Introduction
In representative democracies, the act of voting enables citizens to shape policy outcomes by selecting representatives who align with their policy preferences (Dahl, Reference Dahl1989). This ensures that the policies implemented by elected officials accurately reflect the desires of the people they represent. Electoral participation is widely acknowledged as crucial for a healthy democracy for several reasons. First, higher participation ensures citizens’ voices are heard, fostering democratic responsiveness (Verba, Reference Verba1996). Second, voter turnout serves as an indicator of democratic quality (Roberts, Reference Roberts2009), democratic performance (Powell, Reference Powell1982), and system legitimacy (Mair, Reference Mair2023). Third, when fewer individuals participate in elections, those who vote may differ in their socio-economic profile or attitudes from those who abstain, potentially resulting in a group of voting citizens who are unrepresentative of the broader population (Lijphart, Reference Lijphart1997; Verba, Nie, and Kim Reference Verba, Nie and Kim1978; Smets and Van Ham, Reference Smets and Van Ham2013). Scholars consistently voice concerns about low turnout exacerbating participation inequality (Lijphart, Reference Lijphart1997), especially amid the sustained decline in voter turnout in Western democracies (Blais and Rubenson, Reference Blais and Rubenson2013; Kostelka and Blais, Reference Kostelka and Blais2021). Such concerns are justified, given studies indicating a growing disparity in electoral participation, with the less educated becoming less likely to vote over time (Armingeon and Schädel, Reference Armingeon and Schädel2015; Dalton, Reference Dalton2017; Dassonneville and Hooghe, Reference Dassonneville and Hooghe2017).
Research consistently shows that individuals with lower formal education, lower economic status, and younger age tend to abstain from voting at higher rates (Blais, Dassonneville, and Kostelka; Lijphart, Reference Lijphart1997; Smets and Van Ham, Reference Smets and Van Ham2013; Nevitte et al., Reference Nevitte, Blais, Gidengil and Nadeau2009). This pattern holds true in both the United States (Leighley and Nagler, Reference Leighley and Nagler2013) and Europe (Gallego, Reference Gallego2007). If certain groups participate in elections to a lesser extent, it could mean they have less political influence, highlighting a lack of political equality among those who vote and abstain, hence the term ‘unequal participation’ (Lijphart, Reference Lijphart1997). This raises concerns about inadequate representation for non-voters and the potential distortion of the political process in favour of specific groups (Verba, Reference Verba2003). Unequal turnout becomes particularly problematic if voters’ policy preferences differ from non-voters, challenging the idea that election results truly represent the choices of all citizens, as they may instead reflect the preferences of specific segments of the population. Considering the decreasing turnout, these differences between voters and non-voters may hold greater significance for democratic representation.
Turnout inequality may not only reflect existing disparities but also reinforce them. A growing body of research shows that voting is habit-forming: individuals who vote are more likely to vote again, while abstention reinforces disengagement (Green and Shachar, Reference Green and Shachar2000; Gerber, Green, and Shachar, Reference Gerber, Green and Shachar2003). Early participation – or non-participation – can crystallise into long-term patterns, particularly as young people experience their first election (Bhatti, Hansen, and Wass, Reference Bhatti, Hansen and Wass2016; Franklin, Reference Franklin2004; Plutzer, Reference Plutzer2002). These habits are difficult to reverse and typically shift only through generational replacement. Once participation gaps emerge, they tend to persist. Moreover, voting itself may influence political attitudes. Casting a ballot can have ‘upstream’ and ‘downstream’ effects on political interest, knowledge, and partisan attachment (Franklin, Reference Franklin2004; Dinas, Reference Dinas2014), which may in turn shape policy preferences or partisan loyalties (Dinas, Reference Dinas2014; De la Cuesta, Reference De la Cuesta2023). Electoral participation, then, may not purely be reflective of pre-existing preferences – it may also shape them. This raises the possibility that turnout gaps could produce or amplify differences in policy views between voters and non-voters, and that these differences, once established, may become self-reinforcing.
