Were Traditional labour Voters Disillusioned with New labour ?

New Labour’s move away from traditional socialist policies towards the centre of the left/right spectrum demonstrated the classic trade-off between middle-class gains and working-class losses. New La bour gained relatively more votes in the middle class than it did in the working class, leading to a marked class dealignment. As Przeworski and Sprague have argued, when a formerly socialist party moves towards the centre in the search for middle-class votes "The workers who would otherwise have voted for a Socialist Party have three avenues open to them: they can vote for bourgeois parties; they can abstain from voting altogether, and in some countries, they can vote for other parties that appeal to t hem as workers" (p61). In this paper the focus is on the second of these three options - abstention. Our hypothesis is that Labour’s traditional constituency in the working class will not have responded with enthusiasm to New Labour’s apparent lack of c oncern with their interests, and may have shown a greater reluctance to turn out and vote for the party.

Turnout was in any event one of the most interesting aspects of the 1997 election. The election saw a striking fall in turnout from its relatively high level of 77.9 in 1992 to 71.6 in 1997. We have discussed the overall fall in turnou t in detail elsewhere (Heath and Taylor 1999). Essentially, we found a number of pieces of evidence which suggested that turnout is higher when the election is expected to be a close one, and lower when it is expected to be an easy victory for one or oth er party. In 1997 the opinion polls were certainly suggesting an easy victory for Labour whereas in 1992 the opinion polls suggested a very close race. Other major landslides such as 1935 and 1983 have also seen low turnouts, while close races such as 1950 and 1951 saw relatively high turnouts. It is important to emphasize that our theory relates to the closeness of the race nationally rather than to the closeness of the race in the individual constituency. The relationship between the marginality of the constituency and the turnout there is logically distinct from that between the closeness of the national race and the national turnout.

However, our concern here is not with the overall decline in turnout but with the possibility that the decline was greater among traditional Labour voters than it was elsewhere. Did Labour's move to the centre and its wooing of new rec ruits from the middle classes mean that traditional supporters were less inclined to turn out? Curtice and Steed (1997) produced some highly suggestive figures from their analysis of the aggregate data. They found that turnout was especially low in Labo ur seats, and they drew a marked contrast between the situations in 1945 and 1997. Commenting on the 1997 result they wrote "On average just 68 per cent turned out to vote in the average Labour seat, compared with 74 per cent in the typical Conservative one, a larger gap than ever before. Labour's heartlands, then, were distinctly lukewarm about their party's surge to victory" (1997, p299). They contrasted this with the situation in 1945: "This is in sharp contrast with 1945 when Labour first swept to a major Commons victory. Turnout then was highest in Labour strongholds in northern cities, in Wales and in the mining and textile areas of Yorkshire and Lancashire" (p322).

Finally, we cannot from the aggregate data distinguish a hypothesis of disillusion with New Labour from one of complacency. Were Labour voters in these constituencies simply even more confident than those in other constituencies that L abour would win easily? Or were safe Labour seats ones where there happened to be large numbers of disillusioned traditional voters?

Survey data can help with some of these problems. We can avoid the ecological fallacy by comparing the turnout of Conservative and Labour supporters (although if we use previous vote rather than party identification we must remember th e likelihood of biases in recall). We can also control for registration differences by looking at the turnout of people who actually were on the register. We can also check more directly on the subjective feelings and perceptions of the electors.

However, there is one well-known problem with the use of survey data for studying turnout. Election surveys the world over tend to show much higher levels of reported turnout than are found in the aggregate data. The discrepancy is li kely to have three components: errors and redundancy in the register (which in effect means that the 'true' turnout rate will be higher than that shown by the aggregate data); non-response bias in the surveys, with non voters also tending to be non-part icipants in election surveys; and actual overclaiming by respondents of their turnout. We cannot do much about the first two of these problems, but we can deal with the third by validating the data against the official records. In this paper we report both self-report and validated results.

We begin by checking whether our survey data replicates Curtice and Steed's finding that turnout was lower in Labour seats. This is done in table 1. (Following usual practice we define a Labour seat as one that Labour won at the previ ous election.)

It is heartening that survey and aggregate results tell the same story. Both self-report and validated turnout data indicate that abstention was several points higher in seats that Labour had won in 1992 than in the seats that the Conservatives had won at that election. Note that the validated data show a larger difference. Moreover the gap of four or five percentage points is reasonably close to that found in the aggregate data (six points).

