A banking company wants to build a neural network to predict who will default on

Question

A banking company wants to build a neural network to predict who will default on 30-year fixed-rate home mortgage loans. Historically, approximately 2.5% of individuals default. Given the small percentage of defaulters, what are some of the problems that may be encountered when fitting a neural network model? Is this a problem specific to neural networks, or is this a problem other modeling techniques have as well? What are some of the solutions that can be implemented to overcome the insufficient minority class problem? Provide two or three examples.

Explanation / Answer

BANKRUPTCY prediction has long been an important and widely studied topic. The main impact of such research is in bank lending. Banks need to predict the possibility of default of a potential counterparty before they extend a loan. This can lead to sounder lending decisions, and therefore result in significant savings.

Neural Netowk for bankruptcy is still active. NNs have generally outperformed the other existing methods. Currently, several of the major commercial loan default prediction products are based on NN.

In spite of the sucess of NN models, there are a number of open issues that should desirable be addressed. Even though a prediction of the default event is by itself very useful, an estimate of the default probability is very desirable. Typically bank have several prediction systems in place. They make a lending decision based on the combination of these predictions. They make a lending decision based on the combination of these prediction. Having a probability of default rather than a prdiciton of default is valuable for them. Even though there are some objective function measures that achieve that, such as cross-entropy error functio, our experience with this objective function has not been vavorable.

The other open issue is to consider macroeconomic indicators as inputs to the NN. The prevailing economic conidtions can have a significant effect on the probability of bankruptcy. There are very few studies that consider these factors in conjection with NN models. This should therefore be a recommended study.

Credit risk has been the subject of much re-search activity, especially after realizing its practical necessity after a number of high profile bank failures in Asia. As a result, the regulators are acknowledging the need and are urging the banks to utilize cutting edge technology to assess the credit risk in their portfolios. Measuring the credit risk accurately also allows banks to engineer future lending transactions, so as to achieve targeted return/risk characteristics.

The traditional approach for banks for credit risk assessment is to produce an internal rating, which takes into account var-ious quantitative as well as subjective factors, such as leverage,earnings, reputation, etc., through a scoring system. The problem with this approach is of course the subjective aspect of the prediction, which makes it difficult to make consistent estimates. Some banks, especially smaller ones, use the ratings issued by the standard credit rating agencies, such as Moody’s and Standard & Poor’s. The problem with these ratings is that they tend to be reactive rather than predictive (for the agencies to change a rating of a debt, they usually wait until they have a considerably high confidence/evidence to support their decision). There is a need, therefore, to develop fairly accurate quantitative prediction models that can serve as very early warning signals for counterparty defaults.

Formalized rganizational procedures can mitigate discrimination by limiting individual discretion. The case of the military (Moskos & Butler 1996), for example, and the public sector more generally (DiPrete & Soule 1986, Moulton 1990) provide examples in which highly rationalized systems of hiring, promotion, and remuneration are associated with an increasing representation of minorities, greater racial diversity in positions of authority, and a smaller racial wage gap. Likewise, in the private sector, formal and systematic protocols for personnel management decisions are associated with increases in the representation of racial minorities (Reskin et al. 1999, Szafran 1982, Mittman 1992), and the use of concrete performance indicators and formalized evaluation systems has been associated with reductions in racial bias in performance evaluations (Krieger 1995, Reskin 2000).

Individual discretion has been associated with the incidence of discrimination in credit markets as well. For example, Squires (1994) finds that credit history irregularities on policy applications were often selectively overlooked in the case of white applicants. Conversely, Gates et al. (2002) report that the use of automated underwriting systems (removing lender discretion) was associated with a nearly 30% increase in the approval rate for minority and low-income clients and at the same time more accurately predicted default than traditional methods. These findings suggest that formalized procedures can help to reduce racial bias in ways that are consistent with goals of organizational efficiency.
At the same time, increased bureaucratization does not necessarily mitigate discriminatory effects.There is evidence that formalized criteria are often selectively enforced, with greater flexibility or leeway applied in the case of majority groups (Wilson et al. 1999, Squires 1994). Likewise, indications of racial bias in performance evaluations cast doubt on the degree to which even formalized assessments of work quality can escape the influence of race (McKay & McDaniel 2006). The degree to which formalization can reduce or eliminate discrimination, thus, remains open to debate, with effects depending on the specific context of implementation.

Taking a broader look at race-targeted employment policies, Holzer & Neumark (2000) investigate the effects of affirmative action on the recruitment and employment of minorities and women. They find that affirmative action is associated with increases in the number of recruitment and screening practices used by employers, increases in the number of minority applicants and employees, and increases in employers’ tendencies to provide training and formal evaluations of employees. Although the use of affirmative action in hiring is associated with somewhat weaker credentials among minority hires, actual job performance appears unaffected.

Related Questions

Navigate

• Browse (All)
• Browse A
• Subjects

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.