Introduction
There is an inherent conflict between the ideals of mandate representation and the realities of coalition government in multiparty systems. The mandate model requires parties to make pledges before an election and to fulfill these pledges if they become part of the government (American Political Science Association 1950; Mansbridge Reference Mansbridge2003; Schedler Reference Schedler1998). Yet these ideals are much more likely to be realized in two-party systems that produce single-party governments. Here, the winning party’s manifesto serves as the blueprint for the government’s program, and voters have an easier time assigning responsibility and judging the performance of the government (Hobolt et al. Reference Hobolt, Tilley and Banducci2013).
Proportional electoral systems, however, regularly produce coalitions that force multiple parties to compromise on their policy ambitions (Lijphart Reference Lijphart2012; Powell Reference Powell2000). The government program is typically negotiated after the election in the form of a coalition agreement (Strøm et al. Reference Strøm, Müller and Bergman2008). Post-election majority-building (Ganghof Reference Ganghof2015) and the need for compromise inevitably result in lower rates of pledge fulfillment for coalition governments as compared to single-party cabinets (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017). More importantly, pledge fulfillment in multiparty cabinets is strongly influenced by coalition-specific characteristics, such as the allocation of ministerial portfolios, the content of the coalition agreement, and the strength of interparty dissent (Praprotnik Reference Praprotnik2017; Schermann and Ennser-Jedenastik Reference Schermann and Ennser-Jedenastik2014; Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017). The reality of coalition governance thus severely constrains the policy ambitions of individual cabinet parties.
We seek to explore whether citizens’ expectations align with these political realities. Specifically, we aim to explain the phenomenon of accepting pledge non-fulfillment, which we consider a precursor to liking a party or voting for it.Footnote 1 How do the negotiation context and outcome influence citizens’ tolerance for unmet electoral promises? And to what extent is this relationship moderated by citizens’ instrumental considerations? We theorize that when party or coalition characteristics are unfavorable to fulfillment, citizens are more likely to accept pledge non-fulfillment. More precisely, we hypothesize that the negotiation context (number of government parties, party size, disagreement between government parties) and the negotiation outcome (prime ministership, distribution of portfolios, coalition agreement) affect the acceptability of pledge non-fulfillment. In addition, we test the role of instrumental considerations and hence whether agreement with the respective pledge cancels out any effects of party and coalition characteristics.
To test our hypotheses, we ran two online survey experiments in Finland and Germany in the spring of 2024. While both countries feature proportional electoral rules that produce multiparty systems and coalition governments, they differ substantially in the prevalence of institutional and partisan veto points. Our findings showed that citizens in both countries consider the context and outcome of coalition negotiations when evaluating how acceptable it is for parties to renege on their promises. Disagreement between coalition partners, not holding the prime ministership, and the absence of a pledge in the coalition agreement were associated with increased voter tolerance for unfulfilled promises. We did not find support in either country for the role of instrumental considerations; that is, citizens make similar judgments regardless of their agreement with a pledge. There were also notable differences between the countries: in Germany, we found no evidence that the number of parties in the coalition or portfolio control affected voter evaluations, while such factors appeared to influence evaluations in Finland. Additionally, the effects were consistently stronger in Finland than in Germany. Thus, while the results were broadly comparable between the two cases, they were not entirely uniform, suggesting that structural factors beyond the characteristics of a particular coalition condition the way in which citizens evaluate parties’ performance in office.
The central contribution of this study is to confront voters’ (legitimate) expectations about promise fulfillment with the real-world constraints of coalition government. Thus, we build on the puzzling findings of previous studies: while coalition settings generally restrict parties’ ability to fulfill their pledges (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017), citizens do not necessarily appear to adopt a ‘compromising mindset’ (Plescia et al. Reference Plescia, Ecker and Meyer2022). Instead, initial evidence suggests that citizens maintain high expectations of parties even in complex governing situations (Fortunato et al. Reference Fortunato, Lin, Stevenson and Wessel Tromborg2021; Matthieß Reference Matthieß2022). For example, junior coalition parties are often punished more severely (Klüver and Spoon Reference Klüver and Spoon2020; Hjermitslev Reference Hjermitslev2020), although fulfilling pledges is systematically more difficult for them compared to senior coalition parties (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017). While senior coalition partners may receive at least modest electoral rewards for fulfilling their pledges, junior partners tend to incur electoral losses regardless of whether they fulfill more or fewer of their promises (Matthieß Reference Matthieß2024, 87). Unlike previous studies that examined citizens’ expectations with respect to abstract pledge fulfillment (Werner Reference Werner2019; Heinisch and Werner Reference Heinisch and Werner2023; Markwat Reference Markwat2020), our study focuses on specific pledges and tests whether citizens’ approval of these pledges moderates the relationship between the political situation and their acceptance of non-fulfillment. Understanding whether and how voters take coalition and party characteristics into account when forming opinions about unfulfilled campaign promises is key to addressing the challenges multiparty democracies face in producing responsive and accountable government.
Investigating the acceptability of pledge non-fulfillment is therefore essential, especially for countries with multiparty coalition governments, including Finland and Germany, as well as for a large number of both old and newer democracies. More generally, this endeavor is related to the question of whether citizens recognize and approve of the democratic rules in the countries they live in. Proportional systems require parties to be politically flexible and willing to make policy concessions to others, and citizens to be understanding of the limited policy influence any one party can exert. Even if citizens generally – and legitimately – expect promises to be kept (Matthieß and Regel Reference Matthieß, Regel, Kneip, Merkel and Weßels2020), the question is whether they can also take into account the particular circumstances and constraints under which coalition parties operate. If this is the case, there is an opportunity for parties to communicate and explain why certain promises are not being honored, and thus to be transparent and inform citizens about unfulfilled promises. If this is not the case, the result will be a permanent tension between the political reality that sometimes requires promises to remain unmet and the expectations of citizens.
