> ## Documentation Index
> Fetch the complete documentation index at: https://meta.niceshare.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Default Effect

> Default Effect is the higher rate at which a pre-selected option is kept when doing nothing selects it. See one 401(k) cohort and one organ question.

<Info>
  **Category**: Effects<br />
  **Type**: Keeping the option that doing nothing selects<br />
  **Origin**: Brigitte C. Madrian and Dennis F. Shea, NBER Working Paper 7682 (May 2000), published in the *Quarterly Journal of Economics* (November 2001); Eric J. Johnson and Daniel Goldstein, *Science* (21 November 2003)<br />
  **Also known as**: A no-action default
</Info>

<Note>
  **Quick Answer** — The Default Effect is the higher rate at which people keep an option when that option is already selected and doing nothing keeps it. Madrian and Shea measured it in one company's 401(k) plan. At 3 to 15 months of tenure, 85.9 percent of 5,801 employees hired under automatic enrollment were in the plan, against 37.4 percent of 4,249 employees hired earlier who had to enroll themselves. Johnson and Goldstein's online question, put to 161 people, found 42 percent consenting when donation started unchecked, 82 percent when it started checked, and 79 percent when the form had no default.
</Note>

## What is Default Effect?

The Default Effect is the higher rate at which people keep an option when that option is already selected and doing nothing keeps it, compared with the rate at which they choose the same option when they must mark it themselves.

A permission slip that arrives with "I agree" already marked, and the same slip with an empty box, are the everyday picture. The trip does not change. Setting the pen down is the whole difference. A printer that ships with double-sided printing already checked does the same work. The paper and the ink are the same. The checkmark is the difference.

> Most public policy choices have a no-action default, that is, a condition is imposed when an individual fails to make a decision.

That is Eric J. Johnson and Daniel Goldstein, writing in 2003. The condition is whatever the form does when the person does not act. [Status Quo Bias](/effects/status-quo-bias) is a preference for the state already in place, including when keeping it takes effort. A pre-checked box can be new. In their organ question, people were told to assume they had just moved to a state, so the checkmark was not a long personal history. [Loss Aversion](/effects/loss-aversion) is one weight they name for leaving a reference point. [Endowment Effect](/effects/endowment-effect) is a higher price on what a person already holds. [Ambiguity Aversion](/effects/ambiguity-aversion) is a preference for a chance that has been written down. A default is a mark on a form, not a missing probability.

### Default Effect in 3 Depths

* **Beginner**: Take one permission slip with "I agree" already marked, and the same slip with an empty box. Write down what is true if the person sets the pen down.
* **Practitioner**: When you can, publish three versions of the same question: yes already marked, no already marked, and a blank that requires a mark. In the 161-person organ question, the blank sat at 79 percent and the pre-checked yes sat at 82 percent. The pre-checked no sat at 42 percent. The paper says the first two did not differ significantly.
* **Advanced**: A jump in enrollment can sit beside a lower contribution. In the tenure-matched 401(k) comparison, 85.9 percent were in the plan and participants contributed 4.4 percent of pay on average. The earlier cohort had 37.4 percent in the plan and participants contributed 7.3 percent. A 2019 meta-analysis of opt-out against opt-in puts the average at d = 0.68, and the studies do not share one size.

## Origin

The Default Effect, as this page uses the name, is the no-action option on a form. Madrian and Shea used the phrase for a savings package. Johnson and Goldstein set the same idea out for organ donation and named three paths.

Brigitte C. Madrian and Dennis F. Shea, "The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior," NBER Working Paper 7682 (May 2000), is the paper read for the figures below. It was published in the *Quarterly Journal of Economics*, volume 116, number 4 (November 2001), pages 1149–1187. Section VIII of the working paper is titled "The 'Default' Effect of Automatic Enrollment." They define default behavior as staying in the plan at the contribution rate and the fund the company filled in. The company is a publicly traded Fortune 500 firm in health care and insurance. The paper does not print its name. The case below keeps the counts inside that working paper.

