The accuracy of electron backscatter calculations at megavoltage energies is important for many medical physics applications. In this study, EGSnrc calculations of megavoltage electron backscatter (1–22 MeV) are performed and compared to the data from 21 experiments published between 1954 and 1993 for 25 single elements with atomic numbers from 3 to 92. Typical experimental uncertainties are 15%. For EGSnrc simulations, an ideal detector is assumed, and the most accurate electron physics options are employed, for a combined statistical and systematic uncertainty of 3%. The quantities compared are the backscatter coefficient and the energy spectra (in the backward hemisphere and at specific detector locations). For the backscatter coefficient, the overall agreement is within ±2% in the absolute value of the backscatter coefficient (in per cent), and within 11% of the individual backscatter values. EGSnrc results are systematically on the higher end of the spread of the experimental data, which could be partially from systematic experimental errors discussed in the literature. For the energy spectra, reasonable agreement between simulations and experiments is observed, although there are significant variations in the experimental data. At the lower end of the spectra, simulations are higher than some experimental data, which could be due to reduced experimental sensitivity to lower energy electrons and/or over-estimation by EGSnrc for backscattered secondary electrons. In conclusion, overall good agreement is observed between EGSnrc backscatter calculations and experimental measurements for megavoltage electrons. There is a need for high quality experimental data for the energy spectra of backscattered electrons.