Background: What’s Going On?
While Utah citizens fight to get rid of the broken vote counting method called Instant Runoff Voting (IRV) and the infamous Ranked Choice Voting (RCV) pilot; while Utah city councils—under pressure from constituents—choose to go back to regular voting (as in Draper, UT, for instance); while legislators educate themselves about the pitfalls of RCV; while all that is going on, current Utah legislative session generates a whopper of irony: proposal HB205. What is this bill about: It wants to address the “plurality problem” by bringing to you—wanna guess?—IRV! While avoiding terms like RCV and IRV in the verbiage, its sponsors believe they found the solution: a new toy called Contingency Voting. The House sponsor, Rep. Jordan Teuscher, acknowledges that it is not perfect but claims it “very different” from RCV. (In fairness to the sponsors, their other legislative work deserves respect. HB205 is one, strong exception, nevertheless).
In the following I show that HB205 will de-facto establish RCV/IRV in Utah statewide (unlike just for municipalities as the pilot did), with no opt-out (unlike the pilot), and with no time limit (unlike the pilot). While many have been fighting the pilot. Get the irony?
Update: Bill HB205 passed Utah House with 38-36 votes on February 17, 2023, and is headed to the Senate.
Update 2: HB205 never made it out of Senate Rules Committee. Thanks to all who contacted their Senator and/or committee members with concerns. My bet: this bill will pop up in some form again next year, so please read on to prepare for the next legislative session.
Summary: Why is Contingency Voting Harmful
I claim that CV is a version of IRV with all its brokenness. For three-candidate elections, CV is mathematically equivalent to IRV. I will show below how that is true. For 4+ candidates, CV is not necessarily exactly IRV. CV does reduce the number of rounds to 2. However, it suffers from the same set of terrible paradoxical properties as IRV (which we have shown occurred in Utah, see Table 2 of our report, and as occurred in Alaska): non-monotonicity, participation failure, Condorcet failure, etc. This all harms voters’ trust.
RCV ballot format HB205 relies on is identical to one used in the 2021 municipal IRV elections. We observed substantial ballot issues with that (see Table 1 in the report). The gravity of this issue cannot be debated away. Some towns saw 30% of ballots discarded. What price tag do we put on a vote lost? And are we willing to sacrifice large percentages of voters for the sake of addressing a “plurality problem?” This harms voters.
The two-round nature of CV anticipates that County Clerks (except SLCo) will perform a semi-manual step of extracting preferential information from spreadsheets with 100’s of thousands of rows. Either they do this manually, which is not realistic, or they each write/apply a script (a piece of uncertified software) with no specific quality standards. The potential for error due to both coding and handling seems high to me. This may harm voters.
I will not blame anyone for wanting more information, given these claims. Challenge to you: While reading my arguments, please question everything, poke holes! Reach out with questions and comments until you are fully convinced either way. The more so if you happen to be a legislator who wants to vote with a clear conscience.
1. Contingency Voting is IRV in Disguise
Contingency Voting (CV) relies on ranked ballots (just like RCV/IRV). Unlike IRV, however, it “compresses” multiple rounds of IRV into just two. This leads us to the first inconvenient truth:
For Three-Candidate Elections, CV is Identical to IRV
Three-candidate IRV elections have just 2 rounds. So there is nothing to compress for CV. Mathematically speaking, CV is exactly IRV, full stop. Well, IRV is also known to deliver worst of paradoxes for three-candidate elections (read our analysis of the Alaska Special Election in 2022). This means CV, too, will deliver them (since it is IRV). Remember, three-candidate elections are not rare: out of 16 IRV races in Utah County in 2021, 7 were three-candidate (see Table 1 in our technical report).
A damning example: based on Alaska RCV data, I can confirm that CV would fail to elect the Condorcet winner (Nick Begich) as well as would deliver a non-monotonicity failure (“more votes for Mary Peltola would make Peltola lose,” i.e., literally counting backwards!) and, as a bonus, would incur same spoiler (Sarah Palin) - despite IRV supposedly preventing spoilers. (Read what happened in Alaska in 2022).
For those detail-oriented: Here’s why CV=IRV for three candidates. (You can skip if you are willing to trust me). The key fact is highlighted in the red rectangle: with three candidates, determining the loser (IRV) and determining the top 2 candidates (CV) are equivalent operations (i.e., knowing one implies knowing the other) - the rest unfolds from that.
For 4+ Candidates CV Delivers Paradoxes just like IRV
My previous argument should actually suffice to rest the case: CV is IRV in disguise. Some may argue not all races are three-candidate. So let’s take a look at a now-famous council race (Seat 1) that was held in Moab, UT, in 2021.
The Council race in Moab had a Condorcet failure (see Table 2 of report). I am showing here a screenshot of my code output, listing the individual IRV rounds with intermediate tallies. At the bottom of the screen you can also see a table with head-to-head counts. From this table it can be determined that Luke Wojciechowski (LW) is the Condorcet winner. But IRV failed to elect him in this race (he gets eliminated in Round 2). My handwritten notes then show what CV would do. Yes, CV, too, would eliminate LW and elect Jason Taylor, same as IRV.
