Widening the gap of political inequality? The effect of the COVID-19 pandemic on political engagement

There is extensive evidence that the COVID-19 pandemic has mostly affected the less well-off in society, boosting economic inequality. In contrast, little is known about how much such rising economic disparities affected the involvement of individuals in politics, thereby enhancing political inequality. Extending the research on political inequality to a key and somewhat neglected dimension of citizens ’ involvement with politics - political engagement - this article claims that the COVID-19 depressed engagement and promoted political inequality. The analysis relies on a comparative European approach and on data before and after the emergence of the pandemic. Besides generally finding an overall socioeconomic gap with regard to political engagement, results also suggest that the pandemic somewhat lessened engagement, increasing the gap between the more and less socioeconomically advantaged. Generally, this is not strictly due to a tendency to decrease engagement among the latter but also to increase engagement among the former.


Introduction
Since Aristotle, conceptions of democracy have often included as a requirement some level of socioeconomic equity between individuals.The underlying assumption is that socioeconomic conditions affect individuals' ability to express opinions on political issues (Bartels, 2016;Elsässer and Schäfer, 2023).This implies that the equality of citizens in the political sphere is a defining principle of democracy as a political system (Dahl, 1996).Thus, the democratic ideal is not fully achieved if some groups in society are taken into account in policy decision-making while others are not (Lijphart, 1997;Verba, 2003).Actually, there is evidence showing that this is the case in contemporary liberal democracies: individuals cannot equally make their voices heard, with the most silenced being the least privileged (Schlozman et al., 2018;Houle, 2018;Guntermann, 2021).Far from being solved, political inequity is a challenge for democracies that seems to be accentuating (Dalton, 2017) and, for that reason, deserves further attention.
The pervasive growth of economic inequality in post-industrial societies contributes to this less optimistic scenario (Chancel and Piketty, 2021;Fisher et al., 2022).As widely demonstrated, economic and political inequalities tend to be interdependent and mutually reinforcing, with economic inequality negatively impacting political equality (Bartels, 2016;Giger et al., 2012;Houle, 2018;Schäfer and Schwander, 2019;Schlozman et al., 2018; for a research review, see Elsässer and Schäfer, 2023).Recently, the COVID-19 pandemic seems to have made a further contribution to widening the gap between rich and poor (Adams-Prassl et al., 2020;Darvas, 2021), especially by deteriorating the economic conditions of the less well-off in society (Shibata, 2021).As a consequence, it is believed that the gap in citizens' political voice will tend to be accentuated as a result of the pandemic.
This article aims to contribute to the expanding body of work on political inequality in liberal democracies, mainly by examining how the pandemic episode may have affected it.During the pandemic crisis, the restrictions imposed on individuals' movement and self-restriction of social contact for fear of being infected, have rendered some typical political participation variables, such as turnout, unsuitable for research.As political participation generally implies freedom of movement and social contact, the pandemic constrained its manifestation (Fernandez-Navia et al., 2021;Offe, 2021: 39).In such a context, supporting the study on individuals' political engagement appears to be a more valid perspective of analysis.This is indeed a core perspective on political equality, although quite surprisingly seldom studied (recent exceptions are: Schlozman et al., 2018;Solt, 2008;and, on political efficacy: McBrayer et al., 2022;Oser et al., 2023).
Political engagement regards the psychological connection of citizens with politics, such as identifying with a party, feeling politically efficacious and being interested in politics (Brady et al., 1995).Voters who feel close to a political party, who feel able to understand and influence the political process, and who are interested in politics, will be more likely to be mobilized to political participationi.e., to act politically (such as by voting, participating in a campaign, or in a protest) compared to those who do not have such engagement (Bingham Powel, 1986;Brady et al., 1995;Karp and Banducci, 2008;Schlozman et al., 2018).Furthermore, proximity to a political party, political efficacy and interest likely act as mediators of the effect of socioeconomic status on political participationi.e., in turnout (Sondheimer and Green, 2010, 186).
Insofar as engagement predisposes individuals to take an active part in political life (Bingham Powel, 1986;Brady et al., 1995), it is of the utmost importance in studying inequality in politics.The focus of this piece of research is, therefore, the assessment of the effect of socioeconomic inequality on individuals' unequal political engagement.
The major goal of this research is to test whether the pandemic had a greater effect on those with lower incomes and less schooling, decreasing their levels of political engagement, and thereby widening the gap in political inequality.The research also has two subsidiary goals.One is to assess whether there is actually a socioeconomic gap in political engagement.As identified for other dimensions of political involvement (see, for instance, regarding turnout: Gallego, 2015;Schäfer and Schwander, 2019), it is crucial to primarily ascertain whether a gap is also in place with regard to political engagement (for some aspects of engagement, see : Solt, 2008).Another secondary goal aims at checking whether the COVID-19 contributed to depressing political engagement.There is already evidence that the pandemic affected political participation (e.g., Fernandez-Navia et al., 2021).However, more than ascertaining whether it reduced participation, which is to be expected given the constraints on individuals' freedom of movement, it is important to determine whether it conditioned their predisposition to act politically (for an exception on political efficacy, see McBrayer et al., 2022).A look at political engagement allows conclusions to be drawn in this regard.
Based on data from the European Social Survey (rounds 9 − 2018 and 10 − 2020), this research is carried out from a comparative perspective at the European level, encompassing 17 European countries, and comparing two moments in time: the pre-and post-pandemic emergence.
The organization of this article is as follows.First, a review of the extant literature on socioeconomic and political inequality and on pandemics and inequality is presented, and the hypotheses are formulated.Then, data and methods are described, and the results are presented and discussed.Afterward, concluding remarks and implications of the study are enunciated.

