UBS - Market risk in 2006

Graphs

The movements in the S&P 500 Composite and FTSE 100 and the historical volatility of these indices, which are shown in the charts below, reflect the variations in market conditions in general over the course of 2006.

The year started with active, buoyant markets and heavy trading volumes – equity indices rose, spreads tightened, new issuance was strong and there was considerable merger and acquisition activity. Inflation fears surfaced in May, and volatility increased, followed by a period of uncertainty and challenging market conditions. As market sentiment picked up in fourth quarter, equity markets resumed their upward trend, which continued to the end of the year when many reached all-time highs, and credit spreads tightened once more.

Value at Risk

In mid-2006 we reported that some of our strategic initiatives would naturally lead to an increase in our risk profile, especially in emerging markets where our exposure had been too low in recent years. The impact of these initiatives – combined with the excellent trading conditions in first and fourth quarters – can be seen in the Investment Bank's market risk. For 2006 as a whole, average VaR (10-day, 99% confidence, 5 years of historical data), increased to CHF 420 million, up from CHF 346 million in 2005. Both interest rate and equities VaR were higher than in the preceding year. The integration of Pactual from 1 December 2006 contributed to this increase – without Pactual, 2006 year-end VaR for the investment Bank would have been CHF 445 million, rather than CHF 473 million.

Interest rate risk was, as usual, the largest contributor to overall Investment Bank VaR averaging CHF 417 million, an increase from the 2005 average of CHF 364 million. Credit spread exposure remained the dominant element of interest rate VaR and the main driver of both the fluctuations over the period and the year-on-year increase, but changes in directional interest rate positions also contributed to the variations

Average equities VaR, at CHF 203 million, was higher than the 2005 average of CHF 173 million. Much of this increase was a response to good trading conditions, particularly in the first and fourth quarters.

Average Corporate Center VaR for 2006 was CHF 43 million, a decrease on the 2005 average of CHF 63 million, with reductions in both interest rate and foreign exchange risk. At year end, UBS VaR was lower than Investment Bank VaR as Corporate Center exposures provided some offset to Investment Bank positions. Market risk exposure in the other Business Groups have only a marginal impact on the total UBS VaR.

Changes in VaR limits in 2006

Following the annual limits review in early 2006, the Investment Bank VaR limit was increased and the Corporate Center limit decreased, within an unchanged limit for UBS. The VaR limits for UBS as a whole and for the Investment Bank were increased from 1 December reflecting the integration of Pactual and expected business growth in 2007. The Corporate Center limit was further reduced. The limits in force at year end are shown in the table.

Backtesting

Despite the increased market volatility in May and June 2006, and the resulting fluctuations in backtesting revenues, we had no regulatory backtesting exceptions in 2006. The graph below shows 1-day VaR for Investment Bank portfolios subject to market risk regulatory capital and the corresponding backtesting profits and losses. In the accompanying histogram we show the backtesting revenues alongside the daily "full revenues" – all revenues from business areas which have trading activities.

Investment Bank: Value at Risk (10-day, 99% confidence, 5 years of historical data)
	Year ended 31.12.06				Year ended 31.12.05
CHF million	Min.	Max.	Average	31.12.06	Min.	Max.	Average	31.12.05
Risk type
Equities	144	360	203	232	120	266	173	235
Interest rates	237	607	417	405	223	514	364	269
Foreign exchange	16	65	31	40	11	63	30	23
Energy, metals and commodities ¹	26	102	49	44	6	88	38	46
Diversification effect	²	²	(280)	(248)	²	²	(259)	(218)
Total	331	559	420	473	245	512	346	355
1 Includes base metals and soft commodities risk from 15 March 2006. 2 As the minimum and maximum occur on different days for different risk types, it is not meaningful to calculate a portfolio diversification effect.

UBS: Value at Risk (10-day, 99% confidence, 5 years of historical data)
	As of 31.12.06	Year ended 31.12.06				Year ended 31.12.05
CHF million	Limits	Min.	Max.	Average	31.12.06	Min.	Max.	Average	31.12.05
Business Groups
Investment Bank ^{1, 2}	775	331	559	420	473	245	512	346	355
Global Asset Management ³	30	4	16	9	10	3	13	10	8
Global Wealth Management & Business Banking	25	4	14	10	5	4	14	9	12
Corporate Center ⁴	100	25	69	43	27	32	84	63	62
Diversification effect		⁵	⁵	(54)	(52)	⁵	⁵	(62)	(64)
Total	850	336	565	429	464	255	520	366	373
1 Includes risk managed by Dillon Read Captial Management. 2 Includes Pactual from 1 December 2006. 3 Only covers UBS positions in alternative & quantitative investments. 4 VaR for Corporate Center includes non-trading interest rate exposure in the Treasury book. The sale of Private Banks & GAM was completed on 2 December 2005 and their exposures are excluded from this date. 5 As the minimum and maximum occur on different days for different Business Groups, it is not meaningful to calculate a portfolio diversification effect.

UBS: Value at Risk (1-day, 99% confidence, 5 years of historical data)1
	Year ended 31.12.06				Year ended 31.12.05
CHF million	Min.	Max.	Average	31.12.06	Min.	Max.	Average	31.12.05
Investment Bank ²	129	230	172	160	105	206	150	155
UBS	131	233	173	162	109	211	157	164
1 10-day and 1-day VaR results are separately calculated from underlying positions and historical market moves. They cannot be inferred from each other. 2 Positions in Investment Bank subject to market risk regulatory capital contributed average VaR of CHF 169 million in 2006 and CHF 147 million in 2005.

Stress loss

Stress loss for the Investment Bank, defined as the worst-case outcome from our macro scenarios, was also higher in 2006 than in 2005. As with VaR, credit spread exposures remained the dominant contributor. Stress loss exposure increased towards the end of the year following the integration of Pactual.

VaRiations on a theme

Financial institutions commonly use Value at Risk – VaR – to measure market risk as part of their internal risk control framework, to determine market risk regulatory capital requirements and for external disclosure of their market risk. But there is no industry- standard VaR – the term describes a family of measures and models with different statistical bases and parameters, and capturing different types of market risk events. Firms apply their VaR models to positions arising from different activities, and their external disclosures are not necessarily on the same basis as their internal risk controls or their regulatory capital calculations. All these factors affect the level of reported VaR and can make comparisons between firms difficult.
To aid comparison, the chart below shows the Investment Bank’s average VaR based on a variety of parameters. It clearly shows that the time horizon, the historical look-back period and the confidence interval have a significant impact on reported VaR. UBS uses the same VaR model for internal risk control (including limits), for regulatory capital calculation, and for detailed external disclosure, using a 10-day horizon and five years of historical data, as explained under “Value at Risk”. 1-day VaR, separately generated by the same models from the same historical data, is used for backtesting and is also included in our external disclosures. Our 1-day VaR has always been lower than our 10- day VaR but not in a constant ratio – the relationship depends on both the composition of the portfolio and the pattern of the historical data (in particular trending markets or sharp reversals of earlier moves).

The length of the historical time series affects VaR results but not in a systematic way. A longer time series tends to produce a more stable VaR over time because it is less influenced by short-term trends. Because we use a five year look-back period, our VaR in the last two years has continued to reflect the extreme market moves of 2001 and 2002. It has thus been higher than it would have been using a one year time series.

The chosen confidence interval also affects the result – a higher confidence interval always produces a higher VaR but not according to a statistical formula.