Banks, bonds and deficits

Lots of you pointed out that in my last post the monetary effect of government bond purchases was unclear. Indeed, I had rather skated over it and thereby created confusion. I gave the impression that government can always sterilise spending (broad money creation) by issuing bonds. But this isn't quite true. It all depends on who buys them and how they finance their purchases.

So I thought I'd do a few scenarios to show what happens when different types of private sector actors buy and sell government bonds.

Firstly, let's look at primary issues. Government issues sufficient bonds to sterilise its entire deficit spending. This is normal fiscal behaviour.

Primary issue: bonds entirely bought by banks

I'm going to consider the banks in aggregate here, rather than as individual entities, and I'm going to assume that the same banks also intermediate government spending. As I noted in my previous post, in the absence of pre-funding, governments will run overdrafts at commercial banks when paying private sector agents. In practice, governments usually pre-fund spending commitments with bond issuance. So there are two versions of this scenario.

Spending precedes bond issuance: Government runs overdrafts at the banks. The proceeds of bond issuance pay off these overdrafts. But in this scenario, the banks that have granted the overdrafts also buy the bonds. This is an asset swap for the banks: it replaces loan assets (drawn overdraft facilities) with bonds. There is no change in the banks' liabilities. So M3 rose* when government drew on its overdrafts, as explained in the previous post, but it did not fall when the banks replaced loan assets with bonds. The government's deficit spending is NOT sterilised.

Bond issuance precedes spending: banks buy bonds from government instead of granting overdraft facilities. The proceeds from the bond sale go into the government's deposit accounts (which therefore have POSITIVE balances). Government then draws on that funding to meet spending commitments. This is exactly the same as lending to government and the accounting is identical. Banks create new deposits to the value of the bonds, which raises M3. When the government spends the money created, the liabilities move from bank to bank but there is no change in the aggregate liabilities of the banking sector. So M3 is created as a consequence of bank lending to government (bond purchases), not government spending.

In summary, a primary auction bought entirely by broker-dealers does not sterilise government deficit spending. It is simply an alternative version of bank lending to government, which raises M3.

However, there is a complication. In some versions of primary issuance via banks, banks pay the central bank for their purchases from their reserves, and the central bank reimburses the fiscal authority via its account at the central bank. This is a drain of M0 (the monetary base), not M3**: the commercial banks exchange reserves for government bonds, leaving their liabilities unchanged, and the transfer of reserves to the government's account takes them out of circulation. However, because private sector agents to whom government makes payment don't have central bank reserve accounts, government must still make those payments via commercial banks. In this mechanism it does so via the central bank, which credits the reserve accounts of the payees' banks: the banks then create deposits on behalf of the payees to the amount of the transferred reserves. Both M0 (reserves) and M3 (deposits) therefore rise. The combination of reserve-funded bond purchases and reserve-funded government spending leaves M0 unchanged, but banks still have to sell government bonds to non-banks to sterilise M3.

Primary issue: bonds bought entirely by non-banks: Again, we have two versions - bond issuance after spending, and bond issuance prior to spending.

But there is an additional complication. Non-bank asset purchases are always intermediated through banks, and non-banks may borrow from banks either to finance the purchase itself (unlikely, admittely) or to fund other spending as a consequence of bond purchases. So let's look at how this works.

Spending precedes bond issuance: Government has drawn overdrafts at banks. When non-banks buy the bonds, they pay from their own deposit accounts into government accounts, extinguishing the drawn overdrafts. If non-banks make these payments from existing funds (i.e. balances on their deposit accounts were positive prior to payment and either positive or zero afterwards), then the extinguishing of government overdrafts reduces M3 in exactly the same way as paying off any other bank loan does.

But if non-banks themselves borrow from banks, then the reduction in M3 from the extinguishing of government overdrafts is offset by the rise in M3 due to new lending to non-banks and there is no change in the broad money supply.

