Abstract
This article examines the relationship between party regulation, trust in political parties and partisanship in 24 European democracies in 2010. It tests two rival hypotheses, one suggesting that the regulation of political parties improves party support among electorates, and the other arguing the opposite case that regulation weakens support for parties. These hypotheses are tested using a multi-level modelling strategy that controls for a number of variables that might account for trust in parties and partisanship at the individual level. The evidence suggests that heavy regulation of political parties is associated with low levels of trust in parties and fewer partisans in these countries. This finding persists even in the presence of a variety of control variables. The analysis uses cross-sectional data and so definitive causal relationships cannot be fully identified, but the data implies that regulation by the state in the context of a severe economic recession could be having the effect of stifling support for political parties and weakening civil society.
Introduction
There are two prominent findings from comparative research into political parties in the advanced industrial democracies. The first is that parties are growing ever closer to the state, and are becoming, in effect, part of a state-sponsored cartel (Detterbeck, 2005; Katz and Mair, 1995). The second is that, with some exceptions, parties are losing their voluntary organisations and fewer party members and activists are found in these democracies over time (Katz et al., 1992; Mair and Van Biezen, 2001; Van Biezen et al., 2012). These findings have been linked together by the argument that activism and volunteering in political parties are declining because they are growing closer to the state. This idea is supported by evidence that states that heavily regulate their party systems have fewer party activists and members, other things being equal (Whiteley, 2011). Essentially the argument is that proximity to the state has the effect of smothering voluntary party organisations.
There are good reasons why parties have grown closer to the state, the chief reason being that parties are crucial to the workings of a successful democracy. As is well known, parties fulfil a variety of functions that support and sustain democratic practices and civil society more generally (Webb et al., 2002). This in turn implies that the state is anxious to regulate parties with the aim of ensuring that they fulfil these roles effectively. From this perspective the legitimacy of political parties rests on whether or not their rules on membership, their financial arrangements, regulations for appointing and electing candidates and a variety of other aspects of their behaviour stand up to scrutiny. Unfortunately, this perspective tends to be blind to the problems caused by over-regulation, one of which may be to deter ordinary citizens from participating in voluntary activity in parties and for them to increasingly feel that mainstream parties are remote from their concerns.
It can be argued that while the loss of grassroots volunteers in political parties is a very undesirable development because it weakens civil society, it is not fatal for the effectiveness of parties (Dalton and Wattenberg, 2000; Seyd and Whiteley, 1992, 2002; Webb et al., 2002; Whiteley and Seyd, 2002). This is because party volunteers have always been a small minority of citizens in these countries, and although they disproportionately contribute to political participation (Whiteley, 2009) democracy can survive without them. However, if we look beyond voluntary activity in political parties to partisanship in the wider electorate, that is a different matter. If proximity to the state is having the effect of weakening support for and trust in political parties among citizens in general this is a much more serious matter, since partisanship helps to anchor the political system and contributes to effective government (Whiteley, 2012).
Ever since the concept of party identification was introduced in the US in the 1950s it has played a key role in explaining voting behaviour and also in the maintenance of the stability of party systems over time (Campbell et al., 1960; Lewis-Beck et al., 2008). Originally, the concept of partisanship, developed at the University of Michigan, saw it as a stable, long-term psychological attachment that many individuals possess and which influences their voting behaviour both directly and indirectly. It prompts them to vote for their preferred party, but it also acts as a perceptual screen that filters out dissonant messages that might otherwise persuade individuals to switch to other parties (Campbell et al., 1960). The Michigan model suggests that partisanship will be more stable and grow stronger as the individual grows older.
Subsequent work has criticised this conception of partisanship and shown that party attachments are more dynamic and subject to influence by issues and by incumbents’ performance in office than the Michigan model allows (Achen, 2002; Clarke et al., 2004, 2009; Fiorina, 1981; Franklin and Jackson, 1983). Despite this, partisanship acts as a force for stability in a political system. As the experience of emerging democracies demonstrates, when partisanship is weak in a country, electoral behaviour in that country becomes very volatile and the party system is often fragmented and unstable (O’Donnell and Schmitter, 1986).
