Blog

The motivation for federated credit risk models

Federated learning is a machine learning technique that is receiving increased attention in diverse data driven application domains that have data privacy concerns. The essence of the concept is to train algorithms across decentralized servers, each holding their own local data samples, hence without the need to exchange potentially sensitive information. The construction of a common model is achieved through the exchange of derived data (gradients, parameters, weights etc). This design stands in contrast to traditional model estimation where all data reside (or are brought into one computational environment).

A new logo for the Open Risk Manual

We have updated the logo for the Open Risk Manual.

The new logo aims to make more explicit both the inspiration that the Open Risk Manual project draws from the trail-blazing Wikipedia initiative (and increasing collection of associated Wikimedia projects) and the reliance on the open source ecosystem of software and tools, including the mediawiki software and the important semantic mediawiki extension.

An overview of EU Financial Regulation initiatives

In the European Union there are several ongoing large scale legislative and regulatory projects that transform the context within which individual, firms and the public sector interact economically. While financial and regulatory reform is an ongoing process in all jurisdictions globally, the size and supra-national nature of the European Union makes those projects particularly interesting.

Overview of the Julia-Python-R Universe

A new Open Risk Manual entry offers a side-by-side review of the main open source ecosystems supporting the Data Science domain: Julia, Python, R, sometimes abbreviated as Jupyter.

Motivation

A large component of Quantitative Risk Management relies on data processing and quantitative tools ( aka Data Science ). In recent years open source software targeting Data Science finds increased adoption in diverse applications. The overview of the Julia-Python-R Universe article is a side by side comparison of a wide range of aspects of Python, Julia and R language ecosystems.

Data Quality and Exploratory Data Analysis using Python

In two new Open Risk Academy courses we figure step by step how to use python to work to review risk data from a data quality perspective and how to perform exploratory data analysis with pandas, seaborn and statsmodels:

• Introduction to Risk Data Review
• Exploratory Data Analysis using Pandas, Seaborn and Statsmodels

What is a Risk Taxonomy?

There are various formal definitions of risk taxonomies (and we will go over those below), but it might be useful to first look at a very intuitive example of a risk taxonomy: the classification of fire hazards (also known as fire classes)

Limit frameworks are fundamental tools for risk management

A Limit Framework is a set of policies used by financial institutions (or other firms that actively assume quantifiable risks) to govern in a quantitative manner the maximum risk exposure permitted for an individual, trading desk, business line etc.

Open Source Securitisation

Motivation

After the Great Financial Crisis securitisation has become the poster child of a financial product exhibiting complexity and opaqueness. The issues and lessons learned post-crisis were many, involving all aspects of the securitisation process, from the nature and quality of the underlying assets, the incentives of the various agents involved and the ability of investors to analyze the products they invested in. While the most egregious complications involved various types of re-securitisation and/or the interplay of structured credit derivatives undoubtedly even vanilla securitisation structure has a considerable amount of business logic.

Visualization of large scale economic data sets

Economic data are increasingly being aggregated and disseminated by Statistics Agencies and Central Banks using modern API’s (application programming interfaces) which enable unprecedented accessibility to wider audiences. In turn the availability of relevant information enables more informed decision-making by a variety of actors in both public and private sectors. An excellent example of such a modern facility is the European Central Bank’s Statistical Data Warehouse (SDW), an online economic data repository that provides features to access, find, compare, download and share the ECB’s published statistical information.

Motivation for the comparison

A large component of risk management relies on data processing and quantitative tools. In turn, such information processing pipelines and numerical algorithms must be implemented in computer systems.

ESMA Securitisation Templates are now documented at the Open Risk Manual

The ESMA Securitisation Templates are now fully documented at the Open Risk Manual. Users can browse, search and cross-reference with the rest of the knowledge base.

The challenge with historical credit data

Historical credit data are vital for a host of credit portfolio management activities: Starting with assessment of the performance of different types of credits and all the way to the construction of sophisticated credit risk models. Such is the importance of data inputs that for risk models impacting significant decision-making / external reporting there are even prescribed minimum requirements for the type and quality of necessary historical credit data.

Release of version 0.4.1 of the transitionMatrix package focuses on stressing transition matrices

Further building the open source OpenCPM toolkit this release of transitionMatrix features:

1. Feature: Added functionality for conditioning multi-period transition matrices
2. Training: Example calculation and visualization of conditional matrices
3. Datasets: State space description and CGS mappings for top-6 credit rating agencies

Release of version 0.4 of the transitionMatrix package

Further building the open source OpenCPM toolkit this release of transitionMatrix features:

1. Feature: Added Aalen-Johansen Duration Estimator
2. Documentation: Major overhaul of documentation, now targeting ReadTheDocs distribution
3. Training: Streamlining of all examples
4. Installation: Pypi and wheel installation options
5. Datasets: Synthetic Datasets in long format

Enjoy!

Is the IFRS 9 or CECL standard more volatile? Its all relative

Objective

In this study we compare the volatility of reported profit-and-loss (PnL) for credit portfolios when those are measured (accounted for) following respectively the IFRS 9 and CECL accounting standards.

The objective is to assess the impact of a key methodological difference between the two standards, the so-called Staging approach of IFRS 9. There are further explicit differences in the two standards. Importantly, given the standards are not prescriptive, it is very likely that there will be material differences in interpretation and implementation of the principles (for example on the nature and construction of scenarios). In this study we perform a controlled comparison adopting a ‘ceteris-paribus’ mentality: We assume that all other implementation details are similar and we focus on the impact of the Staging approach.

Release of version 0.4 of the Concentration Library adds Geographic / Industrial concentration indexes

Further building out the OpenCPM set of tools, we release version 0.4 of the Concentration Library, a python library for the computation of various concentration, diversification and inequality indices.

Representing economic activity using pictograms

Visualization can produce significant new insights when applied to quantitative data. It is currently undergoing a renaissance that mirrors other developments in computing and data science. Sophisticated open source libraries such as d3.js or matplotlib, to name but a couple, are enabling an ever wider range of users to distill valuable information from the avalanche of data being produced.

Credit Portfolio PnL volatility under IFRS 9 and CECL

Objective

We explore conceptually a selection of key structural drivers of profit-and-loss (PnL) volatility for credit portfolios when profitability is measured following the principles underpinning the new IFRS 9 / CECL standards

Methodology

We setup stylized calculations for a credit portfolio with the following main parameters and assumptions:

Extending the Open Risk API to include the EBA Portfolio Data Templates

The Open Risk API provides a mechanism to integrate arbitrary collections of risk data and risk modelling resources in the context of assessing and managing financial risk. It is based on two key technologies of the modern Web, RESTful architectures and Semantic Data.

Credit Portfolio Management in the IFRS 9 / CECL and Stress Testing Era

The post-crisis world presents portfolio managers with the significant challenge to assimilate in day-to-day management the variety of conceptual frameworks now simultaneously applicable in the assessment of portfolio credit risk:

• The first major strand is the widespread application of regulatory stress testing methodologies in the estimation of regulatory risk capital requirements
• The second major strand is the introduction of new accounting standards (IFRS 9 / CECL) for the measurement and disclosure of expected credit losses While both Regulatory Stress Testing and IFRS 9 / CECL accounting require investment in analytic capabilities and provide unique new insights, both are aimed at satisfying evolving prudential or investor disclosure requirements. Neither is designed to help credit portfolio managers analyse and steer their portfolios in the bottom-up fashion that is an essential part their mandate.

The above developments are overlaid into pre-existing conceptual and practical frameworks such as