The typical Australian is a 38-year-old married woman, according to the ABS, but is that just indicative of those who filled out the census?
This week the Australian Bureau of Statistics released the first batch of data from its controversial 2016 census. Given all the furore over the census last year, can we trust the data?
In the aftermath of the census website crashing on census night, and the controversy over the ABS’ decision to retain names and addresses of people completing the census, questions have been raised over whether the data collected is be an accurate reflection of the Australian population.
Yesterday, the Australian Bureau of Statistics released the first snapshot of the data it has collected ahead of a full release of the data in June. The ABS yesterday said that according to the data collected the “typical” Australian is a 38-year-old married woman born in Australia with two children, a three-bedroom house and two vehicles.
It will not be known until the release of the full data and two accompanying reports in June on the 2016 census whether the data is reliable. The ABS has suggested in previous submissions to the Senate inquiry into the 2016 census that fewer people refused to complete the census in 2016 than in 2011, but that was based on a preliminary assessment in November.
As of the start of November, the ABS said 10,576 households had refused to submit a census form, fewer than the 13,194 refusals in 2011. As of November last year, no referrals had been made to the Commonwealth Director of Public Prosecutions for failing to comply with ABS orders to complete the census. There were 78 referrals as a result of the 2011 census.
The ABS can opt not to chase people for not having filled out a census form, but usually this is only done for compassionate reasons such as illness or death in the household, or because of language barriers, not because of concern about how the ABS was storing the data.
The ABS said it had undertaken a “comprehensive data verification process” on the names and addresses given and said there was no indication that there were a “significant number” of households either withholding their names and addresses or providing false information to evade the ABS data retention.
In response to public concern about the quality of the data, the ABS announced last year that in addition to the post enumeration survey (PES) — an independent household survey to measure the completeness of the census counts — there would also be an independent assurance panel reporting back to the Chief Statistician to provide extra assurance and transparency on the quality of the census.
The panel of six, chaired by James Cook University vice-chancellor Sandra Harding, was due to report at the end of March, but an ABS spokesperson told Crikey that the report was delayed so that the PES could be incorporated into the report to be released publicly with the rest of the census data in June.
“This report will be made publicly available on 27 June 2017, so that government, the community and other stakeholders can make their own informed judgments about the fitness-of-purpose of 2016 census data. The report is expected to contain embargoed statistical information and thus cannot be released in advance of the 27 June 2017 data release,” the spokesperson said.
In its response to the parliamentary committee’s report on the 2016 census released in late February, the ABS said it would consult the public on proposed changes to future censuses only after completing a privacy impact assessment not less than six months out from the next census. The agency also said would consider its communications strategy for how to inform people about their obligations to complete the census and the penalties for failing to do so.
The ABS’ disastrous handling of the 2016 census should be the canary in the coalmine for government agencies attempting to move their services online, according to the Prime Minister’s special adviser on cybersecurity Alastair MacGibbon.
Along with the parliamentary committee report yesterday, the government released MacGibbon’s report, which will determine the fate of ABS staff and those involved in the CensusFail.
MacGibbon doesn’t call for anyone to get sacked, so it seems no heads will roll, but he paints a picture of an organisation woefully underprepared for what should have been entirely predictable, and ministers and senior government executives were unable to properly communicate with each other and the public about the matter due to their own lack of technical understanding.
During the day on August 9, the online census form was subject to four distributed denial-of-service (DDoS) attacks, and on the fourth attack the website went offline.
MacGibbon found that the ABS had a “library of six incident management documents” designed to prepare the ABS for any possibility but had none to address what happened on census night, meaning decisions undertaken by the ABS were ad hoc and often not sufficient for what was required.
For example, on the night of the incident, the ABS prepared a brief for the government and sent it around to several agencies at 1am, but “due to an administrative error” the Prime Minister’s Office didn’t get the brief until closer to 5am, just before a teleconference on the matter.
