Some models used by US lenders to set loan-loss provisions under new accounting rules fail to produce accurate estimates of credit risk during economic downturns, research reveals. The study backtested six modelling methods with mortgage data from the financial crisis of 2008 and showed that most models were unable to equip banks with appropriate levels of reserves.

The results raise questions for banks as they adapt to the Current Expected Credit Loss (CECL) regime, a new set of standards that

The CECL implementation deadline is approaching. Banks and other financial institutions should be evaluating the likely effects of the standard on them and deciding what their next steps are.

This course will consider a range of key areas, including: challenges & opportunities; CECL quantification; model risk management for CECL; and internal controls framework. 

This course is CPE accredited and delegates will earn 12 points if in attendance for both days of the course. The course will be held under Chatham House Rule with the opportunity for discussion within a practical learning environment. 

As large US banks ponder whether or not to switch to the new standardised approach to counterparty credit risk (SA-CCR) this year, some European lenders are bemoaning being stuck with a June 2021 implementation date.

US prudential regulators have said banks can ditch the less risk sensitive current exposure method (CEM) and start using SA-CCR as early as April this year, with a mandatory compliance deadline of January 1, 2022.

Crucially, the final rule also incorporates SA-CCR into the cleared

COMMENTARY: Keep an eye on AI

New applications for artificial intelligence (AI) technologies such as machine learning continue to surface – making it ever more important for regulators and risk managers to focus on explainability, even for internal-facing uses.

The idea of fully automated trading or asset management has never become reality, any more than the idea of widespread lights-out manufacturing. The coming reality is closer to ‘assisted human operations’ – the teaming of human and machine to cover each other’s weaknesses. And often these hybrid teams are aimed at being used internally rather than externally – processing data, analysing returns and classifying assets, rather than predicting prices and directing investments. Ethical algos are also helping quant managers handle the volumes of new data associated with investment strategies involving environmental, social and governance considerations.

But all this rapid evolution increases the...

Stress tests set by the Federal Reserve for 2020 are tougher for participating banks than those produced by the European Union.

Under the severely adverse scenario drafted by the Fed for this year’s Dodd-Frank Act Stress Tests (DFAST) and Comprehensive Capital Analysis and Review (CCAR), real GDP is projected to shrink about –8.5% from its pre-recession peak. In contrast, the adverse scenario for the EU round of tests projects a start-to-stress real GDP decline of –4.2% for the eurozone. The EU

Jump to In focus: top op risks | Spotlight: Citi fine

In January’s largest operational risk loss, Russia’s VTB Bank lost $535 million in a fraud involving loans to state-owned companies in Mozambique. Between 2013 and 2016, three Mozambique companies borrowed over $2 billion to finance maritime projects, comprising $535 million from VTB, $622 million from Credit Suisse, and an $850 million eurobond arranged by VTB and Credit Suisse. The loans were guaranteed by Mozambique’s government.

In May

Turning a profit in over-the-counter clearing is a tough slog. Some of Europe’s biggest banks – such as Deutsche Bank and RBS – have all but quit the business. Others have pulled back significantly. The big US banks – Citi, JP Morgan, Morgan Stanley and Bank of America – have been dominant for much of the past decade, bolstered by their scale and sizeable balance sheets.

But changes to capital rules, coupled with Brexit uncertainty and fears of concentration risk, may be aiding a European

This issue of The Journal of Risk contains articles that concern regulatory capital requirements, the linkage between granular transactional data and systemic risk and dependencies, and the empirical performance of a variety of hedging portfolios involving investments in oil. In “Hedging incentives for financial institutions”, Frans J. de Weert makes use of a standard continuous-time model to derive hedging incentives for financial institutions that are particularly relevant in the context of asset substitution. The author identifies, in particular, explicit conditions under which a financial firm’s market value is maximized by reducing the volatility of its regulatory capital ratio. This result is consistent with the Basel III regulatory framework, which regards available capital in relation to market value of equity. The next paper in the issue, “An internal default risk model: simulation of default times and recovery rates within the new Fundamental Review of the...

Eurex has been quietly sounding out its clearing members on changes to the way it applies margin add-ons to account for concentration risk, a member survey seen by Risk.net shows. Conversations with the bourse’s largest clearing members suggest the recalibration has attracted diverging views, leading to a delay in its implementation.

The additional margin is designed to account for the increased difficulty a central counterparty (CCP) might face when selling off or neutralising a portfolio of

COMMENTARY: 239 problems

Irritation over slow adoption of the Basel Committee on Banking Supervision’s BCBS 239 standard for effective risk data aggregation and risk reporting is not, unfortunately, a new story – it’s one that Risk.net has been covering virtually from the standard’s launch. Earlier fears of fines and capital add-ons have been, more or less, fulfilled – though most of the fines imposed for failures to organise risk information have not been explicitly linked to the BCBS 239 requirements.

Compliance has been slow despite the standard’s leisurely pace of implementation, and regulators have been imaginative in response; in the US, BCBS 239 has crept in under the umbrella of the Comprehensive Capital Analysis and Review (CCAR) and now regulators in Europe are trying a new tool – fire drills. Both the European Central Bank and Swiss regulator Finma have been throwing unexpected requests for data at the banks they supervise, often at the most...

Running a quantitative finance master’s programme is hard work. Adapting courses to meet the rapidly changing roles quants play in the banking industry is challenge enough for quant academia. Finding the right experts to teach newer skills can be harder still.

In recent years, the biggest structural change most programmes have faced is the migration of machine learning and data science topics from optional modules to a core part of the curriculum, with such skills now considered foundational by

This training course will address in-depth the opportunities and limitations of machine learning in quantitative finance with practical guidance from a variety of expert tutors.

