The Case for K12 Domestic Talent Systems

Carina Initiatives and the Polynera Fund are private philanthropic foundations dedicated to expanding the pipeline of future scientists and innovators. We view the early identification and development of exceptional STEM talent as both a national imperative and an achievable investment with extraordinary social and economic returns. Below, we outline our rationale and proposed policies to make this vision possible.

Introduction

America is underinvesting in its most promising STEM talent, leaving enormous potential untapped. Third graders who score in the 99th percentile in mathematics will account for a disproportionate share of future scientific breakthroughs. Yet among these students, invention rates vary nearly 20-fold depending on the support they receive. Students from lower-income and middle-income families without specialized programs invent at 3 in 1,000. Students from high-income families who receive acceleration and enrichment support invent at 54 in 1,000. We aren’t supporting our most talented kids because we lack the resources (we already spend $15,000 per student on education), but because we've chosen to deprioritize these students rather than provide the support they need. This choice wastes enormous human potential and leaves breakthrough discoveries unrealized. Closing this talent development gap could increase innovation by 2-55% and add $2-52 trillion to GDP over 65 years, making it among the highest-return investments available to policymakers.

America has successfully prioritized talent development before, and we know it works. Following Sputnik in 1957, the U.S. demonstrated what systematic talent investment could achieve. The National Defense Education Act (1958) and early NSF programs modernized curricula, raised standards, expanded access to advanced math and science, and seeded pipelines that produced generations of engineers and scientists. The NSF and other agencies funded selective pipeline programs and scholarships for advanced students at the K12, undergraduate and graduate levels. This strategic investment secured American scientific leadership for decades.

Research on innovation suggests this focus on talent was wise. Breakthroughs in fields come from a small number of innovators at the very top of their disciplines. Research shows that innovation is highly concentrated among top talent - the top 1% scientists account for over 20% of citations, and the top 1% of inventors are five to ten times as productive as the average inventor. In fields like AI, the concentration is even starker: just a few hundred researchers are setting the agenda for the entire discipline. In short, the discoveries that redefine industries and drive economic prosperity come from cultivating and supporting those at the very frontier of talent.

However, we have been neglecting our K-12 talent systems since the 90s.

  • The only recurring federal program for gifted education, the Jacob K. Javits Gifted & Talented Students Education Act, is funded at just ~$16.5M annually, and the FY 2026 budget proposes eliminating it entirely.
  • Many states have cut gifted programs, with 1 in 4 schools eliminating these programs between 2012 and 2018: California eliminated categorical GATE funding in 2013–14; Texas reduced per-pupil gifted funding by 8–10% post-Recession; Massachusetts and Connecticut cut programs multiple times in fiscal downturns.
  • NSF has redirected its K–12 priorities away from talent identification and top STEM pipelines (e.g. discontinuing the Young Scholars Program).

Neglect of domestic talent systems has led the U.S. to increasingly depend on international talent, particularly in frontier fields like AI. With immigration policies becoming more restrictive, the U.S. will need to cultivate our own pipeline of high-ability students.

Other nations offer models worth studying. Countries like China, Russia, and Singapore have demonstrated that systematic talent identification and development programs can successfully nurture mathematical ability at scale. In China, for instance, selective high schools, national math Olympiad pipelines, and elite universities are linked seamlessly. Russia continues to invest in specialized math schools and Olympiad systems. These programs show what's possible when talent development becomes an institutional priority, and they suggest that America's current approach leaves significant potential on the table.

Landscape

Mathematical ability has been an extraordinarily strong predictor of future innovation across disciplines. Raj Chetty’s Lost Einsteins study (2017)  shows that people who were in the top 5% of math ability at 3rd grade account for 25% of all patents. The Study of Mathematically Precocious Youth (SMPY) reinforces this connection, demonstrating that early math achievement predicts not just patents but also publications, doctorates, and other markers of innovation, with no plateau in the relationship up through 3 standard deviations above the mean.

Yet the U.S. is missing most of its innovation potential. Among students who show 99th percentile math ability in early grades, innovation rates vary dramatically based on the supports they receive:

  • 99th percentile students (3rd grade) from households below the 80th percentile in income become inventors at 3 in 1,000
  • 99th percentile students (3rd grade) from households above the 80th percentile in income become inventors at 6.5 in 1,000
  • 99th percentile students (8th grade) who are identified and receive out-of-school supports through SMPY become inventors at 54 in 1,000

Closing the talent development gap would likely have enormous economic returns through increased innovation.

