leader change throughout the governing term, so too might one's support for the party.
H2c: M5S voters that become dissatisfied with M5S leadership are more likely to desert the M5S
Data and methods
To compare what leads voters to still identify with the M5S or have altered their partisan preference, we employ panel data collected in March 2018 (for the National Parliamentary Elections) and September 2020 (intra-term) by ITANES (Italian National Election Study) via computer assisted online survey CAWI. Of key interest is comparing those who continue to prefer M5S during this period of two consecutive M5S governments. Loyalist respondents are coded if they identified M5S as the party they felt closest to and best represents them in both 2018 and 2020 waves.Footnote 2 Deserters are coded if they identified with M5S in 2018 but identified with another in 2020. Disengaged are coded if they identified with M5S in 2018 but identified with no party in 2020.
Hierarchical fixed-effect logistic regression models are used to take account for unobservable differences between regions that might also influence electoral volatility (Valbruzzi and Vignati, Reference Valbruzzi and Vignati2021) or the propensity to support the M5S, such as economic inequality, institutional performance (Albertazzi and Zulianello, Reference Albertazzi and Zulianello2021), or regional governing arrangements. Errors are also clustered at the regional level to account for heteroskedasticity that might arise from the aforementioned regional differences. Models also control for age-group, gender, completing tertiary education, political interest,Footnote 3 and linear and squared perceived left-right position of M5S.Footnote 4 Table 1 provides the operationalization of key variables related to the aforementioned hypotheses. As dependent variable is dynamic in nature, so too, are each of the variables associated with hypotheses.
Coding of key independent variables
Results
First, we can examine the distribution of the dependent variable as depicted in Figure 1. Of the 326 respondents who identified with M5S in the 2018 wave that provided answers in the 2020 waveFootnote 5, 51.84% remained as loyalists (n = 169), 29.2% deserted to another party (n = 95), and 19% became disengaged (n = 62). Just under 10% deserted to parties associated with the left while just under 15% deserted to parties associated with the right. Less than 1% deserted to centrist parties (Italia Viva and Azione). As a reminder, at this stage, the M5S had already governed with Lega (of the right) and was currently governing with the Partito Democratico (of the left).
Party identification of 2018 M5S voters.
Next we can examine descriptive differences between the three groups of interest (loyalists, deserted, disengaged) on the key independent variables in Table 2.Footnote 6 A fourth column of data indicates whether these means differ based on an ANOVA test and a fifth column provides data on the full sample to inform the reader about how those who preferred M5S in 2018 differ from the population at large.
M5S party preference difference 2018–2020 on selected characteristics
First we see that only 5% of sample were unable to place M5S on the left-right spectrum in 2018. Both loyalists and disengaged share this value as well. Interestingly, the fewest amount, 3%, can be found in the deserter subgroup, which would suggest that those who eventually left, might have always been able to place parties on the left-right spectrum, though the differences between these groups are not statistically significant. However, we do see a statistically significant (beyond 0.05) difference between the subgroups on what percent were unable to place M5S in 2020. Again, deserters are the lowest category, with only 8%, still suggesting that deserters might have steadfastly been focused on the left-right positions of parties. Contrast this with those who disengaged, of whom 23% were unable to place the party on the left-right spectrum. This is result is counter to the argument made above, which argued that those who believed that M5S joined the ideological spectrum were likely to disengage. The opposite is found, whereby those who were unable to place this party were also likely to feel disconnected from all parties. Of those identifying with M5S in 2018 and loyally continuing to do so, 11% could not place them on the left-right spectrum, which is similar to the overall sample of respondents. Of note is that both those who identified with M5S in 2018 and in the full sample of respondents find it more difficult to place M5S in 2020 than in 2018. Counter to previous literature (Fortunato and Stevenson, Reference Fortunato and Stevenson2013; Adams et al., Reference Adams, Bernardi and Wlezien2020) that suggests governing provides left-right cues about parties, it would seem that Italians might have gotten confused given the variety of coalitions in which M5S participated.
Next, in examining whether voters felt closer or further from the M5S in 2018 or 2020, on average respondents in 2020 felt closer (negative values mean closer in 2020 than 2018) than in 2018. For those who preferred M5S in 2018, we can see this value to be even higher, with these voters feeling on average about 2 points closer to M5S on a 0–10 scale. Expectedly, significantly so, loyalists felt closer to the party in 2020 as compared to 2018 by nearly 3 points. Those who deserted the party for others were the subgroup that felt that shifts brought the party closer to them the least, by less than 1 point, which is half of the full population sample.
