Sunday, February 5, 2012

The Idiot's Guide to the Federal Reserve Interdistrict Settlement Account

The argument about ECB Target2 imbalances, an argument which began in mid 2011 with this article, continues to be reignited in various arenas. One component of the argument is the comparison of Target2 to the Federal Reserve’s mechanism for settling inter-Fed settlement imbalances; the Interdistrict Settlement Account.

Because the various debaters often have such diverging views on how the Interdistrict accounts are settled, I’m donating this post to the blogosphere with the goal of ensuring that the debate doesn't founder on faulty comparisons of Target2 to the Interdistrict Settlement Account. As such, I'm going to talk a bit about the history of Interdistrict settlement, how the process has changed over time, and how it works now. Once that's done I'll bring the discussion back to Target2 in order to pin down the comparison in a more robust way. I don't claim to know everything about these mechanisms; I've just spent a few weeks educating myself, so if I am wrong on any aspect leave me a comment.

In first learning about these two mechanisms I had a very meta reaction. By meta, I mean the same sort of reaction faced by anyone asking who watches the watchers?  In this particular case, the more appropriate question is “who clears for the clearers?” 

Another thing that makes the Target2 and Interdistrict settlement mechanisms so fascinating is that, within the monetary architecture, they are “core”.  Take out the core and the whole structure disintegrates.

Yet despite their centrality to the superstructure, nobody (except for a few cloistered central bankers) really knows much these mechanisms. It is only now that imbalances are cropping up that we the public find the inclination to catch up on how these arcane but vital systems function.

All sorts of commentators have already explained how Target2 works and the imbalances that have arisen. I won’t touch on these; just search Google for Target2. What follows is a discussion of the Federal Reserve Interdistrict Settlement Account.

Far older than the Target2 mechanism, the Interdistrict Settlement Account (ISA) dates back to 1915. Now is a good time to remind readers that the Federal Reserve System is not one unified institution, but an amalgamation of twelve regional Federal Reserve banks. One of the early goals of the system was to establish a unified national check clearing and settlement mechanism to replace the ad hoc system that had evolved among US banks till then. This new system would require coordination among the twelve district Feds, and the ISA was to be the central coordination device.

The distinction between clearing and settlement needs to be understood from the outset. Clearing is the process of tallying up all debts and credits, netting them, and assigning each member of a clearinghouse arrangement a final amount owed or owing. Settlement is the actual process of settling the debts and credits assigned via the clearing process. This involves transferring the amount owed from debtor to creditor so that all outstanding balances are cancelled out and return to zero. Settlement media might include base money, gold, are any other asset stipulated by the clearinghouse authorities.

The process of clearing doesn't require immediate settlement. For instance, those participating in a clearing mechanism might go for many days building up clearing debts and credits amongst each other before these outstanding balances are finally settled. In the modern Fed’s case, settlement occurs after a year (more on this later). In some cases, as seems to be happen with the ECB’s Target2 mechanism, the final settling of clearing balances is on hold indefinitely, or at least until those participants in the clearing process decide that the system needs to be broken up. 

Within a US Federal Reserve district, all member private banks clear and settle on the books of their district Fed. Clearing and settlement are relatively easy to understand as long as the transacting private banks are within the same district. At the end of the period, the district Fed need only tally the amounts owed/owing, and this can be settled by the transfer of clearing balances (or reserves) from one bank's account at that particular district Fed to the other bank's account. 

Note that the ISA has no role to play in the above intradistrict transactions. But once private banks party to a transaction are in different districts, and we switch from intradistrict to interdistrict, the ISA begins to play a role. Say that a customer of bank A in the Atlanta Fed's district sends a $100 check to a customer of bank C in the Cleveland Fed district. The check is deposited by the customer, whose account at bank C is credited. Bank C sends the check on to the Cleveland Fed, and the Cleveland Fed credits bank C’s reserve account. The Cleveland Fed dispatches the cheque to the Atlanta Fed, which proceeds to debit the reserve account of bank A. The cheque is sent on to bank A which now debits the original check writer. The net result of this chain is that the customers of the banks have settled among each other, as have bank A and C. The final debt has been transferred upwards onto the books of the regional Federal Reserve banks. But this final debt has yet to be settled. In crediting bank C’s reserve account, the Cleveland Fed requires a balancing asset. This is an amount owing from the Atlanta Fed. In debiting bank A’s reserve account, the Atlanta Fed requires a balancing liability. This is an amount owed to the Cleveland Fed. Basically, the Atlanta Fed now owes the Cleveland Fed.

Over the course of a day, huge quantities of interdistrict payments are processed by the district Fed banks. The total quantity of the resulting credits and debts among Federal Reserve banks are cleared via the ISA. At the end of a business day, a district Federal Reserve bank is either a net debtor to the other district Feds via the ISA, or it is a creditor to them. A district bank’s respective position can be quickly gleaned by referring to its balance sheet.  If the entry “Interdistrict Settlement Account” falls in the liability section of the district bank's balance sheet, then it is a debtor, but if it can be found in the assets section, it is a creditor.

 In its first few decades, the ISA was known as the Gold Settlement Account. At the end of the day, each district Federal Reserve bank would telegraph to the Federal Reserve board the total amount due/owed from each other district. All district Feds were required to maintain a balance of $1 million in gold or gold certificates in the Gold Settlement Account. The account was held at the Treasury and administered by the Federal Reserve Board. Once clearing was completed and the net balances totaled, a district Fed that owed another had its gold account in the Gold Settlement Account reduced while the receiving district had their account increased. Thus the final leg of a check payment was completed with settlement in gold. 

