The Evolution of Industrial Agglomerations and Specialization in the Yangtze River Delta from 1990–2018: An Analysis Based on Firm-Level Big Data

Round 1

Reviewer 1 Report

1) What is the relation of your findings to sustainability? In your paper, sustainabilityisn't mentioned even once. There should be a clear link to sustainability in the introduction and the discussion.

2) Revise the phrasing throughout the text: e.g., line 88: "in Europe" (not: "the Europe"), line 131: "which can be / is represented by the EG index" etc.

3) Do you have a reference for "The theory of geographic (territorial) division of labor especially emphasizes the role of natural resource advantages in industrial agglomeration and regional specialization." (line 74-5)?

4) What follows from the fact that you couldn't identify Shanghai as a financial center (p. 13)? You should discuss this point.

Author Response

Response to Reviewer 1 Comments

Dear, reviewer:

Thank you for your comments concerning our manuscript. These comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made the corrections. We hope it meets with approval.

Point 1.What is the relation of your findings to sustainability? In your paper, sustainability isn't mentioned even once. There should be a clear link to sustainability in the introduction and the discussion.

Response 1: In lines:58-59 and lines:457-460, we highlighted the link between our research to sustainability.

lines:58-59“Understanding the spatial-temporal patterns and characteristics of industrial development also contributes to the sustainable development of local industry and economy.”

lines:457-460 “Knowing these overall evolutionary spatiotemporal patterns and characteristics of industrial development in YRD contribute to peripheral cities to better undertake the industrial transfer from core cities and realize the sustainability of regional industrial transformation and upgrading.”

Point 2: Revise the phrasing throughout the text: e.g., line 88: "in Europe" (not: "the Europe"), line 131: "which can be / is represented by the EG index" etc.

Response 2: We removed the article “the” in line 107 and added “is” before “represented” in line 149.

Point 3: Do you have a reference for "The theory of geographical (territorial) division of labor especially emphasizes the role of natural resource advantages in industrial agglomeration and regional specialization." (line 74-5)?

Response 3: The theory of territorial division of labor created by Soviet geographer Baransky are based on the theory of territorial-industrial complexes, these theories are popular in former or present socialist countries. There are so many related references in Russian or Chinese language, but most of them are pretty old. The following are some related references about the theory of territorial division of labor.

1.Saushkin, Y. G. (1976). The role of geography in the definition and solution of the problems associated with the soviet economy. Geoforum, 7(3), 161-166.

2.Wuyang, Y. , & Jinshe, L. . (1987). Theories of territorial division and locational superiorities. Acta Geographica Sinica(3), 201-210.

3.Pilipenko, I. V. (2005). Clusters and Territorial-Industrial Complexes: Similar Approaches or Different Concepts? First Evidence from Analysis of Development of Russian Regions. Ersa Conference Papers.

We added the third reference to reference list.

Point 4: What follows from the fact that you couldn't identify Shanghai as a financial center (p. 13)? You should discuss this point.

Response 4: We have explained more why Shanghai is not identified as a financial center in lines:368-378.

“It was worth noting that although Shanghai (A1) had very high LQ of finance in each year, it was not a hot spot in 2000, 2010, and 2018. That was because, on the one hand, since 1990s, the administrative division within the jurisdiction of Shanghai has been altered frequently, to 2016, all the towns and sub-districts under the jurisdiction of Shanghai were merged to the big Shanghai. For clear comparations of the spatial patterns of specialization in the YRD in different years, we had to merge all the sub-districts under the jurisdiction of Shanghai into one spatial unit, so the spatial relationships of intra-Shanghai cannot be detected. On the other hand, although Shanghai (A1) had very high specialization of finance from 1990 to 2018, some of its surrounding cities (D2 and D3) had low LQ. Influenced by these neighbors, Shanghai was not a hot spot from the perspective of spatial statistics. But Shanghai is undoubtedly the financial center of the YDR and even the whole country.”     

Reviewer 2 Report

This work used 3,053,024 pieces of firm-level big data, the spatial-temporal evolution and spatial patterns of industrial agglomeration and specialization of 9 major industries in the Yangtze River delta were revealed.

The work presented in general has a correct structure and is well written in scientific language.

Some comments or modifications are required

The lines between 149-161 have to be moved to the end of the Introduction section.

Figure 3 has different blue colors and they are very similar, please other colors to define these categories.

Red numbers in Table 3 have to be detailed

Author Response

Response to Reviewer 2 Comments

Dear, reviewer:

Thank you for your comments concerning our manuscript. These comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made the corrections. We hope it meets with approval.

Point 1:The lines between 149-161 have to be moved to the end of the Introduction section.

Response 1: We moved the last two paragraphs in lines between 149-161 to the introduction section in lines:60-73.

Point 2: Figure 3 has different blue colors and they are very similar, please other colors to define these categories.

Response 2: The Figure 3 are redesigned as follows, we changed the blue color in the first picture to black one.

Point 3: Red numbers in Table 3 have to be detailed

Response 3: We provided detailed explanations about the red numbers in Table 3 in lines: 312-313. “The red numbers are the industries whose spatial patterns are significant at the significant level P≤ 0.05.”

Reviewer 3 Report

Review of “The evolution of industrial agglomerations and specialization in the Yangtze River
Delta from 1990-2018: A spatial-temporal analysis based on firm-level big data” by Hu, et al.
Summary:
This manuscript describes advances in spatially and temporally analyzing industrial
agglomeration and specialization as demonstrated using data from the Yangtze River Delta
region of China. While there have been advances in developing indicators for this type of
analysis, there is not yet an indicator which is individually superior. The authors use a
combination of indicators to be able to account for the spatial dimension of industrial
agglomeration using firm-level data, the modifiable areal unit problem, and the spatio-temporal
evolution of industrial agglomeration and specialization, effectively addressing all aspects of the
industrial agglomeration and specialization analysis. They successfully demonstrate this
approach using data from the Yangtze River Delta between 1990 and 2018. The authors find
that the amount of industrial agglomeration is linked to industrial attributes, and the temporal
trajectories of such agglomeration are dependent on industry type.
Broad Comments:
The work presented within this manuscript seems original, of high quality, and an advancement
to its field. The manuscript is, overall, well written and informative, although grammatical
errors and lengthy text made some portions of the manuscript difficult to read. Broad level
suggestions for the authors are presented below, followed by specific comments.
• Section 2 is quite long and contains information that would be more relevant in the
Introduction (which is comparatively quite short). I recommend you either condense
Section 2 and add it as subsections to the Introduction, or (and perhaps this is a better
approach) take the last two or three paragraphs of Section 2 and add it to the
Introduction (as this information is necessary for the Introduction).
• There were some grammatical errors throughout the text that should be addressed
(e.g., missing articles, misuse of nouns in place of verbs, misuse of commas, etc.). The
text was still readable, but some of the errors (especially in Section 2) were distracting. I
commented on them in detail up until the methods section. I focused more on the
science being presented from the methods section onward, so I may have missed some
errors. Thus, it might be a good idea to have the article proofread for grammatical
soundness.
• The results section needs more clarification on how the temporal aspect of the data is
being handled (please see the specific comments below).

