Abstract

In the autumn of 2011, there was significant discussion in the British press about the level of support among women for the Conservatives and for Prime Minister David Cameron, arguing that he had a ‘problem with women’ and suggesting that this had arisen since the general election the previous year. We examine the opinion polls over the period, and find that they mostly point towards an opposite conclusion. Further, there seems little to suggest that attitudes to David Cameron personally were damaging Conservative standing among women, as much of the coverage implied. We consider how the reporting came to give a misleading impression, and why it matters.

In the autumn of 2011, some sections of the British media began to discuss the low (or lower than expected) support that the Conservatives were drawing from women. This quickly came to be encapsulated in the phrase ‘problem with women’. On 2 October, Channel 4 News was asking ‘Do the Conservatives have a problem with women?’(Channel 4 News, 2011), and Yvette Cooper (the Labour Party's Shadow Minister for Women and Equalities) was saying ‘It is typical that David Cameron thinks his only problem with women is spin and presentation’ (Cooper, 2011).

In the next few weeks, ‘Cameron's problem with women’ became a cliché, used or referred to in headlines by, among others, writers in the Independent (‘Cameron's problem with women’—Sieghart, 2011), Daily Mail (‘Calm down, Dave. Your real problem's that women can spot a fake a mile away’—Phillips, 2011), New Statesman (‘David Cameron's problem with women’—New Statesman, 2011), Spectator (‘Like the Conservative Party, I have a problem with women’—Rifkind, 2011), Guardian (‘David Cameron's trouble with women makes Theresa May close to unsackable’—Stratton, 2011) and Telegraph blog (‘David Cameron's “woman problem” is not something a tokenistic SpAd appointment will solve’—Knowles, 2011). By 19 December the Daily Mail was allusively headlining the same phrase to discuss Boris Johnson (‘Does Boris Johnson Have a Woman (Vote) Problem?’—Purnell, 2011).

Clearly the phrase ‘Cameron's problem with women’ has (and was intended to have) connotations going far beyond opinion poll standings; but the underlying assumption throughout was that the opinion polls at the time demonstrated that female support for the Conservatives was in some sense lower than it might have been or ought to have been, and perhaps that there was direct evidence that this was specifically related to women's reactions to David Cameron rather than a more general dislike of the coalition or its policies.

The idea of ‘Cameron's problem with women’ fitted the narrative that reflected opposition attacks on Cameron, theories of differentially severe impacts on women of coalition policies and the Conservatives' own concerns about the attitudes of female voters. Yet although the unanimity of the coverage created the impression that the existence of the ‘problem’ was supported by a solid foundation of uncontested fact, that was far from the case. Systematic examination of the opinion polls at the time finds no consensus on any of the aspects of the ‘problem’ that were stated or suggested.

Polls, news about polls and public opinion

It is now well understood that one of the most significant influences on political public opinion is news reporting by the media. Reporting is necessarily selective since the supply of reportable potential news is effectively infinite, and the choice of what to report exerts a powerful agenda-setting effect (McCombs and Shaw, 1972; Semetko et al., 1991; McCombs, 2004). Furthermore, the media cannot avoid the need to make sense of what they report and to put it into context, which involves choosing frames through which their audiences interpret what they see and draw conclusions from it. The framing of issues by the news media can shape the way the public understands political issues (Goffman, 1974; Iyengar, 1991; Nelson et al., 1997), and an important element in the power of framing effects is the impact of repetition (Chong and Druckman, 2007). Furthermore, politicians are forced to react to the news agenda, to answer questions from interviewers, to explain their policies in new ways or from new perspectives, perhaps even to change their policies, in the hope of receiving more favourable coverage, of offering new frames to the voters or of changing the agenda.

Attitudes towards individual politicians are considered one area of public opinion in which these forces are particularly potent, notably through priming. In this process, the (agenda set) perceived salience of different issues and (the framed interpretations of) what a citizen knows about them can affect the criteria that the citizen uses to evaluate a leader or candidate (Iyengar et al., 1982; Iyengar and Kinder, 1987; Stevens et al., 2011). Moreover, political scientists are now coming to accept that the public's perceptions of the party leaders have an important, perhaps increasingly important, influence on voting in British elections (Mughan, 2000, p. 49; Clarke et al., 2004; Heffernan and Webb, 2005, pp. 48–55).

From the power of the news follows the importance of accuracy in reporting, including in the coverage of opinion polls. Because people care (consciously or unconsciously) about what other people think, reports of the state of public opinion can be potent elements in framing other news stories, and if stories are framed with reference to inaccurate or misleading reports of public opinion, this risks distortion of the political debate. Many studies have argued that opinion poll findings can in some circumstances affect public opinion, for example through ‘bandwagon effects’ (Whiteley, 1986; McAllister and Studlar, 1991) and perhaps also through ‘underdog effects’. This may involve mechanisms such as the spiral of silence (Noelle-Neumann, 1993), and may not necessarily be dependent on published poll findings—Noelle-Neumann theorises that people perceive the public mood through some ‘quasi-statistical organ’. Nevertheless, news reports about polls play an important role. Recent research suggests that fear of social isolation prompts some people to seek out information about public opinion, and this manifests itself in attention paid to media stories about opinion polls (Hayes et al., 2013).

There is evidence that reports showing movements in attitudes, rather than merely static opinions, may be particularly influential (Marsh, 1984; Marsh and O'Brien, 1989). Further, Marsh noted in passing (1984, p. 73) that this may mean that the most influential reports may be those based on the most unreliable polls. The same argument holds where the polls are reliable but the reporting is misleading, drawing conclusions of change which the poll findings do not support. We examine one such instance, as a case study, in which news stories about the government and Prime Minister were repeatedly framed by reference to a shift in public opinion of which the polls gave little evidence.

