United States: Even With Reprieve, Census Delays Add Stress To 2021 Redistricting

On March 15, the U.S. Census Bureau (the Bureau) slightly backtracked from its bombshell announcement in February that Public Law 94-171 redistricting data would be delayed until as late as Sept. 30. The March 15 announcement was that the Bureau would now release a so-called legacy format summary redistricting data file in "mid to late August." What the Bureau will release by Sept. 30 will only differ by (a) the addition of some summary tabulations and (b) a user-friendly front end.¹ In essence, the Bureau is promising to publish, in August, the redistricting data in a format that is the same as or similar to the format in which it published the data in 2011. This means that vendors will be able to access the data, as they did in 2011, to provide summaries of information the same day the data is released. Loading this information into any specialized software may take a bit longer. One challenge is that data for all states will be released on the same day via the web, so resources for vendors will be stretched to the limit. Although undeniably helpful, a "mid to late August" publication date — five months late — still puts into jeopardy state statutory and/or constitutional deadlines to complete congressional or state legislative redistricting. What led to this delay? And what options do states and their political subdivisions have?

The Delay

It should come as no surprise that the administration of Census 2020 suffered from delays related to the COVID-19 pandemic. Well-laid plans — that had been debated and established throughout the decade — were cast aside to deal with the suspension of field operations in the spring and summer of 2020.

At several times over the past few months, the Bureau signaled a delay in the release dates for the two datasets critical for the apportionment of the seats in the U.S. House and for the revision of district lines for Congress, state legislatures and thousands of other representational districts across the nation. The Bureau confirmed that delay in mid-February when it announced a plan to deliver apportionment numbers by April 30 and indicated the redistricting data (known as the PL 94-171 data) would be available by Sept. 30, six months after the statutory deadline of March 31, 2021.² The Census also announced a change from prior decades in that it would not release redistricting data on a rolling basis (i.e., releasing data on a priority schedule to accommodate states with earlier redistricting deadlines), but would instead release data for all the states at the same time. The Bureau attributed that change to "COVID-19-related shifts in data collection and in the data processing schedule ...."³

With the March 15 announcement, what had been a six-month delay will now be a five-month delay. The impact of this several-month delay will be felt around the country as states scramble to reconfigure their congressional and state legislative district boundaries in time for the 2022 election calendars and state-mandated deadlines related to the revision of districts. But it creates even more significant problems for the 14 states whose legislative districts and 12 states whose congressional districts must be redrawn in 2021.⁴ Some of the states in that bucket — such as Maine, Iowa, Colorado, Connecticut, Ohio and Vermont — even have state constitutional or statutory deadlines to redistrict before or around the same time as the "legacy" PL 94-171 data files are set to be released. Other states, such as Virginia, have elections in 2021 but will not be able to redistrict in time for those elections.

So, why the delay?

Differential Privacy

The decision to release all 50 states' PL 94-171 data simultaneously is an impactful one, as production of PL 94-171 files on a "rolling" basis might mitigate the effect of the delay on states with early redistricting deadlines. Once the state-level counts are determined for apportionment purposes, the counts in the PL 94-171 datasets should already be determined as well. Users might logically expect that the first datasets would be available shortly after the release of the state-level counts, e.g., during May or June 2021. However, many users may not know that there will be a significant change in the block-level redistricting data that users will receive.

New for this Census, the Bureau is implementing a process known as Differential Privacy through its Disclosure Avoidance System (DAS). This project grew out of a Bureau determination that because the data it releases is so granular, there was a risk that Census datasets could be used by data scientists employing supercomputing resources to deconstruct these, and other census datasets, to obtain personally identifiable information. To prevent that deconstruction and protect privacy, the Bureau decided to make the 2020 PL 94-171 datasets the first in history that will not contain population counts from the actual enumeration, with the sole exception of the count for the entire state. Every other record at lower levels of geography will have statistical noise added, and the users will never know the actual count.

