The Variable Is the Family

Every homeschooling parent gets asked to justify it eventually. Usually at a family dinner, usually by someone who loves your kids and is genuinely worried. So a few weeks ago I sat down to build a proper file of evidence, the kind you could hand to a skeptical grandparent or a curious neighbor without flinching.

I came back with a case that is narrower, and a great deal stronger, than the one I went looking for.

Here is the short version, and then I will show you the work. Public schools are measurably declining. Families are leaving in record numbers. The institutions have taken positions on children’s bodies that the evidence underneath them does not support, and in more than a thousand districts they have written down that parents need not be told. And the thing that actually produces good outcomes is not the building a child sits in, it is whether the instruction is structured, individual, and watched over by a parent. That last sentence is the argument. Everything below it is the evidence.

What I did not expect was how consistently the evidence points past the school altogether, at something Scripture named three thousand years ago.

Two sections below are heavy going. One covers youth suicide and the school calendar. The other covers what schools are teaching about the body. I have put the limits of that research inside each section rather than at the end, because the limits are part of reading it honestly.

The Decline Is Real, and It Started Before the Pandemic

The class of 2024 posted the worst twelfth-grade math and reading scores the National Assessment of Educational Progress has ever recorded. These are federal numbers, from the test that exists precisely because states grade their own homework.

Sources: the Nation’s Report Card, K-12 Dive, Chalkbeat, and AEI’s 2025 absenteeism report.

Two details matter more than the headline. The first is that the slide shows up in fourth, eighth, and twelfth grade alike, and it began roughly a decade ago, well before anyone had heard of COVID. The lockdowns accelerated something already in motion. They did not start it.

The second is who is falling. Students at the ninetieth percentile held steady. Everyone below them dropped. The gap between the children who were already going to be fine and the children who needed the school to work is widening every year the test is given.

I will give the other side its due. Absenteeism is slowly improving, fourth-grade math ticked up two points between 2022 and 2024, and a real share of the enrollment decline is falling birth rates rather than families walking out. But 45% of graduating seniors cannot do basic math, and no amount of context makes that a good number.

We Pushed Academics Earlier. It Got Worse.

This is the finding that changed how I think about our own five-year-old, and it is the strongest evidence I know of for a Montessori-shaped, child-paced approach.

Somewhere around 1998, American kindergarten stopped being kindergarten. Researchers at the University of Virginia compared kindergarten classrooms in 1998 and 2010 and found the room had been quietly rebuilt:

Science corners, sand tables, water tables, and dress-up areas fell by double digits. Worksheets and textbooks rose. Kindergarten became first grade, and first grade became something no five-year-old was built for. (In fairness, daily recess actually rose nine points.)

Then we ran the experiment on whether starting earlier helps. Vanderbilt randomized 2,990 low-income Tennessee children into the state pre-K program or a waitlist and followed them for years. The early gains faded by the end of kindergarten and had gone negative by third grade. By sixth grade the children who attended had lower test scores, more disciplinary infractions, and more special education placements than the children who stayed home. The effect sizes are small, and critics blame the program’s quality rather than its timing. Both fair. But this was a randomized trial, which is as close to proof as education research gets, and it did not find what everyone expected.

And then there is the finding I cannot stop thinking about. In a study of 407,846 children, researchers writing in the New England Journal of Medicine found that in states with a September 1 cutoff, children born in August were diagnosed with ADHD at 85.1 per 10,000. That is about thirty percent more often than children born in September. The gap did not exist before school entry, and it did not exist in states without that cutoff.

We are diagnosing a meaningful number of children with a disorder whose actual condition is that they are eleven months younger than the child beside them, in a room that has no accommodations for eleven months.

The Crisis Tracks the School Calendar

This is the hardest material in the file, and I want the limits on the table before the findings rather than buried under them.

None of this research compares school to homeschooling. Not one study. It compares school in session to school out of session. What it establishes is that something about conventional schooling, as currently run, carries a measurable weight. It does not establish that home education is the remedy.

With that said, here is what the research shows.

Youth suicide deaths fall sharply in June, July, August and December, and climb again when school resumes. Hansen and Lang’s 2011 analysis found the pattern holds across decades and is specific to school-age children. Adults show no such seasonal trend. Weather, unemployment and seasonal affective patterns do not account for it.

