Introduction
Cross-national research on advanced democracies has produced excellent work on new party formation and entry, their electoral, and parliamentary performance as well as government participation (see, for instance, Hug, Reference Ibenskas2001; Heinisch, Reference Hooghe, Marks, Schakel, Chapman Osterkatz, Niedzwiecki and Shair-Rosenfield2003; Tavits, Reference Tavits2006; Bolin, Reference Bormann and Golder2007; Meguid, Reference Mudde2007; Deschouwer, Reference de Lange and Art2008; McDonnell and Newell, Reference Monroe2011; Spoon, Reference Spoon2011; Akkerman and de Lange, Reference Akkerman and de Lange2012; Biezen and Rashkova, Reference Biezen van and Rashkova2014). However, more fundamental questions about the survival of new parties are surprisingly rarely dealt with (but see on old democracies Collignon, Reference Collignon2018; Lowery et al., Reference Lowery, van Witteloostuijn, Péli, Brasher, Otjes and Gherghina2013; Rose and Mackie, Reference Rose and Mackie1988; on new democracies Bakke and Sitter, Reference Bakke and Sitter2015; Deegan-Krause and Haughton, Reference Deegan-Krause and Haughton2015; Cyr, Reference Cyr2016; Casal Bértoa and Spirova, Reference Casal Bértoa and Spirova2017). Drivers of party death and survival shed light on the fundamental question of how we should theorize what political parties are. Indeed, core party actors are likely to dissolve their party only when they perceive the raison d’être of their organization’s existence to be fundamentally undermined. What defines this raison d’être – the core motivations that underpin a party’s very existence – depends on the conception of parties we start out with (Mudge and Chen, Reference Mudge and Chen2014: 310–312).
This is why in this paper we develop two sets of hypotheses on party death building on two contrasting conceptions of political parties co-existing in the literature for many decades, one originating in Downs’ (Reference Downs1957) economic theory of democracy, the other in Lipset and Rokkan’s (Reference Lipset and Rokkan1967) seminal work on party systems. Are parties best understood as a ‘team of men seeking to control the governing apparatus by gaining office’ (Downs, Reference Downs1957: 25), as careerist vehicles composed of ambitious politicians driven by the pursuit of individual rewards? Or do they bear stronger resemblance to ‘alliances in conflicts over policies and value commitments’ (Lipset and Rokkan, Reference Lipset and Rokkan1967: 5) sustained by followers committed to build support for a collectively shared ideology or group interest?
Comparative party research is strongly shaped by the Downsian conception. As Bawn et al., put it, ‘[c]ontemporary scholarship views a party as a team of politicians whose paramount goal is to win electoral office’ (Reference Bawn, Cohen, Karol, Masket, Noel and Zaller2012: 571; see also Aldrich, Reference Aldrich1995: 4; Müller and Strøm, Reference Müller and Strøm1999). This rationalist, office-driven conception of parties formed and sustained by self-interested individuals underpins – implicitly or explicitly – many large-N studies on new party performance (e.g. Harmel and Robertson, Reference Harmel and Robertson1985; Hug, Reference Hug2001; Tavits, Reference Tavits2006). This also includes recent studies that stress the need to overcome the treatment of parties as unitary actors (often associated with this rationalist conception) and show how intra-organizational factors shape parties’ strategic choices and their long-term evolution, for instance, in terms of the programmatic profile they adopt (Spoon, Reference Spoon2011; Schumacher et al., Reference Schumacher, de Vries and Vis2013; Tavits, Reference Tavits2013). As detailed below, this Downsian conception of parties provides a rationalist, office-driven rationale allowing us to derive hypotheses on party death.
This account contrasts with works following sociological traditions, which approach political parties as societal organizations driven by group goals rather than individual interest-maximization (Lipset and Rokkan, Reference Lipset and Rokkan1967: 5; Monroe, Reference Monroe2001: 20–21; Bawn et al., Reference Bawn, Cohen, Karol, Masket, Noel and Zaller2012: 571; Mudge and Chen, Reference Mudge and Chen2014: 310; see also Panebianco, Reference Panebianco1988; Kitschelt, Reference Kitschelt1989). Works taking this perspective are predominantly qualitative and small-N. In contrast, the interest group literature has applied this conception more broadly to account for mortality or organizations’ anxiety thereof, distinguishing central resource dimensions crucial to an organization’s viability such as a distinctive area of competence (comparable to a party’s ideological niche) (Gray and Lowery, Reference Gray and Lowery1997: 28). We make this sociological conception useful by developing hypotheses on party death centring around a party’s (in)ability to fulfil its representation function in order to test them alongside hypotheses following a Downsian rationale.
To derive hypotheses from both rationales is important as the assumptions about the core motivations that underpin a political party’s very existence, as associated with the Downsian and the sociological conceptions, are usually not tested in themselves. Unlike the study of parties constitutive for fully institutionalized party systems that have lasted many decades and rarely die (hence do not allow to examine party death), the study of the evolution of 144 new parties – irrespective of their origins, durability or ideological profile – over the course of more than four decades in 17 established party systems opens a window of opportunity: it allows us to examine the drivers of party death in an encompassing fashion. It allows us to go beyond the study of parties’ relative success – the dominant focus of existing research – to explore what fundamentally sustains them, by covering these parties’ whole life cycle from their birth to their (potential) ‘death’ (Pedersen, Reference Pedersen1982).