As M3 rises initially when the government draws its overdrafts (because banks create money), this means that debt-financed (leveraged) purchases of government bonds by non-banks do not sterilise government spending, but purchases from existing funds, say by cash-rich pensioners (Government pensioner bonds, for example) do.

Bond issuance precedes spending. As before, when non-banks buy the bonds, they pay from their own deposit accounts into government accounts, increasing the balances in those accounts. Government then has positive balances in its bank accounts and subsequent spending has no effect on M3. But whether M3 is raised by the non-bank bond purchases depends on how they are financed. If non-banks borrow from banks (including temporary overdraft facilities) to fund the purchases, then M3 rises due to new bank lending and government spending is not immediately sterilised, though it would be when the non-banks extinguish the loans or overdrafts. If non-banks fund the purchases from existing funds, it is simply an asset swap for the private sector, payment is intermediated through the banking sector without affecting the size of its aggregate balance sheet and there is no change in M3.

For government spending to be fully sterilised by bond issuance, therefore, bonds must be issued to NON-banks and paid for with existing funds. In practice, primary issuance is primarily bought by banks, so does not sterilise spending.

Central bank monetisation of deficit spending: If the government borrows directly from the central bank to fund deficit spending with no bond issuance, then both M0 and M3 rise ***. The central bank creates new reserves (M0), which initially go into the government's account at the central bank but are then disbursed to reserve accounts at payees' banks as the government spends. Alternatively, the central bank simply creates the reserves as the government spends, placing them in the appropriate reserve accounts and creating an equivalent loan asset (overdraft) for the government. The loan asset is simple balance sheet accounting between two parts of the government, which disappears in consolidated accounting, but the monetary base increase is real. As before, when the new reserves are distributed to bank reserve accounts by the central bank on behalf of government, the banks create deposits in payees' accounts, which increases M3.

Now let's look at the effect on broad money of trading in the secondary market. 

Secondary market effects
Trading in the secondary market does not change the quantity of bonds issued, unlike a primary auction. But it does affect the broad money supply. This is because actors in the secondary markets include both non-banks and banks, and bank trading activity changes the money supply: banks create money when they purchase assets and destroy it when they sell them. 

The chief benefit to a bank of buying bonds from government rather than lending directly is that the bonds are liquid. They can be sold, to other banks or to the general public. When banks do this, the broad money supply changes, as I explain in the following examples. To keep things simple I'm going to pretend that all trades are OTC and bilateral.

Bank buys government bonds from another bank: This is simply an exchange within the banking sector and there is no net change in assets & liabilities. The buying bank increases M3 when it makes the purchase, because it creates money to do so, but the selling bank decreases M3 because a loan obligation (the bond) has been extinguished. The balance sheet of the buying bank increases, that of the selling bank shrinks. There is no net change to the aggregate balance sheet of the banking sector and therefore no change in M3.

Non-bank buys government bonds from a bank (and pays for them from existing funds): The size of the selling bank's balance sheet is unchanged, since the government bond is replaced with an equal cash asset (an asset swap). The size of the non-bank's balance sheet is also unchanged, since a bond purchase from existing funds is also an asset swap. However, the liabilities of the bank from which the non-bank makes the payment reduce, as do its reserves. Overall, therefore, the balance sheet of the banking sector shrinks and M3 falls.

Non-bank buys government bonds from another non-bank: This trade is intermediated through banks. As far as the non-banks are concerned this is simply an asset swap on both sides and there is no change in the aggregate non-bank balance sheet. Importantly, if the purchase is made from existing funds (no leverage) there is no change to the aggregate balance sheet of the banking sector and therefore no change to M3. But if the purchaser borrows from the bank to fund the purchase, even temporarily, M3 rises and remains elevated until the temporary loan is extinguished.

Summary and conclusion
When banks purchase primary bond issues from government, government spending is not sterilised. Sterilisation of the monetary effects of government spending happens when banks sell on government bonds to non-banks, unless non-banks borrow from banks to fund the purchases or cover consequent shortfalls.

