Spatiotemporal Evolution and Driving Forces of Housing Price Differentiation in Qingdao, China: Insights from LISA Path and GTWR Models

Abstract

As China’s urbanization deepens, the spatial structure of residential areas and land use patterns has undergone profound transformations, with the differentiation of housing prices emerging as a key indicator of urban spatial dynamics and socioeconomic stratification. This study examines the spatial and temporal evolution of residential housing prices in Qingdao’s main urban area over a 20-year period, using data from three representative years (2003, 2013, and 2023) to capture key stages of change. It employs Local Indicators of Spatial Association (LISA) spatial and temporal path and leap analyses, as well as Geographically and Temporally Weighted Regression (GTWR) modeling. The results show that Qingdao’s housing price patterns exhibit distinct spatiotemporal heterogeneity, characterized by multi-level transitions, leapfrog dynamics and strong spatial dependence. The urban center and coastal zones demonstrate positive synergistic growth, while some inland and peripheral areas show negative spatial coupling. Evident is the spatial restructuring from a monocentric to a polycentric pattern, driven by shifts in industrial layout, policy incentives, and transportation infrastructure. Key driving factors, such as community attributes, locational conditions, and amenity support, show differentiated impacts across regions and over time. Business agglomeration and educational resources are primary positive drivers in central districts, whereas natural environments and commercial density play a more complex role in peripheral areas. These findings provide empirical evidence to inform our understanding of housing market dynamics and offer insights into urban planning and the design of equitable policies in transitional urban systems.

1. Introduction

At present, with the process of urbanization and rapid economic development, China is experiencing institutional transformation and spatial reconfiguration. The built-up area of cities is expanding exponentially, internal space is being fiercely reorganized, and living space is changing rapidly. The types and grades of urban housing are closely related to the economic and social characteristics of residents [1,2,3], and households across diverse socioeconomic strata make housing purchases in alignment with their financial capacity and practical requirements [4,5,6], and the living space is gradually divided and reconstructed. The different social characteristics of residents and the spatial characteristics of housing differentiation caused the imbalance of supply and demand [7,8], and thus the spatial differentiation of housing prices. The spatial differentiation of housing prices is not only a structural characterization and market reflection of the differentiation of urban living space, but also an important driving mechanism for the development of living space.

Against the backdrop of the continuing boom in the commercial housing market, the pattern of residential spatial differentiation within the city continues to intensify, and the spatial gradient difference in housing prices is not only a spatial projection of the land capitalization process, but also an explicit expression of the imbalance in the spatial allocation of social resources, which is further reinforced by the feedback of price signals to reinforce the pattern of residential spatial differentiation. Therefore, the study of residential spatial differentiation has become an important entry point for analyzing the evolution of urban spatial structure. Multidisciplinary studies such as urban geography’s analysis of residential spatial patterns [9,10], urban sociology’s exploration of residential segregation [11,12], and regional economics’ observation of factor flows [13,14] indicate that examining the spatial differentiation in residential property prices has emerged as a pivotal approach for analyzing the spatial configuration of urban residential areas.

At the theoretical level, spatial political economy, environmental justice theory, and institutional path dependence theory provide multidimensional perspectives for analyzing housing price differentiation. Spatial political economy emphasizes the hierarchical accumulation of the spatial value of capital through the production and reproduction of the built environment [15]; for example, local governments in China rely on the institutional incentives of land finance, and through differentiated land supply strategy and infrastructure investment, direct capital to the politically tilted area agglomeration to form the “policy-capital” complicity in the price core [16,17]. The theory of environmental justice shows how the spatial imbalance of environmental risk distribution can be capitalized into house price differentials, where high-income communities convert climate risks into premium assets through green infrastructure investments, while disadvantaged communities suffer house price discounts due to environmental degradation and lagging facilities, forming a “double deprivation” effect [18]. Institutional path dependence theory further suggests that the historically formed institutional framework shapes long-term economic behavior through the “lock-in effect” [19]. The dual structure of public ownership of land and commercialization of housing has created a “duopoly” market, where local governments regulate price fluctuations through spatial and temporal control of land supply, and developers exploit the location monopoly to make excessive profits, leading to the asymmetric effect of urban renewal projects.

Subsequent studies have advanced the knowledge of spatial differentiation of house prices from the perspective of spatial scale, influencing factors, and methodology. On the spatial scale, scholars have established a multi-level research system ranging from macroscopic comparison of urban agglomerations [20,21,22], mesoscopic linkage of metropolitan areas [23,24], to microanalysis of key cities [25,26,27,28], where capital flows reshape the core of house prices in core cities through the business network [29], and the “siphon-spillover” effect promotes the diffusion of price gradients in regional economic zones [30], while the community scale focuses on the capitalization effect of micro-attributes [31]; the research on the influencing factors is mainly conducted through the two-way interaction model between the macro-mechanisms and the micro-factors [32,33,34]. The structural macro factors such as the city’s economic energy level, administrative control strength and population agglomeration scale [27,35], and the micro variables such as the housing attributes [36], location conditions [37,38,39,40] and service support [32,34], make it possible to explore the influencing effects of the different factors; at the methodological level, the early studies relied on the characteristic price model (HPM) to resolve the marginal contribution of housing attributes [41], but the neglect of spatial autocorrelation leads to biased parameter estimation [42]. The spatial lag model (SLM) and the spatial error model (SEM) of spatial econometrics [43] capture proximity effects by introducing a spatial weight matrix [44], but still imply spatial smoothness assumptions and are difficult to account for the heterogeneity of driving mechanisms. Subsequently, Geographically Weighted Regression (GWR) achieves local estimation by modeling the spatial variation of parameters [45], but reduces the temporal dimension to one independent variable, severing the spatiotemporal interaction effect. Spatiotemporal geographically weighted regression (GTWR) overcomes the limitations of previous studies by constructing spatiotemporal cube weight functions to capture spatial dependence and temporal inertia simultaneously [46]. In addition, LISA spatiotemporal jumps accurately represent the dynamic integrality and local heterogeneity of spatial pattern evolution through indicators such as path length, curvature, and jump type, which extends the depth of the study of spatiotemporal differentiation of house prices.