Sessions will cover key theories, models and more advanced tools in machine learning using a quantitative approach. The course will examine what impact machine learning has on trading, portfolio construction and optimisation as well as focus on deep neural networks, applications of natural language processing , trading strategies and more.

For anyone who shudders at the mere thought of a clearing house failure, the latest bank-sourced data from Credit Benchmark could make for uncomfortable reading.

After improving by more than 5% from November 2017 to September 2019, the credit risk of 30 central counterparties (CCPs) reversed sharply at the end of last year. The 2.6% deterioration seen in October and November was the worst in two years, and compares with a drop of less than 2% following the default of power trader Einar Aas at Nasdaq Clearing in September 2018.

The sudden shift in sentiment defies easy explanation. CCPs have made a concerted effort to improve their risk management since the Nasdaq default, resulting in higher margin requirements and deeper liquidity buffers.

Following changes to its value-at-risk model for client clearing, required initial margin at LCH’s SwapClear service hit ?145.1 billion ($189.6 billion) at the end of the third quarter of 2019, up 46% from a year...

Welcome to the latest edition of Risk.net’s guide to the world’s leading quantitative finance master’s programmes, and ranking of the top 25 courses.

Almost 50 programmes (full list below) feature in the 2020 edition of the guide. Of these, 25 have been ranked according to set criteria including a programme’s selectivity, its research power, and its faculty’s links with the financial industry, among others (jump to How to read the metrics tables).

US programmes continue their dominance in this year’s rankings, but the overall make-up is more diverse: eight of the top 25 programmes are European, while another two are Canadian. Five programmes from Asia-Pacific also feature in this year’s guide, including City University of Hong Kong’s MSc Financial Engineering programme, which makes its debut, as does the Chinese University of Hong Kong, Shenzhen’s Master of Science in Financial Engineering – the first institution based in mainland China to feature in the...

This article accompanies Risk.net’s guide to the world’s leading quantitative finance master’s programmes. The full guide and a ranking of the top 25 programmes will be published tomorrow.

Princeton University’s Master in Finance graduate degree has taken the top spot in Risk.net’s annual ranking of the world’s leading quantitative finance master’s programmes for the second year running.

Almost 50 programmes feature in the 2020 edition of the guide. Of these, 25 have been ranked according to

Exasperated by the patchy uptake of the BCBS 239 principles on risk data, regulators in Europe have been conducting ‘fire drills’ to test compliance, while authorities both in Europe and the US have been quietly knitting them into expected regulatory performance.

Both the European Central Bank and the Swiss Financial Market Supervisory Authority (Finma) have been carrying out fire drills for some time, making surprise requests for risk data to see how quickly banks are able pull it together. A

I am delighted to introduce to you the first issue of The Journal of Computational Finance for 2020. This issue highlights the progress being made in scenario-driven approaches to computational finance, where traditionally models formulated as partial differential equations (for example) would have been used. These new approaches not only alleviate potentially problematic modeling assumptions, but also – perhaps surprisingly – show advantages in terms of computational complexity, even in cases where model-based approaches are applicable. In this issue’s first paper, “A shrinking horizon optimal liquidation framework with lower partial moments criteria”, Hassan Anis and Roy H. Kwon introduce a novel method for solving optimal liquidation problems. The key elements of their algorithm are scenario-based optimization, which bypasses the curse of dimensionality observed in alternative approaches such as decision trees, and a shrinking time horizon, to respect the dynamic...

COMMENTARY: Decomposing climate risk

This year, UK banks and insurers must prepare regulatory stress tests with an unprecedented scope, as the UK Prudential Regulatory Authority’s stress tests in 2021 include climate risk for the first time – but this should be only the first step towards incorporating the hazard into bank risk management practice.

Open for comments until March, the stress test proposal includes testing a static portfolio for the next 30 years at five-year intervals against three climate scenarios, which can broadly be described – in descending order of preference – as “tough but survivable”, “very bad but just about survivable” and (the worst) “current policy goals”. The last scenario will include the anticipated harm if current emissions policies continue, with the damage expected to occur between 2050 and 2080 (but brought forward to 2050 to ensure its inclusion in the 30-year window).

It’s an important step forward, and will, with...

Welcome to the first issue of Volume 8 of The Journal of Financial Market Infrastructures, which contains three papers.

In our first paper, “Supervisory  stress testing for central counterparties:  a macroprudential, two-tier approach”, Edward  Anderson, Fernando Cerezetti and Mark Manning  take stress testing for central counterparties (CCPs) from the microsupervisory level to the macrosupervisory level.1 The underlying motivation is that the robustness of a single CCP does not necessarily imply that an ensemble of all major CCPs globally  will also be robust. From a systemic risk standpoint, the impact of the simultaneous default of two large banks on the broader financial  system can best be studied when all linkages are taken into account. The authors of this paper therefore recommend using supervisory stress tests, involving multiple CCPs, that go beyond the current practice of the multi-CCP fire drills in the United States and...

Chris' current and previous quant roles at the top banks such as Lloyds Banking Group, Credit Suisse and DEPFA Bank as well as being an active publisher for Risk, earned him a top position in our poll of tutors.

Meet Chris and quiz him on anything quant related at the Machine Learning in Finance course in London

This is the sixth profile in our ‘Bank disruptors’ special report. Scroll to the bottom of the introductory article to view others in the series as they go live over the next few days.

When David Hudson was picked to lead the digital transformation of JP Morgan’s markets business in 2016, he built a team to start an agency execution platform for over-the-counter products.

The idea was simple enough: to use the bank’s technology and scale to source liquidity for institutional clients on an