Every parent knows the loneliness of when you first leave the hospital with your newborn in tow. It's a sense of incredulity, a feeling like aren’t-they-going-to-make-me-show-a-permit-or-something that smacks you in the face as you strap that baby into a car seat for the first time. In my case, I kept whipping my head back and forth, thinking someone would yell out that I had forgotten a pamphlet that told me how to be a Dad. That didn’t happen. I just closed the door to my dented Subaru Crosstrek and started to meander home.

On that drive, one that every new parent does well under the speed limit, I realized something: there was no one coming to help. That burbling bundle of cuteness in the backseat was my responsibility. If that kid is going to get what they need, I was the one who was going to make that happen. You are constantly looking for answers on giving them food, helping them walk, and teaching them to not pull on the dog’s ears.

Then one day, the question that defines so much anxiety parents feel starts to loom—education. I want my child to have access to great schools that will help shape her into an intelligent, moral, and hard-working person. That desire, however, presents a problem:

What the hell is a good school anyway?

Before my daughter was born, I was what could only accurately be described as an education system dolt. I had never thought about childhood education because why would I? In my day job, I’m a researcher who focuses on technology markets. My time is spent playing with AI, talking to scientists, and trying to figure out what the future holds. (Lately, I mostly spend my time counting the minutes until I get to see my kid again.)

My daughter has some health concerns that means I have to think far harder and earlier about the education question then most parents. When I went to answer the question of what a good school was, I was surprised at how subpar the resources were. Everything I could find was written for education practitioners, or was clearly motivated by a political ideology, or was overly reliant on anecdotal narratives. And hilariously, there was almost nothing written for parents that didn’t treat us like morons.

I don’t care about politics! Or theory! I only care about my kid and I want to make a smart, risk-adjusted choice that maximizes my kids chances at a successful education. That something to help me make this choice didn’t exist was a little baffling.

So, I got work. I have now interviewed dozens of parents, academic researchers, ed-tech founders, charter school networks, and teachers. Simultaneously, I have poured over an ungodly stack of books and research papers to make sure that my understanding was academically rigorous. All of that effort was geared towards gaining a pragmatic understanding of how education works. This series of essays is the result of those months of effort.

Why bother?

These essays are not meant for researchers or academics, it is meant for parents like you and me. This framing matters because I am only interested in what is possible. Too much of the advice out there assumes infinite parental resources, or drowns in theory while ignoring the reality of local school systems. When you finish this you’ll understand the following:

• The urgency of the problems facing American schooling systems
• An overview of what we know about learning science (and why schools can’t seem to utilize what the science says).
• How schools are funded and who decides where the budget goes
• Why schools are bundle of services that are impossible to deliver 100% on
• Why schools don’t care about academics that much
• The incentives driving the mismanagement of schools
• Why educational technologies haven’t changed anything
• What are the actionable and available choices you can make to improve your kids education today

Think of this as something you can read in less than three hours, and by the end, you will know enough to be confident in the decisions you are making for your child.

Around the time my research project kicked off, I got in touch with the folks at Alpha School. They are a private school network that has been the focus of lots of excellent analysis, and they told me they had a problem. Namely, parents didn’t feel like they had the requisite knowledge of how school systems work to make bold choices for their kids' education. This was the exact problem I had been noodling on. I also wanted to be able to make the right choice for my kid, but needed intellectual scaffolding to help me. We realized that we could help each other. They could pay for my time to write the guide that I always wanted and they could get a handy resource to send to parents who were thinking about changing things up. To their credit, they exerted no editorial oversight of this project and gave me complete creative control. I am grateful for their support.

I called this intro, “No One Is Coming to Save Your Kid” because that is perhaps the most important thing you can take away from this series. The U.S. government has been saying that our education system is broken for 50 years. They haven’t fixed it. They may someday, but that day will be too late for your child. If you want your kid to have a great education, it's up to you to do something about it. You should start right now.

In 2021, American high schoolers had the highest GPAs ever recorded—and the lowest standardized test scores in a decade. This stat should give every parent pause. How on earth can both of these things be true at the same time?

Every president since Jimmy Carter has declared American education "broken." 50 years of alarm bells ranging from Reagan's "crisis caused by low standards" to Obama's "we've let our grades slip" to both Trump and Biden agreeing (perhaps the only thing they agree on) that the system isn't working.

You've probably tuned it out. It sounds like background noise at this point. 50 years of a crisis is no longer a crisis, it's just the way things are. Still, despite the uniformity of the headlines, some of the crisis is overblown, and some of it is exponentially worse than I think most parents understand.

The question this essay answers:

How can you know if (or in what way) the school system is failing your kid? Because the primary tool you rely on to find out—the report card—might be lying to you.

Are the kids actually dumber?

The best tool we have for this is the NAEP Long-Term Trend assessment—a nationally representative test that's been administered to American students since 1971 in reading and 1973 in math. It tests three age groups—9, 13, and 17—which means we can track if kids improved and if those improvements lasted as students moved through the system.

And what you see is… complicated.

Let's start with the youngest kids.

From the early 1970s through 2012, 9-year-olds' reading scores rose from 208 to 221. Math scores climbed even more dramatically, with a 25-point gain. During the exact decades that politicians were declaring American education a catastrophe, the youngest students were actually getting better at reading and math. Slowly, unevenly, but the trajectory was clearly upward.

So far, so encouraging. But watch what happens as those kids get older.

Thirteen-year-olds showed a similar pattern to the 9-year-olds. Their math scores rose 19 points from 1973 to 2012, reading gained 8. But the gains were smaller, and the collapse was worse. By 2023, 13-year-olds' reading scores had fallen all the way back to 1971 levels. In math, they dropped 14 points from the 2012 peak, erasing gains that took three decades to build. Whatever was working in elementary school was already losing force by middle school.

Now look at what should be the most important chart of the three—the one tracking 17-year-olds, the students about to enter college and the workforce.

In reading, 17-year-olds scored 285 in 1971 and 287 in 2012. A two-point gain over 41 years. In math, 304 to 306. Essentially flat. All those improvements in elementary and middle school, the hundreds of billions being spent, produced stronger young learners who somehow became completely average older ones. The gains washed out by the time they mattered most.

And then there's the hatched gray area on the right side of those charts. We don't know what happened to 17-year-olds after 2012, because the Department of Education canceled the test. The 2020 administration was scrapped for COVID. The 2025 test administration was defunded by the DOE. And don’t worry, the federal government has already cut the planned 2029 test entirely. We stopped measuring high school readiness on our longest-running national assessment right when the younger-age scores started falling off a cliff. Whoops!

But there is something those overall averages hide—the gains weren't evenly distributed, and neither were the losses.

During the improvement era, Black 9-year-olds gained 36 points in reading—roughly five times the gain for White students over the same period. Black 13-year-olds gained 24 points in reading versus 9 for White students, cutting the White-Black gap from 39 points to 23. Hispanic students showed similar patterns where they had two to three times the gains of their White peers across both subjects and age groups. The American education system was improving because it was reaching the kids who started furthest behind. The rising tide was real, and it was lifting the boats that needed it most.

Then came the collapse, and it reversed the equation. In the 2020–2022 period alone, Black 9-year-olds lost 13 points in math while White students lost 5—widening the gap from 25 points back to 33. The pattern repeated across age groups and subjects: the students who had gained the most lost the most. Decades of slow, hard-won progress in closing achievement gaps were undone in a few years.

This is the part of the story that should matter to every parent, not just the ones whose children are in struggling schools. The system's best accomplishment over the past half century—narrowing the gap between its highest- and lowest-performing students—is being erased. And the collapse isn't just pulling the bottom down. As we'll see, the ceiling is falling too.

Are American kids even competitive anymore?

The most recent data, from the 2024 NAEP, confirmed that the collapse in scores we’ve discussed wasn't a blip. Forty percent of 4th graders now read below the "Basic" level—the highest share since 2002. Among 12th graders, 45% score below "Basic" in math which is the highest percentage ever recorded. That means nearly half of American high school seniors, on the verge of entering college or the workforce, cannot perform what the federal government considers basic mathematical operations.

It's tempting to blame COVID. School closures were devastating, and their effects are real and measurable. But the National Center for Education Statistics Commissioner Peggy Carr has cautioned against that framing: the pandemic, she argued, "laid bare" an opportunity gap that existed long before anyone had heard of COVID-19. Scores were already declining between 2012 and 2020, before the pandemic hit. COVID just accelerated a decline that was already underway and stripped away the veneer of normalcy that had kept parents and policymakers from noticing.

Still, maybe you're thinking, hasn’t American education been getting worse by international standards?

There’s truth to this sentiment and COVID accelerated this trend. The 2022 PISA results for the U.S. math score was a 13-point drop from 2018, among the lowest the country has ever recorded on this assessment. Only 66% of American students reached Level 2 proficiency in math—the baseline considered necessary for participation in modern society.

Over two-dozen countries now outscore us in math. Singapore, Japan, South Korea, Estonia, and others are pulling away. We remain respectable in reading and roughly average in science, which means we're not uniformly terrible. But in the subject most closely linked to economic competitiveness and technological innovation, American students are falling behind peers in countries with a fraction of our resources.

The international context matters because it sets a ceiling on the "our kids are fine" argument. Even if your child is in a top district, outperforming the national average, they're competing in a global economy against students from systems that are producing stronger math outcomes on average.

Income doesn’t solve everything

Before we get to what all this means for your family, it's important to be honest about where the crisis is worst.

The Education Recovery Scorecard, a joint project of Stanford and Harvard, found that the highest-income school districts are recovering from learning losses nearly four times faster than the poorest ones. If you're in a well-funded suburban district, your schools probably have the tutoring programs, the staffing, and the parental resources to claw back what was lost. If you're in a high-poverty district, the odds are much worse. (Though there are notable exceptions such as in Mississippi).

When you control for poverty and English-language proficiency, much of what looks like a racial achievement gap is really a poverty gap. This isn't to dismiss the reality of racial disparities in education, they're real and they matter. But the single strongest predictor of a child's test scores remains family income. The system is failing poor kids hardest, and the recovery is reaching them last.

Still, if you're an educated parent in a well-resourced district, reading all of this, the natural reaction is: My kid is probably fine, right? Not really. "Fine" is an exceptionally lower bar than you think. And you might not have the information you need to know whether it's even true.

The NAEP Long-Term Trend assessment doesn't just track averages—it reports scores at the 10th, 25th, 50th, 75th, and 90th percentiles. And at the 90th percentile—the high-achieving students, the kids you'd expect to be thriving—13-year-olds' math scores dropped 6 points between 2012 and 2023. Reading showed a similar decline. These are the top performers in the country, and they are learning less than their counterparts did a decade ago.

The fact that the ceiling dropped at all should unsettle any parent who assumes their high-performing child is immune to what's happening in American education. The system isn't optimizing for anyone's potential right now. It's failing some kids more visibly than others—but the decline is everywhere.

In 2010, a high school student with a 3.0 GPA—a solid B average—had an average ACT score of 19. By 2021, a student with that same 3.0 GPA scored a 15 on the ACT. (This was a stat that made me sit up during my research.)

Between 2010 and 2021, the average high school GPA rose from 3.17 to 3.36—the highest ever recorded. Over the exact same period, the average ACT composite score fell from 21.0 to 20.3. Students were getting better grades for worse performance. 2021 was simultaneously the year of the highest GPAs and the lowest standardized test scores in a decade.

Your kid's 3.8 might reflect genuine excellence or it might reflect a system that has redefined what an A means. Without an external benchmark, you have no way to tell the difference. And most parents don’t even know to look for one.

The implications go beyond college admissions. If grades no longer reflect actual learning, then the primary feedback mechanism parents rely on—the report card—is broken. A parent who sees a transcript full of A's and B's has every reason to believe their child is on track. The transcript is more than likely lying to you. Alpha School told me that INSERT TESTING SCORES FROM LEADING PRIVATE SCHOOLS TRANSFERRING TO ALPHA.

And it isn’t just the report card! It is likely that your child may not even be in the correct grade.

I talked with Ryan Delk, the cofounder of Primer, a charter school network of 48 schools about the extent of this problem. He told me that “85% of our students are coming from traditional public schools. The academic performance of a lot of these students, particularly coming from low performing public schools, is extremely bleak. It is as bad or worse than a lot of what the data is that you see publicly.”

In all of my interviews, with teachers and administrators serving both rich and poor students, this sentiment was repeated to me. “Kids are more behind than parents realize.”

The lowered ceiling

The decline at the top isn't just a COVID story. It reflects something more structural—a system that, for the past two decades, has quietly stopped trying to push its best students.

On the international TIMSS assessment—one of the most respected international math and science tests—10% of American 8th graders score at the "advanced" level in mathematics. In Japan, it's 34%. In South Korea, it's over 40%. In Singapore, 38%. We're not just losing to the countries at the bottom. We're also losing at the top. The students who should be our future engineers, researchers, and founders are performing at a level that wouldn't even be notable in a dozen other countries.

Over the past two decades, a combination of policy incentives, budget pressures, and a well-intentioned focus on minimum proficiency gradually pushed gifted education to the margins. Accountability systems rewarded schools for getting struggling students over the proficiency bar—a defensible goal—but created almost no incentive to push high-performers higher. Gifted programs were cut and schools that had offered acceleration pulled back. Today, the average school district receives $3.38 per pupil in state funding for gifted education, and only 20% of districts receive any gifted funding at all.

The research on what happens to high-ability kids in this environment is uncomfortable. Camilla Benbow at Vanderbilt directs the Study of Mathematically Precocious Youth—a 54-year longitudinal study tracking over 5,000 gifted individuals. Her finding was then when gifted kids aren't challenged, they develop habits that actively work against them. They coast on ability. Their tolerance for difficulty stays undeveloped. They learn, very early, that school is something you get through rather than something that stretches you. And by the time they hit a wall—college, graduate school, a demanding career—they don't have the tools to push through it, because no one ever made them.

Meanwhile, one of the interventions with the strongest evidence base—academic acceleration—remains dramatically underused. A major 2020 Vanderbilt study confirmed that accelerated students outperform non-accelerated peers on the SAT and ACT, earn higher grades through college, pursue more advanced careers, and show no negative long-term effects on psychological well-being. The fear that pushing kids harder will somehow damage them is not supported by the data. What damages them is the opposite—years of sitting in classrooms where they already know the material, learning that effort is optional.

If you're a parent of a bright kid who seems to be doing fine, the question this data raises isn't whether your child will succeed. They probably will, by conventional metrics. The question is what they could have become if anyone had pushed them.

