Andrew Leigh

 

SPEECH

How COVID Changed the World of Data

Population Symposium
Australian National University School of Demography

6 October 2023
Canberra

 

I acknowledge the Ngunnawal people as traditional custodians of the ACT and recognise all First Nations people present today. 

I commit myself, as a member of the Albanese Government, to the implementation in full of the Uluru Statement from the Heart, which starts with voting Yes on October 14. 

Thank you to the Australian National University for hosting today’s Symposium and thank you for focusing your efforts on understanding the impact of the COVID pandemic on demography in Australia.

I want to preface my remarks by acknowledging the remarkable ability of the Australian Bureau of Statistics (ABS) and other government agencies to rapidly shift their operational focus during the pandemic.

Demographers deserve credit for helping guide governments, policymakers and the community through the COVID pandemic.

Today, I will tell the data side of the story.

COVID might have shrunk our worlds to frequent Zoom calls in which we took it in turns to remind each other ‘You’re on mute’.

But COVID also opened up a whole new world of data.

In the space of just a couple of years, we have access to more timely and frequent updates, new sources of data and greater integration.

I welcome the opportunity to talk about this extraordinary legacy, the lessons learnt and how we can build on it.

Rocking the demographic boat

The COVID pandemic was not only a rapidly evolving health crisis but an economic crisis too – the most severe economic downturn since the Great Depression (Kennedy 2022).

Restrictions to limit the spread of COVID saw a reduction in spending, business turnover, losses in jobs and hours worked, and supply chain disruptions (ABS 2022a).

By June 2023, the level of GDP is estimated to have suffered a cumulative loss of $116 billion compared to its pre-pandemic trajectory (Figure 1).

 

 

 

Figure 1. Australia’s GDP, actual and pre-COVID trajectory

Source: ABS 2023a and Treasury

The COVID pandemic rocked the demographic boat.

As shown in this chart, Australia’s population growth slowed to 0.1 per cent in 2020–21 – the lowest rate in more than 100 years (Figure 2).

 

Figure 2. Australia’s population growth and components of growth, historical and projected

Source: ABS 2023b and Treasury

Australia’s net overseas migration fell into negative territory for the first time since the end of WWII, with a net loss of 85,000 people in 2020–21 (ABS 2023c).

There was a larger than expected increase in deaths due to COVID and other causes – 10.9 per cent above what was expected in 2022 (ABS 2023d).

Births fluctuated in interesting ways. In the December 2020 quarter, nine months after the pandemic hit, births fell. We know that crises can make couples cautious about starting a family, and this drop likely reflected the uncertainty that many couples felt in the early months of the COVID lockdowns.

But then things turned around. In the March 2021 quarter – nine months after mid-2020, births hit an all-time high (ABS 2023c). We can’t be sure as to why this occurred, but it may be that couples felt a little less anxious about the future as the year unfolded. Lockdown boredom may also have been a factor. Border closures between states and territories reduced internal mobility. The number of interstate moves in the year to March 2021 was 30.2 per cent lower than in 2018-19 (ABS 2023c).

The pandemic also influenced where people wanted to live with an increase in net moves from urban areas towards suburban and regional areas (Figure 3) (Centre for Population 2020).

COVID doubled the net number of people moving to the regions.

You can see a break in the data series on this chart. It’s due to the pandemic’s impact on Medicare address information this series relied on, which I will get to later.

 

 

Figure 3. Net internal migration to regions outside capital cities, quarterly

Source: ABS 2023c and ABS 2021d

 

And in 2020–21, regional areas grew faster than capital cities for the first time since 1993–94 (ABS 2023e) (Figure 4).

Figure 4. Combined capital city and rest of state population growth, annual

Source: ABS 2023e and ABS 2019

Quite frankly, we needed lots of sourdough, lots of quarantinis and lots of data to make sense of it all.

Putting the data spotlights on

The data challenge for government agencies during the pandemic was enormous.

