Think US gun violence is bad? A new study suggests it’s even worse than we think.

America is already known for leading the developed world in gun violence. But a new study finds the problem may be even worse than we think.

The study, from economists Jillian Carr and Jennifer Doleac, looked at new ShotSpotter data, which uses high-tech audio sensors to report gunshots, in Oakland, California, and Washington, DC. It found that only 12 percent of gunfire incidents resulted in a 911 call to report gunshots, and only 2 to 7 percent of incidents resulted in a reported assault with a dangerous weapon.

In other words, shootings are tremendously underreported in the US.

One catch to the research: ShotSpotter is likely picking up some false positives. Past evaluations have suggested the technology has anywhere from 50 to 97 percent accuracy, although Carr and Doleac acknowledge more rigorous research is necessary. But even if the low range is right, there would still be a lot of unreported shootings.

It’s also possible that many shootings go unreported because no one was injured or killed. It’s hard to imagine, after all, that a death went completely unnoticed, and Carr and Doleac note that homicide is reported to and by police “with near-accuracy.”

Still, Carr and Doleac say it’s likely that injuries in some incidents went unreported. For example, in drug-related or gang shootings, “neither the victim or offender is interested in involving the police.”

But it’s important for policymakers to know about these nonlethal shootings. Public policy’s goal isn’t just to reduce lethal crime and violence, but to reduce crime and violence in general. Yet much of the research on crime and gun policy focuses on homicides. Carr and Doleac argue that focus may skew the results of policy research — and could mislead lawmakers who are trying to keep the public safe.

Currently, crime research generally relies on surveys and crime reports from law enforcement. More specifically, the research tends to focus on reports for homicides, since homicide reports tend to have the most accurate data. Researchers use these data sources to try to evaluate the effects of certain policies — if crime reports went down after a policy was implemented, it’s presumed that the policy helped bring crime down (after some statistical checks).

But the ShotSpotter data suggests the traditional sources of crime research — the law enforcement reports — overlook a lot of crime. What’s more, Carr and Doleac suggest that a drop in reports of crime may just mean that people are reporting fewer crimes even as it continues happening. So researchers using the traditional sources may have been picking up how policies affect reports of crime, not necessarily crime itself.

As an example, Carr and Doleac cited their previous study with ShotSpotter, which found that curfews lead to more gun violence. “That study would not have been possible with traditional crime measures, because curfews also affect reporting rates (more police scrutiny could increase reporting, while fewer witnesses out on the streets could decrease reporting),” they write. “Indeed, when [we] conduct the same analysis using reported crime and 911 call data, the estimates suggest the opposite (and incorrect) conclusion: crime reports and 911 calls fall when curfews are in effect.”

Beyond the implications for policy research, the findings are simply alarming. It’s hard to imagine that America’s gun violence problem is actually worse than we think, given that the country, even after the big drop in crime over the past few decades, still fares much worse than its developed peers. But this study shows that may be the case. At the very least, we’re not counting a lot of shootings as shootings.