At a practical level, the spatial differentiation of housing prices has become an important indicator of the rationality of urban spatial structure and the efficiency of resource allocation, which directly affects the realization of urban planning, housing policy, and social equity. Although the purchase restriction policy can curb speculative demand in the core city in the short term, it often triggers the spillover of demand to areas without purchase restriction, which exacerbates the reconstruction of the housing price gradient at the metropolitan scale [47]. Some cities in China have tried to mitigate the premium effect of school district housing through the “equal right to rent and purchase” policy, but its effect is limited by the spatial monopoly of high-quality educational resources, and even the phenomenon of policy dissolution has occurred in local areas [48,49]; the spatial layout of guaranteed housing ignores the accessibility of employment and service support, which may lead to the consolidation of the phenomenon of “residential segregation”—low-income groups are forced to live on the urban periphery, and the high cost of commuting further compresses their disposable income, creating a “spatial poverty trap” [50,51]. Overseas studies have shown that inclusive planning requires the comprehensive use of multi-dimensional policy tools, such as Singapore through the spatial mixing design of the HDB system, which effectively inhibits residential differentiation [52]; European cities rely on the concept of “15-min living circle” through the equalization of public services to weaken the excessive dependence of housing prices on location conditions [53]. Although many studies have examined the differentiation of housing prices in large metropolitan areas, relatively little research has focused on medium-sized cities such as Qingdao, which face unique urbanization challenges. This limits our understanding of how housing prices evolve in cities experiencing rapid growth and transformation that are neither megacities nor rural areas. Furthermore, most existing studies have concentrated on spatial differentiation without adequately addressing the temporal evolution of housing prices. The dynamics of housing price change over time, influenced by factors such as policy shifts, infrastructure development and economic changes, remain under-explored. Understanding how these factors interact over time is crucial for effective urban planning and policy formulation.

Qingdao, positioned as a pivotal coastal metropolis in China, occupies a strategic vantage point within Shandong Province, embracing global engagement and fostering an outward-looking developmental orientation. This paper takes the main urban area of Qingdao as the object of empirical research, reveals the morphological characteristics of the spatial and temporal paths of residential prices and the leap law based on the LISA spatial and temporal leap, and employs the Geographically and Temporally Weighted Regression (GTWR) model to dissect the spatiotemporal driving influences exerted by community attributes, location and transportation factors, as well as ancillary facilities, to construct a full-chain analysis framework of “pattern evolution-mechanism identification-policy response”. The aim is to construct a full-chain analysis framework of “pattern evolution-mechanism identification-policy response”, which provides a scientific basis for optimizing urban spatial resource allocation and suppressing residential spatial differentiation.

2. Research Area and Data Sources

This study has constructed an integrated analytical framework to systematically analyze the spatiotemporal evolution of residential prices in Qingdao and explore their underlying mechanisms (Figure 1). This comprises three core components: data preparation, dynamic spatiotemporal pattern analysis, and mechanism interpretation. Combining quantitative modeling (LISA and GTWR) with theoretical mechanism deduction reveals both the surface characteristics and the underlying logic of price differentiation, offering a comprehensive perspective of “pattern recognition-driving factor identification-mechanism interpretation”.

Qingdao, as the core city of Shandong Peninsula, has a strong comprehensive economic strength, with a total economic volume of 1,492,075,000,000 RMB in 2022, ranking 13th among cities in China and 24th in the country in terms of per capita economic development level, with a high gold content in urban development, attracting a large influx of population, and a nationally high urbanization rate of 78.3%, and is in the strategic position of Shandong Province in the opening up to the world for development. In the year 2023, the population of Qingdao is projected to reach 10,371,500, with 8,121,000 residents inhabiting the urban regions. The urban built-up area is anticipated to span 11,282 square kilometers. Due to the rapid urbanization process and the continuous expansion of living space, the demand for housing has increased dramatically, and the price of housing has shown obvious differences within the region. This study selects the principal urban region of Qingdao as its research scope, encompassing seven administrative districts: Shinan, Shibei, Laoshan, Licang, Chengyang, Jimo, and Huangdao Districts (Figure 2).

In this study, residential neighborhoods are designated as the fundamental unit of analysis. A comprehensive dataset of housing prices has been meticulously compiled for sample neighborhoods located within the primary urban region of Qingdao, spanning the years 2003, 2013, and 2023. The pricing information is aggregated by calculating the mean transaction price per unit area for both newly constructed and pre-owned residences traded within the same calendar month of each respective year.

To accurately depict the spatial pattern of Qingdao’s main urban area, this paper applies ArcGIS to vectorize key spatial components such as transportation network, mountain topography, and water system, and overlays these converted vector data on the latest Qingdao geographic base map to ensure the accurate positioning of the sample neighborhoods. Among them, the basic geographic data such as administrative boundaries and water and green areas were obtained from the Resource and Environmental Science and Data Platform of the Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 18 November 2023), and the urban road network data were selected from the vector layer published by OSM (open street map) (https://www.openstreetmap.org, accessed on 18 November 2023); geographic coordinates (latitude and longitude) of residential areas, average yearly (2003, 2013, and 2023) listing price per unit area, and comprehensive attribute data including architectural features were obtained by relying on web crawler tools on the website of Anjuke (https://qd.anjuke.com/, accessed on 12 November 2023) for collection. The neighborhood boundaries were derived from the area of interest (AOI) boundary data of Baidu map; 5314 residential neighborhoods were selected as the study samples after data processing, and the spatial attribute database of the sample neighborhoods was established.

