Media definitely do matter: Brexit, immigration, climate change and beyond

The notion that the media’s principal role regarding public opinion is reinforcement of pre-existing attitudes – and that this idea is relative inconsequential politically – is pervasive, across many political and social science sub-disciplines, and in non-academic commentary. This article comprehensively challenges the evidential and theoretical underpinnings of this thesis, drawing on a wealth of contemporary survey data and media coverage research, across a range of issues, including climate change, Brexit, immigration, the economy and benefit fraud. It also argues that ‘reinforcement’ is an important and consequential power, and that the processes involved have significant implications for public misperception of salient political issues. It makes the case that the media create attitudinal uncertainty, and can have pervasive but subtle influences on political attitudes, particularly when there are persistent patterns of coverage across a range of media. But also that in contexts like particularly close political contests, such influences can be decidedly consequential.


Introduction
Public debate following the Scottish Independence and Brexit referenda, and Trump's presidential victory, was coloured by concern about how the media refract politicsinvoking consideration of biased coverage or misleading, even 'fake', news. Such concern would trouble us more, if the coverage were thought to have a significant influence on citizens. But there is a strong strand in academic and non-academic commentary suggesting such influence -where it can be detected at all -is small and, therefore, relatively inconsequential politically Iyengar, 2008, 2010), when set against much broader societal, economic and political factors. From a 'minimalist influence thesis' (MIT) perspective, which goes back many decades (Klapper, 1968), the media does not change the opinions of the public, but 'only' (Gitlin, 1978: 216) or 'simply' (Miller and Krosnick, 1996: 79) reinforces pre-existing ones. The main intellectual tenets of this perspective hold that, primarily, citizens actively and selectively engage with media messages (Newton, 2006). And from a variety of sources of political content produced by a free media market (Anderson, 2007;Kuhn, 2007), people choose to expose themselves to those whose output they find congenial, while actively reinterpreting political messages failing to match their predispositions.
This article outlines just how widespread this view still is, across a range of political science domains, and seeks to assess, in the light of a developing research literature and an accumulating body of counter-evidence, whether such a perspective still holds good, or indeed ever did. The analysis argues that the MIT is unduly sanguine about the media's lack of political impact. It illuminates some of its highly significant and underexplored theoretical and empirical weaknesses. The assessment theoretically reinterprets an emerging body of media impact research, exploring its relevance to a range of politically salient topics, including Brexit and Europe, immigration, benefit fraud and climate change. It argues that, cumulatively, this literature and some new evidence suggests the media can have complex, multi-layered, longer term (and definitely not insubstantial) impacts on citizens, not least with regard to the creation of significant, politically important, misperceptions among electorates.
In what follows there is a wide-ranging theoretical reassessment of existing literatures. There have been a range of theoretical and broadly conceptual assessments of media impact Iyengar, 2008, 2010;Gans, 1993;Holbert et al., 2010;Newton, 2006;Street, 2005;Wolfsfeld, 2011), complementing useful meta-analyses of ongoing research in this field (Banducci, 2018;Chong and Druckman, 2012;Goldstein and Ridout, 2004;Neuman and Guggenheim, 2011). However, all of these explorations suffer from a range of highly significant, sometimes overlapping, theoretical and conceptual problems and omissions, which the following analysis addresses. For instance, theoretical assessments suggesting insignificant or politically inconsequential media impact Iyengar, 2008, 2010;Gans, 1993;Newton, 2006) rarely fully explore the serious empirical weaknesses of the core assumptions underpinning the MIT. And the same can be said of meta-analyses when they explore research suggesting limited, or non-existent, media effects.
The assessment that follows explores these signal MIT flaws, but also other weaknesses that both meta-analyses and theoretical explorations share. For instance, neither touch on the significance of the way the public are said to be 'dependent' on the mass media for information, in areas beyond the immediate lived experiences of citizens. We argue that the range of areas where the public are media-dependent is wider -and more politically significant -than the MIT, or indeed recent meta-analyses and theoretical explorations, acknowledge. But we will also argue that both bodies of work too narrowly focus on attitudes, attitude change or attitudinal stasis. In doing so, we argue, they faillike the MIT more generally -to explore attitudinal uncertainty, and the media's role in fomenting it. We explore the associated processes, and the significant political consequences that follow, in regard to a range of prominent issues. But we also look beyond a narrow focus on citizens' attitudes to consider politicians' perception of media impact, and how such perceptions might influence their policy-related behaviours in ways that are politically consequential. Here we theorise 'media impacts' beyond citizen attitude change or opinion reinforcement, in a way previously neglected in the theoretical and meta-analysis literature.
In addition, we undertake a conceptual reconsideration of the reinforcement processes at the heart of the MIT. Neither meta-analyses nor theoretical assessments tend to consider that the ability to reinforce pre-existing attitudes is a significant power in its own right. We explore this ability in the light of political theory around the manner in which power is thought to be exerted, and extend the analysis to encompass the well-evidenced 'agenda setting' impact of the media. Some MIT assessments fail completely to engage with agenda setting (Newton, 2006). But while many meta-analyses (Banducci, 2018;Chong and Druckman, 2012) and some theoretical explorations (Wolfsfeld, 2011) focus heavily upon this power, all wholly fail to theorise the role it plays in the way media power is exerted. Our article seeks to address this significant conceptual gap.