What this means for you

Most parents evaluate their kid's school the way they'd evaluate a restaurant: the neighborhood is good, the test score averages look decent, the neighbors' kids got into solid colleges, property values are high. It all checks out.

The problem is that almost everything parents use to judge school quality measures inputs, not outputs. Spending per pupil, teacher credentials, class sizes, demographic composition—these tell you what goes into the system. A school can have every advantage on paper and still produce surprisingly little actual learning growth. And you as the parent just can't tell, because everyone's getting A's.

This is the core of the problem. The combination of grade inflation, reputation-based evaluation, and the absence of clear external benchmarks creates a situation where parents genuinely believe their children are getting a world-class education—and have no reliable way to verify it. Not catastrophically wrong. Not "my kid can't read" wrong. But "my kid is doing B+ work while being told it's A work, and would be unremarkable in a classroom in Singapore" wrong.

If your kids is in a good school district, they’ll be ok. They'll graduate, get into college, find work. The question you’ll never be able to answer is how much of your kid's potential is being left on the table, and would you even know?

No one is coming to fix this for you. Your district isn't going to send a letter home saying "your child is getting A's but is two years behind a peer in Seoul." The grades will keep coming. The graduation rate will stay high. If you want to know whether your kid is actually learning you're going to have to look for yourself.

Which raises the obvious next question: what does real learning actually look like? What does the science say about how kids absorb, retain, and apply knowledge—and why do so few schools seem to use any of it?

We’ll cover that in Essay 2.

Here is a scene you have likely lived as a parent. Your kid is sitting at the kitchen table, textbook open, highlighter in hand, “studying” for tomorrow’s test. They read through their notes. There might be stickers or doodles on the margins. There are highlights in 4 different colors. When you ask how it’s going, they say, “Good. I feel like I know this.”

Are they right? Is all that work your kid putting in actually effective at getting them to learn?

Your brilliant child might even get an A on that test! But research shows that an A is a remarkably poor measure of durable learning. A landmark study in 2006 found that students who crammed through repeated re-reading—exactly what I did growing up—forgot over half the material within a week. Without specialized techniques, memory savings drop to roughly 20% within six days and essentially zero within a month.

Unfortunately, most American schools teach with a method that we know, scientifically, will have our kids forgetting everything they were tested on. There has been a stunning lack of upgrading of teaching methods on the basis of cognitive science. I do not state this in a way that is judgemental of the teachers in your life. They aren’t even trained in this stuff. A 2016 analysis examined the most commonly used teacher-training textbooks and found that none of them covered all six key evidence-based learning strategies identified in the research literature. Zero! Zilch! Nada!

The system failed the teachers before the teachers could fail the students. Halpern and Hakel (2003) put it more bluntly:

“It would be difficult to design an educational model that is more at odds with the findings of current research about human cognition than the one being used today at most colleges and universities.”

What a quote.

In Essay 1, I showed you that the American school system is producing declining outcomes masked by inflated grades. This essay answers the obvious follow-up: why aren’t our teaching methods effective? What does the science say about how kids actually learn? We will go through:

What the science says about how memory works

How kids should be taught based on the science and what schools do instead

To front-run my critics: the evidence for these strategies is strong, but I want to be honest about its limits. Much of the foundational research was conducted in laboratory settings. A growing body of classroom research confirms the effects translate to real schools, but effect sizes are often smaller and more variable in the wild than in the lab. This is normal for any science-to-practice translation. The directional findings are still robust enough to take extremely seriously by my estimation. But where the evidence is contested or nuanced, I’ll tell you.

This is not about some ivory tower hypothetical or press conference held by a politician. I wrote this essay because I want to do right by my kid, and I want you to have the power to do the same.

How Your Kid’s Brain Actually Works

Before we can talk about what schools get wrong, you need a mental model of how learning works. Without this knowledge, every claim about education sounds equally plausible, and you can’t tell the science from the marketing.

The Three-Part Memory System

Your brain has three memory systems that work in sequence. Sensory memory captures fleeting impressions—what you see, hear, and feel in the moment. Most of it vanishes almost instantly. A tiny fraction gets passed to working memory—your mental scratchpad. This is where you actively think, reason, and process new information. And if the processing goes well, the information gets encoded into long-term memory. This is where you store crucial life information like how the mitochondria is the powerhouse of the cell. Learning is, fundamentally, the process of getting information from working memory into long-term memory in a way that it can be reliably retrieved later.

The critical bottleneck is working memory. It is shockingly limited. You may have heard that people can hold “seven plus or minus two” items in short-term memory—that’s from George Miller’s famous 1956 paper. But subsequent research has revised that number downward. A 2001 review concluded that the true capacity of working memory is closer to four chunks of new information. Four. That’s the fundamental constraint that everything in learning science bumps up against. Anything beyond that and some info is getting deleted.

Because working memory is so limited, how material is presented matters enormously. This is the core insight of cognitive load theory, developed in 1988 and refined over decades of research. The theory distinguishes three types of mental load:

Intrinsic load is the inherent difficulty of the material itself—you can’t simplify long division into something trivially easy. Extraneous load is the unnecessary difficulty created by bad instruction—confusing worksheets, teachers so boring that students frequently fall into comas, poorly sequenced material, or irrelevant information that eats up working memory. And germane load is the productive effort of actually learning where students are making connections and building mental models by integrating new information with what they already know.

The goal of good instruction is to minimize extraneous load and maximize germane load. Think of working memory like a small kitchen counter. If you pile it with every ingredient and tool you own, you can’t actually cook. Good instruction clears the counter and hands you exactly what you need, when you need it. As Sweller and colleagues wrote in their 2019 update to the theory:

“Cognitive load theory uses our knowledge of human cognitive architecture to devise instructional procedures… The architecture consists of a working memory that is very limited in capacity and duration when dealing with novel information, coupled with a long-term memory that is unlimited in capacity and duration.”

This has consequences across every subject. When a student is asked to read a passage about the Civil War but doesn’t have background knowledge about the political context, their working memory gets consumed by basic comprehension—leaving nothing for the higher-order analysis the teacher is actually asking for. Kids without sufficient background knowledge can decode words just fine but can’t comprehend texts, because comprehension itself demands working memory. Your child has to have basic knowledge on how things work to then move to higher abstractions of learning and thinking.

What “Knowing” Something Actually Means

Ultimately, there’s a critical difference between recognition and retrieval. Recognition is the feeling of familiarity—you see something and think, yeah, I know that. Retrieval is the ability to produce it from memory on demand, without any cues. Most study habits—re-reading, highlighting, reviewing notes—build recognition. But real learning requires retrieval.

This is a subtle, emotional distinction for learners. Roediger and Karpicke (2006) demonstrated that students who took practice tests remembered significantly more material a week later than students who spent the same time re-reading—even though the re-readers felt more confident about their knowledge. The students couldn’t even predict this effect. They felt like re-reading was working better even though it wasn’t.

This is a theme we will return to a lot. Much of education is designed around the feeling of learning, without any focus on actually doing the things that improve what the student knows, regardless of how it feels.

So, with this knowledge on how memory works, let’s examine how schools structure their programs.

The Three Strategies Your Kid's School Should Be Using

The following strategies are well-proven and extensively replicated. That's what makes their absence from most classrooms so striking.

1. Retrieval Practice (The Testing Effect)

Try this tonight. Your kid is studying for a test, textbook open, re-reading their notes for the third time. Take the book away. Hand them a blank sheet of paper and say, "Write down everything you remember about this chapter."

They will look at you like you've suggested something cruel. They'll stare at the blank page. They'll write three halting bullet points. They'll say, "I don't know anything." They will not enjoy this and you will be the bad guy.

That’s ok because that discomfort is learning in action.

Being tested on material—even without feedback—produces more long-term learning than spending the same time re-studying. This is one of the most replicated findings in all of cognitive science. A 2006 review showed that testing is an effective way to create knowledge. And this isn't just a lab curiosity. A 2021 systematic review analyzed 49 classroom studies across math, science, psychology, history, and foreign languages—over 5,300 students—and found that 57% showed medium or large benefits from retrieval practice.

What this means for you: If your kid's school gives tests only a few times per semester, they're missing the point. Frequent low-stakes quizzing is learning, and a 2020 meta-analysis of 52 classes with nearly 8,000 students confirmed it works, especially for struggling students, dramatically reducing failure rates.

What most schools do instead: Tests are treated as summative events—the final judgment—rather than formative tools that produce learning. Homework reinforces what was taught that day rather than requiring retrieval of older material. Many teachers would love to change this, but research on implementation barriers consistently finds that time and structural constraints stand in the way, even when teachers want to adopt better practices.

2. Spaced Repetition (The Forgetting Curve)

Imagine two kids preparing for the same biology exam, spending the same total number of hours studying. One crams for three hours the night before. The other reviews the material for thirty minutes across six different evenings over two weeks. On test day, they score about the same. The crammer might even do slightly better.

A month later, one of them remembers almost nothing. The other still knows most of it.

This is the spacing effect, and it's been replicated more thoroughly than almost any other finding in educational psychology. In the 1880s, the German psychologist Hermann Ebbinghaus ran a series of experiments on himself and discovered that memory decays exponentially after learning—what he called the "forgetting curve." But he also found the fix: revisit material at strategically increasing intervals, and each review strengthens the memory trace and slows its decay.

More than a century later, his findings hold up at a massive scale. A 2006 meta-analysis of 184 articles containing 839 assessments found that spaced practice dramatically outperformed massed practice across virtually every condition tested. A follow-up study (Cepeda et al., 2008) with over 1,350 participants confirmed that the benefit of spacing was present at every delay tested, from one week to one year.

What this means for you: When your kid crams the night before a test and gets an A, they'll forget most of it within weeks. The A is an illusion. Real retention requires coming back to material again and again over time. If nobody is scheduling those returns, the forgetting curve does its work silently.

What most schools do instead: Material is taught in blocks. Chapter 3 is covered in week 3, tested in week 4, and never systematically revisited. Teachers who assign cumulative reviews or spiral back to old material are fighting the structure of their own curriculum.

3. Interleaving (Mixed Practice)

Picture your kid's math homework: 20 problems, all on the same concept they learned today. This is called "blocked" practice, and it's how virtually all textbooks and classrooms work. It feels productive. Your kid might even get 20 out of 20 right and feel great about it.

But it's an illusion of mastery. The student knows which strategy to use because every problem requires the same one. They never have to identify which strategy is needed—the harder, more important skill.

Rohrer and Taylor (2007) demonstrated that "interleaved" practice—mixing problem types together—produces better learning and transfer, even though it feels harder in the moment. The problem is even baked into your kids textbooks. A 2020 study found that math textbooks contain almost no interleaved practice.

Interleaving isn't restricted to math finding. In art history, students learn to distinguish painters better by seeing works mixed together rather than grouped by artist. In foreign languages, mixing vocabulary topics outperforms studying them in neat categories. In physics, shuffling problem types on homework beats working through them in blocks. A 2019 meta-analysis confirmed that interleaving benefits are robust, though the size of the effect depends on how similar the interleaved categories are.

I want to be honest: lab effect sizes for interleaving are large, but classroom results are more variable. The principle is sound but execution is what separates success from failure.

What this means for you: If your kid breezes through homework but struggles on tests, the homework structure may be the problem, not your kid. Real mastery means being able to identify which tool to use when the problems are mixed—not just applying the tool you know is coming.

Why All of This Feels Wrong

You might have noticed a pattern. Every strategy above makes learning feel harder in the short term. Retrieval practice feels worse than re-reading. Spaced review feels less productive than cramming. Interleaved homework produces more wrong answers than blocked homework.

This is perhaps the central insight of my writing process for this essay—what feels productive is usually not.

Robert Bjork at UCLA coined the term "desirable difficulties" to describe conditions that slow initial performance but massively boost retention and transfer. As Soderstrom and Bjork (2015) put it: "Conditions of instruction that make performance improve rapidly often fail to support long-term retention and transfer, whereas conditions that create challenges and slow the rate of apparent learning often optimize long-term retention and transfer."

And here's where it gets alarming. In 2019, researchers ran an experiment on introductory physics classes at Harvard. Students in active learning sections learned more—and reported feeling like they learned less. Students in passive lecture sections learned less but felt like they learned more.

This is a wild result. People's lived experiences directly contradicted their tested reality. This mismatch between perceived and actual learning is a systematic bias that warps how students study, how teachers teach, and how parents evaluate schools. Almost every participant in education is inadvertently incentivized to maximize the feeling of learning at the expense of durable knowledge. And the people being harmed—our kids—can't even tell it's happening.

One important caveat: "desirable" is the key word. Not all difficulty is productive. Confusion from bad instruction, missing prerequisites, or anxiety is not the right kind of struggle. And parents who respond to this research by forcing hours of extra study are missing the point. The research doesn't say "learning should be hard." It says learning should involve the right kinds of effort—effortful retrieval of material you've been taught, or the challenge of identifying which strategy to apply when problem types are mixed. That's a crucial difference.

4. Mastery Learning (The Prerequisite Chain)

What if your average kid could outperform 98% of their peers? In 1984, Benjamin Bloom showed this wasn't hypothetical.

Bloom published "The 2 Sigma Problem" after discovering that students who received one-on-one tutoring combined with mastery learning—meaning they had to demonstrate genuine proficiency on each topic before moving to the next—performed two standard deviations above students in conventional classrooms. Two standard deviations. The average tutored student outperformed 98% of the control group.

You've already seen mastery learning at work in your own home, even if you didn't call it that. A student who never fully grasped phonics struggles with decoding. Because they struggle with decoding, they read slowly. Because they read slowly, they can't keep up with grade-level texts. Because they can't keep up with the texts, they fall behind in every subject that requires reading comprehension—which is every subject. One gap in the foundation, and the entire structure tilts.

The same cascade happens in math. When your kid "suddenly" starts struggling in algebra, it's rarely sudden. They probably have undiagnosed gaps in arithmetic or pre-algebra that nobody caught, because the class moved on whether they were ready or not. Your kid has the capability. They may just be missing foundational skills that nobody went back to check.

A 1990 meta-analysis of 108 controlled evaluations found that mastery learning consistently improved student outcomes, with an average effect size of 0.5 standard deviations on locally developed tests—plus benefits for motivation, self-regulation, and attitudes toward learning. There is an honest debate here: the effect size on standardized tests was smaller, and critics argued that some studies had methodological limitations. (There are responses to this criticism.) But even the conservative estimates show meaningful improvement, and the logic of the mechanism—building on shaky foundations produces shaky outcomes—is hard to argue with.