This is important not only theoretically but also empirically. Only new parties that contest more than a few elections can broaden the offer of the party system and may have a direct or indirect impact on policy-making, by entering government or by triggering shifts in the offer of mainstream parties (e.g. Meguid, Reference Meguid2007; Mudde, Reference Mudde2007). Only once knowing which types of new parties tend to die and survive and why, can we truly evaluate the implications of the rise of new parties for representative democracy, in which support for mainstream parties has been suffering over the last decades (e.g. Dalton and Wattenberg, Reference Dalton and Wattenberg2002). More particularly, whether Downsian or sociological drivers of party death finds more support has important empirical implications for which type of change new parties are likely to introduce in established party systems and the way they contribute to representative democracy undergoing significant societal change. Essentially, it reveals whether new parties predominantly exploit on-going processes of dealignment and the decline of mainstream parties (if Downsian arguments find more support) or whether new parties (echoing the sociological rationale) contribute to the formation of new, lasting linkages to societal groups and thereby enhance the representational capacity of established party systems.
In the next section, we develop two sets of hypotheses derived from two distinct conceptions of political party co-existing in the literature. After that, we present the indicators, the data set and methods used to test these hypotheses. Then, we examine – using survival analysis – the patterns of death and survival of 144 organizationally new parties formed in or after 1968 in 17 democracies. We conclude with a discussion of our findings and of avenues of future research.
Theory: two conceptions of political party and drivers of party death
Defining party death
Participating in elections by nominating candidates for public office is a key characteristic of political parties, which sets them apart from organizations such as interest groups (Sartori, Reference Sartori1976). A party survives as long as it takes part in elections, irrespective of the governmental tier – national, regional, or local (Rose and Mackie, Reference Rose and Mackie1988: 539; Spoon, Reference Spoon2011: 16–17; Cyr, Reference Cyr2016: 129; Collignon, Reference Collignon2018). Consequently, we define a party as dead when it permanently ceases to nominate candidates for any electoral contest as a separate, autonomous organization.Footnote 1
Downsian and sociological rationales of party death
To develop a series of systematic hypotheses on drivers of party death, we start from the core motivations of those actors instrumental to forming and sustaining a party (Krouwel and Lucardie, Reference Krouwel and Lucardie2008; Mudge and Chen, Reference Mudge and Chen2014: 310–312). These motivations – or ends – constitute the raison d’etre of a party and allow us to identify conditions most central to a party’s continued survival and conditions whose absence are, in turn, likely to lead to a party’s death. To identify those conditions we use the Downsian conception of parties as careerist vehicles and the sociological conception of parties as ‘alliances in conflicts over policies and value commitments’ directed towards fulfilling important mobilization and representative functions (Lipset and Rokkan, Reference Lipset and Rokkan1967: 5; Enyedi, Reference Enyedi2005: 699) as analytical devices.Footnote 2 Instead of representing rivalling accounts, the two conceptions (each of which is inevitably a simplification of ‘reality’) serve as heuristic tools to generate systematic expectations regarding which particular factors are likely to be immediately important – in light of the respective conception – to account for whether core actors within a party maintain their organization through assuring ongoing electoral participation or not. This perspective starting out from parties’ core motivations stresses the importance of party agency when theorizing the decision to continue investing the resources necessary to build and maintain an active party organization or to cease to do so (Enyedi, Reference Enyedi2005; de Lange and Art, Reference de Lange and Art2011; Bolleyer, Reference Bolleyer2013).Footnote 3
More specifically, each conception allows us to distinguish factors or conditions that are merely desirable for core party actors (and might be important for explaining party strategy, for instance) from those that are likely to motivate core party actors to dissolve their party as they see the latter’s raison d’être as fundamentally undermined. This distinction is important as, for instance, parliamentary seats are desirable and helpful to any party to realize its goals, including those parties predominantly aspiring to represent group interests in line with the sociological conception. However, seats are only one means and parliament only one arena among several towards the end of engaging in representative activities (Saward, Reference Saward2009).Footnote 4 Thus, starting out from a sociological conception, we would expect access to seats to be less decisive for party actors’ determination to sustain their organization than, for example, a party’s societal roots or its ideological identity. If, however, we consider parties as career instruments to self-interested politicians in line with the Downsian conception of parties instead, we would expect the inability to win seats (or access government) to motivate wide-spread defection likely to trigger party death (as the aspiration of winning office is considered constitutive for creating a party in the first place).
