
Five ballot fights, two months of online conversation, and what voters decided on Nov 3.
Key finding: Where X and other platforms disagreed, the other platforms better reflected how people voted. Three times out of three.
About this series
Ballot Signals was a daily narrative intelligence series on five of 2026’s biggest ballot fights. Before the election, we used Rolli IQ to read the public conversation about each one and set it beside the latest public data. On Election Day we named what to watch. This is the check: what the conversation showed, next to what voters decided.
The scorecard
All five measures are now decided. California is still counting ballots, so its vote shares may shift slightly before certification; the outcomes will not.
| Measure | What the conversation showed | Result |
|---|---|---|
| California Prop 40, billionaire tax | X: 90% of engagement on top posts went to Yes | Failed, 47.2% Yes to 52.8% No |
| California Prop 39, voter ID | X: 90% Yes. YouTube: 56% No. | Failed, 39.4% to 60.6% |
| Missouri Amendment 3, abortion-rights repeal | X: 83% Yes. Facebook, Instagram and YouTube: 80% or more No. | Failed, 44.6% to 55.4% |
| Montana I-194, corporate election spending | Facebook: 71% of engagement went to supporters | Passed, 65.5% to 34.5% |
| 41 local AI data center votes | Engagement with local opposition rose more than 70% in five weeks | Limits won 40 of 41 |

California: X said Yes. Voters said No.
Prop 40 would have placed a one-time 5% tax on the wealth of Californians worth more than $1 billion. On X, 9 of every 10 interactions on the top posts went to the Yes side. Voters rejected it, 47.2% to 52.8%, according to CalMatters. Yes did better than the newest poll before the election, which had it at 42%, but not well enough to pass.
Prop 39 would have required voters to show ID. On X, 90% of engagement went to supporters. YouTube leaned the other way, 56% to 44% against. Voters rejected it, 39.4% to 60.6%, close to the last PPIC survey (41% to 58%). On Prop 39, YouTube read the vote better than X.

Missouri: the other platforms read it right
Amendment 3 would have repealed the abortion-rights amendment Missouri voters passed in 2024. A Yes vote repealed it; a No vote kept it. On X, 83% of engagement went to Yes. On Facebook, Instagram and YouTube, 8 in 10 interactions or more went to No. The latest poll was tied. Voters kept the 2024 amendment by 11 points, 55.4% to 44.6%. The platforms that leaned No pointed the right way, though by wider margins than the vote itself.

Montana: the louder side won big
I-194 bars corporations, unions and other groups from spending to influence Montana elections. There was no public poll, so the conversation was one of the only reads available, and it pointed two ways. Supporters drew 71% of Facebook engagement. Opponents were spread across more accounts, and 21 of 25 posts taking a side from Montana-based X accounts leaned against. Voters approved it, 65.5% to 34.5%. In this race, volume carried the vote; breadth did not.

AI data centers: the town hall voted
Communities in 10 states voted on whether to ban, pause or require approval for new AI data centers. In Report 1, we found the online fight had moved from power bills to local town halls, with engagement on local opposition up more than 70% in five weeks. On Nov 3, voters favored limiting data centers in 40 of 41 local measures, according to Ballotpedia. The one exception was Harwood, North Dakota, where voters rejected a review requirement.

What we learned
Read across platforms. Where X split from Facebook, Instagram or YouTube, those platforms better reflected how people voted, three times out of three. That says nothing about which platform moved voters; it says whose audience looked more like the electorate. Each platform has its own audience, and no single feed is the electorate.
In Montana, volume won. With no poll to go on, the side that drew the most engagement won by 31 points. That is one race, and a pattern worth watching rather than a rule.
The gap is the signal. Polls show where voters stand. The conversation shows why, and who is moving it. When the loudest feed and the voters disagree, the useful question is who is talking, and where.
What the conversation can’t tell you
Engagement share is not vote share. A platform where 80% of interactions lean one way says which way its audience leans, not by how much the electorate will. Three races where platforms disagreed is a pattern to watch, not a rule. In Montana, we published two signals that pointed in different directions and named which one held only after the vote. Online conversation shows which arguments are moving and who is carrying them; it works best read alongside polls, not instead of them.

Want this read on your issue? Send us a topic and we’ll show you what Rolli IQ finds. Talk to Rolli →
Read the series
- Report 1: AI data centers
- Report 2: California Prop 40
- Report 3: Missouri Amendment 3
- Report 4: Montana I-194
- Report 5: California Prop 39
- Election Day: What to watch tonight
Method and sources. Online figures come from Ballot Signals Reports 1 to 5 (Rolli IQ, public posts; see each report for windows and method). Engagement is likes, shares and replies on the posts collected. These are public posts, not a sample of voters. Results: Prop 40, CalMatters (AP call Nov 9; counts as reported Nov 9); Prop 39, NBC Los Angeles/AP (as of Nov 6, 83% of expected vote); Missouri Amendment 3, The New York Times via Ballotpedia (95% reporting, unofficial, Nov 9); Montana I-194, Montana Secretary of State (727 of 728 precincts, Nov 9); data centers, Ballotpedia (Nov 6, 2026). Polls as cited in Reports 2, 3 and 5. Ballot Signals takes no side on any measure.