What's new

P2.T7.304. Model Risk


Thread starter #1


304.1. Each of the following statements about a firm's value at risk (VaR) model is true EXCEPT which is false?

a. In comparison to an operational VaR model, a market risk VaR model is both more likely to be driven by internal historical data and to utilize backtesting as a the primarily means of validation
b. Each of market risk VaR, credit risk VaR, and operational risk VaR are similar in their intention to estimate the quantile of a loss distribution
c. If two risk managers employ the same confidence interval, time horizon and high-level approach (e.g., historical or Monte Carlo simulation), their portfolio VaR estimates should be nearly identical
d. The confidence band around a VaR estimate, which is a direct function of its standard error, will tend to narrow (decrease) with more historical data and widen (increase) with higher confidence levels

304.2. Your colleague Brian makes the following four assertions about financial "model risk." Three of his statements are supported by the FRM. But which one of his statements is hard to defend, if not false?

a. All social science models are "wrong" in the sense that the most they can hope to achieve is to be successful approximations and/or abstractions under certain conditions
b. Although with proper training (e.g., the FRM credential), model risk should be easy to identify and measure
c. Model risk includes data input risk, estimation risk (e.g., parameters of model), model selection risk, and implementation risk
d. Valuation risk is often reduced by marketing-to-market rather than marketing-to-model, but always marketing-to-market is not without its problems

304.3. According to Malz, model error (or model "defects") was a significance CAUSE of the substantial inaccuracy of credit rating agency models, notably the models of Moody's and Standard & Poor's, in rating subprime residential mortgage-backed securities (RMBS). He does allow of course for other causes; e.g., conflicts of interest. The model errors became painfully evident in "higher-than-expected defaults rates" that began to materialize in 2007. With respect to the credit agencies' model errors in underestimated default risk of these RMBS, each of the following was a material cause EXCEPT for which was the least material cause?

a. The PSA rates were too optimistic
b. In general, the rating agency models assumed positive future house price appreciation rates
c. These securtization products were often mapped to time series of highly rated corporate bond spread indexes
d. Correlations among regional housing markets were assumed to be low