Party membership in Europe

Abstract

Research on party membership development commonly reports figures aggregated to the country level and/or using only a few time-points. While these choices may be appropriate for certain research questions, they nevertheless hide major differences between parties and conceal short-term fluctuations. Additionally, they are inappropriate for studying individual party trajectories. This is necessary, however, to better describe and ultimately explain the phenomenon of membership decline. The article analyses in total 1653 observations across 47 parties in six western European countries between 1960 and 2010 to test hypotheses pertaining to individual party membership development. Using multilevel modelling and time-series analyses, the results show what aggregated data with few time-points cannot: membership decline is by far not a universal phenomenon. Additionally, membership decline appears to be part of a party’s life-cycle. The more consolidated parties are, the fewer members they have. Few differences between party families are observable.

Keywords

party membership decline time-series cross-sectional data multilevel modelling time-series analysis Europe

Political parties and their members mark the enduring link between civil society and the modern democratic state. This link can be weakened if fewer citizens engage in parties. Existing research indicates the prevalence of party membership decline. Scholars conclude that ‘there is scarcely any other indicator relating to mass politics in Europe that reveals such a strong and consistent trend as that which we now see with respect to the decline of party membership’ (van Biezen et al., 2012: 38).

However, most existing studies analyse membership data aggregated over all parties, only a few time-points, or a mix of both to evaluate the development of membership (see Delwit, 2011; Krouwel, 2012; Mair and van Biezen, 2001; Scarrow, 2000; Scarrow and Gezgor, 2010; van Biezen et al., 2012; Van Haute, 2011; Whiteley, 2011). This strategy has some drawbacks. Firstly, data aggregated on the country level have the disadvantage of potentially hiding major differences between parties of the same system. Secondly, analysing only a few time-points can conceal fluctuations in-between. Annual time-series for a larger number of individual parties can provide a more thorough analysis of western European parties’ allegedly increasing failure to hold on to members.

Employing a unique dataset, this article tests party-level hypotheses pertaining to membership development with individual parties as units of analysis. With this, the article does not address issues of when or why an individual decides to leave a party. It also does not study the parties’ capacity to retain loyal voters as members. Instead, it concentrates on an underdeveloped field of party politics research: party-level theories of membership decline.

The results show that membership decline is by far not a universal phenomenon of parties in western Europe. Additionally, membership decline appears to be an element of a party’s life-cycle: the more consolidated parties are, the fewer members they have. Small differences between party families are also observable.

The first section briefly reviews some of the latest studies on party membership development. In the second section hypotheses are derived from existing theories of party development. The third section introduces the data and describes the methods used, and the fourth section presents the results. A final section summarises the findings and highlights their implications.

Existing research on party membership development

Previous research on party membership can be mostly characterised by the use of aggregated data and/or few time-points. Consider, for example, the selection of recent studies summarised in Table 1. They have at least three commonalities. Firstly, most studies analyse membership figures on an aggregated level. Secondly, most report the data using only a few time-points even when spanning long periods of time in total. Excluding Delwit’s (2011) study, average intervals range between four and eight years. And thirdly, they all report a decline in party membership during the last decades. Exceptions can be found for new democracies and green parties (Delwit, 2011; van Biezen et al., 2012).

Table 1.

Summary of recent studies on party membership development.

Study	Number of time-points	Timespan	Unit of analysis
Scarrow (2000)	5	1950s–1990s	country
Mair and van Biezen (2001)	3	1977–2000	country
Scarrow and Gezgor (2010)	3	1987–2000	country
Delwit (2011)	a) 7 b) 14	a) 1946–2006 b) 1980–2010	a) country b) party
Whiteley (2011)	1	2004	country
Krouwel (2012)	13	1945–2010	party family
van Biezen et al. (2012)	4	1980–2008	country¹

Note: For each study, the longest time-frame and the largest number of time-points analysed is displayed.

Both the aggregation of figures to a higher level as well as a small number of time-points limit the scope of research. Aggregations to a higher level can conceal possible influences by extreme cases. For example, aggregated party membership for Sweden in 1990 yielded 1,119,071, which is equivalent to 17.5% of the electorate at that time (M/E ratio).² However, the social democratic party alone contributed with around 72% because of the party’s close ties with the trade union.

The caveats of aggregates are well-known and they hinder drawing valid conclusions for lower-level units, also known as ecological fallacy (Bryman, 2012: 323; King, 1997: 12ff). Aggregated party membership figures fall short of providing warranted conclusions for individual parties.

An equally problematic issue in studying membership development pertains to analysing only a small number of time-points. Such an analytical strategy takes only a snapshot of the data. This strategy is insensitive to changes and can at best report broader trends; outliers or short-term fluctuations are not captured. Consider, for example, the development of the aggregate M/E ratio in the Netherlands between 2000 and 2009: van Biezen et al. (2012: 45) report figures for 2000 (2.51%) and 2009 (2.48%) that show a marginal decline, but the increase to 2.56% in 2004 is not reflected.³ Considering all three values together over this period of nine years provides not only a more detailed picture of party membership development in the Netherlands but also shows that its respective trend has not been continuously downwards. Apparently, 2004 was special in some respects. Counter the country trend, the social democratic Labour Party (PvdA) and the communist Socialist Party (SP) gained members in that year.⁴

The number of required time-points depends on the research interest. Detailed data may not be necessary in order to map global trends, but as soon as the research question pertains to change over time, the intervals between measurements should ideally be as short as possible.

Party-level theories of membership development

Since the scholarly interest in party organisations experienced a revival during the 1980s, theories on party development evolved fast. Even though these theories do not exclusively focus on the development of party membership, they cover it implicitly. The life-cycle approach explains membership size by a party’s different life-phases or levels of institutionalisation, whereas the modernisation approach suggests a pattern of development according to the ideological family a party belongs to. Although each theory considers a different driving force behind a party’s development, they are complementary because each stresses a different party characteristic.