3. Variable Explanation and Research Approach

3.1. Description of Variables

The spatial variation in urban housing prices arises from a confluence of multiple factors, encompassing the broader macroeconomic context [54], monetary policy frameworks, prevailing interest rate conditions, and fluctuations within the financial market landscape, as well as micro-socioeconomic factors such as population growth, urbanization [55], employment opportunities and income levels of residents [56,57], and regional characteristics such as educational resources, health facilities, transport accessibility, and environmental quality [33,37,40]. The objective of this study is to examine, from a regional perspective, how the intrinsic attributes of a city influence housing prices. Taking into account the availability of data and the practicality of conducting quantitative analysis, we preliminarily pinpointed three principal categories of factors that significantly contribute to the fluctuations in urban housing prices: (1) Community attributes (C): This category encompasses six variable metrics, namely, the age distribution of the residential community, the quality of services and management provided, the availability of parking spaces, housing density levels, the extent of green landscape coverage, and the scale of community living amenities; (2) Location and transportation characteristics (L): Comprising five variable indicators, this category includes the centrality of the location, the surrounding environmental conditions, the proximity to educational institutions, the ease of internal transportation access within the area, and the convenience of external transportation links; (3) Surrounding ancillary facilities (S): This category is made up of four variable indicators, specifically, the breadth of housing-related amenities, the accessibility of medical services, the concentration of commercial resources, and the variety of recreational facilities in the vicinity (

Table 1. Characterization of characteristic variables affecting the spatial differentiation of residential prices in a community.

Classification	Explanatory Variables	Variable Description	Remarks
Community attributes	Degree of newness	Age	2023, the year in which reduced the building of the communities
	Service management	Property fee	Monthly property management fee per square meter of floor area
	Parking space package	Number of parking spaces/total number of households	Average number of parking spaces per household
	Degree of crowding	Volume ratio	Floor area/occupied area
	Green environment	Greening rate	Green area/footprint
	Residential scale	Number of residences	Total residential units
Location and transportation	Central location	The shortest route to large business districts 1	City center business district
	Environmental location	Shortest path to large-scale mountain and sea resources in the city 2	Large-scale mountain and sea resources refer to Laoshan Mountain, Signal Mountain, Badaguan Scenic Area, etc.
	School district attributes	Shortest path to public primary and secondary schools	Shortest path to primary and secondary schools near residences
	External transportation	Shortest path to the nearest expressway entrance/exit	Expressway including Binhai Avenue
Neighborhood support	Commercial support	Number of large commercial service facilities within 1 km radius (10 min walk)	Shopping malls including urban complexes, hypermarkets, community supermarkets, and convenience stores; representing daily life support services, etc.
	Medical support	Number of general hospitals within 1 km radius	Including general hospitals, private hospitals, specialized hospitals, etc.
	Business support	Number of bank outlets and commercial buildings within 1 km radius	Including postal, telecom, mobile, and Unicom service outlets; banks excluding ATM outlets
	Leisure Support	Number of parks, squares, venues, and attractions within 1 km radius	Venues include cultural relics and monuments, residences of celebrities, cultural venues, and sports venues

Note: ¹ 6 business circles are Hong Kong Middle Road Business Circle, Taidong Business Circle, Zhongshan Road Business Circle, LiCun Business Circle, JimeiYa Business Circle, Laoshan Business Circle, Xinduxin Business Circle. ² Large scenic resources refer to Laoshan Mountain, Signal Mountain, Badaguan Scenic Area, and so on.

).

3.2. LISA Spatiotemporal Path

The LISA spatiotemporal path method creates a new paradigm for spatiotemporal co-evolutionary analysis by integrating local spatial autocorrelation and a Markov chain. The spatiotemporal interaction framework constructed by Rey breaks the static limitation of traditional spatial autocorrelation analysis and extends spatial dependence to time series [58]; the spatiotemporal leap matrix developed by Ye and Rey systematically reveals the evolutionary path-dependent features of spatial units by quantifying the state of the region itself and the structure of its neighborhood with the synergistic leap probabilities [59]. The LISA (Local Indicators of Spatial Autocorrelation) time path serves as a dynamic visualization of the positional evolution of a spatial unit within Moran’s I scatterplot framework [60]. Through the graphical depiction of how an attribute value of a spatial unit shifts in tandem with its spatial relocation, this approach unveils the magnitude and orientation of spatiotemporal interactive transformations occurring among neighboring areas [58]. The length index of the LISA time path reflects the dynamic characteristics of the local spatial structure of the city, the curvature index reflects the fluctuating characteristics of the local spatial structure of the city, and the jump direction reflects the integrative characteristics of the evolution of the local spatial structure of the city. The LISA path is computed using the following formula, with detailed derivations provided in Appendix A.

$$d_i={\displaystyle \frac{N{\displaystyle \sum _{t=1}^{T-1}d(L_{i,t}},L_{i,t+1})}{{\displaystyle \sum _{i=1}^{N-1}{\displaystyle {\sum }_{t=i}^{T-1}d}(L_{i,t},L_{i,t+1})}}}$$

(1)

In the formula, $d_i$ is the path length and curvature of the region $i$, respectively; $N$ is the number of study units, with $N$ = 97 in the text; $T$ is the length of the study period; $L_{i,t}$ is the LISA coordinates of region $i$ at time $t$; and $d(L_{i,t},L_{i,t+1})$ is the distance traveled by region $i$ from time $t$ to time $t+1$.

3.3. LISA Spatiotemporal Leaps

LISA spatiotemporal paths describe the geometric characteristics of the temporal trajectories of each spatial unit moving on Moran’s I scatterplot, while spatiotemporal leaps further explain the temporal variations of spatial relationships between neighbors in different localizations. Rey, based on LISA, embedded the Moran’s I scatterplot, in which the attributes of each spatial unit in a given time interval, such as movement distance, direction, coalescence, etc., are embedded in a traditional Markov chain, into a traditional Markov chain, proposing localized Markov shifts and spatiotemporal jumps that are used to reveal the spatial dependence of geographic phenomena [59,61]. The spatiotemporal jump cases are categorized into four types, Type 0, Type 1, Type 2 and Type 3, where Type 0 signifies that the region in question experiences no inter-morphological transformation over time, with all instances of this type positioned along the principal diagonal of the transition matrix. Type 1 denotes a scenario where the region itself undergoes a jump while its surrounding neighborhood remains static, including HH_t→LH_t+1, HL_t→LL_t+1, LH_t+1→HH_t+1, LL_t→HL_t+1; Type 2 indicates that the region itself remains stable, whereas its neighboring area experiences a jump, including HH_t→HL_t+1, HL_t→HH_t+1, LH_t→LL_t+1, and LL_t→LH_t+1; Type 3 indicates that both the region itself and its neighbors undergo leaps, and if the region itself and its neighbors have the same direction of leaps, it is Type 3A, including HH_t→LL_t+1 and LL_t→HH_t+1; if the region itself and its neighbors have opposite directions of leaps, it is Type 3B, including HL_t→LH_t+1 and LH_t→HL_t+1. Rey defines the spatiotemporal flow (SF) of the system with the definition of coalescence (SC), expressed as the path dependence and locking traits in the spatial pattern of the object under investigation, quantified by the proportion of a specific type of leaps occurring during the study period relative to the overall number of potential leaps within the system (m).

