Too much to choose from? The long-term effects of political fragmentation on electoral turnout

Abstract

How does political socialization in a highly fragmented political scene affect propensity to vote? This article focusses on the long-term relationship between the number of political parties and the propensity to turn out in 96 parliamentary elections between 1996 and 2016 of nearly 100,000 individuals in 31 countries. Although intuitively more options might be expected to translate into a greater likelihood of participating in elections, existing research claims that high levels of party fragmentation instead lead to ‘choice overload’ and alienate citizens from voting. Building on the theory of voting as a habit, I show that early adulthood political socialization in a highly fragmented context leaves a footprint of non-voting in subsequent elections. This finding is especially relevant given the recent significant rise in fragmentation of most party systems in Europe, which in light of this research could mean a further decline in turnout rates in many countries in the future.

Keywords

effective number of electoral parties electoral turnout party system fragmentation political socialization voting habit

Introduction

It is a common place in the literature that electoral turnout depends to a great extent on the kind of choice people are offered in elections, and that choice is structured by the party system (Blais, 2000: 30). However, the findings and theories about the impact of party system fragmentation on turnout are contradictory. In some accounts, more parties equals higher mobilization and thus higher turnout, whereas it could also lead to uncertainty, information deficits and lower turnout. The results of empirical tests of the effects of the number of parties on aggregate levels of turnout are inconclusive – some find a positive impact, others negative, while most do not find any significant effect at all (Cancela and Geys, 2016; Geys, 2006; Stockemer, 2017).

However, existing studies on the effects of fragmentation have focussed on the short-term impact at the aggregate level, by correlating the number of parties at an election and turnout rates in that particular election, or the following one. We know much less about the effects of fragmentation in the long run, and how this affects propensity to vote at individual level. In Smets and van Ham’s (2013) meta-analysis of an embarrassment of riches, which analysed a number of factors affecting individual-level propensity to vote, they showed that extant research has produced equally mixed evidence regarding the short-term effects of fragmentation. This article aims to fill this gap by focussing on the long-term effects of fragmentation on individual propensity to turn out in subsequent elections.

Building on the theories of voting as a habit (Franklin, 2004; Plutzer, 2002) and of political socialization (Jennings and Niemi, 1968; Neundorf and Smets, 2017), this article assesses the effects of the number of political parties in the first elections at which one is eligible to vote (i.e. crucial for the formation of voting habit) on the propensity to participate in subsequent elections. Following the idea that the effects of contextual-level variables on turnout are conditional on individual-level characteristics,¹ I test whether fragmentation especially affects the likelihood to turn out of the young. The socializing effects of many contextual-level variables (e.g. election competitiveness, electoral system, general turnout rates, and polarization) have already been tested (Franklin, 2004; Smets and Neundorf, 2014). This study focusses, for the first time to my knowledge, on the long-term effects of fragmentation; thereby aiming to contribute to a rich body of literature that researches the imprint that an individual’s first election leaves on future turnout decisions.

Drawing on Comparative Study of Electoral Systems (2018) Integrated Modules Dataset (CSES IMD) data on 96 parliamentary elections between 1996 and 2016 of nearly 100,000 individuals in 31 countries, I show – using multilevel fixed and random effects logistic regression models – that exposure to a high number of parties in the first elections at which one is eligible to vote leaves a footprint of non-participation in the case of both Western and Central and Eastern Europe (CEE); with an especially strong effect for the latter. The results are robust to several tests: including measures of disproportionality and polarization in contemporary elections, and formative ones. This continues to be the case when controlling for contemporary fragmentation; assessing parliamentary fragmentation, and not electoral; limiting the sample to more recent cohorts; including older citizens socialized under authoritarian rule; and to sensitivity checks of the age at which the effect is strongest. These findings are especially relevant given the recent significant rise in fragmentation of most party systems in Europe, as one of its unforeseen long-term consequences might be a further decline in turnout, which has already been falling for decades in many democracies.

Theoretical framework and hypotheses

Party system fragmentation and turnout

What influence does the number of political parties have on electoral turnout? There are competing theoretical arguments in favour of both a positive and negative relationship. As mentioned above, intuition suggests that the more options one has to choose from, the higher the likelihood of participation. More parties maximize choices on the ballot paper and increase the probability of voters identifying with a particular party (Geys and Heyndels, 2006). For instance, Seidle and Miller (1976) showed that turnout was higher if there were three parties instead of two. It has also been argued that the appearance of smaller single-issue parties has a mobilizing effect, at least for the case of local elections in Norway (Hansen, 1994). Thus, some scholars argue that ‘larger party systems help spur participation’ (Brockington, 2004: 485).

On the other hand, it has been argued that high levels of fragmentation lead to confusion and alienate citizens from casting a ballot (Blais and Dobrzynska, 1998; Huber et al., 2005). With many parties, information becomes diluted and the cognitive costs of voting increase, while party differential is lower. Moreover, not all choice alternatives form part of voters’ ‘consideration sets’. As Vassil et al. (2016: 89) argue, ‘having more bad choices provides few incentives to choose any of them – especially if abstention is an easy option’. After all, human rationality is bounded (Kahneman, 2003) and too many options might lead to choice overload (Chernev et al., 2015; Schwartz, 2004) and, consequently, to choice deferral (Iyengar and Lepper, 2000). Furthermore, the existence of many parties may cause considerable difficulties in forming stable majority coalitions. As coalition bargaining becomes more difficult, it is harder to vote strategically and ensure that the preferred policy will be implemented (Jackman, 1987). Broad coalitions may diminish the legitimacy of the political system and depress political efficacy, altogether discouraging voter participation (Karp and Banducci, 2008). Hence, in Downsian (1957) terms, a high number of parties reduce the probability that an individual’s vote will be decisive, and increases the cognitive costs of making up one’s mind on whom to vote.

