The American Education and Innovation Machine: Strengths, Structural Flaws, and the Research Pipeline

A deep, cited analysis for Bucket Foundation

education-atlas · `docs/deep/01` · companion to `docs/EDUCATION_PROBLEMS.md` (the global quantitative atlas). Where the atlas measures the world against SDG 4 benchmarks, this document goes structural and goes deep on a single country — the United States — and on the one thing the global indicators barely touch: how a nation manufactures its researchers. Every claim below is sourced inline; numbers are traceable to NAEP/NCES, OECD PISA, the National Science Board, the Economic Policy Institute, the National Student Clearinghouse, FREOPP, EdBuild, and peer-reviewed literature. The point of view is Bucket's: education reform that is not grounded in evidence is opinion.

0. Why the United States is the right case to dissect

The global atlas (EDUCATION_PROBLEMS.md) makes the macro case: the deepest problem on earth is a learning crisis (48.3% of 10-year-olds worldwide cannot read a simple text), it is wildly unequal by region and income, and half the world's governments underfund education below the agreed 4%-of-GDP floor. That document also flags, in its honest-limitations section, the problems no global indicator captures — relevance, pedagogy, credentialism, and the purpose of education itself. The United States is the country where those unmeasured problems become most visible and most instructive, for one reason: the U.S. is simultaneously the most successful and the most dysfunctional education system in the developed world. It has the planet's strongest research universities and its most expensive, most unequal, and increasingly debt-financed pipeline into them. It is the cleanest available test of the thesis that money and access are necessary but not sufficient — quality, equity, and purpose are separate, harder problems (a pattern the atlas already detected in South Africa: near-universal access, near-bottom learning).

This document is organized around five questions:

1. What is structurally wrong with American K-12?
2. What is structurally wrong with American higher education?
3. How does the U.S. manufacture researchers — and where does that pipeline fail? (the key angle for Bucket)
4. What does the system actually optimize for, versus what learning needs?
5. Honestly — what does the U.S. get right, and what specifically can an open-knowledge project address?

1. K-12: a strong average resting on a broken foundation

1.1 The funding model is the original sin: local property tax → manufactured inequity

The United States is nearly alone among rich nations in financing public schools substantially from local property taxes. The mechanism is simple and its consequences are structural: a child's school budget is tied to the assessed value of the real estate around their home, so affluent neighborhoods produce well-funded schools and poor neighborhoods produce poorly funded ones — "from the word go," as one analysis of the model puts it (American University, School of Education).

The most rigorous quantification is EdBuild's 2019 study: U.S. school districts that predominantly serve students of color receive $23 billion less in combined state and local funding each year than predominantly white districts, despite educating roughly the same number of students — about $2,200 less per student, affecting roughly 10 million children (EdBuild, *$23 Billion*; The 74). Racially concentrated nonwhite districts receive on average $11,682 per student versus $13,908 in concentrated white districts. Critically, the gap is not just a poverty artifact: poor white districts received nearly $1,500 more per student than districts serving poor nonwhite children (The 74). Gerrymandered district borders compound it. This is inequity that the system produces by design, not inequity it fails to fix — a different and more damning category.

1.2 The flat, stagnant — and now declining — learning signal

For all that the U.S. spends (and it spends a lot — see §5), its measured learning has been flat for decades and is now falling. The single most authoritative domestic measure is NAEP, the National Assessment of Educational Progress ("the Nation's Report Card"), run by NCES.

The headline trend is a post-2020 decline that has not recovered. In the 2024 main NAEP, average reading scores fell 5 points from 2019 (pre-pandemic) levels in both 4th and 8th grade, and math dipped 3 points in 4th grade and 8 points in 8th grade versus 2019 (Urban Institute; K-12 Dive). Five years after the pandemic, the nation remains below its 2019 scores in both subjects and both grades (NAGB / Nation's Report Card). For 12th graders, a record-high share of the Class of 2024 scored "below basic" in both math and reading — the worst in the history of the assessment (NAGB; The 74).