What this means for you: If your kid is struggling, resist the urge to focus on the current unit. Go back. Find where the chain of prerequisites broke. That's where the real work needs to happen.

What most schools do instead: Students advance by calendar, not by mastery. A kid who gets a C in pre-algebra moves into algebra on the same schedule as the kid who got an A. The system assumes the C student is "a little behind." In reality, they're likely missing foundational building blocks that make algebra nearly impossible to learn properly.

The Bloom Problem

Bloom's result raises an obvious question: what is it about one-on-one tutoring that produces such outsized gains? It's not just the presence of another person.

An important asterisk is that Bloom’s two-sigma figure has been debated. A 2011 meta-analysis of human tutoring studies found a mean effect size of d = 0.79—roughly 0.8 standard deviations, not two. That's still enormous (the average tutored student outperforming ~79% of conventionally taught students), but it’s not the figure Bloom originally reported.

The discrepancy comes from Bloom’s original study combining tutoring with mastery learning. When you isolate tutoring alone, the effect is smaller. This actually strengthens the argument rather than weakening it. It’s not just having a warm body sitting next to your kid that makes the difference. It’s the combination of individualization, mastery requirements, and evidence-based strategies that produces the outsized result.

A good tutor naturally implements every strategy we’ve discussed: they assess prerequisites (mastery learning), they quiz frequently (retrieval practice), they come back to old material (spaced repetition), they mix problem types (interleaving), they break things into small steps (cognitive load management), and they calibrate difficulty to the student’s current level. A mediocre tutor who just re-explains the same content won’t get you to two sigma. A system that implements all of these strategies will dramatically elevate outcomes—regardless of whether the precise number is 0.8 or 2.0. In either case, that is certainly better than what we are doing today.

Talent Development vs. Traditional Schooling

Around the same time he published the two-sigma problem, Bloom was conducting a separate but related line of research that would produce his book Developing Talent in Young People. He studied 120 extremely successful individuals across six talent domains and discovered a general process that was universally applicable.

The process that produces world-class performers in every field—music, athletics, mathematics, science—shares common features that are fundamentally incompatible with how schools operate. In talent development, instruction is individualized; students progress at their own pace; mastery is required before advancement; feedback is immediate and specific; practice is deliberate and targeted. In traditional schooling, instruction is group-based; students progress by calendar; feedback is delayed and infrequent; and practice is one-size-fits-all.

Bloom and Sosniak described the contrast in Educational Leadership (1981):

“In general, school learning emphasizes group learning and the subject or skills to be learned. Talent development typically emphasizes the individual and his or her development… The school schedule and standards are largely determined by the age of the child… Each member of the group engages in the same tasks, and it is expected that different students will learn skills to different levels.”

Bloom observed something that will resonate with any parent who has watched their kid’s soccer practice: schools do talent development well in athletics. Coaching is individualized, progression is based on ability, and practice is targeted. But they abandon these principles entirely for academics. As Bloom told an interviewer in 1985:

“They do a very good job in sports. There’s nothing we can tell coaches in high schools and colleges. But when we get beyond sports, things are sporadic, accidental. Students may have a good teacher one year and a very poor one the next… Schools do not seem to have a great tolerance for students who are out of phase with other students in their learning process.”

What this means for you: The question isn’t whether your kid is talented enough to benefit from individualized instruction. Bloom’s research shows that most kids have the potential to perform at dramatically higher levels. The constraint is the delivery system, not the child. The ultimate goal for every parent is to find a combination of resources, systems, and interventions that can collectively produce the effect of Bloom’s research.

The Foundation Problem

I interviewed Justin Skycak, author of The Math Academy Way and the Chief Learning Director at Math Academy, a company that has built a technology platform around these learning science principles. One thing he said stuck with me: knowledge in math is “relentlessly hierarchical.” If you’re missing a foundation block, everything built on top of it is unstable.

Math Academy’s approach uses a “knowledge graph”—a map of every concept and its prerequisites—to ensure no student advances past a gap. When a student struggles, the system traces back to the specific prerequisite that’s missing, rather than re-teaching the advanced concept. This is what a good human tutor does intuitively, but it’s nearly impossible in a classroom of 30 students.

Skycak also offered a key insight about why kids struggle that I think every parent should hear: struggle does not imply inability. In his words, struggle can be caused by missing foundations, ineffective practice methods, insufficient practice, or lack of motivation. The default assumption in schools—that a struggling student just “isn’t a math person”—is contradicted by the evidence.

So if we know how the brain works and we know how to apply this science to individual learning, why isn’t it happening in our schools?

What Your School Gets Wrong (And Why It’s Not Always Their Fault)

Schools don’t ignore learning science out of malice or laziness. The constraints of the system—one teacher per 25–30 students, age-based grouping, fixed schedules, mandated curricula—make it essentially impossible to implement what the research says works. Understanding why the system is stuck matters, because it tells you what can and can’t be fixed without structural change.

The One-to-Many Problem

Every strategy listed above is fundamentally about individualization. A teacher needs to diagnose each student’s specific gaps, calibrate difficulty to their current level, spacing review based on their forgetting patterns, and provide immediate feedback. A teacher with 30 students and 50 minutes cannot do this. It’s simply impossible.

Many teachers know this intuitively and do their best to work around it. Research on barriers to evidence-based practice in classrooms consistently finds that the obstacles are structural: time constraints, class size, mandated curricula, and lack of training. Teachers aren’t resistant to better methods. They would love to have their students do better. But, unfortunately for them, and unfortunately for our kids, they’re working within a system that makes better methods logistically impossible to implement fully.

What Sounds Scientific But Isn’t

While schools struggle to adopt strategies that work, they’ve enthusiastically adopted ideas that don’t. The most prominent is learning styles. The idea that some kids are “visual learners,” others are “auditory learners,” and instruction should be tailored accordingly has been thoroughly debunked, but still persists.

Pashler, McDaniel, Rohrer, and Bjork (2008)—four of the most prominent learning scientists alive—published a comprehensive review and concluded:

“Although the literature on learning styles is enormous, very few studies have even used an experimental methodology capable of testing the validity of learning styles applied to education. Moreover, of those that did use an appropriate method, several found results that flatly contradict the popular meshing hypothesis.”

Since that landmark 2008 review, the evidence has only stacked higher against learning styles. A 2020 study found visual learners outperformed auditory learners regardless of instruction modality — meaning one method was better for everyone. A 2025 review drawing on 17 meta-analyses found the effect size for the matching hypothesis was essentially zero (d = .04). The one apparent dissent, a 2024 meta-analysis, found a small positive effect of matched instruction, but only 26% of their outcome measures showed the crossover interaction that would actually validate the theory. One method was usually just better for everyone. Even those authors concluded it would be "an overreach" to recommend incorporating learning styles into practice.

And yet the myth thrives. A 2016 study found that 67% of teacher-prep programs still required students to build learning styles into their lesson plans. 59% of education textbooks endorsed the concept. Surveys of practicing teachers consistently show 76-93% agreement with the statement that students learn better when instruction matches their preferred style.

Learning styles have become the educational equivalent of homeopathy: no evidence, enormous institutional momentum, and a psychological appeal (who doesn't want to believe they have a unique learning fingerprint?) that makes it nearly impossible to kill.

In my interview with Skycak, he mentioned that he had tested various large language models on this exact question—asking them to explain the benefits of tailoring instruction to learning styles—and found that some “fell right into the trap” of endorsing the concept. That’s how deeply embedded the myth is: even AI trained on the internet’s corpus of knowledge reproduces it.

You might now be confused. “Wait—you’ve spent this whole essay arguing for individualized learning, and now you’re saying tailoring to individual differences is a myth?”

No. What the research debunks is tailoring to style (visual vs. auditory vs. kinesthetic). What the research supports is tailoring to level—each student’s mastery state, the specific prerequisites they’re missing, the amount of practice they need, and their pace of learning. The variable that matters isn’t how a student prefers to receive information but what they already know and what they’re ready to learn next.

Think of it this way: if you go to a doctor, the treatment should be individualized to your specific condition and history—not to whether you prefer pills or injections. Schools should individualize based on diagnosis, not preference.

The Calendar Problem

Perhaps the single most destructive structural feature of traditional schooling may be age-based grouping and calendar-based advancement. Bloom and Sosniak identified this explicitly in 1981:

“The group is central in the school learning process and only minimal adjustments are made for individual children. If the group as a whole has difficulty, the teacher will reteach the task or skill until some portion of the group has learned it. But generally, all the children are not expected to learn a task or skill to the same level.”

Students who haven’t mastered prerequisites are pushed forward. Students who have mastered the material are held back. For the student pushed forward, each new concept is built on a shaky foundation. For the student held back, years of unchallenging material teaches them that effort is optional. It’s a brutal combination.

A Note on What This Essay Doesn’t Cover

Everything we’ve discussed so far is about cognitive learning—how the brain acquires and retains knowledge and skills. This is critical, and schools largely fail at it. But it’s not the whole picture, and pretending otherwise would be dishonest.

Social and emotional development matters too. Durlak and colleagues (2011) conducted a meta-analysis of 213 school-based social-emotional learning programs involving over 270,000 students, published in Child Development. Well-implemented programs produced an 11-percentile-point gain in academic achievement—on top of improvements in social skills, attitudes, and behavior. The cognitive strategies and the social-emotional environment aren’t in competition. They’re complementary.

A student who feels safe, motivated, and connected will benefit more from retrieval practice and spaced repetition than one who doesn’t. A tutor who implements perfect spaced repetition but makes a kid feel stupid will fail. A warm and nurturing classroom that never challenges students cognitively will also fail. You need both.

This essay focuses on the cognitive science because it’s where the gap between what we know and what schools do is widest and most actionable—and because it’s the piece most parents have never been exposed to. But when you’re evaluating a school and the kids seem happy, engaged, and emotionally supported—that’s genuinely valuable. It’s just not sufficient. You should also ask: how does this school handle learning science? The warm feeling and the cognitive rigor are both non-negotiable. If you want your kid to succeed they have to feel loved, and they have to be challenged.

What This Means for Your Family

If you’ve made it this far, you now understand more about the science of learning than most teachers were taught in their training programs. That’s not your achievement and it’s not their failure—it’s an indication of how broken the pipeline between research and practice is. The question now is: what do you do with this knowledge?

What You Can Do at Home

You can’t reform your kid’s school. You probably can’t get the district to restructure its curriculum around spaced repetition and mastery learning. But you can do a remarkable amount at home:

Use spaced repetition tools. Free apps like Anki or Quizlet (in test mode, not just flashcard browsing mode) implement the science automatically. Have your kid add the most important concepts from each week and review them on a schedule.

Make homework retrieval-based. If your kid is studying for a test by re-reading notes, redirect them. Close the book and try to recall the material. Write down everything you remember. Check what you missed. Repeat. This is harder and less pleasant than re-reading so be prepared with some treats.

Mix it up. If your kid has 20 math problems to practice, don’t let them do all 20 of the same type in sequence. Shuffle the problem types. Yes, this will make it feel harder. Yes, they’ll get more wrong in the short term. And yes, they’ll remember more in the long term.

Look for the foundations. If your kid is struggling, resist the urge to focus on the current unit. Go back. Find where the chain of prerequisites broke. That’s where the real work needs to happen.

What Comes Next

This essay has shown you what the science says and why schools can’t seem to use it. But understanding the science is only part of the picture. A natural question remains: if we’ve known this for decades, why hasn’t the system changed?

The answer lies in how schools are funded, who controls the budget, and what incentives actually drive decision-making in education. It turns out that the financial architecture of American schooling creates its own set of constraints that make reform extraordinarily difficult—even when everyone involved has the best intentions.

That’s Essay 3.

In the meantime, you don’t have to wait for systemic change. You now understand what real learning looks like, how to test for it, and where to start supplementing what your kid’s school provides. No one is coming to save your kid. But now you know what saving them actually looks like.

The default assumption of most parents is “well, my school needs more money, and if they got it, then instruction would improve.” The evidence for this idea is shaky at best.

Your school district probably spends around $20,000 per student per year. It is a staggering sum compared to the rest of the world. The U.S. is 5th in the world in per-pupil spending and that money gets us the 35th position in math testing. This gap isn't a problem we can just buy our way out of. The Organisation for Economic Co-operation and Development has identified a cumulative spending threshold of roughly $75,000 per student between ages 6 and 15, beyond which additional spending shows no relationship to improved performance. Most well-funded American districts blew past that number years ago. At our spending level, the questions should be ones of incentives, rather than merely amounts.

In Essay 1, I showed you that American education is producing declining outcomes masked by inflated grades — and that the decline hits even the top performers in wealthy districts. In Essay 2, I showed you what the science says about how kids actually learn — retrieval practice, spaced repetition, interleaving, mastery learning — and why most schools don't use any of it.

This essay answers the obvious follow-up: why don't schools do what works?

The answer is largely in the budget. By determining where the money goes, but more importantly, what the funding structure makes possible and impossible for the people inside the system. We're going to follow the money, and when we do, we'll find two compounding problems: the incentives point teachers at the wrong students, and the resource constraints mean that even with perfect incentives, teachers couldn't implement the science of learning if they tried.

You Already Made a Choice

If you bought a house in a "good school district," you already chose a funding model. You may not have thought of it that way. But the mortgage payment, the property taxes, the bidding war you sparked over the property—all of those were tuition payments.

In 1956, economist Charles Tiebout proposed a simple theory: when local governments fund services through property taxes, families "vote with their feet" — they pick their neighborhood based on the public goods it offers. Decades later, economists confirmed he was right, at least when it comes to schools. American families sort themselves into districts by school quality, and they pay for the privilege through their mortgage.

How much do they pay? A 1999 study compared houses on opposite sides of school attendance boundaries that were in that same neighborhood, had the same commute, but went to a different school. A one standard deviation increase in elementary test scores raised home prices by 2.1%. A follow-up study in Florida found that when the state started publishing school letter grades, housing prices responded to the grades independently of the underlying test scores. Parents weren't actually paying for testing quality. Instead, they were purchasing the signal of quality.

We already have school choice—it's just the price of a house.

Unsurprisingly, the world is more complicated than the "be rich enough to buy into the best school district" strategy may suggest. What the property-tax model actually is, is a system where your kid's curriculum is a function of your neighbors' home values. If you're in a well-off neighborhood, you're on the "winning" end of that equation. But the same mechanism that gives your district more AP classes gives it zero structural guarantee that the money is well spent.