Another line of argument argues that the relationship between the number of parties and turnout is curvilinear (Norris, 2002; Taagepera et al., 2014). According to this approach, turnout rises as the number of parties increase, but only up to a point before gradually falling. Participation is depressed both by excessive fragmentation and by one-party predominance. For instance, Grofman and Selb (2011) showed – for the case of local elections in Switzerland and Spain – that turnout rises with an increase in effective number of electoral parties (ENEP)² up to about 3; after which the chances of an individual casting a ballot start to decrease (see also Taagepera et al., 2014: 409). Notably, this approach combines both the positive and negative effects found by other scholars.

All in all, there is little evidence supporting a linear positive relationship between fragmentation and electoral participation. In their meta-analyses of aggregate-level determinants of turnout, Geys (2006), Cancela and Geys (2016), and Stockemer (2017), found that roughly from a fifth to a third of studies had succeeded in confirming the positive relationship between number of parties and electoral participation. On the other hand, Blais (2006: 118), after analysing the existing literature, concluded that ‘almost all empirical research has found a negative correlation between the number of parties and turnout’. For instance, in the context of CEE, an increase in ENEP of 1 translates into a 2.4% drop in turnout (Kostadinova and Power, 2007: 369). At the individual level, the meta-analysis of Smets and van Ham (2013) does not clarify the picture, as they also found mixed evidence. However, all things considered, the empirical evidence is relatively more robust on the negative effects of fragmentation on turnout.

Voting as a habit: Socialization into participation

This study follows an influential electoral turnout theory, which contends that voting is a habit (Converse, 1969; Dinas, 2012; Franklin, 2004; Plutzer, 2002). People acquire it by participating, or not, in the first few opportunities they have to cast a ballot. Depending on the competitiveness of those elections, the youngest cohorts of the electorate decide to cast a ballot, or not. Their decision accustoms them to voting or not-voting, and this is crucial for future electoral participation, as it leaves a reasonably stable ‘footprint’ on voting behaviour (Franklin, 2004: 60–61). According to this approach, the propensity to vote or to fail to participate in subsequent elections is contingent on the contextual factors specific to those first few elections in which one is eligible. On top of electoral competitiveness (closeness of the race), Franklin (2004) and Smets and Neundorf (2014) used, as key first election-specific factors: measures of party cohesiveness, electoral system, general turnout rates, polarization, and presidential approval rates. General turnout has been found to have a lasting positive impact on a cohorts later electoral participation, while the margin of victory (closeness of the race), majority status (size of the largest party) are said to reduce propensity to turn out in the long-term.

The focus of this article is on party system fragmentation, as this factor has been neglected in the existing studies of habitual voting (probably due to the ambiguity of its effects found in the literature), although it is a feature of the ‘character of elections’ (Franklin, 2004: 6) that might leave a footprint on future behaviour, like any other. This not to say that the effect of the number of parties is independent of other contextual characteristics of the formative elections. For one, the quality of party competition has been claimed to be empirically more important than the number of parties in explaining turnout (Dalton, 2008; Kittilson and Anderson, 2011). At least in the short-term, party polarization is supposed to increase one’s propensity to cast a vote (Crepaz, 1990). Second, turnout rates tend to be higher in more proportional electoral systems (Blais and Aarts, 2006) and thus electoral system features should also be controlled for. Third, compulsory turnout is the single institutional factor that fosters the greatest turnout (Blais and Dobrzynska, 1998; Jackman, 1987), and it must also be taken into account.

In Wass and Blais’ (2017: 463) model of the ‘funnel of causality of turnout’, they classify the institutional and contextual characteristics of particular elections as distant causes (with individual characteristics being more proximate and rational turnout decision being more immediate). This article is based on a further layer of even more distant causes – the contextual characteristics of elections that occurred during one’s formative period – but analogous contextual factors of contemporary elections should also be controlled for, to more accurately estimate the long-term effects.

The idea of habituation to participation and of the ‘footprint’ these crucial first few elections leave on subsequent behaviour is parallel to the argument of the persistence of political identities, values, and behaviour acquired through the process of political socialization (Neundorf and Smets, 2017). People develop their political attitudes and behaviour through agents like parents, family, schools, and peers (Jennings and Niemi, 1968). According to most authors, political socialization starts at the age of about 14 or 15, when adolescents start to recognize the political world (Jennings and Niemi, 1981; Mishler and Rose, 2007). However, some scholars argue that it starts earlier, and stress the early impressionable years of childhood (Easton and Dennis, 1969; van Deth et al., 2011); while others argue that it might be a lifelong learning process (Alwin and Krosnick, 1991). This article focusses on the early adulthood formative period of political socialization related to the first elections at which one is eligible to vote, as this is the moment when young adults not only recognize the political world but also choose to participate in it or not. Nevertheless, in light of the openness of the question of the exact age at which political socialization has stronger effects, sensitivity checks for several age groups (cohorts) will be carried out.

