Education
A system under stress in nearly every developed country. The post-pandemic learning loss, the college cost crisis, the future-of-work mismatch, the AI revolution, and what specific things actually help.
(roughly the same number as the US's entire annual education spending)
(higher in lower-income districts; partially recovered, not fully)
(roughly four times the wage growth over the same period)
A note on framing. Education is a topic where the political conversation has become unusually disconnected from the underlying data. The page below tries to walk through what the structural picture actually looks like - what is working, what is not, what the post-pandemic period exposed, and what the AI-era is starting to change - in a way that respects readers across the political spectrum. There is no clean ideological answer to most of the education questions; there are real trade-offs that different societies have made differently, and the data on what specifically works is more useful than the slogans usually suggest.
The post-pandemic learning loss
The COVID-19 pandemic interrupted in-person schooling worldwide for periods ranging from a few months to over a year, depending on country and grade level. Multiple research traditions have since measured the learning loss, and the headline numbers are striking. The average US student in 2022 had lost the equivalent of about half a year of math learning and roughly a quarter of a year of reading learning relative to what they would have gained in a normal period. Similar patterns appear in UK, Australian, Brazilian, and other countries' data, with magnitudes varying based on how long schools were closed and what alternative arrangements were available.
What worsened the picture: the loss was sharply unequal. Students from lower-income families, in lower-income school districts, and in countries with longer school closures lost more than their peers. Students whose parents could supplement schooling at home with paid tutors, technology access, or one parent staying home with the children fared substantially better. The pandemic widened the achievement gap that already existed, in some cases reversing decades of slow progress on closing it.
The recovery has been slower than expected. Federal and state governments invested heavily in catch-up programs - tutoring, summer school, expanded technology, additional teacher hiring. Some specific programs (high-dosage tutoring is the strongest candidate) have shown measurable effects. But four years after schools reopened, average test scores in the US, UK, and several other countries remain below their 2019 levels. The cohorts who were in the most disrupted grades during the pandemic are now passing through the system without having fully recovered the lost ground. The lifetime economic effect on these cohorts is estimated in the trillions of dollars across affected countries.
What this exposes about the system. The pandemic showed both that the in-person education system is essential (replacing it with remote learning produced enormous losses) and that the system was less resilient than expected (the catch-up has been slower and harder than the recovery investments would predict). The system depends on assumptions - children show up, teachers teach, parents do not have to substitute for schools - that turn out to be more fragile than the pre-pandemic education conversation acknowledged.
The college cost crisis
US college tuition has roughly doubled in real terms since 2000 at public universities and tripled at private universities. Wages have risen by roughly a third over the same period. The result is that the cost of college relative to a typical family's income has roughly doubled, with proportionate increases in student debt, time-to-degree, and the share of college costs that are paid with borrowed money rather than current income.
The drivers are a combination of factors none of which alone explains the picture. State funding for public universities has fallen as a share of those universities' budgets, with tuition rising to fill the gap. Administrative spending has grown faster than instructional spending. The amenities arms race (better dorms, recreation centres, athletic facilities) has consumed significant capital spending. Federal student-loan availability has expanded faster than wages, allowing universities to charge more without immediate price-discipline from buyers. Each of these is a partial cause; together they explain most of the increase.
What this has produced. Outstanding student-loan debt in the US is now about $1.7 trillion - up from about $300 billion in 2000. About 45 million Americans have student debt; the median balance is around $20,000 but the long tail is severe (graduate-degree holders, particularly law and medical school graduates, often carry $100,000+). The debt has measurable effects on home ownership (delayed by an estimated 5-10 years for borrowers), family formation, retirement saving, and career choice. Borrowers who do not graduate (about 30% of those who start) have the worst outcomes - debt without the credential that justifies it.
The political response has been incomplete. The Biden administration's loan-forgiveness program was struck down by the Supreme Court in 2023; partial alternatives have been implemented through different mechanisms. Income-driven repayment plans have been expanded. Some states (Tennessee, New York for SUNY/CUNY, others) have made community college free or reduced cost. Comprehensive reform of how college is financed has not happened. The structural problem - high tuition, easy borrowing, weak incentives for cost discipline - remains.
