Effects, Agenda-Setting and What the Evidence Supports

The media is the institution through which most people learn most of what they know about everything else in this course.

And it is the institution about which the confident claims — from every direction — have most consistently failed to survive measurement.

For a century the field has argued about how much the media changes what people think, and the argument has swung between two extreme positions and back. Neither extreme was right, the truth is more interesting than either, and the reason persuasion effects are so small is one of the most useful findings in the discipline.

The broadcast that supposedly caused a panic, and what actually happened.

On the evening of 30 October 1938, an American radio network broadcast a dramatisation of an alien invasion, presented as a series of interrupting news bulletins. The story that grew up around it is one of the most repeated in the history of media: that millions of listeners, mistaking the drama for real news, fled their homes in terror — roads jammed, people praying in the streets, a nation panicked by a radio programme.

It became the founding parable of media power. If a single broadcast could do that, the media could do anything.

Almost none of it is true.

The audience that night was small. A far more popular programme aired on a rival network at the same time, and survey data from the period shows the great majority of radios that were on were tuned to it. Most people who heard the broadcast understood it was a drama — it was announced as such, and it followed the network's regular schedule.

The evidence for mass panic is thin to non-existent. The dramatic accounts came overwhelmingly from newspapers — an industry with a direct commercial and institutional interest in portraying radio, its new competitor, as dangerous and irresponsible. Later reconstructions found few verifiable instances of anyone actually fleeing, and the studies that established the "panic" relied on a small number of self-reports gathered weeks afterward, from people prompted to describe fear.

So the founding demonstration of overwhelming media power was largely manufactured — by a competing medium — and has been repeated for eighty years because it makes a better story than the truth.

Which is, once again, 7.9.1 in the least expected place, and it is the right place to begin, because the history of media effects research is a history of the field slowly discovering that the media does far less than everyone — advertisers, propagandists, critics and moral reformers alike — was certain it did.

Why is it so hard to show that the media changes minds?

The intuition that it must be powerful is overwhelming. Enormous sums are spent on advertising and political campaigning on the assumption that messages change behaviour. Propagandists believed it. Critics of the media believe it. And yet the measured effects, when the measurement is done well, are persistently small.

Three reasons, and they compound.

People are not blank. They arrive with existing beliefs, loyalties and group memberships, and they process messages through those (see 7.4.1 on considerations). A message that fits is accepted; one that does not is resisted, discounted, or reinterpreted.

People are not captured. They choose what to attend to, increasingly so, and they choose things that fit what they already think — selective exposure — so persuasive messages disproportionately reach people already persuaded.

And people are not alone. Messages are discussed, filtered and reinterpreted through social relationships, which frequently matter more than the message itself.

Each of these blunts direct effects. And the field's central discovery is that once you account for them, the residual capacity of a message to change a mind is small — but that this does not mean the media has no power, because its power operates through a different channel entirely.

The three phases

One — the hypodermic model, and why it was assumed rather than found.

The earliest assumption — never really a tested theory — was that media messages are injected directly into a passive audience, producing more or less uniform effects. The image is a needle, or a bullet.

It arose in a particular context: the propaganda of the First World War, the rise of mass advertising, and the mass mobilisations of the 1930s, all of which seemed to demonstrate that populations could be moved at will.

It was assumed, not established. And when researchers actually began measuring, in the 1940s, it collapsed.

Two — the minimal effects finding, and the study that produced it.

A team studying the 1940 American presidential campaign followed the same voters through the campaign — a panel design (see 7.4.4) — expecting to document the media converting voters.

They found almost no conversion. The overwhelming majority of voters decided as their social characteristics and prior loyalties predicted, before the campaign began. The campaign's main effect was reinforcement — activating existing predispositions and bringing wavering partisans home — not conversion.

And they found something that reoriented the field: the two-step flow. Information and influence flowed from media to opinion leaders — engaged, attentive individuals — and from them, through personal conversation, to everyone else. The influential channel was not the broadcast but the interpersonal relationship, and the media's effect was mediated through people.

This became the minimal effects position: the media mostly reinforces, rarely converts, and works through social relationships rather than directly. It dominated for two decades.

And it was an overcorrection, in the same way the hypodermic model had been.

Three — the return of effects, relocated.

The rediscovery of media power came from asking a different question. The minimal-effects researchers had asked whether the media changes what people think — their opinions and votes. The answer was: rarely.

The new question was whether the media changes what people think about — and the answer was: constantly.

What the media actually does

Agenda-setting: the effect that survives.

The media may not tell people what to think, but it is stunningly successful at telling them what to think about.

The founding study compared the issues a sample of undecided voters considered important with the issues the news media had emphasised. The correlation was very high: the ranking of issues in people's minds matched the ranking of coverage.

