The Stake That Changes Tomorrow: Designing Capital for Children Who Start With Nothing

In 2008, a few thousand mothers in Oklahoma opened a letter telling them the state had opened an investment account in their newborn’s name and put $1,000 in it. Most would never add a cent. The babies, obviously, had no idea. The money was locked away for eighteen years and could not be touched for rent, or nappies, or the electricity bill.

By any ordinary accounting, nothing had happened. A thousand dollars is a fortnight’s rent. It is not wealth.

Then the researchers came back, and kept coming back for eighteen years. The families who got that letter had been picked at random from a much larger group, so the families who got nothing were, statistically speaking, their twins. Over the years that followed, the mothers who had the account showed fewer depressive symptoms and held higher educational expectations for their children. The children showed better early social and emotional development. The research centre that ran the study summarises its own results in a telling way: the benefits accrued above all to financially vulnerable families — the households for whom $1,000 solved nothing at all.

Nothing material had changed. What changed was what a family believed was going to happen next.

The account did not make anyone rich. It made someone expect something. That distinction is the whole argument, and most writing on this subject has it the wrong way round.

The question this policy keeps being asked, and the one it can answer

Child capital accounts — baby bonds, child development accounts, whatever the local name — are almost always argued about as instruments of distribution. Do they narrow the gap between the top and the bottom? How much of the wealth concentration do they undo? It is a natural question, and it is the wrong one, for a reason that has nothing to do with politics.

It is the wrong question because the policy cannot answer it, and everybody involved knows this. The most ambitious design in serious circulation would give the poorest children $50,000 to $60,000. Independent modelling finds it would narrow the United States racial wealth gap considerably, and not one simulation closes it; estimates of what remains range from tens of thousands of dollars to well over a million by mid-century. Measured against distribution, a policy like this is always going to look like a near-miss, no matter how well it works.

But there is a second question, and it is the one the evidence actually speaks to. Does a small, certain stake make life materially better for a person who starts with nothing? That question does not require anyone to look upward for a comparison. It asks about a floor, not a ladder, and it turns out to have a surprisingly precise answer.

Ask how much of the gap this closes and the answer is always disappointing. Ask what it does for the child at the bottom and the answer is unexpectedly specific.

Two different theories wearing the same policy

Almost every case for giving children capital at birth runs two arguments at once, without noticing that they are different arguments pointing toward opposite designs.

The first is a compounding argument. Put capital in early, let markets work for eighteen years, and a modest deposit becomes a substantial sum. This is the version that appears in press releases, and it is the one that fails.

It fails for an arithmetic reason, which the next section sets out, and for a quieter reason worth stating here. Returns are not the same for everyone. Using Norwegian administrative data covering an entire population, Andreas Fagereng and colleagues found that returns to wealth rise systematically with the wealth already held — a difference of roughly 500 basis points a year between the tenth and ninetieth percentiles — and that this persists within individuals and across generations. The practical implication is not a grievance about the wealthy. It is that a model built on a market-average return will overstate what a small, unsophisticated, fee-exposed account actually earns.

That finding has a constructive edge, and it is the better half of it. If small investors systematically do worse, then a single state-run trust holding broad index funds and charging almost nothing may be the only route by which a low-income household ever receives an institutional-grade return. The design answer is a good public trust, not a bigger promise.

The second argument is quieter and far better evidenced. Michael Sherraden’s asset-based welfare theory holds that owning something does what receiving an income does not: it changes how a household orients itself toward the future. Income is consumed and forgotten. An asset is a fact about tomorrow. It makes planning rational, which makes effort rational, which changes behaviour — and it appears to do this at sizes far too small to matter financially. That is precisely what the Oklahoma mothers demonstrated.

The distinction is not academic, because the two theories imply opposite priorities. If compounding is the mechanism, size and investment returns are everything, and every spare dollar should go into bigger endowments. If asset effects are the mechanism, then visibility, universality, early timing and the certainty that the money will actually be there matter far more than the amount — and a modest, well-communicated, automatic account may beat a larger one nobody can see.

