Are political parties necessary?

Electoral Studies 22 (2003) 287–299 www.elsevier.com/locate/electstud Are political parties necessary? Sören Holmberg ∗ Department of Political Science, Göteborg University, Box 711, S-405 30 Göteborg, Sweden Abstract Based on recent survey data from 20 countries, this article examines the ‘contagion’ thesis: that lack of support among the electorate at one level of the political system may spread to other levels. The levels examined are political parties and the party system. The results demonstrate that the degree to which people are attached to a party is related to how they view the need for parties in their country. In countries with widely different democratic systems, people who identify strongly with a party tend to be much more supportive of the idea that parties are necessary to the functioning of the political system than people without strong party attachments. Hence, party identification, a concept originating with the ‘Michigan Four’, is not only a very useful tool in analyzing voting behavior. It can also be applied to studying support for party-based democratic political systems.  2002 Elsevier Science Ltd. All rights reserved. Keywords: Party identification; Party-based democracy; Political party 1. Introduction Systematic cross-national comparative study of elections was one of Warren Miller’s abiding ambitions. Whilst not realised in his lifetime, in his own research, the international research project, The Comparative Study of Electoral Systems (CSES), investigating electoral behaviour in widely different democratic systems in some 30 countries, was embarked upon very much in the spirit of extending our understanding of the behaviour of national electorates through comparative investigation—as frequently urged by Miller. The CSES project enjoyed Miller’s active support, demonstrated by his participation in the meeting, at Ann Arbor in 1995, when the survey ∗ Tel.: +46-31-773-1227; fax: +46-31-773-4599. E-mail address: soren.holmberg@pol.gu.se (S. Holmberg). 0261-3794/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0261-3794(02)00016-1 288 S. Holmberg / Electoral Studies 22 (2003) 287–299 questions were first drafted.1 The analysis presented in this paper, based on early findings from the CSES project, illustrate the leverage that can be achieved on questions about the workings of democratic systems which cannot be answered satisfactorily by limiting the focus to the well-established democracies of the US and Western Europe. To the question ‘Are political parties necessary?’ the answer among a majority of people living in democratic countries around the world is a resounding ‘Yes’. The evidence comes from the answers to questions in the CSES survey asking electors for their opinion about the necessity of political parties in their country. The results reported here are based on findings from the first 20 countries surveyed, which was during the 1996–98 period.2 These results show that a clear majority of electors gave an affirmative answer to the question whether or not parties are necessary to the working of the political system in their country. The majorities range from a high of around 90 per cent in the Netherlands and Norway to lows of 55–57 per cent in Lithuania, the USA, and the Ukraine. In none of these 20 countries was there a majority of the view that parties are not necessary to the working of democracy in their country. The CSES ‘parties-are-necessary’ survey question was phrased in a rather general way, but not so general that it lost touch with political realities. The intention was to get at an ‘in principle’ view of the necessity of political parties without losing the anchorage of parties in national political systems, the habitat in which parties have functioned since their emergence some 100 years ago, or, in a few cases, 200 years ago. The formulation of the question was: ‘Some people say that political parties are necessary to make our political system work in (country). Others think political parties are not needed in (country). Using the scale on this card, where would you place yourself?’. The response scale ran from 1 through 5, with the endpoints representing ‘parties are necessary’ and ‘parties are not needed’. The question was easy to answer for most people in all of the countries considered here: the proportion of Don’t Know responses varied between 1 and 4 per cent in most countries, but it was as high as 18 per cent in Taiwan. 2. The contagion thesis That clear, often high, majorities of people living under representative forms of government in which parties are dominant actors perceive political parties to be necessary to the functioning of their political system is clearly positive news—for 1 Three of the lead investigators on the CSES project, Hans-Dieter Klingemann (Germany), Jacques Thomassen (Netherlands), and the current author (Sweden), had, at some time in their career, come under Miller’s influence, either from working alongside him at Ann Arbor or had been engaged in research under his supervision. 