Can Supplier Concentration Improve Corporate Risk Taking? Moderating Effects of Digital Transformation
Round 1
Reviewer 1 Report (Previous Reviewer 1)
REVIEW OF SUBMISSION TO SUSTAINABILITY –– 1853171 – “CAN SUPPLIER CONCENTRATION IMPROVE CORPORATE RISK-TAKING: MODERATING EFFECTS OF DIGITAL TECHNOLOGY”
Summary of the paper
This paper investigates the association between the level of corporate risk-taking and supplier concentration, using data from the People’s Republic of China. Two hypotheses are postulated. The first hypothesis conjectures that corporate risk-taking is associated with supplier concentration. The statement of hypothesis permits the direction of association to be either positive or negative. The second hypothesis articulates that the degree of (positive) association between corporate risk-taking and supplier concentration is higher for digitised companies.
These hypotheses are tested via Ordinary Least Squares regressions, estimated on a pooled basis. The investigation period is 2007-2020. The principal measure of corporate risk-taking is the temporal standard deviation of industry-adjusted return on assets, calculated over a three-year rolling average. The independent variable of interest is supplier concentration, measured as the percentage of supplies procured from the top five suppliers. For tests of the second hypothesis, digitisation is captured via a binary variable. Digitised companies were identified from scrutiny of their annual reports, using Python software. To test the second hypothesis, the authors estimated the regression model on a stratified basis, separately for digitised versus non-digitised companies. The final sample comprises 17,140 company-year observations, from the manufacturing industry. The regressions models control for the effect, on corporate risk-taking, of financial leverage, firm size, separation of Chief Executive Officer and board chair, board size, board independence, executive remuneration and firm profitability (measured by Return on Assets). The models include industry and year fixed effects.
The results support the first hypothesis. The coefficients of the corporate risk measures are uniformly positive and significant (p<0.01). These results are robust to use of an alternative measure of supplier concentration (Herfindahl-Hirschman index) and use of the two-step Heckman procedure, to control for potential bias in sample selection. (The authors argue that their sample selection procedures are bias towards inclusion of companies that chose to disclosure details of their supplier relationships.) The results also support the second hypothesis. In the models estimated using the stratum of digitised companies, the coefficients of the corporate risk metrics are uniformly positive and significant (p<0.01). The counterpart coefficients, from the models estimated using the stratum of non-digitised observations, are not significant. The authors also report some results, presenting evidence regarding the mechanisms driving support for the second hypothesis.
Critical review
The authors lucidly argue the business motivation. (i.e., the level of corporate risk-taking affects the probability of corporate failure.) They should augment the discussion with argumentation of the academic motivation. I suggest they acknowledge that understanding determinants of corporate risk-taking is necessary to estimate models of investment efficiency (García-Lara, García-Osma and Penalva, 2016).
Suitable themes are identified in the literature review (i.e., external and internal factors associated with the degree of corporate risk-taking and economic consequences of supplier concentration). However, the authors should embellish the discussion to cover the underlying economic mechanisms. There is some discussion of this nature, regarding the second theme. A starting suggestion would be to separate the current introduction into two sections: an introduction and a literature review. The revised introduction should focus on the contribution of the current paper. The literature review should focus on locating the current study within the extant literature.
The approach to developing the first hypothesis is suitable. (i.e., I commend the authors for identifying theories from the industrial organisation literature, presenting opposing predictions, regarding the direction of association between corporate risk taking and supplier concentration.) However, the authors should continue their approach by arguing which mechanism they regard as stronger, in the specific institutional setting of their study (manufacturing companies from the People's Republic of China, during the period 2007-2020). Some discussion of this nature is provided in the sections about sample selection and data collection. This discussion suggests that a positive association is more likely to prevail, in the context of the study.
The argumentation preceding the second hypothesis is also sound. However, the authors should frame this argumentation within a theory. My suggestion is to use agency theory (Jensen and Meckling, 1976). The relationship between managers and suppliers may be regarded as an agency relationship, with suppliers as principals. By reducing information asymmetry, digitisation constitutes an agency mechanism and reduces the extent to which the suppliers are compelled to price-protect. Another problem, with the development of the second hypothesis, is use of the word “improve”. The authors should refer to “increasing”, rather than “improving” corporate risk-taking. The issue of whether increased managerial risk-taking enhances shareholder wealth is complex and beyond the scope of the paper. A third problem is that the second hypothesis is conditional upon the first one. The second hypothesis assumes that the association between corporate risk-taking and supplier concentration is positive. Hence, it follows that the first hypothesis should also be directional, conjecturing a positive association.
