Survival of the fittest. Comparing entry and exit among niche and mainstream parties in advanced democracies, 1945 – 2011 ☆ Electoral Studies

The turnover of political parties is a key mechanism of renewal of electoral choices. We present an organizational ecological theory on party system change, predicting that party system saturation (i.e., the effective number of parties compared to the party system ’ s carrying capacity) differently affects the entry and exit of niche and mainstream parties from lower house elections. Pooled times-series analyses on 352 elections, 509 parties, and 21 established democracies demonstrate that party system saturation indeed increases the likelihood of exit of mainstream parties but not of niche parties. Strikingly, we also find that party system saturation increases the entry of mainstream parties. Hence, an important paradox arises since oversaturation negatively affects their survival chances.


Introduction
National party systems differ considerably in the frequency with which new parties enter and existing parties exit. Some of today's advanced democracies have provided fertile breeding ground for parties such as En Marche in France, Podemos and Ciudadanos in Spain, and Movimento 5 Stelle in Italy. Yet, exit is also widespread. Consider how Italy's Democrazia Cristiana fell apart, or the total disappearance of the Dutch Lijst Pim Fortuyn and Ireland's Progressive Democrats. As noted by Hooghe and Marks (2018, 112) 'the source of dynamism in party systems is new political parties' rather than the transformation of established parties. Hence, this study explores what determines why parties enter the electoral contest (in this study: a lower house election) in the first place and what energizes them to continue doing it in the future.
Extant research has mostly examined entry and exit separately. The party entry literature has provided important insights into how structural and institutional features of a polity affect new party entry (e.g., Hug 2001;Tavits 2006). Recently, this literature has also began to consider more dynamic elements like voter discontent (Sikk 2012;Tavits 2006), party-voter incongruence (Laver and Schilperoord 2007), party collapse (Laroze 2019), voter turnout (Lago and Martínez 2011) and the effective number of parties (Kselman et al. 2016). Nonetheless, how short-term electoral market dynamics affect entry has remained underexplored. Similarly, studies on party exit have shown how party characteristics like party origin, organization and ideological novelty affect party survival (e.g., Bolleyer and Bytzek 2013;Bolleyer et al. 2019). Less is known, however, about how the space (or lack thereof) on the electoral market affects party exit.
Our article fills these voids by offering a twofold contribution. First, we build on organizational ecology, an important tradition in organizational sciences, to explain the entry and exit of political parties (also see Lowery et al., 2013). We rely on an encompassing measure of party system saturation (PSS) recently introduced by Van de Wardt and Van Witteloostuijn (2019). PSS captures the difference between the effective number of electoral parties (ENEP) present and the number predicted based on the party systems' carrying capacity. The latter is set by variables such as cleavages, electoral openness and the heterogeneity of the political supply. Based on the case of the Netherlands, Lowery et al. (2013) have found that parties were more likely to exit and less likely to enter in response to increasing PSS. To test these arguments on a much larger scale, we rely on a cross-national measure of PSS (see Van de Wardt and Van Witteloostuijn, 2019).
Second, contrary to Lowery et al. (2013), we propose that the effect of PSS on entry and exit differs between niche and mainstream parties. We argue that niche parties carve out a niche for themselves, either by emphasizing non-economic issues thereby avoiding the dominant dimension of contestation (e.g., Meguid 2005;Wagner 2012), and/or by taking an extreme position on the general left-right dimension (e.g., Adams et al., 2006). Mainstream parties, in turn, mainly emphasize economic issues and adopt centrist positions. Consequently, they are poorly differentiated from their competitors. Hence, PSS should particularly foster the exit of mainstream parties and deter them from entering.
Our analysis based on 509 parties from 21 established democracies that participated in 352 post-war elections reveal that mainstream parties are indeed likelier to exit when PSS increases. Strikingly, their entry rates increase rather than decrease with PSS. As expected niche party entry and exit is not systematically related with PSS.
In advance, it is useful to clarify, first, that party entry and exit each requires a distinct modelling strategy, which is why the research design is split up in two studies. We follow the state-of-the-art on party entry and exittwo different literaturesby studying entry with the aggregate count of new parties per election (e.g., Hug, 2001;Tavits, 2006) and party exit at the level of party-election combinations (e.g., Lowery et al., 2013). 1 Yet, since PSS varies at the level of elections, it can explain both the count of new parties per election and whether individual parties will survive until the next election. Secondly, we consider a broad and a narrow definition of entry and exit. Our broad definition includes any party as new in its first electoral contest except for a party name change; thus, parties started without the help of members of existing partiers, splinters, mergers as well as divisions are new here. Our narrow definition, in turn, considers mergers and divisions as "old" rather than new parties. We do the same in our exit analyses: the broad definition codes each last electoral contest except for a name change as exit, while the narrow definition considers mergers and divisions as survival rather than exit.

A new perspective on party entry and exit
Organizational ecology studies how competition for scarce resources (e.g., members, customers, funding) determines entry and exit of organizations such as social movements (e.g., Vermeulen 2013) interest groups (Gray and Lowery 1996), and public organizations (Van Witteloostuijn et al. 2018). Organizational ecology's basic assumption is that organizations in any organizational population rely on the same key resources to exist (Hannan and Freeman 1977). The model posits that when this population's density grows, resources become scarcer and the population's carrying capacity is reached. Subsequently, existing organizations will be driven out of business, new organizations will be deterred from entering, or both. Organizational ecology admits that adaptation to competitive pressures may help organizations to survive, but argues that deep and successful adaptation is rare, because organizations resist core changes, and if they ultimately change, they are often too late (Hannan and Freeman 1984). Hence, resource competition among parties is expected to increase party exit (and their replacement by new parties) rather than adaptation by existing parties.
