The 2022 mid-term election cycle brought some interesting results. I mentioned a while back that I’d take a look at those results, with the perspective of hindsight and some data. I’m here to deliver. There’s a little math. And some commentary.
In 2020 the nation conducted a nationwide census. Based on those numbers all legislative and congressional district boundaries were redrawn. [Well, not all. A few states have only one Congressional seat]. These boundaries last 10 years, until the next census. There’s usually a lot kvetching when states roll out their maps. So that’s my main goal here: to objectively assess the district boundaries. How well are the voters’ desires reflected? Or, how badly are things gerrymandered?
A preliminary statement of the obvious. Gerrymandering does not affect presidential elections, or US Senate elections, nor any statewide election, such as Governor.
On a national level, it only affects the lower house of Congress: the House of Representatives. [It also affects states’ legislative districts, but that’s outside this essay’s purview.]
With that out of the way, let’s start at the very top level of the 2022 nationwide results.
Of just over 107 million votes cast for Congressional Representatives on November 6, 2022: 47.41% were for Democratic candidates; 51.56% went for Republicans; 0.54% for Libertarians; and 0.50% for sundry candidates of other parties.
When all the smoke had cleared and the dust settled, the final makeup was 213 Democrats and 222 Republicans. Or 49.0% to Democrats, and 51.0% to Republicans. Overall, I assess this to be a very fair representation of the nation’s intent. That’s overall. There are some incongruent results on a more granular — state-by-state — level.
First, some caveats.
- (1) Some states, most notably California and Washington, only permit two candidates on the November ballot. [Each state does have their own election rules]. In six California races there were two Democrats, and no Republicans on the ballot.
- (2) Across all states there were 35 uncontested races in November. In those, the unrepresented party obviously received zero votes (there are always a handful of 3rd party write-in votes). Of these, 23 had only one Republican candidate, and 12 a Democrat. [The Washington Post, and others, contend that this cost Democrats the national vote — as if that matters. But there’s a reason no one opposed the single candidate: everyone knew the candidate was sacrificial and would garner few votes. Case dismissed].
The upshot of the caveats is that the November vote tallies can’t precisely reflect voters’ intentions by using simple ratios — at least with respect to parties they favored — as some voters were not able to cast a vote for a candidate of a party they preferred. Accepting that this puts some noise in the data … we forge ahead. Nothing is perfect.
[So, for the top level, I assess that the allocation of seats by party tracked the electorate’s preference amazingly close. Much closer than I’d expected. See? I’m wrong sometimes.]
Applying the same algorithm* as in my essay Mr. Gerry, I found the following the states to be the most egregious in mis-representing the voters’ intentions in the allocation of CD seats by party. {* see afterthoughts, below}.
________________________
State | Bias | Favoring |
---|---|---|
California | 8 | DEM |
Illinois | 5 | DEM |
Massachusetts | 3 | DEM |
Connecticut | 2 | DEM |
Florida | 2 | REP |
Georgia | 2 | REP |
Indiana | 2 | REP |
Iowa | 2 | REP |
Maryland | 2 | DEM |
New Jersey | 2 | DEM |
Ohio | 2 | REP |
Oklahoma | 2 | REP |
Tennessee | 2 | REP |
Washington | 2 | DEM |
Wisconsin | 2 | REP |
________________________
On net, summing the entire table for all 50 states, the nationwide results of the election compared to my “fair” model show a +2 bias for Democrats, a -1 for Republicans and a -1 for Libertarians.
(Libertarians lost a possible seat in Texas in 2022. Per 2020 presidential CA vote results, where minor party candidates received votes, and per my model, Greens and Libertarians would have received a seat. However, the CA November election run-off model eliminated all 3rd parties).
The two states with the largest biases, CA and IL, were predictable. Each has many Congressional Districts (53 and 17), and statewide they are overwhelmingly governed by one party.
The absence of Texas on the list (38 CDs and mostly led by one party), and often pointed to as an example of extreme gerrymandering, is somewhat of a surprise. Observers have long bemoaned their districting map, including the esteemed Brennan Center on the current map. They make some good points. But Texas awarded 13 of its 38 seats (34.2%) to the minority party; statewide the Dem candidates garnered 34.7% of all votes cast. A pretty good match. The bias in TX turned out Dem +0, Rep +1, Libertarian -1. Is there some gerrymandering in Texas? Almost assuredly (see Brennan), but you wouldn’t notice by the bottom-line statewide results. In fact, the Texas results are so very close to the actual statewide tallies that one could argue that all the districts which appear gerrymandered are made that way to get a good balance in the result. [1]
Also, a bit unexpectedly, New York is not on the most-biased list (it came in at +1 Dem). Republicans won 11 of New York’s 26 seats, or 42.3%. Statewide Republican candidates received 41.5%. A fair result. [2] [yeah, yeah, George Santos, I know]
I suppose one could use these results as a sort of proxy to determine how much any one state might be gerrymandering. That’s a tough call more than you’d think. Another proxy might be the bias as a percentage of any state’s total CDs. For example, if a state has 6 CDs and statewide votes suggest a 3-3 split, but it ends up 5-1, then that’s a mis-allocation of 2 out of 6, or 33%.
Judged this way (for states with 3 or more CDs) the top mis-assigners are:
- Iowa 50%;
- at 40% are Connecticut and Oklahoma;
- at 33% we have New Mexico and Massachusetts.
- Illinois comes in at 29.4%.
- And a few at 25%: Arkansas, Kansas, Maryland, Nevada, Utah, Wisconsin.
As most in the list have only a few CDs, it’s difficult to determine a consistent gerrymandered bias. California, even with its large overall +8 Dem bias, is only 15% off, since it’s so large. In the largest states on this list, Illinois and Maryland, maps are mixed on suggesting any strong bias in district boundaries.