Theory
A large body of literature examines how citizens react when politicians change their positions, renege on promises, or simply remain ambiguous in their commitments (McGraw Reference McGraw1990; McGraw et al. Reference McGraw, Best and Timpone1995; Hinterleitner and Sager Reference Hinterleitner and Sager2017). In such situations, politicians often engage in blame-avoidance strategies. Studies show that politicians’ explanations of their actions can help mitigate the impact from taking unpopular positions or decisions, although the effects are limited by counter-narratives from political opponents (Grose et al. Reference Grose, Malhotra and Van Houweling2015; Robison Reference Robison2024, Reference Robison2022). We contribute to this literature by examining an institutional context – coalition government – that often requires politicians to change position and renege on pre-election commitments.
Multiparty systems typically produce coalition governments, with several parties sharing executive authority and accountability for implemented policies. The literature on coalition governance in general (for example, Müller and Strøm Reference Müller and Strøm2003), and pledge fulfillment more specifically (for example, Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017), has examined the mechanisms of joint policy-making to understand a coalition party’s influence on policy outcomes. The findings predominantly suggest that a party’s influence is significantly shaped by its size, the status as the prime ministerial party, and the distribution of ministerial portfolios. Shifting the focus to how voters perceive coalition governments, a substantial body of research has emerged, exploring how citizens assign responsibility and evaluate the performance of political parties in such setups (Duch et al. Reference Duch, Przepiorka and Stevenson2015; Bowler et al. Reference Bowler, Gschwend and Indridason2020; Angelova et al. Reference Angelova, König and Proksch2016; Gerstenberg and Lagnado Reference Gerstenberg and Lagnado2012; Lagnado et al. Reference Lagnado, Gerstenberg and Zultan2013). These studies have generally affirmed the relevance of the factors highlighted by research on coalition policy-making.
Our study seeks to synthesize these insights by tracing and advancing the scholarly discourse on the perception of pledges. There have been studies on the perception of pledge fulfillment (Pétry and Duval Reference Pétry and Duval2017; Thomson Reference Thomson2011; Ylisalo Reference Ylisalo2022), and more importantly for our study, studies on citizens’ reactions and expectations. Research on citizens’ reactions has shown a punishment effect if pledges are not fulfilled (Matthieß, Reference Matthieß2020; Matthieß, Reference Matthieß2022; Naurin et al. Reference Naurin, Soroka and Markwat2019). Moreover, Naurin et al. (Reference Naurin, Soroka and Markwat2019) found that citizens who disagree with a pledge tend to penalize the government for fulfilling it. Additionally, a handful of studies have explored whether citizens prefer pledge fulfillment over other ideals of representation and parties’ actions. For example, an experimental study by Werner (Reference Werner2019) in Australia revealed that citizens do not necessarily want parties to fulfill their pledges if doing so conflicts with public opinion.
While voters thus value the fulfillment of commitments, the factors that might lead them to excuse failures remain less understood. Our study addresses this gap by examining citizens’ understanding of the constraints on policy making in proportional systems.
In addition to drawing on previous research on responsibility attribution in multiparty contexts, we build upon the argument for the existence of a ‘compromising mindset’. Wolak (Reference Wolak2020) has demonstrated that Americans consider compromise a core democratic value that surpasses party allegiances. Similarly, Green-Pedersen and Hjermitslev (Reference Green-Pedersen and Hjermitslev2024) analyzed observational survey data and found that citizens in Austria and Denmark appreciate parties that are willing to compromise. Furthermore, their survey experiment in Denmark revealed that the likelihood of voting for a hypothetical party increases when the party is characterized as willing to compromise, even after accounting for factors like ideology.Footnote 2 This argument extends beyond citizens’ vote choice to their acceptance of unfulfilled pledges. If citizens possess a ‘compromising mindset’ they are likely to exhibit greater tolerance for unfulfilled pledges in coalition settings and in more complex policy-making situations.
Below, we first put forward hypotheses derived from the coalition negotiation context, that is, the features of a coalition that are given at the start of the coalition bargaining process (number of parties, party size, interparty disagreement). Next, we present hypotheses based on the negotiation outcome, that is, the institutional arrangements created by the parties as a result of the coalition bargaining process (status as the prime ministerial party, portfolio allocation, coalition agreement). Last, we hypothesize about the interaction between these coalition-related factors and the pledge-specific preferences of respondents.
Negotiation Context
Our initial hypothesis centers on a straightforward heuristic that voters may use to assess the acceptability of unfulfilled promises: the number of parties involved. The pledge literature builds on veto player theory (Tsebelis Reference Tsebelis2002), which suggests that pledge fulfillment is likely to be lower in coalitions with more parties (and those spanning a broader ideological spectrum, see discussion of H3 below). Thus, the more parties involved in a coalition, the higher the level of bargaining complexity (Martin and Vanberg Reference Martin and Vanberg2003), and the more difficult it becomes for any single party to achieve its policy objectives (see also Green-Pedersen et al. Reference Green-Pedersen, Mortensen and So2018), as each party holds veto power in a government coalition, especially in minimal winning coalitions (Müller and Meyer Reference Müller, Meyer, König and Tsebelis2011). As a result, a larger number of parties should lead to increased voter leniency toward unfulfilled pledges.
HYPOTHESIS 1: Voter acceptance of pledge non-fulfillment is higher when there are more parties in government.
In addition to the overall number of parties in cabinet, we expected that citizens’ willingness to accept pledge non-fulfillment would be affected by a party’s agenda-setting and bargaining power. These powers imply that larger parties have a greater ability to exert actual policy influence (Martin and Vanberg Reference Martin and Vanberg2011; Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017; Green-Pedersen et al. Reference Green-Pedersen, Mortensen and So2018). As Fortunato et al. (Reference Fortunato, Lin, Stevenson and Wessel Tromborg2021) show, voters do rely on the size heuristic and thus attribute greater policy influence to larger parties (see also Angelova et al. Reference Angelova, König and Proksch2016). If voters use party size to judge the influence of coalition members on policy making, then their acceptance of non-fulfillment should also take this factor into account. In addition, voters may use the electoral support a party receives as a proxy for the popular support of its election promises. Pledges made by smaller parties could thus be seen as carrying less of a mandate than those made by larger parties. As a result, the perceived imperative to enact pledges should vary with party size.
