Comment by meindnoch

It all boils down to the psychoacoustic model in the encoder. It determines how the bitrate is allocated to various parts of the spectrum, depending on the audio signal.