Say there was a run from banks in a certain Fed district towards banks in another district. Ultimately, this would be reflected in increasing transfers of gold from that Fed's gold settlement account towards the settlement account of other Federal Reserve banks. Once the $1 million dollar mark was breached, the district Fed facing the run was required to top up its account back to the $1 million mark. In other words, the Gold Settlement Account did not allow for permanent overdrafts. Theoretically, this placed a major constraint on the district banks and the system as a whole, for once a Federal Reserve bank ran out of gold, it could no longer settle with the other district banks.  And once it could no longer settle, the private banks in that district would be unable to do business with other districts. The monetary union would simply collapse.
There were various fail safes that relaxed this constraint. Howard Hackley, for instance, in his 1973 book on the laws governing Federal Reserve lending, points out that Federal Reserve banks were allowed to rediscount amongst each other (see pdf version of Hackley here).  This happened on a voluntary basis several times between 1917 and late 1921, says Hackley. By discounting with a surplus district bank, a district bank that was facing outflows would enjoy a counterbalancing credit, ensuring that its balances in the Gold Settlement Account remained topped up.

The rediscounts needn’t be entirely voluntary. Section 11(b) of the Federal Reserve Act (a section which is no longer in existence) allowed the Federal Reserve Board to force district banks to rediscount on other district bank’s behalf. According to Hackley, this only occurred one time. In March 1933, as bank runs swept the nation, several districts were required  to rediscount with the New York Fed as this bank’s gold reserves  had fallen before their minimum statutory requirements.

In 1935, the Gold Reserve Account became known as the Inderdistrict Settlement Account, for which it is known as today.

The legal basis for the Federal Reserve’s clearing mechanism can be found in Section 16 of the Federal Reserve Act. Specifically, section 16(14) stipulates:

The Board of Governors of the Federal Reserve System shall make and promulgate from time to time regulations governing the transfer of funds and charges therefor among Federal reserve banks and their branches, and may at its discretion exercise the functions of a clearing house for such Federal reserve banks, or may designate a Federal reserve bank to exercise such functions, and may also require each such bank to exercise the functions of a clearing house for depository institutions.
Thus the Federal Reserve board was allowed to undertake the function of clearing house for the Federal Reserve banks. Furthermore, it was allowed to “make and promulgate” the rules governing these interdistrict payments, the significance of which seems to be that it could vary the system’s settlement dates and settlement media.

Given the Board’s broad powers to control the interdistrict clearing and settlement infrastructure, it is not a surprise that the ISA would eventually be changed. By the late 1960s, interdistrict payments were growing in leaps and bounds as the financial system grew and evolved, and as a result it became more common for the gold deposit accounts of the district Federal Reserve banks to be in temporary overdraft. This is first mentioned in the minutes of an August 1975 FOMC  meeting by the manager of open market operations, Peter Sternlight:
Mr. Sternlight observed that under existing procedures the Reserve Banks used their gold certificate holdings to settle daily clearings with one another. Because of the large volume of clearings now handled in the Interdistrict Settlement Fund, one or more Reserve Banks might on any one day have an inadequate balance of gold certificates to make settlement. The possibility that adverse clearings might eliminate a Reserve Bank's gold certificate holdings had made it increasingly difficult for Reserve Banks to earmark significant amounts of gold certificates as collateral for Federal Reserve notes.
In that meeting, a change to the clearing and settlement mechanism was proposed:
…it was recommended that the clearings be effected through the use of inter-office accounts among the Federal Reserve Banks that would be settled once each year by increasing or decreasing each Bank's holdings of securities. That approach would obviate the need for monthly reallocations of the System Open Market Account to equalize gold-to-note liability ratios. The new procedure would take effect at the beginning of May--a convenient starting point since the end of April marked a coincidence of a month end and an end-of-week statement date.
The Chairman indicated that the Board planned to act shortly on the related recommendation to discontinue the use of gold certificates as the medium for interdistrict settlements.
Thus began a new procedure of interdistrict clearing and settlement. Rather than require district banks to deposit gold in the ISA and transfer it daily so as to achieve final settlement, shifts in the holdings of the System Open Market Account (SOMA) became the settlement medium. Furthermore, these settlements were to occur on a yearly basis, not a daily basis. Thus, while the ISA would clear each day and allocate Federal Reserve banks amounts due and owing, it would only require once-a-year settlement of these respective balances.

SOMA refers to the portfolio of securities acquired by the Federal Reserve System via open market operations. Prior to 1935, Federal Reserve banks could initiate their own open market purchases and sales, but after changes to the Federal Reserve Act in 1935, the authority for these operations was limited to the Federal Open Market Committee (FOMC). Essentially, the district banks were obliged to buy and sell as the FOMC dictated. All districts were allocated an ownership interest in SOMA. SOMA has historically been comprised almost entirely of Federal government debt, although in recent years this has changed as QEI and QEII brought large quantities of agency debt and agency MBS into the portfolio.

To this day, the above method of managing the ISA continues. The Federal Reserve Financial Accounting Manual (pdf) goes into some depth on this on page 136-138, including the provision of an example.  An average is calculated for each district Fed's daily ISA clearing balance over the course of a year. Sometime in the first week of April, the Board allocates SOMA holdings to district Federal Reserve banks so as to settle this average yearly balance. Note that the end of period clearing balance isn't what is settled, but only the average yearly balance. Thus a bank which is owed $10 billion on April 2 settlement, but was only owed on average $6 billion in the twelve months prior, will receive $6 billion in SOMA allocations as settlement, not $10 billion.

[Note: Some elements of the accounting manual's example might confuse the reader. Specifically, it is confusing that each Federal Reserve bank must have the same gold-to-note ratio as the System as a whole. This example is easier to deal with if it is assumed that the gold-to-note ratio of the district Feds doesn’t change over the years. Therefore, each district Fed’s gold-to-note ratio is always equal to the system average. Proceeding from this assumption, all SOMA adjustments are caused directly by clearing imbalances in the ISA. You can ignore changes that stem from one district bank issuing currency at a faster rate than others, which is interesting, but unnecessarily complicates the example.]

