Monday, November 22, 2021

Detailed Comments On The Voter ID Debate

With the Bill scheduled to go before the House of Representatives tomorrow, I thought I should post some detailed comments about the Electoral Legislation Amendment (Voter Integrity) Bill 2021, aka "Voter ID".  The bill has been introduced to the House of Representatives but I have no knowledge of when it might be scheduled for a second reading vote.  On the one hand the government has struggled to make out the case that Voter ID as they propose it is necessary or would even work all that well in preventing deliberate voter fraud - the public evidence being that voter fraud is a relative non-problem.  On the other, opponents have been so inaccurate and overblown in their replies that the primary way their fearmongering will become true is if it so confuses their own supporters as to become a self-fulfilling prophecy and costs them votes.  That's a fate that would be self-inflicted for the way this issue has been debated.  Overall I would prefer not to be seeing this debate and instead to be seeing real bipartisan progress on long-standing problems with our electoral system such as the lack of savings provisions for unintended informal voting in the Reps and also the incorrect surplus method used in Senate counting.  

We still don't know yet if the proposal is actually going anywhere (quickly or slowly) as much currently appears to hang on the vote of Centre Alliance Senator Stirling Griff, who has to this stage made some fairly supportive noises, though his House of Reps colleague Rebekha Sharkie has seemed somewhat less impressed.  As I write it is reported that Griff will support an inquiry (presumably JSCEM) which would be likely to delay the bill.  It is also possible that disruption from Senators concerned about vaccine mandates will affect the fate of the bill (although there is always scepticism about whether their threats to withdraw their votes will be carried out).  The numbers in the lower house are less critical but without the support of Bob Katter the government could end up depending on either the lack of pairing arrangements for crossbenchers or the support of Craig Kelly.  This said, Kelly was a big fan of voter ID in his former role as the Member for Azerbaijan.

Voter ID was one of the recommendations of the Joint Standing Committee of Electoral Matters in its report on the 2019 election (as it also had been in the 2016 report).  The 2019 recommendation was light on evidence and discussed no relevant expert in favour but instead cited Gideon Rozner of the Institute of Public Affairs.  Rozner had stated twice at a verbal hearing that turnout had increased when Queensland had had a similar voter ID law for the 2015 election.  He provided no evidence that this was the case, and the claim was not queried.  It is, in fact, false by the official definition of turnout (see Queensland 2015 section below) though also probably true in an effective unofficial sense.  Mostly the 2019 report referred back to the 2016 report, and the 2016 report referred back to the 2013 report  which has a much more detailed discussion (pp 112-119).  

The more recent history of the idea is intriguing.  Nothing had been heard further of the JSCEM recommendation until One Nation Senator Malcolm Roberts moved a private member's bill containing both voter ID and Senate ballot scan accuracy testing.  A handful of us went through the motions and put in submissions panning the former and commending the principle of the latter but not the execution, and the Senate Finance Committee recommended the Bill not be proceeded with.  It was surreal (and in the case of Senate accuracy testing very pleasing) to see Government bills for both issues on the notice paper within a fortnight.  Pauline Hanson has claimed credit.

The proposal is that a very wide range of forms of voter ID (see list page 8) will be acceptable, including voter cards mailed out by the AEC.  If a person has no ID they can have their identity attested to by another elector who shows their own ID. Failing that they can cast a declaration vote in which they declare their name and address, and their vote is admitted once it is established that they have not voted elsewhere.  Contrary to countless false claims about the proposal, nobody will be unable to vote on account specifically of not having or bringing any ID.

How Much Multiple Voting Happens In Australia?

Every federal election thousands of voters are marked as having voted more than once.  However many of these apparent multiple votes are resolved as cases of matching clerical error once voters are asked to explain having apparently voted more than once or not at all.  Others result from memory issues (especially among elderly voters), confusion about the electoral process, or serious mental health issues.  Deliberate multiple voting by legally accountable voters appears very rare.

