The fair value of our contracts always represents the net present value of the expected settlement amount, and the expected settlement amount is a function of the interaction of our Notional Amount ("CRV") and our multiple underlyings at any given time. Somewhat unique to this situation is the fact that the underlyings are merely estimable during the life of the contract and only becomes "objectively verifiable" (as discussed in SFAS No. 133 paragraph 57) at the end of the contract's life after which the vehicle is sold (typically at auction some 30 to 60 days later). In our experience, the "Actual Sales Value," when it becomes observable, dominates the settlement of the contract, effectively overruling what the "Black Book Value" might have predicted the settlement value (e.g., final fair value) to be at the end of the contract. Effectively, since the CRV is fixed at the inception of the contract, the essential questions asked by the guidance in SFAS No. 133 are:

The fact that the "Black Book Value" comes into play in less than 10% of the vehicles coming off lease indicates that, from a qualitative point of view, it is the "Actual Sales Value" variable that is predominant in determining the ultimate settlement loss, answering the first question above.

It is fair to ask whether a quantitative analysis would support our qualitative one. For example, we realize that if hypothetically the "Actual Sales Value" systematically exceeded the "Black Book Value" by, say, a consistent $100 every time, a qualitative statement that the "Actual Sales Value" was "predominant" just because it always determined the settlement amount would ring somewhat hollow. In that hypothetical example, it would seem that both underlyings would correlate highly in terms of the degree to which overall they influenced the settlement amount of the contract, and neither underlying would be particularly "predominant."

We populated our regression analysis with actual settlement amounts regressed against hypothetical settlement amounts as if "Black Book Value" were always the determinant of the settlement. For those settlements actually based on "Black Book Value," the two variables (actual and hypothetical) were identical and, by definition, perfectly correlated. For those majority of settlements based on "Actual Sales Value," the regression tested how close those actual settlements resembled what a pure "Black Book Value" settlement would have been. If those actual settlements had been consistently close in dollar amount to a hypothetical "Black Book Value"-determined settlement, such variables (actual vs. hypothetical) would have plotted tightly against the imagined perfectly correlated regression line. But they didn't.

Our actual quantitative analysis (which gave appropriate weighting to the less than 10% of our contracts where the actual paid loss under our contracts did indeed match the hypothetical Black Book Value Based Loss as in the fourth item in the table above), revealed a common variance ("r-squared") of 45% between the hypothetical Black Book Based Losses and the actual losses paid on these contracts. We believe that 45% r-squared may establish a weak relationship between the settlements based on all the underlyings versus a hypothetical settlement based only on "Black Book Value," we do not believe it evidences a "highly correlated" relationship under GAAP. From this we conclude (in accordance with the last sentence of paragraph 10(e) of SFAS No. 133 as clarified further by paragraph 254) that the behavior of all underlyings in combination do not behave in a manner highly correlated with the behavior of "Black Book Value," the component variable that does not qualify for an exceptio n from SFAS No. 133.

Paragraph 254 of SFAS No. 133 states that if the contract's underlyings are a combination of variables ("Black Book Value" and "Actual Sales Value" in our case) and "its settlement (emphasis added) is expected to change in a way that is highly correlated with the way it would change if it was based on an underlying that would not be eligible for one of the exceptions in paragraph 10(e)," then SFAS No. 133 would be applicable. Since settlement of the contract is not expected to change in a way that is highly correlated (i.e. r-squared is 45%) with the way it would change if it was based solely on Black Book (the variable not eligible for the exception), we concluded that the contract is not subject to SFAS No. 133.