However, one reason why abstention might be higher among electors in the Labour seats could simply be that these were more likely to be working-class electors and working-class electors generally tend to have slightly lower turnout than do middle-class electors. Even the Conservative voters in these seats might be more working-class than their fellow Conservatives in safe Conservative seats. In other words, the social geography of Conservative and Labour voters could well explain diff erences in turnout across the different types of constituency.

The general relationship between social group and abstention can be seen in table 2, which gives the self-report figures for 1979 to 1992.

Table 2 shows a rather familiar pattern of turnout differentials : the middle class groups tend to have higher turnout than the working class, and in turn the unemployed usually have lower turnout than the working class as a whole. Thi s pattern is more or less clearly evident in most of the years. In general, the core Conservative groups (the Petty Bourgeoisie + Middle Class Homeowners) have lower abstention rates than the core Labour groups (Working Class Council Tenants + Trade Uni onists + the Unemployed), with the mixed groups lying in between.

So social geography could in principle explain the lower turnout in Labour seats. The crucial question therefore becomes whether these turnout differentials became magnified in 1997 as New Labour moved away from its traditional support ers. Curtice and Steed for example emphasized that the gap was larger in 1997 than it had been before. But it was only 1 percentage point larger, so we are looking for very small effects.

Table 3 looks at the difference between the level of abstention in 1997 and the average for the period as a whole. All the differences are positive, reflecting the low turnout in 1997. But abstention does seem to be especially higher than usual among the working-class council tenants, unionists and unemployed where the 1997 figures are 6, 7 and 8 points higher respectively than the 1979-92 average.

Table 3, therefore, gives a clear indication that, in 1997, abstention was rather larger in Labour's core constituency than it had been in the earlier years. In other words the core Labour groups, such as the unemployed, the council te nants and the working-class unionists, who were relatively less likely to increase their support for Labour, were also relatively less likely to turn out and vote.¹

We need to check whether these possible changes were specific to former Labour voters or were common to both Labour and Conservatives in these working-class groups. If abstention in the working-class groups had increased among both Lab our and Conservative supporters alike, we could hardly attribute the increase to disillusion with the Labour party. Marginality of the constituency would become a more plausible explanation. Cross-tabulating table 3 by previous vote would lead to many v ery small categories, so to explore this question we aggregate our groups into three broad categories - the Conservative core, consisting of the petty bourgeoisie and the middle class homeowners, the Labour core consisting of the working-class council ten ants, trade unionists and unemployed, and an intermediate group comprising all the other categories with more mixed voting traditions.

Table 4 shows a rather interesting pattern. In general, former Conservatives show a clear gradient, with the lowest abstention occurring in their own core of the middle-class groups and the highest abstention occurring in their opponen ts' core working-class groups. In the 1979-92 period as a whole, we find that Conservative abstention was only 7% in their middle-class core rising to 10% in the intermediate groups and 13% in the working-class core. Labour in contrast shows a relativel y flat profile. It is as though membership of the core working-class group manages to counteract the general depressing effect of lower class on turnout. Indeed, there are hints that Labour abstention was actually slightly lower on average, at 10%, in i ts core working-class group than it was in the middle-class core where it was 12% on average.

Again we must emphasize that the differences are small but the pattern makes good intuitive sense. In effect there are two processes occurring: abstention generally tends to be higher in the working class, but at the same time abstenti on by a party's supporters tended to be lowest in its own core group and highest in its rival's core. These two processes lead to a relatively steep class gradient among Conservative supporters, and a weaker gradient in the opposite direction among Labou r supporters.

Did 1997 demonstrate the same pattern? Table 5 shows that the Conservative pattern in 1997 was a straightforward repeat of the Conservative pattern in the previous four elections, although as expected in all three groups abstention is s omewhat higher than previously. The Labour pattern, however, is rather different from previous years: Labour abstention was actually lower in the core middle-class core in 1997 than it had been previously, down by 4 points, while it rose by four points i n the working-class core. The net result was that, in 1997, Labour abstention showed a very similar class gradient to the Conservative one. The inverse pattern that had characterised previous elections completely disappeared.