The life-cycle approach originates from the organisational literature and aims at predicting businesses’ development over time (see Hannan and Freeman, 1984; Whetten, 1987). Businesses experience different life-phases: emergence, establishment and decline. The political science literature also considers a party’s or a social movement’s development as following different stages (see Bolleyer and Bytzek, 2013; Harmel and Svåsand, 1993; Lipset and Rokkan, 1967; Müller-Rommel, 2002; Pedersen, 1982; Stein, 1973; van Biezen, 2005). Most famously, Pedersen (1982: 6f) distinguishes four stages of party development: declaration, authorisation, representation and relevance. While clear transition points from one stage to the next are difficult to identify, a party’s electoral and parliamentary strengths act as guidance for determining these transition points (Harmel and Svåsand, 1993: 71).

The main idea behind the life-cycle approach is that a (party) organisation is facing a set of problems and possibilities at each stage. It acts in response to each challenge and gains experience as it ages and moves along these stages. A party’s age can then be considered a proxy for the stage of development or level of party institutionalisation (see Mainwaring, 1998; van Biezen, 2005).⁵ This is not to suggest a deterministic relationship beyond the parties’ control. Every stage of party institutionalisation simply offers new opportunities, challenges and incentives for the party organisation to act.

As a major resource, membership plays an important role in a party’s life-cycle. At the beginning of a life-cycle, visibility and initial electoral support are crucial for party survival. A large and loyal membership base can assist in that. A party will exert greater efforts to increase its membership size at an early stage of the life-cycle. During later stages when the party is already established, members are less important and the parties’ efforts in attracting them subside. Accordingly, at an early stage of a party’s development, membership size should be increasing. However, as the party ages and institutionalises, fewer people enrol.

Hypothesis 1: The relationship between a party’s level of institutionalisation and its membership size is positive but flattening.

Complementary to that, the modernisation approach suggests that parties of certain party families are experiencing a disproportional decline in political support (see for instance Inglehart, 1977, 2007). Modernisation theories maintain that changes in citizens’ socio-economic environment since the 1950s have induced a shift in values from materialist to post-materialist (Dalton, 2006). Consequences of this shift can be seen in citizens’ voting behaviour but also in their modes of political participation (Dalton, 2006). Specifically, they should be observable in the electorate’s propensity to enrol in parties as a political long-term commitment.

Modernisation theories propose that citizens connect less to parties that do not capture the above new politics values, while new parties representing post-materialist values, such as ecological parties, are on the rise (see for example Müller-Rommel, 2002). The same goes for other niche values that were not captured by existing parties, such as right-wing beliefs. Right-wing beliefs also represent to some extent anti-party sentiments towards existing parties (see for example Mudde, 1996; Poguntke, 1996). This implies that under the modernisation approach, parties belonging to the ecological party family and right-wing party family can be expected to gain in party members.⁶

Parties of other party families, on the other hand, should be heavily impacted by membership decline. Social democratic parties, for instance, should be more strongly exposed to membership decline. Research into social democratic values and their development suggests that economic changes, particularly in the labour market, have lessened the social democratic parties’ appeal to the electorate (see Pontussen, 1995; Rueda, 2005). In addition, traditional welfare state demands have been generally fulfilled in the last decades (see Hirst, 1997; Merkel et al., 2008), making social democratic parties enter a state of identity crisis. These fundamental changes are expected to negatively affect social democratic parties’ membership size. Likewise, with increasing secularization and increasing atheism amongst European citizens over past decades (see for instance Halman and Draulans, 2006), Christian parties should have been disproportionally affected by membership decline.

Therefore, according to modernisation theories, patterns in the development of party membership size as an outlet for political participation should follow party-family belonging.

Hypothesis 2: Since the 1950s, ecological and right-wing parties have been increasing in membership size whereas Christian and social democratic parties have been decreasing.

Data and methods

Unlike previous studies on party membership development, the hypotheses are tested using annual membership data on individual parties between 1960 and 2010. All data originate from the parties.⁷ I am aware of the short-comings that the data include but consider them the best solution (see Delwit, 2011; Mair and van Biezen, 2001).⁸ Combinations of time-series cross-sectional and time-series methods are used to test the hypotheses.

Case selection and sample

Parties are selected if they have been in parliament during three consecutive electoral cycles in the considered period (1960–2010). This means that the tracking of parties’ membership sizes starts when parties have reached Pedersen’s (1982) stage of representation at the earliest but it is not limited thereafter.

The criteria enable approaching a detailed party-within-country sample that includes all parties of a system that once played an enduring role in politics. It follows parties systematically during their life-phases of representation and relevance. This also means that the dataset only includes success stories of parties that emerged and made it into parliament and not those that failed before entry or those that experienced a short lifespan. However, the dataset does include parties that vanished. The selection criteria were applied to six countries in Europe: Denmark, Germany, the Netherlands, Norway, Sweden and the United Kingdom.

The choice for selecting precisely these countries was motivated by patterns of party organisational continuity and by data availability for individual parties (for a similar argument see Katz and Mair, 1992). In all six countries, parties have been maintaining stable party organisations for a long period of time during which members have been playing an important role. Additionally, all parties in those countries have been equally exposed to the economic, political and social changes important for testing the modernisation approach. Finally, practical reasons have also played a role for the selection. Traditionally, Scandinavian countries and Germany have a stronger culture of government transparency and party regulation, respectively (see van Biezen, 2009). Both features foster data availability. Equally, the Netherlands and the United Kingdom have regulations, albeit only introduced later, requiring the publication of party data.

