The ANN index is used to identify whether the spatial distribution of SWOs is clustered, random, or dispersed by calculating the average distance between each organization and its nearest neighbor (Zertsalov, 2018). The formulas are expressed as Equations 1–3: A N N = D 0 ¯ D E ¯ (1) D 0 ¯ = ∑ i = 1 n d i n (2) D E ¯ = 0.5 1 / A (3) where D ₀ represents the observed mean distance between each SWO and its nearest neighbor; D _e denotes the expected mean distance under a random distribution; d _i signifies the distance; n is the total number of SWOs; and A represents the area of the minimum enclosing polygon covering all SWOs. If the ANN index (i.e., the ratio of observed to expected mean distance) is less than 1, the SWOs exhibit a clustered distribution; if the ANN index is greater than 1, the distribution tends toward dispersion.

KDE was employed to characterize the spatial clustering intensity of SWOs. By generating a smooth surface that reflects the distribution density of point features, KDE visually identifies “hotspot” areas and reveals the hierarchical, multicenter structure of SWO development (Dong and Xu, 2020).

The spatial shifts of SWOs were tracked using the center-of-gravity model (El Deeb, 2021) and the SDE (Lefever, 1926). This approach captures the actual distributional dynamics of SWOs rather than relying solely on geometric centroids, enabling a quantitative description of their spatial expansion and directional tendencies across China.

Following the principles of data availability, quantifiability, and comparability, 13 variables were selected across the dimensions of economic development and policy environment, population and social demand, and social resources and locational characteristics (Table 1). These variables served as explanatory factors for the spatiotemporal differentiation of SWOs in China.

The Geodetector (Wang et al., 2016) was employed to assess the explanatory power (q-value) of various socioeconomic factors in the spatial differentiation of SWOs. To ensure statistical robustness, all continuous variables were discretized into six classes using the Natural Breaks (Jenks) method. Building on this, the GWR model was applied to examine the spatial nonstationarity of these factors further. Unlike traditional OLS regression, GWR allows regression coefficients to vary across geographic locations, thereby capturing localized “heterogeneity” in the driving mechanisms underlying SWO distribution.

From a time-series perspective (Figure 1), the number of newly established SWOs in China between 2000 and 2024 exhibited a distinct three-stage pattern: early development, rapid expansion, and subsequent contraction. During this period, a total of 18,473 new organizations were established, with the eastern region accounting for the largest proportion, 52.2%. In the early development phase (2000–2010), the national annual increment remained below 150 units. Beginning in 2011, SWOs entered a period of rapid expansion, with the annual increments peaking at 2,282 in 2021. From 2022 onward, the number of new organizations declined sharply, reaching 544 by 2024.

Alongside quantitative growth, the spatial distribution of SWOs in China has undergone a fundamental transformation. Results from the ANN analysis (Table 2) indicate that in 2000, the ANN value was 0.877 (p > 0.05), representing a nonsignificant random distribution. By 2010, the ANN value declined to 0.31, indicating a shift toward a clustered distribution pattern. By 2024, the ANN value further decreased to 0.157, suggesting that the spatial polarization of SWOs has become increasingly pronounced. The spatial distribution map (Figure 2) shows that SWOs gradually expanded from sparse concentrations in central cities such as Beijing, Shanghai, and Chengdu to the southeastern coastal regions and other central urban cores. This expansion has produced a prominent “Dense Southeast, Sparse Northwest” pattern, with the distribution boundaries closely aligned with the Hu Huanyong Line (Hu’s Line).

The spatial distribution of SWOs exhibits a clear transition from weak, localized clustering to a stable, multicore agglomeration pattern. Overall, the clustering intensity has steadily increased over time, accompanied by growing spatial dependence among cities.

KDE (Figure 3) reveals a progressive evolution of SWO clustering from isolated, low-density patches to large-scale, contiguous agglomerations. In 2000, only sporadic, low-density centers were observed in the Beijing–Tianjin and the Chengdu–Chongqing regions. By 2010, high-density areas began to concentrate around provincial capitals along the eastern coast. By 2024, four dominant high-intensity clusters had clearly emerged, centered on the Pearl River Delta, Yangtze River Delta, the Beijing–Tianjin–Hebei region, and the Chengdu–Chongqing urban agglomeration, forming a stable multicenter spatial structure with evident spatial connectivity.

Global spatial autocorrelation analysis further confirms the strengthening of clustering patterns (Table 3). The Global Moran’s I increased from 0.0084 in 2000 (p > 0.05), indicating a random spatial distribution, to 0.0975 in 2024 (p < 0.01), reflecting a pronounced and steadily intensifying positive spatial autocorrelation. This trend suggests that SWOs increasingly exhibit spatial spillover and proximity effects rather than behaving independently or being randomly distributed.

Hot spot analysis (Figure 4) provides additional evidence of spatial polarization. By 2024, statistically significant hot spots had expanded into a continuous belt along the eastern coastal region, encompassing Jiangsu, Zhejiang, Fujian, and Guangdong, which now constitute the core concentration zone of SWOs nationwide. In contrast, cold spots persistently dominate western regions, including Xinjiang and Tibet, and their spatial extent has expanded over time. Collectively, these patterns highlight pronounced spatial polarization, characterized by “hot spot agglomeration and cold spot contiguity,” closely aligned with the Hu Line.

