Though it is accepted that the 3-d magnetic electrons of transition metals such as nickel are itinerant, at high temperature these itinerant ferromagnets act as if the electrons were localized at lattice sites. In particular, three experimental results conflict with the Stoner itinerant model: 1) The spin band gap does not decrease with temperature as the average magnetization, but much more slowly. 2) Spin waves of short wavelength propagate above the Curie temperature. 3) Magnetic degrees of freedom play a role in determining thermodynamic properties n ear and above TC. The source of these discrepancies is the failure of Stoner theory to take into account magnetization fluctuations. In this paper, I do calculations of single particle and spin wave properties in a generalization of Stoner theory devised by R. E. Prange and V. Korenman to take account of fluctuations. In Stoner theory, electrons interact with an effective magnetic field proportional to the average magnetization, which becomes zero at the phase transition. The basic idea of the generalization of Stoner theory is that electrons are sensitive to their local environment and therefore that electronic and spin wave properties should be calculated in the presence of a local slowly fluctuating magnetization configuration. Only after calculating these properties should the fluctuations be thermally averaged. As a result, electrons interact with an effective magnetic field which is basically proportional to the magnitude of the local magnetization vector and which need not become zero at TC. Single particle properties are calculated by making a transformation to the spatially varying frame of reference of the local magnetization and doing perturbation theory with the magnetization gradients as the small perturbation parameter. We find that the spin eigenstates are approximately in or opposite to the direction of the local magnetization. Even when there is no longer a macroscopic magnetization, an energy gap is maintained between spin-split bands, the bands now being defined in terms of the local magnetization direction. The change in the energy gap from its zero temperature value is proportional only to the average square o f a magnetization gradient, a quantity which may be small even above TC. Thus we can understand that the gap changes only slowly with temperature and that the spin wave does not decay into Stoner single particle excitations even at high temperature. A free energy is found which is very similar in form to the free energy used to compute thermodynamic properties in localized models; thus we find that magnetic degrees of freedom are still important in computing thermodynamic properties above TC. It is the existence of a population difference and energy gap, rather than a macroscopic average magnetization that permits the existence of a spin flip collective excitation. We find a secular equation for the spin wave frequency in the presence of fluctuations which is very similar to the usua1 RPA secular equation, except for small perturbations proportional to the square of magnetization gradients. The corrections to the spin wave frequency and lifetime include the effect of the perturbation of single electron energies by the background, and also of the scattering of the spin wave from single particle spin-conserving excitations and from other spin waves. These corrections are quite small and allow for propagation even above TC. Thus it is a prediction of our theory that one see spin waves even above the critical temperature, so long as an appropriate Population difference maintains a locally ordered magnetization.