Previous studies used socio-economic characteristics and vote choice as proxies to examine policy preference gaps between voters and non-voters. The assumption that different groups inherently have distinct policy preferences was plausible in post-industrialised democracies characterised by strong unionisation and frozen class-based cleavages. However, the increasing multidimensionality of European democracies (Kriesi et al., Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006, Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2008; Rovny and Edwards, Reference Rovny and Edwards2012; Marks et al., Reference Marks, Hooghe, Nelson and Edwards2006; Bakker, Jolly and Polk, Reference Bakker, Jolly and Polk2012; Prosser, Reference Prosser2016) and the decline in union membership (Gray and Caul, Reference Gray and Caul2000) have diminished the reliability of class-based cleavages as indicators of political preferences or choice (O’Grady and Abou-Chadi, Reference O’Grady and Abou-Chadi2019; Caughey, O’Grady and Warshaw, Reference Caughey, O’Grady and Warshaw2019).
In these evolving contexts, citizens hold conflicting preferences across various dimensions of political conflict (Lefkofridi, Wagner, and Willmann, Reference Lefkofridi, Wagner and Willmann2014; Hillen and Steiner, Reference Hillen and Steiner2020; Mudde, Reference Mudde2007) and often resort to trade-offs between their policy preferences (Kirkizh, Froio, and Stier, Reference Kirkizh, Froio and Stier2022), challenging assumptions about automatic support for parties linked to traditional group interests. Consequently, observed differences in participation among socio-economic groups may not necessarily mean that their policy preferences also differ, potentially leading to unequal political influence on desired policy outcomes. Given the complexity of today’s multidimensional political landscape, understanding the potential impact of low voter turnout requires moving beyond traditional indicators such as vote choice or left-right ideological positions.
State of literature
It is argued that those who regularly participate in elections (ie core voters) tend to have stronger predispositions compared to those who only vote from time to time (ie peripheral voters) or those who never vote (ie perpetual non-voters) (Campbell, Reference Campbell1960; Hansford and Gomez, Reference Hansford and Gomez2010; Burnham, Reference Burnham1965).Footnote 1 That is, regular voters differ from both peripheral voters and perpetual non-voters in terms of policy preferences. However, empirical evidence is rather mixed and geographically limited.
In the United States, findings from 1972 show that American voters and non-voters share similar preferences (Wolfinger and Rosenstone, Reference Wolfinger and Rosenstone1980). Addressing the same question, Highton and Wolfinger (Reference Highton and Wolfinger2001), analysing data from 1992 and 1996, find the differences to be modest, albeit present. In contrast, Griffin and Newman (Reference Griffin and Newman2005) find that in 24 out of 44 American states, voters are more conservative than non-voters, while Shaffer (Reference Shaffer1982) finds some conservative bias only on economic issues, not on social issues in the United States. Moreover, although voters and non-voters share similar preferences on most issues, Bennett and Resnick (Reference Bennett and Resnick1990) find that support for government spending is higher among non-voters. Analysing data spanning from 1972 to 2008, Leighley and Nagler (Reference Leighley and Nagler2013) provide evidence that voters are substantially more conservative on redistributive politics, which stands in stark contrast to previous works’ findings (Wolfinger and Rosenstone, Reference Wolfinger and Rosenstone1980; Highton and Wolfinger, Reference Highton and Wolfinger2001).
Evidence is also mixed in Canada. Rubenson et al. (Reference Rubenson, Blais, Fournier, Gidengil and Nevitte2007) find minimal differences in policy preferences between voters and non-voters. In contrast, Godbout and Turgeon (Reference Godbout and Turgeon2019) find that non-voters differ more substantively from voters in Canada in that they hold more conservative policy preferences.
As for the European context, there is evidence from Germany that non-voters and voters do not differ on their left-right, economic, and cultural positions, but they do differ on the issue of immigration (Koch, Meléndez, and Rovira Kaltwasser, Reference Koch, Meléndez and Rovira Kaltwasser2023). At the comparative level, the studies only focused on differences in overall left-right positions. There is evidence that non-voters tend to be more centrist than voters (Rodon, Reference Rodon2017). Moreover, focusing on 36 OECD countries, Dassonneville et al. (Reference Dassonneville, Feitosa, Hooghe and Oser2021) find that the median voter is more right-wing than the median citizen, indicating a right-wing bias among voters.