In the following we theorize the two sets of factors respectively. Each set is derived from one conception of political party, as distinguished by the dominant (not necessarily only) end that actors running a party are assumed to strive for predominantly.Footnote 5 The factors forming part of each account of party death can be divided in two groups, respectively, one time-invariant, the other time-variant. First, each conception (characterized by a specific motivational underpinning) is closely associated with the way an organization is formed, as the latter captures who is involved in a formation for what reasons (e.g. Duverger, Reference Duverger1959; Lucardie, Reference Lucardie2000; Krouwel and Lucardie, Reference Krouwel and Lucardie2008 Lucardie and Ghillebaert, Reference Lucardie and Ghillebaert2008). Formative characteristics defining what a party is for should be immediately relevant for any evaluation of if and when a party has failed to meet its constitutive purpose. Second, by specifying one central end, each conception allows us to theorize those factors that are direct realizations of, or constitutive for, such ends. Table 1 summarizes the two accounts of party death. All factors associated with the Downsian conception and the sociological account, and how they are expected to shape party death, are detailed below.
Table 1 Drivers of party death
The Downsian rationale of party death: hypotheses
Theorizing formative conditions closely aligned with the Downsian notion of parties, Duverger (Reference Duverger1959: 290–291) has famously argued that parties formed by parliamentarians distinguish themselves in terms of their organization, finances, and evolution from parties formed without such ‘insider’ support. Developing this argument further, Krouwel and Lucardie (Reference Krouwel and Lucardie2008: 283) argued that this is the case as in parties formed by parliamentarians, their ‘strategic’ or ‘personal ambitions’ – both short-term orientations – are likely to be dominant. This, in turn, makes long-term investments in a lasting party infrastructure which is then able to outlive the founding elites less likely (should the latter decide to leave). Indeed, parliamentarians forming their own party often do so after defecting from another party because they could not exercise the level of influence they envisioned (Ceron, Reference Ceron2015). While parliamentarians might have ties to societal groups or not, the core of the argument is that the motivational structure underpinning insider formations created with the central involvement of professional politicians is defined by career aspirations central to our Downsian conception of a political party, making long-term structural investments less likely. The lack of such investments increases the risk of party death.
Hypothesis 1.1 (Insider Formation Hypothesis): Parties formed by parliamentarians are more likely to die than those that are not.
Individual rewards in the form of office are the central end for party actors driven by the maximization of such rewards. Consequently, while electoral support signals to party elites that institutional access might be within reach, whether the votes bring tangible rewards from party elites’ point of view depends on their translation into seats that can be allocated to central figures in the party (usually the first national seat a new formation wins is taken over by the party leader). Thus, what matters to elites is the party’s institutional access – both in terms of seats and government – rather than its mere electoral performance (Mayhew, Reference Mayhew1974; Obert and Müller, Reference Obert and Müller2017). While winning many seats too early on can be destabilizing for new parties (Bolleyer, Reference Bolleyer2013), the Downsian rationale suggests that MPs can be expected to make stronger efforts to keep their party going and assure its success, efforts that are directed towards assuring their own reelection, than elites who remained outside public office. Consequently, the more party actors benefit from parliamentary seats (incentivized to sustain their organization), the more the party is likely to profit, thereby reducing the risk of its death.
The same rationale applies to government access as the ultimate locus of power and prestige. For ambitious politicians, seats in parliament might be little more than a stepping stone on the way to the ultimate reward, the take-over of ministerial posts. The literature is divided regarding the merits and perils of government participation for new political parties as organizations: for some parties the access to government had destabilizing or disintegrating effects, on others, strengthening effects (e.g. Deschouwer, Reference Deschouwer2008; Bale and Dunphy, Reference Bale and Dunphy2011; de Lange and Art, Reference de Lange and Art2011; Bolleyer et al., Reference Bolleyer, van Spanje and Wilson2012). Yet starting out from a Downsian conception, what matters is how members of the party elite respond to government access because such access constitutes a major reward (Müller and Strøm, Reference Müller and Strøm1999). Once occupying ministries in national governments or powerful regional governments, office-holders can be assumed to be highly motivated to sustain their party’s core activities to assure re-entry into government, efforts which should make party death less likely.Footnote 6
This leads us to two further hypotheses on the risk of party death derived from the Downsian perspective:
Hypothesis 1.2 (Seat Share Hypothesis): Parties with a higher seat share are less likely to die than parties with lower seat shares.
Hypothesis 1.3 (Government Access Hypothesis): Parties with government access are less likely to die than parties without.
The sociological rationale of party death: hypotheses
If parties are formed as societal organizations to represent and give voice to issues or groups within society as well as the institutional arena, two characteristics in a party’s origin can be expected to be relevant for the risk of party death. First, it is important whether the new formation is supported by a promoter organization already established in civil society in pursuit of a collective goal which the newly formed party carries into the electoral arena. The second formative feature of relevance is whether a party has an ideologically novel profile. Both these formative characteristics are likely to decrease the risk of death as the motivations of their founders are more likely to be ‘ideological’ or ‘altruistic societal’ (Krouwel and Lucardie, Reference Krouwel and Lucardie2008: 283; Spoon, Reference Spoon2011: 27–28), motivations that suggest a commitment to long-term policy goals as opposed to an orientation towards short-term rewards.
