Credit scoring

from Wikipedia, the free encyclopedia

A credit score (from English to score points, score ) is a numerical value based on a statistical analysis that represents a person's creditworthiness . With credit scoring, companies try to determine the creditworthiness of customers or partner companies more or less automatically according to a predetermined procedure.

In a more general sense, scoring refers to the use of a probability value about a certain future behavior of a natural person for the purpose of deciding on the establishment, implementation or termination of a contractual relationship with this person ( Section 31 BDSG ).

On the basis of borrower characteristics such as “customer since”, “place of residence”, “occupation”, “collateral” points are assigned, these are weighted and then combined into a single credit rating in order to make it easier to grant loans with this overall score. If the credit rating is sufficient, a loan can be granted. However, scores can not only be used to make a credit decision, but also to determine interest rates and credit lines .

The motivation is to avoid risks and to receive objectified decisions based on a statistically supported method. The better the underlying scoring model reflects reality, the fewer loan defaults there will be. Scoring models and the features that flow into them must be constantly updated.

The specific rules and algorithms for awarding and weighting points are called "scorecards", after the sporting term with the same name . There are various techniques for developing suitable scorecards, such as logistic regression , discriminant analysis , artificial neural networks and other data mining methods.

Internal and external scores

Credit scores can be based on a company's own data (e.g. personal master data, credit application data) or take into account external data, e.g. from credit agencies.

Internally determined credit scores do not have to match external ratings . U. different failure probabilities result. There can be many reasons for this:

  • different inputs
  • various information aggregation methods
  • different rating scales
  • Banks create “point in time” ratings internally , ie a default forecast for one year after the assessment date , while external ratings are based on a “through the cycle” approach, ie a default forecast over the economic cycle.

Credit scoring in retail banking

This article or paragraph presents the situation in Germany . Help us to describe the situation in other countries.

Credit scoring is used as a statistical method by credit institutions to carry out a risk classification for private standardized installment loans and small loans . Such loans are usually granted unsecured and are based solely on the personal creditworthiness of the borrower (s) .

A quick credit decision is sought when processing installment loans , although the detailed financial situation of the borrower can only be clarified to a limited extent .

There are personal characteristics (such as occupation, employer, marital status, account maintenance in-house, both positive and negative features of the credit reports) and economic conditions used (disposable income and financial circumstances, expected expenditure). If you have your own customers, you can draw on experience in their customer relationships; The credit officer makes the creditworthiness decision in the traditional form after a personal interview , which is based on a subjective, intuitive assessment, but also provides a holistic impression.

The recorded characteristics are standardized through a point evaluation (credit scoring). Evaluation rules that classify the data to be collected and assign a point value (the score ) can be laid down in various procedures. In addition to stand-alone applications, table processing programs or paper-based process descriptions are common.

Admissibility according to § 28b BDSG

On April 1, 2010, an amendment to the Federal Data Protection Act (BDSG) came into force, according to which scoring ( Section 28b BDSG) is permissible for the purpose of deciding on the establishment, implementation or termination of a contractual relationship with the person concerned if the data used is scientifically based recognized mathematical-statistical procedures are demonstrably significant for the decision, in the case of credit agencies the transmission of the data used would be permissible, address data is not exclusively used for the calculation and, in the case of the use of address data, the person concerned has been informed beforehand about the use of this data .

Information according to § 34 BDSG

Since April 1, 2010, companies that use scoring have been obliged , according to Section 34 (2) BDSG, to provide the person concerned with information on the probability values ​​determined in the last 6 months, the types of data used for the calculation and the occurrence and significance of the probability values ​​on a case-by-case basis To be comprehensible in a generally understandable form.

According to Section 34 (4), credit bureaus are obliged to provide relevant information about the probability values ​​transmitted to third parties in the last 12 months.

Situation according to the General Data Protection Regulation

The General Data Protection Regulation ( GDPR ) was passed in April 2016 . Since May 2018, this has been in force in all EU member states without any further implementation in national law. Changes to the nation states are only possible to a limited extent via so-called opening clauses. With regard to scoring, the GDPR will significantly lower the current German data protection in some areas. For example, the obligation to provide information is restricted, geoscoring is permitted again, and disputed claims can be transmitted again. The latter could force consumers to pay bills even in the case of poor or missing services in order to avoid the disadvantages of a negative score.

At the same time as the General Data Protection Regulation came into force, the supplementary national regulation for scoring came into force in Germany with Section 31 of the new Federal Data Protection Act.