Several studies have begun to apply organizational ecology to the entry, exit and adaptation of parties. Van de Wardt et al. (2017) focus on the effect of niche density on party entry and exit, finding that a higher number of competitors in a party system's niche increases exit within that niche, while not deterring entry in that niche. Lowery et al. (2013) compare the ENEP with the carrying capacity of the Dutch party system. They find that if ENEP surpasses carrying capacity, which we define as an oversaturated party system, parties will be more likely to exit and less likely to enter. In a recent study, based on a cross-national and longitudinal measure of party system saturation (PSS), Van de Wardt and Van Witteloostuijn (2019) find no evidence that parties engage in core changes (i.e., to increase the nicheness of their platform or to merge) when PSS increases. This confirms organizational ecology's background assumption of limited adaptation. This study relies on the same measure of PSS, enabling us to test the predictions of Lowery et al. (2013) beyond the Dutch case. Contrary to Lowery et al. (2013), we also propose that PSS strikes back differently on niche and mainstream parties.
PSS is the difference between the ENEP present and the carrying capacity of a party system. In organizational ecology, carrying capacity is defined as the maximum number of organizations that can be sustained in a given environment. In case of parties, the carrying capacity equals the predicted ENEP based on a country's electoral rules (Cox 1997), cleavages (Lipset and Rokkan 1967), the political agenda's issue diversity (Lowery et al., 2013) and the interactions between these variables (Clark and Golder 2006). Hence, cleavages (public demand for parties), the diversity of the political agenda (political supply), and the openness of a country's electoral institutions determine the available room for parties in the system. A party system is oversaturated when more effective parties contest elections than predicted by these variables.
How does PSS relate to other concepts of a party's competitive environment? A first body of literature focusses on the role of political opportunity structure (POS) to explain the electoral breakthrough of specific party families like green (Kitschelt 1988;Kitschelt, 1994) and radical right (Kitschelt and McGann 1997;Rydgren 2004). Within the POS approach, the crucial independent variable is whether there in an ideological niche for a specific ideology. Next, general political opportunities like the openness of the electoral system and the ideological convergence of mainstream parties are considered (e.g., Rydgren 2004). These models are relatively good in explaining the successful entry or electoral success of specific parties or party families under particular circumstances. Yet, since a fundamental part of the model relies on ideology-specific opportunities, it cannot simultaneously be applied to different party families. The latter is the main strength of our indicator of PSS. Below we work with a parsimonious model to measure the carrying capacity for political parties in general that travels across party families, party systems of consolidated democracies and time. Consequently, our model of PSS is relatively high on Sartori's (1970) ladder of abstraction. As he noted, this is appropriate for concepts that are designed to travel across regions (in our case: across consolidated democracies), and thus for the purpose of this paper.
Second, it is relevant to stress that PSS differs from extra-system volatility (the amount of vote shifts from established to new parties): an indicator which has recently been proposed as indicative of the permeability of the electoral market (Mainwaring et al. 2017, 2). While extra-system volatility infers permeability from the success of newcomers, PSS focusses on the available space in the party system. Hence, contrary to the extra-system volatility indicator, PSS is also a potential cause of the electoral success of newcomers rather than only its result. In fact, an election with high extra system volatility likely increases PSS and could subsequently increase the competitiveness and limit the permeability of the party system for future newcomers.
Finally, it is important to note that Laroze (2019) argues that party collapse in the previous elections could serve as a measure of the size of the policy space that could be exploited by new parties. She also shows empirically that the successful entry of new parties is directly related to the exit of parties in the previous elections. Hence, this relationship between party collapse and successful party entry plays lip service to our assumption that party systems have finite carrying capacities. Yet, only taking the degree of party collapse as an indicator of open space neglects that simultaneously the carrying capacity for parties may be expanding or contracting.
In sum, the measure of PSS first introduced by Van de Wardt and Van Witteloostuijn (2019) offers a novel approach to measure the competitiveness of party systems.

