Spatial patterns and their influencing factors for China's catering industry

The study addresses how the spatial distribution of restaurants in China relates to urbanization, economic development, consumption capacity, and cultural openness. Prior work indicates that population concentration, income, and urban form shape restaurant clustering and diversity, with globalization making food consumption both an economic and cultural phenomenon. The authors posit: (1) the number of urban restaurants is mainly affected by the level of economic development and the size of the urban population; (2) restaurant types are additionally influenced by city openness, residents’ habits, and other socio-cultural factors; and (3) spatial distributions are further shaped by city characteristics (e.g., tourism). The study aims to quantify spatial distribution patterns of four restaurant types, evaluate spatial dependence, and statistically test heterogeneous drivers across China to inform urban planning and service sector strategies.

The paper reviews research on restaurant location and clustering, noting links to income and proximity (Eckert & Vojnovic, 2017), CBD-centered clustering and spillovers (Prayag et al., 2012), population-driven diversity (Dong et al., 2019), and density effects on dispersion by quality tier (Mossay et al., 2022). Studies document aggregation from diverse consumer groups and co-location patterns (Kim et al., 2020; Zhang et al., 2021b; Han et al., 2024). Work on food culture and globalization shows that food consumption reflects cultural interactions and openness (Goto et al., 2014; Tricarico & Geissler, 2017; Tian et al., 2021; Yigit, 2022). Tourism, gastronomy, and local foodscapes are emphasized as drivers of urban food spaces and destination appeal (Cohen & Avieli, 2004; Mak et al., 2012; Jiménez-Beltrán et al., 2016; Zhang et al., 2022). Chinese studies using POIs explore spatial patterns and influencing factors of catering industries and urban structures (Xia et al., 2018; Lu et al., 2020; Wang et al., 2022; Zhang et al., 2021a,b). The review highlights a gap in comprehensive, nationwide statistical testing of heterogeneous drivers by restaurant type.

Data: Restaurant POIs were sourced from AMAP (Gaode) in March 2020 (6.09 million POIs), filtered to 4.49 million valid records across 336 mainland Chinese cities (excluding Taiwan, Hong Kong, and Macau). Restaurants were categorized into four types: Chinese restaurants (CRs), Western restaurants (WRs), fast-food restaurants (FFRs), and dessert and drink restaurants (DDRs). Socio-economic indicators (10 total) were compiled from the China Statistical Yearbook 2021 and China City Statistical Yearbook 2021: urban population, population density, urbanization rate, GDP per capita, GDP per unit area, proportion of tertiary industry in GDP, GDP (aggregate), total tourism revenue, total retail sales of consumer goods, and per capita disposable income of urban households. Spatial pattern analysis: Kernel density estimation (Gaussian kernel; bandwidth via reference method; search radius 2 degrees) was implemented in ArcGIS to identify clustering for each restaurant type. Exploratory spatial data analysis included global Moran’s I (adjacency/distance-based weights) to assess overall spatial autocorrelation and local Moran’s I (LISA) using a Queen contiguity weight matrix to detect local clusters and outliers, computed in GeoDa. Modeling drivers: An OLS regression first tested relationships between restaurant counts (by type) and the 10 indicators, checking multicollinearity (VIF < 5, tolerance < 1) and selecting variables passing model tests (Supplementary Tables S1–S2). Geographically weighted regression (GWR) then modeled spatially varying relationships, using Gaussian distance-decay weights and bandwidth chosen by AIC for optimal fit. Model performance (R² and adjusted R²) was compared between OLS and GWR, with GWR outperforming. Spatial distributions of coefficients were mapped in ArcGIS to interpret regional heterogeneity (Supplementary Tables S2–S3).

