Category: Effects
Type: A preference for eliminating one risk when total harm falls by as much, or by less
Origin: W. Kip Viscusi, Wesley A. Magat, and Joel Huber, RAND Journal of Economics (Winter 1987); Jonathan Baron, Rajeev Gowda, and Howard Kunreuther, Risk Analysis (1993)
Also known as: A certainty premium on the step that reaches zero
Type: A preference for eliminating one risk when total harm falls by as much, or by less
Origin: W. Kip Viscusi, Wesley A. Magat, and Joel Huber, RAND Journal of Economics (Winter 1987); Jonathan Baron, Rajeev Gowda, and Howard Kunreuther, Risk Analysis (1993)
Also known as: A certainty premium on the step that reaches zero
Quick Answer — Zero-Risk Bias is a preference for driving one risk to zero even when another plan prevents as many harms, or more. Viscusi, Magat, and Huber measured a certainty premium in 1987: among equal cuts in injury risk, the cut that reached zero cost more than the middle cut in every group, and more than the first cut in three of the four. Baron, Gowda, and Kunreuther gave the pattern this name in 1993. Where complete cleanup of the smaller city prevented one fewer cancer case, 42 percent of the people who saw that wording did not rank the plan worst.
What is Zero-Risk Bias?
Zero-Risk Bias is a preference for eliminating one risk completely, even when another use of the same resources prevents as much total harm, or more. The 1993 paper states the stricter form. Complete cleanup of one site is preferred even though the same money, spent on partial cleanup of two sites, would prevent more cancer cases. An equal-saving form is scored separately. There, the plan that hits zero is ranked best even though the other plans prevent the same number of cases.Zero-risk bias is defined as a preference for cleaning up a waste site completely, even if the total number of lives saved as a result of this strategy is lower than the number that would be saved if the identical resources had been expended toward cleaning up two different waste sites partially.That sentence is Baron, Gowda, and Kunreuther in 1993. The dollar pattern is older, and it has a different name in the paper that measured it. Viscusi, Magat, and Huber called an extra willingness to pay for total elimination a certainty premium. Prospect Theory uses certainty effect for a different comparison: people underweight an outcome that is merely probable, relative to an outcome obtained with certainty, in choices among money gambles. Omission Bias is a harsher rating of harm from an act than of the same harm from not acting. A vaccine risk of zero can look like both patterns. The comparison on this page is cases prevented, or dollars for an equal cut that does or does not hit zero.
Zero-Risk Bias in 3 Depths
- Beginner: Two written cities share a cleanup budget. One plan clears the smaller city’s cancer cases and leaves some in the larger city. Write down how many cases each plan prevents before you prefer the zero.
- Practitioner: For a product with two injury risks, price three equal cuts of 5 injuries per 10,000 bottles. Read the cut that reaches zero against the two cuts that do not.
- Advanced: The 42 percent is a mail ranking of a simplified city, scored as not placing the worse plan last. The $2.41 is a mean extra price in one of four interview groups. Neither figure is a cleanup budget or a store receipt.
