+ All documents
Home > Documents > Policy agendas, and births and deaths of political parties (Online First

Policy agendas, and births and deaths of political parties (Online First

Date post: 24-Jan-2023
Category:
Upload: rug
View: 0 times
Download: 0 times
Share this document with a friend
28
http://ppq.sagepub.com/ Party Politics http://ppq.sagepub.com/content/early/2011/07/01/1354068811407576 The online version of this article can be found at: DOI: 10.1177/1354068811407576 published online 4 July 2011 Party Politics Gherghina David Lowery, Arjen Witteloostuijn, Gabor Peli, Hollly Brasher, Simon Otjes and Sergiu Policy agendas and births and deaths of political parties Published by: http://www.sagepublications.com On behalf of: Political Organizations and Parties Section of the American Political Science Association can be found at: Party Politics Additional services and information for http://ppq.sagepub.com/cgi/alerts Email Alerts: http://ppq.sagepub.com/subscriptions Subscriptions: http://www.sagepub.com/journalsReprints.nav Reprints: http://www.sagepub.com/journalsPermissions.nav Permissions: What is This? - Jul 4, 2011 OnlineFirst Version of Record >> at Tilburg University on September 19, 2012 ppq.sagepub.com Downloaded from
Transcript

http://ppq.sagepub.com/Party Politics

http://ppq.sagepub.com/content/early/2011/07/01/1354068811407576The online version of this article can be found at:

 DOI: 10.1177/1354068811407576

published online 4 July 2011Party PoliticsGherghina

David Lowery, Arjen Witteloostuijn, Gabor Peli, Hollly Brasher, Simon Otjes and SergiuPolicy agendas and births and deaths of political parties

  

Published by:

http://www.sagepublications.com

On behalf of: 

Political Organizations and Parties Section of the American Political Science Association

can be found at:Party PoliticsAdditional services and information for    

  http://ppq.sagepub.com/cgi/alertsEmail Alerts:

 

http://ppq.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

What is This? 

- Jul 4, 2011 OnlineFirst Version of Record>>

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

Article

Policy agendas andbirths and deaths ofpolitical parties

David LoweryUniversity of Leiden, the Netherlands

Arjen WitteloostuijnUniversity of Antwerp, Belgium, Utrecht University and Tilburg University, theNetherlands

Gabor PeliUtrecht School of Economics, the Netherlands, and Antwerp Centre of EvolutionaryDemography (ACED), Belgium

Holly BrasherUniversity of Alabama at Birmingham, USA

Simon OtjesUniversity of Leiden, the Netherlands

Sergiu GherghinaUniversity of Leiden, the Netherlands

AbstractThe standard model of political party density emphasizing the interaction of social cleavagesand district magnitude (M) is incomplete in accounting for number of parties in cases of highvalues of M in an arbitrary way. We explore an alternative model for such cases emphasizingthe slack in the issue agenda available to parties with which to construct viable identities orniches they can employ to mobilize cognitively-limited voters. The model is tested with timeseries data and event history analysis on the sizes of the public policy agenda and the political

Corresponding author:

David Lowery, Pieter de la Court gebouw, Wassenaarseweg 52 2333 AK, Leiden, the Netherlands

Email: [email protected]

Party Politics1–27

ª The Author(s) 2011Reprints and permission:

sagepub.co.uk/journalsPermissions.navDOI: 10.1177/1354068811407576

ppq.sagepub.com

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

party system in the Netherlands, an extreme case of large district magnitude given its singlenational district. Change in the slack of the issue agenda influences the births and deaths ofpolitical parties, a result thatmay also have implications for cases of lower districtmagnitude.

Keywordsparty density, policy agendas, the Netherlands

Paper submitted 5 November 2009; accepted for publication 25 February 2011

The now standard model of the density or number of parties found in political party

systems combines the insights of political sociologists such as Grumm (1958) and Lipset

and Rokkan (1967), who emphasized the role of social cleavages, and those following

Duverger (1954; Lijphart, 1990; Riker, 1982) highlighting the role of electoral institu-

tions, especially district magnitude (M) or the number of legislators elected from a

district (Cox, 1997; Neto and Cox, 1997; Ordeshook and Shvetsova, 1994; Powell,

1982; Taagepera and Grofman, 1985; Taagepera and Shugart, 1993). Despite this very

real success, there are several reasons to be less than fully satisfied with the standard

model. It does not, for example, account for the numbers of political parties found in

political systems with high district magnitudes or values of M. The standard model hides

away these troublesome cases under the blanket of an arbitrary logged specification.

When the logged measures of district magnitude are unpacked using a polynomial

specification, number of parties does not respond to changes in the value of district

magnitude after M values of 15 or so (Lowery et al., 2010), a value considerably lower

than the extreme values of M found in Israel (120) and the Netherlands (150) and, indeed,

to about a fifth of the cases used in the usual cross-sectional tests of the standard model

(Cox, 1997; Neto and Cox, 1997; Ordeshook and Shvetsova, 1994). This suggests that

the model is at least incomplete in terms of the party systems it can account for and that

something other than electoral institutions and/or cleavages may matter.

After first describing the standard model and its difficulties with high values of

district magnitude, we develop an alternative explanation for these cases emphasizing

slack in the issue agenda available to political parties with which to construct viable

electoral identities or niches with which to mobilize myopic, inattentive, and cognitively

limited voters. We then test this alternative using time series data on the size of the public

policy agenda and the political party system in the Netherlands, the most extreme case of

a large M given its single national district. We also assess the comparative importance of

the variables cited in the standard model of party density: social cleavages and electoral

institutions. We conclude by discussing how our model contributes to the broader under-

standing of party system density and how these findings may even have implications for

cases with low district magnitude.

The standard model and it limits

The standard model of the density of party systems combining the insights of work

by political sociologists emphasizing long-standing social cleavages and those of

2 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

institutionalists emphasizing the role of political institutions, especially district

magnitude, is perhaps best represented by the powerful analysis of Cox (1997). But his

model clearly emphasizes the role of district magnitude. Indeed, he argues that strategic

voting on the part of individual citizens largely accounts for the strong relationship

between M and number of effective parties, noting that ‘instrumentally rational voters

eschew wasting their votes on hopeless candidacies, preferring instead to transfer their

support to some candidate with a serious chance of winning’ (Cox, 1997: 30). Still,

his specification – building on Powell (1982), Neto and Cox (1997), Taagepera and

Grofman (1985), and Ordeshook and Shvetsova (1994) – also includes number of

social cleavages in the model in interaction with district magnitude (Cox, 1997; Neto

and Cox, 1997; Ordeshook and Shvetsova, 1994) based on the expectation – at least

as old as Key’s work on party factions in Southern Politics (Canon, 1978) – that

‘political entrepreneurs’ have strong incentives ‘to base separate parties on these

cleavages’.

So, what is wrong about the standard model? One problem concerns the functional

form of the core district magnitude variable routinely used in cross-national tests of the

standard model. The district magnitude term is included in the model in either a decimal

or natural logged form suggesting that the relationship is curvilinear. Unfortunately, this

specification is treated as a measurement convenience and is virtually undefended

theoretically. The closest one can find to a theoretical argument is Ordeshook and

Shvetsova’s (1994: 106–7) summary of Sartori’s observation that:

although we might predict that single member districts imply two-party systems, and that,

say, 15-member districts might imply four or five parties, it is unreasonable to suppose that

120- or 150-member districts (Israel and the Netherlands) will generate 30 or 40 parties,

ceteris paribus. (1986: 67, n. 15)

We have no trouble with this observation per se. Any number of models might suggest

that the relationship between M and number of parties is curvilinear. But none of the

several extant versions of the standard model actually provides such an explanation.

Instead, these exceptions are buried deep in the logged values, where their theoretically

interesting, if exceptional, character is effectively hidden.

To better explore these issues, Lowery and colleagues (2010) re-estimated the

standard model using a polynomial specification to unpack the logged values of M.1

By unpacking the LogM term of the standard model, they determined that the relation-

ship between party density and district magnitude traces an S-shaped response function

that is initially flat, then convex, rising through a period of linear growth, then concave as

the growth of party density slows, and finally flat as there is no further growth in number

of parties as M increases. Such sigmoid functions are the hallmark of the growth of many

different kinds of populations of organizations discussed in the organizational ecology

literature. Organizational ecology examines the evolution of populations of organiza-

tions through processes of selection (Hannan and Freeman, 1977; Hannan et al.,

2007). The approach, as noted by Baum and Oliver (1996: 1378–9), emphasizes ‘how

competition for scarce common resources and mutualism based on complementary func-

tional differences’ affects organizational founding and failure rates. More to the point,

Lowery et al. 3

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

this suggests that the theory of party system density might be viewed as one part of a

more general theory of the evolution of populations of organizations.

