Introduction
Political theorists have long argued that representatives should reflect not only citizens’ policy preferences but also their views on how representation should work (Mansbridge, Reference Mansbridge2003; Wolkenstein and Wratil, Reference Wolkenstein and Wratil2021). This debate centres on two models: the trustee model, where representatives act on personal judgement, and the voter-delegate model, where they follow constituents’ expressed preferences (Dynes et al., Reference Dynes, Hassell and Miles2023; Eulau et al., Reference Eulau1959; Pitkin, Reference Pitkin1967; Wass and Nemcǒk, Reference Wass, Nemcǒk, Cotta and Russo2020). Throughout this paper, we build on this framework distinguishing four ideal-typical styles of representation: voter delegates and trustees, but also party delegates, who see their primary obligation as following the party line, and politicos, who adjust between voters, party, and personal judgement depending on the situation. Empirical research has explored how closely citizens and elites align on the autonomy representatives should have—both from voters and party lines (Campbell and Lovenduski, Reference Campbell and Lovenduski2015; Carman, Reference Carman2007; Esaiasson et al., Reference Esaiasson2019). However, this literature has focused predominantly on policy congruence, that is, the alignment between citizens’ and representatives’ substantive positions on specific issues or issue domains (Ezrow and Xezonakis, Reference Ezrow and Xezonakis2011; Ferland, Reference Ferland2021). By contrast, studies of process congruence—how representatives conduct themselves in office—remain comparatively scarce. As André and Depauw (Reference André and Depauw2017, 380) observe, “[i]t strikes as particularly odd that the scholarly discussion of citizen satisfaction with democracy has disregarded altogether the importance of citizen–elite congruence with regard to what it is that representatives should do in the process of representing citizens.” This gap is difficult to reconcile with the longstanding argument that democratic legitimacy derives not only from responsive policy outcomes but also from the quality and inclusiveness of the representational process itself—that is, from citizens’ sense that they have a meaningful say in how they are represented and governed (e.g., Dahl, Reference Dahl1989). Blumenau et al. (Reference Blumenau, Wolkenstein and Wratil2025) argue that when representatives act in ways that violate citizens’ normative expectations about how representation should work, this undermines the perceived legitimacy of the representative relationship.
There is growing empirical support for the relevance of process preferences in shaping democratic attitudes. André and Depauw (Reference André and Depauw2017) show that while policy congruence has a stronger effect than process congruence on satisfaction with democracy, the two effects are additive. This means that the process gap is theoretically and empirically distinct from the policy gap. Relatedly, Papp et al. (Reference Papp2024) demonstrate a clear negative effect on democratic satisfaction when there is a mismatch between citizens and representatives in terms of process preferences. Bøggild (Reference Bøggild2020) similarly observes that distrust in politicians partly reflects a general dissatisfaction with the mismatch between the representational styles adopted by politicians and citizens’ expectations. A key alleged mechanism underlying process gaps leading to democratic dissatisfaction is citizens’ sense that they do not have a say over the rules of the democratic game (Allen and Birch, Reference Allen and Birch2015; Norris, Reference Norris2011). Nevertheless, the mechanisms driving these effects remain theoretically underdeveloped, and two interrelated problems warrant attention. First, it is unclear whether citizens hold well-defined preferences about representational styles, whether they can accurately identify elite behaviour, and whether perceived gaps translate into diminished democratic support (Bengtsson and Wass, Reference Bengtsson and Wass2010; Carman, Reference Carman2007; Wass and Nemcǒk, Reference Wass, Nemcǒk, Cotta and Russo2020). Unlike policy congruence—which can be experienced through tangible, material outcomes—process congruence presupposes active and sustained engagement with elite conduct. Given persistently low levels of political knowledge (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Rhodes et al., Reference Rhodes, Axsen and Jaccard2014), many citizens may simply be unaware of how they are represented. This suggests that the effect of process congruence on democratic satisfaction is likely to be conditional on political interest: citizens who follow politics more closely should be better able to perceive representational style mismatches and to translate those perceptions into evaluations of democratic quality. Previous studies have already observed that the effect of policy congruence on democratic satisfaction is amplified by political interest (Costa, Reference Costa2021; Mayne and Hakhverdian, Reference Mayne and Hakhverdian2017). A closely related mechanism concerns educational attainment. Education has long been associated with higher levels of political knowledge, cognitive sophistication, and the ability to process abstract political information (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Hakhverdian and Mayne, Reference Hakhverdian and Mayne2012; Turper and Aarts, Reference Turper and Aarts2017). More educated citizens may therefore be better equipped to recognize discrepancies between their own representational preferences and those of political elites, and to incorporate such discrepancies into their evaluations of democratic performance. Unlike political interest, which captures motivation and attention to politics, education reflects a more structural capacity to perceive and interpret political processes. As such, educational attainment may similarly condition the effect of process congruence on satisfaction with democracy, with more educated individuals expected to respond more strongly to both congruence and incongruence in representational style. Second, existing studies often conflate citizens’ views of all MPs, their preferred party’s MPs, or those in government—each potentially influencing democratic attitudes differently (Bøggild, Reference Bøggild2020; Hibbing and Theiss-Morse, Reference Hibbing and Theiss-Morse2002). It remains unclear which of these perceptions most strongly shapes democratic satisfaction. Importantly, these mechanisms are not mutually exclusive: process incongruence may influence democratic evaluations both through diffuse legitimacy perceptions and, among more attentive citizens, through more explicit recognition of representational mismatches.
This paper advances the literature by empirically examining these mechanisms—political interest and educational attainment—based on three waves of repeated cross-sectional data from three Portuguese legislative elections (2005, 2011, and 2015), uniquely enabling us to compare citizens’ and MPs’ representational preferences. Portugal’s closed-list proportional representation system minimizes the importance of the connection between individual citizens and individual MPs (Borghetto et al., Reference Borghetto, Santana-Pereira and Freire2020). MPs therefore have fewer individual incentives than in other systems to be responsive to citizens’ representational style preferences. Hence, if such an effect appears in Portugal, it is likely to be at least as strong in more open, candidate-centred systems. Our findings show that, even in this least-likely Portuguese case, policy congruence is strongly associated with satisfaction with democracy, whereas the independent contribution of process congruence is modest and not statistically robust once policy and ideological congruence are accounted for. We also find that citizens’ own conceptions of representation are related to their satisfaction in their own right: those who prefer a voter-delegate style report somewhat lower satisfaction than those who prefer a trustee style. Contrary to expectations, this weak relationship is no stronger among the politically interested or the more educated. Political interest does not moderate it, and the limited moderation we find—by education—runs in the opposite direction, being somewhat more pronounced among citizens without a university degree. We also show that three alternative measures of process congruence—the match between citizens and (a) all parliamentarians, (b) MPs from government parties, and (c) MPs from citizens’ preferred party—are highly intercorrelated and yield substantively similar results.
Process congruence as a determinant of democratic satisfaction
Concerns over rising democratic dissatisfaction have long occupied political science, from foundational works by Easton (Reference Easton1965), Inglehart (Reference Inglehart1977), and Dahl (Reference Dahl1989) to recent studies (Broderstad, Reference Broderstad2023; Wiebrecht et al., Reference Wiebrecht2023). Democratic satisfaction refers to how well citizens feel democracy functions in practice (König et al., Reference König, Siewert and Ackermann2022), which is distinct from abstract support for democracy as a regime type (Linde and Ekman, Reference Linde and Ekman2003). While governance quality can positively affect democratic satisfaction, it does not necessarily alter more principled support for democracy as an ideal (Claassen and Magalhães, Reference Claassen and Magalhães2023). Democratic satisfaction has been described as occupying a middle ground between stable regime support and more volatile measures such as political trust (Norris, Reference Norris2011), and a recent meta-analysis suggests that democratic dissatisfaction is increasing across multiple countries (Singh and Mayne, Reference Singh and Mayne2023).
