Friday, 11 March 2011

Did AV really decrease turnout in Australia?

There’s a throwaway charge made towards the Alternative Voting system that has always bugged me. It’s the charge that AV will cause the turnout in this country to drop through its introduction, and the “proof” of this is supposedly a single year’s turnout figures in Australia, the 1922 Federal Elections. So, did AV, or IRV, cause low turnout at the Australian Federal Elections?

But is this even proof? Statistically it’s flimsy, looking at only two years of turnout for comparison, near the beginning of a parliament’s lifetime, and not long after a war effort. But I’ve decided to take it on it’s own basis and look deeper...and I think the realities of what can be drawn from the drop in turnout are not quite what the “No2AV” camp would enjoy.

First, let’s get some background on the country and it’s politics prior to AV being implemented.

We know from our own history that times of great internal discomfort lead to higher turnouts. 1917 was a time when the war was having a huge toll on the Australian’s resolve (as it was, no doubt, everywhere), a controversial attempt was made to try and get the people to agree to conscription, and heavy economic penalties for having got involved in the war, kept turn out high, one of the highest Australia has had pre-compulsory voting.

It’s easy to compare this year for the Australian government to our own 1997 or 2010 elections where national events conspired to push the public in to action to reform it’s own government.

There was also high turn out in 1914 and 1913, however, the former “helped” by the outbreak of the first world war that would later come to hurt the Labor party, and the latter by an increase in in turnout by “Liberal” (conservative) voters in response perhaps to the biggest socialist set of reforms of Australia’s then short history in the preceding government.

Before this point turnout was low, once in the low 60%’s and otherwise in the 50%’s. This is probably more to do with the relatively new status of federal Australia and how it operated...it’s legitimacy in the eyes of the people and how well it was in itself supported...rather than an indictment of the FPTP system that was being used.

History lesson over, what about 1919 and 1922? I’ve gone through the elections from 1914 to 1922, seat by seat to work out if there is anything that can be said to be a clear indicator of a trend. Is there an apathy when asked to rank between 3 or more voters, for example, instead of two? Is there a marginality effect in 1919 that isn’t there in 1922?

Well let’s start first with the candidates. In 1914 and 1917 the FPTP system meant very few seats stood more than 2 candidates, and in 12 and 10 seats respectively they stood only 1 candidate uncontested. This fell to 2 uncontested seats in 1919 and 4 in 1922. In 1919, when the AV system was introduced there were only 27 seats contesting more than 2 candidates, and in 1922 this rose to 44 seats, not even 60% of the total number of seats.

So, for a start we know that roughly around 4% of the 12% drop in turnout was down to two uncontested seats.

Next there is the drop in turnout. From the 1919 year to 1922 there were 27 seats that can say they lost turnout after moving from staging a contest between 2 candidates to a contest between 3 or more candidates. However 29 seats lost turn out by keeping exactly the same number of candidates, while 8 lost turnout despite REDUCING the number of candidates fielded in 1922.

What’s most interesting is that of the top 5 biggest drops in turnout, all 5 either decreased the number of candidates standing or remained contesting with only 2. Turnout drops of 26%, 24%, 23%, and none of them to do with any added complication of having to choose between more than two candidates!

Of the 3 seats that can be said to have gained turnout during 1922, the only seats we can be sure of doing so (due to other 5 other seats having been abolished/reformed between the elections), 2 increased their candidate list to 3 candidates standing, and the other REMAINED on 3 candidates.

Then there is the phenomenon of vote counts actually not dropping, despite turnout doing so, suggesting that in reality a lot of these seats gained a significant amount of non-voting population between 1919 and 1922, rather than people choosing to not vote.

For example, Capricornica... an extra 10600 voters, but only an increase of 0.6% in the turnout. Wide Bay, 5400 extra voters but LOST 3.1% in turnout. Darling, increased it’s vote count by 1700 but lost a huge 12.5% of it’s turnout. Kalgoorlie remained stable on voters but lost an even bigger 20.9% of it’s turnout, similarly Batman lost only 200 voters but also a turnout share of 15.9%

The list goes on, in fact there are only 10 constituencies that have anything like a logical correlation between their turnout figure and voting total figures (comparing the drop or gain in turn out to a reasonable matching percentage... 10% of turnout equates to roughly 3000-4000 votes if the constituency size is staying roughly equal).

This isn’t to say what happened was unbelievable, just that there was clearly a population upheaval in some respect between 1919 and 1922 that caused placed to drop turnout so suddenly while increasing their vote totals

On this subject, of those places increasing their vote total (the number of ballots cast), of 17 constituencies 8 increased their number of candidates, 8 remained with the same number of candidates (2 of those with 3 candidates) and 1 had decreased their number of candidates.

On the flip side, the 17 (to be consistent) constituencies that lost the most vote totals consisted of 10 remaining on the same number of candidates (2 of those with 3 candidates), 5 increasing the number of candidates and 2 decreasing their number of candidates.

And when it comes to marginality the obvious expectation seems true. The most marginal seats in 1919 were the best at increasing vote totals, and tended to minimise the loss in turn out. in the seats where the margin was about 5.5% or lower in 1919, out of 16 seats 9 of them gained vote numbers, with an average drop of about 9% in turnout boosted in part by a few modest turnout increases in 2 areas.