Differences between niche and mainstream parties
Below we formulate our hypotheses regarding party entry and exit, which happens when a party contests lower house elections for the first or last time. While party entry is always a voluntary decision, we assume that party exit can be both voluntary and involuntary. Our expectations differ from Lowery et al. (2013) on one crucial aspect, namely that we expect PSS to strike back differently on niche and mainstream parties. Importantly, several competing approaches exist. First, scholars differ as to whether nicheness is to be determined based on the issues parties emphasize or based on their positional extremity. The saliency approach argues that niche parties reject the traditional class-based orientation of politics and that they focus on a limited set of issues. Mainstream parties do the exact opposite and stick to economic issues (Bischof 2017;Meguid 2008;Wagner 2012). In turn, Adams and co-authors (2006) popularized a spatial conception, defining niche parties as those with extreme positons on the general left-right position, while arguing that mainstream parties occupy the political centre. Throughout we assume that parties can either be distinctive because they mobilize non-economic issues, because of their extreme left-right platform, or because of both. Specifically, communist, ecologist, ethnic, Protestant, radical right, regionalist, special issue and the non-Scandinavian agrarian parties are defined as niche. Christian democrat, conservative, liberal, social democrat and Scandinavian agrarian parties as mainstream. 2 Consequently, we differ from the saliency approach, as based on Adams et al. (2006) we code communist/radical left parties as niche because they have an extreme left-right position, while the saliency-approach either discards these parties (e.g., Meguid 2008) or codes them as mainstream (e.g., Bischof 2017). In turn, we differ from Adams et al. (2006) and follow the saliency approach by including regionalist, ethnic and special issue parties as niche based on their emphasis of non-economic issues. 3 Second, there is controversy about whether niche and mainstream parties are to be operationalized based on their party family as above (e. g., Adams et al., 2006;Meguid 2008;Spoon and Klüver 2019) or by means of concrete characteristics that can vary by elections (Bischof 2017;Bischof and Wagner 2017;Meyer and Wagner 2013). As explained by Mair and Mudde (1998), the party family approach would be useful if one wishes to focus on the ideological imprint of parties, thus, on what they are rather than what they do. Political parties are identified as ideological vessels consisting of a whole package of characteristics that together constitute a party's identity. This way of examining parties contrasts with the "policy approach" that focuses on what parties do in one specific election by examining time-variant characteristics like the nicheness or extremity of a party's profile in isolation (e.g., Bischof 2017; Wagner 2012). Besides ideological characteristics, scholars have also proposed to directly measure other alleged characteristics of niche parties like the degree to which they are policy-seeking and dominated by their activists (Bischof and Wagner 2017). 4 We will adopt elements of both approaches. Our main emphasis will lie on party family, as we do not wish to fully detach ideologically differentiation from the genetic imprint of parties. As argued by Downs (1957, 142), ideologies cannot be thrown off as if they were disguises. Even if a mainstream party would enter a niche segment, it has been shown that parties cannot "steal" associative issue ownership (the spontaneous association by voters between issues and parties) from other parties (Tresch et al. 2015). Hence, party family better gauges that niche parties enjoy a first mover advantage on the niche segments of the electoral market (also see De Vries and Hobolt 2020, 54). While we certainly do not suggest that what parties do in one election has no consequences for how they are differentiated, we assume that a party's ideological imprint carries more weight than its platform in one election. Focusing on party family also enables us to consider that niche parties differ from mainstream parties on multiple dimensions (e.g., ideological differentiation, issue ownership, party goals). Hence, it is the whole package that must be considered.
That said, we believe that focusing on party attributes like nicheness also has numerous advantages. First, rather than treating the niche/ mainstream category like an empirical commonality (Wagner 2012), this approach directly taps into the alleged differences between niche and mainstream parties. Hence, we will also directly study the impact of concrete traits ascribed to niche parties (also see Bischof and Wagner 2017). Another strength is that the nicheness approach treats nicheness as a relative concept (Meyer and Miller 2015, 262). Therefore, both our nicheness and positional extremity measure below capture the distinctiveness of the focal party's platform vis-a-vis the other parties in the system.

Party exit hypothesis
For our theoretical argument, it mainly matters that niche parties are better positioned to avoid competition with other parties on the same issues for the same voters than mainstream parties. This is because of their ideological differentiation (either through salience or positions). Mainstream parties, however, tend to focus on economic issues and to moderate their position towards the centre to maximize their vote and office ambitions (Adams et al., 2006;De Vries and Hobolt 2020, 51). While such a catch-all strategy could maximize their votes in the short run, the main drawback is that mainstream parties become poorly differentiated from their competitors, which will also make it more difficult for them to differentiate and effectively target their supporters (Ezrow et al., 2010). Another drawback is that the centrist voters targeted by mainstream parties display less solid associations with political parties, meaning that they can easily lose their voters in the future (Karreth et al., 2013). Hence, despite that mainstream parties are generally more of a constant in party systems than niche parties, a share of their electorate may still be weak partisan identifiers. Moreover, the catchall strategies of mainstream parties could alienate their less centrist core supporters (Karreth et al., 2013) who potentially provide them with an enduring source of support and other resources like time and money (Schofield and Sened, 2005).
