Corporate Bankruptcy Prediction Models: A Comparative Study for the Construction Sector in Greece

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The study discusses corporate bankruptcy prediction models and their relevance in financial research, particularly in the context of Greece.

The study lacks clarity and organization. The ideas are presented in a somewhat fragmented and unstructured manner. It would benefit from a clearer introduction, better transitions between points, and a more organized flow of information to guide the reader.

In any kind of  bankruptcy prediction, you need a large sample size. The sample size in this study so small that you cannot draw any meaningful conclusions.

Moreover, you need to include discussion of machine learning techniques as a tool for bankruptcy prediction. Machine learning tools such as neural works, neuro-fuzzy systems, kNN, gradient boosting approaches are now part of mainstream business school education and they are used to predict bankruptcies.

Conclusions are weak. The study suggests potential future research areas, such as applying the models to non-listed companies or other sectors in the Greek economy. However, it lacks specific research questions or hypotheses that could guide such research, making these suggestions somewhat vague.

Comments on the Quality of English Language

paper needs extensive editing. 

Author Response

Dear Reviewer

Thank you for your constructive comments and recommendations. All these suggestions were taken into account when preparing the new version of the manuscript; relevant modifications are now incorporated into the text. This has, in our opinion, improved the quality of the manuscript.

• The study lacks clarity and organization. The ideas are presented in a somewhat fragmented and unstructured manner. It would benefit from a cleaner introduction, better transitions between points and a more organized flow of information to guide the reader.

We thank the referee for this comment. In the revised version we address this issue and there are modifications towards improving the above in all sections. Please see added text in red letters.

• In any kind of bankruptcy prediction, you need a large sample size. The sample size is so small that you cannot draw any meaningful conclusions.

We thank the referee for this comment. Please see added text in red letters in the Materials section.

•  Moreover you need to include discussion of machine learning techniques as a tool for bankruptcy prediction. Machine learning techniques such as neural networks, neuro-fuzzy systems, KNN, gradient boosting approaches are now part of mainstream business school education and they are used to predict bankruptcies

We thank the referee for this comment. Relevant bibliographic sources have now been added in the revised version of the paper (in red letters) so that the initial number of 65 references has been raised to 83 in the revised version.

•  Conclusions are weak. The study suggests future research areas, such as applying the models to non-listed companies or other sectors in the Greek economy. However, it lacks specific research questions or hypothesis that could guide such research, making these suggestions, somewhat vague.

We thank the referee for this comment. In the revised version we address this issue and there are modifications towards improving the above section. Please see added text in red letters.

Reviewer 2 Report

Comments and Suggestions for Authors

The paper

Corporate Bankruptcy Prediction Models: A Comparative Study for the Construction Sector in Greece

has the following structure:

Abstract

1. Introduction

2.Literature review

3.Materials

4.Methods and Results

5. Conclusions

The Abstract is very narrow, and states the objective of the paper.

 

Introduction presents some literature review, and some thoughts about the construction sector.

Literature review has lots of bullets and presents some about the models.

 The Materials part presents the data. There are some coloured tables (there is no real need for colours, in our opinion). Altman score for healthy and non - healthy companies is presented.

 Methods and Results continues in the same way (some tables with scores).

 

It is not clear:

- the novelty of the paper;

- why are these calculated scores important;

- who can benefits from this research;

- there is no statistical method;

Regarding the data

- 10 companies are analysed, which is insufficient for valid conclusions.

 

There are 65 references, in line with the scope of the paper.

 

We propose to REJECT the paper.

Author Response

Dear Reviewer

Thank you for your constructive comments and recommendations. All these suggestions were taken into account when preparing the new version of the manuscript; relevant modifications are now incorporated into the text. This has, in our opinion, improved the quality of the manuscript.

Below we provide a response-rebuttal to each of the issues that you have raised, indicating how each of the recommendations was addressed and providing the appropriate changes in the text of the revised paper.

We are submitting the new manuscript with changes indicated in the text in red letters  in order to assist identifying corrections..

