Analysis of Preference Flows at the Upper Hunter By-election

The NSW Upper Hunter by-election on 22 May was notable for the unusually low first preference vote for the two major parties. The Nationals polled 31.2%, Labor 21.2%, with the combined vote for the other 11 candidates an unusually high 47.6%.

The by-election was conducted under NSW’s optional preferential voting rules and 63.4% of other candidate ballot papers exhausted their preferences before reaching one of the final two candidates. At the end of the count, exhausted ballot papers represented 30.2% of the first preference vote.

With ballot paper data from the by-election now published, it is possible to examine more closely the two-party preferred flows of preferences from excluded candidates, to determine how many preferences voters completed, and to measure the influence of how-to-vote recommendations on preference flows.

Table 1 breaks down the two-party preferred preference flows by party. There is an entry in the table for each excluded candidate showing the percentage first preference vote the candidate received, then percentage figures showing the percentage of ballot papers flowing to the National candidate as preferences, the percenatage flowing to Labor, and the percentage of ballot papers that exhausted because the voter did not indicate a preference for either National or Labor.

The table is in descending order of first preference vote. The table includes an indication of each candidate’s how-to-vote (HTV) preference recommendation between National and Labor or whether no recommendation was made.

Table 1 – Two-Party Preferred Preference Flows by Party

% Preferences to
Pos Candidate (prefs to) Party % Vote NAT (5) ALP (9) Exhaust
4 McNamara (None) ONP 12.3 19.5 10.2 70.2
12 Gilroy (ALP) SFF 12.0 15.6 21.6 62.9
1 O’Connell (None) IND 8.8 18.5 20.7 60.8
6 Norman (ALP) IND 4.1 13.7 30.5 55.8
8 Abbott (ALP) GRN 3.5 7.5 45.3 47.1
7 Reynolds (None) IND 2.2 12.2 22.9 64.9
3 Pears (None) LDP 1.5 23.2 6.4 70.3
13 Fraser (None) IND 1.4 10.7 12.7 76.6
11 Dello-Iacovo (None) AJP 0.8 14.6 17.4 68.0
10 Blair (None) SAP 0.8 12.0 27.5 60.5
2 Lea (None) IND 0.3 20.5 9.0 70.5
All Excluded 47.6 16.3 20.3 63.5

Exhausted ballot papers in Table 1 represent formal votes where a voter did not make a final preference choice between the Labor and National candidates. The Greens stand out in the table as the only entry where less than half of all ballot papers exhausted. The Green’s exhausted rate was only 47.1%, with Green preferences flowing 45.3% to Labor and only 7.5% to National.

Independent Tracy Norman, whose HTV recommended preferences for Labor, also had a low exhaustion rate and stronger flows of preferences to Labor. Calum Blair of the Sustainable Australia Party did not have a HTV but was the only other candidate where more than a quarter of ballot papers flowed to Labor, and the SFF HTV recommendation for Labor generated only weak preference flows. Parties that made no preference recommendation between National and Labor tended to have higher rates of exhausted preferences.

Turning to how many preferences were completed, Table 2 summarises the number of further preferences completed for each candidate, listed in descending order of vote. The number of preferences recommended on how-to-votes is shown in brackets. Note that National and Labor preferences were never distributed.