Btw, if you are keen to see the full-fledged craziness of the Moab race, please read my post on it here (more comics, too!)
2. Contingency Voting Causes Voter Disenfranchisement
The CV method uses same ballot format (rows are candidates, columns are ranks) as the municipal RCV elections. In the 2021 Utah County RCV data, we have observed substantial amounts of votes discarded across 16 races. I am not referring to “exhausted” ballots, no, I mean discarded, as in: tossed out. In some instances these amounts exceeded 30% (read the Genola story here). While in regular elections percentages of invalid ballots lie well below 1%, in RCV elections these percentages rise to unacceptable levels (see Table 1 of our report). To be precise, on average 6.2% of ballots were discarded per IRV race in Utah County in 2021. Now replace IRV with CV and see the future. On top of this issue comes an additional bonus: further substantial amounts of ballots had “issues” that needed automated rectification (speak: truncation). In other words, on top tossing out portion of votes, remaining votes were subject to changes that modified voters’ original intent (more detail in the report). Mind you, voters never knew!
One could argue: “Well, it’s the silly little voter’s fault. Shame on him!” On the other hand, could it be that many common-sense voters thought it natural to rank two candidates equal, say, first, especially in two-seat races (ballot discarded)? Apparently many also thought it natural to rank one candidate first (love her) and another bottom (can’t stand him), leaving the middle empty (don’t care). This leads to ballot being truncated. HB205 provides some budget for “education” to this end. However, given the scale of statewide primary elections involving 100’s of thousands of voters, I am skeptical these serious problems with RCV ballots will magically disappear. It boils down to the question: are we fixing a smaller problem (plurality) with a bigger problem? We are, aren’t we?
3. Spreadsheets Disaster
I spoke with a County Clerk as well as with the State Director of Elections to clarify one nagging question: We know that IRV is non-additive. This has a logistical implication: IRV ballots must be tallied centrally. Now, because CV reduces the number of rounds to two, HB205 implies a clever way of doing a central-yet-local tallying. In essence, county clerks will bear the brunt of the complexity and it is not clear that their duty will be doable in a safe manner. The following flow diagram describes the CV counting process:
First, county clerks process local ballots to extract the top (rank-1) counts for all candidates. They then send these statistics to the State. The State determines the top two candidates. If neither of these reaches majority (>50%), the State sends back a request to all county clerks: “Here are the top two candidate names (A and B). Please go back to your ballots and tell us how many times A was preferred over B and B over A!” (See? It’s central but not central).
That’s when stuff hits the fan: With the exception of Salt Lake County, no other county has certified equipment to do such calculations. The HB205 proponents claim that clerks can export their ballots to a spreadsheet and subsequently manipulate that spreadsheet to extract the preferential statistics. Given how nonchalant they act about it, I suspect none have actually tested this. Here is a tiny sample of such exported spreadsheet:
Each row corresponds to a single (anonymous) ranked ballot. Each column corresponds to ranked position. For the sake of a mock-up exercise, suppose the State determines that Michelle Miles (MM) and Aaron Bullen (AB) are the top two candidates (1st round). A clerk is now supposed to process each row (100’s of thousands of them) and mark when MM>AB and when AB>MM. Note that some rows will have MM missing, some will have AB missing, some will have both missing. The clerk has to treat all of these cases and provide a correct tally back to the State. I am an experienced programer and would write a script that extracts these statistics with relative ease. But, will my program be comparable to a certified election equipment? I hope not. I could make a programming mistake, e.g., miss counting special cases. Certification bestows the election process a quality standard involving thorough testing. So, who will write these programs (or scripts, or spreadsheet formulas)? Clerks are not sure (in fact, they are opposed to this) as they aren’t necessarily experience programers. The sponsor seems confident the clerks will do that and the effort will be “negligible”(!) (check the HB205 gov ops committee hearing on 01/26/23). It seems the certification question ended up safely under the rug. This situation is bizarre and reckless. Using uncertified voting equipment poses additional risks of voter disenfranchisement and just promises an extra dent in voters’ confidence.
Kudos
If you’ve made it all the way down here (other than by wild scrolling).
The proposed bill HB205 spells trouble for Utah voters for times to come. I strongly believe it must be stopped. Looking forward to hearing your comments. And, if you want to help, please contact your senator and tell them you oppose HB205, before it is too late. Thanks for reading.
Acknowledgements
I thank Carolyn Phippen (Executive Director, Freedom Front of Utah) for her unwavering support in the “battle” for sane elections in Utah.
Work on technical report “Analysis of 2021 IRV Elections in Utah” as well as Special Alaska Election Analysis was done in collaboration with Dr. Warren D. Smith (Director of the Center for Range Voting). Dr. Smith also pointed out the equivalence of IRV and CV for C=3 to me.
I am grateful to the officials, legislators and friends for their time meeting with me on this topic (you know who you are!). In particular, I appreciate Rep. Jordan Teuscher taking his time to hear me out.