A Bayesian approach to the empirical valuation of bond options

In this paper we propose and implement a Bayesian procedure for the empirical valuation of bond options given the observed term structure of interest rates, and given assumptions about the time series behavior of the instantaneous spot rate. The Bayesian approach is motivated by the extreme multicollinearity in the cross-sectional data. The multicollinearity is caused by some local identification problems in the likelihood function. These same singularities motivate the choice of prior. The proposed method is applied to a dataset of Dutch bond prices.