Skip to content

How We Compare Safety Between Countries and Cities

The rules every safety page on this site is held to — published so you can check our work.

Last verified 2026-08-19Next review 2027-02-192 official sources

We publish a safety comparison only when five conditions are all met. If any one fails, we do not publish the page — we do not publish it with caveats, and we do not publish it with a weaker source.

This page exists so that every safety claim on this site can be checked against a stated rule rather than taken on trust.

The five conditions

  1. Same indicator. Both places measured on the same thing — for example intentional homicides per 100,000 people.
  2. Same source family, comparable years. Figures drawn from the same statistical programme, for the same or adjacent years.
  3. Same geographic level. City against city, or country against country. Never a city rate against a national rate.
  4. Published methodology. The source documents how it counts, so its limitations can be stated alongside the number.
  5. More than one indicator, or an explicit statement that only one is available.

If a fair comparison cannot be made, the page does not get made.

Which sources we use, and how far we trust them

TierSourceUsed for
AUNODC; World Bank intentional homicide seriesCountry-level comparison — the only well-standardised cross-border crime statistic
AGovernment travel advisories (U.S. State Department, FCDO, Smartraveller)Sub-national risk, always quoted with their own wording and date
BGlobal Peace Index (Institute for Economics & Peace)Composite ranking — reported as a measure of peacefulness, not of street crime
BNational police statisticsCity or department level where published. Not valid for cross-country comparison — recording practice differs too much
CNumbeo and similar crowd-sourced indicesPerception only. Never presented as a crime rate, never the basis of a "safer than" claim

The standing caveat

Every safety page on this site carries this, and it is not boilerplate:

Crime statistics are not directly comparable between countries. Recording practices, reporting rates and definitions differ. Homicide is the most consistently recorded category and is used here for that reason. Figures are shown with their year and source.

Reporting rate is the part that does most damage to naive comparisons. A country where people trust the police and report thefts will record more thefts than one where they do not. On paper the first country looks worse. The difference is confidence in institutions, not crime.

Claims we will not publish

  • "Paraguay is the safest country in Latin America."
  • "Asunción is as safe as Europe."
  • "Asunción is completely safe" — or any equivalent for any city.
  • "There is no crime in [place]."
  • Any "safer than" claim resting on a Tier C source.
  • Any statement of a national advisory level without also naming the areas assigned a higher one.

These are not hypothetical. Each is in circulation in this industry, and the last one is the most misleading, because it is technically accurate and materially incomplete.

What a defensible statement looks like

Paraguay is currently rated Level 1 — Exercise Normal Precautions — by the U.S. Department of State, which nonetheless assigns Level 2 — Exercise Increased Caution — to Alto Paraná, Amambay, Canindeyú, Concepción and San Pedro, citing criminal groups active along the north-eastern border with Brazil. Safety in Paraguay varies substantially by department. (Verified [date].)

It is longer than "Paraguay is safe". It is also checkable, which is the point.

What this rules out, for now

City-to-city comparisons involving Asunción. We have not sourced Asunción-level crime data with a stated methodology, and we have not sourced comparable city-level data for European or American capitals on a shared basis. Until both exist, condition 3 and condition 4 both fail, and the comparison pages do not get written.

That is a deliberate constraint on our own content plan. Comparison pages are commercially attractive and easy to produce badly, which is exactly why the rule is written down in advance rather than applied case by case.

How to hold us to this

Every safety page on this site names its sources and the date they were checked. If you find a claim here that fails one of the five conditions above, it is a mistake and we want to know — the standard is meant to be enforceable by a reader, not just declared.

Common questions

Why do you not use Numbeo crime indices?
Numbeo measures perception through crowd-sourced surveys. It is a legitimate measure of how safe people feel and not a measure of recorded crime. We do not use it as the basis of any comparative safety claim.
Why can you not compare Asunción with London?
Only if both figures come from the same indicator, the same source family, comparable years and the same geographic level. Comparing a city figure against a national figure produces a number that looks precise and means nothing.

Sources

Every legal, fee and procedural statement on this page is taken from the sources below and was checked on 2026-08-19. Where they change, this page changes.

Related