Does higher computational complexity associate with higher model estimation risk? Some implications for the practice of risk management from term structure modeling
摘要
Quintessential to the practice of risk management is the appropriate modeling of the term structure of interest rates. Arguably the most popular class of dynamic term structure models is the class of affine term structure models (ATSM). Authors contend that the class of ATSM is widely appealing because of its empirical tractability. One property of ATSM that researchers note as attractive is the reduction of the bond pricing problem to the solution of Riccati ordinary differential equations for which the imposition of parameter restrictions yields closed form expressions that have been shown to yield some computational benefits (e.g., reduced optimization times), nevertheless, we show that employing closed form analytic solutions oftentimes brings a higher degree of computational complexity and this tends to associate with higher model estimation risk implying the existence of limitations to the benefits of analytic solutions. The use, and misuse, of such models has important implications for both valuation and decisions made in risk management and so we conclude with some implications for our findings that are relevant to this field.