Multiscale Impacts of Land Environmental Features and Planning on Apartment Resale Prices in Jinan City, China

: As a typical city with a population of 5 to 10 million in China, Jinan has undergone significant increases in land supply during the past decade, resulting in substantial volatility in apartment sale/resale market prices. In this study, we investigated all second-hand apartment transactions from 826 communities of Jinan city and explored the multiscale impacts of land environmental features and planning on apartment resale prices throughout the city. Specifically, central and eastern regions had significantly positive impacts on apartment resale prices, while western regions had significantly negative impacts; education resources had consistently positive impacts throughout the city while shopping, business buildings, and medical resources had insignificant impacts; subway stations had insignificant impacts and bus stations had significant effects only in congestion points and northeastern edges. Our results revealed the formation mechanisms and spatial heterogeneity of apartment resale prices in Jinan. Our work will not only help in the decision making of potential apartment purchasers, but will also be conducive to enhancing the spatial justice of local governments in land supply and planning policies.


Introduction
Global apartment sale/resale markets have exhibited substantial volatility over the past two decades and are characterized by irregular cycles of robust expansion and severe downturns [1].In 1998, the Chinese government stopped its welfare apartment distribution and gradually implemented apartment distribution monetization.From then on, China has officially entered the era of the comprehensive commercialization of apartments.According to data from the National Bureau of Statistics in China, during 1998-2017, the cumulative investment in real estate of China reached 85 trillion CNY, which has greatly improved the living environments of Chinese residents and promoted the development of urbanization in China [2,3].At the same time, apartment sale/resale markets have experienced pronounced instability due to rapid urbanization and intensive government interventions in China [4].
Unlike in Western countries, residential communities in China serve as the fundamental unit of urban living.Each community consists of about dozen of internal apartment buildings and is surrounded by walls or fences, and each apartment building consists of 5-33 floors and dozens to hundreds of apartments.Such a residential community model is shaped by China's considerable population and limited urban land [5,6].With the emergence of more urban subcenters since 2000, the disparities of apartment sale/resale prices in different communities in the same city have become increasingly significant.The demand differences among residents and structural features inside apartments as well as neighborhood attributes adjacent to communities have led to a spatial differentiation of apartment prices in China [7].
Second-hand apartment transactions in China are highly market-sensitive and provide a more accurate reflection of supply and demand than brand-new apartments due to the wide and dense distribution, so their price fluctuation and formation mechanisms have received wide attention from both purchasers and policymakers in China.During second-hand apartment transactions, prospective purchasers not only consider the location and structure of apartments, but also neighborhood facilities and accessibility [8].These facilities, such as transportation systems, educational institutions, adjacent parks, and restaurants, which cater to the daily needs of residents' work and living, offer additional benefits to apartments and have increasingly influenced apartment resale prices.A deep mining of disparity features and dynamical processes in apartment resale prices may enhance the understanding of formation mechanisms and further support the sustainability development of China.
