Can the Development of Digital Inclusive Finance Curb Carbon Emissions?: A Spatial Panel Analysis for China Under the PVAR Approach

Comments 1: [more explanation is needed to explain the link between DIF and improved ecological behavior. A vague reference to an ant forest is made, but this is uncite and, to me, not a strong example of the benefits of DIF. More citations or data demonstrating that households, SMEs, or other market players change their behaviors (e.g., purchase more energy-efficient appliances, reduce physical travel, or adopt greener processes) when they have better access to digital financial services.]

Response 1: [Thank you for pointing this out. Therefore, we have modified the part 3 to explain the link between DIF and improved ecological behavior and citations are added. See lines 154-167.]

Comments 2: [information on what DIF is (e.g., a definition highlighting its key attributes) would improve the readability of the paper.]

Response 2: [Thank for your suggestion. We have, accordingly, added a definition of DIF and its citation in section 1 Introduction (lines 41-43).] 

Comments 3: [The paper is based on the “dual mechanism” (alleviating financing constraints and reducing information asymmetry) of DIF. The text would benefit from more explicit cause-and-effect statements for the dual mechanism. That is, state exactly how each mechanism cuts emissions.]

Response 3: [Thank for this comment. We have, accordingly, modified the part 3 to explain how alleviating financing constraints and reducing information asymmetry of DIF cuts emissions. See lines 167-174.] 

Comments 4: [Finally, the paper needs a copy edit. There are several typos, punctuation errors, and some random and unneeded capitalizations throughout the manuscript.]

Response 4: [Thank for suggestion. We have, accordingly, revised several typos, punctuation errors, and some unneeded capitalizations throughout the manuscript]

Comments 1: [The introduction needs to be improved. There are several facts that are being reported but they are not connected to citations. How can we know which are the sources that are being used to construct the argument?]

Response 1: [Thank for your suggestion. Therefore, we have improved the introduction, the related citations are added and added the definition of DIF (lines 34-37, 41-43 in the revised version).]

Comments 2: [the theoretical section is quite challenging to understand. What is the theory and what are the main authors that have discussed this theory? There is not a single citation in the first paragraph (lines 153-166). The entire section does not seem as a theoretical background for the model to be estimated later.]

Response 2: [Thank for this comment. We have, accordingly, modified the part 3 to explain how alleviating financing constraints and reducing information asymmetry of DIF cuts emissions, related citations are added (lines 154-174 in the revised version).] 

Comments 3: [Equations 1 and 2 are incorrect. In a panel VAR with 2 endogenous variables, we have that the first dependent variable is a function of lagged variables 1 and 2, and the same is true for the second dependent variable. The way these equations were presented suggests that a univariate regression has been estimated. This issue has to be corrected and the authors need to show that they are estimating a Vector Autoregression and not univariate regressions.]

Response 3: [Thank for this comment. In the manuscript, our objective is to investigate the impact of digital inclusive finance and technological innovation level on both the total and intensity of carbon dioxide emissions. In the process of model construction, we aim to highlight the impact of two key variables( digital inclusive finance and technological innovation level ) on the two dependent variables ( total and intensity of carbon dioxide emissions), which might have affected the clarity of the model’s formula expression. In response to your feedback, we have adjusted the PVAR model,that is specifically expanding Σⁿ _j=1α_jy_1,t-j in equation (1) and Σⁿ _j=1α_jy_2,t-j in equation (2) to more clearly express the research intent, thereby more clearly articulating our research objectives. This refinement does not affect the progression of our subsequent research.

The following offers additional explanation of the reasoning and methods of using the PVAR model in this study:

In this paper we use the PVAR model to investigate the impact of digital inclusive finance and technological innovation level on total and intensity of carbon dioxide emissions. We have constructed the PVAR model using seven variables. In formulating our equations, we adopt the model construction approach outlined by Chen et al. (2022) [1], who provided a comprehensive derivation and construction process of the PVAR model. Based on their method, We processed the lag coefficient matrix, introduced a single fixed-effect variable and time dummy variables, and established the final form of our PVAR model. Additionally, we refer to the research( Charfeddine, 2019) [2], which employs the PVAR model to examine the effects of renewable energy and financial development on carbon dioxide (CO₂) emissions and economic growth. Based on these references, we develop a PVAR model tailored to our study, focusing on the influence of digital inclusive finance and technological innovation level on total and intensity of carbon dioxide emissions. To enhance the explanatory power of our model, we also introduce independent variables such as urbanization level, environmental regulation system, and economic development level.