The methodology is sound and appropriate for the purpose of the paper. However, two matters warrant clarification, regarding Equation (1). It seems that return on assets was adjusted on an industry-year, rather than an industry basis. The subscripts of EBIT and total assets, after the summation sign, should be “k”, rather than “i”. (i.e., the summation is over the other k companies in the same industry-year as the treatment company.)
I have some concerns about the control variables. Most importantly, there is no discussion of the economic rationales for selection of the controls. Return on Assets (and components thereof) are on both sides of the regression equation. This is circular and would artificially inflate the R2. I suggest the authors delete Return on Assets, as a control.
Rationales for the sample selection are lucidly argued. I have one suggestion for improvement. Special Treatment companies are a unique institutional feature of the People’s Republic of China (Jiang and Jones, 2018). The authors should embellish the positive impact of this feature, on their research design. (e.g., the Special Treatment classification has enabled the authors to filter distressed companies. Earnings of distressed companies are likely to be distorted. The going concern assumption may not be satisfied (Balcaen and Ooghe, 2006). Furthermore, management may have engaged in opportunistic real and accruals-based earnings management, to avert corporate failure (García-Lara, García-Osma and Neophytou, 2009).)
I have some minor presentational concerns about the descriptive statistics reported in Table 1. Panel (b) is superfluous. The information in this panel is contained in the correlation matrix, reported in Table 3. The authors should reverse the polarity of the t-statistic reported in Panel (b), for conventional presentation.
I have some concerns about the sensitivity analyses, for tests of the first hypothesis, reported in Table 4. Further clarification is warranted, regarding the two-step Heckman procedure, applied to produce the results in Panel (b). The basis for the choice of instrument (industry-year) seems to be that the costs of disclosing information about supplier concentration may vary by industry and year. For example, in industries with high product homogeneity, disclosure by one player may be revealing about demand across the entire industry (Ali, Klasa and Yeung, 2014).)
The analyses using Propensity Score Matching, reported in Panel (c) of Table 4 are very poorly explained. Propensity Score Matching is designed to compare a company in a group with a matched company not in the group but otherwise as similar to its treatment match as possible. What is the basis for being in the treatment group, for the analyses reported in Panel (c)? (e.g., is the purpose to compare digitally transformed companies with non-digitally transformed companies? If so, this does not relate to the first hypothesis.) Were the 200 samples selected from the population, or are they sub-samples of the sample used in this study? If they were selected from the population, surely the authors would be including, in these sensitivity analyses samples, some companies that were excluded from the principal analyses, for sound reasons. What is the size of the samples? My suggestion is to delete Panel (c) of Table 4, completely.
The results presented in Table 6 detract from the paper. Identification of mechanisms, explaining the evidence in Table 5 seems to be an afterthought. I can suggest two possible courses to address this concern. The authors could delete Table 6 and the accompanying discussion. Alternatively, they could embellish the hypothesis development, to include a third hypothesis, about possible mechanisms. I regard the first course as more suitable. The paper makes a standalone contribution, with only two hypotheses. A third hypothesis would be conditional on the second hypothesis, which is conditional upon the first hypothesis.
Recommendation
The paper reflects a high quality. The concerns broached in my report could be readily addressed. Hence, I recommend that the authors be invited to re-submit the paper to Sustainability, after making major changes.
References in this report, not in the paper
Ali A., S. Klasa and E. Yeung E., 2014, “Industry Concentration and Corporate Disclosure Policy”, Journal of Accounting and Economics 58 (2–3), 240–264.
Balcaen, S. and H. Ooghe, 2006, “Thirty-five Years of Studies on Business Failure: An Overview of the Classical Statistical Methodologies and their Related Problems”, British Accounting Review 38 (1), 63-93.
García-Lara J, B. García-Osma and E. Neophytou, 2009, “Earnings Quality in Ex-Post Failed Firms”, Accounting and Business Research 39 (2):119–138
García-Lara, J., García-Osma, B. and F. Penalva, 2016, “Accounting Conservatism and Firm Investment Efficiency”, Journal of Accounting and Economics 61(1), 221-238.