Various models have been developed to explore the relationship between potential influential factors and apartment prices.The hedonic pricing model (HPM) can estimate marginal prices from different factors influencing the decision of potential purchasers [9,10].After embedding ordinary least squares (OLS) regressions, the HPM has become the most popular approach in real estate research due to its simplicity and effectiveness (e.g., [11,12]).However, OLS regressions do not consider spatial autocorrelation, which may generate biases when identifying the determinants of apartment prices.To overcome this drawback, improved models, such as the spatial lag model, spatial error model, and spatial Durbin model, have been developed to capture the spatial autocorrelation of apartment prices and influential factors [13,14].Although geographical phenomena often exhibit spatial heterogeneity due to the complex nature and interaction mechanisms, the regression coefficients in these improved models are typically spatially and temporally invariant.Geographically weighted regressions (GWRs) were introduced to reveal the spatial heterogeneity through allowing the regression coefficients to vary in the spatial domain [15,16] and has been widely utilized in the analysis of apartment sale/resale prices [17][18][19][20][21].More recently, multiscale geographically weighted regressions (MGWRs) [22] incorporated further local variations in all influential factors at multiple spatial scales [23].
Limited research on the apartment sale/resale prices of China has mainly focused on main metropolises with populations of over 10 million (e.g., Shanghai, Wuhan) or on wellknown port cities (e.g., Ningbo, Zhuhai).Huang et al. [24] investigated how apartment prices in Shanghai varied over space in Shanghai and what were the most important factors in determining the market value of a single apartment through GWR modeling, and they found that the attribute effects were different over space and the GWR coefficient of the total area and the distance to the downtown area were distributed regularly.Liang et al. [25] found that with the help of OLS and GWR modeling, adjacency to subway stations in Ningbo city exerted slightly positive impacts on apartment prices in downtown regions and was significantly positive in non-downtown regions despite the fact that many subway stations were still in the planning stage.Liu et al. [26] took the transportation analysis zone of Wuhan city as the spatial unit and utilized MGWR modeling to investigate the spatial scale of the impact of the accessibility of various public facilities on apartment prices.Liu and Strobl [27] used the performance of three hedonic pricing models (OLS, GWR, and MGWR) on modeling thirteen neighborhood features on apartment prices in Zhuhai city and indicated that GWR and MGWR accurately demonstrated local spatial heterogeneity.Lu et al. [28] further demonstrated the good modeling performance of GWR and MGWR in Wuhan City.As a window of communication between China and the world, apartment sale/resale prices in these metropolises or ports have been among the highest in the world, which are very different from ordinary cities in China.
Our study selected Jinan city as the research area.Jinan city is located in the eastern region of China and is the capital of Shandong Province.Jinan city has 10 districts and 2 counties, with a total area of 10,244 square kilometers and a population of 8.2 million.Its core urban area consists of five districts (Lixia, Shizhong, Tianqiao, Huaiyin, and Licheng) and has an elongated shape bounded by mountains in the south and the Yellow river in the north.Jinan is a park city known to be "surrounded by lotus and willow trees and with a city of mountain view and a half-city lake view" since ancient times.Although its GDP is listed among the top 20 cities in China, Jinan's development has significantly lagged behind any main metropolis and well-known ports in China and is rarely affected by international market fluctuations.As a typical megacity with a population of 5 to 10 million in China, Jinan's urban area has undergone significant expansion in the past decade with the urbanization rate up to 75.3% in 2023.Such rapid urban expansion and urbanization resulted in substantial volatility in apartment sale/resale markets in Jinan.The aim of our study was to explore the multiscale impacts of structural features, neighborhood facilities, and accessibility on apartment resale prices in Jinan city.
This article is organized as follows: Section 2 introduces the data sources of apartment resale prices and internal and neighborhood features in Jinan, and examines the correlation among apartment resale prices and various influencing factors.Section 3 uses three hedonic pricing models (OLS, GWR, and MGWR) and compares their modeling performance on apartment resale prices in 826 residential communities of Jinan city.With the help of optimal modeling by MGWR, Section 4 reveals the impacts of location, structural information, neighborhood features, and facility accessibility on apartment resale prices in 826 residential communities of Jinan city.Finally, Section 5 gives some conclusions and possible suggestions and recommendations.The detailed flowchart of our apartment resale price analysis is shown in Figure 1.

