Is Rio Safe? What the Neighbourhood Data Shows

Rio publishes what the rest of Brazil does not

Brazil's national statistics stop at the state. Rio de Janeiro's public security institute goes down to the CISP — the area covered by a single police station — and publishes it monthly, as an open CSV, with a companion table translating station numbers into neighbourhood names.

That granularity is unusual anywhere in the world, and it changes what can be said. A state rate mixes a beachfront strip with districts that have nothing in common with it. A station-area rate does not.

It also comes with a counting rule that explains a lot of apparent contradiction between sources: crimes against the person are counted by victim, crimes against property by case. So one incident with three victims is three homicides and one robbery.

In the tourist zone, violent crime is close to absent

Run the 2025 numbers for the southern beach districts and the result is not what a travel advisory prepares you for.

The station area covering Ipanema and Leblon — roughly 95,000 residents — recorded one intentional homicide in the entire year. Counting the wider category that Brazil uses for headline comparisons, violent deaths of all kinds, the rate lands near 3 per 100,000 against 22.6 for the state.

And it is not the lowest in the city. The area covering Gávea, Jardim Botânico, Lagoa, São Conrado and Vidigal recorded zero intentional homicides in 2025, as did Santa Teresa. Several southern districts sit in that band.

Two limits keep this honest. The population denominators come from the 2022 census while the crimes are from 2025, which inflates rates in districts with few residents and large daytime crowds — the extreme case is a central station area with under 2,000 residents, where the arithmetic produces a figure that means nothing. And a single year is noisy: a gap between two districts that looks dramatic in 2025 is substantially smaller averaged over recent years.

The same district, near the top for phone crime

Now the inversion. Ipanema and Leblon recorded roughly 2,900 phone crimes in 2025 — a rate above 3,000 per 100,000, among the highest in Rio.

The composition is the point. In that district, phones taken without confrontation outnumbered phones taken by force by roughly nine to one: about 2,600 thefts against 287 robberies. In Copacabana the ratio is similar. Move to certain outer districts and it flips — one station area recorded more robberies than thefts, the opposite balance.

So "phone crime in Rio" is two different events wearing one label. In the tourist strip the realistic scenario is a phone leaving your hand or bag without you noticing. Elsewhere it is a confrontation. The same statistic covers both, which is why a city-wide number is useless for deciding how to behave.

Across the whole state, 2025 brought about 25,600 phone robberies and 46,600 phone thefts — theft leading by nearly two to one.

Express kidnapping has a number, and it is small

Sequestro relâmpago — being held for a few hours and driven to cash machines — appears in every foreign advisory about Brazil, always in prose, never with a figure.

The figure exists. In the state of Rio de Janeiro in 2025 there were 137 victims. Kidnapping in general, a broader category, recorded 1,817.

That is not zero and the crime is real. But a risk described in four government documents and quantified in none tends to acquire a size in the reader's head that the data does not support. This is the number.

Three official homicide figures, same state, same year

Worth knowing before you compare two articles and conclude one is lying.

For homicide in Rio de Janeiro in 2025, the federal ministry publishes one figure, and Rio's own institute publishes two — one from its own base and one prepared to the federal standard, which is not identical. Three official numbers, two bodies, one crime, one year.

The institute is transparent about the reason: after the raw stage it runs additional quality processing and aggregation, so its own series and the one it exports do not have to match. None of the three is fraudulent, and the same divergence exists nationally — Brazil publishes two national figures and both are correct.

Reading it yourself, and what will trip you up

The base is genuinely open, which is rare enough to use. Two practical warnings if you download it.

The CSV quotes fields that contain semicolons, and its separator is also a semicolon. Split on the delimiter naively and a share of rows shatter into the wrong number of columns — our first pass produced a state total tens of thousands of times too large before we used a proper CSV parser. If a number comes out absurd, this is why.

And station numbers are not neighbourhood names. The institute publishes the translation table separately; without it, the base is a list of integers.

For the practical side of moving around the city, the useful reading is elsewhere: ride apps and how they are regulated, getting cash without a local card and which ATM to use — express kidnapping is a cash-machine crime, so where and when you withdraw matters more than the headline rate. The strongest mitigation is not needing a machine at all, which is what paying by PIX removes. How much cash you may bring in is a separate rule, and carrying less of it is the simplest mitigation there is. A working phone underpins every one of these instructions, so a local SIM is not a convenience. And what your own government says about Brazil is worth reading with its publication date in view.

Sources

• ISP-RJ, monthly base by police-station area — the raw series behind every district figure here, summed over the twelve months of 2025. Accessed September 3, 2026.
• ISP-RJ, station-to-neighbourhood table — which station number covers which neighbourhoods. Accessed September 3, 2026.
• ISP-RJ, methodology — the counting rule for crimes against the person versus property, and the note on additional processing. Accessed September 3, 2026.
• ISP-RJ, municipal base — state totals for phone robbery and theft. Accessed September 3, 2026.
• Mapa de Segurança Pública 2026 (Ministry of Justice) — the federal figure for Rio, for the three-way comparison. Accessed September 3, 2026.

What we could not verify

• How much of the tourist-zone phone crime involves visitors rather than residents. The base records the offence and the location, never the victim's nationality or residence.
• The remaining share of the divergence between the three homicide figures. The institute's own note explains part of it; a residual we could not account for stays unexplained.
• Current-year population by station area. The denominators available are from the 2022 census, which is what limits the district rates above.
• Whether the composition note behind any single district holds year to year. We summed 2025; the multi-year averages we checked move the ratios substantially, so treat one year as one year.