Jensen, M. and W. Meckling, 1976, “Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure”, Journal of Financial Economics 3 (4), 305-360.
Jiang, Y. and S. Jones, “Corporate Distress Prediction in China – A Machine-Learning Approach”, Accounting and Finance 58 (4), 1,063-1,110.
Author Response
Please refer to the attachment: revised paper and modification description
Author Response File: 
Author Response.docx
Reviewer 2 Report (New Reviewer)
The document is very interesting; however its readability can be improved. Throughout the document, there is inconsistency in references, for example on page 1 "...investment decisions 1" and "Hui (2009) 2". On the same page 1 "...investor-friendly 34", is reference 3 and 4 being called, or reference 34? The authors must correct all this in the document. References 13, 19 and 46 do not exist in the body of the document.
I consider that it is important to put the reference regarding the statistical tests (perhaps books) as well as the software that was used to do the statistical analysis.
Author Response
please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 3 Report (New Reviewer)
Overall, the paper is very well written and the findings are interesting and robust. Notwithstanding, there are some aspects that can be improved, namely:
1. the background (introduction) and literature review are sometimes confusing or not fluid, which does not help authors to properly stress the novelty and contribution of the paper. Then, I would suggest authors improve the organization of the literature over these two points;
2. I would avoid an alternative hypothesis to test the same aspect (H1a and H1b)
3. a final revision would avoid some minor mistakes, such as "To test our first hypothesis on relationship between supplier concentration and corporate risk-taking (Hypothesis 1a or Hypothesis 1b), and the moderating effect of digital transformation in hypothesis 2." (it does not make sense as a phrase); "Additionally, the following are excluded from the analysis: firstly, samples with abnormal financial data and missing data were excluded. Secondly, ST and ST* companies with poor performance were excluded." (a redundancy can be avoided);
4. citations must be reviewed, as it does not fit the journal requirement;
5. I would suggest improving the justification for the control variables;
6. Also, the relevance of the findings would be stressed if better discussed, also using the literature review as must as possible;
7. I am not totally convinced that the proxy used for the digital transformation is a good one since it can be biased by the companies disclosure (is it reliably enough?)... a few papers were used to provide support for it, but I wonder why not financial indicators (intangible assets changes, for instance) is not a better alternative. I would propose to better discuss this aspect, as it is highlighted as the paper contribution.
8. it is not clear how this paper fits the scope of this journal (sustainability). If authors consider that is the case, this aspect needs to be improved and clarified.
9. I am not sure if the stepwise method is the best one to provide the findings based on hypotheses that can be even barely supported by the literature.
10. I would say that the appendix usually appears after references.
Finally, congratulations on the work developed! Good luck!
Author Response
Please check the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report (Previous Reviewer 1)
Comments for author File: Comments.pdf
Author Response
please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report (New Reviewer)
My recommendations were answered.Thank you very much
Reviewer 3 Report (New Reviewer)
There are still some adjustments that should be made, namely:
1. The text still should be reviewed, as some excerpts do not make sense or need to be improved, as in the examples below:
"Such as Brynjolfsson et al. (2011)[30] found that the actual effect of data-driven productivity improvement is about 5% higher than the application of ICT technology [31], which improves the total factor productivity[32]. Hu (2020)[33] found that digital technology accelerates innovation in firms' products and services [4] and enhances innovation capacity [31]."
"Where, Adj_Roa denotes the return on total assets adjusted by industry and annual averages;"
On page 7: "According to the" and "According to the" in two consecutive paragraphs.
2. I think that a reduced level of improvement was made to justify (support) the moderator and independent variables from a theoretical perspective.
3. As long as I can see, references are still not accordingly to the journal requirements.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
REVIEW OF SUBMISSION TO SUSTAINABILITY –– 1853171 – “CAN SUPPLIER CONCENTRATION IMPROVE CORPORATE RISK-TAKING: MODERATING EFFECTS OF DIGITAL TECHNOLOGY”
Summary of the paper
This paper investigates the association between the level of corporate risk-taking and supplier concentration, using data from the People’s Republic of China. Two hypotheses are postulated. The first hypothesis conjectures that corporate risk-taking is associated with supplier concentration. The statement of hypothesis permits the direction of association to be either positive or negative. The second hypothesis articulates that the degree of (positive) association between corporate risk-taking and supplier concentration is higher for digitised companies.