For niche parties we expect the opposite characteristics to apply. As 2 Extant research does not distinguish between Christian-democrat and Protestant parties. Yet, since the latter tend to mobilize more conservative positions on morality issues, we believe that they are clearly differentiated from Christian democratic parties. Further we distinguish Scandinavian agrarian parties from other agrarian parties, as the first are often regarded as mainstream parties. They hold centrist positons and have been in government coalitions. 3 While both Adams et al. (2006) and saliency based work code green parties as niche, in our view, they better fit the saliency-based definition as scholars have argued that they have relatively moderate positons (e.g., Dalton 2009), which would make them mainstream in a spatial sense. It is, however, uncontestable that they mobilized around non-economic issues. 4 A third controversy is whether due to the rise of new issue dimensions, some niche parties have become mainstream. However, Wagner (2012, 850-51) shows that in most countries economic issues remain the primary focus of electoral debates. Notwithstanding that since the 1980s the salience of the liberal-authoritarian dimensions has increased, policy niches have increased at the same time. That is, mainstream parties have certainly increased attention for these issues and shifted to the right, but the radical right has done the same (Wagner and Meyer 2017). Hence, we believe that there is insufficient ground to make claims that from a certain moment onward niche parties like the radical right become mainstream. said, they are clearly differentiated ideologically. Moreover, several studies have argued that these parties are policy-seeking and mostly responsive to their core voters (Adams et al., 2006, Ezrow et al., 2010, Przeworski, 1986Klüver and Spoon, 2016; but see Tromborg, 2015). 5 For niche parties this would be an effective vote-seeking strategy because their supporters are supposed to be more policy-oriented than the voters of mainstream partiers (D'Alimonte, 1999;Kitschelt 1997). Because of their horizontal party organization, the leaders of niche parties also have less leeway to respond to environmental incentives (Adams et al., 2006;Ezrow et al., 2010). Hence, according to these authors, niche parties are confined to stay within their niche and do not respond to public opinion, even though this sets a natural ceiling in terms of votes (Adams et al., 2006). Based on the aforementioned studies, we contend that especially mainstream parties will become vulnerable in an oversaturated party system. They are less distinguishable from competitors and owing to their preference of votes over policy and the power of the party leadership over their activists, they have less of a loyal base to rely on. Lastly, mainstream parties may be more inclined to voluntarily exit an overcrowded electoral market, as electoral success is their primary motivation. We, therefore, propose that PSS will only increase the exit of mainstream parties, while niche parties' chances of exit are not systematically related to PSS.
Party Exit Hypothesis (H1). Party system saturation only increases the likelihood of exit of mainstream parties.
As said above, we will also examine whether the concrete characteristics (ideological nicheness, positional extremity, being office seeking and being dominated by the party activists) commonly ascribed to niche parties indeed mitigate the effect of higher PSS (also see Bischof and Wagner 2017;Schumacher et al. 2013).

Party entry hypothesis
While we can clearly derive from the organizational ecological literature that oversaturated systems revert to their carrying capacity through increased exit rates, it is less clear-cut whether oversaturation also suppresses entry. The traditional model posits that when the population reaches its carrying capacity, resources become scarce, and organizations will be deterred from entering (Hannan and Freeman 1977). Yet, other studies suggest that higher density can, for instance, create better opportunities for learning organizational skills or may ease the cost of collective actionboth of which lead to higher rates of entry (Aldrich 1999). Furthermore, another view citing Colinvaux, a biologist (Gray and Lowery 1996, 3), suggests that prospective entrants might simply be incapable of strategically acting upon the carrying capacity: 'the way an animal breeds has very little to do with how many of it there are … The numbers that may live are set by the environment, and these are quite independent of how fast a species makes babies'. Hence, if this is true, a higher number of organizations will foster organizational death but not lower entry rates.
The party entry literature, however, appears to largely agree that new parties are able to strategically respond to their environment. The theory of strategic entry assumes instrumentally rational elites who must maximize their vote share in the short-run to enjoy the spoils of office and/or to influence policy (Cox 1997;Tavits 2006, 102). Assumed is that new parties will only enter the electoral contest if the benefit of holding political office multiplied by the probability of getting elected is equal to or higher than the cost of entry.
In keeping with the assumption of new parties that optimally strategize in response to their environment, we expect that PSS will influence mainstream party elites' entry propensity. Mainstream parties are less well distinguished from their competitors. Hence, they should be discouraged from entering oversaturated party systems where ideological differentiation is important. Analogous to their exit, we expect no systematic relationship between PSS and niche party entry.
Party Entry Hypothesis (H2). Party system saturation only decreases the entry of mainstream parties.
Contrary to H1, we will not evaluate H2 based on the concrete characteristics of parties, as new parties are hardly included in extant databases in their first electoral contest.

Measuring party system saturation
We compiled a longitudinal dataset containing information between 1945 and 2011 for 21 established Western democracies (see Fig. 1). To measure PSS we use the residuals from a model that regresses the ENEP on a party system's characteristics. We rely on the model of ENEP published in Van de Wardt (2017, Model 2, p. 50). The ENEP, or the carrying capacity of the party system, is determined by societal heterogeneity (measured on the basis of the cleavage that is seen as most consequential within a country) (Clark and Golder 2006), the diversity of the political supply side in terms of the dimensionality (Stoll 2011) and fractionalization (Lowery et al., 2013;Zons 2015) of the party system agenda. Hence, these time-variants factors capture changes in the nature of party competition such as increased dimensionality. Each of these explanations is interacted with the permissiveness of the electoral system to capture the intuition that the latter may act as a brake or catalyst (Clark and Golder 2006). The measure is based on 387 elections that took place in 25 advanced democracies between 1945 and 2011 (which is why our sample stops in 2011). This model explains 31% of the variance. Van de Wardt (2017) shows that the explanatory power of the model drops and determinants of ENEP fail to have their hypothesized effect when the newer democracies of Central and Eastern Europe are added to the analysis. Therefor we focus on consolidated democracies only. Of the 25 consolidated democracies for which we have PSS data, we could only include the 21 that are also in the ParlGov database from which we derive entry and exit data. For further detail, we refer to Van de Wardt's (2017), Van de Wardt and Van Witteloostuijn (2019) and the SI (Section 3.1). Fig. 1 maps the observed and predicted ENEP for each country. The distance between the two lines depicts the degree of under or oversaturation. If the grey line is above the black line, the system is oversaturated; if the black line is above the grey line, the system is undersaturated; and if the two lines overlap, the system is saturated. For instance, the high oversaturation of the Italian party system in 1994 can be understood from the fact that it simultaneously adopted more disproportional electoral rules (as denoted by the black line, the carrying capacity contracts) and witnessed increasing electoral fragmentation (as shown by the grey line) in the aftermath of the Mani Pulite corruption scandal which caused the demise of the Christian Democrats and the entry of new parties. In turn, the heavy undersaturation of New Zealand's party system from the early nineties onward is driven by the 1993 electoral reform from First Past the Post to Mixed Member Proportional system, which substantially widened the carrying capacity. Yet, as shown the observed ENEP has only moderately increased, suggesting that new and existing parties have not profited very much from this opportunity. The Spanish case provides an example of a party system that has been continuously undersaturated throughout the time span of our analyses. This is no surprise, given that until the 2015 elections, the party system was dominated by only two parties: the Partido Popular and the PSOE. However, more recently and beyond the time frame of our analysis, parties like Podemos, Ciudadanos and Vox have exploited this open space. These examples illustrate the face validity of the PSS measure. 5 Tromborg (2015) finds evidence that niche parties moderate their platform when a proximate moderate coalition forms. This suggests that under specific circumstances, niche parties could also respond to electoral incentives.

