City Population Annual Timeseries (UN) - Female/Male
- Rows
- 50,737
- Columns
- 11
- Data range
- 1970-01-01 to 2025-01-01
- Auto-refresh
- Re-checks its source every 30 days
UN city population annual time-series by sex (female/male). Large demographic dataset from UN Statistics Division.
AI analysis
Dataset Analysis: UN City Population Annual Timeseries (Female/Male)
1. Dataset Overview & Research Value
This dataset, sourced from the UN Statistics Division via [datahub.io](https://datahub.io/core/population-city/r/-/data/unsd-citypopulation-year-fm.csv), represents one of the most comprehensive longitudinal demographic resources available at the city level, spanning 55 years (1970–2025) across 188 countries and 4,995 cities. The sex-disaggregated structure (Female/Male split per observation) makes it particularly powerful for gender-demographic correlation studies — enabling analysis of urbanization patterns, sex ratio shifts, and city-level population dynamics over time. The dataset's breadth across both developed and developing nations creates strong potential for cross-referencing with economic, health, and infrastructure indicators. The Source Year column ranging through 2025 suggests the data includes recently published or projected figures, adding near-term forecasting value.
---
2. Data Quality Observations
Overall data quality is mixed but workable. Across 11 columns and 50,737 rows, there are 39,288 null cells total, which is significant but heavily concentrated in just two columns. The Value Footnotes column is the primary concern, with 36,602 nulls (72.1%) — this column is effectively sparse metadata and should be treated as optional/supplementary rather than analytically load-bearing. The Value column (the core population figure) carries 2,686 nulls (5.3%), which warrants attention: these missing population counts could reflect unreported city data, boundary changes, or suppressed figures for small populations, and rows with null Value should be excluded or flagged in any quantitative analysis. Notably, duplicate row counts are pending Phase D recomputation, so deduplication validation — especially given the multi-key structure (Year × Country × City × Sex) — should be confirmed before modeling. No type mismatches are flagged among the join-key columns, which is a positive signal for schema integrity.
---
3. Key Column Distributions
The Value column (city population by sex) is the analytical heart of the dataset and shows extreme right skew (skew = 9.54), which is expected for urban population data — a small number of megacities dwarf the majority. The mean of 226,266 versus a median of 98,959 confirms this skew, and the Q1–Q3 range of 64,310–191,918 captures the bulk of mid-sized cities, while the maximum of ~11.85 million and 5,621 outliers represent major metropolitan centers. Log-transformation will be essential before any regression or correlation work. The Source Year column is notably well-behaved by contrast: mean 2013.6, median 2013, σ of 6.77, and mild negative skew (-0.107), indicating data is concentrated in the 2000s–2010s with modest but manageable temporal spread — and only 38 outliers, suggesting a few very early or very late source references. The Sex column with exactly 2 distinct values and zero nulls confirms clean binary disaggregation throughout.
---
4. Recommended Join Key Columns
The four designated join-key columns — Year, Country or Area, City, and Sex — form a natural composite key for linking to external datasets. For cross-dataset joins: - Country or Area + Year is the most portable join pairing, compatible with country-level economic or health datasets using standard UN country names. - City + Country or Area + Year enables city-level joins but requires careful name normalization (spelling variants, transliterations) given 4,995 distinct city names. - Sex serves as a stratification key rather than a join key, enabling gender-disaggregated correlations when paired with similarly sex-disaggregated external data. - The City type column (2 distinct values) may help align with urban classification schemes in other datasets and is worth preserving in joins.
---
5. Suggested Paired Datasets for Correlation Studies
Several dataset categories would pair powerfully with this data:
| Paired Dataset | Correlation Opportunity | |---|---| | GDP per capita by country/year (World Bank, IMF) | Urbanization rate vs. economic development trajectories | | UN Gender Inequality Index or women's labor force participation | Female urban population share vs. gender equity indicators | | City-level infrastructure / housing indices | Population density vs. infrastructure investment or housing costs | | Climate/emissions data by city or country | Urban population growth vs. CO₂ emissions or urban heat trends | | Health outcomes (WHO mortality, disease prevalence) | Sex-disaggregated urban population vs. gendered health metrics | | Migration and refugee statistics (UNHCR) | City population volatility vs. displacement events |
The strongest immediate pairing would be World Bank country-level economic indicators, as the Country or Area + Year join key is directly compatible and would enable rich urbanization-vs-development correlation analysis across the full 1970–2025 window.
Columns
- Year (date)
- Country or Area (string)
- City (string)
- Sex (string)
- Area (string)
- City type (string)
- Record Type (string)
- Reliability (string)
- Source Year (integer)
- Value (integer)
- Value Footnotes (integer)