HYPOTHESIS 2: Voter acceptance of pledge non-fulfillment is higher for pledges made by smaller parties.
Multiparty government can only function if a minimum level of coalition discipline is upheld by all parties involved (see also Angelova et al. Reference Angelova, König and Proksch2016). One important condition for establishing coalition discipline is that all actors submit to the veto powers of their coalition partners (Müller and Meyer Reference Müller, Meyer, König and Tsebelis2011). This ensures that individual government parties do not enter legislative alliances with opposition parties unilaterally. The parties’ submission to mutual veto powers implies that no policies can be enacted if one coalition party objects. Policies that are jointly agreed will be more likely to be implemented. Intracoalition dissent regarding specific pledges thus becomes a central obstacle to their fulfillment.Footnote 3
While voters would like to see their preferred party fulfill its pledges in general, they typically display some willingness to accept compromise (Plescia et al, Reference Plescia, Ecker and Meyer2022; Green-Pedersen and Hjermitslev Reference Green-Pedersen and Hjermitslev2024). We therefore conjectured that respondents’ acceptance of non-fulfillment would be higher if there was dissent about a policy among coalition parties.
HYPOTHESIS 3: Voter acceptance of pledge non-fulfillment is higher when there is disagreement about the pledge among the coalition parties.
Negotiation Outcome
The biggest and most visible prize in coalition bargaining is the office of the prime minister (PM). While prime ministerial parties are typically the largest coalition party, their power is not merely a function of their size. Rather, PMs are powerful because of institutional, partisan, or personal factors (O’Malley Reference O’Malley2007), and this power translates into higher rates of pledge fulfillment (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017). This aligns with coalition formation and policy-making models, such as that of Huber (Reference Huber1996) who argued that the vote of confidence procedure allows the prime ministerial party to raise the stakes in legislative votes, thereby limiting the influence of other coalition parties on policy.
Given the outsized influence of PMs and their parties in coalition governments, it is logical that voters view them as disproportionately responsible for policy outcomes (Fortunato et al. Reference Fortunato, Lin, Stevenson and Wessel Tromborg2021; Angelova et al. Reference Angelova, König and Proksch2016). Conversely, junior coalition parties are not expected to be able to wield as much influence. We therefore conjectured that voters would be more likely to excuse non-fulfillment if a party does not hold the office of PM.
HYPOTHESIS 4: Voter acceptance of pledge non-fulfillment is higher if the pledge-making party does not hold the office of the prime minister.
Duch et al. (Reference Duch, Przepiorka and Stevenson2015) found that, in collective decision-making scenarios, individuals tend to attribute the greatest responsibility to the proposer, punishing them more severely when outcomes are perceived as unfair. This focus on the proposer persists regardless of the proposer’s voting weight or the presence of veto power among other decision makers (see also Forsyth et al. Reference Forsyth, Zyzniewski and Giammanco2002; Caine and Schlenker Reference Caine and Schlenker1979). Similarly, Duch and Stevenson (Reference Duch and Stevenson2013) demonstrate that responsibility attribution in the context of economic affairs in the UK is predominantly influenced by the ability to propose. Building on the research of Duch et al. (Reference Duch, Przepiorka and Stevenson2015), Angelova et al. (Reference Angelova, König and Proksch2016) affirm the significance of the PM’s role, but also highlight that responsibility attribution to the PM’s party is particularly evident in the portfolios held by the PM’s party. Voters thus do have some knowledge of portfolio allocation among coalition parties. This is in line with Lin et al. (Reference Lin, Stevenson, Tromborg and Fortunato2017) who found that (some) voters can correctly identify the portfolio distribution in a coalition. Given that portfolio control comes with substantial agenda-setting and veto powers (Laver and Shepsle Reference Laver and Shepsle1990; Bäck et al. Reference Bäck, Müller, Angelova and Strobl2022), we theorized that non-fulfillment would be more acceptable if a party did not hold the relevant portfolio.Footnote 4
HYPOTHESIS 5: Voter acceptance of pledge non-fulfillment is higher if the pledge-making party does not control the relevant portfolio.
Beyond the distribution of offices, the coalition agreement is another crucial outcome of the negotiation process between coalition parties following an election (Strøm et al. Reference Strøm, Müller and Bergman2008). In this agreement, parties outline their joint agenda. When there is significant overlap in positions, drafting this joint program becomes easier. However, conflicts typically arise, requiring either compromises or logrolling (that is, a party supports other parties’ proposals in return for their support of its priorities) to resolve them (Martin and Vanberg Reference Martin and Vanberg2011, Reference Martin and Vanberg2014). Although the coalition agreement is not legally binding, it often carries a de facto binding character for the parties involved (Strøm et al. Reference Strøm, Müller and Bergman2008). Research also indicates that commitments included in the coalition agreement are more likely to be fulfilled (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017; Praprotnik Reference Praprotnik2017).
We argue that this de facto binding character is also considered significant by citizens. The coalition agreement is a public document that indicates a high degree of commitment. Pledges included in the coalition agreement have been accepted by all coalition parties, even if they did not originally support them. Therefore, if an election pledge is included in the coalition agreement, citizens are likely to have higher expectations that the pledge will be fulfilled. Conversely, if the pledge is not included, there will be a higher tolerance for its non-fulfillment.
HYPOTHESIS 6: Voter acceptance of pledge non-fulfillment is higher if the pledge was not included in the coalition agreement.
Instrumental Evaluation
All of the above arguments are based on the premise that voters understand (some of) the constraints that coalitions impose on parties’ policy ambitions. However, voters do not just take such institutional or preference-based obstacles into account, they also reason instrumentally. Previous studies conclude that they systematically favor the decision-making venue that is more likely to produce their preferred outcome (Werner Reference Werner2020; Pilet et al. Reference Pilet, Bedock, Talukder and Rangoni2024; Beiser-McGrath et al. Reference Beiser-McGrath, Huber, Bernauer and Koubi2022; Landwehr and Harms Reference Landwehr and Harms2020; Markwat Reference Markwat2020). For instance, Markwat (Reference Markwat2020) shows that, while voters value pledge fulfillment per se, they value congruence between their own preferences and implemented policies more. In a similar vein, Naurin et al. (Reference Naurin, Soroka and Markwat2019) found that voters may punish governments for the fulfillment of pledges with which they disagree.
This tendency toward outcome-based reasoning implies that voters will find non-fulfillment more acceptable if they disagree with the pledge. Such instrumental thinking is potentially at odds with the normative expectation that parties should keep their promises. Yet, for those opposed to a pledge, the constraints of coalition provide a welcome reason to discard the promise-keeping norm and simply embrace their preferred outcome. Those who agree with a policy, however, have much less motivation to look for a reason to accept its non-fulfillment. They will therefore be less responsive to coalition-related constraints. As a result, we expect the effects of the previous hypotheses to be stronger for respondents who disagree with the respective pledge.
HYPOTHESIS 7: The effects posited in H1 through H6 will be smaller for respondents who agree with the pledge than for respondents who disagree with the pledge.
Data and Methods
The data for this study (Ylisalo et al. Reference Ylisalo, Matthieß, Praprotnik and Ennser-Jedenastik2025) originate from two Western European democracies: Finland and Germany. We conducted online survey experiments in both countries, asking respondents to evaluate the acceptability of a fictional party’s failure to fulfill a pledge under various negotiation contexts and outcomes.
Both Finland and Germany elect their parliaments by proportional representation and feature multiparty systems and coalition governments. Thus, respondents in our surveys will be familiar with ‘minority situations’ that require two or more parties to form a majority cabinet, allocate portfolios between parties, and negotiate a coalition agreement that serves as a guideline for policy making throughout the legislative period.
However, beyond these commonalities, there are important differences between the two countries. This begins with the constitutional structure of the state. Finland is a unitary state that centralizes power at the national level, whereas Germany is a highly federalized state that constitutionally distributes power to its regions (see, for example, the two countries’ different ratings in the Regional Authority Index: Shair-Rosenfield et al. Reference Shair-Rosenfield, Schakel, Niedzwiecki, Marks, Hooghe and Sandra2021; Hooghe et al. Reference Hooghe, Marks, Schakel, Chapman-Osterkatz, Niedzwiecki and Shair-Rosenfield2016) and features strong bicameralism (Mueller et al. Reference Mueller, Vatter and Dick2023). The German states thus serve as significant veto players outside the coalition – an element absent in the Finnish context. In addition, Germany’s powerful constitutional court severely constrains its legislature and executive (Vanberg Reference Vanberg2004).
While Germany thus has a greater number of institutional veto players, Finland has more partisan ones. The Finnish party system has traditionally been more fragmented than the German one (Casal Bértoa Reference Casal Bértoa2024), leading to a more complex bargaining context and ultimately a more intricate negotiation outcome. The modal number of government parties in Finland has been four since the 1970s, often including five or even six parties. In contrast, coalition governments in Germany at the national level have typically involved two, and until recently never exceeded three parties.Footnote 5 In addition, post-war Finland has typically had oversized coalitions, whereas minimal winning coalitions have been the norm in Germany (Bergman et al. Reference Bergman, Bäck, Hellström, Bergman, Bäck and Hellström2021, 16).
In summary, Finland and Germany share a reliance on coalition government, but within the group of multiparty democracies, they represent two quite different cases in terms of institutional and partisan veto points. Although we do not specify hypotheses regarding the effects of these variations, our aim is to enhance the external validity of our results by testing them across two distinct cases within coalition settings. If we observe similar reactions from citizens in both contexts, it suggests that the mechanisms we are studying operate effectively regardless of the specific constitutional and negotiation contexts and outcomes to which respondents are accustomed.
Experimental Design
To test the hypotheses described above, we designed a vignette experiment that was fielded in Finland and Germany in the spring of 2024. Both surveys were preregistered with OSF Registries before the collection of data started.Footnote 6 Ethics approval was obtained from the Institutional Review Board of the Faculty of Social Sciences at the University of Vienna (Decisions 1081 and 1143).
The Finnish data were collected using the online panel Kansalaismielipide (‘Citizen Opinion’), which is part of the Finnish Research Infrastructure for Public Opinion and funded by the Research Council of Finland. The experiment was administered in both Finnish and Swedish using the online survey platform Qualtrics. A random sample of panel participants, recruited from the adult general Finnish population, was drawn for this study with no specific quotas. We obtained 2,269 responses from Finnish nationals at the age of 18 and older.Footnote 7 The survey ran from 5 March until 25 March 2024.
The German online survey was conducted by the survey company Bilendi, targeting eligible German nationals aged 18 and older. Participants were sampled from Bilendi’s online panel based on a quota sample stratified by age, gender, region, and education (representative of the German population). A total of 1,359 interviews were completed.Footnote 8 The fieldwork was carried out from 22 April until 25 April 2024. No major events, such as national electoral campaigns, took place in either country during the data collection period.Footnote 9
Before the experiment, respondents were introduced to the six pledges that would later be used in the experimental vignettes (see Table 1). Respondents were asked to express their opinion about each statement using a five-point scale (‘fully agree’, ‘somewhat agree’, ‘neither agree nor disagree’, ‘somewhat disagree’, ‘fully disagree’). The order of these statements was randomized for individual respondents.
Pledges and associated policy areas per country
Half of the pledges in the experimental vignettes are associated with leftist or liberal political views, while the other half correlate with rightist or conservative views.Footnote 10 In addition, to make the settings depicted in the experiment realistic, the statements pertained to issues that had recently been the subject of public debates. Consequently, they touched upon policies that real parties could have promised to carry out in both countries, even though no references to real parties were included in the vignettes.
