The Influencing Factors of a Polycentric Employment System on Jobs-Housing Matching—A Case Study of Hangzhou, China

Round 1

Reviewer 1 Report

The paper is very interesting and the analysis has well described.

I suggest to rich the analysis with some territorial elements. Could be useful to analyse "Towards a European Territorial Reference Framework"
Andreu Ulied (MCRIT) (ed.) to catch some territorial attractors that could improve and influence the dichotomies described.

It would be better to indicate the type of workers involved in the analyzes (sector to which they belong, educational qualifications, family composition, etc.).
The Author indicates the time on which the analysis was conducted (p. 45 location during sleep time (12 PM to 5 AM) and working time (9 AM to 5 PM) )so this suggests that no work shifts is taken into consideration. Make this clearer.
Did the workers know they were involved in the analysis?
Does the author know any aspects related to the quality of life? this is one of the aspects that very much affects urban and residential development in Europe. Is it an element that could also affect the residents of the study area?

Author Response

Thanks for your comments to help us improve the paper.

The paper is very interesting and the analysis has well described. I suggest to rich the analysis with some territorial elements. Could be useful to analyse "Towards a European Territorial Reference Framework"Andreu Ulied (MCRIT) (ed.) to catch some territorial attractors that could improve and influence the dichotomies described.

Thanks for the reviewer’s suggestion, we find some files about the “Towards a European Territorial Reference Framework”, and we think the reviewer may refer to the file” Annex 2: Modelling Results”. In this file, it introduces the MASST Model (Macroeconomic, Sectoral, Social and Territorial Model), in the model there are many territorial attractors that the reviewer think can improve our model. The MASST model give us a lot of inspiration, the model include territorial model, social and sectoral model, local competitive model, macroeconomic model, especially the territorial model gives us a lot of enlightenment in explaining the spatial connections between regions (such as proximity effect) and the territorial structure of regions (urbanization, agglomeration and rural structure of regions). And the variables used in the model are also very illuminating to our research, such as borrowed size, borrowed functions, urban land rent, relative specialization, share of college graduates etc. But in our case, this study is more inclined to intra-urban polycentric problems, may be similar to the FUA (functional urban areas) of Europe, the European Territorial Reference Framework is more inclined to regional polycentric problems, and payed attention to the inter-urban polycentric problems. So we have to study the MASST, the territorial attractors can’t be directly use in China, as Chinese and European cities are quite different in terms of territory, urban development, culture, economy and institution.

it would be better to indicate the type of workers involved in the analyzes (sector to which they belong, educational qualifications, family composition, etc.).

This is a good suggestion, but for the data availability, we could not include these factors, maybe in the future by sample survey we could enrich our research model. The signaling data is anonymous, and we can also know more about the workers, but for the privacy protection reasons, it is not yet possible to collect.

The Author indicates the time on which the analysis was conducted (p. 45 location during sleep time (12 PM to 5 AM) and working time (9 AM to 5 PM) )so this suggests that no work shifts is taken into consideration. Make this clearer. Did the workers know they were involved in the analysis?

 The situation that work shifts haven’t taken into consideration in our data processing, however, this did not affect the results of our study. First, we have a large sample of 7.26 million cellphone used identified by cellphone signaling data; Second, we calculate the repetition rate of users during working hours and residence hours respectively within 20 working days (more than 60%), and the work shifts can be effectively ruled out; Third, only the users whose place of residence and workplace both be identified can ultimately use for data analysis. So, our rules exclude effectively work shifters, and the percentage of night shift workers who reverse day and night will be very small, according to the industry situation of Hangzhou, it will only account for a tiny proportion among our more than 7 million identifications. We do not think this will have a significant impact on the results, at the same time, there have been most literatures using cellphone data for jobs-housing analysis adopting this method (Zhou, Yeh, Li & Yue,2018; Zhang, Zhou & Zhang,2017; Zhou, Chen & Zhang,2016).

The workers didn’t know they were involved in the analysis. The signaling data is from a mobile communication operation company. As cellphone signaling data is anonymous and is collected through a passive process. The records include an anonymous user ID, the latitude and longitude coordinates and the time the signaling event was generated. The latitude and longitude coordinates directly correspond to the location of cellphone tower. The signaling dataset used in this study contains all events, e.g. the start and end of incoming or outgoing call, sending and receiving an SMS, mobile phone switched on or switched off. In additional to these events, the cellphone signaling dataset also contains location update event. When a mobile phone is moving from one cell to another cell in mobile network, the network updates its position and registers the signal as a location update event.  In this process, no investigator is required to participate in, and no personal privacy information will be collected.

