Note on 'Australia's Pandemic Exceptionalism', by Steven Hamilton & Richard Holden

My notes on: Holden, Richard., & Hamilton, Steven. (2024). Australia's Pandemic Exceptionalism.

[This note may contain errors and inaccuracies. It was mostly written just for myself, in preparation for my live podcast conversation with Richard Holden and Steven Hamilton.]

This is really a book about Australia's state capacity—that is, the ability of our governments, at all levels, to achieve their policy aims—told through the prism of the pandemic.

Insofar as that's how you should interpret this book, it offers us a valuable case study, though no deep explanations for Australia's level of state capacity.

Sixteen things I learned

(1) Treasury's complacency around the 1990-91 recession was really bad.

I knew Treasury had under-reacted to the recession of the early 1990s, but I hadn't realised just how complacent it was.

Unemployment peaked at 11%.

The stimulus package that was finally announced was truly too little too late. It was announced in May 1994.

(2) Treasury's complacency in the early 1990s derived from a triumphalism about the economic reforms of the 1980s.

The 1980s reforms were in large part motivated by the 1982-83 recession. The lesson Treasury took from that recession was that the economy was too arthritic, with high tariffs, centralised wage fixing, a fixed exchange rate, and a non-independent central bank.

Come the 1990-91 recession, it assumed that the economy was now sufficiently flexible and so fiscal stimulus wasn't needed. This was incorrect.

(3) The 2008-09 stimulus package was pulled 'off the shelf'. Treasury had designed it in 2004.

The 2008-09 stimulus package totalled more than $50 billion. Its centrepiece was of course the $950 cash transfers to every Australian earning less than $80,000.

It had been designed years earlier.

Ken Henry:

"We started the work in 2004, and we had come to the view that if we ever found ourselves in front of a government confronting the probability of a big economic slowdown, we would say, ‘Yes, you do need to act. And you do need to act early.’ And that payments to households, of all the things that we can think of, are likely to be the most effective. We had a rule of thumb that if you’re going to do something you should be delivering about half a per cent of GDP, at least." (p44)

(4) Infrastructure projects are not an effective means of injecting fiscal stimulus, and have now basically been abandoned.

The economic effects of infrastructure projects operate with long lags.

Ken Henry recalls a Treasury study showing that infrastructure projects begun towards the end of the 1990-91 recession did not generate any economic impact until as late as 1997. (p136)

By the time of the pandemic, Treasury had learnt its lesson: not one dollar of the third-of-a-trillion-dollars of fiscal stimulus announced in 2020 was directed to infrastructure.

(5) The 2020 economic response—not only the 2008 response—learned from the 1990s mistakes.

We're all familiar with how Ken Henry and other senior Treasury officials were haunted by the 1990-91 recession, especially the labour scarring it caused, and how this influenced their aggressive response to the 2008 financial crisis (immortalised in Ken's "Prime Minister, go hard, go early, go households" mantra).

I wasn't quite familiar with the fact that these same lessons were perhaps equally looming over Treasury officials in 2020, and motivated the potent JobKeeper response.

The prime lesson was to keep people connected to their jobs. Allowing those links to disintegrate would cause labour scarring and destroys firm-specific capital.

Ken Henry:

"I know that was in the minds of people in Treasury in 2020 when they were thinking about how to deal with the pandemic. You’ve got to preserve the jobs. You’ve got to preserve that connection between employee and employer. That is the most important thing." (p46)

(6) The key difference between the US and Australian economic responses to the pandemic was that the US focused on stimulus payments, whereas Australia focused on keeping employees attached to their employers.

Australia's response proved, of course, to be the more effective and better-adapted one. In a pandemic, the Keynesian multiplier is muted because people are self-isolating or locked down. At the same time, lockdowns threaten the destruction of businesses.

The effectiveness of Australia and America's respective approaches is shown in their employment-to-population ratios. Australia's returned to its pre-crisis level within 12 months. In contrast, America's e-pop ratio has barely recovered five years later.

