FBI Crime Statistics: Reported Total vs. Estimated Total

Long ago I worked in a state government’s econometrics group. One of the senior economists would say, the only data that is any good is birth numbers and death numbers… and they are always wrong.

In that spirit, it turns out that some part (but which part??) of FBI crime statistics is just made up, using an unconvincing and obsolete formula, that barely made sense when it was instituted 60 years ago. I have to say this is not very encouraging.


[Summary] Table: “FBI Crime Statistics: Percentage Difference Between Reported Total vs. Estimated Total: Ten States With the Most Missing Data, 2017”, in Wall Street Journal, October 20, 2018. p. A2.

The FBI uses estimates to fill in missing crime data for states, sometimes resulting in figures that stray substantially from reported numbers. 18000 police agencies are supposed to submit data to the FBI. Using a method developed in the 1960s, jurisdictions that did not report all 12 months are assumed to be like a nearby community of the same size in the same state that did report: In part this crude method was adopted because accessing the data on the 9-track computer tapes, and the national data required 7-8 tapes, was very difficult. As a result of poorly documented estimation of data, year-to-year comparisons are fgenerally meaningless.

Mississippi: 68% (violent crimes increased form 5084 reported to 8526 adjusted; including 157 reported murders to 245 adjusted murders).

West Virginia: 13%.
Indiana: 9.9%.
South Dakota: 6.7%.
Ohio: 4.3%.
New Hampshire: 3.5%.
Oregon: 3.3%.
Louisiana: 3.0%.
Arkansas: 2.9%.
North Carolina: 2.8%.

Source: FBI.