Origin
The dollar result is Viscusi, Magat, and Huber, “An Investigation of the Rationality of Consumer Valuations of Multiple Health Risks,” RAND Journal of Economics, volume 18, number 4 (Winter 1987), pages 465–479. The counts below are from that article. Shoppers were interviewed at a shopping mall and a hardware store in Greensboro, North Carolina. The paper calls the draw a representative mix of consumers and points to Viscusi and Magat (1987) for that description. This page does not re-read that book. Each person considered one product. The interviewer stated a price of 2 a bottle for a toilet bowl cleaner, and a current risk of 15 injuries of each of two types per 10,000 bottles sold. The analyzed insecticide groups are 672 people with no child under five and 113 people with a child under five. The toilet bowl cleaner groups are 551 and 183 on the same split. Those four counts add to 1,519. The abstract says a sample of over 1,500. The questions asked how much extra the person would pay for a safer bottle. The authors call this the first quasi-market test of certainty premiums. It is a stated price in an interview. It is not a purchase. They list three reasons a last step to zero might cost more. Eliminating every risk on a product can cut anxiety and the cost of thinking about a chance. A person who overestimates a small remaining risk, and who treats zero as zero, will overestimate how large that last cut is. Prospect Theory, Daniel Kahneman and Amos Tversky, Econometrica, volume 47, number 2 (March 1979), is the third reason they name. The abstract of that paper calls the certainty effect a tendency to underweight outcomes that are merely probable, relative to outcomes obtained with certainty. This page does not carry the 1979 gamble counts. The name on this page is Baron, Gowda, and Kunreuther, “Attitudes Toward Managing Hazardous Waste: What Should Be Cleaned Up and Who Should Pay for It?,” Risk Analysis, volume 13, number 2 (1993), pages 183–192. Baron was in the Department of Psychology at the University of Pennsylvania. Gowda was in the Science and Public Policy Program at the University of Oklahoma. Kunreuther was at the Wharton Risk Management and Decision Processes Center. They mailed simplified cases about cleanup and about who should pay. They cite Viscusi, Magat, and Huber, and an in-press paper by Ilana Ritov, Jonathan Baron, and John C. Hershey, “Framing Effects in the Evaluation of Multiple Risk Reduction,” Journal of Risk and Uncertainty. This page does not carry that in-press paper’s counts.Key Points
Four comparisons keep the size of the cut, or the cleanup budget, fixed, and move whether one risk hits zero. A stated bottle price is not a receipt. A rank on a written city is not a cleanup that was funded. A mean can favor the last step while one column only ties the first step. Clearing both risks at once is a separate test from clearing one.1
Read the last equal cut against the earlier cuts
The baseline was 15 injuries of each type per 10,000 bottles. Table 4 prints the extra dollars for cutting both risks by another 5 per 10,000, at three starting levels. For insecticide users with no child under five, the printed increments are 0.34 from 10, and 1.04 (standard error 1.37 (standard error 3.78 (standard error 1.37 to 2.41. The gap from 1.37 is 0.34. This page does not force those two prints into one cent.
2
Keep the four groups apart
The other three columns of Table 4 follow the same shape, with one tie. Insecticide users with a child under five: 0.54, then 0.65, then 0.83. Toilet bowl cleaner with a child under five: 0.24, then 8.09 (standard error 10 starting price. The mean for eliminating both toilet risks in that group is 0.22), against a $2 starting price. The authors say that, for all four injury pairs, at most 11 percent of subjects valued the cut from 10 per 10,000 down to 5 more than the cut from 5 down to 0. A large majority valued the cut that reached zero more highly. The paper does not print that majority as one percentage.
3
Do not add a second premium for clearing both
A separate test asks whether eliminating both risks costs more than the sum of eliminating each risk on its own. For insecticide users with no child under five, eliminating skin poisoning alone has a mean of 0.13), eliminating inhalation alone has a mean of 0.15), and eliminating both has a mean of 0.18). The sum of the separate means is $3.90. Table 6 compares, person by person, the sum of the two separate answers with the answer for eliminating both. Three of the four injury pairs have a positive mean difference, which would mean the sum exceeds the joint price. None of the four mean differences is significant at the 5 percent level. The authors’ closing reading is that there is no additional premium for eliminating multiple risks beyond the sum of the premiums for eliminating each one. They also say that this pattern points away from anxiety about the mere presence of any uncertainty. The driver they name is something about the particular risk, such as overestimating a low probability.
4
Score a zero by the cases it prevents
Baron, Gowda, and Kunreuther gave half the subjects one wording and half the other. The paper does not print how many people were in each half. Two cities have landfills. The larger city has 2,000,000 people and 8 cancer cases a year from leakage. The smaller city has 1,000,000 people and 4 cases. Partial cleanup of both sites cuts the larger city from 8 to 4 and the smaller from 4 to 2, which prevents 6 cases. Concentrating on the larger city cuts it from 8 to 3 and the smaller from 4 to 3, which also prevents 6. Total cleanup of the smaller city, plus a partial cleanup of the larger, cuts the smaller city from 4 to 0. For half the subjects the larger city goes from 8 to 6, so the plan prevents 6. For the other half it goes from 8 to 7, so the plan prevents 5.