Due to these deliberate choices, the sample is not random but it represents the full population of parties within the boundaries of the selection criteria. This entails that a generalisation of the findings through assessing statistical significance does not apply. Yet, the similarities in the parties’ and countries’ characteristics allow the expectation of similar results for comparable parties; that is, parties also operating in advanced industrial democracies in Europe.

The resulting dataset comprises, in total, 1653 observations of party membership sizes for 47 parties between 1960 and 2010.⁹ The parties are listed in Table 2 with their respective year of foundation, party family and time coverage.¹⁰ The number and regularity of time-points in the dataset facilitates accurate analyses and comparability across parties, and this is not hindered by missing values.

Table 2.

Overview of parties included in the data set.

Country	Party name (national)	Abbreviation (national)	Party name (international)	Founded	Family	Time coverage
DK	Socialdemokraterne	A	Social Democrats	1871	soc	1960–2010
	Det Konservative Folkeparti	C	Conservative People’s Party	1915	con	1960–2010
	Socialistisk Folkeparti	F	Socialist People’s Party	1959	com	1960–2010
	Venstre	V	Left, Denmark’s Liberal Party	1910	lib	1960–2010
	Fremskridtspartiet	Z	Progress Party	1972	lib	1972–2004
	Det Radikale Venstre	B	Danish Social Liberal Left Party	1905	lib	1960–2010
	Centrum-Demokraterne	CD	Centre Democrats	1973	soc	1973–2007
	Kristendemokraterne	K	Christian Democrats	1970	chr	1970–2010
	Dansk Folkeparti	O	Danish People’s Party	1995	right	1995–2010
	Enhedslisten	Ø	Red–Green Alliance	1989	com	1989–2010
GE	Christlich Demokratische Union Deutschlands	CDU	Christian Democratic Union of Germany	1945	chr	1960–2010
	Sozialdemokratische Partei Deutschlands	SPD	Social Democratic Party of Germany	1875	soc	1960–2010
	Freie Demokratische Partei	FDP	Free Democratic Party	1948	lib	1960–2010
	Bündnis 90/Die Grünen	GRÜNE	Alliance ’90/The Greens	1977	eco	1977–2010
	Christlich-Soziale Union in Bayern	CSU	Christian Social Union of Bavaria	1945	chr	1960–2010
	Die Linke	LINKE	The Left	1989	com	1989–2010
NL	Volkspartij voor Vrijheid en Democratie	VVD	People’s Party for Freedom and Democracy	1948	lib	1960–2010
	Partij van de Arbeid	PvdA	Labour Party	1946	soc	1960–2010
	Christen-Democratisch Appèl	CDA	Christian Democratic Appeal	1980	chr	1980–2010
	Socialistische Partij	SP	Socialist Party	1971	com	1971–2010
	Politieke Partij Democraten 66	D66	Democrats 66	1966	soc	1966–2010
	GroenLinks	GL	Green Left	1989	eco	1989–2010
	ChristenUnie	CU	ChristianUnion	2001	chr	2001–2010
	Staatkundig Gereformeerde Partij	SGP	Reformed Political Party	1918	con	1960–2010
	Politieke Partij Radikalen	PPR	Political Party of Radicals	1968	soc	1968–1989
	Pacifistisch Socialistische Partij	PSP	Pacifist Socialist Party	1957	com	1960–1989
	Communistische Partij van Nederland	CPN	Communist Party of the Netherlands	1909	com	1960–1989
	Katholieke Volkspartij	KVP	Catholic People’s Party	1945	chr	1960–1980
	Anti-Revolutionaire Partij	ARP	Anti-Revolutionary Party	1879	chr	1960–1979
	Christelijk-Historische Unie	CHU	Christian Historical Union	1908	chr	1960–1979
NO	Arbeiderpartiet	A	Labour Party	1887	soc	1960–2010
	Fremskrittspartiet	Frp	Progress Party	1973	con	1973–2010
	Høyre	H	Conservative Party	1884	con	1960–2010
	Sosialistisk Venstreparti	SV	Socialist Left Party	1973	com	1973–2010
	Senterpartiet	Sp	Centre Party	1920	agr	1960–2010
	Kristelig Folkeparti	KrF	Christian Democratic Party	1933	chr	1960–2010
	Venstre	V	Liberal Party	1884	lib	1960–2010
SE	Sveriges Socialdemokratiska arbetarparti	S	Swedish Social Democratic Party	1889	soc	1960–2010
	Vänsterpartiet	V	Left Party	1917	com	1960–2010
	Centerpartiet	C	Centre Party	1913	agr	1960–2010
	Folkpartiet Liberalerna	FP	Liberal People’s Party	1934	lib	1960–2010
	Moderata samlingspartiet	M	Moderate Party	1904	con	1960–2010
	Kristdemokraterna	KD	Christian Democrats	1964	chr	1964–2010
	Miljöpartiet de Gröna	MP	Green Party	1981	eco	1981–2010
UK	Conservative and Unionist Party	Con		1834	con	1960–2010
	Labour Party	Lab		1900	soc	1960–2010
	Liberal Democrats	LibDem		1943	lib	1960–2010

DK: Denmark; GE: Germany; NL: the Netherlands; NO: Norway; SE: Sweden; UK: United Kingdom; soc: social democratic; con: conservative; com: communist; lib: liberal; chr: Christian; right: right-wing; eco: ecological; agr: agrarian.

The dataset also includes a number of new parties (N = 17) that, according to the life-cycle approach, should display positive growth during their early years.¹¹ All new parties passed the threshold of representation and thus are success stories. They allow for a test of party membership growth during and after this life-cycle phase.