Does the author know any aspects related to the quality of life? this is one of the aspects that very much affects urban and residential development in Europe. Is it an element that could also affect the residents of the study area?

Quality of life factors are indeed more important to the choice of jobs and housing behavior, in particular, the residents of western developed countries pay more attention to the quality of life. In China's big cities, people's living standards are getting better and better, and more and more people tend to choose areas with better quality of life, such as better schools, perfect public service facilities, rich cultural and entertainment facilities, greening and other environmental qualities. Reviewers have put forward a good direction for our study. In the future, we can continue to improve the model and add related life quality indicators.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper deals with the relationship between the urban polycentric spatial structure and jobs-housing matching in Hangzhou city in China. Research data (ex. Cellphone signaling data), data processing and analysis are interesting and also well described. Article structure is clear. However, this reviewer has some concerns, especially, in the discussion.

For example,

・Figure 3 is little complicated because the correlation coefficients and p-values are arranged vertically. Why do not you enclose the p-value in parentheses?

・This reviewer thinks that the explanation about the meanings of influencing factors derived from the results of multiple regression analysis is insufficient.

・In the conclusion part, it is better to discuss the novelty and new findings of your study, compared to the previous research. Although the conclusion is long, most of them are duplicated with the part of results.

Author Response

We would like to thank the reviewer for pointing out the problems to help us make the paper better. According to some concerns of the reviewer,

Figure 3 is little complicated because the correlation coefficients and p-values are arranged vertically. Why do not you enclose the p-value in parentheses?

The problem of Figure 3 probably refers to Table 3, and we think it is a good suggestion, and we revised the table according to the reviewer’s suggestions.

This reviewer thinks that the explanation about the meanings of influencing factors derived from the results of multiple regression analysis is insufficient.

According to the reviewer, the meanings of influencing factors derived from the results of multiple regression analysis is insufficient, we made some improvements as follows:(1) We rewrote the two paragraph, reorganized the structure to make the factors more clearly expressed;(2) We added some factors explanation to make it more clear understand. For example:

The revised the paragraph of workers’ jobs-housing matching is as follows:

As shown in Table 4, model 1 confirms the positive correlation between the size of the employment center and the workers’ jobs-housing matching rate, and the E/R and subway accessibility have negative correlations. Model 2 confirms the negative correlation between the level of industrial specialization and the workers’ jobs-housing matching rate. Model 3 considering the factors of model 1 and model 2, and it improved explanatory powers with the R-squares to 53.3%, the size of employment, E/R and industrial specialization are consistently significant in all model specifications: as expected, the distance of workers from their place of residence becomes father as the size of employment centers decreases, as controlling for the other variables, the area of employment center increases 1 unit, the workers’ jobs-housing matching is expected to rise by 0.286 unit; the higher the E/R, workers are more likely to travel father from their place of residence, as controlling for the other variables, the E/R increases 1 unit, the workers’ jobs-housing matching rate will be reduced by 0.421 unit; the more specialized the employment center is, the more difficult it is for workers to choose a suitable place to live near the workplace, which indicates that the heterogeneity of employment opportunities is more important in explaining the housing selection decision, as controlling for the other variables, the industrial specialization index increases 1 unit, the workers’ jobs-housing matching rate will be reduced by 0.379 units. However, there are some factors that had significant influence in the correlation analysis were not include in these models, such as the employment density, distance from CBD, natural barriers and industrial agglomeration factors, indicating that they are not determinant factors of workers’ jobs-housing matching.

In the conclusion part, it is better to discuss the novelty and new findings of your study, compared to the previous research. Although the conclusion is long, most of them are duplicated with the part of results.

The reviewer thought the conclusion need to be improved, we rewrote the conclusion, and the main changes were as follows:(1) We reorganizes the structure of the conclusion, summarizes it in three parts: polycentric spatial structure, jobs-housing matching characteristics, and influence factor analysis, highlighted the important findings of our research and compared with the western results.(2)We applies the research results to the practical problems of urban development, points out the meaning of the results to the reality, and provides guidance for the sustainable development of urban structure policy making.(3)We analyzes the difference of research results between China and western countries, and analyzes the difference of research results from the perspectives of market economy, institutional factors, policy factors, etc. It shows that the development of urban spatial structure is complex and it is difficult to fully explain the phenomenon by simple factor analysis.

