Abstract
Book Reviewed
Clarke, Harold; Sanders, David; Stewart, Marianne; and Whiteley, Paul (2004) Political Choice in Britain. Oxford: Oxford University Press. ISBN 0-19-926654-9, 371pp.
The unbroken series of British Election Study (BES) national surveys of the electorate began 40 years ago at the time of the 1964 general election. From then on, the publication of the main BES report—recently one on each individual election—was always a major event for academics in the field, awaited, one might say with only slight exaggeration, with bated breath by enthusiasts.
Earlier studies did not disappoint. Not only did they provide masses of useful data but in theoretical terms they defined the agenda for British electoral studies. Political Change in Britain by David Butler and Donald Stokes (1969), which covered the elections of 1964 and 1966 and was extended to include the 1970 election in the second (1974) edition, was a landmark. The authors developed what came to be known as ‘the Butler-Stokes model’ of party choice. This emphasised the role of long-term factors such as class and party identification in determining vote and suggested that short-term factors, including issue opinions and evaluations of party leaders, had only a marginal impact. The emphasis was on the stability of party choice rather than change, which was slow and gradual. Although there was some criticism at the time and the events of the electorally volatile 1970s soon cast doubt on at least some aspects of the model, the work of Butler and Stokes remained, nonetheless, the essential starting point for all subsequent survey-based electoral analysis in Britain.
The next major study, Decade of Dealignment, by Bo Sarlvik and Ivor Crewe (1983), propounded what shortly became a new orthodoxy. Based on surveys at the two 1974 elections and in 1979, the argument was that there had been a substantial weakening of class voting and of the strength of party identification amongst the electorate. Rather than being broadly aligned with parties on the basis of class and socialisation, a large proportion of electors were now ‘dealigned’. The opinions of electors on issues and their evaluations of party performance had become much more important in explaining why people voted as they did.
In turn, this orthodoxy was itself challenged by the team appointed to undertake the 1983 BES. In How Britain Votes, Anthony Heath, Roger Jowell and John Curtice (1985) argued that there had been no class dealignment and that class voting was, as it were, alive and well. This striking and surprising conclusion was based on a redefinition of class categories and the use of a novel measure of class voting, and it led to furious debate in articles and at conferences over the next 10 years or so. Heath and colleagues (1991) returned to the fray in their post-1987 volume. With six named authors, this book (Understanding Political Change) focused on change between 1964 and 1987, but there was little by way of theoretical innovation (other than in regard to very specific topics, such as tactical voting).
The same was true of the Heath team's reports on the 1992 and 1997 elections—Labour's Last Chance? (Heath et al. 1994) and Critical Elections (Evans and Norris 1999). Both, for the first time in the series, were edited volumes. The first, as the title implies, focused on a relatively narrow issue—the electoral decline of the Labour party—and concluded that Labour's prospects of winning an overall majority in parliament in the next election were poor. As it turned out, of course, this was about as wrong as it was possible to be. The second comprised 14 articles by different authors on a variety of topics only loosely linked by a focus on long-term electoral change and the concept of critical elections. Heath appeared only as a joint author of three articles, Curtice as joint author of one and Jowell not at all. Both books, of course, were hugely valuable and interesting, but neither could be said to have prompted new directions in research and it seemed that, for all their excellent work, the Heath team had rather run out of theoretical steam.
Responsibility for the BES at the 2001 election passed to a new team, comprising Harold Clarke, David Sanders, Marianne Stewart and Paul Whiteley (henceforth CSSW), based at the University of Essex. The title of the resulting book, Political Choice in Britain, has strong echoes of Butler and Stokes and it is a major achievement which is a worthy successor to the earlier BES classics. CSSW examine and evaluate theories and models of party choice using the entire series of BES surveys, use the 2001 surveys to put forward a comprehensive explanation of the outcome of that election and provide a detailed consideration of why people vote (or do not). In addition, the book promotes a ‘big idea’ that will assuredly influence future research on electoral behaviour and will significantly modify our understanding and interpretation of elections.
The ‘big idea’ is ‘valence politics’. Butler and Stokes first introduced the concept of valence to distinguish between issues on which the public take different sides (‘position’ issues) from those—such as crime, corruption or peace and prosperity—on which (almost) everyone is agreed about the end to be pursued (‘valence’ issues). When position issues are analysed in relation to party choice, the key questions concern the extent to which voters actually have a position, can locate the positions of the parties and vote for the party that is perceived as being closest to their own position. What is important about valence issues, however, is how voters judge the relative competence of the parties to achieve the desired ends or evaluate the performance of the government in doing so.