Apartment Resale Data: Preprocessing and Correlation Analysis
Lianjia Real Estate Co. Ltd.(Beijing, China) is the largest Chinese real estate service enterprise, established in 2001, and is committed to providing safe, efficient, and comfortable real estate transaction services for Chinese residents.Its online platform (www.lianjia.com, assessed on December 1, 2022) can provide the most comprehensive database on second-hand apartment information and related resale prices in China.In this study, we collected structural features of second-hand apartments and the latest resale prices from the Jinan branch of Lianjia.After excluding transaction data with severely incomplete information and abnormal apartment prices, we obtained 22,554 second-hand apartment transactions, mainly from 826 different communities in Jinan city.For each transaction, structural features including total area, number of bedrooms, age, floor, and decoration status were collected.In this study, these five structure features were investigated.Their detailed description and basic statistical analysis are shown in Table 1.For external influencing factors, since different second-hand apartments in the same residential community have very similar neighborhood features due to the small size of each community in Jinan, we ignored the internal differences in residential communities in the analysis of neighborhood effects and the accessibility of public facilities on apartment resale prices among all 826 residential communities.Figure 2   Compared with internal factors, the neighborhood environment and facility accessibilities of each residential community are core external influential factors of apartment resale price.Convenient transportation and living facilities, high-quality primary and secondary education, and high-quality medical conditions often bring added value to second-hand apartment transactions, while industrial pollution and noise often have a significant negative impact on second-hand apartment transactions.Gaode Map (Mountain View, CA, USA) (https://ditu.amap.com/(accessed on 26 June 2024)) is a leading provider of digital map content, navigation, and location service solutions in China.All neighborhood features of any residential community in Jinan can be captured well from Gaode Map.We considered ten types of neighborhood features, which are described in Table 2.In order to better evaluate the effect of facility accessibilities, any neighborhood attribute around each residential community was weighted, calculated not only by their types in Table 2 but also by the range distance to the community in Table 3.The score of each neighborhood feature was equal to the distance score multiplied by the type score.

Range Distance to the Community (m)
Score 0-300 10 300-600 5 600-1000 3 Applying the Pearson correlation test indicated that apartment resale price was highly related to most of the internal and external influential factors (Figure 3).In detail, apartment resale prices in Jinan city showed obvious positive correlations with total area, decoration status, number of bedrooms, bank, business, park, shopping, and bus stations, but negative correlations with subway stations, factory, and restaurant.The collinearity between total area and bedroom number was clear, and the high correlation with bank and other neighborhood attributes was also observed.
The spatial dependence of apartment resale prices in Jinan was assessed by the Moran's I statistic.After normalization by variance, the range of Moran's I was limited to the interval [−1, 1], where positive values mean positive correlations, negative values mean negative correlations, and the magnitude can measure the degree of high/low values whether attracted or not.The Moran's I value of 0.417 with a very high p-value (Table 4) revealed that the spatial distribution of apartment resale prices in Jinan had a significant positive autocorrelation and regional agglomeration.

