None Of The Economics Changed

After several articles in this series reporting that a famous finding did not hold up, here is one that did, in administrative data, at scale, with a natural experiment behind it. The complication is not whether it works. It is what you have taken responsibility for once it does.

A Note On Scope

Stated first.

This article describes research on choice architecture. It is not investment, pension, tax or legal advice.

The arithmetic later in this article is illustrative, uses figures we invented, and is included to show the magnitude of what a default setting determines. It is not a projection and should not be used for planning.

Anyone setting up a retirement or benefit arrangement should take proper advice, and in Canada the applicable rules differ by plan type and province.

Our Grades For These Claims

Applying the scheme from the first article in this series.

Grade A for the default effect on participation. Administrative records from a real company, a clean before-and-after design in which the economics were held constant, and a finding replicated across many domains since.

Grade A for defaults persisting into the contribution rate and allocation, from the same paper.

Grade B for the mechanism, being inertia plus perceived advice, which the authors describe as what the behaviour appears to result from.

Grade C for the organ donation case, for a reason that has its own section and that is not usually mentioned.

A Note On Method

Everything here is verified to August 2026.

We obtained the published abstract in full from a bibliographic record, and the working paper abstract from the issuing bureau[1][2]. We did not obtain the full paper, and the abstract reports no participation percentages, so neither do we in the authors' own voice.

Specific percentages reach us through a United States patent document quoting the study, which is an unusual source and one we flag prominently at each use[3].

We did not obtain the organ donation paper, nor the study testing whether presumed consent affects actual donation rates, and report the second only by citation.

All arithmetic is ours, illustrative, and uses invented figures.

Closing A Gap From Article Nine

A piece of housekeeping, and the reason this article exists.

The ninth article in this series covered a public dispute over whether nudging works, in which a meta-analysis of 447 experiments and a bias-corrected reanalysis reached opposite conclusions. That article contained a section titled A Note On Defaults Specifically, which recorded that we could not determine whether defaults survived the correction, because we did not obtain the per-category results.

Two observations, ours.

This article does not resolve that. We still do not know how defaults fared in that particular reanalysis, and nothing below should be read as answering it.

What it does instead is go to the primary evidence on defaults directly, which is a different and in some ways better route. The strongest evidence here is not a laboratory experiment pooled into a meta-analysis. It is a company's payroll records before and after a policy change.

The Study

The design, which is unusually clean.

Madrian and Shea published The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior in the Quarterly Journal of Economics, 116(4), 1149–1187, in 2001, following a working paper in 2000[2].

In their words: "we analyze the 401(k) savings behavior of employees in a large U.S. corporation before and after an interesting change in the company 401(k) plan. Before the plan change, employees were required to affirmatively elect participation in the 401(k) plan. After the plan change, employees were automatically and immediately enrolled in the 401(k) plan unless they made a negative election to opt out of the plan."[1]

Then the sentence that makes the study valuable: "Although none of the economic features of the plan changed, this switch to automatic enrollment dramatically changed the savings behavior of employees."[1]

Three observations, ours.

The economics were held constant. Same match, same funds, same limits, same tax treatment. The only thing that moved was which way the form was already filled in.

The outcome is administrative payroll data, not a survey and not a laboratory task. People's actual money, over years.

And it is a before-and-after within one firm, which controls for a great deal that a comparison between companies would not.

The First Finding

The one everybody quotes.

"First, 401(k) participation is significantly higher under automatic enrollment."[1]

Two observations, ours.

The abstract gives no percentage, and neither do we in the authors' voice. A patent document quoting the study describes participation as much higher under automatic enrollment than when an affirmative election is required[3], which is qualitative and consistent.

The mechanism proposed in that same document is worth recording: it suggests the best explanation for low participation without automatic enrolment is that the decision to participate is complicated, which leads employees to defer making a decision and to procrastinate, and that automatic enrolment decreases the perceived complexity by decoupling the participation decision from the savings and investment decision[3]. That connects directly to the twenty-fifth article in this series.