Gender differences in political opinion have received considerable scholarly attention; they also make an appealing topic for journalists, as well as being of practical concern to politicians. Historically, voting in Britain has displayed a ‘gender gap’ under which the Conservatives' advantage over Labour has been greater among women than among men, although this may always have been a simplistic interpretation: certainly, by the 1980s it was more useful to consider a gender-generation gap, which helped to explain why the overall scale of the gender gap had fallen (Norris, 1999). By the last election of Tony Blair's premiership, the overall gap had reversed, so that the Conservatives were now winning votes less effectively from women than from men (Campbell, 2006, pp. 115–119). This prompted concern and debate within the Conservative Party (Campbell et al., 2006), and David Cameron's conscious attempt to ‘feminise’ the Conservative Party was an important element in the project of modernisation on which he embarked after he became leader (Childs and Webb, 2012).

Since the 2010 election, the question of women's attitudes towards Cameron and the Conservatives has been placed firmly on the political agenda by debate over the Coalition's austerity policies. The Fawcett Society unsuccessfully attempted a legal challenge to the 2010 budget under gender equality laws (Press Association, 2010), and has continued to argue that the cuts have had a disproportionate impact on women (Sands, 2012). David Cameron recognised its relevance when he used interviews on 2 October 2011 in the Sunday Times and on BBC1’s Andrew Marr show to apologise for any offence he had caused by two much-criticised incidents in the House of Commons, both putdowns of female MPs. (He had responded to Labour frontbencher Angela Eagle with a catchphrase from a well-known TV advert, ‘Calm down, dear’, and had evaded a question from his own backbencher, Nadine Dorries, by joking that she was ‘extremely frustrated’.) These interviews prompted Yvette Cooper's attack.

The question of whether a new Conservative ‘problem with women’ emerged after 2010 was therefore of political as well as theoretical significance, with the reporting having the potential to affect public attitudes to David Cameron and the Conservatives, and perhaps even to create a ‘problem’ if it did not already exist. A further reason for caution was that the initial story emanated from an interested source—in fact, it seems, from within the Conservative Party itself. In June, the Sunday Times and the Spectator both reported (McDonagh, 2011; Oakeshott and McGee, 2011) that the Conservatives had become concerned about their standing with women after studying the polls during the 2011 local elections, and the Sunday Times used the phrase ‘the party has a problem with women’. Soon, David Cameron himself was said to be ‘poring “almost masochistically”’ over the polling evidence (Kite, 2011). Further evidence of Tory concern came from the leaking of an internal Conservative ‘Restricted—Policy’ memo, published in the Guardian on 13 September (Curtis and Stratton, 2011). This stated that ‘We know from a range of polls that women are significantly more negative about the Government than men’, and went on to present ‘a long list of ideas’, including policy initiatives as well as communications strategies, to rectify the situation (Anon, 2011).

Of course, the press was fully justified in reporting that Cameron believed he had a ‘problem with women’. But they might at the same time have questioned whether that problem was real, and (if so) whether there was more than one possible view of its implications.

The voting intention polls in 2011

The press coverage in the Autumn of 2011 was not entirely in agreement on how the ‘problem’ was being manifested. Between them, the articles stated, implied or at least might have led readers to suppose any or all of the following: that Conservative support among women was significantly lower than at the 2010 election; that Conservative support among women was significantly lower than Conservative support among men; that Conservative support among women since the 2010 election had fallen significantly more than Conservative support among men; or, failing that, at the very least that Conservative support among women was significantly lower than at some previous point since the 2010 election. Some also implied that these were true because of David Cameron's personal performance, or that his personal standing was lower among women than among men.

We need, therefore, to take each of these propositions as hypotheses to be tested against the polling evidence that was then available. Analysis of the 2010 British Election Study suggests that the Conservatives achieved a bigger swing in their favour among women than among men at that election (Childs and Webb, 2012, p. 184) and, while all three party leaders were better liked by women than by men at the time of the election, Cameron was the only one of the three who was more popular than his party (Childs and Webb, 2012, p. 199). But this analysis had not been published in 2011 and, in any case, related only to the election and not to what happened afterwards.

In considering the information on which journalists could base their stories, it is essential to bear in mind the limited evidential value of any single poll, especially in measuring voting intentions. Sample sizes in newspaper polls are small compared with those used in election studies, and fine analysis of subgroups is not possible. For most telephone polls, after discarding ‘don't knows’, adjusting for turnout effects and making allowance for design effects caused by weighting, the effective sample size with which voting intention is measured is generally between 500 and 700 respondents—for the voting intentions of women alone the number is of course about half this. Substantial swings from poll to poll are therefore to be expected from simple sampling variation, even before any ephemeral political effects are taken into account. Moreover, differences in methodology between polling companies mean that there are often small but systematic differences in their findings—‘house effects’ (Smith, 1978; Kellner et al., 2011)—which mean they cannot simply be pooled to give more sensitive measurements; nor are comparisons between polls in different series a valid indication of trend.

We looked at polls from the five established polling companies (those that have published polls regularly since before the 2005 general election): by Ipsos MORI for Reuters, ICM for the Guardian, Populus for The Times and ComRes for the Independent (all four of these poll series being telephone polls) and also online polls by ComRes for the Independent on Sunday/Sunday Mirror and YouGov for the Sunday Times. It is perhaps worth noting that there is no prima facie reason to favour the evidence of polls from one of these companies over another if they disagree: all five were satisfactorily accurate in their predictions of the Conservative vote at the 2005 and 2010 elections, with only negligible differences between them (Worcester et al., 2011, p. 310).