A large body of research links citizens’ satisfaction with democracy to both policy outcomes and policy congruence. For instance, macroeconomic downturns, elevated crime rates, and negative experiences with public services are some key drivers of dissatisfaction (Claassen and Magalhães, Reference Claassen and Magalhães2023; Quaranta and Martini, Reference Quaranta and Martini2016; Weitz-Shapiro, Reference Weitz-Shapiro2008). Yet these objective performance indicators do not necessarily suffice to foster satisfaction among citizens with heterogeneous values and interests (Schmidt, Reference Schmidt2013). Scholars therefore emphasize policy congruence, arguing that satisfaction depends on the extent to which citizens’ preferences align with enacted policies (Soroka and Wlezien, Reference Soroka and Wlezien2010). Ideological proximity between voters and elites, both in general and on specific issues such as redistribution or EU integration, tends to increase satisfaction (Kim, Reference Kim2009; Stecker and Tausendpfund, Reference Stecker and Tausendpfund2016). Crucially, satisfaction is shaped more by individuals’ own alignment with elites (“egotropic” congruence) than by perceptions of elite–public alignment in the aggregate (“sociotropic” congruence) (Mayne and Hakhverdian, Reference Mayne and Hakhverdian2017). In other words, citizens evaluate the system’s performance based primarily on whether their personal preferences are realized, rather than whether the government accurately reflects the will of the broader majority. Yet policy congruence is only part of the story. Process congruence, the alignment between citizens and representatives on how representation should work, can also shape democratic evaluations (Wass and Nemcǒk, Reference Wass, Nemcǒk, Cotta and Russo2020). If popular control is a core component of democracy, then legitimate governance requires not only that politics produces outcomes citizens want, but that it operates according to their procedural preferences (Dahl, Reference Dahl1989). Theorists argue that when representatives act in ways that violate citizens’ normative expectations about how representation should function, the perceived legitimacy of the representational relationship itself is put at risk (Blumenau et al., Reference Blumenau, Wolkenstein and Wratil2025).
In this paper, we contribute to this debate by focusing on the longstanding “mandate-independence controversy” (Miller and Stokes, Reference Miller and Stokes1963; Pitkin, Reference Pitkin1967), which remains central to the literature today (see, e.g., Pedersen, Reference Pedersen2026). At its heart, this controversy is about representational style, which concerns the appropriate degree of autonomy representatives should exercise, and centres on two contrasting positions: trustee and (voter) delegate (Eulau et al., Reference Eulau1959; Pitkin, Reference Pitkin1967; Wass and Nemcǒk, Reference Wass, Nemcǒk, Cotta and Russo2020). The trustee perspective holds that representatives should rely on their own conscience and judgement, acting in what they believe to be the best interests of their constituents. The delegate perspective, by contrast, casts representatives as direct proxies of their constituents, bound to translate voters’ expressed preferences into political action. How, then, should we expect gaps in preferred representational style between citizens and their representatives to affect democratic satisfaction? Two main directions of incongruence are possible. Citizens may perceive their representatives as insufficiently responsive, acting as trustees when citizens would prefer delegates, or, conversely, as insufficiently autonomous, deferring to public opinion when citizens would prefer independent judgement. We argue that these two forms of incongruence operate through distinct psychological mechanisms, and that the relationship between process incongruence and dissatisfaction is therefore asymmetric. For citizens with a delegate preference, perceiving that representatives act as trustees produces a direct sense of exclusion, a feeling that their voice is structurally bypassed in favour of elite judgement. This maps onto the agency mechanism identified in the broader representation literature, whereby responsiveness functions as a democratic norm and its violation signals disrespect for popular sovereignty (Allen and Birch, Reference Allen and Birch2015; Norris, Reference Norris2011). Recent evidence reinforces this: citizens evaluate representatives as more effective at core parliamentary functions—scrutinizing government, setting the legislative agenda—when those representatives are seen as responsive to public wishes (Pedersen, Reference Pedersen2026). For citizens with a trustee preference, the mechanism is different. Such citizens do not simply want representatives to ignore public opinion; they want them to exercise informed, independent judgement on their behalf. For these citizens, a delegate-style MP who blindly follows public opinion may generate a different form of dissatisfaction: not a sense of exclusion, but a sense of recklessness. If those with a preference for trusteeship believe that public opinion is poorly informed, emotionally driven, or susceptible to manipulation, then a representative who defers to it may be seen as abdicating the very responsibility that legitimate representation requires (Barker and Carman, Reference Barker and Carman2012; Eulau et al., Reference Eulau1959). They may agree with Burke’s (Reference Burke1889 [1774], 95) classic admonition that “[y]our representative owes you, not his industry only, but his judgement; and he betrays instead of serving you, if he sacrifices it to your opinion.” Importantly, this preference for trustee-style representation need not reflect an elitist or pessimistic view of citizens’ political capacities. Some citizens simply prefer not to be burdened with the demands of ongoing political engagement and expect their representatives to exercise judgement in their stead or to abide by their party’s programme (Hibbing and Theiss-Morse, Reference Hibbing and Theiss-Morse2002).
Despite the continuous relevance of this debate, empirical research on process congruence remains sparse. Bengtsson and Wass (Reference Bengtsson and Wass2010) identify three assumptions that may help explain this neglect. First, the instructed-delegate model is assumed to dominate public opinion, even though early studies find more mixed views (Sigelman et al., Reference Sigelman, Sigelman and Walkosz1992). Second, voters may lack the capacity or interest to evaluate representational styles or to detect divergence from their own preferences. Third, MPs often lack autonomy from party leadership, raising doubts about whether mass–elite congruence on representation is politically consequential. As Bengtsson and Wass (Reference Bengtsson and Wass2010, 57) put it, studying representational alignment “may feel like an abstract academic exercise, untouched by the reality of a political process characterized by strong political parties.” However, despite these criticisms, existing research shows that citizens do not hold a uniform view on representation, nor are their preferences irrelevant to political attitudes (Mongrain et al., Reference Mongrain, Junius and Brack2025). Several studies suggest that process congruence—the alignment between citizens’ preferred representational style and how they perceive politicians to act—can substantially shape democratic satisfaction. In the United States, process gaps have been found to influence government approval more strongly than policy gaps (Hibbing and Theiss-Morse, Reference Hibbing and Theiss-Morse2002), and dissatisfaction with Congress has been linked to perceived misalignment in representational style (Kimball and Patterson, Reference Kimball and Patterson1997). In Belgium, André and Depauw (Reference André and Depauw2017) document a similar effect, albeit smaller than that of policy congruence. More recently, Bøggild (Reference Bøggild2020) has shown that when citizens who prefer trustee or delegate representation perceive politicians as prioritizing party loyalty, this mismatch in representational style systematically depresses trust in politicians across three countries, namely the United States, the United Kingdom, and Denmark.