We think of the lower bound on returns to investing in 99th percentile math achievers as closing the low/middle income and high income gap and the upper bound as closing the gap between all kids and fully supported SMPY kids.

  • Closing the gap in innovation between low/middle income vs. high income 99th percentile students through acceleration could increase overall innovation by 2%.
  • Closing the gap in innovation between all 99 percentile kids vs. SMPY 99th percentile kids (i.e., those who received acceleration and enrichment support) could increase overall innovation by 55%.
  • These increases in innovation could produce between 0.01pp and 0.32pp increase to GDP over 65 years, or $2T to $52T.

Our hypothesis is that systematically identifying high-ability students and providing them with high quality acceleration and enrichment could meaningfully boost innovation. 

Evidence from SMPY demonstrates that the students who received these supports went on to innovate at extraordinary rates. Research suggests that these interventions are highly impactful, yet they are seriously underdeployed. Thus, expanding best practices in talent identification, acceleration, and enrichment across the country should be a national priority.

Identification

We can't effectively support talent without first identifying it.

The United States has no universal screening for intellectually gifted children, despite research showing that screening increases the number of identified students by 1.4x for high-income students and 3x for low-income students.

Universal screening is far more effective than referral-based identification, which misses large numbers of talented students, especially from low-income backgrounds. In one of the most rigorous studies to date, Card and Giuliano (2016) found that implementing universal screening in a large Florida district increased the share of gifted third graders from 3.3% to 5.5% without lowering standards.

Only 13 states1 currently require universal screening. If scaled nationally, this would mean identifying approximately 850,000 additional gifted students across grades 3–12 who would otherwise remain overlooked.

In the out-of-school space, The Center for Talented Youth (CTY) and similar talent identification programs aspired to create a national talent screen and offer challenging STEM coursework. However, their fee-based opt-in model leaves most gifted students, especially from low- and middle-income backgrounds, undiscovered. Recently, CTY and its partners have faced funding cuts and have either shrunk or closed altogether.

Acceleration

Even when students are identified as exceptional, public resources to cultivate their talent do not exist. Third graders at the 99th percentile in math are typically about 3 grade levels ahead of their classroom peers, meaning they are not being challenged. Around 50% of U.S. public high schools do not offer calculus. Grade-level subject acceleration is rare and often discouraged, with few laws in place to ensure access.

The consequences of this failure are stark. Based on the NAEP distributions, we estimate that low and middle-income students make up 29% of the 99th percentile in math achievement in 4th grade. By 8th grade, that number drops to 13%. Between 4th and 8th grade, we lose more than half of our low and middle-income top students, likely because they aren’t fully challenged and supported.

Acceleration is consistently shown to be the most effective intervention for high-ability learners. It is essentially a form of “Teaching at the Right Level” for students whose level is well above grade level. Meta-analyses find average learning gains of 0.50-0.70 standard deviations from acceleration, amongst the most impactful learning interventions in existence, exceeding even tutoring's 0.35-0.40 SD gains at scale. Card and Giuliano's research demonstrates that accelerated tracks raise test scores, especially for low-income students, with the largest gains accruing to the most advanced students.

Yet acceleration has faced sustained opposition for three decades. The detracking movement began with Jeannie Oakes's 1985 "Keeping Track" and was institutionalized through state mandates in California (1987) and Massachusetts (1993), receiving endorsement from the National Governors Association and major education organizations in 1993. The No Child Left Behind Act (2002) fundamentally restructured federal priorities by creating accountability solely for proficiency, which led to significant cuts in gifted funding. District eliminations accelerated in the 2000s-2010s: Washington D.C. (2005), Evanston, IL (2010), and San Francisco (2014). California eliminated categorical GATE funding in 2013. California's 2023 Mathematics Framework discourages tracking and questions the concept of mathematical talent on equity grounds, while Cambridge, MA eliminated accelerated pathways in 2017-2019 with promised reinstatement repeatedly delayed to 2026-27.

These efforts led to a steep decline in gifted and tracking programs across the country. Between 2012 and 2018, 1 in 4 schools eliminated their gifted program. A RAND report from 2024 shows that only 7% of elementary schools and 39% of middle schools nationwide track by math ability.

