It's based on sale surveys, on the one hand. Fifty per cent of the data that are used are sales data. So if someone sells albums, it can be assumed they will eventually be copied.
The other 50% are broadcasting data. So if someone is regularly broadcast on the radio or elsewhere—it can be a quite limited number of times—that person's name will appear in the broadcasting surveys and money will be sent to that person.
So it is assumed that the artists who sell music and artists who are broadcast on the radio are more likely to be copied than others.