Challenges in Capital Adequacy UHGEMI 3 rd Annual

Challenges in Capital Adequacy UH-GEMI 3 rd Annual Energy Trading & Marketing Conference: Rebuilding the Business Houston, Texas January 20, 2005 Laurie Brooks VP Risk Management and Chief Risk Officer Public Service Enterprise Group UNIVERSITY of HOUSTON Global Energy Management Institute

Capital Adequacy and Capital Allocation Connected? • Capital Adequacy – How much capital is required to achieve the company’s stated goals and objectives? • Capital Allocation – How should corporations allocate capital between competing demands? 2

Capital Adequacy for Energy Transactors 1. Capital for what? Business models: regulated utilities, merchant generators, marketing and trading entities Economic capital vs liquidity adequacy Banking models S&P liquidity survey Measures - Ea. R vs CFa. R, role of stress testing, market risk vs credit risk trade-offs, role of ECE and PFE 2. Why energy is different - impact of following on margin/cash requirements: volatilities sector ratings storability regulatory intervention age and depth of markets contract terms risk mgt tool availability 3. Capital how? Access to capital markets Diversification of cash flows Credit mitigations role of netting and clearing stair stepped margining agts. 3

Capital Use by Activity Utility Merchant Generator Marketer/ Trader Assets Pipes & Wires, Customers Generating Facilities People, IT Protection Insurance, Va. R Plant Cash collateral Maintenance Plant, customer satisfaction Growth Acquisition of service territories New facilities New products, services, markets Multiple Venture capital 4

Market Risk – Trading vs. Non-Trading Activities Non-Trading Purpose Liquidity • • Positions to facilitate marketing Proprietary trading positions • • Positions generated by asset/customer business Strategic “buy and hold” hedges • Liquid, actively funded positions across many markets Holding period measured in days/weeks • • Illiquid or “buy and hold” positions Holding period measured in months/years Price-driven exchange traded or OTC options Short holding period allows linear approximations • • Asset/customer-driven embedded options Long holding period makes non-linearity material • • Short-term volatilities and correlation Jump diffusion, intra-day Va. R – analytical, simulation • • Long-term volatilities and correlation Mean reversion, seasonality simulation, Earnings at Risk • Va. R limit reduction, stop loss limits, hedging with traded instruments • Structured solutions, contract renegotiations, asset sales and purchases Management of regulatory process • • Optionality Valuation Risk Management/ Intervention • • 5

Key Concepts of Capital Adequacy: Three Risk Types The framework for determining capital adequacy for economic value requires an estimation of economic capital and thus quantifying the following significant risks: • Market Risk - Variation of portfolio market value due to a change in a market price or rate, as well as a change in energy demand • Credit Risk - Variation of portfolio market value due to default or a credit downgrade of an issuer or counterparty • Operative Risk (term to address Operations and Operational risk collectively) – Operations - The risk associated with delivering or producing physical energy – Operational - The risk of direct or indirect loss resulting from inadequate or failed internal processes, people, and systems or from external events 6

Key Concepts of Economic Capital Adequacy: Market Risk Modeling Approaches Price Behavior Process Market Exposures Pros/Cons Analytical Closed-form approach for modeling price movements Works well for linear Pros: type exposures • Simple and fast • Easy to change as assumptions change Cons: • Does not capture optionality well • Minimal ability to model complexities over a longer period of time Simulation Robust methodology for mean reversion, jumps, linking, spot, and forward prices Full revaluation at each price iteration better approximates nonlinearity of asset/option positions Comments • Works well for determining shorter-term price moves for a trading portfolio • Can be used as a quick metric to help manage portfolio positions Pros: • • Robust • Captures optionality • Provides a full distribution of outcomes Cons: • Complex to construct the simulation model • Only as good as model input parameters • For historical simulation, values are constrained to conform to history which may be irrelevant due to market, economic, or regulatory changes 7 As the time horizon is extended and the need to model certain energy price characteristics increases, simulation becomes a more suitable solution. Meanwhile, the technical difficulties increase and the model needs to be modified to fit the long-term simulation purpose.

