Do quasar broad-line velocity widths add any information to virial black hole mass estimates?
We examine how much information measured broad-line widths add to virial BH mass estimates for flux limited samples of quasars. We do this by comparing the BH mass estimates to those derived by randomly reassigning the quasar broad-line widths to different objects and re-calculating the BH mass. For 9000 BH masses derived from the H_beta line we find that the distributions of original and randomized BH masses in the M_BH-redshift plane and the M_BH-luminosity plane are formally identical. A 2D KS test does not find a difference at >90% confidence. For the MgII line (32000 quasars) we do find very significant differences between the randomized and original BH masses, but the amplitude of the difference is still small. The difference for the CIV line (14000 quasars) is 2-3sigma and again the amplitude of the difference is small. Subdividing the data into redshift and luminosity bins we find that the median absolute difference in BH mass between the original and randomized data is 0.025, 0.01 and 0.04 dex for H_beta, MgII and CIV respectively. The maximum absolute difference is always <=0.1 dex. We investigate whether our results are sensitive to corrections to MgII virial masses (e.g. Onken & Kollmeier 2008). These corrections do not influence our results. Moreover, the mass residual - Eddington ratio correlation discussed by Onken & Kollmeier is more directly attributable to the slope of the relation between H_beta and MgII line width. Our results imply that the measured quasar broad-line velocity widths provide little extra information, after allowing for the mean velocity width. In this case virial estimates are equivalent to M_BH L^alpha, with L/L_edd L^(1-alpha) (with alpha~0.5). This leaves an unanswered question of why the accretion efficiency changes with luminosity in just the right way to keep the mean broad-line widths fixed as a function of luminosity. (abridged)