On the morning of the last day of 2014, the District’s homicide count stood at 106. By nightfall, it was 105.

Anthony Cornell Jones, who was shot in Southeast in September, was taken off the list.

For three months, the 39-year-old stood as last year’s 81st homicide victim. But authorities ultimately concluded that the person who shot him was acting in self-defense, making the shooting legally justified. Jones, along with two others during the course of 2014, were added, then removed, from the grim tally that chronicles the pulse of killings in the District.

The case shows how a jurisdiction’s homicide count — commonly seen as a measure of violence in cities and counties across the country — is not a precise measure of the number of killings in a calendar year. It excludes some violence, and it sometimes includes violence from previous years.

A city’s homicide count can be used to judge the success of a mayor or police chief. It can stain a city as lost to despair — the way Detroit is now, the way Washington used to be. The FBI publishes a comprehensive list of homicides every year and immediately warns not to compare cities, followed almost immediately with lists comparing cities. The media and politicians are equally guilty of using that one statistic to declare their city safe or dangerous.

The homicide number is widely believed to be the most reliable crime statistic, even though it represents a minuscule percentage of any big city’s overall crime. In the District, fewer than 2 percent of last year’s violent crimes were criminal homicides. New York recorded 321 killings in 2014, out of 104,000 felonies, less than one-half of 1 percent.

A homicide is technically a killing of one person by another, which include accidents and those justified by self-defense. But the police generally count only the deaths being investigated at that moment as crimes. Thus, initially, three fatal shootings in the District went on the list, under the belief the acts were criminal, and were removed after investigators proved otherwise. A robber killed by a man who wrestled away a shotgun from him and fired never made the list, as it seemed justified from the start. Same for the killing of a 13-year-old boy whom police said was accidentally shot by his best friend. All involved one person taking the life of another.

All were ruled homicides. Police said none was a crime.

Fatal shootings by police — there were at least two in the District last year — are usually kept off and added later if the shooting comes back as unjustified, though that is rare.

The list also doesn’t always reflect the number of killings within a jurisdiction, but rather the number of killings investigated by the police agency that keeps the list.

For instance, Prince George’s County police recorded 54 homicides in 2014. That does not include Laurel and Greenbelt, which have their own departments, and each one investigated a single slaying last year. Similarly in Virginia, Loudoun County’s tally of three did not count a killing in Leesburg.

Another quagmire is how the dead show up on the list in the first place. It’s often not when a person is shot, and not even when the person dies. It is when the medical examiner rules the person’s death a homicide. It’s then up to authorities to decide whether to investigate that homicide as a crime.

That’s how month-old Hakeem Brown and hours-old Baby John Doe Alvarado died in 2013 and ended up as the District’s second and 10th homicides of 2014. It’s how 53-year-old Carmen Payne could be strangled in October 2013 and become homicide No. 34 in April 2014. And it’s how James S. Brady, shot in the 1981 assassination attempt on President Ronald Reagan, became the District’s 70th homicide of 2014. A medical examiner ruled that the decades-old bullet caused his death.

Cases are recorded this way so numbers from previous years don’t keep changing. But the list is used to measure a city’s safety and well-being, and violence that occurred in the previous year, or the previous generation, perhaps speaks better to conditions when the guns were fired than when the victim succumbed to the wounds.

Similar arguments are taking place as cities grapple with how to count murders. Those killed in the Sept. 11, 2001, attacks in New York don’t show up on the police department’s homicide list. The FBI created a separate category. Boston didn’t include the three people killed in the 2013 marathon bombings because it was a terrorist attack, not a typical street crime. The District included the 12 killed at the Navy Yard in 2013.

In 2013, Baltimore police reported 235 homicides to the public but 233 to the FBI. The deaths of two young children — one left in a hot car, the other asphyxiated when her mother rolled over on her in bed — fit Baltimore’s definition but not the FBI’s, according to a department spokesman. Suspects in both cases were charged with manslaughter.

In the District, Tonya Reaves, 53, was killed in September while riding a bicycle in Northwest. Police charged the driver with second-degree murder, but Reaves is not among the 105 homicide victims of 2014.

The explanation from D.C. police: Victims in traffic fatality cases are never listed on the homicide chart.