IOR caps: a new instrument of monetary policy?

As the MMT-ers emphasize, in aggregate bank lending is almost never reserve-constrained. Unless a central bank is willing to tolerate arbitrarily high interest rates, it must be willing supply reserves in response to increasing demand.

However, to point out that the banking system is not reserve constrained does not imply that individual banks are not reserve constrained. Macro types tend to assume there is an interest rate and an interbank market to which any bank can turn for reserves at the policy interest rate. But that’s simply not true, at least in the United States. Most banks do not regularly borrow at all on interbank markets, or at the central bank discount window. [1] The largest banks have ready access to interbank loans and use them to finance substantial portions of their balance sheets. But small banks do not. This is understandable. Borrowing in the “Federal Funds market” occurs via bilateral unsecured loan contracts. Small banks are perfectly comfortable lending to large, implicitly backstopped banks. But banks large and small are reluctant to make cheap unsecured loans to the Bank of Palookaville, whose exposure to the local Cadillac dealer is hard to evaluate from a distance. So small banks tend to amass precautionary stashes of reserves and lend them overnight to big banks, who may borrow at will. Prior to the financial crisis, the net reserve position of the largest 20% of US banks was negative. All of these banks’ reserves and then some were borrowed on the interbank market. That has changed since the Federal Reserve has flooded the banking system with cash, but small banks continue to devote a larger fraction of their balance sheets than large banks to reserves and near-reserves. While large banks happily rely upon central-bank guaranteed liquidity, small banks maintain buffers in case that abundance fails to trickle down. Aggregate lending cannot be reserve constrained, but lending by small banks may well be.

If small bank lending is reserve constrained, then policies that redirect flows of reserves from larger banks to smaller banks might be expansionary. At the margin, Bank of Palookaville is more likely to fund a loan if its reserve stockpile is well above what it requires for self-insurance. As the stockpile dwindles, a small bank’s cost of lending must include an increasing charge for the bank’s liquidity risk. There is evidence for this. In an influential paper, Kashyap and Stein famously documented a “bank lending channel” of monetary policy that operates predominantly through smaller banks.

But how could a central bank oppose the general tendency of reserves to flow from smaller to larger banks? Once the Fed buys an asset or makes a loan, the reserves are “out there”, no longer under its control. In the past, affecting patterns of reserve flow might have been difficult. But now the Fed pays interest on reserves, on terms that are entirely at its discretion. Suppose that the Fed were to cap, in absolute dollars, the quantity of reserves on which it is willing to pay interest to any single bank. Then banks with “too many” reserves would look to shed the excess, an unremunerated asset they need to finance. They might reduce the interest they pay (or raise the fees they charge) to depositors, which, at the margin, would cause funds to flow to banks whose reserve level is below the cap. Alternatively, reserve-heavy banks might lend their excess directly to smaller banks. This would allow the two banks would split the interest payment that would otherwise have been foregone, while reversing the usual direction of reserve flow in the interbank market. In either case, smaller banks might find any liquidity constraints they face substantially reduced.

The level at which reserve remuneration is capped would become a new instrument of monetary policy. The most contractionary setting would be where we are now, with no cap at all and the vast majority of excess reserves held by the largest banks. A hyperexpansionary policy would determine the cap by dividing the total quantity of excess reserves by the nearly 7000 US banks, unleashing forces of arbitrage that would inflate the balance sheet of every Podunk bank. Between these two poles there are infinite gradations, a tunable instrument of monetary policy.

Beyond the important but sterile dimension of “contraction” vs. “expansion”, there are other reasons to like this idea. Banking activity in the United States is quantitatively dominated by a small number of very large banks for whom the absence of reserve constraint is a competitive advantage. Reducing or eliminating the reserve constraint faced by small banks would even up the playing field. From a Hayekian perspective, status quo banking has devolved from an enterprise in which dispersed decisionmakers compete for advantage based on context-specific information towards a notionally private form of central planning, a Soviet backed by a few giant data-crunching hierarchies. If your inner Hayek is strong, you should applaud increasing the ability of small bankers to lend to people they know and projects they understand, while bearing much more of the risk than employees of large banks would. Smaller banks lend disproportionately to small and medium-sized enterprises, so if you think small entrepreneurs are at the heart of the economy, shifting activity towards smaller banks should help. Finally, small banks are not too big to fail. If we can’t break up the TBTF banks (I still hope we can), perhaps we can slowly deflate them by creating incentives for activity to migrate elsewhere.

Capping interest on reserves might or might not prove helpful. Quantitatively, I don’t think we have a sense of how much more lending would occur if small banks, like large banks, faced no liquidity constraint. On one hand, “the 99%” of banks currently account for only 22% of activity (measured by balance sheet size) and the smallest 90% of banks account for less than 8%. You can argue that with numbers like these, smaller banks just can’t make a difference. On the other hand, finance flows fast and the hyperconcentration of banking may prove reversible. An IOR cap has no financial cost, can be implemented at will, and would be expansionary at the margin however large or small the effect. Perhaps it is worth a try.

[1] According to bank call reports (June, 2011), only about 40% of US banks borrow funds from the interbank or repo markets. Both the fraction of banks that use interbank finance at all and the degree to which banks finance their assets in the interbank market strongly correlates with size. As of the most recent reports, 71% of banks in the largest size decile reported using some interbank or repo finance, which financed more than 4% of these banks’ balance sheets. Only 6% of banks in the smallest size decile reported any interbank or repo funding at all. Of these few small banks that do borrow reserves, interbank and repo finance accounts for only 2% of their funding. Prior to the financial crisis (September 2006), banks in the top decile relied on interbank and repo markets to fund roughly 9% of their (gigantic) balance sheets, while the smallest decile (all banks) received less than 1% of its funding from these markets.

Call reports are end-of-quarter snapshots, and so will undercount banks that transiently tap the interbank market to cover sporadic reserve shortfalls. But that would be true across deciles, and wouldn’t alter the lower frequency structure of interbank funding.


42 Responses to “IOR caps: a new instrument of monetary policy?”

  1. John Thomas writes:

    Would the big banks and the small banks create a secondary market so that they could get around the cap?

  2. Phil Koop writes:

    Bravo! This is a beautiful post.

  3. Scott Fullwiler writes:


    Two things . .

    1. You are not arguing that small banks are reserve constrained. You are arguing that small banks pay a higher interest rate to borrow reserves in interbank markets, and therefore this higher interest rate leads them to be more cautious expanding their balance sheets–though as general a rule small banks don’t try to grow as fast as large banks anyway. If a small bank decides to grow its balance sheet beyond what it knows it can acquire in deposits, it will estimate the cost of acquiring additional liabilities in interbank markets, and this will affect its net interest margin on such expansion and thus its decision on whether to expand. So, what you’re then saying is under your proposal small banks would have access to interbank markets at a lower rate than otherwise, not that they won’t be reserve constrained anymore.

    2. I don’t have time this morning to think too carefully about your proposal, but it seems to me that setting the fixed qty of absolute dollars paid on IOR would simply return us to a situation in which the Fed had to drain reserves to hit its target rate, and the qty it had to drain would depend upon where it had set this new instrument.