Given the prevalence of mixed findings in the literature and the lack of comparative evidence in Europe beyond the left-right scale, the question of whether voters and non-voters differ in policy preferences in Europe remains unanswered. While the most comprehensive evidence comes from the United States, European democracies differ fundamentally in how politics is structured, making it challenging to draw inferences from the American context. This research note addresses that gap by being the first study to systematically examine differences in policy preferences between voters and non-voters across 29 European democracies. It analyses attitudes not only on the left-right scale but also on economic and social issues and European integration.
Data and methods
I use data from the 2014 and 2019 survey waves of the European Election Study (EES), encompassing 29 countries (see Figure 1 for the list of countries). These surveys (Schmitt et al., Reference Schmitt, Hobolt, van der Brug and Popa2022, Reference Schmitt, Hobolt, Popa, Teperoglou and Parliament2024) are well-suited for this study as they coincide with European Parliament Elections rather than national elections. Asking respondents about their turnout outside national election contexts helps mitigate over-reporting, as social desirability to over-report decreases as election salience decreases.Footnote 2
Reported and actual voter turnout across European democracies.
EES surveys include items capturing the traditional left-right stance, economic attitudes (ie attitudes on redistribution, state intervention in the economy, taxation for public services), social attitudes (ie same-sex marriage, immigration, environment, and law and order), and European integration. Each policy item serves as a dependent variable and is measured on an 11-point scale, with values coded to range from 0 (indicating a liberal position) to 10 (indicating a conservative position). More specifically, lower values on the scale indicate a more left-wing position, support for redistribution, state intervention in the economy, higher taxes for more social spending, same-sex marriage, immigration, protection of privacy rights, and further European integration. Specific question wordings, descriptive information such as distribution of these variables, and survey-level correlations between these items are available in online Appendix C.
The independent variable is the reported voter turnout at the last national election. Figure 1 depicts the reported turnout in the survey (hollow circles) compared to the actual turnout in the elections (circles). The reported turnout ranges from 54% in Poland to 94% in Malta with an average of 79% in the overall sample. Actual turnout varies from 38% in Romania to 93% in Malta with an average of 66% across the entire sample. While there are instances where reported turnout closely aligns with the actual turnout (eg less than one percentage point in Cyprus and Italy), there are also cases where the difference is larger (eg Romania and France with a difference of 43 and 32 percentage points, respectively). This variation could be attributed to factors such as over-reporting or politically interested people being more likely to participate in surveys (Dahlgaard et al., Reference Dahlgaard, Hansen, Hansen and Bhatti2019; Ansolabehere and Hersh, Reference Ansolabehere and Hersh2012).Footnote 3 Overall, Figure 1 shows that there is sufficient variation in the key independent variable, reported voter turnout, allowing for meaningful analysis in the study.
I estimate ordinary least squares models where I regress each item on reported turnout using survey fixed effects. Since contextual and temporal factors also influence turnout and vote choice (De la Cuesta, Reference De la Cuesta2023; Dunaiski, Reference Dunaiski2021), all models include survey-level fixed effects, capturing country-specific and election-specific particularities. Given that the objective of this research is to determine whether voters and non-voters differ in policy preferences, rather than recovering the effect of voting on policy views, these models are estimated without controls for socio-economic characteristics.Footnote 4
Results
Table 1 presents the results of the models that regress each policy preference on reported turnout. The general finding is that voters differ from non-voters on the left-right scale, economic attitudes, social attitudes, and EU integration, with the preferences on environmental protection being an exception. For instance, on the left-right scale, voters are, on average, more right-wing than non-voters (0.22 points). On the economic attitudes, voters are less in favour of wealth redistribution from the rich to the poor (0.25 points), but they are more supportive of state intervention in the economy (0.17 points) and of raising taxes to increase public services (0.28 points). Similarly, on the social attitudes, voters have more conservative views on same-sex marriage (0.17 points) and more restrictive views on immigration (0.11 points), more in favour of restricting privacy rights (ie civil rights) to combat crime (0.34 points). However, they do not differ from non-voters in terms of their views on the environment. Lastly, on European integration, voters are more pro-EU integration (0.24 points), supporting European unification to be pushed further.
Do the policy preferences of voters and non-voters differ?