More substantively, Lowery et al. (2010) found not one world of party system density,

but several. The first is Duverger’s (1954) world of single-member districts where an M

of 1 and elite calculations produce only two parties, unless regional of ethnic enclaves

otherwise provide a niche for specialists. The second is Cox’s (1997) twilight world

of low M in which party numbers are suppressed below what M alone might allow

because of strategic voting, the impact of which ebbs quickly. The third world – from

M 5 to M 15 or so – is the exclusive province of the Lijphart’s (1990) generalized

Mþ1 rule in which only M matters. And, finally, there is the world of cases with M

greater than 15 in which, given the flat response function, or terminal density depen-

dence in the language of organization ecology, neither district magnitude nor its interac-

tion with social cleavages seems to play a role in determining party system density.2

Although quite different, Lowery et al. (2010) note that these four worlds can be sub-

sumed under an organization ecology model in which different kinds of constraints oper-

ate under different environmental conditions. For low- and mid-range values of M, those

constraints certainly include the kinds of variables highlighted by the standard model:

district magnitude and its interaction with social cleavages. Attention to these variables

is fully consistent with organization ecology’s focus on the role of vital environmental

resources that set limits on the carrying capacities of political systems for political par-

ties. But it is some other resource yet to be identified – and not M – that produces density

dependence or the flat response function observed for the 20 percent or so of cases of M

greater than 15 examined in cross-national tests of the standard model. For an organiza-

tion ecology interpretation to be complete, then, it must do more than simply subsume

the standard model by accounting for what it explains since there would then be little

to distinguish the two models. Instead it must also explain what the standard model can-

not: party density in systems with M greater than 15. We address this problem by using

two key theoretical tools of organization ecology. The first is niche theory (Hannan et al.,

2007; Peli and Nooteboom, 1999), which provides us the tools needed to explain niche

formation in the presence of agenda slack. And, second, we use density dependence the-

ory to theoretically account for how intensifying competition for an issue identity attenu-

ates party density growth as issue agenda slack disappears when the system’s carrying

capacity for political parties is approximated (Hannan and Carroll, 1992; Lowery

et al., 2010).

Issue agendas and party vital rates

Why, then, does the density of party systems become unresponsive to changes in social

cleavages and/or district magnitude above M values of 15 or so? In organization ecology,

the key resources that organizations depend on are associated with demand for organiza-

tional services. Our intuition is that, at high values of M, these critical resources are

related to the issues that allow parties to differentiate themselves before their consumers:

voters. Are issues not able to be subsumed under social cleavages as highlighted in the

standard model? While some scholars claim that the nature and role of both cleavages

and issues within political systems are multi-faceted and dynamic, changing over time

4 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

as a function of the strategic position of parties and political entrepreneurs as well as

more exogenous changes in the environment (Daalder, 1984; Powell, 1986; Rohrschnei-

der, 1993), most scholars have emphasized the stability of social cleavages as distinctive

from more transient issues (Lipset and Rokkan, 1967). Indeed, in the extensive cross-

sectional analyses of party system density, cleavages are typically measured by counting

the number of ethnic, language, and/or religious groups in a nation, not more transient

issues such as those associated with housing, environmental, or tax policy. Still, as Cox

(1997) notes by citing Meisel (1974) and Jaensch (1983), with respect to even long-

standing ethnic, language, and/or religious cleavages:

not all social cleavages become politicized, and . . . even fewer become particized (i.e., made

into important lines of partisan division). Both processes – politicization and particization –

typically do not just happen; they require someone to push them along, someone with

resources who can compete against other politicians who may be attempting to prevent the

politicization of that particular cleavage, or to activate others instead. (Cox, 1997: 26)

Given this, we see issues as distinct from social cleavages and of more proximate or

immediate political import, although we will explore this issue empirically later.3

Our intuition that issues matter is, of course, not without precedent. There is a long

tradition within the discipline of examining the role of issues in determining and con-

straining the behaviour of political parties, including extensive research on the Dutch

party system on the role issues have played in the structuring of competition among par-

ties (Aarts et al., 1999) and the rise of new parties (Belanger and Aarts, 2006; Krouwel

and Lucardie, 2008). However, we wish to highlight here not the role of specific issues

for specific parties, as has been done, but the role of the entire issue agenda in determin-

ing how many parties can survive within a political system. Our model of party system

density in high M systems rests on five assumptions that, while not uncontroversial, are

well established in the discipline.

First we assume, far from the strategic voting interpretation of voters as clairvoyant

calculators of electoral odds, that they are poorly informed and generally inattentive to

politics (Converse, 1964; Delli Carpini and Keeter, 1996; Macdonald et al., 1998). In

summarizing this research domain, Luskin (2002: 282) noted that ‘at long last, and

despite some flickering dissent . . . there now seems to be a near-standard that by any-

thing approaching elite standards most citizens think and know jaw-droppingly little

about politics.’ In a world of myopic, inattentive, and cognitively-limited voters, it is not

easy to attract voters’ attention (Tavits, 2008). This will be especially true for new parties

and new issues, which explains why office seekers rely on established parties rather than

striking out on their own. Established parties offer a well-tried issue repertoire, a known

brand, and/or a reputation that helps to overcome the problem of inattentive voters in a

noisy environment (Aldrich, 1995: 49–56).

Second we assume that parties attract voters with issues (Schlesinger, 1984: 383). Or

as Downs (1957: 28) asserted ‘parties formulate policies in order to win elections rather

than win elections to formulate policies’. With this assumption, we hope to bring parties

as organizations into our explanation of the density of party systems, something that is

again consistent with an organization ecology approach to political phenomena, and also

Lowery et al. 5

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

something that reflects Schlesinger’s (1984: 374) objective of linking ‘together the

theory of party competition with that of party organization’ to construct a broader theory

of party politics.4 In Schlesinger’s view political parties are, first of all, organizations

that seek to survive by winning elections with voters doing the winnowing. More to the

point, they develop issues and the policies that address them as crucial products offered

in exchange for voters’ support (Schlesinger, 1984: 383). Finding attractive or at least

electorally-reliable issues is difficult if voters are paying limited or sporadic attention

to, and are poorly informed about, politics. And this problem is likely to be especially

severe in systems with electoral rules favouring many parties, each clamoring for atten-

tion against the competing messages of, potentially, quite similar rivals in their common

efforts to establish viable brands or issue reputations. Thus the numbers of parties actu-

ally competing for and winning office in such systems will likely lag behind the number

that could potentially survive on the basis of the mathematical opportunities provided by

electoral rules alone.

Third we assume that parties – or entrepreneurs of potential parties – operate in a ret-

rospective manner when calculating their chances of being successful in the future. That

is, they look to the prior election to see if there are issues or issue configurations that are

available upon which to develop an independent party identity. Such a retrospective

orientation is consistent with Aldrich’s (1995) analysis of the evolution of party systems

via emphasis on their role in solving problems, especially the problem of a lack of suc-

cess in prior elections. But prior elections also provide the best concrete evidence avail-

able to politicians about the opportunities and constraints they face in addressing their

electoral problems or ambitions. Among these is the volume of available issue space

upon which to build a viable party identity.

And, fourth, in what is perhaps the most unique addition of our consideration of the

role of issues, we assume that the volume of issue space available to political parties with

which to mobilize voters is restricted, even if it is dynamic. That is, while the cardinality

of the issue set facing a nation may ebb and flow over time (Baumgartner and Jones,

1993; Jones and Baumgartner, 2005), there are very real limits to the absolute size of the

issue agenda available to politicians in any one time and place to differentiate themselves

in the eyes of inattentive voters. Parties cannot just make issues up. Indeed, while party

entrepreneurs will look for new issues with which to mobilize voters (Belanger and

Aarts, 2006; Rabinowitz and Macdonald, 1989), the agenda space available for issues

is itself to a considerable degree exogenously determined, temporally sticky, and likely

limited in scope (Baumgartner and Jones, 1993; Jones and Baumgartner, 2005). Indeed,

it can be difficult for newcomers to displace older parties even when new issues appear

(Krouwel and Lucardie, 2008; Meguid, 2005; Rohrschneider, 1993) if the older parties

also are compelled to address an exogenously-driven issue agenda.

Given voter inattention and a limited supply of issues available to parties with which

to distinguish themselves, the density of the party system itself should act as an increas-

ingly powerful drag on further party formation even as the potential number of surviva-

ble parties given electoral rules increases. As the number of parties increases, the issue

space needed to establish a viable electoral identity is more likely to be already occupied.

Importantly, however, changes in the issue agenda can either relax or enhance this con-

straint. That is, while we assume that the issue agenda is limited, it is not fixed

6 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

(Baumgartner and Jones, 1993; Jones and Baumgartner, 2005). It is this variation – our

fifth assumption – that we can use to determine whether issue agenda change limits the

resource space available to politicians and parties, thereby imposing a flat response func-

tion or negative density dependence on the relation between M and numbers of political

parties. If the size of the issue agenda contracts, then new parties may find it even

tougher to form, and older parties may no longer be able to sustain an electorally-

viable profile with inattentive voters. There will not be sufficient distinctive issue sets

upon which to build a discernibly-unique issue identity. In contrast, if the issue agenda

expands to address a wider range of issues in an electorally-meaningful manner, then

new parties should find it easier to establish distinctive identities and older parties will

be less threatened by the competition. A broader, deeper policy agenda should allow a

greater number of parties to construct a distinctive issue set or configuration that might

be attractive to a sufficient set of voters to secure election.