In practice, government spending does increase the broad money supply even if it is fully covered by bond issuance. Bond issuance cannot wholly sterilise it, because banks don't sell their entire holdings of government bonds - they retain a substantial proportion as collateral for repo funding. Opportunity costs for non-banks, too, may lead to incomplete sterilisation due to higher bank lending to non-banks. So it is probably fair to say that bond issuance is not a very effective way of sterilising the monetary effects of government deficit spending. Taxation is the only fully effective means, and even then only if it is efficient, which most tax systems are not. Over time, therefore, we would expect persistent government deficit spending to increase the broad money supply. 

Whether or not the increase in broad money arising from incomplete sterilisation of government deficit spending - or even outright monetisation of government deficit spending by the central bank - is inflationary in my view depends on how the money is distributed and the purpose for which it is used. I think it is fallacious to assume that increasing the quantity of broad money, whether through central bank money creation, government deficit spending or simply through higher bank lending, necessarily raises consumer price inflation.  Looking at the quantity theory of money equation:


with M defined in this case as M3, we can see that an increase in P (inflation) is only one of the possible consequences of a rise in M. There could be an increase in output (Y) with no effect on inflation, or there could be no effect at all because people just sit on the money (V falls). Or, not shown in this equation, asset prices could rise.

I don't propose to discuss this any further here. Better folk than me have had endless debates about it. At JP Koning's suggestion, I refer those who are interested to Tobin's theory of reflux, Yeager's refutation of it, and more recently the extended argument between David Glasner, Nick Rowe and many others. The causes of inflation are endlessly puzzling and supply-of-money theories are hopelessly inadequate. As ever, in monetary economics, it just isn't that simple.

Related reading:

The fiscal theory of monetary expansion
Printing clever weird stuff - aka money - Principles & Interest
When wonks get things wrong - Pieria
Inflation is always and everywhere a political phenomenon - Pieria
Money creation in the modern economy - Bank of England (pdf)

* Throughout this post I've used M3 to mean broad money. US readers should mentally change this to M2 as the Fed no longer produces M3 figures. All the money creation I discuss in this post would form part of M2 as well as M3 - indeed much of it would be M1.

** The use of reserves to buy government bonds in a primary issue was omitted from my previous post. Thanks to all those who alerted me to this alternative.

*** I got this wrong in my last post:

"An alternative to bond issuance for draining M3 is borrowing from the central bank. Yes, I mean it. If a government borrows from the central bank to pay down a commercial bank overdraft, M3 falls. The monetary base M0 rises as a consequence of this borrowing, but this does not affect the economy unless commercial banks increase their lending to households & businesses or the central bank prints more physical currency."

This is not correct and I have removed it from that post. Thanks to Tim Young for explaining it. 


  1. Banks buys government bonds from another bank… couldn't this simply be a matter of an asset swap where seller bank receives reserves and buyer bank receives the bond. No M3 involvement at all; simply a debiting and crediting central bank accounts.

    1. In theory, yes, But settlement is typically T+2 these days - payment is made and bond delivered (assuming it is delivery-versus-payment) two days after the trade is agreed. The trade is recorded on the books when it is made, settlement entries two days later. Therefore the banks must record not only changes in assets but also temporary changes in liabilities.

  2. Ok, so i assume settlement through Euroclear Finland would not involve such lag? From their disclosure report:

    "HEXClear and RM are both delivery versus payment (DVP) systems. Participants cannot dispose freely of securities before fund finality. The transfers of both securities and funds are executed simultaneously. Momentary (split second) discrepancies in the payment and securities transfer system processes will not allow any disposal of funds or securities until both the cash and the securities leg have been settled."

    1. No, this would still be T+2. DVP settlement 2 days after the trade is agreed involves simultaneous transfer of funds and securities.