Hedonic Analysis
Hedonic analysis is a widely used approach to estimate the impacts of internal and external factors on apartment resale prices.Ordinary least squares (OLS) regression is the first tool to be used in the hedonic analysis and can be expressed as where   is the resale price of the ith second-hand apartment,  0 is the interceptreflecting location impact,   represents the value of the kth influencing factor for the ith second-hand apartment,   is the associated regression coefficient, and   is the background noise.The main drawback of OLS lies in the fact that it is a global regression and ignores local effects completely.Geographical weighted regression (GWR) is a local regression which strengthens the expression of spatial heterogeneity by adding spacevarying parameters (  ,   ) as follows: Different neighborhood attributes possibly have different areas of influence, leading to spatial effects on apartment resale prices appearing at different scales.As an improvement in GWR, multiscale geographical weighted regression (MGWR) assigns different bandwidths to each influencing factor as follows: where the label  represents different bandwidths of each influential factor.
In order to make a fair comparison on factor effects on apartment resale prices in Jinan, all possible internal/external influencing factors eliminated scale effects through a normalization process.At the same time, in order to avoid model outfitting and reduce the collinearity between influencing factors, a stepwise OLS regression was utilized; finally, 13 influencing factors were retained (Table 5).Two influencing factors (subway and medical) were omitted.The main internal mechanisms are as follows: The first subway in Jinan started operating in 2019, with a total operating mileage of only 84 km by 2022, so the accessibility of the subway had a very limited impact for Jinan with a total area of 10,244 square kilometers and a population of 8.2 million.In addition, Jinan is the capital of Shandong Province, with nearly 100 million residents, boasting the second highest number of tertiary hospitals within China.Moreover, the Jinan government has made great efforts in the recent ten years to build branches of well-known hospitals near newly built communities, ensuring the rational layout of medical resources throughout the city.Under this scenario, medical conditions had limited impacts on apartment resale prices in Jinan city.The regression coefficient column in Table 5 measures the link between influencing factors and apartment resale prices.It revealed that bank, area, business, bus, bedroom, park, decoration, and education exhibited positive correlations with apartment resale prices, while restaurant, factory, and age exhibited negative correlations.Specifically, the coefficients associated with bank, business, and restaurant had large magnitudes, implying a stronger impact on apartment resale prices.Combining score standards in Tables 2 and 3 with the outcomes of the stepwise OLS regression coefficients in Table 5, it was revealed that for one more commercial building within a 300 m radius of any residential community, the resale unit price of its second-hand apartments would increase by 390 CNY/m 2 ; for every year of increase in apartment age, the apartment resale price would decrease by 80 CNY/m 2 ; and for one more restaurant within a 1 km radius of any community, the apartment resale price would decrease by 40 CNY/m 2 .
We used three regression techniques (OLS, GWR, MGWR) for hedonic analysis on apartment resale prices in Jinan.Modeling performance is demonstrated in Table 6 in terms of six assessment criteria: residual of squares (RSS), coefficient of determination (R 2 ), adjusted coefficient of determination (Adj.R 2 ), Akaike information criterion (AIC), corrected Akaike Information Criterion (AICc), and Bayesian information criterion (BIC).The values of six assessment criteria demonstrate significant advantages of GWR and MGWR over a traditional OLS approach: The decreasing RSS value and increasing R 2 and Adj. 2 values demonstrate that both GWR and MGWR offer superior capacity in explaining the link of internal/external factors and apartment resale prices.Three criteria (AIC, AICc, and BIC) could assess the selection of bandwidths.The substantial decrease in AIC, AICc, and BIC from OLS to GWR and MGWR suggests that local regression models provide a closer fit to substantial spatial effects on apartment resale prices.Table 6 shows that the resale prices of second-hand apartments in Jinan varied very differently with influencing factors.The presence of spatial heterogeneity in apartment resale prices cannot be captured by any global regression model like OLS.By introducing the idea of a kernel function and bandwidth, GWR and MGWR models become effective in capturing local relationships between apartment resale price and spatial neighborhood attributes.Finally, it was observed that MGWR exhibited a superior level of fitness compared to GWR.

Multiscale Impact Analysis
It is worth noting that the GWR only offers insights into the average scale of each influencing factor throughout the city, while the MGWR enables a direct depiction of multi-scale variation among different influential factors, so we adopted MGWR to deep mine complex links between apartment resale prices and spatial neighborhood attributes and so understand the mechanisms of price formation.Standardized MGWR coefficients not only show multiscale characteristics of various influencing factors that vary with space, but also provide their statistical significance (p-value), from which more accurate information about the co-occurrence of spatial heterogeneity and homogeneity can be revealed.For the MGWR analysis of second-hand apartment resale prices in Jinan city, the bandwidths and statistical characteristics of coefficients for each influencing factor are shown in Table 7.By reverting the standardized MGWR coefficients to their original weights on neighborhood attributes, we found that every year of increase in apartment age could lead to a reduction of approximately 144 CNY/m 2 in apartment prices; the presence of an additional bus line with a stop in a 300 m radius of the residential community could lead to an increase of 14 CNY/m 2 in apartment prices; fine decoration led to a premium of 860 CNY/m 2 compared to simple decoration; and the introduction of a new middle school, primary school, and kindergarten within the 300 m radius of any community led to a premium of 136 CNY/m 2 , 36 CNY/m 2 , and 7 CNY/m 2 , respectively.The long-term onechild policy, although it has changed in recent years, has resulted in most Jinan families tending to have only one child.As the birth rate decreases, kindergartens in Jinan are clearly oversupplied, which leads to a very limited impact of kindergartens on apartment prices.China's college entrance examination is widely recognized as the most fiercely competitive in the world.Middle schools play a key role in children's future entry into top tier universities, leading to many families paying attention to whether there is a middle school near the residential community, which in turn affects apartment resale prices.