Thank you for your consideration of our manuscript.

On behalf of the authors.

 

• There are some coloured tables (there is no real need for colours, in our opinion).

We thank the referee for this comment. Colors were removed from all tables in the revised version of the manuscript submitted.

•  It is not clear the novelty of the paper

We thank the referee for this comment. In the revised version we address this issue both in the Introduction and in the Conclusions sections.  Please see added text in red letters.

•  It is not clear why are these calculated scores important

We thank the referee for this comment. In the revised version we address this issue both in the Introduction, the Methods and Results and in the Conclusions sections.  Please see added text in red letters.

•  It is not clear who can benefit from this research?

We thank the referee for this comment. In the revised version we address this issue both in the Introduction and in the Conclusions sections.  Please see added text in red letters.

• There is no statistical method

We thank the referee for this comment. In the revised version we address this issue adding the Appendix.  Please see added text in red letters.

•  10 companies are analysed, which is insufficient for valid conclusions

We thank the referee for this comment. Please see added text in red letters in the Materials section.

Reviewer 3 Report

Comments and Suggestions for Authors

1. This study focuses on testing the efficiency of alternative bankruptcy prediction models (Altman, Ohlson, Zmijewski). They claim that the results showed that Altman's main predictive model as well as the revised models have 'low' overall predictability for all three years before bankruptcy.

  Specify 'low'  more precisely,  i.e., in terms of  number or %. 

2. Table 3, 5  and 7 are hard to understand.

It categorizes two error type (type 1 and type 2)  and then  put  "no of correct' (the number of firms forecasted correctly, I guess)  in each of type 1 or type 2.   Then "no of correct'  means correct type I error ??    

3. p 10 line 407: Please specify  x for  P=1/(1+e^-x). 

   p 12 line 437 : Please specify z  in the equation number (6) 

 

No 

 2. 

Comments on the Quality of English Language

It's fine but can be improved. 

Author Response

Dear Reviewer,

We owe you many thanks foryour constructive comments and recommendations. All these suggestions were taken into account when preparing the new version of the manuscript; relevant modifications are now incorporated into the text. This has, in our opinion, improved the quality of the manuscript.

Below we provide a response-rebuttal to each of the issues that the you have raised, indicating how each of the recommendations was addressed and providing the appropriate changes in the text of the revised paper.

We are submitting the new manuscript with changes indicated in the text in red letters  in order to assist identifying corrections..

Thank you for your consideration of our manuscript.

On behalf of the authors.

 

• The study focuses on testing the efficiency of alternative bankruptcy presiction models (Altman, Ohlson, Zmijewski). They claim that the results showed that Altman’s main predicitive model as well as the revised models have “low” overall predictability for all three years before bankruptcy. Specify “low” more precisely, i.e. in terms of number or %

We thank the referee for this comment. Please see added text in red letters in the Methods and Results section in page 8.

•  Table 3, 5 and 7 are hard to understand. It categorizes two error type (type 1 and type 2) and then put “no of correct” (the number of firms forecasted correctly, I guess) in each of type 1 or type 2. The “no correct” means type I error?

We thank the referee for this comment. Please see added text in red letters in the Methods and Results in the relevant paragraphs before Tables 3, 5 and 7.

•  10 line 407: Please specify x for P=1/(1+e…….-x)
• p 12 line 437: Please specify z in equation number (6)

We thank the referee for this comment. In the revised version we have added specifications in red letters.

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I continue to have issues with the basic premise of this paper. The paper is evaluating whether Altman's model is superior, but it fails to acknowledge that machine learning models have taken over bankruptcy prediction. As a result, the rationale for the study is not clear. Since machine learning models can prove to be a better approach in predictive analytics, what is the contribution of this study? What does it tell us that adds to the existing body of knowledge? Moreover, the study is restricted to one industry only.

