What makes a city attractive?
Every year we see new lists of the “best cities” in the world. They consider everything from safety and schools to restaurants, salaries and cultural life. But who decides what matters most—and how much?
Why do we call it the City Attractiveness Index?
Most city rankings evaluate places using selected indicators such as economic opportunity, safety, affordability, education, and quality of life. These approaches provide useful information, but choosing the indicators and assigning their relative importance inevitably introduces subjective judgments. Surveys also measure what people say they prefer, which may differ from what they actually do.
CAI takes a different approach: instead of deciding what makes a city attractive, we examine the outcome.
People move to cities, stay in them, or leave them for many reasons. Jobs, housing costs, family, climate, infrastructure, and personal circumstances all influence these decisions. CAI seeks to capture their combined demographic effect rather than assigning separate weights to each factor.
This does not mean that everyone moves freely. Housing shortages, financial constraints, displacement, and other circumstances influence where people can live. Consequently, CAI measures realized population attraction under real-world conditions—not an idealized measure of desirability.
CAI estimates this attraction by measuring population growth beyond what would be expected from births and deaths, with adjustments for demographic composition. The index is most useful for comparing cities within the same country, where many national economic, legal, and institutional conditions are shared.
Start with behavior
Look at population changes rather than asking people to rate dozens of city features.
Remove the expected
Allow for births and deaths, including the fact that younger and older cities have different demographic profiles.
Compare fairly
Compare cities primarily within one country, where national laws, visas, economic systems and institutions are more alike.
Why comparisons within the same country matter
A city in one country may grow faster or slower than a city in another because of national economic conditions, borders, immigration rules, conflicts, currency differences or government policy. Those differences can say more about the country than about the individual city. Comparing cities in the same country reduces some of those confounding factors, although regional economics and moving constraints still matter.
For that reason, the explorer shows each city's rank within its country. The global search and sorting features are useful for research, but a worldwide numerical comparison is not our primary claim.
From a choice to a number
Simply counting how fast a city's population grows would be misleading. A young city can expand because many children are born, while an older city may shrink even if people continue moving in. CAI begins with five-year population growth and subtracts the natural change expected from births and deaths. What remains is a migration-related demographic residual—an estimate, not a direct count of newcomers and not proof of individual preferences.
The calculation: data, assumptions and limitations
CAI estimates a migration-related population residual, not the actual number of people who moved or the reasons for their decisions.
1. Measured growth: consistent city footprint
We use European Commission GHSL 2020 and 2025 gridded population estimates (GHS-POP R2023A, 1 km Mollweide) clipped to the same fixed city polygon from our selected 4,975-city collection. This limits bias from shifting urban-centre boundaries. Figures are model estimates, not census headcounts. The annualized geometric growth rate is computed over five years.
2. City age and sex composition
WorldPop's 2020 estimates provide 20 age groups, from infants to 90+, separately for women and men, for each city polygon. All 4,975 age/sex records passed structural validation. Eight cities have differences greater than 1% between the reported WorldPop population and the sum of age groups, and are visibly flagged.
3. Expected births and deaths: age-mix proxy v1
The UN World Population Prospects 2024 demographic indicators workbook supplies national crude birth rates and crude death rates, averaged for calendar years 2020–2025. It does not supply age-specific birth and mortality schedules in the demographic input workbook.
Instead, this model uses an explicit approximation: each city's age structure produces a fertility-exposure index (female ages 15–49, with greater weight for ages 25–34) and a mortality-exposure index (both sexes, rising markedly with age and with an infant adjustment). Each index is compared against the population-weighted pooled urban-age structure of cities in the same country. National crude birth and death rates are scaled by those exposure ratios, within 0.4×–2.5× guardrails to prevent extreme extrapolation.
National rates are in events per 1,000 people per year; dividing by 10 converts them to percentage points annually. The country reference is the cities in this dataset, not the whole country. One-city countries use a ratio of 1 and therefore receive the national crude rates directly. Relative age weights are heuristic and have not been independently fitted to country-level age-specific demographic schedules.
4. Inputs, evidence and limitations
| Input | Actual source | Role |
|---|---|---|
| Population 2020/2025 | GHSL GHS-POP R2023A population rasters | Fixed-polygon annualized growth |
| City boundaries | Selected GHS-UCDB 2024 urban-centre polygons | Identical clipping footprint for both years |
| Age and sex | WorldPop 2020 | Age-mix weights and comparison plots |
| National births and deaths | UN WPP 2024 GEN F01, 2020–2025 | National crude demographic baseline |
These city boundaries need not coincide with administrative city limits. The raster and age datasets are modeled and originate from different sources. Natural growth is a modeled approximation; the resulting residual may reflect migration, census/model error, commuting and temporary-population effects, or other influences. Results are not a measured flow of migrants, do not establish causation, and cannot distinguish voluntary movement from forced displacement.
5. Data checks and scientific limitations
The CAI data checks verify file integrity, unique IDs, complete age bands, fixed-boundary population extraction, national-rate joining, and repeatability of its arithmetic. It does not claim independent validation of the inferred CAI residuals or of the heuristic age-specific relative weights. In particular, a formal age-adjusted estimate will require WPP/other documented country age-specific fertility and mortality schedules, preferably combined with an international plausibility check.