We decided to use hypothetical parties rather than real party labels in our experiment because this approach, in our opinion, was the only way to effectively investigate the mechanisms of interest. There are several reasons for this choice. First, a hypothetical setup provides greater flexibility. It would have been challenging to ask participants to imagine a real-world small opposition party as the largest party holding the prime ministership. Second, using real party labels would have limited the pledges we could use. We would either have needed to choose pledges that aligned with the party’s ideology (for example, stricter immigration policies for a Green party would be unrealistic) or used valence issues like reducing unemployment or improving education. The former would have imposed restrictions and strong party-issue associations, making it difficult to separate the effects of the coalition setting from pre-existing opinions about certain parties at that time. The latter would have prevented us from testing the role of agreement to the extent we are interested, as we require polarizing issues. However, we acknowledge that respondents might associate the presented pledges with actual parties. To mitigate this, we chose six diverse pledges with varying ideological stances, presented in a randomized order. This ensured there was no consistent association of all six pledges with a single party. Any attachment of specific pledges to particular parties by respondents would thus be reduced to noise that should not systematically affect our results across the different pledges. Additionally, we asked respondents about their agreement with each pledge. This accounts for the possibility that their agreement with the pledge – as well as their opinion about a related party – influences their considerations and tolerance for pledge non-fulfillment.
We acknowledge that the use of hypothetical parties comes at a cost in terms of external validity. One potential criticism is that, in an age of partisanship, polarization, and political distrust, voters will respond very differently to real versus hypothetical political scenarios (Bonilla Reference Bonilla2025). However, the countries included in our study exhibit medium to low levels of polarization (Wagner Reference Wagner2021; Kekkonen and Ylä-Anttila Reference Kekkonen and Ylä-Anttila2021; Munzert and Bauer Reference Munzert and Bauer2013), display above-average levels of institutional trust (Torcal Reference Torcal, Zmerli and Van der Meer2017), and Germany at least has experienced a decline in party identification (Arzheimer Reference Arzheimer2006). We therefore contend that voter responses to these hypothetical scenarios will still be informative about real-world behavior.
Regarding the selection of the pledges, we tailored the surveys to include different sets of statements for each country to reflect the distinct public debates and issues relevant in those contexts. The pledges are also realistic, typically referring to adaptations of existing policy instruments (for example, altering levels of benefits of taxes), or even to an earlier policy status quo (for example, reintroducing conscription, abolishing gender quotas). Moreover, the issues were chosen so that they were divisive in light of recent public discussion and opinion polls. This was necessary for two reasons. First, as personal agreement or disagreement with a pledge is an essential part of our theoretical argument, we needed issues on which respondents’ opinions varied. Second, because the statements were presented as unfulfilled electoral pledges at a later stage of the experiment, depicting them as causing interparty controversy would not have been realistic, had they pertained to issues on which partisan conflicts were unlikely.Footnote 11
After a page break, respondents were asked to indicate the importance of six policy areas, the order of which was again randomized. The policy areas corresponded to the statements presented on the previous page but were more generic. No indication was given regarding the connections between the policy proposals and the policy areas. Importance was elicited using a seven-point scale ranging from 1 (‘not at all important’) to 7 (‘very important’). In the German experiment, respondents were also asked to rank the policy areas based on importance. We did not formulate hypotheses regarding the effects of perceived importance, but we used this information to evaluate the robustness of the results. The proposals and the associated policy areas are presented in Table 1.
Before beginning the six rounds of the experimental treatment, we informed participants about the task. They were told that they would be presented with six scenarios of unfulfilled electoral promises and instructed to read each scenario carefully before evaluating the acceptability of the unfulfilled promise.Footnote 12
Thereafter, respondents were presented with six vignettes where the statements previously described as policy proposals were reformulated as electoral pledges.Footnote 13 The respondents were asked to imagine a situation in which a parliamentary election in Finland/Germany was about to take place. In this imaginary situation, at the previous election, a hypothetical party had promised to enact one of the six policies mentioned in Table 1. This hypothetical party then made it into a majority coalition government that remained in office throughout the full legislative term.
Respondents were then presented with the negotiation context and outcome in a tabular format, enabling us to test the effects of H1 through H6. The details of the party and coalition characteristics (attributes) were randomized using the values (levels) displayed in Table 2.Footnote 14 Reflecting differences in party systems, the German version differed from the Finnish version in two respects. First, the number of government parties was either three or five in the Finnish version but either three or two in the German experiment. Providing larger numbers would have been unrealistic in the German context, while a government consisting of only two parties would have been unrealistic in Finland. Second, in the Finnish case, the pledge-making party was depicted as either one of the largest government parties (22%), holding the prime ministership; one of the largest government parties (22%) without the prime ministership; or the smallest government party (5%) without the prime ministership. In the German case, it was shown as either the party of the Federal Chancellor or a junior partner, with no reference to party size.
Overview of attributes and levels
Note: PM = prime minister; H1 …5 = hypothesis.
All respondents were shown each of the six pledges, one per vignette in random order. Each vignette ended with the following question to measure the outcome variable of interest: ‘How acceptable do you find the fact that the promise was not fulfilled?’ Respondents were asked to use an eleven-point scale ranging from 0 (‘not at all acceptable’) to 10 (‘fully acceptable’). The experimental part of the survey ended when a respondent answered this question in the sixth vignette.
As noted above, our aim in the experiment was to include pledges on issues that are reasonably divisive. The distributions of respondents’ stated opinions in the Finnish and German samples are shown in Figures 1 and 2. While the distribution is somewhat skewed with respect to some issues, variation in the responses was notable in each instance. Moreover, because we did not want the same respondents to agree or disagree with each issue, we formulated the statements so that they would expectedly reflect both leftist/liberal and rightist/conservative views. As shown in Appendix 2, the data are indeed in line with this expectation.
Distribution of opinions about policy statements in Finland.
Note: FD = fully disagree, SD = somewhat disagree, NAND = neither agree nor disagree, SA = somewhat agree, FA = fully agree.
Distribution of opinions about policy statements in Germany.
Note: FD = fully disagree, SD = somewhat disagree, NAND = neither agree nor disagree, SA = somewhat agree, FA = fully agree.
To test H7, the variable measuring personal opinion was recoded into a factor with three levels. Specifically, the ‘fully agree’ and ‘somewhat agree’ categories were combined and the resulting category was labeled as ‘agree’. Analogously, the ‘fully disagree’ and ‘somewhat disagree’ categories were combined to create a ‘disagree’ category. The ‘neither agree nor disagree’ category was simply given a more concise label, ‘neutral’.