To derive reasonably robust measurements with which to test the various propositions suggested by the press coverage, we take the average of each company's findings across a four-month period. For each poll we compare the headline voting intention for men and women, averaging each pollster's polls within a month when more than one was conducted, then averaging the monthly figures across the four-month period.2 The possibility of house effects means that to detect change we must compare each company's polls with the previous polls from the same company; similarly, to measure vote share at the 2010 general election we rely on each company's final pre-election ‘prediction’ poll except for Ipsos MORI, who published a post-election estimate of the voting based on an aggregation of polls, which we use instead.

Table 1 shows the Conservative voting intention share among men and women reported by the five companies at the election and across the September to December 2010, June to September 2011 and September to December 2011 periods. As far as testing the general impression which most of the ‘problem with women’ reports conveyed, it is the comparison between the election voting and the poll standings in June to September 2011 which is key. These were the most recent polls at the time that reporting of ‘Cameron's problem’ became endemic, and therefore the best evidence available to the journalists to judge the accuracy of the trends that they were reporting as news.

Table 1.

Conservative voting share among men and women according to five pollsters

Pollster Period Con share (men) % Con share (women) % ‘Gender gap’ 
ComRes Election 2010 39 35 −4 
 September–December 2010 38 36 −2 
 June–September 2011 36 37 +1 
 September–December 2011 38 36 −2 
ICM Election 2010 39 36 −3 
 September–December 2010 37 36 −1 
 June–September 2011 37 37 
 September–December 2011 39 32 −7 
Ipsos MORI Election 2010 38 36 −2 
 September–December 2010 39 36 −3 
 June–September 2011 35 33 −2 
 September–December 2011 36 34 −2 
Populus Election 2010 37 37 
 September–December 2010 37 39 +2 
 June–September 2011 37 34 −3 
 September–December 2011 35 31 −4 
YouGov Election 2010 33 37 +4 
 September–December 2010 38 44 +6 
 June–September 2011 35 38 +3 
 September–December 2011 37 37 
Pollster Period Con share (men) % Con share (women) % ‘Gender gap’ 
ComRes Election 2010 39 35 −4 
 September–December 2010 38 36 −2 
 June–September 2011 36 37 +1 
 September–December 2011 38 36 −2 
ICM Election 2010 39 36 −3 
 September–December 2010 37 36 −1 
 June–September 2011 37 37 
 September–December 2011 39 32 −7 
Ipsos MORI Election 2010 38 36 −2 
 September–December 2010 39 36 −3 
 June–September 2011 35 33 −2 
 September–December 2011 36 34 −2 
Populus Election 2010 37 37 
 September–December 2010 37 39 +2 
 June–September 2011 37 34 −3 
 September–December 2011 35 31 −4 
YouGov Election 2010 33 37 +4 
 September–December 2010 38 44 +6 
 June–September 2011 35 38 +3 
 September–December 2011 37 37 

It is perhaps simplest to present the conclusions in tabular form (Table 2). It will be seen that if we take the June to September 2011 period as the criterion, none of the four propositions is supported to a statistically significant degree by three of the five pollsters; in fact only one is supported by any of them, and that the one with the most ambiguous implications. At the time the press was writing so prolifically about a Conservative ‘problem with women’, the party's support among women does not seem to have been lower than their support among men, had not fallen more than their support among men had since the general election, and was not significantly lower than at the general election. The only proposition to find even partial support was that support in 2011 was lower than it had been towards the end of 2010. Yet the implication conveyed by the press coverage was not that the ‘problem’ consisted simply of the fading of a brief post-election boom in female support, but rather that the Conservatives were significantly weaker among women than among men, and that this was new. So far as voting intention support goes, the evidence was almost entirely contrary to the impression that the coverage must have given.

Table 2.

Summary of pollsters' findings

 Com Res ICM Ipsos MORI Populus You Gov 
Conservative support among women was significantly lower in mid-2011 than at the 2010 election No No Noa Noa No 
Conservative support among women in mid-2011 was significantly lower than Conservative support among men No No No Noa No 
Conservative support among women fell significantly more than Conservative support among men between the 2010 election and mid-2011 No No No Noa No 
Conservative support among women was significantly lower in mid-2011 than in late 2010 No No Noa Yes Yes 
 Com Res ICM Ipsos MORI Populus You Gov 
Conservative support among women was significantly lower in mid-2011 than at the 2010 election No No Noa Noa No 
Conservative support among women in mid-2011 was significantly lower than Conservative support among men No No No Noa No 
Conservative support among women fell significantly more than Conservative support among men between the 2010 election and mid-2011 No No No Noa No 
Conservative support among women was significantly lower in mid-2011 than in late 2010 No No Noa Yes Yes 

aA 3-point difference, but not statistically significant.

It is true that in the September to December 2011 figures the evidence against a Conservative ‘problem’ is slightly less unanimous. (Three pollsters out of five showed a movement in the gender gap since the election, performance among women having been weaker than among men. But only two found support among women significantly lower than it had been at the general election or significantly lower than support among men.) However, this draws on polls that had not yet been conducted when the press coverage was at its height, and may reflect changes in attitudes after ‘Cameron's problem with women’ was already being reported. It certainly cannot justify regarding the ‘problem’ as already established fact by the end of September.

Polls on other political attitudes

The separate question of how public perceptions of David Cameron and his government interacted with Conservative support cannot be addressed from the voting intention polls alone. The relevant questions are mostly not asked in a comparable way by more than one polling company; we therefore confine our analysis in this section to Ipsos MORI data.