A high tax base buys more. It doesn't buy better. As we're about to see, even the most generously funded districts are hemorrhaging money before it reaches the classroom—and the strings attached to what does arrive may be pulling your kid's education in directions you haven't considered.

That's the first way the budget shapes the pedagogy. Your tax base determines whether your school can afford AP Physics, a gifted program, or a second language before middle school. Two kids in the same state, with the same grade, and same "public school" label will have radically different educations based on zip code.

But even if you're in a well-funded district spending top dollar, that doesn’t mean those dollars are being spent on helping your kid achieve all they can be.

Follow the Dollar

Let's trace your $20,000.

Of every dollar your district spends, about 60 cents goes to instruction. The other 40 cents goes to operations, administration, student support services, and a patchwork of other categories. Right away, nearly half of what you're paying never touches a classroom.

And that ratio has been stuck for decades. According to the National Center for Education Statistics, instruction consumed about 61% of current expenditures in the 2020–21 school year. The decline happened between the late 1960s and early 1990s, when the instruction share dropped from roughly 80% to 60%, and it never recovered. Per-pupil spending kept climbing, up 28% in real terms over that period alone, but the growth was consumed by expanding non-instructional services.

The staffing numbers show you where it went. Between 1970 and 2010, public school enrollment grew 7.8%. Non-teaching staff grew 151%. By 2015, there were more non-teaching staff than teachers in American public schools for the first time. Between 2000 and 2015, administrative staff increased 88% while students and teachers increased just 8%.

It is important to be precise about this data because not all of it is waste! Counselors, nurses, special education aides, and psychologists directly serve students, and special education's share of spending rose from 4% to 17% between 1967 and 1991, driven by federal mandates to serve students with disabilities. That's a genuinely important obligation. The concerning growth is in compliance and central administration, roles that don't put anyone in front of a child. But regardless of the reason, the aggregate picture is that a shrinking share of each dollar reaches the classroom.

This results in the person who matters most in the building—the teacher standing in front of your kid—getting squeezed from both directions. Admin takes more and more of their budget, while their pay shrinks. The Economic Policy Institute tracks what they call the "teacher pay penalty": the gap between what teachers earn and what comparably educated professionals earn in other fields. In 2024, that penalty hit a record high of 26.9%. Teachers earned 73 cents on the dollar compared to other college graduates. In 1996, they earned 94 cents. For male teachers, the penalty is 36%.

So here is your $20,000: about $12,000 reaches instruction. The teacher delivering that instruction earns a quarter less than she would in almost any other profession requiring a college degree. And the system keeps spending more every year without any more of it reaching them.

(Your district's expenditure data is public, by the way. The NCES School District Finance Survey lets you look up any district in America. You can find yours and check the instruction percentage. You might be surprised by what you find!)

But the allocation problem—bad as it is—isn't the most important way that funding shapes what your kid learns. That distinction belongs to something subtler and more powerful: the strings attached to the money that does reach the classroom, and the constraints on the teacher who is supposed to spend it.

Imagine You're the Teacher

Before we get to the strings, I want you to understand the person holding them.

According to a 2025 survey of American teachers, you work 49 hours a week—10 hours more than your contract says and 9 hours more than other working adults with your education level. You spend 28 of those hours in front of students. For comparison, teachers in Finland—a country that consistently ranks in the top five on international assessments—spend 18 hours a week teaching and work 32 hours total. They use the other 14 hours for lesson planning, collaboration with colleagues, and professional development. You don't have those hours. You are white knuckling every lesson, frantically trying to plan lessons and grade at the same time.

You have around 25 kids in your classroom. And it may be getting worse: as of the 2024-25 school year, 29% of schools reported increasing class sizes because they couldn't fill open teaching positions. Evidence-based methods like mastery learning require a teacher who can assess and respond to individual students. That's hard with 15 kids. With 28 or 30, it's a fantasy.

Worse, you earn about $73,000 a year. Professionals with your same degree earn about $103,000. More than half of your colleagues report burnout. One in eight teaching positions in America is either unfilled or filled by someone without a full certification—411,549 positions nationally.

This is the part that really here's the part that matters for a kid's education. You know — because you care, and you're reading about it on your own time — that the science of learning says retrieval practice and spaced repetition work better than re-reading and lectures. You want to try interleaving in your math class. But your district has a pacing guide that tells you which chapters to cover each week. You have standardized assessments coming in six weeks that will determine your school's rating, your principal's evaluation, and quite possibly whether you keep your job. The data backs up this view. When surveyed about barriers to implementing better methods, 65% of your peers cited lack of time, 61% said too many competing demands, and 53% said there simply isn't enough flexibility in the school year.

So you teach what the pacing guide says, the way the pacing guide implies, aimed at the students the accountability system counts. The funding system has made knowing better irrelevant. And that's if you're one of the few who know better at all.

Your Kid Is Invisible

If your child is in a well-funded suburban district and already reading above grade level, here is what the accountability system thinks about them—absolutely nothing. They don't exist.

Under the accountability rules created by No Child Left Behind, which measured schools primarily by the percentage of students reaching a "proficient" threshold on standardized tests, schools concentrated resources on students near that cutoff. The bubble kids who could be pushed over the bar got targeted attention. Students well below the bar got left behind. Students already above it became invisible.

A 2010 study of Chicago Public Schools confirmed the pattern, showing that NCLB's proficiency incentives led schools to systematically neglect students at the top and bottom of the distribution—math gains of 0.27 standard deviations in the middle of the distribution, zero gains for the lowest- and highest-achieving students.

For your dedicated, above-average kid, the school gets zero funding credit for moving them from the 80th percentile to the 95th. It gets full credit for moving the kid next to them from "below basic" to "proficient." The funding formula determines who the teacher pays attention to. If your child is already above the proficiency bar, the funding and school accountability system creates almost no incentive to push them further.

This is the mechanism behind the finding from Essay 1 that even at the 90th percentile, 13-year-olds' math scores dropped 7 points between 2012 and 2023. The average school district receives $3.38 per student in state gifted funding. For comparison, Title I—the federal program targeting low-performing and low-income schools—spends about $316 per pupil. Federal special education funding adds another $1,578 per student. Those programs serve incredibly important purposes. But the ratio tells you what the system values: pushing a kid from struggling to proficient is worth 93 times more, per dollar, than pushing a kid from good to great. And only 20% of districts receive any gifted funding at all. The budget decided who matters — and it decided your kid who is doing good enough at school is a lower priority.

Has it improved since NCLB was replaced by ESSA in 2015? According to the Fordham Institute's analysis of all 51 state ESSA accountability plans, only four states—Arkansas, Colorado, Idaho, and Oregon—base at least 50% of school ratings on growth measures for all students. The rest still optimize around proficiency thresholds.

How did we get here? After all, the federal government has historically provided roughly 8–10% of K-12 funding through programs like Title I. That's not a huge share! But the compliance requirements attached to that money reshape the behavior of every school that receives it—which is nearly all of them. Federal standards have resulted in local behavioral distortion. We are 20+ years into a systemic reorientation of teaching, attention, and resources in response to what the accountability metrics reward.

Donald Campbell, a social psychologist, observed something in 1979 that we should have tattooed on the forehead of every education policymaker: "The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor." When you judge a school by one number, the school optimizes for that number, even if it means distorting everything else. Jacob and Levitt estimated that 4–5% of elementary classrooms involve serious standardized test cheating every single year. That's Campbell's Law at its most literal.

These distortions can even reshape what gets taught. A survey found that 62% of school districts had increased time on language arts and math at the elementary level since NCLB's passage, while cutting other subjects—science alone lost an average of 75 minutes per week. Roughly 71% of districts had reduced instructional time in at least one subject area, including history, arts, languages, and music. Daniel Koretz, a Harvard testing expert, wrote an entire book arguing that high-stakes testing "corrupted instruction" and produced "sham increases in scores" with "precious little" real improvement in learning.

The extreme version of this was the Atlanta Public Schools cheating scandal in 2011. 178 educators systematically altered test answers, inflating proficiency rates by as much as 50%. That was criminal, and obviously not what is supposed to happen. But the everyday version is subtler and universal. Schools teaching to the test format, narrowing the curriculum to tested subjects, directing teacher attention to the students who move the accountability needle. This is great for those children on the bubble, but for those who are in need of drastic help, or for those who have the potential to soar even higher, they end up being lower priority.

That's the first problem — the wrong target. The incentive structure points teachers at the students who move the school's rating, not the students who could learn the most. But there's a second problem that compounds on top of it, and it's the one that connects directly to Essay 2.

Why Don't They Just Use What Works?

Even for the bubble kids—the students the system is optimizing for—schools overwhelmingly do not use the instructional methods that the science of learning says are most effective. This was the question Essay 2 left hanging: we know retrieval practice works, we know spaced repetition works, we know interleaving works. Why don't schools use them?

The answer starts in the places where teachers learn to teach. Teacher preparation programs focus on classroom management, lesson planning templates, and educational philosophy. The science of how people actually learn — the stuff from Essay 2 — is largely absent.

What fills the gap is worse than nothing. A 2012 study by Dekker and colleagues surveyed teachers in the UK and Netherlands and found that 93% believed students learn better when taught in their preferred "learning style" — a theory that has been thoroughly debunked by cognitive scientists for decades. A 2020 review by Newton and Salvi synthesized 37 studies covering more than 15,000 educators worldwide and found that 89.1% endorsed learning styles as valid. The overwhelming majority of the profession believes something that is demonstrably false—and was taught to them by the institutions that certified them to teach.

So the teachers who are supposed to implement learning science don't know learning science. But couldn't professional development fix that?

In theory, yes. In practice, TNTP studied this question in 2015, tracking 10,000 teachers across three large school districts and two charter management organizations. Districts were spending an average of $18,000 per teacher per year on professional development. The result was that only 3 in 10 teachers improved their practice over a two-to-three-year period. Separately, economists found that marginal investments in teacher professional development had no statistically significant effect on student achievement in math or reading.

Why doesn't professional development work? Because most of it isn't about instruction. It's about compliance — new testing protocols, updated documentation requirements, software training for the district's latest data system. The professional development budget, like the rest of the education budget, gets eaten by the system's administrative overhead before it reaches pedagogy.

The cruelest part is that even when individual teachers do learn about retrieval practice or interleaving— likely on their own time, through their own reading—the system makes implementation nearly impossible. Their district's pacing guide dictates what gets taught each week. The standardized tests are written in formats that reward recognition (multiple choice) rather than retrieval. Adopting mastery learning, where students don't advance until they've demonstrated understanding, is structurally incompatible with a calendar that says every student must be in the same place at the same time for a state assessment on a fixed date.

There's one more layer to this that I find particularly damning. Even when a method works, it often feels wrong to both teachers and students. Remember the Harvard physics study from Essay 2 — students in active-learning sections scored significantly higher on tests of understanding, but reported feeling like they learned less than students in polished lectures. Actual learning and perceived learning were inversely correlated. Teachers face this same perception gap. Without training in why confusion is productive, and without institutional support to weather the awkwardness that comes from implementing learning science, teachers rationally avoid the methods that work best.

So let's name the two problems clearly.

Problem one is the wrong target. The incentive structure created by accountability systems points teachers toward proficiency thresholds, making high-performing and low-performing students invisible. The budget decides who matters.

Problem two is the wrong tools. Even for the students teachers are focused on, they're using instructional methods that cognitive science has shown to be less effective — because nobody taught them the right ones, the system doesn't give them time to learn, and the pacing and testing infrastructure makes better methods nearly impossible to implement.

A randomized controlled trial in Tanzania tested what happens when you address each problem separately versus both together — giving teachers performance incentives alone, additional resources alone, or both. The combined treatment produced effects significantly greater than the sum of the individual treatments. Neither better incentives nor more resources alone was enough. You need both. American schools have neither.

The funding model and the science of learning are pulling in completely different directions, and the budget wins every time, because the budget is what keeps the doors open.

Every Model Has Strings

If you've made it this far, you might be thinking: OK, the traditional funding model isn't optimal. What about the alternatives? How does their funding change things?

Let's look at all of the options and examine what each model's funding structure actually optimizes for.

A quick taxonomy, since these terms get conflated constantly.

Voucher and ESA programs redirect public dollars to families, who then spend them at schools or on educational services of their choosing — the child leaves the public system but the funding follows.

Charter schools are publicly funded but independently operated — they're still public schools, free to attend, but run outside the traditional district structure.

Private schools are funded entirely by tuition and donations, with no public dollars unless a voucher or ESA program provides them.

Homeschooling is parent-directed education outside any institutional school — the family assumes full responsibility for curriculum, instruction, and assessment, with regulatory requirements varying wildly by state.

Vouchers and ESAs. The recent belle of the ball is school choice programs — vouchers, Education Savings Accounts, charter schools — which have exploded over the past five years. In 2020, only three states had ESA programs. Today, more than 20 states have them, representing roughly $9 billion in total funding. In July 2025, the first-ever federal school choice program was signed into law. COVID shattered parental trust in public schools, and the political landscape shifted accordingly.

The theory is straightforward: if parents can direct education dollars, schools will compete for those dollars, and competition will drive improvement. This funding is inherently limited however — ESAs only redirect the state portion of per-pupil funding. In Iowa, total spending is about $16,000 per student, but the ESA is only about $8,000. Parents are directing roughly half the money, not all of it. Joe Connor, who runs Odyssey — a company that manages ESA programs for six state governments and 80,000 students — told me that most families spend 80-85% of their ESA on tuition, with the rest going to supplemental services. "In older programs like Florida, parents are increasingly unbundling — buying individual tutors, online courses, and specialists rather than picking a single school."

So does choice work? The honest answer is that it depends entirely on the specific design features of the program. Essentially, it comes down to whether the school choice program attracts good schools competing for students, or bad schools desperate for them.

Where choice has clearly worked: The Milwaukee Parental Choice Program, the oldest voucher program in the country, shows modest test score effects at best. But the long-term outcomes tell a different story: 4-7 percentage point increase in high school graduation, higher college enrollment, a 53% reduction in drug convictions, and an 86% reduction in property damage convictions. The DC Opportunity Scholarship Program — a true randomized controlled trial — found 82% of scholarship recipients graduated high school versus 70% of the control group. There's also evidence from Florida that competition from choice programs makes traditional public schools slightly better. In all three cases, quality private schools with existing reputations participated in the programs. Parents had quality signals to choose from.

