Kindergarten — that bastion of macaroni crafts, crayon-eating and life lessons in sharing — is actually a major driver of crime, at least according to data collected by New Hampshire state legislator Bob Kingsbury.
Kingsbury (R-Laconia), 86, recently claimed that analyses he’s been carrying out since 1996 show that communities in his state that have kindergarten programs have up to 400% more crime than localities whose classrooms are free of finger-painting 5-year-olds. Pointing to his hometown of Laconia, the largest of 10 communities in Belknap County, the legislator noted that it has the only kindergarten program in the county and the most crime, including most or all of the county’s rapes, robberies, assaults and murders.
The lawmaker, who opposes New Hampshire’s public kindergarten mandate, promoted his theory at a Belknap County meeting of state legislators last week, stirring enough controversy to provoke responses from the Democratic candidates in New Hampshire’s gubernatorial race: for the record, they support kindergarten.
So, what could account for the association he found between early childhood education and crime? “We’re taking children away from their mothers too soon,” Kingsbury said. He explained his research this way to the Huffington Post:
The sources I have is, I went to the Department of Education and got a list of kindergartens and I went to the safety department and got the crime report. … In general, the towns with a kindergarten have 400 percent more crime than other towns in the same county. In every county, the towns and cities with kindergarten had more crime.
But Kingsbury’s conclusions contradict virtually the entire body of literature on the effects of early childhood education. And his “research” isn’t published, of course. While there’s nothing wrong with investigating counterintuitive hypotheses, like the idea that kindergarten could cause crime, Kingsbury’s analysis makes a number of Science 101 errors that are instructive to examine.
(MORE: The 20 Best- and Worst-Paid College Majors)
To start, scientists who set out to investigate a topic first tend to review the earlier literature. Kingsbury argues that age 5 is too early for children to spend time away from their parents, but a check of previous data reveals that even younger children — preschoolers aged 3 to 4 — enjoy wide-reaching benefits by receiving high-quality education outside the home.
In a 2004 paper [PDF] by Nobel Prize-winning economist James Heckman of the University of Chicago and colleagues, a review of the literature found that overall, preschool and very early childhood education increase children’s educational achievement, raise their rates of future employment, cut welfare dependence and yes, reduce delinquency and crime.
A more recent study, published in the esteemed journal Science last July, followed more than 1,500 poor children born in Chicago between 1979 and 1980. Those who attended preschool starting at age 3 or 4 (the children went to the second-oldest federally funded preschool program in the country) were 22% less likely to be convicted for a felony, 28% less likely to develop alcohol or other drug problems, and 24% more likely to go to college, compared with those who started school later in childhood.
In other words, if kindergarten is linked to crime, it’s because kids start school too late, not too early. Certainly, questions about early childhood education are complex and worth asking: What are the effects of day care for the very youngest children? What is the right age to start kindergarten? Should it should last all day and what should the curriculum include? However, there’s no suggestion in the research that kindergarten per se leads to criminal activity.
Nonetheless, Kingsbury’s data appears to support the opposite conclusion. Why that appearance doesn’t reflect reality comes down to the difference between simple correlation and true causality.
(MORE: How to Cut Crime, Alcoholism and Addiction: It’s Not Elementary, But Preschool)
In life, many things are correlated. For example, you might find that students from communities with more hot tubs in their homes have higher rates of college graduation. But you can’t conclude from a mere statistical association that giving everyone a hot tub will guarantee college success: what’s far more likely is that richer communities have more hot tubs — and income is well known to be linked with higher educational attainment.
Similarly, there’s likely to be a strong correlation between air conditioner sales and ice cream sales, but no one would argue that buying an air conditioner makes you want to eat ice cream, or vice versa. Quite obviously, both effects can be attributed to a third factor: hot weather.
The fact that correlation does not equal cause — and that powerful correlations may be linked with unmeasured factors that are truly causal — is the reason that genuine scientific research often involves complex statistical analysis. Determining causality is extremely difficult in science, and it typically requires experiments that are designed to allow investigators to manipulate the conditions carefully and to rule out any other factors that might be at play.
That’s why, for example, the FDA requires data from randomized controlled trials of a drug before approving it. Without being able to compare outcomes in people who are randomly chosen to receive the drug to those who are randomly given placebo or another comparable treatment, it is difficult to determine whether the new drug hurts or harms. If investigators were to rely only on patient anecdotes of success, they would surely miss instances of failure, or they might mistake normal fluctuations in response to the drug or placebo effects for a true drug response.
(MORE: Do Family Dinners Really Reduce Teen Drug Use?)
In situations such as those involving kindergarten attendance, in which researchers cannot control who is exposed and who is not, studies need to be even more careful and they can never completely determine cause and effect. Without reviewing the existing literature or knowing about other relevant factors that can account for certain results, researchers cannot even statistically adjust, or control for, these possible confounding variables.
Kingsbury’s so-called research, for example, didn’t control for factors like income or population size, which are already known to have a big influence on crime rates, and which could also correlate with the presence of kindergarten programs. Towns with larger populations might have both more crime and more kindergartens — not because sending 5-year-olds to learn their ABCs together creates antisocial behavior, but simply because more people means more crime. Similarly, public kindergartens may be more likely to exist in lower-income communities, where crime rates already tend to be higher, because property is more affordable or because richer neighborhoods may rely more on private childcare arrangements.
Moreover, even if statistical correlations appear in the data, it doesn’t necessarily mean the associations are real. The more you look for patterns, the more you find them, and some correlations in data will occur by sheer chance. This is why research papers include figures like confidence intervals and measures of statistical significance — these calculations, while far from perfect, allow researchers to be reasonably sure that their results aren’t simply a fluke.
Many people find the math and scientific jargon used in scientific research to be intimidating or boring, but it’s there for good reason. While kindergarten may be too early to start teaching kids statistics, lawmakers who want to influence educational policy should have at least a passing familiarity with basic principles of good science. Otherwise, they risk misleading themselves and many others.
MORE: What You Need to Know About Bullying
Maia Szalavitz is a health writer at TIME.com. Find her on Twitter at @maiasz. You can also continue the discussion on TIME Healthland’s Facebook page and on Twitter at @TIMEHealthland.