How tough is the QLD election?

Senator Cherry spent 8 years as the Democrats’ electoral
numbercruncher and old habits die hard. He provides an analysis of the
Queensland electoral statistics, which suggests the official electoral
pendulums overstate the safety margin of the Beattie Government.

And ABC election analyst Antony Green responds and provides his own revised election pendulum below.
There is no question that Peter Beattie’s Labor Government
enters the 2004 State Election with a thumping majority, holding 66 out
of 89 seats in the Parliament. But, the electoral statistics to some
extent overplay the ‘safeness’ of Labor’s margin due to Queensland’s
optional preferential voting system.

The electoral pendulum published by most media outlets (including
yesterday’s ‘Courier Mail’) base the pendulum on the two-party
preferred vote calculated by the Electoral Commission. On this basis,
Labor would lose office with a uniform swing of 9.5%, with Glasshouse
being the 22nd seat to fall.

But the Commission’s data has a flaw in it. It calculates the two party
preferred vote based on the votes left in the count, excluding those
votes that exhausted due to voters not allocating preferences. As a
result of the ‘optional preferential’ voting system,  Labor’s
winning vote was less than 50% of the formal vote in 14 of the 22 most
marginal seats, as as low as 41.7% in Burdekin where 24.4% of the
formal vote exhausted.

Taking Burdekin as an example, Labor’s winning majority of  1648
votes was 55.13% of the remaining two-party preferred vote. But, to
overcome this lead, the Nationals need a swing of just 3.9% of the
total primary vote to win, not the 5.1% shown in the pendulum. Further,
if One Nation and its largely defunct offshoot the City County Alliance
had run a single candidate in Burdekin instead of two candidates, they
would have won the seat with a 6.2% buffer in 2001. Hervey Bay would
also become a line ball seat on this basis, instead of the misleading
7.6% margin estimated by the Commission.

If the One Nation votes fell by 5% at this election in favour of
the Nationals, the lower exhausting rate would see the National win
Burdekin with a 0.1% majority, without winning a single extra vote off
Labor. In fact, if the Nationals win a 5% vote switch from One Nation
(which averaged 18% across Queensland in 2001 compared with 3% in the
most recent  Newspoll and 6% in Morgan poll), the Labor seats of
Noosa, Burleigh, Toowoomba North, Kawana and Burdekin all shift to the
Coalition, with Charters Towers knife edge. That is without any direct
swing from Labor to the Coalition but purely because of a lower level
of conservative exhausted votes.

The elimination of three-cornered contests between the Liberals and the
Nationals also serve to reduce Labor’s margin in Springwood (from 10.3%
to 6.8%) and Glasshouse (9.5% to 5.9%) and to a more limited extent in
Thuringowa.

Finally, the Greens ran in only 8 of the 25 most marginal seats in
2001. The average exhaust rate for the Greens was 43%, with  most
of their vote coming direct off Labor. This suggests that if the Greens
run in any of the other 17 marginal seats, assuming a 5% vote shown in
Newspoll/Morgan, they could pare up to 2% off Labor’s final vote by
increasing the number of exhausted vote . This could affect the outcome
in seats like Noosa, Burleigh, Broadwater, Thuringowa and possibly
Kawana where the Greens did not run last time (if they run this time).

Correcting the pendulum results in the trigger figure falling as follows for Labor to lose 22 seats:

• ECQ two party preferred figure: 9.5%
• Margins calculated as % of formal vote: 8.2%
• Elimination of Lib/Nat contests: 8.2%
• 5% shift from One Nation to Nats: 7.6%

REVISED ELECTORAL PENDULUM:

– First figure: Two party preferred vote as calculated by the ECQ
– Second figure: Vote changes needed calculated as percentage of formal vote
– Effect of single ONP/CCA or Lib/Nat candidate also indicated, as well
as impact of 5% swing from ONP to Nats. (Negative figure assumes
Non-government majority. Second candidate assumed to be Coalition
unless specified).
– Labor would need to lose 22 seats to lose its majority.