The sharpest version of the finding is geographic. NBER working paper 30795 compared regions by when their school year begins. Where school starts in early August, the rise in teen suicide appears in August. Where it starts after Labor Day, the rise appears in September. Move the calendar and the curve moves with it.

Then the pandemic ran an experiment no ethics board would have approved. When schools closed in March 2020, youth suicides fell about 25%, and the return to in-person instruction raised them 12 to 18%. Pediatric emergency and hospital visits for suicidality from 2016 to 2021 traced the same shape. They ran higher in school months, and dropped unexpectedly through the spring 2020 closures, at the exact moment every forecast said adolescent mental health should have collapsed. One Los Angeles pediatric emergency department recorded psychiatric visits 118% higher in school weeks than in break weeks.

Whatever is happening inside that building is heavy enough to show up in a national mortality curve, and light enough on a Tuesday in July that the curve bends back down.

I am not going to put a sales pitch next to that paragraph. What I will say is this. It is a reason to take seriously the possibility that the shape of conventional adolescent schooling (the hours, the sorting, the social pressure, the evaluation) costs some children considerably more than we have been willing to count. That is worth a parent’s attention.

If you or a child you love is struggling, the 988 Suicide and Crisis Lifeline is available in the US by call or text, any hour of the day.

Families Are Voting With Their Feet

Homeschooling grew 4.9% on average in the 2024-25 school year, against roughly 2% a year before the pandemic, and 36% of reporting states hit all-time highs. That is from the Johns Hopkins Homeschool Hub, which builds its numbers from state filings rather than advocacy surveys.

Massachusetts is the clearest picture of the shift. Since 2019, the state’s school-age population has fallen 2.6% and public enrollment has fallen 4.2%. Homeschooling is up 56%. That is not demographics. That is a decision, made by a lot of families, one kitchen table at a time.

What Actually Happens to the Kids

Here the evidence gets thinner and more honest, and I want to be careful, because this is exactly where homeschool advocacy usually oversells.

First, though, a word about what these studies actually measure, because it governs how much weight any of them can carry.

Nearly all of them measure test scores and grades. I used test scores myself at the top of this article, and I want to be precise about why that was fair there and is not sufficient here. A test can convict a system. It cannot vindicate one. When a school system sets its own bar, writes its own instrument, administers it to its own students, and then cannot clear it, that is a failure on its own terms and in its own language. Forty-five percent of seniors below basic in math. No interpretation rescues it.

The reverse does not follow. A good score does not mean a child was educated. Montessori understood this a century ago. The things that actually prepare for real life appear on no standardized test. Concentration. Self-direction. The capacity to work at something hard because you chose it. Honesty. A settled sense of who you are and whose you are. What those instruments mostly measure is how well a child performs recall under time pressure for someone else’s approval. That is a skill. It is not the skill.

So when you read a study below reporting a smaller advantage than the brochures claim, hold it against the right standard. The question was never whether our children fill in bubbles faster. It is who they are at twenty-five.

The best-designed academic study I found is Martin-Chang’s 2011 work, which tested 74 Canadian children ages five to ten. Structured homeschoolers beat their public school peers by half a grade level in math and by as much as 2.2 grade levels in reading. Family income and maternal education did not explain the gap, and, crucially, both groups were volunteers, which strips out the sampling bias that ruins most homeschool research.

But read the whole study, and read this next part carefully, because it is usually quoted as a cudgel and I would rather not swing it. The twelve unschooled children in it scored lowest of anyone, one to four grade levels behind the structured homeschoolers.

Twelve children. That is a sample too small to hang a verdict on anybody’s family, and Martin-Chang measured standardized academic achievement and nothing else. She did not ask whether those twelve were curious, kind, self-directed or content. I have just spent four paragraphs arguing that those are the outcomes that forecast a life, and that a test cannot vindicate a system. It cannot convict a child either, and I am not borrowing the instrument back the moment it suits my argument. Harvard’s flourishing data does not separate unschoolers out. They sit inside the homeschool category that came out ahead on faith, service and forgiveness.