2 The CSES surveys are conducted as post-election studies with random samples of electors. The number of cases varies between about 1000 in Lithuania to over 4000 in New Zealand. These early data were released in Spring 2000. S. Holmberg / Electoral Studies 22 (2003) 287–299 289 representative government, for democratic ideas, and for the parties themselves. Parties may have changed their character from the classic ‘mass’ party (Duverger, 1954) into ‘catch-all’ parties (Kirchheimer, 1966), or ‘electoral-professional’ parties (Panebianco, 1988), even into ‘cartel’ parties (Katz and Mair, 1995) or ‘cadre-catchall’ parties (Olson, 1998). And electors may be less attached to parties in the partisan sense, and less often join up as members, than was once the case, as research from many countries shows (Schmitt and Holmberg, 1995; Dalton, 1999; Mair and van Biezen, 1999; Widfeldt, 1999). Even so, most citizens in democratic countries are not yet ready to jettison political parties, or to envisage the possibility of democratic politics without parties. Our interest here, however, is not to use the ‘parties-are-necessary’ measure to mount a defence of party in democratic politics. Rather, it is to use the measure to open up questions about the health of, and the outlook for, democratic political systems. Russell Dalton (1999) may have been too alarmist when he claimed that the declining strength of party identification across most Western democracies signaled not only a weakening of party in democratic politics but also a progressive disengagement from politics amongst electorates. After noting that the regression slopes for party identification tend to be negative over time in most democratic countries, Dalton (1999, p. 66) concludes: If party attachments reflect citizen support for the system of party-based representative government, then the simultaneous decline in party attachments in nearly all advanced industrial democracies offers a first sign of the publics affective disengagement from politics. Thus, Dalton presupposes that, in a democratic political system in good health, there will be a rather strong positive relationship between party identification and support for the existence of parties. On these grounds, he fears that weakened party attachments may, in the end, erode beliefs in the need for parties as the major actors in democratic politics. Declining party identification could turn out to be contagious, from Dalton’s perspective, leading to declining support for party-based democracy. Dalton clearly subscribes to a revised version of David Easton’s (1965, 1975) theory of political support, leading him to argue that political support is to be analyzed as a multi-dimensional phenomenon (Dalton, 1999, pp. 58–61). Accordingly, he develops a five-fold conceptual framework distinguishing between different levels of political support. The levels are political community, regime principles, regime performance, regime institutions, and political authorities. Popular support at one level may, or may not, affect support at other levels, but there is always the risk that discontent at one level may spread to another level. According to Dalton (1999, p. 59): Negative attitudes toward political officials can exist with little loss in support for the office itself or the institutional structure encompassing this office. As the object of dissatisfaction becomes more general—the performance of the regime or attachment to the political community—the political implications increase. A 290 S. Holmberg / Electoral Studies 22 (2003) 287–299 decline in support for the political process might provoke a basic challenge to constitutional structures or calls for reform in the procedures of government. Weakening ties to political community in a democratic system might foretell eventual revolution, civil war, or loss of democracy. Alarmist predictions perhaps, but Dalton’s claims merit serious examination. This can be done using data from the CSES. In the present analysis, however, only part of Russell Dalton’s ‘contagion thesis’ is examined; that is, the linkage, in terms of attitudinal perspectives, between support for political actors and support for a regime institution. The actors are political parties; the institution is the political system. The issue addressed in this paper, then, is whether the degree of affective attachment to political parties among national electorates has any bearing on more general views about the need for parties in democratic systems of government. If the relationship is positive, especially if strongly positive—in light of the generally weakening of attachments to parties—the outlook for democratic politics may be cause for concern. 3. National patterns Tabulations of the data for the 20 countries examined here are presented in Table 1. The countries are rank ordered in the table by the overall balance between positive and negative opinions about the need for political parties. The overall, summary balance is shown in the last column. These straightforward distributions reveal that a majority of people in every one of the countries listed thought that political parties are necessary. In most instances, very few respondents were of the opinion that parties are not needed, although the balance of opinion is notably less supportive of parties in the USA, Lithuania, and Ukraine; and the balance is slim in Poland, Mexico, and Japan. At the aggregate level, the necessity of political parties was most widely supported in the Netherlands, Norway, and Sweden, three long established democracies with multi-party systems. People in Spain and Taiwan, two relatively new democracies, also exhibit high levels of