Promoting public understanding
of electoral democracy
Book Review by Ron Johnston: Paul Whiteley, Harold D. Clarke, David Sanders and Marianne C. Stewart, Affluence, Austerity and Electoral Change in Britain
Paul Whiteley, Harold D. Clarke, David Sanders and Marianne C. Stewart, Affluence, Austerity and Electoral Change in Britain. Cambridge: Cambridge University Press, 2013. ISBN 978-1-107-02424-3 hardback, £55.00; 978-1-107-64116-7 paperback, £19.99.
The British Election Studies based on sample survey data were launched by David Butler in the 1960s, and every general election since has been the subject of a major investigation building on that pioneering work. Much has changed in the intervening five decades, however, not only in the political and electoral context but also the theory and methodology of voting studies.
The theoretical underpinning of Butler and Stokes’ classic Political Change in Britain (1969), reporting on the first studies, was largely sociological, emphasising the role of class and party identification as influences on voting behaviour, but incorporating a range of other factors in nuanced analyses of local and other variations. The successor team also adopted that foundation, but the title of their summary volume – Särlvik and Crewe, Decade of Dealignment (1983) – suggested its demise, as voter attitudes and behaviour became more volatile. The next team to conduct the surveys argued for the continuing relevance of the class-based model, however: in Heath et al. How Britain Votes (1985) they claimed that social class, housing, and education structured group interests around which political activity was organised, with the parties shaping those interests and views of how the state should respond. By the time of the next election, however, their book Understanding Political Change (1991) included a chapter on ‘the withering away of class?’ and they paid greater attention to voters’ attitudes to specific issues irrespective of their class or other background characteristics – an argument developed earlier by Rose and McAllister in their Voters Begin to Choose (1986).
That BES team’s summary volume – Heath et al., The Rise of New Labour (199?) – summarised the breakdown of the class-voting nexus as British society entered a post-industrial era: their analyses also continued use of what is widely known as the Downsian spatial model of voting, whereby parties are located on ideological scales (such as reduced government spending with lower taxes at one pole and higher taxes and greater welfare expenditure at the other) and voters support the party perceived as closest to their own position.
For the fourth group to continue the studies – covering the 2001, 2005 and 2010 elections (their first two books were Political Choice in Britain  and Performance Politics and the British Voter ) – the sociological theory was at best obsolescent, and in a series of three volumes they have directed their focus very strongly on an alternative – valence theory. This had been referred to in the Butler/Stokes volumes (Stokes was one of the first authors identified with the theory) but which they now brought to centre stage as a replacement for the, as they saw it, failed social class theory. For them there are now few strong and enduring bonds between members of different socio-economic classes and individual parties. Instead, voters’ main goals are to elect governments that deliver on the main political issues of the day – which are almost invariably dominated by economic concerns (employment, inflation and general prosperity). They vote for whichever party they think is best able to deliver on those issues. The parties may differ on the means but not the end, so governments that have overseen a period of prosperity tend to be re-elected whereas those who have not lose, if there is a viable alternative available.
Apart from this major shift in theoretical emphasis, the three studies conducted by this team of four based at the University of Essex have also made major methodological changes, enabled by new methods of collecting, collating for analysis and rapidly disseminating data via the internet. There is still a face-to-face questionnaire survey of a small sample of c.2,000 voters to maintain continuity with the previous studies, but most of the data analysed were obtained through a monthly Continuous Monitoring Survey – inaugurated in 2004 and maintained into 2012, with a sample averaging 1,050 individuals – plus a large (c.17,000) panel study – the Rolling Campaign Panel Survey – whose members responded to internet-delivered survey instruments before, during and immediately after the month-long campaign preceding election day in May 2010. This approach is vigorously defended in the introductory chapter to Affluence, Austerity and Electoral Change in Britain, not just on cost-benefit grounds in terms of data collection but also – and to the authors much more importantly – because it allows the dynamics of voter attitudes and partisan leanings to be explored over the years, months and weeks before election day, when changing contexts can stimulate variable responses in ways that a single cross-section cannot identify. Continuing the CMS after the election allowed voter reactions to the new (coalition) government and economic context to be uncovered – along with analyses of specific events such as the 2011 referendum on changing the voting system from First-Past-The-Post to the Alternative Vote.
