The Virginia House of Delegates is gerrymandered (still)

Politics & Elections
1

There was some court ordered redistricting between the 2017 and 2019 Virginia General Assembly elections which resulted in the Democratic Party gaining control of both the House of Delegates and the state Senate. The rest of this post will focus on the House of Delegates.

A federal court had found that the 11 VA House of Delegates districts were racially gerrymandered. The Republican legislature was unable to draw a map that the Democratic governor would sign into law, so the court appointed a special master to oversee the process (University of California, Irvine political science professor Bernard Grofman).

Since it is generally impossible to redraw only the 11 districts in question, 26 districts in total had to be redrawn in some way, but there are 100 districts in all so quite a bit of skew remains.

Here’s what they started with…

2017 Mean District: D + 9.6%
2017 Median District: R + 0.2%
Skew: R + 9.8%

And here’s how things ended up…

2019 Mean District: D + 9.7%
2019 Median District: D + 4.7%
Skew: R + 5.0%

The redistricting made a huge impact because the skew was so massive in 2017, but even so the House of Delegates map as a whole retains a five point Republican skew. In terms of seats, team R won about 9 more seats than they deserved in 2017, but still won about 5 more seats than they deserved in 2019.

In 2019 team D achieved a trifecta in Virginia (both chambers of the legislature and the governor). One of the things they should do with that trifecta is redistrict the House of Delegates to a fair map. And by fair I mean with as little skew between the median district and the mean district as possible. If they don’t do this they risk a future election where one party wins statewide by 3 or 4 points, but the minority party controls one or both houses of the state legislature.

Data sources: All the data used for this post came directly from the Virginia Department of Elections. My hat is off to whoever is running things over there, because not only do they make this stuff easier to access than the average state does, they also provide a link to all election results in JSON format which makes for extremely easy ingestion of these results into one’s data analysis platform of choice. Margins are represented here by difference in proportion of two party vote.

2017 results
2017 results (JSON)
2019 results
2019 results (JSON)

2

Isn’t one of the problems with any districting that Dems tend to be concentrated in urban areas and Pubs are spread out in the rural areas resulting in urban areas having very few Pub votes to dilute while outside the cities all of the Dem votes are diluted among the Pub votes?

I don’t know if it would work at a state level but that’s why I favor a true proportional system for the US House.

3

The difference between “mean district” and “median district” had never occurred to me. Would reducing this difference to zero be an effective way to end the ill effects of gerrymandering? I can imagine it’d lead to funky-shaped maps, but AIUI the shapes themselves aren’t the problem with gerrymandering.

4

Not really. Someone says soemthing like this in every gerrymandering discussion, but it’s not really an obstacle to drawing fair districts.

The difference in median and mean is pretty much the canonical way of demonstrating the the concept of skewness (e.g. on the Wikipedia page for Skewness).

Reducing skewness as much as possible is necessary but not sufficient for any fair districting plan. It would be possible for state with a large partisan divide to draw a map that reduced skew to zero that also completely shut out the minority.

For example, the California Assembly has 80 districts of over 400,000 constituents each. California also has a partisan lean of like D + 35 in Assembly races. If instead of the current map, registered voters were each randomly assigned to one of the 80 districts. This would result in 80 districts that were all pretty close to 67-35 D districts. The median and mean district would each be 67-35 so there would be zero skew. There would be effectively zero chance for an R candidate to win an Assembly seat. The chance is so small that it is not even worth the time to calculate it, but it would be much, much less likely than the odds of winning Powerball.

I don’t think it would be impossible to draw a map that effectively does the same thing. Using a computer you could go all Banach-Tarski on the state and pretty much produce anything that you want.

It seems a little unfair for a party that received over three million votes to get zero assembly seats so, for this reason, reducing skewness alone is not enough to ensure a fair map.

5

I also did the same analysis for LHOD’s beloved Old North State.

2018 NC House of Representatives election…

Mean District: D + 2.4%
Median District: R + 5.6%
Skew: R + 7.9%

The effect was about eight seats. The composition of the NC House 65 R to 55 D. Without that 8 point R bump it would probably be 57 R to 63 D.

6

This shows the (R) need for gerrymanders, suppression, and disenfranchisement. They know the current party can’t win free and fair elections. Impartially-drawn districts would force the party to drastically change, to be less repulsive. Expect fierce resistance.