Finally, but importantly, of the meta-analyses and theoretical explorations cited above, only Goldstein and Ridout (2004), Wolfsfeld (2011) and Banducci (2018) focus on the media's role in accumulation of knowledge by citizens. But the first two reports do not consider the role of reinforcement processes in the development of the public's misperceptions of important political issues (something also missing from more sanguine and positive assessments of the media's role in knowledge acquisition, see Norris, 2000: 208-228). Banducci (2018: 307) separates conceptually the processes of reinforcement and knowledge acquisition in a way that impedes a full understanding of the politically consequential impact of the former on the latter. But in the analysis that follows, the processes involved are re-explored theoretically in respect to public misperceptions around the issues of immigration and benefit fraud.
The theoretical gaps, problems and omission highlighted here speak, we argue, to the cumulative deficiencies of the MIT. What follows is an integrated, conceptual reassessment, grounded in an appraisal of existing and emerging empirical research, designed to illuminate the weaknesses and narrow remit of the MIT, while suggesting ways to broaden our understanding of what constitutes a politically consequential 'media impact'. Here the focus is often on British politics and the UK media. However, the assessment is based strongly on an international research literature, and the theoretical and substantive themes are relevant in a much broader range of political contexts.

The MIT, its advocates and flaws
In Brexit's aftermath, on Channel 4 News, 30 June 2016, leading journalist Peter Hitchens, when asked about the media's role said, The old lesson on the media -and this has been true since probably the 1930s when the big press barons really first got engaged -is that they can fan and amplify a genuine shift in public opinion, but they can't create one. This view is very similar to one expressed in Parliament by former cabinet minister Peter Lilley: 'Readers who are interested in politics choose their newspaper because it has congenial political views; the rest are largely uninfluenced by an editor's views'. 1 Likewise, discussing UK anti-European sentiment, senior journalist Peter Riddell said the press has 'not been a creating factor; it has been a generally reinforcing factor over a considerable period of time'. 2 Comparable sentiments were expressed about political advertising by PM David Cameron's campaign manager, Lynton Crosby: it is 'of itself, not a very persuasive medium or a mind-changing tool. Its purpose is really to reinforce and trigger existing perceptions'. 3 Finally from Rush Limbaugh on Fox News, 19 February 2017 we got the assertion that 'The media did not make Donald Trump, and they can't destroy him', while Facebook's Mark Zuckerberg stated 'The idea that fake news on Facebook influenced the election in any way, I think, is a pretty crazy idea. You know, voters make decisions based on their lived experience' (BBC, Panorama, 8 May 2017).
These statements are unsurprising, as scholars old and new have taken a comparable line, underlining 'reinforcement' and limited media impact. For instance, Gitlin (1978: 216, quoting Klapper, 1968 suggests 'The media are taken only to "reinforce existing opinions" rather than to change minds'. And while reviewing seminal media impact research Miller and Krosnick (1996: 79) note it 'concluded that the media do not in fact persuade individuals to change their attitudes but rather simply reinforces attitudes already in place'. More recently Newton (2006: 214) submits that 'The minimal effects tradition … stresses the great importance of individual and structural factors in restricting the mass media to a reinforcing role', while Bennett and Iyengar (2008: 724) affirm that 'As media audiences devolve into smaller, like-minded subsets of the electorate, it becomes less likely that media messages will do anything other than reinforce prior predispositions'.
These general assertions extend to researchers focusing on specific topics and subdisciplinary areas, and focusing on new, as well as 'legacy', media. For instance, the reinforcement of pre-existing political attitudes is said to be evident on the web (Himelboim et al., 2013;Stroud, 2008). Anderson and Weymouth (1999: 169), while exposing the persistently critical, and often distorted, misleading and occasionally xenophobic UK press coverage of Europe, still acknowledge -if somewhat grudginglyselective interpretation of such messages by audiences.
Elsewhere, Hulme (2009: 226) suggests the response of citizens to climate change coverage is not passive and sponge-like, but heavily moderated by pre-existing normative, cultural and political attitudes (see also Smith, 2005Smith, : 1474. Finally, in the partypolitical domain Clarke (2012: 171) notes 'the main newspaper effect is to reinforce reader's opinions, rather than change them', while Schmitt-Beck and Farrell (2002: 189) assert 'the more clear their [voters'] political predispositions are, the less likely campaigns are to affect opinion'. Understandably, Brandenburg and Van Egmond (2012: 441) state that 'The pervasiveness of the media is now accepted, but its persuasiveness is often dismissed'.
The processes operating are well-understood. Numerous authors (Anderson, 2007;Kuhn, 2007;Mullainathan and Shleifer, 2005) note that free, competitive media markets provide political message diversity. Resultant plurality 'provides consumer with what they want to see, hear and read' (Newton, 2006: 214), facilitating active selection of congenial political messages by citizens: people choose newspapers aligning with their preexisting partisan leanings and underlying values, thereby selectively exposing themselves to unthreatening political messages. And they selectively interpret those they do see (Klapper, 1968), in activity sometimes termed 'confirmation bias' or 'directed reasoning'. The partisan press, in turn, serves the resultant demand keenly, following their readers as an editorial and commercial imperative (Mullainathan and Shleifer, 2005;Ponsford, 2016). The picture painted explains the voting-behaviour-newspaper-readership alignment, not as testimony to press impact, but rather in terms of citizens actively patronising sources reflecting their pre-existing views.