Hypotheses

In line with the arguments outlined above, and given the largely mixed results in previous works, first, I will assess the relationship between party system fragmentation in the first elections at which one is eligible and participation in subsequent elections with a two-tailed hypothesis of linear effect: (Hypothesis 1a) the higher the party system fragmentation in the first elections at which one is eligible to vote, the lower one’s propensity to vote in subsequent elections (linear negative effect) and (Hypothesis 1b) the higher the party system fragmentation in the first elections at which one is eligible to vote, the higher one’s propensity to vote in subsequent elections (linear positive effect). Second, I will test whether this relationship is positive with a curvilinear negative effect, claiming that (Hypothesis 2) party system fragmentation in the first elections at which one is eligible to vote will have a curvilinear effect on one’s propensity to vote in subsequent elections – starting with a positive effect up to a particular point, after which a further increase in the number of parties will have a negative effect (curvilinear negative effect).

Data and methods

At the individual level, I use data from the CSES IMD, covering 96 parliamentary elections in 31 countries between 1996 and 2016.³ Aggregate data are taken from Constituency-Level Elections Archive (CLEA; Kollman et al., 2019) and Comparative Manifesto Project (CMP; Volkens et al., 2019). The dependent variable considers participation in last parliamentary elections at the individual level and it is binary: coded 1 for reporting having voted, and 0 for abstention (those reporting ‘not knowing’ and not responding were excluded from the analysis).⁴

The key independent variable measures party system fragmentation in the first elections at which one is eligible to vote. I use the index of ENEP proposed by Laakso and Taagepera (1979), as it accounts not only for the number of parties, but also for the relative size of each.⁵ Figure 1 shows the historical evolution of ENEP at national level, in all countries considered. On top of the recent increase in fragmentation in many countries (see, e.g. Austria, the Netherlands, Switzerland, or Spain), it also depicts the high number of parties in the first rounds of elections in the younger CEE democracies (Poland or Slovenia), and the party stability in the United States and Canada. In sum, there is sufficient cross-national and national variation to perform the analyses.

Figure 1.

Historical evolution of ENEP in countries under scrutiny (1945–2018).

I use several measures of ENEP throughout this article. The most important is specified for each respondent of the CSES surveys and takes the mean value of ENEP at the age of 18–21 years old (ENEP at 18–21). It measures the ENEP of the first elections at which the respondent was eligible to vote or takes the mean value if there were several parliamentary elections in the first 4 years after the respondent reached the majority of age.⁶ To explore whether, in line with ‘voting as a habit’ theory, the first few elections at which one is eligible are indeed crucial, I also run several additional models specifying ENEP for different age cohorts as sensitivity checks: mean values of ENEP for 4-year clusters of respondents.

Data from CLEA include ENEP from elections that go back to the beginnings of the 20th century for many countries. This dataset gathers information on democratic elections, and thus, for the case of the CEE post-communist countries, the earliest data available is for 1990. For the established democracies respondents born as early as 1928 are included. However, for the newer democracies, only respondents born after 1971 are included, leaving a large part of the sample outside of the scope of this study (see Table A3 for the cut-off and its impact on the sample). To overcome this, I run robustness checks restricting the samples from the established democracies to younger cohorts only, and additional ones incorporating older respondents from post-authoritarian countries back to the sample.

To test for the possible curvilinear relationship (Hypothesis 2), ENEP at 18–21 squared is included in part of the analyses. Finally, in additional robustness checks, the impact of ENEP at 18 (ENEP in the year in which the respondent reached the majority of age) and ENPP at 18 (taken from CMP) are assessed. Effective number of parliamentary parties (ENPP) measures the number of parties in parliament, and not at electoral level.

The following individual-level variables perform as controls: age, age squared, gender, education, civil status, party identification, satisfaction with democracy, and external political efficacy.⁷ Several contextual controls are added for the contemporary elections: ENEP in elections (from CSES IMD), party polarization in elections⁸ and disproportionality in elections⁹ (both from CMP). ENEP in elections is included to check whether the long-term socializing effect of ENEP at 18–21 is significant over and beyond the short-term effect of ENEP in the particular elections.¹⁰ Finally, party polarization at 18–21 and disproportionality at 18–21 are introduced to control for the context of the formative elections; both were constructed using CMP data, taking the mean value of polarization and disproportionality when the respondent was 18–21 years old. Table A1 in the Appendix shows the summary statistics for all variables.

As the dependent variable is binary, I run logistic regression models. Given the hierarchical nature of data, on top of the country and year fixed effects models, multilevel random intercept models with respondents nested into elections and countries were also estimated. Identification is an important concern in studying the effects of socialization on electoral behaviour, as it is impossible to disentangle the effects of age, period and cohort (Dassonneville, 2017; Neundorf and Niemi, 2014). The multilevel logit models I used allowed me to model the effects of ENEP at 18–21 (here, C – cohort effects), controlling for the linear (and curvilinear) correlation of age (and age squared) with the propensity to turn out (A – age effects), and period effects (P) through year fixed effects, without running into multicollinearity. As ‘no model is able to solve the identification problem because the identification problem is inherent to the real-world processes being modelled’ (Bell and Jones, 2013: 163), I opted to use standard multilevel models rather than the hierarchical APC (age–period–cohort) models (see, e.g. Smets and Neundorf, 2014).

Countries in which voting is compulsory in contemporary elections have been excluded from the analyses (see, e.g. Taagepera et al., 2014). Additional dummy control for compulsory voting in formative elections (at 18–21) is included in all models (to account for the fact that the Netherlands practised it from 1917 to 1967).¹¹ I have also excluded combinations of ENEP at 18–21 and election with less than 30 observations. Furthermore, elections in countries that scored 3.0 or higher in the Freedom House index at the time also have been omitted due to their partially free or non-free status.¹²

Results

Table 1 shows the coefficients of fixed effects logistic regressions with reported electoral participation in pooled 96 elections in 31 countries as the dependent variable. In the first (null) model, ENEP at 18–21 is the only predictor and the coefficient is significant and negative. Model 2 adds controls for individual-level variables. As expected, the propensity to vote increases with age, education, marriage, party identification, satisfaction with democracy, and external political efficacy. Being a male also has a positive effect, whereas age squared reduces the likelihood of participating in elections, as the oldest face higher costs of getting to the polls. The effect of ENEP at 18–21 stays negative and significant.