I would recommend that you should evaluate machine learning techniques that include

A.      The k-Nearest Neighbor (kNN) Model

B.      Decision Tree Classifier

C.      Extreme Boosting (XGBoost) Model

D.      Support Vector Machines (SVM) Model

E.      Random Forest Classifier

F.       Gradient Boosting Approach

G.      Gaussian Naive Bayes classifier:

H.     AdaBoost

There have been studies that use these models to evaluate credit ratings as well as corporate bankruptcy prediction or credit outlook evaluation.

You need to use these studies and restructure your paper.

Comments on the Quality of English Language

It is fine

Author Response

Dear Reviewer

We would like to thank you once again for the constructive comments and recommendations. All these suggestions were considered when preparing this new revised version of the manuscript, further improving its quality.

Below we provide a response-rebuttal to each of the issues that the Referees raised, indicating how each of the recommendations was addressed and providing the appropriate changes in the text of the revised paper.

We are submitting the new manuscript with changes indicated in the text in red letters to assist identifying corrections.

Thank you for your consideration of our manuscript.

On behalf of the authors.

• The paper is evaluating whether Altman’s model is superior, but it fails to acknowledge that machine learning models have taken over bankruptcy prediction. As a result, the rationale of the study is not clear. Since machine learning models can prove to be a better approach is predictive analytics, what is the contribution of this study? What does it tell us that adds to the existing body of knowledge?

We thank the referee for this comment. Please see added text in red letters at the end of the “Literature Review” section:

“As specifically stated by [63] “although some papers have studied credit default and machine learning ([81-85]), new studies, exploring different models, contexts, and datasets, are relevant, since results regarding the superiority of models are still inconclusive.  The debate over the best models for predicting failure will probably continue in the short and medium terms, as new techniques are frequently being suggested and, particularly for the study of corporate bankruptcy, failure events are subject to myriad variables. In this context, for instance, with the advancement of technology, data scraping will allow the observation of new variables that could be relevant inputs to machine learning models and lead to different results”. They also concluded that “the Altman and Ohlson models are still relevant, due not only to their predictive power but also to their simple, practical, and consistent frameworks”. Furthermore, [86] noted that “despite its “old age”, the Altman Z-score is still the standard against which most other bankruptcy or default prediction models are measured and is clearly the most used by financial market practitioners and academic scholars for a variety of purposes”.

 

• Moreover, the study is restricted to one study only.

We thank the referee for this comment. Please see added text in red letters at the end of the “Literature Review” section:

‘Furthermore, [86] noted that “despite its “old age”, the Altman Z-score is still the standard against which most other bankruptcy or default prediction models are measured and is clearly the most used by financial market practitioners and academic scholars for a variety of purposes”. Finally, it should be noted that such bankruptcy prediction studies analyzed either specific economic sectors of a country (indicatively the banking sector [87]; the hotel industry [88 -89]; the trading sector [90]; the cement industry [91]) or specific companies (indicatively [92-93]). The present study concentrated on the construction sector due to its importance for the Greek economy, as illustrated above.’

 

• I would recommend that you should evaluate machine learning techniques that include a) THE k-Nearest Neighbor (kNN) Model, b) Decision Tree Classifier, c) Extreme Boosting (XGBoost) Model, d) Support Vector Machines (SVM) Models, e) Random Forest Classifier, f) Gradient Boosting approach, g) Gaussian Naïve Bayes classifier, h) AdaBoost. There have been studies that use these models to evaluate credit ratings as well as corporate bankruptcy prediction or credit outlook evaluation. You need to use these studies to restructure your paper.

We thank the referee for this comment. Relevant bibliographic sources have now been added in the revised version of the paper (in red letters) so that the number of 83 references of revised in review round 1 manuscript, has been raised to 102 in the revised version of review round 2. Please see added text in red letters.

Reviewer 2 Report

Comments and Suggestions for Authors

There is still no methods/ methodology presented - ANOVA is inserted in APPENDIX, but there is no clear relationship with the content. Why is ANOVA used? The data for the analysed companies is from 2010 - during COVID19 are there no bankruptcies? The idea is interesting, but applied to panel data - the European sector or at least using multiple countries comparison with an updated newest dataset.

