Before online news and social media, the national papers were the main focus of concern about the power without responsibility of the news media. In the heyday of the popular national press, more than thirty million people were reading a national daily paper, more on Sundays. With such a large market the papers could be party political and not risk their markets, and in the 1970s and onwards their partisan tone grew stronger and shriller.Footnote ¹ This was a period of attack journalism, especially on the part of the Conservative tabloids which ran harsh and permanent campaigns against Labour and its leaders. In spite of this, in 1979 more Labour voters read a Conservative paper than a Labour one and in 1983 the Conservative circulation was more than three times that of Labour.Footnote ² While this might have been taken as evidence that the Tory tabloids had little influence, there was general agreement that the national press had a great, sometimes decisive, influence over the political opinions and voting behaviour of the British people. Although partisanship became less caustic and circulations declined, the idea of press power without responsibility persists up to the present day and the size and partisan nature of the daily newspaper market makes the UK a good case for testing the power of the press.

This chapter examines the claim that the national press has a strong influence on how people vote on election day. It starts with some general points about newspaper partisanship and the public reaction to it. It continues with a close look at newspaper influences in the 1992 and 1997 elections, which are particularly good test cases of newspaper influence. The 1992 election was the one where the Sun claimed to have won it by declaring its support for the Conservative Party. In the 1997 election the Sun switched to the Labour Party which won by a landslide. Newspapers rarely change sides, and if papers have any influence over voting, this should be revealed clearly by a disproportionate swing of Sun readers to Labour in 1997. Consequently, the 1997election has been picked out for in-depth research as an acid test case of newspaper influence.

Since 1992 and 1997 are exceptional elections, the third section of the chapter examines the more normal sequence of elections after that to see if they yield different results. The fourth section of the chapter turns to the question of election agenda-setting over the years in the UK. It has been argued that the media may have little power over what we think, but may have influence what we think about – what we think are the most important issues in elections. If, as many argue, the media have an agenda-setting capacity we would have to conclude that the media have an independent power of their own that is far more significant than a minimal ability to reinforce existing opinions.

Newspapers and Voting

The idea that newspapers influence voting rests on the assumptions that citizens follow the lead of their paper, that their paper is the only or main influence on their voting decision, and that the paper gives clear signals about how to vote, not a mixed and rather jumbled set of cues. Previous chapters show that citizens quite often do not follow the lead of their newspaper or the media in general, and that newspapers are not the only media influence on voting decisions. We now turn to the third assumption: that national daily papers give consistent and unambiguous cues about how to vote.

Readers are not always able to identify the partisanship of their paper correctly. One survey finds that a large majority of broadsheet readers get it right, but small to large minorities of tabloid readers get it wrong, are unsure or do not know.Footnote ³ One survey shows that in the case of the Mail, Express and Sun only 37, 44 and 42 per cent respectively understand the party politics of their paper correctly, and that majorities got it wrong or did not know. Confusion of this kind is not limited to the readers of particular papers. When the population as a whole is asked about the partisan and political positions of the national press there is even more misunderstanding. More than two-thirds are unable to correctly place the Guardian, and half are unable to do the same for the Telegraph and Mail.Footnote ⁴ Similar confusion and misunderstanding is revealed in other surveys.Footnote ⁵ It is difficult to see how newspapers can influence political attitudes and voting patterns of readers who do not know which party their newspaper supports.

Confusion is understandable because although papers are often classified simply as Labour, Conservative or neither, they do not always follow a clear and unwavering party-political line. The Sun’s defection to Labour in 1997 is famous but the Express did the same in 2001. The strongest of partisan papers are not infrequently critical of the leaders and the policies of the party they normally support, pursuing policies and issues of their own. They are obliged to report hard news about important state matters and decisive political events and policies, including those that reflect badly on their preferred party, even if they try to spin them. Nor do all papers clearly endorse a given party in their election day editions. Some do, but sometimes weakly, sometimes strongly, sometimes with qualifications. Others preserve their neutrality, or suggest who not to vote for. Some have favoured tactical voting to keep a party out of power. Few papers have displayed consistent partisanship in the last eight elections.Footnote ⁶ The Mail and Telegraph have done so for the Conservatives, and the Mirror for Labour. The Sun and Express have switched from Conservative to Labour and back again. At various times, the Guardian, Times, Financial Times, Independent, Star and Record have offered different advice or none at all.

