Chapter 6- Risk management

Página 1

Credit risk management

5 ways to minimise credit risk:

screening

monitoring

credit rationing

use of collateral and endorsement

diversification

SCREENING

To overcome adverse selections in loan markets banks have to screen out bank credit risks from the good ones in order to increase the shareholder's value. To do this banks acquire info on a potential borrow to measure its credit risk (defaulting)

At retail level most info are collected internally or externally from credit agencies

There are several models to assess risk

Qualitative model

Expert systems is based on the subjective judgement of the manager on the probability of default on the borrower. They are used in the absence of publicly available information on the quality of the borrower. Information can be collected from internal or external sources.

The quality of info used is determined by the size of loan exposure and cost of collecting info. The key information in the overall credit decision can be divided into borrower-specific factors which are idiosyncratic to each borrower. such a reputation arising from long term borrowing-lending relationship, borrower's leverage because in the presence of large amounts of debt, the default is higher due to an increase in the borrower's interest charges which has an adverse impact of the borrower's cash flow effective the volatility of borrower's earnings which in turn is negatively linked on the borrowers ability to meet interest charges and principle charges.

Market specific factors such are the borrower's business cycle in that time, level of interest rates, monetary policies may affect the decision of the borrower's credit decision. Overall it may encourage borrowers to take excessive risks which only encourages risky borrowers to apply.

Credit scoring models

They are mathematical models used to observed he characteristics of the loan's applicant. To either score the probability of default or to class borrowers in different risk default classes.

Credit scoring models have two broad families- linear probability and linear discrimination analysis.