Reviewer 3 Report

The paper is clear and well written. However, it could be improved in some parts. Here some points: 

The introduction makes some mistakes by combining the descriptive and the normative (policy) meanings of polycentricity. The authors could stress the two different meanings (e.g. page 2, lines 42-43, it is not clear why there is a consequential relationship between polycentricity as phenomenon and polycentricity as a planning tool.  It is not really clear why the case study should offer an analysis of Chinese cities, as stated in introduction (page 3, lines 121-122). This is linked with the conclusions, which should provide some policy implications. Otherwise, the article looks more like an exercise, rather than a substantive contribution to the research about urban spatial structure. In other words, the authors should stress more what we can learn from this case study, both for other Chinese cities (if, as stated, this was a goal of the analysis) and in general about megacities overall the world.  The authors apply am "American-based" framework in a different context. Chinese cities are different from US cities, but also institutions and land regulation is different. The authors can explain the main differences. Actually, a discussion about the regulation of urban land (starting from the 80s reforms) in China might enrich the paper and be very useful for the readers. Again, this can be material for the conclusions.  The research strategy is clear. However, the choice of indicators is less clear. The authors might explain the pros (and cons) of the indicators that they used (e.g.: page 5; page 7: not clear the explanation in lines 256-260. The choice of the thresholds should be motivated, for instance when looking at the co-location hypothesis. 

Author Response

Thanks for your comments to help us improve the paper. As the reviewer pointed out some problems, we revised as follows:

The introduction makes some mistakes by combining the descriptive and the normative (policy) meanings of polycentricity. The authors could stress the two different meanings (e.g. page 2, lines 42-43, it is not clear why there is a consequential relationship between polycentricity as phenomenon and polycentricity as a planning tool. 

The mistakes the reviewer pointed out is very important, the relationship between polycentricity as phenomenon and polycentricity as a planning tool is not clear. The phenomenon of polycentricity mostly researched by empirical studies, which have tried to determine the positive effects of polycentrism, but empirical evidence remains elusive. The policy maker believed that the polycentricity as a planning tool can integrate space economic development, promote cohesion and competitiveness, such as the EU Cohesion Policy since 1999 when the notion was included in the European Spatial Development Perspective(ESDP). But several studies have concluded that it is difficult to identify empirical evidence to support the positive claims made in its name. Polycentrism is assumed toolbox to reduce regional disparities, however, it only become an effective policy tool if local and regional governance actors implement it (Gualini, 2008), but polycentric countries display higher regional disparities than monocentric countries(Daniel,2017). So the lines 42-43 revised as:” Polycentric city has become a common phenomenon in post-industrial megacities; in fact, the policy makers regarded polycentricity as an important planning tool to integrate space economic development, enhance urban competitiveness, social cohesion and environmental sustainability, such as the EU Cohesion Policy since 1999. But the polycentricity as phenomenon and as a planning tool are two different meanings, it is not clear why there is a consequential relationship between the two, several studies have concluded that it is difficult to identify empirical evidence to support the positive claims made in policy makers asked(Daniel,2017).”

It is not really clear why the case study should offer an analysis of Chinese cities, as stated in introduction (page 3, lines 121-122). This is linked with the conclusions, which should provide some policy implications. Otherwise, the article looks more like an exercise, rather than a substantive contribution to the research about urban spatial structure. In other words, the authors should stress more what we can learn from this case study, both for other Chinese cities (if, as stated, this was a goal of the analysis) and in general about megacities overall the world.

This is a good suggestion for the paper, the background of case selection and the purpose of research are very important to the discussion of this paper. For why offer a case study of Chinese cities, we believe that there are many differences between Chinese cities and western cities, such as population density, built environment, social system, etc. Therefore, some western research conclusions on the relationship between urban spatial structure and jobs-housing relationships are not necessarily suitable for China. Hangzhou is an important central city in the Yangtze river delta, after more than 30 years of reform and opening up, it has moved from the "West Lake Era" to the “Qiantang River Era”, and the polycentric spatial structure along and across the Qiantang river has taken shape. In order to enhance the substantive contribution to the research on urban spatial structure, we first revised the introduction, the 120-121 revised as:” In the past 40 years’ reform and opening up, Hangzhou had moved from the "West Lake Era" to the “Qiantang River Era”, and the urban structure have begun to develop from monocentric to polycentric. This transform makes Hangzhou a good case study, with rapid urban growth and changes in jobs-housing relationships, this study analyzes the effectiveness of the polycentric spatial structure by commuting connections and focuses on whether polycentric employment can achieve better jobs-housing matching compared to monocentric development in Chinses big cities, which can provide a comparison between eastern and western studies from the perspective of built environment, economic structure, institutional and other forces.”