Regardless of whether their cause is fascist, religious, or environmental, promoter organizations provide access to a pool of committed activists able to fill positions in the new party (e.g. Rose and Mackie, Reference Rose and Mackie1988; Poguntke, Reference Poguntke2002; Art, Reference Art2011). They decrease a party’s dependence on a particular leader or core elite, while party elites affiliated to promoter organizations (which tend to represent broader collective interests that require political representation in the longer run) are less likely to build an organization for the sole purpose of advancing their careers. Both aspects make it more likely for the organization to outlive the same elites. Simultaneously, followers are less likely to defect if their individual interests clash with organizational demands, as affiliations to already established promoter groups function as a first ‘natural pre-selection mechanism’ for an organization that initially has very little capacity to identify and weed out opportunists (Art, Reference Art2011).
A novel ideological profile is another beneficial ‘formative feature’ we expect to decrease the risk of death. Ideological novelty (being a party introducing a new set of issues or a new ideology formerly unrepresented by other parties) helps new parties not only to mobilize initial support but also to cultivate lasting (non-instrumental) loyalties among their followers (e.g. Lucardie, Reference Lucardie2000; Abedi, Reference Abedi2004; Meguid, Reference Meguid2007; Spoon, Reference Spoon2011; Lowery et al., Reference Lowery, van Witteloostuijn, Péli, Brasher, Otjes and Gherghina2013). Adams et al. (Reference Adams, Clark, Ezrow and Glasgow2007: 514–515, 525) have more specifically shown that niche parties – usually characterized by novelty – respond less to shifts in public opinion. This should help them to maintain the ownership of core issues by making it more difficult for competitors to highjack their issues, which, in turn, makes it more likely for novel new parties to occupy a separate niche in their party system in the longer term (Meguid, Reference Meguid2007; Spoon, Reference Spoon2011).
This leads us to two hypotheses on formative party characteristics closely aligned with the sociological conception of parties expected to shape the likelihood of party death in the long run:
Hypothesis 2.1 (Societal Roots Hypothesis): Parties formed with the support of pre-existing societal organizations are less likely to die than parties without.
Hypothesis 2.2 (Novel Formation Hypothesis): Parties formed with a novel ideological profile are less likely to die than parties without.
Irrespective of whether new parties are novel formations, whether a party (old or new) can engage in representative activity meaningful to its members and supporters in the long term is likely to depend on whether it is confronted with the emergence of a credible competitor from the same party family (Ladrech, Reference Ladrech2012: 17–18). If a party’s main purpose is indeed to represent societal constituencies that need representation according to the sociological conception, the fact that the demands of this constituency are represented by another party should increase the risk of party death (rather than incentivizing the strategic adaptation of the party’s representative profile suggested by the Downsian rationale).Footnote 7 While novel formations, by definition, do not have a credible competitor at the time of their formation, such a competitor (e.g. a second anti-immigrant party) might emerge within a few election cycles and thereby weaken its ‘novelty advantage’. Vice versa, a new liberal party does not possess a novelty advantage as it represents an old ideology, but it still can profit from a ‘distinctiveness advantage’ in periods in which no viable liberal party exists in its party system.
Hypothesis 2.3 (Distinctiveness Advantage Hypothesis): Parties that do not face a competitor party belonging to the same party family, hence, are ideologically distinct, are less likely to die than parties that are not.
Definition of new parties and empirical scope of the analysis
We define new parties by referring to their organizational age. Organizationally new parties still need to build a viable, self-sufficient infrastructure consolidated by a (relatively) stable support base, which makes these parties more vulnerable than, and thus distinct from, the group of established or ‘organizationally mature’ parties (e.g. Mair, Reference Mair1990). Parties are newly created if they are built from scratch (‘newly born’), that is, formed without the help of members of existing parties (Hug, Reference Hug2001: 13), through mergers in which organizationally new parties participate and splits from old parties,Footnote 8 that is, parties that faced or still face the challenge to build a viable infrastructure. Mergers between old parties and successor parties – being able to rely on more extensive infrastructures and resources – are excluded.Footnote 9
Following Bolleyer and Bytzek (Reference Bolleyer and Bytzek2013), we identified new formations meeting these criteria from 1968 onwards as party systems in long-lived democracies were considered ‘frozen’, hence ‘stable’, up to the 1960s. However, citizens’ party affiliations underpinning these party systems started to de-align in the latter period of that decade (Lipset and Rokkan, Reference Lipset and Rokkan1967). Consequently, all parties covered faced a growing challenge to stabilize support in an era of mass media communications as well as high levels of citizen disengagement from traditional forms of political participation.
Meanwhile, to assure unit homogeneity, we deliberately focus on organizationally new parties that – from their organizational birth until their (potential) death – operated in already fully consolidated party systems (a situation that is substantially different if the majority of rivalling parties are organizationally new as well) (Meguid, Reference Meguid2007). Simultaneously, the countries included are all established democracies with long electoral histories that allowed us to assess full party life cycles some of which lasted several decades before their eventual death. Following this rationale, we cover parties in 17 countries in Western Europe, North America, and Australasia,Footnote 10 all of which were fully consolidated by 1968, the earliest possible ‘year of birth’ in our party sample.
