Does financial education work? What 76 randomised trials, two large school experiments and two decades of state mandates say

The soundbite and the arithmetic

The 2014 claim came from Fernandes, Lynch and Netemeyer, who pooled 201 studies of financial literacy and education, most of them not experiments, and reported that the interventions among them explained 0.1% of the variance in behaviour Fernandes et al. 2014. Reported as a share of variance, almost any educational effect looks like nothing. Expressed as an effect size, their estimate for the randomised studies was 0.018 SD.

When Kaiser, Lusardi, Menkhoff and Urban re-ran the exercise in 2022 on 76 randomised trials, they could not reproduce the 2014 number and found four coding errors in the original set Kaiser et al. 2022. Using the 2014 averaging method on the 64 trials with behaviour outcomes gives 0.065 SD, "about 3.6 times" the earlier figure. Using the standard method gives 0.100 SD for behaviour and 0.204 for knowledge, "at least three times as large" as the figure behind the soundbite. For children under 14 the pooled knowledge effect is 0.28 SD, from seven studies the authors call imprecise; the behaviour effect for that age group is 0.06 from nine.

A tenth of a standard deviation is modest and cheap: of the 20 trials that reported costs, the mean was $60.40 per participant and the median $22.90, in 2019 dollars Kaiser et al. 2022. A separate meta-analysis of 37 studies of school programmes concluded that, for knowledge, financial education is about as effective as teaching in other school subjects Kaiser & Menkhoff 2020.

What fades and what holds

Does it last? For knowledge, not on its own. Across school programmes the effect is 0.43 SD when measured within a month of the course, 0.21 after a month or more, 0.10 after three months, and from six months on the estimates (0.06, then 0.12) rest on two or three studies and cannot be told from zero Kaiser & Menkhoff 2020.

Behaviour follows a different curve. Across the 76 trials of all ages, the effect is 0.099 SD within six months, 0.071 at six months or more, 0.088 at a year or more, 0.065 at eighteen months and 0.057 at two years or more, the last from seven studies. The authors describe "some decay in effectiveness when measurement is delayed by six months or more" and then no further significant decline Kaiser et al. 2022. One more practical finding: knowledge gains rise with teaching time up to roughly ten to fifteen hours and then flatten, and behaviour effects barely vary with intensity at all Kaiser & Menkhoff 2020. More hours of the same thing is not the lever.

When the lessons meet real money

The most convincing evidence does not rely on asking teenagers whether they save. Three sets of studies followed pupils into adult credit records.

In Brazil, a seventeen-month programme woven into ordinary subjects in 892 public high schools raised financial proficiency by a quarter of a standard deviation, nudged saving and budgeting, and spilled over to parents, who became slightly more likely to save Bruhn et al. 2016. The programme was backed by Brazil's stock exchange and securities regulator among others, and it had a short-run cost: the published abstract reports an increase in students' use of expensive credit for consumer purchases. Eight to nine years later, in Central Bank of Brazil registries, those students were 1.4 percentage points less likely to carry credit-card debt (against 23% of the comparison group), 0.9 points less likely to use overdrafts and 0.9 points less likely to be in arrears, reversing the early finding Bruhn et al., long run. The follow-up is a conference paper that matched only 45% of students, so read the direction, not the decimals.

In Peru, a one-semester course delivered by trained teachers in 300 public secondary schools, run with the banking regulator and the banks' association, raised students' scores by 0.16 SD and their teachers' own financial skills by 0.32 SD, at a marginal cost the author puts at US$4.8 per student Frisancho 2023. Three years on, the published version reports, it "did not affect students' credit or repayment behaviour on the extensive margin", that is, whether they borrowed or repaid at all, but "reduced arrears by 20%" among the few with loans. A seven-year follow-up released as a working paper in 2025 found treated students borrowing somewhat more, on slightly better terms, with no change in repayment CAF 2025.

In the United States, a FINRA Foundation-funded study compared three states that required personal-finance teaching, inside an economics course, from the class of 2007 with their neighbours in credit-bureau data. First-year cohorts gained nothing. By the third cohort, credit scores were higher and 90-day delinquency lower in all three states, most in Texas, whose mandate came with standard materials and testing Urban et al.. The authors' conclusion doubles as a warning: "if a rigorous financial education program is carefully implemented, it can improve the credit scores". Content matters too. In a separate credit-record study, financial-education requirements were followed by less non-student debt, both maths and financial-education requirements by better repayment, and extra economics coursework by more delinquency and bankruptcy Brown et al. 2013.

Games: what they teach, and what nobody has tested

For a family choosing an app, the evidence is clear in one direction and empty in the other.