Dependent variable

There are different ways of measuring party membership. It can be expressed as a party total (M), as a fraction of the number of voters (M/V), or in proportion to the total electorate (M/E). The goal of this article lies in comparing parties within and across countries, making the absolute number of members an unsuitable measure. The M/V ratio, on the other hand, is problematic because it can result in misleading conclusions. An increase can be either the result of more members with a steady party voter base or the consequence of fewer party voters but a steady membership base (Katz et al., 1992: 331; Scarrow, 2000: 87). The score does not reveal which development took place. More so, each of them would lead to very different conclusions about a party’s societal appeal. Therefore, it is preferable to use a relative measure of membership size that allows more unequivocal intra-country, cross-country and temporal comparisons in membership size, such as the most commonly used measure of the M/E ratio.

Figure 1 illustrates aggregate M/E ratios for all six countries between 1960 and 2010. It shows that the six party systems differ markedly in their proportions of party members. The three Scandinavian countries have almost consistently higher aggregate M/E ratios, compared to the other three countries. Sweden has had very high membership rates, partly due to the social democratic party’s strong ties to trade unions until the 1990s. The figure also confirms, once more, earlier findings of declining aggregated membership size.

Figure 1.

Aggregate member/electorate ratios per country.

However, once the trajectories are disaggregated and examined on the party-level, several parties countering the common downward trend can be identified. Figure 2 depicts, for example, Dutch parties that have been maintaining at least a rather stable membership size in the period of study. They include the Green Left (GL) with an increase of 0.2 percentage points between 1990 (0.14) and 2010 (0.16). The Democrats 66 (D66) and SP have had an impressive rise, increasing their M/E ratios by 0.12 (1966: 0.02; 2010: 0.14) and 0.24 (1992: 0.13; 2010: 0.37) percentage points, respectively. Additionally, the figure also shows the development of membership size for the Reformed Political Party (SGP), which has remained remarkably stable over time. Compared to the aggregate-level decline in the Netherlands, as displayed in Figure 1, these counter-trends are noteworthy.

Figure 2.

Party-level member/electorate ratios for selected parties in the Netherlands.

Equally, the downward trend in Danish membership figures on the aggregate-level veils some success stories. Figure 3 shows that the Socialist People’s Party (F) has had a rather turbulent history in terms of its membership size with a considerable fluctuation around an M/E ratio of 0.18. Yet, in most recent years its membership size has almost doubled (2010: 0.41). In addition, the newer Danish parties, Danish People’s Party (O) (1995: 0.01; 2010: 0.25) and Red–Green Alliance (Ø) (1992: 0.02; 2010: 0.11), have been increasing their membership base substantially since their foundation in the 1990s.

Figure 3.

Party-level member/electorate ratios for selected parties in Denmark.

Other countries also include success stories that are usually hidden by the common country trend. Figure 4 presents those parties outside of the Netherlands and Denmark that have been experiencing upward shifts. They include the German Alliance ’90/The Greens (Grüne), which doubled its M/E ratio since 1980 from 0.04 to 0.08, and the Swedish Green Party (MP) that started with a share of 0.03 in 1981 and reached its highest level in 2010 of 0.2. With its impressive rise the Norwegian Progress Party (Frp) (1973: 0.03; 2010: 0.61) also belongs to the group of parties countering the common trend of membership decline.

Figure 4.

Party-level member/electorate ratios for selected parties in other countries.

Since all of these parties still only hold a small fraction of the nation’s electorate as members, as exemplified in their rather low M/E ratios, their success stories are veiled by the larger parties’ losses and thus not observable on the aggregate level. Overall, this sub-group of parties (approximately 23% of the sample) counters the aggregate-level trend of decline and thus also validates the approach of party-level data analysis.

Independent variables

The variable of time is included as a predictor to each model to depict growth. It is clocked in five-year periods to avoid multicollinearity, and since the pattern of membership is not linear, a second-order polynomial is included as well.

To test the hypotheses, information on party age and party family is needed. The time-varying variable party age (age: mean = 54.63; standard deviation = 37.74) is calculated through the party’s year of foundation; it is standardised through z-scores. The dataset comprises 10 Christian parties, nine social democratic parties, eight liberal and eight communist parties, six conservative parties, three ecological parties, two agrarian parties and a single right-wing party (Table 2).

Multilevel modelling

Multilevel modelling appropriately takes into account the nested structure of the data, where each party’s observations between 1960 and 2010 fall on the level-1 part of the model.¹² Parties as a set of observations fall on the level-2 part. The distinction in levels of data allows the analysis of multiple sets of observations (here annual data of parties) simultaneously with time-variant and time-invariant predictors, resulting in party-specific growth curves (see Hox, 2010; Singer and Willett, 2003). It is the most appropriate way to accommodate the nested nature of the unique dataset employed.

Other analytical tools more commonly used in party politics research such as, for example, ordinary least square methods, fail to correctly estimate serially correlated error processes (Beck and Katz, 2011). However, in the present case a party’s value at time-point T₁ most likely depends on its previous value at T₀. Hence, errors are not independent or identically distributed. Multilevel modelling alone does not suffice to correct for autocorrelation (Hox, 2010; Singer and Willett, 2003). Here, time and polynomials of time are added as random effects to account for autocorrelation. This estimates the error covariance structure more appropriately and sufficiently for the current purpose (Hox, 2010: 103).

Except for party family, variables are included in the level-1 part of the model. According to Hypothesis 1, party age is supposed to have a negative effect on membership size when the entire sample is considered but a positive one after parties’ initial appearance in parliament and in their phase of representation. For that, only the first 15 years of new parties’ existence in parliament are analysed. The threshold of 15 years has been selected in accordance with the case-selection criterion of being a party that once played an enduring role in politics. Depending on the length of a nation’s electoral cycle, these correspond at most to 15 years.