During a March 2020 presentation to the NCSL, a Census official explained a belief that Differential Privacy "allows [the Bureau] to inject a precisely calibrated amount of noise into the data to control the privacy risk of any calculation or statistic. ... Differential privacy allows you to quantify a precise level of 'acceptable risk,' and to precisely calibrate where on the privacy/accuracy spectrum the resulting data will be."⁵ The Bureau indicated in a Feb. 8 conference call that it has finalized its version of the DAS, but that version is not yet known to the public and concerns remain about the degree to which the DAS will introduce biases and geographic dislocations of population into the PL 94-171 dataset. However, the Bureau has avoided discussing publicly the fact that nothing other than actual counts will satisfy most redistricting needs.

The Bureau will implement the DAS, state by state, using its Top-Down Algorithm. There are several passes of the data for each state. Of course, the Bureau needs to process every state consistently. It is unknown how long it will take to process any particular state, let alone all states. There were 11 million census blocks in 2010, though this has been reduced substantially for 2020 to about 8 million. With the large number of breakouts in the record (288 separate fields of race and Hispanic origin) and the large number of records, a lot of processing time will be required because of the nature of the algorithm and the steps involved. If the Bureau discovers problems after a review of live data, which can, presumably, only begin after the apportionment counts are complete, it may have to pause, fix the code and start the entire process again. If these potential hiccups are built into the timeline, and reruns are not needed, it is possible the data might be released earlier.

What Options Do States Have?

Even assuming the Bureau delivers reliable data that states can load into their databases and begin working on right away (which is not guaranteed⁶), a mid-to-late August delivery date makes it impossible to redistrict in time for November 2021 elections, and will present a significant challenge to completing redistricting in time for 2022 primaries. This delay will, unfortunately, leave limited options for those jurisdictions with 2021 elections or with constitutional redistricting deadlines.

A jurisdiction could simply postpone 2021 elections into 2022, and/or delay 2022 primaries, to allow more time to redistrict. Or it could hold its 2021 elections under existing lines and conduct future elections under a new plan.⁷

Federal case law appears to give states flexibility to craft solutions for these kinds of delays.⁸ Specifically, federal courts have been reluctant to find Equal Protection Clause or Voting Rights Act (VRA) violations where census or legislative delays caused missed deadlines that required elections held in the first year or so of a new decade to be conducted under the prior decade's district lines. Citing to language in Reynolds v. Sims, 377 U.S. 533, 583-584 (1964) that the Equal Protection Clause only requires a state to have "a reasonably conceived plan for periodic readjustment of legislative representation," those courts have focused the equal protection analysis on the existence of a plan for periodic redistricting, tending not to find that a state's plan violates the law because delays cause the plan to be executed imperfectly. See, e.g., Garcia v. 2011 Legislative Reapportionment Comm'n, 938 F. Supp. 2d 542, 551-552 (E.D. Pa. 2013) (finding delay in release of "usable census data" a justifiable reason for delay in legislative apportionment, resulting in Pennsylvania conducting 2012 state legislative races under 2001 plan); Clark v. Marx, No. 11-2149, 2012 WL 41926, *10 (W.D. La. Jan. 9, 2012); Political Action Conference of Ill. v. Daley, 976 F.2d 335, 341 (7th Cir. 1992); French v. Boner, 963 F.2d 890, 891-892 (6th Cir. 1992).

A slightly different approach was taken by a Mississippi court faced with an equal protection claim in which the state failed to redistrict in time to hold its 2011 state legislative elections under a new plan. The court held that the use of the old plan in 2011 did not violate the federal constitution or the VRA, but it retained jurisdiction in the event the legislature failed to adopt a new plan by the state constitutional deadline at the end of the 2012 legislative session. Miss. State Conference of N.A.A.C.P. v. Barbour, No. 3:11cv159, 2011 WL 1870222, *9 (S.D. Miss.), aff'd, 565 U.S. 972 (2011). But there are limits to the scope of acceptable reasons for redistricting delays. One court, for instance, rejected New York state's argument that asserted a desire to follow a state "tradition" of redistricting in the third year of a decade as insufficient to justify the delay. Flateau v. Anderson, 537 F. Supp. 257, 265 (S.D.N.Y. 1982). Another court forced Maine to undertake congressional districting in 2012 rather than waiting until 2014, as the state had done for several decades.⁹

