A multilevel algorithm for the statistical characterization of the radiation patterns of beam steered conformal arrays is presented. The algorithm can be used to obtain average complex field patterns and power patterns in the presence of random amplitude and phase excitation errors. The computational scheme is based on a hierarchical decomposition of the array into smaller sub-arrays. At the finest level of decomposition, the radiation patterns of single element arrays are computed or measured over a sparse grid of directions. The subsequent computational sequence comprises interpolations and aggregations of sub-array contributions repeated until obtaining the radiation pattern of the array. The proposed algorithm attains a computational complexity substantially lower than that of the direct computation and thus can be employed for Monte Carlo type statistical simulations.