The great American writer and wit Mark Twain is credited with the well-known aphorism: “There are three kinds of lies: lies, damned lies, and statistics.”
The rejection of the Socialist Equality Party’s (SEP) application for registration by the Australian Electoral Commission (AEC) is a case in point. It was based on the statistical testing of the party’s membership list using methodology that is deliberately opaque and inherently biased against parties applying for registration.
The statistical method serves the political purpose of anti-democratic legislation requiring party registration: to shore up the failing two-party system by preventing parties such as the SEP contesting with their name on the ballot paper in the upcoming federal election. As support for the major parties plunged, the Labor Party and the Liberal-National Coalition came together in August 2021 to treble the membership requirement from 500 to 1,500. The transparent aim was to reduce the number of parties able to contest elections under their names. Parties with parliamentary representation—that is, capitalist parties with the greatest resources—were exempted from the registration process.
The extent of the popular hostility towards the major parties is such that the upcoming federal election represents a major political crisis for the ruling class. There is a strong likelihood that neither Labor nor the Coalition will be able to form a stable, majority government. Thus, a great deal is at stake for the political establishment in preventing the registration of third parties, particularly the Socialist Equality Party. The SEP, formed in 1972 as the Socialist Labour League, has a far longer history than most other parties, including the Greens, and has sunk deep roots into the working class, significant layers of which are now looking for a means of fighting against the growing threat of world war and the worsening social crisis.
The entire process of party registration is veiled in secrecy, drawn-out and profoundly anti-democratic. The SEP submitted a list of more than 1,500 members, complete with their names, addresses and dates of birth to the AEC last year. More than four months later, the AEC responded by declaring that the SEP had failed the membership test. The AEC contacted a very small sample of members, out of which several supposedly denied membership. Under the guise of “privacy,” the AEC will not tell the SEP who those individuals are, so that it can verify the Commission’s claim or amend its list.
The AEC’s guide to party registration provides very limited details of the method used to determine whether a party has at least 1,500 members. After removing those not on the electoral roll or determined as being members of other parties, the AEC does not check the entire list, but uses a sampling technique supplied by the Australian Bureau of Statistics (ABS). All that is revealed by the AEC is a list of sample sizes and allowed “denials” of membership, corresponding to the size of the submitted membership list running from 1,500 to the permitted maximum of 1,650.
Significantly, the statistical methods used to generate those figures have never been revealed. As a result, the underlying assumptions remain a secret and the numbers themselves cannot be verified independently. All that is stated is that the ABS regards the testing method as “reasonable.”
Statistics can be a very powerful tool that has a broad range of applications in the sciences, engineering, manufacturing, medicine and many other fields. It is, however, an elementary requirement that details of the statistical method be open to independent examination. If a scientist submitted a paper for publication that failed to supply statistical methodology it would be automatically rejected. Even in the rather dubious field of opinion polling, the statistical methodology is stated.
There is no innocent explanation for the failure of the AEC or the ABS to reveal the statistical methodology. The SEP has consulted several professionals with a background in statistics to examine the AEC’s sampling in an effort to identify the underlying statistical methodology used. They were astonished by the small size of the samples and the tiny number of denials allowed—in the case of the SEP, a sample of 33 with just 2 allowable denials.
Based on the figures available, they determined that it was very likely that the ABS has used an exact hyper-geometric method, commonly employed for testing relatively small populations, to determine the sampling figures supplied to the AEC. Applying that method, it proved possible to model the sample sizes and allowed denials used by the AEC. More importantly, in doing so, some of the underlying assumptions could also be identified.
The confidence level
The first assumption required in all statistical testing is to choose a confidence interval or the level of confidence that the test is reliable and the results are correct. Different confidence levels are set depending on what is being examined and the degree of reliability required. In the case of the AEC, the confidence level used is not just a matter of statistics but is a political choice. The lower the confidence level, the greater the likelihood that a valid membership list will be rejected.
All the professionals consulted came to the same conclusion—that the confidence level used in the AEC sampling was at the very basic 95 percent level. In other words, what the ABS regards as “realistic” is that the result of membership testing could be wrong 5 percent of the time. Put another way, 1 in 20 tests could reject a valid membership list.
