The way organisations perceive financial risk has rapidly evolved over the course of the last decade and finance directors have mainly a series of sweeping regulatory updates to thank for those constantly shifting parameters.
Researchers at the Boston Consulting Group have estimated that financial institutions must now circumnavigate some 200 regulatory revisions per day – from MiFID II and its 1.7m paragraphs of rules on transaction reporting, to Volcker 2.0, GDPR and the EU’s new Benchmark Regulation – each of those revisions take their own toll.
According to one 2017 Thomson Reuters survey, banks were expecting to devote around 30% of their total 2018 compliance budget adjusting to MiFID II alone.
Compound that mounting compliance burden upon a range of increasing data security and cybercrime threats, it is clear why there is a surge in demand for innovative risk analysis methods.
Fortunately, the advent of big data has meant financial directors are more or less spoilt for choice in terms of new modelling capabilities and risk management solutions. Yet by and large, it’s worth pointing out many of the market’s most dynamic risk analysis methods are still grounded in the basic techniques financial directors have been calling upon for decades.
Even cutting-edge fintech platforms are still heavily reliant upon basic techniques like sensitivity analysis, scenario analysis and Monte Carlo simulations as the foundations from which their solutions calculate and assess financial risk. Yet through major developments in data analytics, these existing models have been supercharged to offer risk liability systems that are now able to draw from granular data pools to provide unprecedented reporting and analysis functionality.
Liquidity and credit risk have been at the forefront of financial legislation in recent years – and that emphasis will only deepen in the years to come as companies brace for the impact of Basel III’s full implementation in January 2019. Basel III is set to introduce tighter capital requirements and liquidity rules that place an added burden on financial institutions to focus on risk models that account for cellular datasets that include items like borrower demographic and quality of collateral.
Transparency is a major concern for banks too. After all, regulatory demand for cellular datasets isn’t simply because government bodies want to ensure that financial institutions are holding enough quality capital – but also because regulators will now be demanding banks disclose more data relating to financial risk as a mandatory element of Basel compliance.
One solution financial directors have been looking to in order to fulfil those regulatory obligations is machine learning, although quite a few machine learning financial risk models are still relatively experimental regarding their availability as a commercial service or product for financial institutions. Even so, machine learning and the financial risk models it has spawned are already inspiring existing risk management systems providers to further develop their own products in the here and now.
Machine learning is a method by which computers can be taught to collect and deconstruct data, analyse it and consequently predict an outcome or make a financial determination. This is a solid step up from the tried-and-tested statistical learning methods that most financial platforms rely on to assess risk. Bearing in mind the huge appetite for automation, there are several intriguing models fintechs have been trialling to enhance their existing financial risk analysis solutions.
One machine learning technique developers have been putting to the test is the deployment of artificial neural networks (ANNs). These are mathematical simulations that use various input values to link hidden data layers and supposedly ideal to analyse credit risk because of how they handle the non-linear and interactive effects of variables to make detailed projections about a loan recipient and their ability to take on debts.
Studies on the potential impact of using an ANN model to assess credit risk suggest the technique is marginally more accurate than the standard linear regression model that many standard financial risk management solutions tend to rely upon. On the other hand, ANNs also carry a reputation for being fairly intensive in order to set up and maintain.
Meanwhile, machine learning risk methods like random forest simulators or boosting are able to generate extensive tree predictors that determine the probability of default based on a range of variables and data inputs.
That being said, it’s worth noting that most full-fledged machine learning systems are still in their infancy. For the time being, financial directors keen on leveraging the power of machine learning would do better to look to well-executed hybrids that have started to incorporate the characteristics of these experimental machine learning methods into their existing risk management modelling techniques.
“For the time being, financial directors keen on leveraging the power of machine learning would do better to look to well-executed hybrids that have started to incorporate the characteristics of these experimental machine learning methods into their existing risk management modelling techniques.”