We can also see statistically significant subgroup differences in each of the variables associated with incumbent governance and in the expected directions. On negative views towards the economy, with the most negative being 3 and post positive being -3.5, we can see that loyalists have the most positive view at 0.17, which is more optimistic than deserters at 0.49, and especially disengaged at 0.66. On views of whether government performance has increased or decreased (ranging from -10 to 10) during the period in which M5S is in power, we see that Loyalists hold the most positive view at 4.62, while deserters only see an improvement of 1.58, and in the middle are the disengaged at 2.18. As expected within the full sample that includes much of the opposition supporters, this group is least positive at only 0.86. We can briefly note, however, that while for the economy, disengaged were most negative of the subgroups that no longer prefer M5S, for overall government policy performance, it is the deserters who are most negative. Finally, on changes to approval of the M5S party leadership (same -10 to 10 scale), each subgroup and the full sample appear to hold a more negative view of Di Maio once in power. Loyalists have decreased, on average, at 1.39, which is more of a decrease than the sample average of -1.10. Between the two groups that left the party, it is the deserters for other parties that had the strongest change of view with -4.51, on average, when compared to the -3.61 change of the disengaged.
Looking at these means overall, it was expected that those viewing issues related to incumbency most negatively would be the ones likely to switch to other parties, which at the level of bivariate relationship appears to be supported in the variables of Government Performance and opinion of party leadership. Multivariate analysis will now be used to estimate whether these patterns hold with the introduction of control variables and to estimate effect sizes.
Table 3 presents multinomial logistic results. The columns of Deserter and Disengaged indicate whether a given independent variable, controlling for the others, has an independent effect on a respondent being more or less likely to be in that category than a Loyalist. Take for example, the coefficient of female, which is negative and significant at the 0.1 level in the deserter column. This indicates that females are less likely to be deserters than they are to remain loyalists. Logistic regression coefficients are difficult to interpret, and so substantive effects will be discussed after statistical significance.
Multinomial logistic regression output for M5S loyalists, deserters, or disengaged
Loyalist as base category. Clustered errors by region in parentheses; fixed-effect regional coefficients omitted.
*** p < 0.01, **p < 0.05, *p <0.1.
Two separate models were run. Given that H1a includes those who did not place M5S on the left right dimension and H1b necessarily does, we can observe that there we a greater number of observations for the former than the latter. The results demonstrated here were not expected. As a reminder, H1a posited that those who felt M5S was beyond ideology, but then gained one once in government would become disengaged. In fact, this is opposite to the pattern that we observe in the data. As compared to those who did not place M5S on either wave, those who did place it on the 2020 wave were less likely to disengage as compared to remaining a loyalist. Relatedly, those who were able to place the party in 2018, but could not in 2020 were more likely to disengage then remain loyal. Together, these suggest that one of the attractive characteristics of preferring the M5S is its ideological position.
Yet, in observing the ideology-based variables of the second model, we cannot observe that any particular perceived left-right placement of the M5S nor respondents believing that the party had moved further away from them resulted in any significant difference in likelihood of deserting or defecting as compared to remaining loyal. This might be suggestive of the effectiveness of the blurring strategy of valence populist parties (Zulianello and Larsen, Reference Zulianello and Larsen2024) in that parties that present blurred positions allow for respondents to perceive that they might be ideologically close to them by selectively focusing on the positions that they agree with (Somer-Topcu, Reference Somer-Topcu2015).
Turning to the hypotheses related to incumbency, no evidence is found for perceptions of the economic situation having an independent effect on M5S respondents no longer preferring the party. While there is ample discussion supporting the economic vote in Italy, this suggests that party preferences, outside the election context, might not be affected by this.
As for the other variables related to governance, they have the expected direction. Having a more positive view of the performance of the government with M5S involved makes respondents less to desert or disengage than remain loyalists. A similar pattern can be noted regarding opinions of the party leader. Those whose views of Di Maio increased between 2018 and 2020 are less likely to desert or disengage. Both models validate the hypotheses that suggested those whose views on government performance or party leadership worsened would be more likely to desert, with the additional caveat that they were also more likely to disengage.
To ease substantive interpretation, Table 4 estimates the effect that the variables have on the probability of being in each group. In the first model, we can observe that those unable to place the M5S in 2018 would be less likely to be disengaged by 7% in 2020. Conversely, those who were unable to place M5S in 2020 were 24% more likely to disengage. Again, here we note the importance in being able to place the party ideologically in government. Those that could not were more likely to not affiliate with any party.