Take the 16 from the side with the least marginality and that average drop in turnout goes up to 15%, no turnout increases whatsoever in any of those areas, none of them gaining higher vote totals.

Just splitting the seats in to two halves shows the half with the more marginal seats had a 3% better performance on turnout than those without marignality.

Finally there’s the issue of the longer turnout trend. 12 of the 75 seats, at least, experienced an ongoing turnout drain from 1914 all the way through to 1922. The introduction of AV didn’t alter that direction. Another 22 seats increased or maintained their turnout in 1919 compared to 1914, a good result considering that 1914 was a year of going to war and 1917 was hot off the heels of controversial conscription legislation being defeated.

Due to uncontested seats, and some seats being incomparable through boundary changes, this only leaves 19 seats that it can be said, for sure, that lost turn out after the introduction of AV despite turnout increasing in 1917. 14 of those seats didn’t contest more than 2 candidates in 1919 and 6 still didn’t contest more than 2 in 1922.

From all of this data we can see that marginality still has a strong pull on how turnout changes in line with the national mood, we can also see that constituencies that adopted more candidates were more likely to get more votes cast, even if this resulted in a drop in turnout due to population factors. Consequently the seats more susceptible to lower turnout were those with fewer candidates standing.

And finally we can see that out of all the seats that we have data for, only a maximum of 15 (from 75, including seats we have not enough data for) can actually be identified as having consistently lost turnout while moving towards multiple preferences.

(As a side note, it was interesting to find in my research that Australia had several seats that uses AV as a means to, essentially, run primaries during the election. Multiple candidates from the same party standing against each other, resulting in a head to head between two people from the same party in some cases...another benefit of AV that I’ve talked about in the past, actually evidenced in use!)

Of course this is just what we can measure! There are plenty of other unaccountable things that could be at play that could affect why turnout changed both from a national and a local perspective:

As with 2001 in the UK, a huge turnout drop after the 1997 high turnout, there is no reason why a stabilising country after war simply wasn’t interested in politics, especially during a time where all the parties seemed to be chopping and changing so frequently to find the right identity and level of power. Maybe people just genuinely didn’t care who was winning.

The usual local issues could have their effect...multiple candidates not different enough to warrant choosing between, poor local campaigning, lack of faith in the system from a local perspective. These feelings can change election to election.

Was centralisation of government a problem from the start in terms of endorsement? We know that turnout was in the 50% area at it’s inception, was it only war and it’s aftermath that grabbed the nation enough to be inspired to vote up to the 70% and above mark in the first place?

Did immigration play it’s part, how did the “type” of population that was arriving in it’s largest numbers since the early 1910’s in the country play it’s part in who felt the need to vote and who didn’t?

How did the Spanish Influenza pandemic of 1918/1919 affect the turn out, especially in Tasmania where around a third of the population were affected? (Coincidentally, perhaps, Tasmanian turnout dropped by around a third that year, unlike most other areas in Australia)

Perhaps it was to do with an aspect of the system we can’t measure...the fact that Australians have to rank every candidate standing may have felt counter productive for example. “Vote 2nd preference for the candidate I dislike?!” I can just imagine the absurdity of the notion at the time!

So...the conclusion:

We just don’t know why turn out dropped in 1922 to the level it did! The figures seem to show that if there was a problem it wasn’t in the preferential nature of the voting system not long since introduced. Voting numbers went up, yet turnout fell dramatically; voting dropped in areas where they had to choose between only two candidates just as much where they had to suddenly rank between three or more.

Population increase appears to be key, combined with a lack of will to vote, and the long studied trend of “safe seats” helps bring that turnout down lower where it’s occurring. Along with 2 more uncontested seats than in 1919 that can be seen as the cause of a 1/3rd of the drop in turnout, perhaps the mystery of where the voters went isn’t that big of a mystery at all.

The only thing we can genuinely say is that being confronted with ranking more than 2 candidates was not a key factor in a decrease in turn out, and in that sense...given how close 1919 and 1922 Federal elections were to First Past the Post in their application during those early years...I think it is highly speculative, and more than a slight stretch, to believe it was the preferential nature of the system that caused the significant proportion of the drop in turn out.

Given that it is this preferential nature that is the only thing the UK AV system would share with Australia, not the requirement to order all preferences, nor the particular style of constituencies at that time of history of only usually standing two candidates, there is no reason to believe that introducing AV would lead to a decrease in turnout here in the UK either.

But, in the interest of fairness, I should also say that if turnout does increase under an AV election in the UK it is no more a reason to claim that the voting system has enfranchised voters either. Trends are the only thing that matter in this kind of data, several elections at the very least are needed to compare to similar periods of time in history to determine any difference in voter apathy influence because of the system.

When we’re making claims, whether it’s about the cost of the voting system, how it’ll engage the apathetic, or the likely political outcomes of it’s use such as making statements about it’s coalition making powers...we need to actually base these claims on reality, measurable reality. It’s not enough to have some anecdotal evidence, such as 1922, and inflate it with your own meaningless opinions.

EDIT: Link to the data for these elections...

EDIT: Antony Green, who runs an excellent blog that gives plenty of analysis to bogus "No" campaign claims, has also put an insight up as to the lack of link between AV and fall in turnout.