Dynamic representation: the rise of issue voting?

Transcription

1 A CRITICAL ELECTION? UNDERSTANDING THE 1997 BRITISH ELECTION IN LONG- TERM PERSPECTIVE Eds. Geoffrey Evans and Pippa Norris CHAPTER THIRTEEN Dynamic representation: the rise of issue voting? by Mark Franklin (Trinity College, Hartford CT) and Christina Hughes (University of Houston) Was the election of 1997 a critical election? The outcome was certainly surprising, but surprise has to do with poor predictions, not with underlying changes in the basis of voting choice. If there has been a critical election in Britain since 1945 we would like to argue that it was the election of 1974 that was critical, not that of In point of fact, the major change that has occurred in British politics in recent years is very hard to date because processes of change took several elections to work themselves out. From one perspective, the change can be dated to as early as 1970, which was the first election in which class voting broke down to any substantial degree, yielding a great many Tory votes from individuals who would in previous elections have voted Labour and (more importantly) many Labour votes from those who previously would have voted Tory (Crewe, Sarlvik and Alt, 1977; Franklin 1982). From a different perspective the change can be dated as late as 1983, which was the first election to show that the party system itself was at risk in the new circumstances of British electoral politics. We choose to focus on 1974 as the really critical election in this sequence partly because 1

2 that was the first election in which the new issue basis of voting choice manifested itself in an election outcome that was clearly different from what would have been possible in what Samuel Beer once called the 'collectivist age' (Beer, 1965). We want to argue that the General Election of 1974 marked the start of a new age in British electoral politics, whereas that of 1997 was just a milestone to mark its progress. In this chapter we are going to do more than just defend the contention that the decline of class voting in Britain was the critical development leading eventually to the surprising outcome of the 1997 British General Election (Franklin, 1985). That would be merely to go over old ground. 1 Instead we are going to paint a picture of evolving British politics in elections starting with that of 1974 that, we will argue, is a picture of dynamic adaptation of parties to voters' issue concerns, and of voters to parties' issue stances - a picture quite different in important respects from any that could have been painted during the collectivist age. We will also suggest that, within a broader perspective, the collectivist age itself can be viewed as an episode in what, in the American context, McEuan, Erikson and Stimpson (1995) have called 'dynamic representation'. What happened with the decline of class voting was that previous limits on the scope of British politics were removed, allowing the issue basis of electoral choice to expand from a single 1 The contention has been disputed by Heath, Jowell and Curtis (1985; 1990) on the basis of several different arguments which in turn were disputed by Crewe (1986) giving rise to a response by Heath, Jowell and Curtis (1987). Franklin (1988) has pointed out that there is no disagreement about the fact that there was lower class voting after 1970 than before, merely about whether this drop should be characterized as a 'decline'. 2

3 dimension into more than one dimension so that dynamic representation became possible on a more general basis than before. The removal of these temporary limits also permitted party support to vary much more greatly than it could have done before, laying the basis for upset elections of every kind - including the surprising landslide of Dynamics of representation One seldom-noted requirement for the proper functioning of democratic institutions is public responsiveness to policy. There is, after all, little reason to expect politicians to pay attention to what the public wants if the public does not pay attention to what politicians do. A responsive public would adjust its preference for "more" or "less" policy in reaction to policy itself, much like a thermostat (cf. Wlezien 1995). This argument has been made most forcibly in relation to particular policy domains such as defense spending in the United States (Wlezien 1996) and integration policies in the countries of the European Union (Franklin and Wlezien, 1997). In the context of this chapter, however, we want to take a broader perspective, more along the lines of Stimpson's Moods, Cycles and Swings (19??) in which American public opinion was characterized as exhibiting broad swings in mood over periods measured in decades. Later on, in 'Dynamic Representation' (19??) this insight was linked to Wlezien's concept of thermostatic control by regarding the changes in mood as manifesting a feedback mechanism by which voters react to the changing policy environment that results from the operations of the political process. 3

