We have calculated the integrated ULF wave power in the Pc5 band at two stations, Kevo (part of the International Monitor for Auroral Geomagnetic Effects (IMAGE) magnetometer array in Scandinavia, at auroral zone latitudes), and Cape Dorset (part of the Magnetometer Array for Cusp and Cleft Studies (MACCS) in Arctic Canada, at cusp latitudes), and compared this power against the solar wind velocity for the last six months of 1993, a period characterized by two persistent high-speed solar wind streams. We find for both local noon at Cape Dorset, and for local morning at Kevo, the Pc5 band power (0.002--0.010 Hz) integrated over a six-hour period exhibits a clear power-law dependence on the solar wind velocity. At Cape Dorset we found power &agr; Vsw4, with a correlation coefficient r=0.73, and at Kevo we found power &agr; Vsw6.5, with r=0.74. Much of the remaining variation in Pc5 power is due to temporal patterns evident at both stations in response to recurrent high speed streams. Power was strongest at the leading edge of each high speed stream and subsequently decreased more quickly than Vsw. Our observations suggest that it is insufficient to make estimates of Pc5-range ULF wave power on the basis of Vsw alone: one must consider other physical factors, either intrinsic to the solar wind or related to its interaction with Earth's magnetosphere. The Kelvin-Helmholtz instability is often considered to play a dominant role in this interaction, and the level of instability depends on both velocity and density. By means of a simple simulation using typical density and velocity values during the passage of a high speed stream, we were able to obtain good agreement with the temporal variations we observed. Finally, this study indicates that ground-based pulsation observations can provide reliable proxies of the initial passage of high speed solar wind streams past Earth. ¿ 1998 American Geophysical Union