Agglomeration and Spatial Spillover Effects of Regional Economic Growth in China
Round 1
Reviewer 1 Report
The paper is very interesting, but I feel that it should be revised a little more before being published in a scholarly journal. I put my concerns and suggestions in what below.
The part of introduction should be rewritten and more focused on the aim of the paper. For example you wrote that “Quah [6] identifies a negative correlation between economic growth and GDP per capita convergence among European countries”, but European Union is not Chine. In this case you also should mention the differences between developed and developing countries.
Moreover, I suggest to separate the results and discussion section into two sections: the “results” and “discussion and policy implication”, as it is common in academic writing.
The descriptive statistics of data should be presented and briefly commented as well.
In your study you found that economic agglomeration leads to regional inequality, but greater regional inequality promotes economic. Could you more discuss these findings, the reasons and the policy implication?
Author Response
Response to Reviewer 1 Comments
Point 1: The part of introduction should be rewritten and more focused on the aim of the paper. For example you wrote that “Quah [6] identifies a negative correlation between economic growth and GDP per capita convergence among European countries”, but European Union is not China. In this case you also should mention the differences between developed and developing countries.
Response 1: By following the reviewer’s suggestion, we have rewritten the introduction to focused on the key aim of our paper. The most studies of European Union were removed. In the revised introduction section, we only focus on China’s studies to simplify tedious introduction. Moreover, we also added an explanation for the differences between developed and developing countries. “It's important to note that, the differences between developed and developing countries are likely to be significant and may have been caused by differential pattern of economic development. For example, China implemented a growth pattern of developing country with high depletion, high devotion, high pollution, high consumption, which different from advanced growth pattern of developed countries. However, further investigation may be needed to confirm this difference.”
Point 2: Moreover, I suggest to separate the results and discussion section into two sections: the “results” and “discussion and policy implication”, as it is common in academic writing.
Response 2: By following the reviewer’s suggestion, we have separated the results and discussion section into two sections: the “results” and “discussion and policy implication”. And an extended discussion and policy implication was added in the fifth section.
Point 3: The descriptive statistics of data should be presented and briefly commented as well.
Response 3: By following the reviewer’s suggestion, we added a descriptive statistics table. “Table 1 shows the descriptive statistics for all key variables in the estimation of Eq. (5) and (6), including both measures of regional inequality and market potential.”
Table 1. Descriptive statistics of variables (Source: modified by the authors).
Variable Mean Std.dev Max Min
ln(yr,t/yr,t-1) 0.1122 0.0232 0.1925 0.0012
ln(Ineqr,t/Ineqr,t-1) 0.0548 2.9422 0.0943 -0.0041
ln(MPr,t/MPr,t-1) 0.1457 2.0812 0.5911 -0.1992
ln(Lr,t/Lr,t-1) 0.0029 0.0734 0.0833 -0.0835
ln(Kr,t/Kr,t-1) 0.2947 1.1776 0.5224 -0.7340
ln(Hr,t/Hr,t-1) 0.0072 0.1784 0.1662 -0.1087
ln(dpr) 7.1043 1.2795 9.6249 0.0000
ln(dcr) 5.2464 0.9376 7.9388 0.0000
Point 4: In your study you found that economic agglomeration leads to regional inequality, but greater regional inequality promotes economic. Could you more discuss these findings, the reasons and the policy implication?
Response 4: By following the reviewer’s suggestion, we added more discussions for our findings, the reasons and policy implication in the section of discussion and policy implication.
In our opinion, one of the reasons for this logic framework is economic growth agglomeration of prefectures can help boost the economic efficiency. And economic efficiency also is one of necessary prerequisites for economic growth agglomeration. Then when economic growth agglomeration located in specific and small fraction regions may result in regional inequality. This regional inequality can be considered as a “by-product” of agglomeration development.
We also added many policy implications in the revised paper.
Controlling regional inequality has become a global problem. For example, the goal 10 of Sustainable Development Goals (SDGs) is to reduce inequality within and among countries. To reduce regional inequality, policies should be universal in principle, paying attention to the needs of disadvantaged and marginalized prefectures. Finally, innovations in technology can help reduce the cost of transferring money for the least developed prefectures. Therefore, offering preferential policies of technology application to these prefectures are convenient ways to dwindle regional disparity.
From the perspective of regional policy, the existence of significant spatial spillover implied that decision-makings of local governments affect not only their own prefectures but also neighboring ones, thereby requiring the central government to pay special attention to coordination across local administrative units, such as prefectures. This also revealed to us the need to gradually eliminate barriers of local protectionism in the course of economic growth to improve market accessibility, strengthen regional economic links and interactions, and improve hardware such as interregional road and other regional infrastructures. It is also necessary to improve the regional soft environment to improve the efficiency of economic links and provide platforms and spaces for spillovers. Improving prefecture-level business environment is one of important part of regional soft environment. For developed prefectures, it is necessary to accelerate the active industry transfer to undeveloped prefectures. On the one hand, developed prefectures need to enhance higher level of industrial evolution and eliminate backward industries which mismatched the industrial development direction of these prefectures. On the other hand, these transferred industries for undeveloped prefectures are still relatively advanced. The undeveloped prefectures of accepted the industrial transfers need to choose appropriate industries matching its industrial development condition and to implement a series of preferential policies to accelerate industrial transfer.
In policy, with the more intense market competition, more attention should pay to enhance persistently education level of workers by implementing education priority policy, establishing a modern education system, improving education quality, and promoting education equity. And if the central or local governments invest more in education will help to reduce regional inequality.
We tried our best to improve the manuscript and made some changes in the manuscript. We appreciate for Editors/Reviewers, warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions. *********************************************************************
Author Response File: 
Author Response.docx
Reviewer 2 Report
The paper aims at establishing the roles of market potential and economic inequality on China’s regional economic growth. Empirical analysis builds on the well-known “wage equation” of New Economic Geography (NEG) that is modified in several respects. First, the dependent variable “wage” is replaced by “income per capita”. Second, regional market potential is defined only by other regions’ potential. Thirdly, levels of income/market potential are replaced by growth.
Although the authors claim that their empirical models are embedded in NEG, the link is very weak. This relates to the analysis of the trade-off between the market potential and economic equality as well as the replacement of the wage variable by the income variable. To avoid explaining regional growth of GDP per capita by own region’s GDP per capita growth, the market potential of a region is defined by distance weighted GDP of the other regions. This means that the used market potential variable only captures income spillovers from all other regions. This is not the actual relevant market potential of a region. Moreover, it will only be weak proxy of the true regional market potential if it is used, for instance, to avoid endogeneity problems.
The modification of the NEG wage curve made by the authors entails adverse consequences in spatial econometric modelling. This can be recognized by the eq. (7) and (9) of the SAR panel model (spatial lag panel model). The first term each captures the spatial lag of GDP growth rate. As the authors do not explicitly define the elements wr,s, the exact interpretation is not quite clear. The spatial lag relates to an average or to the sum of GDP growth rates in the neighbourhood (contiguity approach) or proximity (distance approach). However, according to its definition by the authors, the second term each of eq. (7) and (9) represents a least as well an approximation of the spatial lag of regional GDP growth rate. Because of its composition of distance-weighted GDP growth rates of all other regions, it is strongly affected by growth of proximate regions. What is the economic reasoning of using two slightly different spatial lags of the same variable?
The spatial econometric analysis itself does not reflect the state-of-arts with respect to the partition of total effects in direct and indirect effects. LeSage has shown that the regression coefficients do not reflect the own- and cross-partial derivatives of the explanatory variables in case of an endogenous spatial lag (see. e.g. LeSage and Pace 2009). This is due to feedback effects inadequately captured by the regression coefficients. The deficiency can be resolved by an impact analysis where the regression coefficients are adjusted.
The assertion that “the research method of the neoclassical economic growth theory does not consider sp1atial effects (lines 116-117) is not correct (see e.g. Koch and Ertur 2006).
The loglikelihood is no valid criterion for model choice as it prefers more extensive models.
Author Response
Response to Reviewer 2 Comments
Point 1: The paper aims at establishing the roles of market potential and economic inequality on China’s regional economic growth. Empirical analysis builds on the well-known “wage equation” of New Economic Geography (NEG) that is modified in several respects.
First, the dependent variable “wage” is replaced by “income per capita”.
Second, regional market potential is defined only by other regions’ potential. Thirdly, levels of income/market potential are replaced by growth.
Response 1: Thanks for your suggestion. Yes, the well-known “wage equation” of New Economic Geography (NEG) that is modified in several respects in our paper.
Frist, county and prefecture-level wage data are missing in China from 1992 to 2013, therefore, the dependent variable “wage” is replaced by “income per capita”.
Second, recently, the new economic geography model, as formalized in Fujita et al. (1999), attempts to understand the geographic distribution of economic activities as a mechanism of circular causation for the agglomeration of firms and workers, as well as supply and demand. Such a mechanism may give rise to a core-periphery pattern wherein production is concentrated in one region (the core), therefore explaining the diverging growth of different regions. The potential market capacity for the products and services of a region could consequently be measured by summing up the GDPs of all surrounding regions using the spatial distance as weights (Fujita & Thisse, 2002). Therefore, using other regions’ potential market capacity (by summing up the GDPs of all surrounding regions using the spatial distance as weights) to measure regional market potential has a theoretical basis. Currently, two indices are constructed to measure market potential (Hering and Poncet, 2009).
First, we follow the method by Harris (1954) and specify market potential as used in our paper. is the GDP of prefecture s in year t, and drs is Euclidean distance between prefectures. So the market potential of region r is calculated as the sum of all other regions' GDP, discounted by their distance from region r. Certainly, the market potential formula by Harris considers only final consumptions of households, as represented by regional GDP, while ignoring the substantial share of industry output that is sold to downstream industries in other prefectures as intermediate inputs.
Therefore, Amiti and Javorcik (2008) specify the second measure of market potential for an industry as:
where outputjst is the output of industry j in region s in year t. Note that the summation does not include industry output and GDP for region i itself. This is because our focus is on the growth spillover effect from other regions. bij represents the fraction of industry i's output sold to industry j as intermediate inputs, and bi is the fraction for the final consumption of households. Both bij and bi are calculated based on the China national input–output (I/O) table. By the structure of the I/O table, we have. The region market potential index is calculated by summing all industry-level market potentials within a region, . We think the second measure of market potential is more appropriate. However, current input–output (I/O) data at the prefecture level are missing in China. Therefore, the second measure of market potential is now not applicable in China at the prefecture level.
Third, yes, levels of income/market potential are replaced by growth in our paper. The reason is that using growth is more likely to get a stationary panel data. And this replace also is a commonly used process for regional spillover effect analysis and empirical specification on the effect of market potential on economic growth. For example, the studies of Bai et al., (2012) and Tian et al., (2010).
References:
Amiti, M., & Javorcik, B. (2008). Trade costs and location of foreign firms in China. Journal of Development Economics, 85, 129–149.
Bai, C. E. , Ma, H. , & Pan, W. . (2012). Spatial spillover and regional economic growth in china. China Economic Review, 23(4), 982-990.
Fujita, M., Krugman, P., & Venables, A. (1999). The spatial economy: Cities, regions, and international trade. Cambridge, MA: MIT Press.
Fujita, M., & Thisse (2002). The economics of agglomeration. Cambridge: Cambridge University Press.
Harris, C. D. (1954). The market as a factor in the localization of industry in the United States. Annals of the Association of American Geographer, 44, 315–348.
Hering, L., & Poncet, S. (2009). The impact of economic geography on wages: Disentangling the channels of influence. China Economic Review, 20, 1–14.
Tian, L. , Wang, H. H. , & Chen, Y. . (2010). Spatial externalities in china regional economic growth. China Economic Review, 21(S20-S31).
Point 2: Although the authors claim that their empirical models are embedded in NEG, the link is very weak. This relates to the analysis of the trade-off between the market potential and economic equality as well as the replacement of the wage variable by the income variable. To avoid explaining regional growth of GDP per capita by own region’s GDP per capita growth, the market potential of a region is defined by distance weighted GDP of the other regions. This means that the used market potential variable only captures income spillovers from all other regions. This is not the actual relevant market potential of a region. Moreover, it will only be weak proxy of the true regional market potential if it is used, for instance, to avoid endogeneity problems.
Response 2: Yes, our empirical models are embedded in NEG. And the link is very weak because we used a modified “wage equation” of New Economic Geography, and the wage equation only is used to deduce the equation of market potential. The market potential for the products and services of a region could consequently be measured by summing up the GDPs of all surrounding regions using the spatial distance as weights (Fujita & Thisse, 2002). In this process of calculation, spatial distance weight is important factor. And the introduction of spatial distance weight is help to avoid endogeneity problems. In our opinion, in China, market potential not only captures income spillovers from all other regions, but also represents a large proportion of the purchasing power and demand. And in general, if a region has high-level GDP per capita it also implied that this region has strong purchasing power and product and service demand. The key point of this paper is to investigate the spatial spillover effects, therefore, own region’s GDP per capita growth is not a key factor.
