Before Berkeley asks voters to approve a tax or bond, give an independent reviewer one job: try to find ways the measure’s promises can fail while everyone still follows the written rules.
That is “red teaming” in plain English.
A red team is a skeptical reviewer whose job is to try to break a system before the public has to rely on it.
Cybersecurity teams do this before attackers find the weakness. AI developers do it before releasing a model. The same idea belongs in ballot-measure design.
The reviewer asks:
That question would have caught the Measure FF problem before the election.
Measure FF was designed to add substantial new funding for streets and sidewalks.
Its ballot language said the new tax would be available “provided other funding for street maintenance is maintained.”
The measure backed that promise with a $15.3 million maintenance-of-effort floor for qualifying non-FF street funding.
A red-team reviewer should immediately have asked:
The answer became clear the following year.
Berkeley removed the remaining $3.95 million Measure T1 street allocation because:
The safeguard protected the recurring floor.
It did not protect the supplemental T1 money.
Nobody needed to evade FF. The gap was already in the design.
See the full case study: Fungibility Isn’t Fun
Most public oversight comes after adoption.
Auditors check whether money was accounted for properly and spent on legal purposes. Oversight committees review expenditures. Financial reports track implementation.
Those are necessary.
But a poorly designed promise can survive every one of those checks.
A red-team review happens earlier, while Council can still change the measure.
The assignment is adversarial in the useful sense of the word: look for failure modes, unintended incentives, vague promises, weak baselines, and legal ways for the outcome to drift away from what voters think they are approving.
Auditing asks whether the rules were followed. Red-teaming asks whether the rules are strong enough to produce the promised result.
The review does not need to become a hundred-page exercise. A short set of recurring tests would catch many problems.
List every existing source already funding the same purpose:
Then ask which of those sources can legally be reduced after the new measure passes.
Publish the gross new revenue and the expected net increase.
The FF/T1 case shows why this belongs in every review.
If voters are shown recognizable projects, identify how much legal discretion the City retains after passage.
Can a project be:
If so, say that plainly before the vote.
Require a public reprioritization process when a major project is removed or replaced.
A measure may promise “accountability,” “safety,” “infrastructure improvements,” or “independent audits.”
The reviewer should ask what an auditor will actually be able to prove later.
A useful promise needs a baseline, an output, or another observable standard.
“Money was spent legally” is easy to audit.
“Voters received the result they were shown” often is not.
This is the subject of the companion proposal:
Berkeley Ballot Measures Need an Auditor’s Analysis
For a bond, show:
Berkeley’s own debt policy says debt terms should be matched to asset lives whenever practicable.
The red-team question is simple:
Berkeley Debt Management and Disclosure Policy
Every new measure is presented separately.
Households pay them together.
A pre-ballot review should show how the new obligation sits on top of:
That does not decide whether the burden is acceptable.
It gives voters the information needed to decide for themselves.
Berkeley already has an independently elected City Auditor.
That office is the natural institutional home for this kind of skeptical review.
The City Attorney already provides a legal analysis of ballot measures. The Auditor could provide a different function: test whether the design can deliver and verify the promises Council is putting before voters.
The review should happen before Council places the measure on the ballot, while the language can still be fixed.
The Auditor would need access to:
The result should be public and advisory.
Council would still decide policy. Voters would still decide the measure.
They would simply see the failure modes before making that decision.
Call it a Ballot Measure Adversarial Review.
For each major promise, show:
| Question | Review |
|---|---|
| What is being promised? | The claim voters are being asked to rely on |
| How can it fail while the rules are followed? | The loophole or failure mode |
| What safeguard exists? | Current protection |
| What remains exposed? | Residual risk |
| How can the language be fixed? | Specific design change |
| Can the promise be audited later? | Yes, no, or only partially |
This is not another campaign argument.
It is quality control.
The Berkeley Decision Project has already proposed an independent Auditor’s Analysis for ballot measures.
The two ideas solve adjacent problems.
Red-team review asks, before adoption:
Auditor’s Analysis asks:
Together they create a pre-ballot accountability check that Berkeley’s current process lacks.
The FF/T1 experience shows why it matters. FF actually contained an anti-supplanting safeguard. The weakness was subtler: the safeguard protected a narrow baseline while leaving previously committed supplemental funding exposed.
A skeptical reviewer could have spotted that in a paragraph.
The best time to discover how a ballot promise can fail is before voters make it law.
Fungibility Isn’t Fun · Where Did $300 Million Come From, Anyway? · Show Your Work · Berkeley Decision Project home