IPAA National Conference: Enhancing Service Delivery at the Australian Bureau of Statistics

Dr David Gruen
Australian Statistician
Thursday 30th November 2021
IPAA National Conference

Introduction

Thanks to IPAA for inviting me to speak at the National Conference. It’s always a pleasure to address my peers across the public service. It’s important to gather to share our stories, albeit virtually, after such a tough two years. I’m also pleased to be presenting alongside Greg Duncan from Australia Post. The ABS and AusPost have a long relationship, strengthened by our collaboration on the 2021 Census.

Speaking of long relationships, let me acknowledge the Traditional Custodians of the land upon which we’re meeting today. For me in Canberra, that is the Ngunnawal people and I pay my respects to their Elders past and present and extend that respect to members of the Aboriginal and Torres Strait Islander community watching this presentation today.

Let me begin with a confession. Nothing racy, mind you, but when I first saw the topic for today’s session: Connection and Collaboration for Service Transformation, I was a bit stumped.

I don’t normally think of the ABS as a service provider. We don’t administer payments like Centrelink, we don’t facilitate support for people with disability like the NDIA, and we don’t deliver mail and packages like Australia Post.

But while the ABS is not a service provider per se, we do transact with hundreds of people and businesses every day through our survey program. In August this year, we interacted with millions of people completing their Census.

These interactions are critical to achieving our purpose of informing Australia’s important decisions. We need people to provide us with information to create statistics. Achieving high response rates on our surveys and the Census enables us to produce statistics that accurately represent the Australian population, economy, society, or environment.

As we all know though, people are busy; some people don’t know what the ABS does or why responding to a survey is important; and it is getting harder to cut through the large volume of information (and increasingly misinformation) that confronts us all every day.

So, how do we, at the ABS, encourage people to interact with us and provide their personal and business information? As Australian Statistician, I have legal powers under the Census and Statistics Act to compel people to “answer questions” and “fill up forms”, but that’s a big stick we don’t use unless we have to.

We much prefer to foster willing cooperation with respondents, so we are putting considerable effort into making interacting with the ABS easier. We deploy a range of strategies and techniques to do this, including:

  • using digital technology,
  • accessing existing data from other agencies, and increasingly the private sector,
  • collaborating with experts and not trying to do it all on our own, and
  • consulting with clients, respondents, and the community.

I’ll talk about three examples today to elucidate how these things work in practice:

  • The 2021 Census,
  • Innovation in our survey program, and
  • Better use of existing data and big data.

The 2021 Census

Unless you were living under a rock or were overseas a couple of months ago – which I strongly suspect you weren’t! – you’ll know we ran the 18th national Census in August.

With the outage of the 2016 online Census on our minds, we set ourselves three high-level objectives for the 2021 Census: to be smooth running; well-supported; and result in high-quality data.

Smooth Running

By “smooth running” we wanted it to be secure and quick and easy for people to complete their Census.

We worked with PwC and Amazon Web Services to build a new Census Digital Service from the ground up. The Digital Service included the online Census form, the Census website, and back-end systems that allowed us to monitor what was happening as forms were submitted.

We worked closely the Australian Cyber Security Centre and other experts to ensure the Digital Service had robust cyber security protections.

We used user-centred design to develop the Digital Service. Most users these days expect to be able to complete tasks and/or access services online. So, the Digital Service had a range of self-service options. Online, people could:

  • request a paper form if they didn’t want to use the online form,
  • report they wouldn’t be at home on Census night,
  • control their password and request a password reset, and
  • login without a code – households that hadn’t received their instructions with a login code in the mail could complete and submit their Census form using a PIN sent to their mobile phone.

Login without a code turned out to be extremely popular – it was used by 1.5 million households to submit their Census forms and was an innovation that many people found extremely helpful. An obvious lesson for 2026. ¹

Well-supported

Our second objective was to have governments, businesses, and the community have confidence in the Census and to achieve a high level of community participation. To achieve this objective, we used extensive and sophisticated communication and engagement strategies, and drew on expert assistance.

Privacy Impact Assessments

A completely understandable key area of concern for the community is protection of personal information. We recognise and respect this concern, and took a privacy by design approach for the Census, which included commissioning two independent Privacy Impact Assessments (PIAs).² One was an assessment of the privacy settings of the Census overall, and the other looked at the use of administrative data as a contingency (that is, if we didn’t achieve the required response rate, we may have had to use other data sources to fill gaps). We agreed to the recommendations from these PIAs and published them and our responses on our website in July 2020, over a year before the Census.

Engagement Strategies

We used a range of engagement strategies to reach as many people as possible, including migrant communities, Aboriginal and Torres Strait Islander peoples, people experiencing homelessness, people in aged care, people travelling, defence personnel, FIFO workers, and people in quarantine (including athletes returning from the Tokyo Olympics!). We worked with many community groups, asking them to reach out via their existing networks.