These hypotheses are tested via Ordinary Least Squares regressions, estimated on a pooled basis. The investigation period is 2007-2020. The principal measure of corporate risk-taking is the temporal standard deviation of industry-adjusted return on assets, calculated over a three-year rolling average. The independent variable of interest is supplier concentration, measured as the percentage of supplies procured from the top five suppliers. For tests of the second hypothesis, digitisation is captured via a binary variable. Digitised companies were identified from scrutiny of their annual reports, using Python software. To test the second hypothesis, the authors estimated the regression model on a stratified basis, separately for digitised versus non-digitised companies. The final sample comprises 17,140 company-year observations, from the manufacturing industry. The regressions models control for the effect, on corporate risk-taking, of financial leverage, firm size, separation of Chief Executive Officer and board chair, board size, board independence, executive remuneration and firm profitability (measured by Return on Assets). The models include industry and year fixed effects.
The results support the first hypothesis. The coefficients of the corporate risk measures are uniformly positive and significant (p<0.01). These results are robust to use of an alternative measure of supplier concentration (Herfindahl-Hirschman index) and use of the two-step Heckman procedure, to control for potential bias in sample selection. (The authors argue that their sample selection procedures are bias towards inclusion of companies that chose to disclosure details of their supplier relationships.) The results also support the second hypothesis. In the models estimated using the stratum of digitised companies, the coefficients of the corporate risk metrics are uniformly positive and significant (p<0.01). The counterpart coefficients, from the models estimated using the stratum of non-digitised observations, are not significant. The authors also report some results, presenting evidence regarding the mechanisms driving support for the second hypothesis.
Critical review
The authors lucidly argue the business motivation. (i.e., the level of corporate risk-taking affects the probability of corporate failure.) They should augment the discussion with argumentation of the academic motivation. I suggest they acknowledge that understanding determinants of corporate risk-taking is necessary to estimate models of investment efficiency (Biddle, Hilary and Verdi, 2009; García-Lara, García-Osma and Penalva, 2016).
Suitable themes are identified in the literature review (i.e., external and internal factors associated with the degree of corporate risk-taking and economic consequences of supplier concentration). However, the authors should embellish the discussion to cover the underlying economic mechanisms. There is some discussion of this nature, regarding the second theme. A starting suggestion would be to separate the current introduction into two sections: an introduction and a literature review. The revised introduction should focus on the contribution of the current paper. The literature review should focus on locating the current study within the extant literature.
The approach to developing the first hypothesis is suitable. (i.e., I commend the authors for identifying theories from the industrial organisation literature, presenting opposing predictions, regarding the direction of association between corporate risk taking and supplier concentration.) However, the authors should continue their approach by arguing which mechanism they regard as stronger, in the specific institutional setting of their study (manufacturing companies from the People's Republic of China, during the period 2007-2020). Some discussion of this nature is provided in the sections about sample selection and data collection. This discussion suggests that a positive association is more likely to prevail, in the context of the study.
The argumentation preceding the second hypothesis is also sound. However, the authors should frame this argumentation within a theory. My suggestion is to use agency theory (Jensen and Meckling, 1976). The relationship between managers and suppliers may be regarded as an agency relationship, with suppliers as principals. By reducing information asymmetry, digitisation constitutes an agency mechanism and reduces the extent to which the suppliers are compelled to price-protect. Another problem, with the development of the second hypothesis, is use of the word “improve”. The authors should refer to “increasing”, rather than “improving” corporate risk-taking. The issue of whether increased managerial risk-taking enhances shareholder wealth is complex and beyond the scope of the paper. A third problem is that the second hypothesis is conditional upon the first one. The second hypothesis assumes that the association between corporate risk-taking and supplier concentration is positive. Hence, it follows that the first hypothesis should also be directional, conjecturing a positive association.
The methodology is sound and appropriate for the purpose of the paper. However, two matters warrant clarification, regarding Equation (1). It seems that return on assets was adjusted on an industry-year, rather than an industry basis. The subscripts of EBIT and total assets, after the summation sign, should be “k”, rather than “i”. (i.e., the summation is over the other k companies in the same industry-year as the treatment company.)
I have some concerns about the control variables. Most importantly, there is no discussion of the economic rationales for selection of the controls. Return on Assets (and components thereof) are on both sides of the regression equation. This is circular and would artificially inflate the R2. I suggest the authors delete Return on Assets, as a control.