Impacts of Location and Structure Features
Applying MGWR-based hedonic analysis on apartment resale prices in Jinan, Figure 4 demonstrates the distribution of the MGWR intercept and coefficients associated with structural features.According to the distribution of the MGWR intercept, in central and eastern regions of Jinan, the community location has significantly positive impacts on apartment resale prices, while in western regions, the community location has significantly negative impacts.The central region is the historical downtown region of Jinan and a traditional hub for politics, economy, and culture in Jinan; the eastern region is the incoming central business district that has been heavily invested in and constructed by the Jinan government in recent years; and the headquarters of large banks and enterprises in Shandong Province have gradually relocated.These factors led to a significant positive impact of location on apartment prices in central and eastern regions of Jinan.In contrast, residential communities in western regions emerged on a large scale with the construction of the Jinan high-speed railway station in the last ten years.Due to the lack of sufficient job opportunities in western regions, these communities were often referred to as a "sleeping city" and so residents in Jinan lacked the willingness to purchase second-hand apartments in western regions.This led to significantly negative impacts of community location on apartment resale prices.Among all structural features of second-hand apartments, the age of apartments demonstrated consistently negative impacts and was the only feature with statistical significance in almost all residential communities in Jinan.This indicated that the age of the apartments plays an undeniable role in the formation of apartment resale prices, aligning with empirical rules of second-hand apartment transaction markets where newer apartments tend to command higher prices and older apartments tend to command lower prices due to the difficulty in obtaining bank loans.Apartment age impacts in the western regions and northeastern edges of Jinan urban areas were much less than those in central and eastern regions.Residential communities in the western regions and northeastern edges were built only in the past decade, so the apartment resale price in these regions was not sensitive to apartment age.However, a large number of old residential communities were distributed in central and eastern regions, resulting in apartment age playing a key role in price formation and leading to a faster decease in apartment price with age.It is worth noting that in central Jinan, there was a small circular region where age impact was not statistically significant.This circular region is around Jinan railway station.Various unfavorable environmental factors made apartment age impacts become negligible.
In the second-hand apartment transaction market in Jinan, any apartment with a larger area means a higher net price and then a lower demand, resulting in the unit area price of a large-area apartment always being slightly lower than that of a small-area apartment.Figure 4 demonstrates that apartment area had no statistically significant impacts on most communities, except for two specific regions.The apartment area in the southern protrusion had a significant positive impact on resale price.This region is adjacent to the southern mountainous area of Jinan and is far from downtown Jinan.Residential communities distributed in this region are always surrounded by green mountains and clear waters and have a lower unit area price.The apartments in these residential communities were purchased mostly by seniors who had good economic conditions and pursued a comfortable living environment, so apartment area had significantly positive impacts on resale prices.The other region located in central Jinan, which demonstrated that apartment area had significantly negative impacts on resale prices, is the historical downtown region of Jinan.It is not only a traditional hub for politics, economy, and culture in Jinan, but also owns many iconic spring attractions in Jinan.Purchasing an apartment in this region is a goal of all Jinan residents, so unit area price reached the highest in all communities of Jinan (Figure 2).The combination of limited budget and the highest unit price resulted in apartment area having significantly negative impacts on resale prices.
MGWR-based hedonic analysis revealed that residents in Jinan consistently considered the number of bedrooms in any apartment almost throughout Jinan city.The positivity of the MGWR coefficient associated with the bedroom factor indicated that more bedrooms would significantly increase the unit area price of apartments.Moreover, the magnitude of the MGWR coefficient associated with the bedroom factor was significantly larger than that of age and total area factors.Due to the 40-year-long onechild policy, middle-aged residents in Jinan not only faced the responsibility of raising 1-2 children, but also had to take care of both parents.Due to the need to allocate a bedroom to each family member, the number of bedrooms was particularly important in secondhand apartment resale.It is worth noting that the number of bedrooms had higher impacts in the southeastern region than in the northeastern region.Due to the favorable environment and the proximity to the central business district of Jinan, most residents in southeastern regions are from a rich class who are not sensitive to the transaction price.The demand for more bedrooms was driven by the pursuit of living quality, resulting in a larger increase in apartment prices.In contrast, the traditional industrial zone in Jinan is located in the northeast; due to the poor environment and limited budget from the middleclass, the demand for the number of bedrooms was a necessity and had a smaller impact on the increase in apartment prices.However, in western and southwestern regions, the number of bedrooms had no statistically significant impacts, even negative impacts on apartment prices.Due to the low unit area price in these regions, residents can afford large-area apartments relatively easy and so the number of bedrooms is no longer a limiting factor in purchasing decisions.
Newly built apartment buildings are always equipped with elevators, so floor factors had no statistically significant impacts on apartment prices, except for some central regions and the southern protrusion of Jinan, where many old residential communities existed.The apartment buildings in these old communities had not been equipped with elevators, so high floors had significantly negative impacts on apartment prices.After purchasing a second-hand apartment, most residents in Jinan tended to redecorate it.This led to the decoration factor exerting no statistically significant impacts almost throughout the city, except for some regions with investment in decoration.