Table 1.

Fixed effects of ENEP at 18–21 years old on turning out.

	1	2	3	4	5	6	7	8	9
ENEP at 18–21 years old	–0.21***	–0.14***	–0.07***	–0.06***	–0.07***	–0.07***	–0.06***	–0.06***	–0.19***
ENEP at 18–21 years old	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.03)
ENEP in elections				–0.16***	–0.16***	–0.16***	–0.15***	–0.15***	–0.27***
ENEP in elections				(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.04)
Polarization at 18–21 years old					0.02	0.02	0.02	0.02	0.02
Polarization at 18–21 years old					(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Polarization in elections						–0.04	–0.03	0.00	0.03
Polarization in elections						(0.04)	(0.04)	(0.04)	(0.04)
Disproportionality at 18–21 years old							–0.01**	–0.01**	–0.01**
Disproportionality at 18–21 years old							(0.00)	(0.00)	(0.00)
Disproportionality in elections								–0.05***	–0.05***
Disproportionality in elections								(0.01)	(0.01)
ENEP 18–21 # ENEP in elections									0.02***
ENEP 18–21 # ENEP in elections									(0.01)
Age		0.04***	0.04***	0.04***	0.04***	0.04***	0.04***	0.04***	0.05***
Age		(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Age # Age		–0.00***	–0.00***	–0.00***	–0.00***	–0.00***	–0.00***	–0.00***	–0.00***
Age # Age		(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Male		0.04*	0.02	0.02	0.02	0.02	0.02	0.02	0.02
Male		(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Education (ref. no/primary incomp.)
Primary/lower secondary		–0.06	0.31***	0.32***	0.32***	0.32***	0.31***	0.32***	0.32***
Primary/lower secondary		(0.08)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)
Higher secondary education		0.25**	0.69***	0.69***	0.69***	0.70***	0.69***	0.69***	0.69***
Higher secondary education		(0.08)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)
Post-secondary (non-university)		0.68***	0.85***	0.87***	0.87***	0.87***	0.87***	0.87***	0.87***
Post-secondary (non-university)		(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)
University education		0.85***	1.25***	1.26***	1.26***	1.26***	1.25***	1.25***	1.26***
University education		(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)	(0.09)
Civil status (married)		0.41***	0.34***	0.34***	0.34***	0.34***	0.34***	0.34***	0.34***
Civil status (married)		(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Party ID (yes)		0.92***	0.99***	0.99***	0.99***	0.99***	0.99***	0.99***	0.99***
Party ID (yes)		(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Satisfaction democracy (yes)		0.39***	0.35***	0.34***	0.34***	0.34***	0.34***	0.34***	0.34***
Satisfaction democracy (yes)		(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
External political efficacy (yes)		0.60***	0.61***	0.61***	0.61***	0.61***	0.61***	0.61***	0.61***
External political efficacy (yes)		(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)
Country FE	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	98,998	98,998	98,998	98,998	98,998	98,998	98,998	98,998	98,998
Elections	96	96	96	96	96	96	96	96	96
Countries	31	31	31	31	31	31	31	31	31
Log likelihood	–40,802.71	–36,083.09	–34,003.73	–33,972.81	–33,972.09	–33,971.71	–33,967.07	–33,958.26	–33,950.00
AIC	81,609.41	72,194.17	68,135.46	68,075.62	68,076.18	68,077.42	68,070.14	68,054.51	68,040.01

Source: Own elaboration based on CSES IMD, CLEA, and CMP. Logistic regression coefficients; standard errors in parentheses; countries with compulsory voting are excluded; all models control also for a dummy of compulsory voting at 18–21 years old (not shown).

ENEP: effective number of electoral parties; AIC: Akaike information criterion.

p < 0.05, **p < 0.01, ***p < 0.001.

In model 3, fixed effects of countries and years of survey (from 1996 to 2016) are added to account for the heterogeneity of country-level determinants of turnout as well as for the dynamics of turnout over time (particularly, its overall decline). The negative effect of ENEP at 18–21 is robust to include both fixed effects. The gender coefficient ceases to be significant. This model (3) will serve as the baseline for the analyses that follow. Figure 2 displays the marginal effects of ENEP at 18–21 on turnout. The likelihood of participating in elections drops from 86.8% for those who were socialized when ENEP at 18–21 was the lowest (2.0) to 78.7% for those who entered the electoral arena when the party system was highly fragmented (12.4).

Figure 2.

Marginal effects of ENEP at 18–21 years old on turning out.

Table A4 in the Appendix lists which countries have had high fragmentation (ENEP over 6.0) at some point in their electoral history (after 1945). While it is true that this list includes many countries from CEE (Poland, Slovakia, Latvia, Estonia, Slovenia, Croatia, Hungary, Romania and the Czech Republic), it also shows high fragmentation in the more established democracies of Western Europe (Switzerland, France, the Netherlands, and Finland) and Israel. All in all, the negative effect of fragmentation in the formative elections does not seem to be driven by outliers¹³ or any particular subgroup of countries. To rule out this possibility, I run separate models by regions (see Figure A1 in the Appendix). The effect is significant and negative for CEE and the established democracies of Western Europe (France, Germany, Great Britain, Austria, Ireland, the Netherlands, and Switzerland). It is also negative but not significant for Scandinavia, South Africa, and Israel. In other regions, it is positive but never significant. Separate models excluding each group of countries pinpoint that the strongest effect is to be found in CEE countries (see Figure A2 in the Appendix).

