Experimentally determined rate of the reactor electron antineutrino capture on protons is about 6% lower than expectation. This cannot be explained as arising from the known oscillations among three neutrinos; instead, it might indicate existence of additional, sterile, light neutrinos. Unlike other indications of sterile neutrinos, the reactor anomaly is statistically well established, however, it depends crucially on the calculated reactor spectrum. I will review how the reactor spectrum is determined, and point out possible uncertainties. These include correction to the beta-decay spectrum caused by the nuclear finite size and by the weak magnetism, and in particular corrections caused by first forbidden beta decays that account for ~ 25% of all decays. In addition, recently discovered shoulder or bump in the reactor spectrum, seen in Daya-Bay, RENO, and Double-Chooz experiments, was missed in previous experiments and was not predicted in the calculated spectrum. This is a further indication that theoretical uncertainties of the reactor spectrum were underestimated. Whether light sterile neutrinos exist need to be directly tested. However, reactor anomaly cannot be counted as a strong argument for their existence.