Data had to be timely to respond to a rapidly changing environment.

It had to be delivered at a higher frequency.

And data had to be continuous and granular to help us understand the different things happening in different states and territories.

Treasury recognised early on that existing data assets weren’t going to be sufficient for handling the speed and scale of dual health and economic shocks (Hambur, Montaigne, Parsons and Whalan 2022).

A Treasury paper published in 2022 likened using existing data during COVID to the ‘streetlight effect’ where a drunk only searches for their lost keys under the lamp post, because that’s where the light is best.

When it came to forecasting during the pandemic, historical data – traditionally the foundation for models – suddenly seemed inadequate.

The usual patterns and trends, which underpinned models, were disrupted, making usual methodologies less reliable.

Then there was the unpredictable nature of the virus.

New strains emerged, with different levels of transmissibility and severity, complicating predictive models.

The only plausible option involved shining new light on the rapidly evolving economic and demographic situation using new sources of data.

Faster insights

By mid-March 2020 – about the same time toilet paper madness was unrolling – the ABS already had introduced a range of COVID related products.

This included preliminary data as well as smaller, rapid-response surveys to deliver more timely insights (ABS 2021a, ABS n.d).

The ABS introduced the Business Impacts of COVID-19 (now named Business Conditions and Sentiments) and the Household Impacts of COVID-19 rapid-response surveys.

These surveys fast-tracked the time between the field and publication. The Business Impacts of COVID-19 survey provided information with a delay of 7 days from interviewing to publication, while the Household Impacts of COVID-19 was released around 17 days after the end of the collection period.

Alternate sources

The ABS also made greater use of alternate data sources such as non-survey administrative data and transactions data to develop indicators (ABS 2021a, ABS n.d).

In one example, the ABS partnered with other agencies to provide new data through two integrated longitudinal assets – the Person Level Integrated Data Asset (PLIDA, formerly known as, MADIP) and the Business Longitudinal Analysis Data Environment (BLADE).

These data assets proved valuable because they could be updated frequently, making quicker and more sophisticated analysis possible. 

The PLIDA and BLADE core data was linked providing new insights into the interactions between businesses and their employees (Gruen 2022a).

These insights allowed governments and policymakers to determine COVID’s impact on the business community and examine the economic recovery including employment patterns.

Data from the Australian Immunisation Register (AIR) are also linked to PLIDA each week and Provisional Death Registrations data is linked and updated monthly.

During the pandemic, the Department of Health and Aged Care used this data to generate insights for Australia’s COVID Vaccine and Treatment Strategy. State health departments and primary health networks are also data users.

New sources

The pandemic also saw the emergence of new official data sources.

It made practical sense for the ABS to draw on data already generated in our increasingly digital world (Gruen 2022a). 

It makes even more sense when that source provided previously unavailable statistics, faster and with greater coverage than before.

The Single-Touch Payroll is a prime example.

From April 2020, the ABS worked with the Australian Taxation Office to receive expedited access to the high-frequency Single-Touch Payroll data. 

This access was significant for a few reasons.

First, it provided new and almost real-time insights into the labour market based on more than 10 million employees in Australia (Gruen 2022a).

The teamwork resulted in the Weekly Payroll Jobs publication and the more recent Monthly Employee Earnings Indicator publication.

Second, the access enabled detailed disaggregation including geospatial analysis which just wasn’t possible from the existing monthly Labour Force Survey.

And third, it filled a major void.

Treasury’s data experts said during the COVID pandemic the ‘unemployment rate was a less informative measure of labour market health than usual (because many workers had been stood down temporarily or were working reduced hours) and other indicators such as hours worked gained more prominence’ (Hambur, Montaigne, Parsons and Whalan 2022).

They said, ‘the timeliness and broad coverage of Single Touch Payroll made it one of the most valuable datasets in assessing the health of the labour market in real time… and in modelling fiscal policy response options.’