  4. KidDynamite writes:

    admit it – you are a paid shill who is short the Bank of Palookaville!

    interesting post, Steve

  5. […] interfluidity » IOR caps: a new instrument of monetary policy? […]

  6. JKH writes:


    I must say I’m finding it difficult to parse all of this, due to a somewhat vague impression of some conflicting statements about directionality/distribution between large and small in different parts of the text. That’s just my preliminary impression, but I think it would be an assist if there were a high level summary for fed funds bought/sold distribution, excess reserve distribution, for small versus large banks, and for pre crisis versus post crisis (the pre crisis excess reserve distribution would be moot, but for completeness). Such a summary would not be for data detail, so much as for logical argument’s sake as small relates to large. Perhaps added as a comment?

    That said, a few points:

    A given asset position in the form of either fed funds sold or excess reserves held at the Fed serves essentially the same liquidity management function – it is a call on settlement balances. Excess reserves per se are obviously the most direct form of call, but not a whole lot different than ON fed funds sold.

    Outsized excess reserves in aggregate are of course a feature of the financial crisis.

    Payment of interest on outsized reserves is an issue for Fed Reserve interest rate control. That’s not to say whether the current rate next to the zero bound should be 25 basis points or zero. That particular question has received outsized attention IMO. There is simply not enough at stake in terms of banking system aggregate interest margin – say around $ 4 billion pre tax today – to make it a huge issue. That opportunity cost is pittance in terms of driving system wide pricing behaviour. Arguably, the zero bound reserve compensation question matters the least in the range of all such questions in all possible interest rate environments. The question becomes far more critical at higher levels of targeted Fed funds rates. MMT is well equipped to explain why that is the case.

    There is already in the system a variation on your capping proposal. GSE’s hold balances at the Fed on which they receive no compensation – their cap is zero. That has distorted fed funds pricing. The reason ostensibly is that, notwithstanding arbitrage opportunities, the banks that do receive compensation on their excess balances are not willing to fully engage their balance sheets in such arbitrage, given optical constraints on non-risk weighted leverage. So fed funds has declined below the rate paid on reserves.

    Your proposal would have a similar and perhaps additional marginal effect on fed funds pricing in today’s environment. But like the situation with GSE’s, it would make very little difference to risk pricing.

    And I’m with Scott Fullwiler on his instinct here – the idea would seem to be unworkable at higher levels for the target fed funds rate – simply because the economic importance of ensuring the integrity of the target rate then becomes more material. Slippage of the target rate down to zero would be fast and furious.

  7. JKH writes:

    “While large banks happily rely upon central-bank guaranteed liquidity, small banks maintain buffers in case that abundance fails to trickle down.”

    I think this is wrong.

    Pre-crisis, small banks held liquidity in the form of fed funds sold, among other types of liquid assets. Given the normal operation of interest rate markets and the very tight marginal setting for aggregate excess reserves in the banking system, there would simply be no reason for small banks to hold outsized excess reserve balances pre-crisis. Besides, the evidence would have pointed to systematic window borrowing by large banks pre-crisis, given the low level of system excess reserves. There is no evidence that this happened.

    Post crisis, under your explanation/data, small banks presumably would have expanded their most liquid assets to a combination of fed funds sold and excess reserve balances at the Fed.

    Looking at all banks, reserve balance management and fed funds bought/sold management are operational pieces within a much larger scope of the function of bank liquidity management. The fact that large banks are net takers of fed funds from smaller banks suggests nothing about their gross or net liquidity position, let alone their “reliance” on “central bank guaranteed liquidity”. The fact is that large banks have hundreds of billions of liquid assets on their balance sheets, quite apart from reserve balances and gross fed funds sold. Liquidity management, like capital management, is an operationally complex and widely scoped function across differentiated assets. It cannot be reduced entirely to a sole focus on excess reserve balances and fed funds bought and sold.

    Small bank and/or large bank lending is not necessarily reserve constrained, micro or macro. If anything, it is more accurate to suggest lending is liquidity constrained (but in a very qualified way as below), where liquidity has the much wider scope than just fed funds sold or excess reserve balances. That is to say that, at a purely technical level, maintaining either fed funds sold or excess reserve balance positions is a sufficient but not absolutely necessary requirement for maintaining access to settlement balances. But even that broader interpretation of liquidity and its effect on lending must be highly qualified, according to conditions in the prevailing environment both outside and inside the banking organization. By definition, a bank would only be liquidity constrained at an operational level, where it chose to withdraw lending capacity from the market, if it were running into considerable difficulties maintaining a self-imposed or regulatory-imposed liquidity policy, which in turn would probably reflect considerable difficulty in maintaining a self-imposed or regulatory-imposed capital policy or credit risk policy.

    With respect to your cap idea, banks with over-cap reserve balances would first seek alternative liquid assets such as fed funds sold, treasury bills, etc., driving those rates down in the process. The cost of a bank’s liquidity policy would be higher to the degree that rates earned on liquid assets would be lower. Liquid asset portfolios would be adjusted by mix in order to equilibrate the cost that originated with a zero rate on over-cap reserve balances. But that would not be enough to unhinge what is a bank’s normal attempt to exercise risk management and capital management discipline in allocating capital to risk assets. The argument comes back to the same general one about the difference between reserve balance management and capital management – the argument that goes all the way back to the falseness of the reserve multiplier concept. With respect to the distribution of these aspects between large and small banks, I don’t see how that is a particular issue with respect to the cap proposal. Small banks should not be more susceptible to breaking tradition in looking at capital ratios before implementing newly aggressive risk lending programs, merely because they have reason to adjust their liquid asset portfolio mix while accounting for a marginally higher cost of liquidity.

  8. Indy writes:

    I can’t help thinking that if this rule went into effect, a large bank would try to “spin off” enough of its own “subsidiary” small banks to hold the excess reserves precisely at interest-bearing quantities. At the absurd extreme the new small bank’s only real customer is the big bank, where the big bank deposits some of its non-interest bearing reserves, and the small bank just loans it all back with interest it gets from the Fed’s IOR program. In the alternative, the big bank essentially corrupts and captures preexisting small banks to play this single role for it.

  9. JKH writes:

    Here are some rough 2010 banking stats I dug up on Montana, assuming that particular state isn’t exactly littered with soaring bank towers, and may be representative of small bank balance sheet origination. The numbers are banking system totals to the end of 2010, rounded for simplification, in $ billions:

    Cash, etc. 2
    Securities 6
    Loans, etc. 12
    Total Assets 20

    Borrowed 2
    Deposits 16
    Equity 2
    Total 20

    These are 2010 numbers, but the interesting thing is that both the cash and the securities lines roughly doubled over the period of the financial crisis to that point, with the rest of the asset side not doing that much. As I said, liquidity management goes well beyond fed funds sold and reserve balances, even for small banks.

    And it appears these banks participated in QE1 and that agency holdings have increased materially as part of the securities portfolio.

    Interestingly, Fed funds purchased are 1.4 billion, up only marginally over the crisis period, and fed funds sold are only .25 billion, roughly flat over the crisis.

    So Montana has been a net BUYER of fed funds.

    Interestingly also, only if you go back to the period 1986 – 1992, is Montana a substantial and persistent net SELLER of fed funds.

    From the data, it also appears that reserve balances (not broken out) are up, but not nearly as much as securities.

    On the liability side, deposits are up substantially, with borrowed funds and equity up as well.