What the study does support is narrower and far more useful: structure is the variable, not the address. Moving a desk to a kitchen table changes nothing by itself. A child who is shown a path, given something worth mastering, and held to it grows. A child left entirely to drift usually drifts.

That finding is the reason Forage Academy is built the way it is. Structure is the hardest thing for a homeschooling parent to sustain. Not the love and not the hours, but the scaffolding, across four children and seven subjects, on a Tuesday, while the baby is teething. So that is what we built: a skill tree for each subject, mastery that has to be earned and then re-earned rather than ticked off, checkpoints a child has to actually pass, and a guide that answers a question with a better question. We built it as a prepared environment rather than a conveyor belt, which is the Montessori distinction that matters here. The child still chooses the path. The structure only makes sure there is one.

On college, Cogan’s 2010 study followed 76 homeschooled students at one Midwestern university:

The GPA advantage survived controls for demographics and pre-college factors. The graduation difference did not, and this is one university and seventy-six students. Encouraging, not conclusive.

And notice what it is even measuring: grades and graduation rates, which is to say the system’s own scoreboard, kept by the system’s own scorekeepers. Homeschoolers win on it anyway. That is worth something. It is still not the thing.

The study I would actually put in front of a Christian grandparent is Harvard’s. The university’s Human Flourishing Program followed 12,288 children of nurses from 1999 into adulthood and published the schooling comparison in 2021. Four findings cleared the study’s strictest statistical threshold, and three of them were about character. Homeschooled children grew into adults who were 51% more likely to attend religious services frequently, volunteered more, and scored notably higher on forgiveness. They also reported a greater sense of mission in life, used less marijuana, and had fewer lifetime sexual partners.

The study also flagged a possible higher risk of PTSD, which it turns out is the weakest result in the paper. It passed a conventional p-value and failed the study’s own Bonferroni-corrected threshold.

What a Bonferroni correction is. When a study tests dozens of outcomes at once, some will look significant by luck alone. The correction raises the bar for every result in proportion to how many were tested, so a finding has to be much stronger to survive it. A result that clears the ordinary threshold but fails this one is exactly the kind that usually evaporates on replication.

The PTSD signal rests on 187 homeschooled participants out of 12,288, all of them children of registered nurses, and their adult outcomes were measured in 2010.

That date is the heart of it, because of what the comparison group has done since. In the CDC’s 2023 survey, 40% of all American high school students reported persistent feelings of sadness or hopelessness, up from 30% in 2013. Twenty percent had seriously considered suicide. Nine percent attempted it in the past year. That is roughly one child in eleven, every year, in the conventional system, and the school calendar research above tells you what time of year. The Harvard study measured homeschoolers against a public school cohort that no longer exists.

Nor did it measure what is driving that. It did not ask about social media, which the CDC’s own 2023 data now ties directly to bullying, hopelessness and suicide risk. It did not ask about diet, which the same survey links to grades, sadness and school connectedness. It did not ask about family structure, screen hours, or how many hours a day a child spends in a peer group they did not choose. Those are the live variables in adolescent mental health, and a homeschooling family has more say over every one of them, not less.

So read the finding for exactly what it is. An uncorrected, unreplicated signal in a small subsample of one profession’s children, measured against a baseline taken before the collapse. What the paper establishes at full confidence is that homeschooled children became more faithful, more generous, and more forgiving adults. That is the result. The PTSD line has been doing work it cannot carry.

It is the most credible source in the whole file, and it measures the thing we actually care about. Not what our children score. Who they become.

The Mechanism Is One Adult, One Child

If the school building were the variable, the research would be far cleaner than it is. It isn’t. So what is doing the work?

In 1984, Benjamin Bloom published a finding in Educational Researcher that people have argued about ever since. Students taught one-to-one with mastery-based methods performed about two standard deviations better than students taught conventionally, putting the average tutored student above roughly 98% of the classroom group. He called it the 2 Sigma Problem, not because tutoring was mysterious, but because nobody could figure out how to deliver it at scale.

Later work has trimmed that number considerably. Education Next’s review of the replication record is worth reading before anyone quotes “two sigma” as settled fact. But even the deflated estimates are large by the standards of education research, and the direction has never been in dispute.