MacGibbon reviewed all the public statements made by ministers and politicians over the course of the next day about CensusFail and determined ministers and senior government executives should be sent to a “cyber bootcamp” to learn how to better talk about cyber security incidents to the public. Much of the confusion in the early hours stemmed from whether the incident was a “hack” — it wasn’t, but some politicians referred to it as such.
ABS and IBM had prepared for DDoS attacks, but the decision to go to what was referred to as “Island Australia” (to block all overseas traffic to the site) was not implemented properly. MacGibbon said geoblocking could have stopped the DDoS attacks, but it was not a strategy ASD recommends.
In fact, ABS and IBM did not consider what would happen when all overseas traffic was blocked from accessing the eCensus form. Some parts of the eCensus system itself, such as password resets, were themselves located offshore and thus became unusable. Vodafone customers in New South Wales and Australians using virtual private networks were also blocked from accessing the census website.
“Island Australia” was only ever meant to be applied for 10 minutes at a time, but IBM and ABS decided to keep it going all day on census day. In addition, the telecommunications providers working with ABS on the census were not included in the “Island Australia” testing, meaning for one of them — Vocus — their system was not configured properly for geoblocking, leading to many of the problems on the day.
MacGibbon said that the incident should serve as a warning for other agencies moving their services online, and recommended the digital transformation committee of cabinet develop cybersecurity shared services to provide consultation across government on cybersecurity.
After directing most of the blame at IBM, Prime Minister Malcolm Turnbull has avoided dragging it out into the courts as the Queensland government did with the Queensland Health payroll debacle and announced that a commercial-in-confidence settlement had been made.
Nov 17, 2016
The Australian Bureau of Statistics says your census data is protected by legislation. And since no one would be mad enough to elect a crazy despot, it will be fine.
The Australian Bureau of Statistics has admitted it could have communicated with the public better about its decision to retain names and addresses from census forms for much longer than in past years.
The ABS waited until the week before Christmas to quietly announce it would retain names and addresses in separate files along with the census data in order to “provide a richer and dynamic statistical picture of Australia”. Crikey first reported the issue in January, and controversy over the proposal grew over May, June and July this year, with the ABS dismissing concerns about the privacy implications of retaining every Australian resident’s personal data for years, and boasting it had spent a lot of time discussing the issue publicly before making the decision in late 2015.
The ABS, now somewhat humbled by the disaster of what is now called Censusfail on census night in August, is admitting that the ABS could have communicated the reason for the change and discussed it better publicly before December last year.
Speaking at the GovInnovate conference in Canberra on Wednesday, the ABS’ general manager for strategy and partnerships, Gemma van Halderen, admitted communication had not been as good as it could have been.
“The value that comes from an independent privacy impact assessment in hindsight, that is something we should have done as well,” she said, but added that the data was ultimately vital for public policy decisions and for determining the makeup of our parliament.
“We’d be a very unfortunate democracy if we were making decisions with no information out there.”
Van Halderen reiterated that legislation surrounding the census mandated the data only be used for statistical purposes, and not for surveillance. Van Halderen joked that ABS had learned from its IBM outsourcing mistake during the online census debacle and the storage of the data had not been outsourced, so it would be better protected.
But Anna Johnston, director of Salinger Privacy, said that legislation could be changed, and while the data might be protected from being used for nefarious purposes right now, all it would take is a Trump-style leader for those protections to fail.
“I don’t think that’s overstating it when you see, for example, Trump has called for the reintroduction of the demographics unit within police services in order to use demographic data to target Muslims in America,” she said.
“If you got that kind of government here — which two weeks ago I would have thought was absurd but now we know could happen anywhere in the world — the fact that your legislation says that shouldn’t happen would not stand in the way of a despotic government.”
The only real way to prevent that from happening would be to prevent the ABS from holding identifiable data, she said.
“The assurances that have been given in the past were stronger because the ABS simply didn’t retain names longer than 18 months, and the purpose for which names and addresses were used were basically just to tick off that everyone had done the census. Now there’s a whole dramatic new proposal for how it is going to be used, and there wasn’t a national conversation about that.”