But these arguments rely on assumptions often with very weak -sometimes nonexistent -evidential support. For example, explicit or implicit in the MIT are the following: the public have well-developed partisan orientations driving media choices (where 'development' is ascribed, unproblematically, to lived experience); this drive is strongly manifest in voting-newspaper alignment; citizens perceive accurately the partisan orientation of their newspapers, and such perceptions drive newspaper choices; and the market faithfully reflects a wide range of political messages. Yet MIT critics have, for decades, pointed to the decline in partisanship supposedly underpinning the selectivity mechanism (Harrop, 1987). Park et al. (2013: xxi) note that a fifth of the UK electorate currently have no party identification. Furthermore, while voter-newspaper alignments are identifiable -notably voting Conservative and buying the UK's Telegraph newspaper -even for this title the percentage of readers not voting for the party the paper unswervingly supports was 30%. And for the Mirror, Daily Mail, Times and Sun newspapers, the corresponding figures are 33%, 41%, 45% and 53% respectively. 4 This lack of alignment suggests that if voter predispositions are driving their newspaper choices, they are often not driving them particularly strongly. But these predispositions may not be capable of driving an individual's choice of newspaper at all, since a very substantial proportion of readers either have no idea (or the wrong idea), about the party their chosen title supports, or think it is politically neutral (Telegraph, 19%; Mirror, 33%; Daily Mail, 37%; Daily Express, 44%; Times, 45%; Sun, 59%). 5 And, in addition, between 39% and 49% of UK newspaper readers do not know where their newspaper sits on a left-right spectrum. 6 Furthermore, among the MIT-related studies outlined above not one offers survey evidence clearly signalling that perceptions of a newspaper's politics drives the title choices of its readers. 7 The notion that the free markets generate political message variety reflecting readership preferences has also been seriously undermined during the most commanding recent UK political events. During the Brexit campaign, the circulation of 'out' messaging newspapers outweighed 'in' by 80% to 20% (Deacon et al., 2016). At the 2015 General Election 57.5% of newspaper circulation backed the Conservatives, against 11.7% for Labour, 4.9% for UK Independence Party (UKIP) and 1.4% for 'coalition continuation' (Ponsford and Turvill, 2015). More striking still, at the 2014 Scottish Independence referendum, 'From a field of 18 daily and 12 Sunday titles … only the Sunday Herald, The National and a few columnists elsewhere were pro-independence' (Blain, 2017).
Finally, we should at least consider circumstances of uniform or near-uniform coverage where, consequently, there is narrower room for everyone to find congenial messages. Noelle-Neumann (1981) calls this 'consonant' coverage, suggesting it increases the likelihood that the messages will affect those exposed (Noelle-Neumann and Mathes, 1987). Consonant structures have an evidenced impact. For instance, in climate change coverage research, Feldman et al. (2012: 8) note that, 'when media cues are especially clear and one sided, this is enough to overwhelm partisan biases in processing'. Furthermore, a 'one-sided message flow' in European coverage has led to changes in political opinions (De Vreese and Boomgarden, 2006), while Peter's (2004: 144) comparative analysis of attitudes towards the European Union (EU) notes, 'Cognitive selection mechanisms did not protect EU citizens from the influence of consonant media coverage, which indicates powerful media effects'. Instances of consonance or near-consonance may be infrequent, but they are not unknown internationally (Eilders, 2002). And in Britain, Berry (2016: 542) notes a 'consistent endorsement of austerity measures, by almost all newspapers' in a post-Crash UK assessment, while assessments of stories in the early stages of Jeremy Corbyn's leadership of the Labour Party suggest a very heavy preponderance of negative coverage across a wide range of media (Cammaerts et al., 2016;Schlosberg, 2016).

Emerging and historic research on media impact
Even if such evidence does not comprehensively undermine the MIT, an emerging corpus of research suggests significant impacts of coverage on political attitudes. A range of qualitative studies suggests significant media effects in a range of domains (Miller et al., 1998;Philo, 2008;Philo and Berry, 2004). And quantitative research suggests media coverage has an evidenced impact on the image of UK political parties (Sanders and Norris, 1998), people's positions on policy (Feldman, 2011), and on levels of public cynicism about the political system (Jackson, 2011). A newspaper's support changing from one party to another also influences the political orientations of readers (Ladd and Lenz, 2009). And economic coverage on television has an impact on public optimism or pessimism about the UK economy (Sanders and Gavin, 2004).
However, it could still be contended that such studies often use experimental or focus group techniques. They may therefore lack naturalism or generalisability, or focus on unrepresentative 'samples of convenience' (see Goldstein and Ridout (2004) and Banducci (2018) meta-analyses for comprehensive critiques of such methodologies). Moreover, the size of the impact on economic perceptions may be quite modest (Sanders and Gavin, 2004) and, based on aggregate level analysis, cannot control for the partisanship driving reinforcement processes. Finally studies relating to unusual circumstances -that is, when newspapers dramatically change the party supported (Ladd and Lenz, 2009) -are, by definition, atypical and do not establish a general media impact.
