Abstract
Turnout decline in Britain is greater than it first appears since changes in the social composition of the electorate have had a positive impact on turnout. This paper finds that whereas a weakening in the strength of party identification is associated with the long-term decline, the political context influences short-term variation. Partisan dealignment is also changing the dynamics of the determinants of turnout. Since non-identifiers are more strongly influenced by the political context than strong identifiers, and there are now more non-identifiers than previously, the political context is becoming a more important factor in determining whether people vote or not.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
In 2001 there were 25,554,856 votes cast whereas in 1997 there 30,496,924. Pippa Norris The British Parliamentary Constituency Database 1992–2001.
There have been many studies that examine the individual level determinants of turnout at one point in time (Clarke et al. 2004: 237–260; Whiteley et al. 2001) and from an over time perspective (Miller 1992; Abramson and Aldrich 1981). There have also been many studies that examine the aggregate level impact of the political context at the constituency level (Whiteley et al. 2001; Denver and Hands 1974, 1985) and at the national level over time (Clarke et al. 2004: 261–268; Franklin 2004). However, there have been very few attempts to synthesise these two approaches together to explain turnout change.
For example, the salience of class in determining turnout may weaken as society becomes more middle class. However, it is also possible that the effect of class may be stronger in some years than in others without any general tendency to change. If this is the case then the source of the variation is unlikely to be due to long-term forces and may be the result of particular short-term factors, possibly relating to the political context. Under these circumstances the assumption of parametric uniformity is still broadly valid but it may be necessary to try and specify additional interaction terms in order to explain the differing effects from year to year.
Including the interaction between class and year does not result in an improvement in the fit to the data (χ2 change eq 7.4, df = 11, p = .767). This suggests that the working class have not become disproportionately less likely to vote with Labour’s shift to the right (Budge et al. 2001). However, there have been some modest changes in the odds of voting over time by tenure (χ2 change eq 24.3; df = 10; p = .007), and rather greater changes by Franchise (χ2 change eq 32.7; df = 8, p ≤ .0005) party ID (χ2 change eq 42.5; df = 11, p ≤ .0005) and age (χ2 change eq 34.4; df = 11; p ≤ .0005). The changes by tenure are relatively unimportant since it is only in 2005 that significant deviations are observed (when the effect was stronger than in other years). The changes by Franchise are also modest (the effect is somewhat weaker in 1970 and 1975 and is somewhat stronger in 1987) suggesting that there is little structure to the variation. In both instances, without sound theoretical justification for including additional interaction terms, the assumption of parametric homogeneity is therefore upheld. The changes by age and party ID are somewhat more interesting. There are significant differences in the effect of the age variable in 2001 and 2005 (the two elections with particularly low turnout). In both these elections the effect of the age term was stronger than usual (meaning that the odds of voting increase more rapidly with each year of age or that young people were even less likely to vote in relation to old people than usual). There are also significant differences in the effect of the party ID variable in 1983 and October 1974, when the effect of party ID was somewhat weaker than normal (meaning that strong identifiers were not so much more likely to vote than non identifiers than at other times), and in 2001, when the effect of party ID was somewhat stronger than normal (meaning that weak identifiers were even less likely to vote than strong identifiers than at other times). These short-term changes suggest that the effect of age and party ID may be related to the political context (Franklin 2004). Since there has been no long-term change, and moreover, since we can be sure that the changes that have taken place are not related to the distribution of the variable in the population, there is no reason to relax the assumption of parametric homogeneity (as a general principle) at this point. However, we will return to these points in the second part of the paper when we consider the political context.
Bootstrapping reveals that these results are robust to the exclusion of any particular election, and are in fact stronger when some elections are omitted.
Linear regression of turnout decline 1955–1997: % turnout = 77.7 − 0.08 year (one sided t test significant at 0.1 level).
Linear regression of turnout decline 1929–2006: % turnout = 79.7 − 0.14 year (one sided t test significant at 0.01 level).
If respondent does not have social class then social class of spouse is used as a proxy.
After 1992 the average age of people on the electoral register was substantially older than it was before. However, rather than reflecting a sudden demographic change this is probably more likely due to council tax avoidance, which meant that in the 1990s many people, particularly the young, did not sign up to the electoral register.
