NSW completes first $1 billion of $11 billion bond buy-back (QE) program

Last week NSW's investment and debt issuance agency, TCorp, announced that it had completed the first $1 billion of its proposed $11 billion of bond buy-backs using NSW's $26 billion Debt Retirement Fund. (You can read more about the DRF here.)

This is akin to more than half a standard RBA quantitative easing (QE) program dedicated to just NSW bonds (ie, given the RBA would buy $20 billion of State government bonds, or "semis", in each of its $100 billion QE programs).

Once NSW spends the first $11 billion, the DRF should be left with $15 billion, or possibly more, depending on the direction of travel for the global financial markets in which the DRF is invested.

We are forecasting that either this NSW government or future governments will end up drawing down on the entire remaining $15 billion in the DRF (ie, using $26 billion in total) for bond buybacks to reduce NSW's record debt levels (and mounting interest-servicing costs), which now exceed $100 billion for the first time.

The DRF can then be replenished with future NSW asset sales and true cash surpluses, across the whole of government budget (ie, accounting for CAPEX), when they materialise. It is not clear whether this is NSW government policy as yet. But we believe that this will become the central case.

One of the interesting features of the two $500 million buy-backs last week was that NSW appeared to switch one of the sellers into $200 million of NSW 2032 bonds. It is reasonable to assume the investor selling $500 million of NSW 2024 bonds was a big bank, which is one of the few stakeholders that would own that much of the bond. The buyer of the 2032 bonds was also therefore very likely a big bank.

The spread to asset swap that the 2032 bonds would have paid the bank was about 10-12 basis points. While this is a huge improvement on the negative spread to swap the bank would have been earning on the 2024 bonds, it is materially below the 15-20 basis point spreads to swap that some banks have said they want from 10-year State government bonds to rationalise buying them.

One driver of this shift may have been a change in capital treatment. Normally, banks only really care about their leveraged return on capital or return on equity (RoE). (Banks running their regulatory liquidity portfolios also focus on revenues to minimise the net interest margin drag from holding these assets, which if they pay no return can be very high.) The more capital you hold against an asset, the lower your leverage and RoE.

There are a few recent developments on this front, which are a little complex to explain but important nonetheless.

The first is that banks that are "standardised", which covers all banks except the four majors and Macquarie, do not have to hold any capital at all against a government bond when they hedge the interest rate risk on these assets. This means that these bonds can pay tiny spreads over the swap rate, and still earn standardised banks a huge RoE.

Since APRA shut the $140 billion Committed Liquidity Facility, banks have been forced to hold government bonds instead of their own internal loans and other banks' bonds as a high quality liquid asset. The standardised banks were active users of the CLF, and are now buying government bonds (and semis specifically) to replace it.

Yet the minimum spread above swap that a standardised bank needs is much lower than the spread above swap that other "advanced" banks (ie, the majors and Macquarie) require because the latter do have to hold nontrivial capital against government bonds when they hedge their interest rate risk under a prudential standard known as APS 117.

This means that the marginal buying demand from regional banks for semis could push spreads to swap tighter.

For the larger advanced banks, which are governed by APS 117, there are a few significant recent developments. The capital applied to hedge interest rate risk under APS 117 uses a rolling value-at-risk (VaR) model. Simplistically speaking, this model selects the 99th percentile loss on a past, rolling basis, which is then used to compute how much capital is needed to protect against that loss. 

The March 2020 pandemic shock resulted in a big increase in the capital advanced banks have to hold against semis---and especially longer-dated semis---as spreads to swap moved 30-40bps wider at the 10-year maturity (a fraction of the circa 100bps spread shock suffered by the major banks' senior bonds, the 250bps shock inflicted on the majors' Tier 2 bonds, and the 530bps shock worn by their AT1 hybrids).

If a bank holds its government bonds in a mark-to-market portfolio, the rolling window for the VaR model is 2 years. This means that March 2020 will drop out of the window next month, prospectively slashing the amount of capital they have to hold against semis, which in turn means they can rationalise acquiring these assets on much tighter spreads to swap to hit a given, say, 7% to 12%, RoE target.

Not many banks hold their high-quality liquid assets in mark-to-market portfolios. Most big banks hold them in accruals or hold-to-maturity books. In these portfolios, they have to use a rolling 6-year window, which still captures March 2020.

