Europe's banking system is not as robust as it may seem

08 October 2018

The financial crisis saw almost the complete meltdown of the global financial system, requiring considerable state interventions to stave off the crisis. While recovery has picked up in Europe, the European banking system is not as robust as it may seem, warns a new report, mounting concerns that current non-performing debt levels could see another crisis unfold.

The financial crisis of 2008 unleashed a severe economic downturn, creating headaches for states as many financial institutions – including Lloyds and RBS in the UK, ING in the Netherlands, and Bankia in Spain – required state backing to stay afloat. The aftermath of the crises saw considerable regulation introduced, with the idea that better risk & compliance and oversight of financial industry players would lower the odds of a future crisis, as well as structures to wind down failed banks and poorly performing loan books.

The effect of the crisis saw many European banks hobbled by debt, with various countries – particularly Greece, Italy and Spain, continuing to note particularly high levels of non-performing loans (NPL) on their loan books.

NPLs in the EU by country

A decade down the line, Europe has since seen its economy recover, with GDP growth on the rise. Greece was the hardest hit during the crisis – the country only recently returned to positive GDP growth. Space and Italy too saw challenging years during the start of the decade, with Italy returning to >1% growth last year, while Spain rebounded sharply in 2015, noting three years of >3% growth. Germany and France have booked positive growth since 2012, although growth has been slow to increase above 1% in France, while Germany returned to almost 2% growth in 2014. Across the EU28 as a whole, growth managed to return to above 2% in 2015, and topped 2.4% last year.

Are banks ready for the next crisis?

Yet crises are by no means over – and the next one could strike at any time, warn economists. In a new report by Oliver Wyman, the authors consider to what extent financial institutions are prepared for a new crises and explore possible tactics they can leverage to avoid future none performing loans from hobbling their books. The report, titled ‘Is Europe Ready for the Next Crisis?’, was conducted by the firm’s Corporate Restructuring service line.

The authors highlight that while growth levels are now positive, debt levels have remained high globally. Debt as a % of global GDP increased from around 195% prior to the crises to a peak of 215% at the height of the crises. However, while dropping in the years to 2014, debt again rose since then, reaching 227% in 2016.

Global Debt

Meanwhile, in Europe’s banking sector, non-performing loans as a % of gross loans are understandably high in crisis hit Greece, but relatively lower in Italy and Spain, at 11.8% and 4.8% respectively. Across Europe as a whole, non-performing loans stood at 4.2%. “The European banking system still suffers from a heavy burden of problem loans. The overall ratio of “non-performing loans (NPL)” has still not fallen below the pre-crisis level, and some countries still carry double-digit NPL ratios,” said Lutz Jaede, Head of Corporate Restructuring at Oliver Wyman.

The study finds that some banks continue to hold the hangover from the previous crises, create risks going forward if another crises at the magnitude of 2008’s event hits the landscape’s frontiers. Being prepared for the next crises is a key aspect of wider oversight currently being implemented in the financial system in Europe, including more stringent capital requirements and elaborate stress testing on economic scenario’s. Banks have further become more bullish in their lending, with criteria for loans becoming less stringent in the past number of years as the new state of affairs matures.

Loan restructuring

Banks have also changed tack when it comes to restructuring if a loan looks like it is going sour; the survey found that banks are more likely to monitor debtors at risk (58%), while special task forces have been developed to restructure if a company appears to be failing to meet its requirements (74%). Banks are also keen to sell on NPF earlier than in the pre-crises years. These and a number of other measures, aim to reduce the risk of loans at the individual level – however, if a structural downturn hits, questions about bank preparedness remains a pressing one.

What instruments do you have in place in order to prepare for a crisis?

According to the respondents surveyed by Oliver Wyman – banking executives and restructuring officers – the overall capability of banks with regards to steering instruments required to monitor and act upon crisis situations is significantly more mature compared to ten years ago. Most of the surveyed managers stated that they have implemented good reporting on profitability and operational performance and that they use a rolling liquidity forecast in combination with a mid-term planning to plan the company’s development. Jaede however placed a side note on confidence, stating “these instruments are most appropriate for managing a steady state but do not anticipate major shifts or long-term challenges.”

Concern that a downturn could affect one market, or the whole market, continue to simmer – largely due to geopolitical uncertainties, cyclical fundamentals, as well as key latent risks that could be triggered by political or environmental conditions turning south. The global consulting firm noted that one way to reduce market risk is for improved transparency from debtors about their fundamentals – allowing banks to move earlier to mitigate losses to the company and the wider economy in a downturn.

Collaboration between banks and corporates to avoid a crisis

The researchers further highlighted that banks in general are not happy with the level of collaboration across key metrics that can be used to model both market collapses, as well as company performances – something with which corporates stand at considerable odds to in terms of their self-judged level of collaboration.

Jaede: “In order to avoid a crisis, it is essential that banks and corporates collaborate closely with regards to early identification of a potential crisis event, transparency on a company’s situations and performance, timely information on financing needs, and proper management and coordination of different financing partners. Our survey shows that banks are not pleased with the performance of their debtors in these areas. The corporates, however, do not feel a significant need for improvement here.”