Van Halderen countered the criticism around the census, suggesting that less than 1% of people the ABS had sought feedback from had said they had an issue with the retention of names and addresses. Johnston indicated that part of that was to do with concerns about the $180 per day fines for non-compliance. Johnston said she had members of the public calling her in tears saying they were Jewish and had family members persecuted in the Holocaust but didn’t want to be fined. She said she had told them to complete the census.
Oct 19, 2016
The ABS says a high response rate to the 2016 census means its data is clean. Not so, writes former Australian Statistician Bill McLennan.
The Australian Bureau of Statistics wanted a response rate of 93.3% for the 2016 census, according to The Mandarin, “to be sure that the information is of sufficient quality”. I am at a loss to see how 93.3% of households filling in a census form, or any other rate for that matter, can assure that information is of sufficient quality.
For example, it could be possible that every one of the 93.3% of houses filling in the census form has omitted one person from its list of residents. Assuming on average there are three people per household, that would mean that 33% of the people in Australia were omitted from the census. Or if every third household left out a person, then 11% of the people would have been missed. (I didn’t take into account the people in the missing 6.7% of households, so the percentage of missing people is slightly different to what I have claimed.)
Similarly, let’s say in an act of deliberate civil disobedience every respondent deliberately gave an incorrect answer to one census questions. Can you imagine what that would do the quality of the census? I can’t either, but I have little doubt that it would reduce its quality significantly.
The real problem the ABS has is that it has no way of making estimates of quality. I say this because, as I understand matters, no testing of note was undertaken before the census being undertaken, which might be able to throw a little light on this issue. Further, the Post Enumeration Survey is just a check on coverage, as there is no attempt being made to measure how well Australians answered the census questions.
Similarly, there has been no discussion of whether or not the answers provided on a paper form are different to the answers provided online. I don’t know what the differences might be, but I would bet there are differences which might possibly be large. How does this impact on quality? Is it the same across the whole population or is it focused into particular areas or sectors of society?
We do know the ABS has announced that a panel of “independent experts” has been asked to examine this issue and report on the quality issue early next year. How this panel will be able to come to any credible conclusions is a mystery to me, as there is no believable evidence available. I only hope that this panel of independent experts comes up with conclusions much closer to the truth than did the one that investigated the obviously dodgy Monthly Labour Force Statistics. It gave the MLFS a “clean bill of health”, but all serious users of labour data don’t give the MLF data much credence these days.
* This article was originally published at The Mandarin
We spend a lot of time telling you what we think. It’s time we got to know you better and listened to what you think.
The Crikey Census will help us protect our journalism and deliver more of what you want. And unlike that other national survey, our systems are secure.
Tell us a little about who you are, what you’re interested in and what you really think about Crikey, our products, our people and our journalism. The whole thing shouldn’t take any longer than five minutes. We’d really appreciate it.
We’ll even reward you for your time: a $25 Dendy Direct voucher allows you to stream all the latest movies and premium TV you can binge. Be among the first 500 respondents and it’s all yours.
Sep 26, 2016
We won't know how many people completed the census until early next year, according to the ABS.
The Australian Bureau of Statistics has revealed that its own gloating about census completion rates cannot be trusted.
The ABS has claimed all that concern about privacy and the website debacle was for nothing, with close to 95% of households completing their forms by last week’s deadline.
But in its submission to the Senate inquiry — which was posted and then removed when a committee officer failed to remove confidential information from the submission — the ABS expressly states that final participation rates in the census would be finalised at the end of the collection phase and that “final dwelling response rate is not known at this time”.
The actual response rate will not be known until early 2017, after processing. A total of 6743 people have flat-out refused to complete the census as of September 20, 2016. About 460,000 households have not yet completed the form, but they have not notified the bureau that they are refusing on principle.
The ABS also revealed how it planned to link the data over time. The ABS said it would be “creating anonymised linkage keys on a project-by-project basis”.