That is the finding underneath everything. One adult who knows this child, adjusting to this child, in real time. It is why homeschooling works when it works. It is also, candidly, the hardest thing for a parent of many to actually provide across seven subjects. That is the entire reason Forage Academy exists.

What They Are Teaching About the Body

Where we stand, and why the statistics do not settle it

Something has to be said before the numbers, because the numbers cannot do the work people want from them. A survey can tell you what is happening to children. It cannot tell you what a child is. No study establishes what is good, and I am not going to pretend a citation can carry a moral claim.

Here is the ground we actually stand on:

And God created man in His own image, in the image of God He created him; male and female He created them. (Genesis 1:27, LSB)

The word behind “image” is tselem, the word used elsewhere for a carved likeness, something struck to represent an original. A child is not an accident of chemistry that later gets assigned a meaning. A child is a likeness, on purpose, of Someone.

We believe that is good news about a child rather than a cage around one. Being given a body you did not design is the same kind of gift as being given a family you did not choose. And we believe the work of teaching a child who they are was handed to parents rather than to institutions: you shall teach them diligently to your sons (Deuteronomy 6:7), and bring them up in the discipline and instruction of the Lord (Ephesians 6:4). The two Greek words in that second phrase are worth knowing. Paideia is the whole formation of a child, everything that shapes them, not merely correction. Nouthesia is literally a placing-in-the-mind. Between them they describe a job that takes years and proximity, which is to say a job that belongs to whoever is actually in the room.

That is a conviction, held on an authority we accept. It is not a conclusion I am deriving from the data below, and I would hold it if the data ran the other way. What follows is not an argument that Scripture is true. It is an account of what is being taught, what it leads toward, and how strong the evidence underneath it actually is. A parent is entitled to know all three before someone else decides for them.

What the children are reporting

In 2023 the CDC’s Youth Risk Behavior Survey asked about transgender identity on a nationally representative basis for the first time, surveying more than 20,000 high schoolers. 3.3% identified as transgender and another 2.2% as questioning. A 2022 Williams Institute estimate built on 2017 and 2019 data had put the figure at 1.4%.

The rest of that survey is what I cannot get past. Among transgender students, 72% reported persistent feelings of sadness or hopelessness, as did 69% of those questioning. Twenty-six percent of trans and questioning students attempted suicide in the year before the survey, against 11% of female students and 5% of male students.

Let me be exact about what I think that means and about what I do not know. Those are children in real anguish. Some of them are in families I know, and some will be in families Forage serves. Whatever an adult concludes about the underlying questions, a quarter of a group of teenagers attempting suicide in a single year is a catastrophe, and the right first response to a child in that condition is love rather than enablement.

The CDC reads its own data as a case for affirmation. I read it as evidence that what we are currently doing is not working, and that the children carrying the most pain are the ones being used hardest as a proof point by every side. Both readings begin from the same numbers. Expect to be told the first is the only compassionate one available.

What the medicine actually rests on

The interventions those conversations lead toward rest on an evidence base that national health authorities have now reviewed repeatedly and rated poorly.

England’s Cass Review, an independent review commissioned by NHS England and published April 10th, 2024, examined 50 studies on puberty blockers and 53 on cross-sex hormones and found the evidence remarkably weak, urging extreme caution. The UK subsequently restricted puberty blockers for gender dysphoria in minors. The November 2025 HHS review rated the evidence for blockers, hormones and surgeries in minors as very low quality, listing infertility, sexual dysfunction, impaired bone density, adverse cognitive effects and regret among the risks, and noting that Finland, Sweden, Norway and England had reached similar conclusions and restricted these interventions. On June 18th, 2025 the Supreme Court upheld Tennessee’s ban on these treatments for minors in United States v. Skrmetti, 6-3, leaving comparable bans standing in roughly 25 states.

Who Decides What Your Child Is Taught

The other half of this is procedural, and for many of the families I talk to it outweighs every test score above. When the school forms a view about your child, are you told?

