Colorado's all-mail voting transition (Colorado, 2013 to 2026)

Colorado began mailing a ballot to every active voter in 2013, turnout rose by an estimated 8 points in the best study, costs per vote and provisional ballots fell, and the main costs are signature rejections that fall hardest on young voters and a 2026 ballot fight over adding ID numbers.

The record

What the law did

• The law was House Bill 13-1303, the "Voter Access and Modernized Elections Act." An earlier draft of this research file states it was signed May 10, 2013, that Colorado was the third state to adopt mail voting for all, and that it ranked second in turnout in 2020, all citing a Secretary of State press release. We tried to reopen that release on October 19, 2026 and the fetch was blocked, so these three points are not re-verified. The bill page on the legislature's site also failed to load. (Colorado Secretary of State press release, May 10, 2023; not opened in this check)
• Under the system, a ballot is mailed to every active registered voter, in-person voting remains at county voting centers (Pew: vote centers were required in all counties by 2014), and voters can return ballots by mail or at drop boxes. Pew's page confirms vote centers and drop boxes. Same-day registration and a shorter residency requirement (30 days to 22) were stated in the earlier draft; Pew's page summary we read did not mention them, so those two are not re-verified. (Pew, Colorado Voting Reforms: Early Results, March 2016; opened)
• The first statewide general election under the system was November 2014. This was a package of changes: mail ballots, vote centers and (per the earlier draft) same-day registration arrived together, so no study can fully separate their effects.

Timeline

Turnout

| Finding | Number | Source | Caution | |---|---|---| | Turnout of registered voters, 2010 and 2014 | 51.7% then 54.7% | Pew, 2016 (opened) | Two different elections. Pew itself says a competitive 2014 Senate race and a long-term rise in midterm turnout may also explain the gain. | | Effect of the law | "approximately 8 percentage points" | Bonica, Grumbach, Hill and Jefferson, Electoral Studies, 2021, abstract reproduced at Election Law Blog (opened) | The article's full text on PubMed Central was blocked. | | Extra ballots | about 900,000, 2014 to 2018 | same abstract | Depends on the study's assumptions | | Who gained most | larger gains for less frequent voters: younger people, manual laborers, people with less schooling | same abstract | Abstract also lists voters of color in the earlier draft; the abstract text we opened did not, so this is not re-verified. | | 2024 turnout | 3,240,549 ballots, 79.5% of 4,074,680 active voters; mail 2,988,864 (92.2%), in person 251,685 (7.8%) | Colorado Secretary of State, Dec. 6, 2024 (opened) | Denominator is active voters, which is not the same as Pew's registered voters. |

• Subgroup numbers from the authors' 2020 working paper (for example 13.2 points for Black voters, 11.2 for Asian American voters, 9.4 overall) are not used here. The working paper was not reopened, and the published overall figure changed to about 8.
• An opposing finding: Luke Keele and Rocio Titiunik studied the town of Basalt, split between Eagle County (all-mail) and Pitkin County (polling places), in the 2010 primary. Their abstract says "all-mail voting decreased turnout in Colorado" under their assumptions. The size of the drop was not in the abstract we opened. This was one town and one primary, three years before the state law, and with no vote centers or same-day registration. (Keele and Titiunik, CESifo Economic Studies, 2018; abstract opened)
• Other states: Thompson, Wu, Yoder and Hall (PNAS, 2020) studied California, Utah and Washington counties and found universal vote-by-mail raised turnout by about 2 points and had no apparent effect on either party's turnout share. The paper does not analyze Colorado data. (Thompson et al., 2020; opened)

Cost and administration (all from Pew, 2016; opened)

• Pew says costs fell in four of the five tracked categories, and rental costs rose. The earlier draft said costs fell about 40% across five categories. Pew's own limits section says that 40% figure is based on five categories while the body reports four, so we do not use the 40% number.
• Survey of more than 1,500 voters: 95% of mail voters and 96% of in-person voters were satisfied or very satisfied. Of respondents, 64% returned ballots in person, usually at a drop box, and 78% of those said it took under 10 minutes. (Pew)
• Pew's stated limits: 2008 was a presidential year, which may lower cost per vote; cost data came from 46 of 64 counties and the 2008 comparison used a separate clerks' survey; matched and unmatched counties differ in college education. Costs were not adjusted for inflation (earlier draft; not re-verified).