The trajectory of the center-of-gravity combines the SDE (Figure 5) collectively illustrates the dynamic spatial patterns of SWO distribution across China. Regarding displacement, the spatial center-of-gravity for SWOs has consistently remained within the border region of Anhui and Henan provinces in Central China. Specifically, in 2000, the center-of-gravity was located in Lu’an City, Anhui Province; by 2010, it had migrated southward to Anqing City, Anhui Province; in 2020, it shifted considerably northwestward to Xinyang City, Henan Province; by 2024, it moved further north to the vicinity of Fuyang City, Anhui Province. This fluctuating trajectory reflects the dynamic adjustments in the relative strength of SWOs across Eastern, Central, and Western China. Notably, the pronounced westward and northward shifts in the later period indicate that the rapid expansion of SWOs in the central and western regions has become a major driver of the national spatial pattern evolution.

Concerning the morphological evolution of the SDE, its major axis has consistently maintained a “Northeast-Southwest” orientation, which closely aligns with the axis of China’s population and socioeconomic development. In terms of spatial extent, the ellipse area underwent a “contraction followed by expansion” process. This pattern suggests that, as social work has progressed nationwide, SWO service networks have diffused from early, localized, high-density clusters to broader geographic areas. Consequently, both the sector’s spatial coverage and its influence on development have been substantially enhanced.

Based on the Geodetector model and Pearson correlation analysis (Figure 6), the study identifies that the spatial differentiation of SWOs is a dynamic process driven by a combination of administrative hierarchy, economic strength, and social demand.

In the factor detection dimension (Figure 6a), X₁₁ consistently emerged as the dominant factor with the highest explanatory power, underscoring the critical role of institutional resource allocation in SWO development in China. Simultaneously, the influence of socioeconomic factors has increased markedly over time. Notably, the q-value of X₇ surged considerably in 2024, while X₃ and X₄ reached q-values of 0.28 and 0.27, respectively. This trend indicates that the spatial logic of SWO distribution is evolving from an early “administrative empowerment” model toward a “demand-pull” approach co-driven by population governance needs and local fiscal capacity. In contrast, the explanatory power of X₁₃ remained extremely low, confirming that the physical geographical environment functions merely as a macro-level backdrop for urbanization and does not directly constrain the distribution of social organizations.

Pearson correlation analysis further clarifies the directional interactions among driving factors (Figure 6b) Except for X₁₃, most factors positively influence SWO numbers. In particular, X₃, X₄, X₇, X_10, and X₁₁ demonstrate strong synergy, with correlation coefficients exceeding 0.6 in 2024. This supports a “trilateral promotion” model that involves policy support, talent reserves, and an administrative hierarchy. Moreover, the correlation between X₇ and SWO numbers increased rapidly, and X₇ also exhibits strong positive associations with X₃ and X₁₁. This suggests that cities with higher administrative levels and robust fiscal strength often pair large-scale migrant populations with intensified governance investments, collectively fostering SWO aggregation and growth. Conversely, X₁₃ displays negative correlations with most economic factors, reflecting how physical environmental constraints, such as elevated construction costs, can inhibit the establishment of social organizations.

To further examine the spatially heterogeneous effects of key driving factors, the GWR model was applied to variables identified as globally significant in the OLS analysis (Table 4): social security and employment expenditure (X₃), average household size (X₉), and the number of colleges offering social work majors (X₁₀). The GWR model substantially outperformed the OLS model in explanatory power, with adjusted R ² values of 0.8630 in 2010, 0.8016 in 2020, and 0.7174 in 2024 (Table 5), indicating strong spatial nonstationarity in the mechanisms driving SWO distribution.

Regarding social security and employment expenditure (X₃), a notable “coefficient paradox” emerges (Figure 7). Although its global OLS coefficient is markedly positive (p < 0.01), local GWR coefficients are predominantly negative across most cities throughout the study period. Specifically, in 2010, strong negative effects were widespread; although this inhibitory intensity weakened by 2024, coefficients largely remained below zero. This pronounced spatial nonstationarity suggests that the global average masks complex, localized dynamics in which increased fiscal input does not linearly translate into organizational expansion.

Average household size (X₉) exhibits the strongest spatial variability among the three factors, with local coefficients alternating between positive and negative across regions. In 2010, negative effects were concentrated primarily in southeastern coastal areas, while many cities in northwestern China displayed weak positive relationships. By 2024, the negative influence of household size expanded markedly across regions southeast of the Hu’s Line, including the middle and lower reaches of the Yangtze River, South China, and parts of Southwest China, indicating a closer spatial association between household miniaturization and SWO agglomeration in these areas.

In contrast, the number of colleges offering social work majors (X₁₀) shows a consistently strong, expanding positive effect. In 2010, high positive coefficients were limited to parts of South and Southwest China. By 2020, these high-impact areas expanded rapidly across most regions, and by 2024, X₁₀ had become the dominant driver in western and Northeast China, where its explanatory power clearly surpassed that of other socioeconomic variables.