$$\begin{array}{l}SF={\displaystyle \frac{F_1+F_2}{m}}\\ SC={\displaystyle \frac{F_0+F_{3A}}{m}}\\ t=1-{\displaystyle \frac{{\displaystyle \sum _i{P}_{i,i}}}{K}}\end{array}$$

(2)

In the formula, $F_0$, $F_1$, $F_2$ and $F_{3A}$ are the number of leaps for Type 0, Type 1, Type 2, and Type 3A, respectively. $P_{i,i}$ is the diagonal element of the Markov transfer matrix. $t$ = 0 means that no region has inter-morphological transfer, and the diagonal element sums to 1; $t$ = 1 means that all regions have morphological jumps.

3.4. Spatiotemporal Geographically Weighted Regression Models

The spatiotemporal geographically weighted regression (GTWR) model, serving as an advanced iteration of the geographically weighted regression (GWR) model, represents a non-stationary regression framework that incorporates both spatial and temporal dimensions. Its fundamental innovation lies in the integration of the temporal factor into the traditional GWR model, thereby enabling the analysis of spatiotemporal variations in regression relationships [46]. The model is designed to use time and space as weight function inputs, and to accurately capture the spatiotemporal heterogeneity in the data by dynamically adjusting the weights to provide more accurate regression results, with the basic formula as follows:

$$\begin{gathered} {{\displaystyle Y}}_i={{\displaystyle \beta }}_0\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)+{\displaystyle \sum _{j=1}^p{{\displaystyle \beta }}_j\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right){{\displaystyle x}}_{ij}}+{{\displaystyle \epsilon }}_i \\ \left(i=1,2,\dots ,n\right) \end{gathered}$$

(3)

In the formula, $\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)$ is the spatiotemporal coordinates of the $i$th sample point; ${{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i$ is the longitude, latitude and time of the $i$th sample point; ${{\displaystyle \beta }}_0\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)$ is the regression constant of the $i$th sample point, i.e., the constant term in the model; ${{\displaystyle x}}_{ij}$ is the value of the $j$th independent variable at the $i$th sample point; ${{\displaystyle \epsilon }}_i$ is the residual; and ${{\displaystyle \beta }}_j\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)$ is the $j$th regression parameter of the $i$th sample point, which is estimated as follows:

$$\hat{\beta }\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)={\left[X^{t}W\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)X\right]}^{-1}{X}^{T}W\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)Y$$

(4)

In the formula, $\hat{\beta }\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)$ is the estimated value of ${{\displaystyle \beta }}_j\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)$; $X$ is the matrix composed of independent variables; $X^t$ is the transpose of the matrix; $Y$ is the matrix composed in the sample; $W\left({{\displaystyle u}}_{i,}{{\displaystyle v}}_{i,}{{\displaystyle t}}_i\right)$ is the spatiotemporal weight matrix. $W$ Select the Gaussian distance function and use the bi-square spatial weight function to obtain the spatiotemporal weight matrix, and the spatiotemporal distance between sample $i$ and sample $j$ is

$$d_{ij}=\sqrt{\delta \left[{\left(u_i-u_j\right)}^2+{\left(v_i-v_j\right)}^2+\mu {\left(t_i-t_j\right)}^2\right]}$$

(5)

In this case, the choice of bandwidth affects the establishment of spatiotemporal weights, and in this paper, we use the AICc law with adaptive bandwidth.

4. Results

4.1. Spatial Pattern Evolution Traits of Residential Property Prices in Qingdao City

4.1.1. Residential Price Differentiation in Qingdao City

In order to gain a deeper understanding of the complex spatiotemporal evolution of residential property prices in Qingdao, it is first necessary to clarify the spatial patterns that emerged between 2003 and 2023. Using sample data from 5314 residential communities, the study found that Qingdao’s residential property prices have shown a sustained upward trend overall, with the average price per square meter rising from 4786 RMB/m² in 2003 to 21,752 RMB/m² in 2023, a cumulative increase of 454.5% (Figure 3). Prices experienced a rapid rise from 2003 to 2013, with an average annual increase of 19.4%. From 2013 to 2023, the growth rate stabilized at an average annual increase of 5.46%.

There are significant regional disparities in residential prices, with central urban areas (Shinan, Shibei, Laoshan and Licang) consistently commanding higher prices than peripheral regions (Chengyang, Jimo and West Coast New District). This forms a spatial gradient characterized by both “central premium” and “peripheral diffusion”. As the city’s spatial structure has been restructured and resources have been optimally allocated, the spatial form of residential prices has evolved from a “single core” to a “multi-core + fan-shaped” pattern and finally to a “mosaic-like composite structure” (Figure 4). In 2003, Shinan District was the traditional price core thanks to its exceptional natural resources; however, in 2013, the relocation of the municipal government to the east caused Laoshan District to experience a rapid increase in prices, forming a “dual-core” structure alongside the southern district and spreading outwards along the eastern shore of Jiaozhou Bay. By 2023, emerging areas such as Licang District and the West Coast New Area had seen accelerated increases in residential prices due to improved transportation and industrial development. This gave rise to multiple growth nodes and formed a highly nested, scattered “mosaic-like” high-price zone [8].