Yet these critiques are much less relevant to research conducted at individual level. One such UK election study used a three-wave panel design, focusing on changing feelings about political parties (Brandenburg and Van Egmond, 2012). Importantly, this approach was tied to an evaluation of press coverage tone between panel waves, while measuring media exposure, and controlling for voters' pre-existing partisanship. The authors acknowledge that changing coverage tone did reinforce the feelings of those already committed (a point to which we will return), but found '… undecided voters to be sensitive enough to such variation that it makes a substantial contribution to the kind of attitude change that could bring about their ultimate vote choice. This is clear evidence of media persuasion'. Comparable evidence of media-driven direct persuasion and opinion changing are also evident in American studies of climate change coverage, where again pre-existing partisanship was carefully controlled for (Feldman et al., 2012) and a mix of cross-sectional and panel designs were deployed (Zhao et al., 2014).
A sense of the political importance of the media need not, however, be derived solely from this new body of evidence. Take, for instance, the 'media dependency' concept (Ball-Rokeach and DeFleur, 1976). This idea revolves around the notion that when people have no direct, immediate personal, 'lived' or vicarious experience of an issue, they are dependent on the media for information, and therefore more likely to be influenced. Early MIT theorising brooked this possibility in relation to international events, where citizen's direct experience is often limited (Gitlin, 1978: 216-217;Klapper, 1968: 85). However, several studies illustrate a level of citizen dependency on the media for information on a range of more proximate and nationally situated issues: for an appreciation of the national economic situation (Gavin, 1997), or an understanding of crime and the criminal justice system (Pickett et al., 2015), corporate reputations (Einwiller et al., 2010) and EU politics (Gavin, 2000).
Here a range of questions emerge. Where do people with no involvement with the benefit system derive their impressions of its operation or the extent of fraud within it?
How do people with little or no immediate contact with EU political machinery come to understand its characteristics or impact? How do people in areas of very low immigration develop an appreciation of migration levels and any problems that accompany them? This last question is especially relevant to the UK Brexit referendum, where perceptions of immigration were an important component in citizen decision-making (Hobolt, 2016). 8 But Lawton and Ackrill (2016) note that Brexit support was most pronounced precisely in those areas where immigrant numbers were lowest. This finding raises the issue of media dependency in such locations -especially when a great many individuals are known to rely on the media for information on the topic (Duffy and Frere-Smith, 2014: 92) -an issue to which we will also return below.
Beyond such questions, the dependency concept is especially relevant to particularly abstract, technically dense issues, such as the causes of national-level economic turbulence (Gavin, 1992) or the processes driving climate change. On the latter, Weber (2010: 334, emphasis ours) notes, Because climate change is so hard to detect and judge accurately based on personal experience, one might argue that its detection should be left to experts, namely climate scientists, and to their social amplifiers, the media and educators. Such delegation makes climate change a phenomenon for whose existence and likely magnitude and time course people have to rely on their beliefs in scientific observation and modeling, in expert judgments, and/or on reports about all of these in the mass media. Indeed, most people's knowledge and exposure to climate change has been almost entirely indirect and virtual, mediated by news coverage and film documentaries of events in distant regions (such as melting ice sheets in Greenland or Antarctica) that ascribe these events to climate change, events and arguments for which people's personal experience does not provide concurring evidence.
What we can conceive, then, are pools of media dependency, that may be shallow and dispersed but are nevertheless reasonably widespread and, therefore, potentially significant cumulatively. Such media dependency is known to augment the impact of messaging on public perception of at least some issues, notably corporate responsibility (Einwiller et al., 2010), criminal justice (Pickett et al., 2015) and immigration (Wolfsfeld, 2011: 105-106). Finally, if perceptions are not necessarily driven by lived experience, the disjunction problematizes the issue of the provenance of at least some pre-existing attitudes, in a way the MIT often fails to consider. Here we are obliged to ask: where do these preexisting attitudes come from? And we cannot necessarily exclude the media from an answer, on the basis of the MIT.

'Simply' or 'only' reinforcing?
Yet even if individuals are media-dependent, they may still use incoming information from this source to bolster pre-existent attitudes. There is, however, a narrowness to this reinforcement emphasis. This narrowness does not help us appreciate other aspects of the media's role in relation to attitudes and attitude development. For in some important instances, the significance of the media lies not in their capacity to change attitudes, but to undermine confidence in existing ones.
This idea is exemplified in the case of global warming messaging, and the media's role in casting doubt on climate science. There is a well-documented history of politicised climate denial mobilisation, globally, involving highly effective media dissemination tactics (Dunlap and McCright, 2010;Jacques et al., 2008). Observers have noted how political actors consciously sought to sow the seeds of doubt in public debate through such means (Boykoff and Boykoff, 2004: 133;Gelbspan, 2004: 41). Various media have carried climate sceptic messages on both sides of the Atlantic (Boykoff, 2008;Boykoff and Boykoff, 2004;Dunlap and Jacques, 2013;Feldman et al., 2012). And certainly in Britain, many studies show that these messages, juxtaposed with reports about mainstream climate science, generate public uncertainty and doubt (Butler and Pigeon, 2009;Dessai and Sims, 2010;Shuckburgh et al., 2012;Philo and Happer, 2013).
