Left: the relation between the stellar mass M_⋆ and the halo mass M_h in VUDS for different M_⋆ and redshifts (red symbols). M_⋆ is derived from SED fitting of the multi-wavelength photometric data using known spectroscopic redshifts; error bars in M_⋆ indicate expected uncertainties of the SED-fitting method. M_h is obtained from HOD modelling of the two-point correlation function in different redshift and mass ranges. The VUDS data are compared to low- and intermediate-redshift measurements from satellite kinematics (Conroy et al. 2007; More et al. 2011), weak lensing (Mandelbaum et al. 2006), and galaxy clustering (Foucaud et al. 2010). The lines represents model predictions derived from abundance matching at various redshift (Moster et al. 2013). Centre: the ratio of stellar mass M_⋆ over halo mass M_h vs. halo mass at z = 3 in the VUDS survey. The colour scheme is the same as for the left panel. Right: evolution of the M_⋆/M_h ratio with time predicted from stellar and halo mass accretion histories for the most massive galaxy population observed at z ~ 3, using the toy model described in the text.