The arithmetic nobody wants to do

Proposals in this field routinely promise more than compounding can deliver. The error is easy to make, because exponential growth feels like it ought to be bigger than it is.

Take a generous design: $2,000 for every newborn, rising to $10,000 for children born below the poverty line, invested at a real return of 6 to 7 per cent a year. After eighteen years the universal account holds roughly $5,700 to $6,800, and the poverty-line account roughly $28,500 to $33,800. To get a $2,000 seed to $25,000 in eighteen years you would need a sustained real return above 15 per cent — for every citizen, for their entire childhood.

Any sensible design makes this worse before it makes it better. A lifecycle glide path that moves the portfolio out of shares and into bonds through the teenage years is prudent, because a market crash the month before an eighteenth birthday would be ruinous. It also lowers the return over the full period. Safety costs compounding; a proposal cannot claim both.

We now have something better than a model, because the Oklahoma accounts have matured. In the first year of distributions, $349,550 was paid out to 150 young adults across 70 colleges, universities and trade schools. The average distribution was about $2,330, and recipients typically took the whole balance.

It is worth sitting with that figure, because the most widely cited projection for exactly this design — McKinsey’s modelling of the new United States federal accounts — puts a $1,000 seed at about $5,560 by age 18, on an assumed 10 per cent average annual return. The measured result came in at well under half the modelled one. Markets did not read the model.

Every number in this debate is a projection except one, and the one real number is less than half the projection.

Read through the compounding frame, $2,330 is a failure. It is a fraction of what was promised and a rounding error against any wealth statistic. That reading is about to look very strange.

The number that matters turns out to be small

There is a separate body of research, rarely cited in these debates, that asks a blunt question: how much money does a household need on hand to stop a bad week becoming a catastrophe?

Columbia University’s Poverty Tracker followed working-age New Yorkers through financial shocks — a lost job, a medical bill, a car that died. Among those not already in difficulty, 24 per cent of people who experienced a shock fell into material hardship, against 10 per cent of those who did not. Then they looked at what protected people, and found something that ought to change how this policy is designed.

●        Households with nothing were the most likely to fall into hardship after a shock.

●        At about $2,000 in liquid savings, the risk of entering most kinds of hardship — financial, billing, food and housing — fell by roughly 40 per cent.

●        Above $2,000, the curve flattened. Having $3,000 was barely different from having $2,000.

●        The exception was medical hardship, where savings offered very little protection — a reminder that an asset is not a health system.

Supporting evidence points the same way. A study of more than 7,500 low-income American tax filers found that those who deposited part of a refund into savings were, six months later, between 21 and 30 per cent less likely to have skipped rent, skipped a bill, delayed medical or dental care, or gone without prescription medication. That study is observational and cannot prove causation, so it should be read as corroboration rather than proof.

Now put the two findings side by side. The protective value of a cash buffer is largely achieved by about $2,000. The Oklahoma accounts paid out an average of $2,330.

The amount that actually showed up is, almost exactly, the amount the hardship research says a person needs. The policy was being measured against the wrong yardstick the entire time.

This reframes what a child capital account is for. It is not a wealth-building instrument that underperformed. It is a mechanism that reliably delivers, to a young adult with no family balance sheet behind them, roughly the sum at which the evidence says the protective benefit of savings is realised — at exactly the age when they leave school, leave care, move out, start work or start training, and have no buffer of any kind.

One popular explanation for why this matters should be handled carefully. The idea that poverty itself consumes mental bandwidth — that scarcity taxes attention and makes good decisions harder — is intuitive and widely repeated. The original evidence, however, has not held up well: several attempted replications in Vietnam, Zambia, Brazil and the United States failed to reproduce the headline cognitive effects, and a 2021 review concluded that the literature does not provide unambiguous support for the hypothesis. The intuition may still be right. It is not yet a foundation to build on, and arguments for this policy do not need it.

What other countries have learned about reaching the bottom

Figures are from official programme sources, national audit bodies and the published studies; full references at the end.

Australia already ran this experiment, at scale

This debate is overwhelmingly American, which is strange, because the largest and longest-running test of the underlying design philosophy is Australian.

