Does party membership really not matter? in search of an electoral model for mass parties

Does Party Membership Really Not Matter? In Search of an Electoral Model for Mass Parties Joan BOTELLA Universitat Autònoma de Barcelona Working Paper n.3 Barcelona 1989 «I prefer five minutes on television rather than ten thousand party members». (A. Guerra, during the 1979 General Election campaign). As shocking as this statement may seem if pronounced by the Vice-President of a Socialist government (and even more if we take into account the fact that PSOE was in opposition in 1979), it certainly expresses what a scholar has called «the prevailing orthodoxy» on the relative influence of party membership and mass media on voters’ behaviour (HILL, 1974: 215). This largely held view, among both politicians and scholars certainly suggests that the days of mass parties are over. The argument is well known, and it includes such elements as the enfranchisement of all citizens, the changes in social and economic structures in Western societies, the growing de-ideologization of politics, the mass media revolution and their use in politics, the public financement of parties and electoral campaigns, and so on. The political implications of these new facts (mass dealignment, increased electoral volatility, etc.) would make traditional mass parties decline: their traditional organization and ways of action (membership-seeking, creation of local branches all over the country, development of a network of «close-but-not-political» organizations, ...) would become more and more inefficient, affected by some kind of «technological obsolescence». Of course, the main variable at stake here is party membership: to what extent is it relevant, in order to explain electoral results. In Section I we’ll review some published research, while in Section II some suggestions will be made. I. Some findings against the theory. It is well known that the previously mentioned general interpretation has come under criticisms. On a general level, Stein Rokkan suggested that communications through mass media did not reach the mass public directly, but circulated in a «twosteps flow», as more traditional communication. Therefore an organised network of opinion leaders (i.e., a mass political organization) would still be necessary to ensure a large diffusion of messages. Some empirical evidence may be presented in support of this hypothesis. In the United States, several studies have tried to measure the impact of local party activity on election results. Thus, it has been suggested that up to 5% of vote in a presidential election may be due to local party action (see CUTRIGHT and ROSSI 1958; see also CUTRIGHT 1963, 1964). This figure may reach 10% of the vote if both parties differ markedly in membership and activity at the local level (KATZ and ELDERSVELD 1961). Several studies on Great Britain (see BOCHEL 1971; BROWN 1958; DENVER and HANDS 1974) conclude that personal campaigning by party members plays an important role through mobilization of potential non-voters rather than through conversion of other parties’ sympathizers. This impact would therefore be more important in local elections (where turnout is lower), while marginality (or safety) of a seat shows a less clear relationship. A. H. Taylor tried to measure campaign activity by the economic amount of local expenses, and found considerable influence on results; on the other hand, the growth of non-voting would leave more room for party action influence on vote, even in general elections (TAYLOR 1972). Hill, while severely criticizing Taylor’s methodology, reaches similar conclusions (HILL 1974: 216-217). More recently, Seyd and Minkin have attributed electoral losses in the Labour Party to reduction of membership and party activity (SEYD and MINKIN, 1979). For Italy, an enormous amount of literature has underlined how high membership figures, structured organization and «capillary» influence of DC and PCI are the key elements in order to explain their electoral preeminence. For socialist parties, two interesting studies reach differing conclusions. While Bartolini finds no relation between changes in membership and electoral fate for Western European Socialist parties (BARTOLINI 1983), Wellhoffer builds a model, using as independent variables party membership, trade-union membership and density of local organizations, and predicts electoral results very close to actual results for Socialist parties of Great Britain, Norway, Sweden and Argentina. (WELLHOFFER 1979). Communist parties are perhaps a more interesting case, given their higher homogeneity and the common fact of their exclusion from governments since 1950. No general study has been published on this subject, but a mere look at Table 1, suggests that there is no apparent relation between electoral results and voter/members ratio. However, in cross-regional analyses within nations strong correlations appear. In 1979, correlation between membership and vote, at the provincial level was .84 for the PCI. In Spain, at a regional level (n=17), the figure was .87 in 1979. In our own research work, we have found correlations between communist (PSUC) membership and vote ranging