The study’s methodological backbone is a first-differenced panel vector autoregression estimated by generalized method of moments, a combination well suited to the kind of data the authors assembled. Because the dataset is unbalanced—meaning not every province contributes observations for every year between 1993 and 2023—the GMM framework allows the model to extract dynamic relationships from incomplete panels without discarding large amounts of information. First differencing, meanwhile, strips out province-specific characteristics that persist over time, such as geography, climate, and enduring institutional features, so that the estimated coefficients reflect how changes in one variable relate to changes in another rather than static correlations between levels.

This design choice matters for interpreting the headline findings. When the authors ran the baseline model without controlling for nationwide time effects, agricultural civilization development appeared to positively predict subsequent changes in all three land use behaviors: production expansion, non-agricultural conversion, and ecological restoration. Industrial civilization development similarly predicted non-agricultural conversion. But once common national shocks—macroeconomic cycles, nationwide policy campaigns, economy-wide price movements—were accounted for, those development-to-land-use linkages weakened substantially. What survived the stricter specification was the reverse direction: land use behaviors feeding back into agricultural civilization development, with production expansion showing a negative association and non-agricultural conversion a positive one.

The asymmetry is theoretically meaningful. It suggests that much of what looks like civilization driving land use in naive models may actually reflect shared national trends moving both series together, while the genuine causal arrow from land to civilization is more robust. In practical terms, how provinces actually use their land—whether they convert farmland to construction or restore degraded ecosystems—appears to shape the trajectory of agricultural civilization more reliably than civilizational indicators shape land use decisions. This resonates with a long line of scholarship stretching back to Ester Boserup’s 1965 challenge to Malthusian thinking, which argued that human demands drive land use transformation and that optimized land use in turn enables civilizational progress. The present findings refine that two-way logic by showing that the feedback loop is not symmetric in strength or in sign.

The negative association between production expansion and subsequent agricultural civilization development deserves particular attention. One plausible reading, consistent with the study’s framing, is that continued expansion of production-oriented land use—extensive cultivation, intensification of farming footprints—may eventually strain the resource base on which agricultural civilization depends, degrading soil, water, and ecological conditions that underpin long-term agrarian prosperity. Conversely, non-agricultural conversion, by shifting land out of farming and into urban, industrial, and infrastructural uses, may accompany structural transformation that raises productivity, income, and the broader developmental indicators captured in the agricultural civilization index. The authors are careful to frame these as dynamic associations within a vector autoregressive system rather than simple causal claims, but the pattern aligns with the idea that land allocation decisions feed back into the quality and sustainability of civilizational development.

Ecological civilization development stands out as the odd variable in the system. Despite the prominence of ecological civilization as a policy concept in China—formalized in national discourse over the past two decades and enshrined in constitutional and planning language—the study finds comparatively limited dynamic linkages between ECD and any of the three land use behaviors. This may reflect the recency of ecological civilization as an operational framework: if its indicators only began moving meaningfully in the later portion of the 1993–2023 window, the panel model has fewer effective observations over which to detect strong feedback. It may also reflect the difficulty of measuring a multidimensional concept—low-carbon development, harmonious human–land relations, conservation orientation—with the kind of provincial-level quantitative proxies available for agricultural and industrial dimensions.

The temporal breakpoint analysis around 2013 provides a natural experiment in policy regime change. The year 2013 marks the launch of sweeping national reforms, including intensified ecological redlining, stricter farmland protection, and the elevation of ecological civilization as a governing principle. The analysis suggests that after this point, China’s land use system shifted gradually away from expansion-oriented behavior toward a more coordinated balance among production, construction, and ecological functions. In the vocabulary of the study’s framework, the post-2013 era shows the three behaviors no longer dominated by a single expansionary logic but adjusting in a more balanced configuration—a finding consistent with the broader observation that Chinese land governance has moved from quantity-driven growth toward quality- and function-driven management.