The remaining models (4–9) in Table 1 add the controls for several contextual measures. To rule out the possibility that it is party system fragmentation related to contemporary elections rather than a footprint of high ENEP during political socialization that drives the negative relationship between fragmentation and turnout, model 4 includes both measures of fragmentation. The correlation between ENEP at 18–21 and ENEP in elections is high (0.48), but even after controlling for the latter, the effect of ENEP at 18–21 stays negative and significant.¹⁴ Likewise, the effect of ENEP in elections is negative and significant. Model 5 includes polarization at 18–21. Its effect is non-significant and it does not alter the negative effects of ENEP at 18–21 nor ENEP in elections. Just as polarization in elections (contemporary ones) added in model 6. Models 7 and 8 add disproportionality at 18–21 and disproportionality in elections. Both factors hinder turnout but they do not alter the negative effects of ENEP at 18–21 nor ENEP in elections.

To disentangle the effects of fragmentation at the time the respondent was 18–21 years old and in contemporary elections, model 9 adds an interaction term between both factors. It may be argued that high fragmentation at first election and continued high fragmentation has a reinforcing effect; but where fragmentation has decreased there is less long-term impact of fragmentation (or vice versa). As the negative coefficients of both measures show, the short- and the long-term effects of fragmentation have a direct independent impact of lowering the likelihood of casting a ballot. The interactive term is positive and significant. Figure 3 plots its marginal effects, revealing that the socializing effect of ENEP at 18–21 lowers the propensity to vote especially in the context of low fragmentation in contemporary elections.

Figure 3.

Marginal effects of ENEP at 18–21 on turning out conditional on ENEP in election.

It has been argued that the relationship between ENEP and turnout is curvilinear. Nevertheless, this is not the case for fragmentation in formative elections (see Table A5 in the Appendix, which follows the pattern of Table 1, adding the squared term). The coefficients of both ENEP at 18–21 and ENEP at 18–21 squared are insignificant after controlling for country and year fixed effects (from model 3 on). Just as in the linear models of Table 1, ENEP in elections has a consistent negative effect (even when controlling for its square term, which has a significant and positive impact); both measures of polarization are insignificant, and disproportionality both in formative and contemporary elections has a negative effect.

The results of random effects models are consistent with the evidence gathered in those with fixed effects. Tables A6 and A7 in the Appendix reproduce models from Tables 1 and A4, this time estimating random intercept effects hierarchically (nesting respondents in elections and then in countries). Variance at both levels of aggregation is significant. Consistently with the FE models, only the linear effect of ENEP at 18–21 is significant and negative (Table A6). Including ENEP at 18–21 squared makes both measures of fragmentation insignificant (Table A7). Contrary to the FE models, including random intercepts render the effects of ENEP in elections insignificant. Polarization, once again, has no effect and disproportionality has a negative impact. In sum, Hypothesis 1a is supported by the empirical evidence, while Hypotheses 1b and 2 need to be rejected. I find a negative and linear effect of ENEP in the first elections at which one is eligible to vote on the propensity to participate in subsequent elections.

Robustness checks

To confirm that, in line with the voting habit theory, the crucial effect – the one that leaves a footprint of non-voting – occurs at the time of the first elections when one is eligible to vote, I run several age sensitivity checks following the pattern of model 3 from Table 1. The only modification was changing ENEP at 18–21 to measures of ENEP at different ages. Some of them related to voting as a habit and political socialization theories – ENEP at 14–17, 15–18, 16–19, 17–20, 19–22, 20–23, 21–24, 22–25, and 26–29 years old. Others – ENEP at 2–5, 6–9, 10–13, 30–33, 34–37, 38–41, 42–45, 46–49, 50–53, 54–57, 58–61, and 62–65 years old – are used as a contrast to the argument. Table 2 shows the marginal effects of ENEP at different ages and the coefficients from 22 separate models.¹⁵ The socializing negative effects of ENEP are significant between the ages of 14 and 29. The biggest difference (of 8.8 pp.) is to be found for ENEP at 20–23 years old, followed by ENEP at 19–22 and at the examined ENEP at 18–21. This is in line with the argument that it is the first few occasions that one has to vote that are important for the formation of a habit of voting and not only the first election at which one is eligible to vote.¹⁶ The magnitude of the effect gradually reduces for those of a younger age, and it becomes insignificant below the age of 13. Similarly, the effect vanishes with an increase in age, becoming insignificant over the age of 29. Thus, the overall results confirm that it is the high fragmentation at early adulthood socialization years that leaves the strongest negative imprint on the likelihood to turn out in future elections.¹⁷

Table 2.

Comparison of marginal effects of ENEP at different ages on turning out..