  10. beowulf writes:

    Don’t Forget the FDIC I:
    “The new FDIC assessment regime does not raise much more money than the old rule, but the burden is now carried more proportionately by the big banks. This pound of flesh was extracted from Congress by the community bankers to win approval of Dodd-Frank…
    Low risk core domestic deposits of all banks, large and small, are taxed in single digits and not more than before Dodd-Frank. But the large banks now must also pay insurance premiums on debt liabilities. Debt, repo assets and foreign deposits are all now part of the FDIC assessment base, so the broad BKH observation that FDIC is a tax on repo transactions is correct. Banks can reduce the assessment by up to 5bp based on the amount of non-deposit funding, but the assessment is a tax on all liabilities less capital…”

    Don’t Forget the FDIC II:
    “Because the majority rules that the FDIC is not a NAFI, the United States is now directly liable for the FDIC’s contractual commitments. Mr. Slattery and future plaintiffs like him can now sue the United States in the Court of Federal Claims and will receive money damages directly from the Judgment Fund… Under the majority’s holding, the FDIC need not exhaust any of its assets because the government is directly liable.”

  11. John — There already exist secondary markets for reserves: that’s what the Federal Funds market is. But the Fed pays interest on reserves to the party that holds the reserves. There only way around that is for a big bank to let a small bank hold the reserves but promise to remit back the interest. In other words, to deposit the reserves with the small bank.

    I may be missing what you are suggesting. I can think of alternative ways to simulate a deposit relationship, i.e. rather than an ordinary deposit, big bank and small bank could temporarily exchange some asset for reserves and settle cash flows like a total return swap. But the economic effect is the same: the liquid assets must be in the custody of the smaller bank, and the interest on reserves gets split in some negotiated fashion between the large and the small bank.

  12. Phil — Thanks!

  13. Scott — I think that smaller banks are reserved constrained more strongly than you suggest. It is certainly true that small bank interbank borrowing is more costly. There is a negotiated credit spread, although empirically it is not very large. (See Furfine, Banks as Monitors Of Other Banks.)

    But credit markets in general, both as in theory and in practice, don’t always clear by price. Because information asymmetries are difficult to price, and because the act of trying to compensate for poor credit quality with a spread itself reduces credit quality, credit is a service that is often simply rationed. There are households to whom no bank will offer an unsecured loan, regardless of price, because charging a 100% credit spread to the population 50% of whom would fail to repay a dime just ensures that nearly off of the population fails to pay. Sufficiently uncreditworthy borrowers must restrict themselves to collateralized borrowing (payday loans, etc), or black-market borrowing with extralegal means of enforcing debt contracts (loan sharks). However usurious the rate, they’ll not find a credit card.

    Going-concern banks are not like households with poor credit ratings. They don’t have histories of failing to honor debt. But they are highly leveraged black boxes, almost entirely opaque to their counterparties. Under the status quo equilibrium, my sense of the market is that smaller banks can borrow in the interbank market occasionally and for short periods to cover reserve shortfalls, but they cannot use the interbank market as a frequent or permanent source of funds. (My sense of the market may be wrong! I don’t know of any studies addressing this question head-on. It’d be a hard kind of study to do.) Large banks, on the other hand, do use the interbank markets as a source of ongoing finance.

    An intended effect of this proposal would be to change the Nash equilibrium small banks face in the Federal Funds market. Since the general practice is that small banks see to their liquidity via precautionary buffers, frequent or longer-term borrowings of Federal Funds are stigmatized, and are taken to be informative of stress. Since frequent or longer-term borrowing are taken to signal stress, cheap interbank finance is not a reliable funding source for smaller banks.

    If, by Fed nudge, longer-term interbank borrowing by small banks became a common practice, then it would no longer be colorably informative of distress and it would no longer be stigmatized. Therefore, smaller banks could rely on Federal Funds as a source of on-demand finance, just as large banks do.

  14. Scott & JKH — You both make the same point, and it is a good one.

    Under my proposed caps, large banks will not suspend all risk-management and deposit arbitrary sums to any Podunk banks. There will be some discipline of risk management. Much of this might be via diversification: each large bank can spread their deposits over thousands of small banks, so they may rely on statistical markers of credit risk and determine for different populations of banks a minimum deposit rate that would be acceptable. But even with diversification, it’s possible that there will be a sizable population of small banks to which larger banks are simply unwilling to lend. In that case, if the cap is too small, the interbank rate would be bid, “fast and furious” as JKH puts it, to zero.

    I have a few responses. One is to acknowledge the basic truth of the point: the central bank’s cap, along with the IOR rate, reserve requirements, and the quantity of outstanding reserves, would become another variable that would impact the effective interbank rate. All else equal, lowering the cap on IOR would at the margin imply a reduction in reserves outstanding if the Federal Funds rate is to be held constant.

    But “fast and furious” to zero is a bit of an overstatement. Between the status quo cap of infinity and a cap that enforced reserve equality, there is a lot of room for tuning. The Fed, for example, could set a cap at a rate low enough to impact the most reserve-heavy banks, but not so high that evading the cap would require lending to unfamiliar banks of questionable credit. In that case there would be an unproblematic, positive interbank rate. The cap-inspired lending would find creditworthy borrowers and a rate would be negotiated. It would still be the case that this rate would be lower than the rate that would have obtained absent the cap. But that’s not a problem, as the Fed can compensate by altering the headline rate of IOR to compensate for the downdraft imparted on rates.

    In general, this policy would put a wedge between the headline IOR rate and the effective interbank rate. But that doesn’t mean that the Fed loses control over the effective rate. It does not need a perfect correspondence between the administered and market rate. After all, until very recently the Fed was perfectly competent at targeting the interbank rate when the administered rate was stuck at zero.

    JKH points out, correctly, that the effect of this policy would become exaggerated when IOR is high, while at 25bps, the effect might be modest and some banks might eat the cost. That’s all true. Even at 25bps though, at the margin the effect would be to distribute reserves, either via deposits or interbank lending, towards smaller banks. Again, I don’t want to claim this would be a strong effect, but there’s reason to be sure it will be a weak effect, even at 25bps. Banks do love arbitrages, and do hate to forgo a low risk subsidy. In my view, there really no way to know other than to try, and one could try quite gradually to avoid making some terrible error. The Fed could set a high and slowly drop it and monitor the effect. There was a time when repos and open market operations were new, and the Fed had to develop its competence there as well.

    When, sometime in the future, interest rates rise, the Fed would have a variety of choices. It could treat an IOR cap as zero-bound unconventional policy, and simply eliminate it prior to raising rates. It could increase the level of the cap to compensate for its increasing bite. It could leave the cap constant but carefully monitor the increasing wedge between IOR and the effective rate and set IOR accordingly. Finally, if the Fed thinks that too strong an incentive to to lend reserves is counterproductive but a modest incentive is helpful, the Fed could offer tiered interest on reserves, the headline IOR below the cap and, say 50 bps less beyond the cap. This kind of tiering is an common commercial practice, although usually banks pay up for higher balances but here the Fed would pay a lower rate.