We provided resources to help people complete the Census, including:

  • instruction information in multiple languages;
  • face-to-face support, for example, fill-in-the-form sessions (in person and on our Facebook page);
  • pop-up hubs where possible in a COVID-safe way;
  • braille and large print forms;
  • instructional videos and MP3 audio recordings of questions on the website; and
  • a short form for people experiencing homelessness.³

COVID-Safe Operations

The health and safety of the community and our staff during the Census was paramount. We had a Census COVID-Safe plan, which we tested in late 2020 when we ran a major Census test in different parts of Australia. We adapted our plans and data collection activities with the frequently changing COVID-19 arrangements and restrictions. So, for example, we got special permission from some state health authorities to have Census field officers enter hospitals or aged care facilities.

Collaboration with Experts

We have always worked with others to deliver the Census, but this time the collaboration was much more extensive. We worked with more than 50 public and private sector partners. As well as the ones already mentioned:

  • Australia Post delivered 8.3 million instruction letters, 2.7 million reminder letters, and returned 2.5 million paper forms to our processing centre in Dandenong, Victoria;
  • Services Australia ran our contact centre and fielded 650,000 calls over the data collection period, and responded to 24,000 general and technical emails;
  • KPMG provided independent assurance of the whole Census program; and
  • Adecco managed recruitment of our Census field staff.

Given the multifaceted nature of a modern Census, it would have been folly for the ABS to attempt to manage all aspects on our own. Collaboration is a two-way street, though, and our partners benefitted from the experience of working on the Census too. We needed their expertise, and they learnt from us because many of them hadn’t been involved in something as large and complex as the Census before. Our partners were motivated by the challenge and wanted to be part of something of major significance and public good.

High quality data

Our third objective was to produce high quality data. We’re still in the processing phase but we’re confident of exceeding our target response rate of 95% which bodes well.

We have established an independent assurance panel, to assess aspects of the quality of Census outputs, having considered issues related to Census design, enumeration, and processing, using the quality of outputs from the 2011 and 2016 Censuses as benchmarks. The panel will publish a report in June 2022 to coincide with the first release of Census data so governments, the community and other stakeholders can make informed judgments about the fitness-of-purpose of 2021 Census data.

Let me also provide some numbers to highlight the scale of the exercise:

  • We received 10.1 million responses, equating to approximately 25.6m people. 
  • About 7.7 million (or more than three-quarters) of those responses were received electronically.
  • There were 2 million Self Service Requests online; 4 million end of form feedback messages (most of which were positive); 19 million website visitors, and 195 million webpage hits.
  • During the peak time on Census night, the online form received 142 online submissions per second with a peak of 249 logins per second.
  • In the 24-hour period on Census day (10 August) the service scaled to enable 2.5 million households to submit forms.
  • 130,000 malicious IP addresses were blocked. In total, there were nearly 1 billion attempted cyber-attacks, which were all repelled.
  • For the whole period the Digital Service was up and running there were no service interruptions or loss of availability.

These results are all the more pleasing because we achieved them while more than half the population, including our two most populous states, were in lockdown, meaning we could not conduct our follow up field work there in the usual way by knocking on doors to remind those who had not yet completed the Census to do so.

We are all of course very pleased with how smoothly the Census went, and that the community embraced it, but it does take years of planning and preparation. Planning for the 2021 Census started before the 2016 Census, and we have already started planning for the 2026 Census. Building and maintaining relationships with partners and community groups is an ongoing endeavour.

Innovation in our survey program

Let me turn to innovation in our survey program. As well as the five-yearly Census, the ABS conducts hundreds of household and business surveys every year. We seek to continually improve how we engage to reduce the burden we place on respondents and to achieve high response rates. Over the last couple of years, we have been trying different things.

I’ll tell you about one important example.

Each month, the ABS recruits around 3,500 households into the Monthly Population Survey. In total we have about 28,000 households in the survey at any one time. Key outputs from the survey are Labour Force statistics, including the employment, unemployment, and labour force participation rates.

People in the survey are required to complete the survey monthly for 8 months. Our normal process is to have an interviewer call the household and ask a series of questions every month to record the employment circumstances of the people in the household. For example, in the first month of the survey a member of the household may have been unemployed, but by the third month they may have found a job; or they may have been working part time and changed to full time, etc.

However, interviewer collection like this is expensive, and not always convenient for the respondent, so we have been using behavioural insights to:

  • increase digital adoption,
  • improve engagement with respondents,
  • increase response, and
  • reduce costs.

Subtle changes to what we tell survey respondents, as well as when and how we tell them, improves engagement. For example, when a household is first selected in a survey, we send them an approach letter. We don’t know who lives at each dwelling, so our previous practice was to address the envelopes to ‘The Resident’. We found putting an Obligation ID number on the envelope instead of ‘The Resident’ increased response. We also found that shortening and simplifying the letters encouraged more people to fill in the survey online. Sending a reminder letter in the leadup to each new monthly survey period encouraged more people to register. These relatively small and easy changes have made a big difference.

