The first rule of science journalism is to read the study before you write about it. Alas, that hasn’t stopped media outlets from routinely misreporting, exaggerating or exercising insufficient skepticism about scientific research, particularly in the service of clickbait headlines and extra views.

A recent study from the American Journal of Epidemiology on whether the introduction of ridesharing has had an effect on alcohol-related crash fatalities was the latest victim of this kind of sloppy reporting. The Washington Post announced: “Is Uber reducing drunk driving? New study says no.” CNN declared: “Uber doesn’t decrease drunk driving, study says.” Fortune writes: “A New Study Says Uber Has Had No Impact on Drunk Driving.” Other outlets published similar stories.

But alcohol-related fatalities are not the same thing as drunk driving rates. According to the National Highway Traffic Safety Administration, nearly 10,000 Americans die each year in crashes involving a drunken driver; about two-thirds of that total are the drunken drivers themselves. But according to the FBI‘s Uniform Crime Reporting Program, there are annually about 1.1 million arrests for driving under the influence, which itself is just a fraction of the Centers for Disease Control and Prevention’s estimate of 121 million incidents each year in which intoxicated drivers aren’t caught. Astoundingly, according to one analysis, drunk drivers average just one arrest per 27,000 miles driven while intoxicated.

Ideally, society would like each of these three numbers to fall, but first, we must be able tell them apart. The AJE study’s authors make clear that they “did not examine Uber’s association with other traffic outcomes, including drunk driving incidences and nonfatal crashes.” This leads one to the conclusion that these journalists – or at least, those writing the headlines – may not have actually read the study at all.

When it comes to whether services like Uber and Lyft reduce drunk driving overall, logic suggests that more available and convenient transportation options likely would make it easier for many to plan a night out without getting behind the wheel, and reduce the incentives to drive under the influence. The CDC already lists taking a taxi as an important preventative measure and ridesharing options are usually cheaper and very often more convenient than getting a taxi. As these services increase in popularity – particularly among millennials, who both use ridesharing more and have a greater propensity to drive drunk – one would expect a corresponding decline in the number of DUI arrests and alcohol-related fatalities.

There isn’t much research on the subject, but most observations to date seem to support the supposition. A 2015 study published by Temple University’s Fox School of Business concluded the introduction of UberX in California led to a reduction in the rate of motor-vehicles homicides per quarter of between 3.5 and 5.6 percent. Another study by Mothers Against Drunk Driving, in partnership with Uber, also looked at the introduction of UberX in California and found that alcohol-related crashes by drivers under age 30 fell 6.5 percent, or 59.21 fewer crashes per month.

In June 2016, Providence College published a study which found that “DUIs are 15 to 62 percent lower after the entry of Uber” and the introduction of the service “is associated with a 6 percent decline in the fatal accident rate.” More recently, when Uber and Lyft were pushed out of Austin, Texas, DUI arrests spiked by 7.5 percent.

Given that background literature, it’s important to note some significant limitations in the approach used by the AJE study’s authors. They looked at data from 2005 to 2014 for the top 100 metropolitan statistical areas (MSAs) in which Uber has entered the market. Of course, in many of those MSAs, the company may be operating in the largest city or cities, but not across the whole metropolitan area. Also notable is that in most of the MSAs the study examines, Uber was introduced at some point in 2014, the same year the authors’ data ends.

Additionally, many of these jurisdictions also did not have friendly regulatory climates for ridesharing in the period the authors examined. Aside from California and Colorado, where state-level pre-emption laws were passed, most ridesharing regulation through 2014 was done at the city level. It was fairly common at the time for transportation network companies to have uncertain legal status and for jurisdictions to impose hostile regulations, issue cease and desist orders or hold sting operations to block Uber and Lyft from operating. Additionally, carpool services like UberPOOL and Lyft Line, which are significantly cheaper, had not yet become widely available. Today, ridesharing is cheaper, more popular and fully legal in most major cities.

It also may not be that surprising the AJE study didn’t line up with results from other research that focused on California. Uber was founded in San Francisco and launched there in 2009. Lyft launched in 2012. TNCs have been legal statewide in California since the California Public Utilities Commission’s initial rulemaking in 2013. California is the oldest and probably strongest ridesharing market. If ridesharing has an effect on alcohol-related fatalities or drunk driving more generally, it would show up there first.

In much of the rest of the country, ridesharing is not as well-established. According to Pew, as of December 2015, only 15 percent of U.S. adults had used a ridesharing service. Of those, only 17 percent reported they use it more than once or twice a month. In short, outside of millennials in major urban centers, ridesharing hasn’t yet caught on in a big way.

More research looking at more recent data is needed to better understand the effects of ridesharing on drunk driving rates. And with each new report, whatever its conclusion, one hopes science journalists will bring more care and a healthy skepticism to the table. In the meantime, this study alone isn’t a compelling reason to dismiss other evidence supporting the positive effects of ridesharing on reducing drunk driving.