Regional heterogeneity adds another layer of nuance. Eastern China, the most urbanized and economically advanced region, exhibits relatively stable adjustments in the interaction system, suggesting mature institutional capacity to absorb shocks and translate policy signals into land use responses. Central China, often the grain belt and the frontier of industrial transfer from the coast, shows stronger sensitivity, with land use behaviors responding more sharply to changes in civilizational indicators. Western China, with its fragile ecosystems, lower population density, and greater reliance on state-led investment, displays greater volatility, implying that land use decisions there are more susceptible to swings in policy emphasis and external shocks. These regional differences argue against one-size-fits-all land governance and support the study’s call for regionally differentiated strategies.

The impulse response and forecast error variance decomposition analyses flesh out the system’s internal dynamics. Impulse responses reveal that shocks to any variable tend to dissipate relatively quickly—the system is characterized by short-lived responses rather than persistent, self-reinforcing spirals. This is a double-edged finding: it means disruptive land use shocks do not compound indefinitely, but it also means policy interventions may need to be sustained rather than one-off to produce lasting change. Variance decomposition, which attributes the forecast uncertainty of each variable to shocks in the others, shows differentiated transmission patterns: some land use behaviors are driven more by their own inertia, while others are meaningfully influenced by civilizational variables, and the relative importance differs across regions and periods.

Granger causality tests, though often misread as proof of causation, serve here as a screening device to identify which directional relationships merit attention in the richer PVAR framework. Their role is to establish predictive precedence—whether past values of one variable improve forecasts of another—before the model quantifies the magnitude and persistence of those linkages. The combination of these tools, applied to a three-decade provincial panel, gives the study unusual temporal depth for land use transition research, which has often relied on cross-sectional comparisons or shorter panels.

The conceptual contribution lies in operationalizing civilization development as a multidimensional rather than monolithic phenomenon. Classical stage theories divide human history into hunting-gathering, agricultural, industrial, and ecological phases, but most empirical work treats development as a single economic trajectory. By constructing separate indices for agricultural, industrial, and ecological civilization development and letting each interact dynamically with distinct land use behaviors, the study reveals that these dimensions do not move in lockstep with land systems. Agricultural civilization development maintains the strongest and most persistent bidirectional ties, industrial civilization development primarily stimulates non-agricultural conversion, and ecological civilization development remains comparatively detached—each with distinct policy implications.

For land governance, the practical upshots are concrete. The robustness of land-to-civilization feedback suggests that managing land use is not merely an instrument for sectoral goals but a lever for shaping the developmental trajectory itself. The post-2013 transition indicates that policy regimes can bend the system’s direction, but the short-lived nature of responses implies durability requires sustained institutional commitment. The regional heterogeneity implies that Eastern provinces may benefit from refinement and consolidation of existing coordination mechanisms, Central provinces from buffers against overreaction to civilizational fluctuations, and Western provinces from stabilization measures that dampen volatility in ecologically sensitive land. More broadly, the study demonstrates that land use transition theory, which has traditionally emphasized socioeconomic and ecological drivers, gains explanatory power when civilization development is treated as a dynamic, multidimensional force whose feedback with land systems evolves across decades, regions, and policy eras.

The three-decade window also captures China’s most dramatic land transformations: the explosive urban expansion of the 1990s and 2000s, the farmland consolidation and grain-for-green restoration programs, the construction boom following WTO accession, and the ecological turn of the 2010s. That a single statistical framework can accommodate such heterogeneity—and still detect structured, asymmetric, regionally patterned relationships—underscores both the value of long panels and the complexity of the human–land system the authors set out to measure. Their framework offers a template that could be adapted to other rapidly transforming economies where civilization-scale development goals and land allocation decisions are locked in mutual, but unequal, feedback.

Subject of Research: Dynamic feedback between multidimensional civilization development and urban–rural land use behaviors in China

Article Title: Dynamic interactions between multidimensional civilization development and urban–rural land use behaviors in China

Article References: Chen, X., Ling, L., Chen, Y., Kang, J., & Xia, W. (2026). Dynamic interactions between multidimensional civilization development and urban–rural land use behaviors in China. Frontiers in Forests and Global Change, 9, Article 1936306. https://doi.org/10.3389/ffgc.2026.1936306

Image Credits: AI Generated

DOI: 10.3389/ffgc.2026.1936306

Keywords: land use transition, civilization development, ecological civilization, panel vector autoregression, China, urban–rural transformation, sustainable land governance, Granger causality, regional heterogeneity, ecological restoration, cropland conversion, PVAR model