Chile

What the record shows

Santiago is a metropolis that does not administratively exist, and this record is a catalogue of what that costs. Greater Santiago has no metropolitan municipality, no metropolitan police force and no metropolitan data catalogue, because there is no government that could run one. Thirty-two comunas fund their own services from their own budgets. The pattern that follows is consistent enough to be a rule: everything here that works at the scale of the city belongs to a national or regional body, and everything the comunas do themselves stops at a boundary.

The clearest case is surveillance. Las Condes reports the highest security spending of any Chilean municipality and runs more than 3,300 cameras watched around the clock, with drone patrols, a plate-reading ring and a facial-recognition contract — all approved by one comuna's council and all stopping where that comuna ends. Santiago cannot buy surveillance as a city because it is not one, and the only outside scrutiny the programme drew came from a court ruling on five residents' complaints. The bike share was assembled the same way and shows what the arrangement costs over time: it opened in one comuna in October 2013, eleven comunas ran a joint tender the following July and each signed its own contract, no comuna pays anything, and the system has since shrunk from fourteen comunas to five. A concession that costs the city nothing also buys the city nothing — each comuna's coverage lasts exactly as long as the operator finds it profitable, and service can be withdrawn by an app notification rather than by a public decision. The data position is the same absence in another register: the one catalogue that exists belongs to the historic core, one comuna of thirty-two, and of its 136 public map layers only 28 carry any licence statement at all — not one of them a recognised licence, and most of those 28 non-commercial. A non-commercial licence on a public dataset is a refusal dressed as a grant.

What works, works because somebody above the comunas owns it. The winter air regime is the most distinctive mechanism in this record and possibly in the atlas: from May to August, Santiago manages its air by forecast rather than by reading. Ten automatic stations feed a regression model that predicts the next day's maximum particulate average, and if the prediction crosses a threshold fixed in the decontamination plan, a presidentially appointed regional delegate declares an episode — with enforcement split between the environment superintendency for large industry, the health authority for heaters, transport authorities and the police for the driving ban, and the forestry body for agricultural burning. Episode hours in fine particulates fell from 915 in 2015 to 314 in 2025. A threshold can be fixed in law and still leave the decision human: this one acts on a prediction of tomorrow, signed by an official the president appoints.

The plate ration hangs off that same forecast, and it is the reason this scheme clears a bar that Delhi's odd-even does not. Chilean officers check plates by eye at the kerb exactly as Delhi's did. What makes the difference is that the ration is tiered by a vehicle's vintage class — catalytic cars registered before September 2011 lose two digits a day, older vehicles without the seal lose more — and is re-set by an air-quality forecast, so two measured layers sit under a manual stop. It also has the rarest thing in its genre: an independent peer-reviewed evaluation, using the index thresholds that trigger the warnings as instruments in a regression-discontinuity design, which found the temporary bans cut car trips by 6 to 9 per cent in peak hours and 7 to 8 per cent off-peak.

The metro settles a question this atlas keeps asking. Lines 6 and 3 opened unattended in November 2017 and January 2019, and the label has held for eight years for a physical reason rather than a contractual one: the trains built for them have no driver's cab, only a concealed console for taking a train over by hand when something fails. There is no seat from which an attended service could run. The same signalling family went onto the older lines, which kept their drivers, because there the point was capacity rather than automation. The test of a driverless claim is not the grade a supplier certifies but whether the trains were built with a cab.

The rest of the record is about what a measurement is allowed to trigger, and the answer is usually a person. Aguas Andinas has cut supply to more than a million customers repeatedly when Andean storms push raw turbidity past what a treatment plant can handle, and built reserves that took the city's autonomy from about four hours in 2011 to 37 today — storage that turned most turbidity events into a non-event, until the largest storm in decades halved the buffer and the cut happened anyway. The switch to reserves follows a declared alert and an operator's decision, not an automatic trigger. The fare card is unusually honest about its own blind spot: roughly a third of boardings on Santiago's buses never become a tap, so the authority measures the gap by a separate published method instead of inferring ridership from the card, publishes the index — 45.8 per cent unpaid in the first half of 2023, falling for a fifth consecutive period to 34.9 per cent in the second half of 2025 — and says in the open-data tables themselves what they exclude. And the smart-meter rollout is the record's outright failure, for a reason that had nothing to do with metering: a 2018 law moved the meter out of the household's ownership and into the distributor's, more than 250,000 units went in, and the programme stopped. The sums were trivial and the politics were not. If the entries below agree on one lesson, it is that fragmentation does not prevent good systems in Santiago — it just means that every good system belongs to someone other than the city.