Where the Smart People Live: Percent of Population with Bachelor and Graduate Degrees per Square Mile
It’s an interesting study, however, it immediately struck me that using ‘degrees per square mile’ as a basis for the analysis was flawed in that it biased the data towards those cities with high population densities. Both SF and NY have high population densities when compared to LA or SJ, which I suspected would have similar populations with a college degree. In returning to the original post, I was pleased to see that this was recognized in the commentary:
Now, you can look at these graphs and say, “this merely reflects the overall population density in these cities” and you would be on to something…
Here is the original analysis (excel) performed by EO, which was used to determine the relative density of people with a bachelor or graduate degree in major metropolitan cities. I commend EO for performing this analysis, however I would suggest that we take a step back and reevaluate what the goal of this analysis is. As stated in the premise of EO:
The theory that there is economic value to having smart people together rests on the assumption that smart people collaborate with each other. You could have a whole bunch of smart people in one place, but if they don’t interact with each other, what’s the value?
Based on that, where the smart people live and would likely collaborate due to close proximity should take into account the number of individuals that hold a college degree as a percentage of the population and the relative density of each city. Towards this end, it is useful to see that the percentage of the population that hold a college or graduate degree for each city displays far less variance than would be suggested by the original analysis. In order to determine this, I simply divided the number of degrees by the population of each city.
You can download my excel file here.
From this percent of the population that have a bachelor or graduate degree, we can move to an evaluation of the function of proximity, via city land area, by multiplying the percent given above by the inverse of the land area. This rewards those cities that have less land area and therefore, presumably, force those smart people together. Thus, by the following equation – degrees / population * 100% * (1 / land area) – we can obtain a relative determination of the percent of the population that holds a degree per square mile.
I believe this provides a better relative representation of the density of individuals with a higher education as a function of the population and urban density. This is exemplified by a comparison between Providence, RI and San Francisco, CA (the top two), where 18.66% of the population of Providence has a degree compared to 41.13% of San Francisco (2.20x greater than Providence). However, the land area of San Francisco is 2.52x greater than Providence (46.7 sq. mi. vs. 18.5 sq. mi., respectively). As a result, Providence has a greater percentage of its population with a degree per square mile than San Francisco (1.0089 %/sq. mi. vs. 0.8808 %/sq. mi., respectively).
As noted in the comments of EO’s post, this analysis primarily addresses major metropolitan cities and does not necessarily reflect major academic communities (e.g. Cambridge, MA)