The purpose of this paper is to investigate the relationship between government regulation of political parties, trust in political parties and the strength of partisanship among voters in the advanced industrial democracies. The aim is to determine if the regulatory environment is weakening both partisanship and also trust in parties in European democracies. This exercise uses data from the European Social Surveys conducted since 2002, when the first survey was carried out. 1 The paper begins by examining trends in partisanship alongside changes in trust in political parties over time, and then goes on to examine two rival hypotheses about the relationship between parties and state regulation. The first is that regulation strengthens parties and increases levels of trust in them over time. The second is that regulation has the opposite effect of decreasing trust and weakening partisanship. We discuss the reasoning behind these rival hypotheses before going on to model the effects of regulation on partisanship and trust in parties using a multi-level modelling strategy. This exercise uses data from 24 countries derived from the 2010 European Social Survey (ESS), together with indicators of regulation from the Institute for Democracy and Electoral Assistance (Biezen and Borz, 2009) database on party finance and the Leiden database on party law in modern Europe. 2
Changes in partisanship and trust in parties in Europe over time
Insight into changes in public attitudes to political parties can be obtained from the ESS. This is ‘an academically-driven social survey designed to chart and explain the interaction between Europe’s changing institutions and the attitudes, beliefs and behaviour patterns of its diverse populations’. It was established in 2001 and conducted the first round of surveys in 2002, and these have been repeated every other year since with many common questions. The survey has paid particular attention to political parties by repeating questions on partisanship, voting and party membership in each round since 2002. Some 21 countries have been surveyed in all five rounds between 2002 and 2010, and the percentage of respondents in these countries who said that they felt close to a political party appears in Figure 1.

Changes in partisanship in 21 European countries 2002 to 2010.
It is apparent from the responses in Figure 1 that partisanship has been weakening in these countries over time. In the first of the ESS surveys in 2002 just under 52 per cent of respondents claimed to feel close to a political party. By 2010 this figure had declined to just fewer than 46 per cent. That said, the decline is not uniform, with significant changes occurring after 2006 and most notably in 2010. It is noteworthy that this period coincides with the arrival of the ‘great recession’ in Europe that started in late 2007 and early 2008 and had serious implications for the economies of the Euro zone countries, a topic we will return to below.
Figure 2 examines changes in another important indicator of public attitudes to political parties, that is, trust in parties. The trust question did not appear in the first survey but it has been asked in all of the subsequent surveys. Trust in parties is measured using an eleven-point scale where 0 means no trust at all and 10 means complete trust. The mean scores for the countries since 2004 tell a similar story to that of partisanship. The citizens of these 21 European democracies are less likely to trust political parties in 2010 than they were in 2004.

Changes in trust in political parties in 21 European countries, 2004 to 2010. The question asked was ‘Please tell me on a score of 0–10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust. Do you trust political parties?’
The averages in Figures 1 and 2 conceal considerable variations in partisanship and party trust in these countries over time. Figure 3 shows how partisanship changed in each of the countries between 2002 and 2010. It declined in 10 of them and increased in 10 while remaining stable in one. The overall scores declined because the loss of partisans in some countries considerably outweighed the increase in others. These data are not consistent with the evidence on declining voluntary activity in parties that has occurred more or less everywhere (Van Biezen et al., 2012). However, one noticeable feature of Figure 3 is that the countries that have experienced the largest reductions in partisanship are those that have faced the greatest economic difficulties in recent years. These include Greece, Spain, Portugal and the Irish Republic, all of which have been very severely hit by the ‘great recession’ (Galbraith, 2012; Krugman, 2012; Reinhart and Rogoff, 2009; ). In contrast, countries like Germany and Switzerland, which have experienced relatively mild recessions, have seen increases in partisanship over time.

Changes in partisanship in 21 European countries, 2004 to 2010.
Changes in trust in political parties charted in Figure 4 show a similar pattern to the changes in partisanship in Figure 3. In this case 12 of the countries have experienced a decline in trust and nine have experienced an increase. In a repeat of the pattern found in Figure 3, the citizens of countries such as Greece, Spain, Portugal and the Ukraine have experienced particularly large reductions in trust, notably Greece. While there is no evidence of a ubiquitous decline in partisanship and party trust in all of these European countries, it is evident that rather large reductions have occurred in some countries and this explains the patterns observed in Figures 1 and 2. Clearly when modelling the determinants of partisanship and party trust it is important to take into account the performance of the political system in delivering what the voters want. An analysis of the relationship between regulation and voluntary party activity will need to take into account the potential impact of the economic crisis. In the section below we examine the rival hypotheses about the relationship between partisanship, trust in parties and party regulation.

Changes in levels of trust in political parties in 21 European countries, 2004 to 2010.
Party regulation by the state
The UK Electoral Commission is an independent body funded by government for the purpose of regulating political parties and elections in Britain. One of its publications is targeted at party volunteers and in this document it defines campaign spending in some detail:
‘Advertising of any kind. For example, street banners, websites or YouTube videos.’