support for the notion that parties are necessary to the functioning of their political system. At the other end of the spectrum are electors in political systems with very different histories: the USA., the oldest democracy in the world, and in Lithuania and Ukraine, two of the newest democratic systems. In the middle of the ranking is a mixed bag of countries: in some, party-led democratic politics is well established (Britain, Australia, and New Zealand); in others, the democratic regime is only of some 10 years standing (Romania, Czech Republic, Hungary, and Poland); and in yet others (Mexico and Argentina), democratic politics has been punctuated by periods of authoritarian rule. Clearly, then, there is no obvious match between the age of a democratic system and popular support for the notion that political parties are necessary to the functioning of a democratic political system. That the publics in the East European countries tend not to occupy a high placing in the country ranking is not surprising. We should not expect parties, and party-led politics, to be rapidly embraced with enthusiasm by everybody in such new democ- S. Holmberg / Electoral Studies 22 (2003) 287–299 291 Table 1 Distribution of opinion on the necessity of parties in 20 countries Parties are necessary Parties are not needed Country 1 2 3 4 5 Netherlands Norway Spain Taiwan Sweden Germany Britain Romania Czech Republic Argentina Hungary Israel Australia New Zealand Poland Mexico Japan USA Lithuania Ukraine 61 68 58 59 61 58 43 65 39 66 45 53 43 39 38 56 37 25 31 43 29 21 25 23 20 22 33 14 35 12 27 20 28 32 29 13 28 31 24 14 7 8 9 10 12 14 19 10 20 10 19 16 20 20 21 15 21 26 24 17 2 2 4 3 4 2 3 5 4 6 4 5 5 6 7 7 9 12 9 8 1 1 4 5 3 4 2 6 2 6 5 6 4 3 5 9 5 6 12 18 Opinion balance: (1+2)⫺(4 + 5) +87 +86 +75 +74 +74 +74 +71 +68 +68 +66 +63 +62 +62 +62 +55 +53 +51 +38 +34 +31 Notes: Entries in columns 1–5 are row percentages. The question asked was: “Some people say that the political parties are necessary to make our political system work in [country]. Others say that no political parties are needed in [country]. Using the scale on this card, where would you place yourself?”. The scale runs from 1 to 5, coded as 1 for “political parties are necessary for the functioning of our political system” and 5 for “political parties are not needed in [country]’; 3 on the scale is a neutral mid-point. Source: Comparative Study of Electoral Systems. racies. Parties and party systems take time to become embedded in a political system and in people’s minds. Nonetheless, the overall balance of opinion in those countries is positive, if more narrowly so in Lithuania and Ukraine. The United States’ case is another matter. On the one hand, the USA might be viewed as an outlier, which is distinctive in terms of the individualistic and candidate-oriented traditions of American politics. For sure, US parties have functioned more as electoral machines, and are ideologically less well defined, than parties in West European countries. Hence, it might be argued, Americans perceive parties as less necessary to the functioning of their political system. On the other hand, it might be argued that as Americans have a particularly long experience of political parties, the much slimmer balance of support for the idea that parties are necessary to the functioning of the political system foreshadows what may be in store for parties in other old, and notso-old, democracies. The distributions in Table 1, however, tell us little about the relationship between 292 S. Holmberg / Electoral Studies 22 (2003) 287–299 affective attachment to parties and the standing of parties in the political system of these countries. It is the individual-level relationship that we are primarily interested in; that is, the relationship, among electors, between levels of party identification on the one hand, and opinions about the need for parties on the other. If that relationship is positive, fairly strong, and looks much the same across different political systems, Dalton’s thesis passes our test. That is, it would provide support for Dalton’s claim that declining levels of party identification may, in the long run, trigger declining public support for party-based democracy. 4. Party identification The CSES measure of party identification is based on a set of linked questions with a filter component. The first question asks respondents: ‘Do you usually think of yourself as close to any particular political party?’. If the answer is ‘Yes’, respondents were then asked: ‘Do you feel very close to this party, somewhat close, or not very close?’. Those who answered ‘Very close’ to this follow-up question are classified as strong party identifiers; those who answered with any response other than ‘Very close’ are classified as weak party identifiers. In the case of respondents who answered ‘No’ or ‘Don’t know’ to the first question, the follow-up question was: ‘Do you feel yourself a little closer to one of the political parties than the others?’