ENEP at age	Propensity to vote at		Difference (high–low)	Coefficient
ENEP at age	Lowest ENEP	Highest ENEP	Difference (high–low)	Coefficient
2–5	0.873	0.828	−0.045	−0.04
6–9	0.864	0.850	−0.014	−0.01
10–13	0.864	0.837	−0.027	−0.03
14–17	0.866	0.812	–0.054	–0.05***
15–18	0.867	0.804	–0.063	–0.05***
16–19	0.867	0.801	–0.066	–0.06***
17–20	0.868	0.790	–0.078	–0.06***
18–21	0.868	0.787	–0.081	–0.07***
19–22	0.868	0.785	–0.083	–0.07***
20–23	0.869	0.781	–0.088	–0.07***
21–24	0.866	0.793	–0.073	–0.06***
22–25	0.866	0.795	–0.071	–0.06***
26–29	0.864	0.835	–0.029	–0.04*
30–33	0.863	0.855	−0.008	−0.01
34–37	0.868	0.869	0.001	0.00
38–41	0.875	0.882	0.007	0.01
42–45	0.880	0.902	0.022	0.03
46–49	0.894	0.909	0.015	0.03
50–53	0.905	0.902	−0.003	−0.00
54–57	0.904	0.921	0.017	0.03
58–61	0.904	0.921	0.017	0.03
62–65	0.908	0.930	0.022	0.04

Source: Own elaboration based on CSES IMD and CLEA.

ENEP: effective number of electoral parties; all models follow the design of the FE model 3 from Table 1.

p < 0.05, ***p < 0.001.

Another possible shortcoming of this study – already mentioned in the data section – is that the samples from the more established democracies include respondents born as early as 1928, while those from the newer democracies of the post-communist countries are reduced to include only those born after 1971. To rule out the possibility of the unbalanced sampling driving the results, I run two robustness checks. The first restricts the sample to younger cohorts only in all countries. Table A8 in the Appendix shows that, even after limiting the sample to those born in or after 1945 and 1971, the effects of ENEP at 18–21 are negative and significant. Second, including older citizens from post-authoritarian countries in the analyses does not change the key results. Table A9 repeats the analysis of Table 1, incorporating those born before 1971 and socialized in the CEE non-democratic regimes, and in Portugal, Spain, South Korea, and South Africa – that is, nearly 1/4 of the sample (see Table A3 in the Appendix). Although past experiences do matter for present behaviour, in this article, a decision to focus only on ENEP in democratic elections was made. Thus, ENEP in the founding elections (Klingemann, 2019) is matched to both the youngest electors at that time (18–21) and to those older citizens who experienced elections in non-democratic regimes, assuming that they also were newcomers to competitive elections.¹⁸ The negative effect of ENEP at 18–21 is robust to this specification. This particular framework allows us to test whether fragmentation in the first election in which one is eligible affects the young to a greater degree and – as shown by the negative coefficient of the interaction term in model 10 – indeed, it does.¹⁹

Finally, the effects of ENEP at 18 and ENPP at 18 are assessed. The results of both fixed and random effects models in Tables A10 and A11 in the Appendix confirm the key finding of this article: the long-term negative effects on turnout of political socialization in highly fragmented contexts.

Conclusion

This article aimed at shedding light on the relationship between party system fragmentation and electoral turnout. Intuitively, it might be expected that the greater the number of parties the greater the choice electors are offered and, thus, the higher the propensity to turn out should be. However, there is more empirical evidence of a negative effect of the number of parties on turnout. High fractionalization of the political scene tends to confuse and alienate citizens from voting. A large number of parties is also responsible for complex coalition formation which lowers the decisiveness of the elections and hinders the political efficacy of the citizens.

Most studies focus on the aggregate short-term effects of the number of parties on turnout. In this article, building on the theories of voting as a habit and of political socialization, I assessed the long-term effects of fragmentation on the individual-level propensity to turn out in subsequent elections. The results show a robust negative and linear effect of fragmentation in the first elections at which one is eligible to vote on the propensity to participate in subsequent elections. These long-term effects of political socialization in a highly fragmented context are at play even if fragmentation has decreased over time. They are also robust to accounting for the quality of the choice offered, for the characteristics of the electoral system, and they are significant in both Western and Eastern parts of Europe, with the strongest effect found for the CEE post-communist countries (mainly due to the initial instability of party systems in that region).

Nonetheless, these findings should not be interpreted as a contradiction to the positive effect of proportional representation electoral systems on aggregate levels of turnout consistently shown in the literature (Blais and Carty, 1990; Blais and Dobrzynska, 1998). Neither should they be treated as an argument against consensual multiparty systems. Having more choice is not something that is bad for turnout. Proportionality is good in moderation (Gallagher and Mitchell, 2005: 13) and too much choice in terms of quantity of political offer might lead to confusion, choice overload, and in the end to choice deferral. Abstention is always an easy option, especially so for the youngest voters who have not yet developed the habit of voting. Facing too many options to choose from might lead to abstention, which affects decisions on whether to cast a ballot in the future.

Take the example of the 1991 Polish elections, in which 29 parties entered parliament (the Beer Lover’s Party among them), with the winning party securing only 12.3% of the vote and nine parties gaining just one seat. Such fragmentation brought about severe difficulties in the formation of the governing coalition (the smallest had to comprise at least six parties). During the 2 years that followed until the next elections, Poland had three prime ministers, of which one had failed to form a government. Such unstable and fragmented party systems undermine the accountability of representatives and hinder the ability of voters to control the elected; diluting meaningful information available to voters and lowering their trust in politicians and political parties. In such contexts, many might choose to exit the electoral arena (or not enter it at all).