Of course there are lots of problems with the use of retrospective measures, and indeed of contemporary measures of turnout. So this cannot be conclusive evidence. Another possibility is to look at strength of party identification, wh ich is closely related to turnout (Heath and Taylor, 1999, table 9.7) and which is also a more direct measure of enthusiasm for Labour. This also has the advantage of including both voters and non voters, so it gives a fuller picture.

Core Middle

In the years from 1979 to 1992 the same clear patterns are evident as with abstention : there tends to be stronger partisanship for a party within its own core group. Thus among Conservatives in 1979 83% in the core middle-class group s were strong supporters, compared with 76% in the working class core. The reverse pattern held for Labour, with 85% of their supporters in the core working class being strong supporters, falling to 70% in the core middle class. These differences are no t very large, but they persist in every single study from 1979 to 1992. The same patterns would probably be shown if we looked at inter-election loyalty rates.

However, 1997 presents a challenge to this general pattern, but only on the Labour side. The Conservative side is quite interesting, showing a sharp dip in strength of support. But the dip happens in all social groups alike. So even though Conservative supporters in the core middle-class groups were less enthusiastic than usual, they remained in 1997 more enthusiastic than Conservative supporters in the working-class groups.

Core Middle

New Labour, however, sees quite a marked turnaround. In 1997 there was no gradient at all. What we see is enthusiasm increasing in the core middle-class groups, up four points from its previous average to 72%, but declining five point s in the core working-class groups from to 71%. This thus corroborates the story told by the abstention data. In other words there seems to have been a definite weakening, both absolute and relative, of Labour enthusiasm in its core social groups. But the effect is small and we must not exaggerate its importance. However, it does confirm the picture told by the class voting figures and by the

turnout figures. It cannot though simply be described as disillusion: after all, Labour enthusiasm in its core group in 1997 was actually stronger than Conservative enthusiasm in its middle-class core. It is on the Conservative side that disillusion shows itself most clearly.

Finally, we need to check whether these results are affected by the marginality of the constituencies. We therefore fit a general model that includes both social and constituency-level variables. The results are shown in table 8.

Winning party in 1992

Table 8 shows the results, both when we regress self-reported abstention and when we regress the official records of abstention on our explanatory variables. The pattern of coefficients and their levels of significance are very sim ilar in the two models, and so we can have some confidence in the results.

At level 2, on the constituency side, we note that neither marginality nor the winning party in 1992 were significantly related to abstention in 1997, although with the self-report data the parameter estimate is fairly close to signific ance. Even if we restrict ourselves to Labour seats, marginality fails to reach conventional levels of statistical significance. This is consistent with the work of Pattie and Johnston (1998) who found that, once individual-level variables were included in the model, marginality failed to have a significant relationship with turnout. If we drop marginality from the model, the pattern of the other parameter estimates is left virtually unaltered, and so it is unlikely that the story told be the cross-tabu lations in tables 4 and 5 is affected by constituency marginality.

However, in both models there is a highly significant relationship between turnout the percentage of professionals and managers in the constituency. Incidentally, if this latter variable is dropped from the analysis we find that the pa rameter estimate for Labour seats does become statistically significant in both models. The two variables will of course be correlated, but it is interesting that it is the constituency class composition that does the work. The explanation for this uncl ear. It could of course be that there are further unmeasured variables at the individual level such as sense of political efficacy although we should not rule out the possibility that, where there is a large core of professional and managerial electors, their norms of civic duty are more powerful (cf. Eagles and Erfle 1989).

On the individual-level side we find that the working-class core groups - council tenants, trade union members and the unemployed have in general significantly higher abstention rates (There are some slight differences in this respect b etween the two models but the general pattern is clear in both). If we replace our combined variable of social group membership with separate measures of housing tenure, class, TU membership and economic activity (which would be more conventional) we fai l to make a significant improvement in fit. We have also added further individual-level variables to the model, such as age, length of residence in the neighbourhood, and marital status. Adding these does not lead to any appreciable change to the story.

Finally, we have tested for an interaction between the individual-level variables and previous vote, in order to test whether Labour voters differed from former Conservative voters. We find no interaction, confirming the story of Table 5.