The interplay between socioeconomic and political inequality
Prior research provided ample evidence of a negative effect of socioeconomic inequality on political equality, consistently concluding that the less advantaged tend to have less political voice (Beramendi et al., 2022;Bartels, 2016;Giger et al., 2012;Houle, 2018;Schäfer and Schwander, 2019;Schlozman et al., 2018; for a research review, see Elsässer and Schäfer, 2023).It also concludes that they are less knowledgeable (Bartels, 2016), less politically sophisticated (Rosset and Kurella, 2021), and tend to perceive themselves as not being politically represented (Rennwald and Pontusson, 2022).Research has in particular shown that the less well-off are less likely to vote compared to the better-off (Erikson, 2015;Gallego, 2015;Houle, 2018;Lijphart, 1997;Schäfer and Schwander, 2019;Solt, 2008; though this is not fully consensual, see eg. : Guntermann, 2021).Likewise, those with lower education also tend to vote less (Armingeon and Schädel, 2015;Dassonneville and Hooghe, 2017;Gallego, 2015;Sondheimer and Green, 2010).This is of relevance, as the increase in turnout with a view to approaching universal participation is considered fundamental to mitigating political inequalities (Lijphart, 1997).Illustrating this, the income gap in policy responsiveness seems to decrease as turnout increases (Peters and Ensink, 2015).
Similarly, in more unequal societies, political participation tends to be particularly low among low-income groups (Ritter and Solt, 2019;Schäfer and Schwander, 2019).Such societies have the potential to depress the levels of political interest, political discussion, and turnout (Solt, 2008), as well as the participation in protest activities of all social strata apart from the richest (Beramendi et al., 2022;Solt, 2015).In summary, either at the micro or the macro levels, socioeconomic inequality tends to depress political engagement and participation, particularly punishing the less privileged.
The fact that the more disadvantaged are less engaged in politics and participate less is consequential, as they have different political preferences than the more privileged (Schwander, 2019).Interrelatedly, the politically active may have dissimilar preferences to those who are inactive in politics (Beramendi et al., 2022: 16-17).Thus, whether or not individuals are heard regarding their preferences is of relevance, as it affects their corresponding representation.
The effect that the limited voice of those in the lowest strata of society has on political representation has been widely recognized.Mostly focusing on the USA, such studies have concluded that political decisions are systematically more responsive to top-income groups (Bartels, 2016;Persson and Sundell, 2024;Schakel and Van Der Pas, 2021).The poor tend to be more distant from the ideological stances of legislators than upper-income groups and to be less well represented, especially regarding economic issues (Giger et al., 2012;Rosset and Kurella, 2021;Lupu and Warner, 2022).As evidence of its tenacious nature, unequal representation is observed even in democracies with a high degree of social equality, such as Norway (Schakel and Van Der Pas, 2021;Mathisen, 2022).
In light of the above, the mechanism by which socioeconomic inequality interplays with political inequality may be straightforwardly summarized.Individuals belonging to the lowest socioeconomic strata tend to reconcile low levels of education, resources and skills to intervene politically (Erikson, 2015: 20-23).As a consequence, they tend to participate less in the political sphere (Brady et al., 1995;Houle, 2018;Ritter and Solt, 2019;Schlozman et al., 2018;Verba, 2003).Since parties are disproportionately responsive to the citizens who are more active in politics (Peters and Ensink, 2015), the resulting outcome is that the preferences of the least well-off tend to be less taken into account by governments and are thus often underrepresented, which reinforces their disadvantaged condition (Houle, 2018;Guntermann, 2021).In other words, socioeconomic inequality fosters political inequality, which fosters socioeconomic inequality, and so on, in a never-ending loop (Bartels, 2016).This pervasive, systemic and self-reinforcing mechanism is distancing more and more of the less privileged from political life (Dalton, 2017;Dassonneville and Hooghe, 2017).
Contributing to this gloomy scenario, there is ample evidence that levels of economic inequality in most post-industrial societies resumed an upward trajectory during the late 20th and early 21st centuries (Chancel and Piketty, 2021;Fisher et al., 2022).Such a trend reinforces the political underrepresentation of the poor (Bartels, 2016;Erikson, 2015;Schäfer and Schwander, 2019).More recently, the COVID-19 pandemic has also contributed to widening the gap between rich and poor due to a distributional impact that asymmetrically affected individuals according to socioeconomic groups (Adams-Prassl et al., 2020;Darvas, 2021).The crisis will have especially affected the socioeconomically most vulnerable (Shibata, 2021), promoting an increase in wage inequality (Darvas, 2021).Following the deteriorating socioeconomic conditions, political inequality is expected to have increased.However, so far, there is scarce information on how much the COVID-19 pandemic affected political inequality.

