Despite the increase, American streets remain safe relative to recent years. The rate of murder and manslaughter per 100,000 people (excluding negligent manslaughter) increased to 4.9 last year from 4.4 in 2014, which was the fewest of any year since at least 1961. In 1980, the rate exceeded 10 killings per 100,000.
Still, what to make of the sudden increase in homicides is not clear. Some criminologists say the data is evidence against the "Ferguson effect" -- a popular theory that suggests homicides have increased because police have become reluctant to interact with potential criminals on the street. According to this argument, cops fear becoming involved in a violent altercation that could result in protests such as those in Ferguson, Mo., Baltimore and Charlotte, N.C.
Police can deter potential criminals not just by being a watchful presence on patrol. They can stop people who appear to be involved in criminal activity, talking with them to gather information or to disperse people who are fighting. Police can also search civilians for firearms, knives or tools for breaking locks and windows.
These activities make committing all kinds of crimes more difficult, not just homicide. If the increase in homicides were due to hesitance on the part of police to stop civilians, some criminologists say they would expect an increase in other "street crimes," including burglary and robbery.
However, just 3.1 percent more cars were stolen last year, and the number of robberies increased just 1.4 percent. The number of burglaries declined 7.8 percent, according to the new federal data.
"People who are talking about Ferguson effects -- why do you find it in homicide and not burglary?" asked Franklin Zimring, a criminologist at the University of California, Berkeley in an interview earlier this year. "That's just not very good logic, and it's not consistent."
Other experts are not convinced by that argument. Alfred Blumstein, a criminologist at Carnegie Mellon University, is one of those who believes that changes in police officers' behavior might account for the increase in lethal violence last year. He argues that a subtle shift in police behavior might not noticeably affect minor street crimes, since so many of them are committed.
"Undoubtedly, there’s a skittishness on the part of police," he said. "They don’t want to be caught in some compromising situation and put their own careers, if not their liberty, at risk."
Research does suggest, however, that an increase in police presence reduces crime rates, especially for street crimes.
For instance, economists Jonathan Klick and Alexander Tabarrok found that when the Bush administration issued warnings about potential terrorism, police in Washington intensified their presence on the streets and that the additional manpower resulted in significantly fewer crimes.
Working with other researchers, Klick studied the district surrounding the University of Pennsylvania that is patrolled by both the Philadelphia Police Department and the school's privately funded police agency. Compared to otherwise similar blocks patrolled only by Philadelphia police just outside the district, there were significantly fewer robberies, burglaries, motor vehicle thefts and aggravated assaults on streets where there were additional patrols from the university. The researchers could not detect any difference in killings, though, possibly because homicide is so rare in general.
"The pure street crimes -- burglary, robbery and, to some extent, auto theft -- are better indications" of police activity, Zimring said.
At the same time, the federal data are notoriously unreliable. Each police and sheriff's department is responsible for reporting its own numbers, and smaller departments often lack the resources to do a thorough job of collecting the data. Elsewhere, differences in definitions of offenses -- along with political reasons for understating the extent of crime -- can skew the figures.
"My hunch is it's mostly a lot of noise," Blumstein said.