A flawed new version of Risk Limiting Audit bill moves forward in the General Assembly

Last month I wrote about S.B.472 for Risk Limiting Audits, despite all but unanimous testimony that the bill was highly flawed, a flawed substitute was passed shortly thereafter by the General Administration and Elections Committee. Last week that revised bill became available on the web, allowing public scrutiny. There is a Word version and a .pdf version, the .pdf version has a fiscal note and an Office of Legislative (OLR) analysis  (Read it here.)  Read our testimony on the original bill and that of others (Here.)

It is revised from the original with some improvements, yet remains ambiguous and confusing, although now it may be closer to an actual Risk Limiting Audit. I question if the person(s) who wrote this updated bill had or took advantage of review by anyone familiar with the science of Risk Limiting Audits. Considering the fiscal note and the OLR Analysis, it seems to me that the description in some cases reads things into the bill that do not seem to be reflected in the bill. In particular, it assumes costs and roles for UConn that are not clear in the bill. Costs and roles that would be consistent with the prototype last year, but not obvious in the bill that would become law.

There are some welcome changes:

  • This version delays a full RLA until 2024 rather than 2022. This will allow time for planning, especially important since we will have a new Secretary of the State (SOTS) in January 2023. This also provides time for future revisions to the law prior to Nov 2024.
  • This version makes a complete new section for RLAs from the existing and continuing 5% audit which still applies to all primaries and odd-year elections. This makes the bill easier to read and perhaps contributes to its improvements, clarity and perhaps facilitating clarity in future improvements. Yet it copies text that is appropriate to the existing audits but not for RLAs.
  • It seems to have better, if imperfect, definitions of RLAs, closer torecognized standard definitions.
  • It arguably requires that manifest creation be open to the public.

Yet, in my reading, there remain new and old problems:

  • It continues the lack of specific notice requirements. In every very case where something is open to the public, there should a clear notice requirement specifying how far in advance each event should be notices (e.g. 48 hours) and where the notice will appear (e.g. a specific place on the SOTS Web.)
  • Like the original it has no requirements for close public observation, software independence, and public access to information critical to verifying the audit. It should clearly require standard, recognized statistical evidence, the review of samples of actual paper ballots, and public access to algorithms and data necessary to independently verify the results of the audit.
  • It continues the ambiguous requirements that UConn report on the audit, send the report to the SOTS and the SOTS file it with the State Elections Enforcement Commission (SEEC). The weakness is that the current law and this bill provide no deadline for such filing. Many reports of audits since 2008 have yet to be published by UConn, published by the SOTS, or filed with the SEEC. Also in practice these are not independent reports. The UConn Voter Center is dependent on the SOTS for its budget and in practice draft reports are reviewed by the SOTS Office and revised by UConn before they are accepted.
  • It continues the ambiguous language that candidates can use evidence from the audit to challenge an election in court until seven days after the audit is complete, yet provides no definition of completion of the audit. It would seem that an audit is not complete until a report is filled and published by the SEEC. If so, given history, candidates could challenge an election for years after. See Sec. 4 Section 9-323 (page 11) and Sec. 5. Section 9-324 (page 13) and Sec. 6. Section 9-328 (page 15) or Sec. 7. Subsection (a) of section 9-29a (page 16).
  • It requires that presidential electors, all statewide offices including one U.S. Senator, a randomly chosen Congressional race, and 5% of General Assembly offices be audited in even years. Precluding audits of any other offices in even years, such as Judge of Probate and Registrar of voters from any audit.
  • It continues the current flaw in the law that exempts questions from audit. I wonder if anyone would have any concerns if the Early Voting or No-Excuse Absentee Voting questions were to pass or fail by small margins? Would those be concerns be magnified realizing they were exempt from the audit?
  • Recanvassed races (CT’s hand and machine recounts) are included in the RLAs. General Assembly races that have been recanvassed are automatically included in the 5% that are subject to RLA. I wonder if they realize that such races, since they are close, will often end in a complete hand recount? If the intent is to hand-recount all such races, why not do the hand recount instead of the recanvass?
  • The timing of ballot manifest creation seems to be part of the process of the initial close of the polls on election night and is specified as open to the public – an improvement over the original bill. However, other laws allow counting to be interrupted and completed within 48 hours, without any public notice requirements for the date, time and location of where counting will resume.
  • While the non-binding Fiscal Note and OLR Analysis added below the bill seems to indicate that UConn will be part of conducting the audit, the bill mirrors the requirements of the existing 5% audit, where the SOTS creates the procedures for the RLA, the Registrars conduct the RLA submitting results to UConn, who submits a report to the SOTS.
    • It is hard to imagine the Registrars (two independent D and R election officials in each of 169 towns) coordinating a state wide RLA, including escalation, and results without some body coordinating the state wide calculations, decision making, and declaring the final result.
    • It seems the Registrars determine if the audit meets the risk limit or not and then based on that the SOTS shall order a manual recount. See Section 1 (e).

Perhaps some of these concerns can be cured/overridden by procedures issued by the SOTS. See Section 1 (d). However one should question if the SOTS can write binding procedures, opinions, or declarations which contradict the text of state law.


Leave a Reply

You must be logged in to post a comment.