7

Your own cite challenges your comments on skewness:

Do you have stats for a large number of districtings? Are some of them believed to be “objective”, i.e. not influenced by partisan tinkering? It certainly seems likely that clumpings (e.g. the urban-rural divide Saint Cad mentions) will deter the median=mode equation you seek. In fact, to achieve the districtings you call “fair”, you may be the one resorting to drawing salamander-shaped districts!

All discussions of gerrymandering eventually evolve into a realization of the efficacy of proportional representation. Anyway, with today’s high mobilities the motives for compact districts will become as obsolete as the horse-and-buggy.

:eek: Now, who’s exaggerating? :slight_smile:

8

Aha! Having just skimmed that Banach-Tarski wiki page, I suddenly now see how the Many Worlds Interpretation of Quantum Stuff could be plausible! :slight_smile:

9

Nonparametric skew suffices in this case because nonparametric skew is what we’re trying to avoid. It is fundamentally unfair for districts to be drawn in such a way that as a whole the people of state could vote for one party in their state legislature by 8.5 points or more and yet have that legislative chamber controlled by the minority party. Yet the is precisely what happens in states like Virginia and Wisconsin. Reducing nonparametric skew deals with this problem, and drawing salamander shaped districts to achieve this is not fundamentally unfair.

The TLDR version of this is that the median seat controls the legislature so the median seat should reflect the state as whole for things to be fair.

Proportional representation may indeed be the bee’s knees, but it would require an amendment to the state constitution in many states. It’s not a practical solution in most cases. Redrawing district lines both through legislative efforts and through the courts is achievable. It’s what allowed the majority to finally take the the Virginia House of Delegates in 2019, and it’s what that same House of Delegates should try to further now that they have control so that the minority party can’t take back control in an election they lose by 4 points in the future.

Here’s a couple posts from an old thread about Wisconsin. The gist of them is that control of the Wisconsin legislature is extremely resistant to the will of the people. A statewide D + 8.5 win resulted in a legislature that is 36D to 63R. The median seat (and control of the legislature) won’t be in play unless an election is somewhere around D + 20.

10

Yes, the shapes are one of the problems with gerrymandering.

For one, it makes it harder for people to know who is representing them. The more geographically cohesive the districts are, the more of a chance there is that the local news can implicitly convey that info. But that goes out the window with crazy shapes. And if they’re crazy enough, even if a nearby neighbor can tell you who their representatives are, yours have an increased likelihood of being different.

Basically, it’s a function of the perimeter-to-area ratio. For a district of constant area, as you increase the perimeter, you increase the number of people living close to the perimeter, thereby increasing the number of people who aren’t sure which district they’re in.

For another, complicated shapes of districts make it harder for a campaign to reach people in the district. If you have to cover more metro areas, any TV/radio/newspaper advertising has to be duplicated in each metro area, and more of that advertising is wasted on people outside your district. And while you can better target the people in your district by direct mail, you encounter the same problem at the zip code level. (You can use the list of registered voters in your district, but some jurisdictions charge hefty fees for that list.) And for targeted Internet advertising, you’ve got to be able to afford an outfit that can do that for you.

Both of these things make it harder to challenge an incumbent, so they add to the natural incumbency advantage.

11

Imagine a hypothetical state; call it Bizarrabama. Through the state run two rivers, fast and wide. The rivers are so wide that bridges are few and far between. The state is effectively divided into three regions: they call them North Bizarrabama, Central Bizarrabama, South Bizarrabama. Because the bridges are so hard to reach, the South Bizarrabama Fire Dept. serves all of South Bizarrabama, but none of Central Bizarrabama, and so on. There’s a South Bizarrabama High School Football League, and so on. Southerners would rather drive 100 miles to another Southern town, rather than cross one of the rickety bridges.

The state is entitled to three Congressmen and — fortuitously — the three regions each have precisely ⅓ of the state’s population. The D/R split is 58/42, 56/44, 30/70 in the three regions.

The (admittedly contrived) Bizarrabama state legislature decides that — surprise! — each region will elect one Congressman. The mean district is R+4. The median district is D+8.

TL;DR: Blatantly partisan gerrymandering is Bad. Having courts throw out districtings derived by fair-minded people because they fail a mathematical test will, rightfully, annoy many voters. Those who care about Congressional or State districting should spend their energy pushing for Proportional Representation which would address several other problems as well.

