Investors’ Aspirations toward Social Impact: A Portfolio-Based Analysis

Round 1

Reviewer 1 Report

Dear Author(s),

Thank you for giving me the possibility to read and evaluate your work. The paper is interesting to read and addresses a stimulating topic which is fully coherent with the focus of Sustainability: the alignment of investors’ aspirations toward social impact and their investment strategies, in terms of number of investees with high social aspirations.

Impact investing practices are still not completely exploited by extant literature, so any attempt to shed light on the issue is to be welcomed. The paper is well written and attention has been paid to the clarity of expression and readability. I also appreciate your methodological section. However, despite these positives, the paper can be a little further enhanced. In the following lines, there are few issues that I have identified after my revision of the manuscript.

• The introduction lacks a bit of focus. You mention different concepts and try to define all of them: social impact, impact investing, aspirations. So, I suggest you to check if in the section you encompass the following crucial aspects: (1), purpose of the study; (2) relevance and significance of the study; (3) eventually research question(s); (4) how the research question is addressed (including the methods); (5) contribution of the paper; (6) how the paper is structured. Moreover, I would suggest to eliminate the footnotes.
• Theoretical background. You have done a fairly good job with this section. However, I would suggest to modify Hypotheses 2 and 3, as you refer to the mission, that is the tool you used to measure aspirations. Please consider to align these two Hypotheses to your Hypothesis 1 in which you do not refer to mission statements.
• Data and method. I would suggest to insert in the sample sub-section Table 3 on the country origin of investors of the sample. Moreover, it is advisable to explain better how did you evaluate the mission statements of investors and investees. Did you perform a content analysis, or did you simply insert the missions in the different categories? You mention a study of Gamble, Parker and Moroz (2019), but it is not present in the references.
• Discussion and conclusions. It is fine, maybe consider to separate discussion from conclusions and try to be more specific on how your results advance previous literature.

I hope you will find these comments and suggestions helpful in further developing this study and I wish you all the best for the development of your work!

Author Response

Dear Reviewer,

thank you very much for your suggestions. We strongly appreciated your comments and we worked for addressing each point you raised. Here you find below the answers to each question.

We hope everything is clear to you. We are available for additional refinements if needed.

1. Impact investing practices are still not completely exploited by extant literature, so any attempt to shed light on the issue is to be welcomed.
   • We attempted to improve the conceptualization of impact investing in the 2.Theoretical background section, connected with blended value finance context, highlighting it also in the 7.Discussion section. Also in the 8.Conclusion section, we included implications that directly target impact investing practices from a technical literature standpoint – SFDR 2019/2088.

2. The introduction lacks a bit of focus. You mention different concepts and try to define all of them: social impact, impact investing, aspirations. So, I suggest you to check if in the section you encompass the following crucial aspects: (1), purpose of the study; (2) relevance and significance of the study; (3) eventually research question(s); (4) how the research question is addressed (including the methods); (5) contribution of the paper; (6) how the paper is structured. Moreover, I would suggest to eliminate the footnotes.

• We restructured the 1.Introducition section accordingly, cutting unnecessary parts you can see highlighted in track change, and introducing methodologies adopted to run the regression analysis. Footnotes have been eliminated.

1. Theoretical background. You have done a fairly good job with this section. However, I would suggest to modify Hypotheses 2 and 3, as you refer to the mission, that is the tool you used to measure aspirations. Please consider to align these two Hypotheses to your Hypothesis 1 in which you do not refer to mission statements.

• We have reconsidered this section in order to have a more precise identification of the theoretical framework of blended value, and changed some narratives in hp 1 in order to include

1. Data and method. I would suggest to insert in the sample sub-section Table 3 on the country origin of investors of the sample. Moreover, it is advisable to explain better how did you evaluate the mission statements of investors and investees. Did you perform a content analysis, or did you simply insert the missions in the different categories? You mention a study of Gamble, Parker and Moroz (2019), but it is not present in the references.

• We Shifted that table to Data and Method Section. We included Gamble, Parker, Moroz Paper (2019) in the reference list. We mentioned the content analysis in the text, when explain the Dependent variable.

1. Discussion and conclusions. It is fine, maybe consider to separate discussion from conclusions and try to be more specific on how your results advance previous literature.

We separated the sections – 7. Discussions, 8. Conclusions and addressed them more precisely future research implications in section 7

Reviewer 2 Report

Thanks for the change to review the article, “Investors’ aspirations toward social impact: a portfolio-based analysis.”