K-12 stagnation in many countries
Beyond the pandemic effects, K-12 outcomes in most rich countries have been roughly flat or slightly declining for two decades. The OECD's PISA tests measure 15-year-olds' ability in math, reading, and science across roughly 80 countries every three years. The US scores have been near the OECD average and roughly flat for 25 years; UK scores similar; many continental European countries have seen modest declines. The high-performing East Asian systems (Korea, Japan, Singapore, Hong Kong, parts of China) maintain their lead.
Why the stagnation. Education researchers have a reasonably consistent answer: the things that matter most for student outcomes are the quality of teaching, the support and structure children have at home, the peer environment in school, and the cumulative effect of many years of incremental learning. Improving any of these at scale is hard. Most education reforms either focus on the easy-to-measure surface (testing, accountability, curriculum standards) without changing the underlying teaching quality, or try to change the underlying teaching quality without the political and cultural support to actually do it. The reforms that have produced measurable gains (high-dosage tutoring, sustained early-childhood programs like Perry Preschool, specific phonics-based reading instruction in some districts) have been hard to scale beyond their original settings.
What is changing now. Several specific reforms have gained traction in the post-pandemic period: phonics-based reading instruction (the "science of reading" movement) is being adopted in many US states after decades of debate; high-dosage tutoring is being deployed at much larger scale than before; and early-childhood programs are expanding in many countries. Whether these add up to a meaningful improvement in average outcomes over a decade depends on whether the political and funding support sustains beyond the immediate post-pandemic window.
The future-of-work mismatch
The system that produces educated adults was largely designed in the late 19th and early 20th centuries to prepare workers for an industrial economy. Most of the structure - 12 years of K-12 schooling, four years of college as the standard credential, the academic curriculum that emphasises abstract reasoning and writing - has changed less than the economy it was meant to feed.
The mismatch shows up specifically. Most jobs that pay well and do not require advanced degrees now require some form of post-secondary credential or specific technical skill. The traditional vocational pathways (apprenticeships, community college technical programs, employer-trained skill development) are weaker in the US than in most peer countries and have been weakening further. The four-year college path that became the default for middle-class advancement in the post-WWII period now serves a smaller share of jobs than it did in 1990 - both because more jobs need different skills and because the four-year credential has become more expensive without delivering proportionately more economic return.
Germany and Switzerland have shown what works. Their dual-track education systems (academic for some students, structured technical apprenticeships for others) produce much lower youth unemployment, higher earnings for non-college-educated workers, and a stronger industrial base than US-style college-or-nothing systems. The political conditions that built and sustain those systems - employer involvement in education, cultural respect for technical work, cooperative labour-management traditions - are not easy to replicate in other countries that lack them. But the existence-proof matters: the future-of-work mismatch is solvable if the political conditions for solving it can be built.
Now consider AI. Many of the traditional knowledge-work jobs that justified the four-year college credential are now exposed to AI substitution (covered in the AI piece on this site). The professional skills that AI is least likely to substitute for - judgment under uncertainty, learning from limited data, physical and social skill, creative synthesis, and ethical reasoning - are not skills the typical college curriculum is currently designed to develop. Education systems that adapt to this will produce graduates with substantially better economic prospects than those that do not. Most systems are adapting slowly.
AI and what is changing in classrooms
AI has reached classrooms faster than the school system has adapted to it. Roughly half of US students report using AI tools (especially ChatGPT) for schoolwork; comparable numbers in other rich countries. Teachers report wide variation in policy - from outright bans to enthusiastic integration to grudging tolerance. The first wave (2023-2024) was characterised by panic about cheating; the second wave (2025+) is starting to be characterised by serious thinking about what AI changes about teaching and learning.
What AI does well in education. Producing first drafts and outlines that students can react to. Explaining concepts in multiple ways and at different levels of detail. Personalising practice problems for individual students at their actual level. Acting as a patient one-on-one tutor for students whose schools cannot provide that. Sal Khan's work and similar AI-tutor projects have shown measurable benefits in trials. The promise is real.