Specific Comments:
Abstract:
1. Page 1, line 19: It would be good to state at least the country where the Yangtze river
delta is located when it is first mentioned in the abstract. This will give geographical
context to those who are not familiar with this area.
Section 1. Introduction:
2. Page 2, lines 50-51: Are there any particular reasons why the studies have favored
macro-level data instead of firm-level data? Are there inherent challenges with
handling firm-level data that are not present with macro-level data? If so, it would be
good to address them here.
3. Page 2, line 51: It would be useful to the reader to give one or two examples of
traditional indicators. As it is, it is not clear what these indicators are or what makes
them traditional. Are they traditional because they are only associated with the
traditional practice of only studying macro-level data? I notice that this is covered in
more detail in Section 2, but it would still be nice to just name some examples (even just
within this sentence: “What’s more, many traditional indicators (e.g., …) measuring
industrial…”) to give readers context.
4. Page 2 (end of Introduction): The summary sentence is clear as to the advantages of
incorporating firm-level data into industrial agglomeration studies, but it would be nice
if the Introduction ended with a sentence describing what contributions your study
makes to this field. For example, are you going to be addressing some of the questions
posed earlier in the Introduction? Are you going to be focusing on traditional indicators
(this question may or may not apply depending on your answer to the previous
comment)? I noticed that this is addressed at the end of Section 2, but it is really more
relevant to address it here at the end of the Introduction. I think reorganizing these
sections as suggested in the first bullet point under the “Broad Comments” above will
help address this.
Section 2. Theoretical and empirical controversies
5. Page 2, line 68: It would be good to give his first and last name for formality and to
provide more context for those not as familiar with the field.
6. Page 2, line 88: There appears to be an extra “the” before “Europe” in the sentence
“…manufacturing in the Europe and found…”.
7. Page 3, line 91: Should the text read “…trend in opposite directions…” instead of
“…trend in opposite direction…”? Or perhaps it could be reworded to “…would trend in
a direction opposite to the decrease…”.
8. Page 3, line 99: change “concentration” to “concentrated” (use verb instead of noun)
9. Page 3, line 99: remove “of” from “Most of empirical studies from China’s…”
10. Page 3, lines 96-121: In this paragraph, several methods for analyzing agglomeration
(e.g., distance-based method, Duranton-Overman index, Gini coefficient) are mentioned
with varying results by geographical location. Is one of these methods more widely
accepted for this type of analysis? Does choice of method vary by region?
11. Page 3, line 110: Need to give the full name for the abbreviation when you use it for the
first time in the manuscript – “…dependence of the Information and Communications
Technology (ICT) industry…”
12. Page 3, line 110: What do you mean by “space” exactly? Location, spatial relationship
with other companies, land availability? It would be worth using a more descriptive
word in place of “space”.
13. Page 3, line 112: The wording of the end of the sentence is not quite correct. Perhaps it
should read, “… and thus tend to be spatially agglomerated…”?
14. Page 3, line 113: The word “concentration” should be changed to “concentrated” and
there should be a “the” before “knowledge” (i.e., “…tend to be geographically
concentrated due to the knowledge spillover…”)
15. Page 3, line 114: There should be a “that” before “come” (i.e., “the empirical studies
that come from the United States…”)
16. Page 3, line 115: The comma should be a period (i.e., “Brülhart (2005) conducted a
comparative study among different industries. The results showed…”)
17. Page 3, line 116: The word “agglomeration” should be “agglomerated” (i.e., “…that
service sectors were more geographically agglomerated than manufacturing and
agriculture…”)
18. Page 3, line 117: There needs to be an “a” before “higher degree” (i.e., “…that
manufacturing has a higher degree…”)
19. Page 3, line 118: The text should either read “…and the agglomeration effect…” or “and
agglomeration effects…”
20. Page 3, line 119: The text should either read “…tend to be geographical in
concentration” or “…tend to be geographically concentrated”
21. Page 3, lines 119-120: I suggest changing “industries related to textile industry” to
“industries related to textiles…”
22. Page 3, line 131: There needs to be an “are” before “represented” and a “the” before
“Ellison-Glaeser index” (i.e., “indicators which are represented by the Ellison-Glaeser
index…”)
23. Page 3, line 133 – 134: Why is it a problem that the Ellison-Glaeser index cannot take
the MAUP into consideration? It would be nice to have some context as to why this is
considered a “defect” of the method.
24. Page 4, line 137: The comma after “space” should be a semi-colon
25. Page 4, line 138: The sentence does not end in a grammatically correct way. Perhaps
the authors could change it to “…into consideration but can also provide a statistical
test.”
26. Page 4, lines 147-148: The wording of this sentence is not quite correct; I think you may
be missing a verb or something.
27. Page 4, line 148: At the end of this paragraph (or in the next one), you should address
what your study is doing to advance the field. Are you looking at a new combination of
indicators, or are you introducing new methods for spatial analysis? Later in the
methods section of the paper, you use the EG index despite its inability to consider the
MAUP. Does using the EG index in combination with the LQ, Moran’s I, and Getis-Ord Gi
mitigate this limitation of EG? If so, it would be good to specifically state this to
emphasize the originality of your work.
28. Page 4, line 152: Should the first question read: “Since 1990, what industries…”?
Section 3. Study area, data Sources, and research methods
29. Page 4, line 165: The comma after “Figure 1” should be a period and “the YRD is in the”
should be changed to “the YRD is on the…”
30. Page 5, line 203: There should be the word “where” before “each point” (i.e., “…Figure
2, where each point represents…”)
31. Page 6, line 215: It would be clearer if you also said that the spatial units are referring to
level 2 units.
32. Page 6, line 217: By “whole region” do you mean all the regions combined (the entire
area) or do you mean for a particular level 1 unit? It would be good to clarify this.
33. Page 6, lines 220-221: What would it mean for an EG < 0? Presumably it would indicate
that there is no industrial agglomeration, but what would that mean practically
speaking? Also, is there an upper limit for EG?
Section 4. Results
34. Page 7, line 244: How are you grouping the data within each interval, and how is time
accounted for in this method? For example, are you averaging yearly values to get a
mean EG index for each decade? If so, did you consider a moving average method?
What is the variation of EG indices like within a given decade? It would be good to at
least provide a little more detail on how the values are derived for each time period.
35. Page 7, Table 1: It is not immediately clear to me what the column HHI is referring to. I
am assuming that G and EG are referring to the Gini coefficient and EG index,
respectively, so does that mean that HHI is referring to the Herfindahl coefficient? It
would be good to clarify this is the table caption or table column header.
36. Page 7, Figure 3: It seems like a lot of industries experienced drastic changes in their EG,
as evidenced by sharp changes in slope of their time-series curves. Was this sort of
apparent temporal correlation considered across industries? This may be out of the
scope of the paper, but it would be interesting to know if something happened in the
economy or otherwise development of this region that would have influenced EG index
values during this time.
37. Page 9, line 282: You need a reference for the Arcgis software here.
38. Page 13, line 365: How was this merging accomplished? Was any weighting involved?
Section 5. Discussions
39. Page 16, line 407: The end of the sentence does not make sentence grammatically. I
suggest rewording it to the following: “…industries will be more spatially diverse.”
40. Page 16, line 408: Remove the second use of the word “model” so that the text reads
“…economic geography model of ``too-region, too-good’’, a reduction…”
41. Page 16, line 415: “Industries which are directly related…”
42. Page 16, lines 419-420: “…and tend to be spatially dispersed.”

Author Response

Response to Reviewer 3 Comments

Dear, reviewer:

Thank you for your valuable comments and many grammatical errors rectify concerning our manuscript. We really appreciate your patience and meticulous work. These comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made the corrections.

Broad Comments: The work presented within this manuscript seems original, of high quality, and an advancement to its field. The manuscript is, overall, well written and informative, although grammatical errors and lengthy text made some portions of the manuscript difficult to read. Broad level suggestions for the authors are presented below, followed by specific comments.