Applications
Use the cases prevented, or the extra dollars for an equal cut, as lines you can read. Keep a stated bottle price inside the interview that asked for it. Keep a mail rank inside the wording that person saw.Write the cases before you fund a zero
Before a plan that clears one site wins, write the cases each plan prevents. In the equal-saving wording, complete cleanup of the smaller city prevents 6, the same as the other two plans. Eighteen percent of the subjects who saw that wording ranked the complete-cleanup plan best.
Price the step that hits zero on its own line
Split a risk reduction into equal steps and mark the step that reaches zero. For insecticide users with no child under five, Table 4 prints 0.34, and 2.41 is the cut from 5 to 0. It is a mean stated extra price for a bottle, not a receipt.
Test the sum against clearing both risks
Ask whether wiping out every risk on one product costs more than wiping out each risk alone. In the no-child insecticide means, 1.85 is 3.78. Across four pairs, none of the mean gaps between the sum and the joint price was significant at the 5 percent level.
Keep a mail rank off a cleanup budget
The 42 percent did not rank a plan worst. The plan prevented 5 cancer cases, and the other plans prevented 6. Eleven percent ranked that plan best. The authors say the samples were not representative, and that a mail questionnaire cannot show why a person answered. Those ranks are not a share of a city’s budget.
Case Study
Jonathan Baron, Rajeev Gowda, and Howard Kunreuther mailed a questionnaire about hazardous-waste cleanup. The article is in Risk Analysis, volume 13, number 2 (1993), pages 183–192. The cases were simplified on purpose. The authors say they were contrasting principles, not describing a real site. The people who received the mailing included chief executives of oil and chemical companies, economists, environmentalists, people who had attended a 1990 hazardous-waste conference, retired and semi-retired judges who served as arbitrators, and legislators. Return rates differed by group. The authors say the samples were not representative of those groups. This page does not add those returns into one head count. The legislator line prints 104 of 1,170, and it also prints 67 returns from a first mailing and 27 from a second. This page does not reconcile 104 with 67 and 27. The zero-risk question is Case 5. Two cities, one of 2,000,000 people and one of 1,000,000, have leaking landfills. Leakage will cause 8 cancer cases a year in the larger city and 4 in the smaller. Funds are limited. Three plans are ranked. Partial cleanup of both sites prevents 6 cases. A plan that concentrates on the larger city also prevents 6. A plan that totally cleans the smaller city’s landfill, and only partly cleans the larger city’s, prevents 6 in one wording and 5 in the other. The paper says half the subjects saw each wording. It does not print the size of either half. In the wording where every plan prevents 6 cases, zero-risk bias is defined as ranking the complete cleanup of the smaller city as the best of the three. Eighteen percent of the subjects who saw that wording did so. In the wording where that plan prevents 5 and the other plans prevent 6, the bias is defined as not ranking the complete-cleanup plan worst. It should be worst if the score is cases prevented. Forty-two percent of the subjects who saw that wording did not rank it worst. Eleven percent ranked it best. The bias was related to group in both versions (chi-square greater than 20, p < .01). Legislators, judges, and environmentalists showed it most often. Experts and economists showed it least. The paper does not print the percentage inside each group. The lesson is a ranking of written cities, scored against the number of cancer cases the plans prevent. The boundary is the mailing. The authors write that a questionnaire of this kind cannot show the basis of an answer, and that conclusions about biases remain tentative unless more careful laboratory studies support them. Eighteen percent and 42 percent are shares of the people who received that wording. They are not shares of residents, and they are not shares of cleanups that were carried out.Boundaries and Failure Modes