Key Concepts of Economic Capital Adequacy: Credit Risk Expected Loss – Represents the average loss that a company could expect to incur over a given horizon Unexpected Loss – Measures the uncertainty of losses around the expected loss Probability Portfolio Expected Loss (Mean) Expected Loss (Loss Provisions) Credit Economic Capital (Unexpected Loss) Distribution of Portfolio Credit Losses Over a One-Year Time Horizon 8 Confidence Level

CA Framework – Key Concepts of Economic Capital Adequacy: Operative Risk – Scorecard Approach • Can be used for operations and operational risk to identify risks, determine frequency and range of costs, and assesses the effectiveness of controls and mitigation techniques in place. It is subjective, but now that the SEC has mandated the COSO framework for Sarbanes Oxley 404 compliance, standards will be set. In particular, the Capability Maturity Model can be adapted to set standards for a scorecard approach and is already used by many audit firms. Additionally, a company may want to use CCRO Best Practices from earlier white papers as a qualitative assessment of where companies stand with regard to CCRO recommendations. • Regardless of the scorecard criteria used, a scorecard approach can form the basis for continuous improvement processes for internal controls to mitigate operative risk. It can also reflect improvement in the risk-control environment in reducing the severity and frequency of future losses. 9

CA Framework – Key Concepts of Economic Capital Adequacy: Operative Risk – Risk Taxonomy • The risk taxonomy is a system for organizing types of operative risks by serving as a family tree, aggregating risks by various characteristics. The level of aggregation at which each characteristic presents itself may be determined individually. • There is no standardized risk taxonomy, but certain characteristics should be used to create the groupings: – Risk classes (people, processes, systems, asset damages) – the broadest classes of risks – Subcategories – could include whether the risk is internal or external, a type of fraud, or a natural disaster – Risk activity examples – specific activities or events that could cause a loss, such as rogue trading, hurricane, model risk, or pipeline rupture. 10

Key Concepts of Liquidity Adequacy • • Fixed Payments - This would include, but is not limited to; fixed charges such as debt service, dividends, debt/equity retirement and current portion of committed, maintenance and non-discretionary capital expenditures. Contingent Liquidity – Contingent liquidity is synonymous with unexpected change or variation in liquidity. While economic capital protects against losses in the company’s economic value, contingent liquidity is held to support the risk of unexpected reduction in cash. Includes: – Cash Flow at Risk – Trigger events: • Downgrade event – Loss of threshold – Adequate assurance • Debt/equity trigger – Contingency events: • Operational/Operations Risk • Credit/counterparty termination default 11

CA Framework – Key Concepts – Combined Capital Methodology Description Advantages Disadvantages Assumption Simple Sum Derive economic • Easy to implement capital for credit, • Most conservative market, and operative view of risk, then sum them • Overestimates risk Correlation assumed to • Results in the lowest be perfect among risk level of capital components adequacy Modern Portfolio Theory From historical data, determine an explicit correlation among credit, market, and operative risk economic capital Attempts to represent the actual correlation among risks, rather than a conservative assumption Requires a time series of credit, market, and operative risk economic capital that is reasonably robust Monte Carlo Simulation Using consistent parameters, simulate risk factors to produce a joint distribution of outcomes The most robust perspective of risks and their interaction if modeled correctly • Requires a large Material risk inputs can amount of research, be parameterized analytical, and accurately technical resources • Ensuring assumptions are correct is critical 12 Assumes that some risks are uncorrelated, allowing for lower risk and improved capital adequacy

CA Framework – Key Concepts – Correlation Math Refresher In a two asset portfolio with equal investment in assets A and B, the Va. R of the portfolio (at 95% confidence) Va. RA+B = 1. 65 * AB where AB is the standard deviation of returns of the portfolio: where AB is the correlation between A&B (do the returns move together? ) Remember (a+b)2 =a 2+2 ab+b 2 and Then if AB =1 So Portfolio Va. R = Va. RA + Va. RB! If AB=0, (Square root sum of squares) The truth 0 < AB < 1 lies somewhere in between and: < AB Square root sum of squares 13 < A+ B Simple Sum

