Background

The Wisconsin Supreme Court has accepted a request to take up a case challenging the existing state legislative maps. Petitioners in the lawsuit make several arguments. Some of the arguments ask the court to find that “partisan gerrymandering violates” the state constitution’s “guarantee of equal protection,” as well as “free speech and association rights” and the “Maintenance of Free Government provision.”

Less ambitiously, perhaps, petitioners also argue that the “current legislative districts are unconstitutionally noncontiguous.” This argument, if accepted, could allow the court to throw out the current map without making reference to partisan advantage.

The relevant section of the Wisconsin Constitution is Article IV, Section 4, quoted below in its entirety (emphasis mine):

“[As amended Nov. 1881 and Nov. 1982] The members of the assembly shall be chosen biennially, by single districts, on the Tuesday succeeding the first Monday of November in even-numbered years, by the qualified electors of the several districts, such districts to be bounded by county, precinct, town or ward lines, to consist of contiguous territory and be in as compact form as practicable.”

Likewise, Senate districts are required to be composed of “convenient contiguous territory.”

Every square inch of land in Wisconsin is part of a city, town, or village, and many of those municipalities aren’t contiguous. In 1992, a federal court ruled that “literal contiguity” was not required by the state constitution. The judges wrote, “. . . [F]or the achievement of perfect contiguity and compactness would imply ruthless disregard for other elements of homogeneity; would require breaking up counties, towns, villages, wards, even neighborhoods. If compactness and contiguity are proxies for homogeneity of political interests, so is making district boundaries follow (so far as possible) rather than cross the boundaries of the other political subdivisions in the state.”

In other words, unconnected pieces of the same municipality could be placed into the same Assembly district. The 1992 ruling prioritized keeping municipalities intact over strict geographic contiguity.

In the early 1990s, it was Republicans, not Democrats, arguing for the enforcement of literal contiguity. Now, the roles are reversed. And, of course, the Wisconsin Supreme Court—not the federal district court then or any other federal court—has final authority as to the meaning of the Wisconsin constitution.

How many districts are noncontiguous?

Thus far, the specific contiguity-related statistics I’ve read have all been provided by the petitioners, who maintain that 55 Assembly and 21 Senate districts are noncontiguous. In support, their Memorandum of Law simply links to the legislative district maps provided by the Wisconsin Legislative Technology Services Bureau (see footnotes 1 and 21).

The remainder of this blog post is an explanation of an independent and repeatable method for measuring district contiguity across the state. My calculations yield slightly different results from those provided by the petitioners. While we both identify the same 21 Senate districts as being noncontiguous, I only find 52 Assembly districts with unconnected pieces.

Measuring contiguity

The basic building blocks of the current Assembly map are the census blocks used in the 2020 census. There are 203,059 total census blocks in the state of Wisconsin. Of those, 58,111 are unpopulated. Here’s what the statewide census block map looks like. We would need to zoom way in to clearly see the individual blocks in more populated areas. You can download this GIS file directly from the US Census Bureau here. It’s important to use the official Census Bureau GIS data because they are high quality and complete.

Notice the census blocks that just cover water. These aren’t actually assigned to an Assembly district, but everything else is. The files showing which census blocks constitute each state legislative district are available from the Census Bureau here.

They look like this, with a column showing each census block number and another column showing a legislative district assignment. In Wisconsin, this is what a redistricting plan actually looks like.

After joining the Assembly district codes to the census block shapes, we can map all the individual blocks that make up a district. Here is the 18th district, on the west side of Milwaukee and stretching into eastern Wauwatosa. The large square in the center is Washington Park.

See here for more details and access to the source data files.

We could conclude that this district is entirely contiguous using the eye-test alone. Sometimes, it’s more difficult to tell. In any case, examining all 132 state legislative districts separately would quickly become tedious. Fortunately, we can also measure this more formally using a set of open-source GIS and network analysis tools.

My code performing this analysis is available here in the files “AnalyzeAssembly.R” and “AnalyzeSenate.R.”

First, I subset all the census blocks in a given district. For each census block, I identify all of its neighboring blocks. Then, I turn that block adjacency list into a graph. Each graph consists of one or more components.

In a fully contiguous graph, such as that of Assembly District 18, there is just 1 component.

The 88th Assembly District, on the other hand, consists of 3 components. In other words, there are 3 sections of AD-88 that don’t touch the other sections.

Literal islands

Recall that water-only census blocks aren’t assigned to a legislative district. To accommodate actual islands, I allowed the network analysis to identify graph components using the aquatic blocks as connectors, but I do not plot these blocks themselves. For example, here is the adjacency graph for AD-1 (Door County).

Notice that the offshore islands are assigned to component #1, even though they obviously can’t physically touch the mainland. So this district is fully contiguous.