Family and health

The ABS wasn’t the only government agency stepping up to the challenges of measuring COVID’s impact.

The data work of health departments is a topic for another time but I want to acknowledge a couple of other great data examples.

The Australian Institute of Family Studies introduced the Families in Australia Survey series which ran from May 2020 to December 2021.

The survey series explored how families’ lives changed during the pandemic, covering topics such as changes to workforce participation, work patterns, child care, household forms and family supports.

The Australian Institute of Health and Welfare established the COVID-19 Register – a national and linked data set.

The Register combines de-identified COVID case data with other data sets, including immunisation, hospitals and deaths data.

Importantly, it will be updated as more data become available, allowing a long-term view of the effects of COVID on health outcomes and health service usage.

Contribution from the Australian National University

Today I’ve focused on the government side of things, but I want to acknowledge this university for its research and data collection during COVID. This includes the COVID‑19 Impact Monitoring series, which looked at the impact of COVID on wellbeing and gambling participation, as well as research on topics like long‑COVID, changes in volunteering and the determinants of vaccine uptake (ANU n.d.).

When you combine all the data and research, there are many more lamp posts in the data park to help people find their lost keys.

The realm of demographic data in Australia

Mortality

Prior to COVID, summary mortality information was reported with a 6-month delay from when the death was registered to publication (ABS 2023j). The ABS expedited its mortality data reporting to capture the immediate and detailed COVID impacts.

The reporting became essential for understanding COVID’s direct impact and its indirect consequences on the population.

In June 2020, the ABS released the first Provisional Mortality Statistics report to account for the beginning of the pandemic (ABS 2020).

The ABS continued to produce the report monthly as a way of tracking mortality through the pandemic. This meant that deaths were now reported 3 months after they occurred and on a more frequent and detailed basis.

So the consequence of COVID has been to reduce the delay in deaths reporting and support the tracking of mortality during the pandemic and beyond.

These reports showed that the pandemic could impact mortality in unexpected ways.

For example, in 2022, we saw deaths due to other causes rise, highlighting some of the direct and indirect impacts of the pandemic on our health (ABS 2023f).

Measuring excess mortality

The ABS released an article in July on measuring Australia’s excess mortality during the COVID pandemic (ABS 2023d).

The article provides excess mortality estimates for Australia and the states and territories from 2020 to March 2023.

You can see the variations across the states over the Omicron period especially, with Western Australia recording the lowest and Tasmania the highest in this map (Figure 5).

 

 

Figure 5. Excess mortality (%), Omicron period (January 2022 – March 2023)

Source: ABS 2023d

The ABS found that excess mortality was 9.1 per cent above the expected number of deaths in the first quarter of 2023.

In 2022, deaths were 10.9 per cent above expected.

This contrasted with the years 2020 and 2021, when Australia had relatively few COVID cases, and significant mobility restrictions. Analysing ABS mortality data for those years, Oxford’s Philip Clarke and I published research in BMJ: Global Health that studied the impact of lockdowns on mortality (Clarke and Leigh 2022).

We found the mortality rate in Australia was 5.9 per cent lower in 2020–2021 than in 2015–2019.

Contrary to those who claimed lockdowns would increase the death rate, we found lockdowns and social distancing reduced short‑term mortality.

Breaking down mortality

The ABS has published other articles on COVID mortality so we can see how it has been distributed across population groups (ABS 2023h).

The articles have presented COVID mortality data by demographic characteristics. 

This breakdown helps us track the incidence of direct COVID mortality and indirect mortality, and offered insights into demographics variations.

Delivering the W.D. Borrie Lecture last year, I explored some of the health inequalities created by the COVID pandemic. It turns out that although Britain had a higher level of COVID mortality than Australia, the socioeconomic patterns of COVID mortality in the two countries are very similar (Leigh 2022).

Births

Collecting some kinds of demographic data became a lot harder during COVID.