12

If the Virginia House of Delegates reduces the number of seats it contains to three, that argument will be relevant.

In Virginia, the Democratic party controls both houses of the legislature and the governor’s chair. They can draw fair districts and push for proportional representation. This is not a one or the other situation, but only one of the things is realistically accomplishable in the next two years.

They have the power to do one thing but not the other and doing the one thing does not prevent anyone from pushing for the other.

13

  1. Is this really likely?

  2. There ain’t enough rolleyes for your exceedingly contrived, extremely small n example.

14

High mobility? Are you now going to bus people out of towns to go vote in rural areas?

15

He’s got you this time, Septy. But hey, nice try.

Hope you didn’t put down a deposit on those buses, but maybe we can use them to transport the illegal voters to Michigan and Pennsylvania. The three million illegal voters last time took up all the available buses in America, but luckily, nobody noticed.

16

As Jefferson said, it’s an uphill battle to get people to abolish the forms to which they are accustomed, and that’s even more true if the alternative is unfamiliar to them.

Convincing the public at large of the virtues of proportional representation is going to have to be a long-term project, if it’s ever going to succeed. First, you’ve got to make people aware of it, and you’ve got a long way to go there.

So there’s no earthly reason for any Virginia politician to waste one iota of her time in the 2020-2021 session to push for proportional representation, rather than simply fixing the immediate problem of gerrymandered House of Delegates and state Senate districts.

17

By redistricting 26 of Virginia’s 100 House of Delegates districts, not only did the court cut the skewness of the entire HoD in half, but the new map of the 26 districts had districts with much more normal shapes than before.

18

I thought my meaning was obvious. People are telecommuting, or car-commuting long distances. For a young worker to move 100+ miles for a new job is more common these days. Districts following geographical lines is less relevant today.

:confused: I was NOT arguing that egregious gerrymandering examples do not exist. :smack: I am ASKING for the “skewness” stats on districts, if any, thought to be drawn without partisan motives.

19

I truly do not understand what you are saying here. Should people have to vote where they work? Otherwise, your commuting reference makes no sense at all:

  1. I live at point A.
  2. I work at point B (40 miles away).
  3. I’m assigned to vote at point C (25 miles from A and 55 from B) doubling my drive on election day.

And I’m supposed to be okay with this because I’m used to commuting?

20

Connecticut is pretty good.

Seats: 151
Mean District: D + 9.9
Median District: D + 8.9
Skew: R + 1

Plus, somewhere around 20 competitive districts.

Here’s the whole data table inside a spoiler tag.


