1. Click here for BBC coverage of General Elections 1945-1997

2. Click here for the BBC's coverage of the 2001 General Election Results 

3.  Click here for the BBC's coverage of the 2005 General Election Results

4. Click here for IPSOS MORI General Election Data 1974-2005

5. . Click here for IPSOS MORI 2005 General Election Data

The analysis of voting behaviour is known also as "psephology" deriving from the Greek "psephos" [a pebble] with which the ancient Athenians indicated their voting decisions. Psephologists in the UK distinguish between the period  of 1945-1970 which they characterise as the era of electoral stability, two party dominance, party identification and class alignment  and the period from 1970 to the present day which is described as the era of declining party identification/partisan dealignment and class dealignment although there are also important arguments as to whether the general elections of 1997 and 2001 ushered in a realignment of UK voting behaviour.

In this document I shall focus especially on the ways in which the party identification model of voting behaviour was used in Butler and Stokes' "Political Change in Britain" [1969: second edition 1974] to explain relationships between social class and voting behaviour between 1945 and 1970.Working class voters and middle class voters were shown to vote mainly for the Labour and Conservative parties respectively although there were also significant percentages of deviant voters who did not vote predictably according to their social class. My discussion of "Political Change in Britain" may strike beginning students as a bit "long winded" but I do believe that an early appreciation of the significance of this model will pay dividends later when students begin to study the subsequent processes of partisan dealignment and class dealignment.

In the period 1945-1970 there were also observable correlations between voting behaviour and age, gender, region, religion and ethnicity. I discuss briefly the reasons for these correlations and also reproduce IPSOS MORI data on these correlations for the General Elections Oct 1974-2005 in the hope that students will find it useful to have this data "all in one place".

• From 1945 to 1970: An Overview

Let us begin with a broad consideration of the period of 1945-1970 which has been described as a period of relative electoral stability dominated by the two major political parties: the Conservative Party and the Labour Party. Thus between 1945-1970 the Conservative and Labour Parties regularly gained approximately 90% of the votes cast in general elections which under the conditions of the "first past the post" electoral system translated into approximately 98% of the parliamentary seats while the Ulster Unionist Party [UUP] gained a further 10-12 seats and could be relied upon to regularly support the Conservative Party in parliament. Neither the Liberal Party nor the Nationalist parties offered any real challenge to this 2 Party dominance. Further aspects of electoral stability were that opinion poll fluctuations and by election swings were relatively narrow and surveys indicated that relatively few voters switched their party allegiance between general elections although more switched between voting and non- voting. The relative stability of opinion polls suggested that most voters were  influenced far less by short term political issues than by long term social structural factors to be discussed below and variations in general election results could be said to be determined by the relatively small percentages of so-called floating voters who did switch party allegiance between general elections.

• Political Change in Britain [David Butler and Donald Stokes 1969:  Second Edition 1974]

Psephologists at the time demonstrated that voting behaviour was clearly correlated with a range of social variables including social class, age, gender, region, religion and "race" or ethnicity and that social class was the most significant influence on voting behaviour which enabled P.G.J. Pulzer to write in Political Representation and Election (1967) that "Class is the basis of British party policies: all else is embellishment and detail",  a conclusion which was endorsed fully by David Butler and Donald Stokes in their famous study "Political Change in Britain [1969: second edition 1974].

These patterns of voting behaviour and in particular the relationships between social class and voting behaviour were explained in terms of the so-called Party Identification Model of voting behaviour which originated in the electoral research conducted in the 1940s and 1950s at the University of Michigan and was refined further in Butler and Stokes' Political Change in Britain. This is a long detailed and complex study and I can aim here only to outline the core elements of its model of party identification and in particular its analysis of relationships between voting behaviour and social class.

These main elements can be stated as follows.

• Approximately 90% of respondents in Butler and Stokes' surveys stated that they did identify with either the Conservative Party, the Labour Party or, to a lesser extent the Liberal Party and the respondents'  party identifications usually remained relatively stable over the course of several elections and often throughout voters' lives sometimes hardening with age.

• Their party identification was correlated strongly with their actual voting behaviour  such that, for example, in the Local Elections of 1963 85% of Conservative identifiers, 95% of Labour identifiers and 88% of Liberal identifiers voted in accordance with their stated party identification.

• The vast majority of respondents were also prepared to assign themselves either to the working class or to the middle class or, in a small minority of cases , to the upper class.