  15. JKH writes:

    There are now two different schools to which “MMT” could apply – the Modern Monetary Theorists, and the “Market Monetarists”. Ironically, they have polar opposite views of the effectiveness of central bank quantitative monetary policy:

  16. JKH writes:


    I think there’s much evidence that under the pre-crisis regime the demand for reserves was highly inelastic around the equilibrium point where the fed effective rate equalled the target rate. This also corresponds to intuition IMO in the sense that banks respond expeditiously to the cost of excess or deficient reserves in their reserve management operations in that regime. (I tend to think of this phenomenon in functional/axis reverse terms, with the price banks are willing to pay or receive to remedy deficient or excess reserves respectively as being extremely “elastic” to the supply of reserves around the same point.)

    Similarly, it appears that the presence of uncompensated and non-arbitraged excess reserves held by the GSEs has contributed to a meaningful decline in the effective Fed funds rate, relative to a likely potential trading range of zero to 25 basis points.

    So I wonder about the heightened sensitivity of the funds rate to a system in which a portion of excess reserves is uncompensated.

    In any event, the result would be a system that attempted to calibrate the trading level of the funds rate relative to a policy parameter, such as 25 basis points in the zero bound case. The effective calibration is on the level of the interest rate – which is separate from the issue of whether or not reserves per se have much to do at all with risk lending.

  17. JKH — Your point is well taken that liquidity management goes well beyond reserves. However, all reserves have to be held by someone. If the result of this proposal was mostly simple swaps, whereunder larger banks would shed reserves by purchasing Treasuries and other very liquid assets from smaller banks, then I expect it would have very little effect. But I’m almost certain that the stockpile of reserves on large bank balance sheets is sufficiently large that a cap could easily do much more than displace near substitutes, but would alter the quantity of liquid assets. The balance sheet size disparity is so large that small changes to the big-bank consilidated reserve position implies large changes (as a fraction of assets) to small bank balance sheets.

    I’ve looked at a net liquidity measure a bit broader than simple reserves. (closeliq = reserves_held + ff_sold_rev_repo – ff_bought_repo.) The stylized facts are:

    1) At all points in time, smaller banks are more liquid by this measure than larger banks. Liquidity pretty reliably decreases over time.

    2) Looking at end-of-June snapshots, banks of all sizes show drops in liquidity from 2007 to 2008.

    3) The smallest quintile of banks never recovers its (rather exaggerated) precrisis liquidity. Quintiles 2 & 3 restore approximately their original liquidity in 2009. Prior to the crisis, the fourth quintile held almost no net liquidity by this measure, and the net liquidity of the largest quintile was negative, as ff_bought_repo exceeded (reserves_held + ff_sold_rev_repo).

    Here’s a quick graph. The graph is scaled by total assets, so keep in mind that in absolute dollars, the Quintile 5 absolutely towers over all the rest. Qunitiles 1 through 4 combined account for only 6% of total assets.

    You can find my by-quintile close liquidity spreadsheet here, if you want to play.

    I’ve played with this data a bit, but I can’t say I’ve learned that much. It’d require more than looking at snapshots. I confirmed what I already knew, which is that small banks hod more net liquidity than large banks, and I learned a fact I expected would be true, that the net near-reserves position of large banks, has reversed from its historical negative position to a positive position, presumably as a result of the Fed’s balance sheet expansion and perhaps a greater precautionary inclination. Really, those reserves had nowhere else to go but big bank balance sheets. I find it interesting that the smaller three quintiles, whose combines balance sheet is a small fraction of the Fed’s post-crisis reserve expansion, did not see a big increase in liquidity over the period (the first quintile saw a large decline). But it’s not clear how one should interpret this. One could argue that small banks are uninterested in accumulated liquidity, so my story is wrong. Or one can argue that small banks find attracting liquidity costly, despite QE etc. Or one can argue that small banks have a target balance sheet liquidity that hasn’t much changed, and that reserve infusions translated to balance sheet expansions to hold liquidity fairly steady. Obviously, the last story is my favorite, because it suggests that liquid deposits to smaller banks will inspire lending.

    (The balance sheets of all quintiles grew from 6/2007 – 6/2011. Interestingly the top and bottom quintile grew fastest. The bottom quintile saw balance sheet growth despite a sharp reduction in my liquidity measure relative to 2007, which argues against my suggestion of liquidity constraint. However, the drop in liquidity was almost entirely in 2008 – 2009, and may have been involuntary. Again, data at this granularity is really hard to interpret. Fell across the board in 2008 – 2009, which is only possible if aggregate balance sheets expanded more than reserves grew, and may have been a mechanical effect of nonbanks seeking precautionary liquidity and drawing down committed credit lines during the crisis.)

    If you want more fine-graned analyses, here is a spreadsheet showing disaggregated cash and reserve positions, with banks ranked into per-period percentiles, from 2001 – 2011 (Q2).

    [Note that fedfunds and repo markets are aggregated. Very recent call reports ask that disaggregated numbers be reported, but they are not reliably. I compute ff_sold_rev_repo as greatest( rb987 + rb989, rb987, rb989, r1350 ) and ff_bought_repo as greatest( rb993 + rb995, rb993, rb995, r2800 ), that is the greatest of the several different ways these numbers might be reported. Working with bulk call report data, especially in time series is very, very messy.]

    But again, I’ve dabbled, but I’ve not designed or run any kind of experiment that would convincingly support or refute my core conjecture that pushing liquidity towards small banks could lead to more and better quality lending.

  18. JKH —

    The effective calibration is on the level of the interest rate – which is separate from the issue of whether or not reserves per se have much to do at all with risk lending.

    I’m not sure but I think that we are agreeing: Targeting the interest rate is a different thing than trying to alter the distribution of liquid assets, but the existence of uncompensated reserves in the system will put meaningful, as you say, downward pressure on rates which the Fed would need to be cognizant of.

    And all of that is completely orthogonal to the question of whether this scheme would actually “work”, in the sense of inspiring real activity.

  19. Indy — I don’t think the scheme you suggest would be too big a problem, or alternatively if it would be a problem, we have far bigger regulatory fish to fry than worrying about my proposal.

    First, an important clarification: Caps would have to be per bank holding company, not per entity within a shared holding company. If that’s not the case, then the evasion you suggest would be trivial.

    The next question is whether large banks can create and lend to captive banks that are effectively under their control but not formally a part of a common group to evade the scheme. That is supposed to be forbidden.

    Bank lending is supposed to be arms-length, because related-party bank lending invites control fraud. Risky or outright fraudulent loans to related parties can yield fringe benefits or kickbacks worth more than the capital the bank puts at risk. The controversial, now-unfortunately-somewhat-relaxed (I think) near-ban on private-equity bank investors is motivated by this concern. If you can get the leverage for your leveraged buyouts from an affiliated bank, the money you stand to lose (the direct PE capital plus bank capital) may be small relative to state-backstopped depositor funds you borrow. An arms-length banker ought to be discouraged from participating in this kind of thing because she gains nothing but modest interest from the risky LBO deals but takes a hit before the FDIC does when things go sour. But if the banker has more potential upside than modest interest when the deal goes forward, she might not be so cautious.

    The sort of captive banks that you propose would run afoul of the same regulatory proscription. Either a bank is independently controlled, in which case it can use funds it borrows in an arms-length transaction as it pleases, or it is not, in which case it is a covert special purpose vehicle of the large bank for which people should go to jail.

  20. KD — The Bank of Palookaville can be very persuasive, lemme tell ya. (Thanks!)

  21. beowulf — Differential assessments by FDIC on large and/or systemically important banks are certainly something that I support for reasons that dovetail with some of the side benefits of this proposal. I’d rather we move to a less concentrated, less “TBTFy” banking system.