Although authors made an effort to improve their article, there still remained quite a critical point to be solved for the publication.

Major issues

1. Please provide a theoretical framework for blended value finance. The issues summarized in Table 1 appear to have been arbitrarily written by the author. References for each philanthropy to have a theoretical perspective and the implications of each citation should be summarized in order to be recognized as a contributing point.
2. Please refrain from direct verbal citations in CEO conversations.
3. Please change the delivery through footnote to in-text as much as possible.
4. Please cite a lot of S-level journals and show the discussion appropriately.
5. Even if estimation is possible with GLM, is it possible with a sample of 75 investors? And since 50% of the sample represents the United States, there is a high probability that there will be a bias?
6. Since the VIF for Spi_inv is 9 and the impact is 6, multicollinearity is highly suspected in estimation.
7. Has mean-centering been applied to moderation?
8. Why is the dependent variable impact_score missing in correlation?
9. SPI_investor does not have a variable with a high coefficient in correlation even though the VIF is high. What is the reason? can not understand.
10. To do a chi-squared test, derive a p-value by comparing it with the chi2 value of the basic model, which puts only the control variable.
11. Table 6 is not well understood. Although the VIF is high and some variables are highly correlated, there are many things that are significant. Usually these studies are predicted to have significant outliers or biases. I hope the authors will once again consider these issues. I would also recommend trying the univariate test, but the sample is already small, so I wonder if it would make sense to try the t-test mann whitney.
12. There are some things that the notation of p-value in in-text is slightly inappropriate. We usually consider things less than the 0.05 level as significant, but we don't have to put "=" to show the value.
13. The discussion should be rewritten academically. And I can't understand the implications. What theoretical and managerial contributions are there in this study?
14. Don't put unnecessary italics. Readability is poor.

Author Response

Dear Reviewer,

thank you very much for your suggestions. We strongly appreciated your comments and we worked for addressing each point you raised, from the theoretical framework to the multicollinearity issues. When checking again the data for multicollinearity issues, we noticed we have made a mistake in the variables’ list considered for the correlation matrix and the VIF. To be consistent, we run again the regression analysis, and no mistakes were made in that occasion, so tht we confirm the model at Table 6. We apologize for the inconvenience. Here you find below the answers to each question.

We hope everything is clear to you. We are available for additional refinements if needed.

1. Please provide a theoretical framework for blended value finance. The issues summarized in Table 1 appear to have been arbitrarily written by the author. References for each philanthropy to have a theoretical perspective and the implications of each citation should be summarized in order to be recognized as a contributing point.

• We reframed the section by focusing more on the development of a theoretical framework able to correctly introduce the blended value finance context. We excluded table 1, and we proposed a new graph explaining the new mechanisms

1. Please refrain from direct verbal citations in CEO conversations.

• We excluded the citations from the CEO of Blackrock

1. Please change the delivery through footnote to in-text as much as possible.

• We excluded unnecessary footnotes

1. Please cite a lot of S-level journals and show the discussion appropriately.

• Most of the papers cited are from scientific journals, mostly part of the management and organizational behavior fields.

1. Even if estimation is possible with GLM, is it possible with a sample of 75 investors? And since 50% of the sample represents the United States, there is a high probability that there will be a bias?

• We carefully checked the methodological papers to address your question: Papke and Wooldridge (1996) report the econometric specifications standing behind the adoption of GLM estimation, not mentioning sample limitations for the use of the specific methodology. In general, literature has not agreed yet to a minimum number to conduct regression analysis. Considering additional sources such as (Green, S. B. (1991). How many subjects does it take to do a regression analysis. Multivariate behavioral research, 26(3), 499-510), it confirms that there are different perspectives on the sample dimension granting minimum reliability for regression analysis, but samples greater than 50 observations seem overall highly accepted.

