Skip to main content Accessibility help

The Undoing of Winston Churchill: Mobilization and Conversion in the 1945 Realignment of British Voters

  • Mark Franklin and Matthew Ladner

We explore the reasons for the unexpected defeat of Winston Churchill's Conservatives by Labour in the British general election of 1945. Was the outcome a result of Churchill's election campaign errors, as many have supposed, or did the coming-of-age of a new political generation make it a foregone conclusion? Much controversy in the partisanship literature centres on whether electoral realignments result primarily from conversion of existing voters or from mobilization of previously non-voting individuals. In particular, the 1930s US realignment has been the focus of considerable debate. In this article we shed new light on realignment processes by examining the 1945 British realignment that brought the Labour party to power. We find that, in this more straightforward case, the critical impetus came from new voters rather than from converts. Our findings raise questions that need to be confronted in the analysis of other realignments, such as that accompanying the American New Deal. They also shed new light on a much-interpreted episode in British electoral history.

Hide All

1 See, for example, Cole, G. D. H., A History of the Labour Party since 1914 (London: Routledge and Kegan Paul, 1948); McCallum, R. and Readman, A., The British General Election of 1945 (Oxford: Oxford University Press, 1947); Taylor, A. J. P., English History 1914–1945 (Oxford: Clarendon Press, 1965); Urwin, D., Western Europe Since 1945 (New York: Longmans, 1968).

2 Attlee, Clement, As It Happened (New York: Viking, 1954), p. 148; SirChurchill, Winston, The Second World War: Triumph and Tragedy (London: Macmillan, 1954), pp. 674–5.

3 Robbins, K., Churchill (London: Longmans, 1992), pp. 147–9.

4 Charmley, John, Churchill: The End of Glory (London: Hodder and Stoughton, 1993).

5 There is of course some question as to whether 1945 should be regarded as a realigning election. During the period covered by Table 1 (see below) there were clearly several elections that might be considered realigning from one perspective or another. The 1945 election, however, was the first at which Labour won an outright plurality of the vote after a larger increase than any since the extension of the franchise in 1918. Moreover, the number of those voting Labour (even if not the percentage of the two-party vote) reached a plateau in 1945 below which it did not fall for almost forty years thereafter. So 1945 can be regarded as a realigning election in Britain in the same way that 1932 was a realigning election in the United States: it inaugurated a new era in British politics which lasted at least until the 1970s. McCallum and Readman pointed out at the time that, since 1830, on only two occasions (in 1832 and 1906) had there been larger turnovers of seats in the House of Commons (The British General Election of 1945, p. 247).

6 McCallum, and Readman, , The British General Election of 1945, p. 269; Cole, , History of the Labour Party, p. 431.

7 Pelling, H., The British Labour Governments of 1945–51 (New York: St Martin's Press, 1984), p. 30.

8 Lindsay, T. and Harrington, M., The Conservative Party 1918–1979 (London: Macmillan, 1979), p. 145.

9 Lloyd, T., Empire to Welfare State: English History 1906–1976 (Oxford: Oxford University Press, 1976), p. 267.

10 Attlee, , As it Happened, p. 144.

11 Blake, Robert, The Conservative Party from Peel to Thatcher (London: Methuen, 1985), p. 255. See also Lloyd, , Empire to Welfare State, pp. 266–8.

12 Taylor, , English History, p. 596.

13 Robbins, , Churchill, pp. 147–9.

14 McCallum, and Readman, , Election of 1945, pp. 266–71.

15 Butler, David and Stokes, Donald, Political Change in Britain, 2nd edn (London: Macmillan, 1974), pp. 172–92.

16 Andersen, Kristi, ‘Generation, Partisan Shift, and Realignment: A Glance Back at the New Deal’, in Nie, N., Verba, S. and Petrocik, J., The Changing American Voter (Cambridge, Mass.: Harvard University Press, 1976), pp. 7495.

