The SDRtThing2Do committee is working the Brown County Fair to pitch Amendment T, the proposal to get rid of gerrymandering by empowering an independent redistricting commission to draw our legislative maps. Spokesman Matt Sibley from Huron took three minutes this noon to explain how Amendment T works… and perhaps amazingly, doesn’t say the word gerrymandering once:
Sibley emphasizes that allowing legislators to draw their own districts constitutes a conflict of interest. Last week, Amendment T opponent Jason Ravnsborg from Yankton told SDPB Radio that there is not conflict of interest, and that if we voters don’t like how legislators draw their own maps, “we can vote them out of office.” Ravnsborg misses Sibley’s point: if legislators can take advantage of this conflict of interest to draw lines that separate them from voters who don’t like them and surround them with voters who do, then the voters who are ill-served by those conflictually drawn boundaries never get a chance to vote the turkeys who drew those boundaries out.
Important technical notes on Amendment T:
- As Sibley says, Amendment T leaves in place the decennial redistricting schedule. We’ll redraw the legislative map every ten years, the year after each census. But Amendment T also provides that we would do one special redistricting in 2017. If we pass T, we will draw new Legislative districts for the 2018 election.
- Sibley’s statement that the independent redistricting commission would consist of three Republicans, three Democrats, and three others isn’t perfectly accurate. As I pointed out in my first read of Amendment T in July 2015, the exact language of the proposal—”No more than three members of the commission shall be members of the same political party”—does not guarantee a 3–3–3 split. The Board of Elections could name two Republicans, two Democrats, a Libertarian, a Constitutionist, a Green (you guys are petitioning to form a new party, right?), and two independents. My opponent in the District 3 Senate race has feigned concern that that provision could allow partisan rigging of the independent redistricting commission; I remain convinced that the mapping criteria written into Amendment V make it far harder for even a rigged commission to rig the election map than it is for partisan legislators under the current system.
Mr. Ravnsborg saying that it is not a conflict of interest because the voter can vote them out of office if they don’t like the redistricting is asinine. By that logic, there never would be a conflict of interest for a legislator because they were voted in to office and haven’t been voted out yet. Only after they have been voted out would there be an actual conflict of interest, but then the conflict is removed because they are no longer in the legislature with the redistricting power. Thus, legislators never have a conflict of interest. Problem? What problem?
That is how some of them seem to operate so it is no surprise that this attitude is out there.
non biased redistricting is very good. But who choses the committee?
Amendment T oblishes gerrymandering which, we all learned way back in high school government class, is a bad thing.
Amendment T is good. Gerrymandering is bad.
Why not codify an algorithm for redistricting? That would be the most impartial method, immune from the whims of politicians.
Kind regards,
David
Darin, again, we mustn’t expect logic from Republicans planning to run for office in South Dakota.
Roger, the Board of Elections picks the nine committee members. Any registered voter not holding public or party office can apply. The Board of Elections whittle applications down to a pool of 30—ten from each of the two largest political parties, ten from the rest—then picks nine from that pool to draw the map.
David, Amendment T authorizes the IRC to draw the map according to these criteria:
That language leaves open to possibility that the IRC could choose to use an algorithm, even an automated algorithm, to draw that map. I’m not sure they could use Brian Olson’s computer scheme (demonstrated at http://bdistricting.com/2010/), since Olson’s method looks purely at population, compactness, and contiguity but does not consider community and geography, but they could adapt that algorithm.
Hi Cory,
Great find! I presume we would be looking at the 35-district South Dakota Senate Map and using the districts the same way we do now (2 House reps, 1 Senator). [His House District map creates 70 distinct house districts, which is beyond the scope of Amendment T.]
On first blush, his algorithm is amazing at optimizing what he wanted to optimize: keeping all districts within 220 people of each other, and bringing the average voter 3 miles closer to the center of their district.
But in practice, it looks a little clunky… especially along the Missouri River. Does Pierre really need to be separated into 3 different districts? (light-blue, gray, and pink) And it looks like we have slivers of many small counties that are separated off, which probably doesn’t make a lot of sense administratively. (Why can’t all of 3900 people in Lyman County vote in the same district?)
He didn’t design the algorithm with the needs of SD in mind, but with a little tweaking, I think we have something. The code is open source, so tweaking can be done. Unfortunately, we need to find someone comfortable with linux, C, Java, and python… and by his own acknowledgement “The code isn’t pretty or easy”. But maybe the IRC could just come up with their own criteria and commission him to make the tweaks.
Kind regards,
David
PS He also links 8 other algorithms here.
