Performance evaluation of algorithms for Black-Derman-Toy lattice

Within the framework of sensitivity of the optimal value of the portfolio management problem described in Dupaeová and Bertocchi (1996), Dupacová and Bertocchi (1997) with respect to lattice calibration, we compare Bjerksund and Stensland approximation algorithm, Kang Pan-Zenios algorithm and a modified Kang Pan-Zenios algorithm to generate short-rate interest rates tree according to Black-Derman-Toy model. Numerical testing of the behaviour of the three algorithms are given. The necessary inputs for Black-Derman-Toy model are yield curve and log-yield volatilities: we provide an evidence on the relatively large sensitivity of the parameters of the fitted lattice on the chosen volatility curve. The reported numerical experience is based on data from the Italian bond market.