Note: Standard errors in parentheses. *
$p \lt 0.05$
, **
$p \lt 0.01$
, ***
$p \lt 0.001$
. Lower values indicate left-wing/liberal views.
Although the differences between voters and non-voters are statistically significant in the pooled analyses in Table 1, with the exception of environment, the substantive interpretation of such effects and their magnitudes requires additional analyses. In what follows, I report (1) p-values for survey-level associations to see the prevalence of these differences, (2) Cohen’s d for these associations to interpret the magnitude of effect sizes, and (3) a simulated hypothetical full turnout scenario, under the assumption of preference stability, to examine whether and the extent to which the electorate’s preference would be different.
First, I check whether the effect is indeed larger at the survey level, but inconsistent directions across surveys mitigate or cancel it out on average. Survey-level coefficient plots, available in Appendix E, suggest that this is not the case. Moreover, given the large sample size, detecting small effects becomes inherently easier even though these effects would not be detected at the survey level. Figure 2 plots the distribution of p-values for the coefficients representing being a voter from these survey-level estimations. The majority of associations have p-values above the 0.05 threshold (to the right of the vertical dashed line), suggesting that in most surveys, voters and non-voters share similar preferences. That is, the presence of disparities is not typical and tends to be sporadic when observed.Footnote 5
Distribution of p-values from survey-level models.
Note: P-values are from survey-level models (see Appendix E for details).
Second, I standardise these gaps using Cohen’s d to interpret the magnitude of differences in policy preferences.Footnote
6
Typically, a Cohen’s d smaller than 0.2 is considered trivial, between 0.2 and 0.5 small, between 0.5 and 0.8 medium, and larger than 0.8 large. Figure 3 displays d statistics for survey-level estimates for each item, with horizontal lines indicating values of 0.2 and 0.5, between which small effects lie. Black dots indicate statistically significant differences (
$p \lt 0.05$
), and light grey dots indicate non-significant ones (
$p \gt 0.05$
). The pattern is clear. First, most associations are trivial (
$d \lt 0.2$
) and fail to reach conventional significance (
$p \lt 0.05$
), echoing the earlier finding that disparities are rare. Even if these effects are detected in larger samples, as seen in the pooled analyses, they would still indicate trivial effects. Second, when statistically significant differences do emerge, their d values usually fall between 0.2 and 0.5 – small but not trivial. For illustrative shifts in preference distribution based on Cohen’s d, see Appendix F.
Effect sizes based on Cohen’s d.
Note: Cohen’s d statistics of survey-level estimates are presented.
Finally, to contextualise the potential implications of this disparity between voters and non-voters, it is necessary to look beyond the observed differences between voters and non-voters.Footnote 7 The bias such differences introduce is a function of the turnout level. For instance, even if the gap is large, the bias would be minimal if the turnout is high, whereas a small difference under low turnout would introduce a larger bias.
$${\rm{BIAS}} = mea{n_{voter}} - \left( {\left( {mea{n_{voter}} \times proportio{n_{voter}}} \right) + \left( {mea{n_{nonvoter}} \times proportio{n_{nonvoter}}} \right)} \right)$$
To quantify this bias, I follow the methodology indicated in equation 1. Here, I weigh the average positions of voters and non-voters based on their respective shares determined by the turnout rate in the sample. This is equivalent to the mean citizen position, including both voters and non-voters. Comparing the weighted citizen position with the average voter position captures the extent of bias arising from voter and non-voter disparities, contingent on the turnout rate. This bias essentially gives us how different the electorate’s preferences would be if non-voters also voted, assuming that their preferences would be similar.Footnote 8
Figure 4 shows that most biases are below 0.3 points under a hypothetical full turnout. Negative values indicate that right-wing (conservative) biases would be reduced, and positive values indicate that left-wing (liberal) biases would be reduced. Black dots indicate the associations that are statistically significant (
$p \lt 0.05$
), while gray dots indicate those that do not reach statistical significance. While these are by no means trivial differences, they are relatively minor. However, it is important to exercise caution before concluding that these biases are unlikely to lead to a misrepresentation of public preferences.
How much would the average electorate position change if non-voters had voted?.
