Why market narratives survive being proven wrong
- TEBI
- 21 hours ago
- 7 min read
Market narratives shape what investors believe long after the events that produced them. New experimental evidence suggests that when a story is debunked, the correction strips out far less of its influence than we assume — while a false number, once corrected, all but vanishes. Stories, it turns out, do not leave when the facts arrive.
A newspaper prints a retraction. A fund manager retracts a forecast. A widely shared claim turns out to be wrong, and the correction circulates in its place. The intuition most of us hold is that once the record has been set straight, the matter is closed. The false thing has been removed, and belief returns to where it started.
New research suggests that intuition holds for numbers but fails for stories. In a working paper circulated in May 2026, Robin Musolff, Christopher Roth and Florian Zimmermann, economists at the Universities of Cologne and Bonn, tested what a correction actually removes. The paper, distributed by CESifo, has not yet been peer reviewed. Its finding is that a debunked statistic leaves almost nothing behind, while a debunked story keeps working on the people who heard it.
What the experiment found
When a false statistic was corrected, and respondents were told with certainty that the information they had seen was irrelevant, its influence on their judgements all but disappeared. Belief moved by roughly one percentage point, against a rational benchmark of zero. A false story, corrected in exactly the same way, went on shifting beliefs by 4.22 percentage points: around three times as much.
That gap is the paper's central finding. The authors call the part of a story that survives its own correction the 'qualitative residue' — the trace a narrative leaves once every factual reason to believe it has been removed. In their words: 'False quantitative signals generate little to no belief updating, whereas false stories lead to substantial residual belief impact.'
The effect sizes are modest in absolute terms, and the researchers say so plainly. A few percentage points on a hundred-point scale is not a dramatic swing. What matters is the shape of the result rather than its size. Any movement at all is a departure from rational updating, because a signal known to be irrelevant should carry no weight whatsoever. The story produces roughly three times that departure.
The residue also proved hard to dislodge. It survived when the correction appeared on the same screen as the story. It survived when respondents had to confirm that they had read the correction. It even survived when the correction was delivered before the story was seen, so that readers knew the information was worthless before they encountered it. In each case the story kept its grip while the number let go.
How the researchers isolated the effect
The design was built to make that grip measurable. In a series of incentivised online experiments, respondents assessed hypothetical companies, each beginning from a neutral position: a score of 50 on a scale running from zero to 100. Each participant then saw a single analyst recommendation, positive or negative, which was equally likely to be relevant or irrelevant to the company in question. Sometimes the recommendation appeared on its own. Sometimes it came wrapped in a short, persuasive story pointing the same way.
Respondents then learned, with certainty, whether the signal had been relevant or irrelevant. This is what makes the benchmark clean. Once a piece of information is known to be irrelevant, it carries nothing, and a rational judge simply returns to the starting point of 50. Any residual movement is bias, and it can be measured exactly.
The baseline experiment ran with 1,500 participants recruited through Prolific, with further experiments of several hundred to a thousand each. The studies were pre-registered, and beliefs were tied to real money so that a careless guess carried a cost.
One further test located the source of the effect. When the stories were stripped of their evaluative content and made emotionally neutral, the residue vanished, falling to a level indistinguishable from a statistic. The pull came not from narrative detail as such, but from persuasive, emotionally charged content: a story that argues for something. The deck had been stacked against stories, each one short and generic and attached to an invented company, and the residue appeared anyway.
<!-- VISUAL: in-article image — Gemini prompt in metadata; caption to reference how the experiments isolated story effects -->
Why stories beat statistics in memory
Why should a story cling on when a number does not? The answer lies in an earlier study by the same research group, published in 2024 in the Quarterly Journal of Economics, one of the field's leading journals. Thomas Graeber, Christopher Roth and Florian Zimmermann set out to track how information decays in memory once it has been received.
Their finding was that the two forms of information fade at very different rates. Over the course of a single day, the impact of a statistic on people's beliefs fell by 73 per cent. The impact of a story fell by only 32 per cent.
There was a crossover. On first hearing, statistics actually moved beliefs more than stories did, by 21.76 percentage points against 18.19. A day later the ranking had reversed: the statistic's influence had dropped to 5.93 percentage points, while the story's held at 12.28. The number wins the moment; the story wins the week.
