Dependable Erection

Monday, December 15, 2008

Numbers bleg

I'm trying to get a sense of how Durham's population is distributed across the different police districts. Census bureau data doesn't easily translate to Durham's 5 police districts, and the city doesn't seem to have that data readily available, either. I'm not looking for demographic breakdowns, either. Just a sense of how many people live in district 1, district 2, etc.

Anyone point me in the right direction?



  • Could you determine it from the police district maps?

    By Blogger Korps, at 10:15 PM  

  • i can't. maybe somebody else could. i've downloaded the "Cartographic Boundary" files from the Census Bureau, but they're in something called "ARC/INFO export" format, which i can't read.

    here's the deal. i have some numbers from Durham 911 regarding service calls of a certain nature in the various police districts. Calls from 1 district are substantially higher than from the other 4, and i'm just trying to get a handle on whether that could be attributed to a difference in population, or if it's worthwhile looking for other reasons.

    By Blogger Barry, at 10:41 PM  

  • Give me a few hours. I'll run them for you....

    By Blogger Michael Bacon, at 2:30 PM  

  • Okay, less than a few hours. Here's the rough, rough numbers, based on Census 2000:

    D1: 30,303
    D2: 51,580
    D3: 59,667
    D4: 43,392
    D5: 3,839
    DSO: 275

    I assume DSO means "Durham Sheriff 's Office," meaning places within the city where the Sheriff's office covers. I think this is limited to schools.

    Now, a word on the methodology. This is very crude, as I did it the fast and loose way. I took a layer with the Durham police beats overlayed on a layer of census block centroids. It so happens the two layers were in different projections, which ArcMap will generally correct for, but which if you want true precision you nail it down. The centroids are going to treat it as if all population on those census blocks (which are basically city blocks in urban areas, but a good bit larger the more rural you get) existed at the point of the centroid. Some spatial error is bound to occur. Also, naturally, these are Census 2000 numbers, which means they're 9 years old.

    Finally, there's the issue I tried to address in my thesis work, that what you're looking at is residential population, and that people don't always make 911 calls from or about where they live. If you normalize crime rates by population, for example, downtown looks like it's a freakin' war zone. This isn't because crime occurs that much higher there -- it's because there's hardly any residential population but a lot of people are down there anyway, so a lot of crimes occur there even though most of the victims don't live there. To address that, you have to use something like the LandScan ambient population estimates, which have all kinds of copyright restrictions on them. More on that in my much delayed journal article I keep meaning to write.

    Anyway, the numbers above should be a decent starting estimate. Just treat them with a grain of salt.

    By Blogger Michael Bacon, at 2:56 PM  

  • thank you, Michael!

    (check your email shortly)

    By Blogger Barry, at 2:59 PM  

Post a Comment

<< Home