Note: Bias is estimated based on equation 1 in the text. Negative values indicate that right-wing (conservative) bias among voters would be reduced, while positive values indicate left-wing (liberal) bias would be reduced under a hypothetical full turnout. Dark dots indicate cases where voter and non-voter differences are statistically significant (
$p \lt 0.05$
) in survey-level models.
The estimated disparities between observed average voter positions and the hypothetical average positions of the electorate if non-voters were to vote suggest biases in the sample, typically extending up to 0.3 points. Among significant associations, in Denmark in 2019, with an 89% turnout rate, the average voter leaned 0.8 points more right-wing than the average non-voter, a bias that would decrease by 0.09 points under full turnout. Furthermore, in Northern Ireland in 2014, where turnout was 60%, voters had a 1.6-point conservative bias on same-sex marriage, which would be reduced by 0.6 points under full turnout. Or, a 1.3-point pro-Europe bias in Germany in 2014, with an 81% turnout, would be reduced by 0.24 points. Conversely, voters in the United Kingdom in 2014 tended to be less pro-Europe by 0.9 points, and this bias would be reduced by 0.32 points under full turnout.
Figure 5 illustrates that, when disparities emerge (highlighted by significant associations in black), these biases, regardless of their direction, tend to be more pronounced when turnout is lower. This suggests that the electorate’s policy preferences are less reflective of citizens’ preferences when turnout is low.
Bias is larger when turnout is lower.
Note: Bias is estimated based on equation 1 in the text, without distinguishing the direction of the bias. Bias on each item is pooled together. Dark dots indicate cases where voter and non-voter differences are statistically significant (
$p \lt 0.05$
) in survey-level models.
Overall, these biases are not negligible, especially since the simulated biases under full turnout rely on reported turnout, which is typically higher than actual turnout (see Figure 1). This means the proportion of non-voters in the simulation is understated, potentially leading to an underestimation of the true distortion. Given that election outcomes are often decided at the margins, even small biases could substantially affect the ideological composition of governments, thereby influencing ideological representation.
This analysis assumes that non-voters’ policy views would remain unchanged if they voted (Grofman, Owen, and Collet, Reference Grofman, Owen and Collet1999). However, the act of voting may increase political engagement and information-seeking, potentially influencing policy preferences (Franklin, Reference Franklin2004). To examine this, Appendix I draws on British Election Study Internet Panel data (Fieldhouse et al., Reference Fieldhouse, Green, Evans, Mellon, Prosser, Bailey, de Geus, Schmitt and van der Eijk2023) and finds that first-time eligibility does not lead to different policy views on general left-right, redistribution, immigration, and European integration. These null findings mirror those of Holbein and Rangel (Reference Holbein and Rangel2020), Jessen, Kuehnle, and Wagner (Reference Jessen, Kuehnle and Wagner2021), and broader meta-analyses of voting’s transformative effects (Holbein et al., Reference Holbein, Rangel, Moore and Croft2023). Even so, the hypothetical higher-turnout scenarios should be interpreted cautiously – as approximations of what the electorate’s average preferences would look like if non-voters voted but held the views they reported at the time of the survey.
To further examine how turnout levels influence preference biases without assuming preference stability, I use panel data to compare policy preferences among perpetual non-voters, peripheral voters, and regular voters. Regular voters consistently vote, perpetual non-voters consistently abstain, and peripheral voters vote intermittently (Bhatti et al., Reference Bhatti, Dahlgaard, Hansen and Hansen2019). Based on panellists’ information on the 2015, 2017, and 2019 General Elections in the British Election Study Internet Panel data, I classify panellists as regular voters if they participated in all three elections, peripheral voters if they participated in at least one election, and perpetual non-voters if they abstained during all three elections. Differences between these groups would further demonstrate that biases increase as turnout declines, reinforcing the findings presented in Figure 5.
Figure 6 presents the results of regression models where policy views are regressed on voter type separately. On every item, regular (N = 23,305) voters differ from peripheral voters (N = 5491), and this difference is even more pronounced when compared to perpetual non-voters (N = 1842). Perpetual non-voters and peripheral voters are more left-wing, more supportive of redistribution and environmental protection, more positive toward the EU, but less supportive of spending and less positive toward immigration. In sum, there is evidence that most of these divergences are starker between perpetual non-voters and regular voters. See Appendix J for detailed analyses and descriptions of each voter type. In line with Figure 5, these results are direct evidence that under higher turnout, the differences in policy views would be smaller.