Two key observations emerge from this literature. First, studies vary in how they measure process congruence: some focus on all MPs (Kimball and Patterson, Reference Kimball and Patterson1997), others on MPs from citizens’ preferred party (Hibbing and Theiss-Morse, Reference Hibbing and Theiss-Morse2002), and still others on government representatives (André and Depauw, Reference André and Depauw2017). These operational choices likely capture distinct mechanisms. Incongruence with one’s preferred party may be especially damaging, as it undermines the perceived legitimacy of the most acceptable political option. Misalignment with parliament as a whole matters because legislatures are expected to mirror the electorate’s diversity (Mansbridge, Reference Mansbridge1999), whereas incongruence with government may be particularly salient given its direct role in policymaking and its visibility in day-to-day decision-making. To account for this heterogeneity, our study employs three measures of process congruence: citizen–parliament (all MPs), citizen–government, and citizen–party. Second, it remains unclear whether citizens have well-defined preferences about representational style, whether they can correctly recognize elite preferences, and whether such gaps affect democratic attitudes (Bengtsson and Wass, Reference Bengtsson and Wass2010; Carman, Reference Carman2007; Wass and Nemcǒk, Reference Wass, Nemcǒk, Cotta and Russo2020). Given generally low levels of political knowledge (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996), many citizens may be unaware of how politicians represent them. Unlike policy congruence, which can be experienced through tangible outcomes, process congruence presupposes active engagement with political behaviour. Studies show that even politically informed citizens often lack detailed policy knowledge (Jensen and Zohlnhöfer, Reference Jensen and Zohlnhöfer2020; Rhodes et al., Reference Rhodes, Axsen and Jaccard2014; Soontjens, Reference Soontjens2021), raising doubts about the extent of mass awareness of representational styles and, by extension, the prevalence of process gaps among the mass public.
How can we theorize the effect of process congruence on democratic satisfaction in light of these studies? First, we hold that even when citizens have no precise knowledge about representatives’ process preferences, when representatives systematically act in ways that violate those expectations—for instance, exercising autonomous judgement when citizens expect responsiveness, or vice versa—this undermines the perceived legitimacy of the representational relationship itself, even if citizens cannot articulate the precise source of their dissatisfaction. The effect need not operate through conscious recognition of a representational style mismatch; it may also work through a more diffuse sense that the rules of the democratic game are not being respected (see Norris, Reference Norris2011). Process (in)congruence may therefore influence democratic satisfaction through two pathways: (1) conscious recognition of representational mismatches among politically attentive citizens, and (2) a more diffuse perception of procedural unfairness among less attentive citizens.
Second, we consider individual-level moderators that potentially condition the effect of process congruence on democratic satisfaction. Political interest represents a key motivational mechanism: citizens who are more attentive to politics are better positioned to perceive representational style mismatches and to translate those perceptions into evaluations of democratic quality. Prior work on policy congruence shows that political interest amplifies its effects, and we expect a similar mechanism in the case of process congruence (Costa, Reference Costa2021; Mayne and Hakhverdian, Reference Mayne and Hakhverdian2017; Stecker and Tausendpfund, Reference Stecker and Tausendpfund2016). Politically engaged individuals are more likely to hold structured opinions (Barber and Pope, Reference Barber and Pope2018; Bartle, Reference Bartle2000), to recognize representational styles, and to react to misalignment. Political interest may therefore not only condition the impact of process gaps but also directly predict democratic satisfaction and trust (Catterberg and Moreno, Reference Catterberg and Moreno2006; Chang, Reference Chang2018). A related but distinct mechanism is educational attainment. Education captures citizens’ cognitive resources and their ability to process complex political information. More educated individuals are more likely to possess the skills required to understand how representation operates in practice and to detect discrepancies between their own preferences and those of political elites. While political interest reflects motivation, education reflects capacity; both may independently condition how process congruence shapes democratic satisfaction. Building on these theoretical expectations, we derive the following hypotheses:
H1: The larger the process gap between an individual’s preferred representational style and that of elected representatives, the more likely they are to be dissatisfied with the state of democracy.
H2: Political interest moderates the relationship between process gaps and democratic satisfaction, such that the negative effect of process gaps is stronger among the politically interested.
H3: Educational attainment moderates the relationship between process gaps and democratic satisfaction, such that the negative effect of process gaps is stronger among more highly educated individuals.
Data and methods
Surveys
To test our hypotheses, we draw on repeated cross-sectional survey data collected by the Centre for Research and Studies in Sociology at Iscte–University Institute of Lisbon (CIES-Iscte) between 2008 and 2018. Specifically, we use the 2008, 2012, and 2016–2018 Portuguese Voters Surveys or PVS (Freire and Viegas, Reference Freire and Viegas2008a; Freire et al., Reference Freire, Viegas and Lisi2012; Freire et al., Reference Freire, Lisi and Tsatsanis2017a), and the 2008, 2012–2013, and 2016–2017 Portuguese MPs Surveys or PMS (Freire and Viegas, Reference Freire and Viegas2008b; Freire et al., Reference Freire, Viegas and Lisi2013; Freire et al., Reference Freire, Lisi and Tsatsanis2017b). These surveys of voters and MPs were fielded at comparable points in time following three legislative elections—those of 2005, 2011, and 2015. While data linking politicians’ and citizens’ representational preferences with democratic attitudes are rare, these Portuguese datasets provide a unique opportunity to investigate the link between process congruence and democratic (dis)satisfaction during a period of significant economic upheaval.
The case of Portugal
Portugal is a unitary, centralized state with one of Europe’s most stable multiparty systems. Four parties—the centre-left Socialist Party (PS), centre-right Social Democratic Party (PSD), left-wing Communist Party (PCP), and right-wing Democratic and Social Centre–People’s Party (CDS–PP)—have been continuously represented in parliament since the mid-1980s. Over time, Portugal has alternated between majority, minority, and coalition governments. Its closed-list proportional representation system minimizes MPs’ incentives to cultivate personal votes, making constituency representation less prominent than in many other systems (Borghetto et al., Reference Borghetto, Santana-Pereira and Freire2020). These institutional features make Portugal a least-likely case for detecting an effect of congruence between citizens and legislators on satisfaction with democracy: as already mentioned, if such an effect appears here, it is likely to be at least as strong in more open, candidate-centred systems. During the survey years, Portugal was deeply affected by the global financial and eurozone crises, culminating in a 2011 bailout negotiated with the “Troika” (the European Central Bank, European Commission, and International Monetary Fund) (Baer et al., Reference Baer, Dias and Duarte2013; Freire, Reference Freire2016). Despite severe austerity measures, Portugal’s party system remained comparatively stable, unlike in other bailout countries where volatility and fragmentation increased. Nonetheless, austerity deepened the legitimacy crisis of democratic institutions (Pequito Teixeira et al., Reference Pequito Teixeira, Tsatsanis and Belchior2016).
Dependent variable
We examine how process congruence affects satisfaction with democracy (SWD), measured using the standard survey item: “On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the way democracy works in Portugal?” As noted by Singh and Mayne (Reference Singh and Mayne2023), this indicator captures a mid-level form of political support, situated between diffuse regime support and evaluations of specific actors or institutions. To model the ordered nature of satisfaction with democracy, we estimate proportional odds (ordered logit) models in which the effect of process congruence is allowed to vary across representative targets (all MPs, government MPs, MPs from one’s preferred party) and moderators (political interest and educational attainment)Footnote 1 on pooled data from all three survey waves, including fixed effects for survey period. We merge the “fairly satisfied” and “very satisfied” categories as only one percent of respondents across all three survey waves reported being “very satisfied” with the way democracy works in Portugal. Figure 1 displays the distribution of responses. As the figure shows, satisfaction with democracy is generally low, especially in 2012, when roughly half of respondents reported being “not at all satisfied.” This sharp decline coincides with the aftermath of the global financial and eurozone crises and Portugal’s 2011 bailout (Freire et al., Reference Freire2014), a period marked by historically low levels of confidence in parliament and government.
Distribution of answers to the satisfaction with democracy item. “On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the way democracy works in Portugal?” Data from the 2008, 2012 and 2016–2018 (first wave) Portuguese Voters Surveys.