Table 2 – Preference Completion Rates by Party

% of Votes with # of Prefs
Party Candidate (# prefs) % Vote 1 2-4 5-12 13
NAT Layzell (1) 31.2 74.9 15.8 2.5 6.9
ALP Drayton (4) 21.2 45.7 39.9 4.6 9.7
ONP McNamara (3) 12.3 49.6 35.3 3.5 11.5
SFF Gilroy (6) 12.0 50.0 25.1 12.2 12.8
IND O’Connell (4) 8.8 33.2 38.5 8.9 19.4
IND Norman (6) 4.1 40.2 25.6 19.7 14.5
GRN Abbott (6) 3.5 31.4 28.6 21.2 18.8
IND Reynolds (1) 2.2 53.5 28.0 6.5 12.0
LDP Pears (3) 1.5 59.6 22.1 5.7 12.6
IND Fraser (None) 1.4 60.6 19.1 7.6 12.7
AJP Dello-Iacovo (1) 0.8 47.1 27.7 6.0 19.1
SAP Blair (None) 0.8 40.8 27.7 9.1 22.4
IND Lea (1) 0.3 51.3 21.2 9.0 18.6
All candidates 54.6 27.9 6.1 11.1
Excluded Candidates 47.6 45.2 30.4 9.9 14.5

The National Party’s 1-only how-to-vote had a huge influence on its voters with three-quarters of National ballot papers marked with only a single first preferences. The lowest rates of 1-only voters were for the Greens (31.4%), Independent Kirsty O’Connell (33.2%) and Independent Tracey Norman (40.2%).

Of all ballot papers, 54.6% had only a first preferences, but a lower 45.2% in total for the 11 candidates whose preferences were distributed. At the other end of the completion rate scale, overall 11.1% of voters numbered all 13 squares, a slightly higher 14.5% in total for the 11 excluded candidates.

For the seven candidates that recommended a number of preferences, the percentage of ballot papers with that number of preferences were –

  • 22.1% of Labor ballot papers had four preferences
  • 19.4 of votes for Independent O’Connell had four preferences
  • 18.4% of One Nation ballot papers had three preferences
  • 13.8% of Green ballot papers had six preferences
  • 12.7% of Independent Norman ballot papers had six preferences
  • 8.2% of LDP ballot papers had three preferences
  • 7.5% of SFF ballot papers had six preferences

Moving on from the number of preferences completed, Table 3 summarises the influence of how-to-vote material. For each candidate that made a recommendation, the table shows the percentage of votes for the party that matched the second preference, the third preferences and so on. The most preferences recommended on any HTV was six.

Table 3 – Percentage of Ballot Papers in Concordance with Party How-to-Votes

% matching HTV to preference number
Party Candidate (# prefs) % Vote 2 3 4 5 6
ALP Drayton (4) 21.2 23.7 22.1 18.2 .. ..
ONP McNamara (3) 12.3 24.9 11.1 .. .. ..
SFF Gilroy (6) 12.0 19.1 7.8 6.7 5.9 5.4
IND O’Connell (4) 8.8 26.7 15.1 13.0 .. ..
IND Norman (6) 4.1 28.1 11.9 10.4 8.9 8.5
GRN Abbott (6) 3.5 22.2 18.6 14.1 10.1 10.0
LDP Pears (3) 1.5 9.0 2.7 .. .. ..

Voter concordance with a how-to-vote recommendation doesn’t only measure whether voters followed the recommendation. Many voters would never have received a how-to-vote, so the percentages in Table 3 also measure whether voters can work out or guess the same sequence as printed on the HTV.

The how-to-votes for each party can be found on my Upper Hunter Election Guide page.

I make the following observations on how-to vote concordance rates.

  • Labor’s HTV went Norman 2, O’Connell 3 and SFF 4. 18.2% of Labor ballot papers completed there ballot papers with this sequence.
  • One Nation’s HTV was 2 SFF, 3 LDP. Clearly many voters who did not receive a HTV picked the second preference but not the third.
  • The SFF how-to-vote went 2 One Nation, 3 LDP, 4 Independent Reynolds then 5 Labor and 6 National. Voters either followed or picked the One Nation second preference but with much lower concordance rates beyond that.
  • Independents O’Connell and Norman had a well announced second preference swap, and voters clearly knew about it with more than a quarter of voters correctly swapping the second preference. But there was less concordance beyond the second preference. O’Connell’s HTV was to 3,4 to other Independents, while Norman went 3 Labor, 4 Green, 5 SFF and 6 Reynolds.
  • The Green HTV went 2 Norman, 3 O’Connell, 4 Labor, 5 Animal Justice and 6 Fraser. One in ten Green voters picked these five further preferences correctly.
  • The LDP HTV had One Nation second but more LDP voters picked the National as second preference over One Nation. This suggests not many voters saw the HTV.