Preferences of perpetual non-voters (PNV), peripheral voters (PV), and regular voters (RV).
Note: Predictions are based on models from Appendix J. 95% confidence intervals are indicated.
Conclusion
Representative democracies rely on citizens’ electoral participation to ensure policies reflect public preferences. By systematically comparing policy preferences of voters and non-voters across the left-right spectrum, economic and social policies, and on European integration in 29 European democracies, this study provides the first large-scale evidence in European democracies that voters and non-voters usually share similar policy preferences, though occasional meaningful disparities emerge. This provides optimistic results in that election results can usually serve as a reliable proxy for citizen preferences, even amid low turnout levels. However, when disparities emerge, they are not trivial in terms of effect size, resulting in meaningful differences between voter and non-voter preferences, even when comparing voters and non-voters within the same socio-economic characteristics (see Appendix D). Especially under low turnout, representation biases arise, skewing the electorate’s policy preferences away from those of the broader citizenry. This highlights the potential representation costs of declining turnout.
Moreover, do these differences, when they emerge, cause a misrepresentation of citizens’ preferences? The answer is nuanced. While high turnout can enhance the quality of representation by mitigating potential biases stemming from differences between voters and non-voters, the answer depends on both the turnout level and the extent of difference between voters and non-voters. If non-voters do not differ significantly from voters even in low-turnout elections, then low turnout may not pose a problem for representation. However, if low turnout is accompanied by existing disparities between voters and non-voters, it potentially becomes problematic, distorting the citizens’ preferences and rendering the electorate less representative of the citizenry. This would imply that low voter turnout could significantly impact representation by sacrificing representation for non-voters.Footnote 9
Furthermore, this study offers insights into voter types and whether the preferences of perpetual non-voters, peripheral voters, and regular voters differ. Understanding these distinctions helps interpret the sporadic disparities observed. Since these three groups typically hold distinct policy preferences, mobilising peripheral voters can shift the electorate’s overall stance. Their participation in high-turnout elections expands representation compared to low-turnout elections, where the electorate may skew toward regular voters. This raises questions about how accurately elections reflect public preferences and underscores the pivotal role of peripheral voters, making them a key target for party strategists.
Perpetual non-voters abstain even in high-turnout elections, but their potential to skew the outcomes hinges on both their numbers and how strongly their views diverge from those of voters. In the British context, data show that perpetual non-voters make up about 7%; Danish data put the figure at 9% (Bhatti et al., Reference Bhatti, Dahlgaard, Hansen and Hansen2019). At such levels, abstention would meaningfully bias the electorate’s mean only if these citizens held markedly different preferences, which is not the case.Footnote 10 Mobilising this group is difficult but not impossible: in the 2016 referendum on whether to remain or leave the European Union, approximately 25% of perpetual non-voters turned out, versus 69% of peripheral voters and 98% of regular voters.
The sporadic yet meaningful gaps between voters and non-voters should be understood in relation to both voting habits and institutional rules such as compulsory voting. When turnout is low, these differences exacerbate representation bias – that is, the voting electorate diverges more from the broader citizenry. Under compulsory voting, however, even occasional voters and perpetual non-voters are drawn to the polls often enough to develop a voting habit. Hence, compulsory voting likely realigns the core electorate with the preferences of the overall population. Dunaiski (Reference Dunaiski2021), for example, shows that habit-driven turnout differentials vanish under Brazil’s compulsory system, suggesting that mandates accelerate habit formation across the citizenry. In light of this research’s findings, compulsory voting is expected to reduce representation bias by incorporating more of the population’s policy views into election outcomes.