The data set includes all new parties that won a seat in national parliament in these 17 countries between 1968 and 2011 at least once in their lifetime irrespective of vote share and ideological profile, whose vote shares varied between 0 and 37.7%. As parliamentary thresholds differ across countries, we complemented these by all new parties that did not win a seat but a minimum of 2% of the national vote at least once in their ‘life’, increasing inclusiveness in high threshold countries such as Germany or France. This strategy provided us with a highly inclusive sample of 144 parties essential to test factors influencing the risk of party death.Footnote 11
Operationalization of variables
Measurement of dependent variable
The dependent variable measures the length of the party’s life in years from its organizational foundation until its year of death or, if it is still active, up to 2011Footnote 12 drawing on a wide range of primary sources and secondary sources. In essence, death occurs when a party permanently stops nominating candidates for any elections (irrespective of tier) as a separate, autonomous organization.Footnote 13 In operational terms, such ‘death’ can occur through different processes. Predominantly it is linked to the formal dissolution of the party (through a membership meeting or a formal declaration of the leadership). Alternatively, parties can cease to run elections by withdrawing from the electoral arena to exclusively focus on societal activities or merging into other organizations.Footnote 14 Using these criteria, we find that 64 parties or 44 percent of the parties in our sample died during the period under consideration.Footnote 15 Figure 1 shows the timing of death for these parties.
Figure 1 The distribution of party survival times.
The distribution is quite diverse: some parties did not survive even a single year, one died after more than 39 years of continuous existence. The average lifespan is 13.0 years while the std. dev. is 8.4 years. In comparison, looking at the whole sample, we find that the average number of years that a party stays in the sample is 13.4, with the std. dev. of 9.9 years. The highest observed age is 42 years.
Measurement of independent variables
Starting with the variables associated with our Downsian account of party death, a new measure was constructed to capture the variable insider formation (Insider Formation Hypothesis 1.1). We coded each of our parties as 1 if a national parliamentarian (present or former) was involved in its formation and took on a formal role (often but not always its leadership) in the new party. To test the Seat Share Hypothesis (Hypothesis 1.2), we compiled data on the evolution of the number of national seats for the first house of parliament starting with the first national election a party nominated candidates for (earliest 1968) till the last one it participated in or, in case of still active parties, the last national election before 2011. We use the lagged natural logarithm of party’s seat share reflecting the expectation that changes in seat share at lower values have a stronger impact on survival than those at higher values. Participation in national or regional government (Government Access Hypothesis 1.3) is a dummy variable (values 0 and 1) and was coded for each year. Details on data sources for these and other explanatory variables are listed in the Online Appendix, Section 1.
Moving to the variables linked to our sociological account of political party, we distinguished between new formations with and without societal roots (Societal Roots Hypothesis 2.1) relying on the classification of 140 new parties provided by Bolleyer (Reference Bolleyer2013: 43–43, Table 2.2) measuring whether a party’s foundation was supported by one or several identifiable promoter organizations or groups or not. Whether a party is an ideologically novel formation (Novelty Hypothesis 2.2) was measured in two steps. First, we identified which parties in our sample either belonged to the new Green or new right family, the only two genuinely new party families that – according to Mudde (Reference Mudde2007) – established themselves across a wide range of established democracies. To capture whether these new right or Green parties brought something novel and distinct to their party systems (and thus were likely to take ownership of these issues in the longer term), we only coded those of them as ideologically novel formations (1) if they were the first party of that family – in terms of their year of formation – that entered their respective party system. All other parties were coded 0. This was suitable to measure ideological novelty since in numerous countries more than one new Green or new right party have emerged over the last four decades. Appendix 1 discusses the steps in the operationalization process in detail. To operationalize our Distinctiveness Advantage Hypothesis (Hypothesis 2.3), we examine whether any of the parties with at least 1% of the vote represented the same ideological family as any of the new parties in our sample in each previous national parliamentary election throughout the party’s life cycle. If one or more such parties exist, we code the value of the variable indicating the presence of a competitor as 1. The absence of such a competitor is coded as 0. The variable varies in time following the emergence and/or decline of ideologically similar new competitors.