Political engagement and the COVID-19: hypotheses
If we extend research on unequal participation to the study of political engagement, there is good reason to believe that engagement is likely to differ across socioeconomic groups.Actually, there is evidence in the literature on political engagement suggesting such a gap.For instance, high education and income groups seem to be related to the strength of attachment to a party, although weakly (Schlozman et al., 2018: 60-61).Income and education inequality have also been found to be positively correlated with lower political internal and external efficacy and interest (Gallego, 2015: Chap. 6;Oser et al., 2023).In the aggregate, growing economic inequality seems to likewise depress political engagement (political interest and discussion) among all but the most affluent citizens (Solt, 2008).
As none of this previous research comprehensively addressed unequal political engagement across income and education lines, a primary goal is to test whether such a gap exists.Therefore, the first hypothesis is the following: HYPOTHESIS 1: The levels of political engagement tend to be higher among higher income and education groups compared to lower ones.
Prior research has shown that external shocks, such as public health outbreaks, have an effect on how individuals' think and act politically, shaping political attitudes (Albertson and Gadarian, 2015).Likewise, the COVID-19 pandemic implied the triggering of a whole set of imbalances that (differently) affected individuals, including at the political level (Offe, 2021).At this level, research has mainly focused on how it influenced trust in political institutions (e.g., Kritzinger et al., 2021), turnout (e.g., Fernandez-Navia et al., 2021), regime support (e.g., Amat et al., 2020;Erhardt et al., 2022), political participation in general (Borbàth et al., 2021;Belchior et al., 2022), and collective protest in particular (Kriesi and Oana, 2022), signaling the accentuation of the underlying prior gender gap in participation (e.g., Burciu and Hutter, 2023).The way in which the pandemic affected how individuals engage with politics has received almost no attention.As exceptions, one study on the USA shows that it influenced political efficacy, although the direction of the change is uncertain, depending on blame attributions to the government (McBrayer et al., 2022).Another study on Portugal showed that it decreased individual political competences and somewhat increased political interest and information (Belchior et al., 2022: 273-275).
If we more broadly look at the research on natural disasters, findings show that such events can lead citizens to be less supportive of democratic values and practices, unlashing criticism of incumbents (Carlin et al., 2014).The COVID-19 pandemic also seems to have led to a deterioration of democratic preferences (Amat et al., 2020), and generated critical views over the governments' performance, after a brief rally around the flag (or no such effect, in some countries) (e.g., Kritzinger et al., 2021).As citizens' political evaluations are more likely to be negative than positive (Weaver, 1986), even regarding uncontrollable disasters (Arceneaux and Stein, 2006), and will eventually be associated with a more critical view of democratic institutions in the face of external shocks (Carlin et al., 2014; regarding the COVID-19 pandemic see: Amat et al., 2020;Erhardt et al., 2022), the lowering of the level of political engagement after the outbreak of the pandemic is a plausible outcome.This is because engagement is closely and positively correlated with other aspects of individuals' relationships with politics, such as institutional trust or support for the regime (Torcal and Montero, 2006).The depreciation of the latter, as a result of the pandemic context, will probably have repercussions on the disengagement of individuals.Although the effect of the pandemic may be different across dimensions of political engagement, there are no sufficiently strong theoretical reasons to establish specific expectations of them.The second hypothesis thus reads as follows: HYPOTHESIS 2: The advent of the COVID-19 pandemic depressed political engagement.
The concern about the impact of pandemics on inequality is not new.However, past research has almost exclusively focused on economic inequality and only residually on political inequality.Such research shows that, in the absence of long-lasting policies to protect the most socially vulnerable, pandemics are prone to aggravating economic inequality (Furceri et al., 2022;Perugini and Vladisavljevic, 2021).The COVID-19 also seems to have increased economic inequality by worsening the conditions of the poorest (Darvas, 2021;Shibata, 2021).
Given the self-reinforcing relationship between economic and political inequality (Bartels, 2016;Verba, 2003), it is reasonable to believe that by aggravating the condition of the less privileged, the pandemic has also increased political inequality.To the best of our knowledge, to date, only one study has explored how much the pandemic may have affected political inequality, although exclusively based on the Portuguese case (Belchior et al., 2022: Chap. 5.7).It concludes that the gap in political competence and participation between the lower and higher income and education groups generally increased after the pandemic outburst, with the most disadvantaged evidencing a higher deterioration or, at best, a consolidation of their pre-pandemic underprivileged condition.Consonant with this finding, other research concluded that the upsurge of the pandemic led the rich to feel more efficacious than the poor (McBrayer et al., 2022).Based on this scant evidence, the last hypothesis aims at testing whether a gap in political engagement between the more and less privileged has widened since the emergence of the pandemic.It reads as: HYPOTHESIS 3: The gap in political engagement across income and education groups tends to widen with the advent of the COVID-19 pandemic.