Ipsos MORI's monthly satisfaction questions distinguish between attitudes to the government and the Prime Minister, and are also not affected (as voting intentions probably are) by comparisons with the opposition parties and the perceived fitness of the alternative government to take power. Table 3 shows these answers aggregated over the whole post-election period of 2010, and all of 2011. It will be seen that the figures for the June to September and September to December periods of 2011, on which the voting intention analysis concentrated, do not differ materially from those for the whole year.

Table 3.

Satisfaction with the government and Prime Minister

Q. Are you satisfied or dissatisfied with the way the government is running the country? 
 
Q. Are you satisfied or dissatisfied with the way Mr Cameron is doing his job as Prime Minister? 
 
 2010 post-election
 
2011
 
June– September 2011
 
September–December 2011
 
 
 
Government 
 Satisfied 44 37 35 30 35 29 36 29 
 Dissatisfied 43 48 57 62 57 60 58 62 
 Net +1 −11 −22 −32 −22 −31 −22 −33 
David Cameron 
 Satisfied 54 50 44 38 43 38 43 38 
 Dissatisfied 36 36 50 51 50 51 50 52 
 Net +18 +14 −6 −13 −7 −13 −7 −14 
Q. Are you satisfied or dissatisfied with the way the government is running the country? 
 
Q. Are you satisfied or dissatisfied with the way Mr Cameron is doing his job as Prime Minister? 
 
 2010 post-election
 
2011
 
June– September 2011
 
September–December 2011
 
 
 
Government 
 Satisfied 44 37 35 30 35 29 36 29 
 Dissatisfied 43 48 57 62 57 60 58 62 
 Net +1 −11 −22 −32 −22 −31 −22 −33 
David Cameron 
 Satisfied 54 50 44 38 43 38 43 38 
 Dissatisfied 36 36 50 51 50 51 50 52 
 Net +18 +14 −6 −13 −7 −13 −7 −14 

Source: Ipsos MORI.

Base: c. 1000 GB adults aged 18+ each month.

The most obvious feature of the figures is the trend—both government and Prime Minister steadily lost support as the parliament progressed. But there is also a clear difference between men and women across the whole period, with net satisfaction lower for women than for men in each case—furthermore, this gender gap is noticeably bigger than the gender gap in voting intentions in the same polls. In the case of satisfaction with the government the gender gap appears static, perhaps even closing fractionally; for satisfaction with Cameron, on the other hand, the gap seems to have widened a little. Nevertheless, although this movement is just statistically significant and may be an indicator of underlying political forces, it is too small a change in itself to be of much political substance. On the whole we would be justified in saying that women were clearly and consistently less approving of the performance of the government and Prime Minister than men, but there is no evidence to suggest any great change in the situation as time has gone by. If there is a ‘problem with women’, it is not on this evidence one that was new in 2011.

More detailed consideration of views of the Conservatives and of Cameron is offered by Ipsos MORI's party and leader image questions. Unlike the satisfaction questions, these are only occasionally included in the monthly survey, but conveniently the September 2011 survey was one in which they appeared. These too show much more clearly than the voting intentions a less favourable view taken by women than by men (Tables 4 and 5). Of the nine descriptions of the Conservative Party that were tested, women took a significantly more critical view of the party than men on seven, and on none were women significantly more favourable. The story is the same with attributes applied to Cameron personally: five of nine on which women rated Cameron worse, four where there was little to choose between women and men and none on which women were more generous.

Table 4.

Conservative Party image, September 2011

Q. I am going to read out some things both favourable and unfavourable that have been said about various political parties. Which of these, if any, do you think apply to the Conservative Party? 
 
 Men % Women % 
Keeps its promises 31 19 
Understands the problems facing Britain 55 45 
Has a good team of leaders 48 41 
Will promise anything to win votes 59 70 
Divided 49 50 
Extreme 27 30 
Looks after the interests of people like me 37 26 
Fit to govern 57 44 
Out of date 44 40 
None of these 
Don't know 
Average on positive descriptions 46 35 
Average on negative descriptions 45 48 
Q. I am going to read out some things both favourable and unfavourable that have been said about various political parties. Which of these, if any, do you think apply to the Conservative Party? 
 
 Men % Women % 
Keeps its promises 31 19 
Understands the problems facing Britain 55 45 
Has a good team of leaders 48 41 
Will promise anything to win votes 59 70 
Divided 49 50 
Extreme 27 30 
Looks after the interests of people like me 37 26 
Fit to govern 57 44 
Out of date 44 40 
None of these 
Don't know 
Average on positive descriptions 46 35 
Average on negative descriptions 45 48 

Source: Ipsos MORI.

Base: 1008 GB adults aged 18+, 10–12 September 2011.

Table 5.

David Cameron's image, September 2011

Q. I am going to read out some things both favourable and unfavourable that have been said about various politicians. Which of these, if any, do you think apply to David Cameron? 
 
 Men % Women % 
A capable leader 63 56 
Good in a crisis 51 44 
Has sound judgement 49 42 
Understands the problems facing Britain 58 50 
Likeable 58 56 
Out of touch with ordinary people 62 64 
Puts the interests of the country above those of his own party 45 34 
Has a clear vision for Britain 57 56 
More style than substance 52 53 
None of these 
Don't know 
Average on positive descriptions 54 48 
Average on negative descriptions 57 59 
Q. I am going to read out some things both favourable and unfavourable that have been said about various politicians. Which of these, if any, do you think apply to David Cameron? 
 
 Men % Women % 
A capable leader 63 56 
Good in a crisis 51 44 
Has sound judgement 49 42 
Understands the problems facing Britain 58 50 
Likeable 58 56 
Out of touch with ordinary people 62 64 
Puts the interests of the country above those of his own party 45 34 
Has a clear vision for Britain 57 56 
More style than substance 52 53 
None of these 
Don't know 
Average on positive descriptions 54 48 
Average on negative descriptions 57 59 

Source: Ipsos MORI.