The notion of valence has been part of the vocabulary of electoral studies ever since and was further elaborated in a later article by Stokes himself (1992). The idea that voting choices are now more affected by evaluations than used to be the case is also not new. Sarlvik and Crewe, as noted above, were already heading in that direction and I myself used the term ‘judgemental voting’ in the early 1990s to characterise electoral behaviour in the post-dealignment period (Denver 1994, 101–102). Now, however, CSSW define a fully-fledged valence politics model of party choice and argue that this not only provides the best explanation of how electors decide to vote now but, contrary to what most analysts have thought, has always (at least since 1964) done so. They say:
In our view, the most important factor underlying electoral choice is valence—people's judgements of the overall competence of the rival political parties. These judgements, in turn, are arrived at through two principal and related shortcuts: leadership evaluations and party identification (p. 9).
The model always has been as, or more, compelling statistically as either models in the sociological framework or the issue proximity model (p. 63).
For all elections between 1964 and 1997 (although only the analyses for 1966, 1992 and 1997 are reported in detail) CSSW compare the explanatory power of the valence politics model with that of ‘social determinism’—the view that vote is determined by social characteristics, especially class—and the classic issue voting model derived from Downs. They are especially dismissive of the former. Not only were social characteristics weakly related to party choice in 2001—‘those wishing to understand electoral choice in present-day Britain must look elsewhere’ (p. 123)—but, more controversially, it is suggested that social class was never a particularly strong influence on party choice and, even in the 1960s, it was certainly not as centrally important as claimed by Butler and Stokes.
All of this, it must be said, is brilliantly argued and the supporting statistical analyses and marshalling of evidence are hugely impressive. Nonetheless, the argument raises a number of issues which will provoke much discussion amongst election specialists and I will comment on three of these.
To take a simple point first, few would have problems with the evidence that the influence of social characteristics, and class in particular, on party choice has declined, but if it was never very important then one wonders how to explain the fact that the class composition of constituencies used to be a very powerful predictor of election results. In the 1966 election, for example, of the 100 constituencies with the largest percentage of manual workers Labour won 99, and of the top 200 Labour took all but 11 (mostly very rural seats). Of the 100 seats with the largest proportions of professional and managerial workers Labour took only four. The results of survey analysis (not for the first time) seem somehow at odds with the real world and it must be doubted whether a theory such as that of a ‘neighbourhood effect’ can entirely explain the disjunction.
The second issue relates to party identification. Despite all the criticism that has been heaped on the concept and its measurement, and notwithstanding the clear decline in the strength of partisanship in Britain, party identification stubbornly remains a statistically important influence on party choice in elections. In their ‘composite model’ explaining voting in 2001, for example, which incorporates variables from a number of individual models, CCSW find that party identification remains highly significant (pp. 109–110). Initially at least, this would appear to be something of a problem for their argument since, as traditionally conceived, party identification has little to do with judgements or evaluations. Rather, it reflects a sort of ‘tribal’ loyalty—an enduring attachment to a party produced by the individual's socialisation experiences, particularly in the family.
In order to encapsulate party identification within their valence framework, CCSW reconceptualise the notion. Following Fiorina (1981), they suggest that identification is ‘a storehouse of accumulated party and party leader performance evaluations’ (p. 211). ‘Valenced partisanship’ is continually updated as voters acquire new information, react to events and continuously make judgements about the competence of parties, governments and leaders. This interpretation is underpinned by a variety of innovative and sophisticated analyses in an outstanding chapter on ‘The Dynamics of Party Identification’. Crucially, these suggest that there has always been a ‘significant’ or ‘impressive’ level of individual instability in party identification, that this is not an artefact of methodological problems and that changes at both individual and aggregate level can be explained by changes in other relevant evaluations.
The traditional party identification model that CSSW have in mind to debunk appears to be a rather strict version, however—any individual instability in identification appears to be taken as evidence undermining it—and whether or not a given level of party switching is ‘impressive’ is, of course, subjective. In the data analysed it appears that most survey respondents are stable in their party identification—even in the four-wave panel covering 1998 to 2001 (p. 186). This may be because their updating evaluations serve to confirm their previous identification, but might it not also be because at least a portion of them are actually ‘tribal’ voters of the kind described by Butler and Stokes? Everyday experience confirms that such voters still exist and in constituency terms we still refer to Conservative and Labour ‘heartlands’. It may be a little too early, therefore, to write the obituary of traditional party identification theory.