Where it failed: In Louisiana, a rigorous study found that voucher participation lowered math scores by 0.4 standard deviations and increased the probability of failing by 50%. The reason: the private schools that chose to participate were disproportionately low-quality schools that needed the students, not high-quality schools competing to attract them. The program's design created a market that selected for the worst providers. In Indiana, the largest voucher program in the country with 70,000 students, first-year math scores dropped 0.15 standard deviations and the negative effects persisted — a pattern consistent with rapid scaling that outpaces quality supply.

The most comprehensive review of the voucher evidence concluded that "Evidence to date is not sufficient to warrant recommending that vouchers be adopted on a widespread basis." But also: "In some settings, or for some subgroups or outcomes, vouchers can have a substantial positive effect." Program design matters more than program type.

Private schools. What about just paying tuition at a private school? Lubienski and Lubienski analyzed NAEP data and found something that will surprise most parents: the private school achievement advantage disappears entirely when you control for demographics. In some comparisons, public school students slightly outperformed their private school peers in math. The "private school premium" is largely a selection effect — the families who can afford private school tuition bring advantages (education, income, engagement) that would produce good outcomes in almost any setting.

Private schools have their own funding-model distortion. Their revenue depends on tuition, which depends on enrollment, which depends on parent satisfaction. That creates a structural incentive to deliver what parents want — impressive facilities, extensive extracurriculars, strong college placement statistics — which doesn't always align with what produces the most learning. The same gap between actual learning and perceived learning applies here. A market-responsive school has every reason to make parents feel their children are thriving, whether or not the underlying pedagogy is evidence-based.

Charter schools. Stanford's CREDO studies — the most comprehensive charter school research available — found that charter students gain an average of 6 additional days of learning in math and 16 days in reading per year compared to matched traditional public school students. That's modest. But the average hides enormous variance. High-performing networks like KIPP, Success Academy, and Achievement First produce much larger effects — roughly 0.25 standard deviations in math. They do this with highly structured, intensive methods and a culture of relentless focus on results.

The charter funding model creates its own pressures. Schools that depend on enrollment for funding must attract and retain families. This can drive genuine innovation and responsiveness. It can also drive marketing over substance, grade inflation to keep parents happy, or — in the other direction — pushing out students who bring down test scores. The optimization target is enrollment, and enrollment follows parent perception, which we've already established isn't a reliable proxy for actual learning.

Homeschooling. The population of children receiving academic instruction at home — whether through homeschooling or full-time virtual programs — has exploded, from roughly 3.7% of school-age children in 2018-19 to about 5.2% by the 2022-23 school year, with some estimates running even higher. The uncomfortable truth about the evidence for this approach is that it barely exists. The most-cited homeschool researcher, Brian Ray of the National Home Education Research Institute, draws his samples predominantly from HSLDA membership — a highly self-selected, motivated, well-resourced population. There are zero randomized controlled trials of homeschooling outcomes. We genuinely do not know how the average homeschooled student performs, because the average homeschooled student isn't being measured. The one well-controlled study that exists — Martin-Chang et al., matching 37 homeschooled students against 37 public school peers and controlling for income and parental education — found that structured homeschoolers outperformed public school students across every subtest, while unstructured homeschoolers scored the lowest of any group. Even the best evidence we have points to variance, not a verdict.

Homeschooling's funding model distortion is the inverse of public school's. It optimizes entirely for parent preferences and capacity, which creates extraordinary variance. A parent who is educated, has time, and is genuinely engaged can deliver something close to the one-on-one tutoring ideal. A parent who lacks those resources, or who pulled their kid out of school for reasons unrelated to educational quality, may produce something far worse than the public school they left. There is essentially no quality floor, no accountability mechanism, and no way to identify a child who's falling behind.

The pattern. Every model optimizes for something, and every optimization creates its own distortion. Traditional public schools optimize for accountability metrics — proficiency rates — and we've seen what that produces. Voucher and ESA programs optimize for parent choice, which works when quality signals exist and fails when they don't. Private schools optimize for parent satisfaction, which can mean excellence or can mean expensive signaling. Charter schools optimize for enrollment, which can mean innovation or can mean marketing. Homeschooling optimizes for parent capacity, which creates the widest variance of any model.

None of these models automatically align with what the science of learning says produces the best outcomes. And there is no model where you can simply trust the system to get it right without paying attention.

Choose Deliberately

Before I close, there's a counterargument this essay would be dishonest to ignore.

The strongest case for the traditional public school model is social cohesion. Public schools were supposed to be a place where children from every background share a common experience, forming the connective tissue of democratic life. If everyone splinters into niche schools optimized for individual preferences, you might gain academic efficiency and lose something harder to measure but genuinely important. My own education was like this. I grew up in rural Minnesota and my high school had kids from every walk of life, and it exposed me to many different types of people I never would've met normally.

Still, there's an obvious rebuttal. Property-tax-funded schooling already sorts students by zip code and income. Most people don't only have one high school in their town. The integration Mann envisioned is, in many places, a fiction in practice. The system already produces its own version of the fragmentation he feared.

But acknowledging that the current system falls short of the ideal doesn't mean the ideal is worthless. Whether the civic function of shared public schooling outweighs the potential benefits of alternatives is a values question, not an empirical one. I can't resolve it for you. I don't think anyone can. But I think you should understand the weight of it before you decide.

What the data can tell you is that you are already making a funding choice. You have been since the day you chose where to live. And the funding model you defaulted into comes with strings that shape your kid's education in ways you probably haven't considered.

The budget is the pedagogy.

The funding source determines the curriculum, the incentives, and the optimization targets. It determines which students the teacher pays attention to, which subjects get time, and whether the science of learning has any chance of making it into the classroom. In every model, the economic structure tells you what the school will optimize for — and in no model is that "what cognitive science says works best."

Every model has strings. The question isn't whether strings exist. The question is which strings you're willing to live with, and what you're going to do about the ones you're not.

Now that you understand how the funding model shapes what schools offer, a natural question follows. Your kid's school bundles instruction, socialization, childcare, meals, counseling, sports, and arts into one package, for one price, on one schedule. How did that bundle get assembled? Does it have to stay that way?

That's Essay 4.

Here is everything your child received from school today: 50 minutes of math, 50 minutes of English, a science lab, a history lecture, 22 minutes of lunch (consisting of a chicken patty molecularly identical to cardboard, a carton of milk, a fruit cup that is 90% corn syrup, all of it subsidized by the USDA), a vision screening from the school nurse, a meeting with the guidance counselor about feelings, 30 minutes of PE, a fire drill, a pep rally for the football team, and a hallway conversation about Taylor Swift that constituted roughly 80 percent of their social development for the day.

That is "school."

We treat it as a single thing. We talk about whether our kids are getting a "good education." We weigh school choices as if they were singular products that we can easily swap out. But it is a dozen services wrapped in one building, funded by one budget, run by one principal, and delivered on one schedule. Your child's school is simultaneously an academic institution, a restaurant, a health clinic, a therapy office, a sports league, a job training center, a performing arts venue, and a daycare. Very few institutions in American life attempt anything this broad — and none do so with compulsory attendance, public funding, and an all-ages population of children. Your dentist does not also run a basketball league and a tattoo parlor. But your school does all of these things and does it with hundreds of little ones bouncing off the walls.

With that framing, it's amazing that the schools function at all.

That this arrangement feels so natural is credit to how much history shape our perception of what constitutes normal. Most parents, including myself, have never thought to question it. Of course schools serve lunch. Of course schools have counselors. Of course schools run sports programs. Where else would these things go?

In the last essay, I showed you that roughly 60 percent of your school's budget goes to instruction and 40 percent goes to everything else. Now I want you to see what "everything else" actually means. American schools have a list of distinct services, each with its own legacy, its own justification, and its own (often thin) evidence base. In understanding the components of the bundle, you’ll then know what parts best serve your own child, and whether you should even try to separate them. Shoot, even the phrase "helping your child learn" is slippery. Each aspect of this bundle of services aims at different outcomes (test scores, lifetime earnings, behavior, health, socialization) that do not always move together.

My goal for this essay is to convince you to stop treating "does school work?" as a single question with a single answer, and to start asking instead which parts of school work for my child, which parts don't, and what am I now going to do about the difference?

Seeing the bundle

Let's start with the biggest invisible service. Your school is a childcare operation. For most American families, this is the core function. The school day exists, in large part, so parents can work. We do not usually say this out loud because it sounds cynical. It is not cynical! It is honest. Any pandemic parent who tried to "work from home" with children can tell you, with haunted eyes, how important the childcare portion of the service is. When COVID closed schools, mothers who could not telework became 18 percentage points less likely to be employed. Part of the economic stumble from COVID came because the infrastructure that allowed parents to hold jobs disappeared overnight. The childcare function of school is so deeply embedded that we only recognized it when it was taken away.

Then there is nutrition. The National School Lunch Program serves approximately 4.8 billion meals per year at a cost of $17.7 billion. For children from food-insecure households, school meals can constitute the majority of their daily nutritional intake. During COVID, when school meal distribution shifted to grab-and-go sites and bus-route deliveries, the desperate improvisation revealed how many families had been relying on the school cafeteria as a lifeline.

There is health. School nurses, mental health counselors, dental screenings, vision checks, and in some districts full-service school-based health centers that function as primary care clinics. In many communities, the school nurse is the only healthcare professional a child sees regularly.

There is socialization, which is arguably the service children value most and parents worry most about losing. Learning to navigate peer hierarchies, cooperate with people you did not choose, handle conflict, develop friendships, manage social status. None of this appears in any curriculum document, but ask a child what happened at school and they are more likely to mention their friends than their math class.

There is civic formation: the diffuse, unspoken mandate to produce citizens who can function in a democracy. There are sports and extracurricular activities, which for many students define their entire school experience. There is career and technical education, from the auto shop to the coding elective. There is special education, which serves roughly 15 percent of public school students under federally mandated Individualized Education Programs. And there are the core academics themselves: math, English, science, history, foreign language, and the arts.

That is what your property taxes, state income taxes, and federal funding pay for. We are not buying "education." We are paying for a bundle of at least twelve services, many of which have no intrinsic relationship to one another, delivered by a single institution that was never designed to do all of these things at once.

So let's discuss the one thing that we know universally improves the bundle’s performance, teacher quality.

Teacher quality is the one finding that sticks

Imagine two third-grade classrooms in the same school. The kids come from the same neighborhoods. The classrooms have the same textbooks, the same schedule, the same number of desks. The only difference is the adult at the front of the room.

A 2014 study followed one million plus children into adulthood. What they found is the closest thing education economics has to a proven result: a single year with a teacher one standard deviation above average raises a child's lifetime earnings by about 1 percent. Replacing a bottom-5-percent teacher with an average one increases the present value of that classroom's future earnings by approximately $250,000. Over thirteen years of schooling, the cumulative variation in teacher quality your child encounters is, in all likelihood, the single largest in-school determinant of their economic future. No other variable — not the curriculum, not the building, not the technology — comes close.

The complicated thing is that we can't actually tell you what those good teachers are actually doing. That study measured the output of "did the students earn more at 28" without opening the black box that is the classroom. "Teacher quality" is a statistical residual in the model. It is everything about the teacher that the model cannot explain with demographics and prior test scores. It could be content knowledge, or warmth, or a wry sense of humor, or who knows what.

The Gates Foundation spent three years and studied 3,000 teachers trying to answer exactly that question. Their Measures of Effective Teaching project tested every approach they could think of: classroom observations by trained raters, student achievement data, and even student surveys. Harvard's Ronald Ferguson designed those surveys around what he called the 7Cs: Does this teacher care about me? Do they challenge me? When I am confused, do they clarify? Do they captivate my attention? Do they check whether I actually understood, or do they just move on?

It turns out kids know what a good teacher looks like. Student ratings on these dimensions reliably predict achievement gains. A one-standard-deviation increase in how "challenging" a teacher is predicts 31 percent more annual learning. The students in the classroom have better information about teacher quality than the principal down the hall, the school board across town, or the parent at the kitchen table checking grades on their phone.

And notice what all seven aspects of the 7Cs have in common. They are all forms of individual attention. Clarifying requires knowing what this child misunderstood. Challenging requires knowing what this child is ready for. This is where the finding connects back to something we covered in Essay 2: Bloom's famous "2 Sigma" result, which showed that one-on-one tutoring with mastery learning produces a two-standard-deviation improvement over conventional instruction. That finding is about what happens when a teacher can do the things that make teaching effective — assess understanding in real time, correct errors immediately, and then adjust to the individual child. With one student, that is natural. With 28, it is physically impossible.

Now, to be clear, nobody has proven that individual attention is the mechanism that explains the Chetty results. The 7Cs predict achievement gains, and Bloom showed tutoring works, and individual attention is the thread that runs through both, but these are three separate findings that point in the same direction, not one finding that proves a causal chain. Still, the convergence is hard to ignore.

The Tennessee STAR experiment — the only large-scale randomized trial on class size ever conducted in America — adds another data point. Students randomly assigned to classes of 13 to 17 significantly outperformed those in classes of 22 to 25. The gains were largest for minority and low-income students. The most plausible explanation is that smaller classes allow teachers to do more of the things that make them effective — more individual assessment, more real-time correction, more personalized challenge. A brilliant teacher with 30 kids is still brilliant. She just cannot do most of the things that the 7Cs research suggests matter most.

So what does this mean for you, the parent? It means the questions worth asking are not the ones most parents ask. Forget "is the school highly rated?" and "does my child like their teacher?" It is a question about the teacher's ability to, well, teach.

The frustrating part for me as a parent is that the system is designed to prevent you from using this information. No school publishes teacher effectiveness data. Value-added scores exist and they work, but they remain politically radioactive. If your child's teacher is in the 20th percentile of effectiveness, nobody will tell you. If they are in the 95th, nobody will tell you that either. Mostly this happens because schools don’t measure this sort of performance, and if they do, they don’t allow parents to surface the results, otherwise parents would force their children into the “good” classroom.

The system gives you the least transparency about the variable that matters most. Even imperfect information would be better than none — a parent who knows their child's teacher is in the bottom quartile could push for a transfer, seek supplemental tutoring, or simply pay closer attention at home. Instead, you are left with proxies like school ratings and per-pupil spending that tell you more about the neighborhood's income than about your child's actual educational experience. And even well-funded schools suffer from the grade inflation problems I discussed in the first essay, making it near impossible to truly assess whether a school is doing its job.

This pattern — where the most important information is the least visible — will repeat throughout the bundle. Still, teacher quality has the strongest evidence of anything in the bundle, in part because teacher effects are universal enough to study at scale. A great teacher can teach any subject, so researchers can measure their impact across thousands of classrooms. For the individual components of the bundle, the picture gets murkier.

