Researchers have long know that autism is found in clusters. Certain communities and states have rates much higher than the rest of the country — a child born in California is several times more likely to be diagnosed with autism than a child in Alabama, for example. But the question why remains unanswered.

The geographical nature of the disorder seems to imply some sort of local, environmental cause. And a new study suggests just that: it found a strong correlation between autism rates and male reproductive system malformations, which can be caused by environmental toxins.

There is a complex array of factors that can influence autism rates, though: they seem to be affected by issues as diverse as income level, public awareness of the disorder, and diagnostic standards, experts say. In other words, just because there’s more autism in an area doesn’t mean there is something there that it causing more autism. This mess of proximate triggers, risk factors and health care standards can be hard to sort through, but big data is trying to provide some answers.

Researchers know that autism spectrum disorder (ASD) has a strong genetic component, but also are keenly aware that is also probably an environmental factors at play. For example, identical twins are cut from the same genetic cloth—but it’s possible for one of the twins to be autistic and the other not to be.

“We have lots and lots of evidence that genetics contributes to autism, but there’s also a lot of reason to believe that something else is going on too,” says Catherine Lord, who directs the NewYork-Presbyterian Center for Autism and the Developing Brain. “People have been trying very hard to figure out what on Earth that other thing is, and haven’t been very successful so far.”

One team of researchers tackled the problem by sorting a massive amount of data. Andrey Rzhetsky, a professor in the department of human genetics at the University of Chicago, led a team that looked at insurance information for almost 100 million people—amounting to nearly one third of the population of the country—across over 3100 counties in all 50 states. The team focused on ASD diagnoses, but looked at Intellectual Disability (ID) diagnoses, as well, in order to tease out any potential differences in causes of the two disorders.

To get information about environmental factors (which the study laments are “mostly undocumented”) they used an interesting proxy: diagnoses for male genital malformations. Specifically, if an area had a high rate of these kinds of malformations—micropenis, for example, or another disorder called hypospadias—that suggests possible environmental toxins in the area, like lead or pesticides, the study says. As Rzhetsky tells Newsweek, “The idea of the study is to use malformation rates in newborn boys as the canary in the coal mine.” (It was just published in PLOS Computation Biology.)

They also incorporated a large amount of other data, including income levels, ethnicity, the amount of viral infections reported, whether an area is urban or rural, and state diagnostic standards. Their conclusions were unequivocal: the greater amount of malformations in boys, the greater the autism rates in the area, suggesting a correlation between environmental factors and autism diagnoses. In fact, the study reports “an increase in ASD incidence by 283% for every percent increase in the incidence of malformations” in males. Rzhetsky says he was surprised—he didn’t expect the correlation “to be so profound.”

That wasn’t all they found. For example, states with stricter diagnostic standards had a lower amount of ASD diagnoses. In addition, income levels affected autism diagnoses — slightly. As income level went up, so to do diagnosis rates for ASD and ID, but not by much. That’s probably because wealthier families have access to more accurate diagnoses, Rzhetsky suggests.

Some autism experts express skepticism about what the study really shows. “I’m not so impressed,” says Eric Fombonne, a professor at Oregon Health & Science University and an autism researcher who focuses on its epidemiology. His biggest problem with the study falls under what can be called “the ecological fallacy”—basically that correlation is not the same thing as causation, especially when looking at the big picture. For instance it would be a fallacy to assume that just because alcoholism rates are high in a place with high suicide rates, that one is causing the other.

Fombonne also faulted the study for several methodological shortcomings. “The fact they have this significant association [between malformation and autism rates] increases my suspicions of ecological fallacy, because it makes no biological sense at all,” he says.

He says that he doesn’t think environmental causes aren’t a concern — he’s studying one himself — but thinks that anything discovered will probably only apply to “subset” of autism cases.  Future research needs to focus on what he called “gene-gene interactions” and “gene-environment interaction.”

“People who want to believe that environmental toxins play a role…will jump to the conclusion that this [study] is very good evidence that they do,” says autism researcher Angelica Ronald at Birkbeck, University of London. But that’s making assumptions. “I don’t think you can say that it shows environmental toxins play a role. It’s a working hypothesis.”