Calculating demographics using driving distance

In the Using polygon overlays for proportional census estimates recipe in Chapter 2, Structures that Work, we employed a simple buffer around a trail alignment in conjunction with the census data to get estimates of what the demographics were of the people within walking distance of the trail, estimated as a distance of 1 mile. The problem with this approach, of course, is that it assumes that it is an "as the crow flies" estimate. In reality, rivers, large roads, and roadless stretches serve as real barriers to people's movement through space. Using pgRouting's pgr_drivingDistance function, we can realistically simulate people's movement on routable networks and get better estimates. For our use case, we'll keep the simulation a bit simpler than a trail alignment—we'll consider the demographics of a park facility, say, the Cleveland Metroparks Zoo, and potential bike users within 4 miles of it, which we estimate as approximately a 15 minute...