Registration delays in the reporting of births data were especially pronounced in some places during the pandemic as a result of activity restrictions and lockdowns (Centre for Population 2021).

During the pandemic, the Centre for Population – housed in the Australian Treasury – was able to use Medicare newborn entrants data and Medicare pregnancy services data from Services Australia.

This data provided a proxy for births to inform fertility assumptions and improve population projections during the pandemic and beyond.

Migration trends

Migration trends were a major focus during the pandemic.

With international travel restrictions, the ABS provided more detailed and timely data on returning Australians, international students, and the implications for population growth.

For example, additional information was included in Overseas Arrivals and Departures (OAD) to provide more detail on the numbers of arrivals and departures to and from Australia (Gruen 2020).

Additional information included analysis of short-term visitor arrivals and a breakdown of international students for each state and territory.

In response to heightened interest in traveller data, the ABS also created the temporary Overseas Travel Statistics provisional publication (ABS 2021b).

This data was effective in giving governments earlier monthly preliminary information on international travel using border crossing data.

As a result, information for the reference month was available four weeks earlier than what was available in the Overseas Arrivals and Departures release (ABS 2021a).

Given the value of this information, provisional travel data was added to the existing Overseas Arrivals and Departures release from August 2021.

The data give us a dramatic picture (Figure 6).

As shown in this chart, from 20 March 2020, almost all overseas travel was banned, as the WHO declared the COVID outbreak a public health emergency.

The pandemic disrupted international travel with volumes at a record low in 2021.

The Government began to ease travel restrictions in late 2021.

International travel in Australia has seen a resurgence since early 2022, with over 19 million arrivals and departures combined across 2022.

Figure 6. Total overseas arrivals and departures, monthly

 

 

Source: ABS 2023g

The ABS also introduced more frequent revisions of net overseas migration estimates (Gruen 2022b).

This improvement was about providing the most up-to-date information and enabled adjustments to the preliminary model to account for changed traveller behaviour.

Internal movements

As I mentioned earlier, internal movements were fascinating from a demographic point of view.

With borders closing and movement restrictions in place, the ABS quickly adjusted its approach to monitoring internal population dynamics, tracking movements between states as well as urban and rural divides.

COVID challenged traditional understandings of internal migration, not to mention forecasting the distribution of the population and the flows of people between areas.

Remember the demographic boat was rocking – things were still happening fast when the ABS introduced the Provisional Regional Internal Migration Estimates (PRIME) in November 2020 (presented in Figure 3).

These estimates were a way to provide timely insights into internal migration in response to the pandemic and the heightened interest in internal migration data.

However, the publication was suspended after the March 2021 release due to the impact of the COVID vaccination program on Medicare data, which is the primary source for internal migration estimates.

The vaccination program saw the number of people updating their Medicare address data go through the roof – reflected in the number of interstate moves jumping to a series record 164,000 in September 2021 while there were lockdowns and border restrictions in place (Figure 7).

The pandemic also posed some challenges to the approach taken for finalising internal migration estimates from the 2021 Census.

The issue with the Medicare data led the ABS to declare that movements over the second half of 2021 were not reflective of real‑world internal migration.

In addition, the Census data was also affected by lockdown policies in some jurisdictions.

The ABS responded to these challenges by giving greater weight to the Census and Medicare data that was unaffected by these impacts.

This included discounting the Medicare data over 2020–21 in favour of the Census-based estimates and giving more weight to the earlier (2016) Census data than usual.

Figure 7. Interstate movements (based on Medicare address changes), quarterly

Source: ABS 2023c

Looking ahead: demographic data post-COVID

The data legacy is a positive, and now we have the opportunity to act on the lessons learnt.

Life Course Data

COVID demonstrated the potential for more data integration. The Life Course Data initiative, announced in this year’s budget, will connect administrative datasets into a linked longitudinal data asset to support evidence-based policy making at the community level.