The adjusted equation(1) and (2) are presented on lines 220-221 in the revised version.

References（References 1-2 correspond to references 33-34 in the manuscript on lines 651-655 in the revised manuscript）

1. Chen, H. Y.; Yi, J. Z.; Chen, A. B.; Zhou, G. X. Application of PVAR model in the study of influencing factors of carbon emissions. Mathematical Biosciences and Engineering 2022, 19 (12), 13227-13251. https://doi.org/10.3934/mbe.2022619.

2. Charfeddine, L.; Kahia, M. Impact of renewable energy consumption and financial development on CO2 emissions and economic growth in the MENA region: A panel vector autoregressive (PVAR) analysis. Renew. Energy 2019, 139, 198-213. https://doi.org/https://doi.org/10.1016/j.renene.2019.01.010.] 

Comments 4: [In table 1 we find "Error! Reference source not found."]

Response 4: [Thank for this comment. In table 1 the reference is [35] (lines 244), which can be seen in line 565-658 in the revised version.

35. Yang, G. Y.; Tang, L. Exploring the Connotations of New-quality Productivity and its Relationships with Others in the Perspective of Digital Inclusive Finance. Journal of Yunnan Minzu University(Philosophy and Social Sciences Edition) 2024, 41 (05), 84-94. https://doi.org/10.13727/j.cnki.53-1191/c.20240828.003.] 

Comments 5: [The unit root results are not clear. If the variables are stationary after first-differencing why run the test for second-difference? But if non-stationarity poses a problem and the regressions need to be run using variables in differences why the model in (1) and (2) is in levels? How do you run the model?]

Response 5: [Thank for your comment. In 5.1.1. Panel Unit Root Test, this study employs first-order differencing and second-order differencing methods when studying total carbon dioxide emissions and carbon dioxide intensity, respectively. During the stationarity tests, the ADF test achieved stationarity with first-order differenced data, whereas the IPS and LLC tests required second-order differencing to attain stationarity. In the revised version, supplementary expressions have been added to more clearly reflect the results of the unit root tests (lines 252-254 in the revised manuscript).] 

Comments 6: [In section 6.1. the paper states that it is using the PVAR method. But there are no "autoregressive" terms in equations (3) and (4). Therefore, it cannot be a panel "autoregressive" model. It is something else. This needs to be corrected.]

Response 6: [Thank for your comment. In section 6.1., this study conducts a baseline regression model after the PVAR model analysis to examine the impact of digital inclusive finance and the level of technological innovation on total carbon emissions and carbon emission intensity. Equations (3) and (4) are the baseline regression models. The construction of these models serves as a further analysis of the PVAR model’s research findings, hence there are no autoregressive terms. The primary purpose of these models is to quantify results and determine the magnitude and direction of the impact. So, we have made corresponding modifications and improvements to the manuscript to more clearly present the baseline regression models（Lines 375-378 in the revised version).]

Comments 7: [Table 9 should have a note explaining the regression that was implemented. Where are the statistics of the regression? What type of standard errors are used? The same can be said for Table 10.]

Response 7: [Thank for your suggestion. In sections 6.2 and 6.3, the T-statistics are presented in parentheses in the benchmark regression result tables 9, 10, 11, and 12, which provides a straightforward way to assess the significance of the regression coefficients for different independent variables (Liu, S.H., 2024) [1]. However, the manuscript did not include this note below tables 9, 10, 11, and 12. We have now added the necessary annotations below these tables (lines 396, 408, 447, 466 in the revised manuscript).]