During the debate supporters of voter ID have tended to claim that there are many thousands of multiple votes (sometimes over 10,000) and opponents have tended to claim that the actual number is about 20 per election.  However, the former group tend to include all multiple markoffs as multiple votes (although most are clerical errors) while the latter tend to include only those cases referred to the Australian Federal Police.  Because genuine double voting cases resulting from system confusion or memory issues are not normally reported to the AFP, this greatly underestimates how many genuine double votes there are.  I suspect it also underestimates how many "voter fraud" (deliberate and without valid excuse) cases occur.

An especially silly example of the ability of Twitter "left" echo chambers to mangle the facts about anything has been the viral false claim that only 19 people voted twice in the 2019 election:


Over 1000 retweets and over 5000 likes but this is nonsense!  The origin of it is that people searching for information about multiple voting found a Crikey article from 2017  (and Crikey also retweeted it).  Versions of the table from that article circulated, and somehow 19 - the number of AFP referrals for alleged deliberate multiple voting in 2010 - became the number of people who voted twice in total in 2019 (years after the source article was even written).

Here are some tables regarding the actual incidence of multiple voting recently.  Firstly for 1993-2010 from an excellent report by Rodney Smith:


Now the 2010-2016 table reproduced by Crikey:


In all these years, there were over 500, and in more recent cases approaching 2000, self-admitted cases of multiple voting.  However, most were not fraud.  For 2010-2016, over 80% of those were discounted as caused by confusion, poor comprehension or age (typically over 80). In most years very few cases were considered suspicious enough to refer to the AFP, though a number of voters received warning letters without further action.  2013 was an outlier in which the AEC took a once-off limiting view about its level of discretion and hence referred a very large number of cases.

While various cases were referred to the Director of Public Prosecutions from 1993-2004, I can find no evidence of a successful prosecution for multiple voting in a federal election; if there was one I suspect it would be several decades ago.  However, it has recently occurred at state level: in Queensland 2020, a voter was fined $750 for voting twice after she voted by mail in her son's name with his permission (her son did not vote).  She was caught out when evidence was recorded in the course of an unrelated drugs investigation.  Barring unusual circumstances like this, there generally isn't enough evidence to get even the most suspicious multiple  voters into the courts - we know markoff is a fallible process, so it doesn't by itself prove someone voted more than once.

Concerning the 2019 figures, only 2102 voters were recorded with "multiple marks" (source), a figure dramatically below previous years for reasons that I am not sure of (increased electronic markoff is likely to have been a factor).  AEC Commissioner Tom Rogers also said the number of multiple marks was around 0.03% of the total, which would be around 4500 votes (I suspect counting two for every apparent double voting case, consistent with usual experience that most multiple voters vote precisely twice and no more.)   There were apparemtly 743 admitted multiple votes.  Only 24 cases were referred to the AFP (source).  

The 4500 figure appears to be the source of a "0.03%" estimate that has been circulating, but note that that is an overestimate of the number of extra votes for the reasons noted above.  

A case that has created some commentary was Herbert in 2016.  The LNP lost Herbert by 37 votes and decided not to challenge the outcome in court.  Subsequently it turned out that there had been around 200 voters with multiple markoffs.  However, around half were quickly dismissed as clerical errors.  There is a very interesting transcript re this discussing how the AFP was sent to investigate many cases urgently because of the potential for the result of the seat to be affected. Had the AEC believed that there was enough in the multiple voting situation in Herbert to place the outcome in doubt, the AEC would itself have petitioned against the outcome, which could have led to a by-election.  More information would be of interest concerning the basis on which the AEC presumably concluded that the result was safe.

Another close seat where more concrete information is available re multiple voting was the Queensland seat of Chatsworth 2009.  As noted by Antony Green, 30 incidents of apparent multiple voting were investigated with only two found to be genuine, both involving elderly and confused voters who voted both by mail and at a booth.  

I have some experience of multiple marking because I was a "multiple mark" in the 1999 Republic Referendum.  The AEC officials, some of whom knew me, were bemused to find me recorded as voting both at Tolosa Street and Longford but did their job and sent me a letter like they would for anyone else.  I replied stating I had voted in Longford and naming witnesses to the impossibility of me voting at Tolosa Street.  It was most likely a clerical error since it has never happened again, and Tolosa Street was an unlikely booth for any of my political enemies at that time to have voted at.  