. Respondents who answered ‘Yes’ to this second question are classified as without party identification but with a party preference. Respondents who answered ‘No’ or ‘Don’t know’ to both the first and second question are classified as having no identification. The rationale underlying the questions is to tap the same phenomenon as the classic Michigan concept of party identification as a ‘psychological tie’ reflecting ‘an affective orientation’ towards one or other of the parties competing for support in a political system (Campbell et al., 1960, pp. 121–2), a concept with which Warren Miller was closely associated. The original Michigan question format was adjusted, however, to be applicable not only in two-party systems, as in the USA, but also in multi-party systems (Budge et al., 1976; Holmberg, 1994). The data are versatile enough to allow for constructing many different versions of the strength component of party identification. In the analysis reported here, a four-category variable was constructed from the party identification data to distinguish between strong party identifiers, weak party identifiers, non-identifiers but with a party preference, and people who professed neither party identification nor party preference.3 This derived variable measures 3 According to the measure constructed for this analysis, the proportion of party identifiers in the several countries is: Australia 82 per cent, Israel 63 per cent, Ukraine 60 per cent, USA 57 per cent, New Zealand 56 per cent, Norway 53 per cent, Sweden 52 per cent, Poland 50 per cent, UK 49 per cent, Romania 45 per cent, Czech Republic 45 per cent, Mexico 45 per cent, Spain 43 per cent, Japan 41 per cent, Germany 37 per cent, Argentina 37 per cent, Hungary 37 per cent, Taiwan 33 per cent, and Lithuania 32 per cent. A full four-fold party identification variable could not be constructed for The Netherlands; see Table 2. 293 S. Holmberg / Electoral Studies 22 (2003) 287–299 Table 2 Percentage of respondents who think parties are necessary among people with different degrees of party identification; and correlation between strength of party identification and the parties-are-necessary variable Strength of party identification Difference between strong identifiers and nonidentifiers Country strong identifiers weak preference identifiers only no identification Lithuania USA Britain Japan Czech Republic Romania Germany Poland Argentina New Zealand Hungary Sweden Ukraine Netherlands Mexico Taiwan Australia Spain Norway Israel 88 79 88 86 83 88 93 77 93 84 81 87 69 – 80 90 77 90 91 73 69 61 83 72 82 88 87 75 83 76 79 84 62 92 70 88 72 87 91 76 45 43 59 54 58 69 69 55 71 63 62 70 46 84 63 76 63 78 80 68 55 42 77 65 73 89 84 66 81 65 76 78 47 92 69 82 57 82 89 – +43 +36 +29 +32 +25 +19 +24 +22 +22 +21 +19 +17 +23 – +17 +14 +14 +12 +11 +5 Correlation between strength of party identification and parties-arenecessary variable (r) 0.26 0.26 0.25 0.24 0.23 0.23 0.22 0.20 0.19 0.19 0.18 0.18 0.15 0.15 0.14 0.14 0.13 0.13 0.13 0.08 Notes: Entries in columns 1–4 are the percentage of respondents who scored 1 or 2 on the parties-arenecessary scale (see Table 1) among the four party identification groups. See text for wording of party identification questions. A full four-category party identification variable could not be constructed for the Netherlands. Source: Comparative Study of Electoral Systems. varying degrees of party identification in a way that is valid within the different countries. In the case of Sweden, for example, where it is possible to compare (within the same sample of respondents) the CSES variable and the traditional party identification variable used in the Swedish Election Studies since the 1960s, the constructed CSES variable functions with an acceptable degree of reliability.4 4 The correlation (r) between the strength components of the CSES measure of party identification and the traditional Michigan measure is 0.68. 294 S. Holmberg / Electoral Studies 22 (2003) 287–299 The reliability of the CSES party identification measure in comparing level estimates for party identification across different national systems may be more doubtful, however. Words such as ‘close to’ or ‘feel…a little closer to’ used in the interview questions are notoriously difficult to translate to ensure functional equivalence across several countries. Furthermore, the party identification questions were administered in slightly different ways in the national studies, potentially creating problems of comparability. An indication that level estimates of party identification aggregated to the national level might be questionable in some cases is the negative rank order correlation between the strength of party identification amongst national publics and support for the necessity of parties across our CSES countries (Spearman’s rho ⫺0.19). The main reason for this is that the three countries that display a relatively high score on strength of party identification also display a low score on support for the necessity of parties (Ukraine, Australia, and the USA), while three other countries register a low score on strength of party identification but a very high score on the necessity of parties (Spain, Germany, and Taiwan). 