17 Andersen, Kristi, The Creation of a Democratic Majority (Chicago: University of Chicago Press, 1979).

18 See, for example, Erikson, Robert and Tedin, Kent, ‘The 1928–1936 Realignment: The Case for the Conversion Hypothesis’, American Political Science Review, 75 (1981), 951–62; Campbell, J., ‘Sources of the New Deal Realignment: The Contributions of Conversions and Mobilization to Partisan Change’, Western Political Quarterly, 38 (1985), 357–76; Erikson, Robert and Tedin, Kent, ‘Voter Conversion and the New Deal Realignment: A Response to Campbell’, Western Political Quarterly, 39 (1986), 729–32; Campbell, J., ‘Voter Mobilization and the New Deal Realignment: A Rejoinder to Erikson and Tedin’, Western Political Quarterly, 39 (1986), 733–5.

19 Campbell, Angus et al. , The American Voter (New York: Wiley, 1960), pp. 146–7.

20 Lipset, Seymour Martin and Rokkan, Stein, ‘Cleavage Structures, Party Systems and Voter Alignments: An Introduction’, in Lipset, S. and Rokkan, S., eds, Party Systems and Voter Alignments (New York: The Free Press, 1967).

21 Butler, David and Stokes, Donald, Political Change in Britain, 1st edn (London: Macmillan, 1969), pp. 48119.

22 Campbell, et al. , American Voter, p. 150.

23 Jennings, M. Kent and Niemi, Richard, ‘The Transmission of Political Values from Parent to Child’, American Political Science Review, 62 (1968), 169–84; Jennings, M. Kent and Niemi, Richard, Generations and Politics: A Panel Study of Young Adults and their Parents (Princeton, NJ: Princeton University Press, 1981).

24 Butler, and Stokes, , Political Change, pp. 5960.

25 But see Converse, Philip, ‘Of Time and Partisan Stability’, Comparative Political Studies, 2 (1969), 139–71; and Cassel, Carol, ‘A Test of Converse's Theory of Party Support’, Journal of Politics, 55 (1993), 664–81.

26 See, for example, Beck, Paul Allen, ‘A Socialization Theory of Partisan Realignment’, in Niemi, Richard et al. , eds, The Politics of Future Citizens (San Francisco: Jossey Bass, 1974), pp. 199219; Franklin, Mark et al. , Electoral Change: Responses to Evolving Social and Attitudinal Structures in Western Countries (Cambridge: Cambridge University Press, 1992).

27 Franklin, Mark, The Decline of Class Voting in Britain (Oxford: Oxford University Press, 1985); Rose, Richard and McAllister, Ian, The Loyalties of Voters: A Lifetime Learning Model (London: Sage Publications, 1990); Franklin, et al. , Electoral Change.

28 Alford, Robert, Party and Society: The Anglo-American Democracies (Chicago: Rand McNally, 1963), p. 162; Butler, and Stokes, , Political Change, p. 78.

29 Converse, , ‘Of Time and Partisan Stability’; Lipset, and Rokkan, , ‘Cleavage Structures’.

30 Butler, and Stokes, , Political Change, pp. 172–92. The use of a model of stability to explain change is quite innovative, as we shall see. The effects of generational replacement as envisaged by Butler and Stokes are not at all the same as those envisaged by, for example, Beck in his ‘A Socialization Theory’. Beck and others see generational effects in the way new voters reflect the features of the period in which they come of age more strongly than older voters, and then retain these features more enduringly than voters who come of age in other periods. Butler and Stokes were no doubt aware of this phenomenon, but in studying political change in Britain they stressed a different feature of young voters which is to reflect characteristics of the households, neighbourhoods and schools within which they grew up. A full picture of the effects of generational replacement requires both perspectives, but for the purpose of the present article we follow Butler and Stokes in stressing the way in which new voters carry forward to the next generation differences between social groups in such a way as to perpetuate such differences (though with decreasing intensity with the passage of further generations). This mechanism is stressed by Edward Carmines and James Stimson in their Issue Evolution: Race and the Transformation of American Politics (Princeton, NJ: Princeton University Press, 1989), pp. 142–3; and by Franklin, et al. in Electoral Change, pp. 421–2.

