Monday, September 23, 2019

The Repo Man Cometh

The repo market hit some road bumps last week. Trading pressures in this key funding market pushed repo interest rates well above the Fed's target interest rate range. This development caused some observers to worry that it was a 2008-type run on the repo market all over again. Bill Dudley and others, however, noted this was a technical blip, not the beginning of a financial crisis. Moreover it was something the Fed could easily fix with an old fashion tool, temporary open market operations, even if the Fed got off to a slow start doing so last week. 

There have been great Twitter discussions and explanations of this repo market stress, including ones from Nathan Tankus, Bauhinia Capital, Guy LeBas, and George Selgin. There are also have been many good pieces from journalists and think tanks. Here, I want to echo a few of their points and speak to where I hope this experience takes the Fed's operating system in the long run.  Let's begin with what happened.

Stumbling Back into a Corridor System
The standard explanation for what happened has two parts. First, the Fed's quantitative tightening (QT) put the U.S. banking system close to the point of reserve scarcity given the new post-crisis regulations. Second, the Treasury recently auctioned off new securities and at the same time collected corporate tax receipts. These two developments further reduced bank reserves and pushed the banking system back into a reserve-scarce environment. Put differently, the Fed unintentionally stumbled from its floor operating system of the past decade back into a corridor operating system, the framework it intentionally left in late 2008.  

This story can be illustrated using a simple supply and demand model of bank reserves. Using this framework, the figure below on the left shows what the Fed's operating system looked like before the fall of 2008.  This was a simpler time when the Fed kept minimal reserves and banks traded for them on the interbank market. The Fed would conduct open market operations (OMOs) to adjust the supply of bank reserves so that a particular interest rate target was hit. Graphically, this meant moving the red line (reserve supply) along the downward slopping part of the blue line (reserve demand). Changes in the supply of reserves directly influenced the interest rate target. Monetary policy and money were directly linked.

This changed in late 2008 when the Fed expanded the supply of reserve such that the reserves were now on the flat (perfectly elastic) portion of the demand curve. The Fed's floor operating system had emerged. This can be seen below in the figure on the right.  Here, the quantity of reserves could change with no effect on the interest rate target. Monetary policy was now divorced from money. The Fed, in other words, could adjust its balance sheet independent of the stance of monetary policy. This was the appeal of the floor system. 

The Fed's floor system got further refined after the crisis with changes in reserve demand brought about by Dodd-Frank and Basel III. New regulations likes the leverage coverage ratio (LCR), supplemental liquidity ratio (SLR), and resolution planning for global systematically important banks (G-SIBs) increased demand for reserves. This can be seen below with the shifting out of the reserve demand curve (blue line) in the figure on the left. 

Finally, the Fed reduced bank reserves via the shrinking of its balance sheet or QT from 2017 to 2019. The goal was to keep the banking system on the flat part of the reserve demand curve, but with the lowest level of reserves possible. No one knew for sure where that would be, but if it were passed it would be made evident by overnight interest rates rising up. That is exactly what happened this past week with the Treasury sales and collection of corporate tax receipts and can be seen below in the figure to the right. The Fed's temporary OMO's is an attempt by the Fed to offset these recent Treasury actions and push the banking system back onto the flat portion of the demand curve. 

What is the Fed Doing Now?
The Fed is now engaging in temporary open market operations (OMOs) via overnight and term repos to bring the repo rate in line with its interest rate target. Some are calling this a bailout of the banking system, but these responses are what normally happens in a corridor system. It is what the Bank of Canada currently does with its corridor system and it is what the Fed did from the 1920s up until 2008. This is not a bailout, but a temporary provision of liquidity to the banking system. 

The figure below shows that $75 billion of overnight repos this past week are relatively small as a percent of the Fed's total assets. At 1.95% of total assets, they are small compared to the temporary OMOs in the decades leading up to the crisis. If we add in the additional $90 billion in term repo to be disbursed this week, it still is about where the temporary OMOs typically were as a percent of total assets prior to 2008. 

The most striking part of the figure, if anything, is the absence of temporary OMOs over the past decade under the floor system. This absence of repo activity by the Fed apparently has led many observers to forget the Fed engaged in this type of activity for most of its history. Maybe it also led to some repo atrophy at the New York Fed as discussed below. 

Why Did the Fed Stumble? 
So, why did the Fed not see this Treasury-related reduction in bank reserves coming? It should not have been a surprise and the fact that it was has some questioning the competence of the New York Fed. Some are even wondering if the absence of Simon Potter, the former head of the New York Fed's Market Group, was a reason for the misstep. 