The mechanism is recall. After a day, 61.61 per cent of participants could correctly remember a story, against 28.70 per cent for a statistic. A story carries contextual detail, a scene, a reason, a sequence of events, and that detail gives memory more routes back to it. A bare number has few such handholds, and slips away.
Put the two papers side by side and the loop closes. They belong to one continuous research programme: Roth and Zimmermann appear on both, and Musolff assisted on the earlier study. Together they describe a single process. Stories outlast numbers in memory, and by the time a correction arrives, the story is already lodged while the number has already gone.
More confident, less correct
Two further findings sharpen the picture. The first concerns awareness. It would be reassuring to think that noticing a story's pull is enough to resist it, but most of the surviving effect sat with people who recognised that a story had influenced them and still could not fully correct for it. A smaller share, roughly a third of the average effect, came from people who said they had not been influenced at all. Knowing that a story has moved you does not confer the ability to move yourself back.
The second finding came from what people did rather than what they reported. Asked to describe a company in their own words in a voice recording, respondents' emotional tone tracked the direction of a story they knew to be worthless, around twice as strongly as after a false statistic. A classifier reading the audio itself, its intonation and rhythm and energy rather than the words spoken, could hear the residue. People chose neutral language while their delivery gave the story away.
Then there is confidence. Respondents were more sure of the beliefs they formed after false stories than after false numbers, scoring 41.6 against 37.6 on a hundred-point scale, even though those story-shaped beliefs sat further from the truth. Confidence rose more steeply with the size of the error under stories than under statistics. People were most certain at the very moments they were most wrong.
What this means for market narratives
None of this was an investment experiment. The participants were judging invented companies for small sums, not committing their savings; the 2026 paper is still awaiting peer review; and the real-world examples the authors reach for are drawn from politics rather than finance. The application to investing is an inference from the mechanism, not a claim the researchers make.
The inference is this. The information environment investors live in is built overwhelmingly from stories: fund marketing, manager interviews, the narrative of a crisis, the daily explanation of why the market rose or fell. If a debunked story keeps working on the people who heard it, then the market narratives that shape investor behaviour may operate on a channel that later facts and corrections only partly reach, leaving their holders more confident rather than less.
There is a long-running real-world pattern consistent with the idea. Morningstar's Mind the Gap research, which compares the returns investors actually earn with the returns of the funds they own, found that the average dollar in US funds earned 7.0 per cent a year over the decade to December 2024, against the funds' own 8.2 per cent. That shortfall of about 1.2 percentage points a year is given up mainly to the timing of purchases and sales. The gap is consistent with decisions driven by market narratives. It does not, on its own, prove them.
The story you already believe
The research leaves a quiet inversion behind it. The standard advice is to check the facts, and the facts can indeed be checked. But by the time the checking is done, the story has already done its work, and it does not leave when the facts do. The correction removes the reason to believe; it does not remove the belief.
Everyone carries beliefs formed by market narratives whose corrections arrived too late, or arrived and changed nothing, or never arrived at all. The implication of this work is not that some of those impressions are wrong. It is that the ones held with the most confidence may be the ones that deserve the most suspicion.
Resources
Graeber, T., Roth, C. & Zimmermann, F. (2024). Stories, statistics, and memory. The Quarterly Journal of Economics, 139(4), 2181–2225.
Morningstar. (2025). Mind the gap 2025. Morningstar, Inc.
Musolff, R., Roth, C. & Zimmermann, F. (2026). Learning from false stories (CESifo Working Paper No. 12711). CESifo.
Seeing past the story
The stories that move markets are difficult to switch off, and the research suggests they go on working long after the facts have moved on. A durable plan is the best defence against them. How to Fund the Life You Want, by TEBI editor Robin Powell and Jonathan Hollow, sets out an evidence-based way to build and spend your wealth that does not depend on reading each day's narrative correctly. Bloomsbury published the second edition, written for UK readers. Buy it on Amazon.
For readers who would rather have a professional help them tune out the noise, TEBI's Find an Adviser directory lists advisers who have publicly committed to evidence-based investing.