Mental ledger · the same value, rung up differently
Zero-Risk Bias
The tendency to prefer completely eliminating one source of risk over a policy that produces a larger total reduction in risk but leaves some residual risk in place.
The Mental Ledger
where labels override totals
Found a positive but diminishing valuation of risk reductions, plus a distinct 'certainty premium' for total elimination of one risk; increases in risk were valued far more than equivalent decreases (reference-risk effect). No strong extra premium for eliminating multiple risks.W. Kip Viscusi, Wesley A. Magat, Joel Huber · 1987
A substantial minority preferred the complete-elimination option despite its smaller total benefit: about 42% of respondents ranked the dominated zero-risk option better than at least one superior option, demonstrating willingness to forgo greater aggregate risk reduction (6 cases prevented elsewhere) in order to fully zero out one site.Jonathan Baron, Rajeev Gowda, Howard Kunreuther · 1993
The arithmetic is identical down both columns — only the label changed. That difference is the bias.
The accounting rule the mind uses
The dominant explanation is the certainty effect from prospect theory (Kahneman & Tversky, 1979): people overweight outcomes that are certain relative to merely probable ones, so the step from a small residual risk to exactly zero is valued far more than a numerically equal reduction that leaves risk above zero. A complementary affective account holds that complete elimination delivers psychological 'closure' and relief from worry about a specific threat, which a partial reduction cannot. Both predict that the boundary at zero is treated as categorically special rather than as one more point on a probability scale.
Competing account — Schneider et al. (2017) argue the effect is partly shaped by method and context rather than a fixed preference: it appears more in abstract questionnaire tasks than concrete behavioral ones, varies by decision domain, and individuals' zero-risk choices do not correlate across tasks — suggesting situational appropriateness and task framing, not a single stable bias, drive much of the behavior. Protected-values / deontological accounts (Baron's broader work) also frame some elimination preferences as moral rather than purely probabilistic.
Audit — mixed
The basic phenomenon — a premium on reaching zero — has been observed across willingness-to-pay studies and choice tasks since 1987 and is widely cited. However, the most direct methodological scrutiny, Schneider, Streicher, Lermer, Sachs & Frey (2017), ran four studies and concluded the bias is persistent across methods but 'highly sensitive to contextual factors': it emerged more in abstract than concrete tasks, varied by domain, and individuals' zero-risk choices did not correlate across tasks. That makes it real but context-dependent rather than a robust, trait-like effect, and means simple prevalence figures should be read as situation-specific.
Charged in the wild
law / regulation · 1958
Delaney Clause (U.S. food-additive law)
The 1958 Delaney Clause barred any food additive shown to induce cancer in humans or animals at any dose, mandating zero tolerance regardless of actual risk magnitude. It is frequently cited (including in the Wikipedia synthesis of the literature) as an institutional embodiment of zero-risk thinking, since improving detection thresholds forced bans on vanishingly small risks; it was substantially amended for pesticide residues by the Food Quality Protection Act of 1996.
environmental policy · 1980s-1990s
Superfund hazardous-waste cleanup standards
U.S. Superfund site remediation has been criticized for pursuing near-complete cleanup of low residual risks at very high marginal cost-per-life-saved, an allocation pattern the zero-risk literature uses as a real-world parallel to the laboratory finding. This regulatory context directly motivated the Baron, Gowda & Kunreuther (1993) hazardous-waste experiment.
To balance the books
Reframe options in terms of total expected harm prevented (lives, cases, dollars of risk) rather than which risks reach zero, and force an explicit comparison of aggregate reduction across all options — including the cost-per-unit-of-risk-reduced. Asking 'which choice prevents the most harm overall?' rather than 'which makes this risk disappear?' neutralizes the certainty premium.
Baron, Gowda & Kunreuther (1993) showed the bias appears precisely when a salient complete-elimination option competes with a larger-but-partial reduction; Schneider et al. (2017) found the effect shrinks in concrete, well-specified tasks, implying that making aggregate consequences explicit reduces it.
Catch it in the act
Watch for choices framed as 'let's get this one risk to zero / make it fully safe' when the budget or effort could instead cut a larger total amount of harm spread across several risks. The tell is treating the gap between 'almost zero' and 'exactly zero' as more valuable than an equal or bigger reduction elsewhere — e.g., spending heavily to eliminate a tiny contaminant while a larger hazard goes only partly addressed.
Related accounting tricks
- Certainty effectparentThe prospect-theory mechanism (overweighting of certain outcomes) from which zero-risk bias is usually derived; zero-risk bias is essentially the certainty effect applied to risk-reduction choices.
- Protected values / sacred valuesmechanistically-linkedBaron's broader program suggests some elimination preferences reflect deontological 'this risk must be zero' values rather than probability weighting.
- Probability neglectsiblingInsensitivity to the size of a probability; both lead people to treat the move from small risk to zero as larger than its numeric value.
- Omission biaseasily-confusedPreference for inaction over action with the same expected harm; studied by Baron but distinct from preferring complete elimination of one risk.
- Loss aversionsiblingThe reference-risk asymmetry Viscusi et al. found (increases valued more than decreases) is a loss-aversion phenomenon adjacent to the certainty premium.