Responsibility and Performance Relationship in the Banking Industry

Margolis, J. D., & Walsh, J. P. (2003). Misery loves companies: Rethinking social initiatives by business. Administrative science quarterly, 48(2), 268-305.

1

Thank you for the opportunity to read your paper. I agree with the   authors that the issue of the relationship between financial performance and   corporate social responsibility in the financial industry is interesting and   it requires further attention.  Although the issue of the paper   could fit in the special issue and the text is well-written, I am not   convinced that the current version of the manuscript addresses the topic in a   really qualified manner to be publish in the journal. Therefore, I think that   this current piece could be improved if the authors consider the following   comments

Thank you very much for putting in the effort to study our paper and to   provide constructive comments and useful suggestions. We do our best to   account for each of them.

2

INTRODUCTION (and abstract)  My first, and most important, point   concerns the central nature of the contribution. I am fine with the attempt   to explore the relationship between financial performance and corporate   social responsibility in the finance industry, but I am extremely confused   about the paper’s contributions. What is the originality of the paper? How   could you really convince readers about the importance of the large   econometric analyses’ contributions? In addition, does this paper have some   contributions to extant sustainability literature? The major problem here is   that the reader is struggling when it comes to understand the different   hypotheses and the large number of econometric estimations without clearly   reading the contributions of this paper, since this aspect is addressed in a   very superficial manner. 

Thanks for raising this issue. The contribution is in the level of detail   of using ESG information, in the detail of using financial performance   indicators, and in the focus on the banking industry.

We highlight this in the abstract and introduction. As such, we feel our   paper is highly original.

We reduced the number of hypotheses (from six to four). Our econometric   analyses allow us to reveal insights into the mechanisms that relate   financial performance and responsibility. We did rewrite several key   fragments of the text to guide the reader and to highlight the contributions.

3

In the first page, the   authors highlight different international initiatives regarding CSR. They   could improve the motivation, explaining how this paper contributes to other   most recent initiatives like the Circular Economy Action Plan of the EU or   the Commission Action Plan on Financial Sustainable growth.

Thank you very much for suggesting this.

4

Regarding “1.1 Corporate   social responsibility and financial performance”, “1.2. Corporate social   responsibility in the banking industry” and “1.3. Hypotheses”, the authors   should connect these three sections better.  This section is where   the authors should begin to build the case that this research matters. They   mention a mix of works that explore the relationship between financial   performance and CSR but these are disconnected with respect to the hypothesis   that will be empirically tested. Moreover, the manuscript needs to have a   strong theoretical foundation (which is the main theory that supports the   hypotheses: Stakeholder Theory? Resource Dependency Theory?, please see   Rivera et al. (2017) for exploring more theories regarding this relationship)   and it will support the hypotheses. The authors present results from prior   research but fail to connect the empirical literature that cite with   appropriate theory.  They need to improve the justification of the   presented hypotheses.  I suggest that authors focused on the most   interesting hypotheses for the paper in terms of contribution. In the current   version, there are too many hypotheses to address in an article. This fact   distracts the reader’s attention with respect to the contribution of the   paper.  Perhaps it is better to cover a smaller number of hypotheses   in a better way.

We take care to motivate these different elements and to make clear how   they connect.

We do not want to stick to one particular view, but depart from the   spectrum provided by Preston and O’Bannon about the alternative (competing)   views. So, the hypotheses are motivated by the Preston and O’Bannon paper and   are of an exploratory nature. So far, the literature has come up with various   ideas about how CSR and CFP might relate but their underpinnings are not very   strong. Therefore, we keep to the original taxonomy of Preston and O’Bannon,   which also encompasses the spectrum of ideas mentioned in the paper of Rivera   c.s (i.e., their Table 1).

We have an open mind and are not committed to a particular view as none   of these is superior on a priori grounds. For sure, we are not after finding   support for any of these hypotheses. Instead, we try to reject them.

We reorganized the text regarding the hypotheses.

We used the econometric analysis to help us sort what seems to work and   what doesn’t. We reduced the number of hypotheses to be tested (from six to   four).

5

MATERIALS AND METHODS  The authors treat “corporate social   responsibility”, “responsibility”, and the three variables of Asset4   variables “Corporate Governance performance, Environmental performance, and   Social performance” as synonyms in several parts of the paper (for instance   in several hypotheses). This should be clarified. If ESG variables from   asset-4 are proxies of CSR, this should be explained better. In this regard,   the paper could include one of the most accepted definition of CSR (EC, 2011)   “the responsibility of enterprises for their impacts on society”.

Thanks for mentioning this. These constructs indeed proxy for CSR and we   now explicitly mention so.

We refer to the European Union’s view on CSR ([38]).

6

In page 6, authors include   arguments related to “process-based measures” vs. “outcome-based measures”.   However, it is not clear what type of measures the authors include in the   paper and how they could justify the choice. 