ABC election analyst Antony Green responds to John Cherry and provides his own revised electoral pendulum:

I was interested to read John Cherry’s comments yesterday on the
swing need for Peter Beattie to be defeated at the February 7 election.
After the slagging off I received from Crikey the other day, I though a
suitable peace offering might be the little contribution below on
optional preferential voting and my own alternative electoral pendulum
for Queensland.

It is without doubt that Labor received a substantial advantage in 2001
from the operation of Optional Preferential Voting (OPV). As John
pointed out, the result in Burdekin was truly remarkable. Labor polled
36.7% of the primary vote, the three conservative candidates 63.3%
between them, but after the distribution of preferences, Labor won
thanks to the high number of exhausted preferences.

Equally remarkable was Hinchinbrook. After the full distribution of
preferences, 7,192 votes elected the National Party’s Marcus Rowell,
6,436 chose the One Nation candidate, and 6,713 votes had exhausted all
preferences. 33% of all votes ended up expressing no preference for
either of the two remaining candidates. I think we can call this the
‘none of the above’ vote.

For the technically minded, you calculate a percentage by dividing the
numerator by the denominator and multiplying by 100. The numerator is
either the primary or 2-candidate votes, while the denominator is the
formal votes under compulsory preferential voting (CPV) or the votes
remaining in the count under OPV.
Under CPV, your percentage goes up only when preferences are received,
the denominator being fixed at all times. Under OPV, your percentage
also goes up when preferences exhaust, as the denominator is cut by
each exhausted vote. But the key point is, the percentage vote of all
candidates does not go up equally. The percentage of the candidate with
the highest primary vote always goes up fastest with each exhausted
vote. This is why Labor won seats like Burdekin. 58% of preferences
exhausted, and as Labor had the highest primary vote, its vote was
weighted upward.

So can we come up with a measure that removes the exhausted vote
effect. This is effectively to come up with a measure as to what would
have occurred under CPV rather than OPV. The assumption I make is to
take all the votes that did not have preferences, and distribute them
to the two final candidate in the same proportion as the votes that did
have preferences. I think this is a reasonable approach to take, and is
certainly the most defensible assumption when there is in fact no
empirical data to indicate why people exhaust their votes.

Applying this assumption and calculating the percentages, the
differences between OPV percentage and the CPV estimated percentage can
be seen as the advantage gained  by exhausted preferences.

As an example, a hypothetical election of 1,000 votes as follows.

A – 450 (45.0%) / B – 400 (40.0%) / C- 150 (15.0%)
Say of candidate C’s votes, 10 had preferences for A, 40 for B and 100
votes exhausted. Before we go and distribute the preferences, first
exclude the 100 exhausted votes and re-calculate the percentages.

A – 450 (50%) / B – 400 (44.4%) / C – 50 (5.6%)
So you see, A’s vote went up 5% (and achieved a majority without
receiving any preferences), while B’s rose 4.4%. The percentage of the
candidate with the highest primary count is always boosted in this way.
The leading candidate on primary votes is always advantage by OPV.
(That is unless the majority of exhausted preferences would otherwise
have come to them). The final distribution of preferences then gives

A – 460 (51.1%) / B – 440 (48.9%)
Now, what if all C’s votes had expressed preferences in the same
proportion to those votes that did. Of the 50 votes with preferences,
10 (20%) had preferences for A. So 20% of all 150 votes is 30, and by
the same reasoning, 120 for B. My CPV estimate for this election would
now be

A – 480 (48.0%) / B – 520 (52.0%)
A very different result. Under OPV, candidate A ended up with 51.1%,
under my CPV estimate, only 48.0%. If I subtract the estimate from the
actual percentage, you get a measure of advantage received from the
operation of OPV, in this case +3.9%.