‘Unsolicited material sent to voters. For example, letters or leaflets you send that aren’t in response to specific queries.’
‘The manifesto and other documents setting out your party’s policies.’
‘Market research or other methods of finding out how people intend to vote.’
‘Press conferences or other dealings with the media.’
‘Rallies and events, including the cost of people’s attendance, and any goods, services or facilities provided.’
‘Transport in connection with publicising your campaign.’
(Electoral Commission, 2013: 6)
The commission requires parties to name a specific individual as the treasurer who is tasked with recording, reporting and obtaining receipts for all the expenditures associated with these activities. Perhaps in recognition of the amount of work involved, it will allow deputy treasurers to be appointed as well, but only if a party fills in a separate form for each of them. 2 The pre-amble to this document helpfully states, ‘If you do not comply with legal or regulatory requirements you or your organisation may be subject to civil or criminal sanctions’ (Electoral Commission, 2013).
This illustrative example of the details of party regulation can be repeated across Europe, as the Leiden Party Regulation database demonstrates. From the perspective of the state it is reasonable to collect a mass of detailed information to ensure that elections are conducted fairly and freely and in line with legal requirements. But it is also apparent that little or no thought has been given to the incentives for unpaid volunteers to take on such a role, particularly when they could end up with a criminal record if things go wrong.
This vignette illustrates the risks of excessive state regulation. If the state imposes heavy regulatory burdens on unpaid individuals in voluntary organisations without compensating them for their time and efforts it will reduce the incentives for them to be active in those organisations. At the same time the passive membership of parties, that is, individuals who pay their dues but do very little else, are supported and serviced by key activists. Thus, if the incentives to be active are reduced this will have a knock-on effect on passive membership as well. More generally, over-regulation runs the risk of the state capture of political parties so that instead of them remaining independent institutions of civil society they become part of the extended state apparatus (Bartolini and Mair, 2001; Kopecký and Mair, 2003; Van Biezen, 2004; Van Biezen and Kopecký, 2007).
However, our focus in this paper is on partisan attachments and trust in parties in the wider electorate. In this case it can be argued that the regulation of parties has one of two opposite effects. On the one hand, tighter regulation might have the effect of increasing trust in and support for political parties if the public are reassured by such regulation. This is a plausible argument if regulation means that parties will be well managed, free of corruption and deliver the policies that their supporters want. If so, proximity to the state should enhance their reputations and make them more legitimate in the minds of the voters, even if it comes at a cost of weakening volunteering.
The opposite effect is that tight regulation means that the public see parties as mere extensions of the state bureaucracy, or as part of a state-sponsored cartel that is increasingly remote from their concerns. This may not be a problem if the state is seen as well-run and effective in delivering policies, but in the context of the serious economic recession facing Europe, particularly in the Euro zone, states are increasingly seen as being unable to deliver on these policies. It is noteworthy that the average scores on the scales of satisfaction with government and satisfaction with the economy for the 21 countries surveyed in all five rounds of the European Social Survey were 4.45 and 5.10 in 2006, the last pre-recession survey. By 2010 these scores had fallen to 3.83 and 4.02, respectively. Needless to say, the reductions were much larger in some countries, such as Greece and Portugal, than in others. Accordingly, in these circumstances political parties that appear to be part of a state-sponsored cartel will increasingly prove unattractive to many voters, making them appear less trustworthy and weakening partisan attachments. In this view, excessive regulation inhibits support for political parties.
In light of this discussion we go on to model the impact of regulation on parties in order to determine which of these rival accounts actually applies. In order to do this it is necessary to control for a number of different factors that can influence partisanship and trust in parties apart from regulation, and we turn to this next.
The determinants of partisanship and trust in parties
The European Social survey is not a panel and so we cannot model changes over time with the same respondents. However, we can draw on the literature in political science and apply it to the task of explaining variations in partisanship and party trust across Europe. This will be done using the 2010 ESS data with the assistance of four different models of political participation that have been developed in the literature. These are the valence, the cognitive engagement, the civic voluntarism and the social capital models, and they will be described briefly below.
The valence model has been utilised particularly in research into electoral choice (Clarke et al., 2004, 2009; Whiteley et al., 2013). The term ‘valence’ was introduced by Donald Stokes in a seminal article that provided a comprehensive critique of the Downsian spatial model of party competition (Downs, 1957; Stokes, 1963). He wrote, ‘I will call “valence-issues” those that merely involve the linking of the parties with some condition that is positively or negatively valued by the electorate’ (Stokes, 1963: 373).