These five assumptions provide a straightforward test implication. In the language of

the event history analysis that we use in the ultimate tests of the model, we hypothesize

that having fewer than expected political parties given the issue agenda space available

in one election will increase the hazard of new party births in the following election. We

also hypothesize that observing more parties in a political system than expected in a prior

election given the available issue space will increase the hazard rate of party deaths in the

following election.

Testing the expectation

We test these expectations below with time series data on the Dutch case with its extreme

district magnitude of 150. One important advantage of focusing on a single national

party system over time is that we can largely side step at least some of the conceptual

issues associated with distinguishing between long-standing ethnic, language, and/or

religious cleavages and the more short-term impacts of issues. That is, if cleavages rep-

resent more lasting divisions within a polity, focusing on a single polity will effectively

control for them, thereby allowing us to focus more squarely on the influence of issues.

However, we will later consider some additional empirical controls for one possible clea-

vage in the Dutch political system (religion) that might have undergone some meaning-

ful change over the time period we examine.

We proceed in the following way in the subsequent sections. We first describe the

party system in the Netherlands over the post-war period to highlight our two dependent

variables: the number of parties standing in parliamentary elections for the second cham-

ber (Tweede Kamer), which we use in the first analysis, and party births and deaths, used

in the second set of analyses. We then describe the niche model, the theoretical founda-

tion for our model, and how changes in the issue agenda can create space for a new party.

We describe the way in which we estimate issue agenda space. In the two following

empirical sections we first use time series analysis to generate the critical independent

variable for the final analysis, the amount of slack in the party system given the size

of the issue agenda, using historic data on the Dutch policy agenda; and last, we employ

event history analysis (EHA) to assess whether slack in the issue agenda-party system

does indeed influence political party births and deaths.

Lowery et al. 7

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

The Dutch party system

Although the Dutch are rarely viewed as an extreme people (although see Blokker et al.,

2008), their electoral system certainly qualifies for that label. With a single national dis-

trict for the 150 members of the Tweede Kamer, the Netherlands has a district magnitude

of 150. Among the countries typically included in analyses of party density, only Israel,

with a district magnitude of 120, competes. We focus on the Dutch case for several rea-

sons, but one of them is that it is the case that the standard model has the most difficulty

explaining. It is a hard case. With proportional representation, a low threshold for par-

ticipation in the Tweede Kamer, and plausibly retaining two distinct ideological dimen-

sions (i.e., left–right and religious–non-religious), the Dutch party system is well

populated (Andeweg and Irwin, 2005; Koole, 1996), although it is clearly not as crowded

as would be predicted by the generalized Duverger rule alone (Cox, 2006). Still, as seen

in Figure 1, fully 161 different political parties competed in Dutch parliamentary elec-

tions from 1946 through 2006. And with 72.67 percent (117) of these parties only com-

peting in one election, there have been a lot of party births and deaths. In fact, only two of

the 161 parties have competed and won seats in parliament over the entire period: the

social-democratic PvdA and the conservative protestant SGP.

All of the remaining parties were born, died, and/or merged since 1946 (Krouwel and

Lucardie, 2008; Mair, 1999). The latter group includes some notable cases, such as the

Christian-democratic CDA (Christian-Democratic Appeal), which was formed by the

merger of the three major Christian-democratic parties (KVP [Catholic People’s Party],

ARP [Anti-Revolutionary Party] and CHU [Christian-Historical Union]) in 1977 after the

near-collapse of traditional religious cleavages in the Netherlands provided less issue

space for separate denominational parties. While the CDA merger was a significant one,

given that the party or its component parties have been part of the government through

most of the post-war period, mergers are generally rather uncommon, involving only

117

13 113 1 2 1 1 4 2 0 0 2 1 0 0 0 1 2

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Number of elections

Num

ber

of p

artie

s

Figure 1. Number of elections in which 161 parties have competed, 1946–2006

8 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

5 percent or so of all of the parties. Somewhat more common is the founding of new parties

by splitting off from older ones, such as the somewhat more rightwing social-democratic

DS’70 (Democratic Socialists 1970) from the PvdA (Labour Party) in 1970. Most new par-

ties, however, did not arise via direct splits from existing parties, but from entrepreneurial

activity around new issues (Krouwel and Lucardie, 2008). Most of these new parties, as

might be expected, failed after one election. There are very few parties that repeatedly

tried and failed to secure representation in parliament or ultimately succeeded only after

repeated election campaigns. For example, the fundamentalist protestant Reformed Polit-

ical League (GPV) and the socialist SP (Socialist Party) gained seats in parliament only

after many attempts: only after four election attempts for the former and six for the latter.

Many of the successful parties are, of course, still quite small in terms of electoral

support and the numbers of seats they claim in the Tweede Kamer. But in the Dutch con-

text, they can still have considerable import. One or the other of two small parties – D66,

a party emphasizing democratic reform, and ChristenUnie, an economic left-leaning,

fundamentalist Christian party – have been part of the governing coalition for the last

quarter of a century. And, while never part of a governing coalition, it often seems that

Geert Wilders’ anti-immigrant PVV (Party for Freedom) is having a profound impact on

the Dutch policy agenda. More generally, several scholars have noted that the unusual

openness of the Netherlands to new parties has allowed them to bring new issues onto

the policy agenda, especially so when older, more established parties have faced elec-

toral difficulties (Krouwel and Lucardie, 2008; Mair, 1997, 1999).

In examining the pattern of party competition over time, it is also clear that the largely

cross-sectional analyses typical of work on the standard model may give a very false

sense of the stability of party systems in terms of the number of parties that compete.

As seen in the darker line of Figure 2, there has been plenty of variation in the numbers

0

5

10

15

20

25

30

52 56 59 63 67 71 72 77 81 82 86 89 94 98 02 03 06

Predicted parties Actual parties

Figure 2. Predicted and actual number of parties competing by election, 1952–2006

Lowery et al. 9

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

of parties contesting elections in the post-war period, ranging from a low of 10 in 1956 to

a high of 28 in 1971, which provides us another reason for focusing on the Dutch case.

That is, there is considerable variation in the numbers of parties competing in each elec-

tion, the dependent variable in the next step of this analysis. In contrast, Dutch electoral

rules and social cleavages, the explanatory variables of the standard model, are rather

stable over most of this period. Thus, to explain variation in the number of Dutch parties

competing over time, we will likely have to look elsewhere.

Niche model

We argue that it is issue agenda slack that facilitates niche formation and maintenance

for new parties. Before estimating this slack, we make more explicit the conceptual

framework underlying our operationalization. Analytically speaking, the realized niche

of an organization is the multidimensional resource space domain where the organization

sustains itself in the presence of competitors (Hannnan and Freeman, 1977). The

resource space can be the political space spun by issue-axes along which electoral pre-

ferences distribute (Downs, 1957; Rabinowitz and Macdonald, 1989). The focal scarce

resource for organizations is their audience’s demand for services. In our context, this is

electoral demand for political representation on issues of concern to voters. Potential vot-

ers are thus resource carriers in this representation; their distribution in the issue space

tells us where parties may locate electorally viable niches.

Agenda slack refers to the presence of unoccupied, but potentially accessible, spatial

positions in electoral demand space. When spanning their niches, organizations address

the audience members with their offering (Hannan et al., 2007; Peli and van Witteloos-

tuijn, 2009). Normally, niches heavily overlap at mainstream positions, oftentimes leav-

ing the less lucrative peripheries to be occupied by specialized players. Moreover, the

presence of agenda slack also depends on the structural changes of the issue space. The

space can expand or contract along existing dimensions. For example, small parties may

find niches by extending issue-axes towards radical positions. But the space can also

expand as the result of the emergence of new issues (Peli and Nooteboom, 1999). Issue

dimension proliferation is strongly hindered, however, by institutional constraints and

electoral perception limits, as we mentioned earlier. The opposite process, space contrac-

tion by dimension removal, may well happen if issues are successfully addressed or, for

example, in times of severe political stress, such as wars, terrorist attacks, or economic

crises, the issue-focus narrows. Then, the issue space may collapse even to a single

dimension (‘are you with us or against us?’), causing severe niche crowding. Below,

we will apply this resource-based niche conceptualization to the context of the Dutch

political system and provide an empirical operationalization of agenda slack.

Before that, however, we must note that we would not expect that there is always a

simple one-to-one correspondence to between the issue slack generated by the emer-

gence of a new issue and the birth of a new party. Older parties might incorporate a new

issue or governments might try to preempt the formation of new parties by addressing it.

But parties can be highly constrained in addressing new issues, given their support for

positions on older issues and their existing constituencies. For example, the Democratic

Party in Michigan was historically closely tied to labour unions associated with the auto

10 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

industry in Detroit. As a result, the emergence of the ‘Democratic’ issue of the

environment in the 1970s posed real difficulties for the Democratic Party in Michigan.