Empirical strategy for the Party Exit Hypothesis (H1)
We are interested in how parties adapt to their environment (in this case: PSS) regardless of how successful they are in doing so. Hence, we define entry and exit as (non)-participation in the electoral contest, meaning that a party could enter and exit without ever having gained parliamentary representation. For exit and entry data, we rely on Parl-Gov (Döring and Manow 2015). Parties obtaining at least 1% of the vote or one seat in lower house elections are systematically included in this extensive dataset. Thus, given our aims, we needed to ensure that all party-election observations are included: if party X runs in elections t and t+1 but only in election t manages to acquire at least 1% of the votes or a seat, Parlgov will exclude party X in election t+1, while we will include it at t and t+1. 6 The first dependent variable, Party exit, is dichotomous (0 alive, 1 exit) and measured at the level of party/elections. A party receives a score of 1 if it does not contest any other lower house elections after t. We employ two operationalizations of exit. The first is unrestrictive and follows Lowery et al. (2013): only party name changes are excluded as instances of exit. This means that parent parties exit, even if the party merges, or divides into multiple new parties. This operationalization is consistent with organization ecological theory in the espoused belief that when a party merges or falls apart, it is likely not competitive in its current form. Our second operationalization is narrower and excludes mergers and divisions.
We created a binary indicator denoting niche or mainstream status based on the ParlGov data on party family. Specifically, we coded non-Scandinavian agrarian parties, communist, ecologist, ethnic, Protestant, radical right regionalist and special issue parties as niche and Christian democrat, conservative, liberal, social democrat and Scandinavian agrarian parties as mainstream.
Besides party family, we also focus on concrete party characteristics. Our time-variant nicheness indicator is based on parties' issue emphases in their election manifestos derived from the Comparative Manifestos Project (CMP) (Volkens et al., 2013). We aggregated all issues into Stoll's (2011) seven policy dimensions: economic, cultural-ethnic, religious, post-materialist, foreign policy, democratic-authoritarian, or urban-rural. Following Meyer and Wagner (2013), we calculated how many standard deviations a party's emphasis of a dimension was above the weighted average emphasis of the other parties in the same election. Issue emphases were weighted according to parties' vote shares. Then we subtracted the party's standard deviation on the economic dimension from its largest positive standard deviation on any of the non-economic niche dimensions. 7 As for the other characteristics, to measure the extremity of a party's general left-right position, we calculated the absolute distance in standard deviations of a party's so-called rile score to the mean of the other parties in the system. And the extent to which the party leadership prefers policy over office and that activists dominate the party leadership was derived from an expert survey by Laver and Hunt (1992).
Finally, we controlled for party-level explanations of exit, i.e., the number of consecutive elections that they were not represented in parliament/government (Bolleyer 2013) and the rootedness (Bolleyer and Bytzek 2013) and origin (genuinely new, merger, splinter, division, successor) of their organization (e.g., Beyens et al. 2016). Data on  Table A6 (Model 2) of SI. 6 That we still exclude parties that never obtained 1% of the votes or one seat poses no threat to the validity of our conclusions. A large share of these parties could run for other objectives than electoral success like political satire or entertainment. As such, they offer an unsuitable testing ground for the hypotheses. 7 A positive score on this continuum implies that a party's relative emphasis of the niche segment it emphasizes the most is above its emphasis of the economy; a negative score denotes the reverse; finally, a score of zero suggests equal emphasis. electoral/government participation and party origin from Parlgov (Döring and Manow 2015). Descriptive statistics are provided in Table A3 of the SI.
