Omission Bias

Comparing action and inaction

In August 1997, shortly after Amazon made its public stock offering, Chris put his meager savings together to purchase $2,500 of Amazon stock. The following summer, in 1998, Chris’s Amazon stock is worth slightly more than $20,000, a wonderful outcome but a risky holding. At the same time, Pat, who has followed the amazing progression of Amazon’s stock price over the last year, is considering buying $20,000 worth of its stock for the first time. They each ponder this decision and decide that at that moment Amazon stock is too risky to own. (In fact, Amazon did not have a profitable year until 2003.) Chris sells his total Amazon holding. Pat decides not to buy Amazon and buys less speculative stocks instead.

Anyone owning $20,000 of Amazon stock in August 1998 would have over $3 million in May 2021, a much higher return than for almost any other stock investment during this period. The example calls Chris’s missed gain an error of commission: he sold Amazon stock in 1998. It calls Pat’s missed gain an error of omission: he did not buy Amazon stock in 1998. Both would have earned more by owning Amazon stock in 1998 and holding it for many years. That hindsight comparison does not, by itself, show that either decision was unreasonable given the information available then.

The two decisions are equivalent in their material consequences: both Pat and Chris invested $20,000 in stocks that were not Amazon stocks in 1998, and missed out on the monumental gain over the next two decades. But the two individuals perceived their decisions differently. Chris, who sold his stocks, kicks himself regularly for his stupidity. Pat, who didn’t buy the stocks, rarely thinks about it. If you are like most people, you would feel much more regret had you been Chris than had you been Pat.

Omission bias gives greater weight to harm caused by acting than to comparable harm caused by not acting. Richard argues that the comparison should account for both. The consequences, probabilities and responsibilities matter; the fact that one option involves action is not enough to settle the choice.

This bias shows up far beyond financial decisions. In domestic arguments, a hurtful remark (an error of commission) often draws far more blame than a failure to speak up at all (an omission). At work, a misstep is penalized more severely than a missed opportunity. In national policy debates, lawmakers who change a policy are held more accountable for any negative outcomes than those who simply maintain a status quo that is already causing harm.

One potential reason behind omission bias is that we often do not know about the losses that arise from errors of omission. For example, if you choose not to take the job in San Francisco, you will never know how it would have turned out. You are less likely to experience regret. Whereas if you had decided to go and things had turned out poorly (an error of commission), the losses to you are much easier to see. There is a rich literature beyond the scope of this book on why this omission bias exists. Indeed, some work suggests that actions, or errors of commission, generate more regret in the short term; but inactions, or errors of omission, produce more regret in the long run.

Yinglan Tan, a former student of Richard’s and founding partner of Singapore-based Insignia Venture Partners, describes this tendency to give more importance to errors of commission in his own work as a venture capitalist: “A deal gone wrong has far more lasting effects on the fund than missing a deal on a company that turned out to be a billion dollars.”

Richard argues that government responses to COVID-19 placed too much weight on the risks of acting and too little on the risks of delay. In his vaccine-distribution example, waiting for more confidence about safety and effectiveness also carried a cost in preventable COVID infections and deaths. The comparison he urges includes both kinds of risk; the passage does not estimate the effects of particular approval or distribution decisions.

Richard and Sally faced this asymmetry when considering Sally’s bone marrow transplant. A death caused by treatment could have felt different from a death after declining it. Their discussion with each other and their doctors focused on the consequences of both choices, rather than allowing the action/inaction distinction to determine the decision.

Jeff Frankel, Richard’s colleague at the Harvard Kennedy School and a member of President Bill Clinton’s Council of Economic Advisors, applies the comparison to environmental policy:

Errors of omission vs. commission in environmental policy Jeff Frankel Richard’s maxims are useful corrections to common mistakes in decision analysis that are hidden in what is widely taken to be unassailable common sense. What could be more universally accepted as wise words to live by, in the medical profession and beyond, than the apparent truism, “First, do no harm.” This principle, based on a line in the Hippocratic oath, seems to say, “errors of omission should be weighted zero, relative to errors of commission.” For example, if a doctor contemplates a surgical procedure for a dying patient that has a 90 percent chance of fixing the problem and a 10 percent chance of resulting in immediate death, a literal interpretation of “do no harm” would be to forego the surgery and let the patient die in the near future. Richard would of course say that the right answer from the probability-weighted average of outcomes is to go ahead with the surgery. By the way, I have no problem if someone wants to justify “do no harm” as a three-word summary of the idea that surgeons might sometimes be too quick to operate. It would serve as a recommendation to lean toward conservative treatment, thereby counteracting the temptation to “apply your hammer to everything that looks like a nail.” Public policy applications of Richard’s maxim “Errors of commission should be weighted the same as errors of omission” arise in environmental policy. Some, especially in Europe, have a fear of new and unfamiliar technologies in general. The claim that the burden of proof lies with the innovation, rather than symmetrically with the status quo, sometimes goes under the name of the “precautionary principle.” It helps explain the tendency to forget to compare the worst-case risks of the new technology with the known downsides of the old technologies. One example is genetically modified organisms (GMOs). This policy issue first came to my attention as a member of President Clinton’s Council of Economic Advisers. It is true that a fundamentally new technology tends to pose risks that are unknown. Let us say that there is some small probability of doing some harm (an error of commission). That is no excuse for neglecting to weigh in the balance the known risks of the existing technology (an error of omission). In the case of genetically modified crops, costs of doing without them include greater need for insecticides and possible food shortages in poor countries.

Healthcare reform offers a political illustration. In debate over changes such as those passed under President Obama, critics may emphasize problems under the new law while giving less attention to harms under the previous system. Assessing omission bias requires comparing those consequences; opposition to a particular reform is not, by itself, evidence of the bias.