Applying location technology to social issues

In the past, this blog has discussed one aspect of location-based technologies - namely, the ability for businesses to offer services to people who are in close geographical proximity to them. (One example often cited is Foursquare's ability to show special offers for participating locations if you happen to be close to those locations.)

But location - specifically, the ability to assign anything and everything in sight a geographical coordinate - can be used for more than food discount offers, as Physorg notes. (H/T All Points Blog.)

Physorg published a post that describes how professors William Alex Pridemore Tony Grubesic analyzed two layers of data - the locations of licensed alcohol outlets in Cincinnati, Ohio, and the locations of crimes in the same city - and detected a correlation between the two.

"A higher density of alcohol sales outlets in an area means closer proximity and easier availability to an intoxicating substance for residents," Pridemore said. "Perhaps just as importantly, alcohol outlets provide a greater number of potentially deviant places. Convenience stores licensed to sell alcohol may be especially troublesome in this regard, as they often serve not only as sources of alcohol but also as local gathering places with little formal social control."

They used the data to extrapolate some predictions:

Using different suites of spatial regression models, the researchers found that adding one off-premise alcohol sales site per square mile would create 2.3 more simple assaults and 0.6 more aggravated assaults per square mile. Increases in violence associated with restaurants and bars were smaller but still statistically significant, with 1.15 more simple assaults created when adding one restaurant per square mile, and 1.35 more simple assaults per square mile by adding one bar.

But does the data justify the conclusion?

"We could expect a reduction of about one-quarter in simple assaults and nearly one-third in aggravated assaults in our sample of Cincinnati block groups were alcohol outlets removed entirely," Grubesic noted. "These represent substantial reductions and clearly reveal the impact of alcohol outlet density on assault density in our sample."...

"Alcohol outlet density, on the other hand, is much more amenable to policy changes," Grubesic pointed out. "Unlike other negative neighborhood characteristics that often seem intractable, regulating the density of outlets, and to some extent their management, can be readily addressed with a mixture of policies by liquor licensing boards, the police and government agencies that regulate land use."

Perhaps, but if someone concludes from the data that the best policy change would be to outlaw alcohol in Cincinnati entirely, the economic laws of supply and demand need to be considered. Would such an action result in an increase in traffic-related deaths on the roads connecting Cincinnati with Kentucky? Would such a...um...prohibition in alcohol sales result in alternative unregulated alcohol suppliers entering the Cincinnati market, either as a unified syndicate or as competing entities (or gangs) of liquor providers.

That's the danger with analysis - people may analyze one set of data points and reach a conclusion, but ignore how that conclusion could itself affect the data in which they are interested.

Although it would be interesting to compare crime statistics before and after the beginning of Prohibition, and before and after the end of Prohibition, to see if the popular perception of Prohibition as a field day for illegal crime is in fact true. A discussion of crime and Prohibition can be found here.