We cross-correlate the SDSS DR3 quasar sample with FIRST and the Vestergaard et al. black hole (BH) mass sample to compare the mean accretion histories of optical and radio quasars. We find significant statistical evidence that radio quasars have a h
igher mean Eddington ratio Lambda at z > 2 with respect to optical quasars, while the situation is clearly reverse at z < 1. At z > 2 radio quasars happen to be less massive than optical quasars; however, as redshift decreases radio quasars appear in increasingly more massive BHs with respect to optical quasars. These two trends imply that radio sources are not a mere random subsample of optical quasars. No clear correlation between radio activity and BH mass and/or accretion rate is evident from our data, pointing to other BH properties, possibly the spin, as the driver of radio activity. We have checked that our main results do not depend on any evident bias. We perform detailed modelling of reasonable accretion histories for optical and radio quasars, finding that radio quasars grow by a factor of a few, at the most, since z ~ 4. The comparison between the predicted mass function of active radio quasars and the observed optical luminosity function of radio quasars, implies a significantly lower probability for lower mass BHs to be radio loud at all epochs, in agreement with what is observed in the local universe.
