3. Problem loans and early warning systems
In order to prevent borrowers bankruptcy, banks developed
insolvency-forecasting models.
Altman (1968) was the first to design an
insolvency-forecasting model based on multiple linear discriminant
analysis. He studied five financial ratios of 66 American companies and built
a Z-score function to forecast the defeasance of a company. He
studied 22 ratios (liquidity, solvency, gear ...) taken from the most
frequently used by American banks to assess the creditworthiness of
companies.
His model was as follows:
Z = 0.012 X1 + 0.014 X2 + 0.033 X3 + 0.006 X4 + 0.999 X5
where X1 = Working capital / Total of assets
X2 = Reserves / Total of liabilities
X3 = EBIT9 / Total of assets
X4 = Market value of shares / Total of debts
X5 = Turnover / Total of assets
The forecasting model was reliable at 80 % for only one and
sometimes two years. Above that timeline the model became reliable only at
40%.
Another study on a sample of 111 companies by Altman, Haldeman
and Narayanan (1977)
led to the Zeta model. The Altman model was improved later by
Conan and Holder (1979)
and by the Banque de France (1983), which build models specific
to industry sectors.
Such models are highly reliable since they are built from
companies operating in the same activity sector and have similarities of
financial structure.
9 Earning Before Interest and Taxes
14 MBA in Banking and Finance
Other methods were used to forecast insolvencies such as logistic
regression (Boisselier and
Dufour); a neuronal approach was also developed (Beauville and
Zollinger, 1995).
Most of these models were assuming that a scanning of
a company's financial statements three to five years before could
predict its failure. Other authors criticized these models arguing that
other non-financial information were relevant in credit rating models (Grunert,
Norden and Weber, 2002) and that banks were reluctant to let them
drive by a mechanist model (Treacy and Carey, 1998).
Despite all the critical notes, early warning systems are
still growing since the Basel 2
Committee encouraged banks to build Internal Ratings-Based (IRB)
systems. Those systems
are based upon the banks' own estimations of credit
risk. The risk components include measurements of the probability of
default (PD), loss given default (LGD) and exposure at default (EAD). It is
important to underline that those rating systems now integrate both
quantitative and qualitative information and even if they cannot eliminate
problem loans, they remain a key and relevant tool of decision-making.
All these models were applied on banks lending activities to
provide a solution to the issue of problem loans. Unfortunately, banks
still face problem loans. Information asymmetry will always exist since a
borrower's financial soundness can be affected by an external shock that may
occur (Minsky, 1982 & 1985). Interest rate has appeared to be an
ineffective screening device as insolvency forecasting models and IRB systems
only provide a probability of failure
but cannot ensure whether the failure will occur. Williamson's
conclusion that the optimal contract between a lender and a borrower is a debt
contract and the lender only monitors in
the event of default can also not avoid problem loans and loan
losses.
| 