Stokes went on to explain that valence issues arise when there is broad agreement about the desired policy outcomes among the public, such as low unemployment and inflation, strong economic growth and efficient public services. Voters will support the party that appears to offer the best chance of delivering a competent performance in these areas (Clarke et al., 2004: 8). The valence model concentrates on policy delivery and is predominantly a theory about which party wins elections. However, it has wider implications in the contemporary situation facing many European states. Serious recession, and an apparent inability of governments and the major political parties to deal with the problems arising from this, have the effect of extending valence concerns to the party system in general and even to the legitimacy of democracy. Systemic failure to deliver on the key policies that matter to voters can have the effect of weakening support for mainstream political parties of all types. Clearly, both partisanship and party trust are likely to be influenced by the politics of performance in the present economic climate facing Europe.
The key measures relevant for estimating the valence model are public evaluations of the delivery of effective policies on the economy, and also in relation to public services such as health care and education. There are indicators for these in the ESS 2010 survey, which can be combined into a valence policy scale, and this is discussed more fully in the Appendix. Given the fact that valence considerations may help to account for wider support for the political system, then public evaluations of the state of democracy and the performance of the government in general are also included as predictors in the analysis.
In contrast to the valence model, the central idea of cognitive engagement theory is the proposition that participation depends on the individual’s access to political information and on their ability and willingness to use that information to make informed choices about politics and government (Clarke et al., 2004; Dalton, 2005; Norris, 2000). Cognitively engaged individuals are close to the classical Greek conceptions of the good citizen, who is an informed member of the community, fully participates in politics and understands the issues and complexities of government. Such individuals are interested in politics and civic affairs, are politically knowledgeable and have a clear understanding of the principles and practice of democracy. At the same time they are also likely to be critical citizens in the sense of evaluating the performance of incumbent and opposition parties and leaders (Norris, 1999). So the model has links with valence theory. A perception that the state may be failing to deliver in terms of the expectations of its citizens is likely to mobilise individuals to participate in unorthodox ways, for example by protesting (Muller, 1979). It may also reduce their willingness to support or trust mainstream political parties in general if they feel that these are not delivering on their promises.
The key variables for estimating the cognitive engagement model are education, media consumption, interest in politics and political knowledge. Education is measured in the ESS in terms of the respondent’s number of years of full-time education. Equally, media consumption is measured by the extent to which citizens follow politics and public affairs in newspapers, on television and on the radio. Interest in politics is measured with a variable that captures the extent to which individuals are motivated to pay attention to government and politics. Political knowledge relates to the citizen’s understanding of the way the political system works and about policies that are relevant to their concerns. Unfortunately, political knowledge is not measured in the European Social Surveys.
Turning next to the civic voluntarism model, this is perhaps the most well-known and widely applied model of political participation in political science. It has its origins in the work of Sidney Verba and his colleagues on participation in the US (Verba and Nie, 1972). It was subsequently applied to the task of explaining participation in a number of different countries (Barnes and Kaase, 1979; Parry et al., 1992; Verba et al., 1978, 1995). The central idea of the model is that participation is driven by the individual’s resources that underpin their involvement in politics and society. The authors define resources in terms of ‘time, money and civic skills’ (Verba et al., 1995: 271). Thus, high-status individuals rich in these resources will participate more extensively than low-status individuals who lack such resources.
In the model the psychological engagement of individuals with politics can also play an important role in influencing participation, and is defined largely in terms of the individual’s sense of political efficacy (Verba et al., 1995: 272). This derives in large part from the individual’s resources, since those who are well off and work in high-status occupations are more likely to develop a sense of efficacy than individuals who lack these attributes. In addition, the model suggests that citizens can be mobilised to participate by other people in their social networks. This process of mobilisation is defined as, ‘requests for participation that come to individuals at work, in church, or in organizations – especially those that come from friends, relatives, or acquaintances’ (Verba et al., 1995: 272).
The 2010 ESS has fairly good measures of individual resources, making it possible to identify the respondent’s occupational status, their educational background, income and the time they have available for outside activities. However, the survey lacks indicators of political efficacy and also mobilisation, although arguably these are secondary factors when it comes to explaining, as far as the model is concerned, why people participate. The authors applied the model to the task of explaining partisanship in five countries (Verba et al., 1978). In this research they argued that group resources arising, for example, from the individual’s membership of a trade union could offset their lack of individual resources. In particular they suggest that, ‘organization – and we might add ideology – is the weapon of the weak’ (Verba et al., 1978: 15). So low-status individuals who are involved in voluntary organisations might be willing to participate as much as high-status individuals. A similar point can be made about individuals who have strong ideological beliefs.
