Statistical analysis certainly has its place in law enforcement. This is, after all, the information age. So the operation of the justice system should be driven by anecdotes, but by hard data.

There are, however, several potential problems with this. For one thing, data can often be erroneous, due to numerous factors. In addition, no matter how good the data is, it cannot become an excuse for violating people's privacy rights.

Consider, for example, the question of prescription drug charges in New York City.

By analyzing data on Medicaid reimbursements, research analysts believe they can track suspicious distribution patterns involving prescription drugs and specific doctors and pharmacies.

New York City intends to use this data to target enforcement resources toward what appear to be the most frequent offenders. In other words, doctors who are writing what appear to be too many prescriptions, and pharmacies that appear to fulfill too many, may find themselves under suspicion because of broad statistical correlations.

These correlations do not necessarily mean a crime was committed. They are not tangible physical evidence. Instead, they reflect a form of scrutiny-by-spreadsheet that was unheard of not that long ago.

The city thinks it's on to something with this data mining. It is also being used in other types of investigations, such as real estate data analyzed by the city's Financial Crime Task Force. Seeking to target fraud, the task force is looking at data elements associated with suspicious sales, such as frequent changes of title or erratic pricing.

But again, there is a big difference between broad statistical trends and evidence of a specific crime.

Source: "Beyond Performance Measurement," Governing, 12-28-11