The Second Finding

The one that matters more, quoted in full.

"Second, a substantial fraction of 401(k) participants hired under automatic enrollment retain both the default contribution rate and fund allocation even though few employees hired before automatic enrollment picked this particular outcome."[2]

Read the second clause carefully, because it is the whole point. Few people choosing freely picked that combination. Under automatic enrolment, a substantial fraction ended up there.

A patent document quoting the study gives figures: an average of 75 percent of participants hired under automatic enrollment contributed at the default rate of 3 percent, 80 percent invested in the default money market fund, and about 61 percent did nothing at all to change from the default[3].

These figures come from a United States patent, not from the paper, which is an unusual source for a research citation. We report them because they are specific and attributed, and we flag the source clearly rather than presenting them as the authors' own.

Why The Second Matters More

Our own analysis of why this reframes the whole intervention.

The first finding is about getting people through a door. The second is about what happens to them after they walk through it.

Three consequences.

The default is not a nudge toward a decision. On this evidence it frequently becomes the decision, and stays that way for years.

Which means whoever sets it has not merely encouraged participation. They have selected a savings rate and an asset allocation on behalf of most participants, and the abstract tells us the combination they selected is one few people would have chosen for themselves.

And that transfers responsibility. An opt-in plan leaves the outcome with the employee. A default plan moves a substantial part of it to whoever typed the number into the document, which is a real obligation and is usually not experienced as one.

Two Mechanisms, Named By The Authors

Their own explanation, which is more interesting than the usual one.

"This 'default' behavior appears to result from participant inertia and from employee perceptions of the default as investment advice."[2]

Two mechanisms, and they are not the same.

Inertia is the familiar one: changing anything takes effort, and the effort is never quite worth it today.

Perceived advice is different in kind. It says people are not merely failing to act. They are drawing an inference: somebody chose this number, that somebody presumably knew something, and departing from it would be second-guessing them.

Two observations, ours.

The word appears is the authors' hedge and we keep it. This is an interpretation of a behavioural pattern, not a separately demonstrated mechanism.

And the two have opposite implications for what to do. If it is inertia, reducing friction helps. If it is perceived advice, reducing friction does nothing, because the person is not stuck; they have concluded the default is the right answer.

The Default As Advice

The second mechanism has its own literature, and it deserves a section.

A reference list identifies McKenzie, Liersch and Finkelstein (2006), Recommendations implicit in policy defaults, in Psychological Science, 17, 414–420[4]. We did not obtain this paper and report only its title, which states its thesis.

Three consequences, ours.

If a default carries an implicit recommendation, then setting one is a communication, not just a procedure. You are saying something whether or not you meant to.

Which means a badly chosen default is worse than no default, because it does not merely fail to help. It actively advises.

And it means the standard defence, that people are free to change it, is weaker than it sounds. Someone who reads the default as advice is not being obstructed; they are being persuaded, and freedom to change is not much use against a recommendation you have accepted.

What The Number Silently Decides

The magnitude, with our own arithmetic. The figures below are invented and illustrative, not a projection, and not advice.

Take a salary of $60,000, held flat for simplicity, over 30 years, at a 5 percent annual return, with contributions at year end.

At a 3 percent default: $1,800 a year, reaching about $119,590.

At 5 percent: $3,000 a year, about $199,317.

At 6 percent: $3,600 a year, about $239,180.

At 10 percent: $6,000 a year, about $398,633.

Two observations.

The gap between a 3 percent default and a 6 percent one is about $119,590 on these assumptions, and on the study's second finding it is decided for most participants by whoever set the number.

And note that the direction is not automatically favourable. Someone who would have chosen 8 percent, auto-enrolled at 3 and left there by inertia or by reading it as advice, is saving less than they otherwise would have. The default helps the person who would have saved nothing and can hurt the person who would have saved more.

The Organ Donation Case

The other famous example, and where we part company with how it is usually told.