Eric J. Johnson and Daniel Goldstein, "Do Defaults Save Lives?," *Science*, volume 302, number 5649 (21 November 2003), pages 1338–1339, is the organ paper. Both authors were at the Center for Decision Sciences at Columbia University. They write that a default can matter in three ways. Decision-makers might believe the policy-maker is suggesting the pre-selected action. Accepting the default takes no effort, and they cite Samuelson and Zeckhauser (1988) for the further claim that the physical effort of a form can raise acceptance. A default often stands for the existing state, and a change is a trade-off. They cite Tversky and Kahneman, *Quarterly Journal of Economics* (1991), for losses looming larger than gains of the same size. Those three paths are their account. The counts below are what the studies measured.

An oft-cited Gallup poll, as they report it, found that 85 percent of Americans approved of organ donation, that less than half had made a decision, and that 28 percent had signed a donor card. Approval is not a signature. That poll is not a test of a default.

Their first study was an online experiment. Note 24 points to supporting material for the methods. This page uses only the printed article. One hundred sixty-one respondents were asked whether they would be donors, on one of three questions. In the opt-in condition they were told to assume they had just moved to a new state where the default was not to be a donor, and they could confirm or change that status. The opt-out condition was the same, except the default was to be a donor. The neutral condition required a choice and had no prior default. People changed the choice largely with a mouse click, which the authors say largely removes effort as the explanation of this particular gap. The figure prints 42 percent consenting under opt-in, 82 percent under opt-out, and 79 percent under neutral. The authors say revealed donation agreement was about twice as high when opting out as when opting in, that opt-out did not differ significantly from neutral, and that only opt-in, then the usual U.S. practice, was significantly lower. The article prints no per-condition head count, no age, and no p-value beside that sentence. The figure draws error bars and does not print their values. These are answers to a hypothetical move, not signatures on a state registry.

The country figure is a different object. It prints effective consent rates: the share who had opted in where consent must be explicit, or the share who had not opted out where consent is presumed. Johnson and Goldstein say they supplemented registry data reported by Gäbel by contacting central registries. The figure labels four explicit-consent rates: Denmark 4.25, the Netherlands 27.5, the United Kingdom 17.17, and Germany 12. It labels seven presumed-consent rates: Austria 99.98, Belgium 98, France 99.91, Hungary 99.97, Poland 99.5, Portugal 99.64, and Sweden 85.9. The authors write that the explicit-consent rates were lower, that the two distributions do not overlap, and that nearly 60 percentage points separate the groups. On the figure, the highest explicit rate is the Netherlands at 27.5 and the lowest presumed rate is Sweden at 85.9. These are register counts. They are not organs transplanted, and the countries were not assigned a law.

## Key Points

Four comparisons keep the option fixed and move the mark. A blank form is a third cell. Enrollment is not the contribution and not the fund. A change in who may join is not the same as a pre-checked box. An average across studies is not the size in every domain.

<Steps>
  <Step title="Compare the pre-checked box with the empty box and with a blank">
    The organ question holds the request fixed and moves the mark. Opt-in, the unchecked default, drew 42 percent. Opt-out, the checked default, drew 82 percent. The blank, which forced a choice, drew 79 percent. The authors say opt-out and the blank did not differ significantly. A form that requires a mark is not, in this study, a large step down from a pre-checked yes. The gap that remains is the pre-checked no. Because the change was largely a mouse click, this gap is not the cost of a stamp. The 161 people are one pool across three questions. The article does not print how many saw each question.
  </Step>

  <Step title="Count the rate and the fund, not only enrollment">
    Madrian and Shea call default behavior the package the company filled in: enrolled, at a 3 percent contribution, invested entirely in the money market fund. At matched tenure, participation was 85.9 percent in the automatically enrolled cohort and 37.4 percent in the earlier cohort. Among participants, the average contribution was 4.4 percent of pay against 7.3 percent. On 30 June 1999, 61.1 percent of 5,801 automatically enrolled employees were still on the full package. Among the 4,983 participants in that column, 71.2 percent were. The authors say only 1 percent of the earlier cohorts were at a 3 percent contribution invested entirely in the money market fund. A higher enrollment rate is not a higher savings rate.
  </Step>