4.1.2. Spatial and Temporal Dynamics of Housing Prices in Qingdao

The Moran index of housing prices in Qingdao is calculated by GeoDa 1.22 software from 2003 to 2023 to characterize its spatial agglomeration. During the study period, the Moran index value exceeded 0 and successfully passed the significance test, revealing a notable positive spatial autocorrelation in residential property prices. This finding suggests the presence of substantial spatial dependence among residential prices in Qingdao, with the spatial distribution displaying pronounced clustering patterns.

(1)

Spatial and temporal trends in population density

The relative length, curvature, and direction of movement of the LISA time path reflect the multi-level differentiation characteristics of Qingdao’s housing prices in the spatial and temporal dimensions, reflecting the intrinsic correlation between the spatial distribution structure of prices and the regional functional division of labor. Moreover, the dynamic traits of the local price structure are examined (Figure 5). Utilizing the ArcGIS natural breakpoint method, the relative lengths of the LISA time paths are categorized into two distinct groups. Throughout the research period, Qingdao city exhibits 41 roads with relative lengths surpassing the mean value, constituting 44.09% of the total. In terms of spatial distribution, these roads are predominantly located along the coastline of Jiaozhou Bay and are sparsely scattered along the boundary with Jiaozhou city, indicating a relatively vibrant local spatial configuration. There are 52 streets with a relative length below the average, accounting for 55.91% of the total number; they are mainly distributed in the inland part of the city, with a more stable local spatial structure (Figure 5a). In general, Qingdao has a highly dynamic and unstable local spatial structure in areas with a better level of economic development, and a stable local spatial structure in areas with a relatively backwards level of economic development.

By applying the natural breakpoint method in ArcGIS to categorize the curvature values of the LISA time path, it was found that 71 street units exhibit a curvature exceeding 1, representing 76.34% of the total township units. This indicates that housing prices in these units are, to a certain degree, influenced by adjacent spatial areas, demonstrating significant fluctuations in the direction of spatial dependence. There are 73 township units with lower curvature in the LISA time path, accounting for 78.49% of the total number of township units, i.e., most of the residential prices of streets have relatively stable spatial dependence and direction of change in house prices. In the urban area, there are 14 streets with higher than average curvature of the LISA time path, and 22 townships, including Zhonghan Street, Sifang Street, and Helen Road Street, have a curvature of the LISA time path of less than 1. This is mainly due to the fact that the prices of these townships and neighboring areas are subject to the mutual constraints of spatial dependence of fluctuations at the same time, resulting in a smaller degree of curvature (Figure 5b).

Of all the townships examined, 74 (representing 79.57% of the total) exhibit a pattern of synergistic growth aligned with the directional movement of the LISA (Local Indicators of Spatial Association) time path concerning urban housing prices in Qingdao. This suggests that urban housing prices demonstrate a notable level of spatial integration in their developmental trajectory. Specifically, there are 45 townships with positive synergistic development, accounting for 60.81% of the total number of streets with synergistic development, which are mainly distributed in the peripheral areas of the city and show positive synergistic growth in a leapfrog manner with neighboring regions. Among the observed cases of street-level synergistic development, 29 instances (39.19% of the total) exhibit negative synergy. These cases are predominantly located in the central urban region and inland areas of the West Coast New Area. Adjacent zones display a pattern of negative synergistic growth characterized by leapfrog distribution. And Taidong Street, Licun Street, Fushan Road Street and Lingzhushan Street show the opposite direction of leapfrogging with neighboring regions (Figure 5c). The evolution process of urban residential price space in Qingdao reflects, to a certain extent, the synergy effect differences between peripheral and central regions, as well as the integration and local heterogeneous characteristics of the spatial structure.

(2)

Spatial Depiction of the LISA Space-Time Transition

During the period 2003–2023, the most common type of jump is Type 3, which indicates that both the region itself and its neighborhood undergo a jump, with a probability of 51.6%, suggesting that urban housing prices in Qingdao have a high degree of volatility. This is followed by Type 0, which indicates that neither the region itself nor its neighborhood undergoes a morphological jump over time, with a probability of 46.2%, and Type 1 and Type 2, which indicate that only one of the regions itself or their neighborhoods undergoes a jump, with probabilities of 0.5% and 1.6% respectively. Overall, the probability that house prices in Qingdao city have not jumped is 47.8%, and the degree of spatiotemporal cohesion is 56.5%, indicating that the dynamics of house prices in the spatial pattern of each region is more significant and shows greater flexibility. This observation indicates that residential property prices in Qingdao exhibit significant susceptibility to fluctuations in adjacent areas, with inter-regional price variations demonstrating notable correlations and reciprocal influences (

Table 2. Time-lapse matrix of housing prices in Qingdao City.

Period	t/t + 10	HH	LH	LL	HL	Type	Quantity	Ratio (%)	SF	SC
2003–2023	HH	0.425	0.032	0	0.005	Type0	86	46.2%	0.022	0.565
	LH	0	0	0.005	0	Type1	1	0.5%
	HL	0.005	0	0.005	0.005	Type2	3	1.6%
	LL	0.102	0.054	0.312	0.048	Type3	96	51.6%

).

(3)

Characterization of the spatial and temporal dynamics of house prices

Based on the Moran index calculated by GeoDa software from 2003–2023, Qingdao’s housing prices show a significant positive spatial correlation (I > 0 and passes the test), indicating strong dependence and agglomeration characteristics in their spatial distribution. Further analysis of the LISA time path elucidates the spatiotemporal disparities in the evolution of price trends. The relative length indicator shows that 44.09% of the townships are more dynamic than the mean, mainly concentrated in economically active areas such as the coast of Jiaozhou Bay, while 55.91% of the townships maintain a more stable structure. The curvature analysis shows that there are significant directional fluctuations in the paths of 76.34% of the townships, but 78.49% of the units maintain the stability of spatial dependence, forming a dynamic and stable coupling structure. The spatiotemporal jump matrix shows that Type3 jumps (synchronous jumps between itself and its neighbors) account for 51.6%, Type0 steady state accounts for 46.2%, and the overall cohesion reaches 56.5%, highlighting the path-dependent characteristics of price evolution. Among them, the peripheral regions show positive synergistic growth (60.81% of synergistic municipalities), while 39.19% of negative synergistic units in the central urban area reveal the risk of localized decline in the process of spatial integration.