       Second, in the conclusion, we added some comparison analysis between Hangzhou and western research results.

The authors apply am "American-based" framework in a different context. Chinese cities are different from US cities, but also institutions and land regulation is different. The authors can explain the main differences. Actually, a discussion about the regulation of urban land (starting from the 80s reforms) in China might enrich the paper and be very useful for the readers. Again, this can be material for the conclusions. 

This is a good suggestion. In the conclusions, we added some in the conclusions, that is “This study found that there are many differences between Chinese cities and Western cities, especially American cities, in terms of polycentric development and jobs-housing matching characteristics. This differences may be caused by the institutional systems between China and Western. The urban development of the western cities is dominated by the free market economy, the co-location hypothesis that people can make ‘rational’ choices of workplace and place of residence  according to market rules. Although since 1980s China has experienced housing market reforms, its government-led “up-down” planning behavior still has a greater impact on urban development. For example, urban internal renewal policies have forced some inner city residents to be relocated the suburbs, and the suburban new industrial space policy has formed some isolated industrial zones, which affected the relationship between jobs-housing relationships. Therefore, the Hangzhou case study uses Western urban research method to reflect the current spatial structure of China's cities and provide strategic guidance for the future development of sustainable urban spatial structure for China.”.

The research strategy is clear. However, the choice of indicators is less clear. The authors might explain the pros (and cons) of the indicators that they used (e.g.: page 5; page 7: not clear the explanation in lines 256-260. The choice of the thresholds should be motivated, for instance when looking at the co-location hypothesis. 

Frist, the reviewer probably refers to the jobs-housing matching index in page 5. According to the co-location hypothesis, a better jobs-housing balance can be achieved by the mutual local adjustment between jobs and housing in the process of suburbanization (Gordon et al., 1991). But there is no good index to measure this jobs-housing match, the self-containment indexes (Cervero,1989), commuting distance and commuting time cannot specifically measure how balanced the resident workers have access to job within ‘reasonable’ travel distance or time, the index of the paper give an intuitive measure results. Then about the thresholds, there is no ‘standard balance’, for example, Cervero(1989) used ceiling of 1.5,Peng(1997) used the range of 1.2-2.8. This paper used average value of all the centers, just take the average value as the dividing line of jobs-housing matching. So we revised the line 208-209 as: ” According to the co-location hypothesis, a better jobs-housing balance can be achieved by the mutual local adjustment between jobs and housing [17],but most indicators empirical studies used cannot specifically measure how balanced the resident workers have access to job within ‘reasonable’ travel distance or time, such as self-containment indexes[16], commuting distance and commuting time. ”

Second, the LQ indicator is not clear explanation. LQ is an important indicator of industrial agglomeration, it gives an approach to examine extent to which the employment centers are specialized in various industries, and to analyze which sectors have the greatest propensity to agglomerate in employment centers. Changes in urban structure are associated with qualitative and/or quantitative changes in agglomeration economic functions (Anas et al., 1998). The decentralization of high-order service activities, offices, industries etc. away from the CBD lead to the change of metropolitan structure, and there are two aspects of the change, that is scatteration and polycentricity, the former is the process that employment is dispersed generally across the metropolitan area, the role of agglomeration economies is in decline, whereas the latter including one or more professional economic nodes other than the CBD. So line 254-260 revised as:” LQ is an important indicator of industrial agglomeration, if the LQ of a certain industry in an employment center is greater than 1, the center is considered to be specialized in the industry, it gives an approach to analyze which sectors have the greatest propensity to agglomerate in employment centers. Anas et al.(1998) believed that urban structure change associated with qualitative and/or quantitative changes in agglomeration economic functions, the decentralization of high-order service activities, offices, industries etc. away from central city lead to the change of metropolitan structure, and there are two aspects of the change, that is ‘scatteration’ and ‘polycentricity’, the former is the process that employment is dispersed generally across the metropolitan area which the role of agglomeration economies is in decline, whereas the latter including one or more professional economic nodes other than the CBD[35].”