  <Step title="Separate the checkmark from a rule that changes who may join">
    On 1 April 1998 the company made two changes. Everyone became eligible to contribute immediately, and the match still waited one year. Only people hired on or after that date were enrolled unless they opted out. The cohort hired in the prior year became eligible that day and still had to enroll themselves. At 15 to 27 months of tenure, participation was 48.7 percent in the older cohort (N = 3,275, observed 30 June 1998) and 49.4 percent in that prior-year cohort (N = 4,247, observed 30 June 1999). The authors say earlier eligibility added little, and they also say the comparison is not fully clean, because that prior-year group became eligible on 1 April 1998 rather than on the date of hire. The match formula did not change: after one year, the first 6 percent of pay was matched at 50 percent. The default contribution of 3 percent sits below that 6 percent threshold.
  </Step>

  <Step title="Read the average as an average, with a wide scatter">
    Jon M. Jachimowicz, Shannon Duncan, Elke U. Weber, and Eric J. Johnson, *Behavioural Public Policy*, volume 3, number 2 (2019), pages 159–186, pool 58 default studies that met their criteria, with a combined n of 73,675. Opt-out defaults led to greater uptake of the pre-selected option than opt-in defaults, d = 0.68, 95 percent confidence interval 0.53 to 0.83, p \< 0.001. That comparison is opt-out against opt-in. It is not a comparison with a blank form. In a model that codes why a default might work, endorsement, a default read as the designer's advice, has b = 0.32 (standard error 0.15, p = 0.038). Endowment, their code for a default read as the status quo, has b = 0.31 (standard error 0.15, p = 0.044). Ease has b = −0.05 (standard error 0.15, p = 0.75). These coefficients move the effect size. They are not percentage points of enrollment. After those codes, I² is 92.32 percent. The studies do not share one effect. The authors say several studies find no significant effect and two find a negative effect. They do not name those two in the sentences used here.
  </Step>
</Steps>

## Applications

Use the mark as a line you can read. Keep each percentage inside the question, the cohort, and the outcome that produced it.

<CardGroup cols={2}>
  <Card title="Write the sentence that is true if the pen stays down" icon="file-contract">
    Before a form goes out, write what happens when the reader does nothing. In the organ question, doing nothing meant "not a donor" or "a donor," depending on the condition, and the blank form was a third cell at 79 percent. The 42, the 82, and the 79 are agreement with a hypothetical move, among 161 people. They are not registry signatures.
  </Card>

  <Card title="Publish the contribution and the fund beside enrollment" icon="piggy-bank">
    If a savings plan enrolls people at 3 percent in a money market fund, report that package, not only the share who remain enrolled. In the automatically enrolled cohort, 61.1 percent of 5,801 eligible employees were still on both the rate and the fund. Participants' average contribution was 4.4 percent of pay, against 7.3 percent among participants in the earlier cohort at the same tenure.
  </Card>

  <Card title="Hold the match still when you test the checkmark" icon="scale-balanced">
    The company's match still required one year, and it still matched the first 6 percent of pay at 50 percent. The cohort that gained earlier eligibility without automatic enrollment moved from a comparison rate of 48.7 percent to 49.4 percent. The tenure-matched enrollment gap, 37.4 percent against 85.9 percent, is the remaining contrast. It is two hire windows at one firm, not a random assignment.
  </Card>

  <Card title="Read a register rate separately from a transplant rate" icon="heart-pulse">
    Denmark at 4.25 and Austria at 99.98 are effective consent rates on the country figure. Johnson and Goldstein's own count of cadaveric donations, from 1991 to 2001, is a different series: a regression moves the rate from 14.1 to 16.4 donors per million, a 16.3 percent difference, P \< 0.02. Do not treat either figure as a count of lives saved. The boundaries section says what sits between a register and a transplant.
  </Card>
</CardGroup>

