DRA 2020 - New Update

[Boobs]

Hi folks. DRA has been updated to have more powerful analytical tools. Let's take a looksie!



Here's a quick Indiana map I whipped up for a sample. You can see at the top right two new buttons, Analytics and Advanced.

Analytics takes you to this screen:


A silly little radar graph of the analytical measures DRA uses. The larger your radar graph values, the better. As you can see, this isn't too great of a map.



First off, we've got proportionality, which is a very simplistic measure. Basically, it's the statewide Democratic vote - presumably using whichever measure you select when creating your map (I used the 2012-2016 composite for this map) - subtracted by the % of seats leaning Democratic. 2/9 (22.2%) of the seats lean Democratic whereas the expected number, disregarding geography, would be 4 democratic seats.

 

Secondly, competitiveness. It is the sum of normalised values, with districts between 40-60% for a party counting towards the sum. Two of my districts fall barely within this range (5, at 40.02% Dem and 8 at 40.6% Dem) so the low competitiveness score is understandable.



Third, minority rights. This isn't a substitute for VRA compliance, as it disclaims, but it uses a basic formula to discern how many districts with minority representation may be expected, both as specific groups and as a minority-majority measure. I've got one opportunity minority coalition district in Indianapolis, and it's probably likely Carson gets reelected.



Next, compactness. DRA uses two measures - Polsby-Popper, which is the proportion of a district's area to the area of a circle with a circumference equivalent to the district's perimeter (1 being most compact) and Reock, which is the proportion of a district's area to the area of the smallest possible circle encircling the district completely (1 being most compact again.)

The values are then normalized with historical values and averaged to get the score.



Finally, splitting. There's a relatively sophisticated formula for measure those, but they're normalized and then scored. The measure does penalize small cuts to a county, even for population equivalence.

It also tells you which counties are split that are smaller than a whole district. This is a rule that is encouraged by redistricting laws such as that of Ohio.

Now, onto the Advanced tab. It's pretty cool and has some neat graphs.



A graph representing district partisanship. The declination represents an "angle" at which the two parties's vote-rank measures meet. A larger angle implies that one party's votes are unfairly packed - something that is achieved in this Indiana map.




A seat-vote curve. My favourite. Shows you the change in seat % as a party's vote increases. For example, you can see that Democrats need to win 65% of the vote in Indiana to win all districts.

Votes bias is the percent deviation from 50% that Democrats need to win statewide in order to win 50% of the seats - here, Democrats need 53.34% of the statewide popular vote in order to be expected to win 50% of the seats.

Seats bias is the percent deviation in seats from 50% that Democrats are expected to win if they receive 50% of the statewide vote - here, for example, Democrats are expected to win 33.63% of the seats if they win 50% of the vote.



The bias measures summarise a lot of the previous measures. Most of these are self-explanatory.



The responsiveness metrics measure the change at the current average of votes. At the current D vote, a 1% change in margin (Ds up 0.5%, Rs down 0.5%, would only result in 0.13 change in predicted district wins. A more responsive map would have a district be more competitive at the current D/R vote shares, which would raise the responsiveness metric.


So that's about it. A pretty great update, I'd say - it helps introduce a lot of quantitative measures of district fairness that should help us judge maps better.