Index	District	D votes	R votes	D margin
1	141	0	7237	-1
2	149	0	6971	-1
3	112	0	6932	-1
4	108	0	6383	-1
5	70	0	5901	-1
6	68	2949	7261	-0.422331048
7	80	3338	7089	-0.359739139
8	131	3371	7048	-0.352912947
9	78	2900	5818	-0.334709796
10	105	3346	6697	-0.333665239
11	113	3347	6233	-0.30125261
12	76	4190	7615	-0.2901313
13	22	2989	5297	-0.278542119
14	87	4308	7428	-0.265848671
15	63	3443	5868	-0.260444635
16	89	4367	7348	-0.254460094
17	8	4597	7610	-0.246825592
18	57	3978	6538	-0.24343857
19	52	3307	5433	-0.243249428
20	86	3961	6506	-0.243145123
21	44	2774	4441	-0.231046431
22	71	3158	5018	-0.227495108
23	122	4430	7006	-0.225253585
24	69	4501	7085	-0.223027792
25	61	4078	5926	-0.18472611
26	107	4736	6556	-0.161176054
27	67	3911	5301	-0.150890143
28	32	4655	6298	-0.150004565
29	62	5092	6871	-0.148708518
30	47	4343	5829	-0.146087298
31	66	5172	6900	-0.143141153
32	42	3616	4810	-0.141704249
33	125	4959	6522	-0.136137967
34	81	4286	5586	-0.131685575
35	55	5219	6668	-0.121897872
36	117	4626	5823	-0.114556417
37	151	4786	5918	-0.105754858
38	23	5566	6854	-0.103703704
39	59	3608	4411	-0.100137174
40	119	5044	6063	-0.091743945
41	114	5033	6022	-0.089461782
42	14	4976	5929	-0.087391105
43	90	4789	5683	-0.085370512
44	123	4868	5705	-0.079163908
45	51	3577	4110	-0.069337843
46	132	5432	6186	-0.064899294
47	74	2877	3168	-0.048138958
48	35	5461	5979	-0.04527972
49	45	4035	4410	-0.044404973
50	34	5183	5631	-0.041427779
51	106	5297	5697	-0.036383482
52	30	5499	5848	-0.030757028
53	77	4396	4640	-0.027003099
54	38	5088	5362	-0.026220096
55	134	5251	5516	-0.024612241
56	143	5586	5833	-0.021630616
57	37	5244	5473	-0.02136792
58	111	5888	6048	-0.013404826
59	17	5929	6004	-0.006285092
60	27	5215	5171	0.004236472
61	31	6303	6192	0.008883553
62	64	5665	5455	0.018884892
63	120	5222	5028	0.018926829
64	101	6241	6009	0.018938776
65	36	6338	6078	0.020940722
66	65	3388	3214	0.02635565
67	138	4658	4395	0.029051143
68	53	5508	5110	0.037483519
69	103	5188	4680	0.05147953
70	83	4765	4251	0.05700976
71	139	3925	3467	0.061958874
72	2	5245	4623	0.063032023
73	150	4859	4272	0.064286497
74	60	5454	4606	0.084294235
75	58	4079	3438	0.08527338
76	135	6449	5393	0.089174126
77	142	5790	4797	0.093794276
78	48	5641	4628	0.098646412
79	118	5404	4421	0.100050891
80	102	5856	4753	0.103968329
81	85	4763	3857	0.105104408
82	79	3927	3156	0.108852181
83	29	5817	4622	0.114474567
84	104	3785	3002	0.115367615
85	21	6323	4918	0.12498888
86	43	6369	4913	0.129055132
87	99	4502	3407	0.138449867
88	50	5662	4282	0.138777152
89	73	3828	2883	0.14081359
90	28	6273	4691	0.144290405
91	56	4846	3602	0.147253788
92	40	3572	2557	0.165606135
93	100	5033	3534	0.174973736
94	13	5799	4053	0.177222899
95	82	5401	3699	0.187032967
96	24	3945	2699	0.187537628
97	147	6187	4153	0.196711799
98	16	7625	4935	0.214171975
99	12	5327	3345	0.228551661
100	136	7945	4704	0.256225789
101	109	3555	2066	0.264899484
102	33	5356	3096	0.267392333
103	133	5845	3241	0.286594761
104	46	3422	1846	0.299164768
105	41	5526	2961	0.302226935
106	26	3440	1601	0.36480857
107	88	6218	2786	0.381163927
108	137	6076	2662	0.390707256
109	115	4814	2016	0.40966325
110	121	5376	2164	0.425994695
111	146	5913	2257	0.44749082
112	110	2367	880	0.457961195
113	18	7608	2783	0.464344144
114	116	3643	1299	0.474301902
115	127	3965	1341	0.494534489
116	148	5013	1604	0.515188152
117	72	3362	1072	0.51646369
118	96	6183	1598	0.58925588
119	140	3844	974	0.595682856
120	97	4890	1207	0.604067574
121	129	4507	976	0.643990516
122	145	4580	965	0.651938683
123	126	5362	997	0.686428684
124	5	6474	1144	0.699658703
125	6	3563	516	0.746996813
126	128	2366	317	0.763697354
127	95	3166	424	0.763788301
128	94	6043	743	0.781019747
129	4	2733	335	0.781616688
130	124	3971	458	0.793181305
131	130	3202	362	0.796857464
132	92	6221	627	0.816880841
133	93	4782	310	0.878240377
134	75	2491	0	1
135	3	2975	0	1
136	25	3075	0	1
137	84	3513	0	1
138	7	4098	0	1
139	39	4136	0	1
140	49	4325	0	1
141	1	4419	0	1
142	11	4547	0	1
143	10	5236	0	1
144	54	5956	0	1
145	9	6042	0	1
146	20	7220	0	1
147	144	7242	0	1
148	91	7841	0	1
149	98	8572	0	1
150	15	9268	0	1
151	19	9515	0	1