Thus, parties may annex new issues, but they may also be highly constrained in

addressing new issues and new issue configurations by their existing positions on other

issues, something that provides an opportunity, but by no means a certainty, for new

parties to emerge.

Even more broadly, there may not be a one-to-one correspondence to a new issue and

the emergence or death of parties because what surely matters most is the overall

dimensionality of the issue agenda (Albright, 2010). As dimensionality increases with

the addition of a new issue, there are many more possible configurations of party

positions across all issue arrays. Thus the emergence of the immigration issue in the

Netherlands potentially created the space for both an anti-immigrant, economically right

wing party and an anti-immigrant, economically left wing party as well as pro-immigrant

version of both. Add then conservative and liberal social positions to these potential

party configurations and we now have many more possibilities. Indeed, there is a further

complication in that parties can vary in their emphases of the issues they address. Thus if

parties are more than single-issue causes, more issues simply provide for more combina-

tions and permutations across all of the issue space, thereby creating more opportunities

for parties. Conversely, contracting the agenda space reduces the opportunities for par-

ties to maintain a distinctive identity across all of the issues that they must address.

Estimating issue agenda slack

The next step is to develop a measure of the amount of issue agenda-party slack in the

Dutch political system. We do so by examining the residuals generated from regressing

an election year measure of the density of Dutch party system on an election year mea-

sure of the heterogeneity of the Dutch political agenda, controlling for other explanations

of the number of parties competing in a given election. The dependent variable is the

number of parties competing in each election, as observed in Figure 2. We also examined

as alternatives two fractionalization measures – both party vote and party seat share – as

is more common in the literature on party system density. But both alternatives were

essentially constants over the post-war period while number of parties, we have seen,

is highly variable.

Turning to the critical independent variable, we employ Breeman and colleagues’

(2009) Herfindahl index of the density of the Dutch policy agenda based on coding of

the government’s statements each year, as proclaimed on Prinsjesdag in the Troonrede

(Queen’s speech), constructed using a Dutch variant of the Baumgartner and Jones

(1993) policy agendas coding scheme using 19 major topics. The Queen’s speech – writ-

ten by the coalition government – reports the government’s achievements and major

events of the previous year as well as the government’s goals and policy decisions for

the year to come. These annual speeches can involve symbolic language as they address

hopes and concerns about the well-being of the country, but are also surprisingly sub-

stantive and specific in terms of attention to a variety of policy topics.5

In assessing the validity of this measure, two critical issues must be considered. The

first is the distinction between long-standing cleavages of the type considered in extant

Lowery et al. 11

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

cross-national analyses of party system density and the kinds of short-term issues that are

relevant for our theory of party systems. First, again, by focusing on a single national

case over time, we are plausibly controlling for many of the long-standing ethnic, lin-

guistic, and religious divisions that are the focus of the cross-national literature. And the

kinds of issues examined in the coding of the Queen’s speech – and in the policy agendas

literature more generally – do not speak to the kinds of long-standing ethnic, language,

and religious divisions found in work on cleavages. The coding categories included

issues associated with planning, public administration, foreign policy, foreign trade, sci-

ence and technology defence, commerce, housing, social security, justice and crime,

transport, energy, environment, education, labour, agriculture and fisheries, health,

immigration, and macro-economy. One can imagine, we suppose, that some long-

standing cleavage might be reflected in a few of these issues in some places at some

times, such as ethnic divisions in Iraq being reflected in foreign policy issues with

respect to Iran. But it is difficult to see how changes in attention to commerce or trans-

port issues would consistently reflect long-standing ethnic, language, or religious clea-

vages in a manner that would confound our attention to issues, again, especially so

since we examine a single polity over time.

Furthermore, attention to these several issues changes markedly over time, sometimes

dramatically so, something that would not be expected if they largely reflected long-

standing ethnic, linguistic, and religious cleavages. As is fully evident in the analysis

of the Queen speech measure by Breeman et al. (2009), labour, defence, and macro-

economic statements in Queen speech rise in the 1950s peak in the 1960s, and then stea-

dily decline thereafter. Statements on the environment and social security rose in the

1970s and 1980s and have since levelled off. And attention to several issue areas – com-

merce, science and technology, transportation, and planning – is extremely erratic, rising

and falling as specific policy problems emerge and disappear. Given this volatility and

variation therein across issues, it just does not seem plausible to view these issues as rep-

resenting cleavages in the sense of persistent divisions in society, a conceptualization

that is typical of the cross-national literature on party system density. In short, we believe

that our measures tap more short-term issues rather than reflecting long-term social

cleavages.

Turning to the second major issue with respect to the validity of our measure, we must

ask whether attention to issues is truly exogenous. That is, might the party composition

of the government itself determine what issues are discussed in the Queen’s speeches?

Consistent with the larger literature on policy agendas (Baumgartner and Jones, 1993;

Jones and Baumgartner, 2005), we believe that the structure of the policy agendas in

terms of the kinds and numbers of issues that are addressed is largely exogenously deter-

mined. That is, agendas seem to be influenced neither by which specific parties are mem-

bers of the governing coalition nor the demands of election year timing (Breeman et al.,

2009; see also Walgrave et al., 2006). While both variables almost certainly influence

how governments respond to issues, which issues governments must address seems

determined by the hard reality of real events. Furthermore, the numerous topics in the

Queen’s speeches normally – in most years and on most issues – do not constitute radical

departures from patterns of past attention to issues; rather, they cohere into positions

along the extant policy or issue dimensions that must be addressed by any party in

12 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

government. Thus the themes mentioned in the quite condensed Troonrede of about

2,000 to 3,000 words are normally those that require party policy positioning by the gov-

ernment of some sort even if the issue set is itself exogenously determined by events in

the real world.

This strong claim is supported by two prior analyses of the Dutch Queen’s speech

data. Breeman et al. (2009) examined the similarities between the number of statements

on each of the issue areas in year-by-year pairs over the whole time series. In a few

paired years, relatively low levels of correlation were observed. But these were clearly

externally driven, such as the greater attention given to foreign affairs issues when Indo-

nesia gained independence and in 1992 when the conflict in the former Yugoslavia and

the Maastricht treaty dominated the attention of the government. Overall, however, the

average correlation in the numbers of statements given to the several issues between two

years within a period during which a single government was in power (n¼ 38, r¼ 0.824)

hardly differed – substantively and statistically – from the comparable average correla-

tion for pairs of years in which the government and the coalition partners therein chan-

ged (n¼ 13, r¼ 0.809). Attention to issues – if not how governments in the Netherlands

address those issues – does not change as the parties in control of the government

changed.6

This last point is especially evident in Breeman et al.’s (2008) statement-by-statement

analysis of all of the Queen’s speech sentences on one of the issues in our measure of

attention that might be thought to be the most ideological and thus more closely related

to the party composition of government – public administration. They found that the

number of such statements increased linearly with the expansion of the welfare state.

During this linear increase, there are clear differences over time in how governments

speak about public administration, with a shift from attention to structuring government

services to providing more oversight. But both are largely responses to exogenous forces

associated with the growth and then aging of the Dutch welfare state. More to the point,

with only one exception (a few more sentences addressing the openness and transparency

of government when D66 was in the government), they found that the numbers of state-

ments addressing public administration year-to-year seemed unrelated to the party com-

position of the government. This does not mean, of course, that parties do not matter.

What did change with patterns of party control was the tone or how governments talked

about these issues. Centre-right governments were, as would be expected, far more crit-

ical in how they talked about public services than were center-left governments. But both

were compelled to address the issue of an expanding and then aging welfare state

whether they cared to or not. Thus we are confident that our measure of the issue agenda

facing Dutch governments over the post-war period is valid and is not confounded by the

party composition of those governments.

Dutch policy agendas vary over time. For our purpose, the most important character-

istic of the agenda is how crowded it is. The trend towards more crowding is evident in

Figure 3, which reports election year values by averaging the values over the prior three

years and the year of election of a Herfindahl index of policy attention, calculated by

summing the squared proportions of an item – in this case, the proportions of statements

in each speech addressing one of the 19 major policy topics. Thus a value of 1 would

indicate that attention was given to only one of the 19 policy topics and a value near

Lowery et al. 13

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

0 would indicate that attention was spread evenly across all of the issues. There is a rather

marked linear decline in the annual Herfindahl values reported in the figure. More recent

governments gave more even-handed attention to a wider array of policy issues. This

suggests that the policy agenda in the Netherlands has become much more crowded in

recent decades. For our purposes, we expect that this more crowded issue agenda has cre-

ated opportunities for parties to develop viable issue niches within the minds of other-

wise inattentive voters. Thus a decreasing concentration of topics in the speeches

indicates increasing issue space size, and so potential agenda slack. However, the domi-

nant pattern of a growing issue agenda crowding is not monotonic. When issue concen-

tration increases, there should be less issue slack for parties to employ in building

distinctive identities in the minds of voters. More importantly for our purpose, the tigh-

tening or relaxing of the issue agenda constraint should enable us to see if it is the nature

of the issue agenda itself that imposes density dependence on the responsiveness of party

system density to the usual suspects of district magnitude and social cleavages included

in cross-sectional tests of the standard model.

