| Risk-neutral derivative valuation often requires the identification of an unobservable market
price of risk. A popular approach to this problem assumes a risk specification and estimates
its parameters using the prices of traded securities. Derivatives are then valued using the
parameterized price of risk. This paper shows that the results of such an approach are quite
sensitive to the assumed risk specification. First, I calibrate several widely used specifications
of the price of interest rate risk using a common dataset of discount bond prices. I then value
coupon bonds and bond options and find that values vary significantly across risk specifications.
Second, I calibrate the risk specifications using Eurodollar futures options data and find that a
term structure of risk tightens in-sample fit and reduces out-of-sample valuation error. These
results highlight the importance of testing assumptions about the price of risk when using
risk-neutral valuation. |