If you watched film and television from an alien spacecraft, you'd wonder how the same government that created a CDC which needs 6 weeks to tell the public that lettuce has E. coli on it can also create "Mission Impossible" plots and shows where FBI agents win every time.

The first part is real, the last is certainly fiction. A new study finds that not only is a government database wrong, it is wrong to an extent that is worrisome when we want people to have more trust in law enforcement agencies. 

The analysis of the FBI’s National Incident-Based Reporting System for crime reporting in Massachusetts compared 480 randomly chosen incidents with what local authorities wrote and there were discrepancies 16 percent of the time. There were arrests in over 33 percent of the cases and a summons to appear in court in over 18 percent, but the concern was the errors they found.


Image: iStock.com/RyanJLane

Over 13 percent of the cases entered into NIBRS were false negative, they did not include an arrest or summons, and the authors linked that to delayed arrests. Almost 41 percent of sexual assault cases were false negatives, 32 percent of intimidation cases, over 19 percent of aggravated assaults, and almost 16 percent of simple assaults. The reason may be they were not updated due to competing responsibilities. Law enforcement involves a lot of paperwork and new crimes happen, causing documents to not be updated, or there may not be awareness an arrest happened later.

 More than 50 percent of the summonses issued in the random sample were not recorded in NIBRS and the authors believe that may be easily fixable. To make the paperwork process move faster, NIBRS uses one field for arrests and summonses, which may cause underreporting of summonses. Cops differentiate arrests from summoning someone to appear in court even if a database uses the same field.

Whatever the reason, a national database used by more and more police officers relies on accuracy. The authors rightly don't blame law enforcement, the false positives were in the statistical noise range so simple human error that can occur, but if the FBI wants more usage than the 29 percent of law enforcement that uses it now, they can convene with existing users and figure out how to improve accuracy without sacrificing time on the job for the people on the job.