The editorial lines of newspapers are probably more consistent than their news content, but papers have their own interests, material or ideal, and pursue them independently. Opinion-piece and leader columnists generally stick to their party but are given to urging them to do things differently. Journalists and reporters commonly switch without friction between papers of a different colour, claiming that this does not matter for a professional.Footnote ⁷ The result is what one writer has called the ‘chronic instability of editorial opinion’.Footnote ⁸ Consequently, the national papers are rarely either blue or red through and through, but various shades and even mixtures of these colours appear quite regularly on different pages so that the overall effect is a muted set of hues, including a non-partisan white.

It may not matter much if a paper is strong and consistent in its party-political support if it is not trusted and, by and large, trust in the national press in Britain is lower than TV news (BBC, ITV, Channel 4) with the tabloids much lower than the broadsheets. In 2022 the Sun was at the bottom of the table with 12 per cent trust. The Mirror and the Mail were at 22 and 23 per cent respectively and even the BBC, at the top of fifteen news providers, was trusted by only 55 per cent.Footnote ⁹ Since we tend not to believe what we do not trust, this casts some doubt on whether distrusted newspapers have much influence.

This may help to explain why there is not always a close alignment between the politics of a paper and the politics of its readers. In 1997, 28 per cent of Conservative tabloid readers voted Labour as did 22 per cent of Conservative broadsheet readers.Footnote ¹⁰ On the other side of the coin 6 per cent of Labour tabloid readers voted Conservative, as did 22 per cent of Labour broadsheet readers. In fact, almost as many Conservative tabloid readers did not vote Conservative (41 per cent) as did (43 per cent). While half (51 per cent) of Conservative broadsheet readers voted Conservative, more than a third (36 per cent) did not. Other studies find a fairly loose association between the party allegiance of the national press, and the voting patterns of readers finding that minorities, sometimes quite large, identify with and vote for a party their paper does not support.Footnote ¹¹

Even if we were to find a closer correspondence between newspaper endorsements and reader voting we would still be left with the chicken-and-egg problem of cause and effect. If papers give their readers what they want to read, and readers select a paper that gives them what they want to read, how can we be sure that the paper is leading and the reader following? That is the tricky question that studies of newspaper effects on voting must answer.

Unravelling the Chicken-and-Egg Problem

The existence of large minorities, sometimes majorities, of newspaper readers who do not understand, care or take any notice of the politics of their newspaper leaves us with its own problem. On the one hand, it suggests that papers have a limited influence over at least some of their readers who go their own way, whatever their paper tells them. On the other, the existence of these readers demonstrates that many people do not self-select their newspaper for its politics and are, therefore, possibly open to political influence by the paper they read.

At the same time, the existence of large groups in the population who read a paper that is inconstant with their own political leanings and voting patterns offers a way of unravelling the chicken-and-egg problem. This involves using the effects of cross-pressures and reinforcement. It may be that people are more likely to follow their own voting inclinations if these are reinforced by the paper they read. Conversely, they may be less likely to follow their own inclinations if they are cross-pressured by their paper. If so, those who are inclined to vote Conservative or identify with the party are more likely to vote Conservative if they are reinforced by reading a Conservative paper. Conversely, if they read a Labour paper they are less likely to vote Conservative. For the same reason Labour voting would be higher among Labour supporters who are reinforced by their paper than those who are cross-pressured.

We also need a control group of people who do not read a paper regularly. Among them, those with a Conservative identification should have a lower Conservative vote than the reinforced, but a higher Conservative vote than the cross-pressured. The same patterns should appear among Labour identifiers who read a Labour paper or no paper at all. Thus, we have a set of six expectations for party voting according to party identification, cross-pressures and reinforcement among newspaper readers and non-readers. To make the test more watertight it must also control the characteristics of individual voters that are normally associated with both newspaper readership and voting. In this study the control variables are age, sex, education, income, trade union membership, social class and interest in politics. And since each general election has its own special characteristics, the two in 1992 and 1997 are compared. The 1992 election is the one the Sun claimed to have won, and 1997 is the one where it switched its endorsement from Conservative to Labour.

Newspaper effects are unlikely to occur in an election campaign lasting six weeks, and are more likely to build up as a result of a steady drum-beat of reporting and party support over a long period of time. The study takes account of this by using the long-term British Household Panel Study conducted by the University of Essex that has tracked the same sample of 5,500 households and some 8,000 individuals since the study began in 1991. This classifies papers according to their long-term party support, rather than their election day endorsement, and tracks newspaper readership in 1991 against voting in 1992, and newspaper reading in 1996 with voting in 1997. In this way predicted causes precede predicted effects, and it recognises that the causes may take time to have an effect.

