I just had an article published in the the Value Examiner about the proper way to compute risk premiums for small businesses. The leading data providers for valuators have believed that a method based on arithmetic means is proper. I disagree and the article goes into depth as to why. The quick version is that people do not buy small closely held businesses with a one year investment horizon but rather expected holding periods of 5 years or longer. Longer holding periods have lower expected rates of return and that is justified by the relatively lower volatility that goes along with those longer periods. The risk/reward relationship holds.
The model used to justify the arithmetic mean also assumes that, like a coin toss, every period is independent from the next. That is not true. Periods of higher than average returns tend to be followed by periods of lower than average returns. The nature of markets is that getting 10 up years in a row is less probable than a model based on independence would predict. This is because markets tend to bubble and crash if they rise for prolonged periods. We saw it in technology stocks in the late 90's or in Japan in the 80's.
If you see company size risk premiums that are higher than 4%, the valuator is not doing a competent appraisal.