On June 27th, 2025, the Supreme Court ruled 6-3 in Mahmoud v. Taylor that parents in Montgomery County, Maryland were entitled to a preliminary injunction restoring notice and opt-outs for a set of LGBTQ-themed storybooks used from pre-K through fifth grade. The district had provided both, then ended them in March 2023 when too many parents used them. Justice Alito wrote that the right of parents to direct their children’s religious upbringing “would be an empty promise if it did not follow those children into the public school classroom.”

Separately, as of May 2024, Parents Defending Education counted 1,062 districts across 38 states and DC with written policies authorizing or requiring staff to withhold a child’s stated gender identity from that child’s parents. That covers 18,658 schools and nearly eleven million students. The count comes from an advocacy group, so weigh it accordingly, though the figures were cited in Supreme Court filings.

Notice what Mahmoud actually won, though. A notice and an exemption. It did not change what is taught. For families who would rather build a curriculum on their own convictions than negotiate exemptions from someone else’s, an opt-out form is a thin kind of victory.

The Variable Underneath

Line these studies up and look at what each one actually isolated, because something happens.

Martin-Chang set out to measure homeschooling and found that structure was the variable, and structure is the thing the parents supply. Bloom set out to measure tutoring and found that one adult adjusting to one child in real time beat every scalable alternative anyone has built since, and for most of human history that was a child’s parents. Harvard set out to measure school types, and the three findings that survived its strictest statistical test were faith, service and forgiveness. No school in the country teaches those. Every family either transmits them or fails to. The Tennessee trial found that moving children out of the home earlier made things worse by sixth grade. The relative-age research found that institutions medicate what a parent would simply recognize as too young a child. The CDC ties adolescent distress to social media use and to diet, which are both household decisions.

And the most thorough review of the homeschool literature reports that once you control for family background, the measured homeschooling effect shrinks considerably, with family background doing most of the work. That finding is almost always quoted as a rebuttal to people like me. Read it again. It is not a rebuttal. It is the result.

Every attempt to isolate the school dissolves into the family. The instrument keeps pointing past the classroom, to the kitchen.

These words, which I am commanding you today, shall be on your heart. You shall teach them diligently to your sons and shall speak of them when you sit in your house and when you walk by the way and when you lie down and when you rise up. (Deuteronomy 6:6-7, LSB)

Notice the schedule in that verse. Sitting, walking, lying down, rising up. There is no bell in it and no building. The model of education Scripture gives is a household, all day, conducted by parents who are themselves under instruction. The charge is not to outsource that. It is to carry it.

Then there is the verb. “Teach them diligently” translates shanan, and everywhere else in the Old Testament, eight of its nine appearances, it means to sharpen. To whet a blade. To put a point on an arrow. This passage is the only place the translators reach for a word about teaching, and you can see why they had to. What the Hebrew actually says is that you sharpen these words into your children. Not that you cover the material. Not that you deliver the curriculum. You hone, and you repeat, and you keep drawing the edge across the stone, sitting and walking and lying down and rising up, until the thing is sharp.

That is not a description of a school day. It is a description of a childhood. Three thousand years before anyone ran a regression, the variable the regressions keep finding had already been named.

When the best available evidence on how children turn out is gathered honestly and allowed to point wherever it points, it points at the family. A healthy one, rooted in faith, where the parents understand they are stewarding someone else’s children and not merely raising their own. That is not a coincidence a Christian should need talking into.

What this costs, and what we built

Almost nothing in modern life is arranged to make this possible. Two incomes are assumed. The hours are not yours. And the part that breaks first is never the love and never the willingness. It is the scaffolding, in faithful families, with good intentions, and a house full of children on a Tuesday.

Two things about that. The first is that provision is God’s department, not ours. Families who have stepped into this have generally found the way made, and I am not going to argue anyone out of that trust.

The second is the reason this company exists. We took the most advanced technology available and pointed it at the one job we could honestly lift off a parent: the structure. Not the teaching. Not the authority. Not the discipleship. Those were never ours to take, and a tool that tried to take them would be the exact thing this article has spent four thousand words warning you about.

The AI is not the teacher. You are. We built you a trustworthy, faith-rooted assistant.