    But, even if you buy my (speculative) conjecture that small bank lending is affected by liquidity flows in ways that big bank activity is not, it’s not clear that differential assessments would be stimulative. They might, if large banks try to recoup the assessments via higher fees or poorer deposit rates. But if the incidence of the fees falls to shareholders or borrowers or market-making spreads or derivative counterparties, differential assessments won’t constitute expansionary monetary policy as I claim my caps would.

    (Re your second paragraph, the case you cite looks interesting, but I’m not quite sure how enforceability against the government of claims against FDIC relates to this post.)

  22. JKH —

    Coming back to my response to Scott, this is a core issue:

    Small banks should not be more susceptible to breaking tradition in looking at capital ratios before implementing newly aggressive risk lending programs, merely because they have reason to adjust their liquid asset portfolio mix while accounting for a marginally higher cost of liquidity.

    My conjecture here (and the reasoning behind the Kashyap and Stein study, it’s not just me) depends on the existence of significant frictions for small banks seeking liquidity. If there are no frictions at all — if small banks can borrow in for long periods and in arbitrary quantities at the interbank rate — then my conjecture is bullshit. If banks face only a “marginally higher” cost of liquidity than large banks, then this proposal would probably do very little.

    But I think a marginally higher cost is an insufficient characterization of the difference between small and large banks’ access to liquidity. I think small banks have access to inexpensive on-demand liquidity only if they tap it infrequently and transiently. I think small banks that try to use interbank borrowing as a permanent source of funds experience sharply rising costs and risk being rationed out of the market, provoking the liquidity problem they intended to avoid. So, I think smaller banks insist on precautionary buffers of liquid assets, which implies that lending is constrained by liquidity. (I think I’ve seen empirical evidence of this — that lending activity in a geographic area is responsive to deposit flows — but I don’t remember the studies.) If smaller banks tether loan growth to liquidity, and if the status quo interbank market is not an adequate source of liquidity, then shifting deposit flows or increasing small-bank access to the interbank market would increase activity.

    It’s a big if, and I could be wrong. But I think I’m right. And if I’m wrong, if the interbank market is nearly frictionless, capping IOR would alter the distribution of reserves in the banking system but otherwise do no harm. (The cap would only put downward pressure on the interbank rate if big banks are reluctant to let their reserve holdings be intermediated by smaller banks, that is, if small banks face significant frictions.)

  23. JKH writes:

    My apologies for so many disjointed comments here, but the reserve “transmission” mechanism under discussion here is a pretty delicate one.

    Going back to basics, banks settle transactions of all types through reserve accounts. The idea that MMT has been the blogosphere leader on is that at the macro level, the central bank supplies the reserves that are required in order to settle transactions in general at the target interest rate. This also implies that individual banks in general will be able to settle their transactions, provided they are operating viability otherwise.

    An individual bank that begins to experience liquidity difficulties may not be able to do this. But that is considered to be an exception to the general case in which reserves are in adequate supply for most banks to settle transactions in an environment where the central bank is meeting its target interest rate.

    The associated idea is that in general banks do not require stocks of reserves in order to make loans. What they require is an operational capability to balance their reserve settlement account to the desired level in the normal course of business. A bank in Montana can make a loan, and balance its reserve account by drawing down fed funds sold, increasing fed funds bought, raising deposits, or selling securities. That is liquidity management. Alternatively, it can draw down a pre-existing surplus balance in its reserve account, but such a surplus is not a necessity for lending. It is in that sense that at both the macro and micro level, banks do not require pre-existing reserve balances for lending. The function of normal liquidity management is to ensure access to reserve balances when required via the effect that all other types of asset-liability transactions have on reserve balances. And it is important to note that direct transactions in the fed funds market are only one method of accessing reserve balances. Other types of transactions executed in all areas of the balance sheet can have a similar effect on reserve balances, provided counterparties are sourcing their funds from other banks. The fed funds market just happens to be the market that banks use to transact directly in reserve balances with each other.

    What banks do require in order to take on risk in lending (or other forms) is adequate capital. This is simply a fact of banking life, viewed from the regulatory perspective, the market perspective, or an internal management perspective. And in this sense, banks do require pre-existing stocks of UNALLOCATED book capital in order to take on risk in lending or otherwise. This can result from internally raised (retained earnings) or externally raised (equity issues) capital. But note that externally raised capital is a unique, heterogeneous market. Unlike fed funds for liquidity, there is no homogenous market for direct equity injections, and there certainly is no fundamental by-product market for equity injections in the way that many types of customer transactions can routinely source reserve settlement balances for banks. Reserve balances and capital balances are very different in their temporal appearance.

    As discussed above then, there are many ways in which a small bank can manage its liquidity position in order to be positioned operationally to ensure access to settlement balances once the capital decision has been made and the actual lending transaction is drawn down for funds. The bank doesn’t necessarily need reserve balances, or fed funds sold to draw down on, or fed funds bought to rely on. It has other sources of liquidity on both the asset and liability sides of its balance sheet.

  24. JKH writes:


    Thanks for that data explanation.

    (BTW, in one of your earlier posts, I linked to some data that showed that the aggregate system reserve increase attributable to QE2 was largely matched, remarkably, by a nearly equivalent increase in reserves held by foreign banks operating in the US, at least at that time. I’ve not investigated that further, or seen reference to it, or explanation of it, elsewhere.)

  25. JKH — I get all of that. Believe me, I’m very aware of the distinction between capital and liquidity, of the fact that loan extension per se does not require reserves, and of the arguments why for the most part it is better to think of lending as capital constrained rather than liquidity constrained.

    But I am challenging that orthodoxy (heterodoxy?) with respect to small banks. I think it’s not entirely true. Banks have many means of managing their liquidity position, but I think that small banks face sufficient friction that they cannot treat liquidity for settlement as an afterthought. I don’t at all mean to suggest that the interbank reserve market is the only or primary means by which small banks might attract liquidity. On the contrary, I think small banks rely substantially on attracting customer deposits rather than on secondary liquidity markets of any sort, that they sell Federal Funds and run down FF sold as necessary, but purchase Federal Funds only rarely.

    Broadly, I think the relationship of small banks to interbank borrowing is quite similar to the relationship of large banks and the Fed’s discount window. In theory, it is always there, but in practice, the cost of frequent access in terms of reputation, regulation, and risk is substantial. So a bank might be well capitalized and have lending opportunities that it would be willing to finance, if it could borrow liquidity at the policy rate, and yet it might forego those loans for fear that it would be costly or impossible to settle transfers and cover reserve requirements should those loans engender external payments.

  26. JKH writes:


    Maybe another way to think about your proposal is to examine it in the context of a continuum between two extreme points – the existing full cap regime for reserve compensation versus the (Scott Sumner / full Monty) zero cap regime where reserve compensation is eliminated entirely. One wonders about both the interest rate sensitivity and the activity sensitivity along that continuum. I guess my instinctive position roughly is that the interest rate sensitivity is relatively acute as soon as you open it up to non-compensated balances, but that the activity sensitivity would be fairly mild. Another way of saying this is that my expectation would be that the difference between your proposal and the Sumner extreme is not great, on my assuming heightened interest rate sensitivity and muted activity sensitivity at the point of opening up the uncompensated subset of reserves. Your view I think is that these sensitivities would be meaningful. I guess that’s what makes a market.