5. Party identification and the necessity of parties Our focus is on the extent to which people’s opinions about the need for parties is related to the strength of their party identification. Table 2 shows the level of support for ‘parties-are-necessary’ among respondents with different levels of attachment to parties in their political system (columns 1–4). These results show that higher proportions of party identifiers (even weak identifiers) believe in the necessity of parties compared with those who express only a ‘preference’ for a party or have ‘no identification’ at all with a party. Also shown, in the last column, is the correlation between the strength of party identification and the parties-are-necessary variable. In each case, the relationship is positive: across all the countries examined, there clearly is a relationship between the strength of party identification and a recognition of the necessity of parties to the functioning of the democratic system in the respondent’s country. Thus, there is empirical support for Dalton’s thesis: there is an individual level relationship between people’s degree of party attachment and their opinion about the necessity of parties. The correlation between party identification and the parties-are-necessary variable is positive in all 20 countries. The strength of the relationship is not particularly impressive, however. The mean across all countries is 0.18, and the correlation for individual countries ranges from a high of 0.26 in Lithuania and the USA to a low of 0.08 in Israel. Given some uncertainty about the validity of the party identification measure, which might also extend to the parties-are-necessary variable, perhaps we should not expect strong relationships. If we compare opinion about parties between extreme groups, however, such as strong party identifiers and non-identifiers, some impressive percentage point differences are revealed. In Lithuania, for example, the opinion that parties are necessary is supported by 88 per cent of strong party identifiers but by only 45 per cent of those with neither party preference nor party identification—a 43 percentage point difference. The corresponding numbers for the USA S. Holmberg / Electoral Studies 22 (2003) 287–299 295 are almost as impressive: among strong party identifiers, 79 per cent were positive about the necessity of parties, whereas among people with neither party identification nor party preference only 43 per cent supported the idea of parties—a difference of 36 percentage points. Note, also, that among the well-established democracies, the USA stands out (with the exception of Australia) as having the lowest proportion of strong party identifiers who consider parties to be necessary. In all, across the countries examined, the average difference between strong identifiers and those with neither party identification nor party preference is 21 percentage points, a not inconsiderable difference by most social science tests. 6. Intervening variables It may be, of course, that the relationship between our two variables of interest— strength of party identification and parties-are-necessary—is influenced by other variables. To test for this possibility, several theoretically relevant variables were included in a multiple regression model, with ‘parties-are-necessary’ as the dependent variable. In addition to strength of party identification, the model included gender, age, educational level, a knowledge index, and a trust item.5 The model was tested in a subset of 11 countries. As the relationship between strength of party identification and ‘parties-are-necessary’ is maintained after applying these controls, the results of the multiple regression analysis are only summarised here. Overall, across the 11 countries, the model explains 12 per cent of the variance (r2) in the ‘parties-are-necessary’ variable, ranging between a high of 20 per cent of the variance in the USA to a low of 6 per cent in Spain. Whereas gender has a weak bivariate relationship with ‘parties-are-necessary’ in most countries, with women tending to be somewhat less positive than men, gender effects disappear into insignificance in most countries in the multiple regression model. The same is true for education: the bivariate correlations, with highly educated people tending to be somewhat more positive to the idea of parties than people with less education, become insignificant in most countries when other variables are controlled for. Political knowledge, however, retains its positive relationship with the parties-are-necessary variable, even after the controls are applied. In most countries, politically knowledgeable people are more likely to think that parties are necessary than people with less political knowledge. The case of age, however, stands out as different: the relationship between strength of party identification and attitudes towards the necessity of parties does not look much the same across our sample of countries. None the less, in all the countries examined here, the multiple regression analyses reveals that the strength of party identification has clear and significant effects on attitudes towards the necessity of parties. We look at the significance of age in more detail in the next section. 5 The knowledge index is based on three factual questions that appear in most CSES surveys; the content of the questions varies according to country. 