Cost of action vs cost of inaction
Claim assessed: We judge harm caused by an action more harshly than the same harm caused by inaction.
Published , updated · how we work
In short
When you consider doing something, you usually weigh the risks of acting against a “safe” option: doing nothing. But inaction is not neutral. It has costs too; they are just harder to see. The model suggests putting both on the table: what it costs you to act, and what it costs you not to act. Research explains why this step is needed. We tend to judge harm caused by an action more harshly than the same harm caused by inaction. Yet when we look back on our lives, we more often regret what we didn’t do.
What it says
Harm from action weighs more. Ritov and Baron (1990) gave students a hypothetical scenario. A disease kills 10 in 10,000 children. A vaccine prevents it completely, but can itself kill some number of children through side effects. Participants had to say what the maximum acceptable risk from the vaccine should be for the government to make it compulsory.
Any risk below 10 in 10,000 would have meant fewer deaths than without the vaccine. Yet only 9% of participants accepted a risk of 9 in 10,000 or more. 23% said no risk should be tolerated, and 57% chose a threshold between 1 and 8 in 10,000. Many explained that vaccinating would mean causing the deaths of some children, which they considered wrong even if it saved more.
The authors call the phenomenon omission bias: preferring harm caused by omission to equal or lesser harm caused by an action. They describe it as an overgeneralization of the distinction between doing and not doing, applied to a case in which the distinction is irrelevant.
How solid the effect is. Yeung, Yay and Feldman (2022) ran a preregistered meta-analysis, meaning the method was announced in advance: 21 samples from 13 articles, with 49 effects. The average effect was g = 0.45, with a wide interval of 0.14–0.77. The g index shows how large the difference is; by Cohen’s conventions, 0.5 means a medium effect. The result depends on what is measured, though. The effect is clear for moral judgement (g = 0.71): harm done through an action seems more wrong than the same harm done through inaction. It is weaker for how much someone is to blame (g = 0.32). For actual decisions, as in the vaccine scenario, it did not come out reliably (g = 0.30, with an interval that includes zero), but there were only 4 studies there. Corrections for unpublished studies gave mixed results: some keep the effect, others shrink it to almost zero.
In the short term we regret actions, in the long term inactions. Kahneman and Tversky (1982) had shown that, after a bad outcome, people think the one who acted regrets it more. Gilovich and Medvec (1994) showed that the picture changes over time. The figures below are from their study, as presented in the replication by Yeung and Feldman (2022):
- A lifetime’s regrets. 60 adults in Syracuse, in the US, chosen at random from the phone book, were asked what they regret more when looking back: the things they did or the things they didn’t do. 45 of them, or 75%, chose the things they didn’t do.
- Two students, two choices. In one scenario, one student transfers to another university and another stays. Both come to regret their decision. Of 80 participants, 76% said the one who transferred regrets it more right away, and 64% said the one who stayed regrets it more in the long run.
Gilovich and Medvec (1995) propose several explanations. People do more to repair mistakes made through action than those made through inaction, so regret after an action fades. Meanwhile, over time we become more and more sure we would have succeeded had we acted, because we forget the risks and uncertainty of the moment.
Example
An example built for this text: for a year you’ve been meaning to ask for a raise, but you keep putting it off. The cost of acting is visible right away: an awkward conversation and the risk of a refusal. The cost of not acting shows up nowhere at the moment. It is the pay gap, month after month, plus the fact that no one finds out you wanted it. If you only write down the first column, doing nothing always looks like the safe option.
How to apply it
The steps below are a practical approach we propose. The studies cited did not test them.
- Write inaction down as a separate option. Instead of “do I do X or not?”, write two columns: “if I do X” and “if I do nothing”. For each, note what happens in a month and in a few years.
- Compare like with like. Omission bias means that harm caused by an action seems worse than the same harm caused by inaction (Ritov and Baron, 1990). Ask yourself: if the outcome were the same, would I judge it the same whether it came from what I did or from what I didn’t do?
- Think about the regret of years from now too. In the short term, a mistaken action hurts more. In the long term, the people in Gilovich and Medvec’s study more often said they regret what they hadn’t done. Ask what you will regret in ten years, not just tomorrow.
- Don’t turn this into “always act”. Sometimes doing nothing really is the better choice. The aim is to compare fairly, not to favour action.
Limits and nuances
- The scenarios are hypothetical. Ritov and Baron’s study used an imaginary scenario, with 53 students in the first experiment. What people say in a questionnaire may differ from what they do when the decision is real.
- The regret pattern replicated, but more weakly. Yeung and Feldman (2022) repeated four of the 1994 studies with 988 online participants from the US. They confirmed the results in three of them, but with weaker effects: on the lifetime-regrets question, 63% chose the things they hadn’t done, compared with 75% in the original. In the fourth, they did not find the pattern. The authors discuss several possible explanations, including changes in how people understand “action” and “inaction”.
- The original studies were small. There were 60 people on the lifetime-regrets question and 80 in the two-student scenario, all from the US.
- Omission bias is reliable mainly for judgements. The 2022 meta-analysis finds it clearly when people judge how wrong something is, but not yet for actual decisions, where there were few studies.
- The difference between doing and not doing is not always irrelevant. Ritov and Baron speak of an overgeneralization. In many moral and legal situations, the distinction rightly matters. The problem arises when you apply it where outcomes are what count.
- What we read. We read in full Experiment 1 of Ritov and Baron’s article (1990), from Jonathan Baron’s website. We read the replication by Yeung and Feldman (2022), in its final accepted version on OSF, including the tables with the results of the original 1994 study. Of the meta-analysis by Yeung, Yay and Feldman (2022), we read the final preprint on OSF: the abstract, the main results and the publication-bias analysis. The articles by Gilovich and Medvec (1994, 1995) have no free copies: their figures come from the replication, and the explanations from the replication’s summary.
- Revised on 7 October 2026. We replaced the figures from the 1995 article, which we could not read, with those from the original 1994 study, as reported by the replication. We added the omission-bias meta-analysis, the replication’s results in the diagram and this list.
Sources
- Ilana Ritov, Jonathan Baron (1990). Reluctance to vaccinate: Omission bias and ambiguity
- Thomas Gilovich, Victoria Husted Medvec (1994). The temporal pattern to the experience of regret
- Thomas Gilovich, Victoria Husted Medvec (1995). The experience of regret: What, when, and why
- Siu Kit Yeung, Gilad Feldman (2022). Revisiting the Temporal Pattern of Regret in Action Versus Inaction: Replication of Gilovich and Medvec (1994) With Extensions Examining Responsibility
- Siu Kit Yeung, Tijen Yay, Gilad Feldman (2022). Action and Inaction in Moral Judgments and Decisions: Meta-Analysis of Omission Bias Omission-Commission Asymmetries
See also: Endowment effect and status quo bias, First-order negative, second-order positive, Opportunity cost, Sunk cost