The Measured Screen

CTV dollars still grow — the clock and migration guarantee a 2028 pulse on a higher floor — but SHARE stalls in the low-to-mid 20s% rather than climbing toward broadcast's. Binding constraints: the four premium refusers (Netflix/Amazon/Disney+/Apple) hold, so demand piles into the fixed accepter set (Hulu/Roku/YouTube/FAST), CPMs spike Sept-Nov, and buyers substitute toward open-exchange/FAST or hold budget on broadcast for the high-turnout cohort. The effectiveness debate stays adverse enough that the 'waiting-for-proof' dollar stays on linear. This is Continuation's mirror — same dominant drivers, the counter-forces binding harder — a SHARE stall layered on still-rising dollars, never a dollar stall.

CTV's share jumps past the flattening-shoulder expectation into the low-30s% of total political spend by 2028 because a ceiling constraint breaks the right way. The likeliest unlock is a premium holdout (Netflix most plausible) opening to political ads, flooding high-CPM inventory, relieving the Sept-Nov scarcity spike, and pulling presidential dollars onto streaming; secondarily a 2026 actual printing well above projection, or AI-driven creative abundance making high-frequency CTV buys cheap enough that budgets shift faster than the base case. This is the upside the outside view's deceleration assumption misses because a structural lid lifts — the cleanest single corpus-flagged discontinuity in CTV's favor.

The arbitrage tailwind partly flips to a headwind, or a structural event reprices the channel. NOT a full federal Honest Ads Act (base rate too low), but a narrower shock: a wave of enforceable state disclosure laws on Washington's court-surviving template, the Dec-2025 AI-preemption EO whipsawing the only legal layer reaching streaming, or a Fox-Roku close (~$22B, 1H-2027 guidance) reshaping the political-ad archive/ACR layer. The EU TTPA shows the mechanism — Google (2024) and Meta (Oct-2025) EXITED rather than de-target — so even a partial US analog raises compliance cost and bends CTV below continuation; but McManus and the First Amendment foreclose the EU targeting-ban path, keeping this a tail. The band also carries the council's residual unmodeled structural break.

The corpus's principal demand-side downside. A credible, independent, CTV-SPECIFIC incrementality study finds political streaming ads don't move votes — extending Kalla-Broockman's ~zero and the Coppock-Green-Porter -0.04pp digital-video null into the streaming era — and enters the trade record loudly enough that buyers re-price the channel. Because much of CTV's premium-CPM value rests on TARGETED PERSUASION (not just reach), a robust null deflates the share trajectory: budget waiting for proof stays on cheaper reach buys or reverts to broadcast's known GRPs, and high-CPM inventory loses pricing power. CTV does NOT shrink (reach value is real and independent of persuasion) but its share growth stalls below continuation. It stays small because spend flowed for three cycles DESPITE the adverse literature — the buy-side runs on reach logic and the vendor layer is paid on dollars spent.