Model risks are present in both appraised price estimates and the assessment of realization risks:
Second order price risks. Model assumptions that prices change in a linear way with an underlying factor can lead to material differences when curvature is present. This occurs with bonds and can be taken account of by using the measure convexity. Callable bonds with embedded options exhibit negative convexity as prices approach the call price. Tranches of mortgage backed securities provide particular challenges depending on pre-payment rates.
These risks are also present with options where this is referred to as gamma risk where the rate of change does not simply vary with the underlying factor but may even reverse sign. Price distribution. Most methods used to assess downside risks from particular trading positions are dependent on price changes being essentially random and having a “normal” distribution. This is not always the case. Some distributions are skewed, with a “long tail” in one direction, while others have a distribution close to that of the normal distribution but have “fat tails”. Fat tails occur when the incidence of occurrences at the extremes is greater than would be expected from a normal distribution.
A further implicit assumption is that price changes are not serially correlated, that is that price changes in one period are independent of price changes in prior periods. This assumption does not always hold.
Reversion to mean. A further assumption often made is that of reversion to mean. This means that past relationships between factors will continue to apply in the future. Examples of these include the following:
Price volatility. Methods to determine the likelihood of losses in a given period exceeding a particular level depends on the volatility of prices of the instruments concerned. A common assumption is that future volatility levels will revert back to historic levels or if they do change over time they change slowly.
Correlations. The level of potential losses for a portfolio of financial assets depends not only on price changes of individual instruments but the way in which such changes are related to one another. The prices of two instruments may tend to move together, or if the price of one instrument rises the price of the other tends to fall. These relationships are core to overall portfolio risks. These relationships are captured in a quantitative way through the use of correlation coefficients calculated on the basis of historic prices.
These correlation coefficients are assumed to remain stable over time. Risk premiums. The same considerations apply to the market equity risk premium and stock specific betas. Trading volumes. In estimating the time necessary to unwind a position the historic average daily trading volume is usually assumed.