Retail Credit Score Factor
Contents
Definition
An Retail Credit Score Factor is any indicator, attribute, characteristic that can be used (usually after being converted to a numerical form and possibly in combination with other factors) as input to the construction of a numerical Retail Credit Score as part of an Retail Credit Scorecard building procedure.
These factors typically serve as predictors of borrower credit behaviour, that is, they are are considered as correlated with a potential Credit Event
Conceptual Risk Drivers
There is a wide variety of possible Credit Data sources and it keeps expanding with the availability of Alternative Credit Data. Grouping of data inputs into conceptual categories is useful as an organizing principle. In particular, using the Five Cs Of Credit Analysis which traditionally applies to SME Lending is also useful for the conceptual organization of various possible retail credit factors:
- Capacity, a retail borrower's ability to repay their debt, on the basis of their projected income profile and their other expenditures (including other debt). Exemplified by sourcing and calculating the Debt to Income Ratio
- Capital, the asset base (net worth) of the retail borrower / household and the degree to which it is available to support a given amount of debt
- Character, a borrower's overall behavioural profile towards meeting obligations / repayment of debt, exemplified by their Credit History
- Collateral, any assets the borrower pledges as security for their borrowed funds (when applicable)
- Conditions, referring to the purpose of the loan, its lending conditions (clauses, terms) in relation to the prevailing economic environment
Variable Data Types
The type of variables used as credit score factors can also vary significantly. The following is a list of some relevant dimensions of variability:
- Quantitative versus Qualitative Risk Factors is a major dichotomy.
- Qualitative risk factors will typically unstructured data, captured in text, have an element of subjectivity and in general can be hard to process with quantitative methods
- Quantitative risk factos will be typically structured data, captured in some well defined data type. In turn Quantitative data can be separated into subcategories, e.g.:
- Financial Data (that have traditionally been the backbone of credit analysis) versus Non-Financial Quantitative Data
- Numerical Data (data with a continuous range) versus Categorical Data (data expressed in discrete category choices)
NB: It is important to distinguish qualitative assessments from categorical data. The primarily issue with the former is the potentially uncontrolled Human Judgement involved and the format in which it is delivered (while keeping in mind that numerical or categorical data may also suffer and hide subjectivity)
Other possible classifications of credit score factors that are more relevant at the modelling stage (See also Feature Engineering):
- Using the Absolute Value of factors, versus their ranking in a Peer Group assessment
- Using the Absolute Value versus using Ratios or Percentage
Issues and Challenges
- Deciding on factors relevant and useful for retail credit analysis
- Sensitivity of demographic data and issues with bias / fair lending practices
- Behavioural predictors apparently work well for shorter periods whereas accounting data work better over longer periods.
See Also
- The List of Retail Credit Score Factors has a detailed, annotated catalog of Retail Credit Data that can be used as credit risk factors
- Adverse Action Code