Linking data across different subject-matter areas and levels of government will bring many benefits.

It will enable more comprehensive analysis of the characteristics, programs and service interactions of individuals, households and families (Gruen 2023). 

Census 2026

The ABS is exploring new ways to enhance the Census using administrative data.

The 2021 Census was supported by integrated administrative data that helped in assuring its quality and led to more accurate area population counts (ABS 2023i).

More recently, new income and benefits data added to the 2021 Census for its final data release supported new analysis and insights on Australian pensioners.

Work to understand administrative data has led to the release of a Census-like population and housing data snapshots.

These experimental datasets have been released as stand-alone products and over time are expected to provide additional information on sub-national population and internal migration patterns.

Closing remarks

The COVID pandemic delivered lasting change to Australia’s statistical systems.

• Before the pandemic, deaths were reported 6 months after the end of the reporting period. Now, they are reported with a 3‑month delay.
• Before the pandemic, the delay in reporting overseas arrivals and departures was 6 weeks. Now, provisional data is available in only 2 weeks.
• Before the pandemic, Australia was one of only two OECD countries (the other being New Zealand) to report inflation on a quarterly basis. Now, we have a monthly CPI indicator.
• As a result of the pandemic, the ABS created the Weekly Payroll Jobs publication and the Monthly Employee Earnings Indicator publication.

Yet there may be more to be done. Right now, the reporting delay is 2 weeks for unemployment figures, 3 and a half weeks for inflation,[1] and just over 2 months for national income accounts and  house price data.

By contrast, share market data are available in real time. And in the latest issue of the Australian Economic Review, a series of papers explored innovations in ‘nowcasting’.

The Queensland Treasury Corporation (QTC) uses a bunch of data from Bloomberg and Macrobond, as well as the Citi Australian Data Change Index and the QTC-constructed Financial Conditions Index to provide daily insights into key indicators of the Australian economy including inflation, GDP growth and the unemployment rate (Anthonisz 2023).

Looking locally, de Silva, Yanotti, Sinclair and Angelopoulos find unconventional real-time data such as pedestrian counts could be good high-frequency predictors of local economies, such as suburbs or central business districts (CBDs) (de Silva, Yanotti, Sinclair and Angelopoulos 2023).

And finally, in response to rising house prices, Shi and Phillips created a real-time “thermometer” of “housing fever” in Australia and New Zealand. Using data on mortgage interest rates, rents, disposable income, employment and housing supply the authors diagnose in real-time bubbles as they begin to emerge (Shi and Phillips 2023).

These intriguing examples show what’s possible using new data sources and machine learning algorithms. It is vital that government continues to watch developments in this space. In producing statistics, there is always a trade-off between quality and timeliness. In another crisis, policymakers may want data more quickly, even if it comes at the cost of perfection.

The Government sees data as a big part of building a better future – we intend on releasing our Data and Digital Government Strategy by the end of the year.

Demographic data is like a guidebook for government. It informs policies, directs resource allocation, and underpins strategic planning.

As communities grow, shrink, age or evolve, it’s this data that offers us direction – it tells a government where to build schools, where hospitals are needed or where jobs might be created.

When unexpected things happen, like the COVID pandemic, it’s also this guidebook that helps governments make good decisions.

COVID left a lasting data legacy. I thank those of you who contributed to it.

References

Anthonisz M 2023 Nowcasting Key Australian Macroeconomic Variables, Australian Economic Review, vol. 56, no. 3.

Australian Bureau of Statistics (ABS) 2023a Australian National Accounts: National Income, Expenditure and Product, June 2023 [Released 6 September 2023].

Australian Bureau of Statistics (ABS) 2023b National, state and territory population, September 2022 [Released 16 March 2023].

Australian Bureau of Statistics (ABS) 2023c National, state and territory population, March 2023 [Released 14 September 2023].

Australian Bureau of Statistics (ABS) 2023d Measuring Australia's excess mortality during the COVID-19 pandemic until the first quarter 2023 [Released 19 July 2023].