The unglamorous interventions that actually work

Here is a finding that will surprise most parents: some of the strongest causal evidence for any non-academic service in the bundle belongs to the cheapest and least prestigious interventions.

One study found that improvements in school lunch quality led to higher test scores. In Norway, a 2018 study of a school breakfast program as a natural experiment found it increased long-term earnings and educational attainment. And a randomized trial providing free eyeglasses to students with poor vision in Florida's Title I elementary schools found achievement gains of 0.1 to 0.16 standard deviations at roughly $434 per classroom.

Feeding kids and making sure they can see the board. These are not glamorous interventions. They do not require new pedagogical frameworks or expensive software platforms or professional development retreats. But they have better causal evidence behind them than most of what schools spend their time and money on. If you wanted to design a school from scratch based purely on what the evidence supports, you would start with a good teacher, a good meal, and a pair of glasses. I asked one Chief Scientist at one of the largest ed-tech companies in the country if it would be more effective to have a good lunch or a good app and her answer, to her credit, was "a good lunch and a good friend at school is more important than any app will ever be." She is not guessing. A 2020 study used causal methods to show that a single good friendship link raises a student's GPA by a quarter of a standard deviation—an effect size that most ed-tech interventions would kill for. A 2024 meta-analysis of 15 studies confirmed the pattern. Friendship consistently predicts academic performance regardless of region or methodology, while research on the flip side shows that loneliness in school predicts depression, anxiety, and dropout.

But let’s assume your kid has lunch, glasses, and a best buddy they share a secret handshake with. What does the science say about what portion of the bundle’s services will help them the most? Let's give an example of what usually happens when you try to examine one portion of the school bundle.

Arts education is a case study in how correlation gets sold as causation

If you have ever heard someone argue that music lessons make kids better at math, you have encountered one of the most persistent misrepresentations of research in American education. (I have fallen prey to this urban legend, my poor baby was relentlessly subjected to Mozart in her first year). It is worth understanding this myth not just because it is wrong, but because the way it is wrong illustrates a problem that runs through almost every component of the bundle.

The best evidence comes from the Houston Arts Access Initiative, a cluster-randomized controlled trial involving 10,548 students across 42 schools. Arts education significantly improved writing scores—by 0.13 standard deviations, roughly equivalent to moving from the 50th to the 55th percentile. That's a meaningful bump from a well-designed experiment.

What they did not find was any significant effect on reading, math, or science scores. Zilch! Zero!

Harvard's REAP project went further, conducting systematic meta-analyses specifically designed to test whether studying art transfers to non-arts cognitive performance. Their conclusion was that there was little evidence for broad transfer. The only reliable causal connections they found were music instruction improving spatial reasoning and classroom drama improving verbal skills.

So where does the "music makes you smarter" claim come from? Correlational studies. (You'll see these everywhere in educational research). Scientists observe that students who participate in arts programs tend to score higher on standardized tests. From this observation, they conclude (or, more often, imply) that the arts participation caused the higher scores. But this is like concluding that joining a gym makes you healthy. People who join gyms are, in fact, healthier. But much of that is because health-conscious people join gyms, not because the membership itself transforms couch potatoes. Similarly, the kinds of students who enroll in arts programs tend to come from higher-income families, attend better-resourced schools, and have more engaged parents. The arts participation is a marker of advantage, not — or at least not primarily — a cause of achievement.

I am not saying arts education is worthless. I plan on having my child in piano lessons, just like I did growing up. After all, that Houston study found genuine benefits. There is an intrinsic value to aesthetic experience and creative self-expression that no standardized test measures. But the specific claim that parents hear most often, the one that gets deployed every time a school board threatens to cut the music program, is not supported by rigorous evidence. And the fact that correlational data is routinely presented as if it were causal, not just in arts education but across the entire bundle, is a problem that parents deserve to understand. You can find similar arguments ranging from physical education to technical education. Each has suggestive evidence, but none of it survives the kind of rigorous testing that would let you say with confidence: this is helping my child learn.

For the majority of the bundle, the honest assessment is some version of: we have suggestive evidence, but we cannot cleanly isolate the effect, and the version of this service your child actually receives may not resemble the version that was studied. Teacher quality has strong causal evidence, but parents cannot observe it. Nutrition and vision have strong evidence, but they are among the cheapest and least prestigious services in the bundle. Arts, SEL, sports, and extracurriculars rest on correlational evidence that does not always survive rigorous testing.

Why is this? Why, after decades of educational research and billions of dollars in spending, can we not answer the basic question of which components of the school day are helping your child and which are just filling time?

The answer is the bundle itself.

Why the bundle makes all of this nearly impossible to figure out

When a single institution simultaneously delivers instruction, nutrition, socialization, mental health support, physical activity, and custodial care, isolating any one component's contribution to student outcomes requires assumptions that are rarely met and often untestable.

Todd and Wolpin (2003), in the definitive methodological paper on education production functions, demonstrated that cognitive achievement depends on "the entire history of family and school inputs and on innate ability." To estimate the effect of any single input, you need data on all the others, across all prior years, including inputs that are unobservable (innate ability, parental effort, or peer quality for example). The commonly used statistical models handle this by imposing assumptions about how past inputs decay over time, but these assumptions cannot be empirically verified.

Even if we could isolate individual components, we would face a second problem: schools produce multiple outputs that do not correlate with each other, and we mostly measure only one of them.

A 2018 study showed that teacher effects on test scores and teacher effects on non-test-score outcomes (attendance, suspensions, course grades, on-time grade progression) are "weakly correlated." A teacher who dramatically reduces suspensions and improves student behavior might show zero measurable effect on standardized math scores. A teacher who raises test scores might have no effect on whether students show up to school or stay out of trouble.

This means we are evaluating the bundle by measuring one of its many outputs and ignoring the rest. A school that excels at socialization, reduces violence, and keeps kids healthy might look mediocre on a state report card that measures only reading and math proficiency but could be exactly what a community that struggles with truancy needs. The metrics we use to evaluate "school" are too narrow for an institution this broad.

The bundle makes the school eat itself

Here is one final finding that connects the identification problem directly to the bundle. Researchers at Brown University studied how much instructional time schools actually deliver, conducting a detailed case study of the Providence, Rhode Island public school district. They found that elementary students lose roughly 16 percent of allocated instructional time, middle school students lose about 21 percent, and high school students lose approximately 25 percent. For a high school student, that is the equivalent of roughly 45 school days per year.

Where does the time go? The biggest drivers are unexcused student absences and teacher absences covered (or not covered) by substitutes. But even when everyone shows up, the hours erode. In a companion study, Kraft observed classrooms with a stopwatch and found that the typical Providence classroom experienced over 2,000 external interruptions per year — roughly 15 per school day. That includes intercom announcements, phone calls, staff and student visitors, and the single largest category: late-arriving students whose entrance disrupts the room. Collectively, the researchers estimated they cost 10 to 20 full school days of instruction per year.

This is a story about what happens when you ask one institution to do twelve things. The fire drills are required by law. The counselor pullouts serve real student needs. The testing windows are mandated by the state. The assemblies serve the sports and extracurricular programs. Each interruption has its own justification. But collectively, they consume a quarter of the time that was supposed to be spent on instruction. My critique of this bundle is not solely epistemological; the bundle makes instruction itself harder. The services compete for the same finite resource of your child's time.

None of this was designed

Every component of the bundle was added for a specific reason. The reasons were political, cultural, or military. They were almost never scientific.

School nurses arrived in the 1890s to control contagious disease during mass urbanization. PE was mandated because the military worried boys could not fight. The National School Lunch Act of 1946 was simultaneously a child welfare program and a farm subsidy—Truman signed it to feed kids and "encourage the domestic consumption of nutritious agricultural commodities." Vocational education came from industrialization anxiety. School counseling exploded after Sputnik because the National Defense Education Act needed a talent pipeline for defense and aerospace. Special education was mandated in 1975 because schools had been excluding 1.8 million children with disabilities entirely. Each addition was in response to a real crisis. Each part of the bundle was, mostly, reasonable and justifiable. And yet not one of them was added because someone studied whether it would help children learn more.

Even more damning, essentially none have ever been removed. The bundle only grows. Congress promised to fund 40 percent of the additional cost of special education. Federal funding has never exceeded 18 percent. The gap comes out of the same budget that pays for everything else. This is the pattern at every layer—a new mandate arrives, a new service gets wedged into the building, and the time and money to support it get carved from whatever was already there.

Again, just to be very clear, I am not arguing that we need to cut special education or that our kids need to be calculator maxing math devotees, practicing equations at the expense of all else that brings childhood richness and meaning. What I am saying is that we have asked schools to do everything for everyone so we shouldn’t be surprised that the core promise is slipping.

As David Tyack and Larry Cuban documented in Tinkering Toward Utopia, Americans have translated their cultural anxieties and hopes into dramatic demands for educational reform, treating schools as panaceas for every social problem. The Institute for Family Studies calls this institutional "mission creep" so extensive that schools function as social services hubs that happen to teach some reading and math on the side.

Picking apart the bundle

Jim Barksdale, CEO of Netscape, had a line he liked to repeat: "There's only two ways I know of to make money — bundling, and unbundling." He meant it as an observation about technology companies. But it describes the history of American education with uncomfortable precision. For 170 years, our schools have only ever bundled. They have never once unbundled.

In every other industry, the pattern eventually reverses. Cable television bundled 190 channels into a single package. The bundle forced non-sports fans to subsidize sports fans, niche viewers to subsidize mainstream ones, and everyone to pay for content they never consumed. Then Netflix arrived, and the bundle cracked. You’ve likely heard this story before.

But what most people are unfamiliar with is what followed, and it is the part that matters most for schools. Economists modeled what would happen if cable went fully à la carte. Their finding was that consumers would not actually be better off. The reason is that unbundling triggers renegotiation of prices—content providers who lose guaranteed distribution demand more per subscriber, and those costs get passed through.

Unbundling the cable package did not give consumers freedom. It gave them subscription fatigue. The bundles broke. Then they re-formed, under new names, on new terms, run by new companies.

The school bundle has the same economic structure as the cable bundle—a dozen services, wildly different in quality and relevance, packaged together because the distribution technology (a building, a bus route, a tax base) made bundling cheaper than buying them separately. But schools have something cable never did—lock-in mechanisms so deep that even the people trapped by them cannot see the walls.

Why you just can’t unbundle

I finished the research for this essay frustrated. Fine, the bundle is arbitrary, so therefore I’ll do…something. Maybe get a tutor? I thought about homeschooling for 5 minutes (hard pass). Maybe a microschool? Unfortunately, here is why all that is harder than it sounds.

The school is your childcare. The average cost of full-time childcare in the United States is $13,128 per year according to Child Care Aware of America—roughly $1,100 per month per child. Public school absorbs nearly all of that cost for seven hours a day, 180 days a year. It is the largest implicit subsidy most American families receive, and you only notice it when it disappears. Any alternative you consider — a learning pod, a tutor, a homeschool co-op — has to replace that subsidy. During COVID, families who tried learning pods paid an average of $300 or more per week, and half of participating families earned over $125,000 a year. The supervision problem is the lock that holds the tightest, and it is the one education reformers talk about least. And those numbers are national averages! In every major metro, childcare costs are usually much higher.

Your child's credential is still tethered to the default path. While homeschoolers can earn diplomas through the GED, it isn’t as straightforward as just “take the test.” In practice, the non-traditional paths are narrow and penalized. Heckman, Humphries, and Mader found that GED holders earn roughly the same as high school dropouts — about $3,100 per month compared to $4,700 for diploma holders. The gap is not purely about the credential. Heckman argues it reflects differences in non-cognitive skills like persistence and reliability, traits that the traditional school path develops (or at least filters for) and that alternative paths do not reliably signal. Homeschooled students face a different version of the same problem where their diplomas are accepted, but they carry a higher documentation burden — transcripts, portfolios, test scores, letters — to prove what the traditional diploma proves by default. The Carnegie unit system, adopted in 1906, still defines the default where one credit equals 120 hours of seat time. Forty-two states now allow competency-based alternatives in some form, but the default remains the default, and defaults are powerful. Most families follow the path of least resistance, which is the path that runs through a school building.

Your child's friendships are the invisible infrastructure. Ask children what school is for and they are more likely to mention their friends than their teachers. They are not wrong. A longitudinal study of 1,227 children in South Korea found that the quality of peer relationships at age seven was the single most important predictor of internalizing mental health problems in adolescence. Your child's social network is, for many children, the most developmentally consequential thing school provides.

Which is precisely what makes switching so costly. A meta-analysis of 26 studies found that mobile students' achievement exceeded only 40 percent of their non-mobile peers, a gap equivalent to three to four months of learning. And when EdChoice surveyed parents about barriers to trying alternatives like homeschooling, socialization was the concern that surfaced most consistently. Parents intuit what the research confirms; relationships are load-bearing. Breaking them is a developmental cost, and for many families, it is the reason they never seriously consider leaving.

These lock-in mechanisms operate simultaneously. The childcare trap keeps you from considering alternatives. The credential fusion keeps you from assembling your own. The social cost keeps you from switching even when you find one. Cable television just had simple unit economics to keep you subscribed to a hundred channels. Schools have regulatory, social, and financial forms of lock-in, all stacked on top of each other, reinforced by 170 years of cultural customs. It is why the school bundle has outlasted every other bundle in American economic history.

But barriers erode. And the early data suggests these are eroding now.

What this means for you

If you have read this far, you may be experiencing something between helplessness and vertigo. You send your child to an institution every day that bundles together a dozen services, most of which lack strong causal evidence, assembled through a century of political decisions you had no part in, organized in a way that makes rigorous evaluation of any single component nearly impossible, and locked in by barriers so thick that even seeing them clearly does not make them easy to escape. The system gives you the least information about the variable that matters most (teacher quality) and the most information about variables that matter least (test scores in two subjects, facilities, and extracurricular offerings).

That is the honest situation. Sitting with it is uncomfortable for me. But it is also clarifying, because once you see the bundle for what it is, you can start asking questions that most parents never ask.

Some parents are further along. Arizona launched the nation's first Education Savings Account program in 2011 with 144 students. By March 2026, after expanding to universal eligibility, enrollment had exploded to over 102,000 students, with $886 million in annual spending. The RAND Corporation's 2025 evaluation found that roughly 60 percent of ESA spending goes to private school tuition—families swapping one bundle for another. But the remaining 40 percent flows to individual components like curriculum, tutoring, online learning, therapy, and specialized services. Most importantly, the longer families stay in the program, the more they unbundle. The share going to private school tuition decreases over time, while spending on individual components increases. When parents can see the bundle and have the resources to take it apart, they do.