The results are wholly consistent with the expectations. Conservatives who read Conservative papers, are more likely to vote Conservative than those who do not read any paper. Socialists who read Labour papers are more likely to vote Labour than those who read a Conservative paper. Socialists who read no paper are more likely to vote Labour than those who read a Conservative paper, but are less likely to vote this way than socialists who read a Labour paper. Unfortunately, there are too few Conservatives reading Labour papers in the panel study to draw conclusions about them, but the results confirm five of the six expectations.

This suggests that there is indeed a newspaper effect on voting, but the conclusion requires three important qualifications. First, the newspaper effect is small to statistically insignificant and heavily outweighed by age, trade union membership and party identification. These are the main determinants of party voting, not the paper read, which is of little significance. Second, reading a Labour paper is somewhat more important for increasing the Labour vote among Labour identifiers and those with no party ID. Third, though small in both years, the newspaper effect was even smaller in 1997 than in 1992. The reasons for this will become evident later in the chapter.

The conclusions that newspaper effects on party voting are small to insignificant is confirmed by most other British elections. Starting in 1958, when Conservatives papers outnumbered Labour by three to one, Harrop concludes that: ‘Although this exercise is not based directly on evidence from the 1983 campaign, the results do strongly suggest that the Labour Party is unable to blame Fleet Street for its debacle in June.’Footnote ¹² Later studies are able to use time series data to track voting behaviour and newspaper habits over the whole duration of a Parliament, and regression analysis in order to hold constant an assortment of other variables. This enables them to answer the question of whether citizens change their newspaper if they change their voting preference and whether they change their voting preference if they change their newspaper? One study finds that

A second study covers the period from 1992 to 1997. It finds ‘little evidence that newspapers had much impact on the aggregate outcome of election … At best we have found, in line with our previous research, that newspapers have but a limited influence on the voting behaviour of their reader.’Footnote ¹⁴ A third study states that the effect of the Sun’s switch ‘was trivial in the overall context of the forces shaping Labour’s election victory … Like the rest of the press, the Sun was following opinion rather than creating it. In 1997, for sure, it was not “The Sun Wot Won It”.’Footnote ¹⁵ A fourth reports finds that

An Electoral Commission report on the 2005 election also makes the same point that this book has laboured in previous chapters: almost nine out of ten citizens use TV as their main source of news, 50 per cent use the radio and 43 per cent use newspapers, and the influence of partisan papers must be set against that of non-partisan TV, radio and newspapers of record.Footnote ¹⁷ Studies of, or speculations about, newspaper influence that do not take full account of all the news used by the population may be misleading – indeed, are likely to be so, especially where the low trust rating of many newspapers, especially the tabloids, is concerned.

The Acid Test that Fails, Even as an Acid Test

Because the Sun changed its endorsement in a loudly proclaimed way in 1997, and because this is a relatively rare example of a national paper doing so, closer attention has been paid to the possible newspaper effect in that election than in any other. If newspapers influence voting, then this should show up clearly in a larger than average swing to Labour among Sun readers. However, the study of cross-pressures and reinforcements discussed earlier tried to uncover a Sun effect by testing the effects of Conservative and Labour papers separately from the Sun to see if it had a special influence. The figures show that, on the contrary, the swing to Labour among Sun readers was even smaller than that among the readers of other papers.Footnote ¹⁸ Another study finds that ‘the pattern of vote switching of The Sun or any-ex-Tory newspaper proved to be much like that of those who did not read a paper at all’.Footnote ¹⁹

What happened in the 1997 election is probably revealed most clearly in a set of simple percentage figures presented by an Ipsos MORI poll. These measure voting intentions according to the swings to the Conservatives and the swing to Labour among newspaper readers and a wide variety of demographic measures. Both swings are helpful because a large number of those who voted Conservative in 1992 did not vote, or did not vote Conservative at all, in 1997, so the overall swing against the Conservatives is larger than the swing to Labour. The results, presented in a brief form in Table 6.1, show three patterns of voting changes among newspaper readers and non-readers alike.