What AI does not do well in education. Building the foundational skills that come from struggle and persistence (writing one's own first draft, working through a math problem on paper, memorising vocabulary). Replacing the social-emotional learning that happens in classrooms with peers. Substituting for the relationship between a teacher who knows a student well and that student. Identifying when a student is struggling for non-academic reasons. These are the things that good schools have always done and that no AI can fully replace.
The integration question is not whether AI will be in classrooms - it already is - but how to integrate it without losing what schools do well. The most promising experiments combine AI as a tutor and practice tool with strong in-person teaching that focuses on the dimensions AI cannot substitute for. Schools that have done this thoughtfully report better outcomes than schools that have either banned AI or wholesale embraced it. The right balance varies by age, subject, and student. Most school systems are still figuring this out.
How country systems actually compare
Different countries' education systems perform very differently and reflect different political and cultural choices. The numbers below are rough.
The takeaway: different systems work for different reasons, and the relationship between spending and outcomes is loose. The countries that have improved most consistently (Estonia, Poland, Portugal, parts of the UK) have done so through specific focused reforms over many years rather than through any single dramatic policy change. The US case - high spending, average outcomes, severe equity gaps, world-class top tier - is one of the most distinctive in international comparison.
The paths from here
Education is one of those topics where policy moves slowly and consequences accumulate over decades. Each path below is one realistic shape the next 10-20 years could take.
Continued slow drift in K-12
Average outcomes stay roughly flat. Equity gaps remain wide. Reform energy is concentrated on the most visible recent priorities (literacy instruction, post-pandemic recovery, AI integration) without producing transformational change. The system continues to produce graduates with roughly the current distribution of preparation, and the labour market continues to reward those at the top while struggling to absorb those at the bottom.
Will it happen? This is the base case absent meaningful reform. Most education-policy efforts produce incremental change. The structural conditions that have held average outcomes flat for two decades have not weakened.
College costs reform meaningfully
Some combination of state-level free-college expansions, federal student-loan structural reform, employer-funded education credentials, and university cost discipline brings the cost-versus-wage ratio back toward something more sustainable. Total student debt grows more slowly. The political consensus that college is necessary for middle-class life is partly broken.
Will it happen? Slowly and partially. State-level free community college has spread. Federal student-loan reform has been incremental. The deeper structural drivers (administrative bloat, amenity arms race, easy federal lending) have not been fundamentally addressed. A more comprehensive reform requires political conditions that have not emerged.
Vocational and apprenticeship pathways expand
Renewed political and economic interest in non-college pathways produces a meaningful expansion of US apprenticeships, technical credentials, and employer-funded training programs. The system moves modestly toward the German-Swiss model, with corresponding improvements in non-college earnings and youth-employment outcomes. The cultural respect for technical work that has been weak in the US for decades partly recovers.
Will it happen? Some movement is happening. Apprenticeship enrolments have grown modestly. Several states have invested in technical education. Federal investment in workforce development has been mixed. Cultural attitudes toward non-college work are slowly shifting. Whether this adds up to meaningful structural change depends on sustained political support.
AI tutors transform individual learning
AI-aided personalised tutoring becomes standard across schools and homes. Children get one-on-one tutoring at the level that previously required wealthy families paying private tutors. The achievement gap that has been driven partly by tutoring availability narrows. Average learning gains accelerate as the tutoring effect (which has the strongest evidence base of any educational intervention) becomes universally accessible.
Will it happen? Some version is plausible by the early 2030s. The technology is moving in this direction. The implementation challenge is real - integrating AI tools without undermining what good teachers do, and ensuring access for students who lack home technology. Early data is encouraging. The political and funding support to scale this is uneven across countries.
A serious reform of how education is funded and structured
One or more US states (or countries with similar structural problems) undertakes comprehensive reform: standardising school funding, expanding pre-K, reforming teacher recruitment and pay, reducing administrative overhead, and aligning curriculum with the actual labour market. The reform is sustained over multiple administrations and produces measurable outcomes.
Will it happen? Possible but rare. Comprehensive education reform requires political consensus across multiple election cycles. The cases of sustained reform that have produced results (England 2010-2020, Poland 2000-2010, parts of US states like Mississippi's recent literacy reform) are notable for being exceptions. Whether the post-pandemic period produces enough urgency for comparable reforms in more places is open.