Data and methods
The analysis relies on data from the European Social Survey (ESS) before and after the outbreak of the pandemic (round 9 − 2018 and round 10 -2020) whose fieldwork for the sample of countries was, respectively, August 2018 -January 2020 and May 2021 -May 2022.Relying on data relating to the second year of the spread of the coronavirus in Europe allows for a valid assessment of the pandemic consequences on political inequality, given that this is when its cumulative effects (health and socioeconomic) are being felt with greater intensity in Europe. 1 All countries that in both waves had fully equivalent questions to operationalize the dependent and independent variables were included in the analysis.The sample encompasses 17 European countries. 2 Note that the observational nature of the research design impedes the straightforward inference of causality.That is, although it is fair to assume that people's levels of education and income precede their political engagement, the reverse causation direction is also conceivable.

Dependent variables
The analysis is based on three dependent variables to measure political engagement: closeness to a political party, political efficacy and political interest. 3Although the three make it possible to jointly measure the level of political engagement (Brady et al., 1995), each of them will be analyzed autonomously.This is due to the low internal consistency of aggregating them (Cronbach Alpha = 0.494), as well as the advantage of achieving some cross-validation of results by measuring different aspects of the same concept.
Given that the original scales of the dependent variables are different and party closeness is dichotomous, in order to enable the comparison between the three dependent variables while maintaining a parsimonious analysis, the option is the dichotomization of the remaining two.Thus, all dependent variables vary between 0, corresponding to low political engagement, and 1, corresponding to high political engagement.To ensure that results are not affected by the option of dichotomizing those two variables, the analysis was also carried out with the non-dichotomized variables.The results are very consistent with those achieved after dichotomization (see Tables A7 to A10 in the Online Appendix).
Closeness to a political party.The question assessing whether individuals are close to a party is the following: 'Is there a particular political party you feel closer to than all the other parties?'The response items from the original scale are considered in the analysis: 'does not feel close to any party' = 0 and 'feels close to a party' = 1.
Political Efficacy.Both internal and external indicators of political efficacy are used.As commonly assumed, internal efficacy refers to individuals' beliefs in their own ability to comprehend politics and to competently engage in political actions.In turn, external efficacy refers to individuals' beliefs that their participation will influence political leaders and institutional decisions (see e.g., Karp and Banducci, 2008;McBrayer et al., 2022: 98).Despite being correlated, these are distinct concepts.The decision to treat them together is supported by the assumption that the variance of both -internal and external efficacy -is similar across different socioeconomic groups, as suggested by previous research on political inequality, which also used the two concepts together (Schlozman et al., 2018: 59).Also favorable to this option is the need to work with simple and robust measurements.
Two variables of internal efficacy and two of external efficacy are used.For the former, the variables are: the level of agreement with being 'Confident in own ability to participate in politics' 4 ('not at all confident' = 1; 'completely confident' = 5) and with being 'Able to take active role in political group' 5 ('not at all able' = 1; 'completely able' = 5).For external efficacy, the variables are the level of agreement with the 'Political system allows people to have influence on politics' 6 and  Political information is also an important variable in the operationalization of political engagement (Schlozman et al., 2018).However, it was not possible to include it in the analysis given the unavailability of a suitable question in the ESS.This is not considered a drawback, as the variables that are included offer a dependable measurement of political engagement.

4
The question is: "How confident are you in your own ability to participate in politics?" 5 The question is: "How able do you think you are to take an active role in a group involved with political issues?"6 The question is: "How much would you say that the political system in [country] allows people like you to have an influence on politics?"with the 'Political system allows people to have a say in what government does' 7 (the scale is for both: 'not at all' = 1; 'a great deal' = 5).These variables were aggregated into a composite variable, both for the period before and after the outbreak of the pandemic.The Cronbach Alpha is, respectively: 0.749 and 0.784. 8This composite variable was then recoded so that values below the median (meaning low efficacy) were coded as 0, and values above the median (meaning high efficacy) were coded as 1.
Political interest.This measurement relies on the question: 'How interested would you say you are in politics?Are you very interested, quite interested, hardly interested, or, not at all interested?'.This scale was recoded so that 0 corresponds to disinterest in politics ('not at all' plus 'hardly interested') and 1 to interest in politics ('very' plus 'quite interested').

Correlates between the dependent variables
As the dependent variables measure different facets of political engagement, they are expectedly interrelated among each other (as before seen: Karp and Banducci, 2008;Oser et al., 2023: 2;Schlozman et al., 2018: 60-61).In order to clarify how much this is the case, associations and correlations were estimated among all the variables underlying political engagement (using their original scales; see Table A1 in the Online Appendix).Not surprisingly, the variables are either associated or positively correlated.As the relationships have weak to moderate magnitude, it can be said that, despite being related, these variables are independent of each other.

Independent variables
The analysis is based on two main explanatory factors: socioeconomic status, measured by income and education; and the pandemic effect, measured by time and by how much the respondent's health and economy were affected.

