Notes (unaudited)

a) Fair Value Hierarchy

Determination of Fair Values from Quoted Market Prices or Valuation Techniques

Material changes in level 3 instruments

Financial assets measured with valuation techniques using significant non-market observable inputs (level 3) mainly include instruments linked to the US residential and commercial real estate markets, US reference linked notes, US student loan ABS, other non-real estate ABS, leveraged finance deals, and structured rates and credit trades. Level 3 financial liabilities mainly include hybrid financial liabilities, instruments linked to the US residential sub-prime market, and structured rates and credit trades. Level 3 assets and liabilities include cash trades and synthetic trades in the form of derivatives.

The reduction of level 3 items recognized as Trading portfolio assets and Trading portfolio assets pledged as collateral by CHF 10.2 billion in second quarter 2008 mainly relates to the sale of US RMBS to a fund managed by BlackRock (CHF 3.5 billion), other sales of RMBS level 3 assets (CHF 1.2 billion), reclassifications from and into level 2, and write-downs. The reclassifications from level 3 to level 2 due to increased availability of reliable external inputs or prices include US reference-linked notes asset pools (CHF 1.9 billion), instruments linked to the US commercial real estate market (CHF 1.1 billion), instruments linked to non-US residential mortgage markets (CHF 2.6 billion) and other ABS (CHF 0.9 billion). Leveraged finance loans of CHF 2.4 billion were reclassified from level 2 to level 3 due to continued lack of activity in the respective markets.

The decrease in level 3 items recognized as Positive and Negative replacement values mainly relates to reductions in gross values of structured rates and credit trades including bespoke CDOs due to credit spread changes, and to a minor extent, reclassifications to level 2.

b) Valuation Techniques and Inputs

Where possible, financial instruments are marked at prices quoted in active markets. In the current market environment, such price information is typically not available for all instruments linked to the US residential mortgage market, and UBS applies valuation techniques to measure such instruments. Valuation techniques use "market observable inputs", where available, derived from similar assets in similar and active markets, from recent transaction prices for comparable items or from other observable market data. For positions where observable reference data are not available for some or all parameters, UBS estimates the non-market observable inputs used in its valuation models.

For the period ended 30 June 2008, UBS used valuation models primarily for super senior RMBS CDO tranches referenced to sub-prime RMBSs. The model used to value some of these positions projects losses on the underlying mortgage pools and applies the implications of these projected lifetime losses through to the RMBS and then to the CDO structure. The primary inputs to the model are monthly statistical data on delinquency rates, foreclosure rates and actual losses that describe the current performance of the underlying mortgage pools. These are received near the end of each month and relate to the preceding month's cash flows on the mortgages underlying each RMBS. The other key factor input to the model is an estimate of loss given default, which is a non-market observable input.

In fourth quarter 2007 and first half 2008, UBS used relevant ABX market indices to calibrate its loss projections to ensure that the super senior RMBS CDO model is consistent with observed levels of the indices in the market. Despite the various limitations in the comparability of these indices to UBS's own positions, it was felt that this was the best approach in view of the further deterioration in liquidity and resultant lack of observed transactions to which the model could be calibrated.

The valuation model also considers the impact of variability in projected lifetime loss levels and applies a discount rate for expected cash flows derived from relevant market index prices to value expected cash flows. The external ratings of the RMBSs underlying the CDO tranches or the CDO tranches themselves are inputs to the valuation model only to the extent that they indicate the likely timing of potential "events of default".

The valuation model incorporates the potential timing and impact of such default events based on an analysis of the contractual rights of various parties to the transaction and the estimated performance of the underlying collateral. There is no single market standard for valuation models in this area, such models have inherent limitations, and different assumptions and inputs would generate different results. The super senior RMBS CDO valuation model is used to value a portion of UBS's net long exposures to super senior RMBS CDOs and in cases where UBS holds a gross long position in a super senior RMBS CDO hedged one-to-one with an offsetting short position (since this valuation is necessary to calculate any related credit valuation adjustments).

In cases where liquidation of the RMBS CDO is deemed imminent, and where it is possible to obtain reliable pricing of the underlying instruments, the super senior RMBS CDO valuation model is superseded. Instead, valuation in these cases is based on the estimated aggregate proceeds of the liquidation (using current fair value estimates of the underlying instruments) less any estimated expenses associated with the liquidation.