As states abandoned acceleration, the private sector stepped in. Companies like Art of Problem Solving, Russian School of Math, and Think Academy have opened over a hundred learning centers and offer classes in-person and online to nearly a hundred thousand students. These programs disproportionately serve children of STEM professionals. 90% of these learning centers are in 8 metropolitan areas with significant tech presence: San Francisco Bay Area, New York, Boston, D.C., Los Angeles, Chicago, Seattle, and Houston.

These programs achieve results far beyond the best in-school programs. Companies like Art of Problem Solving have developed accelerated math curricula that cover 50% more content than Common Core at a significantly deeper level of rigor. Students are tracked exclusively by ability. In these programs, for many students, the question isn't Algebra or Geometry by 8th grade, it's Algebra or Geometry by 5th or 6th grade.

The results are striking. Math competitions are the best proxy measure we have for accelerated math achievement. Although only about 10% of students live within 25 miles of one of these centers, areas with high center density produce the vast majority of top competition performers.

  • Among the 115 top scorers in the 2025 Kangaroo Math competition (5th grade), approximately 25% attend one of the acceleration centers, 50% or more attend some form of learning center, and 60% live within 25 miles of an acceleration center.
  • By high school, approximately 75% of the 2025 USA Junior Math Olympiad qualifiers live within 25 miles of an acceleration center.
  • At the very top level of high school math, all 66 U.S. International Math Olympiad team members over the last 11 years have taken Art of Problem Solving courses, winning 65 medals and taking 380 courses in total.

While it's possible that 75% of our country's math talent naturally concentrates in geographies representing only 10% of students, the NAEP data suggests otherwise. We can directly observe low and middle-income students falling out of the top 1% between 4th and 8th grade, and these are precisely the students unlikely to access acceleration programs due to prohibitive costs and geography.

It is of the utmost strategic importance to U.S. scientific endeavor that we get all of our top math students in top programs. Right now we are neglecting at least 30% of them (the low and middle-income students who start in the 99th percentile but disappear) and potentially as many as 90% if geographic access to quality acceleration is the key determinant of reaching potential.

Enrichment

While acceleration prepares students to contribute at the frontiers of innovation, enrichment develops the interest and motivation to pursue frontier work.

Elite math camps epitomize talent enrichment programs. Math camps immerse students in extended, intensive problem solving that reveals deep connections and patterns in mathematics, with constant support from counselors and faculty living alongside them. The programs fundamentally change how students see math and encourage them to pursue the most challenging questions in the field. Students who engage in these activities often go on to pursue PhDs. 25% of PROMYS alumni earn PhDs in mathematics, and 22% of International Math Olympiad participants, nearly all of whom participated in similar experiences, earn PhDs in mathematics.

Enrichment works through two mechanisms: surrounding students with talented peers and connecting them with expert mentors.

High-ability students excel when surrounded by a community of peers with similar talent. Glenn Ellison's studies of the American Mathematics Competitions (AMC) show that students are far more likely to persist and advance in math when they are part of local clusters of high performers. The absence of such peer communities contributes to attrition, particularly among women. These findings underscore that gifted students need environments where advanced ability is the norm rather than the exception.

Mentorship is equally critical in sustaining innovation trajectories.  The Lost Einsteins paper shows not only is there strong correlation between the historical patent rate of a commuting zone and the number of inventors that a zone produces, but innovation among parents' coworkers leads to a 10 times larger increase in innovation in exactly the same technology class. Calaway also highlights that early mentorship of exceptional math students by high school math teachers causes them to win honors, attend selective universities, major in STEM fields, and have careers that disproportionately spur economic growth.

Despite the importance of enrichment in talent development, it has been highly neglected over the last 30 years.

Math camps in the U.S. emerged directly from Cold War anxieties following Sputnik's 1957 launch, when programs like Arnold Ross's summer mathematics program began expanding with new NSF funding aimed at strengthening American scientific competitiveness. Through the post-Sputnik push, the NSF funded summer programs via SSTP (Summer Science Training Program). In 1988, as the Cold War was ending, concerns about the scientific "pipeline" led to SSTP's resurrection as the Young Scholars Program, which funded 114 summer programs reaching around 5,000 students annually. However, the NSF ended Young Scholars in 1996 despite acknowledging its success, due to changing strategic priorities and legal troubles over a lawsuit challenging minority-focused Summer Science Camps.

