Project 2025’s Mandate for Leadership—published in 2023 by the Heritage Foundation as a “playbook” for the next Republican presidential administration—included a recommendation to defund Head Start.

So far, the Trump administration has not acted on this recommendation. During the late-January freeze on federal funding, some Head Start centers experienced brief interruptions, and on April 1 the number of regional offices providing administrative support to Head Start centers was reduced from 10 to 5. But the basic idea that the federal government should fund free childcare for low-income families remains in place—for now.

Before more serious action is taken, it’s important to check how Project 2025’s recommendation stacks up to rigorous evidence.

An example from abroad illustrates what might happen here if Head Start disappeared. In 2007, when Queensland, Australia, shut down its free preschool program, four-year-old children’s vocabularies shrank by 5 percentile points, and their kindergarten readiness fell by 9 percentile points, relative to four-year-olds in other Australian states that kept state preschools running. In addition, research by Elise Chor and colleagues shows, about 7 percent of four-year-olds’ mothers stayed home who would have worked if Queensland hadn’t defunded free preschool.

Why would the United States risk similar consequences? We’re already beating ourselves up over 2024 results from the National Assessment of Educational Progress, which showed that children’s math scores haven’t recovered since the pandemic, while reading scores have actually gotten worse. And for more than three decades, our public policies have told low-income mothers that the government would rather see them working for pay than at home with their children. Since 1996, Temporary Assistance to Needy Families (TANF) has required recipients to work for benefits—unless they cannot find affordable childcare, a requirement that Head Start helps them to meet. Even earlier, the 1988 Family Services Act attached some work requirements to TANF’s predecessor program, Aid to Families with Dependent Children. And today’s Congress is considering work requirements for Medicaid.

So where does the idea of canceling Head Start come from? Let’s take a closer look.

The Case Against Head Start Was Made in a Single Paragraph

Project 2025’s case against Head Start appears in chapter 14 of Mandate for Leadership, which addresses the Department of Health and Human Services. The chapter’s author, Roger Severino, directed that department’s Office for Civil Rights during the first Trump administration and is now the Heritage Foundation’s vice president for domestic policy.

Severino wrote just one paragraph about Head Start. Here it is, in full:

Office of Head Start (OHS)

Eliminate the Head Start program. Head Start, originally established and funded to support low-income families, is fraught with scandal and abuse. With a budget of more than $11 billion [in 2022], the program should function to protect and educate minors. Sadly, it has done exactly the opposite. In fact, “approximately 1 in 4 grant recipients had incidents in which children were abused, left unsupervised, or released to an unauthorized person between October 2015 and May 2020.” Research has demonstrated that federal Head Start centers, which provide preschool care to children from low-income families, have little or no long-term academic value for children. Given its unaddressed crisis of rampant abuse and lack of positive outcomes, this program should be eliminated along with the entire OHS. At the very least, the program’s COVID-19 vaccine and mask requirements should be rescinded.

The Office of Head Start actually funds two free-care programs for low-income families—not just Head Start proper, but Early Head Start as well:

• Head Start proper, which launched in 1965, funds free preschool centers for three- and four-year-olds.
• Early Head Start, which launched in 1996, provides free services to expectant mothers and to children from birth to age three. (Some three-year-olds attend Head Start proper instead.) In 2023, about two-thirds (63 percent) of Early Head Start children received center care, while the rest received services at home.

In fiscal year 2023, Head Start served 544,549 children at a cost of $12,000 per child, and Early Head Start served 186,585 children at a cost of $19,000 per child. The total budget of the Office of Head Start was $12 billion, amounting to 0.7 percent of federal discretionary spending.

Given the disruptive implications of ending childcare programs that serve more than 730,000 children and their parents, Head Start merits more than a paragraph of discussion. Before the government considers action on the program, it deserves a thorough look at the evidence.

How Common Is Abuse and Neglect at Head Start?

Let’s start with the claim that Head Start has an “unaddressed crisis of rampant abuse.” In support of this claim, Severino cited a 2022 report on Head Start’s safety record by Suzanne Murrin, then a deputy inspector general in the Department of Health and Human Services. Murrin did indeed report that “during the period of review [October 2015 and May 2020], 27 percent of Head Start grant recipients”—about one in four, as Severino claimed—“received an adverse finding for child abuse, lack of supervision, or unauthorized release.”

