20% of early retirees regret their decision to stop working, according to a MedicareFAQ survey of 569 retired Americans (“74% say they were able to retire early, but of those who did, 20% regret that decision”). The mirror side — regret about working too long — is much rarer, but no US survey measures it directly, so this pair is proxy- graded: a UK Which? survey (May 2025, n=1,188) found roughly 5% felt they had retired too late, “half as many” as the ~10% who felt they had stopped work too early. Even at that rough level, action (retiring early) generates more regret than inaction (continuing to work), reversing the typical Gilovich pattern. The reason on the action side is almost entirely financial — 86% of the MedicareFAQ respondents wished they had saved more, and 75% of early retirees in a separate Manulife study said they regretted not saving enough.
Hurwitz and Mitchell (NBER Working Paper 30696, 2022) sharpen the direction of retirement-timing regret with a controlled experiment of 1,764 Americans aged 50 and older. In their sample, 37% regretted not working longer and 57% regretted not saving more — both point the same way as the MedicareFAQ finding: the dominant timing regret is retiring too early, not too late. (This 37% is NBER’s figure; the MedicareFAQ article itself does not report a “worked longer” number.) Börsch-Supan et al. (2023) reinforce the underlying cause: 58.5% of those aged 60-79 wished they had saved more. Working longer is often a proxy for the saving people failed to do, which is why “I should have worked longer” and “I should have saved more” track together.
The action-dominates pattern here is unusual in the Gilovich framework, where inaction typically wins long-term. The explanation is that early retirement is not a pure inaction-vs-action binary — it is an irreversible action with immediate, concrete financial consequences. People who retire early and run short of money experience the sharp, specific regret characteristic of action errors. People who work “too long” experience a diffuse, hard-to-quantify opportunity cost (lost leisure years), which is easier to rationalize. The low working-too-long rate may also reflect survivorship bias: those who worked until health failed may not be surveyed.