And the design was strengthened over time to establish direction. Panel studies, and eventually controlled experiments in which participants viewed news broadcasts edited to emphasise different issues, found that manipulating coverage changed which issues viewers rated as important — which is a causal claim, established experimentally (see 7.4.2), and it holds.

Two extensions matter.

Priming. By making an issue salient, the media changes the criteria by which people evaluate — for instance, judging a leader on the issue currently prominent. The agenda does not change the opinion of the leader directly; it changes which considerations the opinion is built from.

And the second level: attribute agenda-setting. The media influences not only which objects are salient but which attributes of them are — not just that the economy is important, but that unemployment rather than inflation is the aspect that matters.

Agenda-setting is the best-established media effect, and it is not persuasion. It is control over attention — which, in a world of unlimited things to attend to, is a very great power indeed.

Framing: the same facts, a different conclusion.

A frame is a way of organising a story — what is included, what is emphasised, what the problem is taken to be, who is responsible, and what would count as a solution.

The same facts, framed differently, produce different responses. A classic experimental demonstration: participants shown a report of a rally by a controversial group, framed as a question of free speech, were more tolerant of it than those shown the same rally framed as a question of public order — identical event, different frame, different conclusion.

The mechanism is that a frame activates some considerations rather than others (see 7.4.1, 7.6.4 on how the same information supports different conclusions).

And the framing literature has its own replication caution. The largest framing effects come from artificial studies presenting a single frame with no competition. In the real world, frames compete, and when people are exposed to opposing frames — as they usually are — the net effect is substantially smaller, and can cancel. Competitive framing effects are real but modest, which is the pattern this whole lesson keeps finding.

Cultivation: the long, slow effect.

A separate tradition argued that the effect of heavy media consumption is not on any particular opinion but on the general picture of reality. Heavy television viewers, on this account, come to see the world as more dangerous, more crime-ridden and more violent than it is — a mean world perception — because they absorb the statistical world of television rather than the world.

The evidence is real but the causal reading is contested. Heavy viewers do hold these perceptions; whether viewing produces the perception or anxious people watch more is the standard problem (see 7.6.2), and the measured associations are small and shrink under controls. Cultivation is best treated as a real but modest and hard-to-identify effect, not the dominant force its strongest advocates claimed.

And the persuasion finding, which is the most useful and most counterintuitive result.

The best modern evidence on political persuasion comes from field experiments — randomised trials of campaign contact, advertising and messaging, measuring actual behaviour (see 7.4.2).

The central finding is that the persuasive effect of campaign contact in general elections is, on average, close to zero.

A major meta-analysis of field experiments found that the best estimate of the average effect of campaign contact on candidate choice in general elections was indistinguishable from nothing — with the important qualifications that effects are larger in primaries and ballot-measure campaigns (where partisan loyalty is not doing the work), larger for unfamiliar issues, and that early persuasive effects decay rapidly, often within days.

Read that carefully, because it is genuinely surprising. Vast sums are spent on general-election persuasion, and the average measurable effect on vote choice is about zero. The money is not wasted on mobilisation — getting existing supporters to turn out is a real and measurable effect — but conversion, the thing everyone assumes the media does, is the thing it does least.

This is the minimal-effects finding, re-established with far better methods sixty years later. People are not blank, not captured, and not alone, and the compound result is that direct persuasion is weak — while agenda-setting, the control of attention, is strong.

The two findings are not in tension. They are the whole picture. The media has little power to change your conclusion and enormous power to determine what you are reaching a conclusion about.

Two structural questions the effects literature brackets, but which matter.

Who decides what gets covered? The effects tradition studies what audiences do with content; a separate tradition studies where content comes from.

News is manufactured under constraints — deadlines, budgets, the need for a reliable supply of stories — which produces systematic patterns: reliance on official sources, who become the routine definers of what is news; news values that favour the novel, the dramatic, the personal, the negative and the elite; and the reproduction of official framings because officials are the accessible, credible, cheap source (see 7.4.4 on official statistics as institutional products; this is the same point about news).

A propaganda-model version argues these constraints — ownership, advertising dependence, source reliance — systematically filter news toward the interests of the powerful, without requiring any conspiracy, purely through the structure of the industry. The mechanism is real; the strong claim that it produces uniform outcomes is contradicted by the visible variation between outlets and countries, which the model handles poorly.

And does representation shape belief? The evidence on media representations of groups — which groups appear, in what roles, doing what — establishes that representations are patterned (content analysis; see 7.5.3), and that the patterned representations can affect perceptions. What is harder to establish is the size of the effect on behaviour and belief in the real world, for all the reasons this lesson has given — which is not an argument that representation does not matter, but a caution against assuming the effect is large without measuring it.