Superannuation is a compulsory, universal, market-invested, default-enrolled asset-accumulation system. Employers must contribute 12 per cent of wages, a rate reached in July 2025 after three decades of legislated increases. Anyone who makes no choice is placed in a regulated default product. As at 30 June 2026 the system held about $3.2 trillion across 25.2 million member accounts, 14.9 million of them in those defaults. Automatic enrolment, no application, professional management, regulated low-cost defaults — every feature asset-building advocates argue for has been running here, nationally, for a generation.

It shows that regulated defaults protect the people who never engage. When the prudential regulator began publishing an annual performance test in 2021, thirteen MySuper products failed it. By 2025 none did, and the number of members sitting in products that failed had fallen from about one million to 8,500. Underperforming funds improved, merged or closed. Nobody in those funds had to notice, complain or switch. That is the whole point: the person least equipped to monitor a financial product was the person the reform protected, and it worked without requiring anything of them.

It also shows what happens when the buffer and the long-term asset are the same pot of money. When early access to superannuation was permitted during the pandemic, $36.4 billion was withdrawn across 4.8 million payments, averaging $7,638. More than half a million people emptied their accounts entirely. Those who withdrew were disproportionately young and low-paid. Faced with an immediate crisis and one asset, people did the rational thing and spent it — and the people who did so were the ones with the least to spare.

Lock the money up and a family in crisis cannot reach it. Leave it open and the crisis takes it. The answer is not a better lock. It is a second, smaller pot.

Which is the strongest practical argument in this whole field: pair a small, genuinely liquid emergency buffer with a protected long-term account, so a car repair does not consume a childhood of saving. The United States has legislated a version — emergency savings accounts linked to pensions, capped at $2,500, with the employer match routed into the protected retirement account. The cap is, notably, almost exactly the figure the hardship research points to. Take-up has been close to nil: around three per cent of small plans offer one, and roughly nine in ten large plans say they are not considering it. A good mechanism nobody adopts helps nobody.

Four objections that deserve real answers

1. Is this the best way to help the same people? A programme on the scale proposed in the United States would cost around $60 billion a year — money not spent on childcare, housing supply, early intervention or simply larger payments to poor families now. Several of those have stronger causal evidence behind them, and a family struggling this month is not helped by an asset their child receives in 2044. The honest reply is that these do different jobs: transfers raise current consumption, an asset provides a buffer at a specific transition. The second honest reply is about durability — transfers are easy to cut, and an asset with a named owner is harder to take back. Though as Washington DC demonstrated, not impossible.

2. Restricting what young people may do with their own money is paternalism, and should be defended as such. Most proposals permit only education, a first home, a registered business or a retirement rollover, and several require a financial-education module before release. There are decent reasons for this. But a scheme that exists to give young people from poor households some independence begins by telling them, and only them, what their independence is for. And if the buffer finding is right, the most valuable use of the money may be precisely the unglamorous one the restrictions forbid: covering a crisis so it does not become a spiral.

3. A national trust holding an entire birth cohort’s share portfolio is a very large shareholder. Who votes those shares? Over decades a public master trust of the size proposed would become a significant owner of the listed market, with real influence over corporate governance. That might be benign or even desirable. It is not nothing, and no proposal in this field has seriously addressed it.

4. The claim that financial education does not work is weaker than it is usually presented. The standard citation is the 2014 meta-analysis by Fernandes, Lynch and Netemeyer, which found that financial-literacy interventions explained about 0.1 per cent of the variance in later financial behaviour. Its central insight — that teaching compound interest to someone with no capital is close to useless — is sound, and the authors proposed a “real but narrower role” for just-in-time education. But two things are usually mangled in the retelling. Their decay finding is that at delays of 18.5 months or more there is no significant effect of even 24 hours of instruction — a statement about hours of teaching, not a 24-month expiry date. And a 2022 meta-analysis by Kaiser, Lusardi, Menkhoff and Urban, covering 76 randomised experiments and more than 160,000 people, found effects on behaviour three to five times larger, argued that “variance explained” makes meaningful effects look trivial, and reported that they could not replicate the earlier result and had found coding errors in it. This is a live dispute, not a settled finding.