First, the direction of the swings in all cases is the same: against the Conservatives and too Labour. The figures for Sunday papers (not presented here for reasons of space) are, if anything, more uniform, with each category falling within a two to three percentage point range. Second, there is some variation in how much each category moved, but not much. The size of the swing is almost the same for no paper, all papers and other papers, and the swing against the Conservatives was the same for Sun readers as for all paper readers. Third, the two exceptions to the general pattern are the slightly larger swings against the Conservatives and to Labour among consistently Conservative paper readers, but these do not suggest newspaper effects. Why would the readers of the loyal Tory papers register the largest swing against the Conservatives and the second largest swing to Labour? Why would the consistently Labour paper register the lowest swing against the Conservatives and a no more than average swing to Labour? If there is a newspaper effect why are the swings among those who do not read a paper identical to those who do?

Part of this might be explained by the difference between net and aggregate swings. If there were large swings in different directions according to party endorsements they might cancel each out, leaving a false impression of no newspaper effect. But this cannot be the case in 1997 because all newspapers register swings in the same direction, so they cannot cancel each other out. The only exception suggesting a newspaper effect in Table 6.1 is the Sun’s larger than average swing against the Conservatives and to Labour. This at least suggests a Sun effect, so perhaps after all there is a small newspaper effect, but for the Sun alone. We return to this possibility a little later.

The suggestion, frequently advanced in previous chapters, has been that the underlying causes of both newspaper reading and voting are the individual characteristics of citizens, especially their age, sex, education, income, social status, ethnicity and religion, plus their political identity. Fortunately, the Ipsos MORI poll that produced the figures in Table 6.1 also presents figures for swings broken down by fifty-eight demographic categories covering sex, age, social class, region, phone ownership, work status, housing tenure and trade union membership.

If demographic characteristics are the key, then substantial variations between different groups in the electorate would be expected. The Ipsos MORI poll finds the complete opposite. Every one of its fifty-eight categories is within a percentage point or so of the national average. Not only does every single category swing against the Conservatives and to Labour but every swing is within a few percentage points of all the others. Table 6.2 presents a few of the figures as examples.

Irrespective of demographic features and irrespective of the paper read, every group in the population acted in much the same way in the 1997 election. Neither the press nor the usual social and economic characteristics of voters made much difference to the way that many deserted the Conservatives and rather fewer opted for Labour. In the 1997 election the great majority of the electorate behaved as one.

This presents us with a new puzzle. Party voting in the 1997 election did not vary according to either the newspaper read or demographic groups. What explains this highly unusual state of affairs that resulted in a Labour landslide? The answer is simple and obvious. Elections are political events and, curious though it may seem, the result was the outcome of political circumstances.Footnote ²⁰ The polls had shown a clear Labour lead ever since Black Wednesday, 16 November 1992, only a few weeks after a new Conservative government had taken office after the 1992 election. Black Wednesday was the day the government was forced by economic pressure to withdraw the pound sterling from the European Exchange Rate Mechanism. It cost the country an estimated £3.4 billion and its cost the Conservative government its priceless reputation for economic competence. Its opinion poll ratings dropped almost overnight and remained low for the next four years. It was unprecedented for a party and its leader (Blair) to sustain such a high level of approval for such a long and unbroken time. Many voters formed their voting decision in 1992 and stuck to it to the last. As one journalist put it, ‘this was never going to be the kind of election in which the press could sway’.Footnote ²¹

On top of that the government added to its unpopularity with scandals, internal rancour, indecision, incompetence and an uncharacteristically poor campaign that revealed deep splits over the EU. Labour managed an unusually slick campaign led by the young and charismatic Blair and a group of wily campaign managers. In retrospect and with the advantage of 20:20 vision the 1997 election was a foregone conclusions because a great number of voters had made up their mind four years earlier.

This does not explain the slightly larger than average swings, shown in Table 6.1, to Labour among readers of the consistently Conservative press (16.7 per cent) plus a larger than average swing to Labour (12.7 per cent). In contrast, the Mirror, a faithful Labour paper had a smaller than average swing against the government, and an average swing to Labour. In Table 6.2 the ABC1 social status group had the largest swing against the Conservatives and the second largest swing to Labour. Does this show that true-blue papers are good at manufacturing Labour voters, or that the upper social strata are ripe for conversion to the socialist cause? Of course not.

The answer to these, and other anomalies in the figures, is simple. Tory papers have a larger than average proportion of Conservative readers and, therefore, a larger pool of voters who could turn against the party in the right circumstances. With few Conservative voters among Mirror readers, there is less room for change and so the paper records a slightly smaller swing against the Conservatives and only an average one to Labour. The social class of ABC1s are similarly Conservative in most elections, with a large pool of potential switchers who could do what the rest of the country was doing by deserting the Conservatives and voting Labour. But because it is a big leap from not voting Conservative to voting Labour, some of the Tory faithful could not bring themselves to vote at all, or voted for the half-way house Liberal Democrats. The variations from average swing figures in the tables reveal not newspaper or social and economic effects but a simple ceiling effect that sets limits on how many were available to swing.

