The Registry, Not the Gunmen, Is Haiti's Real Election Clock

On Nov. 11, the Provisional Electoral Council (CEP) had 934,194 voters on its register. The National Identification Office (ONI) had, by late August, issued 5,837,560 national ID cards. Divide one by the other and about 16% of card-holders are enrolled to vote. The register closes on May 20, 2027, 190 days after that count, and the first round is now set for July 18, 2027. The arithmetic of those 190 days says more about whether the date holds than any map of gang territory.

Card-holders are a proxy for the eligible electorate, not a census of it, so read the 16% as an order of magnitude.

Is the vote really waiting on the gunmen?

The prevailing theory is sequential. The Salvadoran-backed offensive and the Gang Suppression Force clear ground, the CEP opens centers behind them, and a vote follows. On this reading the slip from Dec. 13, 2026 to July 18, 2027 is a security delay, and the remedy is more firepower, faster.

The theory predicts that enrollment should accelerate wherever clerks can work, because registration does not need a front line. It has done the opposite. By the CEP's published counts, the register reached 628,755 on Sept. 21 and 817,991 on Oct. 13. Between Oct. 2, when it stood at 728,437, and Oct. 13, it grew by roughly 8,100 a day. From Oct. 13 to Nov. 11 it grew by about 4,000 a day. These daily rates are our calculation from CEP and press counts, not CEP-reported figures.

What does a slowing register do to the calendar?

Run the two rates forward as illustration, not forecast. At 4,000 a day, the register reaches about 1.7 million by May 20, roughly 29% of card-holders. At the faster average since July, about 8,200 a day, it reaches about 2.5 million, or 43%. The CEP's former president had spoken of a target of 3 to 4 million. Even the optimistic path falls short of the low end of that range.

A election with a register at 29% of card-holders is not a delayed election. It is a contested one. Excluded majorities hand losing candidates, including those facing G7 sanctions and those who simply lose, a ready claim that the roll was captured or rigged. Whoever controls the CEP's data pipeline then controls the narrative. The pressure to push the date again, or to hold a vote nobody accepts, builds quietly in the gap between the enrollment curve and the calendar.

The geography sharpens the point. The Ouest department holds about 34% of the population but accounted for only 134,954 of the 817,991 registrants on Oct. 13, or 16.5%. That is half its demographic weight. The Ouest is where the gangs are, so security plainly matters there. But it is one department, and the national slowdown is not explained by it alone.

One data-quality flag: Rezo Nòdwès's headline pairs the 934,194 enrolled voters with 1,447,000 'support registrations', and calls the relationship between the lists an enigma. We have not reconciled those figures and do not rely on the larger one. Anyone designing incentives around enrollment counts should audit what is being counted first.

Are the displaced really out of reach?

The displaced-voter problem is usually presented as a wall: more than a million people scattered by violence, unreachable by electoral staff. The IOM's Round 14 count is 1.49 million internally displaced people. By IOM's breakdown, about 79% of them live outside the capital, in departments where registration centers are operating. Many of the displaced are therefore reachable by a clerk with a laptop and a camera. What stops them is documentary, not kinetic.

The Haitian Times reports 6,583,753 people enrolled with ONI and 5,837,560 cards issued. That leaves 746,193 people with an ONI record and no card. Since the card is the precondition for registering to vote, these are citizens already identified by the state and stranded one document short of the register. A successful offensive does not print them a card. Territory cleared in Artibonite or the Centre does not help a displaced family in a host commune that has no card to show.

What would a card-to-register surge cost, and why might it fail?

The mechanism is simple. Combine two PeaceBot governance instruments, civil_registry_pop_up and electoral_commission_rebuild, into a single ONI/CEP card-to-register surge:

• Clear the card backlog. The anchor price is the Japan–OAS arrangement: $3 million for 350,000 biometric cards, about $8.60 each. Applied to 746,193 undelivered cards, the bill is about $6.4 million, or roughly HTG 837 million at the BRH reference rate of 130.73 HTG per dollar. This is our extrapolation and assumes similar unit costs at scale.
• Go to the people. Station pop-up delivery and registration desks at the 230-plus IDP sites and in the host communes around them, pairing card handover with enrollment on the same visit.
• Publish a weekly dashboard of registrations against card-holders by department, so the Ouest gap and any padding are visible to everyone.
• Condition donor tranches on a floor of 8,000 verified enrollments a day, the pace the register briefly achieved in early October.

PeaceBot's feasibility and cost scores for these instruments could not be retrieved for this column, so treat them as unscored. The costing above rests on the Japan–OAS unit price alone.

The strongest objection is that the backlog is concentrated precisely where the state cannot deliver. If a large share of the 746,193 uncollected cards sit in gang-held parts of the Ouest, a pop-up desk in a host commune will not reach them, and the 8,000-a-day floor becomes unreachable. A volume target can also invite padding, which the support-registration puzzle suggests is a live risk. Both objections are serious. The response is to weight the conditionality toward verified biometric matches rather than raw counts, and to publish the departmental split so donors can see whether the shortfall is logistical or territorial. If the dashboard shows the backlog is mostly inside gang territory, that is real evidence for the security-first theory, and it is evidence the current approach never collects.

That is the point. Today the July date is defended or doubted by assertion. A register that grows past 2.5 million by May tells us the date is achievable. One that crawls at 4,000 a day tells us it is not, regardless of what the offensive achieves. The clock is in the CEP's enrollment ledger, and for now nobody is reading it.