Point 1: Section 2 is quite long and contains information that would be more relevant in the Introduction (which is comparatively quite short). I recommend you either condense Section 2 and add it as subsections to the Introduction, or (and perhaps this is a better approach) take the last two or three paragraphs of Section 2 and add it to the Introduction (as this information is necessary for the Introduction).

Response 1: The last two paragraphs of section 2 are moved to the end of the introduction section to make the structure of the paper more balance. The revision can be seen in lines:60-73.

Point 2:  There were some grammatical errors throughout the text that should be addressed (e.g., missing articles, misuse of nouns in place of verbs, misuse of commas, etc.). The text was still readable, but some of the errors (especially in Section 2) were distracting. I commented on them in detail up until the methods section. I focused more on the science being presented from the methods section onward, so I may have missed some errors. Thus, it might be a good idea to have the article proofread for grammatical soundness.

Response 2: we have tried our best to correct the grammatical errors, the misnomers and the punctuation mistakes.

Point 3:  The results section needs more clarification on how the temporal aspect of the data is being handled (please see the specific comments below).

Response 3: The conception of “spatio-temporal” from the perspective of geography refers to the evolution of something in space over time. The “spatio-temporal” analysis of this paper is carried out from two dimensions of time and space, namely, the evolution of industrial agglomeration and specialization in space over time. The objective of this paper is to explore the evolution and spatial patterns of industrial agglomeration and specialization.  We used the Moran’s I index in different years (1990, 2000, 2010, 2018) to reveal the  evolution of industrial specialization from 1990-2018 and then used the hot spot analysis to detect the spatial patterns of industrial specialization in space over time.: The conception of “spatio-temporal” from the perspective of geography refers to the evolution of something in space over time. The “spatio-temporal” analysis of this paper is carried out from two dimensions of time and space, namely, the evolution of industrial agglomeration and specialization in space over time. The objective of this paper is to explore the evolution and spatial patterns of industrial agglomeration and specialization.  We used the Moran’s I index in different years (1990, 2000, 2010, 2018) to reveal the  evolution of industrial specialization from 1990-2018 and then used the hot spot analysis to detect the spatial patterns of industrial specialization in space over time.

Specific Comments:

Abstract:
Point 1: Page 1, line 19: It would be good to state at least the country where the Yangtze river delta is located when it is first mentioned in the abstract. This will give geographical context to those who are not familiar with this area.

Response 1: we added the country information where the Yangtze river delta is in abstract.

Section 1. Introduction:

Point 2: Page 2, lines 50-51: Are there any particular reasons why the studies have favored macro-level data instead of firm-level data? Are there inherent challenges with handling firm-level data that are not present with macro-level data? If so, it would be good to address them here.

Response 2: The macro-level data are easy to access and analysis, however, they lack detailed locational information. The firm-level data contain more locational information therefore facilitate an in-depth  analysis of spatial patterns, point density, and spatial evolution of economic activities. Much of the existing studies has used macro-level data due to the difficulty of obtaining firm-level data.

we added reasons why the studies on firm-level data are rare, in lines:50-52

“although firm-level data contains more location information than the macro-level statistical data, the firm-level data are always hard to collect and process, it’s the biggest reason why the firm-level data studies are rare.”

Point 3: Page 2, line 51: It would be useful to the reader to give one or two examples of traditional indicators. As it is, it is not clear what these indicators are or what makes them traditional. Are they traditional because they are only associated with the traditional practice of only studying macro-level data? I notice that this is covered in more detail in Section 2, but it would still be nice to just name some examples (even just within this sentence: “What’s more, many traditional indicators (e.g., …) measuring industrial…”) to give readers context.

Response 3: We added some examples of traditional indicators in line:53.

“many traditional indicators( e.g. Location quotient and Gini coefficient) measuring ….”
Point 4: Page 2 (end of Introduction): The summary sentence is clear as to the advantages of incorporating firm-level data into industrial agglomeration studies, but it would be nice if the Introduction ended with a sentence describing what contributions your study makes to this field. For example, are you going to be addressing some of the questions posed earlier in the Introduction? Are you going to be focusing on traditional indicators (this question may or may not apply depending on your answer to the previous comment)? I noticed that this is addressed at the end of Section 2, but it is really more relevant to address it here at the end of the Introduction. I think reorganizing these sections as suggested in the first bullet point under the “Broad Comments” above will help address this.

Response 4: the last two paragraphs of section 2 are moved to the end of the introduction section.

Section 2. Theoretical and empirical controversies

Point 5: Page 2, line 68: It would be good to give his first and last name for formality and to provide more context for those not as familiar with the field.

Response 5:  We added the full name of “Alfred Marshall ” in line: 87.

Point 6: Page 2, line 88: There appears to be an extra “the” before “Europe” in the sentence “…manufacturing in the Europe and found…”.

Response 6:   we deleted the “the” before “Europe” in line: 107.
Point 7: Page 3, line 91: Should the text read “…trend in opposite directions…” instead of “…trend in opposite direction…”? Or perhaps it could be reworded to “…would trend in a direction opposite to the decrease…”.

Response 7:  we replaced “trend in opposite direction” with “trend in opposite directions” in line: 110.

Point 8: Page 3, line 99: change “concentration” to “concentrated” (use verb instead of noun)

Response 8: We replaced “concentration” with “concentrated” in line:117.

Point 9: Page 3, line 99: remove “of” from “Most of empirical studies from China’s…”

Response 9: We remove the “of ” from “Most of empirical studies from China’s…” in line: 117.
Point 10: Page 3, lines 96-121: In this paragraph, several methods for analyzing agglomeration (e.g., distance-based method, Duranton-Overman index, Gini coefficient) are mentioned with varying results by geographical location. Is one of these methods more widely accepted for this type of analysis? Does choice of method vary by region?

Response 10: Overall, distance-based methods (such as Duranton-Overman index, Ripley’s k-functions) are more suitable for analyzing industrial agglomeration at multi-spatial scales.  LQ are local indictors can be used to detect the spatial patterns of industrial agglomeration at a single spatial scale. It is hard to say which method is more popular than others, it is up to your research purposes and the data being used.

Point 11: Page 3, line 110: Need to give the full name for the abbreviation when you use it for the first time in the manuscript – “…dependence of the Information and Communications Technology (ICT) industry…”

Response 11:we give the full name of ICT industry before its abbreviation in lines:128-129.

Point 12: Page 3, line 110: What do you mean by “space” exactly? Location, spatial relationship with other companies, land availability? It would be worth using a more descriptive word in place of “space”.

Response 12: We replaced “space” with “location” in line:129.
Point 13: Page 3, line 112: The wording of the end of the sentence is not quite correct. Perhaps it should read, “… and thus tend to be spatially agglomerated…”?

Response 13: We replaced “spatial agglomeration” with “spatially agglomerated” in line:134.
Point 14: Page 3, line 113: The word “concentration” should be changed to “concentrated” and there should be a “the” before “knowledge” (i.e., “…tend to be geographically concentrated due to the knowledge spillover…”)

Response 14: we replaced “concentration” with “concentrated” and added “the” before “knowledge spillover”.
Point 15: Page 3, line 114: There should be a “that” before “come” (i.e., “the empirical studies that come from the United States…”)

Response 15:  We added “that” before “come”.

Point 16: Page 3, line 115: The comma should be a period (i.e., “Brülhart (2005) conducted a comparative study among different industries. The results showed…”)

Response 16:  We replaced the comma with period.
Point 17: Page 3, line 116: The word “agglomeration” should be “agglomerated” (i.e., “…that service sectors were more geographically agglomerated than manufacturing and agriculture…”)

Response 17:  We replaced the “agglomeration” with “agglomerated” in line:135.