The oldest version of this argument was never about the gap

Asset-based welfare is usually filed as a progressive policy. Its oldest form is not.

The phrase “property-owning democracy” was coined in the 1920s by Noel Skelton, a Scottish Conservative MP. His argument was not that some people had too much. It was that a person who owns nothing is not fully free — that having no stake leaves someone permanently dependent on an employer, a landlord or the state, and that a society of such people cannot sustain self-government. The remedy was to spread ownership rather than expand relief. The British Conservative tradition carried that argument through the post-war decades and into right-to-buy. George W. Bush campaigned on an “ownership society”. Singapore’s Central Provident Fund was built by a government nobody would call left-wing.

What is striking is that this older argument is the one the evidence supports. Skelton was making a claim about the condition of the propertyless person — their security, their independence, their capacity to plan — not about the distance between them and anyone else. That is an absolute claim about a floor, and it is exactly what the Oklahoma mothers and the hardship research describe.

The case for giving a child a stake does not require anyone to be worse off, or even for anyone to be compared. It requires only that we think a person with nothing behind them is less free than a person with something.

Amartya Sen put the general version of this well: what matters is not what people own but what owning it lets them do and be. An asset too small to make anyone rich can still be large enough to let someone turn down bad work, leave a bad house, finish a course, or absorb a shock without borrowing at forty per cent. Those are capabilities, and they are what this policy actually distributes.

Measure the floor

If the goal is the floor rather than the gap, several things follow immediately, and they are not the things the distributional framing suggests.

Bigger is not obviously better. If the protective benefit of a buffer plateaus around $2,000, then the marginal value of raising an endowment from $3,000 to $10,000 is far lower than the marginal value of making sure every child actually has $2,000. Universality and reliability beat generosity. That is a genuinely counterintuitive conclusion and it only becomes visible once you stop measuring against the top.

Certainty matters more than size. An asset only changes how a family plans if the family believes it will be there. A programme that might be defunded delivers none of the psychological benefit and all of the cost.

Timing is a design variable. The money is most valuable at the moment a young person has the least behind them — leaving school, leaving care, moving out. Paying it then is worth more than paying more later.

There is a cost to this reframe and it should be admitted. Floor language is vaguer than wealth language, harder to audit, and far easier for a government to claim without delivering. “We improved wellbeing” survives any amount of failure. If the discipline no longer comes from a distributional target, it has to come from pre-committed, published outcome measures: hardship entry rates, post-school transitions, use of high-cost credit. Those are measurable. The question is whether a programme will agree in advance to be measured by them.

Six questions to ask of any child capital scheme

1.  Does every child get one without anyone applying? Every scheme that required a parent to act has lost the children it most wanted to reach. Birth registration, not a form.

2.  What does the least-engaged family receive? Whatever the default is, that is the real programme. The parents who never open the letter are the parents whose children need it most.

3.  Will there be at least a couple of thousand dollars at the end, reliably? That is the level at which the hardship evidence says protection is realised. Hitting it for everyone matters more than exceeding it for some.

4.  Can the family reach money in a crisis without destroying the asset? A small liquid buffer beside a protected core. Australia showed what happens when there is only one pot — and the people who emptied it were the ones with least to spare.

5.  Does the asset disqualify the family from anything? Savings that cost a household its benefits are not savings. That exemption belongs in the statute, not the press release.

6.  What has been promised, and will anyone measure it? Promise a share of the wealth distribution and the scheme will be judged against a target it cannot hit. Promise a floor — fewer young people entering hardship, fewer borrowing at punitive rates, more completing training — and it can be held to something real, and it can pass.

The mothers in Oklahoma were not made wealthy. Eighteen years later their children collected an average of $2,330, which would not buy a reliable second-hand car. Measured against the wealth of the people at the top, that is nothing at all. Measured against having nothing, which is what those young people would otherwise have had, it is roughly the amount that research says keeps a bad month from becoming a bad decade — and something had already happened in those households a long time before the money was ever spent.

Sources

1. Sherraden, M. (1991), Assets and the Poor: A New American Welfare Policy — the founding statement of asset-based welfare theory.

