Understanding LSAP losses

There was an article on Business Desk yesterday that led with the suggestion that the LSAP losses now totalled almost $20 billion.

As soon as the article appeared I emailed the author and pointed out that the two numbers she was using could not be added together and that the best estimate of the direct fiscal losses were still around $10 billion. We had a few email exchanges and a telephone conversation, by which point she accepted that her number was wrong, but didn’t fully understand why she was wrong (apparently several other journalists were also confused), and indicated that there would be a further article forthcoming, using quotes I (and others?) had provided. I didn’t have the time for anything in depth yesterday but suggested I might write a post today that attempted a fuller and more intuitive explanation.

Unfortunately I didn’t print off the original piece and the article online has now been changed. The original opening has been modified to remove the explicit $20 billion assertion (it is still in the heading but with a question mark) and this note has been added at the start.

As it stands, the article now begins

But it isn’t “before” it at all. Instead, the number labelled as the mark-to-market losses on the LSAP portfolio is just, in effect, a subset of the Treasury estimate referred to in the first paragraph.

Note that everything in this post, in Jenny Ruth’s article, and in the Treasury papers referred to, deals only with the direct fiscal losses to the taxpayer from the LSAP programme. It does not deal with any possible offsetting savings by lowering general Treasury issuance costs or with possible macroeconomic benefits (or costs), not because either of these might not be important, but simply because the issue at hand is about the direct fiscal costs only.

Jenny Ruth’s article drew on the Treasury paper to the Minister of Finance, dated 6 December 2022, which I discussed (together with a follow up note from the Reserve Bank) in this recent post on the idea of tiering returns on the current very large level of settlement cash balances. The paper itself is included in the huge recent OIA release from the Minister of Finance, hosted on Treasury’s website. So too is one other Treasury paper I will draw from.

Here is the $10.5 billion number from the Treasury paper Ruth draws on

Note that there is a footnote at the end of that sentence. Here is the footnote, some of which was withheld

This, on its own, really should have been enough to suggest to any reader that the two numbers were approximating to the same thing, and thus could not be added together. (All that said, this paper was about tiering and so did not go into any great depth on how the various LSAP loss numbers were derived.)

But here it is worth taking a step back.

For the last 18 months or more I have been banging on about the rising scale of the direct fiscal losses associated with the LSAP. To illustrate that point I regularly used, both here and on Twitter, a chart of the line item on the Reserve Bank’s monthly published balance sheet of the Bank’s claim under the Crown indemnity the Minister of Finance has given them for any LSAP losses. I used that because (a) it was available monthly, and it was an official number and (b) because, while there were no realisations (bonds sold off again) it was a reasonable best estimate (the market price) at the time of what the actual direct fiscal losses were.

But, as I have pointed out, and as The Treasury itself has frequently pointed out, the indemnity claim itself does not represent the cost to the wider Crown finances. The indemnity claim is just a transfer from the Treasury to the Reserve Bank. Had the indemnity not been given, the Reserve Bank could still have done the LSAP. Instead of a large claim on The Treasury for the indemnity it would instead be showing large losses itself and deeply negative equity (as is the situation with the RBA, which did large scale bond buying, without an indemnity). Either way, taxpayers are worse off but they are worse off because of the LSAP (and subsequent movement in market prices for bonds) not because of the indemnity. The Reserve Bank balance sheet is consolidated into the Crown accounts, and transfers among entities included in that consolidation don’t make us taxpayers any worse (or better) off. They are really just intra-group transfers.

It is also worth noting – as many early sceptics did in their responses – that if the LSAP bonds were to be held to maturity by the Reserve Bank no indemnity claim would have been payable at all. The market price of the bonds approaching matiruty would have tended towards the face value of the bonds.

But that is not the same as saying there would have been no direct fiscal losses from the decision to do the LSAP.

What matters in assessing the direct fiscal costs is not the flows between the Reserve Bank and the Treasury (intra-group transfers) but the flows between the government as a whole (here, Treasury and Reserve Bank combined) and the private sector. The LSAP changed the private sector’s lending to the government from bond holdings to settlement cash balances at the Reserve Bank. This is the story I’ve been telling for ages that the LSAP is just a big asset swap. Treasury says the same thing (from that same 6 December paper)

In what follows, I’m treating the Reseve Bank and Treasury as one consolidated entity (whole-of-Crown). In decision making terms it doesn’t work that way (the RB MPC made independent decisions to do the LSAP), but I think the substance may be clearer if I do.

Take a very simplified example in which these are the only transactions whole-of-Crown does:

Step 1

  • the government issues $50 billion of long-term bonds to the private sector to finance $50 billion of spending,
  • the net effect of those two transactions is no change to the level of settlement cash balances (spending boosts settlement cash, but issuing bonds drains those balances)
  • in this scenario, the government now has $50 billion of long-term debt outstanding and no extra (RB) settlement cash balances outstanding

Step 2 (the LSAP)

  • the Reserve Bank (branch of government) buys back the $50 billions of bonds on the open market,
  • paying for its purchases boosts the total level of settlement cash by the full purchase price (for simplicity, assume also $50 billion)
  • having done this, the government (whole of Crown) now has a) no long-term debt outstanding (one division might have issued the debt while another has repurchased it, but they net to zero) and (b) $50 billion of extra settlement cash liabilities outstanding.

The government could have financed that $50 billion at the market interest rate on the long-term bonds.  Instead, it is now financing itself with settlement cash balances (in economic substance, borrowing from banks) on which its Reserve Bank arm pays the full OCR interest rate, reviewed every six weeks or so.

