Before we apply the model to the old districts, let’s look at the demographics two different ways. The chart below shows each of TX’s 2020 districts, with the population broken down by race/ethnicity (Black, Hispanic, Asian, white-non-Hispanic and other) and education (College Grad and non-College Grad). Each bar also has a dot representing the (logarithmic) population density1 of the district. The scale for that dot is on the right-side axis of the chart. For reference, a log density of 5 represents about 150 people per square mile and a log density of 8 represents about 3000 people per square mile.
We’ve ordered the districts by the modeled D-share which is helpful for understanding how the model responds to demographics and density.
Below we look at these demographics a different way. We place each district according to its proportion of college graduates and non-white citizens of voting age. We also indicate (logarithmic) population density via the size of the circle and modeled D-edge (D-share - 50%) via color. This makes it easier to see that the model predicts larger D vote share as the district becomes more educated, more non-white and more dense.
Below we chart the 2020 election results and compare them to the model. This is so we have some sense that the model makes sense and some sense of how approximate we might expect it to be.
The election result is on the x-axis and our model estimate on the y-axis, and a line representing where districts would fall on this scatter-plot if the model and election result agreed precisely. We’ve also included confidence intervals from the model. In districts left of the line, the model overrestimated the D vote share and underestimated it in districts to the right.
The model seems to capture the trends pretty well, though it has some significant misses2. We’re most interested in districts between 45% and 55%, and in that range the model does pretty well.
Below we put the same information in a chart in district-number order, just so it’s easier to check a particular district.
| State | District | Demographic Model (Blue Ripple) | 2020 Election |
|---|---|---|---|
| TX | 1 | 45 | 27 |
| TX | 2 | 60 | 43 |
| TX | 3 | 58 | 44 |
| TX | 4 | 41 | 23 |
| TX | 5 | 50 | 37 |
| TX | 6 | 57 | 45 |
| TX | 7 | 62 | 52 |
| TX | 8 | 47 | 26 |
| TX | 9 | 71 | 78 |
| TX | 10 | 53 | 46 |
| TX | 11 | 44 | 19 |
| TX | 12 | 52 | 34 |
| TX | 13 | 43 | 19 |
| TX | 14 | 56 | 38 |
| TX | 15 | 58 | 51 |
| TX | 16 | 62 | 65 |
| TX | 17 | 51 | 42 |
| TX | 18 | 69 | 76 |
| TX | 19 | 47 | 23 |
| TX | 20 | 61 | 66 |
| TX | 21 | 51 | 47 |
| TX | 22 | 61 | 46 |
| TX | 23 | 51 | 48 |
| TX | 24 | 59 | 49 |
| TX | 25 | 47 | 43 |
| TX | 26 | 54 | 38 |
| TX | 27 | 50 | 36 |
| TX | 28 | 56 | 60 |
| TX | 29 | 65 | 72 |
| TX | 30 | 71 | 81 |
| TX | 31 | 53 | 45 |
| TX | 32 | 60 | 53 |
| TX | 33 | 63 | 73 |
| TX | 34 | 57 | 57 |
| TX | 35 | 58 | 69 |
| TX | 36 | 45 | 25 |
We use logarithms here because density varies tremendously over districts, from tens to hundreds of thousands of people per square mile. We use population-weighting because the resulting average more closely expresses the density of where people actually live. For example, consider a district made up of a high-density city where 90% of the population live and then large but low-density exurbs where the other 10% live. Most people in that district live at high density and we want our density to reflect that even though the unweighted average density (people/district size) might be smaller.↩︎
We’re not convinced that we are handling population density as well as we could. Even on a log scale, the variation is large and very different state-to-state. We’ve worked on this some without any simple solution. If we make significant progress there, we’ll update this note.↩︎