Point 18: Page 3, line 117: There needs to be an “a” before “higher degree” (i.e., “…that manufacturing has a higher degree…”)

Response 18: we added an “a” before “higher degree” in line:136.
Point 19: Page 3, line 118: The text should either read “…and the agglomeration effect…” or “and agglomeration effects…”

Response 19: We replaced “agglomeration effect” with “agglomeration effects” in line: 137.
Point 20: Page 3, line 119: The text should either read “…tend to be geographical in concentration” or “…tend to be geographically concentrated”

Response 20:  we replaced “geographical concentration” with “geographically concentrated” in line:138.
Point 21: Page 3, lines 119-120: I suggest changing “industries related to textile industry” to “industries related to textiles…”

Response 21:  we changed the “industries related to textile industry” to “industries related to textiles” in line:138.
Point 22: Page 3, line 131: There needs to be an “are” before “represented” and a “the” before “Ellison-Glaeser index” (i.e., “indicators which are represented by the Ellison-Glaeser index…”)

Response 22:  We added an “are” before “represented” and a “the” before “Ellison-Glaeser index” in line:149.
Point 23: Page 3, line 133 – 134: Why is it a problem that the Ellison-Glaeser index cannot take the MAUP into consideration? It would be nice to have some context as to why this is considered a “defect” of the method.

Response 23:  we highlighted the defect of the Ellison-Glaeser index in lines:151-152.

“However, the Ellison-Glaeser index applies to measure industrial agglomeration at a single spatial scale but cannot take the modifiable areal unit problem (MAUP) into consideration.”

Point 24: Page 4, line 137: The comma after “space” should be a semi-colon

Response 24:  We replaced the comma after “space”  with  a semi-colon.

Point 25: Page 4, line 138: The sentence does not end in a grammatically correct way. Perhaps the authors could change it to “…into consideration but can also provide a statistical test.”

Response 25:  We replaced the “but can provides statistical test” with “but can also provide a statistical test ” in line: 157.
Point 26: Page 4, lines 147-148: The wording of this sentence is not quite correct; I think you may be missing a verb or something.

Response 26:  We added a verb “are” before “based on” in line:166.
Point 27: Page 4, line 148: At the end of this paragraph (or in the next one), you should address what your study is doing to advance the field. Are you looking at a new combination of indicators, or are you introducing new methods for spatial analysis? Later in the methods section of the paper, you use the EG index despite its inability to consider the MAUP. Does using the EG index in combination with the LQ, Moran’s I, and Getis-Ord Gi mitigate this limitation of EG? If so, it would be good to specifically state this to emphasize the originality of your work.

Response 27: The methodological innovations and contributions of this article reflect in the following two aspects:

Firstly, most of previous studies mainly used statistical data collected at macro-level administrative units; studies which are based on firm-level data are rare. In this paper we used more granular  firm-level location-based data to investigate spatial and temporal patterns of industrial agglomeration and specialization over a relative long-time span (1990-2018) in the Yangtze River Delta.

Secondly, findings from this article contribute to ongoing debates in the existing literature on industrial geography, especially spatial patterns and characteristics of industrial agglomeration across space and transition over time. This paper concludes that whether an industry tends to spatial cluster or disperse is highly related to its industry attributes. Industries that are directly related to production tend to be geographically concentrated, while most industries that serve for production tend to be spatially dispersed. What’s more, scholars’ opinions are divided on whether manufacturing tends to be more geographically concentrated or not. This paper found that manufacturing facilities showed a tendency of transition from spatial agglomeration to spatial dispersion. Another theoretical debate is about  whether manufacturing has a higher degree of spatial agglomeration than service sectors.  Our findings support the point that service sectors are more spatially dispersed than manufacturing. So, the main contribution of this paper is to solve some academic disputes to some extent.

We refined our discussion and build more connection with previous studies in discussion section in lines :410-460.

Point 28: Page 4, line 152: Should the first question read: “Since 1990, what industries…”?

Response 28:  we added “what” before “industries”.
Section 3. Study area, data Sources, and research methods 29. Page 4, line 165: The comma after Point 29: “Figure 1” should be a period and “the YRD is in the” should be changed to “the YRD is on the…”

Response 29:  These errors are solved.
Point 30: Page 5, line 203: There should be the word “where” before “each point” (i.e., “…Figure 2, where each point represents…”)

Response 30:  We have added a “where” before “each point” .
Point 31: Page 6, line 215: It would be clearer if you also said that the spatial units are referring to level 2 units.

Response 31:  We replaced the “spatial units” with “the level 2 spatial units” in line:216.
Point 32: Page 6, line 217: By “whole region” do you mean all the regions combined (the entire area) or do you mean for a particular level 1 unit? It would be good to clarify this.

Response 32:  the “whole region” means “the entire area”, we replaced the “whole region” with “the entire area”, in line:218.
Point 33: Page 6, lines 220-221: What would it mean for an EG < 0? Presumably it would indicate that there is no industrial agglomeration, but what would that mean practically speaking? Also, is there an upper limit for EG?

Response 33:  The EG index can be negative or positive, EG < 0 means there is no industrial agglomeration; EG=0 for a random distribution, Generally, 0<EG<0.02 indicates a low agglomeration, 0.02<EG<0.05 indicates a moderate industrial agglomeration, and 0.05≤EG indicates a high industrial agglomeration. But no existing studies show there is an upper limit for EG.

 We added more explanation about the EG index in lines:220-223.

“The EG index can be negative or positive, EG < 0 means there is no industrial agglomeration; EG=0 for a random distribution. Generally, 0<EG<0.02 indicates a low agglomeration, 0.02<EG<0.05 indicates a moderate industrial agglomeration, and 0.05≤EG indicates a high industrial agglomeration.”

Section 4. Results
Point 34: 34. Page 7, line 244: How are you grouping the data within each interval, and how is time accounted for in this method? For example, are you averaging yearly values to get a mean EG index for each decade? If so, did you consider a moving average method? What is the variation of EG indices like within a given decade? It would be good to at least provide a little more detail on how the values are derived for each time period.

Response 34:  We did not calculate the mean EG index for each decade, we just calculated the EG index in 1990, 2000, 2010 and 2018 and used the EG index in each year to measure industrial agglomeration and use the variance of EG index to show the evolution of industrial agglomeration from 1990-2018. Whereas itis a good idea to use the moving average method to calculate the mean EG index for each decade, the reasons why we did not use it are as follows: Firstly, the time span is 1990-2018, so it is hard to divide them into decade evenly. Secondly, it is a super big task to calculate the EG index in each year, because the EG index consists of sub-indicators including G and HHI.

Point 35: Page 7, Table 1: It is not immediately clear to me what the column HHI is referring to. I am assuming that G and EG are referring to the Gini coefficient and EG index, respectively, so does that mean that HHI is referring to the Herfindahl coefficient? It would be good to clarify this is the table caption or table column header.

Response 35:  We added explanations for HHI and G in table column header in Table 3.

Point 36: Page 7, Figure 3: It seems like a lot of industries experienced drastic changes in their EG, as evidenced by sharp changes in slope of their time-series curves. Was this sort of apparent temporal correlation considered across industries? This may be out of the scope of the paper, but it would be interesting to know if something happened in the economy or otherwise development of this region that would have influenced EG index values during this time.

Response 36:  It’s really a very interesting question, but we cannot give you a perfect explanation, because the purpose of this paper is not to explore the mechanisms driving the evolution of industrial agglomeration.

Point 37: Page 9, line 282: You need a reference for the Arcgis software here.

Response 37:   We added a new reference for the Arcgis software in reference lists.

