The Bank of Canada Review, Summer 2009 has been released, with the articles:
- Collateral Management in the LVTS by Canadian Financial Institutions
- The Complexities of Financial Risk Management and Systemic Risks
- The Changing Pace of Labour Reallocation in Canada: Causes and Consequences
- BoC-GEM: Modelling the World Economy
The question of Collateral Management, addressed in the first article, sprang to prominence at the height of the Credit Crunch in 4Q08, when the “eligibility premium” – the yield differential between two securities identical in all respects except that one was eligible to be used as collateral with the central bank, the other not – exploded from its normally immeasurably small levels.
The LVTS [Large Value Transfer System] is a real-time, electronic wire transfer system that processes large-value, time-critical payments quickly and continuously throughout the day. Participants in the LVTS use claims on the Bank of Canada to settle net payment obligations. To secure the payments that are sent through the LVTS, collateral is required.
…
The Bank originally accepted only Government of Canada (GoC) securities as collateral, but since it expanded the list in November 2001 to include a larger variety of securities (e.g., municipal securities and commercial paper), pools of collateral pledged by individual FIs to the LVTS have diversified significantly. Thus, while GoC-issued securities constituted about 55 per cent of the discounted value of securities pledged in 2002, they made up less than 30 per cent in early 2007
…
The results of this study are important for policymakers such as the Bank of Canada, which is concerned both about the effi cient functioning of fixedincome markets and about the credit risk it ultimately bears in insuring LVTS settlement. Given these new insights into the behaviour of FIs, future changes in collateral policies, in particular those regarding the eligibility of assets as collateral, can be designed more effectively.Ongoing monitoring of and research into collateral management practices is required to keep abreast of
the changing behaviours at fi nancial institutions and within an evolving financial environment. Future
research will examine collateral management in more detail, with a particular focus on changes resulting from the recent fi nancial crisis and the ensuing increase in Government of Canada debt issuance.
The second article is also of interest, although I suspect that it is merely another volley in the battle for the BoC to extend its influence to macro-prudential regulation and frustrate the designs of OSFI upon the turf:
Banking theory has made very limited inroads into the theory and practice of risk management, where modelling has been dominated by the frictionless, efficient-market model masquerading under the title of financial engineering.
…
For example, in 1998, Salomon Brothers (as related in Bookstaber 2007, Chapter 5) were using a model of the yield curve, the so-called two-plus model (two random factors plus a constant—with the constant signalling shifts in Federal Reserve policy). The model had worked well to produce a steady stream of arbitrage profits over several years. In 1998, these profits changed to a stream of losses as the fixed-income arbitrage group struggled with what seemed to be a change in the underlying model. It seemed that another random factor had appeared, leaving the group holding residual risks, which were causing large losses. The risk manager struggled to help the group, but in the end, it was shut down. The exit had to be disguised and undertaken over several weeks, since Salomon’s large positions in the market were affecting bond liquidity and could entice arbitrageurs to exploit the company. The worst-case scenario would have occurred if Salomon’s sales had driven down prices, leading other traders to dump bonds and driving prices even further down, thus exacerbating Salomon’s losses. Bookstaber argues that this exit by Salomon’s large bond-arbitrage group made the market less liquid and increased the difficulties faced by Long-Term Capital Management (LTCM) later in the year, when its bond-arbitrage position became untenable after the Russian bond default (another unmodelled risk).
…
In all the above models, three major risks stem from model misspecifi cation through either: (i) choosing the
wrong number of random factors; (ii) inappropriate random factor distributions (e.g., normal, symmetric
distributions rather than skewed distributions), and/or (iii) using poor parameter estimates for the coefficients or factor loadings on risky factors. These risks should be tested regularly by back-testing the models llooking for systematic deviations from the model using actual data), and checking the history of trades and the profi t/loss outcomes on exposures. Because all models are merely approximations, losses and profits on exposures should be expected. In a well specified and calibrated model, however, the history of profits and losses will expose biases. Any detected biases should be examined, and appropriate action taken. Although this is easy to state as a general principle, in reality, the management and estimation of risks is far from perfect, especially in periods of high volatility, where correlations can change rapidly.