Having successfully defended the approach and briefly outlined the political context in the 2005-2010 inter-election period, the remainder of the book comprises a series of linked essays testing models based on the valence theory, but with some recognition of alternative theories – mainly to conclude that they provide little explanatory power. The presentation is detailed, with a large number of valuable graphics plus tables reporting rigorous statistical analyses, details of some of which will be beyond the general reader but the conclusions drawn are clearly stated.
The first analytical chapter considers the period 1997-2007 – Blair’s premiership – and establishes the main features of the argument that occupies much of the rest of the book: people supported Labour because they believed in its economic competence (relative to the Conservatives) and its positions on the other major issues, plus their positive attitudes to Blair (for most of the time) – leader images are presented as important short-cuts to party evaluations for many voters. The next chapter turns to 2005-2010, in particular the period after Gordon Brown became Prime Minister in 2007: confidence in Labour’s competence declined, Brown rapidly developed a negative image – and defeat at the polls looked increasingly likely.
All of these preliminary analyses are presented as very clear confirmations of the valence theory’s relevance, indeed predominance, in providing accounts of attitudinal trends. The sociological theory is found very largely wanting, and the spatial theory – where both parties and the electorate are divided on policy goals as well as means – also has little apparent impact. The analyses are rigorous and the conclusions valid, but one aspect of the model’s operationalization raises queries. A key variable in these – and most of the book’s later – analyses is party identification. This is a well-established concept drawn from American political science – used by Butler and Stokes in the original BES – which depicts voters as having partisan identifications, enduring links with a particular party having determined – through family history, class locations and other influences – that it is the one they will usually support. As earlier BES analyses had shown, the class basis of British electoral behaviour has declined very significantly and such long-term attachments no longer dominate many people’s political attitudes and behaviour. They are more volatile in their evaluations of parties, and to reflect that Whiteley et al. adopt an approach from rational choice theory that they describe as a ‘running tally’. People’s attitudes to parties change over time – sometimes over short periods only in the light of particular events (such as the Iraq war and Labour in 2003 and student fees and the Liberal Democrats in 2011). Positive evaluations can strengthen voters’ feelings about a party, negative evaluations can weaken them – and even lead to a change in favoured party. Those changes are not identified directly. Instead, the survey instruments ask respondents ‘Generally speaking, do you think of yourself as Labour, Conservative, Liberal Democrat or what?’ and those who pick a specific party are said to identify with it. This is not party identification as many interpret it, therefore (or partisanship as frequently deployed in the book), but rather a measure of ‘current party preference’ and it is unfortunate that the same terminology is used for two separate concepts.
The next chapters turn to the 2010 election, covering in turn: the campaign period – with a major focus on leader images, and particular reference to the three televised debates and their short-term impact on voting intentions; and then voting at the contest itself. Once again, the conclusions are clear: the valence model provides the major contributions to appreciating the patterns of party choice, with party identification, leader images, and party positions on the respondents’ main issues the main determinants.
The conclusions are consistent with theory, model and analytical results. But alternative interpretations can be suggested. For example, the sociological model is comprehensively dismissed because most of the variables representing it – age, education, gender, income, social class, union membership, employment sector etc. – are only weakly linked to party choice in the regression equations relative to the strength of the links associated with valence model variables – leader images, partisanship, party policy positions. This seems somewhat at odds with some of their findings. Table 5.1, for example, shows that when only the socio-demographic variables are included in a model they correctly classify 71.5 per cent of voters according to whether they voted for Labour or another party; when only the valence model variables are included, that percentage is 87.7; and when all variables are included it is 88.8. Is the first group of variables really irrelevant? And why, if the sociological model is obsolescent, are ‘several sociodemographic variables [incorporated] as statistical controls’ in some of the models (p.202).
This suggests some unexplored collinearity. In one place the authors refer to the well-known ‘funnel of causality’ model, whereby some variables are prior to others as determinants of voting choice (in the classic sociological model, for example, social class would precede – largely determine – party identification). In the valence models, therefore, people from one class may be more likely to identify with one party rather than the others, more likely to prefer one of the party leaders than the others, and so forth – but these relationships get submerged in the model-fitting that includes all of the variables in one analysis. On p.89 the authors summarise their analyses of the 2007-2010 period in a flow diagram that recognises such a ‘funnel of causality’ – but it is never tested as such. Like almost all analysts of British electoral behaviour (the present author included), Whiteley et al. pay passing recognition of the funnel argument, but do not test it with the appropriate method – structural equations. If they were deployed then, for example, the direct and indirect influences of class and the other sociological variables could be evaluated; perhaps the sociological model is not as obsolescent as we are led to believe?