## Case Study

The firm is a publicly traded Fortune 500 company in health care and insurance. The working paper does not name it. The plan began in 1985. In March 1999 the company had major locations in 38 states, the District of Columbia, and Puerto Rico. Madrian and Shea follow employees from June 1997 through June 1999. On 1 April 1998 two rules changed. Employees could contribute before they had a year of service, still with no match until that year. People hired on or after 1 April 1998 were enrolled unless they declined. If they did not decline, the contribution was 3 percent of pay, placed entirely in the money market fund. They could change the rate and the fund. A change took effect on the second bi-weekly paycheck, so the first paycheck was the default if they had not opted out. The match, the investment menu, vesting, and the rules for loans and hardship withdrawals stayed as they were. Before the change, an eligible employee had to authorize a payroll deduction, pick a rate between 1 and 15 percent, and pick funds. After one year of service, the first 6 percent of pay drew a 50 percent match.

The question is what the new hires did, set beside hires who faced the same tenure and had to enroll themselves. Three cohorts organize the tables. The old cohort was hired from 1 April 1996 through 31 March 1997. The window cohort was hired from 1 April 1997 through 31 March 1998, became eligible on 1 April 1998, and was not enrolled automatically. The new cohort was hired from 1 April 1998 through 31 March 1999 and was enrolled unless they opted out. Most analyses drop employees aged 65 and over, and the one-year windows are kept so that a new hire observed on 30 June 1999 has 3 to 15 months of tenure.

Table 5 holds that tenure fixed. On 30 June 1998 the window cohort, then at 3 to 15 months of tenure, had a participation rate of 37.4 percent (N = 4,249). On 30 June 1999 the new cohort, also at 3 to 15 months, had a rate of 85.9 percent (N = 5,801). The raw gap is 48.5 percentage points. On the June 1998 date the window cohort had been eligible only since 1 April 1998, so time-since-eligible is shorter than tenure. The authors' check on eligibility uses a later snapshot. At 15 to 27 months of tenure, the old cohort was at 48.7 percent (N = 3,275) and the window cohort was at 49.4 percent (N = 4,247). They read that pair as a small role for earlier eligibility. The window cohort at that later date, still not automatically enrolled, remains far below the new cohort's 85.9 percent.

The same table compresses some gaps and does not print a head count for each row. In the window column, white employees were at 42.7 percent and black employees at 21.7 percent. In the new column those cells are 88.2 and 81.3. Employees paid under $20,000 were at 12.5 percent in the window column and 79.5 percent in the new column. Employees paid $80,000 or more were at 68.3 and 94.2 percent. Women were at 35.9 and 86.0 percent. Men were at 42.3 and 85.7 percent. The column sizes are 4,249 and 5,801. The row sizes are not printed.

The contribution rate tells a second story. Figure 4, as the authors describe it, puts 76 percent of participants in the new cohort at the 3 percent default, against around 10 percent of participants in the other cohorts. Slightly more than a third of participants in those earlier cohorts chose 6 percent, the match threshold. Table 6 compares participants at 3 to 15 months of tenure: the average contribution is 7.3 percent in the window cohort (N = 1,589, on 30 June 1998) and 4.4 percent in the new cohort (N = 4,983, on 30 June 1999). The body text once prints 7.4 percent for that window average. The table cell is 7.3.

Including non-participation as a zero contribution, the share of employees at exactly 3 percent is the composition check. The gap in non-participation between the two tenure-matched cohorts is 48.5 percentage points. The authors add the 4.4 percent of the window cohort who were already at a 3 percent contribution and predict that 52.9 percent of the new cohort would sit at 3 percent if the only change were the new participants landing on the default. The measured share is 65.1 percent. The extra 12.2 points are their evidence that some people who would have joined anyway also landed on 3 percent. This 4.4 percent is a share of the window cohort at exactly 3 percent. It is not the 4.4 percent average contribution among new participants, and it is not the 7.3 percent average.