Process congruence
The process congruence measure is based on a series of three dilemmas (each implying a choice between two “principals”) used to establish voters’ or MPs’ preferred style of representation. The three dilemmas are the following:
-
Q1. How should a Member of Parliament vote in parliament (ideally): If the voters in his/her constituency have one opinion and his/her party takes a different position?
-
– MP should vote according to his/her party’s opinion.
-
– MP should vote according to his/her voters’ opinion.
-
-
Q2. How should a Member of Parliament vote in parliament (ideally): If his/her own opinion on an issue does not correspond with the opinion of the voters in his/her constituency?
-
– MP should vote according to his/her own opinion.
-
– MP should vote according to the opinion of the voters in his/her constituency.
-
-
Q3. How should a Member of Parliament vote in parliament (ideally): If his/her own opinion on an issue does not correspond with his/her party’s position?
-
– MP should vote according to his/her own opinion.
-
– MP should vote according to his/her party’s opinion.
-
According to their answers to these three items, respondents can be classified as having a preference for one of four representational styles (see Close et al., Reference Close, Legein and Little2024), namely (1) party delegates (respondents who gave priority to the party in items Q1 and Q3, i.e., who see MPs as primarily bound by the party line), (2) voter delegates (respondents who gave priority to voters in items Q1 and Q2, i.e., who expect MPs to follow constituency opinion), (3) trustees (respondents who gave priority to the representative in items Q2 and Q3, i.e., who view MPs as using their own judgement), or (4) politicos (respondents who do not consistently choose the same principal when it is present in different dilemmas and thus appear to favour a context-dependent mix of principals). The representational preferences from a total of 337 politicians (115 in 2008, 93 in 2012–2013, and 129 in 2016–2017)Footnote 2 and 5,079 citizens (1,134 in 2008, 1,048 in 2012, and 2,897 in 2016) were collected. The distribution of MPs per party is presented in Figure 2.
Distribution of MPs across major parties. Data from the 2008, 2012–2013, and 2016–2017 Portuguese MPs Surveys. In the 2016–2017 survey, one MP was classified as “Other party” (bringing the total to 337). BE = Bloco de Esquerda, CDS–PP = Centro Democrático e Social–Partido Popular, CDU = Coligação Democrática Unitária, PS = Partido Socialista, PSD = Partido Social Democrata.
Table 1 reports the distribution of representational style preferences among citizens and MPs in all three survey periods. As the table shows, citizens clearly favour voter delegates, whereas MPs are much more likely to express a preference for trusteeship or a party-delegate style of representation. In other words, most voters would like elected representatives to follow public opinion closely, while politicians tend to value their own independence and to recognize more strongly the demands and constraints of party discipline. The data indicate that MPs do not generally endorse the idea of binding mandates. Although the literature contains some conflicting evidence, previous research tends to find similar preference distributions (Andeweg and Thomassen, Reference Andeweg and Thomassen2005; Barker and Carman, Reference Barker and Carman2012; Bowler, Reference Bowler2017; Carman, Reference Carman2006; contra Ruiz Barreda and Rodríguez, Reference Ruiz Barreda and Rodríguez2019; Méndez-Lago and Martínez, Reference Méndez-Lago and Martínez2002; Mongrain et al., Reference Mongrain, Junius and Brack2025). Table 2 reveals little variation across parties in these preferences among voters: the largest difference is 5.2 percentage points in support for the party-delegate model, between BE (3.7%) and CDU (8.9%) voters. Among politicians, only CDU (PCP–PEV) MPs stand out, being considerably more likely to prefer the party-delegate role (84.6%) over trusteeship (11.5%). In all other parties, MPs prefer trusteeship over the party-delegate role. Table 3 further shows that Portuguese citizens do not appear to favour different representational styles depending on whether they supported a winning or a losing party. The absence of systematic differences between winners and losers helps alleviate concerns that citizens might be “opportunistic” or biased in their judgements about the fairness or desirability of procedures, as suggested by previous research (Doherty and Wolak, Reference Doherty and Wolak2012; Esaiasson and Öhberg, Reference Esaiasson and Öhberg2020; Esaiasson et al., Reference Esaiasson2019).
Distribution of representational style preferences among citizens and MPs
Table 1. Long description
The table shows the distribution of representational style preferences among citizens and MPs for the years 2008, 2012-2013, 2016-2017, and all years combined. The preferences are categorized into voter-delegate, party-delegate, trustee, and politico. Row 1: Voter-delegate, 2008 Citizens 73.19 percent, MPs 14.78 percent; 2012-2013 Citizens 72.14 percent, MPs 11.83 percent; 2016-2017 Citizens 68.76 percent, MPs 27.91 percent; All Citizens 70.45 percent, MPs 18.99 percent. Row 2: Party-delegate, 2008 Citizens 10.41 percent, MPs 30.43 percent; 2012-2013 Citizens 7.92 percent, MPs 24.73 percent; 2016-2017 Citizens 3.87 percent, MPs 26.36 percent; All Citizens 6.16 percent, MPs 27.30 percent. Row 3: Trustee, 2008 Citizens 11.90 percent, MPs 45.22 percent; 2012-2013 Citizens 13.84 percent, MPs 58.06 percent; 2016-2017 Citizens 22.82 percent, MPs 38.76 percent; All Citizens 18.53 percent, MPs 46.29 percent. Row 4: Politico, 2008 Citizens 4.50 percent, MPs 9.57 percent; 2012-2013 Citizens 6.11 percent, MPs 5.38 percent; 2016-2017 Citizens 4.56 percent, MPs 6.98 percent; All Citizens 4.86 percent, MPs 7.42 percent.
Note: Data from the 2008, 2012 and 2016–2018 (first wave) Portuguese Voters Surveys, and the 2008, 2012–2013, and 2016–2017 Portuguese MPs Surveys.
Distribution of representational style preferences among citizens and MPs by party identification
Table 2. Long description
A table comparing representational style preferences among citizens and MPs by party identification. The table has 6 columns and 5 rows, including headers. Column headers are Preference, Citizens (%), and MPs (%). Row labels are Voter-delegate, Party-delegate, Trustee, and Politico. Under Citizens (%), the subheaders are BE, CDS-PP, CDU, PS, PSD, and None. Under MPs (%), the subheaders are BE, CDS-PP, CDU, PS, and PSD. Row 1: Voter-delegate, BE 74.69, CDS-PP 70.55, CDU 69.40, PS 70.33, PSD 70.73, None 70.14, BE 26.92, CDS-PP 18.52, CDU 3.85, PS 17.69, PSD 22.73. Row 2: Party-delegate, BE 3.70, CDS-PP 7.53, CDU 8.90, PS 8.26, PSD 7.38, None 4.62, BE 15.38, CDS-PP 29.63, CDU 84.62, PS 20.41, PSD 25.45. Row 3: Trustee, BE 17.59, CDS-PP 19.18, CDU 17.08, PS 15.87, PSD 17.88, None 20.38, BE 57.69, CDS-PP 37.04, CDU 11.54, PS 52.38, PSD 45.45. Row 4: Politico, BE 4.01, CDS-PP 2.74, CDU 4.63, PS 5.54, PSD 4.02, None 4.86, BE 0.00, CDS-PP 14.81, CDU 0.00, PS 9.52, PSD 6.36.
Note: Data from the 2008, 2012 and 2016–2018 (first wave) Portuguese Voters Surveys, and the 2008, 2012–2013, and 2016–2017 Portuguese MPs Surveys.