There have been various amendments to the franchise in Britain over the last 100 years or so. In 1918 women over 30 were given the vote for the first time, and all men over 21 were eligible to vote (prior to that there were property requirements for men). In 1928 full adult franchise as we know it today was established when the age of maturity for women was reduced so that it was the same as for men.
Turnout is measured by whether respondents reported having voted in the election, or not. The base is all those on the electoral register. Reported turnout consistently overestimates actual turnout or validated turnout. Swaddle and Heath (1989: 539) suggest that four main reasons account for this discrepancy: misreporting by survey respondents, response bias, failure to trace all movers, and redundancy in the electoral register. However, Heath and Taylor (1999) show a fairly constant relationship between official turnout and reported turnout across the years, with the latter tracking the former albeit at a higher level. If the relationship between the two measures is constant, then this will affect the constant term in the regressions but will not affect the parameter estimates of interest.
Electoral cohorts are defined by the first General Election that an individual was eligible to vote in and the aggregate turnout in this election is computed for each individual in that cohort. The range of values for electoral cohort are centred on zero. People born before 1904 (who were thus eligible to vote before full franchise was implemented) score zero. The dummy term for Franchise thus tells us whether people born before full franchise was implemented turnout differently from those born after, and the interaction with sex tells us whether this was different for women (who did not have the vote) and men.
Party identification is measured using the standard four point scale (0 = None; 1 = Not very strong; 2 = Fairly strong; 3 = Very strong).
This is done using the following function: \( P_{v} = (e^{{\log x_{0} + \log x_{i} }} )/(1 + e^{{\log x_{0} + \log x_{i} }} ) \), where log x 0 refers to the constant (1964) term and log x i refers to the dummy term for a particular election year. Thus, the probability of voting in 2005 (calculated from the log odds presented in Model 1 in Table 1) is \( e^{{2.08 - 0.75}} /1 + e^{{2.08 - 0.75}} = 3.79/4.79 = 0.79 \). Or, put another way reported turnout among survey respondents was 79% (Table 2). This corresponds to our observed known value from the survey estimates. We can then calculate the level of turnout in each year assuming that various demographic variables had stayed the same since 1964 by comparing the magnitude of the coefficient for each dummy term after we control for age, sex, class and tenure etc. (Model 2) to the constant 1964 term in Model 1. This implicitly holds demographic variables at the same distribution as in 1964 and allows us to see what turnout would have looked like in subsequent years had there been no change. Thus, the probability of voting in 2005, controlling for demographics (calculated from the log odds presented in Model 2 in Table 1) is \( e^{{2.08 - 1.03}} /1 + e^{{2.08 - 1.03}} = 2.83/3.83 = 0.74 \). Or, put another way, had the age, class and tenure etc. distribution of the population been the same in 2005 as it was in 1964 reported turnout among survey respondents would have been 74%, or 5 percentage points lower than what we actually observed. These calculations are repeated for each block wise step.
In the 2001 and 2005 survey the question on the perceived ideological difference was asked in the mailback part of the survey. Non responses are set to the mean for the year in question.
This fits well with conventional accounts of the difference between the parties. In the 1970s and late 1990s Harold Wilson and Tony Blair respectively, steered the Labour Party away from the Left and closer to the Centre ground, in both cases reducing the gap between the two main parties. By contrast, the 1980s was characterised by sharp conflict between Mrs. Thatcher’s brand of neo-conservative liberalism on the one hand, and Labour’s militant left on the other. This, combined with the narrow width of the confidence intervals, suggests that on the whole the electorate understood fairly well the ideological platforms of the two major parties and how they differed from each other, and that there was not much divergence in opinion.
The interaction between party identification and the expected closeness of the contest is not significant but this result is not surprising given the data constraints of having only 12 level 2 data points.
On the whole the predictions slightly over estimate the observed level of reported turnout (by about 2 percentage points), but otherwise they capture the dynamics of turnout change over time fairly accurately. From Table 4 we can see that predicted turnout in 1964 was 90.3% and predicted turnout in 2005 was 81.7%, (a decline of 8.6 percentage points which is not far off the 10 point observed decline).
References
Abramson, Paul, & Aldrich, John (1981). The decline of electoral participation in America. The American Political Science Review, 76, 502–521.
Achen, Christopher (2002). Parental socialization and rational party identification. Political Behavior, 24, 151–170.
Brians, Craig (1997). Residential mobility, voter registration, and electoral participation in Canada. Political Research Quarterly, 50, 215–227.
Budge, Ian, Klingemann, Hans-Dieter, Volkens, Andrea, Bara, Judith, & Tanenbaum, Eric (2001). Mapping policy preferences: Estimates for parties, electors, and governments. Oxford: Oxford University Press.
Butler, David, & Stokes, Donald (1969). Political change in Britain. London: Macmillan.
Campbell, Angus, Converse, Philip, Miller, Warren, & Stokes, Donald (1960). The American voter. New York: John Wiley & Sons.
Campbell, Angus, Converse, Philip, Miller, Warren, & Stokes, Donald (1966). Elections and the political order. New York: John Wiley & Sons.
Clarke, Harold, Sanders, David, Stewart, Marianne, & Whiteley, Paul (2004). Political choice in Britain. Oxford: Oxford University Press.
Converse, Philip (1969). Of time and Partisan Stability. Comparative Political Studies, 2, 139–172.
Converse, Philip (1976). The Dynamics of Party Support: Cohort-Analyzing Party Identification. Beverley Hills, CA: Sage Publications.
Crewe, Ivor, Sarlvik, Bo, & Alt, James (1977). Partisan dealignment in Britain 1964–74. British Journal of Political Science, 7, 129–190.
Dalton, Russell J. (2000). The decline of party identification. In Russell J. Dalton, & Martin P. Wattenberg (Eds.), Parties without partisans: Political change in advanced industrial democracies. Oxford: Oxford University Press.
Dalton, Russell J., & Wattenberg, Martin P. (Eds.). (2000). Parties without partisans: Political change in advanced industrial democracies. Oxford: Oxford University Press.
Denver, David, & Hands, Gordon (1974). Marginality and turnout in British general elections. British Journal of Political Science, 4, 17–37.
Denver, David, & Hands, Gordon (1985). Marginality and turnout in British general elections in the 1970s. British Journal of Political Science, 15, 381–388.
Downs, Anthony (1957). An economic theory of democracy. New York: Harper and Row.
Fiorina, Morris (1981). Retrospective voting in American national elections. Newhaven: Yale University Press.
Franklin, Mark N. (2004). Voter turnout and the dynamics of electoral competition in established democracies since 1945. Cambridge: Cambridge University Press.
Heath, Anthony, Jowell, Roger, & Curtice, John (1985). How Britain votes. Oxford: Pergamon Press.
Heath, Anthony, Jowell, Roger, & Curtice, John (2001). The rise of new labour. Oxford: Oxford University Press.
Heath, Anthony, Jowell, Roger, Curtice, John, & Taylor, Bridget (1994). Labour’s last chance? The 1992 election and beyond. Aldershot: Dartmouth.
Heath, Anthony, & Taylor, Bridget (1999). New sources of abstention? In Geoffrey Evans, & Pippa Norris (Eds.), Critical elections British parties and voters in long-term perspective. London: Sage.
Key, Valdimer Orlando (1964). Politics, Parties and Pressure Groups. New York: Crowell
Miller, Warren (1992). The puzzle transformed: Explaining declining turnout. Political Behavior, 14, 1–43.
Pattie, Charles, & Johnston, Ron (1998). Voter turnout and constituency marginality: Geography and rational choice. Area, 30, 38–48.
Pattie, Charles, & Johnston, Ron (2001). A low turnout landslide: Abstention and the British general election of 1997. Political Studies, 49, 286–305.
Rosenstone, Steven J., & Hansen, John Mark (1993). Mobilization, participation, and democracy in America. New York: Macmillan.
Stewart, Marianne, & Clarke, Harold (1998). The dynamics of party identification in federal systems: The Canadian case. American Journal of Political Science, 57, 368–377.