This story would be less interesting if there wasn't a modern twist: the ISA seems to no longer be settling. First noted on the Money View, a quick visual inspection of interdistrict settlement balances would indicate that the New York Fed is running up a larger owing balance than what would be expected by the rules stipulated in the accounting manual.

Interdistrict data, available on FRED, goes back to late 2002 and is available on a weekly basis. Let's take a look at the 2011 settlement. From end of March 2010 to March 2011, the New York Fed’s average end-of-week interdistrict settlement balance due comes to about $147 billion. That's a large amount owed to the FRBNY. The system seems to have settled somewhere between April 13, 2011, and April 20, 2011. This seems obvious given some of the large changes in district balances between those two dates. Because we only have weekly data (end of Wednesday), we don't have sufficient granularity to know exactly on what day settlement occurred.

The New York Fed's closing ISA balance was around $288 billion as of April 13. To settle, $147 billion worth of SOMA securities should have been transferred from owing district Federal Reserve banks to the NY Fed, and its interdistrict account should have fallen to around $140 billion ($288b minus $147b). But the next reporting day, April 20, shows that the account fell to just $224 billion, a shortfall of around $83 billion. Thus it seem that the NY Fed did not receive enough SOMA securities to settle average outstanding ISA balances as dictated by the accounting manual.

The Richmond and San Francisco Feds, on the other hand, seem to have disproportionally benefited. Much of the SOMA securities that should have been transferred by these two districts so as to balance the ISA were retained by them, and they were allowed to keep unusually large outstanding ISA debits. From my calculations, San Francisco's debit balance should have fallen by an additional $17.5 billion, and Richmond's by an additional $27.5 billion.

Now it could be the case that the poor resolution of the data – it is weekly, not daily – does not give enough information. Perhaps between April 13 and April 20 settlement was achieved as per the manual's requirements – say on April 15 – but large post-settlement payments from Richmond and San Francisco to New York between April 15 and April 20 caused the ISA accounts to again diverge, giving the semblance of a lack of settlement. 

The alternative explanation is that the Board has suspended the manual's requirements for the time being. San Francisco is the district bank for Wells Fargo, the US’s fourth largest bank, while Richmond serves Bank of America, its largest. New York is the hub for almost all the other top banks. Flows from Bank of America and Wells Fargo to, say Citigroup and JP Morgan in New York, might require such large transfers in SOMA security allocations that the district banks in Richmond and San Francisco are being temporarily exempt from full settlement. If this is the case, it is odd that the Fed wouldn't have issued a public notice to that effect.   

This brings up the question of a modern US monetary dissolution. We already touched on this briefly in our discussion of the Gold Settlement Account, and how interdistrict rediscounts softened the hard limits of gold settlement. Say that the Richmond Fed experiences such large outflows that, come April settlement, it lacks the SOMA securities necessary for it to settle its account. As a result, the other district banks cease accepting cheques and payments from banks located in the Richmond district, since they don't anticipate that the Richmond Fed will be capable of settling its account. Banks located in the Richmond district would now be effectively frozen out of the US payments system. Well, not entirely, since it would probably be the case that cheques and payments from the Richmond district passed at a discount to par. In effect, the Richmond dollar would be worth less than the US dollar.

The reason a scenario like this is unrealistic is because the Fed is required by law to maintain par clearing between Federal Reserve banks. According to Section 16(14) of the Federal Reserve Act:
Every Federal reserve bank shall receive on deposit at par from depository institutions or from Federal reserve banks checks and other items, including negotiable orders of withdrawal and share drafts and drafts drawn upon any of its depositors, and when remitted by a Federal reserve bank, checks and other items, including negotiable orders of withdrawal and share drafts and drafts drawn by any depositor in any other Federal reserve bank or depository institution upon funds to the credit of said depositor in said reserve bank or depository institution.
Par clearing means that all dollars are treated equally, no matter which private bank or district bank has created them. Given its authority in setting settlement and clearing laws as per Section 16(13), the Board can effect modifications to make the system more flexible so as to ensure the system operates at par. Perhaps in permitting San Francisco and Richmond to accumulate larger debts on New York than would otherwise be the case, the Board is rendering the system “softer” to prevent any possibility of non-par clearing.
Let's return briefly to the main point of contention in the Target2 debate: that the ECB imposes far less constraints on those central banks that make up the Eurosystem than the Federal Reserve Board imposes on the district banks that make up the Federal Reserve System. 

It is true that the Fed has historically called for some form of settlement, whether this be daily gold settlement or yearly settlement in SOMA securities, while the Target2 balances never settle. But Fed settlement is not required by law, it is only a tradition. The Fed can change the system’s clearing and settlement operations so as to ease constraints on district banks facing outflows. Indeed, this could be the case right now. Thus, when push comes to shove the Federal Reserve is not really that different from the ECB when it comes to interdistrict/intra-eurosystem clearing and settlement. The focus of both institutions is on ensuring the par value of all monetary instruments, not on disciplining individual members.

Postscript: A few misconceptions corrected.

In criticising Hans Werner Sinn, Willem Buiter says:
…the Interdistrict Settlement Account System of the Federal Reserve System in the US does not imply a substantially reduced scope for persistent imbalances in credit flows within the US currency area. While it is true that interdistrict imbalances need to be settled once a year with gold-backed securities or Treasury bills, individual district banks can purchase these securities in the open market and finance these purchases with base money creation. Therefore the US system does not effectively constrain interdistrict credit flows.
Buiter is right the ISA need not prevent persistent imbalances. But he claims that the reason for this is that district Federal Reserve banks can purchase securities in the open market. This isn't true because only the FOMC is permitted to purchase or sell securities (Read Section 12A of the Federal Reserve Act). It does so on the entire system's behalf. After a security is purchased, a district Fed is allocated a portion of the security, but it cannot buy said security outright on its own initiative. So in the end, what relaxes the constraints on interdistrict flows is not district open market operations, but the Board’s decision to change settlement procedures so as to protect the integrity of the system as a whole.