The direct fiscal cost of the decision to do the LSAP is the difference in the debt servicing costs in these two scenarios (how much whole-of-Crown pays the private sector).   Transfers between the government and the Reserve Bank themselves don’t add to or subtract from those numbers at all.

When the LSAP was underway, the Reserve Bank was probably buying back the bonds at yields to maturity of around 1 per cent.   In other words, the Crown was giving up, say, 1 per cent fixed rate financing and replacing it with variable rate financing.  Had the OCR averaged below that (approximate) 1 per cent over the remaining life of the bonds, there would have been a net direct fiscal saving.   If that seemed unlikely even in 2020, it wasn’t impossible (for the first year or more the OCR was 0.25 per cent).

But it hasn’t happened (or, to be strictly accurate, since we do not know what disasters might lie just around the corner, it now seems most unlikely to happen).  The OCR is now 5.25 per cent.  From here, in its latest OCR review the Reserve Bank indicated it isn’t sure where the OCR will go next.   The market is looking towards rate cuts at some point, but the market does not expect anything like an average of 1 per cent or less.  There will, almost certainly, be substantial net direct fiscal costs from the LSAP programme.   

That number was what The Treasury was estimating (reporting in early December) at about $10.5 billion.  Bond yields, changing from day to day, encapsulate implicit market expectations of future short-term rates, including the future path of the OCR over the remaining life of the relevant bond.  Thus, the mark-to-market value of a bond position is going to approximate to the estimated net direct fiscal losses from a decision to purchase those bonds (as Treasury notes in the first quote above).  They are two ways of looking at the same thing, not different losses that can be added together.

Current market bond yields are a little below those at the end of October, so Treasury’s best estimate now of the net direct fiscal losses would probably be a little lower.   Since we don’t know exactly the date as at which they came up with the $10.5 billion or the precise details of their methodology we can’t be quite sure how much lower, but $10 billion or perhaps a touch below look to be about right.  Those estimates too will, in principle, change every day.

As Jenny Ruth’s article notes, the Reserve Bank balance sheet for 31 March will be published on Tuesday, and we will get an update on the Bank’s claim on the Treasury under the indemnity.   However, the claim under the indemnity no longer approximates the total expected net fiscal direct losses (or the Reserve Bank’s own losses).  That is (mostly) because some of the LSAP bonds have been sold (back to Treasury), under a steady announced monthly programme, crystallising some of the Reserve Bank’s losses, and triggering payments from the government to the Reserve Bank under the indemnity (some of the shorter-term bonds have also been held to maturity).   I am not aware of where we might find an up-to-date time series of these payments, but in that big Treasury OIA release there is a 24 November 2022 paper to the Minister of Finance on monthly indemnity payments.  It includes a table of actual indemnity payments to date

indemnity 1


and one of expected payments over the followng year (on market rates prevailing when the estimates were done)

indemnity 2

Jenny Ruth reported (correctly) the Bank’s outstanding indemnity claim as at the end of February of around $8.75bn. But to get the total Reserve Bank losses under the indemnity one would have to add to that (a) the $438 million of known indemnity payments already made, and (b) any other indemnity payments between November and February, estimated in the Treasury papers then at another $1.1bn or so. In other words, a total a bit over $10 billion. Note that the entire interest rate yield curve for medium to long term bonds is lower now than it was in Oct/Nov (or in February) so accurate estimates now would be a bit lower – probably a touch below $10 billion.

If I think back to yesterday’s conversation I think one issue that may be playing on people’s mind is “yes, we know about the settlement cash payments you describe, but…..surely there are still those mark to market losses showing on the RB balance sheet”.

And that is true, but if the Treasury were to engage in full mark to market accounting of the value of its debts (which few entities do, or are required to be accounting standards, but which is nonetheless the economic substance), what would have happened to the market value of the debt Treasury had issued? It would have increased massively – low interest outstandings are very valuable when the market rate is so much higher. In fact, the market value of the increased value of the Treasury’s liability in respect of the LSAP bonds would be equal to the reduced market value of the Reserve Bank’s holdings of those LSAP bonds. The two net out – reverting to my oversimplified scenario it is as if the whole-of-Crown no longer has that debt outstanding at all.

The indemnity (claim and payments) matters a lot to the Reserve Bank. It doesn’t much matter to us, or to the Treasury. We face whatever gains (or losses) the LSAP gives rise to whether or not there was an indemnity. The remaining claim – latest estimate out on Tuesday – is (from our perspective) little more than an intra-group memo item. But as taxpayers we don’t care much about either the RB or the Treasury finances individually, but about the consolidated financial position, which is what determines the losses we ultimately have to bear.

Finally, because some LSAP bonds are being held to maturity (and there are losses arising from them, albeit often smaller) there will never be a single government accounts line item capturing the full net direct fiscal costs of the LSAP programme.

I hope this post has made things a bit clearer. I’m happy to engage with anyone on the points in it, or any that are relevant that I may have missed.

UPDATE (17/4): Jenny Ruth’s follow-up story written on Friday is here. Note that this line she was given by Treasury is wrong or at best misleading.

If the Reserve Bank sold the LSAP bonds on-market that would certainly and automatically drain settlement cash levels. If the Reserve Bank sold the bonds back to the Treasury that transaction alone would not affect banks’ settlement account balances at all (it would just deplete the government’s account at the RB). The effect on settlement cash would arise only as and when the Treasury issued new bonds on-market to replenish the government’s account and (in effect) replace the bonds the RB had sold back to them.

It is a small point, but since the Treasury line is now on the public record and there is evident lack of clarity in some quarters about the LSAP issues it is worth being as clear as possible. As per the post above, transfers/flows between the RB and Treasurydo not directly affect banks or the wider public. For many purposes (but not all) there are just intra-group transfers among branches of whole-of-government.