  3. Frances, I'm still having trouble with the accounting logic in such bank-to-bank transaction. Assuming the seller bank originally bought the bond from the government by crediting the governments account at the central bank – thus only changing the composition of the bank's asset side on its balance sheet, reserves for a bond.

    Hence, when the same bank now agrees to sell the bond to another bank, I can't see a direct reason why that would affect seller bank's liability side in terms of M3 (quote FC: "…but the selling bank decreases M3 because a loan obligation (the bond) has been extinguished"), even with a T+2 arrangement.

    Sorry for being nitpicky here!

    1. Johan,

      One of the things that accounting has to show is the relationship between transacting entities. If each bank involved in the transaction simply swapped assets around, no transaction BETWEEN the banks would be recorded. Therefore the banks account for a purchase/sale pair between them in exactly the same way as they would purchases and sales with non-bank customers. The liability effect (M3) nets out, but it still has to be recorded.

      Also my point about settlement accounting stands. Failure to record the liability side of the transaction breaches double entry accounting rules when settlement is made.

    2. This comment has been removed by the author.

    3. Yes, I certainly agree that the relationship between transacting entities ought to be recorded. However, I'm rather curious as to what kind of accounting record would suffice. Seller bank's balance sheet before the transaction is the following (overly simplistic, but follows from my previous reply): ASSETS = 100 government bonds | LIABILITIES = 100 customer's deposits. Buyer bank's balance sheet: ASSETS = 100 reserves | LIABILITIES = 100 customer deposits.

      So what escapes me is how the accounting entries would continue now that there's an agreement to sell/purchase the bonds. How can the seller bank decrease M3 in this situation; by what kind of accounting entries? I know there's a commitment to pay for the bond and likewise a commitment to deliver the bonds… until the transaction is finalized. But if buyer bank's accounting entry results in an increase on the liability side on the ledger, then what would seller bank's accounting entry be – especially since both start out with 100 in customer deposit liabilities each (before the transaction)?

      There must be something missing here. Thank you for your patience thus far.

    4. Johan,

      The accounting entries are exactly the same as for the first leg of a repo.

      When the deal is made, but before settlement:

      Buying bank

      DR Securities (asset)
      CR selling bank deposit account (liability - NEW DEPOSIT)

      Selling bank

      CR securities (asset)
      DR buying bank deposit account (DEPOSIT DESTRUCTION)

      At settlement:

      Buying bank

      DR selling bank deposit account
      CR reserves

      Selling bank

      CR buying bank deposit account
      DR reserves


      DR buying bank reserve account
      CR selling bank reserve account

      The new deposit created by the buying bank when it does the deal extinguishes its debt obligation with the selling bank when the deal is settled. Reserve movement does not happen until settlement. So M3 is destroyed by the selling bank and created by the buying bank and the reserve movement restores the position to what it was before the deal. The net M3 position is unchanged. But that doesn't mean you can fail to make the accounting entries. To do so would breach double entry accounting rules at both banks when the deal is made (ownership of securities is transferred on agreement not on settlement).

    5. Thank you Frances for the clarification! Found the missing piece in that both banks must have accounts with each other (or an arrangement to that effect). All makes sense now.

  4. > Frances

    Your scenario of spending preceding bond issuance does not sound right. Overdrafts to account for the typical size of government deficit spending does not sound like a plausible description of the deficit spending process.

    1. It may not be how governments choose to do their deficit spending, but it is entirely possible for them to do it that way. And as I explained in the previous post, in the absence of pre-funding it is de facto how government spending must be done. You assume that there must be pre-funding (bond issuance). That is actually not essential and in the past has not always been done - hence some governments (including the Bank of England) running overdraft accounts at their central banks. In the even more distant past before central banks were invented, governments borrowed directly from banks.