Perhaps the most controversial aspect of CSSW's operationalisation of valence politics is the key role that they ascribe to evaluations of party leaders. A long tradition of electoral analysis in Britain, going back to Butler and Stokes, has generally downplayed the influence of party leaders on voting choices. This tradition is upheld in a recent piece by John Bartle and Ivor Crewe, who conclude that ‘leaders have not had much of an impact on election outcomes net of prior variables’ (2002, 93). The subtle distinction between impacting on the election outcome and on voters’ party choices means that the two analyses are not exactly comparable but, even so, the contrasting emphasis of the two is clear.
CSSW first explain why, especially nowadays, leaders might be expected to be an important influence on voting—the increased ‘presidentialisation’ of the role of prime minister, media focus on leaders, the fact that election campaigns are leader-centred and the relative ease with which voters can form opinions about leaders. Not many would dissent from this. Secondly, they show that there is a high level of consistency between how people evaluate leaders and how they vote. Again, this is no great surprise and is entirely consistent with the traditional party identification model. Finally, however, on the bases of a series of multivariate analyses, CSSW argue that leadership evaluations remain highly significant even after other variables, including party identification, have been taken into account. They use BES data to show that this holds true for the 1966, 1992 and 1997 elections and Gallup data (focusing on the ‘best prime minister question’) to show that it also applies in the period from 1992 to 2001. They also analyse the 2001 BES surveys (focusing on the extent to which respondents liked or disliked the leaders) to document the importance of leadership evaluations in the last election. The latter analysis is backed up by a series of diagnostic tests, all of which appear to support the main conclusion.
The fact that the Liberal Democrat leader, Charles Kennedy, ended up being the most liked of the party leaders in 2001 (p. 92) whilst his party trailed well behind the others in terms of votes might give one pause for thought, but it has to be said that the assorted high-powered statistical analyses on which CSSW's claims are based appear to provide powerful supporting evidence. There may be technical and methodological difficulties of which I am unaware; alternative explanations may be possible. For the moment, however, a clearly revisionist view of the influence of party leaders on voting choice has been set out with great rigour (and vigour). In the piece referred to above, Bartle and Crewe say, in respect of the influence of leaders in elections:
Journalists believe it. Politicians believe it. Even voters themselves believe it—at least of other voters. Unfortunately, no one cares what political scientists believe—only what they can demonstrate (2002, 93).
CSSW have provided a very convincing demonstration and brought political scientists into line with journalists, politicians and voters (and also, one might add, party strategists). It is now up to critics of their position to respond.
As well as reviewing the bases of party choice over the whole period from 1964, CSSW pay particular attention to the 2001 election. Unsurprisingly, they argue that valence considerations largely explain the election outcome—Labour triumphed because the party was judged more competent than its rivals (especially on the economy), focused on the right issues and had the most popular leader. Media commentators might have ventured the same opinion without the need for election surveys and the statistical pyrotechnics displayed here, of course, but CSSW provide an empirically-based account that is irrefutable.
In this context, the book also includes an innovative and ingenious chapter on the impact of the election campaign. For some time, the orthodox view among academics—in line with the model of stable voting propounded by Butler and Stokes—was that campaigns make little difference to voting choices. Despite growing evidence to the contrary, the BES study of the 1997 election (Evans and Norris 1999), for example, makes no mention at all of the election campaign. If party choice is a product of evaluations and judgements, however, then one might hypothesise that campaigns would be more important, since voters are more open to persuasion. So it proves. Making use of both the pre-election and post-election survey waves, CSSW find that the party choice of individual respondents was significantly influenced by the campaign contacts made by parties (although recall of such contacts may be subject to selective retention) and also by the amount spent by the parties in the constituencies concerned during the campaign (which is entirely exogenous to the surveys). Campaign effects may not be massive but they are nonetheless real. The old orthodoxy, one would think, must surely now be abandoned.
In another departure from previous BES reports, the current volume provides a detailed account of non-voting. Given that turnout in the 2001 election slumped dramatically and that this subsequently became a subject of intense interest amongst the political classes, the decision to include questions in the surveys that would allow extended analysis of the topic was either prescient or lucky. In some ways, however, the treatment of turnout here is rather disappointing. No less than six models—all with fancy names, such as ‘civic voluntarism’, ‘cognitive mobilisation’ and ‘general incentives’—purporting to explain political participation in general are described and discussed. This is done at a rather abstract level which is difficult to relate to the simple question of why some people voted and others did not. The models are then duly specified in detail and applied to voting in 2001. None provides a very satisfactory results (the largest pseudo-r2 is 0.16) but all have something to contribute to explaining the decision to vote or not. None of this leaves us very much the wiser. It is not clear that knowing that people who are more interested in the election, know more about politics and pay more attention to the campaign on television are more likely to vote takes us very far in understanding non-voting.