The demise has left the U.S. with a shortage of math camps. The programs that have survived like Ross, PROMYS, Hampshire College Summer Studies in Mathematics, and Canada/USA Mathcamp cannot support demand. The camps receive 10 qualified applications for every seat and run at such slim margins with so few qualified instructors that they are unable to expand to meet demand.

The story in the math competition space is similar. The American Mathematics Competitions, founded in 1950 as a Cold War-era effort to identify and develop mathematical talent, grew from a local New York contest of 6,000 students to a national program with over 300,000 annual participants by the 2000s, serving as the gateway to the International Mathematical Olympiad and influencing college admissions at elite institutions like MIT and Carnegie Mellon. However, Ian Calaway's research documents a troubling decline of 50% in school participation since the 1980s. This decline is primarily among low-income schools, where exceptional math students go unidentified due to lack of teacher-mentors who facilitate these competitions, with missing students disproportionately concentrated in rural states.

Opportunities

When the Study of Mathematically Precocious Youth (SMPY) launched in 1971, its vision was to identify youth with exceptional mathematical reasoning ability and provide them with accelerated educational opportunities to help them flourish. It was an active intervention, not merely observational, designed to discover optimal mechanisms for promoting intellectual development among the gifted.

SMPY launched a national talent search for 12-13 year olds, partnering with school districts to identify students performing at the very top of grade-level achievement tests. Families received letters explaining their child had scored unusually high and was invited to participate.

Since no acceleration programs existed outside schools at the time, SMPY created their own, building and running educational services for eight years before transferring them to Johns Hopkins' Office of Talent Identification and Development (later CTY).

A group of funders, including Carina and Polynera, have come together to reimagine SMPY's vision for today's landscape and generate data on its effectiveness.

National Math Stars (NMS) was created in 2024 to ensure students with extraordinary mathematical talent have access to the resources, community, and guidance they need to fully realize their potential.

Like SMPY, NMS partners with schools, districts, and states to identify top performers by awarding the top 2% of math achievers based on standardized tests. Awardees are then invited to apply for selection into the program, which includes talent development supports worth ~$15K annually. Selection is based on demonstrating 1 in 1000 numerical, logical, or spatial reasoning skills.

Unlike SMPY, NMS runs a leaner model by leveraging the acceleration and enrichment ecosystem that has emerged over the last 20 years. This offers distinct advantages: NMS maintains low operational overhead as essentially a talent identification and scholarship administration organization, while leveraging providers' proven expertise in working with younger students, achieving extraordinary outcomes, and teaching virtually.

Early results from NMS are promising:

  • Program launches in Texas and the Midwest have surfaced talent from 100% of schools in Iowa, 50% in Texas, and 12% across five midwestern states.
  • Program participants in the 2025 cohort average 148 IQ and $70,000 household income.
  • After nine months, students outperformed peers with the same starting math achievement on NWEA MAP standardized tests by 1.2 standard deviations in learning gains.

NMS has started to demonstrate that acceleration and enrichment providers have become more effective and accessible than ever. The early results are promising, but NMS currently reaches just hundreds of students when we should be serving thousands. Policy change is the only path to making these resources available to all talented kids across the country.

Policy Recommendations

We propose a tiered approach to talent development, with interventions scaled to different levels of mathematical ability. Each tier addresses critical gaps in identification, acceleration, and enrichment while building on proven models.

**NOTE: Impact analysis in this section is still in early stages. We are fairly confident that the NPV of potential returns is in the 10s of trillions of dollars and the NPV of costs is in the 10s of billions of dollars, but there’s more work to be done to achieve better precision.

