Chile's Theft Count Fell 20 Per Cent. The Insured Rate Didn't Move.

Chile lost 46,242 vehicles in 2022. It lost 36,868 in 2025 — down 20.27 per cent. Over the same three years the rate at which insured vehicles are stolen has not moved at all: 0.6455 per cent, three years running. Insured thefts rose from 13,733 to 14,276, up 3.95 per cent, while the national count fell by a fifth.

None of those numbers is wrong. They are the same country described from two different vantage points, and the gap between them is where a year gets mis-timed and a forecast gets built on the wrong lever.

The count came down because coverage went up

An insured theft rate is a ratio, and both halves of it moved in 2025. The numerator — insured vehicles stolen — rose 3.95 per cent. The denominator — the insured parc — grew 4.66 per cent, from roughly 2.11 million vehicles to 2,211,320. A slightly larger numerator over a slightly faster-growing denominator produces a flat rate, and flat is what the series shows.

That is the mechanical reason the insured count can rise while the country’s total falls. The insured slice of the fleet is expanding faster than theft is contracting, so more of the same crime lands inside the measured universe each year. The national series is measuring crime. The insured series is measuring crime divided by the growth of a financial product.

Pheonix Research’s Chile vehicle theft recovery report puts the consequence plainly: a rising count with a flat rate is what coverage growth looks like; a flat count with a rising rate is what a deteriorating market looks like; and Chile is currently producing the first of those. Anyone reading the annual count as a trend in crime will mis-time this market by a year or two.

A record number of policies, bought at a falling price

What drove the denominator is visible in the insurance statistics, and it is not a story about risk. In 2025 the industry issued 4,182,519 voluntary vehicle policies — a record, up 10.79 per cent on 2024’s 3,775,047 — for UF 33 million of premium, about 26 per cent of the whole general-insurance premium.

At the same time the average premium fell, from roughly UF 14 in 2024 to about UF 11 in 2025 — a 21.43 per cent decline. A record volume sold at a falling average price is the shape of a market competing on distribution rather than on risk, and it matters here because the growth this report’s base case depends on most is being purchased on price.

It is also concentrated. Four insurers wrote 77.9 per cent of those policies: Bci Seguros 1,498,372 (35.8 per cent), HDI 913,944 (21.9 per cent), Suramericana 467,042 (11.2 per cent) and Zurich Chile 373,997 (9.0 per cent). Detection technology in Chile is funded out of member premium, which means a national recovery capability is decided by four balance sheets — faster to move than a statutory system, and with no floor under it if one of them changes strategy.

Four published numbers describe one segment, and they do not tie

Before any of this is used to argue a forecast, it is worth noticing what the published record does to itself. Four figures, all for 2025, all for voluntary vehicle cover:

The 1.89 ratio is the problem. A single-vehicle annual policy cannot produce 1.89 policies per insured vehicle, and the two candidate explanations point in opposite directions for anyone modelling affordability. Either the policy count carries renewals, part-year certificates and endorsements written inside the year — in which case it is a flow being compared with a stock — or the parc is a point-in-time figure set against a full-year total. Neither explanation is published, and the report does not invent one.

What it does instead is choose, and say why. It uses the parc, because the parc is the model’s denominator, and it states the consequence: at UF 14.92 per insured vehicle the funding pool per unit is nearly twice what it is at UF 7.89 per policy, and every affordability argument about the recovery instrument turns on which unit is used. The penetration figure carries a smaller version of the same defect — the association states 32.3 per cent, its own two numbers divide to 31.81 per cent. Half a point changes nothing in the model, but a series whose published percentage and published ratio disagree is a series to handle rather than quote.

Which lever the market is actually long

Recovery value in this market is a product of four inputs: insurance penetration, the theft rate per insured vehicle, the recovery rate, and average insured vehicle value. The base case in the report takes gross recovery value from USD 122.1 million in 2025 to USD 201.5 million in 2033 at 6.47 per cent, and it is worth decomposing what that move is made of, because it is not what the headline suggests.

Recovered vehicles go from 6,424 to 9,052 — up 40.91 per cent. The average value of a recovered vehicle goes from USD 19,000 to USD 22,261 — up 17.17 per cent. Multiplied, that is 1.651 times, which is exactly the path from 122.1 to 201.5. Of the USD 79.4 million of added value, 62.83 per cent is more vehicles coming back and 37.18 per cent is each one being worth more. Most of the forecast is volume, and most of that volume is penetration — a coverage story wearing a recovery story’s clothes.

That distinction is not academic, because the two fail separately. And the cheapest available falsifier is the lever least connected to crime: if policy issuance stalls at the 2026 renewal, the denominator stops growing, the largest contributor to the base case stops contributing, and nothing has to happen to crime for the forecast to lose its main engine. A 21.43 per cent decline in the average premium is not a durable competitive position to be selling growth on top of.

