Background

The Wisconsin Supreme Court enjoined the further use of Wisconsin’s existing state legislative maps in any future election when it ruled on December 22, 2023 that the current maps are unconstitutional because they include noncontiguous districts. The state constitution requires “[Assembly] districts to be bounded by county, precinct, town or ward lines, to consist of contiguous territory and be in as compact form as practicable” (Wis. Const., art. IV, sec. IV).

Previously, this provision had been interpreted to allow physically separated sections of a single municipality to be placed in a single district, even if this meant that the district, taken as a whole, lacked contiguity.

The 2023 ruling rejected this flexible definition of contiguity in favor of a strictly literal rule. “[F]or a district to be composed of contiguous territory, its territory must be touching such that one could travel from one point in the district to any other point in the district without crossing district lines.” Clarke v. Wisconsin Elections Commission, 2023 WI 79, ¶ 66. Literal islands don’t violate this requirement. “A district can still be contiguous if it contains territory with portions of land separated by water.” Id., ¶ 27.

Extent of noncontiguity

Seven proposed redistricting plans were submitted to the Supreme Court on January 12, 2024, as discussed in an earlier post. One of them—that of the Democratic Senator Respondents—includes several Assembly districts with noncontiguous land areas.

In an Expert Report submitted to the Wisconsin Supreme Court in support of the remedial maps proposed by the Democratic Senator Respondents, Kenneth Mayer maintains that these districts should not be counted as noncontiguous. “The remaining cases of apparent non-contiguity stem from ward fragments resulting from errors in the underlying Census data, which the parties have agreed should not be counted as noncontiguous” (p. 7 of the report). I discuss below whether this data stipulation has any relevance to territorial contiguity.

The noncontiguous Assembly districts are the 44th, 45th, 47th, 48th, 91st, 92nd, and 98th. In total, the unconnected segments include 34 census blocks, of which two are populated according to the 2020 census count. The populated blocks are “550250008001000” in AD48 (pop. 88) and “550350008021018” in AD92 (pop. 14). The remaining blocks are generally tiny strips of land and in any event are unpopulated.

The redistricting plans submitted to the Wisconsin Supreme Court are defined by block assignment files—spreadsheets which list each census block in Wisconsin, along with the district to which it is assigned. I obtained the block assignment files for this plan from links shared by Senate Minority Leader Hesselbein on X on January 12^th.

I measure district noncontiguity by matching the block assignment files to the Census Bureau’s block GIS file. Then, I measure the adjacency of each census block and identify the components of the resulting network. Replication code is available here.

The following graphics show the noncontiguous sections of each Assembly district. Each block in the main component of the district is outlined in blue. Any disconnected blocks are outlined in red. Click here to view a web page with interactive web maps.

The Joint Stipulation

On January 2, the parties to the case filed a joint stipulation describing a series of deviations, agreed to by all parties, from the official redistricting data.

The stipulation mainly deals with 216 ward fragments, themselves including nearly 300 blocks, with incorrect ward or municipality labels. The incorrect labels seem to originate with the original Census Bureau data. This matters because the parties don’t want to be penalized for splitting a ward (or municipality) when, really, it is the label that is incorrect. In the stipulation, the parties agreed on a consistent way to correct and handle these ward and municipality designations.

The position of the Democratic Senator Respondents is that this stipulation also means that the specified blocks should not be counted as physically noncontiguous with the rest of the district.

Here are the relevant paragraphs from the Joint Stipulation. In his report (p. 7 & n.6), Mayer cites paragraphs 8 and 9. I also include paragraph 7, which may be relevant.

7. All parties agree that the Franklin ward and the 215 additional ward fragments identified in Appendix A do not reflect true municipal-ward “islands,” that is, noncontiguous territory which is separated by the territory of another municipality from the major part of the municipality to which it belongs, see Wis. Stat. § 5.15(1)(b), (2)(f)(3).

8. All parties agree that detaching any of the 216 ward fragments identified in Appendix A from the rest of the ward to which it is assigned in the August 2021 Redistricting Dataset will not count as a ward split when evaluating a proposed remedial map.

9. All parties agree that any of the 216 ward fragments identified in Appendix A will be considered part of the municipality to which the August 2021 Redistricting Dataset and the 2020 Census Redistricting Data assigned it, regardless of whether that assignment may have been due to error in the U.S. Census data, when evaluating a proposed remedial map.

Paragraphs 8 and 9 address whether or not incorrectly labelled blocks should be counted as ward or municipality splits. This has nothing to do with physical contiguity.

The meaning of paragraph 7 is less clear to me. What does it mean to say that the ward fragments are not actually “separated by the territory of another municipality from the major part of the municipality to which it belongs”? What if the ward fragment is separated by another district from the main body of the district to which it belongs?

One thing is clear. The Joint Stipulation does not dispute the actual physical location of any of the census blocks. Using the Census Bureau’s census block GIS file to draw maps defined by the Democratic Senator Respondents’ block assignment files will still result in situations where landbound portions of some districts are unconnected to the rest of the district.