4 To put it simply, voters have policy demands that are catered to by parties and candidates. Parties win office in order to meet those demands. If the demands are satisfied then voters do not necessarily want more of the same; and if more of the same is rammed down their throats they can react quite dramatically against those who do the ramming. The outcome of the British General Election of 1997 is best understood in such terms. By continuing to produce Thatcherite policy long after Thatcher had been replaced as Leader, the Tory party showed itself unresponsive to the changing mood of the British electorate and the electorate reacted by deposing them. In a sense, much of British electoral history can be understood in similar terms. The decline of class voting can be seen, as one of the present authors argued in 1985, as a consequence of the delayed realization that the socialist project in Britain had essentially been completed in 1950 and that subsequent Labour governments had not had anything substantive to add that would bring about a 'socialist Britain' (Franklin, 1985: 174). The earlier rise of the Labour Party can do doubt be seen in terms of a reaction against the limitations of 19th Century Liberalism. One could continue to trace the swing of the political pendulum backwards in time indefinitely in such terms. Still, there is an important difference between the swings prior to what Crewe, Sarlvik and Alt (1977) called the 'decade of dealignment' and those that followed. Before 1970, swings that occurred from one election to the next were limited by the fact that established voters 4

5 were 'immunized' against change (Butler and Stokes, 1974). In those days, large changes in the fortunes of political parties had to wait upon the slow emergence of new generations of voters. Small changes from one election to the next could be accomplished by those who were as yet unimmunized, but landslides were possible only in elections that saw large numbers of new voters (see Franklin and Ladner, 1995, for an analysis of the Labour victory of 1945 in these terms). After 1974, however, increasingly larger swings became possible because increasingly fewer voters were immunized against change (Franklin, 1984). Moreover, after 1974, the terms of political discourse started to show their liberation from class concerns, so that not only could the swing of the pendulum become more dramatic in extent, it could also respond to dramatically new issue concerns. The first politician to benefit from this enlargement of the issue space was the leader of the previously tiny Liberal Party, Mr. Thorpe, in This story is not one that is amenable to empirical testing on the basis of data collected in the British Election Studies, which are the foundation of the present volume. However, in the remainder of this chapter we will show that the impact of issues on party choice were consistent with the general picture given above. The evolving issue space of British politics In characterizing the years after World War II as the 'collectivist age' in Britain, Samuel Beer (1965) had in mind a political discourse 5

6 in which other axes (particularly the libertarian-authoritarian axis) had been squeezed out by the dominance of class-oriented concerns. Both the major parties competed along a dimension that assumed a high level of government involvement in social and economic life, differing only in terms of who should be the major beneficiaries of Government largesse. With the decline of class voting, this dominance of classoriented concerns started to erode and new issues sprang up to compete for voter support. In particular, concern for the rights of minorities and women, and devolution of government powers of all kinds, started to distinguish a 'new left' electorate from an 'old left' more concerned with traditional issues of pensions, job protection, poverty, and the like. On the right of the political spectrum, a new concern for school choice and privatization of all kinds started to distinguish a 'new right' electorate from an 'old right' with traditional right-wing concerns about law and order, defense spending, and overly generous handouts to the poor. In other work (Franklin, 1988), one of the present authors has shown how these new issues moved progressively between 1974 and 1983 to distinguish themselves on a dimension at right angles to the traditional left-right dimension. Figure 1 shows the resulting issue space in 1983, defined by factor analysis of 39 issues about which respondents were asked in the British Election Study of that year (see Franklin, 1988, for details of how the analysis was performed). The way to think of this illustration is as depicting issues in proximity to each other when the positions concerned tended to be held by the same people (respondents to the 1983 British Election Study 6

7 survey), and at a distance from each other when the positions concerned tended to be held by different people. Thus race equality is depicted about as far as it possibly could be from prison sentencing, because those who want tougher prison sentences are not at all the same people who are concerned that society should treat different 7

