It is widely understood that galaxies use, throughout the Hubble time, only a small fraction of the baryons associated to their dark matter halos to form stars. Such low baryon-to-stars conversion efficiencies are expected in galaxy formation scenarios where stellar & AGN feedback play a key role in regulating star formation in galaxies, respectively at the low- and high-mass end.

In this talk I will show how we can constrain this scenario using galaxy dynamics. Both robust determinations of disc dynamical scaling relations (e.g. Tully-Fisher, mass-size) and accurate measurements of dark matter halo masses from HI rotation curves of spirals and from the kinematics of globular clusters around ellipticals, provide compelling evidence that the population of massive spirals has systematically larger baryon-to-stars conversion efficiencies than ellipticals. In fact, we see that the baryon-to-stars conversion efficiency monotonically increases with mass for late-type galaxies, while it shows a clear turn over at about L* only for early-type galaxies. Thus, while massive early types are compatible with standard stellar-to-halo mass relations based on abundance matching, massive late types are systematically discrepant from it.

I will discuss the possible repercussions that these results have, highlighting in particular what they imply in terms of AGN feedback and merging in galaxies of different types. Finally I will show that current state-of-the-art cosmological hydrodynamical simulations (EAGLE, TNG) still struggle to reproduce what we observe for the most massive discs.