If the civic voluntarism model concentrates on individual resources, the social capital model tends to concentrate on community resources. Putnam defines social capital as ‘features of social organization, such as trust, norms and networks that can improve the efficiency of society by facilitating co-ordinated actions’ (Putnam, 1993: 167). The core idea of social capital theory is that if individuals are embedded in dense social networks that involve frequent interactions with other people in a voluntary capacity this will foster interpersonal trust or social capital. Social capital can then be used to solve common community problems. In this sense it is like other types of capital that can be used to make society more productive and the economy more efficient.
For many researchers trust is the key indicator of social capital (Brehm and Rahn, 1997; Fukuyama, 1995; Putnam, 1993, 2000; Whiteley, 1999). Trust is important since it allows individuals to engage in cooperative activities with strangers beyond their immediate family or community. There is a lively debate about the origins of social capital (Whiteley, 1999), but the dominant model argues that interactions between individuals in voluntary associations generate interpersonal trust. Communities characterised by high levels of social capital have dense networks of civic engagement and appear to have better health and education, less crime and higher rates of political participation (Putnam, 1993, 2000). By implication members of such communities will have more trust in political parties.
The 2010 ESS contains a battery of three items that measure interpersonal trust, and from these a latent variable can be constructed. In addition there is a question asking respondents if they have worked in voluntary organisations during the previous 12 months, which captures one of the drivers of social capital. A third item asks how frequently respondents meet with friends, relatives or colleagues, which provides a broad measure of social interaction, so a number of key components of the social capital model – interpersonal trust, voluntary activity and social interactions – are reasonably well measured in the survey.
In light of this brief review of four theoretical models that can be used to explain support for and trust in political parties, we estimate models of party attachment using the 2010 ESS data, in the next section.
Modelling partisanship and trust in parties in 2010
There are two models estimated in this section: one examines trust in political parties and the other examines individual partisanship. The pooled scores on the ‘trust in parties’ scale appears in Figure 5, and it is apparent that trust was rather low in these 24 countries in 2010 with a mean score of only 3.17 on the scale. Some 19 per cent of respondents gave political parties a trust score of 0, and beyond a score of 5 trust scores tail off rather rapidly. The result is that only half of 1 per cent of respondents assigned a score of 10 on the trust scale.

Trust in political parties in 24 countries in the 2010 European Social Survey.
Figure 6 contains the distribution of scores on the ‘strength of partisanship’ scale. To measure this, respondents were asked if they felt close to one political party rather than another; the positive responses to this question appeared in Figure 1. If they replied ‘yes’ to this question a follow-up question then asked how close they were to that party, and the responses to both questions are combined in Figure 6. With a total of 58 per cent of respondents saying that they did not feel close to any party and a further 1 per cent saying that they did have a preference for a party but did not feel at all close to it, then almost 60 per cent of Europeans can be described as non-partisan in 2010.

The strength of partisanship in 24 countries in the European Social Survey, 2010.
Trust in political parties and strength of partisanship are both modelled in Table 1 using the various indicators of the valence, cognitive engagement, civic voluntarism and social capital models. The table contains unstandardised regression coefficients in the first and third columns and standardised coefficients in the second and fourth columns. It is important to note that with more than 45,000 respondents it is relatively easy for coefficients to attain statistical significance in these models, and so the standardised coefficients play a particularly useful role in distinguishing between the variables in terms of their importance for explaining trust and partisanship.
Models of party trust and partisanship in 24 countries in 2010.
Note: p<0.01=***; p<0.05=**;p<0.10=*.
Source: European Social Survey (2010) (n = 45,872).
The goodness of fit of the party trust model is reasonable for an individual-level analysis (R2 is 0.45). The valence model is the most important for understanding party trust, with all of the variables attaining statistical significance and having large standardised effects. Easily the most important variable, as the standardised coefficients show, is satisfaction with government. This was closely followed by the valence issues scale, and then the satisfaction with democracy variable. Given that the valence model appears to be dominant for explaining trust in parties, it is easy to see why trust has declined in the countries that have experienced the worst of the recession. This is fairly strong evidence to suggest that a severe economic crisis coupled with an inability of governments to cope with the crisis has directly affected civil society in Europe. It has made people less likely to trust political parties. The effect works through policy dissatisfaction, but also goes beyond that to include concerns about the performance of governments, and also with democracy itself.