Base: 1008 GB adults aged 18+, 10–12 September 2011.

However, while the gulf in opinions between men and women on some of the questions is wide, the difference is considerably bigger on positive than negative descriptions.

Taking the average of these four comparisons (two positive and two negative) we find a gender gap on image of only five percentage points. As the voting intentions gap was higher with these respondents than the Ipsos MORI average (the Conservatives scored 39% among men and 32% among women in this poll), the difference in the image profiles is therefore well in line with the difference in voting intentions after all.

Cameron or the party?

Let us now consider how far the evidence supports blaming David Cameron for any shortcomings in the Conservatives' ratings among women. Returning to Table 3, we can see that throughout the period, David Cameron's rating as Prime Minister was consistently better than that of the government he led. That is almost invariably true of any Prime Minister, and may owe something to a greater reluctance among some of the public to criticise people than to criticise institutions; it may or may not indicate that attitudes towards the Prime Minister are not the government's main image problem. Nevertheless, for our purposes we need to note that this is true for men and women alike. In fact in 2010 and 2011 Cameron's satisfaction score surpassed the government's score by a wider margin among women than men.

Similarly in the image questions (Tables 4 and 5), while the ratings are consistently worse among women this is no truer of responses about Cameron than about the Conservatives. The gender gap is in fact slightly more marked in the responses about the party than in the responses about Cameron. Furthermore, if women's negative attitudes towards Cameron were driven by personal factors such as his patronising behaviour towards female MPs, as some of his critics were suggesting, we would surely expect to find this reflected most strongly in the ‘empathetic’ assessments of him rather than in assessments of his competence. Yet women differed most strongly from men in doubting his capability as a leader, sound judgement and whether he was good in a crisis, while being almost as likely as men to find him likeable and not significantly more likely to feel him ‘out of touch’. Nor was this gender gap a new one: his ratings among women were already lower than among men in an Ipsos MORI poll, using similar though not identical questions, conducted almost immediately after the election.

We can draw similar conclusions from a question perhaps better suited to distinguishing between attitudes to Cameron and his party. In January 2011 and again in October 2012, Ipsos MORI asked their respondents—in a single question——whether they liked Cameron and whether they liked the Conservative Party. Here, therefore, to the fullest extent, they were asked to disentangle their views of the party from the leader, rather than reflecting their views of one in their responses about the other. If Cameron was in any sense a liability to his party it ought to be evident here. In fact, as far as the gender gap is concerned, no such situation emerges (Table 6): the views of men and women are almost identical (and, furthermore, in both sexes twice as many say they like Cameron but not the Conservatives as say the opposite). This would seem to confirm the impression that, however deep the Conservatives' ‘problem with women’, it is probably both unjust and a misunderstanding of the causes of the situation to portray it as being ‘Cameron's problem’.3

Table 6.

Like Him? Like His Party?

Q. Which of these statements comes closest to your views of David Cameron and the Conservative Party? 
 
 January 2011
 
October 2012
 
 Men % Women % Men % Women % 
I like David Cameron and I like the Conservative Party 27 27 26 24 
I like David Cameron but I do not like the Conservative Party 17 17 17 16 
I do not like David Cameron but I like the Conservative Party 11 10 
I do not like David Cameron and I do not like the Conservative Party 42 41 43 45 
Don't know 
Q. Which of these statements comes closest to your views of David Cameron and the Conservative Party? 
 
 January 2011
 
October 2012
 
 Men % Women % Men % Women % 
I like David Cameron and I like the Conservative Party 27 27 26 24 
I like David Cameron but I do not like the Conservative Party 17 17 17 16 
I do not like David Cameron but I like the Conservative Party 11 10 
I do not like David Cameron and I do not like the Conservative Party 42 41 43 45 
Don't know 

Source: Ipsos MORI.

Base: 1162 GB adults aged 18+, 21–24 January 2011; 1005 GB adults aged 18+, 20–24 October 2012.

The press coverage

Why, given the evidence to the contrary in most of the polls, did the press think the Conservatives had a ‘problem with women’? The explanation seems to fall into two categories. In some cases, the reports were reasonable interpretations of the findings of one poll or poll series, but did not take account of contrary evidence from other sources (which other reporters might have used to question the story, but did not). In other cases, however, the evidence cited was not adequate to support the conclusion drawn on any reading.

In the first reports in the Sunday Times and the Spectator (McDonagh, 2011; Oakeshott and McGee, 2011), inspired by leaks about concern within the party, the polling evidence referred to was a fall in support since late 2010 in YouGov's polls: ‘45 per cent of women backed the Tories, according to YouGov, against just 34 per cent of men. Now that gap has vanished’ (McDonagh, 2011).

Expanding further on this, the leaked Conservative memo stated that ‘We know from a range of polls that women are significantly more negative about the Government than men’—but, tellingly, it went on to say that ‘We don't at present have a finer-grained analysis than this’, suggesting perhaps that the conclusion had been drawn from the figures in the published newspaper polls rather from the party's own private polls (Anon, 2011).

The YouGov trend, showing a fall in Conservative support among women between late 2010 and mid-2011, seems to have been the sole solid foundation to justify the early coverage. This movement can be seen in Table 1, though the size of the fall was exaggerated by taking the extreme highs and lows in individual polls rather than longer term averages: our table finds a 6-point fall rather than the 11-point plunge of the press reports. Even after this fall YouGov still had support among women a point higher than at the general election, and three points higher than among men. If this paints a true picture of the movement of public support, one might have thought that it was the disproportionate rise in women's support after the election in 2010, rather than the reversion to the norm in 2011, which most required explaining and merited most Tory attention. Moreover, the fall in women's support that YouGov found took place between October 2010 and March 2011, and had long re-stabilised by the Autumn, reasonable excuse perhaps for Conservative panic at the time of the local elections (in early May), less so for stories reporting a Conservative ‘problem’ as news in October.