The most important structural fact inside that decline is divergence: the losses are concentrated at the bottom. Between 2022 and 2024, learning losses for students in the bottom 10% grew 70% larger, and for the bottom 25% grew 25% larger (Hechinger Report). The modest 4th-grade math gains were driven entirely by high-achieving students; the reading deterioration was driven by low-achieving students. The American school system, under stress, widens its own gaps rather than closing them.

A genuine bright spot, in fairness: the 2026 NAEP Long-Term Trend assessment (released June 2026) showed 9-year-olds gaining 4 points in both reading and math since 2022, recovering to pre-COVID reading levels — and the gains were powered overwhelmingly by the lowest performers (the 25th-percentile group rose 7 points in math, 6 in reading) (NPR; The 74). But the same release showed 13-year-olds stagnating (K-12 Dive). Recovery, where it exists, is partial, age-specific, and fragile.

1.3 Internationally: middling, despite paying for excellence

On PISA 2022 (OECD, 15-year-olds), the U.S. scored 465 in math — below the OECD average of 472 and statistically lower than 25 of the 80 participating systems — while scoring above the OECD average in reading (476 avg) and science (485 avg), trailing only 5 systems in reading and 9 in science (NCES PISA 2022 results; OECD Country Note: United States). The U.S. is not failing on PISA — but for a country with by far the highest per-pupil spending in the OECD, average math performance behind dozens of countries is a poor return. The reading/science strength also masks the internal inequality: U.S. PISA means hide one of the widest socioeconomic achievement spreads in the developed world.

1.4 The factory model, high-stakes testing, and curriculum narrowing

Three pedagogical pathologies compound the funding problem:

• The age-batch "factory" model. Children are sorted into grades by birth year and moved through a fixed-pace conveyor regardless of whether they have mastered the prior material — the assembly-line design inherited from the 19th-century industrial era. A student who hasn't mastered fractions is advanced anyway; a student who mastered them in October waits until June. This is the structural reason the NAEP "below basic" tail keeps growing: the system advances students past gaps rather than closing them. (This is the "pedagogy" problem the global atlas explicitly says no indicator captures — §5 of EDUCATION_PROBLEMS.md.)

• High-stakes standardized testing. Since No Child Left Behind (2002), federal accountability has tied school fates to standardized test scores. The well-documented downstream effect is curriculum narrowing: instructional time shifts toward tested subjects (reading, math) and away from science, social studies, art, music, and recess, and toward test-preparation rather than understanding. The system optimizes the metric, not the learning the metric was meant to proxy — a textbook case of Campbell's/Goodhart's law in education policy.

• Measuring sorting, not mastery. High-stakes tests are designed to rank students and schools, not to certify that a given child has mastered a given concept. This is the same optimization target that recurs at every level of the U.S. system (§4).

1.5 The teacher problem: a record and widening pay penalty

You cannot run a quality system while systematically underpaying the people who deliver it, and the U.S. does exactly that. The teacher pay penalty — the gap between weekly wages of public school teachers and other similarly educated college graduates — hit a record 26.9% in 2024: teachers earned just 73.1 cents on the dollar relative to comparable professionals, up from a 6.1% penalty in 1996 when teachers earned 93.9 cents on the dollar (EPI; EPI press). Teachers are paid less than comparable graduates in every state, from −10.0% in Rhode Island to −38.5% in Colorado. Teachers' real weekly wages actually fell 6.1% since 1996 while other graduates' rose 26.9%. The predictable consequences are recruitment and retention crises and chronic vacancies, especially in high-poverty, STEM, and special-education roles — and shortages fall hardest on the same low-income districts the funding model already starves.

K-12 summary. The U.S. has built a system whose central financing mechanism manufactures inequality, whose pedagogy advances students past their gaps, whose accountability regime optimizes a sorting metric over mastery, and which underpays its teachers more than at any point in three decades. The result is a flat-to-declining learning signal whose declines concentrate among exactly the students the system was supposed to lift.