Overall I suspect the number of multiple or otherwise fraudulent votes where the voter does not have a valid excuse each election might be a few to several hundred, typically a handful per division.  Some day some very close seat may be voided and rerun as a result.  To put the number, whatever it is, in context, around 400,000 voters have their vote disallowed for failing to number all the boxes in the proper sequence. 

Queensland 2015

The 2015 Queensland election is the major test Australia has had of similar voter ID laws.  Opponents point to a drop in turnout.  Supporters point to a low rate of voters turning up without ID to vote, and also sometimes claim turnout increased (a la Rozner).  Who is correct?

The answer is that turnout fell slightly as a proportion of enrolled voters (from 91.00% to 89.89%) but most likely increased as proportion of voters entitled to enrol.  At around this time, the proportion of people eligible to vote who were enrolled was increasing across the country as automatic enrolment procedures swelled the rolls, but increasingly this added voters who were less likely to actually vote.  I don't have the exact figures but through roughly the time frame between the 2012 and 2015 elections the percentage of eligible Queenslanders who were enrolled with the AEC increased by about 3%.  So it's likely a higher percentage of eligible-to-enrol Queenslanders voted in the 2015 state election compared to 2012 - even though the official turnout figure was lower.  (In my view, to discourage debacles such as the SMH's premature and false claim of a record low turnout in 2019 - which the SMH has still not retracted - two separate official turnout figures should be reported.)

The unconvincing nature of pointing to Queensland's 1.11% official drop in turnout (January 2015) can be seen by following how other states and territories fared in official turnout terms immediately after.  Following a few years where nothing much happened in turnout terms in state and federal elections (except WA 2013 where there was an increase following a drop at the previous election) Queensland 2015 was the start of a string of significant (and often much larger) turnout drops:

Qld (Jan 2015) -1.11% (Voter ID added)
NSW (March 2015) -2.06%
Federal (July 2016) -2.05%
NT (Aug 2016) -2.89%
ACT (Oct 2016) -0.81%
WA (March 2017) -2.31%
Qld (Nov 2017) -2.37% (Voter ID removed)
Tas (March 2018) -2.15%
SA (Mar 2018) -0.98%
Vic (Nov 2018) -2.86%

Turnout as a percentage of enrolled voters fell in every state and territory in the following cycle, and fell again and by more in Queensland after voter ID was removed, so the claim that the voter ID laws caused the modest downturn in official turnout in Qld 2015 (which was an increase in overall participation anyway) is one that requires further argument.

One might argue that turnout should have been up because of the unusual nature of the election - both a massive swing and an expected close election, with enthusiasm for chucking out Campbell Newman contributing to a surge of enrolments as the rolls closed.  On the other hand, the election was in late January, a good reason why turnout might have dropped.   So there is not clear evidence that the official turnout drop, given increased enrolment, was unusual.  

Another place we might look for evidence is in changes by division.  Here, a detailed paper about voter ID at the election by Graeme Orr and Tracey Arklay has noted that divisions with high Indigenous populations and also divisions with low population density tended unsurprisingly to have relatively high rates of "uncertain ID" voting.  (Supporters of voter ID point to the very low level of "uncertain ID" voting overall, about 0.7%, as evidence of near-universal compliance.)  The high rates of uncertain ID voting in the divisions listed (Mt Isa, Cairns, Warrego, Cook, Keppel, Townsville, Thuringowa, Barron River), however, don't correspond to an unusual decline in turnout.  Turnout was down in all these divisions, but by less than the state average (0.82%) and with only two of the eight (Mt Isa down 1.83% and Cairns down 2.59%) exceeding the state average.  There may well have been confounding factors including anti-Newman factors working to boost turnout in these divisions compared to others, but again, proving that voter ID deflated turnout in Queensland 2015, especially in disadvantaged communities, is not that simple.

What multiple votes would or would not be stopped?

