Credit Risk Analysis
Credit risk is one of the most significant aspects of financial intermediation activity and it can be defined as the possibility that an unexpected variation of counterparty creditworthiness may generate a corresponding unexpected variation of the current value of its credit exposure.
It should be noted that the word “unexpected” is emphasized in this explanation, and that is because there are essentially two risk components, the Expected Loss and the Unexpected Loss, and only the last one represents the real source of risk. In fact, since it is an ex-ante estimation, the expected loss shall not constitute a veritable risk, but rather a cost expectation that is already included in prudential provisions of the financial intermediary.
In the light of the above, it is a foregone statement to say that in order to minimize unexpected losses it is of paramount importance for the financial intermediary to carry out a careful management of credit risk through a proper assessment of the counterparts and a proper allocation of loans.
As already said, the necessary condition for a risk to occur is that it may be an unexpected variation of a credit rating, and this risk it is not just about the possibility of default of a counterpart, but also a degradation of creditworthiness thereof.
In fact, credit risk is not just about the risk of default, but it is possible to distinguish different types:
- Risk of downgrading and therefore the possibility that the creditworthiness of a counterpart will deteriorate;
- Country risk if the counterpart has its registered office in a country with a high-risk profile;
- Spread risk if, with the same default, the spread and therefore the degree of coverage required increases;
- Risk of exposure if the exposure to a counterpart increases when the default occurs;
- Concentration risk if there is a portfolio with a low level of diversification;
- Recovery risk that occurs if the recovery rate of an exposure turns out to be lower than expected.
Let’s see below how credit risk is measured, and how the use of Big Data, Machine Learning and Alternative Data can positively impact this activity.
Credit Scoring models
Credit risk measurement takes the form of a scoring process to establish the creditworthiness of a counterpart. In fact, scoring models are used to make forecasts regarding the probability or otherwise of default of a counterpart, in order to decide whether to grant a loan or not.
Scoring models assign a score to the counterpart and, as a consequence, a specific rating class. At the basis of the scoring process there are economic-financial assessments, if it is about a company, or socio-demographic assessments in the case of an individual.
However, as constant feature we see the use of historical reports like the analysis of the credit life of an individual or of the financial statements as regards a company. For this reason, decisions are very often made on past events and not on the actual state currently carried out by the counterpart.
Building a scoring model means having a sample of reliable or insolvent behaviour of counterparts and defining those variables that have characterized the default event. But sometimes information concerning the history is lacking or unavailable and therefore the intermediary is in a situation of limited data. Furthermore, continuing changes in the regulations and market fluctuations can make the findings which emerge from the historical analysis far from incisive.
To solve such issues, or at least mitigate their impacts, Big Data and the analysis of alternative data come to our aid. If analyzed and managed correctly, all data present on the web and on social networks allow to get information in advance about changes of the market consequently on possible changes in the level of solvency of a counterpart. The historical data alone is no longer effective if not supported by the analysis of current data.
Alternative Data in Credit Risk Analysis
The use of Big Data and Machine Learning techniques allow to intercept and analyze large amounts and different types of constantly updated data. The application of these techniques to credit scoring models allow to greatly improve their performances.
Furthermore, through the use of Machine Learning, maximum benefit can be obtained from the correct use of information set available, in order to acquire a competitive advantage in optimizing decision-making processes.
In particular, to intercept new information sets and support the analysis of traditional historical data, it is necessary to use Alternative Data, i.e. data obtained from non-traditional sources that can improve credit scoring performance in the analysis and management of credit risk.
Analysis with Alternative Data allows to intercept in advance possible situations or events in the reference field that may impact the solvency of customers or suppliers. In addition, Alternative Data also allow the detection of so-called weak signals, i.e. apparently random or initially irrelevant information which, connected to other information, can be recognized as significant in order to predict the level of solvency of a counterpart.