Buiter goes on to say that:
The Interdistrict Settlement Account must be settled once a year with gold-backed securities or Federal treasury bills. This would represent a constraint on inter-district credit flows only if the stock of Federal Treasury bills allocated to the individual Federal Reserve banks was exogenous. However, individual regional reserve banks can always buy Federal treasury bills from banks or other holders of the stuff in their own districts, financing this with an increase in base money.
The allocation is indeed exogenous because it is assigned by a formula set by the Board, and the FOMC conducts all purchases. Individual regional reserve banks can't buy assets, and even if they could, they must do so on the open market, and not from a narrow group of banks within a certain geographical area.

As for Peter Garber's Mechanics of Intra Euro Capital Flight, I do like his explanation, although he says that settlement of interdistrict balances is achieved by transfers of gold, whereas nowadays it is via reallocations of SOMA.

While that system does allow for Target-like balances resulting from the creation of outside money, i.e. money used for acquiring a net inflow of goods and/or assets from other districts, it has never experienced excessive flows comparable to those now taking place in the Eurozone as a result of the European sovereign debt crisis, wiping out the refinancing credit in its sub-districts and making the local central banks net borrowers of central bank money. As far as we know, Target-like problems never were an issue in US history.
As I’ve pointed out, the Federal Reserve System experienced significant imbalances between 1917 and 1921 and in 1933. In the former, a number of districts voluntarily rediscounted with the district facing outflows so as to soften the gold constraint. Forced rediscounting in 1933 saved the New York Fed from the same. Nowadays, it appears again as if the Federal Reserve system is relaxing requirements so at to allow some districts to continue as net borrowers of money for longer than they would otherwise be allowed. 

Sinn also points out that:
According to the official statements of the Federal Reserve, the assets whose ownership shares are transferred are gold certificates. Gold certificates are securities collateralized by gold, issued by the US Treasury, that bear the right to be exchanged for gold on demand. They are safe, marketable securities, which cannot be created by the District Fed itself. However, when a policy of credit easing was adopted during the crisis, this practice was abandoned and US Treasury securities were permitted for settling the balances, accepting in the end even mortgage-backed securities.
In actuality, the credit crisis did not result in the Federal Reserve switching from interdistrict settlement in gold to settlement in securities like US government bonds. This switch happened back in the 1970s.


  1. Great post--very useful, hard-to-get info.

    Here's a question: what are the implications of settlement using the SOMA for District bank cash flow? Don't they get interest on those holdings? So a surplus bank, after settlement, gets more interest allocated from the SOMA?

    Related question: Similarly, do the Target2 balances generate interest income for surplus NCBs, and interest expenses for deficit NCBs?

  2. Hi Tom.

    Yes, the resettlement of SOMA means that surplus banks get more assets allocated to them and therefore they earn increased interest income relative to debtor districts.

    This is similiar to the Target2 balances. Debtors have to pay the main refinancing rate set by the ECB.

    See Whitaker on this.

  3. Thank you for this useful comprehensive discussion of the Fed Reserve Bank settlement arrangements. I suppose the bottom line is the point made by Bijlsma and Lukkezen of Breugel, that all Reserve Banks are owned by the federal government, which I take to imply that the US government would always back a district Fed. Settlement would be assured if called for.

    In the eurosystem, while each national central banks is supposed to answer to the ECB, its credit is only as good as its government. Hence, at least in the case of Greece, settlement could not be achieved if it were called. So its Target2 debt stays on its books with no contract of understanding about repayment.

  4. Hi John, what a pleasure. I much enjoyed your Target2 paper.

    I did like the Bruegel piece. They made a small mistake in that the district Reserve banks are actually owned by their member private banks. So if the Federal Reserve system was to be dissolved and there existed significant imbalances between districts, final settlement might not be guaranteed. That's because a district Fed's credit is only as good as the credit of its member private banks. If these member were asked to re-capitalize the district bank to enable it to achieve settlement, but they were unable to do so, it would cause problems.

  5. Hi!

    It seems to me, that you miss in your, otherwise very interesting post, an important point:

    You write: "Say that the Richmond Fed experiences such large outflows that, come April settlement, it lacks the SOMA securities necessary for it to settle its account"

    It seems to me, that such a situation isn't a possible, because the combined value of the assets in SOMA is (roughly) equivalent to the combined central bank money in the system. Remember: the central bank money is created as a result of open market operations and at the same time the assets, which are bought bei the FED in these operations, are added to SOMA.
    So, the whole goal of the ISA-settlement, is, in my opinion, to restore the equivalence between the SOMA-assets and the liquidity on the level of the district banks. It is very alike to the gold-to-note you mentioned above, only that in this case, the ratio is 1. But if this system works properly, then it isn't possible that one district bank hasn't SOMA-assets to cover the outflowing liquidity, because both flows (SOMA und liquidity) have the same amount und are synchronized around the system

  6. Hi Anon,

    Let me see if we agree or not phrase by phrase.

    "the combined value of the assets in SOMA is (roughly) equivalent to the combined central bank money in the system"


    "Remember: the central bank money is created as a result of open market operations and at the same time the assets, which are bought bei the FED in these operations, are added to SOMA."


    The next step I'd add is that SOMA assets are managed by the Federal Reserve board on behalf of district Reserve banks. Each district is allocated some initial portion of those assets. In other words, SOMA assets are held on the balance sheet of district Reserve banks.

    Next, this allocation changes each year as imbalances are settled by transfers of SOMA assets.

    That's why I wrote that the Richmond Fed might run out of SOMA assets to settle. If imbalances continue for a long period of time, it could be that the Richmond Fed no longer has enough assets because they've already transferred their entire SOMA allocation to other district bank so as to achieve settlement in prior periods.

