The aim of this study is to determine a robust prediction algorithm that can be used to correct the measured specific absorption rate (SAR) in a homogeneous phantom when its complex permittivity deviates from standardized reference values. Results are analyzed over a frequency range of 30-6000 MHz. Both measurements and numerical simulations are presented. Several antenna sizes and distances to the phantom are investigated so as to study a large range of SAR distributions. It is demonstrated that the prediction algorithm, while developed using dipole antennas, also works well for mobile telephone models. Employing the prediction algorithm reduces the SAR measurement uncertainty, thereby improving the reproducibility of SAR compliance assessment between laboratories. Another benefit of the algorithm is that it enables the use of broadband tissue-equivalent liquids, whose dielectric parameters are not currently within the tight tolerances of existing standards. The use of broadband liquids reduces the cost of SAR measurement. The method presented in this paper is of benefit to the IEEE 1528 and IEC 62209 measurement standards.