Availability cascade
A belief snowballs as repetition makes it feel true and dissent grows costly.
What it means
An availability cascade is a self-reinforcing cycle in which a belief gains public credibility through repetition rather than evidence. Kuran and Sunstein identify an informational mechanism, in which each repetition reads as independent confirmation; a reputational mechanism, in which private doubts go unvoiced to avoid social cost; and an availability mechanism, in which sheer repetition in media and conversation keeps the idea salient, so it comes to feel more probable and important — though the retrieval-ease account of why this happens is itself contested. The result can be wildly disproportionate alarm over small risks while larger ones are ignored, and policy made in response to vivid but unrepresentative cases. It matters as an account of how collective beliefs can detach from evidence through ordinary social dynamics — though it is a synthesis of parts of unequal evidential strength rather than a measured effect, its self-silencing leg is weak, and critics argue that risk perceptions track cultural worldviews at least as much as availability.
Why it accelerates
A cascade runs on a bookkeeping error. Everyone downstream counts each person who repeats a claim as an independent witness, but most repeaters are only passing on what they heard. Aggregate confidence rises while the evidence behind it stays the same size, or shrinks. The two engines compound: the informational one persuades the genuinely uncertain, the reputational one recruits the privately doubtful, and from outside their contributions look identical. Doubters cannot tell how many other doubters exist, because doubt is exactly what goes unspoken. That is also why cascades are brittle. Because the consensus rests on very little private information, one credible defector, or one hard number, can reverse it quickly. Sudden collapse of an apparently settled belief is the signature of a cascade, not evidence against it.
What the evidence shows
The cascade is a synthesis of parts, and the parts are not equally solid. The repetition leg is strong: a meta-analysis of 51 studies puts the effect of mere repetition on judged truth at roughly d = 0.4 to 0.5, depending on how the comparison is drawn, and repetition lifts implausible statements as much as plausible ones. Sequential-choice experiments reliably produce herding on other people's visible decisions. On Twitter, verified false rumours travelled farther and faster than true ones, though whether falsehood itself is the cause remains disputed. The weaker legs matter. Meta-analysis of the self-silencing mechanism finds a correlation of only about .10 between the perceived opinion climate and willingness to speak, rising to about .34 in face-to-face talk on issues people feel directly. The ease-of-retrieval reading of availability is shakier than it looks: meta-analysis finds the effect but not evidence that ease is what produces it, and a direct replication in a large national sample failed where most classic effects held. Treat the cascade as a plausible assembly, not a measured constant.
Where it shows up
Risk regulation is the home ground: a vivid case, amplified, draws spending far out of proportion to the hazard while duller and larger risks go unfunded. But the shape recurs wherever beliefs are formed by listening. Markets cascade into and out of assets, and a bank run is the limiting case, where a cascade tips into self-fulfilment: once enough people believe others will withdraw, withdrawing becomes correct, and the belief stops needing to be wrong to spread. Inside firms, a metric becomes what everyone knows because it has appeared in enough decks, and the person who asks for the original number looks obstructive. Public health runs both ways: cascades have produced unwarranted alarm about safe interventions and unwarranted calm about real ones.
Where the account breaks down
The concept is easier to apply backwards than forwards. Calling an episode a cascade requires knowing the belief was disproportionate, which is usually the very thing in dispute. Applied casually it becomes a way to dismiss any popular concern you do not share. Kahan and colleagues press this point against Sunstein directly: risk perceptions track cultural worldviews, not just availability, and experts diverge along similar lines, so 'the public cascaded, the experts were right' assumes its conclusion. Two further limits. Cascades often converge on the truth, because social learning usually works and the mechanism does not care which way it runs. And some alarms dismissed as hysteria were later vindicated. The useful question is not whether a belief spread socially, but whether anyone in the chain checked.
Examples
A single alarming report about a trace contaminant can spiral into national panic and costly regulation as coverage feeds on itself and questioning the scare becomes socially risky.
A rumour that a local school is unsafe spreads through parent group chats; each repetition makes it feel established, and the parent who asks for evidence is treated as callous, so nobody does.
Office talk that a reorganisation means layoffs snowballs: everyone hears it from three colleagues, treats that as confirmation, and voicing doubt starts to look naive, so the fear hardens without a single fact.
Analysts cite one another on a stock's obvious upside until the thesis is everywhere and nobody can name the original model; the dissenting note never gets written.
A surgical technique becomes standard because every conference talk assumes it works; trainees learn it as settled, and the surgeon asking for the trial data is heard as behind the times.
First described in Timur Kuran & Cass Sunstein (1999).
Key references
- Matthes, J., Knoll, J., & von Sikorski, C. (2018). The 'spiral of silence' revisited: A meta-analysis on the relationship between perceptions of opinion support and political opinion expression. Communication Research, 45(1), 3-33. doi.org/10.1177/0093650217745429
- Weingarten, E., & Hutchinson, J. W. (2018). Does ease mediate the ease-of-retrieval effect? A meta-analysis. Psychological Bulletin, 144(3), 227-283. doi.org/10.1037/bul0000122
- Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146-1151. doi.org/10.1126/science.aap9559
- Dechene, A., Stahl, C., Hansen, J., & Wanke, M. (2010). The truth about the truth: A meta-analytic review of the truth effect. Personality and Social Psychology Review, 14(2), 238-257. doi.org/10.1177/1088868309352251
- Kahan, D. M., Slovic, P., Braman, D., & Gastil, J. (2006). Fear of democracy: A cultural evaluation of Sunstein on risk. Harvard Law Review, 119, 1071-1109. papers.ssrn.com/sol3/papers.cfm?abstract_id=801964
- Kuran, T., & Sunstein, C. R. (1999). Availability cascades and risk regulation. Stanford Law Review, 51(4), 683-768. doi.org/10.2307/1229439
- Fazio, L. K., Rand, D. G., & Pennycook, G. (2019). Repetition increases perceived truth equally for plausible and implausible statements. Psychonomic Bulletin & Review, 26(5), 1705-1710. doi.org/10.3758/s13423-019-01651-4
- Anderson, L. R., & Holt, C. A. (1997). Information cascades in the laboratory. American Economic Review, 87(5), 847-862. econpapers.repec.org/RePEc:aea:aecrev:v:87:y:1997:i:5:p:847-62
- Yeager, D. S., Krosnick, J. A., Visser, P. S., Holbrook, A. L., & Tahk, A. M. (2019). Moderation of classic social psychological effects by demographics in the U.S. adult population: New opportunities for theoretical advancement. Journal of Personality and Social Psychology, 117(6), e84-e99. doi.org/10.1037/pspa0000171
- Juul, J. L., & Ugander, J. (2021). Comparing information diffusion mechanisms by matching on cascade size. Proceedings of the National Academy of Sciences, 118(46), e2100786118. doi.org/10.1073/pnas.2100786118