By the 2026–27 school year, at least 17 states will be running universal or near-universal school choice programs, and roughly half the nation's students will be eligible for some form of component-level funding. The infrastructure for unbundling is being built, right now, whether your state is participating or not.

The place to begin is with a simple exercise: write down what your child actually receives from school. You will likely hit eight to twelve categories. The list itself is the point, because it makes visible a bundle that is designed to be invisible. Once you can see it, you can ask which of those services your child actually needs from this particular institution, which could come from somewhere else, and which you are not even using.

The most revealing line will be "academic instruction," because it is the one the institution claims as its core purpose and the one it can tell you the least about. Your school's daily schedule says 6.5 hours. But how much of that is actual instruction — delivered by a present teacher, undisturbed by interruptions, with your child in the room? Most schools do not track this number. If yours cannot tell you how much instruction your child actually received this year, it cannot tell you whether its academic program is working. The difference between a strong and weak instructional year is not measured in curriculum choices. It is measured in whether a qualified teacher was actually in front of your child, actually teaching, for enough hours to compound into learning.

None of this means the bundle is worthless — only that its value depends on which part you are actually utilizing on, and that answer varies enormously. For a dual-income household, childcare is the core service; academics are almost secondary. For a family in a high-poverty district, school meals may be genuinely irreplaceable. For a child with an IEP, the special education services may be the reason the school matters most. The point of seeing the bundle clearly was never to dismiss it. It was to stop treating "does school work?" as a single question with a single answer, and to start asking: which parts work for my child, which parts don't, and what am I going to do about the difference?

The lesson from every other industry that has gone through unbundling this is that the pieces do not stay scattered. Cable unbundled, and then it re-bundled — under new names, on new terms, run by new companies. If the same pattern holds for education, and the Arizona data suggests it is already beginning, then someone will reassemble the pieces of your child's school day into something new. The question for you is what do you actually want school to do for your child? When you drop them off in the morning, which of those twelve services are you paying for? And when someone offers to reassemble the bundle on better terms, will you know what to ask for?

That is Essay 5.

Palo Alto School District is, by every metric parents are trained to care about, one of the best public school systems in America. It has a math proficiency of 79%, versus a California average of 34%. Its AP pass rate is a superhuman 95% while all its students are averaging an SAT score of 1410. For this performance, parents pay a median of $3.4 million to live near these schools, making Palo Alto the second-least affordable city in America—a price-to-income ratio of 19 to 1.

Between 2009 and 2015, two suicide clusters struck Palo Alto's high schools. At least eight students died. The CDC investigated and found that Palo Alto's youth suicide rate from 2003 to 2015 was 14.1 per 100,000—nearly three times the national average and the highest in Santa Clara County.

Is that a good school?

This is not a rhetorical flourish. It is the question this essay is built around. Because if the highest-performing districts in America are producing students who are simultaneously acing the SAT and falling apart psychologically, then the metrics we use to evaluate schools are measuring the wrong thing. And if parents are spending millions of dollars to buy homes near schools that look extraordinary on paper and produce extraordinary suffering in practice, then something is broken in the way we think about what school is supposed to do.

In Essay 4, I showed you that "school" is not one thing—it's a bundle of a dozen services crammed into one building. This essay answers the question that naturally follows: now that you know the bundle exists, which part of it should you actually optimize for?

The answer is academic mastery. It isn’t grades or test scores, or even college placement. As a parent, you should care about a deep, genuine understanding of the most difficult materials. In achieving that, your kids learn the things that compound and protect your child over the course of a lifetime.

However, there’s a catch. The way most high-performing schools pursue academic outcomes actively undermines genuine mastery while producing impressive-looking numbers. The distinction between performative achievement and genuine academic mastery is the throughline of everything that follows.

It's not just Palo Alto

Before you dismiss the opening as an extreme edge case, you should know that Palo Alto isn’t an isolated and tragic tale. It’s a data point in a well-documented pattern.

Two decades of research out of Columbia have found that affluent suburban adolescents show depression and anxiety rates 1.5 to 2.5 times the national norms. One in five affluent girls show clinically significant depressive symptoms—three times the normative rate. Perhaps most strikingly, suburban 10th graders show higher substance abuse than inner-city peers across every category measured. The two primary risk factors were excessive achievement pressure and emotional isolation from parents. By 2013, the research had become strong enough to designate students in high-achieving schools as a formal at-risk group—ranking excessive achievement pressure among the top four risk factors for adolescent mental health, alongside poverty, trauma, and discrimination.

Stanford's Challenge Success program has surveyed tens of thousands of high school students. Their findings helped us understand what students in these high-performing schools experience. With that significant academic pressure that came from going to the “best” schools, only about a third felt confident coping with it, and yet 52% of 15,000 California students said most of their assignments were busywork. A related yearlong ethnographic study at a top California high school found that many of the most "successful" students weren't learning at all. They were doing school—gaming the system for grades while retaining almost nothing. A meta-analysis of 48 studies confirmed the pattern with a positive association between academic pressure and mental health problems, with both rising in parallel.

None of this is an argument against rigor. Encouraging kids to be their best is normal and healthy parenting. What I’m arguing against in this essay is the rigor of academic gamesmanship. The Palo Alto pattern—high achievement plus high distress—is not the inevitable cost of pushing kids hard. It's the specific cost of pushing kids toward performance rather than mastery. That's a distinction the research takes very seriously, and one that most parents have never heard articulated.

What is the actual job of a school?

In the last essay, I showed you that your kid's school is simultaneously a classroom, a cafeteria, a health clinic, a therapy office, a sports league, and a daycare. The natural next question is: which of those things is actually the job?

ou can't optimize for everything at once. A school that produces incredibly healthy kids who can't read is not a good school. A school with zero truancy and zero learning is a well-attended failure. Every one of those other services—nutrition, health, socialization, supervision—is genuinely valuable. But none of them is the reason schools exist. And when you start evaluating schools, you need to be clear about what you're actually evaluating.

The job is academic mastery. And I can make that case two ways.

Only school does this

Your kid can get a meal from a lunch program, a food bank, or your kitchen. They can get health care from a pediatrician, a school nurse, or a community clinic. But systematic academic instruction is different. Can it happen outside a school? Yes—homeschooled students perform comparably on standardized tests when researchers control for family background, and structured curricula exist for families who want to sequence instruction at home. But the research is thin. Homeschool studies are plagued by selection bias (these families tend to be wealthier and more educated), microschool outcomes have barely been studied, and no alternative has been evaluated at anything close to the scale of the public school system. A meta-analysis of tutoring interventions found that even the most intensive non-school alternative—targeted one-on-one tutoring—works within the context of school-based learning, not as a replacement for it. If you are building reading comprehension systematically from phonics through literary analysis, or mathematical reasoning from arithmetic through calculus, it requires a coordinated, multi-year instructional sequence. Schools are not the only institution that can deliver that. They are the only institution that does deliver it, for tens of millions of children, with an evidence base we can actually evaluate. Everything else in the school bundle happens somewhere else. Academic instruction at scale does not.

And we know that instructional time matters. One study analyzed instructional time data across 50 countries and found that one additional hour of weekly instruction in a subject raises test scores by approximately 0.06 standard deviations. That sounds small until it starts to compound over a school year. A complementary study using student fixed-effects confirmed the finding and added a critical nuance—the effect of additional instructional time is moderated by classroom quality. More time in a good classroom helps a lot. More time in a bad one helps less.

Which makes the mission creep problem genuinely alarming. After No Child Left Behind, research documented that schools systematically reallocated instructional time away from science and social studies toward tested reading and math—and the schools that did this most aggressively were the ones serving the most disadvantaged students. The result was a measurable narrowing of the curriculum; No Child Left Behind produced 4th-grade math gains of about 0.23 standard deviations, but at the cost of hollowing out the content knowledge those students would need later. Schools traded the subjects that build background knowledge—history, science, geography—for test-prep drills in the subjects being measured.

I'm not saying schools should stop feeding kids or providing counseling or testing math. I'm saying that the instructional time data gives us a precise way to measure the cost: every hour reallocated from academic instruction to non-academic programming represents a measurable loss in learning—approximately 0.06 SD per weekly hour, compounding across subjects and years. When schools cut social studies by 32% or science by 33%, the research tells us exactly that our kids lost background knowledge that reading comprehension depends on, content that builds the cognitive skills that actually predict life outcomes, and instructional time that no other institution is going to replace. We can and should test, but the gamesmanship that the testing regime in the U.S. incentivizes harms our kids.

Academic mastery is the thing that produces all the other benefits

Education research has spent decades documenting that more schooling leads to higher income, better health, and lower crime rates. That’s fairly obvious. But the critical nuance that most of those studies gloss over is that it's not the schooling that produces the benefits. It's the learning.

This distinction sounds pedantic until you see the data.

The 2018 World Development Report—the World Bank's flagship annual publication—documented what it called a global "learning crisis." Hundreds of millions of children attend school regularly and yet, learn almost nothing. In sub-Saharan Africa, 86% of children could not read proficiently by age 10 despite being enrolled in school. These countries expanded enrollment dramatically over two decades, but the economic and health benefits that were supposed to follow haven’t yet materialized—partially because the schooling wasn't producing learning.

The World Bank formalized this distinction into a metric called Learning-Adjusted Years of Schooling, or LAYS, which adjusts raw enrollment by what students actually learn. A 2022 study confirmed that LAYS is a robust predictor of economic growth, while years of schooling without learning produces zero measurable growth effect. Two countries with identical enrollment rates can have vastly different LAYS scores—and it's the LAYS score, not the enrollment rate, that predicts prosperity.

The same pattern holds at the individual level. A landmark study which linked a million students to their adult tax records, confirmed that students assigned to high-value-added teachers—teachers who produced more learning, not just more seat time—earned 1.3% more at age 28, were more likely to attend college, and were more likely to live in higher-income neighborhoods. Replacing a bottom-5% teacher with an average one increases the present value of a classroom's lifetime earnings by approximately $250,000. Now, “value-added” is measured by test score gains. And yes, later in this essay I'll argue that test scores are an imperfect proxy for genuine mastery. Value-added works as measurement of mastery because it isolates growth: it measures whether a student learned something they didn't know before, while controlling for where they started. That's closer to mastery than a raw SAT score, which conflates learning with prior advantage.

Crime data tells the same story. One study found that attending a higher-quality school—not just a school, but a better school—reduced serious criminal arrests by roughly 15% for high-risk youth by age 30. School quality was the variable that predicted whether young people ended up in prison.

This is why the "benefits of education" framing gets it wrong. It makes it sound like all a school has to do is keep your kid enrolled. But the convergence of evidence across these studies—from cross-country growth models to individual-level tax records to criminal justice outcomes—say that schools produce life-changing outcomes to the extent that students actually master academic material.

So when you're evaluating whether a school is doing its job, the question has to be if your kid is learning. Every study in this section points in the same direction. The income, the health, the resilience, the freedom—all of it flows downstream from academic mastery, not from enrollment or extracurriculars.

The thing that actually matters

We now arrive at the core of this essay. Everything above establishes that academic outcomes matter. What follows is the argument that how your child pursues those outcomes matters just as much—and that the way most high-performing schools structure achievement is psychologically destructive.

The distinction between "performing well" and "actually learning" is not some vibe. It is a 40-year research program with hundreds of studies across dozens of countries. Psychologists call the two orientations performance goals and mastery goals.

The foundational research

The core finding, established in a 1984 paper and confirmed hundreds of times since then, is that when a student sits down to study, they are doing one of two fundamentally different things. Either they're trying to understand the material— a mastery goal—or they're trying to prove their ability to other people—a performance goal. One is driven by curiosity and the intrinsic satisfaction of figuring something out. The other is driven by external validation.

And the most important discovery in the literature is that the school determines which one your kid adopts. The most important study in the field identified six classroom design features (task structure, authority, recognition, grouping, evaluation, and time allocation) that predict whether students orient toward mastery or performance. The finding was not that some kids are naturally mastery-oriented and others performance-oriented. It was that the classroom environment produces the orientation. Design the environment one way and kids focus on understanding. Design it another way and they focus on looking smart.

A meta-analysis of 243 studies covering 91,087 participants confirmed the pattern. Performance goals that emphasized normative comparison—how am I doing relative to everyone else?—correlated negatively with actual learning (r = -.14). Mastery goals showed consistently positive relationships with deep processing, interest, and persistence. In parent language: "trying to get into Stanford" and "trying to actually understand calculus" produce fundamentally different outcomes.

What performance-oriented environments do to kids

A 1999 study that should be required reading for every school administrator in America found that when students perceived their school as emphasizing mastery goals, they reported higher well-being across every measure. When students perceived their school as emphasizing performance goals—grades, rankings, relative ability—they reported increased anxiety, depression, and disruptive behavior.

This was a study where I had a mental double-take. Yes, it was the school's goal orientation that predicted the student's mental health. A study of 656 students found that kids in performance-oriented classrooms were significantly more likely to use self-handicapping strategies—procrastination, feigning illness, spreading their efforts thin across activities so they'd always have an excuse for underperformance. These are the kids who say "I didn't even study" before a test. That's not laziness. It's a rational defensive strategy in an environment where looking smart matters more than being smart. If you try hard and fail, you're dumb. If you don't try and fail, your ability is still an open question.

The cheating data is even more damning. Research on achievement goals and academic dishonesty found that students with performance orientations showed a 41.7% dishonesty rate, compared to 19.2% for mastery-oriented students. When the system tells a student that the grade matters more than the learning, some percentage of students will conclude that the most efficient path to the grade doesn't involve learning at all.

And then there's what happens to self-worth. When a student's identity becomes contingent on academic achievement—when they feel, at a gut level, that they are only as valuable as their last test score—the psychological consequences are severe. Higher anxiety. Higher depression. Vulnerability to imposter syndrome. Fragility in the face of any setback, because the setback doesn't just mean they got a problem wrong—it means they're less. This is the Palo Alto pathology, described precisely by the research, where students are simultaneously high-achieving and psychologically fragile, because their self-worth rests on a performance that can never be good enough.

What mastery-oriented environments do to kids

The reverse pattern is just as well documented.