The remaining chapters move beyond the 2010 election. Chapter 6 uses the CMS to assess voter attitudes after the coalition government had been formed and its austerity programme implemented, once again validating the valence model: which party respondents supported reflected their evaluations of the economy and the performance of the parties and their leaders. The next chapter turns to the AV referendum: some voted according to their opinions on this form of electoral reform; others relied on the party leaders’ positions and whether they trusted them.
The eighth chapter is innovative in that it revisits the valence model but with a new intervening variable – life satisfaction. The underpinning argument is that citizens may not respond favourably to some policy initiatives, so if they want to be re-elected governments should focus on policies that have the largest positive impact on public well-being. Questions asking people how satisfied they are ‘with their life as a whole’ are used to address this, alongside others asking whether CMS respondents had any direct experience of a particular policy area – such as treatment by the NHS – and whether they were satisfied with the outcome. Analyses indicate that those most satisfied are most likely to support the governing party – but they also tend to feel more satisfied if the party they voted for is in government, an endogeneity problem that requires further analysis (the funnel of causality again).
Finally, Chapter 9 turns to the future, assessing electoral prospects in the context of valence and austerity politics, and setting the British situation in a wider European context. In many ways this is the most difficult of chapters for readers unused to the modelling strategies deployed, but the conclusions are clear – and largely uncontroversial: whether one or both of the parties that formed the government in 2010 will retain power after the 2015 contest will depend on the success of the government’s economic policies and their impact on subjective well-being. The policies may succeed (i.e. in reducing the deficit, keeping inflation low and reducing unemployment) on their own terms but may not be judged to have by voters who feel that their well-being has not improved – real wages have declined for many since 2010.
This conclusion continues the stress on the importance of valence politics, linked to subjective well-being, but …. One of the difficulties with the valence model relative to its competitors, especially the spatial model, is one of cause-and-effect. Is the valence model winning out in the analyses reported in Affluence, Austerity and Electoral Change in Britain because there is no alternative? If the parties converged in true Downsian fashion over the last few decades – thereby enabling Blair’s 1997 and 2001 substantial victories – then if they are indistinguishable ideologically another means of discriminating among them is needed, hence the valence argument. But if the parties diverge – and some of Miliband’s policy proposals as the long campaign for 2015 got under way in late 2013 suggest that is now happening, with Cameron moving rightwards on several issues to counter the UKIP threat – then maybe the spatial model will become more relevant?
This is an incredibly rich book in its analytical contents: a great deal of testing has been packed in to it – and all undertaken and written-up within little more than 18 months of the election. The authors claim to have made six major innovations: their theoretical and empirical extension of the valence model; the use of ‘fast and frugal heuristic’ models to account for the short-cuts voters take when making electoral choices (as with the use of leader images); the detailed analyses of campaign dynamics; the CMS data; the study of the AV referendum; and the introduction of subjective well-being as a linking variable within the valence model. The sheer volume of work reported almost overwhelms: certainly it cannot all be fully considered in a single review. There are many points of detail where the adopted strategies can be challenged – as with the use of tactical voting as an independent variable with no recognition of the local context: indeed, apart from some use of regional dummy variables – largely unjustified; much research has shown that the important variations occur at much finer-grained scales – the important geographical variation in the decision-making contexts goes largely ignored.
In its own terms, this book is a major success, overwhelmingly validating the authors’ chosen approach to understanding recent British political attitudes and voting behaviour over the three BES that they have conducted and reported on. As such, it will have a major impact. In the short run this might be on the authors’ own terms, but the many issues that it raises, theoretical and operational as well as analytical, may mean that in the longer term it acts as a major stimulus to rethinking the relevance of the various theories. To achieve that is a major contribution, as is the massive innovation that the authors have made in data collection – as well as its early dissemination to all would-be users of their data. The next generation of BES studies will build upon this firm foundation, and as the series enters its sixth decade with a new team of investigators, its contribution to this jewel in the crown of British analytical social science – sustained by the ESRC – will become even further embedded in contemporary practice.
University of Bristol