Runge-Kutta timestep schemes have been used recently in an attempt to increase the maximum stable timestep for IMPES (implicit pressure, explicit saturation) simulation of oil recovery. It has been claimed that anm-stage method will increase the real stability boundary bym2. A rigorous stability analysis for a Buckley-Leverett problem shows that this claim is false, and that previous stabilized IMPES schemes are never more efficient than ordinary IMPES. Test calculations support these results.