This criticism seems off to me. Yes, Simon Potter's presence would have been nice this past week, but the Market Group is more than one person. There are other talented people in this group who are also well informed. I also find it a reach to blame the market inexperience of New York Fed President John Williams. This too puts too much weight on one person and ignores the collective wisdom of the New York Fed staff. 

I think a simpler explanation is that a floor system is just hard to run in the United States. First, no one knew for sure when the reserve scarcity point would hit because no one knew exactly how the regulations would manifest themselves in each bank. Even if they did know this regulatory part, they still would not know for certain if they had hit reserve scarcity until overnight rates shot up. 

Second, the New York Fed's slow response to the repo market stress might itself also be the fault of the floor system. The floor system has meant no need for temporary OMOs for almost a decade. Maybe the New York Fed was simply out of practice--no repo muscle memory--and caught off guard. 

Third, as noted by George Selgin, the floor system encourages the Treasury to use its TGA balance at the Fed. For a fixed Fed balance sheet, this growth in the TGA requires a reduction in bank reserves. If the Treasury had instead used the TT&L accounts this could have been avoided. But the floor system discourages their use. Maybe the Fed's choice of a floor system needs a rethink. 

Longterm Solution for the Fed
This repo market bump in the road should encourage the Fed to rethink the longterm future of its operating system. The floor system was supposed to bring greater interest rate control to the Fed, but this experience suggest that this is not the case. Moreover, as I noted in a previous post, the Bank of Canada appears to have better interest rate control with its symmetric corridor system. This, in my view, should be the longrun destination for the Fed's operating system. 

As George Selgin notes, the Fed's adoption of a standing repo facility could be a step in that direction.  There will also need to be tweaks to the regulatory front as well. The sooner we start this journey the better. 

The Repo Man Cometh
Some commentators worry these funding problems will continue to plague the Fed's floor system in the future given the Fed's desire to keep its balance sheet as small as possible while still being on the flat portion of the reserve demand curve. Since it is hard to know exactly where that sweet spot is on the curve, I agree that we may see strains again in the repo market. Eventually, this discomfort should move us toward adopting the standing repo facility, but until that time get used to seeing more temporary OMOs and Bill Dudley explainers. The Fed's repo man will come again.

P.S. There was some discussion as to who was holding the remaining reserves. This chart partly answers that question.

Source: FRED Data


  1. Great post David! I knew I could count on you for a lucid explanation of what happened.

    A follow up question... I've always (naively) wondered why the Fed bothers with repo. Why not just buy and sell treasuries as needed to achieve the targeted FF rate? What does this agreement to repurchase at a certain date allow the Fed to do that it can't do with standard open market operations?

    1. The SOMA account (Fed's holdings of Treasuries) will probably be expanded. The Fed did not do this last week since the Fed made a big deal of contracting its balance sheet.

      On another note, reserves available in banking system have also been reduced in the last year. Reverse repos have increased by $100bn. Reverse repos are for foreign central banks (BOJ is largest) who accumulate reserves in their Fed account and then lend them back to the Fed. Currency in circulation has also increased by $100bn.

  2. Very helpful post.

    "...the Bank of Canada appears to have better interest rate control with its symmetric corridor system. This, in my view, should be the longrun destination for the Fed's operating system."

    Yes, but don't forget that the Bank of Canada also had better interest rate control than the Fed when it ran its floor system from 2009-10. (See the chart in the your earlier post).

    So sure, Canada is a good example of how to run a corridor system. But also a good example of how to run a floor. To implement a good Canadian-style floor system, the Fed should start by mimicking the Bank of Canada and paying interest to *all* reserve holders, i.e. GSEs. That way the leaky floor problem would be fixed.

    1. "To implement a good Canadian-style floor system, the Fed should start by mimicking the Bank of Canada and paying interest to *all* reserve holders, i.e. GSEs."

      bingo ... along with foreign bank regulatory arbitrage within the Fed system ...

      more might be written about how this aspect affects the point at which nominally excess system reserves can become prematurely binding on rates in the US floor system

    2. JP, yep, Canada is a great example of the floor system should work: temporary and only in the liquidity crunch times. :)

    3. Not sure if you've read some of my previous posts, David, but I think the BoC should permanently adopt the floor. With corridor systems, liquidity is kept artificially scarce and supplied at above marginal cost. This forces participants to incur wasteful shoe leather costs. With a floor, liquidity is plentiful and produced at its marginal cost. It does away with the shoe leather costs.

    4. JP:
      Oh, I know your views but firmly disagree with them. There are many more margins of comparison than just the one you listed when it comes to evaluating operating systems. Ulrich Bindseil, for example, list nine different criteria in his KC Fed Symposium paper and comes down on the side of a leaner balance sheet. And he is a leading authority on this topic.