• The definition and measurement of social class presents enormous problems but in early studies of voting behaviour social class is usually measured by occupation as in the schema developed by the Institute of Marketing  in which individuals  are assigned to "social classes" A, B, Cl, C2, D and E defined as follows: A=Higher professional, managerial and administrative; B =Intermediate professional, managerial and administrative; C1 =Supervisory, clerical and other non-manual; C"=Skilled manual; D=Semi-skilled and unskilled manual; E= Residual including casual workers and people dependent wholly on state benefits

• Butler and Stokes used a modified version of this schema in which they  subdivided the C1 grouping into "skilled or supervisory non -manual" and "lower non -manual categories" because they particularly wished to investigate the possibility of significant variation sin party identification within the broad C1 category.  In some later studies much more attention would be given to the definition and measurement of social class, as we shall see in the following document.

The following first three tables have been slightly adapted from Political Change in Britain .

• When the respondents self assigned social class position[ their subjective social class position] was compared to their social class position as measured by Butler and Stokes' modified I.O.M. scale [their objective social class position] it was found that subjective and objective social class positions were strongly but far from perfectly  correlated.

Table 1

Class-self-image by occupational head of household 1963 [=Objective social class position]
Self-assigned social class position [= subjective social class position]	Higher Managerial [1]	Lower Managerial[ 2]	Supervisory non-manual [3]	Lower non -manual [4]	Skilled manual [5]	Unskilled manual [6]	Residual, on pension or other state benefits [7] [omitted]
Middle Class	78%	65%	60%	32%	17%	9%
Working Class	22%	35%	40%	68%	83%	91%

• There was a high correlation between respondents' objective social class position and their party identification

Table 2

• There was a high correlation between respondents' subjective social class position and their party identification or partisan self-image.

Table 3

• It can be seen clearly in the following table there was a strong correlation between voters' objective social class position and their actual voting behaviour from 1945 to 1970 as is indicated in the following data from selected general elections. Assuming for simplicity that we can measure social class by occupation and that the division between manual and non-manual workers corresponds to the division between the working class and the middle class,  the link between voting behaviour and social class was at its strongest in the General Elections of 1950 and 1951 when approx. 2/3  of working class voters voted Labour and 1/3 Tory, while approx 3/4 of the middle class voted Tory and 1/5 Labour. Class voting remained fairly high throughout the 1960s while support for the Liberals was very low and fairly evenly distributed across the social classes, although with a slight middle class bias.  

  [However as the data for the 1983 General Election show, the link between social class and voting behaviour had weakened by 1983, especially because working class support for the Labour Party had declined significantly but also because middle class voters had to some extent deserted the Conservative party for the Liberal-SDP Alliance. We shall consider such matters in the next document]

But why did so many voters vote in accordance with their social class position between 1945-1970?

• Butler and Stokes argued that most voters had only limited knowledge and understanding of key political issues of the early 1960s such as the state of the UK economy or the desirability of otherwise of UK entry into the EEC [as it then was; they could only rarely describe in any detail the policy differences between the political parties; and their political opinions were often ideologically inconsistent in the sense that they could only rarely be combined into composite ideological positions which were recognisably "left wing" or "right wing" or "centrist."  Therefore for most but not all voters the voting decision could not be explained as an individual response to perceived differences in party policies. Instead voting decisions could be better explained via the influences of long term social structural factors: it is in this sense that the Party Identification model came to be described as a sociological model of voting behaviour [as distinct from the more individualistic models of voting behaviour which were developed from the 1970s onwards. However Butler and Stokes did not deny totally the influences on voting behaviour  of short term and medium issues, policies and events but these were considered to be much less influential than long term social structural factors.

• Butler and Stokes argued that voters were heavily influenced by long term processes of political socialisation operative especially in the family but also in the work place and the wider community which presented them with generalised broad images of the political parties. Thus Labour might be presented as the party of the disadvantaged, of the trade unions, of the working class of nationalisation or of the welfare state while the Conservative Party might be presented as the party of private enterprise, of private property and the nation as a whole.

• There were also, obviously important social class differences in these processes of political socialisation n which working class people and middle class people to identify especially with the Labour Party and the Conservative Party respectively and to vote accordingly. [ Within these class differentiated socialisation processes members of each social class would be provided both with positive images of "their party" and negative images of opposing parties].

