Using dispersion relations, we derive the complete virtual QED contributions to Bhabha scattering due to vacuum polarization effects. We apply our result to hadronic corrections and to heavy lepton and top quark loop insertions. We give the first complete estimate of their net numerical effects for both small and large angle scattering at typical beam energies of meson factories, the CERN Large Electron-Positron Collider, and the International Linear Collider. With a typical amount of 1-3 per mil they are of relevance for precision experiments. The determination of the luminosity at lepton and hadron colliders is a necessary task, since in many cases the normalization of the measured cross sections is an observable of direct phenomenological interest. In practice, this task can only be solved by selecting a particular reference process, which is expected to generate large statistics, be as free as possible of systematic ambiguities and predicted by the theory to suitable accuracy. As far as lepton colliders are concerned, the above criteria are fulfilled by Bhabha scattering, i.e., the e e ÿ ! e e ÿ process, where a precision under the per mil level can be achieved on both the theory and the experimental sides [1][2][3].In the last few years, there has been major progress in the evaluation of the corrections at the next-to-next-to-leading order accuracy. In fact, the two-loop QED corrections were first evaluated in the massless case in [4]. The photonic corrections to massive Bhabha scattering with enhancing powers of ln s=m 2 e were soon derived from that [5]. The missing constant term [6] plus the corrections with electron loop insertions [7,8] followed later. Recently, the heavy fermion (or N f 2) corrections were derived in the limit m 2 e m 2 s, jtj, juj [8,9], where m is the mass of the heavy fermion, and soon after also for arbitrary m, with m 2