Next Article in Journal
Enhancing Efficiency: Halton Draws in the Generalized True Random Effects Model
Previous Article in Journal
Impact of Areal Factors on Students’ Travel Mode Choices: A Bayesian Spatial Analysis
 
 
Article
Peer-Review Record

Exploring the Role of Global Value Chain Position in Economic Models for Bankruptcy Forecasting

Econometrics 2024, 12(4), 31; https://doi.org/10.3390/econometrics12040031
by Mélanie Croquet 1, Loredana Cultrera 1, Dimitri Laroutis 2, Laetitia Pozniak 1 and Guillaume Vermeylen 3,*
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3:
Econometrics 2024, 12(4), 31; https://doi.org/10.3390/econometrics12040031
Submission received: 14 August 2024 / Revised: 5 October 2024 / Accepted: 28 October 2024 / Published: 5 November 2024

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is an interesting study of bankruptcy using AI techniques to predict bankruptcy, finding that AI provides the best predictions. I have the following comments:

- The study uses a sample of Belgian bankruptcies, are there any differences between Belgian bankruptcy law and SMEs compared to other countries already studied, these differences are referred to on page 11 but not expanded on? This could be emphasised more in the introduction.

- The data ends 2012, so covers most of the financial crisis era, was this because of the large number of bankruptcies then or another reason? In general the nature of the data could do with more discussion.

- How was the non-bankruptcy data chosen, was it a random selection approach? Also I think this is a panel data model, was the Logit model estimated with fixed effects? 

- Some summary statistics of the data should be added and discussed to give the reader a feel for the data, possibly in an appendices.

Comments on the Quality of English Language

The English is good, but could do with a final proof reading.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The use of non-parametric methods of analysis, including those used in this paper, is currently gaining popularity. The combination of parametric and non-parametric estimations indicates that the authors use various possibilities to test their assumptions and, accordingly, allows us to conclude that the results obtained are reliable.

Predictors were chosen correctly; the variables that showed correlation between each other could have caused concern, but the authors took into account the possibility of error and excluded indicators that could cause multicollinearity by calculating the VIF-coefficient.

It should be noted that the necessity and relevance of such a study is dictated by the questionable methods developed to date, which do not allow a more or less accurate assessment of the possibility of bankruptcy of small and medium-sized enterprises. As a consequence, the paper under review closes to some extent the relevant gap in the scientific literature, making a notable contribution to its resolution.

In my opinion, the study and its results are interesting and reliable, can be applied for the specified purposes.

As comments and suggestions, I would like to point out:

1. С. 94-99 the description of the structure should be corrected - section numbers do not correspond to reality.

2. It would be useful to specify what software tools the authors used in calculations and model building.

3. It is necessary to specify the limitations of the study.

4. It is important to present the directions of further development of the work.

I wish you success in your future endeavours! 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors

The manuscript “Exploring the Role of Global Value Chain Position in Economic Models for Bankruptcy Forecasting” conducts a comparative analysis of statistical and artificial intelligence techniques for predicting bankruptcy in small and medium-sized enterprises using a database of 7,104 Belgian SMEs. While the study is valuable, there are some limitations and drawbacks in the manuscript that should be considered:

1.     The problem statement is not clear in the abstract.  It also doesn't explain why only Belgian SMEs were chosen. The abstract suggests that the firm's position in the value chain affects bankruptcy prediction, but it doesn't provide details on how this was measured or the extent of its impact..

 

2.     The introduction has some issues. It overemphasizes the history and general context of bankruptcy without clearly connecting it to the specific focus of the study on SMEs. It doesn't adequately explain why the gap is significant. The introduction need to clearly link the ideas.

 

3.     The introduction could benefit from a clearer connection between the proposed solution and the specific problems it aims to solve. To strengthen the introduction, it would be beneficial to include studies that have successfully applied AI techniques, to complex, real-world data forecasting. The studies, including Advancing bankruptcy forecasting with hybrid machine learning techniques: Insights from an unbalanced Polish dataset;  deep learning based modeling of groundwater storage change; cdlstm: a novel model for climate change forecasting; smotednn: novel model for air pollution forecasting and aqi classification" demonstrate the effectiveness of AI in processing and forecasting the data in diverse and challenging environments.

 

4.     The exclusion of firms with missing or inappropriate variables may introduce selection bias, potentially skewing the results. The methodology mentions the use of 50 financial ratios but then reduces them to 30 through multicollinearity checks and further to a smaller subset, yet it does not clearly explain the criteria or rationale behind this reduction process, which could affect the robustness of the model. The use of four different variable selection techniques (statistical and AI-based) is mentioned, but there is no discussion on how these methods compare to each other. other or why these specific techniques were chosen.

 

5.     To enhance the methodology, consider incorporating Principal Component Analysis (PCA) for dimensionality reduction and to address multicollinearity. PCA simplifies the model by focusing on uncorrelated components, improving prediction accuracy.

6.     Random Forest and XGBoost can be employed for their ability to capture complex, non-linear relationships and automatically select important features. These ensemble methods can complement PCA by identifying critical variables while handling interactions.

7.     Limitations and the future scope should be added with more clarity.

 8.     An experiment environment with computational complexity should be added.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The revisions have been done well.

Comments on the Quality of English Language

The English is well written but could do with a final proof reading.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors

Thank you for submitting your revised manuscript. There are still several critical issues that have not been sufficiently addressed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors

Dear Authors

Thank you for your thoughtful revisions. I am pleased to see that you have successfully addressed the crucial comments regarding feature extraction, and have provided more detailed information on the model and computational processes. The improvements in these areas significantly enhance the clarity and depth of your work.

This version is much improved and meets the expectations set during the review. It looks ready for the next steps.

Back to TopTop