31 See Butler, and Stokes, , Political Change, p. 174.

32 For a discussion of the relative implications of the enfranchisement of the working class and the later enfranchisement of women, see Downs, Antony, An Economic Theory of Democracy (New York: Harper, 1957), pp. 127–32.

33 Of course, the discrepancy will be partly due to the number of constituencies that Labour was unable to contest. An anonymous referee has suggested that older voters enfranchised in 1918 might have already ‘had’ a partisanship in much the same way that many people with no connection to either university ‘know’ that they support Oxford or Cambridge in the Boat Race.

34 If we assume that new voters in these years will have been socialized in Labour households in proportion to the number of Labour voters, then about a third of new voters will have had such a socialization (see Table 1). However, a larger proportion of Labour voters in the years after 1918 will have been of child-bearing age than in the population as a whole (if young voters are more likely to have adapted to new conditions) and, indeed, the working class at this time had larger families than the middle class. Moreover, some of the 1945 new voters will have acquired Labour partisanship because of neighbourhood and school influences even if their parents never voted Labour. So it is not unlikely that half the entering cohorts during this period will have been socialized in Labour environments. Indeed, given the close-knit nature of working-class life and the fact that the working class was so much larger than the middle class, the proportion might well have been higher still. See, for example, Hoggart, Richard, The Uses of Literacy (London: Penguin, 1957). Evidence from our own survey (see below) is unreliable because of the extent of missing answers to questions about parental partisanship; but among new 1945 voters who recalled their parents' class and at least one parent's partisanship (N = 312), two-thirds had a working-class childhood and half recalled their parents voting Labour. Of these, 82 per cent were voting as their parents had (88 per cent among those whose parents had both voted Labour). The big difference among older voters was fewer Labour parents (a third of the 1918–35 cohort and under a fifth of the pre-1918 cohort remembered having one or more).

35 Robbins, , Churchill, p. 148.

36 Charmley, , End of Glory, pp. 632–9.

37 Charmley, , End of Glory, pp. 640–6.

38 The small number of older voters supporting the Liberals in Table 3 below, to the extent that it is not simply a sampling artefact, is indicative of this.

39 Reported in Butler, David, The Electoral System in Britain since 1918 (Oxford: Clarendon Press, 1963), p. 186.

40 The swing was 13.6 per cent, compared with an overall swing of 13.9 per cent at the general election of 1945.

41 If the large Tory vote in these constituencies in 1935 had been disproportionately due to converts in an election when other parties were still discredited from the events of 1931, we would expect a large number of switchers back to their previous parties. This may not be what happened, but it is certainly risky to place much weight on results from three by-elections.

42 Butler, , The Electoral System, p. 187; Norris, P., By-Elections in Britain (Oxford: Clarendon Press, 1990), pp. 224–5. Of course the major parties did not oppose each other in the wartime by-elections, so expectations should hardly have been normal. Votes taken from the Tories in these elections went to the Commonwealth party–often seen as a stalking horse for Labour, but with no official connection. See Lloyd, , Empire to Welfare State, p. 259.

43 There were also two by-elections in 1941 which showed a small swing towards the Conservatives (perhaps due to patriotism at a time when the Battle of Britain was at its height). Inclusion of these results does not substantially change our findings.

44 The twenty-six by-election contests analysed by Butler in The Electoral System were ones that the Tories lost. In these constituencies the total swing away from the Tories was greater than average during the 1935–45 period, so the large extent of swing shown at the by-elections themselves is not in fact a large proportion of the total swing in these constituencies.

45 Benney, Mark, Gray, A. and Pear, R., How People Vote: A Study of Voting Behaviour in Greenwich (London: Routledge and Kegan Paul, 1956), p. 109.

