When the clock strikes midnight on New Year’s Eve, people across the country celebrate by firing guns into the air. For a variety of reasons, it’s not clear how many Americans actually do this. It happens enough, though, that many law enforcement agencies issue pleas and warnings in an effort to discourage celebratory gunfire.

The practice may seem prototypically American — an exuberant, reckless outgrowth of gun culture — but it’s a problem around the world, from the Middle East to India to the Balkans to even relatively gun-averse places like France. Bullets fired in celebration do have to come down somewhere, and depending on their trajectory they can injure or kill, though this is statistically rare.

In Milwaukee, a system that tracks gunfire in select areas of the city sheds some light on people who get fire guns into the air on New Year’s. ShotSpotter, a system developed by Silicon Valley company SST Inc., uses microphones placed on rooftops throughout the coverage area to detect and triangulate gunshot sounds. In Milwaukee, this system covers about 11 square miles on the city’s near south side and central north side. The audio is analyzed through the ShotSpotter’s proprietary software, and location information is relayed to Milwaukee police dispatchers, ideally within a minute or less.

A reporter for Forbes obtained and released data on ShotSpotter alerts from Milwaukee and six other cities that use the system. SST treats the data it generates as proprietary and not subject to open-records laws, but dispatch reports from police departments are public record. The data obtained from Milwaukee cover 10,285 incidents between January 1, 2013 and September 29, 2015 when officers were dispatched in response to a ShotSpotter alert.

On each New Year’s Eve in that time period, the data show a dramatic spike in ShotSpotter alerts. Of course, some of these alerts could be false alarms caused by fireworks, but the company’s technology has gotten better at distinguishing between gunshots and other loud, abrupt sounds. Celebratory gunfire is also known to be a problem with Independence Day, but that time period in summer doesn’t show a similar jump in alerts. This is in part because people typically mark New Year’s over a single late night and early morning, whereas people tend to celebrate the Fourth of July over the course of a few days.

The ShotSpotter data indicate that officers responded to 82 incidents on Jan. 1, 2013 and 65 on the same date in 2014. In summer 2014, Milwaukee expanded the system from covering three square miles to its current 11, and the number of incidents went up to 121 on Jan. 1, 2015. For comparison, the next-highest number of ShotSpotter alerts on a single day during this 33-month period was August 22, 2015, during which officers responded to 38 of the alerts.

Milwaukee police spokesperson Sgt. Tim Gauerke could not provide specific information on how many celebratory gunfire incidents take place in the city on New Year’s.

“We don’t have a way of determining celebratory gunfire versus other gunfire — except for shots fired at midnight on New Year’s, [which are] assumed to be celebratory,” Gauerke said. “Unless an arrest is made and the suspect speaks to their motive, we won’t know if it was celebratory.”

Celebratory gunfire on New Year’s Eve has killed at least one person in Milwaukee in the last two decades — a 13-year-old girl died after being struck by a rifle round in 1999.

New Year’s Eve tends to be a long night for SST employees who work on ShotSpotter systems, which is used by about 95 police departments around the country. On New Year’s and the period around July 4, CEO Ralph Clark said, the company tends to staff up, provide food for employees, and brace for a wave of information flooding the system.

“We kind of beg our clients’ indulgence,” he said of these fireworks- and gunfire-heavy occasions. “We’re probably not seen at our very, very, very best during those periods of time. More things slip through.”

Admittedly, with so many gunfire alerts coming in on New Year’s Eve, it can be tough for police to actually go out and find or arrest the people engaging in celebratory gunfire. But Clark said the audio from ShotSpotter could help law enforcement zero in on some of the worst offenders, like people firing off rapid successions of shots from semi-automatic weapons.

Clark added, of course, that these holiday periods are not easy for police either. In his experience, police departments can cut down on celebratory gunfire by understanding where it’s happening, and going out and having conversations with residents in those areas.

“You can’t arrest your way out of any of this stuff,” Clark said. “The real biggest lever is changing the norms of the people involved in the behavior.”

Jeri Bonavia, executive director of Milwaukee-based nonprofit Wisconsin Anti-Violence Effort, strikes a similar note. The best solution to celebratory gunfire, she said, is “making this socially unacceptable and having people talk to their friends and family about, ‘this is not a good way to celebrate.'”

Bonavia also decried the lack of data available about injuries and deaths from celebratory gunfire, and from other types of gun incidents that aren’t homicide or suicide.

On a nationwide level, no one really knows whether celebratory gunfire is going up or down. The last major scientific study on the issue seems to have been conducted in 2003, conducted by the Centers for Disease Control and Prevention, and that focused on Puerto Rico. (CDC and U.S. Department of Justice employees contacted for this report could not point to a more recent major study of celebratory gunfire.)

More recent federal data suggests that this kind of shooting rarely kills people United States, but these data are limited too. A CDC report analyzing accidental gun deaths in 2013 found no deaths resulting from celebratory gunfire, but it covered only 17 states (including Wisconsin) and one calendar year. Additionally, the CDC classifies such deaths under a category called “Unintentional Firearm Death,” which includes several other types of incidents, including hunting accidents and people accidentally getting shot while cleaning guns. This leaves mostly anecdotal evidence of individual deaths and indicators in admittedly noisy data sets gathered by ShotSpotter.

While most known gun deaths are suicides, the number of accidental gun deaths that occur in the United States might be greater than previously thought. A recent investigation by the Associated Press and USA Today found the CDC was undercounting the number of children who die in gun accidents.

In Bonavia’s opinion, the official accounting of celebratory gunfire deaths and injuries may well be similarly lacking.

“I’m aware of the incidents and tragedies when they occur, just because we follow the news about that, but as far as hard facts on it, there just really isn’t good data when it comes to gun violence across our country,” she said

“It’s very frustrating to try and combat gun violence in all its many facets when we’re kind of kept blind about a lot of different things.”