Why the US mistake? Partly it seems (and this is more my interpretation) they were too trigger-happy because they over-learned the lessons of 2008-09, where they did not provide enough stimulus but should have, and misapplied them in 2020 (where they were less applicable).

Partly, it was also because of their political system and the need to pass policy responses through two independent houses of parliament (a constraint Australia does not have when seeking to inject fiscal stimulus). Claudia Sahm explains how this led to the emphasis on stimulus:

"I believe you have to do the checks because that’s the buy-in … There are a lot of Americans who hate unemployed people, so you don’t get the unemployment benefits without the checks. That’s the importance of going broad – you get the buy-in." (p59)

(7) From a budgetary perspective, it was as if the pandemic never happened. We successfully executed the V-shaped recovery.

Australia's net debt before the pandemic (2018-19)?

19.2% of GDP.

Australia's net debt after the pandemic (2022-23)?

19.2% of GDP.

No difference.

(8) Australia's economic response was enabled to a remarkable degree by its economic infrastructure.

To put this in perspective, it's helpful to contrast Australia with the US.

Two crazy US anecdotes:

1. For its boost to unemployment insurance (UI), the US picked a flat increase of $600 per week for all workers, regardless of their prior income. This number led to the median UI recipient receiving 145% of their prior earnings, perversely incentivising workers to quit to secure higher unemployment payments. So why did the US pick that number? Answer: several states' computer systems didn't allow them to use multiplication. The systems could only handle addition. And the states couldn't reprogram them fast enough (they had to pull former employees out of retirement to do so). Adding $600 to every recipient's check was projected to take the median benefit up to 110%, so that's what they did.
2. The Paycheck Protection Program (PPP), which was the US equivalent of JobKeeper, was administered via the Small Business Administration. But the SBA was small and lacked capacity. Quote from page 140: "the SBA typically had been tasked with assisting businesses in smaller scale and geographically localised natural disasters. Applications would have to be made via commercial banks that had a relationship with the SBA and processed on a physical terminal inside the branch. The program was poorly targeted as the SBA had no real means of assessing eligibility, and is said to have resulted in a high rate of fraud."

In short, America's economic plumbing severely constrained policymakers.

In contrast, here are two Australian examples:

1. The ATO smoothly handled early super withdrawals. Quote from page 139: "The policy was announced on 22 March 2020 and, less than one month later on 20 April, participants could log onto their myGov account, tick a few boxes, and then within just three days $10 000 would appear in their bank account. The system processed more than 520 000 withdrawal requests in the first three days. And the government wasn’t only implementing this program but a raft of others simultaneously."
2. JobKeeper—the centrepiece of the 2020 fiscal response—was able to be delivered because the ATO could ensure high integrity (contrast this with America's PPP, which was beleaguered by fraud). It could insure high integrity thanks to Single Touch Payroll (below).

(9) Single Touch Payroll (STP) was the "unsung hero" of the pandemic.

The ATO's newfangled electronic reporting system, STP, finished rolling out just six months before the pandemic began.

It integrates with payroll and accounting software used by most businesses. Every time an employer processes a pay run, STP reports salaries and wages, tax withholding and super liabilities to the ATO, all invisibly and without the business needing to do anything.

Its intended purpose was twofold: reduce the reporting burden on businesses and ensure tax compliance.

It was serendipitously co-opted during the pandemic to provide the financial plumbing for JobKeeper: it enabled the ATO to verify employee retention and wages at the same fortnightly frequency at which it provided the JobKeeper payments.

(10) The ABS was the unsung departmental hero of the pandemic.

The ABS, under the leadership of David Gruen, seized the pandemic as an opportunity to innovate.