That sounds pretty bad, but let’s unpack what a rate of one in four grant recipients really means. It’s easy to get the impression that one in four children was endangered, but a “grant recipient” was not an individual child, or even a single childcare center. Instead, as defined by the report, “grant recipients” included 1,611 different “public agencies, for-profit organizations, nonprofit organizations, school systems, and Tribal governments,” which managed more than 20,000 Head Start centers serving about 861,000 children in 2021. That’s about 4 million “child years” over the period—four years and eight months—covered by the report.

Over that time period, there were just over 1,000 reported incidents at Head Start centers (see Table 1). In other words, each year about one child in 4,000 suffered a reported incident at their Head Start center.

That’s one child in 4,000—not one in four.

It’s important not to minimize these incidents. Although only 42 children were released to an unauthorized person (one of the examples highlighted by Severino), 533 children were left unsupervised, and 374 experienced “physical abuse or corporal punishment, defined as hitting, spanking, shaking, slapping, twisting, pulling, squeezing, or biting a child.” The report found another 124 incidents that had not been reported, and there were almost surely more.

On the whole, we agree with the report that “the Administration for Children and Families should improve oversight of Head Start to better protect children’s safety.” But a rate of one reported incident for every 4,000 children contradicts Severino’s claim that abuse at Head Start centers is “rampant.”

We also question his claim that safety issues have gone “unaddressed.” In an Appendix to the Inspector General’s report, January Contreras, assistant secretary for the Administration for Children and Families, which oversees Head Start, concurred with the report’s recommendations and described several measures that the administration was taking to better monitor safety, including a new online training course for Head Start staff called iLookOut and an effort to gather additional safety data from state agencies.

It’s hard to know how effective these steps have been—especially given the paradox that better reporting typically pushes reported incidents up rather than down—but the government is monitoring Head Start’s safety and instituting reforms. The “crisis,” if there is one, has not gone “unaddressed.”

On Average, Children Are Safer at Head Start than at Home

Even if Head Start’s safety record were worse, it’s hard to see how shuttering the program would make kids safer. To assess the safety implications of closing Head Start, we’d need to look at children’s alternatives. Project 2025 didn’t address that, and neither did the inspector general’s report, since when it was published in 2022 no one had suggested—not so publicly, at least—that the program should be ended.

If Head Start closed, some families would seek other childcare centers, which raises several questions. Does the country have enough capacity to absorb the 545,000 children in Head Start centers or the 115,000 Early Head Start children who receive center care? What would it cost to serve those children, and who would pay for it? And are other childcare centers safer, or less safe, than federally regulated Head Start centers?

Many families who couldn’t find free care would keep their children at home during the day, and we know something about the safety of Head Start children’s home environments. The best evidence comes from the Head Start Impact Study—a randomized controlled trial that assigned 4,667 eligible three- and four-year-olds to be either admitted or denied access to Head Start. At the start of the study, in fall 2002, 45 percent of participating parents reported that they had spanked their children at least once in the previous week. (About 5 percent of parents didn’t answer this question.) Seven percent of parents reported parenting styles that the study classified as harsh and “authoritarian,” and 7 percent reported practices classified as “neglectful.” These numbers are much higher than the one in 4,000 who experienced some form of abuse or neglect at a Head Start center.

We do not mean to stigmatize low-income parents. Many middle-class parents spank their children, too. In the 2017 Monitoring The Future survey of 35-year-old parents who had at least a high school diploma, 37 percent reported spanking their children at least once in the past year. But that was in one year; remember that 45 percent of parents applying to Head Start reported spanking their children at least once in the past week. More broadly, surveys show that spanking and other forms of corporal punishment are more common among preschool-aged children, and more common in families enduring the stress of poverty—the very families that Head Start was designed to help. On average, many of those children are safer at a Head Start center than they are at home.

Head Start Makes Homes Safer

Not only is Head Start safer than some recipients’ homes, but participation in Head Start also makes some of those homes safer.