These processes are significant on two counts. First, they illustrate the MIT's conceptual weakness in its narrow focus on attitude change. For here it is not attitudinal alteration (nor even attitude reinforcement) that is the issue, but attitudinal uncertainty. Second, mobilisation fomenting doubt on climate science is no one-off. The associated strategies and tactics were borrowed from the tobacco industry, where one notorious internal memo from an executive read, 'Doubt is our product, since it is the best means of competing with the "body of facts" that exist in the minds of the general public' (Oreskes and Conway, 2010: 34). As Oreskes and Conway extensively evidence, it was borrowed because it was so singularly successful. They convincingly illustrate how the same robustly effective public relations techniques were used subsequently to seed uncertainty around the science of ozone depletion, acid rain, pesticide pollution and passive smoking. By the same token, Boyce (2007: 157-160, 179, 186-187) notes the uncertainty generated in Britain by politically driven messaging around the measles, mumps, and rubella (MMR) vaccine issue.
However, there is an additional danger in too narrow a focus on reinforcement, and it is that this restricted view may lead researchers to an overly sanguine view of the media's political impact. Here the issue is whether reinforcement processes and concomitant lack of attitude change necessarily suggest the media do not matter, or the processes have little or no political significance. Some eminent media scholars seem to think so. For instance, Bennett and Iyengar (2008: 725) suggest 'The increasing level of selective exposure based on partisan preference thus presages a new era of minimal consequences, at least insofar as persuasive effects are concerned', and reiterate this later stating 'Most social science scholarship … views reinforcement as being less consequential than attitude formation or change' (Bennett and Iyengar, 2010: 35).
But the idea that reinforcement is inconsequential was seriously questioned even in an early seminal defence of the MIT: It must not be assumed, however, that reinforcement which involves no change in opinion, is synonymous with lack of effect. Reinforcement is an effect, and an extremely important one. It strengthens the resolve in question, produces a type of immunity to counterpropaganda, and nurses straying sheep back into the fold. (Klapper, 1968: 86, emphasis original) As a more recent assessment rightly notes, 'classic social influence scholarship has identified persuasion as consisting of not only attitude change, but also attitude formation and attitude reinforcement' (Holbert et al., 2010: 17, our emphasis).
The idea that 'reinforcement' is politically significant and decidedly consequential, has long been acknowledged by political theorists seeking to define the mechanisms by which power is exerted. In the classic depiction, a political actor, 'A', has power insofar as 'A devotes his energies to creating and reinforcing social and political values and institutional practices that limit the scope of the political process to public considerations of only those issues which are comparatively innocuous to A' (Bachrach and Baratz, in Hay, 2002: 174, my emphasis). Brandenburg and Van Egmond (2012: 457) note, 'In an increasingly fluid political landscape, this [reinforcing pre-existing attitudes] may actually be a media effect of increasing value to parties'. But the notion that the media might also have the capacity to reinforce pre-existing anti-EU, antiimmigration or climate sceptic views, in significant portions of the electorate, further underscores the weakness of the argument that the processes involved are inconsequential. And we can rightly inquire whether reinforcing something like our anti-immigration views is speaking to the better angels of our nature.
Even if such capacities to reinforce sentiment -or, indeed, to persuade the undecided -affects only a small number of people, the processes have potentially profound political consequences. David Cameron's ability to call a Brexit referendum depended on getting a working majority. And the difference between a majority and none at all, hinged on a few tens of thousands of voters in a few marginal constituencies. The importance of the voters in these constituencies was not lost on Cameron's 2015 election campaign director, Lynton Crosby who acknowledges targeting very significant resources at them (see Note 3). Here the media's capacity to, in Klapper's words, 'strengthen the resolve' of Conservative voters, or nurses straying Conservatives 'back into the fold' -or, alternatively, persuade small numbers of undecided people to vote Conservative -was potentially the difference between Cameron being Prime Minister, or again a coalition partner; that is, between being in a strong position to call a referendum, or not.
This example of the media's potential importance in extremely close run campaigns is also not unique. Labour Party leader Neil Kinnock met his chief pollster Bob Worcester immediately after the 1992 general election defeat. Kinnock noted that '21' (the number of seats by which his party lost) was etched in his heart. But Worcester bid him disregard this number, since 1,240 should be written on your heart because that is the total number of votes that swung it in the Tories' bottom eleven seats. So you lost, my friend, not by twenty-one seats but actually by 1,240 votes. (Delaney, 2015: 158) Comparably, in the 2017 UK general election, the difference between a minority Conservative government and minority Labour one was 2,227 votes, Agerholm (2017) suggests. The broader significance of narrow electoral margins is acknowledged by Cambridge Analytica's CEO, Alexander Nix, who stated, 'elections are won by small but crucial numbers of votes -putting the right message in front of the right person at the right moment is more important than ever' (Channel 4 News, 19 March 2018), while the Managing Director of the same organisation notes, 'He [Donald Trump] won by 40,000 votes in three states. The margins were tiny' (Channel 4 News, 20 March 2018). But even if all these numbers are underestimates, they underscore the theoretical significance of the media's capacity to reinforce existing partisan sentiments -or, indeed, sway small numbers of the undecided. In this conception, the media's power to help reinforce the pre-existing attitudes of relatively few voters is consequential, and indeed may matter decisively.