On the other hand, the analysis of changes in turnout over time is lucid and entirely convincing. The pattern of aggregate turnout in post-war elections is largely explained by the expected closeness of the election in question and the ideological distance between the parties. These factors do not entirely explain the slump in 2001, however, and CSSW go on to consider age differences in the propensity to vote. These differences are well-established and are usually explained in terms of a life-cycle effect: as people grow older they acquire more responsibilities, become more involved in their communities and so on. Here again, however, CSSW take a distinctive view. Using the BES surveys since 1970 they argue that there is little evidence of a consistent life-cycle effect. Rather, there is a clear generational effect—the ‘Thatcher’ and ‘Blair’ generations who came of voting age in 1979 and after are less likely to vote than older generations and the reason for this is that they have a weaker sense of civic duty. They are less likely than previous generations to think that every citizen has a duty to vote or that non-voting is a serious neglect of one's duty. This suggests that turnout in future elections may continue to be relatively low. It is likely that there will be closer contests and the parties’ policies may become more distinct but, especially if there is no life-cycle effect, it is not easy to see how the post-1979 generation could develop an increased sense of civic duty.
The last substantive chapter of the book moves away from voting to focus on the pessimistic view advocated by some commentators that Britain is experiencing something of a crisis of democracy. The electorate, it is believed, is more apathetic, cynical and disillusioned than ever before. What the chapter demonstrates splendidly is how soundly-based empirical political science can debunk popular myths. Although the decline in sense of civic duty is a worry, it turns out that over 40 years there has been little change in citizens’ levels of political engagement and there is no evidence of progressive disillusion with politics or the political system. Evaluations of elections, parties, parliament and other institutions have not become more negative and, indeed, levels of satisfaction with democracy have risen. In short, there is no need for panic as yet.
Political Choice in Britain is not a book for the statistically faint-hearted. Throughout, models and hypotheses are rigorously specified and then tested using advanced statistical techniques of which binary logistic regression is perhaps the most familiar. Few readers, one suspects, will be entirely au fait with the Akaike Information Criterion, the Bayesian Information Criterion or Mixed Markov latent class models; not many will fully appreciate passing references to ‘parameterisation penalty’, ‘simultaneity bias’ or ‘nearest neighbour’ regression analysis. Sometimes entire passages are impenetrable. Thus:
The simplest model for a long-memory process is: (1-L)dYt = ∈ t where ∈ t is a random shock at time t, L is a backshift operator, and d is a fractional differencing parameter … If d equals 1.0, then the process is a random walk and, hence, nonstationary. If d is greater than or equal to 0.5, but less than 1.0, a long memory process is nonstationary, although it is ultimately mean reverting. As noted above, nonstationary processes invite so-called ‘spurious regressions’ and hence, it is important to determine whether time series variables are nonstationary. One way of doing this is to estimate the value of their d parameters (p. 206).
Got that?
This highlights a dilemma regularly faced by academics who write on elections and electoral behaviour. Should they keep the statistics simple so that their work is widely accessible and non-specialists can follow the evidence as well as the argument? Alternatively, should they use the most appropriate techniques even though they may be difficult and unfamiliar? CSSW obviously take the latter course and no quarter is given even to those who are semi-proficient in statistics. As a consequence, non-professionals in the area would struggle to make much of the analyses around which the book is built. Most undergraduates would flounder, interested media commentators would be dismayed and lay people mystified.
The book is manifestly not written for them, however. It is for fellow-professionals and, as such, can be fairly described as providing a masterclass in empirical political science. To be sure, the careful and detailed specifications of models, the careful laying out of results together with a multiplicity of statistical tests and diagnostics can become a little tedious, even for the keenest elections buff. But this is how it ought to be done. Quite simply, this is a brilliantly executed work which is a superb example of political science at its very best.
After the Labour landslide in the 1997 general election, Philip Gould, strategy and polling adviser to Tony Blair and the Labour party, ridiculed (some) British psephologists. Having noted the lessons that could be drawn from previous election studies, he commented, with heavy irony: ‘Needless to say we based our strategy on precisely the opposite precepts and just managed to scrape home’ (Gould 1998, 3).
The work of CSSW should put an end to remarks of this kind. Although few may fully understand the statistics and some may find the terminology unusual, the fact that party choice is now best explained in valence terms will ring true with those professionally involved in politics as well as others. It is not, of course, the last word—there never is a last word in electoral analysis since another election is always just around the corner. What can be said, however, is that Political Choice in Britain will undoubtedly be the starting point for much future research on British electoral behaviour.