Our model also includes several control variables to address rival explanations, espe-

cially those associated with the standard model, of the variation in number of parties that

have competed in Dutch elections. The first two control variables address the electoral

institutions that lie at the heart of the standard model – changes in electoral institutions.

There have been only two significant changes in the electoral structure over this period.

In 1956, the electoral threshold for participating in parliament was in effect lowered. The

Dutch election law has no explicit threshold per se. Thus all a party needs to do to win a

seat in parliament is to secure enough votes for a single seat. Therefore, the expansion of

the Tweede Kamer from 100 to 150 seats, which was part of the 1956 constitutional

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

52 56 59 63 67 71 72 77 81 82 86 89 94 98 02 03 06

Figure 3. Herfinhdahl index of policy concentration in annual Queen speeches, 1952–2006

14 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

revision, represented in practice a change in the electoral threshold. Since 1956, a party

only needs 0.67 percent instead of 1.00 percent of the vote to obtain a single seat.

A dummy variable scored 1 after 1956 was added to the model to account for this change

in electoral legislation. Second, the Kieswet – or election law – was changed in 1994,

making it considerably more difficult for new parties to register for elections, something

that has been found in cross-national analysis to influence party entrance (Tavits, 2006).

A dummy variable scored 1 after 1994 was added to account for this change.

The second set of variables in the standard model address social cleavages, usually

measured by religious, ethnic, or language fractionalization. We have already dis-

cussed one way in which we distinguish these from issues, highlighting our focus

on a single case over time as a means to control for persistent cleavages, the nature

of the issues in our measure in distinction to the kinds of divisions addressed in the

literature on cleavages, and the volatility of attention to these issues in distinction to

the presumed persistence of long-standing cleavages. Further, it is not clear that such

cleavages should still matter in the Netherlands, even if we could draw a clear link

between them and the kinds of issues in our measure of issue crowding. While the

religious cleavage between Catholics and Protestants, along with the less clearly

defined secular pillar, once unarguably structured much of Dutch politics, the political

relevance of that cleavage has diminished to a very considerable degree since the

1960s (Andeweg and Irwin, 2005). More recently, however, the divide between

Muslims and others has been of political import. Still, this cleavage was arguably

placed on the agenda by Pim Fortuyn only during the 2002 election7 (Belanger and

Aarts, 2006; Pellikaan et al., 2007), which is near the end of our time series. We

believe that, through case selection and the nature of our issue measure, it taps issue

agenda diversity, not changes in long-standing cleavages. Still, with a cautious philo-

sophy that wearing suspenders and a belt is always safer than relying on only one or

the other, we also included in the estimating models additional controls for the reli-

gious based cleavages. Indeed, we measured election year social fractionalization with

several measures of religious-based cleavages.8 All produce essentially the same (null)

results. In the empirical results reported here, we opted for the simplest fractionaliza-

tion measure with the longest temporal consistency across the post-war period.9

A third set of control variables address government and coalition stability. We expect

that periods of government or coalition instability will encourage new parties to join the

electoral fray. Simply put, governments and coalitions that are stable for long periods

may discourage entrepreneurs from taking a chance on new parties. In contrast, instabil-

ity offers opportunities (Tavits, 2007). Government stability is operationalized as the

number of new cabinets that were formed during the last five years preceding an election.

The higher the value, the more frequently new cabinets have been installed by the Queen

due to an election or a cabinet crisis. Coalition stability is similarly measured, but

assesses the frequency of changes in the party composition of the governing coalition

over the last five years. Thus it does not count those cases in which the same coalition

was re-installed after an election or a cabinet crisis.

With 17 election observations, the time series model was estimated using Ordinary

Least Squares (OLS) regression after determining that there was little evidence of serial

correlation.10 In terms of collinearity, there was some evidence that the Queen’s speech

Lowery et al. 15

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

agenda variable was modestly related to the party application rules dummy (r ¼�0.535). Successive exclusion of variables, however, did not alter the results. And as

might be expected, there is a correlation of 0.670 between government and coalition sta-

bility. Again, the coefficients generated by the model did not change in any substantive

manner when only one or the other was included in the specification.

The results reported in the first column of Table 1 indicate that issue density, as

measured by the Herfindahl index of the proportions of attention paid to 19 broad

issues, is strongly related to the number of political parties competing in the 17

post-war elections we examine. Given the inverse coding of the Herfindahl index,

the negative sign of the estimate indicates that, as expected, numbers of parties

increase with issue agenda size. Further, none of the control variables generated esti-

mates that were discernibly different from zero. Neither the changes in the electoral

rules undertaken over this period nor religious fractionalization, the variables cited

in the standard model, influenced the number of parties competing during the post-

war election. Nor does government/cabinet stability seem to matter. We were not

surprised by these null results given that, with an M of 150 and little sustained

change in the values of the control variables, it seemed unlikely that the control

variables would matter.

Nevertheless, we should not over-interpret their lack of significance. As seen in

column two of Table 1, while none of the control variables produced statistically-

discernible estimates when the agenda variable was excluded from the model, the

R-square value of 0.268 is not close to zero. Further, as seen in column three, inclu-

sion of only the agenda variable in the model, while still producing a significant and

negative estimate, produces an R-square value (0.368) that is more than half of that

reported for the full model (0.516). So it seems likely that the control variables are

Table 1. OLS analysis of number of parties competing in elections, 1952–2006

Independent variable Dependent variable: no. of parties competing in election

Size of policy Queen speech �168.303 ** — �118.824 **agenda Herfindahl index (�2.267) (�2.514)Social Religion �88.764 82.244 —cleavages fractonalization (�0.471) (0.406)Electoral Electoral �7.239 3.487 —institutions threshold (�0.884) (0.445)

Party application �4.055 1.886 —rules (�1.024) (0.542)

Political Government �5.716 �2.688 —stablility stablility (�1.560) (0672)

Coaltion 4.517 3.641 —stability (1.572) (1.090)

Constant 114.859 �40.843 32.646R-square 0.516 0.368 0.283N 17 17 17

Note: Figures in parentheses are t-values.*p < 0.10; **p < 0.05; ***p < .0.01, two tailed tests.

16 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

adding something to the model even if we cannot identify precisely which is playing

a meaningful role. Simply put, with only 17 cases, even modest levels of collinearity

are likely precluding us disentangling the effects of the control variables. For this

reason, we use the results from the full model reported in column one of Table 1

to estimate how policy agenda size influences the density of the party system. But

whatever else is unclear in these results, we have strong support for our expectation

that issues and the way parties use them are important. Or rather, we think that it is

slack in the issue agenda space that either restrains or provides the niche space

needed to sustain the formation of new parties, making it more or less difficult for

established parties to maintain a viable identity in the face of voter inattention and

indifference. Our final test, then, must examine how issue agenda slack influences

the birth and deaths of parties. Our key independent variable is policy agenda slack,

the values of which are simply the residuals from the model reported in Table 1.

That is, Figure 2, in which we earlier reported (the darker line) the actual number

parties competing in each of the post-war elections, also reports using a lighter line

the predicted number of parties competing in these elections based on the estimates

of the first model reported in Table 1. The residuals or differences between the

actual and predicted values are reported in Figure 4, where positive values indicate

that there were fewer political parties competing than would be expected given the

volume of the issue agenda alone or that there was slack in the system available to

sustain more parties. Negative values indicate that, given the size of the policy

agenda, there remains little room for more parties and that existing parties may face

difficulties sustaining a viable electoral niche.

8

6

4

2

0

2

4

6

8

10

12

14

Slack in party system

52 56 59 63 67 71 72 77 81 82 86 89 94 98 02 03 06

Figure 4. Estimated issue party slack, 1952–2004

Lowery et al. 17

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

Event history analysis of party births and deaths

We now have all of the elements needed to test our hypotheses that political party births

and deaths are associated with changes in the slack available in the issue agenda-party

space using event history analysis (EHA). Our dependent variables are party birth or for-

mation and party death or disbanding.11 For each political party the election in which it

first competed provides our indicator of party birth. The election after the last election in

which it competed serves as our marker of party death. Importantly, although there are

relatively few mergers in our data, they require special attention. We treat mergers as

births of new organizations and the disbanding of their independent predecessors as party

deaths. This should pose no problem for party deaths. As noted earlier, the best example

of a merger is provided by the CDA. Given the fact that the new party formed largely

under the constraint of a rapid dissolution of the traditional denominational structure

of Dutch political and social life in the 1960s, the process that led to the disappearance

of the component parties is consistent with our emphasis on insufficient policy agenda

slack with which to sustain a viable electoral identity. Treating the merged CDA as a

party birth is more problematic, however, since the model suggests that births result from

slack in the policy agenda space, not through forced marriages among weakened party

organizations. Still, there are few such mergers, and their inclusion – given the underly-

ing logic of the hypothesis – should militate against our finding support for it rather than

making it easier.