The credentialing system unbundles
The four-year college degree as a default labour-market signal weakens. Employers increasingly hire on demonstrated skills, portfolio-based credentials, and shorter targeted programs. Online learning platforms, professional certifications, and employer-issued credentials all gain weight. The economic value of a generic four-year degree falls; the economic value of specific demonstrated competencies rises.
Will it happen? Already partly happening. Major employers (Google, IBM, Apple, several others) have publicly removed degree requirements for many roles. The actual hiring practice still favours degree-holders; the gap between rhetoric and reality is real. The trajectory is genuinely toward unbundling but the pace is slower than the publicity suggests.
A widening gap between students whose families can supplement and those who cannot
The class effect on educational outcomes - which has been substantial for decades and worsened during the pandemic - continues to grow. Wealthy families afford private schools, supplement with tutors, choose neighbourhoods for schools, and pay for college without debt. Less wealthy families cannot. The educational system becomes increasingly stratified along class lines, with the public schools serving a different population than they did 30 years ago.
Will it happen? Largely already happening. The class gap in educational outcomes (covered in the wealth-inequality piece on this site) has been widening for a generation. Reversing this requires policy interventions that have not been politically achievable. The trajectory is toward deeper stratification absent specific countervailing policy.
The realistic forecast is, again, a mix. The base case is slow drift in average outcomes, with localised improvements in specific places that have invested in serious reform. AI-aided learning is the most likely transformational variable. Vocational and apprenticeship pathways may expand modestly. The college cost crisis will continue to produce political pressure without comprehensive reform. The class gap is likely to widen further before any countervailing force becomes politically possible.
Where serious analysts disagree
Education research is unusually robust in some areas (the effect of high-dosage tutoring, the importance of early-childhood development, the lower returns on specific kinds of school-spending increases) and unusually contested in others. Each reading below is held by named scholars worth engaging directly.
The system is structurally broken and reform is needed
The K-12 and higher education systems in the US specifically (and in many other rich countries) suffer from deep structural problems: misaligned incentives between schools and students, weak accountability, slow adoption of evidence-based practices, regulatory and political capture by educator interests rather than student outcomes, and a college-financing structure that rewards higher prices and longer enrolment regardless of value delivered. Marginal reform is insufficient; structural change is needed.
Held by: Caroline Hoxby (Stanford), Roland Fryer (Harvard), and a body of education-economics work focused on accountability and incentive design. Their case has been controversial in education-policy circles but has produced specific reform agendas with measurable results in some implementations.
The early-childhood foundation is the highest-leverage intervention
The structural improvements that produce sustained gains start before kindergarten. High-quality early-childhood programs (the kind documented in the Perry Preschool, Abecedarian, and similar long-running studies) have produced measurable gains in long-term outcomes - educational attainment, earnings, health, lower criminal-justice involvement - that no later intervention has matched. The economic return is among the highest of any public investment ever measured. The political coalition for serious early-childhood expansion has been weak relative to the evidence.
Held by: James Heckman (Chicago, Nobel Prize in economics), and a substantial body of human-capital research. The empirical case is among the strongest in social science. The political case has been harder to build because the benefits accumulate over decades while the costs are immediate.
Vocational and technical paths need expansion
The over-emphasis on four-year college as the standard middle-class credential has produced a generation of people who took on debt for credentials they did not need for the work they ended up doing, while leaving non-college-bound students without the structured technical pathways that could have produced good careers. The German-Swiss model demonstrates what is possible. The US specifically has been slow to take this seriously despite repeated rounds of awareness.
Held by: David Autor (MIT), and a body of labour-economics work on the future of employment. The case is strong in the data; the political coalition for non-college pathways has been hampered by the cultural assumption that not-college means failure - an assumption that does not match the data on actual life outcomes for people in skilled trades and technical fields.
AI will transform education within a decade
The combination of AI tutors that can deliver personalised one-on-one teaching, AI tools that can identify student difficulties early, and AI-aided assessment that can adapt to individual progress will produce gains in educational outcomes larger than any policy reform of the last 50 years. Schools that integrate these tools well will produce visibly better outcomes. The technology is improving fast enough that the effect should be measurable by the late 2020s.