Zero-Risk Bias is the wrong name when a stated bottle price is retold as a purchase, when “not worst” is retold as “ranked best,” and when a sure amount of money is given this page’s name. The first boundary is the form of the answer. The Greensboro figures are extra dollars a person said they would pay for one bottle, after an interviewer stated a risk of 15 injuries per 10,000 bottles. The 8.09 for eliminating both insecticide risks in the group with a child under five, and that mean has a standard error of $1.13. The same paper reports what it calls reference risk effects: increases in risk were valued much more than decreases. This page does not carry the dollar amounts for those increases. Loss Aversion is a greater weight on a loss than on a matched gain. The reference-risk result is a neighbor of that weight. It is not the step that reaches zero. The second boundary is the score on the cities. Forty-two percent is the share who did not place the 5-case plan last. Eleven percent placed it first. Eighteen percent placed the equal-saving zero plan first. The paper does not print how many people saw each half. It says the samples were not representative, and it says the bias conclusion from this mailing is tentative. A later laboratory study is not on this page. The in-press Ritov, Baron, and Hershey paper is cited and not re-read. The third boundary is the export. The authors’ own reading of the bottle prices is that a large certainty premium attaches to eliminating a particular risk, and that eliminating both risks does not add a further premium beyond the sum. They say this pattern does not support anxiety about any uncertainty at all as the driver. A rule that spends the budget to hit zero on one site, because zero feels finished, is the misuse. The check is the cases the other plan prevents, and the price of an equal cut that stops short of zero.Common Misconceptions
Three retellings travel with the zero. The last step is treated as several times the earlier step in every group. The mail rank is treated as if most people put the worse plan first. The money gamble and the missing act are given this page’s name.The last step always costs several times as much
The last step always costs several times as much
Table 4 is four columns, not one ratio. In the no-child insecticide column the printed steps are 0.34, and 0.99, the same as the first step, and the middle step is 5.71 belongs to insecticide users with a child under five. It is not a multiplier for every product, and it is not a receipt.
Most people ranked the worse cleanup best
Most people ranked the worse cleanup best
In the wording where complete cleanup prevents 5 cases and the other plans prevent 6, 42 percent did not rank the complete-cleanup plan worst. Eleven percent ranked it best. In the wording where every plan prevents 6, 18 percent ranked the complete cleanup best. Those are the people who saw that wording. The authors say the samples were not representative of the groups that were mailed.
Zero-Risk Bias is the certainty effect, or omission bias
Zero-Risk Bias is the certainty effect, or omission bias
The certainty effect, in Kahneman and Tversky’s 1979 abstract, is underweighting a merely probable money outcome relative to a certain one. The certainty premium in the 1987 interviews is extra stated dollars for an equal cut in injury risk that reaches zero. Omission Bias rates harm from an act more harshly than the same harm from not acting. Ambiguity Aversion prefers a known chance to an unknown chance with the same prize. Status Quo Bias prefers the state already in place. A plan that clears one site is often a change, not the state already occupied.
Related Concepts
Zero-Risk Bias sits beside a sure sum of money, beside a harsher rating of an act, beside a heavier weight on a loss, beside a known chance, and beside the state already in place.Prospect Theory
A model of choices under risk. Its certainty effect underweights a merely probable money outcome relative to a certain one. The 1987 bottle prices are a different measurement.
Omission Bias
A harsher rating of harm from an act than of the same harm from not acting. A vaccine risk of zero can look similar. The comparison here is cases prevented, or dollars for an equal cut.
Loss Aversion
A greater weight on a loss than on a matched gain. The same 1987 interviews report a stronger valuation of risk increases than of decreases. Those dollar amounts are not on this page.
Ambiguity Aversion
A preference for a known chance over an unknown chance with the same prize. The risks on this page were stated as counts per 10,000 bottles, or as cancer cases a year.
Status Quo Bias
A preference for the state already in place. Ranking a complete cleanup first is a preference for one change over other changes, scored by cases prevented.