Again, the best evidence comes from the Head Start Impact Study. In addition to surveying families with children admitted to Head Start—the treatment group—the Head Start Impact Study also surveyed a control group of families who had applied to Head Start and were denied. Families in the treatment and control groups were comparable because the decision to admit some and deny others had been made, following the study design, entirely at random. By comparing the treatment and control groups, we can estimate the effect of Head Start access on various outcomes, including spanking. Our analysis of the data agrees with the official report’s conclusion that admission to Head Start reduced parents’ spanking of children admitted at age three by about 5 percentage points, though it had no effect on their spanking of children admitted at age four.

Results from the Early Head Start Research and Evaluation Project—another randomized controlled trial, this time of the birth-to-three program—showed much the same thing. In a 2013 analysis by Cheri Vogel and colleagues, parents with access to Early Head Start were 6 to 7 percentage points less likely to spank their two- and three-year-old children than parents who had applied to Early Head Start but had been randomly denied. Early Head Start also increased immunization rates and improved family functioning overall; parents with access to Early Head Start were more likely than parents in the control group to teach and read to their children, to enforce a regular bedtime, and to be engaged rather than detached during play. Analysis of Head Start Impact Study data by Alexander Gelber and Adam Isen similarly finds that Head Start increases parents’ positive interactions with their children, including how much time they spend with them in educational activities.

After looking closely at the evidence, it’s hard to endorse the idea that Head Start suffers from “rampant abuse” or that shutting it down would make children safer. Despite Head Start’s imperfect safety record, eliminating the program would almost surely make children less safe.

Keeping Head Start open, while continuing to improve oversight and supervision, including quick removal of the few workers who mistreat children, would serve children’s safety and wellbeing much more effectively.

Head Start Children Pull Ahead in the Short Term

Severino’s second argument was that “research has demonstrated that federal Head Start centers . . . have little or no long-term academic value for children.”

The adjective “long-term” is doing some heavy lifting here. In the short term, it’s quite clear that Head Start and Early Head Start have cognitive benefits. Both randomized trials showed that children enrolled in Head Start and Early Head Start scored significantly higher than control-group children on tests of “pre-literacy” skills—such as pointing to the picture that matches a spoken word, naming and sounding out letters, and so on. For children who entered the Head Start Impact Study at age three or four, participation in Head Start raised scores by about 7 percentile points on average, with benefits ranging from 4 to 11 percentile points, depending on the skill. For children in the Early Head Start Research and Evaluation Program, access to Early Head Start raised scores by 4 to 5.5 percentile points. These results put Head Start in the most effective third, and Early Head Start in the upper half, of educational interventions that have been rigorously evaluated. (The average effect of an educational intervention is about 4 percentile points, and the most effective third of interventions have effects of at least 7 percentile points.)

These effects probably understate what Head Start actually accomplishes. In the Head Start Impact study, families in the control group, who were denied access to their first-choice Head Start center, didn’t just sit still. According to analysis by Fuhua Zhai and colleagues, 15 percent of control-group children found their way into another Head Start center, and 30 percent found a childcare center that was not run by Head Start. But 55 percent of control-group children ended up being cared for by parents or other relatives—and children in the Head Start treatment group far outperformed those children. After a year, children in the Head Start group scored 11 to 19 percentile points higher than children in parent or relative care. These effects, which would put Head Start in the most effective fifth of all educational interventions, are obscured when the analysis includes both control-group children in parental care and those in center care—many of whom did about as well, or not too much worse, than children in the Head Start treatment group.

The contrast between Head Start and family care has significant implications. If Head Start abruptly shut down, relatively few of the 700,000-plus children now in Head Start and Early Head Start would find other free centers before they entered kindergarten. Most of these kids would end up where half of the control group in the Head Start Impact Study went: to informal care by a parent or other adult in their own home or another home-based setting. Informal care providers generally do not need to meet licensing requirements, including basic health and safety standards. And compared to Head Start teachers, they often lack the time, resources, and experience needed to build children’s early literacy and number skills in preparation for kindergarten.