Measuring income and education groups
Comparing different income groups and, to a much lesser extent, education attainment levels, have been the most common strategies used to measure social stratification in political equality research (Elsässer and Schäfer, 2023).Although less used, education not only fosters political participation via its effect on individuals' skills and resources (Verba, 2003: 668-669) but also relates to income level.Compared to income, education is even considered more valid in estimating the socioeconomic condition of individuals (Schlozman et al., 2018), as well as more reliable insofar as it is not as touched by the problem of non-responses (Dassonneville and Hooghe, 2017).For these reasons, recent research has more frequently encompassed education to measure socioeconomic status (Darvas, 2021;Dassonneville and Hooghe, 2017;Mathisen, 2022;Rennwald and Pontusson, 2022;Schakel and Van Der Pas, 2021).Thus, besides relying on income groups, the present research also considers education groups to measure socioeconomic status.By combining both income and education, this research aims to move one step forward in political inequality studies that have mostly focused on one measure of stratification at a time (Elsässer and Schäfer, 2023: 13).
Income and education are operationalized in order to ensure cross-country comparisons.Income is based on the ESS question assessing the household's net income, which classifies income into decile intervals relative to the distribution of national income.Following recent research 7 The question is: "How much would you say the political system in [country] allows people like you to have a say in what the government does?" 8 In order to ensure the consistency of the variables underlying both concepts -internal and external efficacy -an exploratory factor analysis was also carried out.The results point that the four variables are organized into a single factor, in both years (KMO = 0,634 in 2018 and 0,670 in 2021).
The effect of the COVID-19 pandemic on political engagement (Rosset and Kurella, 2021), respondents were then classified according to their individual income level.To this purpose, the median of the intervals was calculated and, given that the last decile has no upper bound, the procedure of Hout's was applied (Hout, 2004).In order to take the size of the household into account, the resulting income values were then weighted by the number of people in the household (by dividing the values by the square root of the number of household members).The resulting distribution of income deciles was then recategorized into quintiles and terciles (the latter for the purpose of running robustness tests).Relying on quintiles and terciles allows the necessary comparison between high-and low-income groups and the cross-validation of results.
To compare the least educated with the most educated respondents, the analysis is based on the ESS question measuring the level of education in United Nations Educational, Scientific and Cultural Organization's International Standard Classification of Education.This question allows to create three groups (as done by Aaldering, 2017).One corresponds to the least educated, that is, to citizens who have no formal education or who have had up to primary or lower secondary education.Another group regards the most educated, including citizens with a higher vocational or university degree.Respondents positioned in the remaining categories (corresponding to upper secondary education levels) are coded as having a medium education level.
As the aim is to compare the higher and lower income and education groups, it is necessary that such groups have in each case a similar dimension.If income groups are the same size, derived from the calculation of quintiles and terciles, in the case of education this is not exactly the case.The education groups above-mentioned have, however, approximately the same percentage of respondents, with the least and most educated being in both years a little less than a third (31,6 and 32,7 percent in 2018, and 32,1 and 28,5 percent in 2021).

The pandemic effect
The extent to which the pandemic contributed to widening the gap in political engagement is measured in three ways.The first is by comparing the moments before and after the onset of the pandemic (that is, 2018 and 2021).Although relevant, this comparison does not allow for validly assessing the effect of the pandemic since other uncontrolled extraneous factors may interfere with the variation of the dependent variables.Therefore, time is not sufficient to measure the effect of the pandemic.
For this reason, the analysis also uses the pandemic's impact on people's health and economy (making full use of the questions in the ESS).The former supports whether the individual has had COVID-19 disease.The question is: 'Have you had coronavirus?' whose responses were coded as 1 if the respondent stated 'Yes, I tested positive for coronavirus', and as 0 if he/she stated 'No, I have not had coronavirus' (the ambiguous answer: 'Yes, I think I had coronavirus but was not tested/ did not test positive' was discarded).
Regarding the individual's economy, the following question is used: 'Please tell me if any of the following happened to you as a result of the coronavirus pandemic?Please include anything that has happened at any time since the start of the pandemic, even if it is no longer impacting you'.Since what matters is whether the respondent was affected by the pandemic at an economic level, code 1 was assigned if he/she selected any of the following response items: 'I was made redundant/ lost my job'; 'The income from my job was reduced'; 'I was furloughed'; and 'I was forced to take unpaid leave/holiday'.Code 0 was assigned if none of these items were selected.It should be noted that an overwhelming majority of respondents (88,4 percent) did not report having experienced any of these economic effects and that among the minority who did, one event was the most commonly mentioned (by 9,4 percent of the respondents; only 1,8, 0,4 and 0,1 percent mentioned, respectively, two, three, and four of these events).

Control variables
Since the purpose of this research is not to identify the factors predicting political engagement but rather to determine if, at this level, a socioeconomic gap prevails, the analysis is controlled by a limited number of variables.As usual in prior research, the controls included are age and gender (e.g., Rennwald and Pontusson, 2022).Table A2 in the Online Appendix presents descriptive data on the variables included in the analysis.