Conclusions

Once again, we must emphasize the small changes that occurred in the patterns of abstention and strength of partisanship that occurred in 1997. It would be far too strong to argue that supporters in Labour’s traditional core constituen cy were disillusioned by New Labour’s move towards the centre of the political spectrum, but there are some strong hints from our data that the move to the centre was, albeit in a rather modest way, responsible for muted enthusiasm among traditional suppo rters. Our analysis thus confirms the story that Curtice and Steed had told from the aggregate data. It is also perhaps instructive to consider the magnitude of the electoral cost to Labour of this muted enthusiasm. How many votes in its traditional c onstituency might Labour have lost as a result of the changed patterns of abstention? To be sure, New Labour gained more middle-class recruits than it might otherwise have done from its move to the centre, but were these gains offset in part by working-c lass abstention?

We cannot make any definitive calculation of the gains and losses, but we can illustrate the scale and nature of the problem with a simple simulation. In the first column of table 9 we show the reported levels of turnout among previous Labour voters in 1992. This repeats the information from table 4, but for simplicity we have combined the categories into our three - the middle-class groups that represent the Conservative core, the working-class groups that represent the Labour core, and mixed groups that lie in between. This first column shows the even social distribution of Labour abstention that had characterised the 1979-92 period. The second column then shows the reported abstention in 1997 and here of course we see the mark ed social gradient that we noted in table 5.

Our next step is to calculate some hypothetical figures for 1997 that might have obtained if, say, John Smith had remained leader and the Labour Party had not made its move to the centre. We take the mixed group as the reference categor y and calculate the 1992 odds ratios for each group. We then solve so that the 1997 Labour abstention among the mixed group is the same as its actual level in 1997 and so that all the odds ratios are the same as they had been in 1992. This then gives us hypothetical figures for 1997. These hypothetical figures in essence tell us how the different groups might have abstained if there had been an across-the-board increase in Labour abstention that was everywhere the same as it actually was among the mixe d group. We can think of this as representing the kind of turnout that Labour would have obtained if it had not made the determined effort to increase its appeal to middle-income middle Britain.

What we find of course from our hypothetical calculations is that Labour turnout would have been higher in its core working-class groups and lower in the core middle-class groups - up four points in the core working class and down four points in the core middle class.. However, the crucial electoral consideration is the actual numbers of former Labour voters in these different groups: a gain of four points among a large group of potential voters could in theory outweigh a loss of four points in a smaller group. However, rather surprisingly, we find that former Labour voters were slightly more numerous in 1997 in the middle-class core than they were in the working-class core.

The explanation for this is not hard to find : while the working-class core contained a higher proportion of former Labour voters than did the middle-class core, the working class core - council tenants, union members and the unemploy ed - had shrunk to a much smaller proportion of the electorate while the middle-class core had grown. Social change, therefore, had cancelled out the advantages of Labour’s traditional strength in the working class. The shrinking size of its traditional core groups had of course been one key element in the thinking that lay behind Labour revisionism (see in particular Radice 1992).

Aldrich, John H (1993) Rational choice and turnout. American Journal of Political Science 37:246-78

Curtice, John and Michael Steed (1997) The results analysed. Appendix 2 in David Butler and Dennis Kavanagh, The British General Election of 1997. London: Macmillan.

Denver, David and Gordon Hands (1974) Marginality and turnout in British general elections. British Journal of Political Science 4: 17-35

Denver, David and Gordon Hands (1985) Marginality and turnout in British general elections in the 1970s. British Journal of Political Science 15:381-88

Denver, David and Gordon Hands (1997) Turnout. In Britain Votes, 1997, eds Pippa Norris and Neil Gavin. Oxford: Oxford University Press.

Eagles M and Erfle S (1989) Community cohesion and voter turnout in English parliamentary constituencies, British Journal of Political Sience 19: 115-25.

Heath, Anthony and Bridget Taylor (1999) Turnout: new sources of abstention? In G Evans and P Norris (eds) Critical Elections. London: Sage.

Pattie, Charles and Ron Johnston (1998) Voter turnout at the British general election of 1992: rational choice, social standing or political efficacy? European Journal of Political Research 33:263-83.

Przeworski Adam and John Sprague (1986) Paper Stones: A history of Electoral Socialism. Chicago: University of Chicago Press.

Radice, Giles (1992) Southern discomfort. Fabian pamphlet 555. London: The Fabian Society.

Swaddle, Kevin and Anthony Heath (1989) Official and reported turnout in the British general election of 1987, British Journal of Political Science 19:537-51.