US super senior RMBS CDO

Write-downs of super senior US RMBS CDO positions (subprime, and to a lesser extent Alt-A and prime) during the second quarter of 2008 reflected worsening remittance data as well as declines in the ABX indices to which the valuation model is calibrated. No significant changes to the RMBS CDO valuation model occurred during this period, although approximately two thirds (by market value) of the super senior RMBS CDOs are now valued using the liquidation-based approach described above. The two primary unobservable factors in the valuation model are the loss projections on the underlying mortgage pools and the risk premium component of the discount rate. To assess the sensitivity of the super senior RMBS CDO valuations to the loss projections, a 10% adverse change in all mortgage pool loss projections (that is, from 25% loss to 27.5% loss, where 25% is the average ABX implied loss rate for sub-prime mortgage pools) across all relevant RMBS collateral is considered. Holding all other elements of the model constant, this adverse change in loss projections would result in an additional write-down of approximately USD 436 million (CHF 445 million). The current risk premium assumption in the valuation model is 11.1% (implying a discount rate of Libor plus 11.1%). An increase in the risk premium of 100 basis points, holding other aspects of the model constant, is estimated to result in an additional write-down of approximately USD 92 million (CHF 94 million). These estimates are intended to convey information on the sensitivity of the model-based valuation to unobservable inputs; they are not intended as risk assessments. In the interest of completeness, these sensitivity estimates include both RMBS super senior CDOs valued using the valuation model and those valued on a liquidation basis (corresponding figures for the model-only population are USD 131 million (CHF 134 million) and USD 39 million (CHF 40 million)).

During June 2008, UBS exited multiple long super senior RMBS CDO positions. Transaction pricing was consistent with marks produced by the RMBS super senior CDO valuation model.

Credit valuation adjustments on monoline credit protection

Credit valuation adjustments (CVA) for monoline credit protection are based on a methodology that uses five year credit default swap spreads on the monolines as a key input in determining an implied level of expected loss.1 Where a monoline has no observable credit default swap spread, a judgement is made on the most comparable monoline and its spreads are used instead. During the second quarter of 2008, this methodology was re-calibrated to market conditions based on the increasing prevalence of CDS trading with up-front cash exchanges and declines in potential recovery rates implied by recovery swap contract pricing. In second quarter 2008, UBS changed its estimates related to input factors, which increased the credit valuation adjustments brought forward by approximately USD 670 million (CHF 684 million).

To assess the sensitivity of the CVA calculation to alternative assumptions, the impact of a 10% increase in monoline credit default swap spreads (e.g. from 700 basis points to 770 basis points for a specific monoline) is considered. At 30 June 2008, such an increase would have resulted in an increase in the monoline credit valuation adjustment of approximately USD 157 million (CHF 160 million). The sensitivity of the monoline credit valuation adjustment to a decrease of 1 percentage point in the monoline recovery rate assumptions (e.g. from 40% to 39% for a specific monoline; conditional on default occurring) is estimated at USD 29 million (CHF 30 million).

In addition, the credit valuation adjustments related to transactions referencing RMBS CDOs are sensitive to the same unobservable inputs highlighted in the preceding discussion of US super senior RMBS CDOs. Holding all other parameters constant, the sensitivity of the monoline credit valuation adjustment to a 10% adverse change in loss projections related to the collateral underlying referenced RMBS CDOs is estimated at USD 70 million (CHF 72 million). The sensitivity of the monoline credit valuation adjustment to an increase of 100 basis points in the discount margin used in valuing RMBS CDOs, holding all other parameters constant, is estimated at USD 53 million (CHF 54 million).

Credit valuation adjustments are intended to achieve a fair value of the underlying contracts and accordingly are based on publicly available information. Potential restructurings of individual monolines may result in economic outcomes that differ from the CVA amounts shown above, although we would expect the market variables that are key inputs to the CVA methodology to react promptly to public information regarding any such restructurings. Overall exposures to monolines would also be adjusted for any cash or other forms of settlement received against outstanding liabilities.

Student loan auction rate certificates (ARCs)

Where student loan ARCs have been classified as level 3 due to impaired market liquidity, they have been valued as floating rate notes with three pricing inputs - the coupon, the current discount margin or spread, and the maturity. The coupon is generally assumed to equal the maximum rate allowed under the terms of the instrument, the current discount margin is based on an assessment of observable yields on instruments bearing comparable risks, and the maturity is based on an assessment of the term of the underlying instrument and the potential for restructuring the ARC. Model assumptions differ according to the characteristics of the underlying ARC (i.e. taxable versus tax-exempt instruments, government insured versus privately insured). The primary unobservable input to the valuation is the maturity assumption, currently set at five years for the majority of the ARC instruments. If this assumption is increased by one year (that is, to six years) the impact would be an additional writedown of approximately USD 218 million (CHF 223 million). A 100 basis points increase in the discount margin would reduce ARC valuations by approximately USD 343 million (CHF 350 million). The sensitivity numbers in this paragraph relate to student loan ARCs recognized on UBS's balance sheet at 30 June 2008. Refer to Note 14 for details about future acquisitions of student loan ARCs.

1 In one case, UBS has assessed a credit valuation adjustment of 100% and declared the hedge completely ineffective. The sensitivity of the now unhedged CDO valuations to unobserved inputs is included in the preceding discussion of US super senior RMBS CDOs.

c) Deferred Day-1 Profit or Loss

The table reflects financial instruments for which fair value is determined using valuation models where not all inputs are market-observable. Such financial instruments are initially recognized in UBS's Financial Statements at their transaction price although the values obtained from the relevant valuation model on day-1 may differ. The table shows the aggregate difference yet to be recognized in profit or loss at the beginning and end of the period and a reconciliation of changes in the balance of this difference (movement of deferred day-1 profit or loss).