Any project like this requires certain assumptions and is open to certain errors.

One assumption is that people’s driving has a clear epicenter around the county in which they obtained their license plate. For certain counties, this is very likely to be true; for others it may be harder to discern an epicenter. Denver is an example of a county in which whose residents will travel in all directions more or less equally, but her suburban counties are likely to see a strong “pull” toward the urban center.

Along the same lines, observational bias is likely to be a significant source of error. Bias toward not only urban areas in general but specifically toward gathering places is likely to occur simply because that is a good place to make observations.

Outliers will occur. Not everyone remains close to their home county. However, it is relatively rare, for example, to spot a Mesa county plate in Denver.

The possibility of “bad” data from other means is also present

In most cases, a large dataset will improve results. However, it’s unlikely that this application will ever be able to get the quantity of data required to make this more than a fun exercise

I’d also like to address privacy concerns. To a certain extent, this project relies on collecting personal data. However, because only the letter combination (and not the number) is collected, it cannot be construed as personally identifiable information.

This project relies on a few third-party resources. While the original version used Google Maps, this version of the application uses Mapbox and takes advantage of OpenStreetMap data. The open Mapquest API is used for geolocation calls (finding coordinates from an address or intersection, and reverse geocoding the “average” point to acquire county information).