Yu Li and Wei Qi of the Institute of Geographic Sciences and Natural Resources Research at the Chinese Academy of Sciences, together with Raya Muttarak of the University of Bologna, drew on the nationally representative 1% census micro-sample data from 2010 and 2020. These micro-samples, authorized by the National Bureau of Statistics of China, allow researchers to move beyond provincial averages and estimate adoption rates at the prefecture level, separately for urban and rural populations within each prefecture. That granularity matters, because a single average for a prefecture can hide two very different worlds: a gas-connected city and surrounding villages still burning coal or crop residues.

The headline numbers tell a story of two transitions running at different speeds. In urban areas, the share of households relying on clean cooking fuels rose from 71.0% in 2010 to 94.2% in 2020, a level the authors describe as near-universal reliance and one that consolidated alongside sustained socioeconomic development. Rural areas moved too, climbing from 22.9% to 63.1%, but that still leaves more than a third of rural households outside the clean-fuel fold. In relative terms the rural gain was enormous, nearly tripling adoption in a decade, yet the absolute distance from the urban benchmark remains wide.

That distance narrowed, but not evenly. The mean absolute gap between urban and rural adoption rates within prefectures fell from 48.1 percentage points in 2010 to 31.1 percentage points in 2020. A narrowing gap is good news, and it reflects genuine rural progress rather than urban stagnation. But the authors emphasize that the decline in inequality was not uniform across the map. Disadvantages persisted in northern and western prefectures, where rural adoption rates remained stubbornly low even as coastal and southern regions converged toward their urban neighbors. In other words, the clean cooking transition is not simply an urban-versus-rural divide; it is an urban-rural divide whose depth varies by region.

To understand what drives these divergent trajectories, the team turned to machine learning. They trained leakage-free XGBoost models, a gradient-boosted decision tree method prized for its ability to capture nonlinear relationships and interactions among many predictors, and evaluated them with repeated nested cross-validation. Nested cross-validation separates model selection from performance estimation, reducing the risk of optimistic bias, while the leakage-free design prevents information from the test folds from contaminating the training process. The target variables were the urban and rural adoption rates and the gaps between them, and the predictors were prefecture-level contextual features spanning socioeconomic and demographic conditions.

The modeling results carry two important caveats that the authors are careful to state. First, population density and migration emerged as consistently important model features, appearing as influential predictors across the outcomes examined. Densely settled places, with their economies of scale for pipeline gas networks and their pull on labor and investment, appear structurally advantaged in the transition, while areas shaped by out-migration face a different set of constraints, from shrinking demand for local infrastructure to remittance-dependent household budgets. Second, and crucially, predictive performance varied across outcomes: the models explained some measures of adoption and inequality better than others. Contextual variables, in short, are informative but not determinative, and no single factor explains why one rural prefecture electrifies its kitchens while another does not.

The study’s framing sits within a well-established literature on the energy ladder, the idea that households climb from traditional biomass through transitional fuels to modern clean energy as incomes rise. China’s experience complicates the simple version of that ladder. Urban households have effectively reached the top rung, but rural progress, while real, has plateaued well short of universality. The persistence of solid fuel use in the countryside is not merely an inconvenience; cooking with coal, wood, and crop residues indoors is associated with household air pollution, a major health burden, and it falls disproportionately on the women and elderly people who spend the most time at the stove. A transition that stalls at 63% rural adoption therefore leaves a substantial equity and public health gap embedded in the national energy statistics.

Why would northern and western prefectures lag? The data point to structural context rather than any single cause. Population density and migration, the two consistently important features, cut in a direction that disadvantages exactly these regions: they tend to be less densely settled and more affected by labor outflows to eastern cities. Infrastructure economics reinforce the pattern, since extending piped gas or robust distribution grids across dispersed rural settlements costs far more per household than serving compact urban blocks. Affordability compounds the problem where incomes are lower, because even when clean fuel is physically available, the recurring cost of gas or electricity can deter households from abandoning free or cheap local biomass. The authors do not claim to have isolated causal mechanisms, and their machine learning approach identifies associations, not causes, but the pattern is consistent with a transition that follows the path of least infrastructural and economic resistance.

The policy implication the authors draw is an integrated urban-rural strategy rather than a purely rural one. They point to three pillars: shared infrastructure, affordability protection, and place-specific support for lagging transitions. Shared infrastructure means designing gas grids, electricity distribution, and delivery networks so that urban expansion can be leveraged to serve adjacent rural communities rather than stopping at the city boundary. Affordability protection acknowledges that the last third of rural adopters are likely the hardest to reach and the least able to pay, requiring subsidies or tariff designs that keep clean fuel competitive with the biomass it replaces. Place-specific support recognizes that the northern and western prefectures where disadvantages persist will not respond to the same policy levers that worked in the wealthier east, and that a uniform national playbook will leave the same regions behind in the next decade as it did in the last.

The decade between the 2010 and 2020 census rounds was, by any standard, transformative for Chinese household energy. Urban near-universality was achieved, rural adoption more than doubled, and the urban-rural gap shrank by seventeen percentage points on average. But the study’s prefecture-level lens shows that averages can flatter a transition. Behind the national numbers lie persistent regional pockets where rural households remain dependent on polluting fuels, and where the drivers of adoption, density and migration among them, are not moving in the transition’s favor. As China pursues carbon neutrality and rural revitalization in tandem, the authors’ central finding stands as a warning and a guide: sustainability transitions do not distribute their benefits automatically, and closing the last, hardest gap will require deliberately connecting urban infrastructure, rural incomes, and the specific geographies where the energy ladder still has its lowest rungs.

Subject of Research: Urban-rural inequality in household clean cooking fuel adoption in China

Article Title: Inequality in clean cooking adoption across urban and rural China

Article References: Li, Y., Qi, W., & Muttarak, R. (2026). Inequality in clean cooking adoption across urban and rural China. npj Urban Sustainability. https://doi.org/10.1038/s42949-026-00478-y

Image Credits: AI Generated

DOI: 10.1038/s42949-026-00478-y

Keywords: clean cooking, China, energy transition, urban-rural inequality, household energy, census data, XGBoost, machine learning, prefecture-level analysis, rural development, energy policy, household air pollution