Beyond Business Intelligence

Predictive Science for Today's Bottom Line

PREDYCT ERM is a cash flow based model that marks all financial investments to market. PREDYCT ERM marks insurance liabilities to model, i.e., the value of the liabilities is the expected value of their future payments over the life of the obligation present valued to the horizon. Earnings in PREDYCT ERM are defined as changes in the value of net worth (assets minus liabilities) between today and 1-year horizon. For investments, this is not dissimilar to statutory accounting principles where such factors as realized and unrealized gains/losses are either added to or netted from a firm’s surplus. The difference, however, is that statutory accounting principles do not run these credits and debits through the income statement. PREDYCT ERM, on the other hand, converts the firm’s income statement from accounting values to mark-to-market values so that realistic rates of RAROC can be attributed in the current accounting period.

**Fixed income portfolio**

Predyct ERM calculates market value of each individual instrument for every simulation scenario. The resulting portfolio value at the horizon comes as an average of all scenarios. PREDYCT ERM uses zero-coupon curves and assumes continuous compounding then estimating an asset value. At this moment we utilize the following curves: USD, EUR, GBP, JPY, AUD, HKD, and KRW. Predyct ERM simulates individual interest rates for a specified set of maturities and captures both parallel and non-parallel shifts in the yield curves. Therefore interest rate risk incorporated in fixed income portfolio will come naturally from the resulting distribution of scenario outcomes. Simulation parameters, like volatilities and correlations between rates are calibrated from the 10 year historical data for all of the above zero-coupon curves.

*Risk free (Sovereign) Bond:*

We calculate today’s value of the bond by discounting each coupon as well as the principle payments with the current zero rates as provided by Bloomberg. The expected horizon value is calculated by discounting coupon and principle payments to the horizon using the simulated zero rates. For coupon payments that are within the time horizon, Predyct ERM assumes that these cash flows earn the interest at the constant risk - free short - Predyct ERM rate adding to the horizon value of the instrument.

*Corporate Bond:*

Predyct ERM assessment of corporate bonds is similar to that of the risk-free bonds. The major difference is that Predyct ERM utilizes corporate curves instead of zero curves for discounting when calculating today and horizon values. We also presume that credit rating of the security can migrate over the horizon period, thus changing the security’s value. Credit risk, therefore, is an uncertainty in future value of the assets associated with credit rating migration in each simulation scenario. Predyct ERM models the change in the credit quality of the bond over the specified horizon through the use of Merton model which links the return on a company’s stock with its probability of being upgraded or downgraded within one-year period. Recovery rates of the defaulted bonds are calibrated from the historical experience. At this moment Predyct ERM assumes that credit spreads stay constant over the horizon period.

Many sovereign and corporate bonds carry a call provision which grants the issuer an option to retire (or “call”) the bond prior to its maturity. The callable bond value therefore equals the “optionless” bond value, less the call option value. Predyct ERM utilizes the Hull-White model of interest rate evolution to calculate a value of the callable bond for each simulation scenario.

*Municipal Bonds:*

Predyct ERM valuation of municipal bonds is similar to that of corporate bonds, but instead of corporate curves Predyct ERM uses municipal curves as provided by Bloomberg. Since most of municipal bonds are tax-exempt, Predyct ERM adjusts their coupons proportionally to tax rate.

*Preferred stock:*

Usually insurance companies hold only a small portion of the investment portfolios (up to 2 percent) in preferred stocks. More than 90 percent of them are of callable non-convertible cumulative types which are similar to corporate bonds. Predyct ERM treats preferred stock as callable corporate bonds with the identical maturity, face value, coupon percentage and rating. If stock does not have a maturity, Predyct ERM assigns it the longest maturity tPredyct ERM available for a given currency. Given that preferred stocks are subordinate to even the least senior bonds, Predyct ERM sets their recovery rate to zero.

*ABS/MBS:*

When calculating risk for ABS/MBS securities, one must take into consideration that these instruments (unlike bonds) carry a prepayment provision granted to the borrower. Predyct ERM does not implement prepayment models, instead we rely on analytics provided by CMS BondEdge®. We use Bondedge software and extensive database to

- get today prices of the ABS/MBS securities
- calculate expected prices at the horizon given that yield curves do not change over one-year period
- produce forward key rate durations and convexities to adjust horizon prices for potential non-parallel shifts in yield curves
- calculate interest and principal cash flows from the ABS/MBS portfolio over the one-year period. Again we assume that they earn the interest at the constant risk-free short-tPredyct ERM rate up to the horizon

**Equity portfolio**

Predyct ERM assumes that firm’s return could be sufficiently explained by the index return of the industry classification to which the firm belongs, with a residual part that can be explained solely by information unique and specific to the firm. PREDYCT ERM utilizes the following industry indexes, as defined by Dow-Jones:

- Basic Materials
- Consumer Cyclical
- Energy
- Financial
- Aggregate Index
- Healthcare
- Industrial
- Consumer Noncyclical
- Technology
- Telecommunications

Predyct ERM simulates fluctuations of the return on each of these indexes in every scenario. Volatilities and correlations required for simulations have been obtained from historical price series for each of the above indexes. Note that Predyct ERM does not predict expected returns; we can make assumptions only about the size of fluctuations and their coherence across different indexes. We suggest that Predyct ERM users provide their estimates for the anticipated returns on each of the above industry classifications.

Firm - specific risk can generally be considered to be a function of company asset size; to capture this risk we utilize data on historical volatility of the firm’s stock.

The simulations result in a distribution of equity portfolio value over one year horizon; required risk measures could then be calculated from this distribution.

*Portfolio Management*

The portfolio manager can use Predyct ERM to assess the RAROC and capital attribution to each CUSIP (or portfolio of CUSIPS). The portfolio manager can add or subtract financial instruments (or portions of instruments) to assess the impact to the portfolio’s risk and return. Using this procedure the portfolio manager can use Predyct ERM to construct an efficient portfolio defined as being a combination of investments that offer either the highest possible return at a given level of risk or the lowest possible risk at a given level of return. This application can be applied at the investment level only (meaning just financial instruments) or to include all of the institutions’ liabilities.