Stokes, Donald (1992). Valence politics. In Dennis Kavanagh (Ed.), Electoral politics. Oxford: Clarendon Press.
Swaddle, Kevin, & Heath, Anthony (1989). Official and reported turnout in the British general election of 1987. British Journal of Political Science, 19, 537–551.
Verba, Sidney, & Nie, Norman H. (1972). Participation in America. New York: Harper and Row.
Verba, Sidney, Schlozman, Kay, & Brady, Henry (1995). Voice and equality: Civic voluntarism in American politics. Cambridge, Mass: Harvard University Press.
Wattenberg, Martin P. (2000). The decline of party identification. In Russell J. Daltonand, & Martin P. Wattenberg (Eds.), Parties without partisans: Political change in advanced industrial democracies. Oxford: Oxford University Press.
Wattenberg, Martin P. (2002). Where have all the voters gone? Cambridge, Mass: Harvard University Press.
Whiteley, Paul, Clarke, Harold, Sanders, David, Stewart, Marianne (2001). Turnout. In Pippa Norris (Ed.), Britain votes 2001. Oxford: Oxford University Press.
Acknowledgements
I would like to thank David Sanders, Paul Whiteley, Anthony Heath, David Denver, John Bartle, Michael Lewis-Beck, Jouni Kuha and three anonymous reviewers for their helpful comments and suggestions. I am also grateful to the ESRC (PTA-026-27-0486) and the British Academy for research funding.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
British General Election Study, 2005
The 2005 British General Election Study was funded by the ESRC. The study was carried out by University of Essex and University of Dallas at Texas. The sample was drawn from the Postcode Address file. The fieldwork was carried out by National Centre for Social Research. The survey consisted of two waves. The first wave took place before the election. It had a response rate of 61% and achieved 3,589 interviews. The second wave took place immediately after the election. It had a response rate of 88% (of Wave 1) and achieved 2,959 interviews. In addition a top-up survey was also carried out which had a response rate of 54% and achieved 1,202 interviews. The data used in the paper comes from the post election panel, which has a valid sample size of 3,431 respondents on the electoral register. Further details are available at the BES website (http://www.essex.ac.uk/bes).
British General Election Study, 2001
The 2001 British General Election Study was funded by the ESRC. The study was carried out by University of Essex and University of Dallas at Texas. The sample was drawn from the Postcode Address file. The fieldwork was carried out by NOP. The survey consisted of two waves. The first wave took place before the election. It had a response rate of 53% and achieved 3,220 interviews. The second wave took place immediately after the election. It had a response rate of 74% (of Wave 1) and achieved 2,359 interviews. In addition a top-up survey was also carried out which had a response rate of 47% and achieved 681 interviews. The data used in the paper comes from the post election panel, which has a valid sample size of 2,081 respondents on the electoral register. Further details are available in Clarke et al. (2004: 329–336).
British General Election Study, 1997
The 1997 British General Election Study was funded by the ESRC and the Gatsby Charitable Foundation, and the Commission for Racial Equality. The study was carried out by CREST. The sample was drawn from the Postcode Address file. The fieldwork was carried out by Social and Community Planning Research in May to June 1997. The issued sample was 6,540, of which 5,814 were eligible addresses. There were 3,615 interviews giving a response rate of 63%. Further details of the survey are given in Heath et al. (2001: 170).
British Elections, 1963–1992
The combined British General Election study dataset, 1963–1992 was compiled by the Data Archive, University of Essex. The original studies were carried out by Nuffield College, 1963–1970, University of Essex, 1974–1979, and Nuffield College, 1983–1992. The samples were drawn from the Electoral Register. The achieved sample sizes were 2,009 (1963); 1,769 (1964); 1,874 (1966); 1,843 (1970); 2,462 (1974F); 2,365 (1974O); 1,893 (1979); 3,955 (1983); 3826 (1987); 3534 (1992). Further details are available in Heath et al. (1994: 302–308).
Rights and permissions
About this article
Cite this article
Heath, O. Explaining Turnout Decline in Britain, 1964–2005: Party Identification and the Political Context. Polit Behav 29, 493–516 (2007). https://doi.org/10.1007/s11109-007-9039-4
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11109-007-9039-4