The proposed voter ID bill would be unlikely to stop any of the following:

* A voter who votes multiple times because of a mental health issue.

* A voter who votes multiple times because they forget they have already voted.

* A voter who votes multiple times because they are confused about the system.

Its targets would be: voters who deliberately (and without a mental health related compulsion) vote multiple times in their own name, and voters who impersonate other voters.  The latter can be divided into overt impersonators (who impersonate a voter who they would have expected to vote) and covert impersonators (who impersonate a voter they know won't vote).  The latter can in turn be divided into consensual and non-consensual cases.  

These voters are together a small portion of all multiple voters but exactly how small isn't clear, because of gaps between the numbers of voters the AEC accounts for in other ways in the tables and those it decides to refer to the AFP.  Also, covert impersonators aren't getting detected as multiple votes at present, except in the Queensland case mentioned - there may be more of them than is obvious.  But voter ID will not necessarily stop:

* a consensual covert impersonator who is given non-photographic ID of the voter (such as an electricity bill), or a non-consensual covert impersonator who has access to the same - though this won't always be possible (eg obviously male-looking voters impersonating a more or less exclusively female name could get caught)

* any overt impersonator, or non-consensual covert impersonator, who is willing to create and use fake ID (which given the range of ID allowed, won't be particularly difficult)

A voter can even still present ID and vote as themselves deliberately at multiple booths.  The risk of prosecution exists but they might get away with it by claiming confusion or simply claiming that someone else must have faked or stolen an ID document.  Unless there is evidence linking them to locations where they cast multiple votes, how would the case against them be proven?  

Given that the voter ID required for an ordinary vote isn't photo ID, one might ask why not ask for a higher standard?  One reason is that pretty much any level of ID can be faked if someone is determined enough.  Another is that encouraging voters to cart important ID documents around with them unnecessarily isn't a good idea.  (This too I know from experience - I once lost a passport while unnecessarily carrying it for ID purposes.  It somehow fell into and lodged itself in the end of a blanket and was only found months later, after I had had to replace it.)

Who might be deterred?

Opponents of the bill have likened it to "voter suppression", but I think it's the opponents who will be doing most of the suppressing here.  It is not comparable to the USA with its voluntary voting system where voters might be severely discouraged from voting at all.  In Australia booth attendance (or voting by some other means) is compulsory.  In theory at least, no voters would be "disenfranchised".

It's still possible to imagine some voters missing out.  For instance, a voter falsely believes that if they have no ID they will not be allowed to vote (perhaps because they have swallowed too much nonsense spread by the bill's opponents, which raises the question of why those opponents are making such claims in the first place and risking damaging their own side's vote.)   Another scenario is this: a disorganised voter heads to the booth at 5:40 pm, realises they forgot to bring ID with them, goes home to get it because they want to do the right thing, underestimates the traffic and doesn't get back to the booth in time.  

Other voters who might end up not voting include voters with some mental health issues who might be paranoid about showing ID, voters who are ashamed to show up without any ID, or voters who became hopelessly confused or frustrated during the declaration vote process.  Overall, I doubt the number of such voters would be noticeable amid all the other turnout noise.  There is, however, some risk that polling staff will not all be across the procedures and that some voters without ID may be incorrectly screened out, especially if implementation is rushed.

An issue that has not received enough attention (except from eg Antony Green) is the matter of how declaration votes will be handled.  This and other loose ends might be usefully referred to a specific JSCEM inquiry rather than bringing in a change at short notice and with the likelihood of teething problems (such as the false positives for double voting in Queensland 2015 that Antony refers to.) 

Existing step against multiple voting

The Parliament has already addressed own-name multiple voting by setting up a special roll on which suspected multiple voters can be placed thus requiring them to cast a declaration vote.  That won't catch impersonators, but there's no evidence as to whether overt impersonation at least is much of a problem.   I like this solution because it does something to combat whatever problem might exist without having any obvious downside.  A problem with bringing in two solutions at once, as would occur if Voter ID was now added, is that whatever progress is observed cannot be clearly attributed to one alone so you don't really know what worked and what didn't.    