Author Response

Dear Reviewer

We would like to thank you once again for the constructive comments and recommendations. All these suggestions were considered when preparing this new revised version of the manuscript, further improving its quality.

Below we provide a response-rebuttal to each of the issues that the Referees raised, indicating how each of the recommendations was addressed and providing the appropriate changes in the text of the revised paper.

We are submitting the new manuscript with changes indicated in the text in red letters to assist identifying corrections.

Thank you for your consideration of our manuscript.

On behalf of the authors.

• There is still no methods/methodology presented – ANOVA is inserted in the Appendix, but there is no clear relationship with the content. Why is ANOVA used?

We thank the referee for this comment. Please see added text in red letters in the “Methods and Results” section.

 

• The data for the analyzed companies is from 2010 – during COVD are there no bankruptcies? The idea is interesting but applied to panel data – the European Sector or at lease using multiple countries comparison with an updated newest dataset.
• We thank the referee for this comment.

Please see added text in red letters:

a) in the “Materials” section:

"Since then (2010). Greece entered a prolonged recession period, that coupled other crises that occurred (i.e. COVID pandemic, energy crisis, etc.) resulted to a domino of company failures Thus the pre-Memorandum period sample choice offered a better representation of the country’s normal economic activity (strengthening the importance of the results obtained). Furthermore, the prolonged duration of the financial crisis prevented the study of European countries panel data that would include Greece."

b) in the “Literature Review” section:

"Finally, it should be noted that such bankruptcy prediction studies analyzed either specific economic sectors of a country (indicatively, the banking sector [87]; the hotel industry [88 -89]; the trading sector [90]; the cement industry [91]) or specific companies (indicatively, [92-93]). The present study concentrated on the construction sector due to its importance for the Greek economy, as illustrated above. The Altman Z score model has widely been applied all over the world; indicatively, for Greece [53]; for Pakistan [94]; for China [95]; for USA [89]; for India ([96-97]; for Indonesia [98]; for Bangladesh [87], [91]; for Malaysia [90]."

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

When I googled research about corporate credit outlook, I came across some papers.  You need to update the literature review.

 

Author Response

Dear Reviewer

We would like to thank you once again for the constructive comments and recommendations. All these suggestions were considered when preparing this new revised version of the manuscript, further improving its quality.

Below we provide a response-rebuttal to each of the issues that the Referees raised, indicating how each of the recommendations was addressed and providing the appropriate changes in the text of the revised paper.

We are submitting the new manuscript with changes indicated in the text in red letters to assist identifying corrections.

Thank you for your consideration of our manuscript.

Our best wishes for a healthy and happy New Year

On behalf of the authors.

 

• When I googled research about corporate credit outlook, I came across some papers. You need to update the literature review

We thank the referee for this comment. Relevant bibliographic sources have now been added in the revised version of the paper (in red letters) so that the number of  102 references of the revised manuscript in review round 2, has now been raised to 124 references in the revised version submitted for review round 3, an increase of 21,6% over the previous review round 2. Please also note that

• in review round 2, as requested by the reviewers, the number of references has also been upgraded from 83 references in the revised in review round 1 manuscript, to 102 in the revised version of review round 2 (an increase of 22,3% over the previous review round 1) and
• in review round 1, as requested by the reviewers, the number of references has also been upgraded from 65 references in the manuscript initially submitted, to 83 in the revised version of review round 1 (an increase of 27,7% over the initial submission)

Therefore, the initial number of 65 references in the manuscript initially submitted, has finally been upgraded (after the 3 review rounds) to 124 in the revised version of review round 3 (a total increase of 90,8% over the initial submission)

Please see added text in red letters.

Reviewer 2 Report

Comments and Suggestions for Authors

No new comment.

Author Response

Dear Reviewer

Thank you for your consideration of our manuscript.

We are submitting the new manuscript with changes indicated in the text in red letters to assist identifying corrections.

Our best wishes for a healthy and happy New Year

On behalf of the authors.