If a federal court finds that a redistricting delay violates federal law, the next question would be whether special elections could be ordered as a remedy. The Supreme Court provided guidance on this issue following the 2010 cycle in North Carolina v. Covington, 137 S.Ct. 1624, 198 L.Ed.2d 110 (2017). In Covington, the district court held that 28 majority-minority districts were racial gerrymanders and ordered the legislature to adopt a remedial plan for post-2016 elections. But soon after the 2016 elections, the district court entered an order shortening the normal two-year term for the 28 districts to one year and ordered a special election in 2017 under the new plan. Id. at 1625. The Supreme Court reversed, and held that when a district court considers imposing a special election as a remedy, it must undertake a "careful case-specific analysis" balancing such factors as "the severity and nature of the particular constitutional violation, the extent of the likely disruption to the ordinary processes of governance if early elections are imposed, and the need to act with proper judicial restraint when intruding on state sovereignty." Id. at 1625-1626. A court's decision not to impose a special election could leave a plaintiff without an effective remedy, which could raise a standing issue (standing requires that an injury be redressable).

The March 15 announcement was included in a responsive pleading in the recent litigation filed by the state of Ohio seeking to compel production of the PL 94-171 datasets by the statutory deadline. Alabama has also filed suit against the Census to challenge the delay as well as the use of Differential Privacy. But regardless of the outcome of that litigation, and whichever delay-mitigation strategies states choose to employ, one thing is clear: State legislatures, independent redistricting commissions and other redistricting authorities will have their hands full come Q4. They should consider taking the time in advance to prepare, as much as possible, for a busy and compressed redistricting cycle. And in jurisdictions where the delay will prevent compliance with constitutional deadlines, options should be explored to seek extensions of those deadlines through legislative or judicial action.

Authorship Credit: Patrick T. Lewis and Clark H. Bensen*

* Bensen is the founder of Polidata, LLC.

Footnotes

1 U.S. Census Bureau, U.S. Census Bureau Releases Statement on Release of Legacy Format Summary Redistricting Data File, Rel. No. CB21-RTQ.09 (Mar. 15, 2021), at https://www.census.gov/newsroom/press-releases/2021/statement-legacy-format-redistricting.html.
2 U.S. Census Bureau, Census Bureau Statement on Redistricting Data Timeline, Rel. No. CB21-CN.14 (Feb. 12, 2021), at https://www.census.gov/newsroom/press-releases/2021/statement-redistricting-data-timeline.html.
3 Id.
4 A complete list of state redistricting deadlines is published by the National Council of State Legislatures (NCSL). NCSL, State Redistricting Deadlines, https://www.ncsl.org/research/redistricting/state-redistricting-deadlines637224581.aspx.
5 Michael Hawes, Differential Privacy and the 2020 Decennial Census, U.S. Bureau of the Census (Mar. 5, 2020), at https://www2.census.gov/about/policies/2020-03-05-differential-privacy.pdf.
6 Upon release of the PL 94-171 datasets for the 2010 Census, errors of geographic misassignment or miscoding, some substantial, were noted in several states. See https://www.census.gov/programs-surveys/decennial-census/data/errata-notes.html.
7 In such a situation, there may be a challenge under one-person, one-vote, as the districts might have been used for more than a "term of ten years." See U.S. Const. art. I, §2, cl. 3, as amended by U.S. Const. amend. XIV, §2,
8 Notably, however, some state courts — including in North Carolina and Pennsylvania — have recently applied state constitutional provisions giving a right to "free and equal elections" to invalidate conduct that federal courts had declined to strike down under the federal constitution.
9 See Desena v. Maine, 793 F. Supp. 2d 456 (D. Me. 2011) (finding constitutional violations "should be cured at the earliest practicable date").