Crime as altitude – feedback

Contrarian friend Gus Reed doesn’t think altitude maps add much to our understanding of complex social issues:

These graphs don’t meet the minimum standard for clarity. Your pal Edward Tufte would be appalled. What is the scale of the z-dimension? Are we to suppose that the high peak for narcotics is on the same scale as the high peak for prostitution? Absolute numbers? Percentages? Logarithmic? I’m suspicious that McCune is mixing his units.

And I don’t like the fundamental assumption that it’s OK to smooth this kind of data. Consider the three hills of prostitution down by Fisherman’s Wharf. They are almost certainly hotels, yet they have volume and give the impression that prostitution is a spreading neighborhood crime wave. Better to use some more accurate and consistent visualization like dots (as on the SFPD KML files from which this is taken).

Furthermore, as some commenter about halfway down McCune’s post says, the unit isn’t crime, but incident reports. It would be instructive to map donut shops, say, as a reality check. Or distance from police station. Are they maps of crime, or just maps of cops?

This kind of simplistic use of numbers is fraught with pitfalls. It makes a complicated phenomenon into a simple problem of location. It usually turns out that peaks are located over poor neighborhoods, and it encourages a two-dimensional approach to crime. It’s easily fudgeable. Give me some numbers — “the use of hyperbole on Boulardrie Island” — and I can make it look very bad for you. At the very least, we are entitled to know what the z-scale is.

Chris Macormick, more a fan of visual thinking than Gus, makes the same point about underlying poverty, and offers this poverty map of the Bay area, by Catherine Mulbrandon of, as corroboration:

Zoom in on downtown SanFrancisco, and the correlation is evident, albeit at low resolution:

Chris writes:

Altitude maps of crime are interesting as a way to see patterns easier than perhaps numerical rates do. However they disguise other patterns in the same way, eg. class, policing.

Disgaggregated crime statistics show spatial variations in crime, whether they’re presented visually or numerically. But when comparing income distribution in SF to crime distribution some loose associations are available:

  • Prostitution is tightly localized in several poor areas
  • Narcotics is more widespread but also in poor areas
  • Theft is more widely distributed and extends in rich areas
  • olicing is probably more prevalent in poor areas, especially for the victimless crimes of sex and drugs, and
  • he crimes depicted are by definition lower class crimes.
  • I would like to see a map of tax evasion, price-fixing, insider-trading, embezzlement, tax fraud, environmental pollution and other cozy crimes of the powerful… Just saying.

    Visual data: crime as altitude

    Doug McCune uses San Francisco Police Dept. crime reports to map crime in that city as altitude. Narcotics:


    Various criminal activity:

    What would an altitude map of Halifax crime look like?

    Or better still, a North American altitude map of multiple sclerosis, a disease that concentrates in northern latitudes (with Nova Scotia a likely mountain range)? Any data-and-graphics-savvy medical researchers out there want to take this on?

    Hat tip: Flowing Data.