But in any case, YouGov's findings at this period are out on their own. While they found that Conservative support among women had reached a 44% share in the last third of 2010, three of the other four companies put the same figure at 36%, and found little movement in the loyalties of either men or women by the end of 2010. The main evidence of a Conservative problem that was cited, therefore, was only that the party had not held on to an increase in female support which just one of the polling companies had found in the first place.

This coverage, at least, was clear in attributing its evidence to a particular company's polls, perhaps allowing the alert reader to surmise that other polls might find differently. More insidious is where contested polling evidence is reported as known fact. When differing polls are so treated it can create direct contradictions. According to the Independent, ‘As recently as the last election, women were more likely than men to vote Tory’ (Sieghart, 2011), but the New Statesman thought that ‘The Tories’ disproportionately low support among women prevented them from winning a majority at the last general election’ (New Statesman, 2011).

Finally, there were a number of reports which adduced further evidence to amplify the existing story, but did so without regard to basic rules of poll interpretation. Many made classic errors of poll reporting against which pollsters have often warned in the past (e.g. Worcester, 1991, pp. 184–188). One story (Moss, 2011) reported a statistically insignificant one-point fall in Conservative support among women as ‘Mr Cameron's female support slipping away’ (Moss, 2011); a three-point movement in the opposite direction the previous month had not been commented upon. Others confused voting intentions with approval ratings, comparing one with the other to report a trend that doubled the real scale of the movement4 (Roberts, 2011), or interpreted a single poll as indicating a change when no comparable earlier data existed (Mulholland, 2011). Many of the stories, too, personalised the coverage and subtly changed its meaning by headlining or reporting in terms of support for David Cameron when their data referred only to the parties.

Discussion and conclusions

During the course of 2011, several journalists seem to have received leaks from within the Conservative hierarchy indicating that the party was worried about its standing with women, and the internal memo obtained by the Guardian seems to confirm those leaks as true. The Conservatives remain divided on how to treat women's issues, and such leaks might have been intended to manipulate the political situation in several different ways, or simply to embarrass. Naturally the media reported this story, and the use that the Labour Party made of it to attack the government. But they also took the leaked diagnosis as fact, and our investigation of the figures from the published opinion polls—all of which were available to journalists at the time since all of the pollsters publish gender breakdowns of their findings on their websites, in line with British Polling Council rules——suggests that the premise was at the very least open to question. Further, the press certainly helped to characterise the situation as ‘Cameron's problem with women’, apparently personalising the causes of the supposed problem in a way that the polling data seem to contradict.

A little investigation would have revealed several pertinent facts. First, that although some findings such as the satisfaction scores showed greater discontent among women than men, this was less clearly reflected in voting intentions—while voting intentions are not the be-all and end-all of political opinion and a Conservative concern about the potential effect of being poorly rated by women on other measures would be perfectly reasonable, best reporting practice would clearly distinguish between the two.

Secondly, the evidence that we have considered contains nothing that points toward David Cameron having been in 2011 a particular liability for his party. Possibly there is other evidence in one-off polls that have not been drawn to our attention which would support this, but if so the coverage that linked headlines on a Conservative ‘problem’ with discussion of Cameron's character and behaviour failed to quote it.

Thirdly, as far as voting intentions were concerned, the three key interpretations that most readers would be likely to put on the idea of a Conservative ‘problem with women’ were all unsupported by the polls, properly read, up to the end of September 2011: Conservative support among women was not lower than their support among men, there had not been a big fall in women's support for the Tories since the election, and women's support had not declined relative to that of men since the election. There was not even significant disagreement between the pollsters on this.

Fourthly, there was one specific proposition that did find support, but only from some of the polls. This was the YouGov finding that Tory support among women had risen considerably between the election and the end of 2010, but had fallen back again by the middle of 2011. Because of the frequency of YouGov's polling, the combined sample sizes involved are entirely adequate to be sure this was not merely a freak of sampling variation: taken in isolation, the YouGov polls could be considered adequate evidence. However, they cannot be taken in isolation. Of the other four pollsters, three find an entirely contradictory pattern of little movement; the fourth, Populus, had intermediate results that are not firmly in either camp. Again, taken purely in isolation if YouGov had not polled, the other four companies together would amount to a sufficient weight of evidence to conclude that no such movement occurred.

This contradiction between the polls is a very clear case of a ‘house effect’. YouGov's findings were, for a period, systematically different from those of the other companies. It is almost inevitable that such differences will occasionally arise. Each company conducts its polls and makes its calculations in the way it judges best suited to making an accurate measurement, but their judgements on which is the best method do differ. But when the resulting findings differ as a consequence, clearly at least one is wrong.

Of course, it is not realistic to expect the newspapers to explore every nuance and contradiction of the polls and never to report a finding without consulting half a dozen competing sources. The Sunday Times quite reasonably turned to their own YouGov polls to complement their reporting of disquiet at Conservative headquarters. But other newspapers might easily have found that other pollsters disagreed and that the evidence was by no means so clear cut as the YouGov figures alone suggested.

They might also have questioned the political implications that were being read into the YouGov figures: the rise in support among women after the election was as dramatic as the subsequent fall away from that peak, and one might suppose that understanding such a swing would depend as much on knowing what women felt the Tories were doing right in 2010 as on what they were doing wrong in 2011. But none of the coverage considered this aspect of the situation, which would be crucial to its use in framing other news stories.