Distribution of representational style preferences among citizens by winner/loser status
Table 3. Long description
The table presents the distribution of representational style preferences among citizens and MPs for the years 2008, 2012-2013, 2016-2017, and overall. It includes preferences for voter-delegate, party-delegate, trustee, and Politico styles. The table has 5 rows and 9 columns. Row 1: Voter-delegate, 73.63, 71.24, 72.20, 71.63, 68.53, 69.51, 70.27, 70.50. Row 2: Party-delegate, 9.82, 11.71, 7.73, 7.09, 3.93, 4.65, 5.84, 7.62. Row 3: Trustee, 11.78, 13.04, 13.61, 17.73, 23.27, 20.41, 19.05, 17.29. Row 4: Politico, 4.77, 4.01, 6.46, 3.55, 4.27, 5.43, 4.84, 4.59. The table shows the percentage preferences for each style across different periods and groups.
Note: Data from the 2008, 2012 and 2016–2018 (first wave) Portuguese Voters Surveys.
To measure the process gap, we follow André and Depauw (Reference André and Depauw2017) by relying on two dilemma items (i.e., Q1 and Q2) that gauge how MPs should deal with diverging interests between voters and either an MP’s own judgement or the position of the MP’s party. This approach differs from measures of perceived process congruence, which ask citizens to evaluate how well politicians are currently performing their representational role (Hibbing and Theiss-Morse, Reference Hibbing and Theiss-Morse2002). Perceived congruence risks conflating normative expectations with performance evaluations that are likely shaped by prior attitudes such as political trust, partisan affect, or general satisfaction with democracy. By measuring congruence in normative preferences rather than perceptions of elite behaviour, we are able to assess whether citizens and representatives share the same representational vision without relying on potentially endogenous judgements about political performance. Based on these two dilemmas, a three-point scale was constructed, “ranging from a top-down to a bottom-up conception of the representative process” (André and Depauw, Reference André and Depauw2017, 385). A score of 2 was assigned to respondents who indicated that the opinion of voters should be prioritized in both scenarios, a score of 1 when the opinion of voters was chosen only once, and a score of 0 when the opinion of voters was not chosen in either scenario. Process congruence was then measured between (1) citizens and all elected members of parliament (citizen–MPs hereafter), (2) citizens and government MPs (citizen–government), and (3) citizens and MPs belonging to their preferred party (citizen–party). As in André and Depauw (Reference André and Depauw2017), process congruence was operationalized as the absolute difference between each voter’s position on the three-point representational preference scale and the average MP position, divided by the standard deviation of MPs’ distributions within parliament, government, or their party.Footnote 3 As such, congruence is conceptualized as a one-to-many relationship (see Golder and Stramski, Reference Golder and Stramski2010, 91, fn. 1). In other words, we obtain a measure of how the representational preference of each individual voter (one “principal”) is congruent with that of multiple politicians (many “agents”): (1) all politicians in parliament, (2) politicians in government, and (3) politicians from the citizen’s preferred party.
In order to measure citizen–MPs congruence and citizen–government congruence in times of coalition government, we adapted André and Depauw’s (Reference André and Depauw2017) equation to account for the size of parties in parliament or inside the governmental coalition by weighting the mean position of MPs from each party by its seat share as recorded by ParlGov (Döring and Manow, Reference Döring and Manow2024).Footnote 4 Therefore, in those instances, process congruence (see Equation 3) was measured as the absolute difference between the weighted average of MPs’ average positions on the representation scale within each party (see Equation 1) and each voter’s position, divided by the weighted standard deviation of MPs’ average positions within each party (see Equation 2).
$${\overline {{\rm{MP}}} _w} = {{\sum\nolimits_{i = 1}^n {\mkern 1mu} {w_i}{{\overline {{\rm{MP}}} }_i}} \over {\sum\nolimits_{i = 1}^n {\mkern 1mu} {w_i}}}$$
where the numerator is the sum of the weights (
${w}_{i}$
) times the average MP position (
${\overline{\mathrm{M}\mathrm{P}}}_{i}$
) for party
$i$
; and the denominator is the sum of the weights.
$$s{d_w} = \sqrt {{{\sum\nolimits_{i = 1}^N {w_i}{{\left( {{{\overline {{\rm{MP}}} }_i} - {{\overline {{\rm{MP}}} }_w}} \right)}^2}} \over {(N' - 1)\sum\nolimits_{i = 1}^N {w_i} \times 1/N'}}}$$
where the numerator is the sum of the weights (
${w}_{i}$
) times the squared difference between the average MP’s position (
${\overline{\mathrm{M}\mathrm{P}}}_{i}$
) for party
$i$
and the weighted mean (
${\overline{\mathrm{M}\mathrm{P}}}_{w}$
); and the denominator is the number of weights (
$N\mathrm{\text{'}}$
) minus 1 times the sum of the weights divided by the number of weights.
$$PG = {{\left| {V - {{\overline {{\rm{MP}}} }_w}} \right|} \over {s{d_w}}}$$
where the numerator is the absolute difference between the voter’s position (
$V$
) and the weighted average of MPs’ average positions within each party (
${\overline{\mathrm{M}\mathrm{P}}}_{w}$
); and the denominator is the weighted standard deviation of MPs’ average positions within each party (
$s{d}_{w}$
).
Lower values of process congruence indicate greater correspondence between the views of voters and those of politicians. To ease interpretation, we inverted the measure so that higher values indicate greater congruence. For the same purpose, congruence scores were also rescaled from 0 to 1. Figure 3 shows the distribution of process congruence scores.Footnote 5 We can observe that citizens are more congruent with their preferred party than with the government and all MPs. When we look at congruence scores according to voters’ preferred style of representation in Figure 4, we can see that congruence scores are much lower among citizens who prefer politicians to act as voter delegates over other representational styles. This is not surprising considering the important mismatch already apparent in Table 1.
Distribution of process congruence scores. Data from the 2008, 2012 and 2016–2018 (first wave) Portuguese Voters Surveys, and the 2008, 2012–2013, and 2016–2017 Portuguese MPs Surveys.
Average congruence score by congruence measure and preferred style. To assess whether average process-congruence scores differed across preferred representational styles, we estimated separate one-way ANOVA models for each congruence measure (citizen–MPs, citizen–government, citizen–party), with the individual-level congruence score as the dependent variable and preferred style as a four-category factor. We then conducted Tukey’s HSD post hoc tests to adjust pairwise comparisons for multiple testing and used compact letter displays to summarize the results: within each congruence measure, bars that share a letter do not differ significantly from one another, whereas bars with different letters differ significantly at p < .05. Data from the 2008, 2012 and 2016–2018 (first wave) Portuguese Voters Surveys, and the 2008, 2012–2013, and 2016–2017 Portuguese MPs Surveys.
Moderating effects
As discussed earlier, we expect politically interested citizens to be most affected by the extent of the process gap. Because they pay more attention to the work and discourse of politicians, interested voters should theoretically have a better sense of the nature of the representational linkage between elected representatives and the electorate. Interest in politics was measured on a four-point scale ranging from “not at all interested” to “very interested.” In addition, we consider educational attainment as a proxy for citizens’ cognitive resources and their ability to process political information. Highly educated individuals are more likely to understand how representation operates in practice and to detect discrepancies between their own preferences and those of political elites. Education was operationalized as a binary indicator distinguishing respondents with and without a university degree. To assess moderation, we estimate a single model per representative target in which each congruence dimension is interacted simultaneously with both political interest and educational attainment. Entering both sets of interactions in the same specification ensures that the moderating role attributed to political interest is not confounded by educational attainment—and vice versa—given that the two are generally correlated (Le and Nguyen, Reference Le and Nguyen2021). Estimating each interaction in isolation, with the other characteristic entered only additively, would leave open the possibility that an apparent interest interaction merely reflects education (or the reverse).