8 races more equally. On the other hand, such concerns tend to be held by very much the same people who believe that prisons should be reformed - and so those two variables are depicted very close to each other. The orientation of the variables on the basis of factor analysis is entirely arbitrary, however, and we have manually rotated the space until the horizontal dimension resembles the familiar leftright spectrum, with social issues on the left and a willingness to increase defense spending at the cost of social spending on the right. Other issues then take up the positions they need to take up in order to retain the proximities discovered by factor analysis. Franklin (1988) also demonstrates that new votes for the Conservative Party in 1979 came primarily from the new right (though also from the new left) quadrants of the issue space. Indeed, the Tories actually lost votes in what should have been their old right stronghold. So Mrs Thatcher's electoral victory in 1979 was the result of capitalizing on new issue concerns - issue concerns made possible by the decline of class voting and the consequential loss of hegemony for collectivist notions. Figure 2 shows locations in the issue space of party leaders, based on the issue locations of voters who thought highly of them, in elections from 1974 to 1983 (see Franklin 1988 for details of these analyses). The plot shows clearly the changing perceptions of where the leaders stood in the two-dimensional issue space, with the Liberal Party being the first beneficiary of 'new' issue concerns, though its position was rapidly taken by Mrs. Thatcher who, by 1983, had moved to dominate the upper half of the issue space while her party continued to command the right-hand half. 8

9 a b a. Liberal/Alliance leaders suppressed to simplify presentation b. How can the events of the following fifteen years be understood in similar terms? The first thing we need to establish is whether the space developed further after 1983, or whether it remained essentially unchanged. Figure 3 shows the locations of issues in 1997, on the same basis as they were shown for 1983 in Figure 1, by locating them according to their proximity with each other in two-dimensional space 9

10 FIGURE 3 The issue space of the British electorate in 1997, represented in two dimensions defined by factor loadings 0.6 N E W 0.4 Race equality Tax car polluters Privatise education Workers' lot is fair 0.2 Aid to Africa Gay rights Prison reform Positive race discrimination Privatise industry Privatise medicine Spend on defence 0 Sex equality -0.2 Fewer abortions Curb TU power Let cars pollute Regulate TUs Curb business power 0 L D Workers say Spend on education Spend on poverty Spend on health Create jobs Leave EU Death penalty -0.8 Redistribute wealth Strict life sentences Tougher sentencing L E F T R I G H T 10

11 as determined by factor analysis of the answers given by respondents to the British Election Study of that year. It shows what appears to be precisely the same issue space, with many variables occupying almost precisely the same locations, to within sampling error. On the left, race equality, spend on poverty, and spend on health are unmoved, as are privatise medicine, regulate Trade Unions and Tougher prison sentences on the right. Other variables appear to have moved in understandable - even predictable - ways. Those who want to leave the Common Market have become those who want to leave the European Union and have moved from the far left to the far right of the lower half of the issue space. Some issues that were central to the left as a whole in 1983 (the idea that government should create jobs, for instance) has moved far down into the old left quadrant by While some issues that were well up in the new left quadrant (sex equality) or well down in the old left quadrant (curb business power) have moved to occupy the middle ground in left-wing opinion. Similar changes have occurred on the right, where privatization and anti-abortion sentiments have moved towards the central ground. But the most important way in which the issue space in 1997 differs from that in 1983 is that the now-old dimension has continued the trend established before 1983 (Franklin, 1988) of expanding relative to the left-right dimension. In 1983 the new-old issue space spanned a range of about -0.5 to +0.3, while in 1997 it spanned a range from almost -0.8 to a 50% increase in its total span. The range encompassed by the left-right dimension, meanwhile, remained about 11

12 constant at 0.6 to Indeed, in the factor analysis conducted on 1997 data the new-old dimension was the first factor extracted in the analysis, explaining 56% of the variance jointly accounted for by the two factors (see Appendix for details of the factor and regression analyses conducted for this chapter). 2 Accounting for voting choice How have the issues discussed above helped to account for party choice? One way to answer this question is to place the parties and candidates within the 1997 issue space, as we showed them in This is done by finding the average score on each factor for each respondent who supported a particular party or who thought a particular candidate would make a good Prime Minister, and use those averages as X-Y coordinates to plot the positions of the parties and candidates on the basis of the locations of those who support them. The resulting plot, shown in Figure 4, displays a number of contrasts with the picture, shown in Figure 2, for In the first place, there has been a dramatic reversal of positions of Labour and Conservative party leaders in terms of the New-Old dimension. Mrs Thatcher had, by 1983, moved to dominate the upper quadrant of that dimension, but Mr. Major in 1997 was supported 2 Of course this new dominance of new-old issues could be partly the result of changes in the questions asked, but if those who planned the 1997 election study asked more questions that were distinguished on new-old terms, this was presumably because they judged those questions to be the more important ones. 12

