Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • harvard-cite-them-right
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
From classical business failure prediction models to business financial models for resilience: using advanced statistical methodologies
University of Borås, Faculty of Textiles, Engineering and Business. (Textile Management)ORCID iD: 0000-0003-2015-6275
Trilobyte Statistical Software Ltd., Czech Republic.
College of Technology and Computer Science, East Carolina University, United States.
Department of Material Engineering, University of Liberec, Czech Republic.
2015 (English)In: : ICORS, 2015, p. 39-40Conference paper, Oral presentation with published abstract (Refereed)
Abstract [en]

Over the last 35 years business failure prediction using various methods like univariate analysis, multi-variate analysis, credit risk models etc. has become a major research domain within corporate finance (Balcaen and Ooghe 2006). These mathematical models are increasingly accepted by financial institutions, governments and the European Union in the Basel Accords (Basel II/III). However, most classic statistical failure prediction models are developed without comprehensive understanding of the nature of company failure with often arbitrarily variables chosen in an ad-hoc manner (Beaver 1967b, Cybinski 2001). In this context, the paper uses advanced statistical methodology to propose a robust business financial modeling technique. Data on 18 key financial parameters were collected for 198 US-based public companies along with their expert credit rating for 2012-13. Firstly, a correlation study was performed between Altman scores and widely accepted expert rating based on stock exchange activities. Secondly, “stage-wise” regression was conducted to select the statistically most significant candidate ratios (from 153 to 9) (Hastie et al. 2007a). Thirdly, linear regression model was employed to model the credit rating and also to reduce the candidate variables (from 9 predictor ratios to five those were statistically significant). Fourth, the significant variables were used to construct the decision plane for the linear discriminant model using support vector machine classifier (SVM-C) estimation procedure (Scholkopf et al. 1995, Vapnik 1998). Binary response variable was obtained by dividing the ratings into two groups: high rating (or “good companies”) and low rating (or “not so good companies”) by choosing an arbitrary threshold rating value. Finally a logistic regression model helped to define the probability of having high rating (i.e. greater than 5) for a given company. Findings were manifold. The correlation between the Z-score and rating was poor (0.0223 and 0.0133 respectively for manufacturing and service companies). The linear regression models, on the other hand, showed high correlation coefficient (0.64 and 0.71 respectively) between predicted and actual expert ratings. With a few exceptions, in the heavy industry sectors, data was homogeneous (found using predicted residual method). The equation of the new discriminating hyper plane created by the SVM classification model (termed as Investor Inclination Index - I3 model) was proposed which means that expert ratings can be more significantly correlated to a set of candidate financial ratios predicting it. These are: (i) Cost of Goods Sold/Total Operating Expenses, (ii) Earnings Before Interest and Taxes/Total Liabilities, (iii) Earnings Before Interest/Total Revenue, (iv) Retained Earnings/Total Revenue, and (v) Working Capital/Research and Development Expense. The paper contributes by updating the original Altman discriminant model by using a data-driven predictor selection strategy to create a general methodology for building financial models providing economic meaningfulness to the credit rating used for assessing company’s performance and health. A wider use of validated financial models will encourage corporate businesses and even SMEs to evaluate themselves internally thus allowing them to identify possible threats and improve credit rating. Future research aims to provide explicit economic meaningfulness to the individual predictor ratios so that companies can create a strategic resource model (SRM) by interpreting the I3 model to determine how to create a decision support aid for the company’s business management. Also authors aim to extend the contribution by developing a tendency-dynamic status to the financial predictor for incorporating a time-series behavior for pattern recognition. This will provide economic meaningfulness to the financial models; predict financial risk and means to be resilient.

Place, publisher, year, edition, pages
2015. p. 39-40
Keywords [en]
Business Financial Models, Business Failure, Stagewise Regression, Support Vector Machine, Expert Rating
National Category
Probability Theory and Statistics Economics and Business
Research subject
Business and IT
Identifiers
URN: urn:nbn:se:hb:diva-531OAI: oai:DiVA.org:hb-531DiVA, id: diva2:842299
Conference
International Conference on Robust Statistics, Kolkata, India January 12 - 16, 2015.
Available from: 2015-07-17 Created: 2015-07-15 Last updated: 2018-04-28Bibliographically approved

Open Access in DiVA

No full text in DiVA

Authority records

Pal, Rudrajeet

Search in DiVA

By author/editor
Pal, Rudrajeet
By organisation
Faculty of Textiles, Engineering and Business
Probability Theory and StatisticsEconomics and Business

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 580 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • harvard-cite-them-right
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf