Comparison between our inferred f_M − M_* (top) and M_vir − M_* (bottom) relations and an idealised toy model of mass assembly. The model considers a population of galaxies (blue circles) lying in the z = 0 relations (blue curves), which are then traced back to z = 0.9 (squares), assuming theoretical stellar and halo mass growths. Galaxy downsizing and halo growth histories yield z = 0.9 relations (squares) with slopes different from those observed at z = 0 (solid pink curves). The only way to reconcile the stellar mass growth of galaxies with our inferred relations at z = 0.9 is if low-mass haloes grew more than high-mass haloes since z = 0.9, contrary to CDM expectations. This suggests that the galaxies populating our z = 0.9 relations do not follow this toy model and are unlikely to be the progenitors of the z = 0 population.