But within a Few Years, Other Children Catch Up

Although Head Start builds skills before children enter kindergarten, the catch—which Severino highlighted with his emphasis on “long-term academic benefits”—is that the lead built up by Head Start children doesn’t last very long.

We see this in both randomized Head Start trials. In the Head Start Impact Study, children in the treatment group finished the Head Start year with a lead in skills over the control group. The leads ranged from 3 to 11 percentile points and were statistically significant in 7 out of 10 early-language, literacy, and pre-writing skills in English. Children who had started Head Start at age three also had a significant 4-percentile-point lead in one out of two tested math skills.

By the end of kindergarten, though, children in the Head Start treatment group had no significant lead in any skill related to math, reading, or writing in English. Later, on tests given in 1st and 3rd grade, the treatment group rarely showed a significant lead, and the few leads that the evaluators reported as “significant” might well have been statistical accidents rather than true benefits of Head Start. In 1st and 3rd grade, children who entered the Head Start treatment group at age four were tested in 20 different skills and only showed a significant lead over the control group on two of them. Children who entered the treatment group at three were tested on the same 20 skills and only showed a significant lead on one. None of this is convincing evidence for the benefits of Head Start, since even if there were no benefits, simply by chance one test out of 10 would produce a “significant” result (defined as p < .10 in the Head Start Impact Study reports). Another reason not to believe the significant results is that there was no consistency in the results; the tests giving “significant” results were not the same in 1st grade and 3rd grade, nor were they the same for children entering at age three as for children entering at age four.

Results from the Early Head Start Research and Evaluation Program also showed the control group catching up. At age three, the treatment group children were ahead of the control group on many skills, and much of the their lead persisted until children started kindergarten at age five. But when children were retested in 5th grade, the treatment group’s lead was gone. Out of 49 impact estimates, only one was significant at the .10 level, and even that one cannot be taken seriously. Even if there were no true impacts, we would expect five significant results. The investigators went on to report results for various subgroups, but Paul von Hippel and Brendan Schuetze have shown that subgroup analyses are usually misleading, especially when there is no main effect.

This phenomenon—which is common for early childhood interventions, not just Head Start—is often described as “fade-out,” but it’s more accurate to call it “catch-up.” The issue is not that Head Start children somehow forget what they learned in preschool but that similar children who didn’t attend Head Start eventually catch up.

But if other children catch up after Head Start ends, how could that be that Head Start’s fault? And how would shutting it down improve the situation? After all, if the first runner on a team in a relay race built up a substantial lead, you wouldn’t blame that runner if her other teammates let the opponents catch up. Instead, wouldn’t you keep your lead-off runner for the next race? Wouldn’t you want more runners like that? Wouldn’t you focus on training the later runners not to drop the baton?

And what might that mean for sustaining the benefits of Head Start?

The Relay Race: How Can We Sustain the Benefits of Head Start?

No one fully understands the reasons for catch-up, but there are at least two compelling explanations.

One is that much of what’s taught in Head Start and other preschool programs gets repeated in kindergarten. In North Carolina, Lora Cohen-Vogel and colleagues have found that more than a third of the math and reading content children are taught in kindergarten—skills like counting to 20, writing their own name from left to right, and recognizing how many objects are in a small group—are redundant with what children learn in preschool. More generally, Mimi Engel and colleagues report that kindergarten teachers cover a lot of material that children already know. In national data, kindergarten teachers report teaching basic counting and shapes an average of 13 days per month—even though 90 percent of children have already mastered these topics by the time kindergarten begins. If kindergarten teachers spent less time on what children already know and more time challenging them with new content, almost everyone would learn more. That would set a higher baseline for 1st grade, which could offer a more challenging curriculum as well. And so on up the line.

The other explanation is that not everyone attends preschool. It would be easier to reduce redundancy between kindergarten and preschool if all low-income children attended Head Start or something like it, but Head Start has never received enough funding to serve all children who qualify. Some states and districts offer free preschool outside the Head Start program, but only about 40 percent of low-income three- and four-year-olds attend any kind of preschool, public or private. As long as preschool access is limited, kindergarten teachers won’t be able to count on poor children having the skills that preschool imparts, nor to move ahead as fast as preschool graduates can handle. A local field experiment by John List and Haruka Uchida found that children who attended a free preschool in Chicago Heights were more likely to keep the benefits if more of their classmates had attended that preschool as well.

