Spatial Non-Stationarity of Influencing Factors of China’s County Economic Development Base on a Multiscale Geographically Weighted Regression Model

Round 1

Reviewer 1 Report (Previous Reviewer 3)

No further comments. 

Author Response

We are very grateful to the reviewers for the recognition and acceptance of our article, and thank for your previous valuable comments.

Reviewer 2 Report (New Reviewer)

The article was solidly prepared and I don't have more serious objections (except for the definitions). It fits into the broad trend of studies of this type developed since the 1990s. The results are coupled with a discussion, but I guess that's the way it can be. All in all, it seems to require fairly minor revisions. I include them below:

1. Keywords: "Economic development", "Influencing factors" I think are too general.

2. Introduction: discussion of the definition of "spatial non-stationarity" is missing. There is a lot of literature (e.g., https://www.tandfonline.com/doi/pdf/10.1080/02693799608902100), including using the same methods the authors propose to study China (https://academic.oup.com/icesjms/article/67/1/145/595559).

3. The methodological part: there is no information on what software was used. I would have emphasized more clearly that the ENT variable is tourism, which obviously has business significance. I would also clarify how the PUB variable relates to the ENT variable, whether there is any overlap.

4. Figure 7 is key, containing the most important results. What I don't understand is why some red and blue areas are marked with circles, while others are not.

5. In the summary it would be better to elaborate on the abbreviations (in parentheses), this will make it easier for those readers who do not have time to delve into the analytical procedure.

6. In the summary, it would be useful to supplement the information with specific figures from Figure 6.

7. The overall methodology and scope of the data used are in principle unobjectionable. The same is true of the design of the whole, it is logical and clear. Unfortunately, I am unable to assess the correctness of carrying out detailed statistical-spatial methods, this is not my specialty.

8. Graphics (maps, charts) were prepared quite carefully, it will only be necessary to standardize the scales of the maps for printing, reduce/enlarge the charts to make it uniform too.

9. The text needs careful editing, the description is not always well understood and clear.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report (New Reviewer)

Based on multi-source big data such as Tencent User Density and Point of Interest data, this article calculates different influencing factors and uses a multiscale geographically weighted regression model to explore the spatial non-stationary impact on the economic development of Chinese counties. This paper carries out adequate research work and draws some important conclusions, but overall it still has the following shortcomings for the authors’ consideration.

 

1. The contribution of the article is still unclear. Although the authors have outlined the research background and literature review in the introduction section, they fail to identify the shortcomings of previous studies, resulting in the originality and contribution of the article appearing insufficient. I think the authors could further sort out the shortcomings of the previous literature and highlight the contribution of this study.

 

2. The paper's topic is to analyse the spatial non-stationarity of influencing factors of China’s county economic development, but it does not explain in detail. What is Spatial Non-stationarity, and what are the key issues that the analysis of spatial non-stationarity can help the reader to understand? From the existing research, it is not very clear how the analysis of the article relates to spatial non-stationarity. It is recommended that the authors further refine the scientific question of the article.

 

3. Although the authors provide a detailed dissection of economic development patterns and influencing factors through multiscale geographically weighted regressionl, relatively few policy implications are generated in response to the model results. In addition to the features reflected in the model itself, the authors should further strengthen the analysis and discussion of results and their policy implications, thus corroborating the marginal contribution of the article.

 

 

4. The full text and detailed information should be further checked. For example, Figure 4 should be supplemented with information on horizontal and vertical coordinates; otherwise, the reader will not be able to understand the intent.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report (New Reviewer)

The author addressed all the comments. I recommend this article could be accepted in its current form.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This research does a very solid work to study the spatial heterogeneity of influencing factors of China’s county level economic development. Multi-sources data are collected and processed, OLS, GWR, and MGWR models are employed and compared. More specifically, results of MGWR are discussed from different aspects. Both of the comparison of methods and the discussion of the results are scientific sound.

One point that authors may need to consider is that human activity can include business, infrastructure, and land use factors, so it may be better to change “human activity factors” to “social media factors” or some sorts of this kind. (line 134)

There are two sections 3.1 in part 3.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors present an analysis of the influence of some factors on county’s GDP in China including spatial effects. The methodology used is innovative (MGWR), the study is well conducted and the paper is well written and interesting. This is a good piece of research with nice potential to be published in IJGI.

Nevertheless, I have two major concerns with the paper, although they are very related. The first one is about the interpretation of results. Specifically, the arguments given to the opposite signs of some coefficients (ROAD, REALTY, ALAND and CLAND). I would expect positive or non-significant effects of these variables on GDP, but not negative. The arguments in p.14-15 can explain the non-significant effect, but not the negative one. For example, I can agree that “the role of real estate in promoting the economic development of the Yangtze River Delta urban agglomeration is gradually weakening” (L. 441-44), but this is not enough reason to justify that this factor negatively affects economic development. Similar uncomplete reasons are given for the unexpected results for ROAD. The opposite results obtained locally with REALTY and CLAND are interesting. Are those variables highly correlated? If this is the case, a substitution effect among them could explain the opposite signs obtained with one and another variable.

