In the prediction models of tropospheric scintillation on earth-satellite paths from Karasawa, Yamada, and Allnutt and ITU-R, the frequency dependence of scintillation is expressed as a power law with a different exponent for each model. In this paper, this is verified using a collection of measurement results from different satellite links in Europe, the U.S., and Japan at frequencies from 4 to 50 GHz and elevation angles from 2.5 to 52°. It shows that the exponent of the power law varies widely among the results from the different sites. Possible explanations of this are: 1) the frequency dependence of scintillation due to cloudy turbulence is different from that due to clear-sky turbulence and this kind of scintillation may be present to different extents in the various databases due to climatic differences and different clear-sky selection procedures or 2) angle-of-arrival fluctuations due to turbulence have a different frequency dependence and this effect may have some impact on the measured scintillation at some of the sites.