Because the confident claims about media power — from advertisers, from critics, from moral reformers, from propagandists — are mostly wrong in the same direction, and the true picture is more precise and more useful.

What the evidence supports.

Agenda-setting is strong. The media determines what a population attends to, and that is established causally.

Framing effects are real and modest, and shrink when frames compete — which they do in reality.

Direct persuasion is weak — the average effect of general-election campaign contact on vote choice is about zero — and mobilisation of existing supporters is real.

And effects work through people: prior belief, selective exposure and social relationships all blunt direct effects, which is why the two-step flow keeps reappearing under new names.

Four questions for any claim about media power.

Is it a claim about what people think, or what they think about? The first is usually weak; the second is usually strong.

Was the effect measured on behaviour, or on a survey response in an artificial setting? Single-frame studies overstate; competitive and behavioural designs deflate.

Does the claim account for selection? Heavy consumers of anything differ from light consumers before any effect.

And who benefits from the claim of media power? The panic story was manufactured by a rival medium; claims of vast advertising effects are made by people selling advertising; claims of vast propaganda effects justify censorship. The belief in media power is itself something with interests behind it.

One closing observation, which sets up the next lesson. All of this evidence — the minimal effects, the agenda-setting, the weak persuasion — was established in the era of a few mass channels reaching a broad common audience.

The conditions that produced those findings — a shared informational environment, limited channels, professional gatekeepers — have substantially dissolved. Whether the findings survive the change is the question 9.6.2 has to answer, and it is genuinely open.

The 1938 broadcast that supposedly caused mass panic largely did not: the audience was small, most understood it was drama, and the "panic" was manufactured by newspapers with an interest in portraying radio as dangerous — the founding parable of media power, itself a fabrication.

Direct persuasion is hard to demonstrate because people are not blank (they process through prior belief), not captured (selective exposure), and not alone (social mediation) — and these compound.

Three phases. The hypodermic model assumed uniform direct effects and was never tested; it collapsed when measurement began. The minimal-effects finding from 1940s panel studies found the media mostly reinforces rather than converts, and works through a two-step flow via opinion leaders — and was itself an overcorrection. And effects returned, relocated: from what people think to what they think about.

Agenda-setting is the best-established effect and is not persuasion — the media tells people what to think about, established causally by experiments manipulating coverage. Priming changes the criteria of evaluation; attribute agenda-setting changes which aspects are salient.

Framing — the same facts organised differently — produces different conclusions, but the large effects come from single-frame studies; competitive framing, as in reality, is modest. Cultivation — the mean-world perception of heavy viewers — is real but small and confounded by selection.

And the persuasion finding is the counterintuitive result: field experiments find the average effect of general-election campaign contact on vote choice close to zero, larger in primaries and ballot measures, and rapidly decaying — the minimal-effects finding re-established with far better methods. Mobilisation is real; conversion is what the media does least.

Persuasion is weak and agenda-setting is strong, and these are not in tension: little power over your conclusion, enormous power over what you conclude about.

Two structural questions: news is manufactured under constraints favouring official sources and dramatic news values; and representations are patterned, though their real-world effect on belief is harder to size than to assume.

Hypodermic / magic bullet model — the untested assumption of uniform direct media effects.

Minimal effects — the finding that media mainly reinforces existing predispositions rather than converting.

Reinforcement — activation of existing beliefs rather than change.

Two-step flow / opinion leaders — influence passing from media to attentive individuals to others through conversation.

Selective exposure — the tendency to attend to content that fits existing beliefs.

Agenda-setting — media influence over which issues are considered important; strong and causal.

Priming — changing the criteria by which people evaluate by changing what is salient.

Framing — organising a story so as to activate particular considerations; large in single-frame studies, modest under competition.

Cultivation / mean world — the perception of a more dangerous world among heavy consumers; small and confounded.

Persuasion (field-experimental) — the near-zero average effect of general-election campaign contact on vote choice; mobilisation excepted.

News values / manufacture of news — the constraints and preferences shaping what becomes news.

One — separate the two questions. Take a claim that "the media made people believe X". Ask whether it is a claim about opinion or about attention. The second is usually the defensible one.

Two — find the frame. Take one news story and identify what it treats as the problem, who it holds responsible, and what would count as a solution. Then write the same facts under a different frame.

Three — check the agenda. For one week, note the issues the news treats as most important, and ask whether their prominence tracks any change in the world or only in coverage.

Four — apply the selection question. Take any claim that media consumption causes an attitude. Ask whether people with the attitude might consume more of the media. Both directions are usually plausible.

Five — ask who benefits. For any strong claim about media power, identify who gains from the claim being believed. The answer is rarely neutral.