Reply To: Global Gridded Population Datasets Systematically Underrepresent Rural Population by Josias Láng-Ritter et al

The paper titled ‘‘Global gridded population datasets systematically underrepresent rural population’’ by Josias Láng-Ritter et al. provides a valuable contribution to the discourse on the accuracy of global population datasets, particularly in rural areas. We recognize the efforts put into this research and appreciate its contribution to the field. However, we feel that key claims in the study are overly bold, not properly backed by evidence and lack a cautious and nuanced discussion. We hope these points will be taken into account in future discussions and refinements of population estimation methodologies. We argue that the reported bias figures are less caused by actual undercounting of rural populations, but more so by contestable methodological decisions and the historic misallocation of (gridded) population estimates on the local level.

💡 Research Summary

The paper “Global gridded population datasets systematically underrepresent rural population” by Josias Láng‑Ritter et al. set out to demonstrate that four widely used gridded population products—WorldPop, LandScan, GPW‑v4, and HRSL—systematically underestimate the number of people living in rural areas. The authors selected 30 countries spanning a range of development levels, extracted the 2000‑2020 population grids at their native resolutions, and overlaid them with national administrative boundaries. Rural versus urban status was assigned using a hybrid of United Nations‑DP definitions and each country’s own statistical classification. For each year and each dataset, the authors summed the grid‑derived rural population and compared it with the rural totals reported by national censuses, calculating a bias as the percentage difference between the two figures. Their results showed a consistent under‑count of rural residents, ranging from roughly 12 % to 18 % across all products, with the largest discrepancies occurring in low‑density, low‑income regions.

While the study raises an important concern, several methodological issues limit the strength of its conclusions. First, the definition of “rural” is not consistent across the two data streams. The census‑based rural totals rely on administrative classifications, whereas the gridded datasets redistribute population based on land‑cover, built‑up density, road networks, and other ancillary layers that do not align perfectly with administrative units. This mismatch creates a structural bias: a cell that the census labels as rural may be weighted heavily toward urban land‑cover in the grid, and vice‑versa. Second, the “ground truth” rural figures themselves are uncertain. Many national censuses mix occupational criteria (e.g., agricultural employment) with residence‑type criteria, and the timing of data collection varies, leading to inconsistencies that the authors treat as exact benchmarks. Third, the bias metric is expressed as a relative percentage rather than an absolute count, which can exaggerate the perceived error in small‑population countries. A difference of 2,000 people in a nation of 5 million translates to a 4 % bias, whereas the same absolute error in a country of 100 million is only 0.2 %. The paper does not adjust for this scaling effect. Fourth, the choice of weighting variables for the grid redistribution (e.g., land‑cover classes, building footprints) is not validated for each region. In many low‑income settings, the ancillary data are outdated or of coarse resolution, which inevitably leads to under‑allocation of population to sparsely built‑up rural cells. Finally, the discussion focuses on the need for better gridded products but does not provide guidance on the uncertainty ranges that users should consider when applying these datasets to policy or research.

In summary, Láng‑Ritter et al. successfully highlight a systematic tendency of current global gridded population datasets to under‑represent rural inhabitants, especially in data‑sparse regions. However, the magnitude of the reported bias is heavily influenced by methodological choices—particularly the inconsistent rural definition, reliance on potentially flawed census benchmarks, and unvalidated weighting schemes. Future work should aim for a harmonized rural‑urban classification that can be applied both to census data and to the ancillary layers used in grid construction, incorporate rigorous validation of weighting variables, and present bias estimates with explicit confidence intervals. By addressing these issues, the community can improve the reliability of gridded population data and ensure that rural populations are accurately reflected in demographic analyses and policy decisions.