Asset4 basically is a processed based measure. Information   disclosure regarding the non-financial performance of firms, especially   regarding their external effects, has not been integrated in mandatory /   compulsory accounting and reporting standards as of yet. See section 2.1

7

With respect to the   aggregation of data concern (page 7), on one hand, there are academic   articles that overcome this limitation (see for instance Escrig-Olmedo et al.   2017). On the other hand, when the authors use governance, environmental or   social performance, even these measures at a more detailed level, implicitly,   they also are aggregating information. 

Escrig-Olmedo c.s. indeed highlight the issue but they do not claim to   overcome them and especially mention on page 159: “…the limitations in selecting one database over other choices could   condition the use of one set of indicators against others…..”. 

8

With respect to the first   two hypothesis, the method used (presented in Table 2) has a large number of   limitations. I recommend from the beginning of the paper focus the scope on   the financial industry and remove these two hypotheses. 

Thanks for this suggestion. We have removed the first two hypotheses in   the revised version.

9

I recommend that the authors   use a concrete model or a system of equation to test the hypotheses. Could   the authors use directly the equations 3 and 4 to test the hypotheses?

We have revised the section explaining our methods (section 3.2) based on   your recommendation.

10

Please justify with previous   studies the best model and variables for the objective of the paper. In that   case, we should expect that the proxies used for each variable should have   the same sign. This information could be presented in a table. 

See also our response to comment #4 above. This study is really of an   exploratory nature and we definitely are not after confirmation of a   particular view. None of the views systematically yields ‘best’ results when we   test them.

11

If the authors consider that   “Corporate Governance score”, “Environmental score”, and “Social score” are   not proxies of “CSR score”, then, in model (3) and (4) they should delate   “CSR score” and include “Corporate Governance performance”, “Environmental   performance”, and “Social performance”.

Thanks for bringing this to our attention. We have revised the equations   accordingly.

12

The issue of collinearity is   not fully explored in the models that include interaction variables.   Additionally, I recommend that the authors compute variance inflation   factors. 

As you would agree with us, the possibility of high correlations with   interaction variables is normal and expected. This is also case in our   models. Since we use a dummy variable interacting with responsibility scores   whose mean values by country/year and country/industry are included as   instruments in responsibility equation, we didn’t have to do much to mitigate   this problem. We report VIFs for our main model in a new appendix and establishe   that there are no any concerns for multicollineary among our main variables.  

13

RESULTS The results of   regressions are different depending on the proxy used. The authors should   explain this better and clarify is they find support or not for each   hypothesis. The authors should justify the model used and display only the   result of the selected model and explain better the results. The rest of the   analysis, could be summarized and considered as a robustness analysis. Also,   they might extend the discussion of practical and theoretical implications of   the study.

We have revised the relevant sections.

14

The authors should highlight other limitations, which can be derived from   the proxies used or the specification of the model. Specifically, the   omission of the variables like R&D variables (Margolis and Walsh, 2003)   and intangible variables (Surroca, Tribó, Waddock, 2010) could generated a   misspecification problem (McWilliams and Siegel, 2000). The authors also   could mention future studies considering this type of variables.

Thank you very much. All variables already are intangible (namely   financial assets), and there is no database which has information about   R&D specific activity in the banking industry. This is motivated when we   discuss the focus on banks. The diagnostics of the estimation results too do   not signal there would be misspecification. We feel the comment in particular   is highly relevant for multi-industry studies.

15

Another limitation is simultaneous effect between financial performance   and corporate social responsibility. Future works could better examine the cause-effect   relationship between both variables using lags in the specification of the   models.

 I hope the authors find these comments constructive and motivating   for continuing the work on this manuscript. PS.References must be   completed with DOI

Thanks for suggesting this. We appreciate your thoughtful reflections a   lot. We also include your suggestion into our limitations listed in our   conclusions.

16

Thanks for suggesting these. We now do refer to the paper by   Escrig-Olmedo et al.; the views of Margolis and Walsh were already accounted   for (references 39, 43).

1

Although in the title, the authors use the word “finance industry,” this   study is mainly about banks. I would suggest to indicate that intention   directly. Also, in the paper, the authors seem to use banks and financial   firms interchangeably. But I do observe occasions that “financial firms” are   broader than “banks,” which include insurance companies, financial trading   companies, real estate firms and banks. The authors need to clearly craft the   writing to minimize the confusion.

Thanks for highlighting this. We now   concentrate on the banking industry based on the regressions analysis in the   revised version.

2

This is a rather long paper. I would suggest   the author to reduce its lengths and increase its readability. For example,   in the literature, the investigation about the general CSR-CFP relation can   be significantly reduced. The focus should be highlighting the unique of bank   industry to illustrate why focusing on this industry can shed new light in   the crowded literature.

Thanks for suggesting this. We have revised   this section.

