So an ongoing project that is taking a significant amount of my time (and thus taking time away from blogging) is executing a full population projection for North Dakota for the next 50 years. There seem to be some special issues when looking at the state level, for a small population state. 

One of the major problems is simply data availability. There are many suppression flags when looking at a state like North Dakota, especially when you try to look at, or in my case derive, age-specific rates. I get why they are suppressed, so that is not really a complaint, it is that there is often not enough information to make meaningful interpolations. 

One of the big issue for me is estimation of the key demographic forces. As an example consider the migration specification:

where I model migration as a function of previous migration and relative economic performance. I really like the specification and think that it is helpful for at least the major sectors at the state level for North Dakota. However, deriving migration at the age level is a real issue in this case. 

Essentially I am estimating an age specific pattern for migration and then seeing how stable it is over time, and then how correlated any changes are with specific economic activity in North Dakota. After that the goal is to take any future total migration estimates and attempt to apply the same pattern to it. It is net migration but there is not much else that can be done with the available data.