Unequal political engagement in Europe
As a first descriptive step, Figure 1 presents the distribution of the dependent variables in relation to income and education groups. 9Consistent with prior findings (Gallego, 2015: Chap. 6;Oser et al., 2023), political engagement tends to increase as individuals' socioeconomic status increases, particularly in terms of political efficacy.Party closeness and political interest are flatter across income and education groups, with the former being the least stratified (as found for the strength of party identification: Schlozman et al., 2018: 60-61).With the advent of the COVID-19 crisis, a tendency towards increasing distance in political engagement can be observed between top and bottom socioeconomic groups, although differences are small and there are exceptions regarding education (very similar results are obtained when all levels of education are considered, as well as for income deciles or terciles, as Figures A2 and A3 in the Online Appendix show).
To estimate the statistical importance of that distance over time, the mean and mean differences were calculated for income quintiles and education groups (in Table 1).The income and education dummies in the models were coded as 1 for the quintile and education group at stake, and 0 for the remaining quintiles and education groups.The results are consistent with the three hypotheses.First, differences among both socioeconomic groups are mostly statistically significant for the three variables measuring engagement (only three out of 36 are not), which denotes that the income or education group to which the individual belongs differentiates his/her level of political engagement.Second, political engagement appears to have generally decreased in 2021 for all three variables and socioeconomic groups.Finally, the pandemic appears to have widened the gap between rich and poor in terms of political engagement.Between 2018 and 2021, differences across the first-and fifth-income quintiles increased for party closeness and political efficacy, while interest in politics showed a slight decrease.Very similar results are obtained if terciles are considered (see Table A3 in the Online Appendix).As for education, differences between the top and bottom groups only became wider in 2021 with regard to party closeness.
In order to fully test the hypotheses, multilevel logistic regression models are now carried out on the pooled dataset (in order to compare the moment before and after the outbreak of the pandemic), including random intercepts for countries (to account for country-level variance).Given the clustered structure of the data (respondents within countries), the analysis supports generalized linear mixed models with robust standard errors (Hox et al., 2018: Chap.2). 10 Table 2 jointly examines the extent to which there is a socioeconomic gap in political engagement, whether the pandemic depressed engagement, and whether it reinforced an income and education gap in 9 The distribution of mean income and education in the sample of countries, shown in Figure A1 in the Online Appendix, demonstrates that, although likely related, both variables are independent at the aggregate level.For the overall sample of countries, the correlation between income and education is statistically significant in both years and stable over time, at a Pearson's r = 0,30.
At the microlevel, the correlations are also statistically significant for all countries, ranging between a Pearson's r = 0,24 (in Estonia) and 0,48 (in Portugal).10 To confirm whether multilevel modeling was necessary, intercept-only models were first run.Intra-class correlation coefficients (ICC) indicate that a non-negligeable amount of the variance is explained at the country level (ICC for party closeness is 5,1 percent; for political efficacy is 22,0 percent; and for political interest is 11,6 percent), thus advising the use of multilevel analysis (Hox et al., 2018: chap.2).
The effect of the COVID-19 pandemic on political engagement engagement.The effect of the pandemic is strictly measured using a year dummy (0 = 2018 and 1 = 2021).
To evaluate whether the pandemic has increased the gap in political engagement, interactions between the year dummy and socioeconomic status were added in model 2. Most of the interactions reach statistical significance for the lower socioeconomic groups, demonstrating that the pandemic year influenced to some extent the individual's level of engagement with politics, especially across the lowest education group.Figure 2 depicts these interactions, clarifying that the decrease in engagement in 2021 has more decisively hit the less educated compared to the more educated, widening the gap between both (consonant with H3).Education as a stronger predictor of political inequality compared to income has already been demonstrated in representational studies (Schakel and Van Der Pas, 2021).The probability of individuals in the lower education group being close to a party and being interested in politics drops, respectively, by 13,0 and 2,9 percentage points.Conversely, the decrease for individuals belonging to the high education group is substantially smaller, at 4,4 and 0,8 percentage points, respectively.In addition, while political efficacy only marginally decreases among the poorly educated (0,7 percentage points), among the highly educated it exhibits a small increase (1,3 percentage points).Consistently, the probability of individuals feeling politically efficacious slightly decreases for the poor and increases for the rich in  2. A4 in the Online Appendix).Although the differences are not always substantial, the consistency of the tendencies is noteworthy, suggesting a pandemic effect on amplifying political inequality.

