The FBI released its Uniform Crime Report (UCR) figures for 2017 last month and what little coverage the release received discussed small drops in both murder and violent crime nationally in 2017 relative to 2016.  The FBI figures showed a 0.7 percent decline in murder from 2016 to 2017 and a 0.2 percent decline in all violent crimes (murder, rape, robbery and assault). The Justice Department’s press release bragged that “after two consecutive, historic increases in violent crime, in the first year of the Trump Administration the nationwide violent crime rate began to decline.”

If one were asked “were more people murdered in the United States in 2016 or 2017?” the logical answer would seem to be 2016. The correct answer, however, is we don’t know.

The FBI reports that there were 17,413 people murdered in 2016 and 17,284 in 2017, so a small decline in murders seems like an open and shut case. But the FBI provides a footnote next to 2016’s entry noting that “The crime figures have been adjusted.”

This is a usual procedure that occurs every year with the previous year’s figures being adjusted up or down. They have to do this because there are over 18,000 law enforcement agencies that report data of some sort to the FBI every year under the UCR program. Not all of those agencies report a full year’s worth of data and sometimes agencies don’t report anything at all.

The way data is collected requires the FBI to estimate crime counts for some jurisdictions. They spell out the methodology here, but the procedure is summed up as:

These tables contain statistics for the entire United States. Because not all law enforcement agencies provide data for complete reporting periods, the FBI includes estimated crime numbers in these presentations. The FBI computes estimates for participating agencies that do not provide 12 months of complete data. For agencies supplying 3 to 11 months of data, the national UCR Program estimates for the missing data by following a standard estimation procedure using the data provided by the agency.

You can clearly see the estimations in the state level data table. Very little is estimated for most states, but Mississippi – for example – reports just 60 percent of the violent crime the FBI estimates took place there in any given year. As a result the FBI estimates that 88 murders occurred in Mississippi in 2017 for which there is no actual data at this time.

Mississippi is the nation’s most egregious offender of non-reporting, but the effect adds up across 18,000 agencies. If some agencies report data after the previous year’s UCR has been published then those numbers are revised when the new year’s report comes out. And if the previous and current year’s figures are close then it can make the difference in the narrative of whether there was an increase or decrease in crime.

The below table shows how the initial and final UCR counts for murder has changed every year since 2000. The initial figure for 2016 was revised upwards by 163 murders which is higher than the average of +88 since 2000, but not unusually so. Murder was only revised downward in two years since 2000, so it would be logical to assume there’s a decent chance that next year’s revision will change last year from a small murder decrease to a small murder increase.

Capture

This is even more true when looking at changes in violent crime where the average revision (+3,255) is larger than the initially reported change in crime between 2016 and 2017.

Capture2

What we really need with these crime data releases are margins of error that highlight how the data collection methodology and why the figures should be treated with a grain of salt. With decreasing crime potentially being used as a political football it’s important to know how uncertain we should be in the precision of those figures.