That's the conclusion of a new St. Louis Fed study by David Wheelock and Paul Wilson. In the two decades between the mid-80's and 00's, the number of commercial banks fell by 50% while the average size per institution surged by an inflation-adjusted 500%. The problem, systematically speaking, is that banks receive increasing returns for getting bigger.

The story behind the rise of big banks seems to be largely driven by technology:

Moreover, reductions in the cost of acquiring quantifiable information about potential borrowers have eroded some of the benefits of small scale and close proximity to borrowers that enabled small banks traditionally to out-compete larger banks for some customers, such as small businesses

But while much of the talk around the Too Big To Fail problem has been about the potential harm inflicted by large institutions on the rest of us, Wheelock and Wilson point out that there will be costs associated with not allowing banks to choose what size is right for their operations:

The fact that most banks faced increasing returns as recently as 2006 suggests that the U.S. banking industry will continue to consolidate and the average size of U.S. banks is likely to continue to grow unless impeded by regulatory intervention. Our results thus indicate that while regulatory limits on the size of banks may be justified to ensure competitive markets or to limit the number of institutions deemed too-big-to-fail, preventing banks from attaining economies of scale is a potential cost of such intervention.

Unfortunately they provide no details on what these costs actually are. But this doesn't even strike me as a big concern since the administration's TBFT proposals don't rely on size caps and are more related to bank risk-taking.

Addendum: Noam sent his (good) thoughts on the above which are worth reproducing:

Interesting item, but that paper (which I admittedly haven’t read) seems to miss the point. We don’t really care about median bank size – or we do, but that’s not what the TBTF debate is about. It’s about a handful of enormous outliers. You could argue that limiting the size of the median bank would have costs in terms of preventing it from reaching its efficient scale. Seems hard to argue that that’s true of the megabanks – that somehow you’re much more efficient with a $1 trillion balance sheet than with a $200 billion balance sheet. (Setting aside the implicit taxpayer guarantee, of course.)