296 S. Holmberg / Electoral Studies 22 (2003) 287–299 7. The young and the old Age is of particular interest in this analysis for two reasons. First, people’s age is necessarily relevant when analyzing something with a bearing on the future. In this instance, the presumption is that what young people think tells us more about things to come than what old people think. Secondly, the relationship between age and opinions about the necessity of parties turns out to be very different across the countries examined here. In some countries, including most of the new democracies in Eastern Europe, young people are more positive about the necessity of parties than old people. In many of the older-established Western democracies, on the contrary, and especially in the USA, young people are less supportive of the idea of parties than older people. The results of the analysis are presented in Table 3. Diffusion theory suggests that young people are generally less imprinted by old habits and ideas than old people, and that opinion moulding is a slow process (Rogers Table 3 Percentage of respondents who think parties are necessary amongst different age groups, and correlation between age groups and the parties-are-necessary variable Age groups Difference between 18–30 and 61+ age groups Country 18–30 31–60 61+ Romania Hungary Lithuania Poland Israel Mexico Ukraine Netherlands Germany Taiwan Czech Republic Sweden Spain Norway New Zealand Britain Australia Japan USA 86 75 60 69 73 69 56 92 82 80 73 80 79 88 68 68 62 51 41 80 74 54 68 72 68 58 90 80 82 73 81 86 89 71 78 71 63 54 72 68 54 62 69 67 54 87 79 82 76 80 81 89 75 81 78 75 69 +14 +7 +6 +7 +4 +2 +2 +5 +3 ⫺2 ⫺3 ±0 ⫺2 ⫺1 ⫺7 ⫺13 ⫺16 ⫺24 ⫺28 Correlation between age groups and parties-are-necessary variable (r) 0.13 0.08 0.08 0.06 0.05 0.02 0.01 0.00 ⫺0.01 ⫺0.01 ⫺0.02 ⫺0.03 ⫺0.04 ⫺0.05 ⫺0.09 ⫺0.12 ⫺0.14 ⫺0.19 ⫺0.20 Notes: Entries in columns 1–3 are the percentage of respondents who scored 1 or 2 on the parties-arenecessary scale (see Table 1) among the three age groups. Age groups could not be calculated for Argentina. Source: Comparative Study of Electoral Systems. S. Holmberg / Electoral Studies 22 (2003) 287–299 297 and Shoemaker, 1971; Eyestone, 1977). Hence, we should not be surprised by what we see in most of the new East European democracies: old people are somewhat less positive than young people about the idea that parties are necessary to the functioning of the political system in their country. This is most visible in Romania, but it is indicated also in Hungary, Lithuania, Poland, and in the Ukraine—although not in the Czech Republic. The difference in the levels of agreement that ‘parties are necessary’ between the oldest and youngest age group (shown in the fourth column), are not large, but, apart from the Czech Republic, all point in the same direction: generally, in the new East European regimes, it is the young who are most ready to embrace democratic politics as party politics whereas the old appear to be somewhat more wary. In most of the well-established democracies, however, that pattern is reversed: in, notably, New Zealand, Britain, Australia, Japan, and the USA, the idea that parties are necessary to the functioning of a democratic political system is more widely supported among the old than the young. This may seem paradoxical. It is well attested, however, that in the established, or ‘settled’ democracies, party identification tends to strengthen as people age (Campbell et al., 1960, pp. 153–156; Miller and Shanks, 1996, pp. 131–132); and the analysis reported in Table 2 demonstrates that attachment to party is related to the conviction that parties are necessary to the functioning of democratic politics.6 Seen in this light, the results for the established democracies are less surprising. It is more surprising, however, that the relationship between age and the necessity of parties is more pronounced in some established democracies than in others. The relationship is strongest in the USA, with Japan close behind, and followed by the ‘Anglo-Saxon’ countries (Australia, Britain, New Zealand). In the USA, a minority of young people (41 per cent) considers that parties are necessary to their politics whereas a clear majority of older Americans (69 per cent) is of the view that parties are necessary to the functioning of the US democratic system. In all the other established democracies examined in the analysis, age shows only a very slim relationship, or no relationship at all, with attitudes towards the necessity of parties. Among the Dutch, Germans, Swedes and Norwegians. What may be most instructive in terms of age is a comparison between the two polar groups in this analysis: the youngest age group in the Netherlands and the USA. Almost every young person, an astounding 92 per cent, in the Netherlands agrees with the idea that parties are necessary to the functioning of their political system; among young Americans, the comparable figure is only 41 per cent. Indeed, young Americans constitute the only age group in any country in our analysis where only a minority agreed with the notion that parties are necessary. It would be easy to dismiss the result for the USA as yet another example of American ‘exceptionalism’ (Lipset, 1996). But it may be more serious than that. If a majority of 6 However, the relationship between age and the necessity of parties is not a mere reflection of the relationship between age and the strength of party identification. The impact of age on attitudes towards the idea of parties was independent of the strength of party identification in the USA, Australia, New Zealand, and Britain, as well as in Poland and Romania. 298 S. Holmberg / Electoral Studies 22 (2003) 287–299 young people in the oldest democracy in the world is not convinced that political parties are necessary, we should perhaps take notice. And especially so since many international trends in behaviour and life styles start among young Americans. 