"Criminologists say an annual ranking of cities by crime rates is guilty of statistical offenses."

WSJ Numbers Guy Carl Bialik has this story and this blog post on criticisms of CQ Press's City Crime Rankings.

For just one example, suppose two metropolitan areas have similar crime rates and profiles, with the core city having a higher crime rate than the suburban collar (a very common profile).  Suppose the city limits are tightly drawn in one area to include only downtown and a few old neighborhoods nearby and widely drawn in the other to include the suburban ring.  Then the statistics will show the divided area to have one high-crime jurisdiction and another low-crime jurisdiction, while the consolidated area has one medium-crime jurisdiction.  But criminals don't heed city limits signs, so the risk of being victimized is the same in both areas.

Any time someone publishes a ranking list, whether it be crime, schools, or whatever, it seems to get more attention than it deserves.

Simplistic statistics go beyond annoying into dangerous when people use them to support arguments over policy.  A common but obvious fallacy is to take the statistic that states that have abolished the death penalty have lower murder rates, on average, than those that have not and use that statistic to claim it disproves deterrence.  Even a retired Supreme Court justice, who really should know better, fell into this trap recently.