Point 3: The modification of the NEG wage curve made by the authors entails adverse consequences in spatial econometric modelling. This can be recognized by the eq. (7) and (9) of the SAR panel model (spatial lag panel model). The first term each captures the spatial lag of GDP growth rate. As the authors do not explicitly define the elements wr,s, the exact interpretation is not quite clear. The spatial lag relates to an average or to the sum of GDP growth rates in the neighbourhood (contiguity approach) or proximity (distance approach). However, according to its definition by the authors, the second term each of eq. (7) and (9) represents a least as well an approximation of the spatial lag of regional GDP growth rate. Because of its composition of distance-weighted GDP growth rates of all other regions, it is strongly affected by growth of proximate regions. What is the economic reasoning of using two slightly different spatial lags of the same variable?Response 3: By following the reviewer’s suggestion, we added a explicitly define for the elements wr,s. wr,s is a 330 × 330 row-normalized spatial weight matrix with zero diagonal elements, which defines relations of proximity between two prefectures (r and s). In this paper, we used a binary contiguity spatial weight matrix model to generate spatial weight matrix.
The first term each of eq. (7) and (9) represents an endogenous interaction term. According to Anselin et al. (2006), the spatial lag model is typically considered as the formal specification for the equilibrium outcome of a spatial or social interaction process, in which the value of the dependent variable for one agent is jointly determined with that of the neighboring agents. is consistent with other literatures, we normalize the matrix according to row standardization to interpret the spatial spillover effects as the sum of all neighboring prefectures. In this article, theoretically, one prefecture’s economic growth rate is spatial related with neighboring prefecture’s economic growth rate.
The second term each of eq. (7) and (9) represents growth rate of market potential. Theoretically, this term only is growth rate of market potential weighed by all neighboring prefectures.
The first term each of eq. (7) and (9) was used to capture the change of neighboring region’s GDP per capita, however, the second term was used to capture the change of all neighboring regions’ GDP scale. In our opinion, these two terms have certain differences.
References:
Anselin L, Le Gallo J, Jayet H (2006) Spatial panel econometrics. In Matyas L, Sevestre P. (eds) The econometrics of panel data, fundamentals and recent developments in theory and practice (3rd edition). Kluwer, Dordrecht, 901-969.
Point 4: The spatial econometric analysis itself does not reflect the state-of-arts with respect to the partition of total effects in direct and indirect effects. LeSage has shown that the regression coefficients do not reflect the own- and cross-partial derivatives of the explanatory variables in case of an endogenous spatial lag (see. e.g. LeSage and Pace 2009). This is due to feedback effects inadequately captured by the regression coefficients. The deficiency can be resolved by an impact analysis where the regression coefficients are adjusted.
Response 4: Thanks for your suggestion. Yes, according to LeSage’s assertion, we also computed the total effects, direct effects and indirect effects for spatial econometric models. However, the SEM spatial fixed model and SEM spatial and time-period fixed model are the most appropriate model in this paper. Direct and total spillovers for the pooled OLS and SEM models coincide with the βi coefficients of the corresponding models. For SEM model, this is because the disturbances do not come into play when considering the partial derivative of the dependent variable with respect to changes in the explanatory variables. There are no indirect spillovers in these models. Therefore, the result of indirect effects of SEM was not reported.
Point 5: The assertion that “the research method of the neoclassical economic growth theory does not consider spatial effects (lines 116-117) is not correct (see e.g. Koch and Ertur 2006).
Response 5: Thanks for your suggestion. Yes, this assertion has removed in the revised manuscript.
Point 6: The loglikelihood is no valid criterion for model choice as it prefers more extensive models.
Response 6: By following the reviewer’s suggestion, we rerun the MATLAB routine, and the LM spatial lag, LM spatial error, Robust LM spatial lag and Robust LM spatial error are used as criterion to choose SAR or SEM model.
We tried our best to improve the manuscript and made some changes in the manuscript. We appreciate for Editors/Reviewers, warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Author Response File: Author Response.docx
Reviewer 3 Report
The central idea of the paper is potentially interesting. In my opinion, the first chapter entitled "Introduction" is too long. The description of the main idea of the paper should be extended and placed in a separate chapter. The authors quote European studies and therefore should explain the differences between the economies of European countries and China. The description of the results and conclusions are too short and can be extended. From the editor's point of view, equations should be written in the same font as the basic text (e.g. 156, 168 lines).
Author Response
Response to Reviewer 3 Comments
Point 1: The central idea of the paper is potentially interesting. In my opinion, the first chapter entitled "Introduction" is too long.
Response 1: By following the reviewer’s suggestion, we have rewritten and shortened the introduction because the first chapter entitled "Introduction" is too long. We separate the introduction section into two sections: the “Introduction” and “Literature review”.
Point 2: The description of the main idea of the paper should be extended and placed in a separate chapter.