There are no doubt myriad statistical applications in which a 95 percent confidence level is legitimate. However, what is being determined by the AEC is not food preferences or voting intentions at the next elections. Rather it is deciding on the basic democratic rights of a party to be identified on the ballot and those of voters to know for whom they are voting.
As one professional remarked: “In the social sciences, the literature often considers a result significant at the 95 percent confidence level. In the physical sciences we generally use 99.99994 percent confidence to establish the reality of a result. And in empirical sciences, the general rule is 99.7 percent confidence is the general heuristic one tends to find. The use of such a weak consideration [by the AEC/ABS] to reject the validity of something so important as an electoral option doesn’t speak well for their aims.”
In the hyper-geometric test, the confidence level, along with another parameter—the margin of error or precision, determine the sample sizes used and, in the case of the AEC, the number of allowed denials. The lower the confidence level, the smaller the sample sizes and number of denials allowed. At a higher confidence level of 99 percent, the figures are substantially higher.
On our calculations, at a 99 percent confidence level, a sample size of 54 and maximum allowed denials of 5 would be used to test the SEP membership list, significantly reducing its chances of failing the test.
This is not simply a theoretical issue. There is at least one case where the choice of confidence level made the difference between a party being registered or not. This is demonstrated in the protracted process lasting over two years beginning in December 2021, in which the AEC moved to deregister the Australian Progressives Party.
The first attempt to deregister the party was unsuccessful, due to the AEC’s own staggering error. After its first membership list was deemed to have failed, the party submitted a slightly amended second list which also failed and the party was deregistered. However, the testing of the second list in February 2022 contained a glaring difference. Of the 1,576 names submitted, the vast majority, 1,391, were deemed not on the electoral roll, as compared to just 44 on the previous list. When challenged, the AEC was compelled to admit that “an error had indeed occurred in formatting the Party’s membership list for testing” and the party registration was reinstated.
Nevertheless, the process of deregistration ground on, recommended by the AEC in June 2022. The party’s membership list submitted in July 2022 was again deemed to have failed. A sample of 33 members was tested and found to have three denials by the smallest margin—one more than the two allowed. The Australian Progressives provided another list in September 2023 which again failed.
The party had already applied for a review of the AEC’s process in November 2022, including a questioning of the methodology used to test membership. The review, which was repeatedly delayed due to elections and by-elections, finally concluded in 2024 and upheld all the AEC’s decisions.
However, one point in the AEC’s lengthy justification of May 2024 to the Australia Progressives stands out. To justify its sampling method, in reference to the July 2022 list, the AEC cited the ABS advice that “with respect to a list of 1,523 members where 2 denials are permitted, the probability of rejecting a valid list is 1.2 percent or below.”
It may appear to be a very small point. But if the probability of rejecting a valid list had been 1 percent or less, 3 denials would have been allowed, and the party would have been registered in 2022.
The choice of confidence level matters. If the statistical methodology were based on a 99 percent, not 95 percent, confidence level, the chance of a party passing the membership test would be significantly higher.
Limiting the size of membership lists
The choice of confidence level biases the AEC membership test against the registration of parties, but the arbitrary upper limit place of 1,650 on the size of the membership lists to be submitted is even more egregious. It markedly decreases the chances of a party being registered.
Commonsense would suggest that the larger the size of the membership list, the greater the chance that a party had 1,500 members, given reasonable care in assembling the list. Translated statistically, as the number of members increases, so does the size of the sample and also very significantly the number of allowed denials. Yet the AEC flatly refuses to test lists of more than 1,650 members.
In one case, however, that of Voteflux.org | Upgrade Democracy, the AEC allowed the party to submit a list greater than 1,650 members. The party applied for registration and submitted two lists—the first of 1,650 in December 2021 and a second of 4,680 in February 2022. Both were deemed to have failed and its application was rejected.
Voteflux challenged the decision, the statistical methods used and subsequently made a Freedom of Information application to secure the internal documents of the AEC and ABS related to the process. The documents reveal aspects of the tortured arguments conjured up by the AEC, in consultation with the ABS, to deny the obvious: namely the high probability that Voteflux indeed had at least 1,500 members.
