In April, the Bank for International Settlements has published three new papers which develop the body of knowledge (the theory forge, as we like to call it) in the topic of Credit Risk analytics, stress testing and Economic Capital quantification. This post will summarise and refer them for you.

Range of practices and issues in economic capital frameworks

This paper emphasises the importance of understanding the relationship between overall economic capital and its building blocks, as well as ensuring that the underlying building blocks (individual risk assessments) are measured in a consistent and coherent fashion. The main body of the paper focuses on issues associated with the overall economic capital process, rather than on the component risks measured by economic capital. Therefore it covers issues related to the use and governance of economic capital, the choice of risk measures, aggregation of risk, and validation of economic capital. In addition, three important building blocks of economic capital (dependency modelling in credit risk, counterparty credit risk and interest rate risk in the banking book) are examined in separate, stand-alone annexes. This list of building blocks is chosen due to the significance and complexity of the topics, and (with the exception of counterparty credit risk) partly because the topics are not covered in Pillar 1 of the Basel II Framework. This list is by no means exhaustive. http://www.bis.org/publ/bcbs152.pdf?noframes=1

Supervisory guidance for assessing banks' financial instrument fair value practices - final paper

The principles in this document cover supervisory expectations regarding bank practices and the supervisory assessment of valuation practices. The principles seek to promote a strong governance process around valuations; the use of reliable inputs and verse information sources; the articulation and communication of valuation uncertainty both within a bank and to external stakeholders; the allocation of sufficient banking and supervisory resources to the valuation process; independent verification and validation processes; consistency in valuation practices for risk management and reporting purposes, where possible; and strong supervisory oversight around bank valuation practices. This guidance applies to all financial instruments that are measured at fair value, both in
normal market conditions and during periods of stress, and regardless of the financial reporting designation within a fair value hierarchy. This guidance does not set forth additional accounting requirements beyond those established by the accounting standard setters.
http://www.bis.org/publ/bcbs153.pdf?noframes=1

Measuring portfolio credit risk correctly: why parameter uncertainty matters

by Nikola Tarashev, Working Papers No 280, April 2009

This paper generalizes the popular asymptotic single risk factor (ASRF) model of portfolio credit risk by allowing for noisy estimates of two key parameters: probability of default (PD) and asset-return correlation. Applied to a stylized empirical framework, the generalized model delivers two alternative measures of tail risk that help underscore the importance of estimation noise. The first measure is a naive value-at-risk (VaR) of the portfolio, which accounts for the credit-risk factor under the assumption that point estimates of the PD and asset-return correlation are equal to the true values of the respective parameters. The second is the correct VaR measure, which accounts not only for the credit-risk factor that influences the naive VaR but also for the uncertainty stemming from estimation noise. A Bayesian inference procedure reveals that ignoring estimation noise leads to a substantial understatement of the correct VaR. In the benchmark specification (where an investor in a homogeneous portfolio estimates the PD and asset-return correlation at 1% and 20%, respectively, on the basis of data covering 200 obligors over 10 years) the correct VaR is 27% higher than the corresponding naive VaR. This result is striking, not least because the underlying stylized empirical framework incorporates a lower bound on the amount of estimation noise. http://www.bis.org/publ/work280.pdf?noframes=1