Some things they did:

1. Historically the ABS published deaths data for the previous year in October of the following year (a 10 month lag). They worked out ways to compress this to 2 months by using doctor certified deaths and coroner assisted deaths. (p159)
2. They integrated with the Australian Immunisation Register to support health policy during the pandemic. Quote from David Gruen: "The person who drove that in the Health Department was a former ABS person who understood integrated data assets. He convinced the Health Department that they didn’t need to sit people down before they got vaccinated and ask them a whole bunch of questions, including about their ethnic origin. We got to a stage where we were updating the link between the Australian Immunisation Register and our integrated data assets weekly. The benefit of that was that as the vaccine rollout was occurring, they then had very good information about which ethnic groups were not getting vaccinated, which allowed them to better target information campaigns." (pages 159-160)
3. They pulled all the data from STP, organised it, and provided the government with real-time information on the severity of the downturn to calibrate its policy responses. Quote from David Gruen: "There was an ABS person who was in-posted into the ATO who had been planning what to do with that data once the ABS got a hold of it, and had thought about the statistical issues that needed to be dealt with. STP gives you a feed of data every time an employer runs their payroll, on different frequencies, on different days. So the challenge was how to turn this mass of data into something useful. So we got it in early April, and they managed to publish meaningful statistical results by the end of April. It’s a huge dataset covering more than ten million employees. They worked incredibly hard. It was all put into the cloud because it was so big. It was a model for how we were going to do things in the future. And the whole thing was stood up in three weeks. The first publication came out in late April showing the huge collapse in the labour market. This was very high-quality information about the sectors of the economy that were suffering the most." (p156)

(11) Single Touch Payroll helped the ABS uncover an additional 270,000 government workers?!

Page 157:

"But STP wasn’t just faster. It was also far more accurate. Demonstrating just how big a leap STP represents over existing labour market indicators, in November 2023 the ABS revealed that using STP they had uncovered an additional 270 000 government workers – 13% more than they had previously been aware of based on the Survey of Employment and Earnings. This meant government workers had in fact received $21 billion more in pay than the $194 billion previously estimated."

(12) Yes, the vaccine procurement strategy was hugely bungled. But I hadn't realised just how inane some of the mistakes were.

The single example that struck me most was the apparent buy-one-from-each-category approach the government took to vaccines.

That is,

• for the mRNA vaccines we chose Pfizer (over Moderna),
• for the viral vector vaccines we chose AstraZeneca (over Johnson & Johnson), and
• for the protein vaccines we choose Vovavax (over Sanofi, and after the UQ vaccine failed).

The elementary problem with this approach is well explained on pages 82-83:

"This implicitly assumed that the success of vaccines within a particular technology was perfectly correlated. In other words, if AstraZeneca’s vaccine succeeded then Johnson & Johnson’s would also succeed with 100% probability. That ignored the differences in approaches among vaccines that used similar underlying technology. It ignored the differences in speed to market among different vaccines. It ignored the differences in manufacturing processes that could lead to different success profiles. It ignored inevitable differences in distribution. And it ignored the potential differences in the eventual efficacy of vaccines using the same underlying technology. In short, the ‘buy one from each category’ approach treated apples and mangoes as if they were interchangeable just because they are both types of fruit."

(13) The costs of the vaccine debacle—about 10% of GDP—make it the largest public policy mistake in Australian history.

To put it in perspective, that's more than half the loss of GDP in the Great Depression, where Australia's GDP dropped 17%.

(14) The vaccine roll-out was also bad.

To pick one strange decision in relation to the roll-out, the government essentially decided GPs shouldn't profit from vaccinating Australians. (See page 98.)

So they were being asked to shift capacity from higher-remunerating appointments to vaccination appointments, at a time when they were already under financial strain.

Instead, they should have been incentivised to provide vaccinations.

(15) Victoria's contact tracing was initially done by telecopier.

Page 125:

"There was near universal acceptance that Victoria’s contact tracing system performed very badly in absolute terms, and also in relative terms when compared with New South Wales. It sounds almost unbelievable, but Victoria’s system in 2020 (not 1980) still involved notifications of new infections coming in by fax – or perhaps we should be more precise and say ‘by telecopier’. Only slowly, almost grudgingly, did Victoria move to a more automated approach to contact tracing – eventually adopting a system for data management developed by IT giant Salesforce. This had been suggested earlier, but rejected by the Victorian government because it was purportedly so overwhelmed by the first wave of COVID-19 that it couldn’t implement the Salesforce system."