Now I have done this for all Queensland electorates. Below is the same
table of marginal seats as used by John Cherry. The number on the left
is the ordering on the conventional pendulum. I have re-ordered the
seats according to my CPV estimate, and included an advantage column.

ANTONY GREEN’S REVISED ELECTORAL PENDULUM:

As you can see, the seats of Burdekin, Thuringowa, Charters Towers and
possibly Burleigh, are in Labor hands in large part because of optional
preferential voting. On this pendulum, the swing for Labor to lose 22
seats and its majority comes down to 6.8%, and quite a few seats that
are currently on margins above 4% suddenly look much more marginal.

One thing to be careful of though is to assume too much of the victory
in 2001 was just optional preferential voting. Sure, Burdekin was a
freak result. The same almost occurred in Gympie much to Labor’s
amazement. But the seat of Burnett is not that dissimilar to Burdekin,
another mid-north coast sugar seat. Burnett only had two candidates, so
no problem with competing conservative candidate. Labor romped home to
win for the first time since 1909 with an 11% swing.

There were 16 seats in 2001 where my advantage/disadvantage measure was
greater than 4%. All were seats with competing conservative candidates.
Given a choice with optional preferential voting, voters expressed a
primary vote intention, but then did not give preferences. If there had
been only two candidates, those same voters would have had no other
option except abstention, voting informal or choosing between Labor and
a single conservative candidate. The evidence from Burnett is that
voters may have been prepared to vote Labor so strong was their anger.
At this election, without competing conservative candidates, it is hard
to see some of the 2001 results being repeated.

Another warning to take from the above table is in the entry for
Indooroopilly. Labor was disadvantaged there. While Labor led on the
primary votes, it’s majority was then built on Green and Democrat
preferences. Overall Labor won 62.4% of preferences from votes with
preferences, but just over a third of all distributed votes exhausted.
If the Greens do well at this election, and their is a high exhausted
rate, then Labor could be the party that suffers from OPV instead of
the the Coalition.

After all, that was one of the problems for Wayne Goss in 1995. Labor
would probably not have had its problems in Mundingburra at that
election, or lost Mulgrave, if the government had not so alienated
Green voters.

Senator Cherry responds to Green’s argument:

“Antony Green’s ‘revised’ electoral pendulum highlights the real
advantage that the optional preferential voting system had for Labor in
2001 with One Nation cannabalising the Coalition vote.

But, it also highlights the risks for Labor in 2004 with One Nation’s
vote in decline and the Greens now contesting upwards of 70 seats,
twice the number in 2004.

Simply put, and despite all the rhetoric about Greens preference deals
in today’s papers, Green voters are incredibly ill-disciplined when it
comes to preference allocations.

In 2001, the Greens directed preferences to Labor in every marginal
seat in Queensland bar two (Greenslopes and Indooroopilly). Yet, in the
six most marginal seats won by Labor which the Greens contested
(Clayfield, Barron River, Glasshouse, Indooroopilly, Gaven, Mudgerraba
and Mt
Ommanney), 65% of Green voters ignored the Greens how to vote card and
only 35% gave their preferences to Labor – 47% allocated no preferences
and 18% gave preferences to the Coalition or others.

Ironically, in Indooroopilly where the Greens didn’t allocate
preferences, the flow to Labor was stronger (52%) and the exhaust rate
was lower (28%) than the seats where they did.

Thus the decision by the Greens to double the number of candidates they
contest is likely to cost Labor around 2-3% of the two party preferred
vote in every seat the Greens contest even if the Greens decide (as
they usually do) to direct preferences to Labor. On the other hand,
their poll results (4% in the latest Morgan and Newpoll) suggest that
the Green vote is not that much moved from the 2001 result.

Labor is still deciding whether to run its ‘just vote one’ campaign
again at this election. But it is likely that regardless of the
campaign, many Queensland voters will again decide to exhaust their
preferences, which could hurt Labor more than the Coalition in this
election.”