Control variables
To assure the robustness of our findings we control for the following institutional variables earlier studies identified as relevant for new party performance. The first is the presence or absence of a powerful regional tier in the political system, giving parties another arena to operate, gain visibility, and access resources on (e.g. Deschouwer, Reference Deschouwer2003; Spoon, Reference Spoon2011; Cyr Reference Cyr2016; Obert and Müller, Reference Obert and Müller2017). To measure the strength of a political system’s regional tier we use the Regional Authority Index (RAI) provided by Hooghe et al. (Reference Hooghe, Marks, Schakel, Chapman Osterkatz, Niedzwiecki and Shair-Rosenfield2015). The RAI captures the authority of regional governments in ten different areas on an annual basis for the period between 1950 and 2010. We include the values of this index lagged by 1 year to account for the time that it takes for any changes in regional authority to influence party survival and death. Second, we control for a party’s access to direct state funding important to maintain basic party functions (e.g. Bakke and Sitter, Reference Bakke and Sitter2015; Casal Bértoa and Spirova, Reference Casal Bértoa and Spirova2017). We construct a time-variant variable measuring each party’s access to funding in each electoral cycle that it participated in during its lifespan. The variable takes the value of 1 if the party won sufficient votes or seats to obtain state funding (organizational or electoral) made available by the party finance regime in place in the respective electoral cycle and a score of 0 otherwise. Finally, electoral system disproportionality may also affect parties’ survival influencing the translation of votes into seats. We use the average district magnitude of the lower tier of the electoral system to measure electoral thresholds. Bormann and Golder (Reference Bormann and Golder2013) is the source of this data. While alternative measures of national electoral thresholds have been proposed (Gallagher and Mitchell, Reference Gallagher and Mitchell2005), there is little research on their properties and suitability as measures of electoral systems. We therefore follow recent studies on new parties (e.g. Tavits, Reference Tavits2006; Biezen and Rashkova, Reference Biezen van and Rashkova2014) and more general studies on the effects of electoral systems (e.g. Carey and Hix, Reference Carey and Hix2011) by using the average district magnitude for capturing potential effects of electoral systems.
The Online Appendix (Section 2) presents the descriptive statistics of all variables used in the analysis and the correlation between three time-invariant variables (societal rootedness, initial ideological novelty, and insider formation).
Empirical analysis
Model choice
We use the proportional hazard (PH) model developed by Cox (Reference Cox1992) to estimate the impact of the variables discussed above on the probability of the death of parties. We choose the Cox model for the analysis as it does not assume a specific probability distribution for the time until an event occurs (Box-Steffensmeier and Zorn, Reference Box-Steffensmeier and Zorn2001). To control for any country specific effects, we use robust standard errors clustered by country.
A key assumption of the Cox model is that of ‘PHs’, which implies that the ratio of hazards is constant over time. This is not to say that the risk of an event is constant, but rather that hazards are proportional over time. We tested this assumption using the Schoenfeld residuals PH test, which is a common test of the assumption of non-proportionality. Following a standard practice when using the Cox model, statistical models include the interaction between a covariate and the natural logarithm of time if the statistical test (we use the 0.05 level of significance) indicates that the effect of this covariate is not time-constant.
We explain the choice of the model and its interpretation in greater detail in the Online Appendix, Section 3.
Findings
Table 2 summarizes the results of the Cox regression models of party death. The first two models test the variables associated with each of our two theoretical conceptions separately. Model 3 includes variables for both perspectives. Control variables are included in all three models. Table 3 presents the substantive effects of the covariate variables. For the covariates whose significant effects are not time-constant, Table 3 shows the first differences (i.e. the percentage change in the hazard rate at the specific point in a party’s life related to the change in the values of the predictor variable) when party age is 1 std. dev. below the mean (3.5 years), at the mean value (13.5 years) and 1 std. dev. above the mean value (23.5 years) (see Online Appendix, Section 7, for the plots of the first differences across the whole range of party age).
Table 2 Cox proportional hazard (PH) models of party death
Regression coefficients of the Cox PH model. Robust standard errors clustered by country in parentheses.
** P< 0.05
* P<0.1.
Table 3 Effects of predictor variables on party death
First differences in the hazard rate of party death for different values of party age, simulated using simPH (Gandrud, Reference Gandrud2015) package in the statistical environment R based on the estimates of Model 3 in Table 2. 95% confidence intervals in parentheses. The values of the continuous predictor variables are changed from 1 std. dev. below the mean to 1 std. dev. above the mean.
Our findings indicate the relevance of factors derived from both conceptions of political parties. Starting with the effects of factors forming part of the Downsian conception of party death, there is support for the notion that party formations involving parliamentarians are more likely to die, in line with our Insider Formation Hypothesis (1.1). Although the coefficient of this variable is negative, the interaction effect with time is statistically significant (Model 3). For very young parties (6 years or less, with 30% of observations falling in this range), insider status decreases the chances of party death. However, for parties that are 12 years or older (half of the observations in the sample have these values of party age), being an insider formation increases the chances of death. The effect is also substantively strong: as Table 3 shows, when the age of a party is 23.5 years (which represents the mean value plus 1 std. dev.), the hazard rate of death increases by more than 500% for insider parties compared to others. While we expected this variable to have a time-invariant effect, these findings nevertheless are in line with our theoretical logic. As the insider parties are formed by parliamentarians to promote their office ambitions, they are unlikely to be dissolved in their early years of life whilst their founders are still active. However, once the founders retire, these parties often struggle with a leadership vacuum as the former often display little interest in recruiting and promoting candidates suitable for leadership who would be able to challenge their position.
National seat share has a negative effect on the chances of death in line with our Seat Share Hypothesis (1.2). Parties with higher seat shares – providing a core payoff for ambitious elites – are less likely to die (although the variable is significant at the 0.1 level of statistical significance). The effect is important substantively: based on Model 3, the change in the values of this variable (logged seat share) from 1 std. dev. below the mean (representing no seats) to 1 std. dev. above the mean (equivalent to 4.4% of seats) decreases the hazard rate of party death by 38%.