    1. Hi JP!

      "The next step I'd add is that SOMA assets are managed by the Federal Reserve board on behalf of district Reserve banks. Each district is allocated some initial portion of those assets. In other words, SOMA assets are held on the balance sheet of district Reserve banks."


      But it seems, that you still don't understand, what i'm trying to point out. Perhaps it's better, I try it with an example.

      For argument's sake, let us assume, that FED consists of only two district banks, say New York and Richmond, and the ISA balance is settled completely every year and not only the daily average.
      So, now, to our example:

      Let's assume, that at the begin of a year, the ISA is settled, and the value of the entire SOMA account ist 1000$. The amount of the central bank money (CB money) in the FED system is, as we've agreed already, also 1000$.
      Now, let's say, the the distribution of these assets/liabilities between the district banks is as follows:


      SOMA: 300 $
      CB money: 300 $
      ISA: 0$

      New York:

      SOMA: 700 $
      CB money: 700 $
      ISA 0$

      Now let's say, during the following year 50$ of the CB money
      flows from Richmond to New York. Then we have the following situation:


      SOMA: 300 $
      CB money: 250 $
      ISA: -50$

      New York:

      SOMA: 700 $
      CB money: 750 $
      ISA: 50$

      Now, the settlement occurs, and 50$ of the SOMA value is transferred from Richmond to New York. The new situation:


      SOMA: 250 $
      CB money: 250 $
      ISA: 0$

      New York:

      SOMA: 750 $
      CB money: 750 $
      ISA: 0$

      At each district bank and at any time the following is true:

      SOMA + ISA = CB money

      So, if, as you assume, Richmond runs out of SOMA at some point of time, then the following would be true (at Richmond):

      ISA = CB money.

      And this means, that, because CB Money is always >=0, ISA is >=0 at this at point of time too,so your scenario with SOMA=0 and ISA<0 (in words: Richmond has still a negative ISA saldo but no SOMA value for settlement) isn't mathematically possible.

      I hope, I was able to make my point clear, it isn't really easy to explain such complex matters in a foreign language :-)

  7. Ok! I see what you mean.

    I agree with you except for one thing. There is a timing issue here. Your formula shows that everything is always equal and immediately balances. But this process is actually a step by step one in which ISA balances can be created before SOMA assets are acquired.

    Say there is capital flight from Richmond to New York. Everyone empties their chequing accounts. New York banks receive these checks in the mail, credit their clients' accounts, then send these cheques on to the New York Fed. The NY Fed credits each banks's reserve account. It then sends the Richmond Fed all these cheques and totals them together, the sum total being Richmond's new ISA debit.

    The Richmond Fed receives these cheques from NY and proceeds to debit all of the Richmond banks' reserve accounts. Outflows were so large that this amount falls below zero. The Richmond Fed requires its member banks to provide collateral (SOMA assets) in order to bring their balances back up. But if outflows are so large, these banks may simply not have enough assets to rebalance their accounts at the Richmond Fed.

    Now it is time for ISA settlement. The Richmond Fed owes $x to the NY Fed in order to resettle payment outflows. But it has not yet received enough SOMA assets from its members in order to meet this settlement.

    You may notice that in this case I am now disagreeing with your statement that I earlier agreed with: "central bank money is created as a result of open market operations and at the same time the assets...are added to SOMA."

    I am providing an example in which central bank liabilities are created before assets are added to SOMA. But if these liabilites were always created at the exact same point in time, then I'd agree with you.

    1. Hi JP!

      Yes, you are right. If at any point of time we have central bank money in the system which isn't covered by the SOMA, then my argument doesn't work anymore. But in this case it would not only the Richmond's problem but of the FED as a whole, because FED has then more liabilities (central bank money) than assets (SOMA).
      I hope really, this your scenario never happens, because otherwise, facing the collapse of the entire banking system, the minor issue of unsettled ISA account, wouldn't be of any interest to anyone :-)
      So I'm glad, that we apparently agree now on all points.
      And at last, I'd like to congratulate you to the work you've done on the ISA matter. All the other participants in the target2 debate address only the H.W. Sinns supposed problem - target2 imbalances, but you are the only one to my knowledge, who adresses his supposed solution - the ISA mechanism.
      Actually, I myself am thinking about adressing this specific point in the german internet: You are aware perhaps, that the target2 debate in Germany continues unabating, with H.W Sinn riding a populist wave and using a much stronger language than he does in his english publications.

      Have a nice day (or perhaps night) :-)


  8. "But in this case it would not only the Richmond's problem but of the FED as a whole, because FED has then more liabilities (central bank money) than assets (SOMA)."

    One point that I didn't bring up in the above piece is that there is no "Fed as a whole". In Europe, the ECB is a stand-alone institution as are the NCBs. But in the US, there are only district banks (NCBs) but no parent institution. The ECB has its own balance sheet and capital - the Fed doesn't.

    This explains another large difference between the Target2 mechanism and the ISA. Target2 liabilities are ultimately claims on the ECB, and a claim on the ECB is ultimately a claim on each government according to the ECB's capital key.

    But ISA liabilities are not claims on the Fed - there is no Fed. They are bilateral claims on other district Reserve banks.

    So if you have large outflows from Richmond to New York, but Cleveland experiences neither net inflows or outflows, the ultimate failure of the Federal Reserve system will lie heaviest on Richmond and New York, not Cleveland. Cleveland will still have enough SOMA assets to "back" each Cleveland dollar. My point being that an outright failure of the system would not be evenly distributed across all districts.

    "And at last, I'd like to congratulate you to the work you've done on the ISA matter."


    "Actually, I myself am thinking about adressing this specific point in the german internet: You are aware perhaps, that the target2 debate in Germany continues unabating, with H.W Sinn riding a populist wave and using a much stronger language than he does in his english publications."