46 Alford, , Party and Society, p. 162.

47 Anderson employed the 1952 election study conducted by Campbell, Converse, Miller and Stokes: the first of the studies they fielded in the course of the project that led to the publication of The American Voter. We employed the 1963 pre-election and 1964 election studies conducted by Butler and Stokes: the first of the studies they fielded in the course of the project that led to the publication of Political Change in Britain. The Butler and Stokes study was made available to us by the Inter-University Consortium for Political and Social Research at the University of Michigan. Of course, neither the ICPSR nor the originators of the data bear any responsibility for the use we have made of it.

48 Evidently the weighting scheme required us to give more weight to older voters, many of whom had died by the time of the later survey. In our study, those aged less than 50 in 1945 were quite well represented in 1963 (being no more than 67 years old at the time of the survey), but those aged 50 or more in 1945 were increasingly under-represented, with virtually no one left who had been over 60. However, because those aged 50 or more in 1945 were born in 1895 or earlier, all of them will have received their political socialization during the period before the extension of the franchise in 1918, and there is no need for us to assume that those aged over 60 in 1945 would have behaved any differently than those aged over 50. By treating all those over 50 in 1945 as a single group and weighting them up by enough to reproduce their actual proportion in the 1945 electorate, we ensure that the diversity of this group in terms of class and other characteristics is retained in our simulated electorate, at the expense of accurately representing the specific age groups over 50. Because of the higher rates of working-class mortality (shown in Butler, and Stokes, , Political Change, pp. 226–7), it was also necessary to weight the dataset by occupational group. This was done with the aid of class and age breakdowns in the 1951 British census (see fn. 55).

49 We do not know which party those who had previously voted Liberal voted for in 1945 if they returned to voting Liberal thereafter. We assume them to have voted Labour because this makes it more difficult rather than less difficult to prove our hypothesis, as we shall see. At the same time, we do not know which if any of those who temporarily shifted from Labour or Conservative in 1945 voted Liberal. Since there were more switchers from Conservative than from Labour, our assumption that such switchers moved to the other major party again makes it harder for us to confirm our hypothesis (see below).

50 These parties received only about 2 per cent of the vote in each of the 1935 and 1945 elections. Even if all of their supporters were involved in complex three-way shifts, which is unlikely, the effect on the fortunes of the major parties would have been minimal. So we assume that shifts to and from minor party voting will have largely cancelled out. Our data also ignores the possibility that respondents might have shifted repeatedly between parties, but this is no great problem since repeated switchers can hardly have been responsible for an enduring realignment. Any imbalance might have been expected to favour Labour, since that was the party that gained in the election and also the party that stood highest in public opinion at the time of our surveys; however, we will see below that our reconstruction does not overcount Labour voters but only the Conservatives. Thus, any imbalance in switching miscounted in this way will make it still more difficult to justify the conversion hypothesis according to the logic used in Section 9 below.

51 One of the most telling criticisms levelled at Kristi Andersen's reconstruction of the New Deal electorate from 1952 survey data was that this reconstruction failed to find almost any blacks or Jews voting Republican in the 1920s, though contemporary accounts tell us that these groups voted overwhelmingly Republican before the New Deal. By deduction, ‘Blacks and Jews… simply projected their contemporary [Democratic] partisanship backward to when most were in fact Republican’ (see Reiter, H., ‘The Perils of Partisan Recall’, Public Opinion Quarterly, 44 (1980), 385–8, p. 388). However, this deduction is by no means watertight. Younger black and Jewish voters might have been voting Democrat in some numbers even before the New Deal without this being noticed, in the absence of survey data, and few of those who had been over 45 years old in 1928 will have survived to be interviewed in 1952.