Reviewer 4 Report

The article studies the relationship between the polycentric spatial structure and jobs-housing matching in Hangzhou. The main interest of the paper is that it does the analysis based on mobile data positioning of almost 4 of each 5 working residents in the city. The potential of such data somehow rises expectations on what further analysis could have been done in addition to this what is presented. I am thinking that it would have been interesting to include patterns of mobility to/from work in terms of routes but also hours. Social class differences in job match and mobility patterns, etc. Also the sustainability issues behind the different matches and means of transportation. In a way, the paper is quite modest, technically well constructed but not too ambitious as authors recognise when they acknowledge that "housing cost, wage level, and family and individual socio-economic attributes are also important factors that affect the workers’ choice of living place in employment areas. However, such factors are not considered in this study, so the research model needs to be further improved in the future." 

Some more concrete (minor) issues are:

line 104 "shorter" than what?

line 122 "better" compare to what?

line 400-1 "The more homogeneous the industry is, the higher the level of specialization" can you elaborate on this?

Also the use of employment center is confusing and should be better explained or even substitute the term.

line 452 "are the worst; contrast, workers" better "are the worst, in contrast, workers"

line 453 "and; residents" better "and residents"

Author Response

Thanks for your comments to help us improve the paper. As you mentioned that the patterns of mobility to/from work in terms of hours are more interesting, but for the characteristics of cellphone signaling data, due to the positioning time and spatial resolution, we can’t accurately calculate the commuting time. The cellphone signaling data has its limitations, what can be located is the position of the cellphone user’s base station. There are maybe 1 or 2 hundreds meters of error between the user's real location and the base station. Anyway, we also can’t accurately know the actual commuting route of the user. In our paper, the commuting distance calculated by the cellphone signaling data is the Euclidean distance between the base stations, so the actual commuting time cannot be accurately calculated.

The reviewer also mentioned that the social class differences and different means of transportation are both sustainability issues behind the jobs-housing matching, as the cellphone signaling data can’t get the social-economic attributes of the users, in the future we have to rely on other ways to fill the gaps in the data, such as sample survey to enrich our research model.

Then the reviewer pointed out some concrete issues,

Line 104 “shorter” than what revised as:” Zhao et al. [14] found that in China’s Beijing, workers who lived in planned suburban sub-center tend to commute shorter distances to central urban area than other sprawling development suburban areas.”. Line 122 “better” compare to what revised as:” better jobs-housing matching compared to monocentric development in Chinses big cities.” line 400-1 "The more homogeneous the industry is, the higher the level of specialization" can you elaborate on this?

It may be a bit inaccurate to use the term “industrial homogeneous”, homogeneous refers to the convergence of products and services, although there are differences in form, but the content, quality, technical content, use value is the same. But the industrial specialization in the paper refers to the gradual separation of enterprises and sectors within the industry to form independent enterprises and new departments, which is also the process of similar products from decentralized production to centralized production. So we changed the term “industrial homogeneous” to “employment opportunity heterogeneity”, and the sentence could be revised as “Industrial specialization has a negative impact on the worker’s jobs-housing balance, the more specialized the employment center is, the more difficult it is for workers to choose a suitable place to live near the workplace, which indicates that the heterogeneity of employment opportunities is more important in explaining the  workers’ housing selection decision.”

Also the use of employment center is confusing and should be better explained or even substitute the term.

There are many definition of employment centers. Giuliano et al. (2007) believed that contemporary metropolitan areas are characterized by employment clustered in ‘centers’, he used the term ‘employment center’ to denote a site of significant geographic concentration of economic activity, including the CBD. Other scholars have different terms, for example McMillen (2004) used the term “employment subcenter”, he believed that a polycentric city as a metropolitan area with a strong central business district and large subcenters. And Cervero(1989) used the notion of ‘suburban employment centers’, Giuliano and Small(1991) used ‘subcenters’, Garreau(1991) used ‘edge cities’, Forestall and Greene(1997) used ‘employment concentrations’, Coffey and Shearmur (2002) used ‘employment poles’. We use the same concept as Giuliano et al.(2007).

line 452 "are the worst; contrast, workers" better "are the worst, in contrast, workers"

we revised according to the reviewer’s opinion.

line 453 "and; residents" better "and residents"

we revised according to the reviewer’s opinion.