Modelling Credit risk for SMEs: Evidence from the US market

Original by Altman, Sabato, 2007, 43 pagesHamster_gagarin_linkedin
hamster writter This summary note was posted on 16 January 2017, by in Credit risk Finance #, #
  • Definition of SME in Basel II is sales less than $50 mio
  • Acknowledge that analysis can still be improved using qualitative variables such as number of employees, legal form of the business, region where main business is taking place  (pointed out by Lehmann (2003) et Grunnet et all (2004))
  • Beaver (1967) used unviariate analysis on 158 firms
  • Use logit regression and compare to equivalent MDA on 1890 firm (120 defaults) from 1994-2002
  • Chen and Shimerda (1981) show that out of more than 100 financial rations almost 50% were found useful in at least one empirical study
  • In some statistical studies, criticism of the forward stepwise selection procedure has been raised as it can yield theoretically implausible models and select irrelevant variables. For this reason, they used a two-step approach, first choosing the most relevant variables for the study and then applying the stepwise selection procedure
  • High variability of financial ratios SME can be due to financial sector they operate in
  • Use logarithmic transformation for all five selected variables in order to reduce the range of possible values and increase the importance given by each one of them
  • Use Wald test to see if each of the predictor is statistically significant
  • Variables used: 1) Ebitda/total assets (ln(1- Ebitda/total assets), 2) short term debt/equity book value and its ln, 3)Retained earnings/Total assets (1-Retained earnings/total assets), 4) cash/total assets and its ln, 5) Ebitda/interest expenses and its ln
  • Validate results on 26 defaults SME for 432 firms from 2003-2004
  • For the Zā€™ā€™ MDA score a score is calculated. Then 30% of the sample with the lowest score is considered rejected in order to check for accuracy of the model to correctly and incorrectly classify the firms (as default and non-defaults) between accepted and rejected clients. The choice of a fix cut-off at 30% is not based on any specific reason
  • Points out that at contrario Nash et all (1989) find that prediction of default is insensitive to the selection of accounting variables and modelling technique.