Journalists need to be careful in their reporting of opinion polls, because they are in a position to check the facts while most of their audience are not. It is not the fault of the polls that measurement of sub-group voting intentions cannot be done to any useful degree of precision within the compass of a single poll—it is something for which the polls are not designed, and a function for which they are unsuitable. Most of a newspaper's readers will be unaware of the technicalities of sampling, and may take figures at face value without allowing for a realistic margin of error. Nor may they read a story closely enough to realise that references to support for David Cameron refer in fact to support for his party. They need to be protected from such misunderstandings, and it is the responsible journalist's job to give them this protection through restraint in reporting. A misleading story about the public mood, used as a frame through which to interpret other political news, has considerable potential to distort public debate and perhaps to change public opinion itself. The same may occur when a story, justifiable in itself but relying on contestable evidence, achieves undue impact because it is not contested; explorations of alternative readings and interpretations by competing news media are healthy.

The overall impact of the coverage in the autumn of 2011 was to give an impression that public opinion had moved in a certain way. This may well not have been true, and was certainly not known with the degree of certainty that the coverage implied, yet—encapsulated in a much-repeated phrase—it was allowed to colour the interpretation of numerous other news stories. It is unclear whether this affected public opinion, but we certainly cannot safely assume that it did not. When the history of women's political attitudes during the 2010 Parliament comes to be written, consideration may need to be given to the impact of the media coverage of the polls as well as to more obvious factors such as policies and personalities.

References

Anon
2011
 
Leaked Memo on Support from Women for the Coalition Government’, accessed at http://www.guardian.co.uk/politics/interactive/2011/sep/13/leaked-memo-women-coalition-government on 29 January 2013
Campbell
R.
Gender and the Vote in Britain: Beyond the Gender Gap?
 , 
2006
Colchester
ECPR Press
Campbell
R.
Childs
S.
Lovenduski
J.
Women's Equality Guarantees and the Conservative Party
The Political Quarterly
 , 
2006
, vol. 
77
 (pg. 
18
-
27
)
Channel 4 News
2011
 
October 2 ‘Do the Conservatives Have a Problem with Women?’ accessed at http://www.youtube.com/watch?V=AMiK8iF25k&feature=youtube_gdata_player on 6 February 2013
Childs
S.
Webb
P.
Sex, Gender and the Conservative Party: From Iron Lady to Kitten Heels
 , 
2012
Basingstoke
Palgrave Macmillan
Chong
D.
Druckman
J. N.
A Theory of Framing and Opinion Formation in Competitive Elite Environments
Journal of Communication
 , 
2007
, vol. 
57
 (pg. 
99
-
118
)
Clarke
H.
Sanders
D.
Stewart
M.
Whiteley
P.
Political Choice in Britain
 , 
2004
Oxford
Oxford University Press
Cooper
Y.
2011
 
‘Typical That Cameron Thinks His Only Problem with Women Is Spin and Presentation. He is Out of Touch—Cooper’, accessed at http://www.labour.org.uk/cameron-is-out-of-touch,2011-10-02 on 29 January 2013
Curtis
P.
Stratton
A.
Revealed: Secret Government Plans to Win Back Women
Guardian
 , 
2011
 
September 13
Goffman
E.
Frame Analysis: An Essay on the Organization of Experience
 , 
1974
Cambridge, MA
Harvard University Press
Hayes
A.
Matthes
J.
Eveland
W.
Stimulating the Quasi-statistical Organ: Fear of Social Isolation Motivates the Quest for Knowledge of the Opinion Climate
Communication Research
 , 
2013
, vol. 
40
 (pg. 
439
-
462
)
Heffernan
R.
Webb
P.
Poguntke
T.
Webb
P.
The British Prime Minister: Much More Than “First Among Equals”
The Presidentialization of Politics: A Comparative Study of Modern Democracies
 , 
2005
Oxford
Oxford University Press
(pg. 
26
-
62
)
Iyengar
S.
Is Anyone Responsible? How Television Frames Political Issues
 , 
1991
Chicago
University of Chicago Press
Iyengar
S.
Kinder
D.
News That Matters: Television and American Opinion
 , 
1987
Chicago
University of Chicago Press
Iyengar
S.
Peters
M.
Kinder
D.
Experimental Demonstrations of the “Not-so-Minimal” Consequences of Television News Programs
American Political Science Review
 , 
1982
, vol. 
76
 (pg. 
848
-
858
)
Kellner
P.
Twyman
J.
Wells
A.
Wring
D.
Mortimore
R.
Atkinson
S.
Polling Voting Intentions
Political Communication in Britain: The Leader Debates, the Campaign and the Media in the 2010 General Election
 , 
2011
Basingstoke
Palgrave Macmillan
(pg. 
94
-
108
)
Kite
M.
2011
 
August 8 ‘Never Mind the Charm. Women Want Jobs, Dave’, Mail Online, accessed at http://www.dailymail.co.uk/debate/article-2023185/Never-mind-charm-David-Cameron-women-want-jobs.html on 20 February 2013
Knowles
D.
2011
 
