It's not easy to figure out how many people have been killed by the police in the United States. The FBI compiles a wealth of information on crime and law enforcement each year as part of its Uniform Crime Reports system. But it only collects two points of data on people killed by law enforcement: The number of people killed by police in justifiable action, and the weapon used in the homicides. (Unjustified homicides are counted, too, of course: as crimes.)

What's more, as USA Today notes in its look at the data on Friday, only 750 of the nation's 17,000-plus law enforcement agencies contribute data to the FBI's justifiable homicide database. That's a huge caveat, particularly given that we can't ascertain how representative the pool of agencies might be. (We reached out the FBI for more information and didn't hear back.)

Regardless, we can learn somethings from the data, when we consider the trends on display and when contrasted with other data.

Here's the annual data on justifiable homicides from 1991 to 2012, the last year for which full data is available. Most of the homicides are committed with handguns -- understandably given that this is the most common firearm issued to officers.

But, particularly given the ongoing debate over the militarization of local police forces as a result of what happened in Fegurson, Missouri this week, there's an interesting trend that isn't immediately obvious in the graph above. If you look at the type of weapon used as a percentage of all of the killings, a few trends emerge. Shotguns are less frequently used in justifiable homicides, and rifle use has increased dramatically. So has the category of "firearms, type not stated."

You'll notice in the first chart that justifiable homicides by police officers spiked in the early 1990s, settling down as crime rates dropped in the late 1990s, and then slightly increasing again.

If you compare the number of justifiable killings to the violent crime rate in the United States, there's an interesting result. As the number of violent crimes has steadily dropped, the ratio of justified homicides to violent crimes has increased. In 1991 there were 1.92 justified homicides for every 10,000 violent crimes. By 2001, it was 2.63. By 2011, it was 3.35.

Of course, the number of police officers has also increased. (The data below excludes civilian police employees.) Since the early 1990s, the number of police officers in the United States increased substantially, up nearly 15 percent between 1995 and 2005. (1995 is the first year for which data was available from the FBI.)

It's natural to assume that this would in itself lead to more justifiable homicides -- more armed police on the streets. Regardless, there's been a slight upward trend in the number of justifiable homicides as a function of the number of police. But given the volatility of the data, it's hard to draw too strong a conclusion from this.

One point of data that the FBI collects thoroughly is the number of law enforcement officers killed in the line of duty. It evaluates those deaths on 47 different metrics, providing a very good overview of these crimes. The good news here is that the number of police killed in the line of duty has generally fallen over the past twenty years.

The FBI data is only through 2012, but as the Post's Radley Balko noted when he was still at Huffington Post, 2013 was likely one of the safest years for law enforcement in decades.

Again, it's hard to evaluate the data at hand with much certainty given how poorly gathered the data on justified police killings is. This is certainly not an accident; law enforcement agencies have an obvious rationale for holding those data as close to their chests as possible. What we can tell is this. Among the agencies reporting to the FBI, they are more frequently using rifles in justifiable homicide incidents, and that, as a function of the national violent crime rate (and possibly the number of police), those killings are up, while their jobs, happily, are getting safer.