Party exit was examined by means of discrete time duration models (Rabe-Hesketh and Skrondal 2008). The data structure is such that a party contributed one party/election observation for each election it was included in our dataset. Thus, a party received a score of 1 on the dependent variable, Party exit, if it will not reappear in any other election after t (0 if reappears). We capture the hazard rate of exit by transforming the number of elections contested until t into n-1 dummy variables. In so doing, we control for the fact that older parties are more institutionalized, helping them to survive (Panebianco 1988). Since elections are clustered within countries and parties, we specified standard errors clustered at both levels. 8

Evaluating the Party Exit Hypothesis (H1)
Table A1-A10 of the SI present the case selection, model specifications, descriptive statistics and regression tables for each analysis in this paper. Fig. 2 visualizes the results. The time dummies are omitted from the table, but Figure A1 of the SI presents the hazard rate of exit, showing that the hazard decreases the more elections a party has contested. As shown in Fig. 2, we find that PSS increases exit both under the broad (log odds = 0.202, p < .05) and narrow definition (log odds = 0.143, p < .1). However, as denoted by the significant interaction PSS*Niche party, the effect differs between niche and mainstream parties. Due to the presence of the interaction term, the "main effect" of saturation can be interpreted as the effect for mainstream parties (Niche party = 0). If PSS increased with one standard deviation, the logged odds of mainstream party exit increase by 0.388 (p < .01, Broad definition), or a factor of 1.474, which is a sizable effect. In turn, the significant interaction between PSS and niche party (b = − 0.389, p < .01, Broad definition) shows, that for niche parties, the effect is about zero (0.388-0.389). Notably, the same pattern emerges under our narrow definition of exit. This provides strong support for H1.
In Fig. 3 we present the marginal effects. As shown in the figure on the left, PSS increases the exit of mainstream parties, but has no effect on niche party survival. The right hand figure presents the marginal effect of being a niche party (compared to being a mainstream party) on the likelihood of exit for increasing values of PSS (x-axis). Should the confidence intervals entrap the zero line, niche and mainstream parties have the same chance of exit; and in cases of positive (negative) values, niche parties are more (less) likely to exit. As shown, when a party system is under saturated with − 2.8 effective parties or less, niche parties have significantly higher odds of exit than mainstream parties. This higher baseline propensity of exit could, for instance, be due to the fact that niche parties are often younger organizations, which increases organizational failure (Freeman et al. 1983). Also we know that if niche parties join coalitions, pressures to moderate will become intense, which combined with their inexperience will likely foster electoral defeat (e.g., Bale and Dunphy 2011) followed by exit. In turn, when the system is oversaturated with 0.1 effective parties or more, this pattern reverses, and mainstream parties become confronted with higher exit chances.  Table A8 (Model 1-4) of SI.
When the system is oversaturated with, say, 4 effective parties, niche parties' probability of exit becomes 0.145 times lower than mainstream parties (see Fig. 3). Fig. 4 shows that except for positional extremity, all traits commonly associated with niche partiesi.e., their programmatic nicheness, dominance of their supporters and preference of policy over office (Bischof and Wagner 2017) -all mitigate the effect of PSS. For parties with a nicheness score below 0, PSS significantly increases exit; when nicheness ranges between 0 and 12, PSS exerts no effect; and when nicheness is above 12, PSS actually reduces the likelihood of exit. Further we see that activist dominance reduces exit. That is, as soon as activist dominance is above 8 (at the midpoint of 10 they are as dominant as the party leadership), PSS no longer increases exit. Finally, we find that if parties are very policy seeking, the effect of PSS on exit dwindles. Hence, except for positional extremity, all characteristics commonly ascribed to niche parties matter. We, however, stress that the non-findings on positional extremity do not suggest that radical left parties are not protected against PSS. In Figure A11 of the SI we show that like the other niche party families, their exit rates are left unaffected by higher PSS. This nicely illustrates the added value of comparing the impact of ideological imprint (based on party family) with what a party does in one particular election.
All controls are in the expected direction. Most importantly, rooted organizations were less likely to exit (e.g., Bolleyer and Bytzek 2013) and those longer excluded from parliamentary representation and office are more likely to do so (Bolleyer 2013).

Empirical strategy for the Party Entry Hypothesis (H2)
We constructed three dependent variables: the number of new parties, new niche parties, and new mainstream parties at election t. Contrary to party exit, we cannot study entry at the party/election-level. This would require comparing all parties that exist as organizations where some decide to contest elections and others not to. However, it is virtually impossible to gather reliable data on all parties existing as organizations. Since we focus on aggregate entry rates, we cannot interact PSS with the type of party as in H1. Hence, here we test H2 by evaluating the effect of PSS on the aggregate entry of mainstream and niche parties in two separate models (Hence, throughout the N is the same). Also contrary to our analysis of party exit, we only evaluate party entry based on party family due to the simple fact that the time variant characteristic used above are unavailable for most new parties. We translated our broad and narrow definition of exit (see above) to our entry analyses. Our narrow definition is in line with the literature on strategic entry that only considers parties started without the help of members of existing parties or splinters as new parties (Hug 2001;Tavits 2006).
Keeping with the theory of strategic entry (Tavits 2006), we controlled for variables tapping into the costs (i.e., registration costs, logged petitions, public party financing) and benefits of entry (corporatism) and a party's likelihood of success (i.e., duration of democracy, changes in unemployment rate and GDP and the logged size of the population). Data derived from Tavits (2006). Descriptive statistics are displayed in Table A4 of the SI.
Since the dependent variables are over-dispersed count variables, we opted for a negative binominal regression (Hall et al. 1984), clustering elections within countries.

