Yesterday, state traffic-safety officials released some alarming-sounding news: In 2014, the year recreational pot stores opened for business in Washington, the number of fatal car crashes involving THC doubled.

And that, of course, is awful—every fatal car crash is awful. “This is particularly affecting young men,” says Shelly Baldwin of the Washington Traffic Safety Commission (WTSC). According to WTSC data, fatal car crashes involving men from age 21 to 25 with THC in their system tripled from six in 2013 to 19 in 2014.

But before we jump to the legalizing-marijuana-means-highway-death conclusions that some headline writers and Republican speechwriters will inevitably exploit, let's take a closer look at the numbers. I want to be very clear: I'm not minimizing the suffering of people who've lost loved ones to car crashes, particularly car crashes involving impaired drivers. But drug-policy reform is a high-stakes game, locally and internationally, so it's important to look at this new data with clear eyes.

Washington, like every other state in the United States, reports its fatal car-crash statistics to the national Fatality Analysis Reporting System (FARS)—a kind of census bureau for traffic-fatality data. While FARS tracks many variables related to fatal crashes (kinds of vehicles, days of the week, ages of people who are killed), it doesn’t allow much nuance when it comes to whether marijuana was involved.

“That database only allows us to answer yes/no about whether marijuana was in the driver’s system or not,” says Baldwin. It doesn’t matter whether the driver has a high concentration of THC in her system, indicating she may have recently smoked pot, or some trace remnants of a cannabis metabolite—such as carboxy-THC—which can hang around in the body long after the psychoactive effects of marijuana have worn off.

That crude yes/no binary, Baldwin says, “was no longer adequate to our needs.” So the WTSC went back through state toxicology lab results since 2008 for more granular detail—which drivers involved in fatal crashes had THC in their systems (which might indicate impairment) and which drivers had other traces of cannabis (metabolites, non-psychoactive cannabinoids, etc.).

“We were pretty surprised by the amount, in 2014, of drivers with active THC involved in fatal crashes,” Baldwin says. (There is debate about how much THC has to be in a person’s bloodstream to indicate impairment—or whether the effects of marijuana can be quantified like they can with alcohol. But we’ll get to that in a moment.)

From 2010 to 2013, the number of drivers involved in fatal crashes who also tested positive for THC hovered between 32 and 38, out of around 600 fatal crashes per year, or about 6 percent. In 2014, after Washington legalized marijuana, the number jumped to 75 drivers out of 619, or 12 percent. Hence last week’s attention-grabbing headlines, like this one from the Oregonian: “Fatal Crashes Involving Marijuana Doubled in Washington After Legalization.”

That statement is accurate but might be misleading. Only half of those 2014 drivers had a THC level over five nanograms per milliliter in their blood—Washington’s “per se” DUI threshold. (In state law, if a driver crosses the per se limit for alcohol or marijuana, she’s guilty of a DUI, regardless of other factors.) The other half of the group was below the legal limit—some with just one ng/ml—leaving just 6 percent of drivers involved in fatal 2014 crashes with illegal levels of THC in their systems.

Complicating matters is the five ng/ml limit itself, which is contested by some researchers. “Lots of studies out there have discussed what the proper per se value should be,” says Brianna Peterson of the Washington State Patrol toxicology lab. “You could find a study that supports five, but you could find other studies that suggest other numbers.”

The lack of consensus is partially due to a lack of research (which is why it’s important for the federal government to change marijuana’s legal status—we can’t have a robust body of research until scientists have freer access to cannabis). “We have decades of studies on alcohol and blood-alcohol content and what that means,” Baldwin says. “We don’t have the decades of study yet to equate THC levels with impairment... It’s impossible to look at THC levels and say, ‘Oh, all these people were driving around stoned.’”

To complicate matters further: Half of all THC-positive drivers in 2014 also tested positive for alcohol, with a majority above the state’s .08 blood-alcohol limit. “We know if you combine alcohol with THC, you really mess with your risk of crashing,” Baldwin says. “At the .08 level, you’re about seven times more likely to get in a crash.”

So while “fatal crashes involving marijuana doubled in Washington after legalization,” it’s difficult to draw conclusions—except for the obvious one, which is that adults in Washington State now have more ready access to pot.

Nevertheless, Baldwin says, the trend is real. Is that troubling? Yes. Does it mean we should let prohibitionist types use WTSC’s data as an argument against further drug reform? No.