In-School Acceleration for Top 10%

Policies:

  • Universal screening: Mandate gifted screening for students scoring above the 90th percentile on standardized math tests, beginning in elementary grades (2nd grade) with periodic re-testing (5th, 9th grades)
  • Automatic enrollment: Automatic enrollment for identified students in accelerated gifted programs using research-backed acceleration curricula
  • School accountability: Include 90th percentile student achievement in accountability metrics

Impact:

  • Identification gains: Universal screening increases identification by 1.4x for high-income students and 3x for low-income students, potentially identifying 850,000 additional gifted students nationally across grades 3-12
  • Learning gains: Enrolling all top 10% students in acceleration programs would yield 0.5-0.7 standard deviation gains in learning
  • Economic growth: Given the connection between learning gains and economic growth, and the skew of those gains attributable to top scorers, we estimate significant gains among the top 10% could lead to a 0.16 percentage point increase in annual GDP growth

Implementation Challenges:

  • Political and systemic barriers: Requires categorical funding protection (recently out of favor) plus extensive curriculum adoption and teacher training that could take decades to deploy nationwide
  • Current landscape: The National Association of Gifted Children ($2M budget) leads advocacy efforts, with support from think tanks including Fordham Institute, AEI, Brookings, Manhattan Institute, and EdChoice. Total annual spending on gifted education advocacy is likely under $10M.

Acceleration Scholarships for Top 1%

Policies:

  • State-level scholarships: Means-adjusted $2,000 scholarship accounts enabling enrollment in online accelerated courses for all students at the 99th percentile, delivered through scholarship-granting organizations (SGOs), state education agencies, or directly to families via ESAs
  • Course waivers: Allow students to opt out of standard school math courses and gain credit for advanced curriculum completed outside school
  • Individualized Gifted Education Plans: Require schools to develop specialized learning plans for academic outliers (modeled on IEPs for special needs students)Based on our analysis, the impact to innovation and the economy would be very large.

Impact:

  • Innovation gains: If this was implemented across all states, we would expect to see roughly a 10% increase in yearly innovation output over the next 65 years..
  • Economic growth: This increase in innovation would cost roughly $700M a year and in turn produce a 0.06pp increase in growth over the next 65 years, resulting in about an ~400:1 return on government spend.

Implementation Challenges:

  • Feasibility: Most readily implemented through state-level ESAs via minor legislative changes in states with existing ESA programs
  • Scalability: Acceleration providers already serving tens of thousands could scale to serve hundreds of thousands
  • Funding: The total national cost is approximately $700M annually, with state-level costs in the tens of millions. The main challenge is whether states would be willing to allocate new funds or have to reallocate existing education budgets.
  • Current landscape: To our knowledge, no organization is currently working on this set of policies.

Acceleration and Enrichment for Top 0.1%

Policies:

  • National scholarships: Means-adjusted $15K annual scholarships for all students demonstrating 1 in 1,000 numerical, logical, or spatial reasoning ability, delivered through scholarship-granting organizations
  • Enrichment funding: Revive support similar to the discontinued NSF Young Scholars Program through direct grants to proven enrichment providers: non-profit independent camps, like Ross, and university-based acceleration programs, like UMTYMP

Impact:

Based on our analysis, the impact to innovation would again be very large.

  • Innovation gains: A national implementation could increase yearly innovation output by 12%-15% over the next 65 years.
  • Economic growth (aggregate): This $500M investment in talent could translate to a 0.08pp increase in economic growth, resulting in 650:1 returns on investment.

Implementation Challenges:

  • Funding: Annual cost of approximately $500M requires partnerships between federal agencies and large philanthropists who see innovation and growth as strategic priorities.
  • NSF institutional capacity: NSF needs to build or rebuild capacity to fund K-12 programs and organizations directly
  • Current landscape: National Math Stars ($5M budget) is the primary organization working to scale solutions  in this space

Conclusion

We have neglected talent for 30 years, and we’re leaving enormous innovation potential unrealized. We know what needs to be done to rebuild our talent pipelines. We need to identify top math students, accelerate them in school, and make sure they have access to high quality enrichment. The investment case is clear. A robust talent system would cost roughly $1-2B a year and would be amongst the highest return investments the government can make. This is totally tractable; it doesn't require systems change, especially at the top 1 and .1 percent.

What we need is for institutions who care about innovation, growth, and scientific progress to come together and make this a priority. We need the NSF to prioritize future development of scientists over broad-based education improvement. We need science funders to allocate a percentage of their budgets to future scientist talent development. We need growth and progress thinkers to think more deeply about domestic talent development. We need education funders to balance their focus on raising the floor with cultivating extraordinary potential. And we need the Department of Defense to treat domestic talent pipelines as a strategic priority, recognizing that scientific breakthroughs emerge from decades-long investments in human capital.

The path forward is clear and achievable. What we need now is public and private funders to make talent development a shared priority and invest steadily over the next decades.