Predicted effect sizes
Displayed values are semielasticities except for
^ , which are marginal effects. All values significant at the 0.05 level, except for *, which is at 0.1 level.
For the continuous variables, semi-elasticities were calculated in Table 4. These estimates the change in likelihood of falling into each category given a change of one standard deviation of the independent variable. Given their different ranges of potential values and observed values, this is one way to potentially compare the strength of the variables to one another. Looking first at the economic variable, there is almost a statistically significant, though substantively minor, effect of having a more negative view of the economy, which increases the probability of becoming disengaged by 3.6%.
Next, in examining the variable of government performance, we observe support for H2b as presented. Those feeling one standard deviation more positive about the government are 10.90% (1st model) more likely to remain as loyalists. This does not pull equally from the other categories. Reversing the sign, we can see that negativefeelings about government performance more strongly influence one's decision to desert than to disengage. This is what wa expected, given that those concerned about government might have been more favourable towards M5S for reasons related to discontent with previous administrations and prefer parties which will govern to their liking. This pattern, with a larger magnitude, is also evident in the second model.
A similar, and stronger, effect on probabilities can be observed on the opinion of the party leader. Those who view Di Maio one standard deviation more negatively in 2020 than in 2018 are 20% less likely to remain loyal to the party and more three times more likely (∼15.5/4.4) to desert the party for another than to completely disengage from any party preference. In line with the literature on the personalization of politics in Italy and how short-term party preferences can shift based on such matters, we see this effect in both models. The party leader effect more strongly predicts M5S losing party preference than perceived government performance, economic performance, or ideological shifts of M5S (Bellucci, Reference Bellucci2007, Reference Bellucci2014; Bellucci and De Angelis, Reference Bellucci and De Angelis2013; Bellucci et al., Reference Bellucci, Garzia and Lewis-Beck2015).
Incumbency effects on populists in power
While the case of the M5S was discussed as a case of a valence populist party in power, populist parties in nearby countries have also faced reduced polling results after their time serving in government. In Spain, Podemos challenged the existing dominant parties of both centre-left and centre-right, offering a populist-left alternative. Similar to the Italian case, part of its success lay in is charismatic leadership (Kioupkiolis Reference Kioupkiolis2016). In 2016, it secured 21.2% of the vote and 69 seats in the Cortes Generales. Thereafter, it served as a support party for the centre-left Sanchez I cabinet. During this time, voters came to associate the party with the governing socialists (Bergman et al., Reference Bergman, Krauss and Thürk2025). Its vote-share dropped to below 15% in the subsequent 2019 elections. It became a formal coalition partner under Sanchez II in 2019. Yet, in joining the government, moderations and compromise resulted in the party lacking a clear brand, which has left it open to challenges from other outside parties (Brown, Reference Brown2024). After the 2023 election, it held only five seats within the Cortes Generales.
Greece also experienced a period of populists in government. After the January 2015 election, anti-austerity parties of the populist left (SYRIZA) and right (ANEL) joined in coalition: Tsipras I. The parties remained in coalition after the September 2015 election (Tsipras II), yet both parties suffered declining polls, as the government capitulated over the terms of an economic rescue plan including tax hikes, which led to further increasing resentment towards Tsipras. At the policy level, the populist-right base of ANEL were not pleased with the governments extension of citizenship and LGBT rights (Tsatsanis et al., Reference Tsatsanis, Teperoglou and Seriatos2020). ANEL did not even contest the 2019 election, which saw SYRIZA lose over a third of its seats and steady declines in subsequent elections.
The valence populist OL'aNO (Zulianello and Larsen, Reference Zulianello and Larsen2024) led two governments in Slovakia from March 2020 to May 2023; 45 of the 150 seats won in the National Council in 2020 dropped to 16 after the 2023 elections. With a similar rise as the M5S, they came to power as an anti-corruption populist of the centre mobilizing non-voters (Haughton et al., Reference Haughton, Rybář and Deegan-Krause2022). Yet, once in office, governance was classified as ‘chaotic’ (Havlík et al., Reference Havlík2024, p. 60) with disagreements over policy generating friction between coalition partners. Voters were persuaded to leave the party on the account of instability and ineffective leadership (Haughton et al., Reference Haughton2024).
What these findings from other countries and the novel analysis presented above have in common is the primacy of issues associated with governance (i.e. government policies and leadership qualities) and populist party decline. Thus, while the literature on voting for populist parties often focuses on either populist or ideological attitudes (Rooduijn, Reference Rooduijn2018), these might not be determinative of continued support or party preference within a governing term. Instead, partisan attachment towards populist parties in office is based on deliverables and government outputs.
