In recent decades, many authors have underlined the erosion of political support (Dalton, 2004) as well as the decline in citizens’ trust in governments and institutions (Bedock, 2017; Dalton, 1999, 2004; Norris, 2011; Pharr et al., 2000). In addition to distrust towards specific actors of the political system and its institutions, other scholars argue that support for democracy as a political regime is declining (Foa et al., 2020; Foa and Mounk, 2017), which raises questions about to what extent representative democracies are able to resist backsliding.

In such a context, this research article aims to investigate and better understand citizens’ dissatisfaction with their political system and asks the following research question: “how do citizens evaluate the political system?.” Building on existing scholarship, this research proposes an alternative way to measure citizens’ evaluation of the political system by looking at it from a different perspective.

The political system’s legitimacy is often conceptualized as citizens’ political support and is measured by indicators of “diffuse” and “specific” support such as: support for democracy as a regime (Foa and Mounk, 2017), satisfaction with democracy, and political trust with different meanings and implications. Among these approaches, one of the most widespread measures of the political system’s legitimacy is the indicator of “Satisfaction with democracy” (SWD). Although it is very useful and relevant for the comparative study of political support, this research argues that it remains unsatisfactory to grasp what citizens are dissatisfied with. Thus, to better understand citizens’ dissatisfaction with the political system this piece proposes to mobilize the classic systemic approach developed by Easton (1965), consolidated by Scharpf (1999) and Schmidt (2013), which distinguishes between input, throughput and output legitimacy. However, instead of assessing the legitimacy of the political system from an external viewpoint, this research reverses the perspective by looking at citizens’ evaluation of the three dimensions of the political system. The underlying assumption of such an approach is that citizens can have different evaluations of each component of the political system (the input, throughput, and output) and be dissatisfied with some dimensions of the political system more than others. If such a distinction exists, it would provide a more accurate understanding of what citizens are dissatisfied with and could inform the design of policies aimed at reducing democratic dissatisfaction.

In order to determine whether citizens differentiate between the legitimacy of input, throughput, and output, a latent profile analysis (LPA) is conducted to distinguish between several ‘profiles’ of citizens who share a similar evaluation of the political system. In addition to examining citizens’ profiles in terms of their evaluation of the political system, the research aims to take a closer look at socio-demographic characteristics associated with political underrepresentation such as gender, age, level of education, and occupation (Talukder, 2023; Bühlmann and Schädel, 2012; Carnes, 2016; Stockemer and Sundström, 2019; Childs and Lovenduski, 2013; Hakhverdian, 2015; Gaxie and Godmer, 2007), in order to reveal the differences between these individuals and their fellow citizens with regard to their evaluations of the various dimensions of the political system.

The research is conducted in the context of Belgium, a representative example of consociational democracy (Van Haute and Wauters, 2019; Deschouwer, 2012; Lijphart, 1968) with proportional representation, a high level of partisan fragmentation and large government coalitions. In a context where multiple parties across the ideological spectrum can access parliaments and represent the voice of citizens, and policy outputs are the result of negotiations between coalition parties, one can expect the results of the Belgian case to be a least likely case for finding differences in terms of input and output. One could expect the distinction to be more acute in majoritarian systems in which the difference between the input (understood as being represented by at least one politician/party) and the output (understood as the policies implemented by the government) is sharper. However, the multiple layers of power may lead to different evaluations of the political system’s throughput among citizens, particularly in comparison with majoritarian systems, and to a lesser extent with political systems that use proportional representation.

The results of the study demonstrate that citizens’ evaluations of the political system are not uniform for various subgroups of the population which constitute a significant proportion of the population. The results of a Latent Profile Analysis (LPA) show that six different profiles can be identified among the respondents with varying evaluation of the input, throughput, and output of the political system. In addition, traditionally underrepresented groups of citizens are more likely to evaluate one or more dimensions of the political system’s legitimacy negatively. These elements help us to go beyond the general indicator of SWD to better understand which dimensions of the political system citizens are dissatisfied with. They also support the idea of taking a systemic approach (in addition to the SWD item) in comparative studies and encourage us to consider reforms that would increase citizens’ support for democracy.

