Get the banks to buy the bonds (or how to better align fiscal, monetary and bank regulatory policies)

[This is an edited, updated, and expanded version of a post that first appeared on livewire]

The COVID-shock of 2020 put a spotlight on a design flaw in Australian regulations. At a time when governments are borrowing lots of money, and the RBA is buying government bonds to boost the economy, regulations discourage banks from owning government bonds. Instead, banks have been selling government bonds, essentially working against monetary policy. This is an undesirable outcome. Regulations ought to be changed to encourage banks to hold more bonds.

In March 2020 households were both unwilling and unable to spend as shops shut and people hunkered down. The massive collapse of spending that resulted from this hunkering down showed up as higher savings and, ultimately, bank deposits. The government borrowed these savings by issuing bonds and spent the money — to keep the economy going. The RBA supported this fiscal strategy by both lowering interest rates and buying government bonds.

At a high level, this sounds like a textbook scenario. Households save more and the Government borrows the excess savings and spends them to keep things going. Except that, in the Australian case, regulations impede the process.

Improving alignment with monetary policy

With the cash rate floored, monetary policy is bond buying. ADIs have been working against monetary policy. Between Q3’20 and Q2’21 LCR ADIs sold $101bn of government bonds, which is just over 40% of RBA purchases (for the two quarters to Q1’21, LCR ADIs sold $76bn v. RBA purchases of $110bn, or about 70%).

LCR ADIs sold $101bn of bonds between Q3’20 and Q2’21

Regulation makes government bonds expensive to own, so as yields have declined ADIs have been selling them. That is, ADIs made an economic decision to sell their bonds, as they prefer to hold zero-yielding deposits at the RBA (Exchange Settlement Balances, or ES Cash).

If, instead, ADIs were buying government bonds alongside the RBA, interest rates would be lower, the Australian dollar would be lower, and the economy would be stronger.

More bank buying of government bonds would also boost savers & support fiscal policy. The revenue boost that ADIs obtained from holding assets that actually have a yield could be passed on to income-starved savers. Lower interest rates on government debt would make it a bit more affordable for governments. Australian governments could then choose between spending a bit more (essentially spending the savings from the lower interest rate) or lowering taxes.

Improving alignment with fiscal policy

Change would also support fiscal policy by aligning demand for bonds with the type of spending governments want to undertake. The current regime has banks hold capital in proportion to the historical volatility of a bond. Bonds in HQLA books are swapped, so in practice this means the historical volatility of the spread-to-swap.

Bonds with longer maturities had much greater volatility in 2020. As a result they now require more capital to be held against them. The upshot is that banks prefer bonds with shorter maturities.

This runs counter to the demands of public policy.

Even before the COVID-recession, there was a broad consensus that Australia should spend more on infrastructure. The desirability of boosting demand following the crisis has further encouraged this category of spending. Infrastructure takes a long time to build and lasts a long time. It is prudent to fund long-lived investments with long-dated debt.

Governments would like to issue more longer dated bonds, but regulations reduce ADI demand for long-dated government bonds! This is an issue for State Government bonds (Semis), where ADIs typically hold about 50% of the outstanding stock of bonds.

Of course, governments could fund the infrastructure by issuing short and managing the ‘mismatch’ – but managing interest rate ‘mismatch’ risk is the business and skill of banks and bankers. Good public policy should seek to align risks with skills.

The ideal regulations would not unreasonably discourage ADIs from buying the debt that public policy and sound risk management demand. Aligning ADI-buying with the investment plans of borrowers would transfer interest rate risk to those who either want it, or are best placed to manage it.

What’s the catch?

Regulators make ADIs hold capital to reduce the probability and cost of a taxpayer bailout.

The problematic regulation, APS 117, is designed to ensure that banks have enough capital to protect them against losses arising from changes in interest rates. Bond prices fall when interest rates rise, so capital is held to protect LCR ADIs from falling prices of bonds in HQLA books when interest rates rise.

However, if a bank is in trouble interest rates are likely to plummet, so this doesn’t really stack up. If you think in terms of hedging a bailout, we may actually wish Banks to take more interest rate risk!

Because these regulations only apply to bonds, current regulations create a bias toward private transactions. For example, if an ADI made a direct loan to an Australian government to fund infrastructure, it would not have to hold APS 117 capital against that loan (assuming that the fixed rate loan is hedged with swap). However, if the same loan were made via the purchase of a government bond on public capital markets, APS 117 capital would have to be held against the bond.

This is perverse. A government bond is both a highly liquid asset and subject to all the scrutiny that comes with public capital markets. A direct loan is illiquid, and subject to much less scrutiny. Good public policy should encourage public funding of governments, to encourage both transparency and scrutiny.

Given these objectives, there’s an argument that banks should have to hold less capital against bonds, relative to loans, as their superior liquidity and greater scrutiny means bonds are safer assets. We should, at least, treat them the same. If we are happy enough with the swap-hedge on a direct loan we should consider if a swap-hedge is sufficient a bond that’s held in a bank’s high quality liquid asset (HQLA) book as well.

The different treatment arises from different assumptions about holding periods. A direct loan is treated as a hold-to-maturity asset, so changes in the theoretical value of the loan arising from interest rate volatility are deemed to be unimportant. However, a government bond might be sold at any time, and has an observable price each day, so banks are asked to hold capital to protect against this risk.