Reinforcement: From informing to 'ignoranting' the electorate?
Yet even if such reinforcement is considered to be somehow marginal, there is another dimension to the associated processes which has the potential to be highly significant politically, but which MIT theory fails to explore. One of the media's principal functions is said to be keeping citizens informed (Anderson, 2007;Kuhn, 2007). But an important set of issues revolve around how they inform the public, how reinforcement processes might condition public engagement with the proffered information, and the theoretical implications that follow.
Taking UK immigration and benefit fraud as examples -two politically charged and controversial issues -a clear picture emerges of the information reaching citizens. For instance, Liz Gerard's work illustrates the very negative and often shocking tone of UK newspaper anti-immigration coverage, 2011-2016, and its crescendo preceding the Brexit vote, 9 emphasising 'out of control ' 'growth' and 'soaring', 'rocketing', and 'stampeding' 'influx' of immigrants, involving 'swamping' to 'crisis point' the UK's 'crowded isle'. More systematic assessments of UK press and television coverage also observe persistent focus on 'floods ', 'invasions', 'waves' and 'swarms' of immigrants, or 'large numbers' (Allen and Blinder, 2013;Cheregi, 2015;KhosraviNik et al., 2012;Philo et al., 2013;Spigelman, 2012).
Turning to benefit fraud, throughout the 1990s and 2000s the national UK press gave persistent and growing attention to this theme, and to 'benefit cheats' and 'scroungers' (Lundström, 2011: 320;Sage, 2012: 369). From around 150 stories in 1992-1993, coverage in 2007-2008 reached around 1000 reports per year containing these 'stereotypically negative phrases' (Sage, 2012: 367). This particular thematic emphasis has persisted subsequently across different media. For instance, British television has carried a range of programmes heavily featuring benefit fraud or scrounging themes -like the 155 episodes of the series 'Saints and Scroungers' (BBC 1, September 2009-October 2015 often reflecting the 'skivers and strivers' theme now firmly a part of British political discourse (Valentine and Harris, 2014). Likewise 'Undercover Benefits Cheat (Channel 5, September 2017-October 2017), 'Britain on the Fiddle (BBC1 November 2013 and March-April 2017), plus one-off documentaries like 'Tricks of the Dole Cheats' (Channel 4, Dispatches, August 2012). Furthermore, an assessment of UK press coverage across 2016, using the LexisNexis database on the search terms 'benefit fraud', 'benefit cheat', 'benefit scrounger', plus 'Benefits Street' and 'skivers and strivers', flags a total of 1374 national press stories carrying the terms. However, when the search is extended to include local/regional newspapers, this figure rose to 3073, underscoring Lundström and Sage's sense of persistent and growing coverage. Park et al. (2013: 45) note that the percentage of UK citizens agreeing that 'large numbers of people these days falsely claim benefits' rose from 69% to 82% in the period that Sage (2012) identified a surge in benefit fraud stories. Nevertheless the MIT would likely suggest that such coverage did not alter citizens' attitudes to benefit fraud, but only reinforced existing ones. However, would this conception miss 'collateral' impacts on knowledge about or, rather, misperceptions of benefit fraud levels? Such misperception is clearly evident: UK citizens think £24 in every £100 spent on benefits is claimed fraudulently, meaning 'people estimate that 34 times more benefit money is claimed fraudulently than official estimates [i.e. 70p in every £100]' (Ipsos MORI, 2013b). And £24 is the average, meaning a great many consider the figure is much higher.
Comparable levels of gross misperception are also apparent in respect to immigration levels: British citizens are 'massively wrong' on immigration and 'hugely overestimate' its scale (Ipsos MORI, 2013a: 5, 6), considering 31% of the UK population are immigrants, when it is actually 13%. Again, 31% is the average. Such misperceptions are not obviously a case of 'rational ignorance', where the costs of developing an accurate appreciation exceed the benefits, and where the emphasis is on not knowing, lack of understanding or unawareness (Caplan, 2004;Somin, 2015). Rather, such misperceptions can be considered tangible, and held with a high degree of certainty (Flynn et al., 2016), something certainly the case for UK perceptions of immigration levels (Ipsos MORI, 2014: 24).
Importantly, the MIT's reinforcement and selective interpretation processes are implicated in such misperception development: 'Why do people often hold these false or unsupported beliefs, and why is it sometimes so difficult to convince them otherwise? We argue that political misperceptions are typically rooted in directionally motivated reasoning' (Flynn et al., 2016: 127). At least with respect to perceptions of benefit fraud among the British public, the media are implicated in a manner consistent with this perspective: The focus groups all claimed that levels of fraud were much higher than they are in reality, with some suggesting that up to 70% of claimants were fraudulent. Participants justified these claims by reference to articles they had read in newspapers. (Briant et al., 2011: 4) In the section where Park et al. (2013: 52-54) explore 'What drives attitudes to the welfare state?', they make only one fleeting, incidental and passing reference to the media. And of the seven studies they cite in exploring the issue, only two mention the media, again only fleetingly, incidentally and in passing. But when exploring the sorts of misperceptions outlined above Flynn et al. (2016: 128) quite correctly argue for 'devoting more attention to the role of elites and the media, who seem to play a critical role in creating and spreading misperceptions'.