2. Higher education: a debt-financed signaling engine

2.1 The cost explosion and the $1.7-trillion debt overhang

American higher education has gotten dramatically more expensive faster than incomes. Average published tuition and fees at four-year institutions rose from $5,740 (in-state public) / $24,840 (private nonprofit) in 1994–95 to $11,610 / $43,350 in 2024–25 (EducationData.org). The financing of that increase shifted onto students: federal student loan debt stood at ~$1.693 trillion as of late 2025 (about $1.84 trillion including private loans), spread across 42.8 million federal borrowers, with the average federal balance at a record ~$39,600 per borrower (EducationData.org; Motley Fool / Federal Reserve). This is the macroeconomic signature of a system that converted a public good into a private debt.

2.2 Credentialism and the signaling problem: degrees decoupling from learning

The deeper critique is not cost but what the cost buys. A large strand of economics (Spence's signaling theory; Bryan Caplan's The Case Against Education) argues that much of the wage premium of a degree reflects signaling — the diploma certifies pre-existing traits (intelligence, conscientiousness, conformity) to employers — rather than human capital actually built in the classroom. The global atlas names this exactly: "credentialism (degrees decoupled from competence)" as a central, unmeasured problem (EDUCATION_PROBLEMS.md §5). When the credential, not the learning, is the product, the system can keep raising prices on a good whose intrinsic learning content is not rising — which is precisely the cost/learning divergence we observe.

2.3 Return-on-degree variance: the median works, the tail is brutal

The "is college worth it?" question has a precise data-grounded answer: it depends enormously on the program, and a large minority of programs lose money. FREOPP's comprehensive ROI analysis finds the median bachelor's program returns ~$160,000 in lifetime value — but 28% of bachelor's programs have negative ROI once you adjust for the real risk of non-completion, and roughly a third of all federal Pell Grant and student-loan dollars pay for programs that deliver no positive return (FREOPP). Engineering, computer science, nursing, and economics can return $1M+; many arts, humanities, and low-completion programs return zero or less. The system sells a single signal ("college") that conceals a distribution running from life-changing to financially destructive — and it lends federal money against the bad tail.

2.4 The completion crisis and its inequity

A degree only pays if you finish, and a large share don't. The six-year completion rate for the fall-2018 cohort reached 61.1% — the highest on record, but still meaning nearly one in three students has no degree six years later (National Student Clearinghouse via Higher Ed Dive). The gap is starkly stratified: 48.2% of students from the lowest-income neighborhoods completed within six years versus 75.8% from the highest-income — a 27.6-point income gap (NSC Research Center; Higher Ed Dive). First-generation students complete at 69% vs. 86% for continuing-generation peers (Inside Higher Ed). Non-completion is the worst-case financial outcome: debt incurred, no credential earned — and it falls hardest on exactly the students least able to absorb the loss.

2.5 Administrative bloat and the adjunctification of teaching

Where did the tuition go? Disproportionately into administration, not instruction. Benjamin Ginsberg's The Fall of the Faculty documents that from 1975 to 2005 administrators grew 85% and administrative staff 240%, far outpacing faculty and student growth, to the point that administrators now outnumber faculty on essentially every U.S. campus (Academe Blog; Claremont Review).

Simultaneously, the teaching workforce was casualized. Tenure-track positions have fallen to roughly 36% of faculty, with adjuncts and other non-tenure-track instructors — many without benefits, job security, or research support, some paid little enough to qualify for food stamps — now making up the majority; adjunct and non-tenured faculty number roughly 1 million of 1.5 million total faculty (Academe Blog). The institution shifted resources away from the people who teach and toward the people who administer — while charging students more. This is the higher-ed mirror of the K-12 teacher pay penalty: the core knowledge-transmission labor is the part the system economizes on.

Higher-ed summary. The U.S. built a positional, debt-financed credentialing machine: prices rose far faster than the learning content or the incomes paying for them; a large minority of programs deliver negative returns; a third of students never finish (and the non-finishers are disproportionately poor and first-generation); and the money flowed into administration while the teaching labor was casualized.