Johnson and Goldstein published Do defaults save lives? in Science, 302, in 2003[5]. We did not obtain it.

A commercial source describes the headline comparison: neighbouring countries with shared cultures and comparable healthcare systems producing a 60 or 70 percentage point difference in registered donors, and the variable that moved was the default[6].

Two observations, ours.

If accurate, that is an enormous difference, far larger than anything else in this series, and it is between countries that are otherwise similar.

But read the outcome variable. It is registered donors. That is a consent status, and the next section is about why the distinction matters.

Registered Is Not Donated

The complication, which is rarely mentioned in business writing about defaults. This section is our own.

A person being registered as a potential donor and an organ actually being transplanted are separated by a long chain: circumstances of death, medical suitability, hospital capacity, family consent in practice, and the transplant system's logistics.

The right question is therefore whether presumed consent changes donation rates, not registration rates, and there is a literature on exactly that.

Reference lists identify Abadie, A., and Gay, S., The Impact of Presumed Consent Legislation on Cadaveric Organ Donation: A Cross Country Study, in the Journal of Health Economics, 25, 599–620, with one source dating it 2004 and another 2006[7][8]. That literature also includes work on the Spanish model and on Belgian experience with presumed consent[7].

We did not obtain any of these papers and report no findings from them.

Two observations, ours.

We flag this because a 60 point difference in registration does not entail a 60 point difference in transplants, and the shorthand "defaults save lives" quietly assumes it does.

And because the same distinction generalises. In any business setting, a default changes what is recorded, and whether it changes the underlying outcome is a separate question requiring separate evidence.

Where Else It Has Been Found

The range, from a citing source.

Text citing this literature records defaults influencing decision making in retirement savings, organ donation, green energy consumption, HIV screenings, colorectal cancer screenings, and participation in tobacco cessation programs, with citations to studies across those domains[8].

We obtained none of these studies and report only that they are cited.

Two observations, ours.

That is a wide and varied set of domains, several of them medical and involving real administrative outcomes rather than stated intentions.

And it is the strongest argument for generality in this article. A finding that recurs in payroll systems, energy contracts and screening programmes is not a laboratory artefact.

The Third Option Nobody Mentions

The most useful thing in this article, and it is almost entirely absent from business writing on defaults.

The choice is usually presented as opt-in against opt-out. There is a third arrangement.

Active choice sets no default at all. It simply refuses to let the person proceed without answering the question.

A study abstract reports: "We find that compelling new hires to make active decisions about 401(k) enrollment raises the initial fraction that enroll by 28 percentage points relative to a standard opt-in enrollment procedure."[9]

Three observations, ours.

Twenty-eight percentage points is a substantial gain over opt-in, achieved without anyone deciding anything on the employee's behalf.

It avoids the second finding entirely. There is no default rate to be retained and no default fund to sit in, because there is no default.

And it is available to almost any business, in a way that redesigning a pension plan is not. Any form with a blank that must be filled in rather than a box already ticked is an active choice design.

And A Fourth

A further variant, with its authors' own caveat attached.

Keller and colleagues (2011), in the Journal of Consumer Psychology, describe testing an alternative, active choice policy in which there is no default, but decision makers are required to make a choice, and propose a modified version, that we call enhanced active choice, that favors one alternative by highlighting losses incumbent in the non-preferred alternative[10].

Their recommendation: enhanced active choice as a complement to automatic enrollment or when automatic enrollment is infeasible or unethical[10].

Two observations, ours.

The phrase "or unethical" is doing real work and it is the authors', not ours. They are conceding that there are situations where deciding on someone's behalf is not acceptable, which is a distinction the enthusiastic literature on defaults tends to skip.

And enhanced active choice is not neutral. Highlighting the losses in one option is a persuasive technique. It is more transparent than a pre-ticked box, but it is not the same as simply asking.

The Same Mechanism, Pointed The Other Way

The uncomfortable half, and it needs stating plainly.