It is possible that the few election studies that do find significant, even very large, newspaper effects in 1997 are actually measuring not a newspaper effect but a ceiling effect. But it takes exceptional political circumstances to produce a collective mood that produces the same swing to Labour and from the Conservatives. That is why the 1997 election turns out to be a very bad case to estimate a newspaper effect, least of all a natural experiment that provides us with an acid test of such effects.

The 1992 election is a better test, but close analysis of that shows, as already demonstrated, that it was not the Sun, or any other paper, that won it.

Newspaper Effects in Normal Times

If 1997 is a particularly bad year for a test of newspapers effects what about more normal times? Fortunately, there is plentiful evidence about them because the Sun’s claim to have won it in 1992, followed by the landslide victory for Labour in 1997, promoted new interest in newspaper influence. Starting with 2000 (not an election year), an Ipsos article investigates the question: ‘Do readers believe what the editors want them to?’ It concludes

We can judge this from swing figures, obligingly provided by Ipsos MORI again, for the election to elections of 1997–2001, 2001–2005 and 2005–2010 compared with the figures for 1992–1997 (see Table 6.3).

The first column of the table repeats what we already know about the 1997 election: the swing to Labour is much the same for all papers, irrespective of their partisanship, and most figures cluster around the non-reader and total adult averages, give or take a few percentage points. The same is true of the uniformly small swing against Labour in 2001–2005 and 2005–2010, when almost all papers saw the same fairly small leakage of Labour support over time and away from the landslide victory of 1997. Once again, most newspaper readers move in the same direction and to the same degree, whatever party their paper supports.

The largest figure in the 2005–2010 column is the Sun’s swing of 13.5 per cent to the Conservatives, which happens to coincide with its return to the Conservative cause after the previous three election with Labour. Does this indicate a strong Sun effect, after all? According to MORI polls, it does not.Footnote ²³ Their figures show that Sun reader support for Labour fell from 45 per cent (actual vote) in the election of 2005 to 29 per cent (intended vote) in September 2009. But this happened before the paper officially declared its support for David Cameron and the Tories on 30 September 2009. Once again, the paper seems to have followed its readers rather than leading them. From September 2009 the paper hammered Labour and presented the Conservatives in the best light on its front pages. In spite of this, reader support for the Tories increased by 3 per cent – two percentage points lower than the national average.

This leaves the 1997–2001 column in Table 6.3 where there is a mixture of plus and minus signs. But it is difficult to see how newspaper partisanship accounts for this variation. The Telegraph swings 6 per cent to the Conservatives but the Guardian also swings this way by 6.5 per cent. The Sun, still on Labour’s side, had a smaller swing to the party than those who did not read a paper. The Express, converted to the Labour cause after unbroken support for the Conservatives in fifteen consecutive elections from 1945 to 1997, registered the same swing against Labour as the country as a whole. Like the Sun in the previous election, it seems to have no measurable effect on its readers’ voting choice.

What the figures in Table 6.3 show is not a newspaper effect but, in spite of the best efforts of the partisan press, a slow drift back to more normal voting after the Labour landslide of 1997. This is typical of any government that sooner or later runs into problems and loses support, and of oppositions that rejuvenate themselves with new leaders, policies and public images. After such a large swing to Labour in 1997 it was likely that voters would gradually return to their normal voting patterns as time passed, especially the Tories who had voted Labour in 1997. In the case of the Telegraph, Express and Mail, the size of the swing back to the Conservatives was larger than average because the size of the swing to Labour in 1997 was larger than average.

By 2017 the huge swing of ten years earlier had been erased but commentary on the general election of that year was still referring to the Sun’s claim of twenty-five years earlier. In that election Temple points out, with some disbelief, that ‘[o]f course, the use of the phrase “it’s the Sun wot won it” and the seriousness with which academics treated the claim, implies extraordinary power to our newspapers and their hold over their diminishing readership’.Footnote ²⁴ Once again, however, the verdict on that election was that ‘[t]his was a campaign, in other words, that showed both the determination of powerful gatekeepers in the mainstream media to foster a pro-Tory agenda and the enduring ability of ordinary voters to ignore these voices’.Footnote ²⁵ Is the continuing belief in newspaper effects yet another example of belief preservation, or more simply an unawareness of the weight of research to the contrary?