3. The innovation and research pipeline — the key angle

This is where the U.S. story is most paradoxical and most relevant to Bucket. The same country with the dysfunctions above runs, by a wide margin, the best research enterprise on earth — and it runs it on a pipeline that is structurally cruel and increasingly inefficient at the very stage where new researchers are made.

3.1 The strength is real — say so plainly

The U.S. is still #1 in research, and not by a little. Total U.S. R&D reached ~$940 billion in 2023, the largest national R&D enterprise in the world, per the National Science Board's Science & Engineering Indicators (AAU summary of NSB; NCSES/NSB). U.S. universities dominate global research rankings, are the top destination for international scientific talent, and produce highly cited, highly collaborative science. Universities account for roughly 72% of U.S. Nobel-Prize-producing institutional output, with government and enterprise at ~11% each (Scientometrics, Springer). The basic-research engine — federally funded, university-performed (federal government funds ~18% of total R&D but most of the basic and university research) (NCSES/NSB) — has underwritten the internet, GPS, mRNA vaccines, modern AI, and most of the 20th-century scientific canon. Any honest reform thesis must start by conceding that this part works extraordinarily well.

3.2 But the funding is hyper-concentrated — and our own research-atlas proves it

Bucket's own research-atlas (75 funders, 1.67 million grants, ~$1.04 trillion in trackable awards) makes the concentration quantitative and undeniable (research-atlas/docs/LANDSCAPE.md):

• Two U.S. agencies dominate global science funding. In the atlas, NIH ($570B across 1.13M grants) and NSF ($108B across 202K grants) together account for ~$678B — the vast majority of all trackable research money, dwarfing the EU's CORDIS ($203B), Gates ($99B), Wellcome ($25B), and UKRI ($26B).
• The money lands in a narrow elite of institutions. Of the atlas's top 25 recipient organizations by dollars received, every single one is in the United States — Johns Hopkins ($13.5B), UCSF ($11.0B), U-Washington ($10.4B), Penn ($10.0B), Michigan, Stanford, Duke, UCSD, UCLA, Pittsburgh, Yale, WashU, and a handful of NIH intramural divisions and hospitals. American research excellence is concentrated in perhaps a few dozen institutions; the long tail of U.S. colleges does little to no federally funded research.

This is a double-edged finding. Concentration buys world-leading depth at the top — but it also means the right to do serious research is gated by which institution you can get into and stay funded at, an access problem that mirrors the K-12 and completion inequities above one level up. It is the "who gets to do research" question made concrete.

3.3 The PhD bottleneck: training far more researchers than the system will employ as researchers

The pipeline's central structural flaw is an oversupply of trained researchers relative to research careers. The numbers are stark:

• Only an estimated ~14% of biological-sciences PhDs hold a tenure-track faculty position 5–6 years after graduation (PMC: predoctoral publication / career outcomes).
• In engineering, the average probability of landing a tenure-track position over a 16-year window was ~12.4% — roughly 1 in 8 (PMC: engineering faculty competition).
• Tenure-track positions account for only ~15% of postdoc career outcomes, and about 40% of postdocs leave academia (Science/AAAS).

A widely cited modeling paper framed this as academia's "basic reproductive number": each faculty member trains far more PhDs than there are faculty slots to absorb, so the system is structurally over-producing relative to its own demand for researchers (PMC: *R0 in Academia*). The PhD is, for the large majority, training for a job that does not exist — which can be fine if the destination outside academia is good, but the system is not designed or honest about that; it is implicitly an apprenticeship for a faculty career most apprentices will never get.

3.4 Postdoc precarity: the holding pen

Between the PhD and the (improbable) faculty job sits the postdoc — a stage that has become a years-long, low-paid, benefit-thin holding pattern. Postdocs often earn ~$50,000–$60,000 for highly specialized PhD-level labor, with uncertain futures (Science/AAAS; Undark). The precarity is now so unattractive that PIs increasingly struggle to recruit postdocs at all — a sign the bargain has broken down from the trainee side (Science/AAAS).

