Quantifying how the baryonic matter traces the underlying dark matter distribution is key to both understanding galaxy formation and our ability to constrain the cosmological model .
Using the cross-correlation function of radio and near-infrared galaxies , we present a large-scale clustering analysis of radio galaxies to z \sim 2.2 .
We measure the angular auto-correlation function of K _ { \textrm { s } } < 23.5 galaxies in the VIDEO-XMM3 field with photometric redshifts out to z = 4 using VIDEO and CFHTLS photometry in the near-infrared and optical .
We then use the cross-correlation function of these sources with 766 radio sources at S _ { 1.4 } > 90 \mu Jy to infer linear bias of radio galaxies in four redshift bins .
We find that the bias evolves from b = 0.57 \pm 0.06 at z \sim 0.3 to 8.55 \pm 3.11 at z \sim 2.2 .
Furthermore , we separate the radio sources into subsamples to determine how the bias is dependent on the radio luminosity , and find a bias which is significantly higher than predicted by the simulations of Wilman et al. , and consistent with the lower luminosity but more abundant FR- I population having a similar bias to the highly luminous but rare FR- II s. Our results are suggestive of a higher mass , particularly for FR- I sources than assumed in simulations , especially towards higher redshift .