Covariates
Over the past few decades, research has investigated the impact of various individual- as well as institutional-level variables on citizens’ support for and satisfaction with democracy, and their confidence in political institutions. In this regard, one of the most robust findings in political science is probably the existence of a winner–loser gap in satisfaction with democracy (Anderson et al., Reference Anderson2005; Blais and Gélineau, Reference Blais and Gélineau2007; Best and Seyis, Reference Best and Seyis2021; Henderson, Reference Henderson2008). Voters who supported a winning party or candidate tend to be more satisfied with democracy and more trusting of government than election losers (Dassonneville and McAllister, Reference Dassonneville and McAllister2020; Hooghe, Reference Hooghe, Oscarsson and Holmberg2020; Keele, Reference Keele2005). Therefore, all models control for respondents’ winner/loser status in the 2005, 2011, and 2015 Portuguese legislative elections. We use party identification to distinguish winners and losers instead of reported vote or vote intention as the former signals a stronger commitment and affective attachment to a party than voting behaviour. Additionally, the 2008 and 2012 PVS do not make a clear distinction between respondents who would not vote, would cast a blank/invalid ballot, did not yet know which party they would support, or would refuse to answer.
Second, we account for perceived economic conditions. Both democratic satisfaction and process preferences may be shaped by individuals’ sense of material security (Allen and Birch, Reference Allen and Birch2015), and failing to include these perceptions could introduce bias into our estimates. Prior research demonstrates that both objective macroeconomic indicators and subjective economic evaluations are strong predictors of democratic satisfaction and political trust (Chanley et al., Reference Chanley, Rudolph and Rahn2000; van der Meer, Reference van der Meer and Eric2017; van Erkel and van der Meer, Reference van Erkel and Meer2016). Accordingly, all models incorporate respondents’ assessments of the national economy. These evaluations were originally captured on a five-point scale ranging from “very bad” to “very good.” However, across all three voter surveys, fewer than three percent of respondents rated the economy as “good” or “very good.” To address this skew, we recoded the variable into a binary format: responses indicating the economy was “very bad” or “bad” (77% of respondents) were coded as 0, while all other responses (“neither good nor bad,” “good,” or “very good”; 23%) were coded as 1.
Third, we also include individual ideological orientations as control variables. The literature shows that voters who hold more radical or extreme ideological positions tend to be more distrustful of the political establishment compared to moderate voters (André and Depauw, Reference André and Depauw2017; Kutiyski et al., Reference Kutiyski, Krouwel and van Prooijen2021; but see Krouwel et al., Reference Krouwel2017). To account for ideological extremity among voters (and potential differences between individuals located on the left and on the right of the political spectrum), we use the standard 0–10 left-right scale along with its quadratic term. In other words, we can expect to find an inverted U-shaped relationship between ideology and democratic satisfaction.
Fourth, as noted by previous research, the quality of representation is reflected in the degree of issue and ideological congruence between voters and elected representatives. Policies and programmes that cater to the interests and preferences of citizens should be conducive to greater satisfaction (Ferland, Reference Ferland2017, Reference Ferland2021; Kim, Reference Kim2009): the system is delivering the policies citizens want and politicians appear to be keeping their campaign promises or to be actively listening to the public. To account for this, we control for policy congruence using a set of policy and issue-opinion items asked to both citizens and elected politicians in each pair of surveys. In total, we use 19 identical items from the 2008 surveys, 15 from the 2012–2013 surveys, and 16 from the 2016–2017 surveys. These items include statements such as “For growth it is more important to decrease taxes on business than to increase public spending,” “Stronger measures should be taken to protect the environment,” or “Same-sex marriages should be prohibited by law.” We also control for ideological congruence using citizens’ and politicians’ self-placement on the left–right scale. Following the same approach used for process congruence, we compute the relative distance between each voter’s position—on each policy item and on the left–right scale—and the (weighted) mean position of all MPs in parliament, MPs in government, and MPs belonging to the voter’s preferred party. Therefore, all three congruence dimensions (process, policy, ideology) are included simultaneously within each model specification, ensuring that the estimated effect of each dimension reflects its association with democratic satisfaction net of the others.Footnote 6 To account for the possibility that the effects of policy and ideological congruence are also conditional on individual-level characteristics, we interact each of these measures with both political interest and educational attainment as well. This ensures that any moderating effects attributed to process congruence are not simply capturing a more general pattern whereby politically attentive or more cognitively sophisticated citizens are more responsive to congruence across substantive and ideological dimensions of representation.
Fifth, because process congruence is by construction correlated with citizens’ own representational preferences—congruence scores are systematically lower among the large group of citizens who favour a voter-delegate style (see Figure 4)—all models additionally control for respondents’ preferred representational style, entered as a set of indicators for the four styles (voter-delegate, party-delegate, trustee, and politico, with trustee as the reference category). Entering citizens’ baseline process preferences directly, alongside the congruence measures, serves two purposes: it ensures that any association attributed to process congruence is not confounded with citizens’ underlying orientation toward representation—a generalized demand for responsiveness or autonomy that is itself correlated with congruence—and it lets us examine how those baseline preferences relate to democratic satisfaction in their own right.
Finally, all models include controls for respondents’ age (in years) and sex (0 = female, 1 = male). Existing research reports mixed and context-dependent age and gender differences in democratic satisfaction and support (see, for example, Claassen and Magalhães, Reference Claassen and Magalhães2023; Foa et al., Reference Foa2020; Hansen and Goenaga, Reference Hansen and Goenaga2021; Romero-Vidal and Talukder, Reference Romero-Vidal and Talukder2025; Valgarðsson and Devine, Reference Valgarðsson and Devine2022; Zilinsky, Reference Zilinsky2019); accordingly, we do not formulate directional expectations for either variable.Footnote 7
Results
Tables with the full regression outputs are provided in Section C of the online Appendix. To calculate predicted probabilities, values for continuous and categorical covariates are set at their mean and modal category, respectively.Footnote 8 Turning first to the main effects, process congruence does not appear to be reliably associated with satisfaction with democracy once policy and ideological congruence are taken into account. For congruence between citizens and all MPs and between citizens and government MPs, the point estimates are in the expected direction—higher congruence is associated with a somewhat lower probability of being fully dissatisfied, but these associations are not statistically significant. For congruence between citizens and their preferred party, the point estimate runs in the opposite direction and is likewise not statistically significant. Figure 5 plots how the predicted probabilities of each satisfaction category vary with process congruence for the three targets. To the extent that congruence is associated with satisfaction, any movement appears to occur primarily between the lowest and highest satisfaction categories rather than through changes in the intermediate “not very satisfied” category. In other words, increases in satisfaction associated with greater congruence seem to correspond mainly to a decline in the probability of being fully dissatisfied, with little apparent effect on the likelihood of being not very satisfied. Given the width of the confidence intervals, however, these descriptive patterns should be interpreted as suggestive rather than as evidence of established effects.
Predicted probabilities of level of satisfaction with democracy given process congruence scores. Dashed lines indicate 95% confidence intervals. Data from the 2008, 2012 and 2016–2018 (first wave) Portuguese Voters Surveys, and the 2008, 2012–2013, and 2016–2017 Portuguese MPs Surveys. Estimates are from Tables C1–C3 in the Appendix.