A particular weakness of valence populists in office, such as M5S and OL'aNO, perhaps, is the lack of thick ideological element or clear positional character with free-floating and ‘often inconsistent’ (Zulianello, Reference Zulianello2020, p. 332) policy stances. Centrism might also not be an appropriate characterization at is supposes some ideological positioning (Zulianello and Larsen, Reference Zulianello and Larsen2024). Without such a base from which to draw support, they might be further scrutinized on their governing characteristics, as there is little else to judge them on, especially if their goals of anti-corruption, transparency, moral integrity, and democratic reform (Zulianello, Reference Zulianello2020) are not enacted.
Conclusion
This article set forth to engage with the literature on (valence) populist parties and whether individuals were attracted to such parties for ideological reasons. In particular, the analysis here was focused on partisan loyalty, desertion, or disengagement as it relates to Italy's M5S. This was an extreme case that lost more than half of its support during the term, and thus could potentially provide enough data to subject to multivariate panel analysis. Drawing on the literature on populism and Italian politics, two explanations for decline were presented. The first, which was not supported, argued the ‘beyond’ ideology aspect of the M5S was no longer credible given the party's participation in two cabinets and that individuals who supported M5S would become disengaged with politics. In fact, there is some evidence to the contrary, suggesting that it was those who could no longer place the M5S ideologically in 2020 that had become disengaged politically. While viewing the party as beyond left-right ideology was an attraction during its rising years (Vittori, Reference Vittori2020; Angelucci and Vittori, Reference Angelucci and Vittori2021), once in power, citizens did not view it that way. Descriptively, most respondents were able to place the M5S ideologically.
An alternative perspective argued that it was the penalties of incumbency that reduced M5S support. This argument did have empirical support, in that respondents who shifted their favourability towards party leadership and government performance were also most likely to desert the M5S for other parties. A particular strength of this finding, as compared to other studies that use recalled vote choice or national/regional aggregates, is the use of a panel survey to control for alternative explanations at the individual level and to resist making claims based off ecological fallacies.Footnote 7 All key variables were also dynamic in nature, allowing the analysis of how changes in an individual's opinions affected their party preference.
Performance in government, like any other incumbent, now serves as the basis M5S support – not ideological protest as in 2013 (Passarelli and Tuorto, Reference Passarelli and Tuorto2018). Research on the 2022 election identified that it was policy motivations that influenced M5S desertion to Fratelli D'Italia (Puleo et al., Reference Puleo, Carteny and Piccolino2025). The analysis here confirmed the strength of party leader favourability as a predictor for Italian political beliefs (Bellucci et al., Reference Bellucci, Garzia and Lewis-Beck2015; Garzia, Reference Garzia2017; Bergman, Reference Bergman2020). Further analysis of the connection between perceived competence in office and partisan attachment to populist parties in other cases could particularly serve as a basis for future research and generalizability of these findings.
While this article examined party loyalty and defection, in-depth analysis into which parties’ deserters were attracted to were not hypothesized. Perhaps their MPs also had deserted the M5S to another party and respondents were simply following their elites (Pinto, Reference Pinto2023). While recent research had begun to examine voter attraction towards leading and losing parties (Puleo et al., Reference Puleo, Carteny and Piccolino2025), there is still more to learn about the psychology of remaining loyal in such dynamic circumstances. The study here employed regional fixed-effects and so the findings here examine the variation within each region. Yet, previous research demonstrates that regional attributes might also have an impact on voting behaviour (Bloise et al., Reference Bloise, Chironi and Pianta2021). One might also investigate if there were to be regional aspects towards party preferences as well, potentially being influenced by subnational administrations (Hjermitslev et al., Reference Hjermitslev, Krauss and Thürk2025).
Funding
This research received no specific grant from any public or private funding agency.
Data
The replication dataset is available at http://thedata.harvard.edu/dvn/dv/ipsr-risp; https://doi.org/10.7910/DVN/HTM5FT
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ipo.2025.10081.
Acknowledgements
An earlier version of this manuscript was presented at the APSA 2024 Annual Meeting & Exhibition. Thoughtful comments were provided by Eric Jones, Marco Improta, and CONGRIPS panel attendees. Travel was made possible by Corvinus University of Budapest (2024/280). Thorough comments by two anonymous reviews strengthened this contribution. A final thanks to the editors and Annette Lillethun for their assistance in ensuring the clarity of the presentation.
Competing interests
The author declares none
Open access
