Party Identification Research Paper

1. Intellectual Context

The concept of party identification was originally developed as an integral part of the Michigan social psychological model of voting behavior. Partly through socialization experiences in childhood and in the workplace, individuals come to feel that their own political beliefs and values are ‘closer’ to the core ideological and policy positions of one party rather than another. Some individuals will enter adulthood as fully fledged ‘identifiers’ with a particular party. For others, the process may take longer. But the more frequently that an individual votes for a particular party over time, the more likely it is that s/he will come to ‘feel closer to’ or ‘identify with’ that party: s/he will become a ‘party identifier.’ This, in turn, will reinforce the individual’s tendency to continue to vote for the party, even when other factors might militate against it.

Although party identification is primarily a psychological concept, it also has a direct counterpart in rational choice theory in Downs’ concept of the ‘standing vote.’ According to Downs, voters can develop a habit of voting for one party as a way of reducing the costs of continuously collecting and evaluating political information. Relatively early in their political lives, they make judgements about which party is closest to their own ideological positions and tend subsequently to continue to support that party unless and until some major political trauma causes them to reassess their early judgement. Fiorina’s work on voting in the USA goes some way to combining the psychological notion of affective attachment with Down’s idea of the standing vote. Fiorina argues that party identification is best seen as a running tally, or cumulation, of voters’ retrospective evaluations of the major parties. In short, identification results from a voter’s perception that a particular party has outperformed its rival(s) over time; it disposes the voter to support that party in the current election.

2. Emphases In Current Theory And Research

The concept of party identification has been deployed in four main ways. The first, and most extensive, use is as a ‘nonproximate’ cause of vote choice. Voters are obviously swayed by a variety of different factors in deciding how to cast their votes in elections. These include voters’ perceptions of any triumphs or disasters that occur in the final weeks of a campaign, their perceptions of the candidates’ character; their views of the policy stances of the parties; and their assessments of the parties’ performance in office and in opposition.

For advocates of the party identification approach, a crucial additional factor is the extent to which voters are predisposed—independently of any of the above ‘proximate’ factors—to support one or other of the parties because they ‘identify’ with it. On this account, the effects of proximate causes of the vote can only be properly estimated if the underlying effects of voters’ prior partisan identifications have been fully taken into account. Concomitantly, explanations of voting choices (and election outcomes) that underplay the importance of long-term party identifications risk overstating the significance of the more proximate factors that explain voting choices. In all this, there is no suggestion that party identifiers always vote for the party with which they identify. On the contrary, it is entirely possible for even strong identifiers on occasion to vote against ‘their’ party because other more proximate factors weigh more heavily in their vote calculations. In US presidential elections, for example, between five and twenty percent of identifiers typically vote ‘against’ the party they identify with. For these voters, party policies, party performance or the leadership qualities of the rival candidates (or some other proximate factors) must have weighed more heavily than traditional party loyalties. Party identification theory argues, however, that in subsequent elections ‘defecting identifiers’ are highly likely, other things being equal, to bring their voting behavior back into line with their partisan identity; they return to their traditional partisanship.

The second usage of the concept of party identification follows from the first. If prior identifications dispose citizens to vote in particular ways, then those identifications are also likely to color the way in which voters interpret political information. Party identification in this sense acts as a framing factor—a sort of perceptual filter—which disposes voters to view their party’s policies, its leaders, and its record in a more favorable light than might otherwise be the case. It simultaneously disposes voters to view other parties’ characteristics less favorably. This implies that voting choices are not only affected directly by party identification. These choices are also affected indirectly because of the ‘coloring’ effects of identification on voters’ perceptions of the parties’ competing attractions—perceptions which in turn also influence the way people vote.

A third use of the concept of party identification is made by the political parties themselves in devising electoral strategy. Party strategists are always aware of the need both to mobilize their party’s own identifiers and to appeal to those voters who in the past have not been inclined to lend the party their support. In the early stages of a campaign, the main task is to trigger the support of existing partisans; in later stages it is to convert neutrals and those who identify with competing parties. Measures of party identification give strategists some idea of the size of their ‘core vote.’ This is the minimum level of support that parties can expect to obtain in a general election and against which, once the election outcome is known, their strategists can judge their success in recruiting new supporters. The need to provide participants in the democratic process with accurate information about their performance puts a particular premium on the accuracy of estimates of partisanship levels. For this reason debates about the measurement of party identification (which are discussed below) involve more than just matters of academic interest.