We run these VaR models internally, and our analysis implies that the sudden jump in interest rates in October/November 2021 could have created a new 99th percentile loss event for the banks (subject to their asset positioning), which is even larger than March 2020 if they were long interest rate duration risk. (For APS 117, this portfolio position is assessed on the last business day of every quarter.)

This is profoundly important because it means that depending on a bank's portfolio holdings on the final day of the quarter, they may shift out of March 2020 and into the new October/November 2021 shock. Crucially, in this latter shock, semi spreads to swap decreased, rather than increased, as they did in March 2020. 

This means that in this 99th percentile loss event, one of a bank's best-performing assets would have been semis, which in turn means that the capital you have to hold against them based on the VaR model to protect against spread changes in October/November 2021 is a tiny fraction of what you have to apply if you use the March 2020 shock.

This may explain why the big banks have been prepared to buy longer-dated semis on tighter spreads to swap than they have previously. Because it is not just about the spread to swap: it is just as important to understand the capital and leverage they hold against the asset, and what return on capital that spits out. 

Simplistically, if they are holding 1% capital against a 10-year semi, and that pays a 10bps spread against swap, the expected return on equity is about 10%, which is in the hitting zone for most banks.

A very similar logic applies to the banks' home loan portfolios, as we explained many years ago. Prior to the Financial System Inquiry changing the risk-weighting system (as we had recommended at the time), banks could get away with risk-weights below 10% for low-risk (ie, low LVR) home loans. 

For a bank with a 10% risk-weighted common equity tier one (CET1) capital ratio, that meant that they might hold less than 1% of CET1 for every $1 of low-LVR home loans they were originating. This in turn meant that they were leveraging their capital 100x, which resulted in immensely attractive RoEs for very small net interest margins on home loans.

Even today the major banks only apply circa 25% risk-weights to low-risk home loans, which means they only have to hold ~2.5% capital against $1 of loans, all else being equal. Put differently, they are leveraging their equity 40x when they lend against houses, which normally makes it a very high RoE business despite seemingly charging very low interest rates (eg, 2%).

You would think that a bank can hold much less capital against an interest-rate hedged government bond than they can against illiquid residential mortgages. What we do know is that the banks now have much more choice about the 99th percentile loss shock they have to select (based on their portfolio choices), which could radically reduce their capital requirements when investing in government bonds.

Back in 2014 and 2015 when APRA was first introducing the new Liquidity Coverage Ratios that required banks to hold government bonds sufficient to cover a 30-day run, or liquidity shock, on the banks' funding, they were typically buying 10-year semis on a 5bps to 15bps spread to swap because the capital treatment of semis was much more favourable than it was after March 2020. These days should eventually return again as March 2020 fades into the background.

Our current modelling suggests the banks will have to buy about $408 billion of government bonds (with a range of say $250bn to $500bn) over the next 3-years to meet APRA's liquidity rules. That's the equivalent of more than 4x standard RBA QE programs.

Of course, when the banks buy government bonds, they tend to spend about 70% of their money on semis and only 30% on Commonwealth government bonds because of the much higher spreads available on the former. 

This differs from the RBA's QE programs, which only spent 20% of the RBA's money on semis (and 80% on Commonwealth bonds). Even if we reduce our bond-buying estimate from $408bn to say $250bn, and we apply the banks' semi vs Commonwealth portfolio preferences (ie, 70:30), that gives about $175 billion of semi bond-buying demand from the banking system over the next 3-years. That is the equivalent of 8.75 standard RBA QE programs of $20bn each in terms of future semis demand.

The banks can sometimes also get attractive returns on Commonwealth bonds via something known as the bond-basket arbitrage. While we don't have time to explain that arb here, suffice it to say that when the basis is positive in this arb, which it is not currently, the arb works. 

The basis is expected to turn positive once the RBA ends it QE program in February. But there are strict operational limits on the banks regarding the maximum number of futures contracts they can sensibly and credibly hold going into a roll, which in turn limits the amount of money they can invest in the bond-basket arbitrage. Some estimates put this at less than $10bn per large bank, although the recent illiquidity in futures markets should have materially reduced this number (ie, given the higher operational roll risks).

Access Coolabah's intellectual edge

With the biggest team in investment-grade Australian fixed-income and over $7 billion in FUM, Coolabah Capital Investments publishes unique insights and research on markets and macroeconomics from around the world overlaid leveraging its 14 analysts and 5 portfolio managers. Click the ‘CONTACT’ button below to get in touch.