The fund is the third part of the package. On 30 June 1999, which does not hold tenure fixed, 75.3 percent of the new cohort's participants with a balance held only the money market fund (balance sample N = 5,205). The window cohort's cell is 6.1 percent (N = 2,207). The average share of contributions going to the money market fund is 80.5 percent in the new cohort (N = 4,992 participants) and 8.2 percent in the window cohort (N = 2,107). The authors also say 85 percent of new participants allocated contributions to only one fund. Table 8, restricted to the new cohort and to employees under 65, puts 61.1 percent of 5,801 eligible employees on the full default: in the plan, at 3 percent, and entirely in the money market fund. Conditional on participation, 71.2 percent of 4,983 are on the rate and the fund. The authors say only 1 percent of the window, old, and longer-tenured cohorts held that package. They read the mass as inertia, and also as employees taking the default as the company's advice. The advice reading is their interpretation. The table is the count.

The lesson is a real payroll default with two outcomes that move in different directions. Enrollment at matched tenure rises from 37.4 percent to 85.9 percent. The contribution among participants falls from 7.3 percent to 4.4 percent, and 61.1 percent of the new cohort remain on the 3 percent money-market package. The boundary is the design. Nobody was randomly assigned a form. The workforce is not the U.S. workforce: on 31 March 1999, 77.7 percent of the company's employees were women, against 46.9 percent in the March 1998 Current Population Survey comparison the authors print. The new cohort, in the 30 June 1999 snapshot, is younger and lower paid than the window cohort. The follow-up stops near 15 months of tenure.

## Boundaries and Failure Modes

The Default Effect is the wrong name when a register count is retold as lives saved, when a pre-checked yes is treated as better than a blank form, and when one company's enrollment jump is treated as a rise in the amount saved.

The first boundary is the outcome. The country bars are effective consent. Johnson and Goldstein's regression of cadaveric donations per million, for 1991 through 2001, uses the default as a predictor and controls for propensity to donate, transplant infrastructure, education, and religion, plus year indicators and a time-series adjustment for technology, infrastructure, and public education. They report no differences across years and a 16.3 percent higher donation rate when donation is the default (P \< 0.02), from 14.1 to 16.4 per million. The figure shows means plus or minus one standard error and does not print those errors as numbers. Gimbel and coauthors, as Johnson and Goldstein report them, find a move from 10.8 to 16.9 per million, a 56.5 percent increase, on a broader set of countries that includes many in Eastern Europe with lower rates. The two estimates are not one sample. The authors write that changes in defaults could add thousands of donors a year in the United States, and that each donor can be used for about three transplants. That is their suggestion. It is not a counted total of lives saved. They also list what stands between a register and a transplant: families' objections, doctors' hesitation to use a default, mismatch with recipients, and differences in religion, culture, and infrastructure. They cite a 1998 Dutch national registry, an education campaign, and a mailing of more than 12 million letters in a country of 15.8 million, and they say the mailing failed to change the effective consent rate.

The second boundary is the blank form. In the online question, 82 percent and 79 percent did not differ significantly. A mandate to choose was, in that cell, close to a pre-checked yes. Only the pre-checked no was lower. The mouse click largely takes postage out of this particular result. The authors say the country gap may be larger because changing a real registration takes a form, a phone call, or mail. [Choice Overload](/effects/choice-overload) is a different limit: a long menu can stall a choice. This page does not carry a menu-size experiment.

