This paper presents the Hessenberg method, a new sparsity-based small signal stability analysis program for large interconnected power systems. The Hessenberg method as well as the Arnoldi method computes the partial eigen-solution of large systems. However, the Hessenberg method with pivoting is numerically very stable comparable to the Householder method and thus re-orthogonalization of the krylov vectors is not required. The fractional transformation with a complex shift is used to compute the modes around the shift point. If only the dominant electromechanical oscillation modes are of concern, the modes can be computed fast with the shift point determined by Fourier transforming the time simulation results for transient stability analysis, if available. The program has been successfully tested on the New England 10-machine 39-bus system and Korea Electric Power Co. (KEPCO) system in the year of 2000, which is comprised of 791-bus, 1575-branch, and 215-machines. The method is so efficient that CPU time for computing five eigenvalues of the KEPCO system is 3.4 sec by a PC with 400 MHz Pentium IIprocessor.