[ Mitchell, A. (1999). The esri guide to gis analysis: geographic patterns & relationships. Esri Inc.]

Point 38: Page 13, line 365: How was this merging accomplished? Was any weighting involved?

Response 38:  No weighting involved. We just merge the data of all sub-districts of Shanghai into one spatial unit.

Section 5. Discussions

Point 39: Page 16, line 407: The end of the sentence does not make sentence grammatically. I suggest rewording it to the following: “…industries will be more spatially diverse.”

Response 39:  we replaced the sentence “industries will be more spatial dispersion” with “industries will be more spatially diverse” in line:425.
Point 40: Page 16, line 408: Remove the second use of the word “model” so that the text reads “…economic geography model of ``too-region, too-good’’, a reduction…”

Response 40:  we removed the redundancy word “model” in line:426.
Point 41: Page 16, line 415: “Industries which are directly related…”

Response 41:  We added “are” before “directly” in line:433.
Point 42: Page 16, lines 419-420: “…and tend to be spatially dispersed.”

Response 42:  we replaced the “spatial dispersed” with “spatially dispersed” in line:436.

Reviewer 4 Report

The paper is well-presented and easy to follow its content. However, after careful revision, I have significant concerns regarding the novelty of the work and its contribution. It is not clear, what are the gaps in the knowledge that the authors wanted to fill? And consequently, what are the contributions of the paper? If this topic has been widely studied, what is different and new the authors’ approach? I feel like it is just an application of several indicators to the various industry in a specific place, but I do not see if this is enough to deserve publication. In that regard, the authors should clarify what the research questions were and how their methodological approach is adequate and novel that makes a substantial contribution.

The biggest issue that I found in the submitted article is related to its methodology and the confusion in the use of some concepts by the authors. The article title said that is a “spatio-temporal analysis” and, in the lines 135 – 137, the authors mentioned tools from spatio-temporal statistics that have been used in previous research, such as space-time Ripley’s K-function. However, after reading the manuscript, the authors did not implement, at least, in an exploratory way, any spatio-temporal analysis, which means the inclusion of spatio-temporal autocorrelation structures in the analysis. The authors performed a “multitemporal study”, they evaluated the spatial variation of a property in several time steps, but they did not consider the phenomena dynamics. Authors should correct this issue and tell why they did not use, for example, spatio-temporal Moran’s I, what it was expected.

Authors said that ArcGIS 10.4 was used to calculate the Moran’s I, but they did not explain how the matrix W was selected. In the literature, it is possible to find, at least, 10 ways to define such a matrix, which is the critical step in the analysis of areal data. Indeed, from the spatial econometrics, several types of weighting mechanisms, has been developed.

Discussion and conclusions are limited because there were not research questions.

Author Response

Response to Reviewer 4 Comments

Dear, reviewer:

Thank you for your comments concerning our manuscript. These comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made the corrections. We hope it meets with approval.

Point 1: The paper is well-presented and easy to follow its content. However, after careful revision, I have significant concerns regarding the novelty of the work and its contribution. It is not clear, what are the gaps in the knowledge that the authors wanted to fill? And consequently, what are the contributions of the paper? If this topic has been widely studied, what is different and new the authors’ approach? I feel like it is just an application of several indicators to the various industry in a specific place, but I do not see if this is enough to deserve publication. In that regard, the authors should clarify what the research questions were and how their methodological approach is adequate and novel that makes a substantial contribution.

Response 1: The methodological innovations and contributions of this article reflecte in the following two aspects:

Firstly, most of previous studies mainly used statistical data collected at macro-level administrative units; studies which are based on firm-level data are rare. In this paper we used more granular  firm-level location-based data to investigate spatial and temporal patterns of industrial agglomeration and specialization over a relative long-time span (1990-2018) in the Yangtze River Delta.

Secondly, findings from this article contribute to ongoing debates in the existing literature on industrial geography, especially spatial patterns and characteristics of industrial agglomeration across space and transition over time. This paper concludes that whether an industry tends to spatial cluster or disperse is highly related to its industry attributes. Industries that are directly related to production tend to be geographically concentrated, while most industries that serve for production tend to be spatially dispersed. What’s more, scholars’ opinions are divided on whether manufacturing tends to be more geographically concentrated or not. This paper found that manufacturing facilities showed a tendency of transition from spatial agglomeration to spatial dispersion. Another theoretical debate is about  whether manufacturing has a higher degree of spatial agglomeration than service sectors.  Our findings support the point that service sectors are more spatially dispersed than manufacturing. So, the main contribution of this paper is to solve some academic disputes to some extent.

Revision: We refined our discussion and build more connection with previous studies in discussion section in lines :410-419.

Point 2:The biggest issue that I found in the submitted article is related to its methodology and the confusion in the use of some concepts by the authors. The article title said that is a “spatio-temporal analysis” and, in the lines 135 – 137, the authors mentioned tools from spatio-temporal statistics that have been used in previous research, such as space-time Ripley’s K-function. However, after reading the manuscript, the authors did not implement, at least, in an exploratory way, any spatio-temporal analysis, which means the inclusion of spatio-temporal autocorrelation structures in the analysis. The authors performed a “multitemporal study”, they evaluated the spatial variation of a property in several time steps, but they did not consider the phenomena dynamics. Authors should correct this issue and tell why they did not use, for example, spatio-temporal Moran’s I, what it was expected.

Response 2: The conception of “spatio-temporal” from the perspective of geography refers to the evolution of something in space over time. The “spatio-temporal” analysis of this paper is carried out from two dimensions of time and space, namely, the evolution of industrial agglomeration and specialization in space over time. The objective of this paper is to explore the evolution and spatial patterns of industrial agglomeration and specialization.  We used the Moran’s I index in different years (1990, 2000, 2010, 2018) to reveal the  evolution of industrial specialization from 1990-2018 and then used the hot spot analysis to detect the spatial patterns of industrial specialization in space over time. Although, we do not use a specific spatio-temporal model like space-time Ripley’s K-function to analyze the spatial-temporal evolution and spatial patterns of specialization in YRD, the two dimensions of space and time are all included. The space-time Ripley’s K-function function can be used to detect the evolution trend of industrial specialization over time but cannot detect the spatial patterns. Actually, the method used in this paper is quite popular and more geographically oriented. What’s more, we downplay the emphasis on “spatio-temporal”  and remove it from title.

Revision 2.1 : We reviewed and cited the studies that used the similar methods in the revised manuscript to show the rationality of our method in lines:157-160.

“In addition, there are another class of methods that combines the first-generation non-spatial indicators indexes (such as Location Quotient, Gini coefficient, etc.) and modern spatial statistical methods, such as Moran’s I or Getis-Ord Gi* [48-51], as well as nuclear density analysis [39] to study the spatial-temporal evolution of industrial agglomeration and specialization.”

Revision 2.2: We changed the title to “The evolution of industrial agglomerations and specialization in the Yangtze River Delta from 1990-2018: An analysis based on firm-level big data”

Point 3: Authors said that ArcGIS 10.4 was used to calculate the Moran’s I, but they did not explain how the matrix W was selected. In the literature, it is possible to find, at least, 10 ways to define such a matrix, which is the critical step in the analysis of areal data. Indeed, from the spatial econometrics, several types of weighting mechanisms, has been developed.

Response 3: Due to the fact that there are some isolated spatial units (i.e., R1、R2、R3 in Figure 1) in our data set, so spatial contiguity weights, such as queen contiguity or rook contiguity are not applicable. Instead,  we used inverse distance weight  measured by fixed distance band ( Euclidean distance, the default neighborhood search threshold was 56340.2 meters).