3

The development of the hypotheses should be integrated with the argument   so that the exposition can be tightened. So far, the hypotheses and their   development seem to be detached. In particular, the fifth hypothesis   regarding the effect of global financial crisis is unclear. A hypothesis   should be a testable proposition. Here, I do not see a clear proposition. The   same argument applies to your sixth hypothesis as well.

Thank you very much. We have amended the text   in in line with the suggestions. We want to reiterate here that the last two   hypotheses are of an exploratory nature due to the lack of theory how CSR   relates to crises and institutional development.

4

The comparison between banks and other non-financial firms is not   informative. Banks mainly use deposits as input and produce loans as output.   Their business operations are quite different from other non-financial firms.   In addition, banks can make their loan decisions according to the CSR   profiles of their customers. Thus, banks can influence non-financial firms’   CSR strategies. Therefore, it is unclear whether we could say that banks   possess less responsibility, compared with non-financial firms.

We have removed such direct comparisons, see also comment #1 above.

5

In the identification of 3SLS, the authors use 4 instruments. Lagged   financial performance and private credit to GDP for the financial performance   equation. The means of responsibility scores by country/year and   country/industry for the CSR equation. There are several issues related to   choice of instruments. First, to implement 3SLS, the set of instrument is the   same across equations (see Hayashi, Econometrics) with the assumption of   conditional homeskedasticity. Second, the authors do not justify why the instruments   satisfy the requirement of predetermined regressor. Third, using peer group   averages as instruments is nearly never valid since any endogeneity at the   individual firm level is simply soaked up at the group level (see Gormley and   Matsa (2013), “Common Errors: How to (and Not to) Control for Unobserved   Heterogeneity”, Review of Financial Studies, especially Section 2.3.4). While   it is true that some papers have got away with using peer group averages in   the past (e.g. Cheng et al., which was not published in a finance journal),   it is now well established that peer group instruments are invalid.

The referee is correct on this and we fully agree.   However, it is highly challenging to arrive at the best instruments. In fact,   in our experience, it is very hard to come up with instruments that are not   open to challenge. As we lack the best instruments, we point out that our suboptimal   instruments have been used widely in the literature and are subject to the   issues you raise. We mention this caveat.

6

The analyses of crisis are interesting. Nonetheless, the authors only   explore the differences in banks’ engagement in CSR, which leaves the   audience “so what” question. Global crisis in 2007-2009 is almost ten years   ago. Do banks revert to their pre-crisis level of CSR engagement? To offer   new insights, the current findings are not sufficient.

Given the significant differences that show up in Table 1 – Panel C, we   felt it highly relevant to further investigate the impact of the crisis.   However, as you rightfully pointed out with comment #3, there is no clear   framework which guides us as to what to expect. That the   finance-responsibility nexus differs after the crisis compared to before in   itself is of interest from our perspective as it reveals the degree to which   CSR engagement is embedded in banks. For the period analyzed, it seems it   does not revert to pre-crisis levels in the governance and social domains.

1

I think that this version   of the manuscript has been improved substantially. Nonetheless, I think that   before the publication, the authors need to take into account the following   considerations presented in the previous review:

Thank you very much. We are   very pleased that you feel we made substantial improvements. Thanks for your   support.

We address the additional   comments; see our response below.

2

Comment 2: The authors   should motivate better why is important to explore the relationship between   ESG information and financial performance in the banking industry.

3

Comment 3: The paper needs   a greater conceptual clarity in the hypotheses. In the paper of Preston and   O’Bannon published in 1997, these authors defined hypotheses according to   their different theoretical perspectives. The author of this manuscript need   to justify better the hypotheses.  An   article is more than an econometric exercise to analyze what seems to work   and what doesn’t.

We expand the motivation of   the hypotheses and relate in more detail to the overview of Rivera et al. in   particular.

4

Comment 10: The authors   need to justify better the chosen models.

We expand the motivation of   the type and structure of the models used in the analysis. HALIT

5

Comment 14: Clarification   to the answer: Databases like “Worldscope/Datastream” has proxies to measure   R&D investment by company, and therefore, it could be calculated for the   specific industries.

In the revised version, as   per request of the referees, we analyse the banking sample only and don’t   compare between industries anymore.

6

Keywords: The authors   include the “ESG” concept, but in the main text, at least from the main text   this acronym does not appear as a key concept.

Thanks for flagging this. It   is changed into “responsibility”, which we use as our main concept in this   study.

7

References: Please, follow   the instructions of the journal: “References must be numbered in order of appearance   in the text (including table captions and figure legends) and listed   individually at the end of the manuscript.”

Thanks for pointing this   out, we did amend the references list and checked the text for the numbering   in order of appearance.

1

I think the authors have   done a good job to address the issues raised in my referee report. However,   this paper could benefit from having a professional English editor proofread   the entire manuscript.

Thank you very much. We are   very pleased you feel we did properly address the issues raised in your   referee report. We very much appreciate that you provided such thoughtful and   constructive comments.

We did thoroughly review of   the text.