If preschool availability is a problem, maybe we should be talking about expanding preschool access rather than defunding our largest preschool program. A few U.S. states and cities (for example, Oklahoma, Georgia, Wisconsin, and Boston), as well as Mexico, Japan, South Korea, and nearly all of Europe, have free preschool programs that are “universal” or open (though not required) for all children, rather than for just a fraction of the poorest children. A 2021 review by David Blau found that universal preschool programs often sustained their benefits to participants through elementary school, into high school, and beyond. An evaluation of Wisconsin’s universal preschool program, published by Hyunwoo Yangs a few years after Blau’s review, found benefits sustained through 3rd grade in reading, but not in math. Despite universal access, the benefits of universal pre-K are typically concentrated among the poorest families.

Shutting down Head Start, then, makes little sense, given the evidence—but at least three promising measures might sustain its effects into later grades: 1) broaden access to preschool; 2) don’t repeat topics and skills in kindergarten if children already know them; and 3) keep children who attended preschool together in elementary school.

Sleeper Effects in Adulthood

Despite the fadeout of test score effects, several studies have reported that children who had attended Head Start often did better socially and economically as adults. In a 2009 article, for example, David Deming compared children who attended Head Start in 1984–1990 to their siblings who had not. On tests, he found that the Head Start sibling scored 5 percentile points higher through elementary school, but that their lead had shrunk to 2 percentile points by middle school. Nevertheless, the Head Start sibling was 6 to 7 percentage points less likely to repeat a grade or be diagnosed with a learning disability, 9 percentage points more likely to graduate high school, and 7 percentage points less likely to be unemployed or suffer from poor health as young adults.

Follow-up work by Lauren Bauer and Diane Schanzenbach showed that Head Start siblings were 9 percentage points more likely to complete higher education, had a greater sense of self-esteem and self-control, and displayed more positive behaviors when they became parents themselves. Compared to siblings who hadn’t attended the program, Head Start graduates were more likely to engage in behaviors such as reading to their children, teaching them colors, shapes, numbers, and letters, providing physical affection and praise, and sharing in their children’s favorite activities.

Oh, and they didn’t spank their children as much—perhaps because they hadn’t been spanked as much themselves.

These benefits may even be passed on across generations. Elise Chor’s analysis of the Head Start Impact Study data showed that Head Start access improves children’s math skills and behavior through 3rd grade for children whose own mothers attended Head Start in their youth. Andrew Barr and Chloe Gibbs used data from Head Start’s early years to show that program attendance led to improved outcomes into adulthood among the children of Head Start participants, including greater educational attainment and less teenage pregnancy and criminal engagement.

Because of these long-term benefits, many scholars calculate that Head Start more than pays for itself. The costs of providing free care when children are young gets recouped with interest when those children reach adulthood. Data from the program’s early years indicate that Head Start’s benefits outweigh its costs by a ratio of up to seven to one through its impacts on educational attainment, criminality, and mortality. Using more recent data from the Head Start Impact Study, Patrick Kline and Christopher Walters find that, if nothing else, Head Start moves children out of other publicly funded preschool programs; accounting for cost savings in those alternative public programs alone makes Head Start pay for itself.

Where do these “sleeper effects” in adulthood come from? No one really knows. Scholars used to talk vaguely about “non-cognitive skills” that aren’t measured by reading and math tests. Yet the Early Head Start Research and Evaluation Program measured a variety of social and emotional outcomes—including attentiveness, distractibility, peer relationships, “externalizing” behaviors (acting out), and “internalizing” behaviors (such as withdrawal or anxiety)—and found just as much fadeout as it did for the academic measures. At age five, the children in the Early Head Start group were doing better than the control-group children on many social and emotional outcomes, but by 5th grade the control group had caught up. Head Start Impact Study data show similar patterns, where short-term social-emotional benefits do not persist beyond the first year or two.