It is also apparent that the cognitive engagement model contributes to explaining trust in parties. Media consumption is an important predictor of trust, while interest in politics has the strongest effect in that particular model with a standardised coefficient of +0.16. Clearly the politically engaged are more likely to trust political parties than the disengaged. However, educational attainment also plays a role in the cognitive engagement model and has a negative impact on trust, indicating that the most highly educated are less trusting of parties than the uneducated. This is an anomalous finding in relation to the cognitive engagement model since educational attainment is a positive predictor of participation, in the model (Pattie et al., 2004; Verba et al., 1995).
The impact of the civic voluntarism model on party trust is particularly interesting since the signs of the effects are nearly all negative. Thus, affluent and highly educated respondents are less likely to trust parties, in comparison with their less affluent and less well educated fellow citizens. Given that we are investigating the influence of social status and its correlates on trust in parties in the middle of the largest economic crisis in Europe since the 1930s, these findings suggest that highly resourced individuals are reacting to this crisis more negatively than their poorly resourced counterparts. The exception to this is the negative impact on trust in parties of the number of hours worked in the week. The expectation is that time poverty caused by excessive working hours reduces political participation, and so this is consistent with the civic voluntarism model.
Two of the three variables in the social capital model contribute to explaining trust in political parties. An individual who works in a voluntary organisation and trusts his or her fellow citizens, will be more likely to trust political parties. In a comparison of the two variables, interpersonal trust has a considerably larger impact on trust in parties than does voluntary activity. The social contacts variable, which measures how embedded individuals are in friendship and family networks, does not appear to have an influence on trust in parties. 3
The left–right ideology scale shows that respondents who identify themselves as being on the right of the ideological spectrum are less trusting of parties than respondents on the left. With regard to the other demographic control variables, older citizens are less likely to trust parties than younger citizens, although the quadratic specification shows that this lack of trust declines in importance as respondents get much older. Religious respondents are more trusting than the non-religious, a finding consistent with the civic voluntarism and social capital models, and women are more trusting than men. Finally, ethnic minorities are more trusting of parties than the ethnic majority. Overall, each of the four models makes a contribution to explaining trust in parties but it is apparent that the valence model dominates the picture.
Turning next to the ‘strength of partisanship’ model, this is rather different from the ‘party trust’ model. The goodness of fit is considerably smaller and, although the valence indicators figure in the partisanship model, the effects are rather weaker than in the party trust model. Satisfaction with government and satisfaction with democracy are both positive predictors of partisanship but the valence policy scale is not a predictor. This suggests that partisanship is driven more by diffuse support for the party system rather than by specific policy concerns. The most important effects in the model are associated with the cognitive engagement model, since, of all the variables, interest in politics has easily the largest impact on partisanship. With respect to the social capital model, voluntary activity has a positive impact on partisanship but interpersonal trust and social contacts do not appear to play a role.
Unlike in the party trust model the indicators in the civic voluntarism model all have a positive impact on partisanship. Thus, affluent respondents with high occupational status are more likely to be partisans than poorer, low occupational status respondents. In addition, males are more likely to be partisans than females and older people are more likely to support a party than younger people. Finally, citizenship counts in the sense that non-citizens are less likely to support a party than citizens, but ethnic minorities feel closer to parties than the ethnic majority. In this respect, ethnic minorities have the same impact as in the party trust model. Overall, the models in Table 1 suggest that valence considerations play a key role in explaining trust in parties, and they also play a role in explaining partisanship although it is not the dominant role. The cognitive engagement model stands out as the key to understanding partisanship and, to a lesser extent, the civic voluntarism model.
Up to this point the modelling does not take into account the relationship between party trust and partisanship. Since they are both likely to figure as predictors in the models, we examine this issue next.
Interactions between trust in parties and partisanship
The relationship between trust in parties and partisanship is very likely to be interactive, with partisanship influencing trust in the first model, and trust influencing partisanship in the second. Given this reciprocal relationship, the influence cannot be accurately estimated using an ordinary least squares (OLS) regression (see Kennedy, 2003: 180–190). This is because OLS produces biased estimates in models with two-way relationships. 4 To deal with this problem it is necessary to find instrumental variables that can stand in for the two endogenous variables when they appear as predictors. The idea is that if such an instrumental variable is closely correlated with trust in parties but not with partisanship, then it can stand in for trust in parties as a predictor in the partisanship model without producing biased estimates. Theoretical considerations provide a guide to identifying these instrumental variables.