The third boundary is how far the 401(k) gap travels. The cohorts were hired a year apart. On 30 June 1999 the new cohort's active employees (N = 5,812) had a mean age of 34.5 and mean pay of $34,264. The window cohort (N = 4,257) had a mean age of 36.0 and mean pay of $38,424. Women were 78.0 percent of the new snapshot and 76.1 percent of the window snapshot. White employees were 68.8 and 71.7 percent. Those snapshots are not the participation samples of 5,801 and 4,249. The first paycheck is mechanically the default if the person has not opted out, and the participation sample starts at about three months of tenure, so that first paycheck is not the whole of the 61.1 percent. In the 2019 meta-analysis, consumer settings were associated with larger effects (b = 0.89, standard error 0.23, p = 0.0003) and environmental settings with smaller effects (b = −0.60, standard error 0.26, p = 0.028), in the model that also codes the three channels. Those are coefficients on the effect size. Two studies in the pool go the other way. A green-energy checkbox is not the 401(k) result, and the 401(k) result is not a national savings rate.

## Common Misconceptions

Three retellings travel with the checkmark. The consent chart is treated as lives saved. Automatic enrollment is treated as a rise in the amount saved. The pre-checked box is treated as laziness, or as the same thing as a preference for the status quo.

<AccordionGroup>
  <Accordion title="The country chart shows that presumed consent saves those lives">
    The bars are effective consent rates. Denmark at 4.25 and Austria at 99.98 are shares of a register, not transplants. The donation regression moves a rate from 14.1 to 16.4 donors per million. The phrase "additional thousands of donors a year" is the authors' suggestion for the United States. The online question was a hypothetical move, and opt-out at 82 percent did not differ significantly from the blank at 79 percent.
  </Accordion>

  <Accordion title="Automatic enrollment raises how much people save">
    At matched tenure, participation goes from 37.4 percent of 4,249 employees to 85.9 percent of 5,801. Among participants, the average contribution goes from 7.3 percent of pay to 4.4 percent. On the later date, 61.1 percent of the automatically enrolled cohort are in the plan at 3 percent and entirely in the money market fund. The match, after one year, applies to the first 6 percent of pay. A person who stays at 3 percent is below that threshold. Enrollment and the dollars contributed are different rows.
  </Accordion>

  <Accordion title="The Default Effect is only laziness, or only status quo bias">
    In the online question the pre-checked box was assigned by the wording, and people were asked to assume they had just moved. The change was largely a mouse click, and the opt-in cell was still lower. In the 2019 meta-regression, ease is not a significant moderator of the effect size. Endorsement and a default coded as the status quo are. Status quo bias is a preference for the state already occupied. A default can be a mark the person has never chosen.
  </Accordion>
</AccordionGroup>

## Related Concepts

The Default Effect sits beside a preference for the current state, beside the extra weight of a loss, beside a higher value on what is already held, beside a menu that stalls the choice, and beside a chance that has not been numbered.

<CardGroup cols={3}>
  <Card title="Status Quo Bias" icon="balance-scale" href="/effects/status-quo-bias">
    A preference for the option already in place. A pre-checked box can be new, as in the organ question's hypothetical move.
  </Card>

  <Card title="Loss Aversion" icon="exclamation-triangle" href="/effects/loss-aversion">
    A loss weighs more than a gain of the same size. Johnson and Goldstein name that weight as one path when the default is the reference point.
  </Card>

  <Card title="Endowment Effect" icon="gift" href="/effects/endowment-effect">
    A higher value on what a person already holds. The 2019 meta-analysis codes a related path and finds it moderates the average effect size.
  </Card>

  <Card title="Choice Overload" icon="list" href="/effects/choice-overload">
    A long menu can stall a choice. The default on this page is the mark already on the form, not the length of the menu.
  </Card>

  <Card title="Ambiguity Aversion" icon="circle-question" href="/effects/ambiguity-aversion">
    A preference for a prize whose chance is known. A default is a pre-selected option, whether or not a probability is printed beside it.
  </Card>
</CardGroup>

## One-Line Takeaway

<Tip>Before a form goes out, write the sentence that becomes true if the reader does nothing, and if the form enrolls people, publish the contribution and the fund beside the head count.</Tip>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.