• The Party Identification Model of Voting Behaviour: An Interim Summary

• The Party identification suggests that voters decisions are influenced much more by social structural variables [and especially by their social class position] than by short term issues, policies and events .

• Differing social class positions result in class differences in political socialisation processes operating in the family , the work place and the wider community.

• Class differences in political socialisaton processes result in the transmission of differing broad party images which encourage working class and middle class people to identify mainly with the Labour Party or the Conservative Party respectively.

• Most voters vote in accordance with their party identification.

• That is: for most voters social class differences result in social class differences in social class differences in political socialisation, social class differences in party images, social class differences in party identification and social class differences in voting behaviour. And there you have it!

• Deviant Voters

Given their relatively large numbers if all  working class voters had voted Labour between 1945 and 1970 Labour would have won every single general election. However a substantial minority of working class voters voted Tory, and also a smaller, but still substantial minority of middle class  voters voted Labour, even at the high point of class voting in the General Elections of 1950 and 1951. These cross-class voters, given their relatively small numbers, were  described as "deviant" voters. Why did they vote "deviantly"?

Working Class Conservative Voters : An Introduction

Early explanations (for example that of Bagehot) of working class Conservatism had stressed the importance of deference, and in their study "Angels in Marble", Mackenzie and Silver claimed that in the early60's approximately 50% of working class support for the Conservative Party could indeed  explained in terms of deference, with such voters to be found mainly in agricultural areas and small towns, working in small scale industries with relatively close paternalistic relationships with  their employers, having limited opportunities for the development of working class consciousness, and perceiving the inequalities of the social  order as essentially just and meritocratic. Another grouping of secular working class  Conservatives was also noted; these voters were mainly young, male, relatively well-paid, voiced  limited support for the Trade Union movement, the Welfare State and nationalisation and  essentially supported the principles of private enterprise. They were more pragmatic in their voting behaviour than the deferentials and more likely than the deferentials to switch their support away from the Tories.

As a response to Labour's three successive defeats in the General Elections of 1951, 1955, and 1959 it was widely argued that this may have been due to changes in the nature of the working class with rising working class  affluence leading to an increased likelihood of secular working class Conservative voting. However this line of argument was undermined by the Goldthorpe-Lockwood critique of  the Embourgeoisement theory (The Affluent Worker in the Class Structure.1969) which found  affluent manual workers had been more  likely than the working class as a whole to support Labour in the General election of 1959 .Nevertheless  the support of these affluent workers for the Labour Party was shown to be very fickle, and potentially  volatile suggesting that more and more working class voters might  desert the Labour Party in the future.

The analysis of working class conservatism can be extended in the following document when we consider the more recent  trends in voting behaviour and the new theories of voting behaviour which were developed in response to these trends.

• Age and Voting Behaviour [Data from IPSOS MORI]

Electoral data from 1945 -1970 suggested that older voters were more likely than younger voters to vote Conservative and less likely than younger voters to vote Labour. This was often explained by the notion that as individuals grow older they become more set in their ways, more wedded to traditional values and attitudes and more likely to believe that Conservative governments are more likely to safeguard their financial and personal security which appear increasingly important in later life. However  Butler and Stokes [and other psephologists] explained the relationship between age and voting behaviour in terms of the different processes of political socialisation affecting successive generations. Thus people born in say 1900 would have grown up before a powerful Labour Party existed and might as a result have been socialised to vote Conservative or Liberal rather than Labour. By comparison people born in say 1930 would have been reaching adulthood at the time of the reforming Labour Governments of 1945-51 and may have experienced processes of political socialisation which encouraged them to vote Labour in the 1950s and 1960s when they were still relatively young. and there was also the possibility that even in old age  these voters might still be more likely to vote Labour than previous generations of older voters.

This line of thought leads on to further speculations. Would voters who grew to adulthood in the "swinging" radical 60s be especially likely to vote Labour when young and more likely than previous generations to continue to vote Labour even as they grew older?  How would growing up in the era of Thatcherism affect voting behaviour in the future and so on? Unfortunately it is difficult to answer such questions but at least the following IPSOS MORI data illustrate relationships between age and voting behaviour between Oct. 1974 and 2005. The data do indeed show that broadly speaking older voters are less likely to vote Labour and more likely to vote Conservative than younger voters although there is not an absolutely perfect correlation between age and voting behaviour. particularly in the 2001 General Election.