  27. JKH writes:

    Separate point:

    Is it fair to say that small bank balance sheets in general reflect the strength of retail deposit bases?

    Is it then fair to say that the interest margin economics of maintaining their net liquid positions become relatively more favourable, and that the costing of liquidity is not a straightforward comparison with the wholesale funding requirements of large banks?

    And would that affect the pricing and activity sensitivity of the small bank bid for funds under your proposal?

  28. beowulf writes:

    “(Re your second paragraph, the case you cite looks interesting, but I’m not quite sure how enforceability against the government of claims against FDIC relates to this post.)”

    1. FDIC asset-based fees are at cross purposes with the Fed targeting an overnight rate. In theory, FDIC could set policy rate by raising or lowering its fee rate, but then any FFR north of 0 would be at cross purposes with the FDIC.
    2. There’s no reason for the FDIC to collect these fees since the cost of future bank bailouts will invariably be charged directly to the guys with the printing presses (and the 14th Amend duty to pay its debts).

  29. Scott Fullwiler writes:

    I think the important point JKH raises has to do with settlement of payments. Small banks almost always have sufficient reserve balances to meet RR. If they don’t, they can use sweeps, etc. But by and large, particularly starting in the late 1990s, the sweeps enabled them to effectively meet RR with vault cash. Even if not, RR isn’t really the issue since a decision to expand the balance sheet and thus add reservable liabilities in any significant way doesn’t turn on a dime, and as JKH notes is highly influenced by capital levels. And once that decision is made, there is time to find new lines of fed funds credit (again, the issue there is overwhelmingly the cost, not the qty, IMO, particularly for the amounts we are discussing and that there is a time lead at work) or arrange to sell liquid assets.

    So, what banks then need, by and large, as JKH noted, are reserve balances to settle payments. And if so, then a proposal like this really misses the point (even aside from the other issues, like a highly inelastic demand for reserve balances, that expose shortcomings in my view). In that case, and particularly given SRW’s points on difficulties assessing credit quality that are that much worse on the fly when a bank needs balances now to end the day positive, then the solutions are much simpler–a central bank with a true no-questions-asked standing facility with a very modest penalty rate (for liquidity purposes, not solvency, though that should go without saying) and to set the central bank as the counterparty to all such interbank lending much the same way that the exchanges are counterparties to traders. These are particularly useful given that the central bank is actually desiring to hit an interest rate target (!!) and that when a liquidity crisis does actually hit the central bank takes interbank activity onto its balance sheet anyway. (I proposed this in a paper I wrote in early 2009 for my book, btw.)

  30. JKH @ 26: You are definitely right to compare this to Scott Sumner’s frequent advocacy for zero / neg IOR (not an idea foreign to interfluidity either). I am broadly with Scott on that. I’d be happy to see the Fed give it a try, although I don’t expect it to work magic. (We discussed this in a recent comment thread.)

    Part of where this idea came from was thinking about what could be done taking the Fed’s clear preference for a positive interbank rate as a constraint. Initially, I thought of it as you are suggesting, as an intermediate step between positive and zero IOR. But I’ve come to think that’s the wrong way to think about it, for reasons that you pointed out earlier in the thread: the effect of IOR caps on liquidity distribution increases with the headline interest rate. Ultimately, the two proposals work, if they work, by different mechanisms. Zero/neg IOR would work by reducing the opportunity cost of lending to banks with access to interbank finance and thereby rendering some potential loans near the margin worth funding. Fundamentally, zero/neg IOR works through the price of funds. IOR caps, if they work, would probably do so by overcoming the effective rationing of liquidity to small bank. Although you can put it in terms of price, you have to posit a sharply increasing, nonlinear cost-of-liquidity schedule, and it’s probably more natural to think of it as a quantity constraint.

    I don’t know whether either would work, and I don’t know the answer to the relative effect on interest rates vs activity, although I’m pretty sure the Fed is capable of neutralizing the interest rate effect so I’m not worried about it. Either proposal strikes me as worth trying, because the costs strike me as small or negative, and the effect at the margin is directionally good. These policies might or might not prove quantitatively ineffective, that depends upon how much potential activity is sitting at the margin in terms of pricing or in unfunded loan applications at smaller banks. I don’t know. I think the only way we can know is to try. (Counterintuitively, it would be possible to try both. For example, headline IOR could be reduced to zero, but a “storage fee” could be charged on balances above a cap. In this case, large banks would have an extra margin to respond along — hoarding physical cash — but could still reduce payments/increase charges to depositors or lend reserves to smaller banks.)

  31. JKH — I do think it’s fair to say that small bank balance sheets in general reflect the strength of their local deposit bases, and “the costing of liquidity is not a straightforward comparison with the wholesale funding requirements of large banks”.

    That’s much of what motivates the proposal. I think small banks pretty much live much closer to a gold-standard world than big banks. So I think that “quantitative easing” for small banks can be effective at generating activity, while for big banks it’s an irrelevance as long as the price of reserves in unmoved.

    The effect on interest rates would be determined on the other side of the trade, by the degree to which big banks are willing to either accept the credit risk of small banks or alienate retail depositors. If big banks have no fear at all, the interest effect would be minimal (unless the cap is very low). Small banks would compete for big bank reserves and bid up the deposit rate to something near the level of IOR. If big banks are chary about lending to small banks, then the wedge between IOR and the interbank rate could grow large, as the few “reliable” smaller banks could offer very low interest rates and still attract funds from large banks. In this case, the interbank rate (which is really the big-bank rate) might be forced all the way to zero unless the Fed raised the level of IOR.

    But even if big banks don’t trust small banks, if they are willing to alienate depositors (with low rates/high fees), then there is a risk-free arbitrage. As long as the flow of funds from large to small banks is intermediated depositors, an state guarantee attaches.

    So I’d expect the interest effect to dominate if and only if 1) big banks can’t find trustworthy smaller banks, or view the cost of credit analysis as larger than the value of reserve compensation; and 2) big banks view the reputational and long-term relationship cost of increasing per-dollar revenues from but losing quantity of customer deposits as intolerable, so they do not send funds elsewhere on the backs of customers.

  32. beowulf — I think you are condensing things a bit too much.

    Looking at the banking system in aggregate and imagining that bank profits are held as retained earnings, your reasoning would be 100% correct. But the banking system does not always fail systemically, and bank profits are frequently distributed, so it matters that the FDIC collects premiums.

    When there are individual bank failures but not a systemic failure, if FDIC did not collect fees, that would be a net cost to taxpayers. But under a deposit insurance system, idiosyncratic bank failures do not cost taxpayers anything; they are paid for by transfers from other banks.

    In a systemic crisis, the aggregate banking system balance sheet comes to matter. If banks’ alternative to paying FDIC were retaining earnings, then as you suggest all the fees the banks have paid come to seem irrelevant. The government is on the hook for the difference between overall bank capital and the system’s liabilities in excess of asset values. FDIC fees just move some of the banking system’s economic capital from notionally private accounts to a notionally public trust, but that has no effect on the overall liability. The economic capital of the banking system is the consolidated private capital plus trust fund balance, and the division of funds between those two accounts seems to matter not a whit.