4.2. Examination of Influential Factors Employing GTWR Modeling

4.2.1. GTWR Model Construction

Based on the theoretical framework of the spatiotemporal geographically weighted regression model, the specific coordinates of the geographical center of the $i$th residential district are set as based on the selected indicator variables and their associated parameters, and the subsequent spatiotemporal geographically weighted regression (GTWR) model is formulated.

$$\begin{array}{cc}\hfill yi=& \beta 0(u_i,v_i,t_i)+{\displaystyle \sum _{j=1}^k\beta 1(u_i,v_i,t_i)x_{ij}(age)+{\displaystyle \sum _{j=1}^k\beta 2(u_i,v_i,t_i)x_{ij}(ser)}}\hfill \\ & \begin{array}{c}+{\displaystyle \sum _{j=1}^k\beta 3(u_i,v_i,t_i)x_{ij}(par)+{\displaystyle \sum _{j=1}^k\beta 4(u_i,v_i,t_i)x_{ij}(plo)+{\displaystyle \sum _{j=1}^k\beta 5(u_i,v_i,t_i)x_{ij}(gre)}}}\hfill \\ \begin{array}{c}+{\displaystyle \sum _{j=1}^k\beta 6(u_i,v_i,t_i)x_{ij}(siz)+}{\displaystyle \sum _{j=1}^k\beta 7(u_i,v_i,t_i)x_{ij}(cen)+}{\displaystyle \sum _{j=1}^k\beta 8(u_i,v_i,t_i)x_{ij}(lan)}\hfill \\ \begin{array}{c}+{\displaystyle \sum _{j=1}^k\beta 9(u_i,v_i,t_i)x_{ij}(edu)+}{\displaystyle \sum _{j=1}^k\beta 10(u_i,v_i,t_i)x_{ij}(com)+{\displaystyle \sum _{j=1}^k\beta 11(u_i,v_i,t_i)x_{ij}(hos)}}\\ +{\displaystyle \sum _{j=1}^k\beta 12(u_i,v_i,t_i)x_{ij}(bus)+}{\displaystyle \sum _{j=1}^k\beta 13(u_i,v_i,t_i)x_{ij}(lei)+\epsilon i}\hfill \end{array}\hfill \end{array}\hfill \end{array}\hfill \end{array}$$

(6)

In the ArcGIS software 10.8.1, the spatiotemporal geographically weighted regression (GTWR) tool was utilized for model analysis, with Qingdao housing prices as the dependent variable. SPSS software 26.0 and ArcGIS software were utilized for the covariance diagnosis, significance test, etc. for the 13 indicators in Table 1. Finally, the degree of newness, green environment, residential scale, central location, environmental location, school district attributes and commercial support, business support—a total of 8 indicators for GTWR analysis—were selected.

4.2.2. GTWR Model Regression Results

The recency or novelty level of residential neighborhoods exhibits a statistically significant positive association with housing prices (Figure 6a), especially in the central city, the West Coast New District, and Jimo District, where housing prices are higher due to the presence of supporting facilities, convenient transportation conditions, and modern living conditions. However, in the old city, Laoshan District, and the remote areas of West Coast New District, the effect of novelty on price appreciation is weakened due to the backwardness of infrastructure facilities and the influence of old buildings, and even shows a negative correlation in some areas. The variation observed in the regression coefficient indicates a progressive augmentation in the impact of residential neighborhood novelty on housing prices. Nevertheless, the overall influence remains comparatively consistent over time. The green environment of residential neighborhoods has a significant positive effect on prices (Figure 6b), especially in areas with poor green conditions and high expectations of residents for green improvements; the higher the greening rate, the higher the house price. However, in some areas with rich natural resources themselves, such as Laoshan Scenic Area and Zhushan National Forest Park, the marginal utility of greening the environment gradually decreases, and the effect on housing prices gradually weakens. Overall, greening has a limited driving effect in resource-rich areas, while in areas with poor greening conditions, greening has a much stronger effect on house prices.

The relationship between the scale of dwellings and house prices is negative: the smaller the number of dwellings in a municipality, the higher the house price (Figure 7a). Particularly in areas with good infrastructure, excellent community environment, and favorable location, these areas tend to have higher levels of privacy and exclusivity, contributing to the increase in house prices. The influence of residential neighborhoods’ central positioning on housing prices exhibits an inverse relationship (Figure 7b). As the city expands and a multipolar urban structure emerges, the rate of increase in residential property values in proximity to the city center has decelerated. This phenomenon reflects the inhibiting effect of factors such as scarce land resources and high living costs in the core regions on house price growth. Environmental location has a negative impact on housing prices (Figure 7c), especially in areas close to natural resources, such as Jimo District, where housing prices are more influenced by natural landscapes, while prices in peripheral urban areas are less influenced by environmental location due to the lack of natural resources. The impact of educational resources is also significant, especially in areas rich in educational resources such as Licang District, Chengyang District, and West Coast New District, where house prices are higher the closer the primary and secondary schools are (Figure 7d).

The impact of commercial facilities on prices varies regionally. Overall, commercial facilities have a negative impact on residential prices (Figure 8a), especially in Laoshan District and some peripheral areas of the West Coast New District, where commercial facilities can effectively meet the daily needs of surrounding residents, so the increase in commercial facilities may have a negative effect, reducing the comfort of living and thus inhibiting the growth of residential prices. On the other hand, in core urban areas such as Shinan District and Shibei District, the improvement of business support still has a positive impact on house prices. In addition, the impact of business facilities varies significantly across regions (Figure 8b), with densely populated areas such as Shibei District, Licang District, and Laoshan District having a high density of facilities to improve the quality of life of their residents, In outlying regions, exemplified by Jimo District and the West Coast New District, characterized by diminished commercial activity and a lack of amenities, the impact of commercial facilities on residential property values is relatively modest. Consequently, housing prices in these areas have grown relatively slowly, resulting in lower house price growth in these areas.