    I am aware that there is some politics involved but I don't read German so haven't been able to follow the debate. But if you are going to address this on the german internet feel free to take anything from here and translate.

  9. To J P Koning
    I have been struggling to really understand this issue by reading a lot of official documents for quite a while. Your explanation is by far the best I have read, and I would like to thank you very much for taking the trouble to publish it.

    I would like to ask for your opinion about the following thoughts.

    Professor Sinn seems to want the Target 2 system to work like the Gold Standard, or maybe the Bretton Woods system by imposing automatic contraction on economies in balance of payments difficulty. If (say) Spain is running a balance of payments deficit with (say) Germany, then the Bank of Spain would experience an increase in its target 2 debits, and the Bundesbank would experience an increase in credits (as has happened since 2008). Under Sinn's proposal, this would require the BoS to make a settlement each year to the Buba by transferring assets to the Buba. Presumably these assets would need to be raised from the Spanish government since the BoS has no other form of raising assets. Faced with a borrowing constraint, the Spanish government would have to raise taxes to make the payment. This would deflate the Spainish economy and eventually correct the bop imbalance.

    Sinn seems to think that this is in some ways a copy of the US system. But you have made it clear that the US system does not work in that way at all. If a District Bank has insufficient assets to make ISA payments at the year end to another District Bank, then the system would make many adjustments to ensure that the troubled District Bank would be given time to adjust without deflating its local district economy. There would be changes in SOMA allocations by the Board of Governors, or loans from strong to weak District Banks to ensure that the troubled District Bank would never have to take precipitous action to deflate its local economy. In other words, the US system would act to automatically finance any bop deficit by a member District, and would not force adjustments on the local economy like the Gold STandard was intended to do.

    Sinn has therefore entirely distorted the argument. If the EMU system were made an exact replica of the ISA system in the US, there would be settlements between member central banks each year, but this would be made possible by allocating securities from the ECB (acquired by open market operations or EMU-wide QE operations) to troubled central banks. If this were not enough to allow them to make settlements, then there would be loans arranged from the strong central banks to the weak, so that settlements could occur. In other words, the EMU would be a tue monetary union in which there could never by any pressure on exchange rates because all cross-border flows would be automatically financed by the operation of the ECB --- the euro would be guaranteed to be at par throughout the union. I wonder if Mr Draghi has any of this in mind when he talks about eliminating "convertibility" risks in the euro area?

  10. Hi Gavyn, thanks for stopping by. I'm always surprised that my most esoteric post gets the most traffic.

    I agree with many of your thoughts.

    Regarding Sinn's actual proposals, in this article he brings up the idea that the Spanish government, for instance, might issue senior tax-backed bonds or bonds secured by government property to settle Target2 balances:

    Whatever the case, Sinn's idea seems to be that somehow settlement must be forced and given the lack of resources of debtor national central banks (NCBs) the government of the debtor NCB will be responsible for settling imbalances.

    As you point out later in your comment, the analogy to the Fed doesn't quite fit. As Sinn rightly points out, the Fed's current operating policy is to have a yearly settlement between district banks. He seemed to have originally believed that this settlement was in gold. It is not. SOMA assets are transferred. I think Sinn now realizes this.

    But as recent events show (and events in the 1930s too), settlement isn't something required by the Federal Reserve Act. Settlement can be delayed, put off, and generally softened. The Federal Reserve Board can also require district Banks to rediscount (or lend) to each other in order to balance the system (this is the section 12A powers. In my original article I mistakenly said these powers had been removed). In any case, as you point out the effect of all of this is to create the conditions for the automatic financing of balance of payment deficits between district Reserve banks. For the Fed, maintaining the par value of the US dollar across all districts seems to be more important than maintaining settlement. In other words, yearly settlement will always be sacrificed by the Fed in order to maintain a universal value of the dollar.


    1. (con't)

      "If the EMU system were made an exact replica of the ISA system in the US, there would be settlements between member central banks each year, but this would be made possible by allocating securities from the ECB (acquired by open market operations or EMU-wide QE operations) to troubled central banks."

      A small quibble here: if the EU were to be a replica of the ISA, settlement would be made possible by transferring securities from troubled central banks to strong central banks. After all, those that are in debt need to transfer something (securities, gold, whatever) to their lenders in order to settle the system. [Tangentially, note that in the US, open market operations are allocated to the balance sheets of each district Reserve bank, and these holdings are in turn used for settlement. I believe (not 100% sure) that the ECB, on the other hand, maintains securities acquired in open market operations on its own balance sheet. So I'm not sure if it would be structurally possible for imbalances to be be settled by the ECB using transfers of securities acquired on the open market, since the national central banks (NCB) never hold these securities to begin with. The more likely way to balance the system might be to vary each NCB's ownership of the ECB, ie. if the BoS owes Buba, then it can decrease its ownership of the ECB by x% while Buba increases it's ownership by x%]

      Here's another way that Sinn's Fed analogy doesn't hold up. Say the district balances cannot be settled because one of the district Reserve banks lacks enough assets to settle. Say that for some reason the authorities require settlement rather than delaying it (the more likely case). In the US, this debtor district Reserve bank would never go to its respective State government(s) to get the resources that allow it to achieve settlement. Rather, it would go to the Federal government to get these resources, or it might require its member private banks (all of whom are shareholders) to recapitalize it. If Sinn wants to truly make the analogy to the US Fed, then Sinn's ECB needs to include a federal body (like the US Federal government) that would, in the end, provide for some sort of final settlement in the event of massive imbalances. More importantly, this same federal body would not only provide for final settlement in the case of imbalances, its very presence would help to prevent these imbalances from arising to begin with. A large quantity of ECB imbalances are due to bank runs arising from fears of euro break up, an eventuality a large federal body (like the US Federal government, or a Euro federal body) would go far to alleviate.

