North Dakota labor market dynamics are an interesting set of circumstances. One of the first states to deal with too few workers compared to open positions it also demonstrated that in meaningful ways an unemployment rate can bee too low. One of the more frequent labor data sets to examine is the weekly initial unemployment claims. As a side note, weekly data can be problematic to deal with and is not often of great use at that level of frequency. I aggregate the weekly data in to monthly observations to make implementation a bit easier and the data friendlier.
I did a forecast based on the initial claims data too, doing some tinkering around. There needs to be some adjustment to the model to account for some changes in seasonal peaks but overall it does not do too badly for a basic, first pass model.
The misses are on the peaks and valleys where the amplitude seems to be a bit variable over time. I will make some stabilizing adjustments in the next pass of the model to see if there is any correction that occurs. The initial claims forecast is over the next 24 months. To make it easier to see I dropped the fitted series and zoomed in after 2017 only.
The forecast predicts a similar pattern to the past, but less volatile in amplitude right now. Again, this is a very simple forecast and there are some factors to add in that will make it a more complete, and presumably more accurate forecast.