Reference

1. Liu, S. H.; Li, J. M.; Xiao, Y. Research on the Spatial Effect of Digital Inclusive Finance on Urban Carbon Emissions in China. Journal of Green Science and Technology 2024, 26 (11), 239-246+260. https://doi.org/10.16663/j.cnki.lskj.2024.11.028.

Comments 8: [There seems to be a typo as Table 41 should be Table 11.]

Response 8: [Thank for your suggestion. The typo has been corrected (line 446 in the revised manuscript).]

Comments 1: [Theoretical and Empirical Implications: Theoretical and empirical implications should be clearly articulated based on the research findings.]

Response 1: [Thank you for this useful suggestion. Therefore, we have added a short paragraph in the section 7 to articulate the theoretical and empirical implications based on the research findings(lines 482-487 in the revised version)] 

Comments 2: [Comparison with Prior Research:The literature review effectively engages with prior research; however, the practical findings could be enhanced by explicitly comparing them with previous research results in the findings or conclusion sections. This comparison should discuss whether the results align with or contradict earlier studies and explore the implications arising from these comparisons. Such analysis would improve the overall clarity and significance of the study's contributions.]

Response 2: [Thank for your useful suggestion. Therefore, in the conclusion section, we have, accordingly, compared our findings with those of previous studies, analyzed and explained the reasons for the differences. See lines 500-507, 521 for details in the revised version).]

Comments 3: [Future Research Suggestions: While the authors recommend future research based on the study's limitations, they could also propose additional avenues for future research based on their specific findings.]

Response 3: [Thank for your useful suggestion. Therefore, in the last paragraph of section 7, based on the research findings that digital inclusive finance exhibits heterogeneous characteristics in its impact on carbon emissions, we add a point suggesting future research directions: the analysis of spatial effects, either within a single country or across different countries, could be a potential necessity and direction for further research (lines556-559 in the revised version).] 

Comments 1: [It is an article with adequate theoretical and applied support, however, it is recommended to improve methodological coherence (the objectives of the research are not specified), as well as to review the wording of the document (for example, "Error! Reference source not found" on page 6).]

Response 1: [Thank for your commets and useful suggestion. In this paper, the research objectives are  presented in the introduction and the literature review sections. According to your suggestions, we have made modifications to ensure that the expressions are completely consistent (lines 57-65, and 134-142).

We have checked the file and found no such error in the DOC file, but in the PDF, Table 1 contains the issue, which is “Error! Reference source not found.” We have re-edited this part. If you still find this error in the PDF version, you can request a DOC revised version file from the editorial office that are uploaded by us.

We have also, accordingly, revised several typos and some unneeded space throughout the manuscript.

Thank you very much for helpful comments and suggestions again. ] 

Dear reviewer,

Thank you very much for your thoughtful and constructive comments and suggestion in round 2 and round 1, which are very helpful and valuable to us.

The concerns you raised in the round 2 review, we had also considered before, therefore, we have integrated the comments and suggestions into the latest revision of the manuscript, which will not only enhance the overall quality of the paper, but also make it more accessible and understandable for other readers by making it clear for them to understand the content and methods of this paper.

The specific point-by-point revisions are as follows： 

Comments 1: [It still not clear how are they running the model - with variables in levels or first-differenced.]

Response 1: [Thank you for this useful comment. In this study, we employ first-order difference values as the variables for our model execution. Consequently, we add the first-order difference symbol ∆ front of each variable in the PVAR model (model 1-2) and the benchmark regression model (model 3-4), to explicitly indicate that the model operations are founded on first-order difference variables (lines 222-223, 378-382 in the revised version).] 

Comments 2: [The discussion about results in Table 2 is far from clear.]

Response 2: [Thank for your useful suggestion. Therefore, to make the results of the first-order difference processing clearer in Table 2, we add a first-order difference symbol ∆ in front of each variable to clarify the data is stationary after the first-order difference processing. At the same time, we also apply identical treatment to Table 3. See lines 260-263 in the revised version).]

Comments 3: [Show the ADF results for the levels and first-differencing. If you cannot reject ADF using levels but you reject for 1st-difference then the model is first-difference stationary.]

