# An analysis of statewise uniform population density (according to Craigslist)

So during a study break this afternoon I took a casual little jaunt over to Craigslist to see what those in Vancouver were most recently ranting about. For whatever reason I decided to scroll down the page rather than click the little “Canada” link at the top, and I noticed that a few of the US states had only one little sublisting beneath them (Wyoming, for example).

This got me to wondering: does the way Craigslist create its state listings reflect the uniformity of those states’ population distributions? In other words, for example, if a state only has cities listed as opposed to large areas like “western Wisconsin,” does that reflect the fact that the state has its population “clustered” into small areas and not uniformly distributed throughout the state?

Now of course you know me and you know how I do things, so this wasn’t going to be some simple analysis in which I would merely count up the listings or something and do a rank ordered map thing.
No, no.
It has to be more complicated than that.

So without further ado, here’s what I did:
First I decided to take a look at the sublistings and rank them in order of size. It turns out that there are listings that range from as large as the entire state itself down to just regular cities. Here’s what I’ve got:

- State
- Area (e.g., “northwest CT,” “heartland Florida”)
- County
- Pair of cities (e.g., “Moscow/Pullman”)
- City

Theory: the more uniformly a state’s population is “spread out” in the state, the more likely there will be larger area listings for that state (e.g., just the state listed, or just areas and counties). The less uniformly a state’s population, the more likely that there will be a lot of smaller area listings (like a lot of cities and pairs of cities) rather than large area listings.

Make sense?

Of course, there is the overall population to consider—for example, Wyoming just has “Wyoming” listed ‘cause nobody lives there. But there are slightly more populous states that also have just the state listed. Similarly, there are also slightly more populous states that just have a few cities listed, thus indicating that the small populations of these states are clustered into areas and not uniformly spread out.

So here’s how I quantified “uniformness”—I gave every city listed under every state a value of 1. I then gave ever pair of cities, county, area, and state listed values of .8, .6, .4, and .2, respectively. I then, for every state, summed these numbers and divided by the number of listings. This way, the more uniformly a state’s population is spread out, the closer this final number will be to zero (or .2, rather, because that’s the value I assigned to “state”), and the more clustered the population, the closer this number will be to 1.

Here’s a map with colors! I’m into gradients lately.

If I could find some reference to compare this to I would, but I can’t find one, sorry.

Woo!