• Composition: Of 4.49 million restaurants, CRs account for 67%, FFRs 21%, DDRs 10%, and WRs 2%.
• National spatial gradient: Restaurant numbers decrease from southeast to northwest, mirroring population distribution. High concentrations occur in the Pearl River Delta, Yangtze River Delta, and the Chengdu–Chongqing region. Cities with >100,000 restaurants: Chongqing, Chengdu, Guangzhou (7.29% of total).
• Kernel density: High-density clusters for all types align with major coastal and inland economic hubs; DDR hotspots also reflect dietary habits, climate, and leisure/tourism (e.g., Chengdu–Chongqing, Pearl River Delta, Shanghai area, Beijing).
• Spatial autocorrelation (Global Moran’s I; p < 0.01): CRs 0.188, WRs 0.135, FFRs 0.213, DDRs 0.135, Total 0.251; indicating significant positive clustering.
• LISA: High–High clusters dominate eastern coastal cities; Low–Low clusters in western regions; Low–High bands around core clusters. Specific High–Low outliers identified (vary by type) include cities such as Chengdu, Chongqing, Shenyang, Harbin, Lanzhou, Kunming, Nanning, Changsha, Wuhan, Xi’an, and Karamay.
• GWR-driven heterogeneity by type: • CRs: Positive associations in most cities with urban population (98.81%), GDP per unit area (85.12%), total tourism revenue (80.95%), and total retail sales of consumer goods (96.43%). Negative coefficients concentrate in Xinjiang/Tibet for population and in northern/southeastern areas for tourism revenue. • WRs: Strong positive links with urban population, GDP per capita, GDP per unit area, tertiary industry share, total retail sales, and per capita disposable income; coefficients exhibit clear east–west and northwest–southeast attenuation patterns depending on factor. • FFRs: Positive across all cities with urban population, urbanization rate, GDP per unit area, total retail sales, and per capita disposable income; strongest ties in highly urbanized and rapidly developing regions (e.g., Pearl River Delta) with notable regional variations. • DDRs: Strong correlations with urban population (south > north, except Xinjiang/Tibet), GDP per unit area (positive in 97.02% of cities; strongest in Tibet, Sichuan, Yunnan), and total tourism revenue (strongest in resource-rich western provinces).
• Regional driver patterns: Central/eastern China restaurant numbers are mainly regulated by urban population, GDP per capita, urbanization rate, and tertiary industry share. In west/northwest, GDP per unit area, tourism revenue, per capita disposable income, and total retail sales dominate. In the northeast, total retail sales, GDP per unit area, urban population, and tertiary industry share are primary.

Findings confirm the hypotheses that economic development and urban population underpin restaurant numbers and that additional socio-cultural and openness factors shape restaurant types. The strong positive global and local spatial autocorrelation indicates non-random clustering aligned with major economic corridors and population hubs. Type-specific patterns reflect functional niches: WRs align with higher tertiary industry development and income, FFRs with urbanization and convenience-driven demand, DDRs with leisure/tourism, climate, and purchasing power, and CRs with broad cultural preference plus tourism and consumption capacity. Regional heterogeneity captured by GWR clarifies how driver strengths shift: coastal and eastern urban agglomerations are primarily population-, income-, and service-structure driven, whereas western and southwestern cities rely more on tourism, retail demand, and density of economic activity per land area. These insights address the gap in statistically tested, nationwide assessments and provide actionable guidance for differentiated restaurant planning and urban service optimization.

This study integrates nationwide POI big data with spatial statistics and GWR to reveal multi-scale clustering and spatially varying drivers of China’s catering industry by type. It documents a pronounced southeast–northwest density gradient and identifies key regional determinants: population, economic intensity (GDP per land), service structure, consumption capacity, and tourism. The results support evidence-based spatial planning for restaurant categories at multiple administrative scales and highlight the role of urbanization and consumption upgrading in shaping food service geographies. Future research should incorporate natural and environmental factors (e.g., topography, climate), extend temporal analyses to assess dynamics, and further explore the coupling between food culture, economy, and politics using richer multi-source datasets.

• Geographic scope excludes Taiwan, Hong Kong, and Macau due to data unavailability.
• While March 2020 POI data are stated as not impacted by COVID-19 for spatial patterns, pandemic dynamics may still influence business statuses.
• Socio-economic indicators are limited to official yearbook variables; unobserved factors (e.g., topography, climate, cultural heritage intensity, policy interventions, transport accessibility) may also shape distributions.
• Potential inconsistencies across data sources and POI categorization could introduce classification bias.
• Cross-sectional design limits causal inference; temporal changes and seasonality are not captured.