Rationales for the sample selection are lucidly argued. I have one suggestion for improvement. Special Treatment companies are a unique institutional feature of the People’s Republic of China (Jiang and Jones, 2018). The authors should embellish the positive impact of this feature, on their research design. (e.g., the Special Treatment classification has enabled the authors to filter distressed companies. Earnings of distressed companies are likely to be distorted. The going concern assumption may not be satisfied (Balcaen and Ooghe, 2006). Furthermore, management may have engaged in opportunistic real and accruals-based earnings management, to avert corporate failure (Rosner 2003; Charitou, Lambertides and Trigeorgis, 2007; García-Lara, García-Osma and Neophytou, 2009).)
I have some minor presentational concerns about the descriptive statistics reported in Table 1. Panel (b) is superfluous. The information in this panel is contained in the correlation matrix, reported in Table 3. The authors should reverse the polarity of the t-statistic reported in Panel (b), for conventional presentation.
I have some concerns about the sensitivity analyses, for tests of the first hypothesis, reported in Table 4. Further clarification is warranted, regarding the two-step Heckman procedure, applied to produce the results in Panel (b). The basis for the choice of instrument (industry-year) seems to be that the costs of disclosing information about supplier concentration may vary by industry and year. For example, in industries with high product homogeneity, disclosure by one player may be revealing about demand across the entire industry (Ali, Klasa and Yeung, 2014).)
The analyses using Propensity Score Matching, reported in Panel (c) of Table 4 are very poorly explained. Propensity Score Matching is designed to compare a company in a group with a matched company not in the group but otherwise as similar to its treatment match as possible. What is the basis for being in the treatment group, for the analyses reported in Panel (c)? (e.g., is the purpose to compare digitally transformed companies with non-digitally transformed companies? If so, this does not relate to the first hypothesis.) Were the 200 samples selected from the population, or are they sub-samples of the sample used in this study? If they were selected from the population, surely the authors would be including, in these sensitivity analyses samples, some companies that were excluded from the principal analyses, for sound reasons. What is the size of the samples? My suggestion is to delete Panel (c) of Table 4, completely.
The results presented in Table 6 detract from the paper. Identification of mechanisms, explaining the evidence in Table 5 seems to be an afterthought. I can suggest two possible courses to address this concern. The authors could delete Table 6 and the accompanying discussion. Alternatively, they could embellish the hypothesis development, to include a third hypothesis, about possible mechanisms. I regard the first course as more suitable. The paper makes a standalone contribution, with only two hypotheses. A third hypothesis would be conditional on the second hypothesis, which is conditional upon the first hypothesis.
Recommendation
The paper reflects a high quality. The concerns broached in my report could be readily addressed. Hence, I recommend that the authors be invited to re-submit the paper to Sustainability, after making major changes.
References in this report, not in the paper
Ali A., S. Klasa and E. Yeung E., 2014, “Industry Concentration and Corporate Disclosure Policy”, Journal of Accounting and Economics 58 (2–3), 240–264.
Balcaen, S. and H. Ooghe, 2006, “Thirty-five Years of Studies on Business Failure: An Overview of the Classical Statistical Methodologies and their Related Problems”, British Accounting Review 38 (1), 63-93.
Biddle, G., G. Hilary and R. Verdi, 2009, “How Does Financial Reporting Quality Relate to Investment Efficiency”, Journal of Accounting and Economics 48 (2-3), 112-131.
Charitou A, N. Lambertides and T. Trigeorgis, 2007, “Earnings Behaviour: The Role of Institutional Ownership”, Abacus 43(3): 271–296.
García-Lara J, B. García-Osma and E. Neophytou, 2009, “Earnings Quality in Ex-Post Failed Firms”, Accounting and Business Research 39 (2):119–138
García-Lara, J., García-Osma, B. and F. Penalva, 2016, “Accounting Conservatism and Firm Investment Efficiency”, Journal of Accounting and Economics 61(1), 221-238.
Jensen, M. and W. Meckling, 1976, “Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure”, Journal of Financial Economics 3 (4), 305-360.
Jiang, Y. and S. Jones, “Corporate Distress Prediction in China – A Machine-Learning Approach”, Accounting and Finance 58 (4), 1,063-1,110.
Rosner, E., 2003, “Earnings Manipulation in Failing Firms”, Contemporary Accounting Research 20 (2): 361–408.