It should be noted that temporal variations in the GTWR regression coefficients represent changes in the strength and direction of statistical associations over the selected years rather than providing definitive evidence of temporal causality. While the GTWR model can identify how the influence of different factors evolves over time and space, without longitudinal tracking of causal mechanisms or experimental controls, the results should be interpreted as correlations that may inform, but do not prove, causal relationships.

4.2.3. Drivers Gradually Diversifying

In summary, the influencing factors of residential prices in Qingdao from 2003 to 2023 have undergone a complex development from single-dominant to multi-dominant, reflecting the dynamic process of urban function optimization, social stratification, and market mechanism synergy (

Table 3. Spatial distribution of house price regression coefficients of the GTWR model in 2003, 2013, and 2023.

Driving Factors	Median			Average			p Value
	2003	2013	2003	2003	2013	2023
Degree of newness	0.212	0.921	−0.614	0.190	0.766	−0.580	0.005
Green environment	−0.694	0.479	−0.602	−0.766	0.614	−0.189	0.000
Residential scale	0.722	0.561	−0.194	0.807	0.421	−0.529	0.000
Central location	1.194	−0.052	−0.757	1.192	0.229	−0.513	0.002
School district attributes	0.767	0.409	−0.749	0.623	0.689	−0.657	0.002
Environmental location	0.716	0.383	−0.705	0.752	0.408	−0.503	0.005
Commercial support	−0.306	−0.306	−0.119	−0.306	−0.306	0.301	0.001
Business support	−0.984	−0.603	0.785	−0.938	−0.747	0.791	0.001

, Figure 9). The spatial differentiation of residential prices in 2003 was mainly driven by community characteristics and location conditions, with residential scale, natural environment, and location advantage of the center as the core elements. With its superior geographical location and perfect infrastructure, the Shinan District became the core area of high prices, forming a gradient effect of urban resource concentration and land use efficiency. In 2013, amidst the accelerated pace of urbanization and the swift advancement of Laoshan District, propelled by the municipal government’s eastward relocation, the determinants of residential price disparities evolved from a singular focus to encompass a multifaceted array of composite factors. The degree of newness, educational resources, and location conditions have gradually become the core driving force; high prices in the central city are driven by ecological advantages, educational resources, and business concentration, while peripheral areas such as Lichang District and West Coast New District have gradually formed a sub-price plateau due to infrastructure improvement and industrial upgrading. The transformation of the core of urban functions from a single center to a multi-center pattern reveals the nested relationship between the spatial structure of the city and the spatial differentiation of the housing market; in 2023, housing price differentiation is manifested in the coexistence of the superposition of multi-dimensional factors and the strengthening of regional differences. Community attributes remain the main driver, but the impact of locational conditions diverges in the time dimension. The importance of core location is relatively declining, while peripheral areas have become the main areas of price increase due to improved support facilities and ecological advantages. The positive impact of business and commercial support has gradually strengthened, and the price level of high-price zones complements the high-density allocation of urban resources.

5. Discussion

As the core city of the Bohai Economic Circle, Qingdao’s property market has shown significant price dynamics and spatial pattern reconstruction in recent years. This is not only a spatial and temporal mapping of the optimization and upgrading of the industrial structure and the promotion of the new urbanization strategy, but also a deeper revelation of the multidimensional interaction between the policy regulation mechanism, the law of market supply and demand, and the spatial differentiation of social forces in the formation of housing prices (Figure 10). While the GTWR results reveal significant shifts in the spatial and temporal patterns of housing price determinants, caution is required when interpreting these shifts. While the model’s coefficients describe how the strength of the association between explanatory variables and housing prices changes over time, they cannot establish that these variables are the direct cause of the observed price variations. Nevertheless, when combined with contextual information such as policy changes, infrastructure development and demographic shifts, these patterns can suggest plausible causal pathways. While these interpretations provide valuable insight into potential cause-and-effect relationships, confirming such causality would require further longitudinal or quasi-experimental research.

5.1. Policy Regulation Constraints and Guidelines

As a national coastal city, Qingdao’s housing price growth has been strongly influenced by policy regulation. The government has guided the orderly development of real estate through land market reform and differentiated land supply strategies: on the one hand, the tightening of land supply in the core urban areas has intensified competition for scarce land, pushing up the structural increase in housing prices in areas such as Shinan and Laoshan; on the other hand, peripheral areas such as West Coast New District and Jimo District have attracted the import of industrial population through preferential land prices, forming a “city–industry integration” housing demand growth point and the an integration of housing demand growth. In addition, limitations on purchases and loans as well as other policies have inhibited speculative demand in the short term, but through the city’s ability to increase the rigidity of the support of population inflow, policy regulation on the level of housing prices has shown marginal diminishing returns.

5.2. Structural Tension Between Market Supply and Demand

The implementation of Qingdao’s “multi-center, cluster” spatial strategy has profoundly reconfigured the power mechanism and market pattern of housing prices. From the establishment and colonial period of the triangular group pattern, the national government has sought to the reform and open up before the belt pattern, and finally formed a multi-core group pattern consisting of a bay and two wings. Thereafter, the city government constructed an eastward drive in the Laoshan Jinjialing financial district, in order to support both a business function and residential function that would give birth to high-end residential clusters, and thus promote the growth of a residential current resource monopoly. In the West Coast new area, the pilot free trade zone policy used dividends from film and television culture to bolster IP empowerment, which increased the value of the cities of Lingshan Bay and Tangdao Bay, thus increasing the prices of housing.

The supply and demand relationship in Qingdao’s residential market is characterized by a “core-periphery” split. Core urban areas (Shinan, Shibei, Laoshan, Licang) rely on high-quality educational resources and high-end business agglomeration, the formation of high-income groups of rigid demand, promoting the price of high-end housing to continue to rise; peripheral areas (West Coast New District, Chengyang District, Jimo District) benefit from industrial upgrading and population influx, the demand for housing changes from a single residential function to the “residential-investment” compound attribute change. The demand for housing has changed from a single residential function to a composite attribute of “residential-investment”. The agglomeration of high-tech industries in the Hongdao Economic Zone and the rise of Dongjiakou Lingang Economy have shifted the housing demand of the working population to the outskirts of the city, causing the housing price to show the spread of the circle, and the “marginal urbanization” effect of the expansion of urban space is remarkable.