Example The Risk Management team at PSEG demonstrated the CCRO’s framework using a sample asset portfolio. • • • This example illustrates how the CCRO framework can be used in practice We will walk you through the following implementation steps: – Portfolio setup – Methodology – Pre-simulation – Simulation – Results We will also discuss some of the firm and systems resources required Please refer to pages 61 -67 of the white paper for a full description of the example. 14

Example – Setup We chose to model the asset-level impacts over a year • • • of different risks on a company over time. We modeled market, credit and operative risks jointly in one simulation versus separately – Felt there was better intuition and that we could better justify a choice of the assumptions – Calculation process seemed clear based on this approach – Used a 1 -year holding period and ran 5, 000 trials with a 95% CI We modeled a five-year time horizon, with price changes modeled as follows: – Year 1: spot – Year 2 -5: forward prices We chose a variety of assets and parameters. – Three different generating assets and fuel types – Assets are in three different pools Generating Plant Gas-fired combined cycle Coal-fired, base load Jet kero-fired peaking Power Pool Capacity VOM Heat Rate Fuel Type Book Value ECAR 850 3. 98 7. 25 Natural Gas $510, 448, 931 NEPool 375 2. 51 10. 3 Coal $49, 720, 351 PJM 500 34. 48 15. 7 Jet Kero $11, 094, 684 15

Example – Setup Market Risk Calculations • Unhedged market risk – Minimum [(realized generation over 12 months) + (Expected generation value of the remaining term)] – (Initial expected value of the generation) • Hedged market risk – (Unhedged market risk) + (Realized and unrealized trading profit or loss) 16

Example – Setup Credit Risk Calculations Counterparty A Counterparty B Counterparty C Rating CCC BBB BB 1 -Year Probability of Default 27. 87% 0. 34% 1. 16% Commodity Fuel – coal, natural gas, jet kero Power – NEPool, PJM, Cinergy Fuel and power • Calculated as the sum of credit loss across the twelve months of simulations, as a function of counterparty risk and power pool risk • The company has three counterparties – – Counterparty A is used for fuel procurement Counterparty B is used for power sales Counterparty C is used for speculative trading. The recovery rate is assumed to be 10%. • Each power pool has collateral requirements that are a function of the company’s credit rating, tangible net worth and activity in the pool – Value is calculated under two potential ratings, BBB (credit limit $80, 000) and BB (credit limit $4, 000) 17

Example – Setup Operative Risk Calculations • Operations loss – Sum of lost profit from plants not running at full capacity • Operational loss (if applicable) – Hidden trade on the books whose value is set to the largest negative value of all the trading positions on the book. 18

Example – Setup Liquidity calculations Liquidity risk is defined as the minimum cash flow point in a simulation. § Prior month realized P/L (retained earnings) § Current month generation P/L § Collateral posted § Accounts receivable § Accounts payable § Full margin on mark-to-market § Credit loss § Operations loss § Operational loss Monthly cash flow 19

Example – Setup Hedging affects liquidity in offsetting ways. • Liquidity risk is increased by hedging in the following ways – Creates cash outflows due to full margining on mark-to-market – Creates the possibility of credit loss • Liquidity risk is decreased by hedging in the following ways – Decreases the amount of cash needed to be posted to power pools since that is determined by net activity. – Decreases the distribution of realized P/L from generation The net effect of hedging was a decrease in the liquidity risk. 20

Example – Methodology Three key methodology choices drive our model Method Risk modeling Energy forward prices Daily power prices Pros Joint simulation of • Consistency credit, market, and • More data available to check operative risks (versus micro relationships rather than assumed correlations) portfolio relationship • Can change micro assumption and rerun • Are not assuming answer Correlated Brownian Motion for Energy Forward Prices Cons • Increases memory need and computer time • Necessitates more simplifying assumptions, leading to less accurate estimates of component risks • Most practical method with 3 • Easier to believe forward power pools and 3 types of fuel prices rather than spot prices for 5 years still oversimplifies reality • Would be difficult to jointly • Probably overstates volatility for calibrate more complex model for longer-dated contracts diversity and tenure of portfolio Daily power prices are • normally distributed with mean equal to forward price and • standard deviation equal to historical daily spot standard deviation Allows for analytical determination • of MWs of generation and generation value • No need to do daily simulation Ignores operating constraints on plants Splitting monthly prices into two normal distributions (normal and extreme days) captures peaking value more accurately • Does not allow for fuels to vary by day 21