Australian Bureau of Statistics (ABS) 2023e Regional population, 2021-22 [Released 20 April 2023].

Australian Bureau of Statistics (ABS) 2023f Provisional Mortality Statistics, Jan-May 2023 [Released 25 August 2023].

Australian Bureau of Statistics (ABS) 2023g Overseas Arrivals and Departures, Australia, July 2023 [Released 12 September 2023].

Australian Bureau of Statistics (ABS) 2023h COVID-19 Mortality in Australia: Deaths registered until 31 July 2023 [Released 25 August 2023].

Australian Bureau of Statistics (ABS) 2023i How administrative data improved the quality of the 2021 Census [Released 20 January 2023].

Australian Bureau of Statistics (ABS) 2023j National, state and territory population methodology, March 2023 [Released 14 September 2023].

Australian Bureau of Statistics (ABS) 2022a Effects of COVID-19 strains on the Australian economy [Released 1 June 2022].

Australian Bureau of Statistics (ABS) 2022b Economic gains and losses over the COVID-19 pandemic [Released 7 September 2022].

Australian Bureau of Statistics (ABS) 2021a Changes to ABS statistical work program [Media Release issues 20 April 2021].

Australian Bureau of Statistics (ABS) 2021b Overseas Travel Statistics, Provisional [Released 14 July 2021].

Australian Bureau of Statistics (ABS) 2021c Investigation into application-based positioning data for COVID-19 1504.0 - Methodological News, Mar 2021 [Released 29 March 2021].

Australian Bureau of Statistics (ABS) 2021d Regional internal migration estimates, provisional, March 2021 [Released 3 August 2021].

Australian Bureau of Statistics (ABS) n.d ABS responds to COVID-19 [Internet content accessed September 2023].

Australian Bureau of Statistics (ABS) 2020 Provisional Mortality Statistics, Jan-Mar 2020 [Released 24 June 2020].

Australian Bureau of Statistics (ABS) 2019 Historical population, 2016 [Released 18 April 2019].

Australian National University (ANU) n.d. COVID-19 publications [Internet content accessed September 2023].

Centre for Population 2021 2021 Population Statement Treasury.

Centre for Population 2020 Migration between cities and regions: A quick guide to COVID-19 impacts Treasury.

Clarke P and Leigh A 2022 Understanding the impact of lockdowns on short‑term excess mortality in Australia, BMJ Global Health, vol. 7, article e009032.

de Silva A, Yanotti M, Sinclair S and Angelopoulos S 2023 Place-Based Policies and Nowcasting, Australian Economic Review, vol. 56, no. 3.

Gruen D 2023 Supporting Analysis of The Life Course [Speech to the Life Course Centre Data for Policy Summit 15 August 2023].

Gruen D 2022a Realising the potential of data in government [Speech to the Institute of Public Administration Australia 15 June 2022].

Gruen D 2022b 20th A tale of two pandemics in Australia: 1918-19 Influenza and COVID-19 [Speech to the 20^th Australian Population Association Conference 23 November 2022].

Gruen D 2020 Measuring the Impacts of COVID-19 [Speech to the Australian Business Economists 30 April 2020].

Hambur J, Montaigne M, Parsons S and Whalan E 2022 Looking under the lamp post or shining a new light: New data for unseen challenges Treasury.

Kennedy S 2022 The 2022 Sir Leslie Melville Lecture [Speech delivered to the Australian National University on 27 July 2022], Treasury.

Leigh A 2022 Health inequalities in the COVID pandemic: evidence from Australia, Australian Population Studies, vol. 6, no. 2.

Shi S and Phillips P C B 2023 Housing Fever in Australia 2020-23: Insights from an Econometric Thermometer, Australian Economic Review, vol. 56, no. 3.

END

 

[1] Refers to quarterly CPI not monthly indicator of inflation.