Summary

You can view my adjacency graphs for each Assembly and Senate district here.

Files showing each census block’s district and network-component-within-district are available here. This directory also includes tables summarizing the results of the network analysis for both the Assembly and Senate, along with accompanying documentation.

In total, 52 Assembly districts and 21 Senate districts are noncontiguous. Of those Assembly districts, 41 include some number of residents in at least two unconnected sections, as do 16 of the Senate districts.

The Assembly district with the largest disconnected population is the 47th, to the south of Madison. According to the 2020 census, 3,737 residents live outside the district’s main component. The 47th district also includes the largest number of census blocks outside the main component, both in total (129) and in populated blocks (61).

Nearby Assembly District 80 contains the largest total number of unconnected components, at 38 (not counting the primary component), although only 418 people live in these areas.

In the State Senate, district 16 has the largest population living outside the main component (2,282). The 27th Senate district contains the most unconnected components (32).

Overall, the 2020 census shows 7,953 Wisconsin residents living outside the main component of an Assembly district and 4,872 in an unconnected section of a Senate district.

This week, Republicans in the Wisconsin State Assembly introduced a bill purporting to establish independent redistricting in Wisconsin. After amendments, it passed with unified GOP support and 1 Democratic vote on Thursday night.

The rushed legislative process and lack of full, public hearings left many aspects of the legislation ill-understood, and many Democrats criticized the political calculus behind the bill’s abrupt timing.

Setting aside the purported motives of legislators, this blog post will consider the specific aspects of Assembly Bill 415, as ultimately amended.

AB415 outlines a redistricting process similar to the one used in Iowa. Here is a simplified description:

• The State Legislative Reference Bureau (LRB) draws legislative maps in consultation with an independent redistricting commission. The LRB maps must be drawn without regard for partisan strength and incumbent addresses. They must contain equal populations, meet federal civil rights requirements, be contiguous, compact, preserve municipal boundaries as much as possible, and use locally-drawn wards as their basic buildings blocks (4.007).
• The legislature must take an up-or-down vote on the LRB maps. They cannot offer any “amendments except those of a purely corrective nature” (p7, l16).
• If the legislature fails to pass the plan, or the governor vetoes it, the LRB tries drawing the maps again, taking into account the other body’s complaints, so long as those complains are consistent with the LRB’s nonpartisan criteria.
• The legislature once again votes on the LRB’s new maps, still without the ability to substantively amend them.
• This process of voting and LRB revision continues until the maps are either passed (and signed by the governor), or until “January 31 of the 2^nd year following the federal decennial census” (Amendment 5).

If no LRB plan is passed by this date, the next steps are unclear.

The text of the bill reads, “No plan may be considered and voted on after January 31 of the 2^nd year following the federal decennial census.” Elsewhere, the bill states, “’Plan’ means a plan for legislative reapportionment prepared under this subchapter.”

In one reading, this prohibits the legislature from considering an LRB plan after January 30, but it does not prevent them from passing a plan not created by the LRB. Does this mean that the legislature could simply run out the clock on LRB plans, before passing one entirely of their own devising?

Of course, this strategy would probably require one party to control both the legislature and the governor’s office. What if the legislature and/or governor reaches an impasse, and is unable to pass any map, whether derived by themselves or the LRB?

The bill is silent on this scenario. Currently, either a federal court or the state Supreme Court draws maps for the state, if the political branches of the government fail. Presumably, that status quo would remain in effect. But, again, AB415 offers no clarity.

The parts of this bill that deal with the LRB criteria for drawing maps strike me as quite good. They lay out widely accepted criteria for drawing fair maps, using reasonably clear definitions. Maps drawn using that guidance would be genuinely neutral; although, Wisconsin’s asymmetrical political geography would still give Republican an advantage in a 50-50 election year.

For me, the crux of this bill rests with that January 31 deadline. What happens if no LRB plan is passed after that time?

Is it true that the legislature can pass whatever they want at that point? If so, this means that under unified government (as in 2011) either party can gerrymander to their heart’s content. Under divided government (as we have now) a court intervenes, following a process of their own devising.

This is hardly different than the status quo, which put Wisconsin in its current predicament.

The drafters of Amendment 5 appear to have made an important concession. Before that Amendment was passed, the bill allowed the legislature to amend the LRB’s 3^rd proposal however they pleased, including replacing it entirely. Obviously, this is unpalatable to the minority party.

Amendment 5 removes the ability of the legislature to substantively amend any LRB plan. But it may replace that ability with a simple deadline, after which the legislature may pass whatever partisan plan it desires.

A better law would clear up this confusion. One option would be specifying that the Wisconsin Supreme Court chooses from among the LRB plans if the January 31^st deadline passes unmet. This would go a long way towards removing partisan bias in Wisconsin’s legislative maps.