Skip Navigation
Report

A Review of Robust Post-Election Audits

Key Point: Risk-limiting audits are an effective tool that can be implemented in Michigan, by Michigan election officials, using Michigan-certified voting systems.

Published: November 7, 2019

i. Introduction

In August of 2018, the Michigan Bureau of Elec­tions and the city clerks of Kala­ma­zoo, Lans­ing, and Rochester Hills partnered with the Bren­nan Center for Justice, Professor Ron Rivest and Mayuri Srid­har of MIT, Dr. Philip Stark and Kellie Otto­boni from the Univer­sity of Cali­for­nia, Berke­ley, Jerome Lovato of the U.S. Elec­tion Assist­ance Commis­sion, Veri­fied Voting Found­a­tion, and the Voting System Tech­nical Over­sight Program at Ball State Univer­sity (the “RLA Team”) to conduct the first risk-limit­ing audit (“RLA”) pilot in Michigan.

While the main goal was to provide Michigan elec­tion offi­cials with a hands-on learn­ing exper­i­ence about RLAs, the “gold stand­ard” of post-elec­tion tabu­la­tion audits, parti­cipants gained broadly-applic­able insights about best prac­tices. We hope this report can serve as a resource to elec­tion offi­cials across the coun­try who are consid­er­ing adding RLAs to their elec­tion secur­ity proced­ures or conduct­ing similar pilots. We have designed it to be useful for elec­tion offi­cials with vary­ing levels of famili­ar­ity with RLAs.

The exec­ut­ive summary and the project over­view provide a high-level over­view of the project, intro­duce the differ­ent audit meth­ods used, and list the voting systems and RLA tools used in each juris­dic­tion. A risk-limit­ing audit defin­i­tion and a chart that summar­izes the pilot audit results in each juris­dic­tion are also provided.

Next, elec­tion offi­cials offer their insights on the process, the pilot and their goals moving forward. State elec­tion offi­cials provide the back­ground to this unique part­ner­ship and add their perspect­ive on the project. Import­antly, they discuss lessons learned through pilot­ing differ­ent audit meth­ods and random ballot selec­tion meth­ods, and how they plan to move forward in the future.

Local elec­tion offi­cials then provide a detailed over­view of their indi­vidual exper­i­ences during the pilot. Each discusses their work­load, commu­nic­a­tion strategy, imple­ment­a­tion strategy and lessons learned. They also provide recom­mend­a­tions to state elec­tion offi­cials.

Finally, two members of the RLA Team describe the found­a­tional math­em­at­ics, specific proced­ures, sampling meth­ods and RLA soft­ware tools.

If you have addi­tional ques­tions about the proced­ure, the pilot, or any other aspect of this part­ner­ship, as we hope you do, please do not hesit­ate to contact anyone involved.

ii. Executive Summary

State and local elec­tion offi­cials serve as our demo­cracy’s last line of defense against malevol­ent foreign actors, equip­ment malfunc­tions, and human errors which may impact elec­tion results. Unfor­tu­nately, this respons­ib­il­ity often comes with few resources and heightened public scru­tiny. One smart and effect­ive tool avail­able to elec­tion offi­cials facing this real­ity is the post-elec­tion risk-limit­ing audit (“RLA”).

This special type of audit uses stat­ist­ical meth­ods and a manual review of paper ballots to check the accur­acy of repor­ted elec­tion outcomes. Specific­ally, RLAs are designed to provide assur­ance that the repor­ted winner did in fact win the elec­tion, or in the altern­at­ive, to correct errors caused by cyber­at­tacks, bugs, miscon­fig­ur­a­tion, or human error, if any combin­a­tion of those altered the repor­ted outcome. While the under­ly­ing math may be chal­len­ging for non-math­em­aticians to under­stand, the proced­ures to conduct such audits were shown to be straight­for­ward.

In August of 2018, the Bureau of Elec­tions agreed to part­ner with the RLA Team to conduct the first risk-limit­ing audit pilots in Michigan. The RLA Team worked directly with elec­tion offi­cials to under­stand the relev­ant elec­tion admin­is­trat­ive proced­ures and prac­tices. They used this inform­a­tion to draft audit instruc­tions (called audit proto­cols) for each local­ity. Each parti­cip­at­ing muni­cip­al­ity had one day to conduct their pilot between Decem­ber 3–5, 2018.

This approach produced many prac­tical lessons. Most import­antly, Michig­anders gained confirm­a­tion that RLAs are possible in their state, which relies primar­ily on precinct-based voting on elec­tion day. Many other lessons related to the proced­ure, voting systems, and messaging were drawn from this ground­break­ing collab­or­a­tion. These lessons include:

  • Risk-limit­ing audits are an effect­ive tool that can be imple­men­ted in Michigan, by Michigan elec­tion offi­cials, using Michigan-certi­fied voting systems.
  • Risk-limit­ing audits are not proced­ur­ally diffi­cult to imple­ment by elec­tion offi­cials.
  • Risk-limit­ing audits can be imple­men­ted with minimal changes to pre-elec­tion and Elec­tion Day admin­is­trat­ive proced­ures.

Over­all, Michigan elec­tion offi­cials were impressed with the results from the pilots, espe­cially the poten­tial to greatly improve post-elec­tion audit effi­ciency. While further work and addi­tional pilots are neces­sary, this pilot equipped elec­tion offi­cials with prac­tical inform­a­tion neces­sary to make import­ant elec­tion secur­ity policy decisions in Michigan. They currently plan to take the pilot to the next step by conduct­ing RLAs at the county level after the May 2019 elec­tions.