In order to carry out this exploratory analysis about the evaluation of the legitimacy of the political system by citizens, this research makes the use of the ‘Belgian Voter Panel Survey 2019’ (Michel et al., 2023), data collected by the interuniversity consortium Represent. A representative sample of Belgian citizens (based on age, gender, and level of education criteria) was surveyed twice, first within the 4 weeks preceding the Belgian Federal Elections in May 2019. The survey was conducted online, and respondents were recruited by a private polling company (TNS Kantar). Thanks to the first wave of the survey, in which 7,609 voters were surveyed (3,420 in Flanders, 3,133 in Wallonia and 1,056 in Brussels), it is possible to assess how citizens evaluate the components of the political system. This survey contains multiple items that measure political attitudes and enables us to operationalise citizens’ evaluations of the input, throughput and output of the political system, but also allows us to examine underrepresented citizens’ evaluations of the political system thanks to the large number of respondents. Furthermore, the classical indicator of satisfaction with democracy (SWD) is included in the latent profile analysis to facilitate its comparison with the three dimensions of political legitimacy. More specifically, including it will allow to understand to what extent there are differences between SWD and the various dimension of the political system.² The indicators mobilised to measure the different dimensions of the legitimacy of the political system are presented in Table 1 and are discussed below. In addition, a principal components analysis shows that the various items do not constitute a single component, but at least two relevant components (with an eigen value above 1.00).

Table 1 shows the operationalization of the evaluation of the input by citizens, which is made up of two items measuring citizens’ perception that their concerns and opinions are considered by at least (a) one political party and (b) a politician. These two items makes it possible to grasp to what extent citizens think that their opinions and their concerns are “entering” the black box and might generate (or not) policy outcomes. Citizens who believe that their concerns and opinions are not listened to might evaluate negatively the legitimacy of the political system.

While the distinction between input and output is straightforward, the integration of throughput terminology in the analysis is more complex, as both input and throughput refer to the procedures leading to the output.³ This research distinguishes the two by considering that input legitimacy refers to the aim of collecting citizens’ demands/opinion, while throughput legitimacy is associated with the way decisions are taken. This distinction is further discussed in the empirical part of this research. The evaluation of the throughput of the political system is measured through two items: citizens’ evaluation of the (a) transparency and (b) efficiency of the political system as these two items allow us to tap into two important components of throughput legitimacy. Transparency captures citizens’ perception of how transparent (or not) the political system is when it comes to decision-making which is often criticized by citizens. Efficiency, on the other hand, captures to what extent citizens are convinced (or not) of the decision-making process’ efficiency to generate outputs. This is particularly interesting in the Belgian case given the time needed for government formation and decision-making negotiations (which often involves multiple stakeholders).

Operationalising citizens’ evaluations of the output is rather straightforward: we consider citizens’ satisfaction with public policies implemented by the government at both the federal and the regional levels (which depend on the respondent’s region). Distinguishing between these levels can be important due to disparities in terms of government ideology at the regional level and the high-level of decentralization in Belgium.

Despite the fact that the classical indicator of satisfaction with democracy is often criticised for being relatively abstract and polysemous (Kriesi, 2013; Linde and Ekman, 2003; Valgarðsson and Devine, 2021), the item is very often used to study the evaluation of the legitimacy of the political system. Satisfaction with Democracy is thus included in the analysis to allow for comparison with the three other measures.