from .71 to .79, according to the level of measurement and the elections under study. An interesting comparison may be made between Spanish Socialists and Communists in the province of Sevilla (which combines industrial areas and a strong presence of agrarian workers with radical traditions). Table 2 presents average electoral result and standard deviation of results, for each party in the 1977 elections, according to whether each party had a local branch before the election, or created it after the election, or had not created one by the end of the year. It appears that PCE was more a mass party than the PSOE in two senses: Communist vote is more dependent than Socialistis on having or not local branches and, secondly, Communist results are more homogeneous where it did not have a local presence while Socialist vote shows a contrary profile: its results were more homogeneous where it was stronger. In other words: local influence was much more important to explain Communist vote than it was for PSOE. What all these results clearly seem to suggest is that party membership, organization and activity still make a difference. Of course, the mass media revolution has entirely changed the «communicational ecosystem» of Western societies, and parties have certainly had to adapt themselves to this new environment. But parties still try to attract new members and to keep the old ones; these members are still politically active in electoral and non-electoral periods as well; and a certain amount of research work has found a visible influence of the variables related to party organisation on vote and the outcomes of elections. Is it possible to draw some lines towards a model to analyse vote as a function of party? Section II will be devoted to this problem. II. Suggestions and proposals. It may be useful to take as a starting point Duverger’s Partis politiques: in spite of the strong criticism it has received, it still stands as an important source of reflections and problems. It is often forgotten how largely Duverger draws on previous work and analyses. More precisely, his distinction between «partis de masses» and «partis de notables» rests explicitly on Ferdinand Tönnies contrasting paradigms of «society» and «community» (Gemeinschaft - Gesellschaft), and the distinction suggested by S. Neumann between representation parties and integration parties (NEUMANN 1932). Significantly enough, when Neumann expanded his tipology 25 years later, breaking his integration parties class in two (total integration vs. democratic integration), this new tipology matched exactly the one proposed by Duverger, which included a third (and rather marginal, in his view) class of parties, which Duverger termed as «Bund» - «Ordre». (See Neumann’s introduction to NEUMANN 1956). Certainly, traditional «partis de notables» are not an important part of Western European party systems anymore. Instead, what seems to be the main dividing line today is the one opposing mass parties to «catch-all» parties; and, it has been repeatedly suggested, mass parties would belong to the past rather than the present. However, we might take a slightly different approach to the problem by thinking of «catch all» and «mass» parties not as two different groups of empirically visible realities but in terms of a «continuum» along which we can place actual political parties. The extremes of the «continuum» would be the «pure» definitions (in a weberian, ideal-type sense) of catch all and mass parties respectively. Figure1 In a weberian approach, ideal types do not have to be realistic descriptions of facts but, rather, highly abstract and theory-oriented concepts. In my opinion, the respective nature of party-voters linkages captures the most important and decisive question in order to clearly differentiate our two ideal-types. Temptative definitions of both might be the following: For mass parties, the sole link between party and its followers (i.e., voters) is the organized action of its members. Its electoral results will depend of (and only of) its organized membership and activity. For catch-all parties, members are not relevant. In a pure definition, their only linkage with voters is the presence of party leadership in the mass media, and the key question is the leadership’s ability to meet the political preferences of voters. Party organization may still be useful for other purposes (e.g. recruitment), but it does not have a linkage function. Of course, no real political party fullfills neither of these two definitions. In the real world, every party may be seen as partially catch-all and partially mass party. However, this mixture will change from party to party, and this is the interesting point: catch-all and mass parties are not two different, closed groups of parties; rather, we may measure how far or how near is a given party from the extremes. Thus, we can build (along the lines of those «pure» definitions) a set of indicators concerning such aspects as programs (general, structured programs, vs. issue-centered manifestos); press (presence or absence of party press); financing (party-raised funds vs. public financement); links with interest groups (stable links vs. occasional