Key differences are noted in terms of proactive communication in case of extraordinary incidents, with a score of 2.7 for banks (with five as ‘managed very well’) and 4.0 from corporates. Timely and constructive dialogue on upcoming financial needs too saw divergent scores of 3.2 – 3.9.

Concluding, lead author Jaede said that while the current mood is good driven by solid economic development, the European banking system is not as robust as it may seem. “Executives feel confident in the face of crisis – but at Oliver Wyman, we have our concerns.”

Related: Banks turn to investors as means to improve restructuring of NPLs.


AI can improve operational risk management in banking

17 April 2019

Risk management is an integral part of banking. By taking financial risks, banks are able to generate the profits that are necessary to survive. Risk management aims to control this process by making potential losses more predictable. This makes the bank more robust to external fluctuations.

Whereas profits can be made by accepting certain financial risks, operational risk is intrinsically different. Operational risks only cause losses – financially in terms of bottom-line impact and non-financially in the form of for instance reputational damage. The consequences of operational risk events can have a large impact on an organisation and the financial system as a whole as experienced during the last financial crisis. It is not surprising therefore that operational risk is receiving more attention within the financial sector, with banks trying to minimise the operational risks they take, given the resources available while keeping in mind the strategic goals of the organisation.

In the past decade, there has been major progress in the development of artificial intelligence (AI). AI algorithms excel at data analysis and have evolved to the point where they surpass human performance for a wide variety of tasks. More and more businesses exploit these technological advances to optimise different kinds of processes such as marketing, sales and e-commerce, manufacturing and logistics. In today’s growing data-driven world, this trend is expected to continue on the back of widening opportunities for use cases, including in the area of operational risk management.AI can improve operational risk management in banking

Challenges in operational risk management

In June 2011 the Basel committee published the Principles for the Sound Management of Operational Risk (BCBS), which provides a framework for the development of proper operational risk management. Three years later, a survey was conducted to measure to what extent banks complied with these principles. One of these principles states that banks should write a risk appetite and tolerance statement. Banks reported that this is more challenging for operational risk than for other risk categories and attributed this to the nature and pervasiveness of operational risk. The banks that did comply with this principle often reported the use of backward-looking metrics of operational risk, such as operational losses as a percentage of gross revenue.

The above example underlines the challenges which banks face in the management and measurement of operational risk. Compared to financial risk, operational risk is a more qualitative field of study. Whereas financial risk management has been the main priority of banking for a longer time, operational risk management is much younger resulting in less extensive historical data. Predictive modeling becomes more of a challenge in this situation.

On top of that, the events in operational risk are much more diverse in scope. The Basel committee defines operational risk as the “risk of loss resulting from inadequate or failed internal processes, people and systems or from external events”. Internal fraud, data leakage and reputational damage are very different problems, yet can be very closely related as well. A lot of the data in operational risk consists of textual input which contain qualitative information. The qualitative nature of operational risk is reflected in the Basel framework, which encompasses guidelines for organisational structures, culture and awareness, and qualitative reporting.

The computer as a reader

Artificial intelligence could play a valuable role in upgrading operational risk practices, with in particular machine learning – the field of self-learning computer algorithms – showing promise. Machine learning algorithms can make predictions based on data fed to the algorithm. Recently, major progress was made in the field of natural language processing (NLP). NLP focuses on using textual data for predictions. As an example, an algorithm can learn to rate hotel reviews. By processing large amounts of reviews together with their given ratings, the algorithm can learn to give a rating to a new review it has never seen before.

“There are many opportunities for operational risk management to exploit AI and other related technological advances.”
– Lars de Ruiter and Matthias Geerse, Solid Professionals

In operational risk, many textual reports are written regarding specific risks or risk appetite of the organisation as a whole. These reports are generally written by risk managers – experts in their field. Assuming that risk reports hold information which is absent in historical loss data, this information could be extracted and used for predictive purposes. So why should computers perform this task instead of risk experts? The answer lies within the physical limitations of human being – as well as the natural biases they have to cope with. While humans may struggle to remember a piece of text, a computer algorithm easily processes thousands of books and finds structure within using statistical methods with near-perfect precision.

Towards prediction of operational risk

The performance of NLP algorithms using word embeddings has increased tremendously of late. Using the embedding as a starting point, these algorithms learn the meaning of full sentences and use this information for further predictions. This is more generally called sentiment analysis. A popular application in the financial setting is predicting stock price movement from news articles, or twitter feeds. A recent paper by Denmark’s national bank predicts corporate distress of firms by analysing their annual reports. Taking textual data under scrutiny can make a contribution to quantitative predictive modelling, even in the financial sector where the use of words is very different to day-to-day language.

In summary, there are many opportunities for operational risk management to exploit AI and other related technological advances. One possibility is the classification of risk events. As an example, a computer algorithm can read risk descriptions written by risk managers and classify them according to their impact and frequency. Combining loss data with risk reports, improved prediction of risk events might lead to more accurate prediction of future losses. A different application is the measurement of more abstract concepts such as the financial health or maybe the cultural aspects of a company.

An article by Lars de Ruiter and Matthias Geerse. Both are finance & risk consultants at Solid Professionals, a consultancy from the Netherlands.