    However, it does matter, because banks don’t retain their earnings, they make distributions. Sucking capital into the trust fund diminishes the ability of banks to distribute capital, and therefore increases the size of the cushion that must be exhausted before taxpayers are on the hook. Holding other regulation constant, eliminating deposit insurance fees would be a transfer to bank shareholders, as who would receive greater distributions than they otherwise would have prior to the systemic crisis that eventually wipes them out. Holding other regulation constant, FDIC’s funds increase the effective capital requirements of the banking system.

    Now we don’t have to hold regulations constant, and it would be possible to replicate the effect of FDIC on the systemic balance sheet by eliminating premiums but increasing capital requirements or restricting distributions of capital to shareholders and employees. That might be cleaner, in some sense, but it might also be politically more difficult to enact and sustain. The frame of deposit insurance is politically defensible relative to the frame of telling corporations what they can and can’t do with “their” money.

    And capital requirements can’t replicate the mutual insurance aspect of FDIC viz individual, idiosyncratic bank failures.

  33. Scott — I don’t think the details of whether banks require reserves to meet regulatory requirements or whether they require them simply to settle outflows much matters here. As you say, banks have developed lots of tools for minimizing the bite of reserve requirements. Empirically I believe those tools are used disproportionately by larger banks, but not exclusively. Ultmately it doesn’t matter. There is some level of reserves below which a bank which cannot borrow on demand is in jeopardy of facing serious costs, whether due to regulatory action or failure to settle net flows.

    As you suggest, the most parsimonious way to address this constraint on small banks would be simply to eliminate it: make it possible for small banks, like large banks, to borrow “friction-free” at the policy rate.

    However, this might prove easier said than done, and might prove risky.

    Suppose the Fed were to announce that it would lend uncollateralized to any bank at the interbank rate through its discount window. In theory, that would eliminate small banks’ liquidity constraint. But in practice, it might not. It remains the custom that small banks tend to their own liquidity, and under the current custom, seeking liquidity from the central bank is taken as a signal of weakness by counterparties and regulators. A statement by the Fed, in and of itself, cannot eliminate the “stigma” associated with tapping its facilities. There needs to be a change in actual practice and custom, so that it becomes irrational to view use of the facilities as informative of stress. In the run-up to the crisis, the Fed tried very hard to destigmatize its discount window. It asked the biggest banks to borrow, hoping that would legitimize the facility. It didn’t work. Ultimately, the Fed had to push its funding to banks via TAF, which carried less stigma because the auction system allowed banks to claim they accept funding on the basis of advantageous pricing rather than need. Borrower-initiated use of Fed-sourced funds remains stigmatized among banks large and small to this day.

    One way to think about IOR caps would be as a way of deputizing large private banks to implement a TAF program for smaller banks. Large banks would “push” funds out to lend, as the Fed did, to avoid the opportunity cost of holding them directly. Smaller banks in accepting funds from private banks would not be stigmatized as they might for seeking liquidity from the Fed.

    On the risk side, if uncollateralized, no-questions-asked direct lending by the Fed could be destigmatized, that would increase the Fed’s overall danger. Yes, the risk of the interbank market is effectively socialized in a systemic crisis. But not all crises are systemic, and the Bank of Palookaville might tap a large Fed loan and go under even in good times. In bad times, the fact that the Bank of Palookaville was excluded from the interbank market reduces the total size of that market and therefore the risk position of the Fed.

    In theory (although not in practice, unfortunately), the Fed monitors the risk of large banks carefully. But no one expects the Fed to monitor the risk position of the Bank of Palookaville so carefully. FDIC or OCC or some state agency examines it, and perhaps the Fed should be able to rely upon delegated risk management. But for the most part, contra Mosler, we rely very much on liability-side discipline for small banks. Small bank sharehoders are never bailed out in a failure, and small bank creditors never allow them to take on the sort of leverage that large banks concoct both overtly and covertly. Direct uncollateralized lending to small banks would involve either a significant increase in the asset-side regulatory burden, or a significant increase in risk.

    IOR caps wouldn’t magically solve this problem. But to the degree that reserves flowed to small banks via interbank markets, there would be some private sector due diligence and discrimination among small banks.

    However, to the degree that reserve flows took the form of customer deposit migration, there would be no due diligence: banks’ customers put FDIC on the hook with no credit analysis whatsoever.

    In summary, what you propose — frictionless direct lending at the interbank rate to small banks — could work, it would solve the problem I’m trying to solve. I’d be enthusiastic to try it, if we can devise credible means of risk management and overcome the stigma. IOR caps are in a sense just a means of trying to implement your solution, laundering loans through private banks to overcome stigma and to offer some degree (but not on its own an adequate degree) of risk management.

  34. JKH writes:


    Here’s a point I don’t know that we’ve touched on. My impression is that the small bank balance sheet model (in general) exhibits a relatively large retail deposit base, relatively small reliance on wholesale funding, a relatively large liquid asset portfolio, and a loan book. All in all, that is a relatively liquid balance sheet. To the degree that the bank is well run, with a strong capital position and good risk management, it sources cheap and fairly reliable retail funding and manages its liquidity mostly through the asset side. Deposit growth feeds those sources over the long term. But at the margin, loan draw downs are funded by cash coming in mostly from the asset side, through asset maturities, asset calls, or asset sales. Fed funds sold would be part of the liquid asset portfolio. Finally, remember that small banks generally get in trouble because of their loan portfolios and their capital positions. That becomes a risk to depositors. And the FDIC moves in as required. But they don’t get in trouble from the get go because their retail deposit base is weak. At least that’s my impression of the general case. There may well be exceptions in specific cases.

    If that model is accurate, then there really is a question as to how much value added a large bank can offer to a small bank by offering funds in the interbank market – funds that the large bank deems disposable at a reasonable price because the Fed has suddenly stopped paying IOR on the top tier of its excess reserve balance. There is an element of pushing on a string here. That element presumes that this additional quanta of funds would be desired small banks to a degree that would cause them to bid up.

    I question this not only because it doesn’t reflect the business model outlined in my first paragraph, but also because it omits consideration of an additional risk. This is the fact that such funding is not a reliable source of liquidity and not a reliable source of funding for the longer term. Any small bank that takes on such discretionary funding from large banks will be subject to the risk that this is essentially “hot money” that will be recalled at the first signs of Fed tightening and even preliminary stages of exiting from its QE reserve position. And there may be other reasons why the large bank pulls back. Interbank funding has its place, but not necessarily as the marginal source of funding for a strategic ramp up in lending.

    Beyond that, it is clear from your discussion and comments that the mechanism in order to be effective in migrating reserves from large to small banks will have to demonstrate the right kind of interest rate sensitivity. On that issue, I see no reason why the existing credit risk and interbank interest rate tiering between large, presumably high quality banks and small, presumably riskier banks will change under a capped reserve compensation system. The overall interest rate structure of the market will slide in linear transformation fashion, it seems to me. All interest rates will adjust. But this point is only secondary to the first about liquidity profiles of small banks in general, and the nature of their needs and preferred sources for liquidity management.

  35. JKH — A lot of good points.

    To a first approximation, you are right that new interbank flows would be “hot money”. Big banks would wish to withdraw the funds as soon as the monetary-policy-induced advantage diminished. If smaller banks anticipated the flows to be very transient, these deposits wouldn not inspire much activity at smaller banks. Effectively, they’d let smaller banks share some of the subsidy inherent in IOR > policy rate, but have little other effect.