In line with our Government Access Hypothesis (1.3), the coefficient of the government access variable is negative and significant at the 0.1 level of statistical significance (Model 3). Parties in government at the national and/or regional level are 62% less likely to die than those without presence in government. Thus, despite the moderately high correlation (0.26) between the variables measuring legislative seat share and government access, our model uncovers distinct effects of both variables. It therefore suggests that the elites of new parties value executive office beyond the privileges that they receive as national MPs.
Two factors associated with the sociological account of party death are substantively and statistically significant in the theoretically expected direction. First, as suggested in our Societal Roots Hypothesis (2.1), parties formed with the support of promoter organizations have much stronger chances of survival according to Model 2. Model 3 (and the first differences plot, see Figure 3 in Appendix 7) indicate that the effect of this variable is not time-constant. Specifically, party formations with societal roots have a lower risk of dying when the age of a party is between 3 and 18 years (60% of the sample has these values of party age) (see Appendix, Section 7, for the plots of the first differences). The effect is substantively strong: when the party is 13.5 years old (the mean value of party age), the risk of death is 54% lower than for parties being formed without the support of pre-existing groups. This finding echoes classical work stressing the importance of party ties to mobilized societal groups representing specific interests (e.g. Duverger, Reference Duverger1959; Panebianco, Reference Panebianco1988; Rose and Mackie, Reference Rose and Mackie1988). That societal roots are particularly advantageous in earlier phases of new parties’ life cycles makes sense as the formation of an organization and the building of loyalties are resource-intense and time-consuming. In earlier phases, rooted parties have advantages to achieve this as compared to organizations formed without external support. Over time, the latter formations have the chance to ‘catch up’ building a resilient organization through accessing other resources, which means that the difference between rooted formation and those without societal backup loses relevance for their risk of death in later stages of new party evolution.
Second, the Novel Formation Hypothesis (2.2) is also supported. Specifically, based on Model 3, the risk of death for such parties is 58% lower than that for the other parties. This finding is in line with a wide range of earlier empirical studies stressing the importance of ideology for new parties to carve out their own, separate niche to survive as organizations (Spoon, Reference Spoon2011; Lowery et al., Reference Lowery, van Witteloostuijn, Péli, Brasher, Otjes and Gherghina2013). However, unlike these previous studies, we suggest that ideological novelty helps parties to survive by mobilizing their members and core supporters rather than assuring them with electoral support. Indeed, the ideological novelty hypothesis is supported in Model 3 while controlling for the variables capturing a party’s parliamentary seat share and access to government. Although the latter variable affects the chances of a party’s survival, among the parties with few or no legislative seats, ideologically novel parties are more likely to survive due to the commitment of ideologically motivated activists. Interestingly, unlike the advantages enjoyed by rooted formations or the disadvantages associated with insider formations, the positive effect of novelty does not diminish over time.
In contrast, the emergence of competitor parties that represent the same party family as the new party does not increase the probability of its death. Our Distinctiveness Advantage Hypothesis (2.3) does not find support (this null result holds when using several alternative measures of ideological proximity, as explained in Section 6 of the Appendix). This indicates that being the only credible representative of a particular ideology does not help a party to survive.
Our controls capturing regional, party funding, and electoral institutions on party death are significant, which underscore the robustness of our main findings. First, echoing earlier work (Deschouwer, Reference Deschouwer2003), new parties are more likely to survive in institutional settings with a strong regional tier, but this variable is important only for the parties that are 8 years or younger. Second, the effect of the access to party funding is also time-dependent, although largely in line with what earlier work would lead us to expect. Interestingly, in comparison to the parties that received no state funding, parties entitled to state funding are more likely to die when they are young (up to 5 years) but less likely to die when they are 12 years or older. The positive effect of this variable may be a consequence of the infighting between core actors in very immature parties due to the access to state funding. The negative effect, however, is substantively more important and stresses the importance of financial resources for parties’ long-term maintenance (e.g. Casal Bértoa and Spirova, Reference Casal Bértoa and Spirova2017). Finally, a higher district magnitude (i.e. a more permissive electoral system with lower electoral thresholds) decreases the probability of the death of young parties (5 years or less) but makes the death of older parties (8 years or more) more likely. In the early years of a party’s existence, more proportional electoral institutions tend to decrease the chances of death as the translation of votes into seats is more favourable. Yet once a party grows older in such permissive systems, it has to compete with a higher number of parties, making it less likely that it can stabilize its support in the long run.
Discussion
Why do some political parties that emerge in advanced democracies fade away by ceasing to nominate candidates in elections while others manage to survive? To address this question, we developed two contrasting sets of hypotheses on party death derived from two prominent conceptions of political parties, which allows us to provide an encompassing perspective on this important phenomenon. According to the Downsian conception, political parties are vehicles for career-oriented politicians who seek to maximize their legislative and executive office. The failure of the party to provide these benefits leads to its demise as the rationale of core actors in the party to sustain it disappears. The sociological conception views parties as societal organizations that serve representational functions valued in themselves by their members and supporters. The party is therefore more likely to survive if it was formed by a promoter organization or represent a distinct ideological position, both at the time of its foundation and in the later years of its existence.