We’ve also had to change our survey operations due to COVID-19. With restrictions on movements, we couldn’t send interviewers to people’s homes to conduct surveys or follow-up non-responding households, but we still needed information from these households. Arguably, it was more important to find out what was happening to people during the pandemic than in normal times.

We encouraged respondents to complete online or over the phone. There were postal delays so our interviewers delivered call back letters or cards with hand-written details to letterboxes or doorsteps where they could. We also used White Pages and Yellow Pages data, and the Individual Phone Number Database from the Australian Communication and Media Authority to follow up non-responding households. This was the first time we’d used these data sources. We’d been seeking access to them for some time, but the pandemic provided extra impetus. It’s a good example of how COVID-19 threw up both challenges and opportunities.

Better use of existing data and big data

That leads me to the last example I want to talk about: how we’ve been using existing data to create statistics and insights.

When the pandemic hit Australia early last year, we recognised our unique position to generate new, near real-time information about the economic and social impacts of the pandemic. To produce more timely information, we accelerated the modernisation of our data collection activities and sought alternative data sources, particularly big data, from both public and private providers.

We received Single Touch Payroll (STP) data from the Australian Tax Office (ATO), which allowed us to release fortnightly information on the jobs and wages of over ten million employees across Australia, published with a 17-day lag. This complements the Labour Force Survey (which surveys around 50,000 people monthly) and provides a welcome addition to labour market information during COVID-19.

Single Touch Payroll was a data source we’d been seeking to access for some time, but the arrival of the pandemic raised the stakes and expedited the process.

We also used de-identified, aggregate credit/debit card and payments data from major banks to quickly understand changes in household spending; and transactions data from large retailers to understand the products households were purchasing during the pandemic.

We used our evolving suite of cloud-based ICT services to acquire, store, and analyse Single Touch Payroll, bank, and retail transactions data. Cloud-based services can handle these big datasets, support modern programming languages and analytical techniques, and provide robust data protection controls. It was not that long ago that we did not have systems that could meet these requirements. Cloud technology in particular has opened up important new opportunities for us.

Building on the success of our statistical response to COVID-19, we’re continuing to use big data and cloud technology to develop new monthly indicators to inform economic policy. These new indicators are:

  • Business Turnover using business activity statements data,
  • Household Spending using bank credit/debit card and payments data, and transactions data from large Australian retailers, and
  • Employee Earnings using an expanded range of Single Touch Payroll data.

We released the first iteration of the Monthly Business Turnover Indicator in mid-October, which demonstrated the impacts on businesses of lockdowns during preceding months. We are on track to begin releasing the household spending indicator in the first half of next year.

These indicators will provide governments with a clearer picture of what is happening in the economy and society in closer to real-time, so they can make more timely policy decisions and provide more timely services. The other advantage of accessing big data sources is to reduce the reporting burden on businesses and households. Public service agencies are increasingly moving to a “collect once, use many times” model, which is both effective and efficient.

Conclusion

I’ve given you an overview of some of the things we’ve been doing at the ABS to improve our service offerings. Some are completely new; others are improved ways of doing traditional things. Resources are finite, and it’s the public’s money we are spending, so it is incumbent on us all to continuously review and improve our operating models and ways of doing things.

I’ve talked about nifty IT solutions, such as cloud services, but technology is an enabler, not a driver. Technology can bring ideas to life, but you need the good ideas first, and they come from people.

I’m grateful and honoured to lead the talented, dedicated, and professional public servants at the ABS. They rose to the challenges of continuing to deliver our usual statistical work program, creating new insights, and conducting a Census in the face of COVID-19. And they continue to innovate and find new ways of doing things every day.

My final advice is, if we in public institutions want to meet challenges, take advantage of opportunities, and achieve our objectives, we need to deploy multiple strategies; be prepared innovate and try different ways of doing things; work to genuinely collaborate with partners across public, private, and community sectors; make the effort to understand community expectations and strive to meet them (and if you can't, explain why); and take care of people. Some people find change and working on new things exhilarating and energising. Some find it exhausting. As leaders we need to know which are which, and support all of them.

Footnotes - show all

  1. The online form also had in-built smart sequencing (that is, it responded and adapted as people answered the questions) to reduce the time to complete it.
  2. The PIAs were undertaken by Information Integrity Solutions for the use of administrative data and Galexia for the overall privacy settings for the Census. The recommendations included developing a longer-term privacy strategy for Census, undertaking thorough security testing, reducing the periods for which we keep name and address information after processing, and clarifying the thresholds and triggers for using other data as contingency in the 2021 Census.
  3. The Census website and online form met WCAG 2.0 Level AA online accessibility standards and was compatible with screen readers and audio assistance services. It won ‘Government Website of the Year’ at the 2021 Australian Accessibility Awards.