In December 2023, the Wisconsin Supreme Court threw out the existing state legislative maps. If the state legislature and governor cannot agree on new maps, the Court ruled that they would choose a remedial map from among a list of submissions.¹ Those submissions were due to the Supreme Court on January 12, 2024.

The Court’s majority opinion described a set of criteria that they would use to evaluate the maps. However, they did not define specific metrics. Each party submitting a map was free to choose their own metric when arguing why their plan is best. There are many empirically legitimate ways to measure concepts like compactness and partisan balance. The most important thing is that the measures be applied to each plan in exactly the same way. That is what I have done in the analysis below.

To analyze these plans, I have constructed an open source repository of code and data available here. That repository generates the scorecards below, along with many more metrics. For instance, this file includes the results of each presidential, gubernatorial, US senate, attorney general, and state treasurer race in each proposed district from 2012-2022. I encourage interested readers to explore the complete resource for themselves.

Update 1/18: On January 17, the Court rejected the Petering submission because Matthew Petering was not among the original parties to the case.

Assembly

Population Deviation

This measures how well each plan achieved equal populations across districts. It is the range between the most and least populous districts, divided by the ideal district size. Lower numbers are better. The new submissions fall into two camps. The Law Forward, WILL, and Legislative Republican plans each have population deviations close to 1%. The rest of the plans have deviations closer to 2%.

Majority Minority Districts

Federal law requires that, under certain circumstances, districts be drawn where minority voters have the ability to elect candidates of their own choosing. In Wisconsin, that has generally meant 5 districts where Black adults make up a majority and 2 majority Latino districts. All the plans here accomplished that, with the exception of the Senate Democrats plan, which created a sixth majority Black district.

Contiguity

This is one is a simple pass/fail requirement. Every component (census block) of a district must be touching the rest of the district, with the exception of literal physical islands. The old maps were thrown out in December because the Supreme Court chose to interpret the Wisconsin Constitution’s “contiguous territory” requirement literally.

Geographic Splits

It’s impossible to draw contiguous, equal-population districts without splitting counties, municipalities, and wards. But the Court prefers plans to split fewer of them, if possible. Here I measure the number of municipalities, counties, and wards split into multiple districts, which themselves cross into other municipalities or counties. (I don’t count districts which lie entirely within a larger municipality or county).

Across all geographies, the plan submitted by Legislative Republicans has the most splits. WILL’s plan splits the fewest municipalities and counties, but more wards. The Wright Petitioners’ plan splits no wards. These ward split totals do not include splits in wards which the parties agreed not to count in a joint stipulation filed on January 2nd.

Compactness

There are many ways to measure compactness, but they generally show similar rankings. See this table for multiple measures of compactness. In the scorecard below, I show each plan’s average Reock score. The Reock score is the ratio of the district’s area to the area of the smallest circle than can be drawn around the district. A score of 1 equals perfect compactness, so higher numbers are better.

Five of these seven plans have closely clustered Reock scores, between 0.36 and 0.39. The WILL plan achieves an average Reock score of 0.41, and the Petering (FastMap) plan a score of 0.44.

Partisan Balance

The partisan balance of a district could be measured in many ways. Here, I use a statistical model to predict the results of the 2022 state legislative election had they taken place in these proposed districts. I show both the likely number of seats won by each party, as well as the partisan lean of the 50^th seat (determining majority control).

The most Republican-leaning plan is, unsurprisingly, the plan submitted by the Wisconsin Legislature. It creates 35 Democratic seats to 64 Republican seats. Democrats would need to carry the state by 16.3 points (58.15% of the two-party vote) in order to win a majority of the Assembly.

The WILL plan slightly softens this Republican advantage, creating 39 Democratic seats to 60 Republican.

All the other plans are more favorable to the Democrats. The most Democratic-leaning plan is the one submitted by Law Forward, which would create (per my model), 49 Democratic seats to 50 Republican ones. Democrats would need to win the state by 3.3 points (51.65%) in order to win a majority.

Senate

The scores for the Senate maps are generally similar. Recall that each Wisconsin State Senate districts consists of three adjacent Assembly districts.

Notably, the Senate Democrats plan and the Petering (FastMap) plans both create outright Democratic majorities, according to my model of the 2022 election. Of course, only even or odd-numbered elections take place in a given election year. The 2024 State Senate elections will cover only the even-numbered districts. Of those 16 seats, 6 are currently held by Democrats and 10 by Republicans. The Wright plan includes the most Democratic-leaning even-numbered seats (10), followed by the Evers and Law Forward plans (9 each).

Edit: The plan submitted by the Senate Democrats contains 7 Assembly and 2 Senate districts which are not contiguous according to the Census Bureau’s Census Block GIS file. However, a joint stipulation filed by the parties states that the disconnected census blocks should not actually be counted as noncontiguous. I have accordingly edited these scorecards and this blog post to reflect this. I have also removed these blocks from the list of wards splits in all plans, and I have corrected an error in the calculation of Reock Scores.

Edit 1/18: I have added a footnote explaining the Court’s 1/17 decision to exclude the Petering map.

1. If none of the proposed remedial maps satisfy the Court, they may request that their hired consultants draw one as well. ↩︎