Impacts of Neighborhood Features
MGWR-based hedonic analysis indicated that adjacent educational resources had overall positive impacts on apartment prices (Figure 5), consistent with the empirical rule that apartments near better educational resources tended to have higher demand.However, its impact degree demonstrated a significant regional difference.The northeastern regions are the traditional industrial zones in Jinan, where various factories and high-tech industrial parks provide sufficient job opportunities.At the same time, residential communities in this region have a relatively poor environment from industrial pollution and noise as well as relatively poor living facilities; however, apartments in these communities are affordable for the middle class and lower middle class.In addition to continuously improving the environment, the Jinan Government has been constantly relocating the best primary and middle schools or setting up branch schools (e.g., Shandong Experimental Middle School, SDNU Affiliated Middle School) in the northeastern regions so that adjacent educational resources become the highlight in the second-hand apartment transaction market in this region and further enhance the impact degree.In contrast, central regions are a traditional hub for politics, economy, and culture in Jinan, so they possess high-quality mature educational resources and are well distributed, so an imbalance in education sources can be ignored in apartment price formation in central regions.Residential communities in the southern protrusion are far from the downtown area and are always surrounded by green mountains and clear waters.The apartments in these residential communities were purchased mostly by seniors who had good economic conditions and pursued a comfortable living environment, but did not need education sources, so the impacts of education sources could be relatively small.Parks generally offer local residents a convenient space for both exercise and relaxation.Any residential community adjacent to a green space means good environmental and atmospheric conditions.In the past ten years, Jinan has built 1202 parks with an area of over 500 square meters, including 87 mountain parks, 140 mountain greening upgrades, and more than 300 km of mountain forest greenways.Jinan's largescale restoration of mountain ecology won the China Habitat Environment Example Award.Since parks are not a scarce source in Jinan, our MGWR-based hedonic analysis revealed that only parks in the northern edges had statistically significant impacts on apartment resale prices (Figure 6 left) due to a lack of park sources in the northern edge.Unlike adjacency parks, our MGWR-based hedonic analysis revealed that restaurants caused an overall negative impact on apartment resale prices (Figure 6 right), indicating that the presence of more restaurants would lead to a reduction in the value of adjacency apartments; especially, the impact degree reached a maximum in downtown and its west region (i.e., the historical business zone).The significantly high demand from visitors boosted the catering and retail industries, making the surrounding environment degraded and further resulting in much larger negative impacts.The southern protrusion and eastern region had a significantly lower density of restaurants and so generated less impacts.Our MGWR-based hedonic analysis revealed that both the bank and factory had a significant impact on apartment resale prices (Figure 7).Banks had a positive impact on apartment resale prices, while factories had a negative impact on apartment resale prices.Although the impact degrees of bank and factory shared similar spatial patterns (high in the west and low in the east), their driving mechanism was very different.Central and eastern regions of Jinan belong to a historical downtown zone and an incoming central business district, respectively, so various branches of different banks are distributed densely in these regions.However, western regions were known as the "sleeping city" of Jinan and few business activities were carried out, leading to a sparse distribution of bank branches.This made apartment resale prices more sensitive to financial amenities in western regions than central and eastern regions.The main industrial enterprises and high-tech industrial parks in Jinan are all located in eastern regions, so residents in eastern regions are more concerned about the impact of environmental pollution and noise generated by these enterprises on their living conditions.This resulted in the factory factor having more impacts in eastern regions than western regions.Our MGWR-based hedonic analysis revealed that both shopping and business buildings had a significant impact on apartment resale prices (Figure 8).Large shopping complexes and chain supermarkets in Jinan had entered a mature development stage, and their accessibility was very high.This means that adjacency to shopping was not a significant influential factor.At the same time, only small portions of Jinan residents worked inside business buildings, so its impacts might be ignored.

