Are there scale economies in banking?

Mark Thoma concludes his summary on whether scale economies in banking justify the existence and benefit of large banks with the following:

I'd like to know the source of the scale economies. The paper linked above estimates returns to scale, but not their source. As noted in the introduction (and also noted by David):

Our results indicate that as recently as 2006 banks faced increasing returns to scale, suggesting that scale economies are a plausible (though not necessarily only) reason for the growth in bank size...

Without knowing the source of the changes in costs as banks increase in size, the (non-parametric) results -- results that differ from most previous work -- are hard to evaluate. I've been hoping for good estimates of the size and nature of the economies of scale for banks, but I'm not fully convinced by this evidence.

Mark is correct to try to figure out the specifics that are driving the returns in the study.  From my own experience - sales and trading in loans, bonds, and commodities - scale is found mostly through network effects.  The market players that have the most trading flow and transactional experience in a particular market have the accurate prices and up to date information.  This advantage leads to a higher return by being more likely to earn the bid / offer spread when making markets and being able to price services at a premium with clients wanting information access as well.  While this can be a stable and profitable advantage it was a small part of the profit in the banking areas I have worked in.

I would expect other economies of scale on the deposit taking side of the business as servicing a large number of depositors in a centralised fashion seems like a classic scale business.  The number of clients and transactions adds little marginal cost while the licensing, reputation, and infrastructure represent substantial fixed costs.  This advantage is hived off from the risk taking and higher margin bank  businesses with the internal treasury and risk department charging LIBOR plus a spread for use of funds.  These risky parts of the business still tend to grow fairly large despite the absence of scale economies.  I attribute this to the money politics of the industry as empires grow and shrink based on raw profit.  Growth is revenue rather than cost driven.   This fits well with the ease that hedge funds have found competing against banks in sales and trading functions though funds usually start with few people and no scale.

Another thought that I had from reading David's post, was that even if there are economies of scale that does not justify "too big to fail".  It shows that big can be a public benefit but if that is the case then a system needs to be created so that the naturally big entities created can be wound down effectively if necessary.  I believe this is now widely accepted following the difficult ad hoc responses that became necessary for troubled large institutions in 2008.

Particular banking businesses may need to be large to get the maximum public benefit but for the reasons above I doubt this applies to the financial industry as a whole.  Careful thought should be given so that if a safety net for large institution needs to be maintained it only applies to functions that can justify scale benefits.  If scale can be justified it still does not justify the lack of a mechanism to wind down large institutions when necessary.