3.5 Grant hypercompetition and the aging of independence

The funding side compounds the labor-side bottleneck. NIH grant competition has intensified to the point of dysfunction:

• The NIH-wide R01-equivalent success rate fell to ~13.0% in FY2025 (from 18.7% in FY2024 and 21.6% in FY2023); for first-time applicants the relevant rate is closer to ~15% (Grantsights / NIH data).
• The average age at which a scientist wins their first R01 rose from 35.7 in 1980 to ~43 (PhD) / 45 (MD) by 2016 (Science/AAAS). Researchers don't reach funded independence until their mid-forties.
• ~43% of first-time NIH grantees fail to win further funding, with the highest drop-out rate 4–5 years after the first award (PLOS One / NIAID outcomes).

Two structural pathologies follow. First, short-termism: when success rates are ~13% and renewal depends on a steady stream of publishable results, the rational researcher pursues safe, incremental, fundable projects over slow, risky, paradigm-shifting ones — the opposite of what basic science needs. Second, wasted human capital and a barrier to genius work: a system that makes researchers wait until 43 for independence, then washes out 43% of them, is not optimized to let exceptional people reach the frontier early. This is directly the population Bucket's mission targets — the small number of people who can do genius work — and the current pipeline actively gatekeeps and delays them.

3.6 Brain drain / brain gain

The U.S. research enterprise has historically been a net importer of talent — its universities are the world's top destination for international students and scientists (NCSES/NSB), a "brain gain" that subsidizes American science with foreign-trained talent. This is a genuine strength, but a fragile one: it depends on immigration policy, on the relative attractiveness of U.S. careers (eroding as precarity rises and success rates fall), and on the rest of the world not building competitive systems. The same precarity that pushes domestic PhDs out (§3.3–3.5) erodes the magnetism that brings foreign talent in.

Pipeline summary. The U.S. runs the world's best research output on a researcher-manufacturing process that over-produces PhDs ~7:1 relative to faculty jobs, warehouses them in precarious postdocs, gates funding behind a ~13% success-rate lottery that pushes independence to age 43 and washes out ~43% of first-time grantees, and concentrates nearly all serious research money in a few dozen elite institutions. It produces brilliant science despite a pipeline that wastes and delays its most capable people — which is precisely the inefficiency an open-knowledge model could attack.

4. What the system optimizes for vs. what learning needs

Pull the threads together and one pattern recurs at every level. The American education system is, end to end, optimized for sorting, credentialing, and compliance — not for curiosity, mastery, or agency.

Each layer optimizes a measurable proxy (a score, a degree, a grant, a citation count) over the unmeasurable goal it was meant to stand in for (understanding, capability, discovery). This is Goodhart's law operating across an entire national institution: once the proxy becomes the target, it stops measuring the thing. The global atlas's honest-limitations section already named the casualties — relevance, pedagogy, credentialism, purpose — as the problems no indicator captures. The U.S. is where you can watch all four happening at industrial scale, inside a system that, by its own metrics, looks like it's working.

5. Honest balance: what the U.S. system genuinely gets right

A reform thesis that strawmans its target is worthless. The U.S. system has real, hard-won strengths that any reform must preserve:

• The world's best research universities and basic-research engine (§3.1) — ~$940B/yr R&D, ~72% of Nobel-producing institutions, the top global destination for scientific talent. Nothing else on earth matches the top of this system.
• Breadth and the liberal-arts model. The U.S. undergraduate model deliberately keeps students broad before specializing — a real advantage for interdisciplinary and entrepreneurial work, and a contrast to the early, narrow tracking common abroad.
• Second chances. Community colleges, transfer pathways, adult and returning students, and a relatively forgiving credit system give Americans more routes back into education than most systems. The completion data is grim, but the on-ramps are uniquely plentiful.
• The innovation ecosystem around the universities. The tight coupling of research universities, venture capital, national labs, and industry (Bayh-Dole tech transfer, Stanford↔Silicon Valley, MIT↔Boston biotech) converts research into companies and products better than anywhere else.
• It is not failing on PISA reading/science (§1.3) and the lowest-performing 9-year-olds posted real recent gains (§1.2). The floor, in places, is recovering.