(16) If there's a 'lingua franca' across all domains of policymaking, it is (or at least ought to be) the language of economics.

Clearly there's a huge variance in economic-thinking capability across government departments.

This was one of the main meta-lessons I took from the book.

Two questions I was left wondering

(1) What was the central analytical failure of the vaccine procurement strategy?

On a couple of occasions the analysis in this book is a bit muddy as a product of having two authors who seem to have mostly written their chapters separately.

For me, the most glaring example of this was the attempt to explain the error in thinking that led to the procurement failure.

In Chapter 1, which I believe was written by Richard Holden (based on his writing style), the failure is framed in terms of the Tinbergen rule:

"This flawed thinking would be on display in a much more serious way when it came to vaccine procurement, where the government favoured the UQ and AstraZeneca vaccines at least partly as a form of industry policy, to disastrous effect. We will return to this issue later, but suffice it to say that a basic principle of economics, known as the ‘Tinbergen Rule’, is that each policy objective requires an independent policy instrument. In other words, one can’t kill two policy birds with one policy stone." (p23)

In Chapter 3, which I believe was written by Steven Hamilton, the failure is described in terms of failing to "buy insurance"—that is, we went all in on two vaccines, when we should have bought many of them in case some failed or were less efficacious.

For example, here's (presumably) Hamilton on page 102:

"Developing a domestic manufacturing capability did not in any way preclude the government from also buying insurance – in the form of redundant mRNA orders – against other possibilities, such as the domestically produced AstraZeneca vaccine running into trouble (as it did). Ultimately, the government needed to buy insurance against a whole raft of unforeseen events."

(Moreover, as Hamilton points out, investing in the domestic manufacturing capability was itself a kind of insurance against 'vaccine nationalism' in the northern hemisphere, which indeed turned out to bite Australia when Italy blocked the export of a quarter of a million doses of the AZ vaccine that Australia had contracted for.)

I think Hamilton's insurance analysis is more relevant.

I don't think couching the error, as Holden does, in terms of failing to heed the Tinbergen rule is helpful. For me, the Tinbergen rule applies more to institutions. For instance, you don't want a central bank to possess one instrument—monetary policy—for targeting multiple goals—e.g. inflation and house prices—because over time those goals may trade off.

In the case of one-off decisions like how many and which vaccines you should buy, if you can achieve multiple objectives with the same decision, then go for it.

The threshold question was: are we buying enough insurance? And clearly we weren't.

The insurance was perfectly affordable.

The mRNA vaccines only cost about US$30-39 per dose. If Australia placed orders for enough doses to cover all 25 million Australians for all eight potential vaccines, the total costs would have been only about $16 billion. (pp79-80)

The bottom line is: we easily could have payed for the insurance, and as long as we did, who cares if we also tried to kick some industry policy goals.

But I'd be interested to hear what Holden thinks of my review here, and exactly what he had in mind with his analysis of the Tinbergen rule.

(2) The plot thickens: Treasury was 'in the room' for decisions on vaccine procurement and roll-out.

JobKeeper was a success. It was designed by Treasury and delivered by the ATO.

In contrast, the vaccine procurement and (to a lesser extent) roll-out, the contract tracing app, the sluggish vaccine approvals, and the fetishisation of PCR tests over RATs were failures. They were variously the responsibility of the Department of Health and Australia's medical-regulatory complex (principally the TGA and ATAGI).

The worst of these health mistakes was, of course, the vaccine procurement strategy. (The overarching failure here, again, being the decision to gamble on the AZ and UQ vaccines rather than buying relatively cheap 'insurance' by advance-purchasing, e.g., each of the eight vaccines sponsored by Operation Warp Speed.)

The main inference we're left to draw is that Treasury was more competent and the health officials lacked economic thinking or mathematical literacy.

But here's the mystery: Holden and Hamilton tell us that they were told that Treasury was 'in the room' during the vaccine procurement conversations. (see p101 and p202)

So why didn't Treasury speak up? Or if it did, what happened?

I would be interested to know whether Steve and Richard—or anyone reading this—may have insights into the nature of those conversations.