PPS If Amendment T passes, how does one volunteer for the IRC? That’s my kind of project.
David, yes, Amendment T does not change SDCL 2-2-42, which calls for 35 Senate districts. Amendment T says, “The commission shall establish legislative districts by dividing the state into as many
single-member, legislative districts as there are state senators. House districts shall be established wholly within senatorial districts and shall be either single-member or dual-member districts as the commission shall determine in compliance with federal and state law.” One could argue that the IRC’s constitutional mandate to draw districts overrides the existing statutory mandate for 35 Senate districts and that the IRC would be as free to set the number of districts as the Legislature is now, but I’d like to get a legal opinion on that.
You identify the key failure of Olson’s algorithm: respect for geographical boundaries. T seeks specifically to avoid fragmenting cities and counties (see Brown County for such unnecessary fragmentation right now, with three districts in a county that need only be divided in two).
We could include coders in the IRC, or the IRC could commission Olson to do the nuts and bolts. T says, “The Legislature, under the direction of the commission, shall provide the technical staff and clerical services that the commission needs to prepare its districting plans.” I think that language would allow the IRC to direct the Legislature to offer Olson a contract to help make the map by T’s criteria.
If T passes, any registered voter not holding office can apply to Board of Elections by January 31, 2017. The process should be much easier than running for State Senate. :-)
David Bergan, great minds think alike. I was thinking the same thing about an algorithm’s utility for redistricting.
The problem is you would always have to have the ability of humans to intervene and overrule the results because I don’t think we can write an algorithm that would take into account every contingency, including never committing an error. Since they can’t seem to make a computer that doesn’t lock up from time to time (although they are much better now than in 1995), we also can’t blindly trust a computer program to set our redistricting boundaries. Then, when you reintroduce human oversight, you reintroduce the potential for partisan shenanigans. Dividing up the power and making it public as Amendment T does, should keep the shenanigans to a minimum.
As Cory has pointed out, it appears T contemplates using IT resources to aid the human effort to carry out nonpartisan redistricting. Let’s drag SD into the 21st century!
“Since they can’t seem to make a computer that doesn’t lock up from time to time (although they are much better now than in 1995), we also can’t blindly trust a computer program to set our redistricting boundaries.”
Hi Darin,
This shouldn’t be a concern so long as the computer can not-crash while it’s running the calculation. The program probably doesn’t run for more than an hour on modern computer hardware. It’s not the kind of system that needs to be running 24/7 all decade long, you just do the calculation once every ten years when the census numbers come in, and save the results.
For the sake of trivia, there have been systems that “don’t lock up” since Tandem Computers were launched in 1974. These systems were designed with everything being redundant… so that if there was a fault, the other side could take over. (similar to how an aircraft has 4 independent engines, so that if one engine locks up, the plane doesn’t crash) They were expensive, but necessary for ATM networks, banks, stock exchanges, etc.
Kind regards,
David
PS Coincidentally, when I went to post this comment the first time, I got a “Site Can’t Be Reached” error from dakotafreepress.com. Cory needs to upgrade his server to a Tandem. :)
I hear you David about redundancy, but what I meant to get at is the folly of humans designing a perfect machine (in this case the algorithm) and letting it run for decades. Since we are not all-knowing, it seems likely that unforeseen issues will arise. If we can build in the mechanism for adjustments and bug fixes without reintroducing a strong possibility of partisan influence, then we could go the algorithm route. If the new redistricting commission can agree to use one that they could adjust as problems are revealed, I am in favor of it.
Back on computers, I blame Bill Gates for introducing the operating systems that were so flexible and subject to adaptation for a million and one uses. Anytime you have that sort of almost limitless possibilities means you have almost limitless opportunities for bugs to creep into the system. Analogizing computers and software to everyday hardware for a minute: A quality toaster does one thing and does it reasonably well without fail. If you would also like the toaster to function as an iron for taking the wrinkles out of clothes, it becomes a much more complicated machine. Then, if you want it to also be able to plug and play with your dishwasher, it becomes even more complicated. The more complicated the machine is, the more chance of problems with its operation. It is full employment for coders until they perfect artificial intelligence.
Darin, I’ll lean toward David’s faith in tech on this one. The redistricting algorithm and the laptop we run it on is a toaster, not a self-driving car. The LRC will have nine months in a non-election year to do its job. There’s no live-fire danger from computer error or crash. If the computer crashes, we’ll have time to ask David to lend us his computer or order a new machine from David’s recommended vendor. If the algorithm fails to observe any of the five criteria (wrongly assigning some island of Pennington County to a district in Minnehaha), the commission can catch and correct with patient visual proofing of each map.