Impacts of Transportation Accessibility
The main public transportation system in Jinan consists of subways and buses.The first subway in Jinan started operating in 2019, with a total operating mileage of only 84 km by 2022, so the accessibility of the subway had a very limited impact on residents of most communities in Jinan.The stepwise OLS regression in Section 3 also indicated that the impacts of subways can be ignored, so we only investigated the bus system here.
Although Jinan owns well-constructed and accessible bus routines, due to the elongated shape of the urban area, Jinan is still one of the most congested cities in China.In recent years, Jinan has made great efforts to solve most of the congestion points, so residents in most regions of Jinan do not consider bus accessibility.Two small patches in urban areas in Figure 9, which demonstrate a higher impact degree of bus accessibility, are just the remaining transportation congestion points.At the same time, the bus system in the northeastern edges is the poorest in the whole of Jinan city, leading to greater concern on bus accessibility.Compared with these regions, bus accessibility in southeastern regions had negative impacts.Due to the good environment and adjacency to the central business district, the main residents in southeastern regions are from the rich class, and they do not depend on the bus system and pursue high-quality living more.More bus accessibility possibly increases the population density of the community, resulting in negative impacts of bus accessibility on apartment resale prices.

Discussion
China's real estate has flourished since China ended its welfare apartment policy in 1998.It has not only improved the living conditions of all Chinese residents, but has also become a pillar industry of the national economy.The second-hand apartment transactions in China are highly market-sensitive and provide a more accurate reflection of price fluctuations than brand-new apartments due to the wide and dense distribution, so its price fluctuation and formation mechanism have received wide attention from both purchasers and policymakers in China.Jinan is a typical city with a population of 5 to 10 million in China.In this study, we collected 22,554 second-hand apartment transactions, mainly from 826 different communities in Jinan, and explored the multiscale impacts of five internal structural features and ten external neighborhood features on apartment resale prices.
Our results will help potential apartment purchasers in Jinan in understanding the price formation mechanisms of second-hand apartments and then in making suitable decisions in the negotiation of second-hand apartment transaction prices; e.g., the accessibility to shopping, business buildings, and medical resources did not provide additional value to second-hand apartments; the total area only in the southern protrusion had a significant positive impact on apartment resale price, which was driven by seniors who had good economic conditions and pursued a comfortable living environment; the western regions had significantly negative impacts on apartment price due to the lack of sufficient job opportunities, thus resulting in a long commuting time; the adjacent educational resources had more impacts on apartment prices in the traditional industrial zone of northeastern regions than the southern protrusion; and the bus accessibility had high impacts only on transportation congestion points and northeastern edges.
Our results will also be conducive to enhancing the spatial justice of local governments in land policy and urban public services.According to our impact degree analysis of neighborhood features, in order to further stabilize apartment prices in Jinan, we suggest that the government assign more educational resources to the northeastern traditional industrial zone, build more parks in northern edges, set up more bank branches in western regions, and enhance the environmental protection requirements in the traditional industrial zone.At the same time, according to our impact analysis on apartment structure, we suggest that real estate developers in Jinan should build apartments with more bedrooms in the central and eastern regions and large-area apartments in the southern protrusion and incorporate high-quality educational resources in new residential communities.