Reinforcement

It has been argued that the newspaper effect is not a strong one but a weaker one of reinforcement in which the partisan press can simply shore-up and stiffen the pre-existing opinions of their leaders. Tables 6.1 and 6.3 do not present strong evidence of reinforcement. On the contrary, the consistently Conservative papers show an average or greater swing to Labour in 1997 followed by fairly average swings back to the Conservatives in the following three elections. It might be argued that the swing to Labour would have been even larger in the case of the Tory press had it not reinforced the views of some of their readers, but this introduces the complications of a counterfactual and some rather unconvincing arguments about why larger than average numbers of voters defected from their paper’s endorsement.

However, the reinforcement effect has not been much researched. It has usually been an unresearched fallback position to salvage at least some media effects when nothing stronger has been found. It may be that reinforcement effects are as elusive as direct effects on voting, but it is also possible that reinforcement has an effect on the strength of party-political opinions and identity of the party faithful, rather than on those who contemplate changing their voting choice. We do not know, but lacking this knowledge we are left with little evidence of newspaper reinforcement effects.

Agenda-setting

Agenda-setting theory states that the media cannot tell us what to think but has a powerful influence on what we think about. In modern times this theory originates in the empirical work of Maxwell McCombs, who summarises forty-five years of research by stating that the media are like teachers who repeat their lessons over and over again so that when ‘citizen students’ are asked what are the most important issues facing the country they repeat the lessons they have learned from the media.Footnote ²⁶ As a result, he writes, ‘the news media set the public agenda’.Footnote ²⁷

The evidence for this is variable and inconclusive. One study of sixteen countries using a big data approach finds that agenda-setting varies from one country to another and is contingent on the economic, political, social and media contexts of the countries and of individuals in the population.Footnote ²⁸ This seems to suggest that agenda-setting is dependent on a total context of almost everything. Nevertheless, we are fortunate to have evidence about agenda-setting in the context of the British general elections of 1987, 1997, 2001, 2005, 2015 and 2019.

The first systematic attempt to investigate the matter in the UK was Miller’s panel study of television and elections in the 1987 general election.Footnote ²⁹ He selected TV because it was the most popular source of election news but found little correspondence between the main items of its news and what electors were mainly concerned about. TV news focused on defence, terrorism and crime but voters were mainly concerned with the daily bread-and-butter issues of unemployment, health, education and social services, which had relatively little news coverage. When TV gave defence more air time in the third week of the campaign, viewers did not follow. Miller concludes that ‘Overall, therefore, the agenda set by television was miles away from the agenda of issues that the electorate rated important and wanted discussed.’

In the light of the Sun’s switch to Labour in the next election, there was more interest in agenda-setting, but a MORI poll found that the media and the public agendas went their own ways with little overlap.Footnote ³⁰ Another study at that time found that the priorities of most electors were fairly constant from one election to the other and there was little indication that any shift in these depended on media exposure.Footnote ³¹

A fourth study of the 1997 election conducted a more intensive two-pronged research strategy.Footnote ³² A carefully planned and analysed experiment involved 474 participants, divided into a control group and six experimental groups, who were shown a set of videos dealing with tax, employment, health, pensions, Europe and overseas aid to see what effects these had on the participants’ election priorities. The second strategy used an analysis of 6,072 news articles in national daily and Sunday papers published during the election campaign period, tracing the main items of focus and how they changed in the six-week period. These were compared with the agendas of a large national sample of voters and the ways in which those agendas changed during the campaign period.

The results of both exercises were the same: the public followed its own agenda. The priority given by experimental subjects to tax, health, employment and pensions were unaffected by the news videos. In the national sample of voters the priority given to health, education, employment, law and order and taxation changed by barely a percentage point or two during the campaign. This despite extensive coverage in the media of membership of the European Union, splits in the Conservative Party and political corruption (sleaze), all of which remained the concerns of small minorities in the population. The single exception was a greater priority attached to foreign affairs in the experimental study as a result of the video shown.

The authors put the media failure to set the agenda down to issue salience. It seems that voters are not much influenced by the media agenda where their own personal experience and knowledge is involved, and where the issue is an old and much discussed one that is also attached to party identification: tax, employment, health, education, pensions or social services, for example. These are salient issues of immediate concern. New issues that have not yet impacted on the public consciousness, and abstract and remote technical matters and events in distant parts of the globe, are less salient and may be more amenable to media influence and agenda-setting. Their appearance on the public agenda may also be more fleeting, precisely because they are not matters relating to daily, household existence.