Voter assurance theatre

Finally, an argument for voter ID has been that the system should not only be secure against voter fraud but that voters should feel that it is secure, even if the measures do nothing much to deter fraud or there is really no voter fraud to be secure against.  To me this is redolent of security theatre at airports, or leftover hygiene theatre measures from the early days of the COVID pandemic.  There is some evidence (see APH link below) that a pretty high proportion of voters think that voter fraud happens.  I think voter ID probably would do something useful to curb a certain crankiness about the voting process that has been evident especially on the right among One Nation voters and so on for some time.  However, what is gained on the right will be lost on the left as voter ID will become just another of the string of excuses that are deployed by the sillier end of Labor's support base whenever an election is lost.  The same people who are still banging on about United Australia preferences (notwithstanding that the uneven split of said preferences won the Coalition only one seat while the same with Greens preferences won Labor fifteen) would have a field day alleging that the election was "stolen" by "American-style voter suppression", even if this was not even remotely true.  

I'm also concerned about anything that increases booth workers' workload and detract from other important duties.  It's important for instance that booth workers tell voters to number at least six boxes above the line or twelve below, not just "1-6 above or 1-12 below".  The more booth workers are made to do, the more they might get wrong.

More comments may be added as the debate proceeds.  

See also
















Saturday, November 20, 2021

Labor Dominates In Recent Victorian State Polling

Recently two polls have shown that the Victorian Labor government of Daniel Andrews is currently crushing the Liberal opposition led by the recently returned Matthew Guy.  A Morgan SMS poll showing a 58-42 lead to Labor raised some eyebrows (mostly because SMS polling has in the past been volatile and trashy) but the first branded-as-such Newspoll for the term has reported the same 2PP.  In fact, the Morgan poll's primaries (ALP 43 L-NP 31 Greens 11 others 15) were even friendlier to Labor than Newspoll's (44-36-11-9) and but for the use of respondent preferencing the Morgan poll may well have come out 60-40.  This follows a Resolve poll in late October that had ALP 38 L-NP 34 Greens 10 IND 11 Others 7.  Resolve overestimates Independents, perhaps especially at Labor's expense.  I estimated the 2PP for that poll at 55.5 to Labor.  


Friday, November 5, 2021

Poll Roundup: Albanese Is Still Rating Below Morrison - Is That A Problem?

Since my last federal polling roundup, there has been quite a lot of new polling but the Morrison government's position doesn't seem to have improved.  We have seen:

* Newspoll (early October) 53-47 to Labor (Coalition 37 Labor 37 Green 11 One Nation 2 others 13)

* Newspoll (late October) 54-46 to Labor (Coalition 35 ALP 38 Green 11 One Nation 3 others 13)

* Essential (late October) 49-44 to Labor using Essential's 2PP+ method (Coalition 37 Labor 36 Greens 10 One Nation 3 others 8 undecided 6).  I get 52.2 to Labor by standard last-election preferences with undecided removed.

* Essential (early October) 46-45 to Labor using Essential's 2PP+ method (Coalition 36 Labor 34 Greens 9 One Nation 4 others 8 undecided 9.  I get 51.2 to Labor by standard last-election preferences with undecided removed.

* Morgan (mid-late Sept) 54-46 to Labor (respondent preferences) (Coalition 36 Labor 36 Greens 12.5 PHON 3.5 others 12).  This would be c. 53.1 to Labor by last-election preferences

* Morgan (early Oct) 53-47 to Labor (respondent preferences) (Coalition 37.5 Labor 36 Greens 11.5 PHON 3 others 12).  This would be c. 52.1 to Labor by last-election preferences

* Morgan (mid-late Oct) 54-46 to Labor (respondent preferences) (Coalition 36.5 Labor 35 Green 13.5 PHON 3.5 others 11.5.) This would be c. 52.7 to Labor by last-election preferences 

* Resolve (late October) Coalition 37 ALP 34 Green 11 PHON 3 IND 9 Others 5.  Resolve doesn't estimate a 2PP, and overestimates independents as discussed last time.  I estimate 52.1 to Labor by last-election preferences.