Conclusions
Due to significant increases in land supply during the past decade, the urban area of Jinan city has undergone significant expansion and the urbanization rate has reached up to 75.3% in 2023.Such rapid urban expansion and urbanization resulted in apartment sale/resale markets in Jinan exhibiting substantial volatility.Pearson correlation analysis and the Moran's I value indicated that apartment resale price in Jinan was highly related to most of the internal/external features, and it had a significant positive spatial autocorrelation and regional agglomeration.Further analysis of the standardized MGWR intercept and coefficients revealed the impact degree and spatial heterogeneity of various internal/external features: ➢ For structural features, apartment age had consistently negative impacts on apartment prices, while the number of bedrooms had consistently positive impacts throughout the city; floor and total area had negative impacts for resale prices in historical downtown regions, where many old apartment buildings were without elevators and had a high unit area price, but the total area had positive impacts on resale prices in the southern protrusion due to the good environment and low unit area price.
➢ For neighborhood features, education resources had consistently positive impacts on apartment prices throughout the whole city, and the largest impact degree appeared in the northeastern traditional industrial zone in Jinan.Restaurants had consistently negative impacts on apartment prices throughout the city, and the impact degree reached the maximum in the traditional business zone.Parks only had statistically significant impacts in the northern edges due to too many parks in Jinan.Banks had a positive impact while factories had a negative impact throughout the city, and banks and factories shared similar spatial patterns of impact degrees.It is worth noting that shopping, business buildings, and medical resources had insignificant impacts.➢ For transportation accessibility, subway stations had insignificant impacts and bus stations had significant effects in congestion points and northeastern edges.
Although subway impacts are significant on apartment prices in many cities, our analysis revealed this is not the case in Jinan, due to the limited operating mileage.At present, the Jinan government is vigorously investing in subway construction with a total of seven lines and a total length of 203 km under construction.These subway lines will be expected to operate in 2027 and cause a large change in transportation accessibility and modes.We will conduct further analysis of subway impacts in the future.

Figure 1 .
Figure 1.The flowchart of our apartment resale price analysis.
demonstrates the distribution of transaction unit prices (Unit: CNY/m 2 ) of second-hand apartments in all 826 communities of Jinan city.It ranged from 7196 CNY/m 2 to 48,693 CNY/m 2 , and the highest unit prices appeared in the south part of Lixia district and Shizhong district in Jinan city.

Figure 2 .
Figure 2. The distribution of transaction price (unit: CNY/m 2 ) of second-hand apartments in Jinan.

Figure 4 .
Figure 4. Distribution of MGWR intercept and coefficients associated with structural features.

Figure 5 .
Figure 5. Distribution of MGWR coefficients associated with adjacent educational resources.

Figure 6 .
Figure 6.Distribution of MGWR coefficients associated with adjacent park (left) and restaurant resources (right).

Figure 7 .
Figure 7. Distribution of MGWR coefficients associated with adjacent bank (left) and factory resources (right).

Figure 8 .
Figure 8. Distribution of MGWR coefficients associated with shopping (left) and business buildings (right).

Figure 9 .
Figure 9. Distribution of MGWR coefficients associated with bus resources.

Funding:
The corresponding author was supported by the European Commission Horizon 2020 Framework Program No. 861584 and the Taishan Distinguished Professor Fund No. 20190910.

Table 1 .
Internal structural features of second-hand apartments.

Table 2 .
Neighborhood features in residence communities of Jinan.

Table 3 .
The descriptions and score of range distance.

Table 4 .
Spatial autocorrelation of apartment resale prices.

Table 5 .
Selection of influencing factors through stepwise OLS regression.

Table 7 .
Statistics of MGWR parameter estimates.