Table 4 looks specifically at second preference choices of voters. ‘**’ indicates where the second choice of voters matched their chosen party’s how-to-vote.

Table 4 – Second Preference Rates by Party

Party Candidate % Vote % Exhaust % Second Preferences
NAT Layzell 31.2 74.9 One Nation 6.3
ALP Drayton 21.2 45.7 **Norman 23.7, SFF 7.0
ONP McNamara 12.3 49.6 **SFF 24.9, National 10.9, Labor 5.2
SFF Gilroy 12.0 50.0 **One Nation, 19.1, Labor 8.0, National 6.8
IND O’Connell 8.8 33.2 **Norman 26.6, National 7.4, Labor 6.1, Green 6.0
IND Norman 4.1 40.2 **O’Connell 28.1, National 7.3, Greens 6.1, Labor 5.9
GRN Abbott 3.5 31.4 **Norman 22.2, Labor 14.4, O’Connell 9.2, SAP 6.8
IND Reynolds 2.2 53.5 Labor 15.0, SFF 7.5, O’Connell 5.6, One Nation 5.0, National 5.0
LDP Pears 1.5 59.6 National 11.6, **One Nation 9.0, SFF 5.7
IND Fraser 1.4 60.6 O’Connell 9.2, SFF 6.7
AJP Dello-Iacovo 0.8 47.1 SFF 13.1, Greens 11.3, SAP 6.8
SAP Blair 0.8 40.8 O’Connell 12.0, Greens 11.7, Labor 6.7, SFF 6.1, AJP 5.6, Norman 5.1
IND Lea 0.3 51.3 O’Connell 14.1, National 7.7, Norman 5.8

‘**’ indicates the recommended second preference on the party’s how-to-vote.

The failure of Labor and Green voters to give the other party a second preference is the unusual feature of Table four. Only 3.5% of Labor voters gave the Greens their second preference, and Labor’s HTV did not include the Greens. Only 14.4% of Green voters gave Labor their second preference, unusually low as even when the Greens recommend a second preference for a party other than Labor, it is common for Labor to still have the highest second preferences rate. This difference in Upper Hunter is probably explained by the nature of the by-election campaign, with Labor choosing a candidate that strongly backed coal mining. Antipathy in the campaign between Labor and the Greens clearly shows up in the preference data.

Finally, a note on how exhausted preferences inflate the winning candidate’s margin. On first preferences, the National candidate led by 10.0 percentage points, by 8.2 points after distributing preferences, and by 11.6% on calculating percentages after excluding exhausted ballot papers.

Labor gained 9.6% of the formal vote as preferences to the National’s 7.8% to leave the Nationals leading 39.0% to 30.8% as percentages of the formal vote. The 30.2% of formal votes with exhausted preferences was excluded in the final 2PP% calculation, which added 16.8% to the National final 2-party preferred percentage and only 13.4% for Labor.

So while Labor gained on preferences, the Nationals made even greater gains in the final percentage due to exhausted preferences.

Under full preferential voting, the winning candidate’s 2PP% is 2PP votes divided by Formal votes. Under OPV, it is 2PP votes divided (Formal votes – Exhausted votes). Under full preferential voting calculations, 2PP votes varies and formal votes is fixed. Under OPV, both the numerator in the division (2PP votes) and the denominator (Formal – Exhausted) vary, and the rate of exhausted votes always boosts the percentage of the candidate that leads on first preferences.

Other Links I did some broad research on ballot papers at the 2015 NSW election. The findings in Upper Hunter are very much in line with my earlier article that you can find here.