I recommend the authors to extend the arguments above by comparing these results with previous knowledge. In section 2.2 you mention some previous literature that justifies the selection of variables, but I miss a further review indicating the expected effect of every factor in GDP. In the results section, you can discuss your results by comparing them with others obtained in previous studies, extracted from those in sections 2.2 and maybe others more related to your outcome.

Following this recommendation, you can address the second question that in my opinion you need to answer in the paper, that is, how do your findings contribute to the knowledge of factors influencing economic development, in China and/or everywhere? The answer to this question can be added in the Introduction and/or Conclusions.

I have also a minor methodological question. I understand that optimizing SOC is the criterium to select the optimal bandwidth in MGWR, but do you apply any method to avoid local optima? Are there frequent when using MGWR?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

This paper studies the factors of economic development of Chinese counties using a Multiscale Geographically Weighted Regression (MGWR) Model. The results show that the MGWR model outperforms OLS and GWR models with higher predictive accuracy and a lower degree of spatial autocorrelation in the residuals. The authors also find that the impact of different factors differs across counties, confirming spatial non-stationary. 

Overall Assessment:

I feel that the contribution of the paper is not very clear. It is not surprising that the MGWR model generates a higher predictive accuracy than the other models, because it introduces more parameters. Apart from that, more advanced models, such as geographically neural network weighted regression, have been applied, and the research paper has been published in Int. J. Geo-Inf..  

On the other hand, the authors study the factors that influence the county-level GDP. However, I am not very convinced by the current analysis. I think selection of factors and the interpretation of the results need to be improved. 

Selection of Variables

Page 4, Table 1: The definition of the explanatory variables is not clear. What is the unit of these variables? I am not sure if ‘the sum of annual Tencent user density’ really reflects human activity. How is the user density quantified? Is the density at the county or zip code level?

In Table 1, REALTY is defined as ‘kernel density data calculated by the commercial, residential type of POI data’. However, in the main text, it is defined as ‘reality business’. What does ‘reality business’ mean? Virtual reality business or real estate business? If it is about the real estate business, can the number of commercial or residential buildings reflect the activities in the real estate business or even the real estate market performance? How does the kernel density is calculated? How to interpret the kernel density of POIs?

Similarly, I am not sure if the enterprise type of point of interest can reflect the business activities. I guess the point of interest won’t change much, regardless of an economic recession or boom. What is the benefit of using Tencent data, rather than other classical economic variables, such as population density, income per capita, total employment, unemployment rate, etc.?

Is the dependent variable GDP or GDP density or GDP per capita? Why use GDP in 2020? Which time period is covered by the Tencent user density, enterprise type of POI data, and commercial and residential type of POI data? Are these data also in 2020? Will the findings be robust by using the data in other years?

I would suggest providing more summary statistics on the dependent and independent variables. Besides, what is the correlation coefficient between the explanatory variables? I guess larger counties will have more enterprise POIs, real estate POIs, traffic roads, etc. Would the multicollinearity problem be a concern?

MGWR modeling

GWR or MGWR can capture not only the spatial autocorrelation but also the heteroskedasticity in the data. The authors could also test heteroscedasticity in the residuals by different models.

The authors use the bi-square kernel function to calculate the weight matrix. How about other kernel functions, such as the Gaussian function? Will the finding remain robust by using other kernel functions?

The authors compare the in-sample predictive accuracy and find that the MGWR model outperforms the other two models. How about the out-of-sample predictive accuracy? Maybe the authors can use 90% of the counties for the regression, predict the GDP for the remaining 10% sample, and compare the out-of-sample predictive accuracy.

Interpretation of the result

Page 16, line 533. The authors mention that ‘ This study found that most counties are affected positively by TUD in China. Thus, it is of great significance for local government to strengthen the intensity of human activities to promote the economic development, especially promote the population inflow of western China’.

‘Third, the business factors are closely related to regional economic development. The CORP and REALTY are local variables and most regions of China are significantly affected positively by them. Therefore, government should encourage and support people to star up business, increasing the distribution density of corporation in China.’

I do not agree with these interpretations. A positive coefficient in the MGWR does not have any implication for causality. Counties with a higher GDP tend to have more business activities, better infrastructure, more jobs, etc. In other words, based on regression, it is hard to show whether a higher GDP causes more Tencent users and more corporate and real estate POIs, or in the other way round, more Tencent users and more corporate and real estate POIs that cause a higher GDP. So I do not think the current results can imply that if the government increases Tencent users, the GDP will be improved.

Page 16, line 542 ‘In order to promote the sustainable development of regional economy more effectively in China, government should consider multi-dimensional development goals, including natural ecology, population introduction, enterprise layout, infrastructure construction and land use, etc., and formulate economic development plans which are fitting to the local characteristics in future.’

I do not see the current analysis can support this implication since no sustainable development indicators have been included. I can’t see any variables that reflect the natural ecology either. The only two related variables are average rainfall and the average altitude for each county. But I do not think they can reflect the natural ecology. 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx