Skip to main content
    • Aa
    • Aa

Two Models of Class Voting


Until the late 1960s, most informed observers agreed with Pulzer's well-known claim that ‘class is the basis of British party politics; all else is embellishment and detail’. More recently, however, most analysts have come to believe that there has been a decline in the association between class and party choice, a development that is generally referred to as ‘class dealignment’. Nevertheless, several researchers have challenged this new consensus and argued that there has been little or no trend in the association between class and party.

Hide All

1 Pulzer Peter, Political Representation and Elections in Britain, revised edn (London: Allen & Unwin, 1972), p. 102.

2 Särlvik Bo and Crewe Ivor, Decade of Dealignment (Cambridge: Cambridge University Press, 1983), pp. 8291; Franklin Mark N., The Decline of Class Voting in Britain (Oxford: Oxford University Press, 1985); Kelley Jonathan, McAllister Ian and Mughan Anthony, ‘The Decline of Class Revisited: Class and Party in England, 1964–79’, American Political Science Review, 79 (1985), 719–37; Rose Richard and McAllister Ian, Voters Begin to Choose: From Closed-Class to Open Elections in Britain (London: Sage, 1986).

3 Heath Anthony, Jowell Roger and Curtice John, How Britain Votes (Oxford: Pergamon, 1985); Marshall Gordon, Newby Howard, Rose David and Vogler Caroline, Social Class in Modern Britain (London: Hutchinson, 1988); Heath Anthony, Curtice J., Jowell R., Evans G., Field J. and Witherspoon S., Understanding Political Change (Oxford: Pergamon, 1991).

4 Crewe Ivor, ‘On the Death and Resurrection of Class Voting’, Political Studies, 34 (1986), 620–38; Rose and McAllister, Voters Begin to Choose.

5 Heath et al. , Understanding Political Change, p. 64.

6 Marshall et al. , Social Class in Modern Britain; Weakliem David, ‘Class and Party in Britain, 1964–83’, Sociology, 23 (1989), 285–97.

7 Crewe Ivor, ‘Changing Votes and Unchanging Voters’, Electoral Studies, 11 (1992), 335–45, p. 342.

8 Dunleavy Patrick, ‘Class Dealignment in Britain Revisited’, West European Politics, 10 (1987), 400–19.

9 Rose and McAllister, Voters Begin to Choose, pp. 89.

10 Heath, Jowell and Curtice, How Britain Votes, p. 19.

11 This type of model is often applied to mobility tables, where it is known as ‘quasi-independence’. The basic quasi-independence model can also be estimated as a log-linear model. Most of the models fitted in this Note, however, place restrictions on the size of the latent classes that make it impossible to estimate them as log-linear models.

12 ‘Liberal’ will be use throughout to include the Alliance, Social Democrats and Liberal Democrats.

13 See Thomas George B., Calculus and Analytic Geometry, 4th edn (Reading, Mass.: Addison-Wesley, 1972), pp. 728–31, for a discussion of these methods.

14 Rose and McAllister, Voters Benin to Choose, pp. 38–9.

15 Dunleavy, ‘Class Dealignment in Britain Revisited’, pp. 414–16; Franklin Mark N., ‘How the Decline of Class Voting Opened the Way to Radical Change in British Polities’, British Journal of Political Science, 14 (1984), 483508.

16 Crewe, ‘Changing Votes and Unchanging Voters’, p. 339; Rose and McAllister, Voters Begin to Choose, p. 52.

17 Campbell Angus, Converse P. E., Miller W. E. and Stokes D. E., The American Voter (Chicago: University of Chicago Press, 1976; reprint of 1960 edn), p. 220; see also Butler David and Stokes Donald, Political Change in Britain, 2nd edn (New York: St Martin's, 1974).

18 Rose and McAllister, Voters Begin to Choose, p. 8.

19 See Goodman Leo A., ‘Measures, Models, and Graphical Displays in the Analysis of Cross-Classified Data’, Journal of the American Statistical Association, 86 (1991), 10851111; and Becker Mark P. and Clogg Clifford C., ‘Analysis of Sets of Two-Way Contingency Tables Using Association Models’, Journal of the American Statistical Association, 84 (1989), 142–51, for detailed discussions of these models.

20 Downs Anthony, An Economic Theory of Democracy (New York: Harper & Row, 1957).

21 Because there is no natural scale for these parameters, they are usually standardized by setting the mean equal to zero and the variance equal to one.

22 Goodman, ‘The Analysis of Cross-Classified Data,’ pp. 112–17.

23 Because it is not possible to observe the scale for the underlying dimension, the parameters must be standardized. This is normally done by making the γ and τ scores have mean of zero and variance of one.

24 Crewe, ‘Changing Votes and Unchanging Voters’, p. 339.

25 Gilula Zvi and Haberman Shelby J., ‘Canonical Analysis of Contingency Tables by Maximum Likelihood’, Journal of the American Statistical Association, 81 (1986), 780–8; see Weakliem David L., ‘The Two Lefts? Occupation and Party Choice in France, Italy, and the Netherlands’, American Journal of Sociology, 96 (1991), 1327–61, for an application.

