Density Happens to Energy Use for Transportation
Monday, November 3rd, 2008Today I presented Density Happens at Medialab-Prado, and in the presentation I gave an example of the kind of insights I intend to dig out with this project. Here it is.
Rooting around the Web (you know, holding it upside down and shaking it to see what falls down) I found a set of classic examples of how to turn data around till it looks right. More specifically I was looking for data on how urban density affects gas usage.
First I found the following diagram in the Wikipedia article for “urban density“. It’s a classic example of “let’s plot points and see what happens”, and indeed it vindicates our intuition that density and petrol use are inversely correlated (click for bigger size):
Petrol use is inversely proportional to urban density
The comments on the talk page for this image are quite damming. The data for this diagram must be wrong; the mind boggles to find out that New York and Los Angeles have the same urban density, or that London is denser than Paris. Somehow the measurements are taking into account large empty areas within the administrative boundaries of cities.
However, dodgy data is not all that’s wrong about this diagram. It also misses an opportunity to explore the evidence, extracting qualitative features (cultural, even) from quantitative data. The discussion page offers this alternative diagram (again, you can click for bigger size):
Inverting density reveals different clusters
Plotting petrol consumption against the inverse of density (sq m/person instead of people/sq km) the data reveals that density is not the only factor in gas consumption. Australian cities are about as dense as US cities, but gas consumption in urban Australia is about half what it is in the US.
Other factors than density must be at play, and my hypothesis is that they could be economic, cultural, or both. This may be due to the lower price of gasoline in the US, and the fact that Australians tend to buy, on average, cars with a lower gas consumption. Petrol consumption in Australia remains almost flat over a 2x increase in inverse density, which suggests that maybe Australians have shorter commutes, or tend to roam a local area regardless of the size of their city.
In any case, this is an example of how pivoting the data reveals hitherto hidden variables.