Why Banks are “Special”: The Short Story

No, not that kind of “special.” Though it sure is tempting…

Paul KrugmanScott Sumner (seemingly unlikely bedfellows, but…), and most other mainstream economists want to argue that banks are not special — that there’s no reason for economists to understand and analyze their operations in detail, or incorporate those understandings in their (mental and formal) economic models.

Their essential position is that borrowers’ and lenders’ incentives and reaction functions are symmetrical, interacting opposites. (This is the very essence of the Krugman/Eggertsson 2010 patient/impatient-agents paper.)


1. Banks are transparent intermediaries between real-sector lenders (“savers”) and borrowers.


2. The machinery of market portfolio allocation (i.e. as described by Tobin) allows us to model the economy as if #1 were true. cf. Krugman/Eggertsson.

So, in Steve Randy Waldman’s words, “Everything that matters is captured by the portfolio preference of nonbanks.” We can ignore the banks.

Steve, Cullen Roche, Scott Fullwiler, and many others suggest otherwise (and at length) — that banks are special and that it is necessary to model that specialness.

I’d like to explain, in brief, why they are special:

Banks aren’t optimizing intertemporal consumption preferences. Unlike real-sector actors, when banks lend they’re not “saving.” The incentives and reaction functions of the economy’s dominant lenders are completely orthogonal to those of the economy’s borrowers.

Now you could argue (back to #2, above) that the inexorable forces of the financial markets force banks (at least in aggregate) to act as if they were optimizing intertemporal consumption preferences (for their borrowers? their shareholders? their lending officers? their directors? their creditors? their depositors?). But I think many will agree that that argument is quite a stretch.

If economics, as I have suggested, is ultimately the study of human (individual and group) reaction functions, it seems irresponsible and misguided to ignore the incentives and reaction functions of some of the most powerful and influential actors in the economy, simply brushing them under the rug.

Mainstream theory tells us we don’t need to understand how banks work. To understand why that theory is wrong, you need understand how banks work.

Cross-posted at Angry Bear.






21 responses to “Why Banks are “Special”: The Short Story”

  1. […] Steve Roth — Why Banks are “Special” […]

  2. Peter N Avatar
    Peter N

    One problem is that the bank modeled to demonstrate banks can be ignored bears very little resemblance to Morgan Stanley, Goldman Sachs or Bank of America. If you actually look at their 10-Ks you can easily see this. No bank of Krugmanland could accumulate $3 billion in trading losses in a few weeks through “intermediation”.

  3. Nick Edmonds Avatar

    In my view, the essence of why banks might be said to be special is captured in your recent post on not spending is not investment. The decision by non-banks to hold more deposits is a saving decision. The decision by banks to lend can be seen as being an investment decision (because in most cases whether the borrower invests or not depends critically on whether the bank will lend). The latter does not depend on the former. Having banks in the middle between lenders and borrowers is a clear case where not spending is not the same as investment.

  4. errorr Avatar

    @Peter N

    I think the reply would be that those banks make their profit from miscellaneous added services that have no relative importance to how they act as a true “bank”. I’m not even sure that is wrong even though I’m with SRW and others believing that bank functions are important. The money comes from either services or their own bets that are hardly related to traditional banking. Although you could argue that the shenanigans of the last decade were profit off the extraneous activities that Krugman ignores as the bank try and meet the portfolio preferences of the cash-rich by massive expansion of securitization as things that people thought of as being like treasuries (mbs, etc…) were in fact realized to be not money like. This precipitated the liquidity crunch in 2007 that killed Lehman when the FED didn’t allow money to exogenously expand fast enough by buying treasuries. There still may have been a need for more intervention to make the MBS market more money like but it was already too late and the portfolio preference of the private sector expanded too quickly.

  5. Ramanan Avatar

    Good points Steve.

    Even Wynne Godley said something similar:


    It is a matter of ascertainable fact that the real world is
    characterised by a huge and complex structure of interdependent
    institutions such as governments, firms, banks and households. I
    do not accept that these institutions are “veils” with nothing
    more to do than passively sponsor or facilitate the optimising
    aspirations of individual agents; and wish, rather, to start from
    a conceptual framework which takes cognisance of (something
    remotely approaching) the real world as we know it.

    I really liked your line:

    “The incentives and reaction functions of the economy’s dominant lenders are completely orthogonal to those of the economy’s borrowers”


  6. Ramanan Avatar

    I think you may have to modify the quote I quoted from your post a little to say what you intended to say. (?)

  7. Fed Up Avatar

    Banks are intermediaries.

    When I see this I think, a bank matches a saver with a borrower. That is what I picked up from Rowe, Krugman, and others. It seems to me a modified version of my example above shows that.

    Let’s say I save $100,000 in demand deposits. A new bank wants to add to its existing capital. They sell me $100,000 in bank bonds (bank capital) and then buy treasuries. The reserve requirement is 0%, and the total capital requirement is 10%. I believe that means the capital requirement is 5% for mortgages and is 10% for ordinary loans. This example will be all mortgages and focusing only on the new capital.

    Assets new bank = $100,000 in treasuries
    Liabilities new bank = $0
    Equity new bank = $100,000 of bank bonds

    The bank now makes 20 mortgages for $100,000 each. The 20 people use the demand deposits to buy 20 homes for $100,000 each from 1 home builder.

    Assets new bank = $100,000 in treasuries plus $2,000,000 in loans
    Liabilities new bank = $2,000,000 in demand deposits (DD)
    Equity new bank = $100,000 of bank bonds

    $100,000 / ($2,000,000 * .5) = .10

    The home builder sets up a checking account at the new bank. So 20 checking accounts at the new bank were marked up and then marked down by $100,000 each, while the home builder’s checking account was marked up by the $2,000,000. Now the DD’s can be spent in the real economy.

    Overall, I saved $100,000 in MOA/MOE, and the borrowers “dissaved” $2,000,000 in MOA/MOE for a difference of $1,900,000 in MOA/MOE. That difference is why banks and bank-like entities matter.

    I saved $100,000 in DD’s (or could be currency). The new bank borrowed $100,000 (bank capital), and I lent $100,000. The 20 people borrowed $2,000,000 in DD’s, and the new bank lent $2,000,000. Can the people who receive the $2,000,000 in DD’s be considered to have lent to the new bank and the new bank considered to have borrowed from the $2,000,000? If so and however, the people who have the $2,000,000 can stop saving at any time?

  8. JKH Avatar

    “The incentives and reaction functions of the economy’s dominant lenders are completely orthogonal to those of the economy’s borrowers.”

    I don’t know what that means?

    Still some confusion in the general debate, IMO.

    Krugman argued that banks are not special in the same way Tobin did.

    Neither Krugman nor Tobin argue that financial intermediaries are not special relative to a counterfactual economy. Quite the contrary.

    Godley uses Tobin to build the case for how banks work.

    I don’t think that conflicts with what Tobin says about banks not being special.

    A point of logic:

    X = A + B
    Y = B + C

    X is special relative to Y in respect of A (e.g. banks creating money from lending)
    Y is special relative to X in respect of C (e.g. issuing insurance policies)
    But neither X nor Y is special relative to B (active asset liability management and pricing)

    It is quite possible for the same entity to be both special and not special

    It’s a matter of the classification paradigm

  9. Peter N Avatar
    Peter N


    The connections between old banking activities and other diversified financial companies (calling them banks is more or less misleading, depending on the bank in question) run through repos, securitization and an unlimited ability to restructure (but not, of course eliminate) risk. The 1960s style banks modeled by Tobin are mostly gone, and those left are small and irrelevant . The twenty largest banks in the world have over $44 trillion of assets.

    I suppose risk restructuring and securitization run wild exist to satisfy customer demand (in the way most profitable to the bank). But in all fields there are consumer preferences that are imprudent, unlawful or systemically dangerous to meet, so this is may be a partial explanation. This doesn’t mean you can just invoke the first welfare theorem as proof of the impossibility of any problems. Its preconditions aren’t met and its results insufficient.

  10. Peter N Avatar
    Peter N


    “It is quite possible for the same entity to be both special and not special

    It’s a matter of the classification paradigm”

    Absolutely correct, but you would expect inter-paradigm leakage and the use of activities in one paradigm to bypass safeguards and regulations in another. There are amazingly clever experts on this sort of thing on Wall street, since it is very profitable, and the risk can be transferred to some convenient Muppets.

  11. Ramanan Avatar

    I think if Asymptosis’ point “2” means banks are not veils then it is okay now I wonder what he means.

    If he means banks lending decisions are not dependent on what depositors say then it is okay.

    Moore’s book (calling banks intermediaries) makes the simplified distinction between “price setters and quantity takers” versus “price takers and quantity setters”. Asset allocator such as an equity mutual fund is in the latter category.

    In this classification some non-bank financials are the former.

  12. Asymptosis Avatar


    Very well explained.

    Can your explain your understanding of where Krugman does go wrong, in a sentence or two?

  13. Asymptosis Avatar


    Orthogonal reaction functions:

    If you repay a loan to a real-sector lender (say…me), that money remains in the hands of of a real-sector actor. Only the IOU disappears. How do the lender and borrower react? I probably spend some more on real newly produced goods and services, and you spend less. Let’s call it a wash.

    If you repay your credit-card debt to the financial sector, how do the lender and borrower react? You spend somewhat less on real-sector goods and services, and the credit-card company…doesn’t spend more. Both the IOU and the money disappears.

    Banks have furnaces right next to their printing presses. Households don’t.

    I should have said that the reaction functions of real-sector lenders and financial-sector lenders are orthogonal.

  14. JKH Avatar


    I’m going to have to go back to check all that out, since the blogosphere commentary has created a huge cumulonimbus cloud obscuring Krugman’s lighthouse right now.

    But a couple of things off the top. I think “patient zero” for this sort of thing was Krugman’s interaction with Keen 18 months ago. And I think then then and now PK has committed some pretty basic errors.

    The first is that he definitely maintained back then that “loans create deposits” out of “thin air” was a ridiculous idea, whereas Tobin is clear about this right away. (Interestingly, Tobin makes this clear as the first thing in the essay that Krugman says he had “forgotten about”.) And PK seems to be avoiding that still. I think he’s actually getting trapped in some kind of fallacy of composition of his own making – he’s identifying the risk that a single bank faces in losing a deposit as something that disproves “loans create deposits”. It becomes interesting that Tobin’s argument against the widows cruse starts with that point, which he then expands to the case where the banking system in total also loses the deposit. So Krugman seems to be conflating all of that in his own rejection of “loans create deposits”, even though Tobin understands how “loans create deposits” exists while the widow’s cruse fails.

    The second is that he still seems to be hanging onto the money multiplier idea. I just don’t think he’s in synch at all with the idea that the central bank supplies required reserves after the fact.

    The third is connected somehow to the fact that Tobin-Brainard assumes a sort of Warren Mosler zero interest rate environment as a simplification. It assumes further that governments essentially deficit spend zero interest currency into the system. And it assumes that is the primary instrument of “monetary control”. In other words, it completely bypasses dealing with the issue of a central bank that sets an exogenous positive interest rate on government liabilities. It is a pure monetarist central bank, adjusting the quantity of currency in setting monetary policy. In fact this may be OK for analyzing the issues of monetary control in the form of reserves and commercial bank interest ceilings that Tobin sought to demonstrate. But I think Krugman is drawing all sorts of strange conclusions about being indifferent to whether the public holds currency or the central bank holds it (as reserves) from this very specific modeling case of a zero interest rate, and that is cluttering his general argument.

    Finally, I think his central bank accounting is not very good. This became clear in his debate with Steve Waldman earlier this year. That debate had to do with understanding the relationship between excess reserves and related central bank scenarios for lifting interest rates off the zero bound.

  15. JKH Avatar


    As a weird personal corollary, I think Tobin might be in synch with the following, but for some reason Krugman would only go for the “in context” part:


  16. Nick Edmonds Avatar

    This is a source of great confusion.

    This idea leads people to conclude that bank lending is expansionary (supposedly because it increases overall purchasing power), but non-bank lending is not (supposedly, because it merely transfers existing purchasing power around).

    Suppose I want to buy a car but I have no money. I get a loan from the bank and buy the car. Money up, spending up. Easy.

    Now suppose the bank won’t lend to me. I go to a rich uncle who has loads of money at the bank. Because he likes me, he makes me a loan (transfers to me some of his bank money) and I buy the car. Spending up, but money unchanged.

    The difference between this example and yours has nothing to do with money. It is everything to do with what assumptions we make about how people’s spending respond to the availability of credit.

  17. JKH Avatar


    Re orthogonality:

    If you repay a loan to a non-bank, the money supply remains the same.

    If you repay a loan to a bank, the money supply shrinks.

    You don’t have to be a monetarist to acknowledge that a shrinking money supply is contractionary, OTHER THINGS EQUAL (particularly velocity). I think that loan repayment would have to be matched by a new loan elsewhere for conditions not to have changed on that basis.

    I think that’s consistent with what you’re saying.

    (Monetarists seem to develop arguments that seem to imply that other things are always equal, or almost equal. Just IMO; they would object of course.)

  18. Nick Edmonds Avatar


    One way of looking at this, dare I say it, is in IS/LM terms. Non-bank lending and bank lending can both have an expansionary effect through a change in the IS curve (because they impact on the net savings ratio). Bank lending also has an expansionary effect through a change in the LM curve.

  19. Ramanan Avatar

    @Nick Edmonds

    “This idea leads people to conclude that bank lending is expansionary (supposedly because it increases overall purchasing power), but non-bank lending is not (supposedly, because it merely transfers existing purchasing power around).”

    Good point. And this “loans create deposits” – although not the intention of people using it but nonetheless not clear who believes this notion or not – reinforces this incorrect notion.

  20. Peter N Avatar
    Peter N

    “you repay a loan to a bank, the money supply shrinks.”

    or perhaps “If you repay a loan and, the money supply shrinks, then you’ve repaid it to a bank.”

    But this requires an unambiguous universally agreed definition/classification of money, so maybe correct is –

    “If you take a loan from (repay a loan to) an entity and, the money supply expands (shrinks) according to some definition of money, then you’ve borrowed (repaid) it to a bank-of-creation for money meeting that definition.”

    This allows for other forms of banks and probably can’t be easily falsified through ramifications of a bank’s other businesses. It also allows for different forms of money and money in other currencies.

    Unfortunately, the converse doesn’t appear to be true. You could also be dealing in loanable funds.

  21. ATR Avatar

    Thought I’d throw my two cents on this issue by starting a blog. Curious on your thoughts…