Response 3: [Thank for your useful suggestion. Due to the non-stationarity of the original data, this paper employs first-order difference processing during the ADF test. To clearly indicate this, we add the first-order difference symbol ∆ before each variable in Table 2. After processing, all variables have passed the ADF test, demonstrating that the data is stationary after the first-order differencing (lines260-261 in the revised version).] 

Comments 4: [Then all equations should have a first-difference or delta before each variable so it is clear that you are running all regressions using first-differenced variables.]

Response 4: [Thank for your useful and right comment. To clearly indicate that this paper uses first-order difference variables in baseline regression analysis, we added the first-order difference symbol ∆ before each variable in equation 3-4 during the revision. This modification is intended to clarify that the regression analysis in this paper is based on the first-order difference form of the variables (lines378-379 in the revised version).] 

Comments 5: [There still a "error" in the table.

There are many typos.]

Response 5: [Thank for your useful suggestion. In table 1 the reference is [35] (lines 246-247), which can be seen in line 658-660 in the revised DOC and PDF version.

Yang, G. Y.; Tang, L. Exploring the Connotations of New-quality Productivity and its Relationships with Others in the Perspective of Digital Inclusive Finance. Journal of Yunnan Minzu University(Philosophy and Social Sciences Edition) 2024, 41 (05), 84-94. https://doi.org/10.13727/j.cnki.53-1191/c.20240828.003.

We have also, accordingly, revised several typos, punctuation errors, space and some unneeded capitalizations throughout the manuscript.] 

Comments 1: [If the paper uses differenced variables, this has to be clear throughout the paper.

In some cases, the paper uses variables in levels and in others in first difference. But if variables in levels are nonstationary, then all regressions should use first-differenced variables.]

Response 1: [Thank you for this comment. In the round 2 of point-by-point response(1-4), the issue of differential variables has been clearly and detailedly addressed. I am willing to provide further clarification here. In the panel unit root test of the empirical results analysis in this paper, it has been explained that due to the non-stationarity of the original data, the data in this paper were processed by first-order differencing. At the same time, in the third part of the PVAR model construction and the fifth part of the further analysis of the regression model construction formula, it can be further demonstrated that this paper used first-order differential variables. The model (1-4)  can be seen in line238-239 and 408-409 in the paper.

Here is the relevant responses in the round 2:

 Comments 1: [It still not clear how are they running the model - with variables in levels or first-differenced.]

Response 1: [Thank you for this useful comment. In this study, we employ first-order difference values as the variables for our model execution. Consequently, we add the first-order difference symbol ∆ front of each variable in the PVAR model (model 1-2) and the benchmark regression model (model 3-4), to explicitly indicate that the model operations are founded on first-order difference variables (lines 222-223, 378-382 in the revised version).] 

 Comments 2: [The discussion about results in Table 2 is far from clear.]

Response 2: [Thank for your useful suggestion. Therefore, to make the results of the first-order difference processing clearer in Table 2, we add a first-order difference symbol ∆ in front of each variable to clarify the data is stationary after the first-order difference processing. At the same time, we also apply identical treatment to Table 3. See lines 260-263 in the revised version).]

 Comments 3: [Show the ADF results for the levels and first-differencing. If you cannot reject ADF using levels but you reject for 1st-difference then the model is first-difference stationary.]

Response 3: [Thank for your useful suggestion. Due to the non-stationarity of the original data, this paper employs first-order difference processing during the ADF test. To clearly indicate this, we add the first-order difference symbol ∆ before each variable in Table 2. After processing, all variables have passed the ADF test, demonstrating that the data is stationary after the first-order differencing (lines260-261 in the revised version).]

 Comments 4: [Then all equations should have a first-difference or delta before each variable so it is clear that you are running all regressions using first-differenced variables.]

Response 4: [Thank for your useful and right comment. To clearly indicate that this paper uses first-order difference variables in baseline regression analysis, we added the first-order difference symbol ∆ before each variable in equation 3-4 during the revision. This modification is intended to clarify that the regression analysis in this paper is based on the first-order difference form of the variables (lines378-379 in the revised version).]

 Thank reviewer very much for helpful comments and suggestions again.]