Ionization energies for nickel clusters with four to nine atoms have been calculated and compared to recent experiments. Two different models have been tried, both of them relying on a clear identification of 3d derived and 4s,4p derived orbitals. In the first model, used previously for copper clusters, all metal atoms are described as effective one-electron systems and a 4s,4p electron is ionized. In the second model, 3d ionization is considered by describing one metal atom at the all-electron level. It is shown that the second model is required to explain the absence of odd-even oscillations in the experimental ionization measurements. It is also argued that the fact that the outermost ionization sometimes occurs from a 3d orbital is not in conflict with a model where the 3d orbitals only make small contributions to the binding of the cluster. The calculations show, in agreement with experiment, that the Ni6 cluster is the only one of the investigated clusters which has an ionization energy higher than 6.2 eV and which is thus out of reach for the laser source used in the experiments.