      Regarding Canada, why would you want to change an operating system that is currently working so well? Even if your one criteria were to dominate all the other, the success of the BoC operating system in stabilizing overnight rates around the target does not strike me as something one would want to give up on so easily.

    5. JP: one more thing. I think you are getting at the Friedman rule in your argument. George Selgin provides a good discussion of this in his book. See especially his Canzoneri cite on p. 29.

  3. Many thanks for your timely and helpful post.

    Tomo Nakamaru

  4. Why is the Fed dealing with two segmented markets, first one then the other? No explanation for this and it seems to be a fundamental flaw. Currency banking has to rely on a single representative market for money demand else the monetary zone is bifurcated, a fundamental error. How did we do a fundamental error for ten years, anyone?

  5. There still seems to be a huge amount of excess reserves in the system relative to the pre-recession baseline. I would be surprised if new regulations increased the cash demands by enough to soak up the excess. Is there perhaps a micro effect at work here? Some banks still in the area of very great excess reserves, but others holding only what is required?
    Anyway, very informative and well written post.

    1. Yes, there is a macro effect, the primary dealership program is not a complete market, it is a value added market charges with smoothing out the liquidity demands of treasury. That makes the excess reserves and seigniorage a hedged variable, they are generally calculated.

      This is the second market the fed tends to, and it is a loop in the banking system and causes the Fed to wind back to the overnight and price the loop.

      Look in Econ 1, under supply and demand and feel free to count a corridor between supply and demand. That corridor is real, it exists and cannot be theorized away. Eventually any currency banker is forced to do their job, act as market maker between supply and demand for liquidity. Government finance cannot be a separate market for the Fed.

  6. Economists don't know a debit from a credit. Banks do not loan out existing deposits. They create deposits whenever they lend/invest with the nonbank public. The nonbank public includes everybody except the member and Reserve banks.

    Repos are liquid assets, they are not money. The markets cannot simultaneously convert a bunch of liquid assets back into money. Liquid assets are past evidence of money (monetary savings) already spent/invested ("activated") - a velocity factor.

    It is not just the put and take between the Treasury's General Fund Account and the payment's System, that is disruptive to interest rates. It is also the idling of savings within the banking System that impacts the supply of loanable funds.

    I.e., all commercial bank-held savings are frozen, lost to both consumption and investment, indeed to any type of payment or expenditure. The only way to activate these savings, put them back to work, is for their owners to invest/spend either directly or indirectly outside of the payment's System, e.g., buy T-Bills.

  7. George Selgin is right about "better use of the TT&L program". The TBAC: “In the past the Treasury did impact the money stock figure by allowing commercial banks to hold Treasury funds in Treasury Tax and Loan accounts, but since January 3, 2012, funds in these accounts must be transferred to the Fed by the close of business on the day they are received so the end-of-day balance is always zero”.

    This drains liquidity (and the money stock, if not offset by an increase in Federal Reserve Bank credit), as Treasury issuance increases (restocking into the Treasury's General Fund accounts).

    Prior to this change, under the Treasury Tax and Loan (TT&L) program, tax payments by individuals and businesses were deposited into accounts at depository institutions, rather than directly to the Treasury's accounts at the Federal Reserve. So there was no impact to reserve balances.

  8. This comment has been removed by the author.

  9. A "Standing Repo Facility" works because it avoids the put and take of endogenous money. Collateral swaps at par with exogenous money (Central Bank deposits) is unit of account for unit of account.

  10. It is a specious claim to pontificate that monetary policy is divorced from money.

    Legal reserves (complicit reserves), are driven by payments (per Dr. Richard G. Anderson, the world’s leading guru on Central Bank deposits). Legal reserves are dependent upon “Total Checkable Deposits” 30 days prior (lagged member bank reserve
    requirements). Upwards of 95 percent of all demand drafts clear through this domestic deposit classification.

    The distributive lag effect of monetary flows, volume times transaction’s velocity, determines AD. The distributed lag effects of money flows have been unswerving mathematical constants for > 100 + years.

    Aggregate monetary purchasing power, AD, in turn is unbeknownst predetermined by FOMC directives.

    I discovered this 40 years ago. I predicted both "Black Swans", the flash crash in stocks 5/6/2010 and the flash crash in bonds 10/15/15.

    The money stock (and therefore interest rates), can never be properly managed by any attempt to control the cost of credit.

  11. What do you think of my "idea" that this repo situation is a new type of crisis?

  12. Fantastic post.

    Really enjoyed reading it and it held my attention all the way through! Keep it up.

    Read my Latest Post