support); types of electoral support, in social terms (class, religious group, or some other kind of differentiated support, vs. cross-cleavage support) or in geographic terms (strong geographic contrasts, vs. homogeneous distribution), etc. It is clear anyway, that the closest indicator to the «mass party» pole would be party membership and, more especifically, its relation with vote. However, the measurement and the interpretation of this relation will depend on, i ), hypotheses of influences between membership, vote, and other variables, and ii ) the span of time considered relevant. i ) In a very simple version, general party action on its political and societal environment is what attracts electoral support (fig. 2-1). However, it is also possible that other variables influence simultaneously both party and vote; in this case (fig. 22) the relation between party activity and vote is important, but superficial (spurious). Figure2-1&2-2 ii ) Although party activity may be more easily measured during election campaigns, in our «pure» definition a mass party (and, therefore, its membership) is always active. The picture will then be different if we focus on a short period, or on a longer time-cycle. According to decisions taken on each of these questions, analysis would have a different focus. Figure 3 attempts to summarize the four possible approaches. Figure3 An attempt to measure the influence of party activity on vote during a short period of time would tend to focus on electoral campaigns (upper left cell), as most of the previously mentioned studies have done. In a longer time perspective, however, it would lead to the study of subcultures, of «micro-climates» due to a especially strong and regular party presence and activity (upper right cell). If, on the other hand, research design takes both party membership and vote as depending on other variables, a short term approach would tend to privilege political events (such as a crisis, a change in government, and so on), while a longer time perspective would perhaps underline social (rather than political) factors which would make some contexts favourable (or disfavourable) both for party organization and for its electoral results. A point to be emphasized here is that possibly cross-national comparisons make little sense. The posed relation between membership and vote is obviously undermined by the fact that i), propensity to become a party member varies very markedly between nations and ii), is in fact one of the key variables in the definition of national political cultures while, on the other hand, it may be much more homogeneous within a given country. That would explain why comparative crossnational studies show little or no relation and cross-regional analyses within countries show quite stronger relations, between both variables. An interesting way of looking at the problem is in terms of regression (not merely correlation) analysis. In the simplest model, where only membership (x) influences vote (y) a simple regression equation would read as y = a + bx + U where a and b are the parameters (intercept and slope, respectively) and U, the error term. Leaving U aside, it is tempting to interpret a and b in the following terms: a equals the share of vote in those units without any party organization or membership presence. Therefore we might term in the «national impact» of the party, i.e., the electoral result exclusively due to its general electoral campaign, leaders, etc. On the other hand b may be considered the «multiplying factor» of membership: to what extent, in average, a party member «attracts» additional vote (thus, b may not be negative: a party is voted at least by its members!). Figures 4-1 and 4-2 show an example of the relevance of these parameters: correlation is strong in both, but the type of relation (and, possibly, of party-voter linkage) may be different. (Actually, both are real: fig. 4-1 presents Spanish Communists results, while 4-2 does the same for the catalan region). Figure4 Of course, multiple regression designs should be used in order to include variables as relevant as financement, organizational density, party press, and so on. Some final remarks, in order to discuss how party strategies may be included in this kind of model. First question to be discussed is, as we have already mentioned, the span of time to be taken into account. Conscious party strategies are most visible in electoral, or pre-electoral, situations: this would lead to put an emphasis on the short term. However, parties also follow long-term, strategies. For which of our «pure A» types of parties is each span of time more relevant? At first glance, one would suggest two easy couples, catch-all/short term strategies, and mass/long term strategies. However, the problem is not that simple, and deserves more attention. Secondly, the problem of adaptation to environment (which is crucial for mass parties) requires also some kind of strategy. We should speak, then, about outside directed strategies, and inside directed strategies, aiming at improving party efficiency or perhaps to merely survive in a context which is more and more disfavourable for mass parties. Table1 Table2 Figure 1