The VAR of the aggregate portfolio at the 99 % confidence interval over a 10-day horizon should be disclosed along the same lines and categories as for the trading book (see Table 1). In addition, a discussion of the methodological aspects of the VAR computation model, the aggregation of market risk factors, the treatment of derivatives and the backtesting procedures should be added, along the lines described in section IV.2. This information allows the reader to examine the incremental contribution of the firm's non-trading positions to the firm's overall VAR

However, three additional questions should be addressed with respect to the VAR of the non-trading activities. First, some firms use Earnings-at-Risk rather than VAR models (see for instance Citibank's EAR model described in its 1997 annual report). EAR is difficult to compare with VAR if the firm fails to mark-to-market all or some of its holdings. Second, the relevant time horizon of the VAR should be 10 days for the sake of comparison across firms and activities (consist with the BIS rules). However from an economic perspective and given the structure of the balance sheet, perhaps a longer horizon, e.g. one year, would be more appropriate. Finally, while marking-to-market is important for the relevance of the VAR figures of the non-trading activities, their meaningfulness cannot be judged without considering the liquidity of those positions over the VAR horizon.

b) Asset-Liability Management

It is important that firms disclose information regarding the asset-liability management of their non-trading interest rate book. For the sake of comparability across firms, this should be accomplished following the guidelines in the BIS document "Principles for the Management of Interest Rate Risk" (1997). This suggests that firms should at least provide maturity or average duration based interest rate sensitivities for all positions, categorised by time band as illustrated in Table 8.

Table 8

Whenever the firm uses an exact duration (or a more elaborate interest rate model such as a static or dynamic simulation model) to estimate the interest rate sensitivity of its assets and liabilities, a brief description of the model should be disclosed. The firm should also disclose the underlying assumptions (how many factors to model yield curve changes, what type of factors, what estimation period, etc.) and the way model handles complex securities with uncertain maturities (for instance mortgages) and embedded options. The backtesting procedure of the model for the entire non-trading portfolio and for each constituent currency denominated exposure should be discussed.

The stress testing procedure of the asset-liability model recommended in the BIS January 1997 Principles should be discussed. This will clarify the sensitivity of economic value and of net interest rate margin to sudden large changes in the level and shape of the yield curves denominated in the relevant currencies, including worst case changes. Finally, the annual report should discuss the corrective actions the firm plans to use if such extreme movements did occur (for instance, the recent jumps in most Asian countries short term rates or the impact of the maturity rescheduling of the Russian debt).

c) Hedging policy

The discussion of the market risk exposure of the firm's non-traded activities should end with a brief discussion of the firm's hedging policy. The discussion should include the horizons (long versus short-term hedges), derivative instruments (description of the type of derivatives used, by currency), and markets (% of OTC versus exchange-traded, domestic versus foreign market traded) for which the hedging of market risks is currently implemented. The effective performance of the hedging strategy and its impact on the firm's annual profits should also be briefly explained for the major market risk factors.