Figure 6 summarizes the estimated change in each satisfaction category when moving from the minimum to the maximum of each congruence dimension, together with 95% confidence intervals. For process congruence, these estimated changes are small and, in every case, statistically insignificant. Moving across the observed range of citizen–MPs congruence (a shift of 3.02 standard deviations) is associated with a 5.5 percentage-point lower probability of being “not at all satisfied” and a 4.5-point higher probability of being “fairly/very satisfied”; the corresponding figures for citizen–government congruence (a shift of 3.06 standard deviations) are 3.7 and 3.3 points. These estimates are consistent with the direction predicted by H1, but the considerable uncertainty surrounding them means that a null association cannot be ruled out. For citizen–party congruence (a shift of 4.86 standard deviations), by contrast, the point estimates run counter to H1: an 8.5-percentage-point increase in the probability of being “not at all satisfied” and a 9.1-percentage-point decrease in the probability of being “fairly” or “very satisfied.” These estimates also do not reach statistical significance. By contrast, policy congruence is strongly and robustly related to satisfaction. Moving across its observed range lowers the probability of being “not at all satisfied” by 65.5, 55.4, and 27.2 percentage points for the citizen–MPs, citizen–government, and citizen–party targets respectively, and raises the probability of being “fairly/very satisfied” by 30.9, 31.7, and 18.4 points—roughly an order of magnitude larger than the process-congruence estimates, with confidence intervals well away from zero. Like process congruence, ideological congruence has a small and statistically insignificant effect across all targets.
Percentage-point change (with 95% CIs) in satisfaction with democracy when moving from minimum to maximum congruence by congruence type and representative target. Standard-deviation shift from minimum to maximum congruence indicated in parentheses. Estimates from Model 1.2 (see Table C1 in the Appendix), Model 2.2 (see Table C2), and Model 3.2 (see Table C3).
Figure 6. Long description
The image contains nine bar graphs arranged in a 3x3 grid. Each graph represents the percentage-point difference in satisfaction with democracy when moving from minimum to maximum congruence by congruence type and representative target. The graphs are labeled as Citizen-MPs, Citizen-Government, and Citizen-Party. The horizontal axis of each graph represents different congruence types: Process, Policy, and Ideology. The vertical axis represents the percentage-point difference, ranging from -100 to 50. Each bar graph includes error bars indicating 95% confidence intervals. The graphs show varying levels of satisfaction with democracy across different congruence types and representative targets. For example, in the Citizen-MPs graph, the Policy congruence type shows a significant negative difference, while the Process and Ideology types show smaller differences. Similar patterns are observed in the other graphs, with notable differences in the magnitude and direction of the percentage-point changes.
These results temper the expectation that process congruence operates as a distinct and consequential dimension of representation on a par with policy congruence. Once policy and ideological congruence are held constant, the data do not provide robust evidence that process congruence exerts an independent influence on satisfaction with democracy: its estimated effects are small, inconsistent in sign across targets, and statistically insignificant. This does not establish that process congruence is irrelevant—the absence of a statistically significant association is not proof that the true effect is exactly zero, and the estimates for two of the three targets are in the theoretically expected direction—but the evidence here does not support characterizing process congruence as a stable or independent driver of democratic satisfaction. What the models show clearly is that policy congruence is the dominant correlate of satisfaction.
We now turn to whether the relationship between process congruence and democratic satisfaction—modest as it is on average—varies with political interest or educational attainment, as H2 and H3 anticipate. In other words, perceived (mis)alignment in representational preferences may still meaningfully influence democratic satisfaction among more politically engaged or sophisticated voters. Since we estimated ordinal logistic regression models, the statistical significance and direction of the interaction coefficients cannot be interpreted directly (Mustillo et al., Reference Mustillo, Lizardo and McVeigh2018). We therefore computed tests of second differences (see Mize, Reference Mize2019). First differences capture the change in the predicted probability of an outcome when process congruence moves from its minimum to its maximum, with the moderator held at a specified level. Second differences compare these changes across moderator levels, thereby testing whether the estimated effect of process congruence varies with political interest or education. Since we cannot present every possible combination here, we highlight the most informative and representative results. Graphical representations of these interactions can be found in Section D of the Appendix.
The estimated effect of process congruence does not seem to vary significantly with political interest. Regarding the probability of being fairly or very satisfied with democracy, moving from minimum to maximum citizen–MPs congruence increases the predicted probability by 0.6 percentage points among respondents who are not at all politically interested and by 3.3 points among those who are very interested. Therefore, the resulting second difference is only 2.7 percentage points. The corresponding second differences are 0.4 points for citizen–government congruence and −2.0 points for citizen–party congruence. None of these second differences is statistically significant. More generally, none of the second-difference tests conducted across the three satisfaction categories and three representative targets is statistically significant. Accordingly, there is no evidence that the effect of process congruence systematically strengthens with political interest, and H2 is not supported. Education produces clearer differences, but they run contrary to H3. For citizen–MPs congruence, moving from minimum to maximum congruence increases the predicted probability of being fairly or very satisfied by 4.1 percentage points among respondents without a university degree but decreases it by 5.9 points among university graduates. The resulting second difference is −10.0 percentage points and is statistically significant. The pattern is mirrored for the probability of being not at all satisfied: greater citizen–MPs congruence reduces this probability by 4.8 points among respondents without a university degree but increases it by 4.4 points among university graduates, producing a statistically significant second difference of 9.2 points. A similar pattern appears for citizen–government congruence. Moving from minimum to maximum congruence increases the probability of being fairly or very satisfied by 3.3 percentage points among respondents without a university degree but decreases it by 8.3 points among university graduates, yielding a statistically significant second difference of −11.6 points. For the probability of being not at all satisfied, the corresponding effects are a 4.1-point decrease and a 5.8-point increase, producing a statistically significant second difference of 10.0 points. By contrast, none of the citizen–party second differences reaches conventional statistical significance. Thus, where education significantly moderates the relationship, greater process congruence is associated with comparatively more favourable changes among respondents without a university degree and less favourable changes among university graduates. In other words, the significant congruence
$\times$
education interactions indicate that the gap in predicted probabilities between university- and non-university-educated respondents narrows as congruence increases (see Figure 7), rather than widens as hypothesized in H3.
Interaction between process congruence and political interest/university education. Because each congruence dimension is now interacted with political interest and university education within a single model, both rows of panels are drawn from the same combined specifications, shown in the upper panels (political interest) and lower panels (university education). Estimates from Model 1.2 (see Table C1 in the Appendix) for citizen–MPs congruence, Model 2.2 (see Table C2) for citizen–government congruence, and Model 3.2 (see Table C3) for citizen–party congruence. Estimates for the intermediate SWD category (“not very satisfied”) are omitted for space.
Across all 63 second-difference comparisons for process congruence—54 involving political interest and nine involving education—only the four education comparisons described above reach conventional significance. The tests thus indicate no moderating role for political interest and a limited, target- and outcome-specific role for education that runs contrary to H3. The full second-difference tables, including the corresponding results for policy and ideological congruence, are reported in Section E of the Appendix.
Taken together, the findings offer very limited support for H1. Once policy and ideological congruence are accounted for, moving from minimum to maximum process congruence is generally associated with small changes in democratic satisfaction that vary in direction across representative targets and are statistically indistinguishable from zero. The estimates for the citizen–MPs and citizen–government targets tend to point in the hypothesized direction on average, whereas those for citizen–party congruence point in the opposite direction. H2 receives no support: there is no evidence that politically interested citizens respond more strongly to changes in process congruence. H3 is also unsupported, although not because education is entirely irrelevant. Rather, where the effect of process congruence differs significantly by education, the relationship is comparatively more favourable among respondents without a university degree than among university graduates—the reverse of what H3 predicts. Due to space limitations, the effects of the control variables are presented in Section G of the Appendix.