The recovery rate everyone quotes measures a third of the fleet

Chile’s recovery rate at first year of claim closed 2025 at almost 45 per cent — the second consecutive annual increase, against 54.75 per cent in 2017 and 61.34 per cent in 2016. It is a real turn in a series that had been falling for most of a decade.

It is also measured on 2,211,320 vehicles, 31.8 per cent of the fleet. Of the 36,868 thefts reported in 2025, 22,592 — 61.28 per cent — fall outside the only statistic that says whether a vehicle comes back. The 45 per cent is not a national recovery rate. It is the rate for the third of the fleet that bought a policy, and the majority of Chilean vehicle theft has no recovery measurement at all.

That gap is what makes the country’s two headline numbers sit together so comfortably and mean so little together: a count measured on everything, and a rate measured on a third of it.

The one instrument with a published output and no published cost

The Asociación de Aseguradores de Chile funds a plate-recognition network, gives it to more than 260 municipalities free of charge, and operates it with the Asociación Chilena de Municipalidades, Vespucio Norte and Carabineros. It is the only recovery instrument in the country whose output anyone publishes.

• 1,858 recoveries in 2025 — 28.92 per cent of all insured recoveries in the year.
• More than 5,200 since 2021, of which 35.73 per cent arrived in 2025 alone.
• More than 260 communes covered, with each mobile unit processing over 150 plates an hour.
• Output concentrated in three communes — Maipú 119, Quilicura 95, Pudahuel 89 recovered in 2025, together 16.31 per cent of national network output.

That last line is the competitive fact that matters: coverage is national, yield is not. Productivity clusters where patrol density, plate volume and theft density coincide.

The cost is not published — not the hardware, not the 260 municipal agreements, not the data platform, not the staff. Cost per recovery, the single most useful number in this market, cannot be computed from the public record. Against it sits the enforcement side: one Carabineros anti-carjacking operation in the Metropolitan Region generated more than 600,000 plate reads, against the network’s 5,200-odd recoveries in five years. Reads are abundant; recoveries are not. Detection is not the scarce input in Chile. Interception is.

The Hilux is the whole argument in one model

The association publishes one model-level recovery rate, and it belongs to the highest-volume model in the country: the Toyota Hilux at 31.4 per cent over the twelve months to mid-2025 — 13.6 points below the market, and 30.22 per cent below it in relative terms, the lowest figure in its own table. Toyota was the most-stolen brand in the same window at 13.75 per cent of insured thefts, with the Hilux, RAV4 and Land Cruiser together above 2,250 units.

A pickup worth more dismantled than returned is the clearest evidence in this market that recovery probability is an asset characteristic rather than a system constant. It is also precisely the input a model-level risk price would need — and the reason its absence shows up in pricing.

On 2015 vehicles, private use and a CLP 5 UF deductible, the same Mitsubishi L200 was quoted at between CLP 60,774 and CLP 540,025 a month across Chilean carriers — a 788 per cent spread on an identical asset. A four-firm funder facing more than 1,800 models and 72 brands from 28 countries, by the association’s own count, cannot price theft risk at model level by hand. The spread is arithmetic, not an inefficiency anyone chose. Whoever builds that model — starting from a stolen-model ranking in which the top ten models already carry about 30 per cent of insured thefts — prices the book better than the market currently does.

What nine years of erosion is worth, and what is still missing

The recovery rate has fallen 26.64 per cent in nine years, from 61.34 per cent in 2016 to 45 per cent at full-year 2025 — a trough reached in 2023–24 from which it has now risen twice. Take the 2016 rate and apply it to 2025’s insured thefts: the same 14,276 thefts would have returned 8,757 vehicles instead of 6,424. That is 2,333 vehicles that were not returned, worth USD 44.33 million at the model’s average value — 36.32 per cent of the entire 2025 market, given up in a single year to a declining capability.

And the annual loss sits on top of an accumulated one. 55,000 insured vehicles stolen over the last twelve years remain unlocated, worth USD 1.045 billion at the same average value — 8.56 times the annual value returned. Because recovery is measured at first year of claim only, a vehicle found in year two is recorded as never found, so that stock grows quietly. This is a market with a backlog, not a growth problem.

What to watch instead of the count

The count will keep falling and it will keep being reported as good news. The number that tells you where this market is going is policy issuance — because it is the denominator, because it is being bought at a 21.43 per cent lower average price, and because 62.83 per cent of the forecast sits on top of it. The second number is cost per recovery, and nobody in Chile publishes it.

The full analysis, including the three-path band, the regional allocation of value and the year-by-year base case, is in the Chile vehicle theft recovery market report.

The headline figures and the source series behind them are also summarised in the press release accompanying the report, which carries the three-path band and the sensitivity on the model’s one unpublished input.

Related reading: Global fleet management · Fuel card and fleet payment solutions · Automotive embedded systems · AI in insurance