An account of agenda-setting in the 2001 election compares news article in six main national papers, a similar content analysis of the press releases of the Conservative, Labour and Liberal Democratic parties and data about the public agenda conducted by two public opinion polls at the beginning and end of the campaign.Footnote ³³ All three were disconnected from each other. The media did not follow the agenda that the parties tried to set in their daily press conferences, and the public did not follow either the press or party agendas. The end result was that that the short-term agenda-setting impact of the press on the public appears to have been limited, as was the impact of the parties and public opinion on press coverage. The qualification ‘in the short term’ is important because McCombs specifies ‘the recent past’ to be the period during which the medias agenda-setting capacity operates.Footnote ³⁴

Norris’s study of agenda-setting in the 2005 election goes beyond the media as possible influences on the public’s election agenda to include people-intensive channels (local party contact activity) and new technology channels (internet campaign information). But this widening of channels of political communication produces no evidence of media agenda-setting effects:

The public agenda, it seems, is generally immune to outside influences whether coming from the national mass media, local party activities or internet content.

For the 2015 election Moore and Ramsay collected a large body of data about the content of sixteen mainstream mass media news sources, the communications of the two main parties and also, for the first time in British agenda-setting studies, social media content. The study concludes that

Table 6.4 compares media and popular agendas in the final period of the 2019 general election. The figures in the two columns are not exactly comparable, coming from different sources with different categories. The first column is the content analysis of TV and the press conducted by the University of Loughborough in the last week of the election campaign and the second is the YouGov tracker poll of the popular agenda in that week. There is some overlap of media content with the public agenda if we assume that business/economy/trade is the same as the economy, that health and health care is the same as health and Brexit is the same as Brexit/EU, but even here the percentage figures attached to these items in the two columns do not correspond closely. The other agenda items in the two columns come from different planets, signifying a wide gap between the media and the public agenda.

Discussion

Perhaps it is not surprising that the best research using the best methods and data find little evidence of a newspaper effects on voting decisions. Few papers are always completely loyal to their party all the time and most have delivered mixed and qualified messages. Quite a few tabloid readers mistake or do not know which party their paper supports. And, in any case, they are not generally trusted. If you read the Sun because it’s cheap and cheerful, because it is easy to read and a bit of a laugh, you are not likely to take its voting advice very seriously.Footnote ³⁷ Besides, most people, including tabloid readers, get most of their news from public service TV and BBC radio, so the fact that most daily papers have a conservative and/or Conservative bias and favoured Brexit may not count for much.Footnote ³⁸

On top of this, a vast amount of experimental psychology has uncovered powerful psychological mechanisms that individuals use to preserve their beliefs and opinions. This literature shows that pre-existing beliefs usually, although not always, triumph in the face of opposing views. In fact, opposing views can have a boomerang effect of embedding beliefs more firmly. And lastly, the simple idea that people believe what they read in the papers and that it was ‘the Sun wot won it’ underestimates the serious difficulties encountered in trying to pin down media effects when readers choose their paper and papers are pushed by market forces to provide their reader with what they want to read. It has been argued, though we do not know for sure, that the Sun has switched its endorsement in order to be on the side of most of its readers.

A few studies do find a statistically significant impact in 1997. How can this be explained? One possibility is that research covering a short period before and during the official election campaign miss the fact that the great majority of the population changed their voting intention after Black Wednesday, four years before polling day. Another possibility is the larger pool of voters reading Conservative papers who were available to swing against the Conservatives and to Labour, compared with the much smaller pool who read Labour papers. This would explain why readers of the faithful Conservative press (the Mail, Telegraph and Express) swung more to Labour than readers of Labour papers and the electorate as a whole.

When the figures are disaggregated there is also little evidence to support the idea of reinforcement. The Mail, Telegraph and Express did not prevent a larger than average desertion from the Conservative cause in 1997 and an even larger swing in that direction compared with those who did not read a paper. Once again this is likely to reflect the fact that these three papers had a large proportion of readers who could swing this way, compared with the Mirror with its larger proportion of Labour readers. None of this will become evident without disaggregating by newspaper, classifying them correctly according to their partisan or neutral political stance and conducting a comparison with non-paper readers.