Model Specification
The outcome variable was the acceptability evaluation made at the end of each vignette. We treated it as continuous and used linear regression models. Because each respondent encountered six hypothetical situations and made as many evaluations, we stacked the data so that each evaluation corresponded to one row in the dataset. Hence, in the analyses reported below each observation is a respondent–pledge dyad. To account for the non-independence of the observations, we clustered the standard errors by respondents.
The levels of the experimental attributes were randomized. However, issue preferences were central to our experimental design, and while our selection of pledges was aimed at ensuring that each respondent would agree as well as disagree with some of them, the assignment of such preferences was not random. Randomization ensures that, in the absence of interactions, the coefficients attached to the experimental attributes can be interpreted as average marginal treatment effects (see Kam and Trussler Reference Kam and Trussler2017). The estimates of these effects served as tests of H1–6. To test H7, we interacted personal opinion with the experimental attributes. Given our observed moderator we estimated model specifications where we aimed to control for both measured and unmeasured variables that could plausibly influence the hypothesized moderating effect of issue opinions. Nonetheless, the conditional marginal effects for H7 must not be interpreted as tests of causality.
Analyses
We begin our analysis by examining the distribution of the outcome variable, which is shown in Figure 3. To improve comparability given the different sample sizes, the figure shows percentages instead of counts. Hence, the percentages on the y-axis refer to shares of all evaluations in the respective dataset. The most striking difference was the fact that the Finnish distribution was relatively even, while German respondents chose the middle alternative much more often than their Finnish counterparts. Another noteworthy feature was the higher occurrence of zeros in the German data, which means that German respondents considered the non-fulfillment of a pledge to be completely unacceptable more often. Nonetheless, despite these differences, the mean of the outcome variable was similar in both countries, 4.72 (standard deviation = 3.07) in Finland and 4.13 (standard deviation = 2.74) in Germany.
Distribution of the outcome variable: Acceptability evaluation.
Note: 0 = not at all acceptable, 10 = fully acceptable.
Main Models
We summarize the results of the multivariate analyses using graphs and report the full results, including all fixed effects, in Appendix 4. In addition to the attributes and personal opinions about the pledges, the models included both pledge and vignette dummies. These variables are not included in the graphs, yet it should be noted that most of them were statistically significant.Footnote 15
The results of the models without interactions are summarized, for both countries, in Figure 4. The effects depicted refer to the changes in the outcome variable (measured on a 0–10 scale) that are associated with each respective attribute.Footnote 16 In the Finnish data, the reference category for the number of parties was ‘three’, in the German data ‘two’. The Finnish case allowed us to further distinguish between party size and party status. Due to the way in which the experimental vignettes were constructed, the reference category for both party size and party status was ‘prime ministerial party’ in the Finnish data. In the German data, ‘prime ministerial party’ was the reference category of party status only. Reference categories in the remaining variables were the same across cases. The reference categories for ‘no portfolio control’ and ‘not included in coalition agreement’ variables were ‘control’ and being ‘included’, respectively. The reference category for the variable personal opinion was ‘disagree’. Therefore, the expected signs of the coefficients attached to the attribute levels were positive, while the expected signs on the levels of personal opinion were negative.
Regression results: Explaining acceptability of non-fulfillment.
Note: Acceptability of non-fulfillment as the outcome variable. The estimates are unstandardized regression coefficients. 95% confidence intervals are shown. Models include pledge and vignette-order dummies (not shown).
*** p < 0.001; * p < 0.05.
The overall message conveyed by Figure 4 is that citizens considered both the negotiation context and outcome when evaluating the acceptability of non-fulfillment. H1–6 were supported in at least one of the cases and sometimes in both. To begin with, regarding the negotiation context, respondents found it more acceptable for a pledge to remain unfulfilled if there were more parties in government (H1). This effect held true only in the Finnish case. The Finnish case further allowed for distinguishing government parties by size. The data suggested that party size affects evaluations beyond parties’ roles, as it was more acceptable for a small government party to fail to fulfill its pledges compared to a large government party that does not hold the position of PM (H2).Footnote 17 Comparison with a prime ministerial party yielded a stronger effect that amounted to 26 percent of a standard deviation of the outcome variable. Disagreement between government parties also positively affected the acceptability of non-fulfillment in both studies (H3). Turning to our second set of variables, concerning the negotiation outcome, respondents found it more acceptable for policy plans to remain unfinished if the party did not hold the position of PM (H4). Only in the Finnish case did we find support for the idea that acceptability increases when the party does not hold the relevant portfolio (H5). In both case studies, the models supported our hypotheses that a pledge not included in the coalition agreement had higher acceptability of non-fulfillment compared to pledges in the coalition agreement (H6).
The effects of the attribute levels were not entirely uniform between Finland and Germany, warranting a closer examination of both the size of the effects and the differences across these cases. The coefficients tended to be larger in the Finnish data, and all were statistically significant at the conventional 0.05 level. In contrast, this was not the case for the German data, where some coefficients did not even reach statistical significance. For example, regarding H3 and the disagreement over a pledge made by one coalition party, there was a strong effect in Finland, while the effect was much smaller in Germany. Not only was the raw coefficient larger in the Finnish data, but it also implied a more notable change relative to the standard deviation of the outcome variable (36 percent) than in the German data (five percent). In addition, the Finnish data suggested that a larger number of government parties was associated with a higher level of acceptability (albeit the effect was only around three percent of a standard deviation), whereas there was no evidence of any effect in the German data (H1).Footnote 18 Moreover, the coefficient for portfolio control (H5) pointed in the expected direction in both studies; however, while it was statistically significant in the Finnish case, it was not in the German case. To summarize, H3, 4, and 6 received support from both sets of data, while Hypotheses 1 and 5 received support only from the Finnish data. Finally, there was evidence for H2 (only tested in the Finnish case).