    2. Its possible but unlikely, as why choose to borrow at overdraft rates of around 10% when bonds sold to the same banks yield 0.6%.
      Financing with the Bond market also has the advantage that the government has people buying their outstanding bonds as apposed to the government repaying the overdraft themselves.
      In the absence of pre-funding overdrafts are not the only way it must be done. The government could keep its general account at the Bank of England "The Consolidated Fund of the UK" In positive balance, and fund spending from that. I suspect that this is in fact what they do.

    3. If a government chose to fund itself with bank loans (overdrafts) rather than bond issuance the interest rate would of course be the same. It would only be higher - considerably so - for governments that were for some reason shut out of markets. Unless of course the government borrowed from a captive bank, in which case we can assume that the interest rate would be at or near zero whether or not bond issuance was an alternative.

      2) The government can only keep a positive balance in its account by either pre-funding with bonds as we have discussed or by running a persistent budget surplus.

    4. The same thing could be achieved by encouraging banks to buy bonds by a tax incentive.

      How does an overdraft appear in the accounting of a bank's assets. Does it have the same status and value that a Bond has.

    5. Yes, banks can be encouraged to buy bonds with tax incentives. They can be encouraged to lend with tax incentives, too. It's the same thing. Buying bonds and lending are the SAME THING where banks are concerned.

      A drawn overdraft is simply a loan of indeterminate length. It has exactly the same status and value as a callable perpetual bond.

    6. Well that's my point they are the almost same thing so whats the point of having multitudes of departmental overdrafts instead of one bond issuance, and I doubt an overdraft is valued the same as a bond . A callable bond can be re sold to a third party but an loan of intermediate length cannot.

    7. In fact government departments DO have bank credit lines. This was pointed out to you by a commenter on the previous post. Why are we going over this again here?

      Loans can be securitised and used as collateral for reop funding. And if government was routinely using bank loans rather than bonds, the loans themselves would be accepted. Arbitrage would keep the interest rates the same.

      But your remarks are off topic anyway. This post is about the monetary effects of bond issuance. Do you have anything useful to contribute on this subject?

    8. This comment has been removed by the author.

    9. Yes of course I do - how can you hypothesize about whether or not bond issuance has a sterilizing effect when there is nothing to sterilize. Tim Young made a similar point on the earlier post. Government departments may have credit lines , but they do not amount to the deficit

    10. Tim understood the point I was making though. You have not. You are stuck in the belief that governments MUST pre-fund spending with bond issuance. That is gold standard thinking. It is no more true than the belief that banks must receive deposits before they can lend. Governments can choose to pre-fund spending with bond issuance, and indeed most do - though that has not always been the case. But they don't have to. The "null hypothesis" is that they do not.

    11. This comment has been removed by the author.

  5. "However, there is a complication."

    At this juncture,you seem to move into a 'two bank' description. I think you depart from the aggregate bank description because I think M0 is an intra-bank measure that would only apply to a 'two bank' internal division.

    Because it seems to me that changes in M0 are strictly intra-bank, it should not be automatic that M0 would increase when Government spent into the private sector.

    "Central bank monetisation of deficit spending:" Again I think that the discussion is shifting into a two bank description. The same distinction would apply.

    Not a simple post. A well thought out post, thanks.

    1. Roger,

      I'm afraid this is not correct. The overall level of M0 does not change as a result of interbank transactions. It changes when the central bank increases or decreases its liabilities. The liabilities of the central bank are bank reserves and physical currency. If the central bank increases either of these, M0 rises. So if the central bank funds government deficit spending by crediting commercial bank reserve accounts or by physically printing currency, M0 must increase, regardless of what the commercial banks are doing.

      Central bank direct monetisation of deficit spending is QE without the bond purchases.

    2. Yes, I think I made a mistake by equating M0 and bank reserves. The two are nowhere near the same. It would be best if I just withdrew my comment and, perhaps, started over again.