November 15 ‘David Cameron's “Woman Problem” is Not Something a Tokenistic SpAd Appointment Will Solve’ [online], News—Telegraph Blogs, accessed at http://blogs.telegraph.co.uk/news/danielknowles/100117516/david-camerons-woman-problem-is-not-something-a-tokenistic-spad-appointment-will-solve/ on 6 February 2013
Marsh
C.
Back on the Bandwagon: The Effect of Opinion Polls on Public Opinion
British Journal of Political Science
 , 
1984
, vol. 
15
 (pg. 
51
-
74
)
Marsh
C.
O'Brien
J.
Opinion Bandwagons in Attitudes towards the Common Market
Journal of the Market Research Society
 , 
1989
, vol. 
31
 (pg. 
295
-
305
)
McAllister
I.
Studlar
D.
Bandwagon, Underdog, or Projection? Opinion Polls and Electoral Choice in Britain, 1979–1987
The Journal of Politics
 , 
1991
, vol. 
53
 (pg. 
720
-
741
)
McCombs
M.
Setting the Agenda: The News Media and Public Opinion
 , 
2004
Oxford
Blackwell Publishers
McCombs
M.
Shaw
D.
The Agenda-setting Function of Mass Media
Public Opinion Quarterly
 , 
1972
, vol. 
36
 (pg. 
176
-
187
)
McDonagh
M.
What Women Want
Spectator
 , 
2011
 
June 25
Moss
V.
David Cameron Drives Women into the Arms of Ed Miliband
Sunday Mirror
 , 
2011
 
October 16
Mughan
A.
Media and the Presidentialization of Parliamentary Elections
 , 
2000
Basingstoke
Palgrave
Mulholland
H.
Why Conservative Party Support from Women is Falling
Guardian
 , 
2011
 
October 5
Nelson
T.
Oxley
Z.
Clawson
R.
Toward a Psychology of Framing Effects
Political Behavior
 , 
1997
, vol. 
19
 (pg. 
221
-
246
)
New Statesman
Leader: David Cameron's Problem with Women
New Statesman
 , 
2011
 
October 6
Noelle-Neumann
E.
The Spiral of Silence: Public Opinion—Our Social Skin
 , 
1993
2nd edn
Chicago
University of Chicago Press
Norris
P.
Evans
G.
Norris
P.
Gender: A Gender-generation Gap?
Critical Elections
 , 
1999
London
Sage Publications
(pg. 
148
-
163
)
Oakeshott
I.
McGee
S.
Why Women Are Giving Cameron a Kicking
Sunday Times
 , 
2011
 
June 26
Phillips
M.
Calm Down, Dave. Your Real Problem's That Women Can Spot a Fake a Mile Away
Daily Mail
 , 
2011
 
October 3
Press Association
Fawcett Society Loses Court Challenge to Legality of Budget
Guardian
 , 
2010
 
December 6
Purnell
S.
Does Boris Johnson Have a Woman (Vote) Problem?
Daily Mail
 , 
2011
 
December 19
Rifkind
H.
Like the Conservative Party, I have a Problem with Women
Spectator
 , 
2011
 
October 8
Roberts
Y.
Cameron's Patronising Manner Stokes Women's Anger against The Coalition
Observer
 , 
2011
 
October 2
Sands
D.
2012
 
‘The Impact of Austerity on Women’ (Fawcett Society Policy Briefing, March 2012), accessed at http://www.fawcettsociety.org.uk/?Attachment_id=407 on 22 August 2013
Semetko
H.
Blumler
J.
Gurevitch
M.
Weaver
D.
The Formation of Campaign Agendas: A Comparative Analysis of Party and Media Roles in Recent American and British Elections
 , 
1991
Hillsdale
NJ, Lawrence Erlbaum
Sieghart
M. A.
Cameron's Problem with Women
Independent
 , 
2011
 
October 3
Smith
T.
In Search of House Effects: A Comparison of Responses to Various Questions by Different Survey Organizations
Public Opinion Quarterly
 , 
1978
, vol. 
42
 (pg. 
443
-
463
)
Stevens
D.
Banducci
S.
Karp
J.
Vowles
J.
Priming Time for Blair? Media Priming, Iraq, and Leadership Evaluations in Britain
Electoral Studies
 , 
2011
, vol. 
30
 (pg. 
546
-
560
)
Stratton
A.
David Cameron's Trouble with Women Makes Theresa May Close to Unsackable
Guardian
 , 
2011
 
November 11
Whiteley
P.
Crewe
I.
Harrop
M.
The Accuracy and Influence of the Polls in the 1983 General Election
Political Communications: The General Election Campaign of 1983
 , 
1986
Cambridge
Cambridge University Press
(pg. 
312
-
324
)
Worcester
R.
British Public Opinion: A Guide to the History and Methodology of Political Opinion Polling
 , 
1991
Oxford
Basil Blackwell
Worcester
R.
Mortimore
R.
Baines
P.
Gill
M.
Explaining Cameron's Coalition: How It Came About—An Analysis of the 2010 British General Election
 , 
2011
London
Biteback
2
YouGov's Sunday Times polls were weekly (YouGov also polled four times a week for the Sun: these polls have not been included in the analysis, but testing a random sampling of them suggests that doing so would make no significant difference to the findings.) ComRes generally conducted one poll monthly in each of its two series: treating their online and telephone polls separately would not affect the conclusions. The other companies polled once a month, except that there were no Populus/Times polls in November or December 2010, or in August 2011.
3
Intriguingly, the gender pattern of responses to the same questions when asked in this period about Ed Miliband, the Leader of the Opposition, was rather different. Dissatisfaction with Miliband's performance as leader was significantly lower among women than among men, and women were considerably more likely than men to say that they liked him (42% to 32%), but there were only small differences between the sexes on the detailed image attributes. This sort of comparison could have been the basis of a much more penetrating discussion of attitudes to David Cameron than anything that was actually published.
4
The article compared an earlier 45% Conservative voting intention figure with the current 25% who approved of ‘the government's record to date’, instead of with the 36% which was the comparable voting intention measure in the same poll.

Author notes

1
The polling data quoted throughout this article is given in the computer tables reporting the results of each individual poll which can be downloaded from the websites of the various pollsters (www.comres.co.uk, www.icmresearch.com, www.ipsos-mori.com, www.populus.co.uk and yougov.co.uk).