The final usage of the concept of partisanship is as a phenomenon its own right that requires analysis and explanation. To argue, as party identification theory does in relation to any given election, that a major reason why voters vote Republican or Democrat—or Conservative or Labour—is that they are already Republican or Democrat, identifiers does not, on the face of it, seem to explain very much. Indeed, it borders on the tautological. The notion that ‘I vote Republican because I am a Republican identifier’ immediately begs the question as to why I identify with the Republicans in the first place. As implied above, party identification theory’s answer to this question is that partisanship develops as a result of long-term exposure to a variety of socialization processes in the individual’s family, school, workplace, and social milieu.

Note, however, that socialization experiences of this sort fail to provide an obvious explanation as to why some voters switch their identifications (from Republican to Democrat or from Labour to Conservative, for example) midway through their voting lives. Fiorina’s notion of identification as a cumulative updating of retrospective assessments of party performance plays a useful explanatory role in this context. Consider a voter A who, as a result of childhood and early adult socialization experiences, identifies with—and accordingly votes for—the Conservatives at election e1. Assume further that the Conservatives appear so incompetent at the subsequent elections e2 and e3 that A switches his vote to Labour on both occasions. In these circumstances, A may also shift his identification to Labour to reflect the fact that his ‘running tally’ of retrospective performance now favors Labour rather than the Conservatives. Although Fiorina’s analysis does not predict the precise point at which voters will switch their identification from one party to another, it offers a plausible explanation as to how, notwithstanding their early political socialization, voters can change their identifications over time.

3. Methodological Issues: The Problem Of Measurement

A key methodological issue concerns the extent to which the party identification measures developed in the USA in the 1950s can be successfully transplanted to other political contexts. Data from successive US National Election Studies (NES) suggest that roughly three-quarters of US voters are either Republican or Democrat identifiers (the remaining quarter are ‘independents’) and that substantial proportions of voters retain their partisan identities over time. Given that these findings are consistent with the idea of party identification being both widespread and durable, the concept of partisanship has maintained a prominent position in analyses of electoral outcomes and electoral change. Indeed, most published accounts of US voting behavior include either controls for, or estimates of, the effects of partisanship on the vote.

Not surprisingly, when efforts have been made to extend the theories and methods of the NES to other countries, the concept of party identification has been ‘exported’ as a central part of the package. An early example was the inclusion of a party identification question in the 1964 British Election Study (BES) survey. The question was borrowed directly from the NES and it has been employed with only minor variations in question-wording since:

Generally speaking, do you think of yourself as Conservative, Labour, Liberal-Democrat, (Nationalist) or what?

The phrase ‘generally speaking’ is intended to tap the idea that party identification—as opposed to the more immediate and ephemeral notion of current party preference—endures over time; ‘think of yourself’ attempts to capture the respondent’s political identity.

Most analysts of British electoral politics would endorse the idea that party identification plays an important role in explaining UK voting decisions. However, considerable doubt has also been expressed as to whether the traditional partisanship question measures anything very different from voters’ current party political preferences. There are two major reasons for supposing that it does not. First, responses to the standard question outlined above suggest that, since the 1960s, 90–5 percent of British voters have had some sort of party identification. This is clearly a much higher level of ‘measured’ partisanship than in the USA, which should in turn imply much greater stability over time in British voting patterns. Yet electoral volatility since the 1970s has not been noticeably lower in the UK than in the USA. Second, it is clear from British data that major shifts in party identification levels (in 1979, 1983, and 1997) have coincided exactly with major shifts in voting behavior —implying that both, perhaps, are indicators of current political preferences rather than separate phenomena. This conclusion is supported by monthly time-series data collected by Gallup which show that short-term fluctuations in British partisanship correlate very strongly with short-term changes in party support levels. More formal tests, based on the principles of ‘convergent validity’ indicate that, in Britain at least, extant measures of party identification are statistically indistinguishable from measures of voting preference. This in turn implies that models of vote choice which include party identification as an explanatory variable are tautologically mis-specified.

Although these difficulties have not been precisely replicated in all of the countries where partisanship measures have been deployed, similar problems have been encountered in a variety of different national contexts. As a result, many analysts outside the USA have expressed concern that in non-American contexts (where the concept of the ‘political independent’ is less well developed) the standard NES question invites respondents to admit to a partisan identity even when they do not have one. This has led a number of researchers to advocate the use of a ‘filter’ mechanism in measuring partisanship, such as the following:

I am now going to ask you a question about party politics in [country]. Some people think of themselves as usually being a supporter of one political party rather than another. Setting aside your current views about the government and opposition parties, do you usually think of yourself as being a supporter of one particular party or not?

If YES, which one is that?