The analytical strategy to answer the research question is twofold. First, we aim to understand how citizens evaluate the political system and more specifically, whether they distinguish between input, throughput, and output, and whether distinct subgroups of (dis)satisfied citizens can be identified. To this end, we use a Latent Profile Analysis (LPA) to uncover specific profiles regarding the evaluation of the political system’s legitimacy. Rather than relying solely on a variable-centered approach (such as PCA or CFA), we opted for a person-centered approach (Lu, 2025), which identifies subgroups of individuals sharing similar patterns across variables. Such an approach allows us to examine whether different subgroups of the population who evaluate some of the three dimensions differently exist. If such a distinction is found, it would offer a clearer picture of the sources of citizens’ dissatisfaction and could guide the development of policies aimed at mitigating democratic discontent. Nevertheless, we also conducted a Principal Component Analysis and (two and three) Confirmatory Factor Analysis to confirm that the distinction between input, throughput, and output is empirically valid; these results are provided in the Appendix. In a final step, once the profiles are established, we examine their composition.

To answer the research question, a Latent Profile Analysis (LPA), which is part of the Latent Class Analysis (LPA/LCA) family, is conducted. This method makes it possible to classify respondents according to their evaluation of the legitimacy of the political system. The aim of this approach is to identify and form subgroups within the population on the basis of their responses to a defined number of items (see Peugh and Fan, 2013; Oberski, 2016). The indicators used in this analysis are the 6 items regarding the evaluation of the input, the throughput, the output of the political system, and the item measuring satisfaction with democracy. The benefit of this method is that respondents can be classified into a finite number of discrete categories (classes/profiles) that tend to differ in terms of responses on all items (Jackman, 2008, 139). Furthermore, the method assumes that citizens can be classified within each of the profiles with different configurations, with a certain degree of probability (Spurk et al., 2020). The method differs from classical classifications such as the k-means method, in which respondents are either members of a cluster or they are not (Peugh and Fan, 2013). With LPA, profile membership is treated as an unobserved categorical variable and is defined on the basis of the calculation of a probability degree (Spurk et al., 2020).

Although it is not the purpose of this research to elaborate in depth on the mathematical foundations of the method, a generic latent profile analysis model to illustrate the method from Peugh and Fan (2013), is shown below (see also Oberski, 2016 for the mathematical foundations).

This is a model in which μ ik and σ ik 2 represent for the specific profile (k) the means and variances for variable i; and π k indicates the density of the profile or the proportion of respondents who belong to profile k (Peugh and Fan, 2013, 618). Latent profile models have three assumptions: first, that samples from a heterogeneous population produce from the data a density function of a mixture law with K specific profiles. Second, the observed indicators are normally distributed. Third, that the mean vectors μ k related to the specific profiles are the observed means of the specific profiles (k) (Peugh and Fan, 2013).

The model first provides an initial random estimate of membership in a specific profile. Then, using the maximum likelihood method, the maximisation steps adjust these values and produce maximum likelihood estimates for the condition table and the expectation-maximisation algorithm uses these parameters to update the estimated parameters of the condition table until they converge and, as a result, do not change (Oberski, 2016). The latent profile analysis method depends on a number of profiles pre-defined by the researcher. However, when the aim is, as in this case, exploratory, the ideal number of profiles can be determined by means of two criteria: the first is linked to goodness-of-fit statistics and the second is linked to the researcher’s intuition as to the significance of the analysis and the meaning that emerges from the profiles established. If adding an additional profile does not change the results, is difficult to interpret or does not seem to make sense (from a theoretical perspective), it is very likely that the established profiles are not ideal, despite a decrease in terms of goodness-of-fit statistics.