Who loses a ballot: signature checks

• A single judge reviews each signature first. If in doubt, a two-judge team from different parties decides, and a ballot is rejected only if the team agrees it does not match or is missing. Voters can fix the problem by the eighth day after Election Day, including through the TXT2Cure text system. (Colorado Newsline, Sept. 5, 2024; opened)

• Differences across counties: In the 2016 general, an Adams County mail voter was nearly three times as likely to have a ballot rejected as a Weld County voter, per CPR's analysis of federal data. Adams is about 40% Hispanic and, CPR says, Weld does not use signature verification software. In 2018 a Pueblo voter was twice as likely as a Douglas County voter to be rejected, though Pueblo's clerk said a brief machine breakdown may have contributed. (CPR News). CPR reported no rates by race. The earlier draft said no Colorado study had measured it; we did not recheck that.
• The earlier draft also cited a December 2022 lawsuit, Vet Voice Foundation v. Griswold, whose complaint claimed young Black and Hispanic voters' ballots were rejected "approximately 25 times more often." That was the plaintiffs' claim, not a court finding. We did not reopen the Democracy Docket or CPR 2022 pages in this check. Not re-verified.

Fraud and security evidence

• In the November 2024 election, Colorado officials said at least 12 mail ballots were stolen from mailboxes, filled out and returned in Mesa County. Three passed signature verification and were counted; signature checks caught the rest, and some real voters called the clerk. The Mesa County District Attorney was investigating. (NBC News, 2024; opened). A KUNC story on the same event exists but its fetch was blocked.
• The 2024 Secretary of State release mentions a "bipartisan post-election audit" but gives no audit figures or counts of fraud referrals. (Secretary of State; opened)
• The Carter-Baker Commission on Federal Election Reform (2005) said growing numbers of absentee and mail ballots were raising concerns about fraud and recommended stronger anti-fraud efforts, especially for absentee voting, along with photo ID based on REAL ID. We read a Baker Institute summary only, not the report. (Baker Institute; opened)
• We did not find a statewide count of prosecuted mail-ballot fraud in Colorado that we could open. Unverified.

Amendment 84, on the November 2026 ballot

• Amendment 84 (Initiative 362) would put in the constitution a requirement that mail voters give identification information, such as the last four digits of a Social Security number or a driver's license or REAL ID number, in addition to a signature. It needs 55% to pass. Named supporters: El Paso County Treasurer Chuck Broerman and former Deputy Secretary of State Suzanne Taheri. (Colorado Politics, Oct. 2, 2026; opened)
• Supporters say ID numbers are "more consistent and less subjective" than signatures and would raise public confidence. Opponents say it adds a burden without meaningfully improving security and that small mistakes could cause eligible ballots to be uncounted. The article names no opponents and gives no fiscal note. The Legislative Council page lists the Blue Book analysis but we could not read its contents. (Colorado Legislative Council; opened, files not read)