One of the major criticisms of the decision to keep names and addresses was that the ABS’ privacy impact assessment was an internal assessment, not conducted by an independent third party, as recommended by the Privacy Commissioner. The ABS responded by saying that there was “no requirement” for an independent privacy impact assessment.
As Crikey‘s political editor Bernard Keane notes elsewhere today, the ABS throws IBM under the bus several times in the submission because of the failure of the online census form. In 2012 the ABS examined whether the existing platform for the previous census would be appropriate for the “digital first” approach encouraging people to fill out their census online rather than on paper. It wasn’t until 2014 it was decided that a new platform was needed, and due to the limited time available, the ABS decided to issue a “limited tender” to IBM given IBM’s experience in this area.
The timeline provided in the submission of the events of September 9 that brought down the online census form contradict the statements made by Small Business Minister Michael McCormack at the time. McCormack said the ABS had made the decision to shut down the form at 7.45pm, but the ABS says in its submission that IBM attempted to reboot the system at 7.43pm after a distributed denial of service attack at 7.45pm, and it wasn’t until after 8pm that the ABS asked IBM to prevent people filing new census forms.
McCormack also said that the system had been restored at 8.50pm but overload controls kept people from filing forms after that point, but the ABS submission makes no reference to this, stating that at 10.26pm IBM was finally able to reboot the system but it was kept offline to resolve the security concerns held by the ABS.
Amusingly, the ABS notes that while the system was offline, the ABS census Twitter account kept telling people to complete their census forms, and that Twitter “was still attempting” to turn off the automatic retweet subscription as late as 9.20pm on the night of the census debacle.
In the fallout from the online census failure and the ongoing privacy concerns around the handling of names and addresses, the ABS told the committee it would establish an “independent panel” to review the census and would “work proactively with key users and the public ore broadly to ensure that there is a sound understanding of the quality of the Census results”, when the first release of data happens in April next year.
On the Census
Marilyn Chapple writes: Re. “ABS denies census is in crisis, with one in five forms still missing” (Friday). Maybe there is an explanation for many of the missing census forms. I have received four census forms. Three were posted, with three different logins, to three different versions of my address, which is a dual occupancy that seems to confuse all bureaucracies. The third letter came after the census date. Then a few days ago I returned home to find a paper form under the door, with a prepaid return envelope. I completed one form online early on census day, as blind Freddie could have guessed the system would crash. Thank goodness I’ve kept the receipt from the ABS. But I have no idea how I’m going to convince them that there is only one dwelling here, not three.
Iain MacPhail writes: It was with considerable reluctance that we finally sent off the form for the recent census. A month ago we submitted an inquiry to the ABS: “Tomorrow our household will be required to fill out the form for the 2016 Census. I see a census as a good thing, and have always been keen that we should be counted. Until this year. … We are in a quandary. It seems that if we are to preserve our privacy, we must either refuse to complete the census form and face a heavy fine, or join the mass of those prepared to obfuscate. Can you please demonstrate how these concerns are either false or are being addressed?” We have received no acknowledgement. We do not make a habit of refusing to comply with law and regulations, but in this case we felt compelled to leave the census forms incomplete.
The Australian Bureau of Statistics is facing another crisis over its census bungling as it struggles to find willing staff to collect the outstanding census forms, which were due more than a month ago.
Last week, Crikey reported that field staff were being told to go back to residences even when residents had told ABS officers they had already sent off their census forms. Now, with just under two weeks left to collect census forms, the ABS is in full-blown panic mode, attempting to get more field staff to go out in the field to harass the remaining residents who have yet to fill in their forms, and could now potentially face prosecution and $180-per-day fines for non-compliance.
Field staff — who are also being ordered by the ABS to “not discuss their work with anyone” — were offered 40 hours’ worth of collection work each, but a lack of people willing to go harass the general public on behalf of the ABS meant that those hours were run through pretty quickly, and now those staff are being offered extra hours to make up the gap.