The AEC did not apply its test to the whole membership of 4,680, but instead selected the first 1,650 names on the list and used those for its membership test, recording a failure. Voteflux challenged the decision, including the statistical methodology.
As the chief ABS methodologist advised the AEC, the method of testing the first 1,650 names was statistically flawed as, without randomisation, nothing could be concluded about the larger list of 4,680. Rather than suggesting a valid test of the Voteflux membership list, he proposed to reject the party’s registration on the basis that the first list had failed and the second “untested list” was “of only limited persuasive value.”
As far as the AEC was concerned, it was all an unfortunate mistake—the list was over 1,650 and should simply have been returned to the party for revision, as per its guidelines. But that simply begs the question, why not allow the submission of larger lists? The testing procedure as a whole is not mandated by legislation. It is certainly possible to test a larger list. The ABS methodologist himself explained that the Voteflux list could have been tested with a sample of 564 and maximum denials of 399.
The internal documents reveal that the question of money and manpower was raised. The Party Registration Team/Services Australia contracted to the AEC estimated the cost of carrying out the current membership test at $1,082, and of the Voteflux list at $10,153, adding that the larger sample could also create administrative problems and potentially require extra staff.
But the fundamental issue, which was not discussed, was not cost. One could legitimately ask, what price democratic rights? Rather, as is intuitively obvious, larger lists increase a party’s chances of registration.
This is clear from the numbers. Testing the full Voteflux list requires a sample of 564 and maximum denials of 399. In other words, only 3 in 10 people of that sample would need to affirm party membership for Voteflux to be registered. By limiting the list to 1,650, the sample size drops dramatically to 60 and the allowable denials to 9. So by comparison, more than 8 in 10 of the sample must affirm membership for the party to be registered.
The refusal of the AEC to countenance proper testing of Voteflux’s entire list of 4,680 minimised its chances of being registered and was a flagrant breach of its democratic rights. The same applies to any party with a membership larger than 1,650.
The AEC’s membership testing is protected by the wall of secrecy surrounding the statistical methods used by the ABS to determine sample sizes and maximum acceptable denials. Any challenge is dismissed out of hand with the claim that the ABS endorses the method as “reasonable.” Yet no-one is permitted to review or challenge the calculations, or the statistical methodology used, and the assumptions that underlie it.
The same secrecy envelops the entire biased and anti-democratic process of party registration, without reasonable accountability to the public, meaning that a multitude of sins can be covered up. That blunders do occur is demonstrated by the treatment of the Australian Progressives cited above. If the party had not challenged the AEC decision, their deregistation would have been done and dusted at that stage. Lesser mistakes, biases, even outright rigging are hidden by the opaque character of the AEC’s operations.
The avenues for challenging any AEC decision are limited, potentially costly and Byzantine in character. As has been pointed out on the WSWS, the AEC dismissal of the SEP’s challenge to its decision to reject the party’s application was absurd on the face of it. The AEC declared that it had not technically rejected the application but had invited the SEP to submit another list. Until that was done, a review could not proceed. But in submitting a new or amended list, the processing of the application begins anew, which could take months. That is under conditions where the announcement of an election is just weeks away, after which the application is frozen. Whether the SEP resubmitted a new list or not, following the election the registration process begins anew and another list is required. A no-win situation.
No doubt Mark Twain could have written a biting satire about the whole process and the state bureaucrats in the AEC who posture as servants of the public. But it is not enough to simply point out the absurdly bureaucratic and biased character of what the AEC does. The legislation concerning party registration and the manner in which it is applied serve a definite political purpose—to prevent a challenge to the political monopoly of Labor, the Coalition and the Greens.
The SEP has launched a campaign to oppose the rejection of the party registration. We urge all our electoral members and supporters to participate actively in this crucial campaign, including by sending letters of protest and support to the SEP at sep@sep.org.au.
Despite the AEC ruling, the SEP will intervene boldly in the 2025 federal election to provide workers and young people with a real socialist and internationalist perspective, against the disastrous program of militarism, austerity and authoritarianism supported by every other party.