The key feature of the filter is that it allows survey respondents to indicate from the outset that they do not usually think of themselves as being supporters of a particular political party. Survey experiments using filters of this kind have been conducted in Canada, the UK, the USA, and elsewhere. The measured levels of party identification fall considerably as a result. In Canada, switching from the traditional question to the filter approach reduces measured party identification levels from 70 percent to 40 percent of the electorate; in the UK, identification falls from around 90 percent to under 50 percent; and in the USA it falls from roughly 70 percent to 60 percent. The effects of using a filter, in short, are very considerable outside the USA, implying that levels of party identification in other countries may have been significantly overestimated in much academic research where the traditional NES question has been imported for local use. This is obviously important in purely descriptive terms but it also suggests significant measurement error problems in any statistical models that (outside the USA) have used party identification measures.

The problems of measuring party identification effectively do not stop here, however. A critical feature of partisanship is that it is, for the individual voter, an enduring feature of his or her political makeup. Partisanship should therefore be quite stable at the individual level over time, which implies that an individual should respond to the party identification question in a consistent way over time. Since 1991, the British Household Panel Survey (BHPS) has been collecting annual data on party identification in the UK using the sort of ‘filter’ question outlined above. In every wave of the panel, under 50 percent of respondents have indicated that they have a party identification. In the general election year of 1997, for example, 28 percent of the sample were Labour identifiers, 14 percent were Conservative identifiers, and 4 percent Liberal Democrat identifiers. As intimated earlier, these levels are clearly much lower than those recorded on the basis of the traditional NES question wording. But the measurement problem extends much further than this. Party identification is supposed to be a stable and enduring characteristic of individual voters. The BHPS reinterviews the same respondents in successive annual waves of the survey. We can therefore inspect the data to see how far individuals retain the same party identifications over time. As Table 1 shows, even if we consider only a 9.5 percent of respondents consistently responded as Conservative identifiers and only 7.9 percent consistently as Labour identifiers. In waves 5, 6, and 7 (1995–7), only 6.9 percent were consistent Conservative identifiers and 12.5 percent Labour identifiers. The remainder (excluding the trivial number of consistent Liberal Democrats) were either consistent nonidentifiers (33.8 percent) or else switched identifications across waves (45.8 percent). In short, only around 20 percent of British voters in the 1990s were ‘genuine’ party identifiers in the sense that the measuring instrument used: (a) allowed them to indicate from the outset that they had no particular attachment to any party; and (b) required them to declare the same general partisan commitment on three (and only three—the figures are even lower if the length of the rolling window is extended!) successive occasions.

Where does this leave us? Party identification is a theoretical construct that makes a great deal of intuitive sense. It tells us that individuals have long-term, enduring attachments to political parties and this idea in turn helps us to understand why individuals often vote for the same party in successive elections. Yet when we try to measure partisanship, things become far less clear. Generations of American scholars have happily employed the concept, confident that the NES question wording effectively captures the enduring nature of the partisan’s affective commitment to his or her party. However, data from other countries—and especially from Britain—casts considerable doubt on the use of the concept in other national contexts. In the UK, large numbers of voters—much larger than are revealed by the NES question—appear to have no commitment to any political party. And of those that do, less than 20 percent retain that commitment over a three-year period. In short, if party identification denotes an enduring affective attachment to a particular political party, then partisanship in the UK at least is very much a minority predisposition.

Future research which seeks to employ the notion of party identification will need to address two major issues. The first, discussed above, is the need to develop operational empirical measures of the concept that are both durable and have substantive meaning in the specific national contexts in which they used. Critically, these new operational measures must distinguish effectively between voting choice and partisanship. The second issue concerns the impact of the longer-term process of ‘partisan dealignment.’ It is generally acknowledged that, however partisanship is measured, levels of party identification have gradually fallen in most democracies since the 1970s. Electorates have become increasingly volatile and voters increasingly prepared to behave like ‘electoral consumers,’ picking and choosing between parties on the basis of the rival policy and leadership packages on offer. In these circumstances, as Fiorina observed, partisan attachments acquired early in voters’ political lives are likely to be less important in determining voting choices than the more temporary ‘brand loyalties’ associated with the perception that ‘party X currently appears more competent in managerial terms than party Y.’

Bibliography:

1. Butler D, Stokes D 1974 Political Change in Britain: The Evolution of Electoral Choice, 2^nd Macmillan, London
2. Campbell A, Converse P, Miller W E, Stokes D 1966 Elections and the Political Order. Wiley, London
3. Fiorina M P 1981 Retrospective Voting in American National Elections. Yale University Press, New Haven, CT
4. Miller W E, Shanks M 1996 The New American Voter. Harvard University Press, Cambridge, MA
5. Sinnott R 1998 Party attachment in Europe: methodological critique and substantive implications. British Journal of Political Science 28: 627–50