The key independent variable of interest in the EHA models of party births and deaths

is the residual for the predicted number of parties in the prior election, as estimated with

the model presented in Table 1 and reported in Figure 4. We hypothesize that, when the

number of parties in the preceding election is less than we have predicted, other parties

will have more issue agenda space in which to form and, therefore, will have an

increased hazard of formation. When the number of parties in the prior election is fewer

than predicted given the issue space available, parties will not have to struggle so hard to

compete in a crowded issue environment and will have, therefore, a decreased hazard of

disbanding.

The issue slack residual variable is lagged from the prior election. In addition, the

specifications of both the birth and death models include several control variables. The

first is the re-centreed square of the lagged residual variable which is included to test for

non-linear effects of the slack variable. The second is the lagged value of the actual num-

ber of political parties competing in an election which is included to test whether simply

having more (or fewer) parties competing in the last election, irrespective of the issue

space available, may lead political entrepreneurs to suspend (or accelerate) efforts to

form a new party or leaders of older parties to consider (or put off the table) disbanding

it. A third control included in the model of political party deaths is the prior success of

the party as measured by the proportion of seats in the Tweede Kamer it won in the last

election. Parties with greater numbers of seats should have some protection from dis-

banding due to loss of all seats arising from simple random fluctuations in support (Pimm

et al., 1988; Raup, 1991).

We employ a simple Cox proportional hazards model with results presented as hazard

ratios. We begin observation in 1952, the first year for which we have residual values

18 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

from our earlier model. Parties are at risk of birth from that date until 2006, the last date

of our predicted values of the residual from the model reported in Table 1. Parties are at

risk of death beginning with the date of the election in which they first competed.

As seen in Table 2, our hypotheses about the influence of the lagged density of the

party system relative to available issue agenda space on party births and deaths are con-

firmed. The first two columns report the results for party births. The positive and highly

significant value (1.830, p < 0.001) for the lagged residual of the predicted number of

parties indicates that, when the number of political parties competing in the last election

is fewer than expected given available agenda space, the hazard of party births increases.

The corresponding basic coefficient is 0.606, which indicates that a one unit increase in

the residual increases the hazard of party births by 83 percent. The hazard ratio for the

squared version of the lagged value of the predicted number of parties relative to avail-

able agenda space is less than 1.00 (0.910), which indicates that the powerfully positive

nominal effect diminishes somewhat – 0.09 percent (1.00 � 0.910) – for each one unit

change in the standardized value of the residual or slack variable. Thus, while apparent

agenda space encourages party births, it does not do so in a simple linear manner. It

seems that a little breathing room in the policy agenda space quickly encourages new

parties to form. But as the pool is drained of new party aspirants, the rate of new party

formation then weakens somewhat. And last in terms of political party births, the simple

number of parties competing in the last election matters. The hazard ratio of 0.830 indi-

cates that simply having more parties competing in the prior election, irrespective of the

relationship between their numbers and available issue space, tends to suppress party

formation.12

Conversely, as seen in the last two columns of the table, when the number of parties in

a prior election is greater than predicted, and the party system is dense relative to the

issue agenda, parties have a greater hazard of disbanding. With a hazard ratio of

0.910 (p < 0.01) or less than 1.000, having fewer parties than is predicted by the size

of the issue agenda leads to a reduced risk, or less likelihood, of party death or

Table 2. Hazard of party births and deaths predicted by lagged residual values

Independent variable

Party births Party deaths

Hazard ratio Coefficients Hazard ratio Coefficients

Lagged residuals for predicted 1.830*** 0.606 0.910** �0.092Number of parties 0.24 0.131 0.02 0.990**Lagged residuals for predicted 0.910*** �0.096 0.032 �0.015Number of parties squared 0.02 0.025 0.01 0.006Lagged number of parties 0.830*** �0.19 1.01 0.009

0.04 0.045 0.02 0.021Lagged seats won in last election — — 0.5 �0.695

0.28 0.56Wald X 2 X 2 (df3) ¼ 24.86, p ¼ 0.000 X2 (df4) ¼ 22.13, p ¼ 0.000

Note: Standard errors in parentheses.*p < 0.05; **p < 0.01; ***p < 0.001.

Lowery et al. 19

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

disbanding. The chance of a party failing declines by 0.09 percent (1.000 � 0.910) for

each one unit change in the standardized value of the residual variable. The hazard ratio

for the squared value of the issue residual variable (0.990) indicates that this effect, too,

is curvilinear, if only modestly so. In this case, however, both hazard ratios are less than

1, suggesting that the risk of party failure increases at an even more rapid rate – an addi-

tional 0.01 percent (1.00� 0.990) for each standardized unit change in the squared slack

variable – as slack disappears in the issue space. Thus it seems that, while party births are

rather readily encouraged, the new parties then become hard to kill off. Neither of the

control variables in the party deaths model generated discernible estimates. Simply hav-

ing more (or fewer) parties competing in the last election did not influence the risk of

party death, nor did having won more seats in Parliament in the last election seem to pro-

vide any special protection against failure.

Conclusion

In their re-analysis of the data supporting the standard model of party system density

using a polynomial model, Lowery et al. (2010) found not one party system, but four.

These sketched a sigmoid response function of party density to district magnitude, the

hallmark of an organizational ecology process. Their organization ecology model sub-

sumes – albeit linking them in a more plausible manner and sharply bounding the oper-

ation of strategic voting – the first three of those systems: Duverger’s (1954) world of

single-member districts, Cox’s (1997) world of low M where party numbers are sup-

pressed below what district magnitude alone would suggest, and Lijphart’s (1990) world

of the generalized Mþ1 rule in which only M matters for systems with M levels of 5 to

15. But neither model could account for the density dependence or lack of responsive-

ness to M of cases with an M of 15 or more. In this analysis, we developed a model based

upon five plausible assumptions about voters, parties, and the issue agenda to explain

such cases. We tested the model with data from the hard case of the Netherlands, finding

strong support for it. Changes in the slack in the issue agenda-party system have a sig-

nificant impact on the vital rates – births and deaths – of political parties in the

Netherlands.

This suggests that it is the issue agenda itself, and the space it provides parties to build

independent identities, that limits the density of the party system when district magnitude

rises above 15 or so. In short, we have found strong support for a theoretically-grounded

account of the fourth world of party systems noted by Lowery et al. (2010). In combina-

tion with prior results re-examining cross-sectional tests of the standard model bearing

on the other party system worlds (Lowery et al., 2010), we believe that the organizational

ecology interpretation can unite them into a single, coherent account of the density of

party systems. Importantly, this interpretation does not so much replace the standard

model as subsuming it. The population ecology model allows us to both better interpret

earlier results and to now go beyond them to account for cases it could not except by the

measurement slight-of-hand of logging.

Still, while the findings presented here contribute to a better understanding of how the

density of party systems is determined more generally, does agenda size matter for less

extreme cases than those with a single national district? We examined the extreme case

20 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

of the Netherlands, after all, because it is a hard case for the standard model, not because

it was in any sense typical. But our results are relevant for more than only the extreme

cases of Israel and the Netherlands. Lowery et al. (2010) found that density dependence

in the responsiveness of party density to changes in district magnitude sets in when M

rises to 15 or so, including about 20 percent of the cases examined in prior cross-

sectional tests of the standard model. But do our findings have implications for the other

80 percent of cases?

We believe that they do in two ways. The first is another set of cases that the stan-

dard model has trouble accounting for (Cox, 1997: 159): new political systems, such

as those found in eastern and central Europe. In such systems, the issue agenda was

typically extraordinarily complex and party identities are typically weak (Tavits, 2006,

2007, 2008). Accordingly, such systems might be expected to support larger numbers

of parties than they might otherwise do given the Mþ1 rule. The second and broader

set of cases are the stable democracies in which the constraint of M suppresses the

number of parties below the level that might survive based on available issue agenda

space alone. That useful issue agenda space does not disappear, however, and will

surely attract the attention of party entrepreneurs as they seek the attention of inatten-

tive voters. Thus we think it likely that the slack in the policy or issue agenda will

still have a considerable impact on politics even in party systems in which M more

clearly determines how many parties can survive, but most likely by engendering mul-

tiple factions within parties.

As a final thought experiment, then, we can reconsider the opposite extreme of

party density – the one party systems of the Solid South studied by V. O. Key in

Southern Politics (1949). In accounting for variations in party factionalism found in

the southern states, Key highlighted their differences in economic and social hetero-

geneity. Arkansas produced no discernible factions and a quiet, limited policy agenda

because of few policy disagreements among its homogeneous white voters. The more

heterogeneous Florida, in contrast, produced both a more active policy agenda and a

complex form of friends-and-neighbours factionalism. In Key’s view, social heteroge-

neity raises new issues, which in turn provides a rich foundation for competing fac-

tions. While Canon (1978) has attempted to reinterpret Key’s analysis using the

kinds of election rule variables cited in the standard model, we think that Key’s orig-

inal analysis based on social and economic heterogeneity and issue agenda complexity

is far more plausible. It is certainly consistent with the results presented here. It could

well be that changes in levels of policy-issue slack in party systems with low and

moderate levels of M are expressed not via influencing party births and deaths, but

by influencing party faction births and deaths.