52 This question needs to be addressed with a long panel study–longer than is normal in election studies and with more respondents than is normal. Evidence from the 1963–64–66–70 panel embedded in our data is promising but cannot be conclusive. There are insufficient 1970 respondents who were new in 1964 to enable a comparison to be made, and only 8 per cent of voters who were interviewed both in 1964 and 1966 mis-recalled their 1964 vote; but the number was much greater (45 percent) among those respondents ( 13 percent of the sample) who had changed their partisanship between the two elections. Among these switchers, the rate of faulty recall was only 14 per cent for the small number of voters (less than 1 per cent of the sample) who were too young to have voted before 1964. If these findings are generalizable, then less than one-third as many voters misremember their earliest vote than misremember some other past vote. Importantly, this dataset is the same one as that used by Niemi, Katz and Neumann in their critique of the use of recall data (see Niemi, R., Katz, R. and Newman, D., ‘Reconstructing Past Partisanship: The Failure of the Party Identification Recall’, American Journal of Political Science, 24 (1980), 633–51.

53 One indication that we more accurately reconstruct the more distant election is that while fewer than 12 per cent of 1964 voters who voted in both 1964 and 1959 remember voting differently in 1959, fully 21 per cent of those who voted in both 1964 and 1945 remember voting differently in 1945.

54 Note that the age structure of the 1945 electorate is not subject to recall bias since our data are weighted to the census tables.

55 Registrar-General of England and Wales, and Scotland, Annual Abstract of Statistics, reprinted in Butler, David and Butler, Gareth, British Political Facts (London: Macmillan, 1986), p. 325. There was in fact no census between 1931 and 1951, and although data for 1939 were estimated by the Census Office and included in their tables, we preferred to simply employ the results of the 1951 Census. From the total population we subtracted those under 21 and then recomputed the percentages to provide us with figures for the electorate. Only about 3 per cent of the voting-age population was not registered to vote, so there is not much room for bias on this account. However, turnout was only 73 per cent, and so it is possible that the proportion of those voting for the first time as a proportion of all those voting was less than the 20.6 per cent estimated by our procedures. The 20 per cent figure given in the text is a deliberately round number. The actual number could have been as low as 18 per cent or as high as 22 per cent on reasonable assumptions about differential turnout by age.

56 Since we have no survey conducted in 1935 we have no direct evidence about the nature or numbers of those who had died. We can, however, impute their characteristics if we assume them to be much the same as those surviving in the oldest cohorts (see below).

57 We will discuss the accuracy of the reconstruction below.

58 See Niemi, , Katz, and Newman, , ‘Reconstructing Past Partisanship’; Reiter, , ‘Perils of Partisan Recall’; Waldahl, R. and Aardal, B., ‘Can We Trust Recall Data?Scandinavian Political Studies, 5 (1982), 101–16.

59 When we reconstruct the 1945 electorate uniquely from individuals who thought that election to have been the most important election occurring in their lifetimes, we do find more respondents who recalled having switched between the time of their earliest vote and 1945. However, by far the greater number of additional switchers remembered having switched to the Tories before or during the 1945 election campaign. The proportion of those switching to Labour is hardly changed by focusing on those who presumably remembered the election best.

60 Niemi, , Katz, and Newman, , ‘Reconstructing Past Partisanship’.

61 Niemi, , Katz, and Newman, , ‘Reconstructing Past Partisanship’; Reiter, , ‘The Perils of Recall’; Himmelweit, H., Biberian, M. and Stockdale, J., ‘Memory for Past Vote: Implications of Study of Bias in Recall’, British Journal of Political Science, 8 (1978), 365–84.

62 Mackie, T. and Rose, R., The International Almanac of Electoral History (London: Macmillan, 1991), p. 449.

63 This match is, however, far closer than that found in a contemporary Gallup survey conducted in October 1945 (just four months after the election) which estimates a Tory vote of only 31.8 per cent compared to a Labour vote of 56.9 per cent, exaggerating the Labour victory three-fold. The fact that we do far better with our reconstructed data than with a contemporary survey is further evidence that recall of initial partisanship is more accurate than recall of immediate past partisanship. The dataset, which was obtained from the Roper Center, University of Connecticut, in the hopes that it could be used to validate our reconstruction of 1945 partisanship, proved completely inadequate to this task. Not only was the 1945 vote hopelessly misestimated, but so were age and class characteristics; bringing into question the basis of the quota sampling methods employed and rendering the survey useless for validation purposes. Contemporary surveys conducted just before the election could apparently be more accurate (see for example the News Chronicle, 4 July 1945)–perhaps with the benefit of an unreported weighting scheme–but none such could be obtained for secondary analysis.