Field officers are also receiving offers to begin covering other areas of the country where other staff have rejected work in order to ensure everyone has been covered. It is estimated around 20% of forms are still to be collected.
The Greens’ Scott Ludlam, NXT Senator Nick Xenophon and independent Senator Jacqui Lambie are putting a motion to the Senate today to call on the ABS to ditch the fines for people who have refused to complete the census.
The Australian Bureau of Statistics has said that release of internal discussions around questions put to the bureau on how it handles private data would undermine what it is telling the public.
Last month Crikey asked the ABS a set of questions about how census data had been handled in the past, in the wake of controversy over the decision to retain names and addresses from census forms for a longer period of time than in previous years. The ABS refused to answer the questions, providing a generic statement about the ABS’ data security policy. We filed a freedom of information request into how the media inquiry was handled, and in a response today, almost all of the email chain dealing with the request was blacked out.
The text was blacked out because the Australian Bureau of Statistics claims it would be “contrary to the public interest” as it would “cause uncertainty” in the official statement that didn’t directly answer any of the questions we asked.
“Such an outcome could reasonably be expected to have a detrimental impact to the confidence placed in official statements and in turn impact the ability of the ABS to fulfil its legislated functions,” the ABS stated.
Amid the ongoing fallout over the ABS’ bungling of the census, the organisation has been reluctant to give full and detailed responses to questions on the census. The organisation is also now on the hunt for a new media manager.
Sep 9, 2016
A month on from the census, only 80% of us have sent in our forms. And the Australian Bureau of Statistics is getting grumpier about it.
There’s been some interesting follow-up to our revelation yesterday that the Australian Bureau of Statistics is panicking about the rate of return of censuses and telling collectors in the field to step up their harassment of people believed to have not lodged them.
The ABS belatedly got back to us with a long response that we’re happy to run in full. First, the ABS denied that executives had used the word “crisis”, which our sources say has indeed been used internally. Secondly, it said:
“The ABS often makes procedural changes to its Census field operations. Adaptive procedures are common practice in the Census, and an important part of delivering Australia’s largest logistical exercise and getting a full and accurate count of the population.
“During an operation the size of the Census, there will inevitably be a small number of households that have already completed their Census who receive reminder materials or are visited by a Census Field Officer. The most common reason these visits occur is the delay between posting a completed paper form and that form being received at our Secure Data Capture Centre. This can take up to 10 working days, but is usually much quicker. There can be other reasons which the ABS investigates.
“If a person who has completed the Census is contacted by a Census Field Officer, they should let them know they have already completed their form.”
(Which many people have done, to no avail).
According to the bureau, “Final Reminder letters will be issued prior to the online form closing on Friday, 23 September. We ask anyone with a completed paper form to send it back to the ABS in the blue Reply Paid Envelope no later than the 18 September.” And on that vexed issue of how censuses can be “overdue” if the closing date is September 23, the ABS takes a rather extreme position: “Census night was Tuesday, 9 August. Census forms are therefore now due.”
According to the ABS, it has received “over 7.5 million household forms (more than 80 per cent of households)” so far. We’re not sure 80% is looking too flash a month on from the census, which needs to have about 95% returns to be useful. Meanwhile, Crikey readers have been offering their own stories about harassment by collectors.
“Thought we were being over paranoid when we photographed us putting the census forms in the post box but after two visits from the ABS saying we haven’t submitted the census I’m glad we did.”
“Two nights ago a man who said he was from the census office knocked very loudly on the front door. When I questioned what he wanted he said you have not completed your census forms and now you are in trouble. I advised him politely that we had until 23.9.2016 and he immediately said that is a lie and after I again reminded him that we had until the 23.9.2016 he said well that was mistake and you must complete the forms now … He was a very rude individual.
Finally, another ABS source reports that the ABS has now pulled from its internal workspace system census manager Duncan Young’s advice on approaching non-respondents. Perhaps the ABS wasn’t happy about the way it was ending up in Crikey?