Held by: Sal Khan (Khan Academy), Eric Schmidt, and a tradition of education-technology optimism. The case has both substantive merit (the data on tutoring effects is strong, and AI can scale tutoring access) and limits (educational outcomes have proven resistant to technology-driven transformation in past waves). Whether AI actually delivers what previous technologies have not is genuinely uncertain.
The credentialing system is the underlying issue
Most of what college teaches has weak measurable economic return. What it actually does is sort students by ability and conscientiousness through the credentialing process. The four-year credential signals to employers something about the holder - mostly that they were able to complete a multi-year structured commitment - rather than transferring specific skills. This is an expensive and inefficient sorting mechanism. The reform that would help most is breaking the credentialing monopoly and allowing employers to assess candidates on what they can actually do.
Held by: Bryan Caplan (George Mason, "The Case Against Education"), and a tradition of skeptical labour-economics work. The case is uncomfortable in education circles but has substantive support. The implication is that policy aimed at expanding college access without addressing the underlying credentialing problem may produce more borrowing without proportional improvement in outcomes.
None of these readings is fully right or wrong. What can be said from the available evidence: the US system specifically has structural problems that have produced flat outcomes despite high spending; the early-childhood case is strong and underweighted; vocational and apprenticeship pathways are real solutions partly stalled by cultural assumptions; AI may transform individual learning over the next decade if integrated well; and the credentialing question is real and is starting to be addressed by some employers. Most countries have something to learn from each of these and will benefit from picking pieces of each.
What this means for you
Education touches every reader's life directly through their own schooling, their children's, and the workforce around them. A few practical observations:
If you have young children
The early years (zero to five) matter enormously. The home environment - reading aloud, talking to children at length, exposure to broad experiences, stable routines, low chronic stress - matters more than almost any specific educational programme. Quality childcare and pre-K make measurable differences for outcomes a decade later. None of this requires unusual wealth; it requires unusual attention. The single highest-leverage parental investment is in the first five years.
If you have school-age children
Average school quality matters; the specific school matters more than the district average; and the home environment matters more than either. Reading at home, demonstrating that learning is valuable, talking through difficult academic material, providing structure, and being involved enough to know when something is going wrong are all things parents can do without unusual resources. The class gap in outcomes is largely explained by things that happen at home, which is uncomfortable to acknowledge but means there is real room for parents who care to make a difference.
If you are choosing college
The economic return on a four-year degree varies enormously by major, school, and individual. STEM, business, healthcare, and economics degrees from solid schools have the highest returns. Humanities degrees from less selective schools have the lowest. Going to a big-name school is less important than completing a degree at all - the gap between dropouts and graduates is far larger than the gap between graduates of more and less selective schools. Borrowing aggressively for a degree of uncertain value is a worse decision than the political conversation typically acknowledges; community college followed by a transfer is often a better financial path.
If you are not college-bound or post-college
Vocational and technical paths can produce excellent careers with low debt and strong employment prospects, especially in skilled trades (electrical, plumbing, HVAC, welding, construction management), healthcare adjacent roles (nursing, technicians, therapists), and emerging technical roles (cybersecurity, network administration, specific software development paths). The cultural pressure toward four-year college has obscured these paths in ways that have hurt many people who would have done better with a different choice. Apprenticeships, community college technical programs, and employer-funded training are all worth taking seriously.
If you are thinking about your own learning
The skills that AI is least likely to substitute for - judgment under uncertainty, learning from limited data, physical and social skill, creative synthesis, ethical reasoning, the ability to navigate ambiguity - are also the most valuable skills in most professional environments. Continued learning across a lifetime, adapting as the economy changes, becoming the kind of person who can think about hard questions without being paralysed - all of these matter more than any specific credential. The half-life of specific skills is short; the underlying ability to keep learning is durable. This is the real return on education.
The mechanics behind this
The education story sits on top of three deeper mechanisms covered elsewhere on this site. If the analysis above depends on ideas you want to understand first, these fundamentals make the conversation more legible:
