Preparations of the 2016 census

3.1        The 2016 census was held in August, but planning had begun five years earlier in 2011 when the ABS agreed to a set of strategic priorities:

• maximise the count of every dwelling and person in Australia;
• maximise the value of the census to all users;
• protect the privacy of the public; and
• increase the efficiency and sustainability of the census.  

Early planning for the 2016 census

3.2        Preparation for the 2016 census commenced in June 2011 with the establishment of the 2016 census team within the ABS. In July 2012, the ABS held discussions with the Australian Government (government) on the census business case. This culminated in funding being allocated in the May 2013 budget for the census.

This provided the required funding for the next four years to develop and conduct a transformed national Census. The ABS considers that it had adequate funding to conduct a high quality Census in 2016 and sufficient time to properly prepare for the Census.^[1]

3.3        In May 2015, the government budget included investment in the Statistical Business Transformation Project and affirmation of the 2016 online census.^[2]

Budget problems and suspended preparations

3.4        The 2013-14 Australian Government Budget allocated the necessary funding to develop and conduct a transformed national census which the ABS reported was 'adequate funding to conduct a high quality Census'.^[3]

3.5        In February 2015, there were reports in the media that the ABS had considered cancelling the 2016 census and moving to decadal censuses.^[4] The Australian Statistician confirmed later that month that the ABS was considering changing how the census was run, but that as the government was still considering that proposal, he was not in a position to comment outside of noting: 'We have provided some proposals to government around changes to the frequency of the census'.^[5]

3.6        IBM's submission to the inquiry informed the committee:

In or about February 2015, the ABS informed IBM that it was considering not proceeding with the 2016 eCensus (or, indeed, any Census in 2016). IBM understands that the ABS was considering decreasing the frequency of the Census to once every 10 years and running a rolling Australian Population Survey (APS) during the intercensal period.

...

In May 2015, the ABS informed IBM that the existing 5 yearly Census frequency would be maintained and the 2016 eCensus would proceed, with Census Day to be 9 August 2016 as originally planned.^[6]

3.7        The ABS' lengthy and otherwise comprehensive submission to this inquiry makes no mention of these events.

3.8        Census preparations were further complicated by the departure of Australian Statistician Brian Pink in January 2014, with his permanent replacement, Mr David Kalisch, appointed almost a year later in December 2014.

3.9        The period of September 2013 through to August 2016 also saw four different ministers (the Hon Steven Ciobo MP, the Hon Alex Hawke MP, the Hon Kelly O’Dwyer MP and the Hon Michael McCormack MP) take responsibility for the Australian Bureau of Statistics.

3.10      The titles for the responsible minister also changed during this time, and included “Parliamentary Secretary to the Treasurer”, “Assistant Minister to the Treasurer”, “Parliamentary Secretary to the Treasurer” and “Minister for Small Business”.

Scope of questions

3.11      The number of topics included in the census has steadily increased over time. The first census had 19 topics; the majority of which have remained part of every census since. Topics are removed from the census when there is no longer sufficient justification for their inclusion. Blindness/deaf-mutism, for instance, was last collected in 1933.^[7] Since 2006 there have been no changes to the 61 included topics.^[8]

3.12      The ABS sought public submissions on both the nature and content of the 2016 census from November 2012 to May 2013.^[9] This consultation formed part of the ABS' attempt to optimise the 2016 census to ensure that the data collected through the census was of a high quality and relevant in contemporary Australia.^[10] The ABS' submission summarised the results of this consultation:

Over 1000 submissions were received. In these submissions, 54 new topics were suggested and all 61 of the topics in the 2011 Census received strong support for retention. The ABS had more than 80 meetings with key external stakeholders including Australian Government departments, state/territory departments, statistical advisory groups and selected interest groups.^[11]

3.13      The ABS reported that there is significant demand for additional topics to be included in the census, but these demands are balanced against economic and methodological considerations:

There is significant demand for additional topics, such as information on long term health conditions. The inclusion of any new topic needs to consider the significant cost, the burden on households, the limitations of the size of the paper form, and the continuing need for other existing topics.^[12]

3.14      The committee heard concerns that 'there is a general perception that the questions on the census form are out of date and some are becoming irrelevant in the modern era. There is also a need for new questions to be asked'.^[13] Some suggested topics for inclusion included questions about health status, disability, use of recreation time, modes of travel to education and other activities, pet ownership and part-time family arrangements.^[14]

3.15      In contrast, the Institute of Public Affairs noted that much of the information that is included in the census is more efficiently collected elsewhere, and concluded that:

If the Census is to continue into the future, its scope and scale should be heavily scrutinised, and the questions it asks should be restricted only to those questions which can be demonstrated to be absolutely necessary.^[15]

3.16      In August 2015, the ABS published the 'Census of Population & Housing: Nature and Content, Australia, 2016' which revealed that the same topics would be used in the 2016 census as were used in the 2006 and 2011 Censuses.

The move to an eCensus

3.17      The 2016 census was designed to be primarily conducted online, in what the ABS described as 'digital first', which 'aims to provide a more effective, more efficient, more environmentally-friendly census that was intended to be easier for people to complete'.^[16]

3.18      It was predicted that the digital-first approach would deliver savings of $100 million in the running of the 2016 census compared to the 2011 census.^[17] One of the largest costs associated with running the census is the temporary employment and training of thousands of field officers used to deliver and collect census information. The use of an eCensus would reduce the number of temporary workers required.^[18] For households, the online census promised to reduce the time required to complete the census from a benchmark of 37 minutes for paper forms in 2011 to 26 minutes for the new eCensus.^[19]

3.19      Although the 2016 census was not the first census to have an online option for submitting, it would be the first whereby the ABS actively encouraged households to use the online option.

3.20      It was reported to the committee that in 2012 the ABS began investigating the possibility of a primarily online census. As outlined in internal ABS procurement documents prepared in 2015:

In 2012 the ABS initiated a project aimed at the consolidation of eForms solutions across the ABS in order to deliver a single streamlined, flexible and cost effective outcome for the ABS. An internal assessment of the ABS capabilities and an approach to market via a Request for Expressions of Interest in 2012 established that current ABS solutions were on a par with anything that the market place were offering and as a result it was recommended that the ABS focus on progressing its implementation of existing eForms solution, Blaise.^[20]

3.21      The ABS goes on to report that this in-house solution was later determined to be unable to scale sufficiently to meet the needs of the census:

Despite significant effort to progress the advancement and implementation of Blaise, it has become clear that Blaise will be unable to scale to the requirements of the Census in 2016. With less than twelve months until the eForm needs to be deployed for the 2015 Census Dress Rehearsal, which requires the use of the 2016 Census eForm, the ABS must utilise the existing ABS Online Census solution.^[21]

3.22      A later report from an external consultant was more straightforward in explaining this fact and the consequences of it:

Whilst Blaise is the strategic choice for e-collection it has taken a significant period of time to demonstrate that this is not a suitable solution for Census. This has diverted resources and wasted time in the Census Program, an impact that is still seen in the delayed schedule today.^[22]

3.23      Following consideration of existing options, the ABS determined that its existing software system was not appropriate for use in the census, and in 2014, the ABS' Steering Committee responsible for the census program took the decision that Blaise was not a viable solution.^[23] 

3.24      The ABS engaged the firm Capability Driven Acquisition (CapDA) to conduct an independent review of the ABS' ICT capabilities for the 2016 census, and provide advice on the capacity and availability of commercial partners.^[24] CapDA produced the Census 2016: ICT Capacity & Capability (ICT Census) report and provided it to the ABS in May 2014. The ICT Census report agreed with the ABS' assessment that Blaise was unable to scale to meet the demands of the census, and recommended:

...[The] Review Team recommend that [the] ABS positively consider placing the bulk, if not all, the responsibility for development and operation of the eCensus and associated web facing components with the prime partner.^[25]

3.25      The ICT Census report, having examined the internal capacity of the ABS, articulated the case for the use of an outsourced solution saying:

[The] ABS will never have the resources, nor would it be likely to prove cost effective, to develop and sustain the skills/capability and capacity to run eCensus 'in-house' every five-years, as it is counter to the [business as usual] nature of the ABS.^[26]

3.26      The ICT Census report highlighted that the best value for money would be achieved through an open tender process—as is typical for government procurement processes—but that this option may not be viable in this circumstance due to time constraints.^[27] Because of this consideration, the ICT Census report recommended a limited tender:

Consideration should also be given therefore, to a limited tender to reutilise the existing eCensus application. This would potentially involve procuring IBM's services given their existing experience of the application, hosting it and working with the ABS on the eCensus. This route although not ideal from a procurement perspective, would have the benefit of mitigating the increasing risks to what is a far more complicated Census Program than has ever previously been attempted and in what is a much reduced timeframe for a [partner] to come 'on board' than in earlier Census cycles.^[28]

3.27      The ICT Census report also noted that the use of a limited tender to IBM would be appropriate 'in consideration of the limited time frame and the inherent risks in working with any new organisation'.^[29] The relationship between the ABS and IBM is discussed in greater detail in chapter 6.

Limited tender to IBM

3.28      IBM had previously worked with the ABS in 2006 and 2011 to provide online census systems. IBM reported to the committee that:

IBM was engaged by the ABS to provide IT services in relation to the 2006 eCensus and 2011 eCensus. IBM developed the electronic form and the online hosting environment (known as a 'platform' or 'system'). The solutions provided by IBM were successful, and saw the percentage of respondents who completed the electronic form rather than the paper form (the 'online response rate') increase from approximately 9 [per cent] in 2006 to 33 [per cent] in 2011.^[30]

3.29      IBM reports that during both the 2006 and 2011 censuses, it delivered 100 per cent availability throughout the busiest periods.^[31] As part of a separate contract, IBM provided the ABS an information technology solution to convert paper forms completed by households into an electronic database.^[32]

3.30      In September 2014­—following a limited tender process as recommended by the CapDA report—the ABS contracted IBM to develop, deliver, implement and host the online census system.^[33] The procurement documents noted that a limited tender would still ensure a satisfactory value for money outcome as:

Through previous experience the ABS has a detailed understanding of the costs to develop and support this system [eCensus], and thus will be able to use these benchmarks to measure whether value for money is being achieved. IBM are conscious that the ABS has this knowledge and that it will need to provide a value for money price if it is to win the tender. The total cost will be reduced through removing the cost of a full market tender process.^[34]

3.31      Under the terms of the contract between ABS and IBM, all intellectual property rights for the eCensus application are assigned to the ABS.[35] The committee was informed that IBM is not required to pay a royalty fee to the ABS to use the intellectual property developed during the census project.[36] The ABS argued that this arrangement would benefit Australia as the intellectual property could be advanced through further development work undertaken in other countries:

...we were keen to see the system further developed at other people's expense. My understanding is that we included in the contract a provision for IBM to leverage the intellectual property in the hope that the software would be further enhanced through other applications.[37]

3.32      Procurement documents show that the ABS were confident that IBM represented value for money and had the capability to deliver the 2016 census.^[38]

Development of eCensus

3.33      The 2016 census built on the solutions provided by IBM in 2006 and 2011, with the aim of increasing the online response rate to 65 per cent, or around 10 million households.^[39] The ABS later requested that IBM increase the eCensus hosting infrastructure capacity to support an online response rate of 80 per cent.^[40] The contract between IBM and the ABS for the design, development and implementation of the eCensus was priced at $9,606,725.00.^[41]

3.34      The agreed solution saw census data protected using encryption on dedicated infrastructure hosted in Australia, and cloud services used for non-sensitive parts of the system (static help pages, for example).^[42] IBM did not have access to the data provided by households. Data was only able to be decrypted by the ABS who possessed the unique decryption keys.^[43]

3.35      The contract required that the eCensus site be available to the public to access, complete and submit census forms for a minimum of 98 per cent of the 61 day period from 9:00am on 26 July 2016 to midnight on 25 September 2016, as well as for 98 per cent of the four hour peak period from 7:00pm to 11:00pm on 9 August 2016.^[44]

3.36      Revolution IT was contracted to perform load testing on the eCensus website.^[45] The key performance requirement of the eCensus was expressed in the number of forms that could be submitted per second. Load testing is the process of simulating human interaction with the application to ensure that the application can support the expected demand.^[46] The ABS expected a maximum required capacity of 250 forms per second, or approximately 900 000 per hour.^[47]

3.37      Submissions to this inquiry questioned whether the ABS had underestimated the load that would be placed on the eCensus system.^[48] Dr Robert Merkel, a software testing specialist, informed that the committee that website traffic peaks are often far higher than average levels. Even a short service disruption could have precipitated the overload of the eCensus due to its unique nature:

In many cases, when a website is overloaded and inaccessible, people will seek alternative sources for the information or service they attempted to access, or try again after a substantial period. In the case of the e-Census website, this would not have been the case—most people would have repeatedly tried to access the site, either for initial access, or to complete the submission of their Census form. As such, unmet demand would have built up like water behind a dam.^[49]

3.38      It was also pointed out to the committee that it is very difficult to simulate how humans are going to interact with software:

Creating a 'real' load to test effectively with is challenging. It can be very expensive to generate sessions which fully replicate a 'real' load situation. Often loads generated for testing are not sufficiently 'real' to fully load the webservers.^[50]

3.39      The ABS retained the services to UXC Saltbush to undertake a code review and penetration testing. Penetration testing looks for security weaknesses that an attacker might exploit. IBM reports that 'no issues of significance emerged in the course of either testing program'.^[51]

Conducting the census

3.40      A key cost of delivering censuses in the past was the large number of temporary staff that needed to be employed to deliver and collect census forms from every household in Australia. One of the key motivators of moving to an eCensus solution was the unsustainability of this model:

The traditional approach to the 2011 and previous Censuses was reliant on increasing the number of field staff in order to deliver and collect forms or online access codes from every dwelling. Recruiting a sufficient number of staff required for the short-term field operation has become more difficult; and staff costs have escalated. In 2011, there were a number of areas where sufficient field staff could not be recruited and the ABS needed to fly in higher paid ongoing employees. In addition, with decreasing numbers of people in each household and an increased proportion of the population that is in the workforce, Census collectors have had decreasing success in contacting people at their homes. Some dwelling types such as secure apartment buildings are particularly problematic for hand delivery and collection of materials.^[52]

3.41      The 2016 census changed the way census materials were to be delivered to the public and returned to the ABS. It was explained to the committee that:

Australia Post’s mail service was used to deliver and return required materials from the majority of households. The majority of households responded online. Households are able to request a paper form through an automated phone service if they preferred or needed to respond by paper.

...

In some areas of Australia, where the postal service was likely to be unsuitable or insufficient address information was known, Census Field Officers delivered materials to each dwelling, enabling residents to either complete their form online or mail back a paper form. In other areas where a high proportion of residents were expected to need to complete the Census form on paper, all households were delivered paper forms in addition to login numbers (e.g. in areas where there is a higher proportion of older residents).^[53]

3.42      The majority of households received a login code via the post that enabled them to log in to the eCensus website during the collection period. The remainder of households were posted or hand-delivered a paper census form to be completed, and later collected by Census Field Officers.^[54]

Communication of move to eCensus

3.43      It was reported by the ABS that 'a communication campaign designed to drive participation is a key component of every Census'.^[55] The committee heard that awareness raising is necessary because of the relatively long period between Censuses, as well as new arrivals to Australia not being familiar with the process. In addition, there were significant process changes to communicate because of eCensus. The changes to the 2016 census were to be communicated to the public through a variety of vectors:

The 2016 Census national campaign comprised integrated paid advertising, media and public relations, social media and online communication, stakeholder communication, special audience advertising and communication, and the development and distribution of information materials.^[56]

3.44      The committee heard that the census was advertised across television, radio, print, digital and social media; additional advertising was also adapted and translated for Indigenous and culturally and linguistically diverse audiences.^[57]

3.45      The committee heard some concerns regarding the advertising undertaken by the ABS.^[58] Vision Australia reported that some public advertising material was inaccessible to the blind and low vision community.^[59] Interlime observed that 'the advertising campaign was not very effective in engaging with Australians or explaining the importance of the Census'.^[60] It was also argued that the ABS did not adequately explain the changes in how the 2016 census was to be completed.^[61]

3.46      Research conducted before the census showed that in July 2016, 78 per cent of people were aware of the census. Further research conducted between 17 August and 4 September showed that 96 per cent of respondents were aware of the census. Less than one per cent of people who were aware of the census indicated that they did not intend to complete it.^[62]

Telephone assistance

3.47      The Census Inquiry Services (CIS) is established each census to provide telephone and email based support to the public throughout the census period. The 2016 census saw the introduction of the Paper Request Form Services (PRFS) to automatically process requests for paper census forms. The ABS signed a Memorandum of Understanding with the Australian Taxation Office to provide call centre operations for the 2016 census.^[63]

Emphasis on 9 August

3.48      Households are able to complete the Census over a number of months. The Census does not have to be completed on any specific date, but the answers should provide information related to the reference date. In 2016, this date was 9 August.^[64] For example, if a household completed their form on 2 September, they should answer the questions as they would have had they completed the form on 9 August.

3.49      The 2016 census advertising campaign followed a similar format as previous Censuses with a strong emphasis on the reference night. The ABS explained that this this approach was taken in order to:

• keep the message simple;
• mobilise people around an 'event'; and
• ensure that census forms are completed in respect of census night.^[65]

3.50      The committee heard concerns that the tight focus on the reference date left many Australians with the incorrect impression that they must complete the census on 9 August.^[66] IBM, for instance, argued that one of the consequences of the ABS focusing heavily on 9 August in information advertising was that many people may have formed the view that the census had to be completed on 9 August.^[67] The eCensus website was in fact open for a total of 61 days, during which time Australians were able to complete the census online. ID Consulting was also critical of the way the ABS communicated with the public regarding how to complete their census forms:

There was poor communication from the ABS on the date for completion. The online site was always planned to be online from July 25 to September 23, to allow people a large window to complete the form, and give the last few an option to still complete online while being followed up by field officers. But this was widely publicised as “You have until September 23 to complete it”, so people were confused as to why they were being followed up before that date. ABS needs to be clearer that August 9th is the reference date and it needs to be completed as close to the date as possible.^[68]

3.51      The events of 9 August are discussed in detail in chapter 6 of this report, including the performance of the website and telephone services, as well as the conduct of census field officers.

Navigation: Previous Page | Contents | Next Page