Published galaxy power spectra from the 2dFGRS and SDSS are not in good agreement. We revisit this issue by analyzing both the 2dFGRS and SDSS DR5 catalogues using essentially identical technqiues. We confirm that the 2dFGRS exhibits relatively more large scale power than the SDSS, or, equivalently, SDSS has more small scale power. We demonstrate that this difference is due the r-band selected SDSS catalogue being dominated by more strongly clustered red galaxies, due to these galaxies having a stronger scale dependent bias. The power spectra of galaxies of the same rest frame colours from the two surveys match well. It is therefore important to accurately model scale dependent bias to get accurate estimates of cosmological parameters from these power spectra.