Revision: we added the details of creating the adjacency matrix in lines:236-238.

“The matrix was created by fixed distance band method and the distance was calculated using Euclidean distance, the default neighborhood search threshold was 56340.2 meters.”

Point 4: Discussion and conclusions are limited because there were not research questions.

Response 4: We refined our discussion and build more connection with previous studies in discussion section, lines:410-460. 

   “The main contributions of this paper reflect in the following aspects. Firstly, most of previous studies mainly used statistical data collected at macro-level administrative units; studies which are based on firm-level data are rare[3,4]. In this paper we used more granular  firm-level location-based data to investigate spatial and temporal patterns of industrial agglomeration and specialization over a relative long-time span (1990-2018) in the Yangtze River Delta. Secondly, findings from this article contribute to ongoing debates in the existing literature on industrial geography, especially spatial patterns and characteristics of industrial agglomeration across space and transition over time. Currently, scholars’ opinions are divided on what industry tends to geographically concentrated and what industry tends to spatially dispersed? as well as what industries are more specialized than others? In this paper, we found some interesting findings that contribute to solve these disputes to some extent.  

        As early as the 1990s, some scholars noticed the impact of modern transportation and communication technologies on industrial agglomeration. Sassen (1994) and Castells (1996) believed that with the progress of transportation and communication technology, the transportation and transaction costs of enterprises will decrease rapidly, thus, enterprises have more freedom and flexibility in their location choice [22,23], which generates both agglomeration and decentralization forces. Some scholars think that industries will be more geographically concentrated, while others think that industries will be more spatial dispersion. In Krugman’s new economic geography model of “two-region, two-good”, a reduction of transport cost would first foster industrial agglomeration and specialization, but then industrial dispersion emerges in the extremely low transport costs case [51]. In another general equilibrium model of the “two-city system” developed by Takatoshi in 1998, industrial dispersion happens in both higher and lower levels of transportation costs, while agglomeration appears in the intermediate level situation [52]. From a theoretical perspective, this study finds that whether an industry tends to spatial cluster or disperse is highly related to its industry attributes, and different industries have different industrial agglomeration characteristics and evolution trajectory. Industries which are directly related to production are still affected by transportation costs and spillover effects and have a higher level of industrial agglomeration and specialization, while most industries which serve for production have a low industrial agglomeration and specialization and tend to be spatially dispersed. What’s more, scholars’ opinions are divided on whether manufacturing tends to be more geographically concentrated or not, as well as whether manufacturing has higher degree of spatial agglomeration than service sectors. This paper found that manufacturing in YRD showed a tendency of transfer from spatial agglomeration to spatial dispersion, and manufacturing has higher degree of spatial agglomeration than service sectors.  These findings supported previous scholars’ conclusion that “services sectors are more spatially dispersed than manufacturing”[22,33,54]. The finance, scientific research and polytechnic services, as well as information transfer and information technology services have a very low original spatial agglomeration. This finding, at first glance, seems to contradict existing theories that knowledge, technology and capital intensive industries tend to cluster in  locations where these factors are the highest [50]. But it makes sense, when we realize the basic fact that China’s industrialization and industrial upgrading started very late compared with most developed countries. Although these industries have very low initial spatial agglomeration, they are agglomerating to the core cities at an alarming speed.

     From a policy perspective, revealing the spatial-temporal evolution laws of industrial agglomeration and specialization has significant implications for individual cities to integrate themselves into regional production networks and make place-based industrial development policies. We found that since 1990, most industries in the YRD have formed distinct spatial patterns of specialization and spatial division of labor. For example, the manufacturing has transferred from north-central to the north, and most regional core cities have experienced deindustrialization processes, this finding stands in line with recent studies on manufacturing relocation in YRD[56]. High-end industries are clustering to the three biggest core cities of Shanghai (A1), Nanjing (B1) and Hangzhou (C1), while the northern cities of YRD are more specialized in scientific research and polytechnic services. Knowing these overall evolutionary laws of industrial development in YRD contribute to peripheral cities to better undertake the industrial transfer from core cities and realize the sustainability of regional industrial transformation and upgrading.”

Reviewer 5 Report

The manuscript ‘The evolution of industrial agglomerations and specialization in the Yangtze River Delta from 1990-2018: A spatial-temporal analysis based on firm-level big data’ conducted research on the industrial agglomerations and regional specialization by using firm-level big data. Based on this research, the spatial-temporal evolution and spatial patterns of industrial agglomeration and specialization of 9 major industries in the Yangtze River Delta were revealed. Generally, the results are solid and meaningful. I am glad to see this manuscript with a high quality and am expecting this manuscript to be published. In general, I have no major comments but some minor ones. They are listed as follows.

In the manuscript, the authors sometimes use ‘spatial-temporal’ and sometimes ‘spatio-temporal’. This is not perfect strictly. I suggest the authors to check the whole manuscript and to keep this consistent.

In section 3.1 paragraph 2, the authors listed the name and code of the whole cities using texts. Personally, I regard it to be better to replace the texts using a table, as things might be clearer via a table. Also, I suggest the authors to replace the texts of 199-203 using tables.

In section 3.2, is it possible to add the data structure of the experiment data in order to make it easier to understand for readers?

In Figure 2, the authors classified the 9 major industries into 3 categories (as there are 3 figures). Why using 3 figures to show the 9 major industries? Is there any specific reason? If yes, please explain this where necessary.

In section 4.1 line 244, the authors mentioned ‘four intervals, namely, 1990, 2000, 2010 and 2018’. This is not 100% correct strictly, as ‘interval’ usually denotes a period of time, but not a timestamp. Hence, the intervals should be ‘1990-2000’, ‘2000-2010’, ‘2010-2018’. In this case, there are only three intervals, but not four. Therefore, I strongly advise the authors to check this carefully and point out clearly whether it is ‘interval’ or ‘timestamp’, AND keep the results to be consistent with the corresponding unit (i.e., ‘interval’ or ‘timestamp’).

Author Response

Response to Reviewer 5 Comments

Dear, reviewer:

Thank you for your comments concerning our manuscript. These comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made the corrections. We hope it meets with approval.

Point 1. In the manuscript, the authors sometimes use ‘spatial-temporal’ and sometimes ‘spatio-temporal’. This is not perfect strictly. I suggest the authors to check the whole manuscript and to keep this consistent.

 Response 1: We replaced all the “spatio-temporal” with“spatial-temporal”in the whole manuscript.

Point 2.In section 3.1 paragraph 2, the authors listed the name and code of the whole cities using texts. Personally, I regard it to be better to replace the texts using a table, as things might be clearer via a table. Also, I suggest the authors to replace the texts of 199-203 using tables.

 Response 2:  To make the spatial division and spatial relationships clearer, we replaced the original text with a table in line:186. The table is as follows.

Table1：the division of spatial units in YRD

Point 3.In section 3.2, is it possible to add the data structure of the experiment data in order to make it easier to understand for readers?

Response 3:  we add the data structure of the experiment data in table 2, line:202.

The data structures which includes the firm number, firm size, as well as the mean value and standard deviations (SD) of each industry are shown in Table 2.

Table2：the data structures and descriptive statistics of each industry

Point 4.In Figure 2, the authors classified the 9 major industries into 3 categories (as there are 3 figures). Why using 3 figures to show the 9 major industries? Is there any specific reason? If yes, please explain this where necessary.

Response 4: In light of the page limitations of the Sustainability journal, we did not include illustrations for  each industry but put every three industries into one subfigure. The total number of firms in each subfigure is roughly equal.

we explained why using 3 figures to show the 9 major industries in lines:198-201.