(see Figure
Regarding controls, the effect of age is negligible, while that of gender denotes a bias in engagement in favor of men (consistent with prior conclusions of a gender gap in political participation; see e.g., Burciu and Hutter, 2023).Men are about twice as likely to feel politically efficacious and interested in politics.These results are consolidated by the remaining analysis.
As time is not the most valid operationalization for the effect of the pandemic, the analysis was re-run including more straightforward measurements, such as: how much the pandemic affected individuals' lives in terms of health and economy (thus only supporting data for 2021).A first analysis of all the countries in the sample generally concludes that neither having had COVID-19 nor having had economic losses as a result of the pandemic significantly relates to political engagement, neither per se nor differently across socioeconomic groups (see Table A6 and Figure A6 in the Online Appendix).
One good reason for this outcome is that the level of economic inequality prevailing in the country was not taken into account.More unequal societies are associated with decreasing levels of political participation and engagement, particularly among the less advantaged, thus promoting microlevel political inequality (Solt, 2008;Ritter and Solt, 2019;Schäfer and Schwander, 2019).Furthermore, the pandemic seems to have mostly deepened economic inequality in already unequal societies (Perugini and Vladisavljevic, 2021).Accordingly, if there is an effect of the pandemic on increasing political inequality, this is more likely to have occurred in more unequal societies.Hence, the analysis was once again re-run supporting a sub-sample of the most unequal countries, corresponding to a total of six countries11 and just over a third of the sample (Eurostat data for 2020). 12onsistent with the results in Table 2 (and also in Table A4 in the Online Appendix), two main conclusions can be drawn from Model 1 in Table 3.One is that income and education solidly maintain the previously observed effects on political engagement, with top income and education groups relating to increased odds of engagement, thus corroborating H1.The other is that by measuring the impact of the pandemic on people's lives, whether in terms of health or economics, almost no effect is observed (only with regard to having contracted COVID-19 is there a significant coefficient in the expected direction, indicating a decrease of 72 percentage points in interest in politics; the other significant coefficient suggests the opposite: having suffered economic loss due to COVID-19 increases party closeness by 25 percent).That is, similar to the results achieved using the year dummy, having experienced the impact of the pandemic does not seem to systematically affect the extent to which individuals, in general, engage in politics (only very partially supporting H2).
As before, interactions were then included in model 2.Although not all interactions are statistically significant, several are, meaning that experiencing health and economic costs due to the pandemic impacted political engagement across socioeconomic groups, especially concerning party closeness and political interest.
Although the probability of individuals feeling politically efficacious is rather low either before or after the pandemic outbreak, Figure 4 shows that having contracted the disease slightly reinforced the probability of feeling efficacious among those with more income and education (around two percent in both cases), thus also contributing to strengthening the gap between the   3. more and less privileged.This result not only meets the previous one using the year dummy but also converges with the findings of other studies (McBrayer et al., 2022).In addition, one likely consequence of such low probabilities is that citizens may distance themselves from politics even more.As they feel unconsidered by policymakers, politics is not a game they find useful to play (Solt, 2008).This may contribute to an increase in the cycle of political inequality, but also to the reinforcement of political disaffection (Torcal and Montero, 2006).
Finally, Figure 5 reiterates for political interest the trends already observed in the two previous variables.The probability of being interested in politics increased substantially among individuals in the fifth-income quintile as a result of being hit by the pandemic (between eight and 11 percentage points) and had a single-digit increase or decrease among those in the first quintile.Exceptionally, the diagrams relating to education groups do not corroborate the idea of a deepening gap in political engagement as a consequence of the pandemic.A decline in the probability of being interested in politics is observed among the more educated (about four and six percentage points), while the less educated show a maintenance of levels of interest in politics.

Robustness tests
In order to consolidate these findings, robustness tests for models in Tables 2 and 3 were run using different specifications for the socioeconomic stratum.The education group was coded so that 1 corresponded to the stratum with the highest education and 0 to the stratum with the lowest education.Likewise, the income group was coded so that 1 corresponded to the third tercile and 0 to the first.In both cases, the intermediate group was considered missing (corresponding to the middle level of educationi.e., upper secondary education levels -and the second tercile).Generally, the results reiterate that: (1) there is a steady socioeconomic gap with regard to political engagement; (2) the pandemic contributed to reducing the engagement of individuals with politics, but only in some circumstances; and (3) the pandemic contributed to deepening the socioeconomic gap in political engagement among the most unequal countries.The results are shown in Tables A4 and A6 and in Figures A5 and A7 3.
Although the increase in the distance between the more and less privileged generally remains small, the fact that it reflects a change over just two years and that it prevails despite the various operationalizations are per se reasons strong enough to uphold the importance of these results.Furthermore, levels of political inequality across socioeconomic groups have not always been large in past research (see e.g., Persson and Sundell, 2024).

