I took another pass at the idea of making comparisons of different counties in North Dakota (my first post is here). I used some different data. In this case I used the annual growth rate in county labor force from 2000 to 2018, the annual growth rate in county non-farm income from 2000 to 2018, and lastly the annual growth in taxable sales and purchases from 2000 to 2018. This is not an all inclusive list of variables but it is more data points than before. I also use the longer time frame to allow some cyclical swings but also to smooth them out over that time.
I vary the number of groups this time as well. I specify three to six clusters to see how the composition of the groups changes too. In all seriousness the suggestion here is that it is difficult to come up with the right comparison counties in North Dakota. As “homogenous” the population seems to be, the economic data is variable and makes for some issues.
I put the clusters into a map of the state to make it easier to see where the comparisons are. The geography of the clusters is interesting too.
The first point to mention is that the Bakken area is off on its own except for a random inclusion of Sioux county. My suspicion is that it is there for some factors related to growth rates versus levels. Maybe the most interesting for my region of the state is that the frequent comparison or competitor county is Fargo, and they may be a competitor but they are not a peer over this data.
With four clusters we now have Grand Forks and Cass together, the Bakken on its own, and some other interesting combinations of counties. The spread of the county types across the state is very interesting. It suggest that contiguous regions are not necessarily enjoying similar economic experiences.
At five clusters we again find interesting divisions, though the core Bakken counties are the together still. But then we have very diverse groupings of the counties that really also suggest a need for more detailed data about economic circumstances in the counties.
With six clusters the Bakken area is still together. Remember the data does not include specific economic activities, just non-farm income, labor force, and taxable sales and purchases, all as annual growth rates over 18 years.
For me the big takeaway and I process this and consider next steps is the geographic spread of the clusters. I need to think of other data to include, which could be population data I think, and maybe some other economics data such as the distribution of economic activity across the different economic sectors.