      "If this were not enough to allow them to make settlements, then there would be loans arranged from the strong central banks to the weak, so that settlements could occur."

      Yes, this would be like forced rediscounts by the Fed.

      "I wonder if Mr Draghi has any of this in mind when he talks about eliminating "convertibility" risks in the euro area?"

      I think you are right. This is exactly what he means when he talks about convertibility risk.

  11. JP

    Thanks for your very helpful reply.

    Only one riposte:

    The ECB's own central balance sheet is only about 8% of the balance sheet of the ESCB system's aggregate balance sheet, with the remainder being held by the national central banks. I think that the vast majority of open market operations result in securities being held by the NCBs, not the ECB itself.

    Best wishes


  12. Excellent! That little bit of information had been evading me - thanks.

    Note: for anyone stopping by, Gavyn wrote an excellent article on Sinn and the Target2 debate:

  13. JP-

    Great article, thanks. Can I clarify something please? Is the allocation of SOMA assets (debits) pro-rata to ISA balances (credits) pertinent because the purchase of SOMA assets is what created the base money in the first place?

    If so, then what would appear to be happening is that base money is being allocated amongst the reserve banks, not that anything is being settled.

    1. Hi Credit Me,

      Apologies if I do not entirely understand your question, but I'll do my best.

      Note that base money can be created by system open market purchases (additions to SOMA) AND also individual district Reserve banks engaging in discount window lending (including some of the fancier lending programs used over the years).

      In extreme events, a lot of base money may get created via discount window loans by one district Reserve bank, and when these deposits are transferred to another district, that second district Reserve bank is now responsible for those liabilities. To balance the transaction, it receives a credit on the the first Reserve bank (a positive ISA balance).

      The reallocation of SOMA assets settles these credits and debits.

      But in times of "financial peace" there are no overwhelming interdistrict movements - very little base money will be created by discount window lending. If so, then most base money will be created by SOMA purchases conducted by the New York Fed. These SOMA assets and the reserves created by the purchases are allocated to each respective Fed at the end of the day. In this case base money is being allocated amongst reserve banks, as are SOMA assets.

      Anyways, that's how I've decoded things. I could be wrong.

  14. Hi JP,
    thank you for that very helpful article.

    I'd like to come back to your example with Richmond Fed and New York Fed. As you've mentioned above, owing district Feds should have been transferred more money to NY Fed in 2011 as they probably did. And you've said that, when the FOMC/NY Fed conducts monetary policies via Open market operations, the new SOMA assets are allocated between the 12 district Feds.

    Do you know what the distribution coefficient is? (i.e. How many SOMA assets are allocated to Richmond Fed and all of the others when FOMC buys securities on the market? Is there a certain share?)
    My intention is, that if there wasn't something like a share in the Fed system, the SOMA assets could simply be allocated only to the debtor district Feds (Richmond and San Francisco). In April each year, they could hand over these assets to the creditor Feds when the settlement process of the ISA starts.

    Referring that thought to the stipulated settlement TARGET2 balances in Europe, this would mean that open market operations like OMT (outright monetary transactions) which are enacted by the ECB (and conducted by the National CBs) could be conducted only by the debitor Central Banks (Spain, Italy, Greece...). ISA-like settlement of the TARGET balances whouldn't pose a problem anymore, because they could easily transfer the assets back to the creditor CBs when setteling the balances.

    What is your opinion on that thought?

  15. Hi Sebastien,

    "Do you know what the distribution coefficient is?"

    This popped up in another post, see here:

    Basically, foreign assets are distributed daily according to a capital key... domestic assets are distributed daily according to each Reserve bank's participation in the ISA account as of the previous April and the equalization of gold certificates. It's a bit complicated... the link above might help.

    But the US capital key is not like the European one. Each NCB owns a bit of the overlying ECB, but the Reserve banks do not own a bit of the overlying Federal Reserve board. The FRB has no balance sheet of its own... it is just a governing board and nothing more than that. Each Reserve bank has its own capital base, and allocations are determined based on the relative sizes of each base.

    You asked about the OMT. Unfortunately I don't know much about how European open market operations work, who does the purchases, etc. But something to keep in mind would be that even if there were transfers of OMT assets among NCBs to settle Target2 balances, this wouldn't avoid the fact that at the end of the day, NCBs still have fixed exposure to each other via their ownership positions in the ECB (ie. the capital key). This feature is not something that characterizes the relationship between Reserve banks, and thus transfers among Reserve banks are much more "final" than transfers among NCBs ever would be.

  16. This comment has been removed by a blog administrator.

  17. I've just come across this really useful post while reading to reacquaint myself with the subject of TARGET2. It is a key contribution to the discussion. I was sceptical about both Sinn's and Buiter's ideas of how Fed inter-district settlement works since they did not give good references, and like Gavin Davies struggled to get a clear picture from the existing publications, but this post just about fills the gap. Well done JPK. I can't help but feel though that the Fed itself should be publishing explanations of all this.

    A couple of points I would make:

    I am inclined to agree with anonymous. The closed nature of the system should guarantee that any district Fed should always have sufficient SOMA assets to settle the ISA. If the SOMA assets are allocated formulaically between the districts regardless of their banks' participation in Fed OMO's, then bank reserves should be too. A district can only be expected to settle inter-district flows out of existing bank reserves on its books, and therefore must possess enough SOMA assets, or some other asset like gold certificates if the allocation of reserves and SOMA assets does diverge, to settle with the other districts if required. If any bank's payment outflows exceed its holding of reserves, it must obtain more reserves either by asset sales or by borrowing, either from other banks or from the Fed. If it has insufficient saleable assets or eligible collateral, then the bank must either seek a LOLR loan from the Fed against non-standard collateral, or it goes bust. And what applies to an individual bank applies to the collection of banks in the district. If, for example, there is a run on a particular district, this will result in a debit position for that district's Fed in the ISA, matched by an increased holding of SOMA assets. If inter-district settlement (or at least a move towards balance) subsequently occurs, that will involve reallocating an amount of SOMA assets equal to the decrease in that district Fed's ISA debit position.