Finally, Figure 4 shows that personal agreement with a pledge was associated with lower acceptance rates. Moreover, the effects were relatively large; agreeing with the substance of the pledge decreased the acceptability of non-fulfillment by 72 percent of a standard deviation in Finland and by 33 percent in Germany. Whether policies are in line with personal issue preferences thus affected citizens’ acceptability evaluations but did not eliminate the effects of coalition and party characteristics.
We now turn to the conditional marginal effects of the attribute levels. According to H7, they should be smaller for respondents who agree with a pledge than for those who disagree. The conditional marginal effects are shown at different levels of the opinion variable in Figures 3 and 4 for Finland and Germany, respectively. Much of the evidence did not support H7.
To begin with, most of the interactions between personal opinion and experimental attributes were statistically insignificant in both countries, suggesting that there is no overall moderating effect. Moreover, whenever the interactions were statistically significant in the Finnish data, their signs were positive, contrary to our expectation. As the reference category of the opinion variable was ‘disagree’, H7 posited that interaction terms composed of ‘agree’ and an attribute level would have a negative sign. As Figure 5 shows, when it comes to party roles and party size the positive effects of ‘no PM’ and ‘small’ were larger among those who agreed with a pledge than among those who disagreed. Furthermore, the positive effect of ‘no ministerial portfolio’ was larger not only among those who agreed with a pledge but also among those who were neutral, which also applied to the positive effect of ‘resist’. The other interactions did not reach statistical significance in the Finnish data. The only statistically significant interactions in the German data (Figure 6) were those that involved coalition agreement. They had the expected negative signs, indicating that for those who agreed with a pledge or were neutral the effect of ‘not included’ was indeed smaller.
Marginal effects of the attribute levels in Finland.
Note: Conditional on personal opinion about the pledge.
Marginal effects of the attribute levels in Germany.
Note: Conditional on personal opinion about the pledge.
The results were highly robust to several alternative modeling choices, including the removal of personal opinions and fixed effects; the addition of background variables measuring respondents’ socio-economic characteristics and political views; the use of the original five-point scale of the opinion variable; controlling for the perceived importance of the policy area; the exclusion of a subset of observations; and restricting the analysis to one pledge or vignette at a time (see Appendix 5).
Propensity to Vote as the Outcome Variable
To evaluate the effects of party and coalition characteristics beyond acceptability evaluations, we conducted another survey experiment with a modified outcome variable. The survey was fielded before the 2025 German federal election. The experimental design, including the selection of pledges and the wording of the vignettes, was essentially identical to the German version described in Section ‘Experimental Design’. However, instead of rating the acceptability of non-fulfillment, respondents were asked to report their propensity to vote for the hypothetical party whose pledge had remained unfulfilled. The results are reported in detail in Appendix 6. Despite generally smaller effect sizes and changes in the statistical significance of some attribute levels (compared with Figure 4), the results indicated that the characteristics we manipulated were consequential for vote intentions specifically.
Conclusion
Our article investigated citizens’ tolerance of unfulfilled pledges, a common phenomenon, particularly in coalition settings where parties must make policy concessions to provide stable government (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017). We explored whether citizens understand and acknowledge this need for compromise and whether their typically high expectations for pledge fulfillment are moderated in coalition contexts. Specifically, we examined whether citizens exhibited a ‘compromising mindset’ (Wolak Reference Wolak2020), showing increased tolerance for unfulfilled pledges in recognition of the constraints inherent in coalition governments.
The results indicated that citizens were sensitive to at least some of the constraints that parties in coalitions face. In both countries, there was evidence that citizens were more tolerant of pledge non-fulfillment if other coalition partners opposed it, if it was made by junior partners, and if a pledge was not included in the coalition agreement. Additionally, there was higher tolerance among Finnish respondents for parties that did not control the ministry responsible for implementing the pledge. Tentative evidence suggests higher voter acceptance in coalitions with more parties, which we were able to test in the Finnish case. Overall, the effects appeared to be more pronounced in Finland than in Germany. While even the largest effect that we detected amounted to a ten percent change relative to the scale of the outcome variable, even seemingly modest effects may be crucial especially in highly competitive political environments. It is illuminating that in one of our cases, Finland, the vote shares of the three largest parties were nearly identical in the two most recent parliamentary elections. Particularly in such settings, the factors that we have discussed may influence public debates and citizen opinion regarding parties’ performance to a degree that notably alters the constellations of political power.
Our findings regarding citizens’ reactions largely aligned with previous studies on pledge fulfillment, which have provided evidence that junior party status, disagreement between coalition partners, and exclusion from the coalition agreement decrease the likelihood of pledge fulfillment (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017). As for portfolio control, empirical tests of ministerial influence on pledge fulfillment have produced mixed results, suggesting that context plays a key role (Costello and Thomson Reference Costello and Thomson2008; Naurin and Thomson Reference Naurin, Thomson, Cotta and Russo2020).Footnote 19
There was little evidence supporting our hypothesis that agreement with the pledge would weaken the effects of coalition circumstances on citizens’ tolerance. In fact, the statistically significant effects that we observed were generally in the opposite direction. The only exception was found in Germany concerning coalition agreements: those who disagreed with the pledge showed a higher tolerance for its non-fulfillment if it was not included in the coalition agreement.