Two instrumental variables will be utilised for the party trust measure and a further two for the partisanship measure, coupled with a two-stage least squares estimation strategy (see Cameron and Trivedi, 2010: 183–192). The instruments for party trust are trust in the legislature and trust in the legal system. Trust in political parties is quite likely to be correlated with trust in other state institutions such as the legislature and the legal system, but there is no obvious reason why partisanship should be influenced by institutional trust in the same way. 5 In the case of partisanship the instruments are party membership and whether or not the respondent worked for a political party in the previous year. Again it is quite likely that party members and volunteers will have strong party attachments, but there is no reason to suppose that trust in parties in general is going to make people join or work for a political party. 6
The two-stage least squares estimates in Table 2 confirm that there is a reciprocal relationship between trust in parties and the strength of partisanship in the models. Strong partisan attachments make people more likely to trust political parties and, equally, trust in parties has a positive influence on partisanship. The standardised coefficients in the two-stage least squares estimation show that partisanship has a slightly stronger impact on trust in parties than the latter has on the former. These variables measure different aspects of the public’s attitudes to political parties. It is noteworthy that when the variables are added to the two models they do not materially change the size of the other effects appearing in Table 1, so the four different models of participation continue to be influential in the interactive model.
Two-way interactions between trust in parties and partisanship in 2010.
Note: p<0.01=***; p<0.05=**;p<0.10=*.
Source: European Social Survey (2010) (two stage least squares estimates).
Bringing in regulation
Having developed fairly well-specified models of party trust and partisanship, including the interaction between them, we can now go on to incorporate regulation into the analysis. In a paper on party members, Whiteley (2011) used data from the Institute for Democracy and Electoral Assistance (IDEA) database to measures the state regulation of political parties. 7 This database has been greatly expanded and updated and contains many more indicators of regulation. A total of 32 measures of party regulation are included in the present analysis covering such things as donations to political parties, the scope and limits of public funding and the reporting requirements imposed on candidates and parties, particularly in relation to elections. A list of the 32 measures appears in the Appendix, and Figure 7 shows that there is considerable variation in the state regulation of parties across Europe.

IDEA party regulation scores in 24 countries.
Figure 7 shows that Switzerland has the least regulated party system of any of these European states. It has only minimal restrictions on donations to parties, and reporting requirements on party activity and expenditure. Interestingly enough, former communist countries such as Bulgaria and Croatia are among the countries with the most regulated party systems with a plethora of rules and reporting requirements applied to them. The Scandinavian countries such as Norway, Sweden and Denmark tend to be lightly regulated, as is the Netherlands. In contrast, southern European countries such as France, Portugal and Greece tend to be more heavily regulated.
Figure 8 shows the relationship between the mean party trust scores in the 24 countries and the regulation scores. It is strongly negative with a correlation of −0.81, showing that a low level of trust is strongly associated with heavy regulation. Of the two hypotheses discussed earlier, it appears that proximity to the state is associated with low rather than high levels of trust in political parties. The precise causal links are ambiguous since cross-sectional data cannot definitively answer the question as to which came first – regulation or mistrust – but it is evident that these variables are strongly associated. A plausible account of this relationship in the case of ex-communist countries and former dictatorships in southern Europe is that a legacy of mistrust of undemocratic parties still present after democratic transition had taken place ensured that new parties are tightly regulated even when democracy has become consolidated. In other words, a legacy of mistrust of parties from the past promotes heavy regulation.

The relationship between trust in parties and party regulation in 24 countries, 2010.
Figure 9 shows the relationship between the mean scores on the ‘strength of partisanship’ scale and the regulation index. The story is similar to that of Figure 8 except that the relationship is not as strong, with a correlation of −0.64. Overall though, regulation is heaviest in those countries in which there are fewer partisans and where the strength of partisanship is weakest. These figures show that the relationship between voluntary activity in political parties and party regulation highlighted in an earlier paper (Whiteley, 2011) extends to partisanship and also to trust in political parties.

The relationship between the strength of partisanship and party regulation in 24 countries, 2010.
A second source of information about party regulation in Europe is the Party Laws in Modern Europe database maintained at the University of Leiden. 9 This is an extensive database of the laws passed relating to the regulation and control of political parties in Europe since the Second World War. The laws vary in their scope and impact on political parties in the 33 countries in the database, covering parliamentary and extra-parliamentary parties, the activities and behaviour of parties, and their external oversight. As a first approximation we can count the number of laws passed in each of the countries represented in the 2010 European Social Survey, governing the conduct of political parties between 1944 and 2012. This provides a rough measure of the amount of legislative control that has been imposed on political parties in the modern era. Clearly some laws are more important than others and so this is a fairly approximate measure, but it is nonetheless useful for capturing the overall legislative regulatory environment faced by political parties.