To determine the optimal number of latent profiles, we employed a combination of statistical fit indices and the theoretical interpretability of the resulting solutions. We primarily relied on the Bayesian Information Criterion (BIC), a parsimony-adjusted measure calculated as BIC = (−2logL + plog(n)). This statistic is calculated for each analysis that extracts a given number of profiles. The Figure 1 shows all the results concerning the BIC for each analysis in 1, 2, 3, 4, 5, 6, 7, and 8 profiles. Theoretically, each time the BIC statistic decreases with the inclusion of an additional profile, it leads to the conclusion that the model with n-1 profiles is less relevant (as it explains less variance) than the one with n profiles. While theoretically the optimal model is the one with the lowest BIC value, our analysis (Figure 1) showed that the BIC continued to decrease up to the eight-profile solution, but with less importance. Given the continuous decrease in the BIC, we introduced a pragmatic step that combines statistical evidence with theoretical judgment. Crucially, the rate of decrease in the BIC value began to flatten significantly after the five-profile solution. This flattening, often interpreted as the “elbow” in the BIC plot, suggests that the marginal improvement in model fit from subsequent profiles becomes increasingly negligible. The Akaike Information Criterion (AIC) scores followed a similar pattern, confirming this trend. Given this point of diminishing returns, and prioritizing theoretical interpretability and parsimony, we selected the six-profile solution. Solutions with five, seven, and eight profiles were rejected because they produced profiles highly similar to those already established or contained very small, unstable subgroups, which did not offer significant new theoretical insights. All those solutions tend to find several subgroups of citizens who distinguish between the three dimensions of the political system’s legitimacy. The results for these higher-order solutions are provided in Appendix C for full transparency.

Once the number of profiles has been decided, the results are interpreted by the researcher. More specifically, the meaning of the profiles is interpreted in a qualitative way. To interpret the results, it is important to consider both variations within and between profiles. Figure 2 shows the results graphically (on a 0–1 scale to ease interpretation), and the table in Appendix B presents the results numerically (mean, standard error, and confidence intervals). The results outline differences between the evaluation of input, throughput, and output both within and between profiles. The results also underline that the classic indicator of SWD might differ from the evaluation of subdimensions of the political system’s legitimacy. The profiles are interpreted in a more qualitative way below.

Beyond the interpretation of the profiles, we observe diversity in terms of profile membership. The composition of the profiles oscillates between 10 and 15% of the sample (with the exception of profile 4) demonstrating that this issue needs to be investigated carefully, as some citizens have uniform positive/negative evaluations while others have mixed evaluations across specific dimensions. It is also worth noting that profile 5 and profile 2 tend to be among those who distinguish the most between at least two dimensions of the political system’s legitimacy⁴ (while the difference is significant for others but to a lesser extent, see Appendix B), which underlines that the input, throughput, and output framework is relevant to better understand what citizens are dissatisfied with in their political system.

The analyses highlight that citizens have very different assessments of the different dimensions of the legitimacy of the political system and that these evaluations are heterogeneous. To deepen our knowledge of profile membership, the following table illustrates the composition of the profiles by sociodemographic characteristics (Table 2).

In addition to the descriptive statistics, a multinomial regression is carried out with the “Positive” profile as the reference category (the results are presented in Table 3, and the same model taking the ‘average’ (Profile 4) as a reference category is available in Appendix D). This choice allow us to see to what extent some socio-demographic characteristics are associated with belonging with profiles who have a negative evaluation of one or several dimensions of the political system. The profile of the most positives (Profile 6) seems to be the most suitable one as the other profiles have a negative evaluation of at least one of the dimensions of the legitimacy of the political system. The multinomial regression opens avenues of reflection, and the interpretation of the results focuses on describing the results for the three main socio-demographic characteristics linked to descriptive political underrepresentation (gender, level of formal education, and occupation).

The results show some clear patterns for the group of women. Indeed, compared to the profile of people with the most positive evaluation (Profile 6) of the legitimacy of the political system, the probability of belonging to any other profile than the most positive one is significant. However, the probability of belonging to the profile of the citizens who are generally dissatisfied (Profile 3) is the highest, followed by belonging to the profile largest profile (profile 4) but also profile 1 which has the most negative evaluation of the political system. When the reference category is the ‘average’ profile, we see that the women’s group has essentially a positive relative risk ratio for belonging to the generally dissatisfied (Profile 3) and a negative relative risk ratio for belonging to Profile 6 and Profile 2 (positive input and negative throughput and output; see Appendix D).

The level of formal education also plays a role in profile membership, as citizens with a low level of formal education are much more likely to belong to the most negative Profile 1, the Profile 3 characterized by general dissatisfaction toward each dimensions, and those who feel represented by a party/politician, but evaluates the other dimensions negatively (Profile 2). The results are similar if we take the profile 4 category as a reference (see Appendix D).