Evaluating the Party Entry Hypothesis (H2)
In Fig. 5 we again begin with the effect of PSS on all parties. Strikingly, contrary to Lowery et al. (2013), we find that PSS at t-1 produces higher rather than lower entry rates at t, but only under our broad entry definition (log of expected counts = .179, p < .1).
In Fig. 6 we disaggregate entry rates by niche (top) and mainstream parties (bottom), we again find differences. Niche party entry rates do not systematically respond to PSS. 9 For mainstream parties' we find, contrary to H2, that their entry rates significantly increase rather than decrease when PSS was higher in the previous elections (logs of expected counts = .527 against broad definition and 0.484 against narrow definition, p < .05). If PSS increases with one standard deviation, the entry of mainstream parties would increase with a factor of 1.69. This appears a sizable effect. However, Fig. 7 displays the predicted number of newly entering parties (upper-left), niche parties (upper-right) and mainstream parties (bottom-left) along the observed values of PSS based on the broad definition. As indicated by the increasing slope of the line in Fig. 7, the effect of PSS on mainstream party entry is clearly nonlinear. It is problematic that the steeper slope at very high levels of PSS lacks common support in the data (Hainmueller et al. 2019), as shown by the white-barred histogram in the background of Fig. 7. If we were to increase PSS along its interquartile range (from − 0.73 to 0.71), the number of new mainstream parties would only increase with 0.19 new parties. Even though this effect is statistically significant (p < .01), it is more modest than suggested by the overall coefficient. That being said, it remains at odds with H2. 10 Why is that so? One plausible explanation consistent with our findings on exit (H1) could be that higher PSS at t-1 increases the exit of mainstream parties between t-1 and t, which in turn creates space for new mainstream parties at t. Moreover, as discussed below in the example of the UDF, if a party exits, different types of successors may rise from its ashes, which would increase entry rates in the next elections. This is most obvious in case of merges: the predecessors that exit at t will produce a new party at t+1. Additionally, also activists or politicians of disbanded parties can launch new parties. To test whether the effect of PSS t-1 is indeed mediated by the exit of mainstream parties after t-1, Model 2 controls for the exit of mainstream and niche parties after the previous elections. While these exit rates exert no effect on the entry of niche parties, we find that higher exit rates of mainstream parties do increase their entry rates at t. While the effect of PSS shrinks, it remains statistically significant, suggesting that the effect of PSS on mainstream party entry is partly mediated through its effect on exit.
It is unsurprising that the effect of mainstream party exit is larger under the broad definition. This definition also includes mergers and divisions as instance of entry and exit. These types of exit will by definition increase entry rates in the next elections. Hence, perhaps it is more interesting to look at our narrow definition that discards such transformations as cases of entry and exit. Recall that it only considers parties formed without the help of existing parties or splinters as new parties. Again we find that the effect of PSS is mediated by the exit of mainstream parties at t-1. So, mainstream parties that get weeded out in an oversaturated electoral marked will be replaced by new mainstream parties. Against this narrow definition, we also find that mainstream party entry increases in response to niche party exit. Yet, since we have already shown above that there is no relationship between PSS and niche party exit, this is not a mediated effect as above. 11 As for the controls, most variables pertaining to the costs and benefits of entry are insignificant. Along with the null findings on H2, this casts new light on the assumption that within consolidated democracies new parties smoothly act upon costs and benefits of entry and factors enhancing their likelihood of success (Tavits 2006). Caution is, however, required. Like Tavits (2006) our narrow definition of new parties includes both genuinely new and splinter parties; yet, Zons (2015) has  Table A9 of the SI (90% ci). For continuous variables the coefficients denote the effect of a one standard deviation increase. N = 184. 9 As shown In Figure A12 of the SI, this is because the effects for individual niche party families are either insignificant or in different directions. We refrain from going in detail here as H2 only expresses expectations about the empirical pattern for mainstream parties. 10 As discussed in Section 2.4. of the SI, we also find that PSS increases the entry of communist parties. A likely explanation is that after the fall of the Berlin Wall, many of these parties disbanded, while producing radical left offspring parties (March and Mudde, 2005). 11 Also at the aggregate level we find that PSS t-1 only increases mainstream party exit.
shown that both subtypes differently respond to the incentives in the model. Nonetheless, that mainstream parties' entry rates do not decrease with PSS and even slightly increase (while they have poorer survival chances in oversaturated systems), suggests that a large subset of parties does not optimally respond to its environment.

Sensitivity analyses
Several analyses, detailed in the SI (Section 2), were carried out to assert the robustness of our findings. They involved: (1) estimating PSS based on the residuals from Clark and Golder's (2006) study on the ENEP, (2) rerunning the model on different operationalizations of niche and mainstream parties, (3) testing whether the findings hold if we disaggregate the niche party concept into its constituent party families, (4) asserting that the findings are not biased by specific sub dimensions driving PSS, (5) asserting that the results are not biased by potential influential cases in our estimation of PSS, (6) showing that PSS continues to exert an effect if one controls for the carrying capacity around a party's position, and (7) examining H1 with more country-level controls.

Conclusion and discussion
Parties are crucial actors in a democracy. It is therefore important to develop an understanding of what determines their entry and exit. Our analyses on 21 established democracies between 1945 and 2011 demonstrate that party system saturation (PSS) increases the exit of mainstream parties. Strikingly, we found no evidence that mainstream parties' entry rates respond accordingly: they are more rather than less likely to enter oversaturated party systems also if we exclude mergers and divisions as instances of entry and exit.
