Synthetic Household Populations for 109 Countries and Two Cities

Global, Newcastle upon Tyne, and Dar es Salaam experiments provide data artefacts for a sample-free population synthesis framework using large language models. The 18.1 MB dataset includes inputs, outputs, prompts, and evaluation results from experiments across 109 countries and two detailed city case studies. Author Michael Jones published the data on figshare in April 2026 under a CC-BY-4.0 license.

Use Cases

• Benchmarking LLM performance on demographic synthesis tasks based on the described LLM benchmark experiment.
• Developing agent-based models for urban planning based on the household-structured synthetic populations.
• Evaluating sample-free population synthesis methods across diverse demographic contexts based on the global evaluation across 109 countries.
• Studying joint demographic patterns in a UK urban context based on the Newcastle upon Tyne case study using ONS 2021 Census data.

Strengths

• Includes evaluation results for multiple commercial and open-source LLMs (GPT-4o, Grok-3, Llama variants).
• Covers a broad geographic scope with experiments across 109 countries and two detailed city case studies.
• Provides all inputs, outputs, prompts, and configurations needed to reproduce the paper's experiments.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to the selected case study locations and available demographic inputs.

Provenance

Source

Data artefacts supporting the paper 'A large language model framework for sample-free population synthesis'.

Collection Method

Generated via an LLM-agnostic framework using aggregate demographic data (marginal distributions) without microdata or model fine-tuning.

Freshness

Last updated 2026-04-17 11:40:51; freshness should be verified.

Geography

Global (109 countries), Newcastle upon Tyne (UK), and Dar es Salaam (Tanzania).