26 A computer program (CDAS) to fit association models, along with a variety of related models, is available from Scott Eliason, Department of Sociology, University of Iowa.

27 The data are given in Heath et al. , Explaining Political Change, pp. 68–9. The class schema is discussed in Heath, Jowell and Curtice, How Britain Votes, p. 16.

28 The class composition of voters for minor parties, primarily the Scottish Nationalist Party and Plaid Cymru, is significantly different from that of Liberal voters. The Liberal electorate is primarily from the middle class, while voters for minor parties are drawn more evenly from all classes. Because of the small numbers, including minor party voters would complicate the analysis while adding little new information. In 1992, voters for the old Liberal party are also excluded.

29 Crewe, ‘On the Death and Resurrection of Class Voting,’ p. 627; Rose and McAllister, Voters Begin to Choose, pp. 138–9.

30 Most of these models cannot be fitted using standard packages for latent class models. Maximum-likelihood estimates were obtained using the method of scoring. GLIM 3.77 macros for these models are available from the author.

31 Crewe, ‘Changing Votes and Unchanging Voters’, p. 339.

32 The model can produce negative estimates of the number of class loyalists. The variation that could reasonably be expected to occur from chance given these sample sizes is fairly large; consequently, the appearance of some negative estimates is not surprising when the true number of class loyalists is small. Because a negative number of class loyalists would have no sensible interpretation, I have reported negative estimates as zero.

33 Butler and Stokes, Political Change in Britain, pp. 194, 419.

34 The fit of A2 is identical to that of the log-linear model of constant class-party association used by Heath et al. , Understanding Political Change, p. 70, because the two-dimensional model uses the same number of parameters for the class-party association as does the unordered model.

35 It would also be possible to allow the class or party scores to change between elections. Since it is possible to achieve a good fit without permitting such changes, they will not be considered further.

36 Because the dimensions are defined to be orthogonal, conclusions about changes in class voting would be essentially identical if the second dimension were entirely omitted. Nevertheless, the existence of a second dimension, and the lack of change in the strength of that dimension, are interesting in the light of theories that predict the development of new class cleavages that are not reducible to the traditional opposition between middle class and working class. Such theories would predict that changes in class voting should involve both dimensions. See Weakliem, ‘The Two Lefts?’, for a more detailed discussion.

37 Heath Anthony, Jowell Roger and Curtice John, ‘Trendless Fluctuation: A Reply to Crewe’, Political Studies, 35 (1987), 259–77.

38 Evans Geoffrey, Heath Anthony and Payne Clive, ‘Modelling Trends in the Class/Party Relationship 1964–87’, Electoral Studies, 10 (1991), 99117.

39 The linear trend term is also significant when the sample is restricted to the 1964–87 period used by Evans, Heath and Payne.

40 The standard method of hypothesis testing involves fitting a comprehensive model that includes both alternative models as special cases. Because the latent class and association models have different functional forms, this procedure is not possible here. It is, however, possible to apply tests for ‘non-nested’ models developed by Davidson Russell and MacKinnon James G., ‘Several Tests for Model Specification in the Presence of Alternative Hypotheses’, Econometrica, 49 (1981), 781–93, and extended by Weakliem David L., ‘Comparing Non-Nested Models for Count Data’, in Marsden P. V., ed., Sociological Methodology 1992 (Oxford: Basil Blackwell, 1992). The results of these tests support the conclusions in the text: none of the models are fully adequate, but the latent class models are rejected more strongly.

41 Rose and MacAllister, Voters Begin to Choose, pp. 3750; Crewe, ‘On the Death and Resurrection of Class Voting’.

42 Heath, Jowell and Curtice, ‘Trendless Fluctuation’; Crewe, ‘On the Death and Resurrection of Class Voting’.

43 Heath et al. , Understanding Political Change, p. 78.

* Department of Sociology, University of Connecticut. The data are taken from the British Election Surveys, originally collected by D. Butler, D. Stokes, I. Crewe, B. Särlvik, J. Alt, D. Robertson, A. Heath, R. Jowell and J. Curtice. I thank Anthony Heath and Clive Payne for providing the data in the form used here, and David Sanders and three anonymous reviewers for their comments on previous drafts.

Recommend this journal

Email your librarian or administrator to recommend adding this journal to your organisation's collection.

British Journal of Political Science
  • ISSN: 0007-1234
  • EISSN: 1469-2112
  • URL: /core/journals/british-journal-of-political-science
Please enter your name
Please enter a valid email address
Who would you like to send this to? *


Full text views

Total number of HTML views: 2
Total number of PDF views: 115 *
Loading metrics...

Abstract views

Total abstract views: 214 *
Loading metrics...

* Views captured on Cambridge Core between September 2016 - 23rd October 2017. This data will be updated every 24 hours.