The honest summary is the same one the global atlas reached about money and access: the U.S. proves that you can lead the world at the top of a system while failing badly in its middle and bottom. Excellence and inequity coexist; the strengths are concentrated and the failures are distributional.

6. The specific U.S. flaws Bucket's open-knowledge thesis could address

Bucket Foundation's mission — make primary research paid-for-once and citeable-forever, route citation fees to authors not publishers, and exist for the small number of people who can do genius work with AI — maps onto a specific subset of the flaws above. Bucket cannot fix the property-tax funding model or the teacher pay penalty; those are policy problems for governments. But it is unusually well-positioned against these:

1. Credentialism / signaling (§2.2). An open, citeable, contribution-based record of what someone actually discovered or derived is an alternative to the degree-as-signal. If real work can be published, verified, and cited directly, the credential loses some of its monopoly on certifying capability. This is the most direct hit.

1. Research-funding concentration and the "who gets to do research" gate (§3.2). The research-atlas shows research money pooling in a few dozen elite institutions. A model that lets anyone publish and be paid for primary work — independent of institutional affiliation — opens the frontier to people the current gate excludes. The pipeline's deepest inequity is access to doing research at all; open knowledge attacks that directly.

1. Grant short-termism and the gatekeeping of genius work (§3.5). A citation-fee model that pays authors for foundational work that gets used — rather than for work that wins a 13%-success-rate grant lottery — creates an incentive aligned with lasting contribution rather than fundability. It is a different selection pressure, and one better suited to slow, risky, frontier work and to the exceptional individuals Bucket exists for.

1. Purpose and the proxy problem (§4). Every U.S. layer optimizes a proxy (score, degree, grant) over the real goal (understanding, capability, discovery). Bucket's unit of value — an axiom, a real derivation, a law, a primary result that others cite and build on — is the real goal, not a proxy for it. Making the foundational thing itself the rewarded, durable artifact is the structural antidote to Goodhart's law in knowledge work.

1. The publisher capture of citation value. Adjacent to all of the above: the current system routes the economic value of citation to journals and publishers, not to the people who did the work. Re-routing that to authors changes who can afford to keep doing research outside an institution — which is, again, the access problem.

What Bucket should not claim to solve, to stay honest: K-12 funding inequity, the teacher shortage, the student-debt overhang, completion gaps, and the factory/age-batch model. Those are real and central (they are the bulk of this document) but they are state-capacity and policy problems, not knowledge-infrastructure problems. Bucket's lane is the research and credentialing layer — and there its thesis is sharply on target.

Headline U.S. findings (cited)

1. K-12 funding inequity is manufactured by design. Predominantly nonwhite districts get $23 billion / ~$2,200 per student less annually than white districts educating the same number of children — and poor white districts still get ~$1,500/student more than poor nonwhite ones (EdBuild; The 74).
2. NAEP shows a post-2020 decline that concentrates at the bottom. Reading down 5 points vs. 2019 in grades 4 and 8; 8th-grade math down 8 points; bottom-decile losses grew 70% larger 2022→2024; a record share of 12th graders scored "below basic" (Urban Institute; Hechinger; NAGB).
3. The U.S. is middling on PISA despite top spending — 465 in math, below the OECD average of 472 (above average in reading/science) (NCES; OECD).
4. Teacher pay penalty hit a record 26.9% in 2024 — teachers earn 73 cents on the dollar of comparable graduates, in every state (EPI).
5. $1.7T+ student debt; ~1/3 of students don't finish in 6 years; ~28% of bachelor's programs have negative ROI. ~$1.69T federal debt / 42.8M borrowers / ~$39,600 avg balance; 61.1% six-year completion with a 27.6-point income gap (48.2% poorest vs 75.8% richest) (EducationData; Higher Ed Dive; FREOPP).
6. The U.S. leads world research — ~$940B R&D, ~72% of Nobel-producing institutions — but the money is hyper-concentrated: in our own research-atlas, NIH+NSF ≈ $678B and all top-25 recipient orgs are U.S. elite institutions (NSB/NCSES; Scientometrics; research-atlas/docs/LANDSCAPE.md).
7. The PhD pipeline over-produces ~7:1. Only ~14% of bio PhDs (~12% in engineering) reach tenure-track; ~40% of postdocs leave academia (PMC R0; Science/AAAS).
8. Grant hypercompetition delays and washes out researchers. NIH R01 success ~13% (FY2025); average age at first R01 rose from 35.7 (1980) to ~43 (2016); ~43% of first-time grantees never win again (Grantsights; Science/AAAS; PLOS One).