I don’t want to dig too much into the states. There are 50 of them, and that would take quite a while. Maybe later. But Illinois, the land of my birth, and the Cook County Political Machine, piqued my interest. How could the “Land of Lincoln” lose a seat and then get even more imbalanced? [OK, OK, I know. Richard “Boss” Daley has been dead a long time, and his machine has all but expired — but still: Cook County]. From the 100,000 foot level, the Illinois map (above) looks pretty even handed … with two exceptions (CD 13 and 17). At a more granular level, some other stats stand out: there are 11 total CDs in and near Cook County (Chicago). All 11 went to one party, as did CDs 13 and 17.
I’ve discussed some other states before. Maryland’s districts look much more reasonable since the recent redistricting. Which goes to show that unbalanced results can result even when no gerrymandering is apparent from gazing at the CD boundaries.
Some states have recently taken district boundary drawing out of the hands of (often very partisan) legislatures. Supposedly independent commissions drew the boundary lines. It looks like mixed results. In a few (Colorado, Michigan) the results turned out exactly fair. In others (Arizona, California, Washington and even New York’s first cut) missed the mark by non-negligible amounts. Which shows how difficult the task might be: both choosing fair “independent” teams and then actually drawing fair lines. [3]
A few years ago I wrote about racism in America. Yes, there’s racism. Of course. Undeniable. It’s horrible. It’s repugnant. Yet it’s not as bad as we are led to believe. Look where we’ve come in 60 years. Same with many things. Bad things aren’t as bad as we often think; and good things aren’t as good (or lasting) as we like to think. I submit it’s the same with gerrymandering — or at least with CD boundaries drawn so that voters’ interests aren’t fairly represented.
In 2022 the results look to much more fairly represent citizen’s intent than I thought when my research began a few months ago. Maybe we just sort of got lucky with most bias effects more or less canceling out. People will continue to kvetch, no matter what. I tried.
Weirdly drawn boundaries aren’t in and of themselves evidence of such unfairness. And clean looking boundaries are no guarantee against it.
Look to the numbers first. Look fishy? Then look at CD shapes. Double sniff test. Top level and at most state levels the results this time around were pretty fair. Demographics change, and we are “stuck” with these boundaries until the 2032 elections. Maybe I’ll be around then to reflect on how well the boundaries held up in representation fairness.
I am cautiously optimistic that, with states trending toward independent commissions to draw the lines and with state courts growing more willing to strike down blatantly unfair lines, we’ll continue to trend to even more fairness in the decades ahead. [4]
Well, we can hope.
Peace.
Joe Girard © 2023 — notes and afterthoughts below
Thank you for reading. As always, you can add yourself to the notification list for newly published material by clicking here . Or emailing joe@girardmeister.com
Notes on my election fairness model below, after all footnotes and maps.
[1] California and Illinois lost one CD each due to the re-apportionment after the 2020 census. Texas gained 2 CDs. Sometimes people “vote with their feet.”
[2] the New York legislature submitted a CD map whose boundaries that were deemed a “brazen gamble” by the New York Times. In fact, they were ridiculous. These were judged to be unconstitutional by the state’s supreme court. A court appointed ‘special master’ drew a set of much more competitive (and fair, based on results) districts.
[3] Both Washington and New York allow the state legislature to override the independent commission suggestions. In New York, they did; and failed, as mentioned. In Washington, the commission failed to submit a map by the required deadline. It was drawn, just not submitted. A bookkeeping fiasco. In the end, the Supreme Court ruled that the commission’s map, even though submitted late, shall be accepted. It’s impossible to call it a gerrymander, even though, based on statewide party vote tallies, the results would be different by 2 CDs.
[3a] The CD map of Massachusetts looks relatively benign. Yet it yielded a
9-0 split, all for one party, despite any fairness models which would suggest 6-3, or at worst, 7-2. [see below]
[4] The US Supreme Court has declined to take cases of unfair CD boundaries. This, they assert, is a problem for the states to sort out.
[5] Afterthought: a reasonable way to address the representation fairness issue is to add more resolution to the districts by drastically increasing the number of Congressional Districts (and seats). The number of representatives has remained essentially unchanged going back to our nation’s beginning (it’s grown only as states were added).
In 1900 each District had an average of 191,000 residents. In 2020 that grew to 761,000. It’s absolutely impossible to draw districts completely fairly when they each must have the same number of residents (in the same state) when there are so few CDs allocated.
I don’t know (or really care) if the chamber can hold more than 435 people. I assume so, as joint sessions get an extra 100 in there. Plus, many many representatives are not in the chamber for most sessions.
Notes on Joe’s CD election model: My model explained, by example.
Suppose a state has 10 CD seats to assign. Party A gets 52.1%, Party B 43.1%, Party C 4.1% and the rest 0.7%. We begin by multiplying the percentage for each party by the number of seats available. Then we strip off fractions, using only whole integer numbers.
A) 52.1% x 10 = 5.21
B) 43.1% x 10 = 4.31
C) 4.1% x 10 = 0.41
Round 1: Party A gets 5 seats. B gets 4. That’s 9, there is one seat left over.
In our simple example 10% equals one full seat.
Subtract away from A and B what has already been awarded.
A is left with 2.1%, B with 3.1% and C is still at 4.1%. So, C gets the final seat.
Repeat until all seats are awarded. So, in our example, if any further seats remain, C then loses its 4.1% (it’s been assigned a seat). And the next seat, if available, would go to B. (3.1% > 2.1%).
It’s simple, elegant, easy to apply and – this year anyway – seems to have given a pretty good unbiased look at the results. (My assessment on how “unbiased” it is, well — that’s probably biased). Also, it’s not all that different than how the states are allotted CDs to begin with.