Notes

1. They employed the simple version of the standard model presented by Ordeshook and Shvet-

sova (1994: 11) – including district magnitude, ethnic heterogeneity, and their interaction –

because it is a bit simpler than Cox’s, yet produces quite similar results (Cox, 1997: 221).

However, they employed Cox’s data – 54 national cases (circa 1985) – because they provide

us many more observations.

Lowery et al. 21

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

2. Lowery et al. (2010) report that, even with extreme values of 120 and 150, the effective

number of parties found in the Netherlands and Israel are estimated with the polynomial model

to be about the same as those found in countries with M values greater than 15: Austria,

Bolivia, the Czech Republic, Finland, Italy, Luxemburg and Portugal.

3. Indeed, we address the issue of cleavages versus issues in the empirical analysis to follow both

through case selection and the use of control variables in the models we estimate. As we will

see with respect to case selection, we examine a single national system over time rather than

cross-sectional analysis of multiple nations at one point in time. If ethnic, language, and/or

religious cleavages are of a long-standing nature we will, therefore, effectively control for

their influence in our empirical analysis in favour of more short-term issue dynamics.

However, as will be seen, we will include controls in the estimating model for the religious

cleavage in Dutch politics, the one source of long-standing cleavage that plausibly pertains

to Dutch politics.

4. Indeed, we find it rather odd that, as a theory designed to explain the numbers of parties found

in political systems parties, organizations play almost no role in the standard model (Schle-

singer, 1984: 378). Rather, the standard model provides a theory about the interaction of voter

behaviour and electoral institutions with parties being all too often a mere ghostly presence

somewhere in the background.

5. They coded all Queen’s speeches between 1945 and 2007 at the level of the individual sentence

and quasi-sentence. If sentences made reference to more than one topic – enumerative statements

– they were coded at the quasi-sentence level. Purely ceremonial statements spoken at the begin-

ning and end of a Queen’s speech were not coded by content. They counted these sentences as part

of the calculation of the total length of the Queen’s speeches. The full dataset consists of 8,772

coded sentences and quasi-sentences, allocated to one of the 19 main topic categories and into

a more specific sub-topic category when applicable. Coding proved to be highly reliable, with

intercoder reliability scores above 0.90 in all cases and well above that in most cases.

6. For the most part, we believe that there is a direct relationship between issues developing in

the real world and government attention to them. However, the relationship could be indirect

as well. That is, a new party or even an older opposition party might be emphasizing a new

issue because they view it as important. The parties in the government may have no intrinsic

interest in the issue, but may still address it as a means of politically pre-empting the compe-

tition. In this alternative format, the exogenous influence of reality is a step removed from the

endogenous behaviour of the parties in government, but the effect would be the same as we

describe here.

7. The precise timing of the injection of this issue is debatable but, in any case, quite late in our

time series. For instance, Kriesi and colleagues (2008) and Bale (2003) suggest that the issue

was already salient as early as 1994, although there is little evidence that it influenced the

party agendas or manifestos until much later.

8. These are all based on Dutch statistical office data, differing only in the way they approach the

other religion category. In one measure, the other category is treated similar to the other reli-

gious groups as a distinct category. In the second, we removed the other category from our

tally, assuming that the other religious groups are very small and inconsequential for major

electoral trends. The third uses data for Muslims and Hindus (available since 1972) as a dis-

tinct category, and treats the remaining other religious groups as in the first operationalization.

The last operationalization is the same as the previous one except that we excluded the

22 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

remaining other religions. The time series of these four fractionalizations measures are

extremely highly correlated.

9. We did not examine the interaction of the electoral law changes and social cleavages given the

limited degrees of freedom and the fact that neither set of controls produced any evidence of

having an independent impact on number of parties when considered nominally.

10. Given the frequency of party deaths and the typically short life-spans of the involved parties, it

seems likely that only first order serial correlation would be a potential problem. But regres-

sing the residuals on their lagged values produced a t-value of only 0.03 (p ¼ 0.974) and an

R-square value below 0.01.

11. Some might argue that number of party births and deaths is not all that interesting. What

matters is how large they are and whether they are in government. Clearly, both sets of

variables are important, if for different reasons. For purposes of understanding party electoral

competition, the first matters a great deal, while the second is important when considering the

policy produced by government. For this reason, the extensive literature on party system

density (see Cox, 1997) typically looks at both the effective number of parties competing and

the effective number of parties in government. At the same time, the two are closely related

empirically. Indeed, Lowery et al. (2010) generated essentially identical results when using

the two different measures to test their polynomial version of the standard model. Despite this,

we would suggest that both of these standard measures based on party fractionalization scores

miss the point to some degree by confounding the size of the party – either competing for seats

or winning seats – with the number of parties – either competing for or winning seats – when

the theories guiding this research are squarely focused on the later. The theories speak to num-

ber of parties competing or in parliament, not number adjusted for size. Thus we think that our

attention to numbers in the prior section of this paper provides a more direct reflection of the

theories underlying this kind of research. And with respect to births and deaths, a party cannot

be represented in government if is not yet born or has died. These are critical demographic

events in the life of a political party that shape its opportunities to be in government.

12. Interestingly, this last finding indicates a density effect as yet undetected in the standard orga-

nization ecology literature. This standard density dependence interpretation claims that

resource competition blocks entry at high densities (Hannan and Carroll, 1992). By softening

competition, agenda slack should facilitate entry independently from incumbent party density.

The fact that high density combined with agenda slack still attenuates entry indicates an extra

density effect. In our context, limited electoral perception might work against density growth:

people are attentive to a limited number of party offerings, as they are attentive only to a lim-

ited issue agenda.

References

Aarts, Kees, Stuart Elaine Macdonald, and George Rabinowitz (1999) ‘Issues and Party Compe-

tition in the Netherlands’, Comparative Political Studies 32: 63–97.

Albright, Jeremy J. (2010) ‘The Multidimentional Nature of Party Competition’, Party Politics

16(6): 699–719.

Aldrich, John H. (1995) Why Parties? Chicago, IL: University of Chicago Press.

Andeweg, Rudy B. and Galen A. Irwin (2005) Governance and Politics of the Netherlands, 2nd

edn. Basingstoke: Palgrave Macmillan.

Lowery et al. 23

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

Bale, Tim (2003) ‘Cinderella and Her Ugly Sisters: The Mainstream and Extreme Right in

Europe’s Bipolarising Party System’, West European Politics 26: 67–90.

Baum, Joel A. C. and Christine Oliver (1996) ‘Toward and Institutional Ecology of Organization

Founding’, Academy of Management Journal 39(5): 1378–427.

Baumgartner, Frank R. and Bryan D. Jones (1993) Agendas and Instability in American Politics.

Chicago, IL: University of Chicago Press.

Belanger, Eric and Kees Aarts (2006) ‘Explaining the Rise of the LPF: Issues, Discontent, and the

2002 Dutch Election’, Acta Politica 41: 4–20.

Blokker, Jan, Jan Blokker, Jr., and Bas Blokker (2008) Nederland in Twaalf Moorden [Netherlands

in Twelve Murders]. Amsterdam: Uitgeverij Contact.

Breeman, Gerhard, David Lowery, Caelesta Poppelaars, Sandra Resodihardjo, Arco Timmermans,

and Jouke de Vries (2009) ‘Political Attention in a Coalition System: Analyzing Queen’s

Speeches in the Netherlands 1947–2006’, Acta Politica 44(1): 1–27.

Breeman, Gerard, Arco Timmermans, David Lowery, Caelesta Poppelaars, and Sandra Resodi-

hardjo (2008) ‘De Politieke Aandachtcyclus voor Openbaar Bestuur en Democratie’ [The

Cycle of Political Attention for Public Administration and Democracy], Beleid, Politiek en

Maatschappij 35(3): 172–89.

Canon, Bradley C. (1978) ‘Factionalism in the South: A Test of Theory and a Revisitation of V. O.

Key’, American Journal of Political Science 22(4): 833–48.

Converse, Philip E. (1964) ‘The Nature of Mass Belief Systems in Mass Public’, in David E Apter

(ed.) Ideology and Discontent, pp. 206–61. New York: Free Press.

Cox, Gary W. (1997) Making Votes Count. Cambridge: Cambridge University Press.

Cox, Gary W. (2006) ‘How Electoral Reform Might Affect the Number of Political Parties in The

Netherlands’, Acta Politica 41: 133–45.

Daalder, Hans (1984) ‘In Search of the Center of European Party Systems’, American Political

Science Review 78: 1.

Delli Carpini, Michael X. and Scott Keeter (1996) What Americans Know About Politics and why

it Matters. New Haven, CT: Yale University Press.

Downs, Anthony (1957) An Economic Theory of Democracy. New York: Harper and Row.

