Nonlinear Hierarchical Effects of Housing Prices and Built Environment Based on Multiscale Life Circle—A Case Study of Chengdu

Round 1

Reviewer 1 Report

Determining the planning scale of urban life circles and their built environment indicators is indeed crucial for the reflection and control of residential differentiation. By using the GBDT machine learning method in combination with urban crowdsourcing data, the paper aims to leverage data-driven approaches to quantify and analyze these indicators. These indicators can provide valuable insights into the quality and characteristics of the urban environment, helping urban planners and policymakers make informed decisions to improve urban living conditions and address potential challenges. Based on the manuscript's high quality, as well as the quality of the work behind it, I highly recommend accepting it for publication.

Author Response

Thank you very much for your recognition of my manuscript.

Reviewer 2 Report

 

This paper explores the nonlinear relationships between assorted variables drawn from the built environment and aggregated house price information across three different scales, focusing on the case of Chengdu and utilizing a machine learning approach. The authors highlight scale and hierarchical effects within the relationship, indicating that house prices are more reactive to changes in the built environment at smaller scales. The research concerns a well-established domain, namely, the relationship between the characteristics of the built environment and house prices. Although the reported findings do not provide considerable new insights into relationships between house prices and the built environment characteristics, the machine-learning methodology applied may interest some readers. However, to be considered for publication, the authors are recommended to address the following major and minor points:

Major:

1. Certain well-established terms remain undefined. For instance, ‘residential differentiation’ is used throughout the manuscript without introduction. Generally, this term refers to the distribution of different types of housing and residential areas within a city or region. However, its application in various contexts suggests that the authors may be referring to something different. Consider the sentence in row 525, ‘[…] built environment indicators not only significantly influence residential differentiation, […]’. Here, how can an indicator that measures a ‘thing’ influence the very ‘thing’ it measures?
2. Only the first of the posed research questions, "What kind of relationship exists between built environment indicators and housing prices?” aligns with the applied research design. The subsequent questions, i.e., “Can a 15-minute life circle balance the built environment within all residential areas?” and “Can the role of built environment indicators at multiple scales explain the phenomenon of residential differentiation?” have not been examined and cannot be satisfactorily answered. The authors assert that their work offers ‘strong theoretical support for life circle planning’, however, the work neither represents a theoretical piece nor demonstrates strong theoretical support. It is strongly recommended that the authors critically re-evaluate the manuscript to ascertain which claims are substantiated by their data and research.
3. Dependent Variable: Even though measurements are aggregated, conclusions are presented as if data were gathered at an individual level. For example, the dataset is split into 'common residence' and 'high-end residence', but as per the description and Figure 1, house price data represent the mean price of various houses within a given location. Therefore, distinguishing between high-end residences and common residences is not feasible, as an area may feature both types and such mixed areas would be classified as average. In this scenario, a mean value would not accurately reflect whether it is a high-end or common residence.
4. Independent Variable: The choice of variables appears unconventional and inadequately justified. The authors are strongly urged to i) offer better theoretical justification for their variable choices, ii) cite relevant work in the field of house price research, and iii) more thoroughly discuss their limitations.
1. For example, the indicator for traffic accessibility (TA) is based on the total length of the road network. However, while the total length of a road network in a certain area can indirectly relate to traffic/transport accessibility, it's not a directly used measure for TA in relation to house price research. A long network of roads doesn't inherently guarantee improved accessibility. For instance, a rural area might have a long network of roads, but if those roads lead to few destinations or amenities, then the practical accessibility for residents might be low. In house price research, more direct measures of accessibility—like proximity to important amenities, travel time to key destinations, proximity to transit stations, and walkability—are usually favored. The authors are advised to either rename their TA indicator and clarify that this measurement refers weakly to transport accessibility and adjust all drawn conclusions on this variable, or to change their analysis by including established measurements and justifying their selection.
2. The indicator for population vitality (PV) combines aggregated movement activity with sport tracking data. The latter introduces various biases into the measurement, ranging from capturing data solely from Keep App users, to biases related to income, gender, and age regarding the usage of such an application. The authors are strongly urged to highlight these limitations. This is also applicable to shopping vitality (SV), which the authors acknowledge (see row 66 ‘has become a trend among young people’), but the implied limitations of such data have not been sufficiently recognized.
3. The authors are advised to argue why variables conventionally used in hedonic house price modelling have not been included, e.g., ‘structural characteristics’, ‘location characteristics’, ‘market characteristics’. For example, network centrality, or how central a property is in relation to the city has strong explanatory power.

6. The authors claim their work to be a multitemporal analysis. However, the incorporation of 'time' remains unclear.
7. It is ambiguous how the authors have addressed potential issues of spatial autocorrelation. For example, if spatial autocorrelation exists within the house price data, a spatial cross-validation would be necessary. The authors are recommended to address this concern, provide a test on spatial autocorrelation, and adjust their cross-validation strategy accordingly.

Minor:

1. Subheadings 3.1.1 and 3.2.1 are identical.
2. The authors might want to consider citing existing work on multi-scale urban analyses investigating relationships between the built environment and house prices published in this journal. For instance: https://doi.org/10.3390/ijgi12060249 or https://doi.org/10.3390/ijgi11050283.

 

 

 

1. While grammatically correct, the language used in the manuscript primarily relies on non-specific field terminology. For instance, the concept of ‘life circles’, rooted in local policy, essentially signifies the widely recognized term 'geographic catchment area'. To enhance comparability, authors are encouraged to use established terminologies where applicable.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

I would like to start by acknowledging the academic significance of your paper and the dedicated effort invested in it. However, during the review process, I have identified areas that could benefit from further enhancement. Below, I present my recommendations for revising your paper:

1. 1.While your paper presents a comprehensive analysis of Chengdu, I kindly request a more explicit elucidation of the reasons behind choosing Chengdu as your case city. Clarifying the rationale will assist readers in comprehending your research motivations. Furthermore, I advise a thorough exploration of the extent to which the conclusions drawn from Chengdu's data can be applied to China as a whole. If the conclusions lack general applicability, please ensure to explicitly address this aspect in the discussion section.

2. 2.Your paper singles out the hedonic model and contrasts it with other linear models. However, based on my literature review, designating the hedonic model as a nonlinear model might not be entirely accurate. I suggest reassessing this classification and comparing the hedonic model alongside other linear models to provide a more precise representation of your research. Since the approach mentioned in the cited references involves nonlinear transformations followed by linear regression, it is advisable to avoid presenting these two methods separately.

3. 3.I have observed citations of external literature and comparisons with other theories in your conclusion section. Nevertheless, I recommend minimizing direct citations of other works in the conclusion, unless your study has indeed revealed substantial findings that necessitate a comparison with other research. Expressing agreement or contrast with the research of others is best suited for the discussion section, allowing for a more cohesive presentation of your perspectives and reflections within the relevant academic discourse.

Several portions of the text seem to reflect the direct output of translation tools. Adding a touch of manual refinement could enhance coherence and academic authenticity.

Author Response

Please see the attachment.

Author Response File: Author Response.docx