A commercial source puts it well: "The same psychology that auto-enrolls people into saving also powers pre-checked upsells, free trials that silently convert, and privacy settings that share data unless you dig into a menu to stop them." And: "The mechanism is neutral; the designer's intent is not."[6]

Three observations, ours.

Every argument for auto-enrolment is structurally identical to an argument for a pre-ticked upsell. Both rely on the same inertia and the same perceived recommendation.

So the difference is not in the technique. It is in whether the default is the outcome the person would have chosen had they engaged, and that is an empirical question about them rather than a preference of yours.

And that gives a workable test, ours: would you be comfortable telling the customer, in the same sentence, that this is pre-selected and why? A retirement default survives that. A pre-ticked insurance add-on usually does not.

Setting Defaults In Your Own Firm

The application. Ours, and none of it tested.

Most businesses set defaults constantly without noticing: the payment terms on the invoice template, the service tier presented first, the renewal that continues unless cancelled, the option pre-selected on a form.

Three observations.

On this evidence, those settings are not neutral starting points. They are likely to become the outcome for a substantial share of people, and to be read as a recommendation by some of them.

Which means they deserve the attention a decision gets, rather than being inherited from whoever built the template. The default payment term on your invoice was probably chosen by nobody.

And where the interests of the two parties diverge, active choice is the honest instrument. It captures a large share of the participation benefit, on the finding above, without your firm deciding the answer.

What To Do

Treat every default as a decision you have made. On this evidence a substantial fraction of people will retain it, so setting it is choosing the outcome rather than proposing a starting point.

Attend to the second finding, not just the first. Participation is the easy part; the rate and the allocation are what the default silently fixes, and they were a combination few people chose freely.

Remember a default can lower an outcome. Someone who would have saved eight percent, enrolled at three and left there, is worse off than they would have been.

Consider active choice. Requiring an answer with no default is reported to raise enrolment by 28 percentage points over opt-in, and it decides nothing on the person's behalf.

Ask whether the default is advising. The authors name perceived advice as a mechanism, which means a badly chosen default does not merely fail to help; it recommends.

Distinguish what is recorded from what happens. A large difference in registered organ donors is not the same as a difference in transplants, and the same gap exists in any business metric a default touches.

Audit the defaults nobody chose. Invoice terms, renewal settings and pre-selected tiers were usually inherited rather than decided.

Apply the disclosure test. If you would not comfortably tell the customer in the same sentence that this is pre-selected and why, the mechanism is being pointed the wrong way.

The Limits Of This Analysis

Several caveats matter. This article describes research on choice architecture and is not investment, pension, tax or legal advice. Everything is verified to August 2026. We obtained the published abstract and the working paper abstract but not the full paper; the abstract reports no participation percentages and we state none in the authors' voice. The specific figures of 75, 80, 61 and 3 percent come from a United States patent document quoting the study, which is an unusual source; we flag it at each use and do not present those figures as the authors' own. We did not obtain the organ donation paper, and the 60 to 70 point registration difference comes from a commercial website. We did not obtain the presumed consent literature, including the cross-country study on actual donation rates, and report it only by citation with no findings; one source dates it 2004 and another 2006. We did not obtain the recommendations-implicit-in-defaults paper, the active choice paper, the enhanced active choice paper, or any of the studies in other domains, and report titles, quoted abstracts and citations only. All arithmetic is ours, uses invented figures, is illustrative rather than a projection, and is not advice. The analysis of why the second finding matters more, the registered-versus-actual distinction, the disclosure test and the section on defaults in a firm are our own reasoning, not findings. This article does not resolve the question left open in the ninth article of this series about how defaults fared in that particular bias-corrected reanalysis.

Related Reading

Behavioural Finance Same Data, Opposite Answers The article that left this gap open, and why the per-category question still is. Behavioural Finance Yes, Yes, No The other literature on deferring a decision, and what structure does about it. No. 28 Three Different Things Called One Name Two independent meta-analyses seven years apart both report d = 0.31. But framing effect names three separate phenomena, and one paper's title asserts...