This study adds party system fragmentation to the other characteristics of elections (e.g. competitiveness) that have been proven to leave a stable footprint on electoral participation patterns in light of the ‘voting as habit’ theory (Franklin, 2004; Smets and Neundorf, 2014). These findings also offer additional evidence, by focussing on the long-term effects, for the approach defending the negative impact of fragmentation on turnout. Nonetheless, we still lack proper evidence of the mechanisms behind this negative effect. Existing research uses two main plausible mechanisms I did not address directly in this article: that high fragmentation decreases turnout because it translates into higher cognitive cost of voting and because it leads to complex coalition bargaining processes, ending up with broad coalition governments that depress political efficacy. The evidence supporting these mechanisms is mixed and often they are used as an ‘alternative explanation’ due to lack of evidence for other plausible mechanisms (Blais and Dobrzynska, 1998: 250). Thus, further research might focus on providing empirical support for the causal mechanisms at play in the long-term.

Supplemental Material

sj-pdf-1-pol-10.1177_0263395720971210 – Supplemental material for Too much to choose from? The long-term effects of political fragmentation on electoral turnout

Supplemental material, sj-pdf-1-pol-10.1177_0263395720971210 for Too much to choose from? The long-term effects of political fragmentation on electoral turnout by Piotr Zagórski in Politics

Footnotes

Acknowledgements

Preliminary versions of this paper were presented at the 26th International Conference of Europeanists (CES) at the Carlos III University in Madrid in June 2019, the ECPR General Conference in Wrocław in September 2019 and at the Electoral Behaviour Seminar at the Autonomous University of Madrid in November 2019. The author would like to thank Laurie Beaudonnet, Kamil Marcinkiewicz, Miroslav Nemčok, Robert Liñeira, and Guillermo Cordero for their helpful comments on earlier versions of this article. He is also grateful to the anonymous reviewers and the editor for their insightful comments, and to José Ramón Montero and Andrés Santana for their constant advice.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Piotr Zagórski

Supplementary Information

Additional supplementary information may be found with the online version of this article.

Notes

Author biography

Piotr Zagórski is Adjunct Professor of Political Science and doctoral candidate at the Autonomous University of Madrid (UAM). He holds an MA in Sociology from the University of Granada and an MA in Democracy and Government from UAM. His research interests lie in the field of electoral behaviour, with a focus on the impact of historical legacies on electoral turnout and the reasons behind the vote for European populist parties. He has published, among others, in East European Politics and Societies and Government and Opposition.

References

Alwin

Krosnick

(1991) Aging, cohorts and the stability of sociopolitical orientations over the life span. American Journal of Sociology 97(1): 169–195.

Anduiza

(1999) Individuos o sistemas?: Las razones de la abstención en Europa Occidental. Madrid: Centro de Investigaciones Sociológicas, Siglo Veintiuno de España Editores.

Bell

Jones

(2013) The impossibility of separating age, period and cohort effects. Social Science & Medicine 93: 163–165.

Belli

Traugott

Young

McGonagle

(1999) Reducing vote overreporting in surveys: Social desirability, memory failure, and source monitoring. Public Opinion Quarterly 63(1): 90.

Blais

(2000) To Vote or Not to Vote: The Merits and Limits of Rational Choice Theory. Pittsburgh, PA: University of Pittsburgh Press.

Blais

(2006) What affects voter turnout. Annual Review of Political Science 9(1): 111–125.

Blais

Aarts

(2006) Electoral systems and turnout. Acta Politica 41(2): 180–196.

Blais

Carty

(1990) Does proportional representation foster voter turnout? European Journal of Political Research 18(2): 167–181.

Blais

Dobrzynska

(1998) Turnout in electoral democracies. European Journal of Political Research 33(2): 239–261.

10.

Brockington

(2004) The paradox of proportional representation: The effect of party systems and coalitions on individuals’ electoral participation. Political Studies 52(3): 469–490.

11.

Cancela

Geys

(2016) Explaining voter turnout: A meta-analysis of national and subnational elections. Electoral Studies 42(2): 264–275.

12.

Chernev

Böckenholt

Goodman

(2015) Choice overload: A conceptual review and meta-analysis. Journal of Consumer Psychology 25: 333–358.

13.

Converse

(1969) Of time and partisan stability. Comparative Political Studies 2(2): 139–171.

14.

Crepaz

(1990) The impact of party polarization and postmaterialism on voter turnout. European Journal of Political Research 18(2): 183–205.

15.

Dalton

(2008) The quantity and the quality of party systems: Party system polarization, its measurement, and its consequences. Comparative Political Studies 41(7): 899–920.

16.

Dassonneville

(2017) Age and voting. In: Arzheimer

Evans

Lewis-Beck

(eds) The Sage Handbook of Electoral Behaviour. London: SAGE Publishing, pp.137–158.

17.

Dinas

(2012) The formation of voting habits. Journal of Elections, Public Opinion and Parties 22(4): 431–456.

18.

Downs

(1957) An Economic Theory of Democracy. New York: Harper Collins.

19.

Easton

Dennis

(1969) Children in the Political System: Origins of Political Legitimacy. New York: McGraw-Hill Education.

20.

Franklin

(2004) Voter Turnout and the Dynamics of Electoral Competition in Established Democracies since 1945. Cambridge: Cambridge University Press.

21.

Gallagher

(1991) Proportionality, disproportionality and electoral systems. Electoral Studies 10(1): 33–51.

22.

Gallagher

Mitchell

(2005) The Politics of Electoral Systems. Oxford: Oxford University Press.

23.

Gallego

(2015) Unequal Political Participation Worldwide. New York: Cambridge University Press & CIS.

24.

Geys

(2006) Explaining voter turnout: A review of aggregate-level research. Electoral Studies 25(4): 637–663.

25.

Geys

Heyndels

(2006) Disentangling the effects of voter fragmentation on voter turnout. Economics and Politics 18(3): 367–387.

26.

