Which would you rather have, a big, expensive toaster that only made toast one way and sometimes burnt the toast, or a smaller, more efficient, less costly toaster that could be adjusted to make toast exactly as was needed each time?
Seems like an easy question. Less costly and more efficient beats expensive and inefficient every day, and twice on Sundays.
Arizona had one of the most successful elections of the 2020 cycle. In the face of a global pandemic and record-breaking turnout, the State delivered a safe and secure election. For more than a decade, Arizona has used paper ballots and post-election audits as a way of assuring the integrity of the vote. It worked well in 2020—but it can work even better.
The Arizona Legislature is considering reforms to the election system. The bills currently under consideration contain a number of laudable proposals. One that merits universal approval is the concept of a “risk-limiting audit” or RLA. RLAs are generally the less costly, more efficient toaster of the election system, and they merit bipartisan approval. Everyone wants to be sure that the election results are accurate. As we outlined in a lengthy report earlier this year , RLAs are a cheap and effective way to do that.
What are RLAs and how do they work? Put simply, they’re a way of sampling the vote after an election so that you can know that the reported results are accurate. As we wrote in the report, they are “an easy and efficient method for verifying the accuracy of unofficial election outcomes (i.e., winners and losers). Through hand counting a statistically meaningful sample of paper ballots cast in an election, an RLA can provide confidence that the election outcome was correct.”
These audits give the public confidence that the hardware, software and procedures used to tally votes found the real winners. In other words, if someone is worried that the vote might have been hacked, or that the machines used have been tampered with, the RLA can put those concerns to rest—or verify them if they are true. That’s why the National Academies of Sciences, Engineering, and Medicine , election officials , statisticians  and cybersecurity experts  have all endorsed these types of audits.
To function, RLAs use statistical methods to determine how many paper ballots to sample in order to verify that the reported election results are accurate. In plain terms, that means that more ballots are counted in a close race, while a race with a larger margin of victory would require fewer ballots to be counted. We don’t need to make a lot of toast, if you will, if the reported result is 80-20—a small sample can tell us that. By contrast, if the results are close—as they were in some elections in Arizona in 2020—we sample more ballots.
If testing of the sample is consistent with the original reported vote total, it is almost certain that the initially declared winner actually won the race. If, on the other hand, the sample has substantial discrepancies with the original tally, the audit continues until the evidence becomes clearer. In some, rare, cases that may mean that the audit is continued until all the ballots have been manually counted.
Today, the audits in use in Arizona always require review of the same percentage of ballots when the race is close and when the election is a landslide. That’s not cost-effective. It’s spending money that sometimes isn’t necessary. And it isn’t efficient—it’s putting a lot more work into the audit than might be required to be sure of the result.
And, as we noted, RLAs aren’t partisan and they aren’t controversial. Pilot projects for risk-limiting audits are underway in Nevada , and Georgia  recently implemented RLAs after conducting pilots for over a year. RLAs are also in use in Rhode Island , Colorado  and Virginia . Those states are getting accurate results—and they are generally getting them at a lower cost.
While, in the end, we think that Arizona should make RLAs universal for all of its voting, we also think that it is right to be cautious with something as important as an election. And so, we offer a modest proposal. The bill pending in the Arizona Legislature should be modified slightly to give election officials the same option that they now have in Ohio —the option to choose between a traditional full hand count audit, and an RLA. With that freedom, election officials could select the option that better suited their needs and Arizona could, gently, begin to implement the RLA system, developing experience along the way.
And, in the end, we might just find ourselves with a more efficient, cheaper way of making election toast. Wouldn’t that be something?
Image credit: Steve Heap
- “we outlined in a lengthy report earlier this year”: https://www.brennancenter.org/sites/default/files/2021-01/AZ%20RLA%20WP%20-%20FINAL2%20-%2002.01.21.pdf
- “National Academies of Sciences, Engineering, and Medicine”: https://www.nationalacademies.org/news/2018/09/securing-the-vote-new-report
- “election officials”: https://coloradosun.com/2020/11/16/2020-election-risk-limiting-audit-colorado/
- “statisticians”: https://www.stat.berkeley.edu/~stark/Preprints/gentle12.pdf
- “cybersecurity experts”: https://www.brennancenter.org/sites/default/files/2019-11/2019_011_RLA_Analysis_FINAL_0.pdf
- “Nevada”: https://www.leg.state.nv.us/App/NELIS/REL/80th2019/Bill/6130/Text
- “Georgia”: https://sos.ga.gov/index.php/elections/historic_first_statewide_audit_of_paper_ballots_upholds_result_of_presidential_race
- “Rhode Island”: https://elections.ri.gov/elections/results/RLA.php
- “Colorado”: https://coloradosun.com/2020/11/16/2020-election-risk-limiting-audit-colorado/
- “Virginia”: https://www.elections.virginia.gov/resultsreports/election-security/rla/
- “that they now have in Ohio”: http://codes.ohio.gov/orc/3505.331