Frequently Asked Questions

Do defaults work?

On this evidence, strongly. A company switched its retirement plan from opt-in to automatic and, in the authors' words, although none of the economic features changed, savings behaviour changed dramatically. The outcome is payroll data over years, not a survey.

What is the second finding?

That a substantial fraction of auto-enrolled participants retained both the default contribution rate and the default fund allocation, even though few employees hired before the change had picked that particular combination. The default did not just open a door; it became the outcome.

Can a default make someone worse off?

Yes. Someone who would have chosen a higher contribution rate, enrolled at a low default and left there, saves less than they otherwise would have. On our own illustrative arithmetic the gap between a 3 percent and a 6 percent default over thirty years is substantial, and it is decided by whoever set the number.

Why does the organ donation example need care?

Because the reported difference is in registered donors, which is a consent status. Whether it translates into actual transplants is a separate question with its own literature, which we did not obtain. A large difference in registration does not entail a large difference in donation.

Is there an option between opt-in and opt-out?

Yes, and it is the most useful thing here. Active choice sets no default and simply requires an answer. One study reports it raising enrolment by 28 percentage points over opt-in, while deciding nothing on anyone's behalf.

How do I know if I am using this fairly?

The same mechanism powers auto-enrolment and pre-ticked upsells, and the technique does not distinguish them. Our own test: would you be comfortable telling the customer in the same sentence that this is pre-selected and why? One of those two survives it.