Grofman

Selb

(2011) Turnout and the (effective) number of parties at the national and district levels: A puzzle-solving approach. Party Politics 17(1): 93–117.

27.

Hansen

(1994) Local elections and local government performance. Scandinavian Political Studies 17(1): 1–30.

28.

Hermet

Rose

Rouquié

(eds) (1978) Elections without Choice. London; Basingstoke: The MacMillan Press.

29.

Huber

Kernell

Leoni

(2005) Institutional context, cognitive resources and party attachments across democracies. Political Analysis 13(4): 365–386.

30.

Iyengar

Lepper

(2000) When choice is demotivating: Can one desire too much of a good thing? Journal of Personality and Social Psychology 79(6): 995–1006.

31.

Jackman

(1987) Political-institutions and voter turnout in the industrial democracies. American Political Science Review 81(2): 405–423.

32.

Jennings

Niemi

(1968) The transmission of political values from parent to child. American Political Science Review 62(1): 169–184.

33.

Jennings

Niemi

(1981) Generations and Politics: A Panel Study of Young Adults and Their Parents. Princeton, NJ: Princeton University Press.

34.

Kahneman

(2003) Maps of bounded rationality: Psychology of behavioral economics. The American Economic Review 93(5): 1449–1475.

35.

Karp

Banducci

(2008) Political efficacy and participation in twenty-seven democracies: How electoral systems shape political behaviour. British Journal of Political Science 37(2): 311–334.

36.

Kittilson

Anderson

(2011) Electoral supply and voter turnout. In: Dalton

Anderson

(eds) Citizens, Context and Choice: How Context Shapes Citizens’ Electoral Choices. Oxford: Oxford University Press, pp.33–54.

37.

Klingemann

(2019) Founding elections. In: Merkel

Kollmorgen

Wagener

(eds) Handbook of Political, Social, and Economic Transformation. Oxford: Oxford University Press, pp.502–508.

38.

Kollman

Hicken

Caramani

Backer

Lublin

(2019) Constituency-Level Elections Archive (CLEA) (Data file and codebook). Ann Arbor, MI: Center for Political Studies, University of Michigan (Producer and Distributor). Available at: http://www.electiondataarchive.org

39.

Kostadinova

Power

(2007) Does democratization depress participation? Political Research Quarterly 60(3): 363–377.

40.

Laakso

Taagepera

(1979) Effective number of parties: A measure with application to west Europe. Comparative Political Studies 12(1): 3–27.

41.

Mishler

Rose

(2007) Generation, age, and time: The dynamics of political learning during Russia’s transformation. American Journal of Political Science 51(4): 822–834.

42.

Neundorf

Niemi

(2014) Beyond political socialization: New approaches to age, period, cohort analysis. Electoral Studies 33(1): 1–6.

43.

Neundorf

Smets

(2017) Political socialization and the making of citizens. In: Oxford Handbooks Online. Available at: https://www.oxfordhandbooks.com/view/10.1093/oxfordhb/9780199935307.001.0001/oxfordhb-9780199935307-e-98

44.

Norris

(2002) Democratic Phoenix: Reinventing Political Activism. New York: Cambridge University Press.

45.

Plutzer

(2002) Becoming a habitual voter: Inertia, resources, and growth in young adulthood. American Political Science Review 96(1): 41–56.

46.

Pop-Eleches

Tucker

(2017) Communism’s Shadow. Historical Legacies and Contemporary Political Attitudes. Princeton, NJ: Princeton University Press.

47.

Schwartz

(2004) The Paradox of Choice: Why More Is Less. New York: Ecco.

48.

Seidle

Miller

(1976) Turnout, rational abstention and campaign effort. Public Choice 27(1): 121–126.

49.

Simpser

Slater

Wittenberg

(2018) Dead but not gone: Contemporary legacies of communism, imperialism, and authoritarianism. Annual Review of Political Science 21(1): 419–439.

50.

Smets

Neundorf

(2014) The hierarchies of age-period-cohort research: Political context and the development of generational turnout patterns. Electoral Studies 33(1): 41–51.

51.

Smets

van Ham

(2013) The embarrassment of riches? A meta-analysis of individual-level research on voter turnout. Electoral Studies 32(2): 344–359.

52.

Stockemer

(2017) What affects voter turnout? A review article/meta-analysis of aggregate research. Government and Opposition 52(4): 698–722.

53.

Taagepera

Selb

Grofman

(2014) How turnout depends on the number of parties: A logical model. Journal of Elections, Public Opinion and Parties 24(4): 393–413.

54.

The Cal Systems (2018) CSES integrated module dataset (IMD) (Dataset and documentation). Available at: www.cses.org (accessed 4 December 2018).

55.

van Deth

Abendschön

Vollmar

(2011) Early political socialization. Political Psychology 32(1): 147–174.

56.

Vassil

Solvak

Ehin

(2016) More choice, higher turnout? The impact of consideration set size and homogeneity on political participation. Journal of Elections, Public Opinion and Parties 26(1): 78–95.

57.

Volkens

Krause

Lehmann

Matthieß

Merz

Regel

, et al. (2019) The Manifesto Data Collection (Manifesto project (MRG/CMP/MARPOR)), version 2019B. Berlin: Wissenschaftszentrum Berlin Für Sozialforschung (WZB). Available at: https://manifesto-project.wzb.eu/datasets/MPDS2019b

58.

Wass

Blais

(2017) Turnout. In: Arzheimer

Evans

Lewis-Beck

(eds) The Sage Handbook of Electoral Behaviour. London: SAGE Publishing, pp.459–488.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.80 MB