One amazing thing about data is its variation. Not within a data set, but between data sets. Data can be meticulously collected on a single pig, or on the whole world. It can be collected long ago, rarely, often, or continuously, as with “big data”. It can be collected with lots of variables that take a long tedious time to enter (with more chance for error as the survey goes on), or just a “yes/no” response. It can be collected at a fine scale for a whole nation, at a finer scale for just one town or “study area”, or at a very broad scale for a very broad area. The researchers that present that data can present it as a global or regional average, and often do. It is hard to make sense of thousands of data points, and easy to make sense on one or a dozen. We can infer different things based on those scale decisions.
Sometimes a conclusion that is obvious at a national scale holds true all the way down to every household, like “Americans like to have plumbing in their homes”. Other times a conclusion that is obvious at the national scale falls apart as we get more local. As we get more local, we might see some things we would never see at the national scale. Such as immense use of transit, walking and biking for commutes.
A few months ago, I downloaded the “journey to work” results from the Census for the year 2000. The 2000 census offered the data for the whole nation, at the blockgroup level (~1,500 residents), and indicated the numbers driving, taking transit, biking and walking. This is better data that I have been able to find from the 2010 ACS. If you know where I can get data this good form the 2010 census, please let me know. The data was based on a sample, so there were some errors in these numbers, but not a large one. The data was just on commutes to work, which leaves out a lot of other reasons people travel. The journey to work is important, however, in that most people do it when their ways are the most congested. The whole point of traffic engineering in the 20th century has been the elimination of congestion, so I see why the Census would measure this in particular.
I found that 3.5 million Americans in 37 states live where over 10% of them can commute by bike to their work. This naturally means that their jobs are close enough and safe enough to be biked to. A lot of these people don’t live in big cities so much as college towns. 40 thousand lucky souls in 6 states live where over 50% of the commutes are by bike.
I wanted to present some maps of what that looked like, and realized it was kind of boring just showing biking. Especially since the 50% threshold was so rare. So I made some frankenstein maps of y2k census data with 2010 transportation infrastructure. These show the 10% and 50% thresholds for walking and transit as well at the same scale, for several cities. I’m sure there are some places where more than one of the modes had a 10% threshold, but this is just quick and dirty presentation. I can mess with the legend later.
I won’t label them yet, as the transportation lines should make it obvious which cities they are. Click to embiggen.
The problem with a lot of local studies is that they are only local, not national. How do I know a study of biking in Dallas holds for the entire nation. It is well and good to say that a bike trail did this miraculous thing, but I can’t trust that conclusion until I know that it has done that all over the country in every situation. Or I at least have a general idea what the probability or magnitude of the benefit is. That was a motivating force behind my book, to look at the finest scale data I could on a national level. To go beyond the averages and case studies to see what was happening both locally and nationally.