These findings have several important implications. They provide strong evidence that citizens accept pledge non-fulfillment when parties are constrained by the realities of coalition government, thus contributing to the literature on compromising mindsets (Green-Pedersen and Hjermitslev Reference Green-Pedersen and Hjermitslev2024; Wolak Reference Wolak2020). Our study is the first to investigate whether citizens consider different coalition and party characteristics when evaluating pledge non-fulfillment. The results supported the expectation that citizens do account for coalition contexts, showing greater tolerance for non-fulfillment when the bargaining situation is more complex, and the pledge-making party is in a weaker institutional position. This finding reflects the intricate political realities of coalition governance, where constraints on fulfilling pledges are significant (Thomson et al. Reference Thomson, Royed, Naurin, Artés, Costello and Laurenz Ennser-Jedenastik2017). However, the ‘acceptability premium’ that benefits junior coalition partners, for instance, can be depleted by a series of unfulfilled pledges, ultimately resulting in electoral punishments (Klüver and Spoon Reference Klüver and Spoon2020; Hjermitslev Reference Hjermitslev2020). Nevertheless, our results suggest that the mechanisms underlying the empirically observed tendency of junior partners to suffer electorally warrant further investigation. Specifically, vignette-based designs could manipulate information about a party’s overall rate of pledge fulfillment in relation to its status within the coalition and assess the impact on vote intention. Recognizing that citizens factor coalition constraints into their judgments may present an opportunity for parties and politicians to explain the constraints under which they operate more transparently. If parties can communicate these challenges sincerely, they may garner understanding from the electorate and thus contribute to a more accountable system of government. However, whether this communication strategy would be effective is likely to depend on additional factors beyond the scope of this study, such as the sender (for example, media, individual politicians, the party) and the recipient (for example, the electorate or party base) of the message. For now, it is important to observe that such a potential window of opportunity exists.Footnote 20
Furthermore, the finding that agreement with a pledge does not negate the effect of coalition circumstances on citizens’ tolerance of pledge non-fulfillment is intriguing. It suggests that citizens consider more than just instrumental factors; even if pledge non-fulfillment means a policy loss, they account for the coalition setting. Interestingly, those who agreed with the pledge were sometimes even more tolerant, possibly indicating that they are more open to explanations for why the pledge could not be fulfilled. In real-world terms, this could provide parties with the opportunity to explain to their supporters, or supporters of a specific policy, why they were unable to fulfill their promises, particularly if the coalition configuration was complex. One alternative explanation for this result is that individuals who agreed with the pledge may have identified more with the party, given that our experimental design involved hypothetical parties. As party identification tends to produce positive biases in the perception of pledge fulfillment (Belchior Reference Belchior2019; Bonilla Reference Bonilla2025), it remains unclear to what extent it affects citizens’ tolerance for non-fulfillment. Moreover, in our experiments we provided information about coalition and party characteristics in a highly accessible fashion. In real-world settings, voters typically have some knowledge of relative party sizes, portfolio allocation, and party policy positions (Lin et al. Reference Lin, Stevenson, Tromborg and Fortunato2017; Angelova et al. Reference Angelova, König and Proksch2016; Meyer and Strobl Reference Meyer and Strobl2016). However, the overall level of information that many citizens possess is limited. Therefore, the effects of party identification and political sophistication remain important topics for future research.
We fielded our experiments in two countries with extensive experience of coalition governments. Our primary purpose was to see whether citizens in both countries, whose institutions and political history exhibit some similarities but also crucial dissimilarities, respond to coalition architecture similarly. Accordingly, we designed the experiments so that they fitted the country-specific circumstances in relation to the levels of the experimental attributes as well as the pledges shown to the respondents. The fact that coalition and party characteristics mattered in both settings corroborated our expectation that citizens indeed take contextual features into account. However, the results were generally stronger in the Finnish setting, and the moderating effects of issue opinions were not consistent in both countries.
A tentative explanation for these differences draws on the fact that, by Western European standards, the Finnish party system has been highly fragmented for several decades, which has given rise to broad-based and ideologically heterogeneous government coalitions. Therefore, Finnish voters may be more accustomed to interparty compromising and accept it as part of the game. By contrast, the number of coalition partners tends to be significantly lower in Germany, and (until recently), these coalitions were ideologically more aligned. Notably, evidence suggests that German citizens are generally skeptical of pledge non-fulfillment for the sake of coalition compromise (Matthieß and Regel Reference Matthieß, Regel, Kneip, Merkel and Weßels2020). Furthermore, Germany’s constitutional architecture creates numerous additional constraints and veto players, such as those imposed by the federal structure, the upper chamber, and the powerful constitutional court. Given that these factors also influence the policy-making process, future research could explore whether citizens consider these external constraints – and perhaps even weigh them more heavily – when evaluating their tolerance for unfulfilled election pledges. In addition, the average level of trust in the political system has been relatively high in Finland and the other Nordic countries when compared to Germany (Grönlund and Setälä Reference Grönlund and Setälä2012; Torcal Reference Torcal, Zmerli and Van der Meer2017), which may translate into a stronger propensity to acknowledge that sometimes parties cannot carry out their pledges. Formal analyses of the effects of cross-country differences in such structural and attitudinal factors are, however, left for future research.
Finally, alongside this positive view there is a more skeptical perspective on citizens’ tolerance of pledge non-fulfillment that deserves attention. Excessive tolerance might pose a problem for democratic accountability. A dose of skepticism is crucial for holding those in power to account. This involves some degree of intolerance toward failure and broken promises. Skilled politicians might deflect blame for their shortcomings by pointing fingers at coalition partners, and while this may sometimes be justified, it is not always a valid excuse. Some parties fail to deliver on their promises not merely due to coalition constraints, but because of incompetence or otherwise poor government performance, thus justifying citizen skepticism and voter punishment. Even within coalitions, citizens can and should expect that pledge fulfillment is the norm rather than the exception. Ultimately, determining the ‘right’ level of tolerance for non-fulfillment is a delicate balance, one that should be informed by the need for compromise while ensuring political accountability.
This study focused on voters’ acceptance of pledge non-fulfillment (with consistent results replicated in a German sample examining vote intention). Future research could build on these insights by employing multistep experimental designs that first assess voters’ acceptance of unfulfilled pledges, and then examine how this shapes affective evaluations and vote choices in the context of real-world parties and political environments.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0007123425100860
Data availability statement
Replication data for this paper can be found in Harvard Dataverse at https://doi.org/10.7910/DVN/LQJNM1.
Acknowledgements
The authors wish to thank Evelina Lundgren, Michael Meffert, and the three anonymous reviewers for their detailed comments on earlier versions of the manuscript. This work also benefited considerably from the feedback provided by the participants of the ‘Parties, governments, and policy making’ workshop at the 2024 Nordic Political Science Association Congress, the ‘Voter evaluations and coalition governments’ panel at the 2024 European Political Science Association Conference, and the Political Science Research Seminar at the University of Turku.
Financial support
Support for this research was provided by the Research Council of Finland (Decision no. 340660).
Competing interests
None.
Open access
