11:03 PM Credit scoring - Oxford Scholarship | ||||
#credit scoring #2.1 IntroductionA sound credit risk management is built upon a good-quality portfolio of performing assets. The pricing of the loans has to reflect the risk. A good selection strategy aims to avoid high losses. Credit scoring is a credit risk management technique that analyzes the borrower's risk. In its early meaning, “credit scores” were assigned to each customer to indicate its risk level. A good credit scoring model has to be highly discriminative: high scores reflect almost no risk and low scores correspond to very high risk, (or the opposite, depending on the sign condition). The more highly discriminative the scoring system, the better are the customers ranked from high to low risk. In the calibration phase, risk measures are assigned to each score or score bucket. The quality of the credit scores risk ranking and calibration can be verified by analyzing ex-post observed credit losses per score. Credit scores are often segmented into homogeneous pools. Segmented scores are discrete risk estimates that are also known as risk classes and ratings. Ratings will be discussed in the next chapter. In the past, credit scoring focused on measuring the risk that a customer would not fulfill his/her financial obligations and run into payment arrears. More recently, credit scoring evolved to loss and exposure risk as well. Scoring techniques are nowadays used throughout the whole life cycle of a credit as a decision support tool or automated decision algorithm for large customer bases. Increasing competition, electronic sale channels and recent banking regulation have been important catalysts for the application of (semi-) automated scoring systems. Since their inception, credit scoring techniques have been implemented in a variety of different, yet related settings. A first example is credit approval. Originally, the credit approval decision was made using a purely judgmental approach by merely inspecting the application form details of the applicant. (p.94) In retail, one then commonly focused on the values of the 5 Cs of a customer [133. 475 ]: Character: measures the borrower's character and integrity (e.g. reputation, honesty, …) Capital: measures the difference between the borrower's assets (e.g. car, house, …) and liabilities (e.g. renting expenses, …) Collateral: measures the collateral provided in case payment problems occur (e.g. house, car, …) Capacity: measures the borrower's ability to pay (e.g. job status, income, …) Condition: measures the borrower's circumstances (e.g. market conditions, competitive pressure, seasonal character, …). Note that this expert-based approach towards credit scoring is still used nowadays in credit portfolios where only limited information and data is available. The early success of application scorecards drew the attention of the academics and researchers to develop advanced statistical and machine-learning techniques that apply a wide range of explanatory variables or characteristics. An application scorecard then assigns subscores to each of the values of these characteristics. These subscores are determined based on the relationship between the values of the characteristics and the default behavior, and are aggregated into one overall application score reflecting the total default risk posed by the customer. An example of an application scorecard is given in Table 2.1. Consider a new application of a customer with age 35, income 12,000, and residential status with parents. Given the above scorecard this customer is assigned 330 points. These points are then compared against a cut-off and a credit decision is made. For example, when assuming a cut-off of 300 (400), the above loan is granted (rejected). When the score of a customer is close to the cut-off, it may be an indication that the scorecard is very unsure as to whether to label the customer as good or bad. This is why one can define a grey zone around the cut-off, which will require further (human) investigation for customers falling into that region. This chapter is organized as follows. section 2.2 discusses the use of scores during different stages of the customer cycle, while section 2.3 compares scoring functions based on their characteristics concerning risk type, risk entity and the score source. Credit bureaus are a popular external reference source for scoring and are discussed in section 2.4. The concept (p.95) Table 2.1 Example application scorecard: a customer with age of 35, income of 12,000 and residential status with parents is assigned a total score of 120 + 140 + 70 = 330 points.
| ||||
|
Total comments: 0 | |