References

1. Madrian, B. C., & Shea, D. F. (2000). The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior. National Bureau of Economic Research Working Paper No. 7682, working paper abstract, on the authors analyzing the 401(k) savings behavior of employees in a large U.S. corporation before and after a change in the company plan; on employees having been required to affirmatively elect participation before the change and being automatically and immediately enrolled unless they made a negative election to opt out after it; on none of the economic features of the plan having changed while the switch to automatic enrollment dramatically changed the savings behavior of employees; and on the first key finding that 401(k) participation is significantly higher under automatic enrollment. Note: the issuing bureau's own copy of the working paper abstract; we did not obtain the full paper, and the abstract states no participation percentages. nber.org
2. Bibliographic record for Madrian, B. C., & Shea, D. F. (2001), The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior, The Quarterly Journal of Economics, 116(4), 1149–1187, reproducing the published abstract, on 401(k) participation being significantly higher under automatic enrollment; on a substantial fraction of participants hired under automatic enrollment retaining both the default contribution rate and fund allocation even though few employees hired before automatic enrollment picked this particular outcome; and on this default behavior appearing to result from participant inertia and from employee perceptions of the default as investment advice. Note: a bibliographic record; access to the full text is restricted and we did not obtain it. ideas.repec.org
3. United States patent document citing Madrian and Shea (2000), on an average of seventy five percent of plan participants hired under automatic enrollment contributing at the default rate of three percent, eighty percent investing in the default money market fund, and about sixty one percent doing nothing to change their savings or investment behavior from the default; on plan participation being much higher under automatic enrollment than when an affirmative election is required; and on the suggested explanation that the participation decision is complicated, leading employees to defer and procrastinate, with automatic enrollment decreasing perceived complexity by decoupling the participation decision from the savings and investment decision. Note: a patent document, not a research source. These figures are quoted here as that document attributes them and are not presented as the authors' own; we could not verify them against the paper. image-ppubs.uspto.gov
4. Reference list in Keller and colleagues (2011), Journal of Consumer Psychology, identifying McKenzie, C. R. M., Liersch, M. J., & Finkelstein, S. R. (2006), Recommendations implicit in policy defaults, Psychological Science, 17, 414–420; and Carroll, Choi, Laibson, Madrian and Metrick (2009) on active choice. Note: citations only. We did not obtain the McKenzie paper and report only its title, which states its thesis. myscp.onlinelibrary.wiley.com
5. Reference lists identifying Johnson, E. J., & Goldstein, D. (2003), Do defaults save lives?, Science, 302, 1338–1339. Note: a citation only. We did not obtain this paper and report no findings from it in the authors' own words. business.columbia.edu
6. Commercial behavioural design website, on Madrian and Shea having tracked a large company that switched from requiring employees to opt into the 401(k) to enrolling them automatically, with participation jumping sharply and a large share of auto-enrolled employees sticking with both the default contribution rate and the default fund; on neighbouring countries with shared cultures and comparable healthcare systems producing a 60 or 70 percentage point difference in registered donors, with the default being the variable that moved; on a default rate set too low potentially leaving people saving too little; and on the same psychology that auto-enrolls people into saving also powering pre-checked upsells, free trials that silently convert, and privacy settings that share data unless the user intervenes, with the mechanism being neutral while the designer's intent is not. Note: a commercial website, not peer-reviewed. The organ donation figure is reported as this source states it and we did not verify it. yukaichou.com
7. Working paper on presumed consent legislation, reference list identifying Abadie, A., and Gay, S., The Impact of Presumed Consent Legislation on Cadaveric Organ Donation: A Cross Country Study; Johnson, E. J., and Goldstein, D. (2003); Madrian and Shea (2001); Matesanz, R. (2001) on a decade of continuous improvement in cadaveric organ donation and the Spanish model; and Michielsen, P. (1996) on ten years' experience of presumed consent to organ donation in Belgium; and noting that Madrian and Shea found the participation rate higher under default enrollment with many workers retaining the default contribution rate and fund allocation. Note: a working paper. We did not obtain any of the presumed consent studies and report no findings from them. nber.org
8. Publisher record with third-party citing text, giving the citation for Abadie, A., & Gay, S., The Impact of Presumed Consent Legislation on Cadaveric Organ Donation: A Cross Country Study, Journal of Health Economics, 25, 599–620, dated 2006 in the citing text and 2004 in another source; and recording that defaults have been shown to influence decision making in retirement savings, organ donation, green energy consumption, HIV screenings, colorectal cancer screenings, and participation in tobacco cessation programs, with citations across those domains. Note: a publisher record with third-party indexed text; we obtained none of the studies described and note the date discrepancy without resolving it. researchgate.net
9. Carroll, G. D., Choi, J. J., Laibson, D., Madrian, B. C., & Metrick, A., Optimal Defaults and Active Decisions, published abstract via PubMed, on the finding that compelling new hires to make active decisions about 401(k) enrollment raises the initial fraction that enroll by 28 percentage points relative to a standard opt-in enrollment procedure. Note: we obtained the abstract, which truncates shortly after this sentence; we did not obtain the paper. pubmed.ncbi.nlm.nih.gov
10. Keller, P. A., and colleagues (2011). Enhanced active choice: A new method to motivate behavior change. Journal of Consumer Psychology, publisher record, on opt-out policies in which the alternative preferred by the policy maker is made the default having gained great traction but potentially not being feasible in health care settings; on the authors presenting a series of studies including two field experiments testing an alternative active choice policy in which there is no default but decision makers are required to make a choice; on their proposing and testing a modified version termed enhanced active choice that favors one alternative by highlighting losses incumbent in the non-preferred alternative; and on their recommending enhanced active choice as a complement to automatic enrollment or when automatic enrollment is infeasible or unethical. Note: a publisher record; we obtained the abstract only. myscp.onlinelibrary.wiley.com

This article describes research on choice architecture and is not investment, pension, tax or legal advice. No paper discussed was obtained in full. The specific contribution and allocation percentages come from a United States patent document quoting the study, not from the paper, and are flagged as such throughout. The organ donation registration figure comes from a commercial website. The literature on whether presumed consent affects actual donation rates was not obtained and is reported by citation only. All arithmetic is the authors' own, uses invented figures, and is illustrative rather than a projection.