A brief comparison of methods used to calculate the electronic work function of polycrystalline metal surfaces is presented. We show that the ab initio methods are unable to give values in acceptable agreement with experiments, whereas simple models based on the free-electron gas approach and classical electrodynamics work well, even in complicated electronic systems like transition metals, lanthanides, or actinides. This observation is made on the basis of previously published computations by the authors and new results for lanthanides and actinides. Surface valence transitions from trivalent to divalent for Sm and Tm are indicated. Heavy actinides work functions are presented. The results are compared with available experimental and theoretical values, and an attempt to address the reasons for ab initio failure is undertaken on the basis of a concept of quantum decoherence.