    That brings me to my second point, which I tried to emphasise in my own post on TARGET2 ( ), and in comments on other posts, which I believe has still not sunk in. In my opinion, the real problem for the eurozone is not the inter-district balances, but the OMO collateral. If the eurozone was a genuine monetary union overseen by a genuinely nationally-disinterested monetary authority, many of the peripheral country banks would have run out of eligible collateral by about 2010 and been forced to contract or be taken over by stronger banks, presumably from the TARGET2 surplus countries. The trouble has been that the eurozone remains too national to accept that Darwinian logic of a monetary union, so the ECB repeatedly lowered the quality standard for eligible collateral to allow the peripheral country banks to survive. In particular, given the strong home bias of peripheral country bank asset portfolios, that meant continuing to accept peripheral country government bonds as eligible collateral from peripheral country banks. Clearly, the riskiness and correlation of such collateral with the fortunes of its owners makes it very poor security. I dare say that the Bundesbank would not consider taking possession of such OMO loan assets in settlement of its TARGET2 credit as much of an improvement in its position. In short, inter-district settlement or the lack of it is not an important issue in the USA because the USA is a genuine monetary union, and – as far as I know – the Fed is not under political pressure to accept inadequate collateral from banks in any particular district.

    1. Well if it isn't Rebel. Welcome. The econ blogosphere hasn't been the same since you stopped updating your blog.

      You pointed out that you agree with anon above. With the benefit of half-a-year of thinking about this, I can explain things better now.

      Reserves are created not just through open market purchases (SOMA) but also Federal Reserve loans, or discounts. SOMA is centralized, and all allocations of SOMA result in a formulaic apportionment of reserves to district Reserve banks, as you and anon point out.

      But Fed lending programs are decentralized. Each Reserve bank undertakes its own discounting program. In lending, these district Reserve banks create reserves locally which must be accepted globally, ie. by the System as a whole. In return for loans, district Reserve banks ask for collateral from the borrowing commercial banks in their district. I'll call this "local collateral".

      Which means that the the total quantity of System reserves is "backed" by a certain quantity of SOMA assets as well as a certain amount of "local" collateral.

      That explains why districts might not have sufficient SOMA to settle ISA debits. Say there is significant capital flight out of one Fed district. In settling its resulting ISA debit the district Fed uses up all its SOMA assets. At the same time, it lends to member banks, creating more reserves, which are immediately moved to other districts. There are no more SOMA assets the district Fed can use to settle its ISA debit. But it does have "local" collateral. But this local collateral is not the accepted settlement medium for ISA balances. So you have an ISA default of sorts.

      This hard limit might be relaxed by delaying ISA settlement or allowing "local collateral" to become a settlement medium. It seemed to me that in 2011 the Fed delayed ISA settlement in order to soften the System, but in conversations with people at the Fed they assure me they never delayed ISA settlement.

      "If the eurozone was a genuine monetary union overseen by a genuinely nationally-disinterested monetary authority, many of the peripheral country banks would have run out of eligible collateral by about 2010 and been forced to contract or be taken over by stronger banks"

      I couldn't agree more. If healthy foreign banks had taken over Greek banks back in 2010 and Spanish banks in 2011, then the Target2 problem would never have arisen. People would have been content to keep their deposits in German owned banks in Greece and Spain. In delaying bank resolution, the NCBs accepted what was probably bad collateral which, as you point out, the Bundesbank will be wary of accepting should the possibility of settlement be broached.

      In the US they have FDIC which quickly winds up insolvent banks and sells them off. This prevents district Reserve banks from perpetually lending to insolvent local banks and creating "bad" reserves that when transferred to other districts, might lead to an ISA default. Perhaps Europe needs an FDIC?

    2. Thanks for responding on such an old post, JPK.

      The trouble is, as you say, we are largely "thinking" through how this should work, rather than being informed by easily-found documented explanations - I mention that just in case anyone from the Fed reads this!

      What you say makes sense. I had not considered the possibility that some Fed lending operations (discount window lending only, I think, not repos - see here: ) are decentralised or that some loans are not part of the SOMA portfolio (again, I think that repo loans are included in the SOMA). In that case, I can see that, to the extent that bank reserves were obtained from the district Feds, there are reserves that are not matched by transferrable SOMA assets. Even so, the fact that the district Feds do hold good collateral against all the reserves provided to local banks does mean that, unlike in the eurozone, the safety of unsettled inter-district exposures is not in question. Hopefully, anonymous will have another look at this post some time.

      I agree that the eurozone (if not the EU) needs an FDIC. It will be a sign that the European monetary union has matured if a central authority can close a regional bank without raising concerns about national disadvantage and advantage. As we saw with the recent putative merger of BAE with EADS, though, we are long way from that degree of European commonality.

    3. No problem, Rebel. I'm always surprised at the amount of traffic this arcane post generates. I think that demonstrates that there's a need for clarity on the issue, and the Fed hasn't really been forthcoming. Lubik and Rhodes, two Richmond Fed researchers, put out a short note on the ISA, but what is needed is a much more extended explanation (see here (pdf))

      "discount window lending only, I think, not repos"

      Sounds right to me. Repos are considered purchases/sales, these go into SOMA.

  18. Serious people are reading your blog, JP!

  19. Hi JP- thanks for this helpful summary from 2012. I came across your article after noticing that the NY Fed's ISA increased rapidly in April and is over $500 billion as of 29 July 2020 I don't know the system well enough to know what has caused the increase or whether it's significant. I'd be very interested to know your thoughts on it.