The Stock Market Game, a fifteen-week classroom simulation, was tested by randomising classrooms in autumn 2008: investor knowledge rose by 0.43 SD in grades 4–5, 0.45 in grades 6–8 and 0.39 in grades 9–10, with some evidence of maths gains Hinojosa et al. 2009. The study was funded by the FINRA Investor Education Foundation, the effects are for pupils who actually played, and no student behaviour was measured. In a Finnish study of 640 ninth-graders in 42 schools, pupils who used one or more of three games alongside ordinary economics lessons scored 0.34 SD higher on a knowledge test than pupils who did not, but "the effects on self-reported financial behaviors were weak", and only one of the three games could be randomised Kalmi & Rahko 2022. An online game-based course tested in a randomised trial of 2,220 students in four countries raised financial literacy by 0.313 SD; the accessible record reports no behaviour outcome Cannistrà et al. 2024. In England, Change the Game, a programme of game-based and story-based sessions that this cohort joined in Year 2, was tested across 45 primary schools randomised as whole schools: a knowledge effect of d = 0.37 after three years, with gains on the evaluators' ability, connection and mindset scales. Its behaviour measure, five questions on how pupils would spend or save in given scenarios, moved by d = 0.01, and the evaluators list "not effective in impacting these outcomes" among the possible readings King's College London 2025. The evaluation was commissioned by the charity that runs the programme.

The pattern holds across every children's study we could find: games raise what children know by roughly a third to a half of a standard deviation, around the average for classroom programmes, and none has shown a convincing change in what children do. That is partly a measurement problem; as the Finnish authors note, "student behaviors are hard to measure for under-aged students who do not make important economic decisions independently" Kalmi & Rahko 2022. It is also a design gap: a 2025 review of 45 studies of gamified personal finance found that only 9 of the 22 with formal statistical tests had a non-game comparison group Ahmad et al. 2025.

For a product like Earnwick, the supportable claim is therefore the narrower one: play is a tested way to teach the ideas, and whether it changes habits has not been tested, for Earnwick or for any game we found. None of the studies above tested Earnwick.

Where 15-year-olds stand

In PISA 2022, the OECD's test of 15-year-olds in 20 countries and economies, the average across the 14 OECD members taking part was 498, and 17.9% of their students fell below Level 2, the baseline at which a teenager can apply basic money concepts to everyday situations OECD 2024.

The share below that baseline runs from about one in nine in Denmark to nearly half in Malaysia; the chart marks the four systems that missed a sampling standard. The same survey shows what schools cover: about seven in ten students had learned "wage", "budget" and "bank loan" at school in the past year and still knew them, but only a quarter could say the same of "compound interest" and fewer than one in five of "diversification" OECD 2024. The United Kingdom and Türkiye did not take the financial-literacy option.

Who gets taught at all

Whether a child receives any of this depends mostly on where they live.

In the United States, by Next Gen Personal Finance's count, which includes only states that make every high-school student take a standalone personal-finance course, eight states guaranteed one in 2020, 17 in 2022 (16 in NGPF's own 2024 report) and 30 by 2025; as of 20 October 2026, 11 of the 30 have fully implemented it NGPF. The Council for Economic Education counts 39 states "requiring" personal finance in 2026 because it includes content embedded in other courses CEE 2026. NGPF is a curriculum provider with a stake in the count, which is why the chart names it.

In England, the citizenship programme of study published in 2013 lists money, budgeting and managing risk at key stage 3 and credit, debt, savings and pensions at key stage 4 DfE 2013. What children recall is thinner: in the Money and Pensions Service's 2022 survey of 4,740 children aged 7 to 17, 33% remembered learning about money at school and finding it useful, 24% had received the key elements of a financial education at home, and 47% had either, in practice unchanged from 48% in 2019 MaPS 2022.

In Türkiye, financial literacy is one of nine literacies in the Türkiye Yüzyılı Maarif Modeli announced on 26 April 2024, a thread running through every subject rather than a separate course, phased in from preschool, grade 1, grade 5 and grade 9 MEB 2024. For 24 March 2026 the ministry scheduled a common financial-literacy lesson in the first period of every primary, middle and high school, prepared with the Capital Markets Board and including warnings about illegal betting and online gambling MEB 2026. Since a presidential circular in May 2025, 22 May is Financial Literacy Day Circular 2025/9. None of this has been evaluated yet. The closest Turkish trial is adjacent rather than financial: an eight-week classroom curriculum on patience, randomised across 37 Istanbul primary schools in 2013, made children's choices markedly more patient a month later, with effects lasting about a year and, for some groups, nearly three Alan & Ertac, J-PAL. Adults scored 12.5 out of 21 in the OECD's 2016 survey, against an all-country mean of 13.2 OECD/INFE 2016; Türkiye has not joined the later surveys or the PISA money test, so there is no national measure for children.