How Much Does Rural Living Predict Broadband Speed?

Posted by Mischa Fisher in Economics   
Tue 08 December 2015

I was recently brushing up on the current status of the telecommunications industry in the United States, and I became curious about how much a state's rural population predicted its overall average internet connection speed levels.

Pulling data on average speeds from (here), which sources Akamai's State of the Internet report, and data on the rural/urban population by state from Iowa State University's site (here), reveals the below plot:

While noticeable, the effect here is pretty minor overall. It would take a shift of about 20% of the state's population into an urban environment to shift average speeds up by a single Mbps.

How About Globally?

Curious about how well this predicts things generally, I did the same thing looking at global data. Wikipedia has a concise list of countries by internet connection speeds (here), and the World Bank maintains a time series list of urban/rural population (here).

Pulling those two datasets together, reveals the following:

Visually this looks similar, although the linear regression slope is a little steeper. It would take a shift of about 10% of an average country's population to shift the speed up by a Mbps.


At the end of the day, this exercise is likely too simple to shine much light on the phenomenon. Economies of scale in terms of population density would, all else equal, suggest that as density goes up, so would speed. But the 'average' speed by state hides the interesting variation that goes on within the county and city level that I imagine would more apparently show the swings based on not just population density, but, also regulatory factors such as the ease of installation and market entry, access to public conduit and utility maps, etc. etc. Looking at the data more locally would probably be more meaningful than trying to examine aggregate state or country data, because there is likely some regional smoothing that goes on in the 'average speed' metric that underestimates the effect.