5.3. The Reconfiguration Effect of Social Space

Residential price differentiation is essentially a spatial mapping of social stratification. Qingdao’s middle- and high-income groups’ pursuit of mountain and seascape resources and high-quality school districts has led to the “resource monopoly” price plate of high-end residential areas such as Fushan Bay and Jinjialing, while migrant workers and young groups in need of housing are limited by their ability to pay, accelerating their agglomeration in Licang and Chengyang, creating a spatial mismatch of “job and housing separation”. This separation is further reinforced by the reform of the household registration system and the introduction of talent policy, which is an high-level talent purchase subsidy policy designed to introduce innovative elements to the Blue Silicon Valley, thus pushing up regional housing prices. Meanwhile, the traditional manufacturing industry has had to relocate the local working class, which has exacerbated the structural downturn in the housing market of the old industrial areas (such as the Sifang, Cangkou).

5.4. Natural Environment

Ecological constraints have also become an important variable: the ecological red line control of Jiaozhou Bay restricts land development in the bay area, causing the supply of residential units to be concentrated in the vertical space, and the price premium of high-rise residential units is significant; the low-density development of ecologically sensitive areas such as Laoshan Mountain and Dazhushan Mountain has supported the rigidity of villa product prices with the scarcity of ecological resources. This dynamic coupling of “function-space-price” reveals the internal logic of Qingdao’s housing market from single-center agglomeration to multi-center group development.

The study’s findings shed light on the spatial and temporal dynamics of housing prices in Qingdao, demonstrating the impact of infrastructure development, policy shifts, and socioeconomic changes on residential price trends. A comparative analysis with other Chinese coastal cities, such as Shenzhen, Xiamen, and Dalian, reveals similar price differentiation patterns, particularly in rapidly developing urban areas where housing demand is driven by infrastructure projects such as transportation networks and commercial hubs. However, differences in local policies, such as land supply management, housing purchase restrictions, and market intervention measures, lead to variations in the magnitude and speed of these price shifts. Cities like Shenzhen and Xiamen, for example, have experienced more aggressive land supply strategies and property market regulations, which have impacted price stability and affordability, particularly in central urban districts. Comparing Qingdao’s experiences with those of other cities sheds light on the common challenges that coastal urban areas face in balancing rapid development with the provision of affordable housing and can inform the development of more effective policies in other regions.

6. Conclusions

Focusing on the urban area of Qingdao as the case study, this research conducts a comprehensive examination of the spatiotemporal dynamics in residential property prices and the evolution of their underlying determinants from 2003 to 2023. The findings reveal that housing price dynamics are influenced not only by traditional factors such as location and community attributes, but also by broader urban developments such as infrastructure improvements, government policies, and shifts in industrial functions. The observed temporal shifts in price changes, particularly in peripheral districts, are likely driven by ongoing transportation network expansions, urban functional relocations, and the integration of new commercial developments. This change reflects the nested relationship between productive and reproductive spaces, and the mutual reinforcement of urban functional layout and class spatial distribution.

The spatial disparity in housing prices across Qingdao stems not merely from the tangible spatial allocation of resources but also reflects the spatial manifestation and perpetuation of urban functionalities and socioeconomic dynamics. The pattern of price differentiation reflects the economic agglomeration effect of resources and the “circular” characteristic of social resource acquisition, i.e., the differences in resource dependence and spatial distribution of different income classes, forming a multi-level residential space pattern. Among them, the scarcity of land supply in the core city pushes up the price of housing, while the peripheral areas attract population influx through the industrial–urban integration policy, forming a gradient spillover effect, and the extension of the metro network and the construction of Jiaodong International Airport reconstruct the locational value of Chengyang, Jiaozhou, and other emerging sectors; The demand for high-end housing is concentrated in the core city’s educational resources and business clusters, while the peripheral areas are driven by the upgrading of industries to form a “residential-investment” composite demand; middle and high-income groups to the mountain and sea scenery and the pursuit of school districts gave birth to the “resource monopoly” price plateau, while the foreign population and new demand for groups to Lichang, Chengyang, and other cost depressions, intensifying the “separation of work and residence”.

The findings suggest that central districts exhibit stability due to their concentration of educational, commercial, and transport resources. In contrast, peripheral regions experience greater volatility due to emerging infrastructure and commercial activity. These dynamics emphasize the importance of urban planners adopting a more integrated and flexible approach to managing urban growth and housing prices. Achieving a balance between residential, commercial, and recreational development in rapidly expanding areas can mitigate negative effects such as overcrowding and excessive commercial development. Based on these insights, we propose an urban planning framework to inform decision-making and reduce residential price disparities. Firstly, urban planners should prioritize balanced development in peripheral areas where infrastructure and commercial development are still emerging. Promoting mixed-use developments that integrate housing, commerce, and public amenities can help to reduce the negative impacts of commercial overconcentration. Secondly, zoning policies should be adaptable in order to respond to changes in market conditions and prevent imbalances between residential and commercial zones. Finally, governments can incentivize investment in underdeveloped areas through tax benefits and subsidies to promote sustainable growth and equalize opportunities across regions. Implementing these recommendations will ensure urban growth is equitable, preventing further price disparities and supporting sustainable urban development.

This study reveals the spatiotemporal differentiation characteristics of residential prices in Qingdao through the coupled application of LISA spatiotemporal leap and GTWR model, which provides empirical evidence to analyze the spatiotemporal heterogeneity of urban residential prices, but it is difficult to capture the dynamic pattern of short-term price fluctuations due to the large temporal granularity of the data. Future research can introduce high-frequency transaction data and build a real-time monitoring model by combining machine learning technology to more comprehensively analyze the temporal change characteristics of residential prices, which can support the scientific regulation of the housing market and the optimization of real estate policies and promote the healthy development of the urban housing market.