• We agree on the fact that the sample present a relevant portion of observations that is from the US. This is the result of our research design, thoroughly conducted starting from a precise context of investors such as those that voluntarily sign to the GIIN. The GIIN gathers information on the population of investors that self-select within an impact investing network, and we exploited Crunchbase as a source to expand the details of information from our population. Since we are aware of this, we have included in our model a control dummy variable called US_investor that takes values of 1 for investor located in US, while 0 elsewhere

1. Since the VIF for Spi_inv is 9 and the impact is 6, multicollinearity is highly suspected in estimation.

• We agree with you about the suspects of multicollinearity so that we looked at the literature and checked again the data. We considered the following sources to deepen on the thresholds:
  1. James G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning: With Applications in R. 1st ed. 2013, Corr. 7th printing 2017 edition. Springer; 2013.
  2. Menard S. Applied Logistic Regression Analysis. 2nd edition. SAGE Publications, Inc; 2001.
  3. O’brien, R. M. (2007). A caution regarding rules of thumb for variance inflation factors. Quality & quantity, 41(5), 673-690.
• Collinearity is considered to be problematic when higher than 10; and it causes concerns when it is higher than 5. Accordingly, we checked again the data, run again the VIF command on Stata, and we realized that the previous table unintentionally considered an uncoherent variable list, for which we apologize. In the new version of the paper, Table 5 at page 13, you can find the new results of the VIF, which present no value higher than 5, and no levels of tolerance lower than 0.2, so that our model does not present collinearity issues.
• To be consistent, we checked again the regression analysis, and we confirm the results in the model in Table 6.

1. Has mean-centering been applied to moderation?
• No, we did not apply mean-centering for the following reasoning: we considered the following source to articulate the answer to your question: Echambadi, R., & Hess, J. D. (2007). Mean-centering does not alleviate collinearity problems in moderated multiple regression models. Marketing Science, 26(3), 438-445.

• From that methodological paper we summarize the following: Extant literature evidence that mean-centering can reduce the covariance between the linear and the interaction terms, suggesting that it reduces collinearity. Authors of the paper proved that “mean-centering neither changes the computational precision of parameters, the sampling accuracy of main effects, simple effects, interaction effects, nor the R².” In addition, they “also show that the determinants of the cross product matrix are identical for uncentered and mean-centered data”. They evidence that the collinearity issue in moderated regression does not change by mean-centering. Several marketing researchers use to “mean-center their moderated regression data hoping that this will improve the precision of estimates from ill conditioned, collinear data, but unfortunately, this hope is futile”

From thes additional source we completed the answer to your question: Echambadi, R., & Hess, J. D. (2007). Mean-centering does not alleviate collinearity problems in moderated multiple regression models. Marketing Science, 26(3), 438-445; Gatignon, H., & Vosgerau, J. (2005). Moderating effects: The myth of mean centering. Fontainbleau Cedex: INSEAD.

• We summarize: authors evidence that “mean-centering changes lower order regression coefficients but not the highest order coefficients, does not change the fit of regression models, does not impact the power to detect moderating effects, and does not alter the reliability of product terms”

1. Why is the dependent variable impact_score missing in correlation?

• The correlation matrix serves to understand whether the combination of independent variables (predictors) are correctly selected in order to provide information on the effects on the dependent variable. We thus checked to what extent predictors are correlated or not. For these reasons, we did not include impact score. When couples of variables have considerably high levels of correlation, one of them can be excluded in order to avoid issues of multicollinearity. The threshold is 0.7 to consider the exclusion of certain independent variables from the model, as reported in Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., ... & Lautenbach, S. (2013). Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography, 36(1), 27-46.
• At page 13, Table 4 reports the new correlation matrix including the variable impact score. As occurred for the calculation of the VIF at point 6, the previous table included some mistakes in the variable list, so that we changed the coefficients for all the variables. We apologize for this inconvenience.

9. SPI_investor does not have a variable with a high coefficient in correlation even though the VIF is high. What is the reason? can not understand.
   • As explained in point 6, and 8, we corrected some mistakes in the variables list, resulting in new versions of the correlation matrix and VIF, at page 13. We apologize for the inconvenience.

1. To do a chi-squared test, derive a p-value by comparing it with the chi2 value of the basic model, which puts only the control variable.

• We reported the chi-square of the models at the bottom of Table 6.

1. Table 6 is not well understood. Although the VIF is high and some variables are highly correlated, there are many things that are significant. Usually these studies are predicted to have significant outliers or biases. I hope the authors will once again consider these issues. I would also recommend trying the univariate test, but the sample is already small, so I wonder if it would make sense to try the t-test mann whitney.

• Following your suggestions, we included a section 6. Robustness Check, in which we conducted the replication of the model on the sample of just non-US investor in order to understand whether the whole sample is biased for US-trends. The results show that the non-US investor sample replicates the tendencies of the main sample, in each of the three hypotheses tested, confirming that the model does not suffer from US-based biases.
• We conducted the Mann-Whitney test to understand whether the sample of investors located outside US are significantly different from those located outside US. In Table 8 reported in the section 6. Robustness Check, results confirm that the two groups are not statistically significant evidencing that our sample presents homogeneous observations, with the absence of potential US-based biases in our sample.

