Start hereA 2022 meta-analysis found that nudges had a small-to-medium effect. Re-analyses of the same data found it shrinking towards zero once publication bias was taken into account. So how much do nudges change?
About the book
Nudge: Improving Decisions about Health, Wealth, and Happiness is by two authors. Richard H. Thaler received the 2017 Nobel memorial prize in economics for his contributions to behavioural economics. Cass R. Sunstein is a Harvard professor who ran the White House Office of Information and Regulatory Affairs from 2009 to 2012.
Yale University Press published it in 2008, and a revised Final Edition followed in 2021. It is written for general readers, and the authors stress that its ideas apply to firms as well as governments.
The Final Edition adds chapters on smart disclosure and sludge and rewrites the one on organ donation. But, in the authors’ own words, it makes no attempt to bring readers up to date on the nudge research of recent years.
Humans, not Econs
The authors contrast real people, whom they call Humans, with Econs: idealised choosers with complete information, unlimited cognitive abilities and complete self-control.
Humans rely on a fast, automatic way of thinking, the gut reaction, as well as a slower, reflective one, conscious thought. They make predictable mistakes as a result. Those predictable mistakes are why, in the authors’ view, a gentle push can help.
There is no neutral design
Whoever arranges a set of choices, whether a canteen, a form or a pension plan, is what they call a choice architect. In their view no choice is ever presented in a neutral way, so the question is how to influence people, not whether.
What a nudge is
In the book’s definition, a nudge is any aspect of the choice architecture that changes behaviour in a predictable way. It does so without forbidding any options or significantly changing people’s economic incentives. Putting fruit at eye level counts; banning junk food does not.
Libertarian paternalism
This is the authors’ name for their stance. It means steering people towards choices that make their lives better, as judged by themselves, while leaving them free to choose otherwise.
They describe it as a relatively weak, soft and nonintrusive kind of paternalism. In the Final Edition preface they admit the term is one “which only its authors love”.
Defaults are powerful
Because people tend to stick with whatever happens if they do nothing, the default option has a large effect. Their showcase is retirement saving: automatic enrolment, and their own Save More Tomorrow plan, in which workers commit now to saving part of future pay rises.
Did automatic enrolment raise pension sign-ups?
Yes, where it was studied. A large US company switched new employees to automatic enrolment in its 401(k) plan, with the right to opt out. Participation was then significantly higher than when employees had to sign up.
Many of those enrolled automatically also kept both the default contribution rate and the default fund. The authors put that down to inertia, and to employees taking the default as advice from the company.
Did Save More Tomorrow raise saving rates?
At its first employer, yes. In the plan, workers commit in advance to putting part of future pay rises into retirement savings.
- At the first company to offer it, 78% of those offered the plan joined, and 80% of them stayed in through the fourth pay rise. Their average saving rate rose from 3.5% to 13.6% over 40 months.
- A Chicago Booth Review article gives 11.6% after 28 months, an earlier point in the same plan.
This is one firm, not a general rate. A later field experiment co-written by one of the plan’s designers tested simply offering a delayed start. It did not get more people to agree to save more, and it lowered savings over the following months. When the later start was tied to a fresh-start date such as a birthday, more people chose it and contributions rose over the next eight months.
Are defaults the strongest nudges?
Strong on average, uneven in practice. A 2019 meta-analysis of 58 default studies found a considerable average effect (d = 0.68). But several studies found no significant effect, and two found the opposite of the intended one. In the 2022 meta-analysis of nudges, defaults had the largest average effect of any technique before adjusting for publication bias (d = 0.62).
After adjustment, the two re-analyses read defaults differently. One found the evidence for interventions that change how choices are structured, defaults among them, undecided. The other named defaults as an exception to its conclusion that large, consistent nudge effects should not be expected.
Try it: spot the pre-ticked box or the pre-selected plan, and ask who chose it and why. Defaults may steer choices partly because people read them as advice; that is how the pension study’s authors explained why so many employees kept the default.
Sludge and the limits of nudging
The Final Edition adds the opposite of a nudge: sludge, the paperwork and friction that make wise choices harder. The authors think every organisation should seek it out and remove it.
It also stresses the limits of the approach. On climate change, for example, they write that the problem cannot be solved with nudges alone.
Try it: for anything you want other people to do, from filling in a form to signing up for an event, count the steps. Removing the steps that are not needed is the Final Edition’s own recommendation.
What holds up, and what the book overstates
What the sources show for each idea
The core observation holds: how a choice is set up changes what people choose, and defaults in particular can have large effects. What is now argued over is how big most nudges are. Published studies overstate them, and after correcting for that, estimates of the average effect range from small to close to zero.
How big are nudge effects on average?
Researchers disagree, and the disagreement is about publication bias: studies with strong results are more likely to be published.