2. SEED for Oklahoma Kids, Center for Social Development, Washington University in St. Louis — study overview, randomised design and findings across waves. https://csd.wustl.edu/sok-studies-1025/

3. Center for Social Development, “Financially Vulnerable Families Reap Multiple Benefits From Child Development Accounts” — CSD’s own framing of where the SEED OK effects concentrate. https://csd.wustl.edu/19-40/

4. Huang, J. et al., “Impacts of Child Development Accounts on maternal depressive symptoms: Evidence from a randomized statewide policy experiment”, Social Science & Medicine. https://www.sciencedirect.com/science/article/abs/pii/S0277953614002548

5. Center for Social Development, first-year SEED OK distributions — $349,550 distributed to 150 young adults across 70 institutions, averaging about $2,330. Note this reports how funds were used; it does not report a treatment-versus-control effect on enrolment. https://csd.wustl.edu/sok-2025-distributions/

6. Robin Hood Poverty Tracker (Columbia University), Liquid Assets, Financial Shocks, and Entrances into Material Hardship, August 2023 — 24 per cent of those experiencing a shock entered hardship against 10 per cent of those who did not; about $2,000 in liquid assets reduced the likelihood of entering most hardship types by roughly 40 per cent, with the benefit flattening above that level and offering little protection against medical hardship. https://robinhood.org/wp-content/uploads/2024/01/poverty-tracker_2023.08_spotlight-on-liquid-assets-financial-shocks-and-entrances-into-material-hardship.pdf

7. Despard, M. et al. (2017), “Do Tax-Time Savings Deposits Reduce Hardship Among Low-Income Filers?”, Social Service Review — 7,537 low-income filers; depositors 21 to 30 per cent less likely to report five of six hardship outcomes six months later. Observational, not causal. https://www.journals.uchicago.edu/doi/full/10.1086/689357

8. Fagereng, A., Guiso, L., Malacrino, D. & Pistaferri, L. (2020), “Heterogeneity and Persistence in Returns to Wealth”, Econometrica 88(1) — returns rise with wealth, with roughly 500 basis points between the 10th and 90th percentiles. https://www.nber.org/papers/w22822

9. Mani, A., Mullainathan, S., Shafir, E. & Zhao, J. (2013), “Poverty Impedes Cognitive Function”, Science 341 — the original scarcity-bandwidth finding. https://www.science.org/doi/10.1126/science.1238041

10. Review of scarcity theory in Theory and Decision (2021) — concluding that the literature “does not provide unambiguous evidence for the hypothesis that poverty reduces mental bandwidth”, with failed or null replications in Vietnam, Zambia, Brazil and the United States. https://link.springer.com/article/10.1007/s11238-021-09802-7

11. Committee for a Responsible Federal Budget, analysis of Senator Cory Booker’s American Opportunity Accounts — approximately $60 billion a year, reaching about $46,000 by age 18 via a $1,000 seed plus annual deposits of up to $2,000. https://www.crfb.org/blogs/cory-bookers-baby-bonds-plan

12. Urban Institute, What Do We Know About Baby Bonds? and Modeling a National Baby Bonds Program — simulations narrowing but never closing the racial wealth gap. https://www.urban.org/sites/default/files/2023-09/What%20Do%20We%20Know%20About%20Baby%20Bonds.pdf

13. UK Child Trust Fund: National Audit Office, Investigation into Child Trust Funds, HC 1197 (14 March 2023) — 45 per cent of matured accounts unclaimed as at 5 April 2021, worth £394 million, averaging £2,721. Note this covers only the first seven months of maturities and excludes 14 providers. https://www.nao.org.uk/wp-content/uploads/2023/03/investigation-into-child-trust-funds.pdf

14. House of Commons Library, Child Trust Funds, Briefing CBP-9988 — scheme design, the £250 and £250 payments later reduced to £50 and £100, the 1.7 million accounts opened by HMRC (28 per cent of 6.3 million), and 760,000 accounts unclaimed as at April 2025. https://researchbriefings.files.parliament.uk/documents/CBP-9988/CBP-9988.pdf

