Opportunities and challenges

"A ship is safe in harbour, but that's not what ships are built for" is a quote attributed to the American writer and author John August Shedd. Similarly, companies are safest when they are not engaged in entrepreneurial activities that involve various inherent risks. However, this contradicts the original purpose of companies.

One of these inherent risks is credit risk. It comprises "[...] the financial effects resulting from the fact that business partners do not fulfil their agreed services, do not fulfil them in full or do not fulfil them on the agreed dates." (Positionspapier VdT 2023, S.16). Managing credit or counterparty risk is one of the central tasks of a treasury organisation (see VdT 2023 position paper, p.5 & p.16) and, according to a recently published KPMG study, one of the biggest concerns in a multi-crisis environment that challenges many treasurers today (see Resilient Treasury April 2024, p. 6).

The focus is often on managing credit risks vis-à-vis financial institutions, for example in the form of deposits or the fulfilment of contractual obligations for derivative financial instruments. The management of "classic credit risks" from deliveries and services, on the other hand, is often a sub-area of the accounting department or, in mature organisations, even a separate department within the finance division. In both cases, the risk strategy should set the framework for measuring and analysing risks, define the criteria, control and monitoring mechanisms that are relevant to decision-making and specify reporting and escalation formats and channels.

Before any decision is made on a new contractual relationship, the risk must be analysed, which is often expressed as the probability of default of the contractual partner in conjunction with the exposure and the loss given default of the contractual partner. This risk analysis must be regularly updated in the ongoing business in the form of monitoring.

Consequently, the credit risk analysis serves both as a starting point for the credit risk process and as a recurring activity in the context of ongoing monitoring. At the centre of this ongoing analysis is the collection, processing and evaluation of relevant data. It is crucial to select the right data and sources for the models to be used in each case. These models, in turn, must also be constantly checked for up-to-dateness and precision and recalibrated if necessary. In this context, however, the availability of suitable data, the risk position and, in some cases, the available budget also often determine the model to be used.

Source: KPMG AG

The following remarks primarily relate to the determination of the probability of default.

In the simplest case, data provided by rating agencies can be standardised to a uniform measure with the aid of translation tables. For example, a Baa rating from Moody's corresponds to a BBB rating from Standard & Poor's or Fitch and harbours a moderate credit risk (https://www.boerse-frankfurt.de/wissen/wertpapiere/anleihen/rating-matrix) with an average probability of default of 1.9% p.a. (https://www.deltavalue.de/credit-rating/). It should be noted that ratings are generally prepared for capital market-oriented companies or companies that want to present themselves to a mostly international investor audience.

Credit agencies or credit insurers can act as information service providers for small or medium-sized companies. Here too, the data is provided in various formats, for example a Creditreform creditworthiness index of 208 corresponds to an annual probability of default of 0.12% (https://www.creditreform.de/loesungen/bonitaet-risikobewertung). A Dun & Bradstreet Failure Score between 1510 and 1569, on the other hand, means an average default risk of 0.09% per year (https://www.dnb.com/resources/financial-stress-score-definition-information.html). Regional expertise should also be taken into account when deciding on a data provider. Schufa, for example, has a high level of market coverage in Germany, while Kreditschutzverband is the leader in the Austrian market (https://www.finanz.at/kredit/kreditschutzverband/).

Advanced approaches use scorecards in which various criteria, such as balance sheet ratios or macroeconomic factors, are weighted and summarised into a probability of default using a corresponding credit risk model. In addition to the service providers mentioned above, possible sources for this data are the respective commercial registers. These provide annual financial statements or balance sheets. However, it is also conceivable to connect market data providers such as Refinitv or Bloomberg and, for example, take into account interest rates, CDS spreads, unemployment figures or inflation figures in the analysis, which can either provide direct information about the company or about the economic framework in which it operates.

Recently, additional data sources have also come into focus, such as content in social media and other publicly available information on the internet. However, there is also a risk of obtaining a subjectively distorted picture that may have been created on the basis of herd effects, for example if many people indiscriminately subscribe to an opinion that is then interpreted as relevant or correct and is not necessarily rational. On the other hand, information could also be available here that indicates risks far earlier than is the case with conventional data sources, which are often published with a time delay. The insolvency of Silicon Valley Bank is a good example of this. On 9 March 2023, before the start of trading, investors were already intensively exchanging information on Twitter, presumably fuelling the bank run that caused the bank to collapse on 10 March 2023 (https://fortune.com/2023/05/02/did-twitter-cause-silicon-valley-bank-run-financial-panic-economists-research/).