Such a media focus is especially appropriate in relation to UK immigration and benefit fraud when significant elements of the public are, as noted above, dependent on the media on these issues (see Lawton and Ackrill, 2016), and when perceptions of migrant numbers are associated with negative attitudes towards immigration (Sides and Citrin, 2007: 491-492), and also when immigration is of signal importance to people in their Brexit voting intensions (Hobolt, 2016) (see Note 8).
Finally it is noteworthy that, prior to the Brexit vote, anti-immigration news stories began to connect with a benefit fraud discourse:

Impact without persuasion or reinforcement -The 'third person effect'
But even if it is still considered that the MIT's reinforcement processes have few implications for power exertion or for public (mis)perception of social and political issues, the media may yet possess systemic influence. This influence is not via any actual capacity to change citizens' attitudes or behaviours, but through politicians' perceptions that the media might do so -a form of what has been termed the 'third-person effect' (Davison, 1983). The influence is exerted where there is fear that hostile coverage will impact on those exposed to it, and such reporting will subsequently erode the personal standing of a political actor, or the image of their organisation more generally.
The fact that politicians are wary of the media's potential impact on citizens is acknowledged in the international literature. Cohen et al. (2008) demonstrate that politicians often presume the media's impact on voters. Other researchers note that in a wide range of countries there is a perception among politicians that the media set the political agenda, and can make or break politicians (Lengauer et al., 2013;Maurer and Arendt, 2016;Van Aelst et al., 2008).
This mind-set extends to the United Kingdom, where Davis (2009: 214) showed that 'Just under half of the politicians asked, including 10 of the 16 (shadow) ministers, stated that journalists and the media had a key role to play in the rise and fall of ministers and in leadership contests'. This notion is entirely consistent with the observation of Lord Leveson who, after a rigorous assessment of UK press behaviour, declared: The press are in a unique position as they carry a very large megaphone: if people cooperate, that megaphone can be used to enhance careers; for those who complain or challenge titles, the megaphone can be used to destroy them (Leveson, 2012: 707).
As one British tabloid journalist notes, 'You could put the fear of God into an MP just by phoning and saying: Hi, I'm a reporter from the News of the World'. 10 The jumpiness of politicians may be because, as a senior journalist concedes, the press 'don't just expose, they also distort, so it is a nasty prospect' (Watson and Hickman, 2012: 119). Such journalism can also be extremely intrusive, with one veteran journalist noting, 'If the [Daily] Mail go for you, they get every phone number you have dialled, every schoolmate, everything on your credit card, every call from your phone and from your mobile. Everything' (Davies, 2008: 273).
The extent to which such a climate influences the behaviour of politicians is difficult to evidence, although Cohen et al. (2008) note that the greater the presumption of media power, the more strenuously politicians seek media attention. However, more tangible and politically non-trivial examples are discernible. On a range of issues UK politicians and institutional actors appear to have been influenced by their consideration of media responses. For instance, Andreadis and Smith (2007: 55) note, 'You can barely fill a taxi with senior mainstream politicians from Western Europe who do not believe action to mitigate and adapt to climate change is necessary. But most are frightened of sticking their neck out'. Such fear is unsurprising in a UK context, given the likely press reaction to the policies needed to mitigate the threat -including carbon pricing, road pricing, increased taxation, industry subsidies, plus regulation of, and state intervention in, markets (Giddens, 2009).
Further examples have emerged from Tony Blair's Labour government. Blair was leery of moves towards a proportional representation electoral system, saying '… if we change the rules on elections they [the press] will know they [the Conservatives] can never get back again. And then the whole mass of the Tory media will be deployed against me. I don't want that' (Ashdown, 2000: 522; see also 336). Blair was also said to be 'wary of joining the euro because of the Euro-sceptic press. It was another hurdle which would have to be faced'(see Note 2). And in evidence to the Leveson Inquiry, Blair showed comparable wariness about the prospect of effectively constraining the press through regulation: My view, rightly or wrongly, was that if in those circumstances I had said 'I've decided what I'm going to do is take on the media and change the law in relation to the media', my view is -and I think it's still my view, actually -is that you would have had to clear the decks. This would have been an absolute, major confrontation, you would have had virtually every part of the media against you in doing it. And I felt the price you would pay for that would actually push out a lot of the things I cared more about. Although I think this is an immensely important question, I don't in the end, not for me, at any rate, as prime minister, was it more important than the health service, or schools or law and order. 11 But instances of consequential inaction also extend beyond government. On finding concrete evidence of widespread instances of the press breaching the Data Protection Act, the Information Commissioner's Office -a UK body responsible for policing such actions -declined to prosecute those involved. As their senior investigator stated in evidence, under oath, to the Leveson Inquiry, 'The decision not to pursue any journalist was based solely on fear -fear of the power, wealth and influence of the press and the fear of the backlash that could follow if the press turned against the ICO'. 12 Finally, the UK Parliament's Culture, Media and Sport Select Committee decided not to issue a warrant forcing News Corp executive Rebecca Wade to give testimony about phone hacking because, as one Committee member said, 'if we went for her, called her back, subpoenaed her, they would go for us -which meant effectively that they would delve into our personal lives in order to punish them and I think that's part of the reason we didn't do it' (Newman, 2010).