“To reach a balance between the length of page and the visual effect, we grouped every three industries into one subfigure, the total number of firms in each subfigure as balanced as possible.”

Point 5. In section 4.1 line 244, the authors mentioned ‘four intervals, namely, 1990, 2000, 2010 and 2018’. This is not 100% correct strictly, as ‘interval’ usually denotes a period of time, but not a timestamp. Hence, the intervals should be ‘1990-2000’, ‘2000-2010’, ‘2010-2018’. In this case, there are only three intervals, but not four. Therefore, I strongly advise the authors to check this carefully and point out clearly whether it is ‘interval’ or ‘timestamp’, AND keep the results to be consistent with the corresponding unit (i.e., ‘interval’ or ‘timestamp’).

Response 5: The “timestamp” is what we want to express, we replaced the ‘intervals’ with ‘timestamps’  in line:284.

Round 2

Reviewer 3 Report

The authors' thorough revisions have significantly improved the manuscript.  All questions/suggestions I had from the first round were clearly addressed, and I have no further questions.

Author Response

Point 1. English language and style are fine/minor spell check required.

Response 1: We have checked our language and style carefully.

Reviewer 4 Report

I appreciate the answers given by the authors regarding my concerns about the submitted article and the clarifications about the contribution of the paper. However, they have not appropriately addressed my comments about the applied methods, which makes not valid many of the presented results. In this sense, my evaluation is to reject for publication the paper. Please find below some general comments of my evaluation:

The authors should be cautious when they are using tools from other fields. Notably, the paper has serious issues from the statistical methods implemented. The research is more than click-on Next, next, etc. in a software. To do that, it requires an in-depth knowledge of the methods behind.  In geography, as well as, in Statistics, the term Spatio-temporal is different from the term multi-temporal. We have to use adequate words to refer to the used concepts. MDPI Sustainability publishes high-quality scientific research. When I referred to spatio-temporal Ripley's K function, I did not mean that the authors must have been used. My comment was that in methodologically that function is more up-to-date than the simple global Moran's I. Then, why did authors use something old? Again, what was the contribution of their method? Assuming that Moran's I in several time steps is valid for the problem in analysis. Authors did not use the method correctly. First, the weighting matrix has to be defined using an analytical method when there is no knowledge in the nature of the phenomenon under study. For instance, a simple review of the next reference would have helped.

Dray, S., Legendre, P., & Peres-Neto, P. R. (2006). Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). ecological modelling, 196(3-4), 483-493.

An inappropriate matrix produces erroneous of the spatial autocorrelation and then in the analysis of the spatial pattern. Thus, just using the matrix that by default, ArcGIS implements are a colossal mistake. Finally, I suggested reviewing some Spatial Econometrics theory because I know that when we are facing a problem with economic variables in the space, there are several mechanisms of defining the matrix W that have shown a good description for economic issues. For instance, the Negative exponential model or the inverse distance. I suggest to the authors to read Luc Anselin and James LeSage for a detailed explanation. Based on all the above, many of the results are not right, and then the findings are basically in doubt.

Author Response

Response to Reviewer 4 Comments

Dear, reviewer:

Thank you for your insightful comments, which inspired us to rethink our research methodology. We have reconsidered your comments both in round 1 and round 2 and have made some corrections. In the first round of the review process, we received a total of five reviewers’ comments. At present, four reviewers provided positive evaluations. We thank you again for evaluating our work and will greatly appreciate it if you reconsider your decision about our manuscript.

Round1,point 1:The paper is well-presented and easy to follow its content. However, after careful revision, I have significant concerns regarding the novelty of the work and its contribution. It is not clear, what are the gaps in the knowledge that the authors wanted to fill? And consequently, what are the contributions of the paper? If this topic has been widely studied, what is different and new the authors’ approach? I feel like it is just an application of several indicators to the various industry in a specific place, but I do not see if this is enough to deserve publication. In that regard, the authors should clarify what the research questions were and how their methodological approach is adequate and novel that makes a substantial contribution.

Round 2,point1: I appreciate the answers given by the authors regarding my concerns about the submitted article and the clarifications about the contribution of the paper. However, they have not appropriately addressed my comments about the applied methods, which makes not valid many of the presented results. In this sense, my evaluation is to reject for publication the paper. Please find below some general comments of my evaluation:

Responds 1: the applied methods

 The methods in this study were chosen to serve our research objective that is to reveal the patterns and characteristics in the spatial evolution of industrial agglomeration and specialization in the Yangtze River Delta from 1990-2018. The research questions in this paper are as follows: 1. Since 1990, what industries in the Yangtze River Delta have become more geographically concentrated or more spatially dispersed? 2. What are the characteristics and evolutionary trajectories of industrial agglomeration for different industries? 3. What are the spatial patterns of specialization for different industries in the Yangtze River Delta?

In this paper, we used the EG index to measure the agglomeration of 9 major industries and then we used the location quotient in combination with spatial statistics (Moran's I and Getis-Ord Gi*) to detect the evolutionary trajectory and the spatial pattern of each industry’s specialization. Indicators measuring regional specialization can be divided into global and local indices. Global index allows for exploring the overall situation of each industry’s specialization over the entire study area (80 spatial units), while local index can capture the variation of an industry’s specialization across each spatial unit. The global Moran's I or Getis-Ord Gi* statistic is one of the most commonly used analytical methods to examine the spatial distribution patterns and is also widely applied to the study of industrial specialization.

The following are some case studies in which global Moran's I or Getis-Ord Gi* statistic is used to examine spatial patterns of industrial specialization.

Carroll, M. C. , Reid, N. , & Smith, B. W. . (2008). Location quotients versus spatial autocorrelation in identifying potential cluster regions. The Annals of Regional Science, 42(2), 449-463.

In this paper both location quotients and Gi∗ are used to identify the potential cluster regions in the transportation equipment industry of four states in the Midwestern USA.

Feser, E. , Sweeney, S. , & Renski, H. . (2005). A descriptive analysis of discrete U.S. industrial complexes. Journal of Regional Science, 45(2), 395-419.

 In this paper, Feser et al used Gi*   to identify industries clustered in different areas of the USA.

Lafourcade, M. , & Mion, G. . (2007). Concentration, agglomeration and the size of plants. Regional Science & Urban Economics, 37(1), 46-68.

In this paper, Lafourcade used Moran’s I and EG index to study whether the geographic distribution of manufacturing activities depends on the size of plants.

Sohn, J. . (2004). Do birds of a feather flock together ?: economic linkage and geographic proximity. Annals of Regional Science, 38(1), 47-73.

In this paper, the Moran’s I was used to identify the intraindustry-intercounty distribution pattern of feather industry.

In fact, Sustainability has published numerous papers in which the Global Moran’s I statistic and the Getis-Ord Gi* statistic are used to detect the spatial patterns of geographic features. Examples of new articles are as follows:

Spatial Distribution Pattern of the Headquarters of Listed Firms in China Sustainability 2018, 10(7), 2564; https://doi.org/10.3390/su10072564 A Network-Constrained Integrated Method for Detecting Spatial Cluster and Risk Location of Traffic Crash: A Case Study from Wuhan, China Sustainability 2015, 7(3), 2662-2677; https://doi.org/10.3390/su7032662 Spatial Hotspot Pattern Analysis of Provincial Environmental Pollution Incidents and Related Regional Sustainable Management in China in the Period 1995–2012 Sustainability 2015, 7(10), 14385-14407; https://doi.org/10.3390/su71014385

So, we have reasons to believe that our research methods are appropriate, and they are effective in measuring the spatial pattern of industry specialization and its evolution. We appreciate that you recommended an article to us. But the methods used in that article are about spatial weighting matrix and are not appropriate for analyzing the spatial patterns of industry specialization.

