C. Conselice: Galaxy Interactions and Mergers at High Redshifts
When do galaxies merge? The merger fraction does evolve lowly up to z=1.2 but at around 2-3, 50% of all high-mass systems are mergers. The small ones again have only slightly higer merging rate. So at z=1 most of the high-mass objects were in place.
Using the same methodology to find mergers on numerically simulated data (with C. Mihos), they derive absolute merger rates (per volume) and a sharp drop after z=1 is found for all masses. A typical massive elliptical galaxy (today) will have undergone 3-5 major mergers since z=3.
A significant fraction (maybe the majority) of SF at z<1 is produced by interactions and mergers. I think this is last point is still debated and there have been contradicting results, e.g. showing that interactions dot not really increase SFR as much as one would think.