A clearer example of the limits of press power is provided in the 2016 American presidential election. In that year more than three-quarters (76.6) of the 653 newspapers and magazines that reported the news endorsed Hilary Clinton, 12.6 per cent endorsed no candidate, 3 per cent were for Trump and 4.9 per cent recommended not voting for Trump. Of the 100 papers with the largest circulations, over half (57 per cent), with a total circulation of 13.1 million, were for Clinton, and two with a circulation of 36,000 supported Trump.Footnote ³⁹ In 2020 Biden was endorsed by 45 per cent of the top 105 papers, including of eight of the biggest ten, Trump by 7 per cent, and yet Trump won 46.9 per cent of the vote.Footnote ⁴⁰ Newspaper endorsements count for little in the national elections of both the USA and the UK.

Agenda-setting theory fares no better when tested against British national elections. In none of the elections of 1987, 1997, 2001, 2005, 2015 and 2019 did the public agenda follow those of the media. Perhaps this should not surprise us either. Because they live in close proximity in the Westminster village, journalists and politicians develop their own particular interests and short-term obsessions with the faults of leading politicians, political scandals, the electoral horse race, most recent polls and issues that flare up and die.Footnote ⁴¹ The electorate may sometimes be interested in reading about these things but there is little evidence that they affect voting.

Why, when agenda-setting has been confirmed often in the USA, is there so little trace of it in the UK? We do not know and can only speculate. One possibility is the importance of public service news in the UK and its small role in the USA. The more the electorate uses and trusts the balanced, impartial and accurate news of the public services, the less influential the press is likely to be. The possible irony here is that the more impartial and accurate a news source, the more it is trusted but the less it will exercise a partisan influence. The more partisan a news source, the less it is trusted and less it can exercise a partisan influence.

A further point emerges from these conclusions about voting and agenda-setting in British elections. There is now a disconnect between the political science of voting, political attitudes and behaviour, on the one hand, and media impact studies, on the other. Few election studies now use media variables in their explanations of voting, and studies of attitudes and behaviour almost never do. For example, the broad-ranging British Election Study of 2001 has little to say about the media, having found them to be of little importance in the previous elections.Footnote ⁴² Similarly, the twenty-seven chapters of The Routledge Handbook of Elections, Voting Behavior and Public Opinion barely touches on the media, and the one essay that does states

And this in spite of the fact that the study of political attitudes and behaviour is a data-rich and widely investigated field of research.

There is also a disconnect between media studies finding large or massive media effects and the experimental psychology of belief preservation. Very little of the media effects literature refers to this vast body of literature stretching back over six decades, in spite of the fact that the literature is virtually unanimous about the great difficulties of trying to change opinions.

All this raises the question, if the news media do not determine voting behaviour or election agendas, what does? The answer from the research is clear and consistent. In the election of 2019, for example, the variables most closely associated with party voting were age, sex (for the 18–24 age group) and education.Footnote ⁴⁴ In the EU referendum voting ran most clearly along age and education lines and, to a lesser extent, employment status and income.Footnote ⁴⁵ These are among the usual social, political and economic variables that form the standard model that emerges from most election studies and from the large volume of research on other forms of political attitudes and behaviour. They are also the underlying variables that explain how much news individuals get, from what sources and how they react to that news. The media are minor players; the driving force is provided by the demographic variables of the standard model.

This may be true but it still does not explain what determines the attitudes and beliefs that underlie both the choice of news media and their supposed effects. The answer that it is the variables of the standard model are part of the answer but not the whole answer. Age in and of itself, for example, cannot explain why older people prefer to get their news from TV and papers. There is nothing about age, as such, that leads people to make these choices. Nor can age explain why older people were much more likely to vote for Brexit or for the Conservative party in 2017. But the life circumstances typical of older age groups can explain a lot. They tend to have ingrained habits, are more likely to have developed lasting political attitudes, have a rather different set of life experiences than other cohorts, are less computer savvy and many have more time to watch TV and read newspapers. They live on pensions, not wages or salaries, and they are more likely to have paid off their mortgage and to have health issues, all of which might affect their voting behaviour.

Age, in other words, is important not in itself but because it stands for and is a shorthand way of referring to a range of other things that do have an impact on voting behaviour. Similarly, income in and of itself cannot explain attitudes and behaviour but the life circumstances of the rich and the poor can. When income is used as an explanatory variable it is assumed that it is not money pure and simple that counts, but all the things that go with it or the lack of it. There are many middle-class socialists and working-class Tories. Money, class, sex and minority status are indicators of (proxies for) a whole set of social, political and economic features that are often tied to them. What the variables of the standard model stand for, and how these condition media effects, is revisited in Chapter 8.