Moody’s offers a dynamic modelling system in its RiskCalc solution, which uses a generalised additive model that assigns weights to various risk drivers and automatically maps the score of those risks to produce detailed reports charting a probability of default or expected default frequency. Moody’s RiskCalc model and its smart system are particularly robust on a business-to-business level in terms of their ability to predict private firm defaults.
Other industry players are already rolling out similar products that incorporate big data to mitigate accounting risk. The revised credit loss model unveiled under IFRS 9 in January 2018 has created demand for risk management solutions that can improve projections of accounting charges before a loan has been issued – and products like the Loan Origination module from Temenos are helping financial directors to produce more detailed reports that have initiated more dynamic product pricing structures amongst financial institutions.
Temenos’ accounting application enables organisations to instantly and automatically decision loans to applicants from any origination channel based on the predefined policies and tolerances of that institution’s choosing through a range of granular matrices. Teams can customise ratios and aggregates to build custom risk and decision models or use more commonly used accounting risk models to improve decision making.
Meanwhile, Finastra’s suite of end-to-end lending software solutions have been able to use similar machine learning technology to streamline loan booking and generate instant, 360-degree snapshots of charges so that institutions are able to adjust pricing and portfolios in real-time. Finastra’s Total Lending solution is designed to connect loan origination with risk management through minute data inputs that automatically generate risk analysis reports using peer-group comparisons, covenant monitoring, customisable ratios and more.
As modelling and analysis tools, these sorts of products undeniably have the power to offer enhanced financial risk projections thanks to advances in technology and big data. On the flip side, it’s worth noting products offering more advanced financial modelling or machine-based learning tend to be provided as part of a wider service or risk management suite that may or may not be compatible with existing systems.
The advanced data capabilities of new systems have drastically improved financial risk modelling solutions – but those same advances are also creating unprecedented new institutional threats. According to think tank the Center for Strategic and International Studies, the global cost of cybercrime exceeded $600bn in 2017. That figure represents a jump of more than $100bn over the last three years, and as the Internet of Things continues to infiltrate new business processes and influence day-to-day operations, organisations are finding themselves more exposed to that threat than ever before.
“According to think tank the Center for Strategic and International Studies, the global cost of cybercrime exceeded $600bn in 2017.”
Cyber wargames are fast-becoming one of the most popular tools with which to model and stress test an institution’s ability to withstand that exposure. Wargaming is a solution designed to enhance risk-informed decision making through rigorous, experimental learning. It sees a range of organisational stakeholders play out engaging and plausible scenarios as part of a Human-in-the-Loop game that enables institutions to dynamically model participant and adversarial actions to experience failure and identify critical vulnerabilities and without facing real-world financial risk.
Although cyber wargaming is a relatively new method of financial risk analysis, there are already plenty of industry players out there offering comprehensive simulation services for institutions of all shapes and sizes.
For starters, Deloitte’s Cyber Risk Services encompass a comprehensive suite of analysis tools with which to prevent cyber threats – including the provision of full-service cyber wargaming simulations. Deloitte’s team are able to assist companies in planning table top exercises to test readiness and identify points of weakness that can be improved upon in advance of a real cyberattack. Simulations can either be customised based upon an organisation’s previously identified financial risks, or smaller institutions can draw from a range of pre-built exercises in order to test readiness.
Yet financial risk modelling must extend far beyond cyber wargames if organisations are to maintain a data-secure business. After all, freshly introduced penalties under GDPR could hit financial institutions with a fine of up to €20m or 4% of worldwide turnover for failing to observe new individual rights concerning the processing of data, consent, data retention and the right to be forgotten. Pair those obligations alongside burdens like Basel III and MiFID II, and it’s fair to say that financial directors have now got a range of concerns to address in terms of financial risk exposure.
Yet through machine learning integrations, all-in-one, automated solutions and new simulation models, navigating these exposures is easier than ever before – and so long as service providers continue to adapt to new regulations and fresh cyberthreats, there are plenty of methods and techniques for financial directors to choose from.