Notice that the criteria aren’t absolute. “To the extent practicable” allows some human-sense revisions we’ll make to the algorithm’s mathematical perfections that will still satisfy the spirit of Amendment T.
I’m still comfortable maintaining the human element as the proofreader and overrider of the algorithm’s choices, thank to the last criterion (unnumbered in statute, but sixth by my count)—”Party registration and voting history shall be excluded from the redistricting process. The places of residence of incumbents or candidates shall not be identified or considered”—and because the humans doing the checking will be nine people with no partisan majority rather than the majority party of the Legislature.
That said, accountability could be a good reason to start with an algorithm. We can guarantee a computer won’t consider party registration, voting history, or places of residence of politicians. We just don’t put that data in the algorithm. When the computer spits out its first draft map, we can publish that map as our benchmark. We can ask the IRC to document and explain every deviation it makes from that map. Explaining a deviation from the original algorithm draft will be easy if there’s an outright error (“the computer failed to assign White Owl to a district”) or a small geographical fix (“the computer cut these two blocks of Hartford out of District 9, and they only make a difference of 50 people, so we decided to put geographical integrity above exactly equal population”). But if some partisan schemer on the IRC, or maybe even just some jilted lover or business partner, decides to cut her nemesis out of a district, she’ll have to convince at least four other members on the committee to go along with it, keep any dissenters from figuring out her real motivation for the change, and come up with a plausible public explanation for why she rejected a seemingly sensible boundary from the original algorithm.
Algorithm first draft, human adjustments—I like this plan! Boy, if I weren’t going to win a Senate seat, I’d sign up for the IRC! And I’d nominate David, Darin, and a whole chunk of the blog commentariat to join me, since we’re going to understand these essential details better than anyone else in the state!
Hi Cory,
If I were making the program, it would be a javascript app on a webpage. The webpage would show South Dakota in Google Maps, and have toggle-able layers on top of it to show county boundaries, current legislative district boundaries, municipal boundaries, and census tracts. Maybe even precincts and ward boundaries? (These would be radio buttons so that there’s only one kind of boundary visible at any time.)
Then on top of that would be another layer showing the algorithm’s output. The output would be 35 different colors for the different new districts, similar to Brian Olson’s coloring. This layer would be determined by a set of sliders, one for each of the following:
– Equal Population
– Minimum Average Distance to Center of District (e.g. “geographically compact and contiguous”)
– Respect City Boundaries (e.g. “respect communities of interest”, “use municipal boundaries”)
– Respect County Boundaries (e.g. “use county boundaries”)
– Respect Census Tracts (e.g. “use undivided census tracts”)
Also there would be a checkmark for:
– prevent districts from crossing the Missouri River (e.g. “use visible geographic features”)
There are 5 sliders that give those variables their weight. Each slider is a percent, and the total of all the sliders must equal 100%. So, a user could use 20-20-20-20-20-N. (The N means no checkmark on the Missouri River.) Or if they think census tracts are pointless, 25-25-25-25-0-Y. Or put a bit more emphasis on equal population and minimum distance to district center (the only 2 factors Brian Olson controlled for) with 30-30-10-15-5-Y.
Point being, a good algorithm would accommodate all the IRC’s criteria. Each board member plays with the sliders on the website and argues for what looks like the most balanced. Then at the end of the last meeting, the board takes a vote on something like the map from 30-30-10-15-5-Y vs the map from 35-25-15-15-0-Y.
If you can’t tell, I’m champing at the bit to write this program. Would it be too soon to submit an application to the IRC today?
Kind regards,
David
Live sliders—awesome! How long would it take the map to refresh on each slider adjustment?
I don’t think Amendment T allows us to deem census tracts or other criteria “pointless”. We can give them lower priority—and notice that five positive criteria are not listed in any binding order—but we still have to use each “to the extent practicable.” I’m willing to posit that we would not have to throw out any of the criteria to make a mathematical mapping work. We would just have to tell the computer some priority scheme, such as…
Notice that in that draft scheme, I’m not saying compactness doesn’t matter. We still need to avoid long, skinny, snaky districts and portions of districts to ensure gerrymandering does not sneak back in. But when conflicts between criteria happen, we have to have a way to resolve those criteria. Within that scheme, we could also impose a meta-criterion that the algorithm should seek a mapping that minimizes criterion conflicts.
David, yes, it is too early to apply, since no one is authorized to take apps until we pass T. But start lining up some hard-core computer and debate nerds, because clearly computer and debate nerds are the best people for this job.