Response 2: Thanks for your suggestion. However, the main idea of the paper cannot support to a separate chapter. In fact, in the sections, introduction, literature review and materials and methods, we have mentioned the main idea of the paper clearly. These descriptions have contributed to readers to understand the main goal of this paper. And in view of your opinion, we have adjusted the structure of manuscript. A literature review section was considered as a separate chapter.
Point 3: The authors quote European studies and therefore should explain the differences between the economies of European countries and China.
Response 3: By following the reviewer’s suggestion, we have added an explanation in the materials and methods section “It's important to note that, the differences between developed and developing countries are likely to be significant and may have been caused by differential pattern of economic development. For example, China implemented a growth pattern of developing country with high depletion, high devotion, high pollution, high consumption, which different from advanced growth pattern of developed countries. However, further investigation may be required to confirm this difference.”
Point 4: The description of the results and conclusions are too short and can be extended.
Response 4: By following the reviewer’s suggestion, we have extended the results and discussion and policy implication.
Point 5: From the editor's point of view, equations should be written in the same font as the basic text (e.g. 156, 168 lines).
Response 5: By following the reviewer’s suggestion, we have revised the font of equations in lines 156 and 168.
We tried our best to improve the manuscript and made some changes in the manuscript. We appreciate for Editors/Reviewers, warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
I missed the conclusion section.
Author Response
Response to Reviewer 1 Comments
Point 1: I missed the conclusion section.
Response 1: By following the reviewer’s suggestion, we have rewritten the conclusion section.
The main goals of the current study are to explore the economic growth agglomeration versus spatial inequality trade-off, investigate direct and indirect spatial spillover effects as well as the drivers of China’s economic growth at the prefecture level in China from 1992 to 2013.
Our analysis reveals the following main conclusions. Firstly, the results show that economic agglomeration, represented by market potential, has a significant influence on economic growth at the prefecture level in China. Meanwhile, economic agglomeration aggravates regional economic inequality, but economic inequality within a controllable range is found to have a positive influence on economic growth. Thus, there is a trade-off between economic growth and economic agglomeration.
Secondly, there are significantly positive spatial spillover across Chinese prefectures, and the spatial spillover effect between prefectures has played an important role in the economic development of China. We furthermore found the direct spillover effects, such as market potential, have the most significant and strongest positive influence on prefecture-level economic growth. We also found that in addition to the influence of direct spillover effects, economic growth of prefectures is inseparable from the random impacts of surrounding prefectures, and they are also affected by indirect spatial spillover effects.
Thirdly, the econometric result shows that prefecture-level economic growth in China not enter the stage of high-quality development from 1992 to 2013, even it has not yet rid itself of the influence of factors of production. The labor participation rate has a significant positive influence on the economic growth of prefectures. However, the effect of human capital on economic growth was significantly weaker than the labor participation rate. This suggests that China’s prefecture-level economic growth is more dependent on the labor force and the demographic dividend in the period of 1992-2013. Contrary to expectations, the influence of per capita fixed-asset investment on prefecture economic growth was slightly weak. The cost and locational factors to enter foreign markets also have an important influence on regional economic growth.
Finally, important policy implications on economic growth of China at the prefecture level were proposed.
We tried our best to improve the manuscript and made some changes in the manuscript. We appreciate for Editors/Reviewers, warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Author Response File: Author Response.docx
Reviewer 2 Report
As some methodical weaknesses are remedied, the paper has been improved. Because the final interpretation is mainly based on the results of the SEM model, the lack of an impact analysis for the SAR model may be acceptable. The problem of using very similar regressior in the SAR model, the spatial endogenous lag and the market potentail with an exclusion of the own regions, remains. However, this problem does not occur in the preferred SEM model.
Author Response
Response to Reviewer 2 Comments
Point 1: As some methodical weaknesses are remedied, the paper has been improved. Because the final interpretation is mainly based on the results of the SEM model, the lack of an impact analysis for the SAR model may be acceptable. The problem of using very similar regressior in the SAR model, the spatial endogenous lag and the market potentail with an exclusion of the own regions, remains. However, this problem does not occur in the preferred SEM model.
Response 1: Thanks for your suggestion. Yes, in the preferred SEM model this problem does not occur. We also followed your comments to improve the quality of this paper marked by tracking mode of Word.
We tried our best to improve the manuscript and made some changes in the manuscript. We appreciate for Editors/Reviewers, warm work earnestly, and hope that the correction will meet with approval.
Once again, thank you very much for your comments and suggestions.
Author Response File: Author Response.docx