1. There are some things that the notation of p-value in in-text is slightly inappropriate. We usually consider things less than the 0.05 level as significant, but we don't have to put "=" to show the value.

• We changed the notations of the p-value

1. The discussion should be rewritten academically. And I can't understand the implications. What theoretical and managerial contributions are there in this study?

• We restructured the discussion and conclusion parts, respectively in sections 7 and 8. We improved the understanding of the concepts proposed, future research opportunities, policy and practitioners’ implications.

1. Don't put unnecessary italics. Readability is poor.

• We excluded unnecessary italics

Reviewer 3 Report

The article approaches a very interesting and actual topic and describes in detail the subject. However, please consider the following improvements:

1. please revise all in-text citations and Bibliography section according to the Journal indications: ”In the text, reference numbers should be placed in square brackets [ ], and placed before the punctuation; for example [1], [1–3] or [1,3]. For embedded citations in the text with pagination, use both parentheses and brackets to indicate the reference number and page numbers; for example [5] (p. 10). or [6] (pp. 101–105).”
2. add in abstract the method you used 
3. rephrase the last phrase of the abstract (L27)
4. add the sources in the following paragraphs: 39-L45, L223-L231; L232-L243;
5. revise the extra spaces in the text (L60, L65, L75) and between paragraphs (L155 and L166)
6. please revise the introduction. It is too long as it is and contains parts that should be moved to the ”Results” section (L98-L107) or  the ”Method” Section (L82-L97)
7. please add sources for the tables and figures included in the article. 
8. please explain in detail the sample and the software you used

Author Response

Dear Reviewer,

thank you very much for your suggestions. We strongly appreciated your comments and we worked for addressing each point you raised. Here you find below the answers to each question.

We hope everything is clear to you. We are available for additional refinements if needed.

1. please revise all in-text citations and Bibliography section according to the Journal indications: ”In the text, reference numbers should be placed in square brackets [ ], and placed before the punctuation; for example [1], [1–3] or [1,3]. For embedded citations in the text with pagination, use both parentheses and brackets to indicate the reference number and page numbers; for example [5] (p. 10). or [6] (pp. 101–105).”

• We changed all the biographical section considering the Journal indications

1. add in abstract the method you used 

• We added the method used– Fractional Logistic Regression model

1. rephrase the last phrase of the abstract (L27)

• We adjusted the last sentence of the abstract, and we made it more punctual

1. add the sources in the following paragraphs: 39-L45, L223-L231; L232-L243;

• We included the references and sources now positioned respectively at L50, L884, L893

1. revise the extra spaces in the text (L60, L65, L75) and between paragraphs (L155 and L166)

• We went throughout the text to remove all extra spaces

1. please revise the introduction. It is too long as it is and contains parts that should be moved to the ”Results” section (L98-L107) or  the ”Method” Section (L82-L97)

• We attempted to revise the introduction according to your suggestions, making it more concise, and shorter.

1. please add sources for the tables and figures included in the article. 

• Each table and figure present a dedicated name. We added the source for Figure 1, 2, 3

1. please explain in detail the sample and the software you used

• We included details of the geographical localization of investors involved in the sample in section 4 – Table 2 and included the software we used in the subsection 5.2

Round 2

Reviewer 2 Report

Thanks for your effort to enhance the qualification. 

Good luck.

Author Response

Dear Editor,

Thank you very much for giving us the chance to further update our paper.

In this process of minor reviews, we have noticed that Reviewer 2 did not provide indications to make further edits to the paper. Thus, we evidenced the following indication, as a part of your comments:

"The formatting rules need to be carefully followed, such as: the headings of figures are not unified, Tables and Figures must be mentioned in the text, and so on. The figure of the graph 1 a legend is missing.”

Here it follows the recap of changes made:

• We renamed Graph 1 into Figure 1, following journal guidelines indications. We have also included a legend better explaining dashed and solid lines.
• We solved formatting issues in the text, relabeling tables 1, 2, 3, 4, 5, 6, 7, carefully indicating them in the text.

In addition, and if necessary, we would like to conduct an English editing to the text. Just in case you find it relevant, we are available to go through it.

We look forward to hearing from you soon

Best

Leonardo Boni, Laura Toschi, Riccardo Fini

Author Response File: Author Response.docx