How accurate are the "most dangerous city or neighborhood" rankings?
Last year, a company called Neighborhood Scout put out a report that called an area of St. Louis near 14th Street and the Rev. Dr. Martin Luther King Drive the “14th most dangerous” neighborhood in the country.
In fact, it’s not even a “neighborhood.” As someone who has covered crime for several years at the St. Louis Post-Dispatch, I can confirm that it does not appear as such on any of the crime data reports for police districts and city neighborhoods published regularly by the city. The area spotlighted by Neighborhood Scout would be better described as a census tract.
We ignored the report because we had better stories to write. The 2009 rankings, however published first on AOL’s personal finance site WalletPop got wide attention across the country. That may be one reason the site seems to be planning to make them an annual event. This month, WalletPop published the 2010 “figures” from Neighborhood Scout. St. Louis wasn’t even on the list.
The WalletPop story claims that the neighborhood crime rankings were compiled from “exclusive data” developed by Schiller’s team “based on FBI data from all 17,000 local law enforcement agencies.”
The story goes on to say, tellingly, that the rankings represent “the top 25 most dangerous neighborhoods with the highest predicted rates of violent crime in America".
Neighborhood Scout isn’t the only self-appointed expert in crime rankings. Another company, for instance, annually uses repackaged data from the FBI’s annual crime report, Crime in the United States, to come up with “Most Dangerous City rankings.”
Such rankings are “bad science,” says Richard Rosenfeld, a criminologist at the University of Missouri-St. Louis and current president of the American Society of Criminology.
Rosenfeld, an expert on crime statistics, says the methodology for establishing rankings should be transparent and the fact that companies won’t explain how they arrive at their conclusions should alert anyone who is tempted to take them seriously. Rosenfeld added that Neighborhood Scout researchers should have asked cities for real neighborhood crime data to test whether their predictive models were accurate but said there is no evidence they did so. If they had, “I think they’d be in much less trouble with police departments.”
Link:http://thecrimereport.org/2010/10/17/mean-streets/#more-48134