The categorical variable party family is recoded into several dummies and added to the level-2 model as a cross-level interaction with time. Since the dataset only includes a single right-wing party (the Danish People’s Party; O), the analysis is restricted to parties of three party families. The model including ecological party family as a mediator is expected to show a growth pattern for membership size, while the models containing the social democratic and Christian dummies should both report a stronger pattern of decline. All models are estimated using maximum likelihood in order to also compare fixed effects across models (Singer and Willett, 2003).¹³

Time-series analysis

In the second step, individual parties’ trajectories are investigated separately. Percentage changes are computed that summarise the change in M/E ratios from one year to the next. For each party, i, the percentage change, pc, in its M/E ratio, ME, over two consecutive time-points is calculated with formula (1):

{p c}_{i, t = 2} = ({M E}_{i, t = 2} - {M E}_{i, t = 1}) / {M E}_{i, t = 2} * 100

This means that a time-series of 51 party-years (1960–2010) produces 50 percentage changes in total. The average of all percentage changes per party, over a time-series, marks the average annual growth, i.e. the growth rate. By way of averaging annual percentage changes per party age, or according to party family, Hypothesis 1 and Hypothesis 2 are tested, respectively.

For the time-series analysis, party age is recoded into three different age groups: ‘old’ refers to parties founded before the end of World War II (N = 22); ‘middle’ refers to parties founded after World War II (N = 18); and ‘young’ refers to parties founded after 1980 (N = 7). The decision for the last cut-off point was guided by a trade-off between only covering the youngest parties while simultaneously including a reasonable number of parties. A 30-year span prior to 2010 seemed reasonable to meet the trade-off. Restricting the group of young parties to a foundation year as late as 1990 would have excluded five of the seven parties.

The expectation is again that older parties are associated with a more negative growth rate. According to Hypothesis 2, ecological parties are expected to display positive growth rates whereas social democratic and Christian parties are supposed to have a negative growth rate.

Influential cases

For all of the multilevel models an analysis of influential groups is conducted.¹⁴ Influential groups – or outliers – are defined as cases displaying unusual combinations of values for the dependent and independent variable. In cases of regression outliers, the regression line is substantially influenced (Andersen, 2008). The Swedish Social Democratic Party (S) meets these characteristics and thus is excluded from the multilevel analysis. In addition, for similar reasons the time-series analyses exclude two individual observations: the Swedish S’s value in 1991 and Dutch D66’s value in 1974. In those years the parties’ percentage changes were extraordinarily high with 210 and –1926, respectively, because of heavy drops in the membership rate from one year to the next. The reasons for these decreases lie in exceptional contextual circumstances that warrant excluding these cases in order to single out the general trend.¹⁵

Results

The results are displayed in Table 3, showing the unconditional growth model (model 1) and the growth model with a second-order polynomial (model 2). According to the overall fit statistics, model 2 is the better fit for the data. The coefficients indicate that the included parties’ trajectories in membership size are on average U-shaped. The estimated turning point lies at period 12, which equals years 2015–2019¹⁶ and means that the estimated upward shifts in membership size emerge in the future. Supporting existing research, this development means that party-level data of M/E ratios show an average decline since the 1960s.

Table 3.

Multilevel models; member/electorate ratio as dependent variable.

			df = 1554	df = 1553	df = 1552	df = 184	df = 1553	df = 1553	df = 1553
			Growth models		Life-cycle		Party-family effect
			Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
Fixed effects
	intercept	γ₀₀	1.53 (0.29)	1.79 (0.36)	1.71 (0.30)	0.37 (0.21)	1.87 (0.35)	1.53 (0.36)	1.89 (0.42)
	TIME	γ₁₀	–0.14 (0.03)	–0.24 (0.3)	–0.22 (0.1)	–0.01 (0.02)	–0.24 (0.32)	–0.21 (0.09)	–0.01 (0.05)
	TIME^2	γ₂₀		0.01 (0.03)	0.01 (0.01)		0.01 (0.03)	0.01 (0.01)	0.01 (0.01)
	AGE	γ₃₀			–0.05 (0.04)	0.08 (0.08)
	ECO	γ₄₀					–1.68 (0.98)
	SOC	γ₅₀						1.06 (0.64)
	CHR	γ₆₀							–0.5 (0.52)
	ECO*TIME	γ₇₀					0.13 (0.09)
	SOC*TIME	γ₈₀						–0.08 (0.05)
	CHR*TIME	γ₉₀							0.06 (0.04)
Variance components
Level 1	Within-unit	σ² _e	0.12	0.05	0.06	0.00	0.05	0.05	0.05
Level 2	var intercept	σ² ₀	3.73	5.65	3.67	0.36	5.10	5.04	6.55
	var slope	σ² ₁	0.05	4.00	0.42	0.00	4.67	0.30	13.28
	cov intercept	σ₀₁	–0.88	–0.85	–0.87	–0.92	–0.84	–0.84	–0.86
Quadratic term	var slope	σ² ₂		0.03	0.00		0.04	0.00	0.11
	cov intercept	σ₀₂		0.82	0.79		–0.81	0.68	0.84
	cov linear slope	σ₁₂		–1.00	–0.99		–1.00	–0.97	–1.00
Goodness of fit
	logLik		–800	–253	–200	–283	–1871	–187	–282
	AIC		1614	526	421	552	3767	398	588
	BIC		1646	580	480	529	3832	462	653

Note: standard errors are in parentheses.

ECO: ecological; SOC: social democratic; CHR: Christian; AIC: Akaike information criterion; BIC: Bayesian information criterion.