Do abstainers lack ideological preferences? Non-voters, both perpetual non-voters and peripheral voters, are less likely to place themselves on the left-right spectrum, suggesting a reduced inclination to view politics through an ideological lens. Yet, many non-voters express preferences on political issues, indicating that European citizens typically possess attitudes rather than mere non-attitudes (see Appendix K). The absence of an ideological orientation and partisan attachments, combined with the presence of issue preferences, likely explains peripheral voters’ electoral volatility. Given that volatility is linked to parties’ ideological responsiveness to the average voter (Dassonneville, Reference Dassonneville2018), peripheral voters’ intermittent participation might contribute to increased responsiveness from parties. Party strategists aiming to mobilise non-voters should focus on specific issues rather than broad ideological and partisan campaigns.
Because policymakers respond primarily to voters’ preferences (Dassonneville et al., Reference Dassonneville, Feitosa, Hooghe and Oser2021; Hakhverdian, Reference Hakhverdian2012), election results can overweight the preferences of those who vote. Elites, believing these signals reflect ‘public opinion’, then tailor policies to voters, reinforcing their own misperceptions (Broockman and Skovron, Reference Broockman and Skovron2018; Pereira, Reference Pereira2021). This feedback loop can demobilise non-voters and amplify even small differences between voters and non-voters, distorting representation beyond the raw preference differences documented in this research.
Suggestive evidence indicates that gaps between voters and non-voters on certain policy preferences might be driven by the mobilisation or demobilisation of citizens with differing preferences. Appendix L shows that, in line with De la Cuesta (Reference De la Cuesta2023), political context influences voting experiences or who is mobilised. For example, there tends to be a right-wing bias among voters only when the outgoing government is centrist or right-wing. However, this bias is absent under left-wing outgoing governments. This is consistent with top-down mobilization, where parties shape citizens’ policy preferences (Moral and Best, Reference Moral and Best2023), and corroborates the importance of political context in shaping voting experience (De la Cuesta, Reference De la Cuesta2023). While not conclusive for each policy, they invite further research on how party behaviour shapes voter–non-voter differences.
While this research note descriptively documented that differences might arise sporadically, future work should focus on the drivers of such differences. Institutional features, such as electoral rules (proportional representation vs first-past-the-post), political system (presidential versus parliamentary), and party competitiveness or polarisation, as well as contemporaneous events such as austerity debates in the United Kingdom or the refugee crisis on the continent, can amplify or mute these gaps and therefore warrant attention in future research.
One potential concern is that reported turnout may not be a reliable measure, or respondents may retrospectively align their reported turnout with their policy preferences, potentially introducing bias (Abelson, Loftus, and Greenwald, Reference Abelson, Loftus and Greenwald1992; Presser, Reference Presser1990; Selb and Munzert, Reference Selb and Munzert2013). However, supplementary analyses provided in Appendix M, which compare results based on reported versus validated turnout, reveal substantial similarity, alleviating the concerns of measurement bias.
Another potential concern is that issue opinions may not be stable, and that the observed differences could simply reflect random noise rather than meaningful differences in underlying attitudes. However, Appendix N addresses this concern by using a reduced-dimension approach to assess whether voters and non-voters differ in their latent predispositions. The results from these analyses counter this concern.
A note of caution is warranted. The findings that no or only small differences between the two groups in most cases should not be interpreted as diminishing the significance of voting or downplaying the potential consequences of low turnout. Voting is a fundamental aspect of democratic participation and plays a vital role in the quality of democracy. Electoral participation can foster democratic citizenship and improve democratic functioning. For instance, research shows that voting can increase satisfaction with democracy (Kostelka and Blais, Reference Kostelka and Blais2018), enhance external efficacy (Finkel, Reference Finkel1985), foster civic duty (Feitosa, Blais, and Dassonneville, Reference Feitosa, Blais and Dassonneville2020), and strengthen partisan attachments (Dinas, Reference Dinas2014) that (1) make citizens more politically engaged (McAllister, Reference McAllister2020) and (2) stabilise the political system (Dalton, Reference Dalton2016).
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1475676525100388.
Acknowledgements
I would like to thank André Blais, Ruth Dassonneville, Markus Wagner, Isabella Rebasso, Alexander Dalheimer, Elena Heinz and Jochem Vanagt for their comments and suggestions throughout the development of this paper. The research was supported by the European Research Council (grant: 101044069). I also thank the editors of EJPR and the anonymous reviewers for their constructive comments.
Funding statement
None.
Competing interests
None.
Open access