Credit scoring from Schufa

The SCHUFA offers its customers since 1997, together with the credit information on individual consumers a score value based on the stored in their data. This is a value from 1 to 100, which is assigned to the respective consumer and indicates the probability of a loan default. The lower the value, the greater the probability of failure. The score value depends on the purpose for which it is requested - for example, insurance companies receive different score values ​​than mobile network providers. Every consumer can prohibit the transfer of scores to his person at Schufa. It is unclear whether this application will have a negative impact on a later credit decision. Since the beginning of 2007, you can see your own basic score value in% values ​​in your own information (online).

The SCHUFA industry scores were revised in 2001. The scoring method is based on the logistic regression model , which models the probability of occurrence of a random event with two possible outcomes. For the 2001 procedure, around 6.7 million anonymised data records were evaluated over a “maturity period” of 15 months. There are 7 different types of industry score. These are: mortgage bank, mail order, trade, telecommunications, cooperative banks and savings banks, banks and the Schufa business line.

Since April 1st, 2010 consumers have been able to use according to § 34 Abs. 4 BDSG a. F. receive information on the historical probability values ​​- that is, SCHUFA scores that were reported to SCHUFA contractual partners within the past 12 months.

The features that the SCHUFA takes into account include actual, for example, entries about outstanding credits and about dunning or enforcement notices that have become legally valid , but also statistical scoring.

SCHUFA keeps the exact calculation formula of its scoring system under lock and key and has so far opposed all requests to disclose it.

Up until 2001, obtaining personal information was included in the scoring as a negative feature; After massive protests, the Schufa stopped this practice according to its own information.

A risk classification is determined from the machine-determined data requested by the lender and the credit decision is prepared.

Credit rating in corporate banking

In the corporate area, the economic data is analyzed in greater detail; The focus here is on the analysis of the annual financial statements for information processing and evaluation. Statements on trends are made and qualitative, future-oriented factors are taken into account (e.g. the potential of human capital ). A rating is then carried out. As a result, the risk factors considered by banks are similar to those of the major rating agencies. They take into account the financial situation, market position and management quality. A long-term relationship with the borrower ( house bank relationship ) can give the banks an information advantage over rating agencies that only have external information.

Advantages and disadvantages

The credit scoring model has advantages and disadvantages compared to conventional methods:

advantages

  • Standardization , personal preferences of the loan officer are switched off
  • Empirically validatable (objectively comprehensible)
  • EDP-technical refinement possible
  • For the lenders (who do not necessarily have to be banks, since the process can also be used for goods financing), the loan decision process becomes more economical through automation.
  • Accelerate the credit decision
  • Time and cost savings

disadvantage

  • The personal experience of the loan officer is not included. A long-term business relationship with the borrower is often an information advantage. The loan officer decides on his holistic impression of the borrower. However, some technical scoring solutions take such data into account.
  • The data may be problematic ( data protection )
  • The decision may be made on the basis of outdated or incorrect data (data quality)
  • Passing on or trading with data is possible
  • Query without customer consent
  • insufficient consideration of qualitative personal data
  • constant update necessary

costs

There are two types of costs to consider in connection with the credit check:

  • Costs from error 1. Type: Lending to customers with poor credit ratings, resulting in a loan default.
  • Costs from the error type 2: no lending to borrowers with good credit ratings, lost interest income

See also

literature

  • Kerstin Dittert: Scoring: The look into the crystal ball (= Digitization series of publications. Volume 1). epubli, Berlin 2017, ISBN 978-3741884849 .
  • Andreas Henking, Christian Bluhm, Ludwig Fahrmeir : Credit Risk Measurement. Statistical basics, methods and modeling. Springer, Berlin et al. 2006, ISBN 3-540-32145-4 .
  • Rothmann, Robert; Sterbik-Lamina, Jaro; Peissl, Walter (2014): “Credit Scoring in Austria”; Institute for Technology Assessment of the Austrian Academy of Sciences (ITA / ÖAW); Study on behalf of the Federal Chamber of Labor (AK Vienna); ITA project report no .: A66. ISSN  1819-1320

Web links

Individual evidence

  1. ↑ Scoring: Preventing a step backwards in data protection | VZBV. Retrieved February 1, 2017 .
  2. ↑ FinanceScout24: Schufa score: Do you know your score & the meaning? In: FinanceScout24 . May 7, 2015 ( financescout24.de [accessed November 14, 2016]).
  3. ↑ R. Hüls, A. Henking (2003): “With Scoring to More Earnings” in “Bank und Markt”, issue 03/2003