Turning to the hypothesis tests, Hypothesis 1 proposed that party age as a proxy for party institutionalisation is negatively related to membership size. The results are displayed in model 3 and are supportive. With every increase in party age by one standard deviation (sd), the M/E ratio drops by 0.05. In addition, controlling for age also alters the coefficient on time compared to model 2. This further supports varying effects of party age. Model 4 reports the results for the subset of newly founded parties during the period of study.¹⁷ As expected, it shows a positive coefficient on party age, which suggests that new parties grow in membership size during their first life-cycle phase of representation.

To further substantiate these findings, percentage changes per age group are inspected. Figure 5 shows the development of percentage changes per age group. The group of old parties almost consistently yields, on average, negative percentage changes in membership ratios between 1960 and 2010. In comparison, parties founded after World War II but before 1980, labelled ‘middle’, had a mixed history. In the late 1960s they had high losses driven by the Danish SF’s loss of more than a third of its members, followed by several years of straight gains for the entire age group (mid-1970s). The 1980s and 1990s can be characterised again as declining years for middle-aged parties. But since the late 1990s, those parties have mostly been gaining again in their aggregate share of party members. The group of young parties also experienced larger fluctuations throughout, yet recorded almost exclusively positive values over the last 20 years.

Figure 5.

Parties’ percentage changes per age group.

The graphs support the hypothesis that party age predicts the development of membership size. Older parties yield almost consistently negative percentage changes, while the other age groups do remarkably better. Additionally, old parties yield a growth rate of –3.46 (sd = 2.03), middle-aged parties yield –0.43 (sd = 5.70) and young parties yield 0.60 (sd = 16.04). This means that parties founded before 1945 have annually lost, on average, 3.46% in their M/E ratio over the period 1960–2010. Comparison of means tests between these averages show that the difference between ‘old’ and ‘middle’ is statistically significant (p-value = 0.00). Other differences in group means were not significant. Nevertheless, the hypothesis pertaining to life-cycle effects of parties finds some support in the data. Older parties have significantly more negative growth rates than middle-aged parties. If party age is considered a proxy for party institutionalisation, these results suggest that the more consolidated parties are, the fewer members they have.¹⁸

Finally, the multilevel analyses of mediating effects of party-family belonging (Hypothesis 2) are displayed in models 5, 6 and 7 of Table 3. The combinations of coefficients across interaction models are not all as expected. They indicate that negative effects for ecological parties are flattening over time. It means that being an ecological party has had a negative effect on the M/E ratio that is rapidly diminishing over time. The results also show that negative effects for Christian parties are flattening over time. Both findings are not in line with the expectation. However, the model including the social democratic party dummy is more supportive of Hypothesis 2. Being a social democratic party has had a positive effect on the M/E ratio that is decreasing as time proceeds. Additionally, including the cross-level interaction improves the overall fit in the model on social democratic parties. Letting party-family belonging exert varying effects over time fits the data better in this case. Overall, testing party-family belonging as a mediator to the development of membership sizes produces mixed results.

Turning to the time-series analysis of Hypothesis 2, Figures 6 and 7 report the aggregated percentage change for two sets of party families: firstly, Christian, social democratic and ecological parties; and secondly, conservative, liberal and communist parties. For the three hypothesised party families, Figure 6 shows interesting patterns. Parties belonging to the ecological party family have experienced considerable fluctuation especially in the earlier years. The extreme peaks are a result of strong membership recruitment of the German Grüne and Swedish MP between 1981 and 1982, immediately followed by heavy losses in the MP between 1982 and 1983 when it lost 3300 of its 5800 members. Despite these fluctuations in the early years of ecological parties, it should be noted that this is the only party family which has mostly experienced positive percentage changes since the mid-1990s.

Figure 6.

Parties’ percentage changes per party-family belonging (Christian; Social Democratic; Ecological).

Figure 7.

Parties’ percentage changes per party-family belonging (Communist; Conservative; Liberal).

In comparison, the social democratic party family only had straight gains during the mid-1960s and 1970s. Thereafter, the average percentage changes are continuously negative, up to a value of –8 (1995). Only the very last change between 2009 and 2010 is positive again due to the British Labour Party’s gains in membership of around 0.08 percentage points in M/E ratios. Likewise, the graph shows that Christian parties have continuously experienced negative percentage changes since the 1980s. It means that the 10 Christian parties have also constantly lost more members than they could newly recruit since the 1980s. However, since then, other party families have also experienced such a decline. Figure 7 shows a similar pattern for parties belonging to liberal and conservative party families.

Inspecting the growth rates per party, as displayed in Table 4, confirms these results. Only ecological parties have a positive growth rate of 1.3% annually between 1981 and 2010, while all other party families yield negative values between 1960 and 2010. As far as the social democratic parties are concerned, their growth rates range roughly amongst those of the other party families, once the two outliers of S in 1991 and D66 in 1974 are discarded. Likewise, the Christian parties’ growth rates are not beyond the ordinary in comparison to other party families. Growth rates do not differ significantly. Overall, these results for Hypothesis 2 are again mixed.

Table 4.

Growth rates and standard deviations per party family.

	Chr	Soc	Lib	Com	Con	Eco
Growth rate	–2.15	–1.68	–2.97	–1.26	–3.15	1.30
Standard deviation	4.29	4.57	3.81	6.61	8.91	18.22

Chr: Christian; soc: social democratic; lib: liberal; com: communist; con: conservative; eco: ecological.

The multilevel models returned mixed results and the time-series analyses showed that several party families are negatively affected in their M/E ratios. Yet, results of both analytical steps indicate that ecological parties are gaining, on average, in membership, which is in line with Delwit’s (2011) findings. If they are related to the first hypothesis pertaining to age, an overlap emerges: ecological parties tend to be younger than parties of other party families. Two out of the three ecological parties covered by the data here belong to the group of ‘young’ parties founded after 1980.