However, this doesn’t fit with reality. It is a regulatory requirement that HQLA books are a held in proportion to the size of an ADIs’ business. As a result, HQLA books seldom decline in size, and are therefore basically hold-to-maturity books. Notwithstanding the recent selling of bonds as ADIs move into ES Cash, the total size of system HQLA assets has remains about the same. Moreover, ADIs would not have sold government bonds to move into zero-yielding ES Cash if regulations were different.

Liquid Asset stacks are non-decreasing

Who pays?

Money doesn’t come ‘out of thin air’ — so where does the higher interest income for ADIs & savers, and lower interest cost for taxpayers come from?

It is mostly a transfer from those who currently invest in Australian government bonds. If there were more buyers, yields would be lower and existing holders would either have to accept a lower interest rate or sell the bond. So it’s basically a transfer from government bond investors (around two fifths of government bonds are held by foreigners) to Australian taxpayers.

There’s also a little bit more risk to taxpayers, as ADIs would be operating with slightly less capital. So Australian taxpayers get a bit more money and accept a bit more risk. The Commonwealth Government did much the same thing in 2009 when they guaranteed State and Territory Government debt. It’s worth noting that the Commonwealth Budget didn’t recognise a contingent liability for this guarantee.

What if they have to sell?

The main argument, then, is that capital should be held against the possibility that the value of bonds in an HQLA book falls at just the point that they must be sold. After all, bonds in liquid asset portfolios are notionally held so that they can be sold for cash in a crisis.

But banks are very unlikely to sell HQLA in a crisis — they will repo the bonds into the RBA.

Banks hold just under $260bn of government bonds just now. Even on the best day, there is no bid for hundreds of billions of dollars of government bonds. And in a crisis we wouldn’t want banks to sell down their enormous liquid asset books, as it would destroy the liquidity of the government bond market.

A liquid risk-free benchmark is a public good that should be, and would be, protected.

So, the only buyer of theses bonds is the RBA. In practice, the banks wouldn’t even sell the bonds to the RBA. They would enter into a long-dated agreement to swap the bonds for cash today, with an agreement to reverse the transaction at a later date (a repurchase agreement, or repo).

What’s the point of all the HQLA?

Given that banks are unlikely to sell HQLA in a crisis, the only purpose of ADI HQLA holdings is to make sure the taxpayer gets collateral of the highest quality when they back-stop the banking system (via the RBA).

We should think of this requirement as a tax. The Australian Government knows that it will have to backstop the banks if there’s a problem. To reduce the expected cost to taxpayers of that implicit Government Guarantee, regulators require banks to hold some of their assets in the form of low-yielding government bonds in special HQLA books. These low-return government bonds displace higher return assets, such as loans to households and businesses.

This is a good aim for public policy.

But it follows that the government bonds banks hold in these portfolios can be treated differently. Otherwise we’re basically adding a tax to a tax.

What would happen if APS 117 were eliminated?

The simplest solution is to eliminate APS 117 capital for government bonds that are held in HQLA books. There are already exemptions from APS 117 for government bonds that are hedged with futures, as they are substantially the same thing. This concept of a ‘good enough’ hedge could be extended to include a government bond that’s hedged with a coupon-matched interest rate swap.

The counter to this is that it may encourage banks to hold the riskiest government bonds (the longest maturity, or the edgiest state and territory government bonds). Should regulators wish to temper this behaviour they could require banks to hold capital in proportion to the (normalised) level of the spread between the bond and swap.

Consider what happened in March 2020. A once-in-one hundred year pandemic caused a liquidity shock. That liquidity shock caused very sharp moves in government bond markets, and in particular the spreads between bond and swap rates. To protect the integrity of the government bond market, Central Banks all around the world started buying government bonds to stabilize government markets. The vigor of this response fully reversed the widening of spreads between government bonds and swaps.

The way APRA currently measures risk, that short period of volatility, that was actively reversed by the RBA, has led to a large increase in the cost (capital charge) for holding government bonds. And that increase in the cost of holding bonds will persist for many years.

In contrast, the level of the spread is now tighter than before that episode (partly due to RBA buying in QE).

Semi ASW spreads are well below historical norm = less risky!

For contrast, consider the case where a slow recession caused a gradual but equivalent widening of spreads. Say this spread widening happened at one basis point per day for two months, and did not revert. In this scenario, the capital charge associated with bank holdings of government bonds would not have changed and may even have declined a little (a move of 1bps per day is lower volatility than the historical norm).

Most people would agree that more capital should be held against bonds with wider spreads, as they are more risky!


In conclusion, the current regulatory framework is working against public policy in many spheres and ought to be changed. Good public policy would incentivize banks to support monetary and fiscal policies a time when these policies are (almost) maxed out. Change would better align bond demand with the type of spending that we want public authorities to undertake. And it would better align risk and skill.

Bonds held in HQLA portfolios are essentially hold-to-maturity, and won’t be sold in a crisis. So the capital banks are required to hold against them should be proportional to measures of forward looking risk, and not backward looking measures of historical volatility. The best forward looking measure of the riskiness of the bond is the (normalised) credit spread.

If regulators are worried about collateral shortages due to mark-to-market losses, they could increase the minimum size of these HQLA portfolios. In practice, banks already do this. Most maintain liquidity stacks that are at least 25% larger than their regulatory minimum.  Also, APRA has just proposed that banks hold an additional 30 percentage points of contingent collateral in the form of self securitised assets — so the real buffer will exceed 50%.

A buffer of that size is unquestionably strong: if banks held an HQLA portfolio with an average maturity of 10yrs, a 50% buffer would cover a 500bps increase in yields (or spreads).

Leave a Comment