What Do These Stones Mean

When Israel crossed the Jordan, Joshua had twelve men carry twelve stones out of the riverbed and stack them on the far bank. The reason he gives is specific, and it has stayed with me through this whole exercise:

When your children ask their fathers in time to come, saying, “What do these stones mean?” then you shall make your children know, saying, “Israel crossed this Jordan on dry land.” (Joshua 4:21-22, LSB)

Read who is in that sentence. The children ask. The fathers answer. Nobody else is assigned a part.

And the verb in the answer is causative. Hoda’tem, the form of yada, to know, that means to cause someone to know. Not expose them to it. Not make it available should they show interest. Cause it. The stones were never a monument for strangers to admire. They were a teaching aid, stacked in advance by parents who knew the question was coming, so that there would be something true to say on the day it came.

That is the whole job, and it has not changed. Which is why the stones had to be ones that would actually hold.

Sources

Every figure above is linked where it appears. Gathered together here for reference.

Public school performance and attendance

• NAEP 2024, twelfth grade. Top five takeaways, National Assessment Governing Board. Reported by K-12 Dive and Chalkbeat.
• Chronic absenteeism, 2025. From Attendance Crisis to Chronic Condition, American Enterprise Institute.

Early childhood and school entry

• Bassok, Latham and Rorem (2016). Is Kindergarten the New First Grade? AERA Open. Compares kindergarten classrooms in 1998 and 2010.
• Tennessee Voluntary Pre-K, randomized trial. Vanderbilt University, followed through sixth grade. Reported by Chalkbeat Tennessee.
• Layton, Barnett, Hicks and Jena (2018). Attention Deficit-Hyperactivity Disorder and Month of School Enrollment. New England Journal of Medicine. 407,846 children.

The school calendar and adolescent mental health

• Hansen and Lang (2011). Commitment or Confinement? Seasonal Patterns in Youth Suicide. Economics of Education Review.
• NBER Working Paper 30795. Teen suicide timing against regional school start dates. NBER.
• School closures and youth suicide, 2020. Journal of Human Resources.
• Pediatric suicidality admissions, 2016 to 2021. JAMA Network Open, via PMC.
• Los Angeles pediatric emergency department, school weeks versus breaks. Psychology Today.
• CDC Youth Risk Behavior Survey 2023. Results summary, plus the MMWR analyses of social media use and of diet, grades and school connectedness.

Homeschooling: growth and outcomes

• Johns Hopkins Homeschool Hub. Homeschool growth, 2024-25. Built from state filings rather than advocacy surveys.
• Massachusetts enrollment and homeschooling since 2019. Reason.
• Martin-Chang, Gould and Meuse (2011). Structured and unstructured homeschooling against public schooling, 74 children. Summary.
• Cogan (2010). Exploring Academic Outcomes of Homeschooled Students. ERIC. 76 students at one university.
• Chen, Hinton and VanderWeele (2021). School types in adolescence and subsequent health and well-being in young adulthood. PLOS ONE. Harvard’s Human Flourishing Program, 12,288 children of nurses, 187 of them homeschooled.
• Kunzman and Gaither (2020). The research review finding that family background does most of the work. Via McCracken and Coleman.

The mechanism

• Bloom (1984). The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring. Educational Researcher 13(6), 4 to 16. PDF.
• The replication record. Two-Sigma Tutoring: Separating Science Fiction from Science Fact, Education Next.

Identity, medicine and parental authority

• CDC YRBS 2023, transgender and questioning students. MMWR supplement. The earlier Williams Institute estimate, via NBC News.
• The Cass Review (2024). Final report, independent review commissioned by NHS England.
• HHS review (November 2025). Pediatric gender dysphoria treatment.
• United States v. Skrmetti, No. 23-477, decided June 18th, 2025. Opinion.
• Mahmoud v. Taylor, No. 24-297, decided June 27th, 2025. Opinion, and the case history from Becket.
• District policies withholding gender information from parents. Parents Defending Education’s count, via Heritage.

Figures were gathered on September 18th, 2026 and are linked to their sources above. If you find an error in one of them, I want to know. Write to us and I will correct it in place and say so.

Scripture quotations taken from the (LSB®) Legacy Standard Bible®, Copyright © 2021 by The Lockman Foundation. Used by permission. All rights reserved. Managed in partnership with Three Sixteen Publishing Inc. LSBible.org and 316publishing.com.