Summary and conclusion

This article tested party-level theories pertaining to party membership development with 47 individual parties from six European countries between 1960 and 2010. While previous aggregate-level findings of an overall decline were validated once more, around 23% of the studied parties could be identified as not having experienced membership decline during the period of study. It appears that there are more success stories than aggregated data could possibly show, considering the large losses of other parties.

The analysis also revealed some patterns in the decline of membership size. Firstly, the results suggest that a party’s age, as a proxy for party institutionalisation, is a determinant of membership decline. Parties founded before 1945 are, on average, more affected in their membership growth rate than parties founded between 1945 and 1980. The results provide mild support for the idea that the more consolidated parties are, the fewer members they hold. This also makes sense because more established parties require fewer grassroots members to spread the message, compared to new parties without an established network. As a party consolidates, it faces different opportunities and challenges and it becomes less dependent on a large membership size.

Secondly, the results indicate that party-family belonging is not a strong factor in explaining membership decline. While some party families appear to be more heavily affected by membership decline than others, few clear-cut patterns were detected. It appears that parties of different ideologies are similarly subjected to membership decline. It means that modernisation theories might be able to explain shifts in citizens’ voting behaviour over the past decades and the overall downward trend in traditional forms of political participation. However, they do not seem very fruitful for explaining differences in membership decline between parties. At the same time, it is noteworthy that the ecological party family stands out as the only group of parties that is growing in membership size.

Overall, these results indicate that party membership decline is a phenomenon that requires further research on the individual party level. In spite of a general decline, there are several exceptions that necessitate more detailed analyses in order to comprehend the underlying causes. For some parties we already know what has caused a steep decline in membership rates. Some parties’ formal disconnection from auxiliary organisations reduced their membership base markedly, as happened in the case of the Swedish S and Norwegian Labour Party (A). It is also known that parties in the UK have changed their reporting practices, which may have resulted in lower membership rates (Webb, 1992). However, beyond these individual stories and one-off anomalies, the preceding analyses call for more research on structural explanations. This is especially the case in light of the number of counter-examples found. The findings should also be supplemented with party data from a wider range of countries, a major shortcoming inherent to the preceding analysis. Nonetheless, this article provides another step towards understanding and above all explaining party membership decline in Europe.

Footnotes

Acknowledgements

An earlier version of this article was presented at the seminar ‘Innovation of Governance’ held at the University of Twente. I would like to thank all participants, but in particular Kees Aarts, Jacques Thomassen and Sedef Turper for their valuable advice. I am also grateful to the anonymous reviewers for their recommendations.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

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

Notes

References

Andersen

(2008) Modern Methods for Robust Regression. London: Sage Publications Ltd.

Bates

Maechler

Bolker

Walker

(2013) Package ‘lme4’: Linear mixed-effects models using Eigen and S4. 1.0-5; 2013-10-25 ed.

Beck

Katz

(2011) Modeling dynamics in time-series–cross-section political economy data. Annual Review of Political Science 14 (1): 331–352.

Bille

(1997) Partier i forandring: en analyse af danske partiorganisationers udvikling 1960-1995. Odense: Odense Universitetsforlag.

Bolleyer

Bytzek

(2013) Origins of party formation and new party success in advanced democracies. European Journal of Political Research 52 (6): 773–796.

Bryman

(2012) Social Research Methods. Oxford: Oxford University Press.

Bundeswahlleiter (2012) Bundestagswahlen [Online]. Wiesbaden. Available at: http://www.bundeswahlleiter.de/de/ (Accessed 05 October 2012).

Compendium voor Politiek en Samenleving (2013) Stemgerechtigden en opkomst [Online]. Leiden. Available at: http://www.politiekcompendium.nl (Accessed 13 February 2013).

Deutscher Bundestag (2012) Rechenschaftsberichte [Online]. Berlin. Available at: http://www.bundestag.de/bundestag/parteienfinanzierung/rechenschaftsberichte/ (Accessed 05 October 2012)

10.

Dalton

(2006) Citizen Politics: Public Opinion and Political Parties in Advanced Industrial Democracies. Washington: CQ Press.

11.

Delwit

(2011) Still in decline? Party membership in Europe. In: Van Haute

(ed) Party Membership in Europe: Exploration into the anthills of party politics. Bruxelles: Université de Bruxelles, pp.25–42.

12.

Documentation Centre Dutch Political Parties (2012) Party membership figures [Online]. Groningen. Available at: http://dnpp.ub.rug.nl/dnpp/ (Accessed 19 April 2012).

13.

Döring and Manow (2013) Parliament and government composition database (ParlGov) [Online]. Bremen. Available at: http://www.parlgov.org (Accessed 13 February 2013).

14.

Folketinget (2012) Demokrati [Online]. Copenhagen. Available at: http://www.ft.dk. (Accessed 17 September 2012).

15.

Halman

Draulans

(2006) How secular is Europe? The British Journal of Sociology 57 (2): 263–288.

16.

Hannan

Freeman

(1984) Structural inertia and organizational change. American Sociological Review 49 (2): 149–164.

17.

Harmel

Svåsand

(1993) Party leadership and party institutionalization: Three phases of development. Western European Politics 16 (2): 67–88.

18.

Hirst

(1997) From Statism to Pluralism. Abingdon: Routledge.

19.

Hox

(2010) Multilevel Analysis: Techniques and Applications. New York: Routledge.

20.

Inglehart

(1977) The Silent Revolution: Changing Values and Political Styles among Western Democracies. Princeton: Princeton University Press.

21.