Finally, citizens who have precarious working conditions (unskilled workers and jobseekers) have a relative risk ratio that is twice as high for belonging to the profiles of the most negative citizens (Profile 1), the generally dissatisfied (Profile 3) and they are the most likely to be among those who feel represented by a party/politician but evaluates the other dimensions negatively (Profile 2). In contrast to those with a low level of formal education, the probability of belonging to the Profile 2 is the highest. Moreover, when we take the “Profile 4” category as a reference (see Appendix D), we observe a higher relative risk ratio of belonging to profile 1, the most negative one toward each dimension of the political system’s legitimacy.

Underrepresented groups of citizens tend to be among the citizens who tend to be the most critical of the political system and are less likely to be among the positives. The results seem thus to corroborate studies according to which ‘under-represented’ groups tend to have a more negative evaluation of the legitimacy of the political system and its institutions (Talukder, 2023; Mayne and Hakhverdian, 2017; Reher, 2016; Ezrow and Xezonakis, 2011; Marié and Talukder, 2021) and the present study allows us to provide a clearer picture of it by focussing on input, throughput, output and satisfaction with democracy. We observe that socio-demographic characteristics are significantly associated with profile membership.

Starting from an exploratory approach, these elements allow us to build up a general picture of citizens’ evaluation of the different dimensions of the political system’s legitimacy. These groups have very different assessments of the legitimacy of the political system, which may be linked to different explanatory elements.⁵ At this stage, we suggest using these results as an entry point for a more systematic analysis of citizens’ (and underrepresented citizens’) evaluation of the political system.

This research contributes to the study of citizens’ political attitudes, specifically their evaluation of the political system. Building on a literature that documents a decline in democratic support, it proposes an alternative approach to better understand citizens’ dissatisfaction with their political systems. Current scholarship largely relies on indicators available in international surveys, such as “Satisfaction with Democracy” (SWD). While convenient, SWD is too broad to capture the specific dimensions of dissatisfaction. This study applies the classic framework of “input, throughput, and output legitimacy” (Schmidt, 2013; Scharpf, 1999; Easton, 1965) to deconstruct SWD from a citizen-centered perspective.

This approach enables a more nuanced analysis, breaking down the evaluation of political system legitimacy into multiple dimensions. Using exploratory latent class (or profile) analysis (LCA/LPA), the study identifies distinct profiles of respondents with varying evaluations of the system’s legitimacy. These profiles include the citizens with a negative evaluation of each dimension (Profile 1) and Profile 3 (which is less negative than Profile 1), the citizens who tend to be satisfied with the input only (Profile 2), the citizens with a positive evaluation of each dimension (Profile 6), and a group of citizens substantively dissatisfied with the throughput of the political system (Profile 5). These findings illustrate that citizens’ evaluations of the political system often extend beyond generalized satisfaction, emphasizing the need for a multidimensional approach.

Despite its exploratory nature, this research offers several significant insights. First, it highlights that many citizens differentiate between dimensions of the political system, such as input, throughput, and output legitimacy. This distinction underscores the limitations of single indicators like SWD, political trust, or democratic support, which fail to provide a comprehensive understanding of citizens’ evaluations. Second, the results reveal that dissatisfaction with the throughput dimension is common in this case study. Reforms targeting procedural aspects of the political system could improve citizens’ evaluations. Unlike output legitimacy, which inherently creates winners and losers, improving input and throughput legitimacy is feasible and could enhance support for democracy. Finally, this study advocates for the use of additional indicators to analyse democratic dissatisfaction, establishing a basis for future research, promoting comparative analyses across cases, and facilitating deeper exploration of the determinants of legitimacy dimensions.

In an era where increasing numbers of citizens globally no longer view democracy as the best form of governance, understanding the nuances of dissatisfaction with political systems is essential. By uncovering these dimensions, this research contributes to the broader discourse on how to strengthen democratic legitimacy and respond to citizens’ concerns.