The valuable company information that can be incorporated into the analysis, such as the historical payment behaviour of customers, should also not be underestimated. The challenges here often lie in the insufficient availability of data, which is often due to fragmented IT landscapes. Another question arises with regard to interpretation. If a counterparty has never paid late or even defaulted, can this counterparty be considered risk-free?

In light of rapidly advancing technological developments, modern approaches integrate artificial intelligence into the analysis process. However, it is necessary to create the right conditions for this. Artificial intelligence must be adequately "trained" using data. A high level of data quality must be ensured, regardless of whether the data is available in-house or obtained externally. Common problems are: Inconsistency, inaccuracy, incompleteness, the presence of duplicates, obsolete information, irrelevance, lack of standardisation and others (cf. https://www.forbes.com/sites/garydrenik/2023/08/15/data-quality-for-good-ai-outcomes/). The first approaches for tools are already providing a remedy by using artificial intelligence to perform quality assurance and data cleansing (https://blog.treasuredata.com/blog/2023/06/13/ai-ml-data-quality/) Large language models have also found their way into credit management, for example in the processing of companies' annual financial statements. They already provide employees with relevant information in compressed form or feed default probability models directly with the necessary data for further processing.

In addition to quality, a sufficiently large number of data points is required to increase the informative value of a forecast produced by an AI model. However, even with sufficient data quality and quantity, there is still the risk that correlations in the historical data are only partially captured, for example the supply chain disruptions caused by the container ship "Ever Grand" running aground in the Suez Canal in March 2021, the ECB's historic interest rate turnaround at the beginning of 2022 or the Houthi attacks on trade routes in the Red Sea, which lead to monthly additional costs in the double-digit million range for Hapag Lloyd alone (https://www.wiwo.de/politik/ausland/rotes-meer-diesen-einfluss-hat-der-konflikt-mit-den-huthi-auf-unternehmen/29597872.html). Such events and their consequences reinforce the multi-crisis environment mentioned in the KPMG study. Here, the utilisation of data from news and social media offers a real opportunity to balance out this historical distortion (https://fastercapital.com/de/inhalt/Trends-bei-der-Kreditrisikoprognose--So-bleiben-Sie-mit-Trends-und-Innovationen-bei-der-Kreditrisikoprognose-auf-dem-Laufenden.html), as the machine can analyse huge amounts of data very quickly and recognise potential correlations that remain hidden to humans or cannot be identified at the same speed. 

The use of artificial intelligence therefore offers advantages, especially when processing mass data. On the other hand, excessive use can lead to blind trust in the technology, although the phenomenon of AI hallucinations has already attracted a great deal of attention and has led to data protectionists filing complaints against Open AI (https://www.zeit.de/digital/datenschutz/2024-04/datenschutz-chatgpt-open-ai-beschwerde). There is also a risk that (sometimes unconscious) prejudices contained in the source data may have an undesirable influence on the result. For example, companies that are currently experiencing a strong media presence could be favoured, although this does not necessarily correlate with creditworthiness. To prevent this, it is advisable to always have AI provide transparency about how decisions were made and which sources were analysed. This continues to pose major challenges for current models and approaches.

Regardless of the type of model used, it must always be checked whether it is still up to date and delivers reliable results, especially in the case of non-linear, complex and unique framework conditions. For example, the expected default risks of many borrowers increased during the COVID pandemic due to supply bottlenecks or impairment of the business basis, for example in the catering industry. However, this development was partially counteracted by adjustments to German insolvency law and intensive financial measures. This experience shows that even short-term, very comprehensive model adjustments may be necessary in order to adequately assess the current reality.

The models used should therefore always be powerful, stable, rational and efficient. Combinations of statistical models and artificial intelligence could contribute to an improvement in credit risk forecasting in the future. In any case, the benefits predicted from the use of the models must be subjected to a comprehensive cost analysis and the framework conditions in the company itself and the market environment must be taken into account.

Those: KPMG Corporate Treasury News, Issue 145, July 2024

Authors: Nils Bothe, Partner, Finance and Treasury Management, Corporate Treasury Advisory, KPMG AG
Philipp Knuth, Senior Manager, Finance and Treasury Management, Corporate Treasury Advisory, KPMG AG