Conclusion
Political equality is a basic democratic principle (Dahl, 1996).To the extent that political equality is not ensured for all groups in society, the democratic principle of political equality is corrupted (Lijphart, 1997;Verba, 2003).Challenging this view, the last few decades have witnessed that 'the unheavenly chorus has been singing with an upper-class accent' (Schlozman et al., 2018: 272).Unequal participation and representation remain quite common among democracies (e.g., Dassonneville and Hooghe, 2017;Lupu and Warner, 2022;Schakel and Van Der Pas, 2021), and even seem to be growing (Dalton, 2017).The upward drift in economic inequality raises the greatest concerns regarding the incremental trend of political inequality.The COVID-19 pandemic was the icing on the cake, as it exacerbated economic inequalities, more severely affecting the less well-off (Shibata, 2021) and the most unequal societies (Perugini and Vladisavljevic, 2021).This scenario should have heightened concerns about the growth of political inequality, given the reciprocal effect of economic and political inequality.However, almost no attention has so far been paid to this issue.This study aimed to make a contribution to filling this gap by focusing on political engagement.To a large extent, the results support the research hypotheses, reinforcing the previous pessimistic view of political inequality.
First, as largely acknowledged in research on participation and representation (e.g., Schäfer andSchwander, 2019, Persson andSundell, 2024), and consistent with similar work on income (Solt, 2008), a socioeconomic gap is also present regarding political engagement (thus supporting H1).There is a steady and robust gap in political engagement across income and education strata among the European countries under study for the three variables measuring engagement.The stratification based on education is particularly strong: an individual in the top education group more than triples the odds of feeling politically efficacious and more than doubles those of being interested in politics.
Second, the advent of COVID-19 depressed political engagement in some circumstances.Compared to 2018, there is some evidence of a drop in individuals' engagement with politics in 2021.The mean differences between both years for individuals' identification with a political party, political efficacy and political interest are generally statistically significant.Of course, this may be a result of a broader tendency that should not be read as an outcome of the pandemic.Though to assess it is behind the purposes of this study, the evidence gathered allows at least to state that the pandemic episode is not entirely innocent in this trend.Although most coefficients operationalizing the pandemic effect in regression models do not reach statistical significance, and some show an unforeseen sign, some correspond to expectations.For example, consistent results were achieved regarding the negative effect of the pandemic on political interest, and, although less consistently, regarding proximity to a political party (allowing to partially support H2).These results suggest that the pandemic may have especially affected less demanding dimensions of political engagement in terms of individual skills, in relation to which the most socioeconomically vulnerable had some relative advantage before the pandemic (as the descriptive data suggest, among these the level of political efficacy was already very low before the outbreak of the pandemic).
Third, the gap in political engagement across income and education groups tends to enlarge with the advent of COVID-19 (which is in line with results achieved for the gender gap; see Burciu and Hutter, 2023).In particular, having been negatively impacted by the pandemic, both in terms of health and economy, it had a greater impact on the lowest income and education groups, discouraging engagement but also, in some cases, increasing engagement among the most privileged.The outcome is the widening of the pre-pandemic gap in political engagement (converging with H3), especially in what concerns being close to a party.The increase in the gap at this level raises added concerns, given that party ties are recognized as a means of mitigating political inequality (Rosset and Kurella, 2021).Although the widening differences between the more and less privileged are generally small, which is reasonable as they regard a change over just two years, their pervasiveness and consistency across different operationalizations give them strong relevance.Indeed, small differences are not uncommon (e.g., Persson and Sundell, 2024).
In a nutshell, this research was able to demonstrate that: a socioeconomic gap exists regarding political engagement; the pandemic somewhat depressed political engagement; and it reinforced the prior income and education gap in engagement.These results have important implications for how citizens' involvement with politics is understood, as well as more broadly for the functioning of democratic institutions.On the one hand, they reinforce the idea of a vicious cycle and the growing character of political inequality, demonstrating that it is also observed upstream of effective participation.The inequality observed in engagement is an important piece in helping to explain the inequality of political participation.On the other hand, this research also allows us to signal the importance that external and uncontrollable events, such as a pandemic, can have on the political involvement of individuals and, more importantly, on the growth of political inequality between socioeconomic groups.Of course, fundamental determinants were left out in the analysis, due to the parsimony that it necessarily required, such as the level of pre-existing socioeconomic and political inequalities in the country, or the volume of public transfers with a view to mitigating the underlying economic crisis.Unconventional forms of political participation were also left out of the analysis, such as protest demonstrations, which were common in several countries during the pandemic period.Further study on political engagement inequality, as well as the effect of external shocks on political inequality, is well needed in order to validate, develop and deepen this line of research.

Figure 1 .
Figure 1.Distribution of the dependent variables by income quintiles and education levels (2018 and 2021).Source: ESS, Round 9 and 10.
Source: ESS, Rounds 9 and 10.Notes: Generalized linear logistic regression models with dependent variables: 1 = close to a political party; 0 = close to no political party; 1 = feel political efficacy; 0 = does not feel political efficacy; and 1 = is interested in politics, 0 = is not interested in politics.Probability distribution: binomial.Link function: logit.Subject effect: country.to higher engagement with politics and the reverse.Belonging to the top education group more than triples the odds of feeling politically efficacious and more than doubles those of being interested in politics.It also increases the chances of being close to a political party by 39 percentage points.Belonging to the fifth-income quintile increases the chances of being politically engaged by between 25 and 71 percentage points compared to belonging to other income groups.

Figure 2 .
Figure 2. Predicted probabilities of political engagement by education groups and year.Note: The estimates are from Model 2 in Table2.

Figure 3 .
Figure 3. Predicted probabilities of being close to a party by socioeconomic groups and pandemic effect.Note: The estimates are from Model 2 in Table3.

Figure 4 .
Figure 4. Predicted probabilities of feeling political efficacy by socioeconomic groups and pandemic effect.Source: The estimates are from Model 2 in Table3.
in the Online Appendix.

Figure 5 .
Figure 5. Predicted probabilities of being interested in politics by socioeconomic groups and pandemic effect.Source: The estimates are from Model 2 in Table3.