What this case shows

1. Making voting easier brought in people who vote least. The best causal estimate puts the turnout effect at about 8 points, 900,000 ballots from 2014 to 2018, with larger gains for young people, blue-collar workers and people with less schooling. Ordinary voters could get a ballot at home without taking time off work. (Bonica et al., abstract)
2. The change was cheaper per vote and cut a burden voters face on Election Day. Cost per vote fell from $15.96 to $9.56 in the 46 counties with data, and provisional ballots fell from 39,361 to 981. Provisional ballots mostly hit people at the wrong precinct, a problem renters and movers have more often (inference; Pew gives the counts only). (Pew)
3. Most voters used it and liked it. 92.2% of 2024 ballots came by mail, and 95% of mail voters in Pew's survey were satisfied. (Secretary of State; Pew)
4. The safeguard is a barrier too. The signature check that guards against stolen ballots rejected 2.8% of 18 to 24 year olds' ballots in 2018 against 0.2% for ages 55 to 64. Voters had eight days to fix it, but clerks estimate only 20 to 30% do. A security tool lands unevenly. (CPR News)
5. Voters have an immediate say on the next change. Amendment 84 puts the question to a statewide vote with a 55% bar. What the ordinary voter can read, and vote on, is the ID rule itself. (Colorado Politics)

The strongest objections, and our answers

1. Objection: Mail ballots create fraud risk that in-person ID checks do not. Mail ballots can be stolen and signed by someone else. The 2005 Carter-Baker Commission raised this concern. In 2024 at least 12 ballots were stolen in Mesa County and 3 were counted. Mitigation: those 3 are a known number out of 3,240,549 ballots, and the system caught most of the rest through signatures and voter reports. We do not claim these counts show the full rate, because no statewide fraud count was found. Critics can fairly say the 3 counted ballots passed the very check meant to stop them, which is the argument for an added ID number.
2. Objection: Signature matching is subjective and rejects eligible voters. The CPR data on age and county differences back this. Mitigation: we show them, and note that CPR found Weld County does not use signature software and that Pueblo's clerk blamed a machine problem. No study we opened measures rejections by race or income. The 25 times claim is a plaintiff's allegation, not a finding.
3. Objection: The 8 point turnout effect may be overstated. Pew warns about a competitive 2014 Senate race and trend. The Basalt study found the opposite sign in 2010. Thompson et al. found only about 2 points in other states. Mitigation: we give all four. Differences in design, place and year explain much of it: Basalt had a primary and no vote centers; the other states did not include Colorado; Bonica et al. compare Colorado voters to voters in similar states.
4. Objection: Lower cost per vote may be an artifact of comparing a presidential year with a midterm. Pew says this itself. Mitigation: we cite it as a limit and do not claim total spending fell.
5. Objection: A package of reforms cannot be credited to the mailed ballot. True; mail ballots, vote centers and (per the earlier draft) same-day registration came together. We say so.
6. Objection: Amendment 84 would make voting safer, or would add a new barrier. Both sides are quoted above. This case does not take a side. Whether an ID number would replace some signature failures or add new ones depends on text we did not read: whether it replaces or adds to the signature check, how typos are handled, and how long the cure period is.

Open questions: what would change our minds

• The May 10, 2013 signing date, the "third state" claim, the "second in turnout in 2020" claim, and the statutory text of HB 13-1303 (including same-day registration and the 22-day residency rule) were not re-verified. The Secretary of State press release and the legislature's bill page could not be opened. Vote totals on the bill were not found.
• Bonica et al.'s full tables (including effects by race) were not read; the PubMed Central copy was blocked. We rely on the abstract.
• The Keele and Titiunik effect size was not in the abstract. The 6-point figure in the earlier draft is unverified.
• Rejection rates after 2020, and by race, income and language, are unknown to us.
• Colorado statewide counts of mail-ballot fraud cases investigated, referred and prosecuted were not found. The KUNC article and Secretary of State audit documents were not read.
• The Mesa County case outcome after the district attorney's investigation is unknown to us.
• The Amendment 84 Blue Book and fiscal note were not read.
• Lawsuit details (Vet Voice Foundation v. Griswold) and the clerks' association response are carried from the earlier draft and not re-verified.
• Turnout as a share of all eligible adults, and inflation-adjusted total election spending, were not found.
• Evidence that would change our minds: a published rejection analysis by race and age after 2020; a verified statewide count of fraudulent mail ballots; a causal estimate that separates the effect of mailed ballots from vote centers.

Sources

Found an error or a better source? We keep every correction on the record with its date and reason. Nothing here is quietly rewritten.