I used the same data that I used for the map yesterday but plotted the ratio against median total earnings. Partly this was to see if there is a pattern to discern in North Dakota counties such as higher median earnings leading to an increasing or decreasing parity ratio. To assist in this examination, I added a linear regression line and a confidence interval and this is what I found.
I included the county names to help us see the pattern. One thing to realize is this is pretty clearly a nonlinear relationship. There really is no reason to believe it would be linear, that is, that the pay ratio would evolve in a predictable pattern as income level changed. As I mentioned in my earlier post there are many different factors impacting median earnings such as education and occupation. With that in mind I relaxed the linearity and got the following picture.
More counties are in the confidence interval and it seems to better capture the shape of the data. To really evaluate the relationship here we would need to control for those various differences in the data. One of the things I find really interesting is the divide between the lower median earning counties as far as the earnings ratio with Sioux county at the low end, just above parity and Slope county above 3. It is not perfectly the case, but as you move higher the spread, for the most part, becomes smaller. I am not completely sure why that is the case, maybe something to do with the number employed or the population numbers. I am also wondering if this would be the same pattern in other states, at least comparable states in the region.