Duverger, Maurice (1954) Political Parties. Trans. Barbara North and Robert North. New York:

Wiley.

Grumm, J. G. (1958) ‘Theories of Electoral Systems’, Midwest Journal of Political Science 2: 357–76.

Hannan, Michael T. and Glenn R. Carroll (1992) Dynamics of Organizational Populations: Den-

sity, Competition, and Legitimation. Oxford: Oxford University Press.

Hannan, Michael T. and John Freeman (1977) ‘The Population Ecology of Organizations’, Amer-

ican Journal of Sociology 82(5): 929–64.

Hannan, Michael T., Laszlo Polos and Glenn R. Carroll (2007) Logics of Organization Theory:

Audiences, Codes, and Ecologies. Princeton, NJ: Princeton University Press.

Jaensch, Dean (1983) The Australian Party System. Sydney: Allen and Unwin.

Jones, Bryan D. and Frank R. Baumgartner (2005) The Politics of Attention. Chicago, IL: Univer-

sity of Chicago Press.

Key, V. O (1949) Southern Politics. New York: Alfred A. Knopf.

Koole, Ruud (1996) Politieke Partijen in Nederland: Ontstaan en Ontwikkeling van Partijen en

Partijstelsel [Political Parties in the Netherlands: Origin and Development of Parties and the

Party System]. Utrecht: Het Spectrum.

24 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

Kriesi, Hanspeter, Edgar Grande, Romain Lacht, and Martin Dolezal (2008) West European Pol-

itics in the Age of Globalization. Oxford: University Press.

Krouwel, Andre and Paul Lucardie (2008) ‘Waiting in the Wings: New Parties in the Netherlands’,

Acta Politica 43: 278–307.

Lijphart, Arend (1990) ‘The Political Consequences of Electoral Laws, 1945–85’, American Polit-

ical Science Review 84(2): 481–496.

Lipset, Seymour Martin and Stein Rokkan (1967) Party Systems and Voter Alignments. New York:

Free Press.

Lowery, David, Simon Otjes, Sergiu Gherghina, Arjen van Witteloosstuijn, Gabor Peli, and Holly

Brasher (2010) ‘Unpacking LogM: Toward a More General Theory of Party System Density’,

American Journal of Political Science 54(4): 921–35.

Luskin, Robert C. (2002) ‘From Denial to Extenuation (and Finally Beyond): Political Sophistica-

tion and Citizen Performance’, in James H. Kuklinski (ed.) Thinking about Political Psychol-

ogy, pp. 281–305. Cambridge: Cambridge University Press.

Macdonald, Stuart Elaine, George Rabinowitz, and Olaf Listhaug (1998) ‘On Attempting to

Rehabilitate the Proximity Model: Sometimes the Patient Just Can’t Be Helped’, Journal

of Politics 60: 653–90.

Mair, Peter (1997) Party System Change: Approaches and Interpretations. Oxford: Clarendon.

Mair, Peter (1999) ‘New Political Parties in Established Political Systems: How Successful are

They?’, in E. Beukel, K. K. Klausen, and P. E. Mouritzen (eds) Elites, Parties, and Democracy,

pp. 207–24. Odense: Odense University Press.

Meguid, Bonnie M. (2005) ‘Competition Between Unequals: The Role of Mainstream Party

Strategy in Niche Party Success’, American Political Science Review 99(3): 347–59.

Meisel, John (1974) Cleavages, Parties, and Values in Canada. Beverly Hills, CA: Sage Publications.

Neto, Octavio Amorim and Gary W. Cox (1997) ‘Electoral Institutions, Cleavage Structures, and

the Number of Parties’, American Journal of Political Science 4: 149–174.

Ordeshook, Peter C. and Olga V. Shvetsova (1994) ‘Ethnic Heterogeneity, District Magnitude, and

the Number of Parties’, American Journal of Political Science 38(1): 100–23.

Peli, Gabor and Bart Nooteboom (1999) ‘Market Partitioning and the Geometry of the Resource

Space’, American Journal of Sociology 104: 1132–53.

Peli, Gabor and Arjen van Witteloostuijn (2009) ‘Optimal Monopoly Market Area Spanning in

Multidimensional Commodity Spaces’, Managerial and Decision Economics 30: 1–14.

Pellikaan, H., S. L. de Lange and T. van der Meer (2007) ‘Fortuyn’s Legacy: Party System Change

in the Netherlands’, Comparative European Politics 5: 282–302.

Pimm, Stuart L. H. Lee Jones, and Jared Diamond (1988) ‘On the Risk of Extinction’, American

Naturalist 132(6): 757–8.

Powell, G. Bingham. 1982. Contemporary Democracies: Participation, Stability and Violence.

Cambridge: Harvard University Press.

Powell, G. Bingham (1986) ‘Extremist Parties and Political Turmoil: Two Puzzles’, American

Journal of Political Science 30(2): 357–78.

Rabinowitz, George and Stuart Elaine Macdonald (1989) ‘A Directional Theory of Voting’, Amer-

ican Political Science Review 83(1): 93–121.

Raup, David M. (1991) Extinction: Bad Luck or Bad Genes? New York: W. W. Norton.

Riker, William H. (1982) ‘The Two-Party System and Duverger’s Law: An Essay on the History of

Political Science’, American Political Science Review 76(4): 753–766.

Lowery et al. 25

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

Rohrschneider, Robert (1993) ‘New Party versus Old Left Realignments: Environmental

Attitudes, Party Policies, and Partisan Affiliations in Four West European Countries’, Journal

of Politics 55(3): 682–701.

Sartori, Giovanni (1986) ‘The Influence of Electoral Systems: Faulty Laws or Faulty Method?’, in

Bernard Grofman and Arend Lijphart (eds) Electoral Laws and Their Political Consequences,

pp. 43–68. New York: Agathon Press.

Schlesinger, Joseph A. (1984) ‘On the Theory of Party Organization’, Journal of Politics 46: 369–

400.

Taagepera, Rein and Bernard Grofman (1985) ‘Rethinking Duverger’s Law: Predicting the Effec-

tive Number of Parties in Plurality and PR systems – Parties Minus Issues Equals One’, Eur-

opean Journal of Political Research 13: 341–52.

Taagepera, Rein and Mathew Soberg Shugart (1993) ‘Predicting the Number of Parties: A Quan-

titative Model of Duverger’s Mechanical Effect’, American Political Science Review 87: 455–

64.

Tavits, Margrit (2006) ‘Party System Change: Testing a Model of New Party Entry’, Party Politics

12: 99–119.

Tavits, Margit (2007) ‘Party Systems in the Making: The Emergence and Success of New Parties

in Democracies’, British Journal of Political Science 38: 113–33.

Tavits, Margit (2008) ‘Policy Positions, Issue Importance, and Party Competition in New Democ-

racies’, Comparative Political Studies 41: 48–72.

Walgrave, S. F. Varone and P. Dumont (2006) ‘Policy With or Without Parties? A Comparative

Analysis of Policy Priorities and Policy Change in Belgium, 1991 to 2000’, Journal of Eur-

opean Public Policy 13: 1021–38.

Author Biographies

David Lowery is a Professor in the Department of Public Administration at the University of Lei-

den in the Netherlands. Dr. Lowery teaches and conducts research on the politics of interest rep-

resentation, state and local politics, research methods, and bureaucratic politics. His most recent

book is Organized Interests and American Government.

Arjen van Witteloostuijn is Research Professor of Economics and Management at the University

of Antwerp, Professor of Institutional Economics at the Utrecht University, and Professor of Orga-

nization and Strategy at Tilburg University. He holds degrees in business, economics and psychol-

ogy. He has published in such journals as the Academy of Management Journal, American

Sociological Review, Journal of Public Administration Research and Theory, and Strategic Man-

agement Journal.

Gabor Peli is an Associate Professor at the Utrecht School of Economics and part-time researcher

at the Antwerp Centre of Evolutionary Demography (ACED). His research interests include orga-

nization theory, the logical representation of social science arguments and East-European political

economic transition.

Holly Brasher is an Associate Professor in the Department of Government at the University of

Alabama at Birmingham. Her research interests include political parties and ideology in the US

Congress, interest groups and lobbying, and public opinion toward parties. She is the author of sev-

eral articles appearing in the American Journal of Political Science, Political Analysis, and Party

Politics as well as a forthcoming book on lobbying from CQ Press.

26 Party Politics

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from

Simon Otjes is a PhD research at the Department of Political Science of Leiden University. His

research interests include new political parties, party positioning and legislative behaviour. He

holds degrees in Political Science and Philosophy of the Social Sciences. He has recently pub-

lished in American Journal of Political Science and Acta Politica.

Sergiu Gherghina is PhD researcher at the Department of Political Science of Leiden University.

His primary research areas are political parties (party organizations) in new democracies, legisla-

tive behaviour, and democratization. He has recently published in the Journal of Legislative Stud-

ies and European Union Politics.

Lowery et al. 27

at Tilburg University on September 19, 2012ppq.sagepub.comDownloaded from


Recommended