Inglehart

(2007) Postmaterialist values and the shift from survival to self-expression values. In: Dalton

Klingemann

(eds) The Oxford Handbook of Political Behavior. Oxford: Oxford University Press, pp.223–239.

22.

Institut for Kultur og Samfund (2012) Danmarks Historien [Online]. Aarhus. Available at: http://danmarkshistorien.dk/forside/ (Accessed 17 September 2012).

23.

Katz

Mair

(eds) (1992) Party Organizations: A Data Handbook on Party Organizations in Western Democracies, 1960-1990. London: SAGE.

24.

Katz

Mair

Bardi

Bille

Deschouwer

Farrell

Koole

Morlino

Müller

Pierre

Poguntke

Sundberg

Svasand

Velde

Webb

Widfeldt

(1992) The membership of political parties in European democracies, 1960-1990. European Journal of Political Research 22(3): 329–345.

25.

King

(1997) A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data. Princeton: Princeton University Press.

26.

Kosiara-Pedersen

(2013) Organisational change within Danish parties 1995-2002. CVAP Working Paper Series 2013/2. Copenhagen: Centre for Voting and Parties, University of Copenhagen.

27.

Krouwel

(2012) Party Transformations in European Democracies. New York: State University of New York Press.

28.

Lipset

Rokkan

(1967) Party Systems and Voter Alignments: Cross-National Perspectives. New York: The Free Press.

29.

Maas

CJM

Hox

(2004) The influence of violations of assumptions on multilevel parameter estimates and their standard errors. Computational Statistics & Data Analysis 46 (3): 427–440.

30.

Mainwaring

(1998) Party systems in the third wave. Journal of Democracy 9 (3): 67–81.

31.

Mair

van Biezen

(2001) Party membership in twenty European democracies, 1980-2000. Party Politics 7 (1): 5–21.

32.

Merkel

Petring

Henkes

Egle

(2008) Social Democracy in Power: The Capacity to Reform. Oxon: Routledge.

33.

Mudde

(1996) The paradox of the anti-party party: Insights from the extreme right. Party Politics 2 (2): 265–276.

34.

Müller-Rommel

(2002) The lifespan and the political performance of green parties in western Europe. Environmental Politics 11 (1): 1–16.

35.

Nordsieck

(2013) Parties and Elections in Europe [Online]. Düsseldorf. Available at: http://www.parties-and-elections.eu/index.html (Accessed 13 February 2013).

36.

Nieuwenhuis

Grotenhuis

Pelzer

(2012) influence.ME: Tools for detecting influential data in mixed effects models. The R Journal 4(2): 38–47.

37.

Pedersen

(1982) Towards a new typology of party lifespans and minor parties. Scandinavian Political Studies 5 (1): 1–16.

38.

Petersson

(2005) De politiska partiernas medlemsutveckling. Rapport till SNS Demokratiråd. Stockholm: Studienförbundet Näringsliv och Samhälle.

39.

Poguntke

(1996) Anti-party sentiment – conceptual thoughts and empirical evidence: Explorations into a minefield. European Journal of Political Research 29 (3): 319–344.

40.

Pontussen

(1995) Explaining the decline of European social democracy: The role of structural economic change. World Politics 47 (4): 495–533.

41.

Rallings

Thrasher

(eds) (2009) British Electoral Facts 1832-2006. London: Total Politics.

42.

Rueda

(2005) Insider–outsider politics in industrialized democracies: The challenge to social democratic parties. American Political Science Review 99 (1): 61–74.

43.

Scarrow

(2000) Parties without members? Party organization in a changing electoral environment. In: Dalton

Wattenberg

(eds) Parties without Partisans: Political Change in Advanced Industrial Democracies. Oxford: Oxford University Press, pp.79–101.

44.

Scarrow

Gezgor

(2010) Declining memberships, changing members? European political party members in a new era. Party Politics 16 (6): 823–843.

45.

Singer

Willett

(2003) Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. Oxford: Oxford University Press.

46.

Statistics Sweden (2013) Population Data [Online]. Stockholm. Available at: http://www.scb.se (Accessed 13 February 2013).

47.

Statistics Norway (2012) Befolkning [Online]. Oslo. Available at: http://www.ssb.no (Accessed 17 September 2012).

48.

Stein

(1973) The Dynamics of Right-Wing Protest: A Political Analysis of Social Credit in Quebec. Toronto: University of Toronto Press.

49.

van Biezen

(2005) On the theory and practice of party formation and adaptation in new democracies. European Journal of Political Research 44 (1): 147–174.

50.

van Biezen

(2009) Party Law in Modern Europe [Online]. Leiden. Available at: http://www.partylaw.leidenuniv.nl/ (Accessed 29 October 2012).

51.

van Biezen

Mair

Poguntke

(2012) Going, going, gone…? The decline of party membership in contemporary Europe. European Journal of Political Research 51 (1): 24–56.

52.

Van Haute

(2011) Party membership: An understudied mode of political participation. In: Van Haute

(ed) Party Membership in Europe: Explorations into the Anthills of Party Politics. Bruxelles: Université de Bruxelles, pp.7–22.

53.

Webb

(1992) The United Kingdom. In: Katz

Mair

(eds) Party Organizations: A Data Handbook on Party Organizations in Western Democracies, 1960-1990. London: SAGE, pp.837–870.

54.

Whetten

(1987) Organizational growth and decline processes. Annual Review of Sociology 13 (1): 335–358.

55.

Whiteley

(2011) Is the party over? The decline of party activism and membership across the democratic world. Party Politics 17 (1): 21–44.

56.

Widfeldt

(1997) Linking Parties with People? Party Membership in Sweden 1960-1997. Göteborg: Göteborg University.