Twenty or more years is a long time to wait to see a financial return from Head Start, but some of the program’s benefits materialize much sooner. For example, studies by Deming, Janet Currie and Duncan Thompson indicate that when Head Start prevents grade repetition, it typically does so by the age of 10. That means the program pays back its cost in full, since repeating a year of school costs about as much as a year of Head Start.

Benefits to parents, when they occur, also show up quickly. In analyses of the Head Start Impact Study, Cuiping Schiman found that Head Start let 7 percent of mothers switch from part-time to full-time work, and Terri Sabol and Lindsay Chase-Lansdale found that Head Start helped 16 to 20 percent of parents who had already started college finish a degree or certificate within four years. Both benefits were limited to parents whose child started Head Start at age three. Note that benefits to parents are another possible explanation for sleeper effects, since parents who are more educated and more advanced in their careers may be better positioned to help their children in young adulthood.

Follow Up Now with Adults who Participated in Head Start Trials Decades Ago

A major limitation of the evidence on Head Start’s long-term benefits is that it’s somewhat dated. To a degree, dated evidence is unavoidable—if you want to know the effects 20-plus years after children participated in Head Start, you’ll inevitably be looking at individuals who attended 20-plus years ago.

But existing evidence on Head Start is more dated than it needs to be. Deming’s sibling study, for example, focused on children who attended the program between 1984 and 1990. Those Head Start graduates are 40 to 45 years old today. Whatever benefits they’ve experienced, the program was different when they participated, and the alternatives available to their non-Head Start siblings, Anamarie Whitaker and colleagues argue, were probably less stimulating than they are today. Head Start’s effects on test scores persisted longer in the old sibling studies than they did in more recent trials, and it’s reasonable to wonder if the sleeper effects detected back then would be hard to replicate today.

There’s a straightforward way to update the evidence on Head Start. Why not follow up with families who participated in randomized trials of Head Start? Children from the Head Start Impact Study, who were three and four in 2002–03, are 25 and 26 today. Children from the Early Head Start Research and Evaluation Project, who were newborn to age three in 1996, are 30 to 32 today. How are those young adults doing? How are the Head Start graduates faring compared to young adults from the control groups? And how are their parents doing? Did mothers who finished college or worked full-time thanks to Head Start childcare reap the benefits of career advancement and economic stability as they got older?

We could answer these questions by finding trial participants and reaching out to them directly, but it would be less expensive to start by matching trial participants to administrative data that’s already been collected—an increasingly common practice in program evaluation. By matching trial participants to data collected by the IRS, various state agencies, credit bureaus, and nonprofits that collect data on the labor market, such as the Burning Glass Institute, we could estimate Head Start’s effects on outcomes such as employment, income, homeownership, marriage, and contact with the criminal justice system.

All the data is out there. We could learn a lot about Head Start’s long-term impacts just by matching it up.

Then Make Decisions Based on Evidence

At a time when the federal government is terminating many research projects, it is vital to recognize the value of rigorous program evaluations. Nearly everything we know about the effects of Head Start comes from federally funded research. The inspector general’s report on Head Start’s safety record was conducted inside the Department of Health and Human Services. The sibling studies relied on data collected by the National Longitudinal Studies, which was sponsored by the U.S. Bureau of Labor Statistics. The Head Start Impact Study and the Early Head Start Research and Evaluation Project were undertaken by external research organizations under contract with the Department of Health and Human Services.

In 2023, OHS had a research budget of $20 million—which sounds like a lot until you realize that it’s less than a fifth of a penny for every dollar the office spends on Head Start and Early Head Start. That’s a small price to pay to find out how well a $12 billion annual program is working.

Why not get full value from the randomized studies that started 20-plus years ago by following up with the families who participated? Then we could make an up-to-date, evidence-based decision about whether Head Start warrants the continued investment of public dollars.

Paul von Hippel is professor and associate dean for research at the University of Texas, Austin, LBJ School of Public Affairs. Elise Chor is assistant professor of political science at Temple University. Leib Lurie is co-founder and CEO emeritus of Kids Read Now, which provides more than a million books annually to elementary students.