Revision 1:We reviewed and cited the studies that used similar methods in the revised manuscript to show the rationality of our method in lines:157-165.

“In addition, there are another class of methods that combines the first-generation non-spatial indicators indexes (such as Location Quotient, Gini coefficient, etc.) and modern spatial statistical methods, such as Moran’s I or Getis-Ord Gi*[48-50], as well as nuclear density analysis [39] to study the spatial evolution of industrial agglomeration and specialization. For example, Carroll et al.( 2008) used both location quotients and Gi∗ to explore the potential cluster regions in the transportation equipment industry[51]. Feser et al ( 2010) also used Gi*  to identify industries clustered in the USA[52]. Lafourcade (2007) used Moran’s I and EG index to study whether the geographic distribution of manufacturing activities depends on the size of plants[53]. Similarly, Sohn (2004) used the Moran’s I to identify the intraindustry-intercounty distribution pattern of feather industry[54].”

Responds 2: innovations and contributions

     The innovations and contributions of our article are mainly manifested in two aspects – the use of more granular firm-level data and extending the theoretical literature on spatial analysis of industrial agglomeration and specialization.

Firstly, most of previous studies mainly used statistical data collected at macro-level administrative units; studies which are based on firm-level data are rare. In this paper we used finer firm-level location-based data to investigate spatial and temporal patterns of industrial agglomeration and specialization over a relative long-time span (1990-2018) in the Yangtze River Delta.

Secondly, although research on industrial agglomeration and specialization has lasted for more than 100 years, there are still a lot of controversies. Currently, scholars’ opinions are divided on what industries tend to be geographically concentrated and what industries tend to be spatially dispersed. What industries are more specialized than others? What trajectories of industrial agglomeration and specialization does an industry follow?  Some empirical studies claim that certain industries tend to be spatially clustered due to the agglomeration effects, while others are more likely to be spatially dispersed due to the reduction of transportation and transaction costs. The first conclusion of this paper is that whether an industry tends to be spatially clustered or dispersed is highly related to its industrial attributes. Industries that are directly related to production tend to be geographically concentrated, while most industries that serve for production tend to be spatially dispersed. What’s more, scholars’ opinions are divided on whether manufacturing tends to be more geographically concentrated or not. This paper found that manufacturing showed a tendency of transition from spatial agglomeration to spatial dispersion and most regional core cities have experienced deindustrialization processes. These findings substantiate recent studies on manufacturing relocation in the YRD[58], which show that conclusions drawn by our method is reliable. Another theoretical debate is about whether manufacturing has a higher degree of spatial agglomeration than service sectors. Our findings support the argument that service sectors are more spatially dispersed than their manufacturing counterparts are. So, the main contribution of this paper is to solve some academic disputes to some extent.

We refined our discussion and build more connections with previous studies in the discussion section in lines :414-465.

Round1,point 2: The biggest issue that I found in the submitted article is related to its methodology and the confusion in the use of some concepts by the authors. The article title said that is a “spatio-temporal analysis” and, in lines 135 – 137, the authors mentioned tools from spatio-temporal statistics that have been used in previous research, such as space-time Ripley’s K-function. However, after reading the manuscript, the authors did not implement, at least, in an exploratory way, any spatio-temporal analysis, which means the inclusion of spatio-temporal autocorrelation structures in the analysis. The authors performed a “multitemporal study”, they evaluated the spatial variation of a property in several time steps, but they did not consider the phenomena dynamics. Authors should correct this issue and tell why they did not use, for example, spatio-temporal Moran’s I, what it was expected.

Round2,point 2: The authors should be cautious when they are using tools from other fields. Notably, the paper has serious issues from the statistical methods implemented. The research is more than click-on Next, next, etc. in a software. To do that, it requires an in-depth knowledge of the methods behind.  In geography, as well as, in Statistics, the term Spatio-temporal is different from the term multi-temporal. We have to use adequate words to refer to the used concepts. MDPI Sustainability publishes high-quality scientific research. When I referred to spatio-temporal Ripley's K function, I did not mean that the authors must have been used. My comment was that in methodologically that function is more up-to-date than the simple global Moran's I. Then, why did authors use something old? Again, what was the contribution of their method? Assuming that Moran's I in several time steps is valid for the problem in analysis.

Response 3.1: For the debate on the term “spatio-temporal analysis”, we delete the term “spatio-temporal analysis” in the title and replace all other terms of “spatio-temporal evolution” in the paper with “spatial evolution”.

Round1, point3: Authors said that ArcGIS 10.4 was used to calculate the Moran’s I, but they did not explain how the matrix W was selected. In the literature, it is possible to find, at least, 10 ways to define such a matrix, which is the critical step in the analysis of areal data. Indeed, from the spatial econometrics, several types of weighting mechanisms, has been developed.

Round2,point3: Authors did not use the method correctly. First, the weighting matrix has to be defined using an analytical method when there is no knowledge in the nature of the phenomenon under study. For instance, a simple review of the next reference would have helped.
Dray, S., Legendre, P., & Peres-Neto, P. R. (2006). Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). ecological modelling, 196(3-4), 483-493.An inappropriate matrix produces erroneous of the spatial autocorrelation and then in the analysis of the spatial pattern. Thus, just using the matrix that by default, ArcGIS implements are a colossal mistake. Finally, I suggested reviewing some Spatial Econometrics theory because I know that when we are facing a problem with economic variables in the space, there are several mechanisms of defining the matrix W that have shown a good description for economic issues. For instance, the Negative exponential model or the inverse distance. I suggest to the authors to read Luc Anselin and James LeSage for a detailed explanation. Based on all the above, many of the results are not right, and then the findings are basically in doubt.

Response3:We appreciate your recommended article, and we have had a better understanding of matrix W from reading your recommended article. But the matrix issue is out of the scope of the paper and it’s extremely hard for us to prove which one is better than others.

We choose the fixed distance band method for the following reasons:

Firstly, due to the fact that there are some isolated spatial units (i.e., R1、R2、R3 in Figure 1) in our data set, so spatial contiguity weights, such as queen contiguity or rook contiguity are not applicable. Instead,  we used inverse distance weight  measured by fixed distance band ( Euclidean distance, the default neighborhood search threshold was 56340.2 meters).

Secondly, we want to explore the evolution of spatial patterns of industrial specialization in different years, so we want to compare them on a comparable scale. The official explanation about the fixed distance band method in Arcgis10.4 software is: “….The fixed distance band method is the default option used by the Hot Spot Analysis (Getis-Ord Gi*) tool. It is often a good option for polygon data when there is a large variation in polygon size (very large polygons at the edge of the study area and very small polygons at the center of the study area, for example), and you want to ensure a consistent scale of analysis.” http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-statistics-toolbox/modeling-spatial-relationships.htm#GUID-F063A8F5-9459-42F9-BF41-4E66FBBCC415.We think this method help us to compare the spatial patterns in different years at a consistent scale of analysis.

We added the details of creating the adjacency matrix in lines:241-243; we also highlighted our defects in the limitation section in lines:466-469.

“The matrix was created by fixed distance band method and the distance was calculated using Euclidean distance, the default neighborhood search threshold was 56340.2 meters.”

“This paper does have some limitations. Firstly, the spatial weighting matrix is critical to spatial analysis and there are many ways to create it [62]. The matrix used in this paper was created by the fixed distance band method and the results are also based on it. We did not explore more about the matrix issue, because it is out of the scope of this paper, and it’s extremely hard to prove which method is the best.”