8. Party identification matters The concept of party identification originated with the ‘Michigan Four’ to account for the stability of voting behaviour amongst the American electorate. Warren Miller, in particular, was engaged with the concept through to the end of his career (Miller, 1999). And we have seen in this analysis that party identification can do more analytic work than contributing to explanations of the behaviour of voters. The relationship between supporting the notion that parties are necessary to the functioning of a democratic political system and strength of party identification is present in all kinds of democracies: the well-established and the more recent; in Western and Eastern Europe; in countries with long democratic roots and in countries with recent experience of authoritarian rule; in countries with well-sedimented traditions of stable representative government and countries in which party politics is volatile or, as yet, incompletely realised. Admittedly, our CSES data are limited to some 20 countries. But in all of them the pattern is much the same: people who strongly identify with a party tend to be much more supportive of the notion that parties are necessary to the functioning of the political system in their country than people without party identification. This is also, by and large, the case among people who are weakly attached to a party. Hence, if Dalton’s reasoning is correct, and we have no evidence to the contrary, party identification is also an important tool for examining the health, and outlook, for democratic political systems. Parties are the major actors in democratic political systems; if their place in the minds of their electorates weakens, we may, indeed, see the eventual downturn in support for democratic regimes. Our results, of course, indicate perhaps a telling irony: it is in the home country of the concept of party identification that the most potent results are evident. In no other established democracy is support for the necessity of parties as weak as in the USA; and in no other established democracy is the relationship between the strength of party identification and attitudes towards the necessity of parties as strong as in America. Thus, the most striking finding of our empirical test of Dalton’s contagion thesis is that its strongest support is found in America. As Dalton suspected, declining attachments to parties may indeed lead to a weakened support for party-based democracy. And if it happens it may happen first and fastest in the USA. Mass political parties were first born in America; they may well die first in America. References Budge, I., Crewe, I., Farlie, D. (Eds.), 1976. Party Identification and Beyond. Wiley, New York. Campbell, A., Converse, P., Miller, W., Stokes, D., 1960. The American Voter. Wiley, New York. S. Holmberg / Electoral Studies 22 (2003) 287–299 299 Dalton, R., 1999. Political support in advanced industrial democracies. In: Norris, P. (Ed.), Critical Citizens. Oxford University Press, Oxford. Duverger, M., 1954. Political Parties: Their Organization and Activity in the Modern State. Methuen, London. Easton, D., 1965. A Systems Analysis of Political Life. Wiley, New York. Easton, D., 1975. A reassessment of the concept of political support. British Journal of Political Science 5, 435–457. Eyestone, R., 1977. Confusion, diffusion, and innovation. American Political Science Review 77, 441– 447. Holmberg, S., 1994. Party identification compared across the Atlantic. In: Jennings, K., Mann, T. (Eds.), Elections at Home and Abroad. Essays in Honor of Warren E. Miller. University of Michigan Press, Ann Arbor. Katz, R., Mair, P., 1995. Changing models of party organization and party democracy: the emergence of the cartel party. Party Politics 1, 5–28. Kirchheimer, O., 1966. The transformation of the Western European party systems. In: LaPalombara, J., Weiner, M. (Eds.), Political Parties and Political Development. Princeton University Press, Princeton. Lipset, S.M., 1996. American Exceptionalism. A Double-Edged Sword. Norton, New York. Mair, P., van Biezen, I., 1999. Trends in Enrolment in Political Parties in European Polities, with Particular Reference to Youth Enrollment. Stockholm: International Institute for Democracy and Electoral Assistance (IDEA). Miller, W., 1999. Temporal order and causal inference. Political Analysis 8, 119–140. Miller, W., Shanks, M., 1996. The New American Voter. Harvard University Press, Cambridge. Olson, D., 1998. Party formation and party system consolidation in the new democracies of Central Europe. In: Hofferbert, R. (Ed.), Parties and Democracy. Blackwell, Oxford. Panebianco, A., 1988. Political Parties: Organization and Power. Cambridge University Press, Cambridge. Rogers, E., Shoemaker, F., 1971. Communication of Innovations: A Cross-Cultural Approach. The Free Press, New York. Schmitt, H., Holmberg, S., 1995. Political parties in decline? In: Klingeman, H.-D., Fuchs, D. (Eds.), Citizens and the State. Oxford University Press, Oxford. The Comparative Study of Electoral Systems, 2000. CSES Office, Institute for Social Research, Ann Arbor. Widfeldt, A., 1999. Linking Parties with People? Party Membership in Sweden 1960–1997. Ashgate, Aldershot.