64 The 1964 election outcome estimated from these data is 46 to 42 in favour of Labour–2 per cent more Labour votes than occurred in fact. The reconstruction of the 1959 election outcome on the basis of recall from 1963 notoriously yields a Labour victory for that year, despite the fact that the Conservatives under Harold Macmillan actually won that election by a handsome margin. We cannot use the difference between these two surveys to validate our reconstruction because the number of additional voters interviewed in 1964 who were old enough to have voted in 1945 were too few for reliable comparisons to be made.

65 It is logically possible, if implausible, that some 1945 Labour voters might have switched to the Tories thereafter without recalling that they had once voted Labour. This possibility was evaluated in other work where it was found that even under the most generous assumptions (for the conversion hypothesis) about the source and destination of any such switchers, the gap between Labour and Tory among older voters did not drop below 2.8 per cent in the Tories' favour. See Franklin, M. and Ladner, M., ‘Partisan Identification and the Labour Party Victory of 1945: Conversion or Mobilization as the Basis for Electoral Realignment in Britain?’ (paper presented at the European Consortium for Political Research Joint Sessions Workshops, Limerick, 1992).

66 Of course it is true that we do overestimate the Conservative vote in 1945, thus making the conversion hypothesis more difficult to confirm; but even if we reduce the Tory vote in the top two rows of Table 3 by the required 3.5 per cent, these cohorts remain on average 3 per cent more Tory than Labour–though this is within the range of sampling error. This does not, however, invalidate our estimate of the shortfall in the number of switchers (see below).

67 During the years before 1945, millions of young men were taken from their homes and socialized with other young men in military conditions. This experience might easily have been a more potent force in generating Labour party support than home and neighbourhood influences. But then the effect should have been felt equally by middle-class and working-class young voters. This did not happen. Indeed, in our reconstructed dataset the tendency of voters to support the party of their class is greater among first-time than among more established voters in 1945. The same finding makes it appear that no general pro-Labour mood simply swept all young voters indiscriminately into the Labour camp (see Section 4, above). Another way to check for this is to contrast the votes of young men with those of young women, who (though many of them did indeed move away from home to help the war effort) were not uprooted to nearly the same extent. Any gender gap in Labour voting will tend to undermine our conclusion that early socialization was important (we are grateful to Ivor Crewe for suggesting this critical test). In fact there is some evidence of such a gap. Female voters in the 21–31 age bracket are 8 per cent less likely to have supported the Labour party than male voters of the same age. However, this gender gap is actually a little smaller than that found among older voters (where it averaged 10 percent). There is thus no evidence that wartime experience had a different effect on men than on women.

68 We have already indicated that contemporary surveys found a strong link between class and vote (see fn. 28).

69 Erikson, and Tedin, , ‘The Case for the Conversion Hypothesis’; Erikson, and Tedin, , ‘Voter Conversion’; Campbell, ‘Sources of the New Deal Realignment’; Campbell, , ‘Voter Mobilization’.

70 Andersen, , Creation of a Democratic Majority. Our findings suggest the need to distinguish between recall of recent past elections and recall of elections that might be considered particularly salient to respondents, either because they are the first election in which respondents voted or for other reasons.

71 Pelling, , British Labour Governments, p. 17.

* Department of Political Science, University of Houston. This is a revised version of a paper presented at the Annual Meeting of the American Political Science Association, Chicago, 1992. The authors would like to thank Ivor Crewe, Robert Erikson, Tim Fackler, Richard Matland, Susan Scarrow, Michael Taylor, Christopher Wlezien and two anonymous referees for helpful comments on earlier 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: 0
Total number of PDF views: 0 *
Loading metrics...

Abstract views

Total abstract views: 0 *
Loading metrics...

* Views captured on Cambridge Core between <date>. This data will be updated every 24 hours.

Usage data cannot currently be displayed