In the event, there is a moderately strong correlation between the number of laws introduced to regulate parties and the scope of regulation measured in the IDEA database in these countries (r = +0.42). This suggests that the two indices are measuring different, but related, aspects of state regulation of parties. The two measures can be used to model individual-level trust in political parties and the strength of partisanship across Europe in 2010.
Party trust, strength of partisanship and state regulation
The aggregate measures in Figures 8 and 9 are interesting, but they do not necessarily show that individual citizens are less trusting of parties or are less partisan as a consequence of regulation. It would be an ecological fallacy to assume that aggregate relationships imply individual-level relationships (Robinson, 1950), and so we need to model the effects of state regulation on individual-level partisanship and trust.
Table 3 contains the two-stage least squares estimates of the models of partisanship and party trust at the individual level, which incorporates the two aggregate-level party regulation measures in a random intercept multi-level modelling specification 8 (see Raudenbusch and Bryk, 2002; Snijders and Bosker, 1999). The impact of regulation at the national level on individuals is captured by the two variables that influence the intercepts of the party trust and strength of partisanship individual-level models. It is apparent that both the IDEA party regulation index and the Leiden party laws index have a significant negative impact on party trust in these countries, controlling for all the other variables in the individual-level model. The goodness of fit of the aggregate trust model is high indicating that the effects are highly significant in both cases.
Multi-level models of trust in parties and partisanship in 2010.
Note: p<0.01=***; p<0.05=**;p<0.10=*.
Source: European Social Survey, 2010 (two stage least squares estimates).
In the case of the strength of partisanship model the IDEA regulation index has a significant negative impact, but the party laws index does not attain significance although the negative sign is consistent with the theoretical argument and reinforces the findings in the party trust model. Overall, in the case of partisanship the aggregate effect is much more modest than for trust in parties, but it is nonetheless consistent with the idea that party regulation inhibits both trust and partisanship.
The individual-level predictors in the models in Table 3 are very similar to those in Table 2, indicating that valence considerations play a key role in explaining trust in political parties whereas the cognitive engagement model tends to be more important in the case of partisanship. Thus, the addition of the regulation variables to the model does not change the conclusion that performance issues have had the effect of weakening support for parties in Europe.
Discussion and conclusions
This paper has linked together two important findings in party research: the growing relationship between political parties and the state on the one hand, and the decline in voluntary support for political parties among electorates on the other. These findings are not the same as those showing that regulation stifles voluntary activity in political parties, because the decline of voluntary parties is much more ubiquitous than party trust or partisanship. Unlike volunteering, partisanship and party trust are not declining at similar levels everywhere. However, support for parties has declined rather precipitously in those countries most strongly hit by recession, which has been the main problem facing European countries over the last five years. Individual-level modelling suggests that a large part of this is due to growing disillusionment with the ability of governments to deliver on the valence issues such as economic prosperity and personal security.
These results show that the party systems in advanced industrial countries are weakened by prolonged recession and by government failure. But having controlled for these effects in the multivariate model it is still the case that over-regulation of political parties weakens their support among European electorates. State capture makes parties more remote from the concerns of individuals, who have less and less of an ability to influence them as ordinary citizens. At the same time this process associates parties with policy failure and growing scepticism about the effectiveness of democracy in these countries. Quite clearly the regulation of political parties in some countries inhibits their attractiveness to the public quite independently of the recession.
There may be other causal links at work such as growing inequality in society as a result of recession and austerity. The consequences of economic failure in Europe have by and large been borne by low-status workers, welfare recipients and young people trying to enter the labour market, rather than by the bankers whose behaviour triggered the recession in the first place (Krugman, 2012). The widespread sense of unfairness this has generated serves to further undermine established mainstream parties in the eyes of many voters, and is having the effect of reshaping the European party system. 9
The policy implications of these findings are hard to untangle using cross-sectional data. A longitudinal study of party support across Europe would greatly enhance our ability to discern whether de-regulation of parties might improve their support in some countries. If regulation is a key factor helping to explain the weakening parties then de-regulation should have a positive influence. On the other hand, if mistrust is the primary cause of party regulation then de-regulation might be ineffective. Future research should focus on acquiring such longitudinal data so that this important question can be addressed.
Footnotes
Appendix: Variables and scales used from the 2010 European Social Survey
The dependent variables in the two models are:
Partisanship – Is there a particular political party you feel closer to than all the other parties? Trust in political parties – Please tell me on a score of 0–10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust. Do you trust political parties?
Acknowledgements
This is a revised version of a paper presented at the conference on Contested Legitimacy: Paradoxes in the Legal Regulation of Political Parties, University of Leiden, the Netherlands, 23–25 January 2013.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