These results concern whether congruence between citizens and political elites is associated with democratic satisfaction. A related but analytically distinct question is whether citizens’ own representational preferences are associated with satisfaction independently of the extent to which those preferences align with elite behaviour. Because citizens’ preferred representational style forms one component of the process-congruence measures, all models include representational preference directly alongside process, policy, and ideological congruence. While the average relationship between process congruence and satisfaction is weak and inconsistent across representative targets, “raw” representational preferences appear to be associated with satisfaction. Citizens who favour a voter-delegate style are, on average, less satisfied with democracy than those who favour a trustee style; this difference is statistically significant in the citizen–government and citizen–party models and points in the same direction in the citizen–MPs model. Section C3 of the Appendix examines this relationship in greater detail.
Conclusion
We find that citizens and politicians in Portugal—as elsewhere (see Mongrain et al., Reference Mongrain, Junius and Brack2025)—vary considerably in their preferences regarding representational styles. Politicians claim greater autonomy from public opinion than voters would like to afford them. Most parliamentarians prioritize their own judgement over voters’ sentiments or recognize the need to toe the party line. This is not surprising as the Portuguese proportional electoral system with closed lists does little to encourage MPs to cultivate close links with their constituents. Indeed, as explained by Belchior et al. (Reference Belchior, Tsatsanis and Teixeira2016, 279), “[p]arties control the electoral process […] and parliamentary activity (which is highly disciplined).” Consequently, we find very low process congruence between the mass of voters favouring delegate-style representation and a legislative body favouring personal autonomy (trustees) and loyalty to party (party delegates).
The main aim of this paper was to assess whether this mismatch between the representational preferences of citizens and elites matters for their satisfaction with democracy. While there are strong arguments in the literature for the importance of process congruence in shaping political attitudes, there is also room for suspicion. Are voters aware of, and do they care about, the process gap? Our evidence is mixed. The gap between citizens’ and elites’ representational preferences is, at most, weakly related to satisfaction with democracy once policy and ideological congruence are accounted for. Yet citizens are not indifferent to representation as such: their own representational orientations are related to how satisfied they are, with delegate-minded citizens somewhat less satisfied than trustee-minded ones. Like Carman (Reference Carman2006), we are not suggesting that voters sit down with a hot cup of tea or coffee to reflect on the nature of representation in their country. In the words of Achen and Bartels (Reference Achen and Bartels2017, 9), most citizens are “too busy with their lives” to devote time and energy to such questions. Nonetheless, in a country where voters and politicians have strongly diverging viewpoints on the matter, we can reasonably expect that informational cues about incongruence percolate throughout the electorate. Voters might feel that they are not being heard or that they have few channels through which to reach, let alone influence, their representatives. To the best of our knowledge, this is the first study to consider the potential moderating effect of interest and education on the connection between process congruence and democratic satisfaction.
Once policy congruence, ideological congruence, and respondents’ own representational preferences are taken into account, we find little robust evidence that satisfaction with democracy increases with process congruence: the estimated associations are small, inconsistent in direction across representative targets, and never reach conventional thresholds of statistical significance. Contrary to expectations, we also do not find that highly politically interested or highly educated citizens are more satisfied (dissatisfied) with democracy when process congruence increases (decreases), nor do we find that the least politically interested or less-educated citizens are the ones least affected in their democratic satisfaction by changes in process congruence.
That neither political interest nor education strengthens the process-congruence relationship in the way we anticipated is itself worth dwelling on. One reading of the broader literature would predict that process congruence should matter most for the politically attentive and cognitively resourced, who are best placed to perceive representational mismatches. We find no such pattern—but, in our data, we also find little evidence of a robust process-congruence effect to be moderated in the first place. Two interpretations are consistent with this. The first is that mismatches in representational style simply do not weigh heavily on citizens’ evaluations of democracy, at least relative to policy congruence. The second is that any effect operates through a diffuse, affective channel (see Norris, Reference Norris2011)—a generalized sense that one’s voice does not matter—that does not depend on close attention to politics and that our observational design, with policy and ideological congruence held constant, may be poorly suited to detect. The Portuguese context, in which roughly seven in ten citizens favour delegate-style representation while fewer than two in ten MPs do, is one of extreme and persistent misalignment; that even this degree of mismatch is not clearly reflected in satisfaction, net of policy and ideology, is a finding worth taking seriously. Distinguishing between these interpretations is a task for future, ideally experimental, research.
We tested for the effects of different measurements of congruence, evaluating citizens’ match in representational style with all parliamentarians, those belonging to government parties, and those from their own preferred party. We show that the three measures of process congruence are highly correlated and do not yield widely different results, which lends confidence that our conclusions do not hinge on the choice of representative target. We refrain, however, from reading substantive significance into the similarity of these estimates: because the process-congruence effects are uniformly small and surrounded by wide uncertainty, they offer little basis for claims about whether citizens attach greater importance to congruence with parliament as a whole, government MPs, or MPs from their preferred party. Establishing whether individual MPs shape voter perceptions in closed-list systems will require designs better able to isolate such effects.
While our findings offer important insights into democratic satisfaction, several limitations of this study must be acknowledged. First, the universe of possible explanations for inconsistencies in the literature is vast and larger than what we can test here. Barker and Carman (Reference Barker and Carman2012), for instance, raise the possibility that voters might ask for different and even diverging styles of representation depending on the type of issue or policy domain being considered. Moreover, elected representatives might simply provide explanations for departing from citizens’ preferred representational style. Explaining why and when politicians deviate from citizens’ preferences has been shown to be positively related to democratic satisfaction across partisan lines (Robison, Reference Robison2024). A congruence gap does not need to lead to democratic disenchantment when politicians convincingly explain why they diverge from citizens’ process and policy preferences (Mansbridge, Reference Mansbridge2003; Wolkenstein and Wratil, Reference Wolkenstein and Wratil2021). One must also acknowledge that congruence in the representational preferences of citizens and MPs is not necessarily reflected in the actual legislative behaviour of elected representatives; what we are effectively looking at are normative conceptions of representation (von Schoultz and Wass, Reference von Schoultz and Wass2016). Second, our study relies on observational measures, which limits causal inference. Reverse causality cannot be ruled out: citizens who are dissatisfied with democracy for other reasons may be more likely to perceive process mismatches, rather than mismatches driving dissatisfaction. A further limitation is that we cannot directly assess how accurately citizens perceive the process preferences of political elites. Future research would benefit from a discrete choice or conjoint experiment presenting respondents with candidates varying simultaneously on both policy positions and representational style. Such a design would allow researchers to estimate the relative weight citizens place on process versus policy alignment and would overcome the perception problem inherent in observational approaches. Bøggild’s (Reference Bøggild2020) experiments, which hold policy outcomes constant while varying representational style, offer such an approach. Furthermore, as with any study focusing on a single country, external validity may be a concern. Future research should consider how the mismatch in representational preferences between citizens and political actors affects voters’ attitudes and behaviours across multiple countries encompassing various electoral systems.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S175577392610071X.
Data availability statement
Replication code can be found in Harvard Dataverse: https://doi.org/10.7910/DVN/EWN6ZQ.
Acknowledgements
First, we wish to thank the anonymous reviewers for their excellent suggestions and comments. We are also grateful to members of the Media, Movements and Politics research group at the University of Antwerp and participants at the Future of Electoral Democracy Conference in Montreal for their feedback on previous versions of this paper.
Financial support
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Competing interests
None.
Ethical standards
This study involved secondary analysis of existing data and did not involve the direct recruitment of human participants.
Use of AI
Generative AI tools were used during manuscript preparation for limited language editing and stylistic refinement.
Open access