- A 2022 meta-analysis of more than 200 studies found an average effect of d = 0.43, which its authors called small to medium, while noting a moderate publication bias. Their own sensitivity analysis suggested the true effect could be as low as d = 0.08 if the bias were severe.
- Re-analysing the same data, one team found no evidence for an effect once publication bias was corrected for (an adjusted estimate of 0.04). Another found adjusted estimates between −0.01 and 0.08, with large variation between studies.
- A 2025 review of 14 meta-analyses found an average of 0.27 that fell to 0.004 after adjustment, while rating most of those meta-analyses low or critically low in quality.
Our reading leans towards real but usually small effects. The clearest evidence is a study of 126 trials run at scale by government nudge units, covering 23 million people, which found a smaller but still highly significant average effect.
Published studies versus nudges at scale
A 2022 study compared trials from academic journals with all the trials run by two large US nudge units. The average effect was 8.7 percentage points (a 33.4% increase) in the journals, against 1.4 percentage points (an 8.0% increase) in the nudge units. It estimated that selective publication, made worse by small samples, explains about 70% of the difference.
Try it: treat a published effect as an upper estimate. When you change a form or a process, measure the result rather than expecting the published number.
Does the book back opt-out organ donation?
The authors say the opposite. In the preface to the 2021 Final Edition, they write that they rewrote the organ donation chapter because readers thought they supported a policy they oppose. They also write that they do not support presumed consent.
The 2009 edition of the book already recommended mandated choice for organ donation, saying it hoped more states would follow Illinois’s lead. In 2009 Thaler argued for mandated choice, where people must state their wishes. He wrote that it may achieve a higher rate of donations than presumed consent while avoiding the objections to it.
Does opt-out raise organ donations?
It changes registrations a lot; its effect on donations is contested. Opt-out sharply raises how many people count as donors: Thaler cites consent rates of 12% in Germany (opt-in) against 99% in Austria (opt-out). Whether it raises actual donations is another matter.
- A 2014 study of 48 countries from 2000 to 2012 found more deceased donors and more kidney transplants in opt-out countries, but fewer living donors.
- A 2019 comparison of 35 OECD countries in 2016 found no significant difference in deceased donors and fewer living donors in opt-out countries.
- A 2024 study that instead followed five countries before and after they switched (Argentina, Chile, Sweden, Uruguay and Wales) found no increase in donation rates on average.
- A 2024 evaluation of England’s switch reported that consent rates in England and Wales had not yet increased.
- A 2026 review of five years of opt-out found that in England, family consent fell from 69% to 57% between 2020 and 2025. Donor numbers there stayed below pre-pandemic levels. In the Netherlands, family consent and donor numbers rose modestly.
The first two compare countries that differ in other ways. The 2014 study used a statistical method to argue that the effect is causal. The 2019 authors concluded that other barriers to donation must be addressed even where consent is presumed.
Our reading leans towards opt-out doing little for actual donations on its own. The before-and-after comparison avoids the differences between countries that the 2014 and 2019 studies cannot rule out, and five years on, England and the Netherlands went in different directions.
Did the urinal fly cut spillage by 80%?
The fly is real; the figure has no published study behind it. The book says the staff of Aad Kieboom, who introduced the fly at Amsterdam’s Schiphol airport, ran trials and found that the etchings reduce spillage by 80 percent.
A 2013 investigation found that no study of the fly’s effect appears to have taken place. Kieboom said he was never aware of any scientific research and called the 80% figure “very empirical”. The inventor of the waterless urinal, interviewed for the same piece, guessed the reduction was probably more like 50%.
Who it’s for, and who can skip it
It suits anyone who designs forms, products, policies or workplaces, and readers curious about how small design choices shape behaviour.
Readers looking for up-to-date evidence on how well nudges work will need to look beyond it. The authors say the Final Edition does not try to cover recent nudge research, and that is where the debate about effect sizes lives. Readers who know Thinking, Fast and Slow will find the psychology familiar and the policy new.
Quick answers
Do nudges work?
Researchers disagree about how much. Our reading leans towards real but usually small effects, and published studies overstate them.
Does opt-out organ donation increase donations?
It raises registrations, but its effect on actual donations is disputed. A 2024 study of five countries before and after they switched found no increase on average.
What is sludge?
The Final Edition's word for the paperwork and friction that make wise choices harder.
What is libertarian paternalism?
The authors' name for steering people towards choices that make their lives better, as judged by themselves, while leaving them free to choose otherwise.
Did Save More Tomorrow work?
At its first employer, participants' average saving rate rose from 3.5% to 13.6% over 40 months. That is one firm, not a general rate.
Use this
Look for the defaults around you, and test changes rather than trusting published effects.
If I do nothing here, what happens, and who decided that?
Quick quiz
Nothing is scored or saved. Recalling it once now is what makes it last.
1. What do the authors say about presumed-consent organ donation?
Answer: They do not support it.Read it again ↑
2. In the 2022 comparison, how big was the average effect in nudge-unit trials?
Answer: 1.4 percentage points.Read it again ↑
3. Where does the urinal fly's 80% figure come from?
Answer: No published study we know of.Read it again ↑
How we checked this · Last checked October 2026
- The book's wording comes from short passages of the 2008 and 2009 printings, found through the Internet Archive's search-inside tool, and from the Final Edition preface as excerpted in Behavioral Scientist.
- We read the full texts of the 2022 PNAS meta-analysis, its correction, the three replies and the authors' answer, the 2014 organ donation study, the Works That Work article, and Thaler's 2009 column in an archived copy.
- For the other studies, including a 2026 review of opt-out in England and the Netherlands, we read the abstracts.
- In October 2026 we searched again for newer reviews of nudge effect sizes and found none.
Sources
- R. H. Thaler and C. R. Sunstein, Nudge: Improving Decisions about Health, Wealth, and Happiness (Yale University Press, 2008; Penguin, 2009): short passages via the Internet Archive’s search-inside tool (2008, 2009).Primary
- R. H. Thaler and C. R. Sunstein, “Nudge: preface to the Final Edition”, Behavioral Scientist, 2 August 2021; Penguin, Nudge (updated edition).Primary
- The Royal Swedish Academy of Sciences, “The Prize in Economic Sciences 2017”, press release, 9 October 2017.Primary
- Harvard Law School, “Cass R. Sunstein”.Primary
- B. C. Madrian and D. F. Shea, “The power of suggestion: inertia in 401(k) participation and savings behavior”, NBER Working Paper 7682 (2000), Quarterly Journal of Economics 116 (2001).Scholarly
- R. H. Thaler and S. Benartzi, “Save More Tomorrow: using behavioral economics to increase employee saving”, Journal of Political Economy 112 (2004).Scholarly
- “Save More Tomorrow”, Chicago Booth Review (2004).Popular
- J. Beshears, H. Dai, K. L. Milkman and S. Benartzi, “Save more later? The effect of the option to choose delayed savings rate increases on retirement wealth”, NBER Retirement Research Center paper (2015), published in part as “Using fresh starts to nudge increased retirement savings”, Organizational Behavior and Human Decision Processes 167 (2021).Scholarly
- J. M. Jachimowicz, S. Duncan, E. U. Weber and E. J. Johnson, “When and why defaults influence decisions: a meta-analysis of default effects”, Behavioural Public Policy 3 (2019).Scholarly
- S. Mertens, M. Herberz, U. J. J. Hahnel and T. Brosch, “The effectiveness of nudging”, PNAS 119 (2022), with its correction and the authors’ reply.Scholarly
- M. Maier et al., “No evidence for nudging after adjusting for publication bias”, PNAS 119 (2022).Scholarly
- B. Szaszi et al., “No reason to expect large and consistent effects of nudge interventions”, PNAS 119 (2022).Scholarly
- J. Z. Bakdash and L. R. Marusich, “Left-truncated effects and overestimated meta-analytic means”, PNAS 119 (2022).Scholarly
- B. Hu et al., “Assessing nudge impact: a comprehensive second-order meta-analysis”, Journal of Behavioral Decision Making 38 (2025).Scholarly
- S. DellaVigna and E. Linos, “RCTs to scale: comprehensive evidence from two nudge units”, Econometrica 90 (2022).Scholarly
- R. H. Thaler, “Opting in vs. opting out”, The New York Times, 26 September 2009 (archived).Primary
- L. Shepherd, R. E. O’Carroll and E. Ferguson, “An international comparison of deceased and living organ donation/transplant rates in opt-in and opt-out systems”, BMC Medicine 12 (2014).Scholarly
- A. Arshad, B. Anderson and A. Sharif, “Comparison of organ donation and transplantation rates between opt-out and opt-in systems”, Kidney International 95 (2019), and its 2023 corrigendum.Scholarly
- M. Dallacker, L. Appelius, A. M. Brandmaier, A. S. Morais and R. Hertwig, “Opt-out defaults do not increase organ donation rates”, Public Health 236 (2024).Scholarly
- L. McLaughlin and N. Mays, “What does the evaluation of the Organ Donation (Deemed Consent) Act 2019 in England tell us about the effectiveness of deemed consent systems for deceased organ donation?”, Transplantation 109 (2025; online 23 October 2024).Scholarly
- N. E. Jansen et al., “Five years after introduction of an opt-out consent system for organ donation; lessons learned from England and the Netherlands”, Transplant International (2026).Scholarly
- B. Evans-Pritchard, “Aiming to reduce cleaning costs”, Works That Work, No. 1 (Winter 2013).Popular