      In my own mind, I keep things straight by assuming that Government, when borrowing directly from the Central Bank, issued a bond to the Central Bank. That assumption places the Central Bank into the aggregate banking system. Then, if we look at a two bank system (Private Bank and Central Bank), the only difference between borrowing from either is that the Private Bank must somehow first 'earn' the money it loans to Government but the Central Bank simply 'prints' the money.

      I remain fuzzy about why you discussed M0 in this post. Perhaps my assumption over-simplifies the situation.

    3. Hi Roger,

      Private banks don't "earn" the money they lend.They create it. Just as the central bank does. Can I suggest you read the Bank of England link in "Related Reading", which explains how banks create money when they lend?

      I discussed M0 because the role of central banks in financing government deficit spending - directly or indirectly - was raised in the comments to the previous post, to which this post is a response.

  6. Of course you are correct. The banks create money identical to the central bank. Only private entities (non-bank) must first earn money before purchasing Government Bonds (allowing the exception of borrowed money that you mention).

    On the other hand, if we look at the loan documents, both bank loan documents and Government Bonds, we see that in every case the money was spent into the economy and thus has been 'earned' by some entity. The fact that bank deposits are much less than total of bank loans and Government Bonds reflects the bifurcation of money into two monetary versions when Government Bonds are purchased.

    I will re-read the Bank of England report more carefully.

    A great discussion here.

  7. This comment has been removed by the author.

  8. I can't wait when starts QE in Europe. Tomorrow we will see what Mario Draghi do

  9. What happens if the central bank buys an existing bond from Warren Buffett?

    What happens if Warren Buffett then just holds the demand deposits in his bank account?

    "Or, not shown in this equation, asset prices could rise." and "MV = PY"

    Is there an MV = PY for financial assets?

  10. Stock market trading is a big term to understand. Briefly the above post has described about it, very nicely.

  11. Hey I am doing, can you suggest me how I can get bank jobs? I have to do too? Or I can choose any other subjects for Post-Graduation. Please reply me, else I need to ask to my professor Aloke Ghosh on this.

  12. Hi Frances. Thanks for pointing me to this and the previous post via Twitter. Both have been really illuminating.

    What I am trying to understand is the role of the government budget deficit in the wider economy. One thing I'm particularly keen to understand is how the UK could be in deficit for 85% of the time since the 1950s - given that this is completely at odds with the rhetoric of politicians and the media.

    One thing that I have been lead to believe (by MMT but also elsewhere) is that the deficit ultimately creates the additional base money required to support economic growth (as well as savings and trade deficit). At first I questioned this because in the UK any deficit is always reflexively matched by borrowing. The implication for me was that while the net spend adds base money, the bond issuance withdraws the money and therefore no new net money is created - in fact this is the point of the debt issuance. This concords with one of your versions of how banks directly by bonds from the government but perhaps not the other. I then learned that the BoE is responsible for forecasting and controlling money requirements and engages in OMO to add or remove base money as required. Presumably, if the economy is growing, there is a general net add of money by the BoE (albeit with variations through the business cycle). I reconciled this the with idea that the deficits provide that base money by assuming that in general, over the long term, the BoE net buys government debt. This creates base money and since the BoE is the public sector, the debt somehow cancels out. The end result is that the deficit has ultimately added base money, albeit via some convoluted accounting mechanisms to get around the fact that the BoE is not allowed to directly fund the government.

    On Twitter, you indicated that the government certainly has to repay the BoE when its debt matures and this is what has thrown me. If the debt needs to be repayed, then this has to come from taxes, which destroys base money. In which case, there is no new net money and the deficit has been completely neutralised.

    So I guess I am struggling to understand if and how the deficit fulfills this role. One thing your posts point to, which I have never been exposed to before, is the idea that some of the deficit may not be neutralised. I suppose these mechanisms are therefore important to the wider economy and something I need to think about a bit more. Any clarifications on my thinking would be much appreciated.


Post a Comment

Popular posts from this blog

WASPI Campaign's legal action is morally wrong


What really happened to Signature Bank NY?