15. HMRC, Annual Savings Statistics (16 September 2026) — 827,000 matured accounts still unclaimed as at 5 April 2026 out of 3,784,000 matured, averaging £2,310. https://www.gov.uk/government/statistics/annual-savings-statistics-2026/commentary-for-annual-savings-statistics-september-2026

16. Grinstein-Weiss, M. et al. (2021), “Household Savings Decisions in Israel’s Child Savings Program”, Journal of Family and Economic Issues 42(2) — income was not significantly associated with active enrolment; parental education strongly predicted choosing a higher-yield fund. https://repository.library.washu.edu/cgi/viewcontent.cgi?article=1020&context=spi_research

17. Connecticut Office of the State Treasurer, CT Baby Bonds — up to $3,200 per child, automatic enrolment via HUSKY, funded from a roughly $393 million reserve, claimable between 18 and 30 subject to residency and a financial-literacy course. First claims fall due around 2041. https://portal.ct.gov/ott/ct-baby-bonds/faq

18. DC Child Wealth Building Act of 2021 (Law 24-53); and DC Fiscal Policy Institute on the FY2026 budget, in which the programme was defunded and the legislation repealed without any child having been enrolled. https://dcfpi.org/all/whats-in-the-fiscal-year-2026-dc-budget/

19. Internal Revenue Service, Instructions for Form 4547, Trump Account Election(s) — the $1,000 pilot contribution for children born 2025 to 2028, the elective (non-automatic) enrolment, and the $5,000 annual contribution limit, within which employer contributions of up to $2,500 sit. https://www.irs.gov/instructions/i4547

20. McKinsey Institute for Economic Mobility, From Access to Assets — modelled outcomes of about $5,560 for a seed-only account, assuming a 10 per cent average annual return. McKinsey presents these as illustrative scenarios, not forecasts. https://www.mckinsey.com/institute-for-economic-mobility/our-insights/from-access-to-assets-how-early-life-wealth-building-can-shape-economic-mobility

21. Fernandes, D., Lynch, J. G. & Netemeyer, R. G. (2014), “Financial Literacy, Financial Education, and Downstream Financial Behaviors”, Management Science 60(8) — 201 studies across 168 papers; no significant effect of even 24 hours of instruction at delays of 18.5 months or more; and the authors’ proposal of a “real but narrower role” for just-in-time education. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2333898

22. Kaiser, T., Lusardi, A., Menkhoff, L. & Urban, C. (2022), “Financial education affects financial knowledge and downstream behaviors”, Journal of Financial Economics 145(2) — 76 randomised experiments, more than 160,000 individuals, behavioural effects three to five times larger than previously estimated, plus a failure to replicate and four identified coding errors. https://www.nber.org/system/files/working_papers/w27057/w27057.pdf

23. APRA, 2025 superannuation performance test results — no MySuper product failed; members in products that did not pass fell from about one million to 8,500. https://www.apra.gov.au/news-and-publications/apra-releases-2025-superannuation-performance-test-results-and-product

24. APRA, COVID-19 Early Release Scheme final report — $36.4 billion paid across 4.8 million payments, averaging $7,638. https://www.apra.gov.au/covid-19-early-release-scheme-issue-36-accessible-version

25. ASFA superannuation statistics (June quarter 2026) — total assets of about $3.2 trillion across 25.2 million member accounts, 14.9 million of them MySuper. https://www.superannuation.asn.au/wp-content/uploads/2026/09/Super-Stats-.pdf

26. Plan Sponsor Council of America survey data on Pension-Linked Emergency Savings Accounts — adoption in the low single digits, with roughly nine in ten large plans not considering one. https://www.psca.org/news/psca-news/2026/8/plan-sponsors-remain-wary-of-in-plan-emergency-savings-accounts/

27. On property-owning democracy: Noel Skelton’s coinage in the 1920s and its development in the British Conservative tradition. https://en.wikipedia.org/wiki/Property-owning_democracy

28. Sen, A. (1999), Development as Freedom — the capability approach, and the argument that what matters is what resources allow people to do and be.