Most important sources

Primary / official data

• NCES — Nation's Report Card (NAEP main & long-term trend): https://nces.ed.gov/nationsreportcard/ ; Fast Facts long-term trend https://nces.ed.gov/fastfacts/display.asp?id=38
• National Assessment Governing Board (NAGB) 2025 releases: https://www.nagb.gov/news-and-events/news-releases/2025/
• OECD PISA 2022 (Volume I) & U.S. Country Note: https://www.oecd.org/en/publications/pisa-2022-results-volume-i-and-ii-country-notesed6fbcc5-en/united-statesa78ba65a-en.html ; NCES PISA 2022 https://nces.ed.gov/programs/coe/pdf/2024/CNU_508c.pdf
• NSF NCSES / National Science Board — Science & Engineering Indicators 2024 (U.S. & Global R&D): https://ncses.nsf.gov/pubs/nsb20243/discovery-u-s-and-global-r-d ; HERD survey https://ncses.nsf.gov/surveys/higher-education-research-development/2023

K-12 structure & equity

• EdBuild, $23 Billion (school funding racial gap): https://edbuild.org/content/23-billion
• Economic Policy Institute, The Teacher Pay Penalty Reached a Record High in 2024: https://www.epi.org/publication/the-teacher-pay-penalty-reached-a-record-high-in-2024-three-decades-of-leaving-public-school-teachers-behind/

Higher ed cost, debt, ROI, completion

• EducationData.org, Student Loan Debt Statistics: https://educationdata.org/student-loan-debt-statistics
• FREOPP, Does College Pay Off? A Comprehensive ROI Analysis (Preston Cooper): https://freopp.org/whitepapers/does-college-pay-off-a-comprehensive-return-on-investment-analysis/
• National Student Clearinghouse Research Center — completion: https://nscresearchcenter.org/yearly-progress-and-completion/
• Benjamin Ginsberg, The Fall of the Faculty (administrative bloat / adjunctification), reviewed: https://academeblog.org/2013/03/28/remarks-on-benjamin-ginsbergs-fall-of-the-faculty/

Research pipeline (peer-reviewed + reporting)

• Larson & Larson, Too Many PhD Graduates… R0 in Academia (PMC): https://pmc.ncbi.nlm.nih.gov/articles/PMC4309283/
• Competition for engineering tenure-track faculty positions (PMC): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11075531/
• Outcomes of early NIH-funded investigators (PLOS One / NIAID): https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0199648
• Science/AAAS — postdoc precarity & NIH grantee age: https://www.science.org/content/article/professors-struggle-recruit-postdocs-calls-structural-change-academia-intensify ; https://www.science.org/content/article/updated-fountain-youth-congressmans-plan-make-nih-grantees-younger
• Springer Scientometrics — Nobel-producing institutions: https://link.springer.com/article/10.1007/s11192-023-04831-1

Internal Bucket evidence

• education-atlas/docs/EDUCATION_PROBLEMS.md — the global quantitative atlas (78,326 obs, 219 countries) this document goes deeper than.
• research-atlas/docs/LANDSCAPE.md — 1.67M grants / ~$1.04T; NIH+NSF dominance and the all-U.S. top-25 recipient concentration cited in §3.2.

Methodological note: this is an evidence-synthesis essay, not a reproducible dataset. Figures are quoted with their sources and as-of dates; where sources disagree (e.g., total student debt $1.69T federal vs. $1.84T incl. private), both are stated. The framing follows the global atlas's standard: name the strengths honestly, ground every flaw in a cited number, and separate the problems a knowledge-infrastructure project can address from the policy problems it cannot.