13 FIGURE 4 Locations of the supporters of various parties (upper case) and leaders (lower case) within the issue space, 1997 N E * Blair W 0.1 * LAB * Ashdown 0 * LIBDEM L -0.2 D Major* * CON L E F T R I G H T primarily by those with the very oldest of old right concerns. Perhaps more importantly, in 1983 the major parties were still located clearly to the left and right of the center of the issue space, but in 1997 we see the Conservatives as being located very close to Mr Major, deep in 13

14 one of the quadrants of the issue space. Mr. Blair, meanwhile, is much more centrally located, dominating the top half of the issue space, while his party dominates the left hand side. This is analogous to the position that Mrs. Thatcher and her party enjoyed in a turning of the tables, indeed. Such a picture of the strategic differences between the parties in terms of issues echoes a view that was widespread among commentators during the election campaign. Strategy aside, what were the actual effects of issues on party choice, and how did they reflect the dynamics of representation in Britain? To answer this question we have run a series of regression analyses to predict left party voting in elections from 1987 to In order to assure comparability between the years, only those variables were employed which had been asked in similar terms in all years. Only nine issue variables fulfilled this criterion. Moreover, the issues that the remaining years had in common underrepresent the issues on the new-old dimension which we have seen to have been increasing in importance. Nevertheless, the findings reported in Table 1 are suggestive. In this table, the EU issue has been abstracted from the rest and placed in a separate column. Class effects are the sum of the effects of parents' class, education, occupation, and union membership. Issue effects are the sum of the effects of eight issue variables (excluding attitude to the European Union). Total issues adds the European Union issue to the rest. See Appendix for details. 14

15 Table 1 Total effects of class and issues on Tory voting, Year R 2 Class Issues EU Total Issues As can be seen, the effects of social class remained pretty constant from 1987 until 1992, but issues took a large jump in importance in 1997, and not only because of a notable increase in the importance of the EU issue. Does this mean that 1997 was a watershed election that gave to issue voting a new level of importance? Probably not. Issue voting and dynamic representation The pattern of effects shown in Table 1 cannot be used to validate our assumption that dynamic representation is occurring in Britain. That would take quite different data. We do not have measures of how satisfied voters are with the responsiveness of their government from election to election. Nor do we have measures of policy outputs or the extent to which those output accord with policy demands. Nevertheless, the pattern is consistent with the story told at the start of this chapter. In particular, the jump in the importance of issues in 1979 should not be taken to mean that issues suddenly obtained a new 15

16 driving force. That jump should be seen in the context of the large apparent drop in the effect of issues in If we assume that by 1992 British voters were in general of the opinion that Tory rule had continued long enough, but many voted for the party nevertheless, on grounds that Tories were more competent to handle the economic difficulties of the day (Sanders 19??), then the drop in the importance of issues that we see in 1992 would be the result. Equally, if in 1997 voters did in large numbers decide that enough was enough, that would translate into a great many issues lining up in the minds of voters as reasons to abandon the Tories. Voters do not necessarily vote their issue preferences, but in 1997 they did. In the next election it is quite likely that issues will return again to rough parity with other reasons for vote choice, and we should not interpret that to mean a sea change in the basis of British voting choice. Nor should we interpret the results for 1997 that way. When voters have had enough of the rule of one party, and decide to 'throw the rascals out' issues will appear to gain in importance. But this does not mean that voters decided to change parties on account of those issue preferences. What were probably much the same preferences in 1992 had no equivalent effect. There is more to voting than class and issues; but, ironically, election studies are not well-suited to studying the dynamics of representation. 16

17 [NOTE: We started to write this draft employing the cumulative file of British Election Studies, It was not immediately clear how debilitating the paucity of issue variables would prove, and we were most reluctant to 'reinvent the wheel' by putting together our own cumulative file of data from 1974 to date. It was not until quite late in the game that it occurred to us that by building on previous work we could eliminate the need to obtain full BES data for years prior to This seems much more feasible than trying to create comparable data for all elections since In the revision to this chapter, therefore, we will replicate the format of Figure 2, showing movement in issue space of parties and party leaders from 1983 through to 1997; and we will supplement this picture with regression analyses that include the factor scores for left-right and new-old issue dimensions where these dimensions are based on analysis of all available issue questions - not just those that are comparable from year to year.] 17