Such instances are not necessarily symptomatic of an overwhelming general malaise facing all political actors, across all the issues they need to consider. But these examples suggest that, on some signally important issues, the media do not have to change the attitudes of the public to exert influence if political actors fear that they might. Such instances are sufficiently prevalent to underline yet another MIT theoretical weakness -its narrow focus on media-driven attitude changes among citizens.

Conclusion
The preceding assessment does not argue that the media are, in any sense, all-powerful. Any nuanced analysis should fully acknowledge that a variety forces -domestic and international -can have important, perhaps even profound impacts on attitudes and attitude development. For instance, Clarke et al.'s (2017) recent, extended and detailed analysis explored why people voted for Brexit. It evidences the disparate forces at play, canvassing a wide range of factors impacting on the decision itself, on the attitude of the public towards the EU and on their earlier support for UKIP. Prominent among them are economic attitudes, perceptions of which party best handled the economy, the image and likability of party leaders, attitudes towards EU control of the UK economy (that is, sovereignty), approval of UK membership of the EU (including the economic benefits and losses that derive from it), perceptions of immigration, emotional reactions to Europe and the salience of Europe as an important issue for people (whether it was on the agenda of citizens).
Many of these themes are highlighted in parallel analyses (Hobolt, 2016). But all have deep social, political and economic roots. However, exploration of these forces and factors by Clarke, Goodwin and Whiteley, and by Hobolt, also seems to exemplify some of the problems our preceding analysis sought to explore. For their use of variables often leaves under-considered the impact of the media upon attitudinal development. It should be acknowledged that some of this research shows only distinctive patterning in coverage, rather than any evidenced impact on attitudes. But such patterning is potentially significant nevertheless (as we have already argued). For instance, alongside research on media portrayals of an uneven balance in the economic costs and benefits of EU membership (Gavin, 2000), there is strong evidence of decades of media criticism of -and lacklustre support for -the EU, which is relevant to (dis)approval of the institution (Anderson and Weymouth, 1999;Hardt-Mautner, 1994;Hawkins, 2012). Research also illuminates persistent and strongly critical coverage of what was often characterised as the EU's exercise of sovereignty over Britain (Britain in Europe, 2000;Daddow, 2012;Dykes and Donnelly, 2012), and there is evidence strongly suggesting that, across a range of media, some party leaders were consistently handled very roughly in pre-Brexit media coverage (Cammaerts et al., 2016;Schlosberg, 2016).
Furthermore, there is overwhelming confirmation of UK coverage emphasising increasing immigration numbers (Allen and Blinder, 2013;Cheregi, 2015;KhosraviNik et al., 2012;Philo et al., 2013;Spigelman, 2012). The same literature (including Liz Gerard (see Note 9) and Balch and Balabanova, 2014) also detects a strong and consistent media emphasis on the threat of immigrants to the UK's economy and security, and its National Health Service (NHS), schools, housing and benefits systems -which is often unfounded (Fitzgerald and Smoczyński, 2015). Such threats are at least potentially relevant to the 'uneasiness' which, according to (Clarke et al., 2017: 158), is one prominent and significant component of the emotional reactions of UK citizens to the EU. Indeed, almost all the factors Clarke, Goodwin and Whiteley considered important were evident in media coverage during the campaign immediately preceding the Brexit vote (Moore and Ramsay, 2017).
Finally, and in addition to these considerations, there is literature evidencing the actual impact of media coverage on economic perceptions, and on attitudes towards which party best handles the economy (Sanders and Gavin, 2004). It finds very strong confirmation of the general agenda-setting power of the media (McCombs, 2005), as well as some supporting evidence on the impact of UK media stories about the EU on the salience for the public of the EU as an issue (Gavin, 2007: 146-149). All such media-related research went almost entirely without consideration, or theoretical exploration.
The preceding analysis has argued that there is ample evidence that the media can impact on attitude formation, especially (but not exclusively) where the public are dependent on coverage, have weak partisan predispositions, or where reporting is uniform or near-uniform across a range of sources. Furthermore, it should be appreciated that the media's capacity to reinforce pre-existing attitudes -whether these attitudes relate to the EU, immigration, benefit fraud or climate change -is a significant power in its own right. Moreover, the forces that drive reinforcement also have potentially important repercussions for public misperception of the issues. So even if reinforcement is the media's only impact, it is as well to remember that the results can, in fact, be consequential, and sometimes profoundly so.
Individually these observations would be enough to seriously qualify the MIT. But, cumulatively, they suggest a political significance for the media that is often seriously underappreciated or unacknowledged. They argue that we need to understand the structure and content of media coverage of prominent political issues, particularly where coverage is heavily structured and persists over time. Such an appreciation is unlikely to develop from researchers' casual observation of coverage. It will only emerge from acquaintance with systematic research on media content and its impact. Such research is less likely to be explored, or even considered, if reinforcement is seen as the primary media impact, and the processes involved are considered relatively inconsequential.