    But that is a function of expectations, and central banks are in the business of expectations management. If the central bank wishes to inspire lending at smaller banks, it can make clear that it intends to keep conditions liquid for smaller banks by capping reserve compensation. No doubt there’d be a wait-and-see period. But if the Fed intends, qua my conversation with Scott, to use IOR caps to simulate for small banks the lack of liquidity constraint that large banks enjoy, it would manage the caps so that a substantial fraction of smaller banks consistently have funds lent or deposited from larger banks. (A substantial fraction, but not all, because the Fed would want to outsource regulation by encouraging some monitoring and discrimination by creditors between smaller banks.) If the Fed were to adopt this kind of regime, well-managed small banks could come to view this source of funds as a fairly stable funding base.

    Plus, there is the second channel through which funds would flow, the client-mediated flows as larger banks discouraged deposits at the margin by paying lower rates and charging higher fees. These flows would come only slowly — we refer to retail deposits as sticky for a reason — but over time they would come, and would add to smaller banks’ traditional source of liquid assets.

    A third point is hysteresis: Once large banks go through the trouble of evaluating and ranking smaller banks as places to park reserves, the investigative work and history they accumulate is a sunk cost. This history would tend to make the money less hot over time than it would immediately: a bank that felt “forced” today to shove reserves into an uncertain partner would pull it when the forcing ends, but a bank that has maintained a lending relationship for some time with another bank might be perfectly happy to maintain the relationship as long as the rate offered is competitive with the interbank rate. Symmetrically, large banks that currently ration smaller banks out of the market for persistent interbank finance may be willing to sell Federal Funds to smaller banks they have evaluated and with which they have developed a history. In the limit, the effect is to wear down the frictions that make small banks different from big banks and to asymptotically approach the big-bank no-reserve-constraint-but-price regime by eliminating the information problems that prevent it for healthy smaller banks.

    One last point on the “hot money” issue. Just as the central bank could prevent interbank funding from behaving as hot money by sustaining caps, it could heat up the money again by unexpectedly reducing or eliminating caps. If the hysteresis described above has not eliminated small-bank borrowing constraints, that would create a liquidity crunch and lead to a sharp pullback in discretionary lending activity at smaller banks. This is the effect that Kashyap and Stein observe with ordinary contractions and small banks. Small banks seem to experience contractionary monetary policy as liquidity scarcity and curtail lending. They may be more likely to get into trouble during contractions, running afoul of regulatory liquidity requirements, although there’s a paper (I think it’s a Peek & Rosengren joint, not sure off-hand) that finds regulators offer forbearance to banks when they deem imperfect compliance is attributable to monetary policy fluctuations.

    In any case, if “hysteresis” doesn’t happen, the effect of this proposal would be to increase the potential traction of monetary policy by magnifying the potential for contraction via a bank lending channel. If “hysteresis” does happen and small banks come to gain access to interbank finance on terms similar to those available to larger banks, then the bank lending channel of monetary policy might be attenuated or eliminated. I have no idea whether hot-money or reduced-information-asymmetry effects would dominate, but it’d be interesting to find out.

    Thanks as usual for thinking this through so carefully with me!

  36. JKH writes:


    My interest is largely in probing the rationale behind the probability that this approach would be effective – not so much that it shouldn’t be tried.

    It’s an interesting idea, and an interesting post – as usual. Thanks for the discussion.

  37. Detroit Dan writes:

    There’s a lot of stretching goin’ on trying to find some way to enlist monetary policy to solve our problems. The thinking seems to be that monetary policy is fairer than fiscal policy. Yet monetary is a blunt instrument that primarily affects banks and is implemented largely by unelected bankers. And when monetary policy goes unconventional, then the opportunity for unfair and unaccountable policy is greatly increased. How about when the Fed bought up all the mortgage backed loans, for example?

  38. Detroit Dan writes:

    Just ran across this: GAO Finds Serious Conflicts at the Fed

    WASHINGTON, Oct. 19 – A new audit of the Federal Reserve released today detailed widespread conflicts of interest involving directors of its regional banks…

    The GAO detailed instance after instance of top executives of corporations and financial institutions using their influence as Federal Reserve directors to financially benefit their firms, and, in at least one instance, themselves…

    The corporate affiliations of Fed directors from such banking and industry giants as General Electric, JP Morgan Chase, and Lehman Brothers pose “reputational risks” to the Federal Reserve System, the report said. Giving the banking industry the power to both elect and serve as Fed directors creates “an appearance of a conflict of interest,” the report added.
    The 108-page report found that at least 18 specific current and former Fed board members were affiliated with banks and companies that received emergency loans from the Federal Reserve during the financial crisis.

    In the dry and understated language of auditors, the report noted that there are no restrictions in Fed rules on directors communicating concerns about their respective banks to the staff of the Federal Reserve. It also said many directors own stock or work directly for banks that are supervised and regulated by the Federal Reserve. The rules, which the Fed has kept secret, let directors tied to banks participate in decisions involving how much interest to charge financial institutions and how much credit to provide healthy banks and institutions in “hazardous” condition. Even when situations arise that run afoul of Fed’s conflict rules and waivers are granted, the GAO said the waivers are kept hidden from the public.

    The report by the non-partisan research arm of Congress did not name but unambiguously described several individual cases involving Fed directors…

  39. DoubleDeuce writes:

    When you mention “interbank lending” do you also consider the repo market? Because the repo market has involvement by much, much more than just banks. For example, a manufacturing firm like mine utilizes the repo market anytime that we have a positive cash balance in our accounts. I guess that my question is when you discuss interbank lending do you mean lending between banks that is funded by much more than the own bank’s cash position?

  40. Sergei writes:

    There might be additional risk-related reasons why small banks tend to be net lenders in the fed funds market rather than net borrowers. The business model of liquidity book can vary from bank to bank but in general one could argue that at least some portion of it is marked to market and can lead to oversized p&l volatility. All banks are exposed to interest rate risk (in the banking book) which demands (regulatory) capital. Finally, small bank might be reluctant to build up capacity and expertise to run complex liquidity books and therefore prefer to rely on super plain vanilla liquidity management via reserves.

    One additional point. The interbank market is in many countries a tiered market with two tiers: payment-heavy large universal banks and small banks. In such systems the former tend to provide payment services to the latter. Obviously for a fee. So small banks rely on payment services provided by large banks while cash collection activities (from the real economy) are realized via net reserves lending.

  41. winterspeak writes:


    the kashyap and stein paper is very weak — and it’s weak in exactly the same way the “safe” national debt/GDP paper, I beleive by sitglitz but I could be wrong — was weak as well.

    It is ridiculous to run a regression to figure out how monetary policy works. You’ll find a correlation, or keep running those numbers until you do, and then come up with a causation and supports whatever you believe. You and I have both seen enough fraudulent economic papers to know how this game is played.

    The discussion of the M-M channel at the start of the paper is completely wrongheaded, as is the discussion of what happens when the Fed drains reserves from the system.

    With respect, I think your rube-goldberg policy suggestion is a besides the point.

  42. […] Thoughtful analysis of the ways interest on excess reserves affects bank lending […]