Example – Pre-Simulation: prior to running our simulations, we calculated a number of initial values. Pre-Simulation Calculations • Initial expected value of the assets – Calculated based on the current forward prices for fuels and power • Expected fuel purchases and expected output to be sold to counterparties – Calculated based on current forward prices • Randomly-generated positions in power and fuels – Constrained to be a quarter of the size of outright positions – Used to simulate a speculative trading operation 22

Example – Simulation: we generated the inputs to credit and operational performance. Market risk simulation* Correlated forward prices - power Correlated forward prices - fuel Generation model Marginal cost of fuel (VOM & heat rate) MTM - A/R A/P on trading contracts Credit risk simulation** Credit excess/loss Probability of default Operational profit/loss Probability of outage Operative risk simulation** Market risk Probability of trader misconduct * 60 product months x 6 products x 12 monthly steps of random standard normal pulls ** 7 risks x 12 monthly steps of uniform random variables pulled 23

Example – Results – Unhedged vs. Hedged Assets Unhedged By hedging assets, market risk is reduced by less than the additional economic capital required for credit risk, increasing economic capital adequacy. Hedged Note: the simulation was also run with all counterparties set at BBB to reflect the average rating of many portfolios. The credit risk remained at zero with a 95% confidence level, while market risk was reduced from $23 million to $6 million. 24

Example – Results – Portfolio Effect Illustration of the mathematical fact: EC = 0 (square root sum of squares) < EC < < 1 (Monte Carlo simulation) < EC =1 (simple sum) Available vs. Required Capital ($ millions) Sq. Root Sum of Squares Monte Carlo Simulation Simple Sum 285. 6 Market Risk 22. 5 Credit Risk 0. 0 23. 2 -13. 4 -11. 8 0. 0 32. 3 33. 9 45. 7 253. 3 251. 7 239. 9 Net Assets - Debt Required Economical Capital Operative Risk Diversification Effect - Across Risks Total Required Economic Capital Adequacy Disclaimer: the closeness of the Monte Carlo (MC) and Square Root Sum of Squares is not representative. In general, one shouldn’t assume that SRSS is a good proxy for MC. Available vs. Required Capital ($ millions) By analyzing capital requirements for unhedged assets as part of a portfolio vs. individually, the example illustrates how diversification reduces the economic capital required for market and operative risks. Total Individual Peaking Assets Total Portfolio Diversified Component Risk Coal Combined. Cycle Net Assets 49. 7 510. 4 11. 1 571. 3 Debt 24. 9 255. 2 5. 5 285. 6 Market Risk 7. 0 27. 6 3. 5 38. 1 22. 5 -15. 7 Credit Risk 0. 0 0. 0 22. 3 3. 4 2. 3 27. 9 23. 2 -4. 7 -11. 1 -2. 9 -1. 6 -15. 6 -11. 8 3. 8 18. 2 28. 1 4. 1 50. 5 33. 9 -16. 5 6. 6 227. 1 1. 4 235. 2 251. 7 Required Economical Capital Operative Risk Diversification Effect - Across Risks Total Required Economic Capital Adequacy 25

Example – Results Why Emerging Practices? • • • These are recommendations for internal use and experimentation for companies to better understand quantify the capital and cash requirements of the merchant energy business; these are not recommendations for external communication or new disclosure. No one is going to implement all of these recommendations over night. Most of us have some capability to begin looking at the components of Capital Adequacy and liquidity requirements through the use of tools that we already have in place but which require extension and modification to achieve the more sophisticated views that result from the white paper recommendations. This should be a controlled evolutionary process - in most cases, the less sophisticated tools that we already have in place generate more conservative answers than the sophisticated approaches do. Why we will implement these ideas over time: • Better than what we have now • Emphasize need to look both long term and short and to look at cash flow as well as earnings and value • Ideas and methodologies useful in decision making 26