Redistricting with 2020 Population Estimates (and 2016/2018 Political Data)

[muon2]

I think it's useful to keep in mind how much the projections based on estimates will change as each new year comes out. Last year jimrtex launched a thread to look at a 2020 map of IA by projecting 2016 estimates to 2020. I used his data and type of methodology to make this map:



At that time the districts had minimal deviations (rounded to the nearest 0.1%): SE +0.1%, NE -0.1%, C 0.0%, W 0.0%.

I updated my spreadsheet to use the 2017 estimates that came out earlier this year. The deviations are now SE +0.2%, NE +0.2%, C+0.3%, W -0.7%. That means the change from last year would be SE +0.1%, NE +0.3%, C +0.3%, W -0.7%. Each 0.1% is about 800 people, so in just one year the changes were enough to cause the 2020 projection for the western CD to dip by over 5000 people compared to the 2020 projection based on the prior year. Think about what deviation in the 2020 projections will be after 3 more years.

IA has very strict rules governing their maps. Based on the number of counties it is statistically likely that no CD will deviate more than 100 persons from the quota even with whole counties. That means it is highly unlikely that any of us will correctly guess today the map that IA will adopt in 2021.

IA is made difficult because it is so thoroughly driven by the cold calculus of geography and population. In states that have humans chop counties to exact equality it is far more likely that we might get close if we can get inside the heads of those who will draw the map.

[muon2]

For example, I took the 7/1/2017 estimates for IA and projected the populations forward to 4/1/2020. I started with the 4-corner model as used for the current map. I'm sure this isn't the best solution to the state requirements, and based on the inaccuracy of the estimates won't be the actual map, but it is whole county and provides all CDs within 100 of the quota.

[muon2]

Dallas County, Iowa has had some major population growth.

Of course it has. Dallas county has been averaging 3.9% growth per year so far this decade. That's why I don't use the current estimate, I forecast a projection to 2020 based on that rate of growth. For example the 7/1/17 estimate is 87,235 but I use a 2020 population of 96,896.

The question here is whether the rate of population growth in Dallas county will be faster or slower over the next 3 years than it has been over the previous 7. Are you suggesting that the growth rate will be significantly higher in the last years of this decade than it currently is?

[muon2]

The wild card is population loss - your method implies that if a county is losing population, it will lose a greater absolute # of people per year. That in turn implies, at least if one considers no in-migration, that a greater percentage of the existing population must be dying or moving away every year. But who knows, that could actually be the more accurate way to model some places like Detroit and to a lesser extent Cleveland.

In a county of declining population, the losses would diminish each year with my method. For example, consider a county with a 10%/year loss rate that starts with 1000 people. In the first year it would lose 100 people bringing it to 900. In the next year it would lose 90 people bringing it to 810. In the third year it would lose 81, and so on. The absolute number lost decreases each year.

Most population changes in nature are based on exponential change, which is what I model for my projections. It turns out that since financial functions require the same sort of exponential change, it's easy to set up my model on a spreadsheet and update each year as new estimates come out.

It's worth noting that some projection models only use a three (or two) year average rate of increase and project it to the end of the decade. In particular when the press reports the likely gains and losses for congressional seats each Dec, they are typically using a projection based on 2 or 3 prior years.

[muon2]

i feel like you should project populations using a more recent time interval like 2014-2017 or 2015-2017 instead of 2011-2017 (north dakota projections for example would be drastically different)

As I noted above there are some professional organizations that use a 3-year growth pattern to forecast the next decade. I did something like that in the prior decade to compensate for the effects of Katrina on LA and TX and avoid 2005. Ideally one wants enough years to get a good average without glitch years, but not so many that it misses real trends that are different from those at the last Census.

[muon2]

I found one small bug in the Virginia population estimates. In the city of Bedford both precincts have no population.

The independent city of Bedford VA was reabsorbed into the county in 2013. The city had estimates made in 2011 and 2012, but not after 2013. However the base population for Bedford county was increased to reflect the added population from the city.