A new paper says yes, and then maybe sort of:
By modeling the market for IPOs as a repeated game with imperfect monitoring, we establish that collusion among underwriters explains the concentration of spreads at 7%, along with other characters of the data on spreads. Furthermore, the structure of optimal spreads in the model explains the existence and quantitative characteristics of underpricing in the market for IPO shares. We estimate the model by deriving moment conditions from both underpricing and spreads. Our estimates indicate that IPOs destroy value on average over the sample period 1985-2007. This result, however, is driven primarily by the dot-com era. Excluding this period, IPOs appear to increase value.
That is from Lowery and Kang at UT Austin, a few years old but I had not seen it before, the link is here. For the pointer I thank Samir Varma. He also points me to this paper on why the reputations of individual star employees may be eclipsing the reputations of institutions at the level of the investment bank.