Research on the elementary-to-middle-school transition found that fifth graders who perceived mastery goals in their elementary school showed sharp declines in mastery orientation when they moved to middle schools that emphasized performance goals. But students who transitioned to reform schools that had deliberately adopted mastery-oriented goal structures maintained their mastery orientation. The environment determines the orientation.

A longitudinal study of 8th and 9th graders found that mastery goals were associated with higher efficacy, higher interest, more positive affect, greater effort, deeper cognitive strategies, and better performance. Students who also held performance goals showed some moderate additional benefits—but only when the mastery foundation was already in place. Performance goals on top of mastery goals can be fine. Performance goals instead of mastery goals are destructive.

A 2016 meta-analysis of 74 motivation intervention studies involving 38,377 students confirmed that these orientations can be deliberately changed. Motivation-based interventions produced a mean effect size of d = 0.49—a substantial effect by educational standards. Schools can shift their goal structures from performance to mastery orientation, and when they do, the outcomes improve.

Why mastery protects and performance destroys

What's the mechanism? Why does mastery orientation protect kids while performance orientation damages them? There are three bodies of research that helped converge on an answer for me.

Self-efficacy. Self-efficacy research identifies mastery experiences as the single strongest source of self-efficacy—the deep belief in your own capability. When a student genuinely masters difficult material, they build a stable sense of competence that doesn't depend on external validation. This is qualitatively different from the confidence that comes from getting an A. An A depends on the grading curve, the teacher's standards, or the difficulty of the test. Genuine mastery depends on you and the material.

Self-determination. Self-Determination Theory, one of the most influential frameworks in motivation psychology, identifies three basic psychological needs: autonomy, competence, and relatedness. Competence—the feeling of mastery—is a basic human need. When it's met through genuine learning, it supports optimal motivation and psychological health. When it's met through external rewards (grades, rankings, praise for performance), it's fragile and contingent. A meta-analysis of 128 experiments found that extrinsic rewards significantly undermine intrinsic motivation. Performance-contingent rewards—the kind that define most schools—decreased free-choice intrinsic motivation with an effect size of d = -0.28. But positive competence feedback—the kind that acknowledges genuine understanding—enhanced motivation.

Self-concept. A meta-analysis of 558 effect sizes covering 274,370 adolescents across 39 countries found that academic self-concept strongly predicts lower depression, with a correlation of r = −0.518. The key nuance is that academic self-concept built on mastery (I understand this material) is stable and protective. Academic self-concept built on performance (I'm ranked higher than my peers) is fragile and collapses under pressure—because it depends on a comparison that can change at any moment.

We can now map this research onto Palo Alto with the district's own data.

When Challenge Success surveyed 1,561 students at Gunn High School, 62% reported being "often or always stressed about schoolwork." Eighty-four percent reported cheating. At Paly, 41% of students were working hard but rarely finding their schoolwork interesting, fun, or valuable—which is exactly what the "doing school" pattern describesIn internal debates over whether to report weighted GPAs on transcripts, school officials acknowledged that the practice could "contribute to unhealthy academic competition" and "breed an unhealthy focus on grades rather than a love of learning."

That last phrase is the performance-mastery distinction, stated in plain English by the people running the school.

I want to be precise about the causal claim. When the CDC released its final report, the most common precipitating factors were recent crises, current mental health problems, and depression—not school culture specifically. Suicide is complex, and no single variable explains a cluster. But the CDC also found that when students were surveyed, they ranked academic stress as a leading risk factor—rating it higher than their parents did. The gap between what students experienced and what adults perceived is itself a signature of the performance-oriented environment. The metrics said everything was working. The students knew otherwise.

Those students had high performance goals. They were optimizing for grades, AP scores, and college placement. What they lacked was mastery—the deep experience of understanding something genuinely difficult, for its own sake, in an environment that values the understanding over the grade. The research predicts exactly what Palo Alto's own data showed: high achievement, high anxiety, high fragility.

The grades problem

This brings us to an uncomfortable finding about the primary feedback mechanism in every school in America.

Two landmark experiments—in 1986 and 1988—tested what happens when you give students different kinds of feedback. Students who received detailed, task-focused comments without any grade showed significantly higher intrinsic motivation and better task performance than students who received grades—with or without comments. The critical finding: when you add a grade to a comment, students ignore the comment and fixate on the grade. The number crowds out the learning.

The downstream research confirms it: grades reliably produce three effects. Less interest in learning. A preference for easier tasks. Shallower thinking. The grade becomes the goal. The learning becomes the obstacle between the student and the goal.

Now connect this to what I showed you in Essay 1. American high schoolers had the highest GPAs ever recorded and the lowest test scores in a decade. Mastery learning helps us understand that this is partially a motivation failure. When grades are the goal, students and schools optimize for grades. When schools inflate grades to keep parents satisfied, the entire system drifts toward performative achievement. Grade inflation is the natural endpoint of a system that rewards performance over mastery.

What should we actually be teaching?

In Essay 2, I covered how kids learn—retrieval practice, spaced repetition, interleaving, mastery learning. This section covers a different question that none of the previous essays address: what should they learn?

The answer turns out to matter enormously, and there's been a quiet catastrophe in American curriculum that most parents know nothing about.

Reading comprehension is not a skill

In 1988, a now-famous experiment gave kids passages about baseball and tested their comprehension. The kids who knew about baseball understood the passages. The kids who didn't know about baseball didn't understand them. Reading ability, as measured by standardized tests, barely mattered. Knowledge of the subject was the dominant factor.

This finding launched an entire body of research showing that there is no such thing as a general "reading comprehension skill." You can only comprehend text about domains where you have background knowledge. Try reading a legal brief or a medical journal article outside your field—you can decode every word but understand almost nothing.

The Knowledge Gap documented the catastrophe of when American schools chose to ignore this research. Starting in the 1990s, and accelerating after No Child Left Behind in 2001, elementary schools across America replaced history, science, and social studies instruction with drilling "reading strategies"—find the main idea, make predictions, identify the author's purpose, make text-to-self connections. Schools cut social studies instruction by 32%, science by 33%, and arts by 35%, while increasing reading and math instruction by 47% and 37% respectively.

The irony is bitter. By cutting content to focus on reading scores, schools eliminated the background knowledge that reading comprehension depends on. They optimized for the metric and destroyed the thing the metric was supposed to measure. A child who spends an hour "finding the main idea" in a passage about the American Revolution, without ever being taught what the American Revolution was, is doing an exercise in nothing.

And the evidence for knowledge-rich alternatives is now overwhelming. The Core Knowledge randomized controlled trial — testing schools that replaced generic reading-strategy drills with a systematic, grade-by-grade curriculum in history, science, and literature — spanned six years and 2,310 lottery applicants. It found that students in Core Knowledge schools improved reading by roughly 16 percentile points. The effects were large enough to eliminate the achievement gap associated with income. Low-income students benefited even more than mixed-income students. The researchers called it "one of the largest effect sizes ever measured in the educational intervention literature."

Louisiana's statewide curriculum reform suggests the same pattern holds at scale—though with the caveat that any statewide policy change carries confounding variables that a randomized trial doesn't. Beginning in 2012, Louisiana adopted knowledge-rich curricula statewide. Fourth-grade reading moved from 42nd to 16th nationally. When the 2024 NAEP scores came in, Louisiana was the only state in the country to beat its pre-pandemic reading scores. The only one. That's not proof on par with the Core Knowledge RCT—teacher training, funding, and political attention all changed alongside the curriculum. But the convergence of a causal study and a real-world implementation pointing in the same direction is about as strong as education evidence gets.

"But what about grit?"

If you've been following the education conversation over the past decade, the strongest counterargument to everything I've argued is probably already forming in your head: doesn't the research say that non-cognitive skills—grit, perseverance, character—matter more than academic ability?

It's a fair objection. The popularized version of this argument has been enormously influential. It's also wrong in important ways.

A Nobel laureate economist showed that non-cognitive skills—persistence, self-regulation, conscientiousness—have high economic returns. Analysis of the Perry Preschool and Abecedarian programs found that IQ gains faded but life outcomes improved, suggesting that something beyond raw cognitive ability was driving long-term benefits. Then came "grit"—passion and perseverance for long-term goals—which predicted outcomes beyond IQ in settings like West Point and the Scripps spelling bee. A bestseller, How Children Succeed, wrapped the findings in narrative, and suddenly every school district in America was talking about "character education."

But then the same thing that always happens when education research gets applied to American schools. Its findings got oversimplified, then scaled badly, then used to justify policy changes the original researchers wouldn't endorse.

So let's start with the actual framework. The original research does not argue that cognitive skills don't matter. The "skill begets skill" model shows that cognitive and non-cognitive skills are complementary—early cognitive investment enables later non-cognitive development. You can't build character on weak cognitive foundations. The conclusion some schools drew—that you can de-emphasize academics to focus on social-emotional learning and get the same long-run benefits—would horrify the researcher whose work they're citing.

Then there's grit itself. A meta-analysis of 88 grit studies covering about 67,000 people found that grit explained only about 4% of the variance in performance. It overlapped 80 to 98% with conscientiousness, a personality trait that psychologists had been studying for decades. Grit, it turned out, was not the new discovery it was presented as—it was a rebranding of something we already knew about, with a much smaller effect size than the popular narrative suggested. Even the researcher who coined the term has cautioned against oversimplifying her findings and has opposed using grit measures in school accountability systems.

KIPP schools, the charter network that most enthusiastically embraced character education, introduced character report cards and made grit a central part of their culture. The results were sobering relative to KIPP's own ambitions. 90% of KIPP students enrolled in college, but only about a third graduated — better than the national rate for similar demographics, but a stark gap given the intensity of the intervention. Whatever KIPP was building, it wasn't enough to sustain students through college completion.

And the growth mindset story—the idea that believing your abilities can improve through effort produces better outcomes—hit a similar wall. A 2018 meta-analysis found that growth mindset accounted for about 1% of academic achievement variance (r = 0.10). The original findings haven't been reliably replicated at scale.

I'm not saying the underlying research is bad. The original work is serious. But the popularized versions were used to justify exactly the wrong conclusion. Namely that schools could de-emphasize academic rigor and instead teach "character" or "mindset" as separate skills. The original researchers would, I think, all agree on my framing of this issue.

The grit has to be in the learning.

And the data supports the claim directly. Research on school goal structures and grit found that students who perceived their schools as mastery-oriented grew grittier over time, while students in performance-oriented environments became less gritty. Productive struggle research shows that sustained engagement with optimally challenging tasks—problems at the edge of a student's ability—develops both perseverance and the capacity to recover from setbacks.

You're looking at the wrong things

By now you've seen that academic mastery matters, that how schools pursue it determines whether kids thrive or break, and that the content of what's taught is as important as the method. But none of this helps unless you can actually identify which schools are getting it right. And the research on how parents choose schools suggests almost nobody can.

Let’s start with what parents believe. PDK/Gallup polling consistently shows that 77% of parents give their own child's school an A or B, while only 18% of Americans say the same about public schools nationally. Everyone thinks their kid's school is the exception. That confidence rests almost entirely on report cards. The Learning Heroes/Gallup "B-flation" study found that 88% of parents believe their child is at or above grade level in reading and 89% in math, while the NAEP says only 26 to 29% of 8th graders actually are. As we showed in Essay 1, report card grades no longer correlate with tested performance—but they remain the primary signal parents trust.

When parents do choose schools actively, they're not choosing what you'd expect. EdChoice polling found the top factors are location, socialization, and familiarity with the assigned school. Academic rigor doesn't crack the top three! And even parents who try to evaluate academics get misled. An MIT study found that parents' school reviews correlated with raw test scores and demographics, but not with value-added—the measure of how much a school actually improves outcomes for the students it serves. After controlling for peer quality, parents showed little preference for instructional effectiveness. They were choosing student bodies, not teaching. A study of Florida's school grading system confirmed the pattern from the housing market side: when the state assigned letter grades to schools, homes near A-rated schools appreciated by about $10,000 relative to B-rated schools — but the price premium tracked the label, not differences in actual test performance. Parents were purchasing the signal of quality, not quality itself.

Your kid, your choice

I want to bring this back to your kitchen table. Maybe it's a Tuesday night and your kid just handed you a report card. It looks good. A's and B's, maybe an A-minus in something hard, a comment from the teacher about what a pleasure your kid is to have in class.

After everything you've read in this essay—the outcomes research, the mastery-performance distinction, the curriculum hollowing, the grade inflation—what does that report card actually tell you?

Maybe everything is fine. Maybe your kid is in a school that teaches real content, that measures genuine understanding, that provides feedback designed to deepen learning rather than sort kids into ranks. If so, protect what you have.

But maybe—like most parents—you have no idea. You see the grades. You see the test scores. And you assume that if all those numbers look right, your child is learning. This essay has shown you why that assumption might be wrong.

Here's what you can actually look for.

Your kid can explain what they learned. Ask them to explain it—to you, to a sibling, in their own words. If they can't, they didn't learn it. They memorized it temporarily, and it's already fading. This is the simplest, most reliable test of mastery, and you can do it tonight at dinner.

Your kid finds some things hard and keeps going. Genuine mastery requires struggle. If everything is easy, your child isn't being challenged. The research is unambiguous on this point: kids who never struggle in school don't develop the psychological tools to handle struggle later. The resilience comes from the difficulty, not despite it.

Your kid is curious about something. It doesn’t have to be about everything all the time! But somewhere in their education, there is a topic they pursue because it's interesting, not because it's on the test. That's the signal that intrinsic motivation is still intact. If school has crushed that curiosity, if every subject is just another performance to get through, the system is failing your kid even if the grades say otherwise.

Your kid doesn't cheat. If your high-achieving child copies homework, shares test answers, or uses shortcuts to maintain their GPA, it’s a symptom. The research on cheating and goal orientation is clear that dishonesty rates are more than double in performance-oriented environments. The system told your kid that the grade matters more than the learning, and your kid believed it.

You can't fix American curriculum policy from your living room. But you can refuse to be fooled by the metrics. The research says that kids who genuinely master hard material—who build deep knowledge, who struggle through difficulty, who learn for understanding rather than for grades—end up healthier, wealthier, more resilient, and more free than kids who were trained to perform. This is not a trade-off between rigor and wellbeing! It's the opposite.

The question for every parent reading this isn't whether your kid is "succeeding." By the metrics most schools use, they probably are. The question is whether the thing your school calls success has any relationship to the thing you actually want for your child.

Now you know what to optimize for. The next essay asks: if genuine academic mastery is this valuable, why don't schools just do it? The answer involves the worst kind of problem—one where everyone is acting rationally and nobody is in charge. That’s essay 6.