With the FBI releasing the 2016 Law Enforcement Officers Killed and Assaulted data this month, reports on police deaths have once again found their way into headlines. Such reports offer a great illustration of how reporting the same facts in different ways can dramatically change how big a problem seems to be. Also interesting is the information that is reported depending on the media outlet. NPR, for example, emphasizes the number of police killed with firearms specifically, rather than the general category of "feloniously." Fox News will emphasize the "War on Cops" narrative while reporting total officer deaths rather than violent officer deaths, presumably because the former is a bigger number. This is misleading, however, since it includes deaths due to things such as heart attacks, accidents, and extreme weather.

Something to watch out for is when percentage changes are reported while excluding some of the raw numbers on which they are based. In this example, CBS News reports that "the number of police killed in the line of duty rose sharply in 2016," with a "56 percent increase in shooting deaths over the previous year." They do report the raw number of police shot to death in 2016 – 64 – but not the previous year's, which was 41. This is an increase of 23 deaths in a country with over 900,000 sworn law enforcement personnel according to the National Law Enforcement Officers Memorial Fund. Reporting that a police officer's chance of being killed went up from 0.00455% to 0.00711% (that is, 41/900,000 to 64/900,000), if it doesn't confuse the reader, will likely make the violent shooting death of a police officer seem like a relatively rare event.

Additional context is provided by looking at trends in officer deaths from gunfire over the last decade (data from Officer Down Memorial Page):

2007: 67
2008: 41
2009: 48
2010: 59
2011: 68
2012: 48
2013: 32
2014: 48
2015: 41
2016: 63

Looking at these numbers, it's difficult to consider 2016 an outlier or any indication of a "war on cops." Further context is provided by looking at longer trends: