Tuesday, August 22, 2017

The IOER Debate Redux

Back in the glory days of macroeconomics blogging there was a lot of electronic ink spilled over interest on excess reserves (IOER). Commentators, including myself, debated whether IOER mattered to the recovery or if it was just another innocuous tool for the Fed to control interest rates. 

I generally argued that the IOER did matter for the economy--it was more than just a new tool. It began with a call I  made in October 2008 that the introduction of IOER that month was likely to be contractionary. In later conversations, I acknowledged that, yes, the Fed does sets the aggregate level of reserves. Even so, I retorted, banks could still influence the composition of all those reserves based on their investing decisions. These decisions, in turn, could be influenced by the level of IOER. That is, if IOER were set high relative to other safe asset yields then banks might decide to invest in excess reserves rather than in other safe assets like treasury bills. This could stall the 'hot potato' process and affect the recovery. For example, imagine the economy starts heating up and, as a result, the demand for loans picks up. Banks facing this increased pressure for money creation might opt to invest in excess reserves instead of loans if the risk-adjusted return on excess reserves were high enough. That could happen by raising IOER sufficiently high. Consequently, IOER mattered to macroeconomic policy and needed to be set appropriately.

The above paragraph roughly summarizes my position during the many IOER debates that took place over the past decade. Needless to say, I got plenty of pushback and there were many spirited debates. These exchanges sharpened my thinking on the topic. Here, for example, is a long write up from Cardiff Garcia at FT Alphaville on one such debate in 2012. Those were fun times, but folks generally moved on to other conversations.  

One person, though, who kept the IOER conversation going is George Selgin. He has written extensively on IOER, most recently in a 60-page testimony to the House Committee on Financial Services. In it, Selgin argues that the Fed has, in fact, set the IOER too high and this has been a drag on the recovery. Along these lines, he presented a chart on page 20 that shows what appears to be a systematic relationship between (1) an IOER and comparable market interest spread and (2) the relative demand for excess reserves. 

The chart was intriguing, but its sample period did not span the whole IOER period. So I wanted to see if the relationship was robust across the period. Also, I thought it would be useful to look at the actual holders of the excess reserves. The figure below shows the combined cash assets of "large domestically-charted banks" and "foreign-related" banks as reported in the Fed's H8 report. These combined cash assets track excess reserves fairly closely. These two types of banks, then, are the main holders of excess reserves.

Following Selgin's example, I plotted the (1) spread between the IOER and the overnight LIBOR and (2) cash assets as percent of total assets. I did so for both the foreign-related and large domestic banks. If the IOER spread does in fact cause banks to hold more excess reserves relative to other assets, then we would expect the banks share of excess reserves in the portfolios to go up with the spread. 

The figure below confirms that this is the case for the foreign-related banks for the period December 2008 - July 2017. The relative yield on excess reserves does seem to influence the real demand for excess reserves. 

The next figure puts these two series together in a scatterplot. The IOER - excess reserve relationship is strong with a R2of 73%.

Next, I looked at domestically-chartered banks. There is still a positive relationship here, but it is weaker as seen in the next two figures.

The last figure shows the relationship is not trivial--it has an R2of 41%--but it is nowhere near the strength of the foreign banks. So for some reason the IOER-Libor spread  creates a stronger incentive for foreign banks to hold comparatively more excess reserves.  That is an interesting observation worthy of future exploration.

The main takeaway, though, from the above figures is that it appears Selgin's claim is correct. A rise in the IOER spread does seem to influence the relative demand for excess reserves, with the effect being  strongest for foreign banks in the United States.This implies the IOER is more than just a new interest rate tool for the Federal Reserve. George Selgin may have just rekindled the IOER debate. 


  1. Doctor Beckworth
    The chart where you compare IOER-LIBOR with Cash/Assets, I have some doubts. The rise in the black line just represents the QE programs (increases) in the numerator combined with the normal expansion of deposits that expands denominator. In a period of absent injections of MB, we see the ratio decreasing, because deposits continue to rise without the CB injections. The difference in rates would mean that whenever fed injects MB there is a liquidity effect that affects LIBOR without afecting IOER (therefore causing an increase in the blue line). So this could be all caused by Fed "injections" of reserves.

    Can you tell me where am I wrong?

    1. There is indeed something to the liquidity effect part of this counterargument. Deposit growth, on the other hand, doesn't seem at all correlated, even allowing for any lag, with Fed assets. Some fancy econometrics may be needed to untangle it all.

    2. I am not claiming that deposits are correlated with reserves Doctor Selgin.
      I know deposits aren't assets, Doctor Beckworth. But if created by loan extension, both assets and liabilities will expand.
      I just claimed that as "loans create deposits.." through increased capital/creditworthiness of borrowers, the ratio will decrease absent of Fed injections

    3. osinterest, deposits enter the liability side of the balance sheet and are therefore not part of the black line in the chart which is cash assets/total bank assets.

      It is true, though, the foreign banks' black line does loosely follow the contours of the QE programs. However, note that the large domestic banks do not follow QE as closely. The third category I did not include in the post, the "small domestically-chartered" banks, follows QE even less so. As a result, this can't be simply a story of the Fed forcing reserves into banks during rounds of QE.

      Different types of banks chose different levels of excess reserves to hold. And per George Selgin's point above about foreign banks not having to pay the FDIC fee and thus making the IOER-LIBOR spread easier for them to capture, it seems the foreign banks were better incentivized by the spread to hold more excess reserves as they became available. And large-domestic banks were better incentivized than the small domestic banks. Put differently, the IOER-LIBOR spread elevated the real demand for bank reserves differently for different types of banks.

    4. losinterest, okay, but your explanation only works for the foreign banks and doesn't explain the other banks. How do you reconcile that? (Feel free to call me David.)

    5. David (my name is also David, but Ricardo David)
      I really dont know how to reconcile that.
      My conjecture would be that "no Fed base injections are created the same way", some may increase american cash assets, some foreign ones. But I think your reply (plus Doctor Selgin's) also make lot of sense.
      Even though I wanted to point out the liquidity effect and the denominator increase also is important!
      Thank you
      Best regards!

    6. Ricardo David, what is needed is a good econometric study of these relationships. That would help clear up matters.

  2. David, thank you for drawing attention to my testimony. Regarding the closer correlation for foreign banks, it is not surprising in light of the fact that US branches of foreign banks are better able to profit from IOER to the extent that, if they offer no insured US deposits, they do not have to pay FDIC insurance premiums, which are assessed according to total assets (including reserves) rather than insured deposits.

  3. Interesting. What other components does "cash assets" include other than reserves? Sales of fed funds?

    I only ask because one potential theory for the relationship in your chart is that when the "IOER minus LIBOR" spread is high, its because GSEs (who can't hold reserves) are desperate to buy fed funds, thus pushing the overnight rate far below IOER. And since banks, in particular foreign ones, are sellers of fed funds, this shows up in their cash assets.

    1. That is a good question. Here is the Fed's definition of 'cash assets': "Includes vault cash, cash items in process of collection, balances due from depository institutions, and balances due from Federal Reserve Banks". George can better answer your question.

    2. JP, the presence of GSE's certainly explains why the effective ffr is almost always below the IOER rate. But I can't see how it would explain the correlations shown in either my or David's posts. The GSEs do not in fact profit by "buying" fed funds. Rather, they lend such funds to banks overnight for a share of banks' IOER earnings. The lent funds show up as loans, not cash.

    3. Doctor Selgin
      But aren't the loans in GSEs assets, correspond to the liability of the bank which "uses it" as an reserve balance earning IOER (asset), which is a liabilty of the FED. Just a roudabout process. No?

  4. Fucking stupid. Lending by commercial banks is inflationary. Lending by the non-banks is non-inflationary. The IOeR allows the commercial banks to outbid the non-banks for wholesale funding in the borrow-short to lend-longer maturity/risk matching process. But the reverse of this operation is impossible. Savers (contrary to the premise underlying the DIDMCA of March 1980, in which DFIs are assumed to be intermediaries and in competition with non-banks) never transfer their savings out of the payment’s system (unless they are hoarding currency or convert to other National currencies). This applies to all investments made directly or indirectly through intermediaries (i.e., non-banks).

    Shifts from time deposits to transaction deposits within the DFIs and the transfer of the ownership of these deposits to the NBFIs, involves a shift in the form of the DFI’s liabilities (from TD to TR) and a shift in the ownership of existing TRs (from savers to NBFIs, et al). The utilization of these TRs by the NBFIs has no effect on the volume of TRs held by the DFIs nor the volume of their earnings assets. I.e., the non-banks are customers of the deposit taking, money creating, DFIs.

    In the context of their lending operations it is only possible to reduce bank assets, and TRs, by retiring bank-held loans, e.g., for the saver-holder to use his funds for the payment of a bank loan, interest on a bank loan for the payment of a bank service, or for the purchase from their banks of any type of commercial bank security obligation, e.g., banks stocks, debentures, etc.

    From a system’s belvedere, “the commercial banks do not loan any existing deposits, demand or time; nor do they loan out the equity of their owners, nor the proceeds from the sale of capital notes or debentures or any other type of security. It is absolutely false to speak of the commercial banks as financial intermediaries not only because they are capable of “creating credit” but also because all savings held in the commercial banks originate within the banking system. The source of time deposits is demand deposits, either directly or indirectly via the currency and undivided profits accounts of the banks…” - L.J.P.

    The financial intermediaries can lend no more (and in practice they lend less) than the volume of savings placed at their disposal; whereas the commercial banks, as a system, can make loans (if monetary policy permits and the opportunity is present) which amount to several times the initial excess reserves held.

    Financial intermediaries (non-banks) lend existing money which has been saved, and all of these savings originate outside the intermediaries; whereas the DFIs lend no existing deposits or savings; they always, as noted, create new money in the lending process, i.e., pay for their new earning assets with new money.

    Saved TRs that are transferred to the NBFIs, etc., are not transferred out of the DFIs; only their ownership is transferred. The reverse process, which is called “disintermediation”, has the opposite effect: the intermediaries shrink in size, but the size of the payment’s system remains the same.

    This was the fundamental and monumental theoretical error Bankrupt u Bernanke executed which deepened and prolonged the GFC, as well as resulted in subpar economic growth.