Vince Phillips
In a small rural county like Chaffee, you would expect elections to be relatively straightforward. Yet in each election cycle, the Chaffee County Clerk’s office may have to print and program as many as 28 different ballot styles for a single General Election.
Here is why the number is so high. Chaffee County sits in two separate State House districts, HD 13 and HD 60. On top of that, we have the City of Salida, the Town of Buena Vista, Poncha Springs, several school districts, fire protection districts, water districts, and other special districts. Every possible combination of these races and jurisdictions creates its own unique ballot. Salida ballot is different than rural Chaffee County, which is different than Buena Vista.
All of those different ballot styles must be individually designed and programmed into the county’s Election Management System by the Clerk’s office. This is not done by the state or the voting machine company. It is done locally, right here in Chaffee County. Every oval, every contest position, every write-in line, and every interpretation rule has to be precisely mapped out for each of the 28 styles.
This level of complexity creates real risk. A single programming mistake on even one ballot style can cause votes to be dropped, misread, or credited to the wrong candidate, especially in local races or write-in contests. Even Colorado’s much-touted Risk-Limiting Audit (RLA) does not solve this problem. In fact, it makes the weakness more obvious. The Secretary of State personally chooses only a handful of “target contests” for the audit, typically just one or two statewide races plus one contest per county. Almost every other race on the ballot is bypassed entirely. Most local races that directly affect our taxes, land use, schools, and county government receive zero statistical protection.
Here is a simple hypothetical that shows how the RLA system can fail:
Suppose Chaffee County casts 15,000 ballots across 28 different ballot styles. Suppose 2 of those 28 styles (totaling just 800 ballots) contain a programming error that drops or mis-credits 20% of the votes in a county commissioner race. The entire RLA sampling for that race might be only 120 ballots drawn randomly from the entire county. Because those two problem styles represent only about 5% of all ballots, the sample is likely to pull less than a handful, or even zero, from the problematic styles. Even if the sample catches a few bad ballots, any discrepancies found would be logged but would not trigger additional rounds of auditing or further investigation for that race. The audit can still “pass” within the required 3–5% risk limit while the localized error goes undetected and uncorrected.
Many people assume Colorado’s Risk-Limiting Audit will catch significant errors wherever they occur. But the process reminds me of something I learned in a physics class back in 1981, The Parable of the Drunken Frenchman: A man drops his keys somewhere along a dark street on his way home from the bar. When he goes back to look for them, he only searches under the bright streetlights, because “that’s where he can see.” The rest of the path stays in darkness.
In Chaffee County the RLA works much the same way. It shines its statistical light brightly on only a handful of “target” contests chosen by the Secretary of State. Most state and local races stay in the shadows. Even when the sample happens to include ballots from a problematic style and auditors spot discrepancies, the process is designed to log them, offer explanations, and move on without expanding the search or triggering further investigation, as long as the targeted contests meet the risk limit.
Officials often point to the clean results of past Risk-Limiting Audits as proof that the voting system is nearly flawless. Yet the audit is deliberately limited in scope. Discrepancies that appear are usually attributed to human error by the audit board or differing interpretations of voter intent. The process is built to reach the required risk limit efficiently, even if that means many potential localized problems never receive deeper scrutiny.
Even worse, the statistical model behind the RLA assumes there is essentially one uniform source of data. It does not adequately account for the reality of up to 28 different ballot styles, each with its own distinct programming and potential failure mode. Programming errors do happen in real elections. When they occur on only one or two styles, the RLA is unlikely to detect them.
In 2022, I wrote to the Mountain Mail about the Chaffee County Clerk race. I documented over 100 write-in ballots cast for County Clerk that were not counted. The ballot images were redacted. When I asked to examine the actual paper ballots to understand why, access was denied. That experience still bothers me. If citizens cannot verify the most basic question, “Did my vote get counted as intended?”, then how much trust should we really have?
The bottom line is this: with up to 28 different ballot styles, each with its own programming and potential failure mode, the current RLA system in Chaffee County is virtually useless for the vast majority of races on the ballot. Officials boast that it provides “high confidence” that the outcomes are correct, but the audit can only reliably catch the most egregious, county-wide errors on a single race. Localized mistakes that affect specific groups of voters are likely to slip through unnoticed.
Add that without independent access to the original paper ballots, there is no way for citizens to know whether the scanned ballot images are accurate. The system tells us to trust the locally-programmed software, the locally-created ballot manifest, and the locally-run audit. That is not verification. That is wishful thinking.
Chaffee County deserves elections that are simple enough to be transparent and verifiable by ordinary citizens. We should not have to accept “trust us” as the final answer when the ballots are complex, the programming is done behind closed doors, and the Risk Limiting Audit does not deliver teh protections promised.
It is time to demand simpler, more transparent systems, and real independent access when questions arise. Because in the end, if we cannot verify the count, we cannot trust the election result.
Support Independent Local Journalism — High Country Advocate was created as a real alternative to regional media that too often silences dissenting voices while taking sides in the political struggle. Producing in-depth, unflinching reporting like this series and others is expensive: servers, editing, research time, and legal review all add up quickly. If these articles have informed you or given you new perspective, please consider supporting HCA with a paid subscription — every subscriber helps keep this reporting strong and independent.
