Kakapo
The Four Kinds of Safety
Myth-buster

The Four Kinds of Safety

A safety score is built from four parts — night, personal, transport and healthcare. But how many genuinely separate things are they? Across 940 cities, night and personal safety turn out to be almost the same number (Pearson r = 0.99), while street crime and healthcare are the loosest pair of all (r = 0.79). "Safety" isn't four things — it's really two: how safe the streets feel, and how well the place is built to help you.

Key findings — free to cite

Across all 940 cities we score on every dimension, night safety and personal safety are almost the same measurement — Pearson r = 0.99. A city that is safe on the street by day is safe on it after dark. They are not two kinds of safety; they are one.

  • At the other end, street crime and healthcare barely predict each other: personal safety vs healthcare is the loosest pair in the whole matrix (r = 0.79). A low-crime city can have thin hospitals; a high-crime one can have world-class ones.
  • The four sub-scores collapse into two clusters. Cluster one is how safe the streets feel — night and personal, moving together at r = 0.99. Cluster two is how well the place is built to help you — transport and healthcare, their own pair at r = 0.96.
  • The two clusters are only loosely linked to each other (the crime-vs-healthcare pairs sit at r = 0.82 and 0.79). That gap is the whole point: a safe-feeling street tells you little about the hospital.
  • The practical read for a traveller: don't treat "safety" as one number with four decorations. It is really two independent questions — is it safe to walk here, and is it built to catch me if something goes wrong — and a place can pass one and fail the other.
  • Computed live from Kakapo's four-part sub-score model — night, personal, transport and healthcare — which no other travel-safety index publishes.

The question: how many things is "safety", really?

Every Kakapo city score is built from four sub-scores — night safety, personal safety, transport safety and healthcare. A companion study, What "Safe" Actually Means, asked how each of those tracks the overall rating. This one asks something more revealing: how do the four relate to each other? If two of them always move in lockstep, they are really one measurement wearing two names. If two of them drift apart, they are genuinely separate things a traveller has to check separately. So we computed the Pearson correlation for all six pairs across every city that has all four sub-scores.

Two of them are basically the same number

The strongest result is almost total. Night safety and personal safety correlate at r = 0.99 — about as close to a perfect one-to-one relationship as real-world data ever gets. Whatever makes a city's daytime streets safe from crime is the same thing that makes them safe after dark; the "night" number and the "personal" number are two readouts of a single underlying reality. The second pair is quieter but clear: transport and healthcare travel together at r = 0.96, an "infrastructure" bond — roads, transit and hospitals all reflect how well a place is built, funded and run.

PairCorrelation (r)Reading
Night ↔ Personal0.99Effectively one measurement. A city that is safe on the street by day is safe on it after dark — the fear of crime and the fear of the walk home are the same signal.
Transport ↔ Healthcare0.96The infrastructure pair. Reliable roads and transit travel with capable hospitals — both reflect how well a place is built, funded and run, not how its streets feel.
Night ↔ Transport0.91Street safety and getting-around safety move together, but not perfectly — the calm-feeling city usually, not always, has the safer roads and transit.
Personal ↔ Transport0.88Day-to-day crime and transport risk broadly align, yet each carries its own danger a headline number can hide.
Night ↔ Healthcare0.82A looser link. How safe the streets feel after dark is only a rough guide to whether a hospital can help you.
Personal ↔ Healthcare0.79The loosest pair in the whole matrix — the genuinely independent signal. Street crime barely predicts medical care: low-crime cities can have thin hospitals, high-crime ones world-class ones.
How to read r. A Pearson r of 1.0 would mean two sub-scores move in perfect lockstep; 0 would mean no linear relationship at all. At 0.99, night and personal safety are effectively the same measurement. At 0.79, street crime and healthcare still lean the same way — but loosely enough that a city can be excellent on one and poor on the other.

The independent signal: healthcare pulls its own way

The two loosest pairs in the matrix both involve healthcare against a crime dimension — night vs healthcare at 0.82, and personal vs healthcare at just 0.79, the weakest link of all six. This is the finding that matters at the airport: how safe a street feels tells you very little about whether a hospital can save your leg. The crime cluster and the care cluster are only loosely tied, which is exactly why a calm, walkable, low-crime destination can quietly be a dangerous place to get hurt — and why medical-evacuation cover is the thing travellers most often skip and most need.

So "safety" is really two questions, not four

Put the six correlations together and the four sub-scores collapse into two groups. Night and personal are one thing: how safe the streets feel. Transport and healthcare are another: how well the place is built to help you. Within each group the numbers move almost as one; between the groups they drift. For a traveller that is a cleaner mental model than a single headline score: ask the two questions separately, because a city can pass one and fail the other.

The outliers: where a tight pairing breaks

The correlations are strong, but individual cities break them — and the breaks are instructive. Splitting the near-perfect night–personal bond takes a genuine crisis: only cities in acute breakdown, like Port-au-Prince and Caracas, show a real gap. Splitting the personal–healthcare pair, by contrast, is common and famous — wealthy cities with a street-crime problem sit on excellent hospitals:

CityPair pulled apartLower subHigher sub
Port-au-Prince, HaitiNight vs PersonalNight 13Personal 26
Caracas, VenezuelaNight vs PersonalNight 5Personal 15
Guatemala City, GuatemalaNight vs PersonalNight 28Personal 37
Johannesburg, South AfricaPersonal vs HealthcarePersonal 24Healthcare 56
San Francisco, United StatesPersonal vs HealthcarePersonal 57Healthcare 86
Cape Town, South AfricaPersonal vs HealthcarePersonal 32Healthcare 60
Paris, FrancePersonal vs HealthcarePersonal 64Healthcare 90
Rio de Janeiro, BrazilPersonal vs HealthcarePersonal 35Healthcare 62

Johannesburg, Cape Town and San Francisco tell the same story from opposite corners of the world: a personal-safety score held down by crime, sitting on top of a healthcare score two or three bands higher. Their overall rating splits the difference and hides both facts. Read the two clusters separately and the real shape of the place appears — risky to be careless in, but well-equipped if you are unlucky.

How we measured it

We queried the live Kakapo index for every city with an overall safety score and all four sub-scores present, deduplicated by city (keeping the largest-population entity and excluding New York's boroughs) — 940 cities in all. For each of the six unique pairs of sub-scores we computed the Pearson correlation coefficient across that full sample, then ranked them from tightest to loosest. Scores draw on national travel advisories from seven governments plus crime, transport and healthcare data; full weighting is at https://kakapo.travel/about/methodology. Scores are comparative model estimates, read live, so the numbers match the rest of the site.

Cite this study

Free to cite in journalism, research or content — all we ask is a link back. Copy the attribution below:

Ready-to-use citation
Kakapo Editorial Team (2026). The Four Kinds of Safety: Which Move Together, Which Are Their Own Thing. Kakapo. https://kakapo.travel/blog/four-kinds-of-safety-2026

Writing about how travel-safety ratings are built? We give journalists custom data pulls at no cost — the full correlation matrix, the pairing breakdown for a specific region, or the city-by-city divergence list. Email [email protected] and we usually reply the same day.

Free to cite & embed

Republish the graphic below — the image links back here automatically. Or pull the raw numbers from the open dataset.

Kakapo publishes real safety scores for every destination, drawing on national travel advisories, local crime data, healthcare infrastructure, and night-safety measures. Kakapo Studies is our research imprint — independent editorial, no advertiser influence on scores. Contact: [email protected].