• In 2001 18-24  year olds were more pro-Conservative and less pro-Labour than 25-34 year olds  although 35-54 year olds and voters 55+ were increasingly likely to vote Conservative and decreasingly likely to vote Labour.

• In 2005 18-24 year olds were again more likely than 25-34 year olds to vote Conservative and but both age groups were equally likely to vote Labour. Again 35-54 year olds and voters 55+ were increasingly likely to vote Conservative and decreasingly likely to vote Labour. 

It has been traditionally argued that women were more likely than men to vote Conservative and less likely than men to vote Labour. The overall size of the gender gap  is measured by the difference between the percentage Conservative-Labour lead among women and the percentage Conservative-Labour lead among men. and , on this basis the gender gap  in voting behaviour  was sometimes substantial between 1945 and 1970 reaching highs of -17 % in 1950 and 1951. These data were usually explained by theories that for much of the C20th, women were less likely to be in paid employment  and less likely to be trade union members or because women were seen as possibly more  "traditional" or  more "deferential" than men in their views as a result gender differences in socialisation or because women, on average, live longer than men and as seen above, age is correlated with Conservative voting.

We may use the following recent IPSOS MORI opinion poll data to clarify the calculation of the gender gap statistic and to assess whether traditional gender differences in voting behaviour have been maintained in recent years. On the basis of these data the traditional negative gender gap was substantial in 1974 [Oct], 1979 and 1983 but relatively smaller  between 1987 and 2001 while in 2005 women were both less likely to vote Conservative and more likely to vote Labour than men thus reversing the traditional gender gap and generating instead a positive gender gap. This reversal of the traditional gender gap continued in 2010 according to the Ipsos Mori data although some other polls do suggest that the traditional gender gap has returned in 2010.

According to Ipsos Mori data this positive [i.e. non -traditional gender gap continued in 2010 when  Women were 5% more likely to vote Conservative than Labour [36% Cons and 31% Labour] but men were 10% more likely to vote Conservative than to vote Labour [38% Cons and 28% Labour ] which results in a positive gender gap of 5 [down from 6 in 2005] illustrating that women's traditional relative Conservatism still did not apply in 2010. {See above link to IPSOS MORI 2010 General Election analysis for further information.]

 [However  it is important  to note  that there may be some uncertainly as to this conclusion since in some YouGov surveys at the time of the 2010 General Election the traditional gender gap does reappear and also a research study cited in Britain at the Polls 2010 : Nicholas Allen and John Bartle {Editors] 2010 reaches the same conclusion . Students should discuss with their teachers how to respond to this divergence in research findings. ]

.

• Religion and Voting Behaviour [Data from IPSOS MORI:  results of their 2001 survey on religious affiliation and party support.]

Religion was a much more significant influence on voting behaviour in mainland Britain from the mid C19th until the early C20th when members of the Church of England were especially likely to support the Conservative Party and Catholics and Nonconformist churches were more likely to support the Liberal Party. It continues to have a very strong influence on voting behaviour in Ireland and also in certain key areas on mainland Britain such as Liverpool and Glasgow where Catholics have traditionally been out of sympathy with traditional Conservative governments' rejection of an independent united Ireland  and Protestants have supported the Conservative defence of an independent and Protestant dominated Northern Ireland. Similar views can also be found in certain key areas on mainland Britain such as Liverpool and Glasgow where religious links with Ireland may override social class loyalties and cause many working class Protestants to vote Conservative and middle class Catholics to vote Labour..

However with the rise of the Labour Party and the growth of social class based voting religion became a less important determinant of voting behaviour. Nevertheless between 1945 and 1970 Catholics  were quite significantly more likely to vote Labour and less likely to vote Conservative than supporters of he Church of England and other Protestant denominations  mainly because mainland British Catholics may often have had family connections with Ireland but also because they were more likely than Protestants to be working class so that their voting behaviour was actually explained more by their class membership than by their religion.  The following data show continuing correlations between voting behaviour and religion in 2001 but these correlations too are likely mainly to reflect the indirect influence of social class.

 

• "Race", Ethnicity and Voting Behaviour

Sociology students will be familiar with the important distinction between "race" [which is a biologically based concept of negligible scientific validity] and the much more useful concept of ethnicity which refers to the cultural, religious and linguistic aspects of different social groups' lives. I focus here on relationships between ethnicity and voting behaviour.

IPSOS MORI Estimates of Voting Behaviour in the General Election of 1997