Our new, organizational ecological perspective on party entry and exit helps explain why mainstream parties, such as Belgium's Vivant, France's Union Pour La Démocratie Française (UDF), and Denmark's Centrum-Demokraterne disappeared from the political scene. In the elections preceding their exit, we find that the ENEP exceeded carrying capacity. Subsequently, each of these parties was confronted with electoral defeat, after which they disbanded. In its last 2002 elections, the UDF, for instance, was confronted with many rivalling mainstream parties (e.g., UMP, Parti Socialiste), some even splinters from itself (Démocratie Libérale and Rassemblement pour la France), making it difficult to set themselves apart ideologically. After its electoral defeat, the party split in two: one fraction supported by UDF's member base started a social liberal party, MoDem, while the other wing founded Nouveau Centre. The UDF itself ceased to exist in 2007. The fact that two new mainstream parties, MoDem and Nouveau Centre, rose from its ashes, despite the oversaturation of the French party system, may appear counterintuitive, but is fully consistent with our finding that oversaturation does not deter but even slightly increases mainstream party entry.
By applying organizational ecological theory to party entry and exit, this study provides several important innovations. First, the party competition literature commonly assumes that entry and exit occurs independently from the presence of competitors. PSS offers an empirical approach to add the relative crowdedness of the electoral market more explicitly to the equation. Second, by showing how PSS interacts with party characteristics, we managed to fruitfully bridge the party competition literature and organizational ecology. Rather than assuming that environmental pressures similarly affect parties, as Lowery et al. (2013) do, we have shown that PSS differently affects the  Table A10 (Model 1, 2 and 4) of SI. entry and exit of niche and mainstream parties. Third, organizational ecology recommends studying organizational entry and exiting simultaneously, as populations of organizations could be moving back to their carrying capacity because of increased exit, inhibited entry, or both. We found that mainstream parties are more likely to exit in response to PSS; yet, PSS also increases their entry.
Even though the effect on entry is only modest, we have explained this seeming paradox with the notion that by increasing the exit of mainstream parties, PSS will also increase the entry of mainstream parties in the next elections. As illustrated in the UDF example, mainstream parties driven out of business can still produce offspring through the activities of its activists or former politicians. Also they can be replaced by genuinely new mainstream parties expecting to benefit from the space they left behind. A prominent example of this type of replacement includes Italy's Democrazia Cristiana (DC). After its dismal showing in the 1992 elections the party was disbanded. This was not only because of the Mani Pulite corruption scandal, but also because of increased electoral competition due to, for instance, the rise of Lega Nord. In fact, our data shows that in 1992 the Italian party system was oversaturated with 2.2 effective parties. Consistent with our findings for H2, this encouraged new mainstream parties to enter the next 1994 elections. While the Partito Populare Italiano (their formal successor) and Centro Christiano Democratico (a splinter) can be seen as DC's own offspring, also a prominent, genuinely new party, Forza Italia, emerged on the scene. This even increased the oversaturation of the system to four effective parties. That an overcrowded market does not discourage but even slightly encourages mainstream party entry suggests that this type of parties does not optimally respond to its environment. As argued, mainstream parties are less well distinguished from their competitors, meaning that they would benefit from entering undersaturated party systems. While merging (producing exit and entry) could still be seen as rational adaptation to an oversaturated electoral market (Van de Wardt and Van Witteloostuijn, 2019), PSS also increases the narrow type of entry (i.e., splinters and genuinely new ones).
Why do mainstream parties depict this suboptimal behaviour? A first explanation could be that career politicians are more likely to self-select into mainstream than niche parties (also see Panebianco 1988;Schumacher et al. 2013, 466). As compared to 'believers' whose participation primarily depends on the party's platform, 'careerists' are more driven by material and status-oriented incentives. Arguably, careerists will be more inclined than believers to start new parties instead of exiting politics if their party is driven out of business, even in an oversaturated party system. In this way, exited mainstream parties thus leave behind the seeds for new mainstream party entries. A second account that could also explain why genuinely new mainstream parties are more likely to enter oversaturated electoral markets could be that whether to pursue one's goals by forming a political party is a complex decision involving information shortage. Party founders must assess their potential electoral viability together with the costs and benefits of entry. While information on the cost of ballot access is readily available, assessing ones potential electoral viability involves considerable uncertainty. According to bounded rationality theory, in case of uncertainty, human beings will rely on proxies (Bendor et al., 2011). One such proxy could be electoral fragmentation where high electoral fragmentation would encourage the entry of new parties because it signals that the electoral market is well permeable (Mainwaring et al. 2017). Hence, when considering the launch of a new mainstream party, elites could infer from the oversaturation of the party system that they can be electorally viable in that voters are volatile and willing to consider a broad range of options. Consequently, higher PSS will increase mainstream party entry rather than decrease it as we originally expected.
While these two explanations could potentially explain the paradox, we stress again that the effect on entry is only weak and that future research in this area is needed. Another challenge for future research would be to develop an approach that would enable to measure the saturation in specific locations in the party system. Yet, measuring the carrying capacity for different party families will be a difficult task, both theoretically and methodologically, as fine-grained explanations and data would be required. This paper has provided an important first step by showing how the carrying capacity of the party system as a whole affects the entry and survival of niche and mainstream parties.