We examined our hypotheses applying survival analysis to a new data set covering 144 new parties in 17 democracies. As to be expected, some arguments derived from either conception hold, although we find somewhat stronger evidence for the Downsian perspective. Specifically, in line with this approach, our empirical analysis suggests that insider parties (i.e. parties formed by parliamentarians) are less likely, while parties with a higher share of legislative seats and access to government office are more likely to survive. The two formative characteristics of the party associated with the sociological perspective – support by a societal promoter organization and ideological novelty – also increased the chances of party survival. Finally, controls capturing federal, party funding and electoral institutions influenced party death and survival indicating the robustness of our findings.
Through the careful theoretical and empirical analysis of party political death, we contribute to a key debate in the party politics literature: the nature of political parties. The dominant view of political parties as coalitions of office-seekers has been recently challenged or complemented by the arguments that see parties as first and foremost representatives of social groups or ideologies (Hanson, Reference Hanson2010; Mudge and Chen, Reference Mudge and Chen2014). The systematic examination of the drivers of the death of new parties as organizations (as opposed to their electoral performance) over more than four decades provides a unique window to address this question. Our analysis strengthens the case for a more balanced understanding of parties by showing that societal and ideological roots of parties matter for their survival even when controlling for their electoral success and access to governmental office.
Our results also emphasize an interesting dynamic regarding the effect of the variables that capture parties’ characteristics related to their formation as opposed to those that capture their ability to achieve their central goals later in their life. In line with classical arguments stressing the importance of the ‘genetic imprint’ that formative features leave on an organization in the long term (Panebianco, Reference Panebianco1988), we find that societal rootedness, ideological novelty, and insider status all have a strong effect on party survival. Thus, the length of parties’ survival can be predicted to large extent at the time of their formation. Parties that are formed by parliamentarians without the support of a promoter organization and novel ideological profile are unlikely to survive more than several electoral terms unless they achieve consistent electoral success and access to government. In contrast, parties with strong societal and ideological roots are likely to survive for relatively long periods of time even without having access to legislative and executive office.
While for methodological reasons we tested our theoretical framework using a sample of new parties in established party systems, it is also useful for understanding the survival and death of parties not included in our empirical analysis. Both theoretical accounts developed here explain why, despite substantial electoral change in established democracies, the death of older parties (those established prior to the 1960s) embedded in established party systems has been a rare occurrence. From the Downsian perspective, the overwhelming survival of these parties can be explained by the fact that they retain substantial legislative representation, often participate in government and only in few cases were formed by parliamentarians. From the sociological perspective, these parties often started their existence as expressions of various societal interests and were also the sole representatives of their respective ideologies. Even if some of them have been reduced to minor legislative parties in the wake of recent economic crises, the sociological perspective suggests that their organizational survival is highly likely.
Finally, with some adaptations, our framework can be made useful for understanding party survival and death in countries with less stable party systems such as in Central and Eastern Europe. While the literature has made important advances in explaining the electoral persistence of parties there (Grzymala-Busse, Reference Grzymala-Busse2006; Tavits, Reference Tavits2013), research on the organizational survival and death of parties has been less prominent (but see Bakke and Sitter, Reference Bakke and Sitter2015; Casal Bértoa and Spirova, Reference Casal Bértoa and Spirova2017; Obert and Müller, Reference Obert and Müller2017). In line with our Downsian account, Obert and Müller (Reference Obert and Müller2017) find that legislative representation reduces the chances of party death, so does access to state funding. However, there is less theoretical and empirical research on the factors related to the sociological account, which could provide new insights into the ‘fleeting parties’ phenomenon observed in many younger democracies (Spirova, Reference Spirova2007; Deegan-Krause and Haughton, Reference Deegan-Krause and Haughton2015). While many of these parties lack clear ideologies (Sikk, Reference Sikk2011), anecdotal evidence suggests that those parties that have strong ties with societal organizations (e.g. many ethnic parties) or have presented a novel ideological profile (e.g. Green parties in the Czech Republic, Estonia, and Lithuania) are less likely to die.
Supplementary materials
To view supplementary material for this article, please visit https://doi.org/10.1017/S1755773918000176
Acknowledgements
The authors are grateful to the referees of the journals and to Will Jennings and David Owen for their helpful feedback on the manuscript. The authors also thank all participants in the ‘Cake for Comments Research Session’ of the Centre for Elections, Media and Participation (CEMaP) at the University of Exeter who discussed an early version of this paper. Many thanks go to Felix-Christopher von Nostitz for his research assistance. This research has received funding from the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement no. 335890 STATORG and the Marie Sklodowska-Curie Actions Programme (Intra-European Individual Fellowship grant number 330446 PARTYINSTABILITY). The support is gratefully acknowledged.
