An increasing amount of empirical evidence documents that city-size distribution within a country follows a power law, often in the form of Zipf’s law. This paper provides new comparative evidence on city size distribution across OECD countries. It uses a database where urban agglomerations are consistently identified across different countries, through an algorithm based on population density and commuting patterns. The paper investigates whether Zipf’s law fits well with data. A robustness check is carried out using a traditional administrative definition of cities. Results show that Zipf’s law describes well city size distribution not only at country level, but also at wider spatial scales. The law does not fit as well with the data when using a traditional administrative definition of cities.