Senate voting changes increases risk of Coalition control: Modelling

In this post, we investigate the impact of the proposed voting changes on splitting of preference flows which may lead to Coalition control of the Senate. This model was produced by James Jansson and submitted to the Joint Standing Committee on Electoral Matters on the 29th February 2016. 

Many claims have been made about vote splitting and optional preferential voting. In particular the Greens and Antony Green have made assertions that a change in preferences will not result in a split of the left vote that will more likely advantage the coalition. These claims are made despite no modelling evidence supporting it. I have decided to create a relatively simple Excel model that shows the impact of vote splitting on a 3 horse race with optional preferential voting.

The model I produced has a particular format and a number of assumptions. Although this is not testing a full senate election, The results are indicative of what could happen under optional preferential voting above-the-line.

The model normally refers to the three parties in this race as Liberal, Labor and The Greens. The purpose of this is to make it familiar and hence easier to understand what is going on in the model. In particular, that there exists a larger party and the two smaller parties are overlapping in their demographic which could potentially split the vote. This model could easily have names switched to represent the opposite ideology represented in the large/small party positions.

Model assumptions:

  • This simulation models a three horse race. The simulation assumes that the minor parties have been eliminated (as is the intention of these reforms)

  • Only include cases in which the Liberal Party primary vote is greater than the Labor vote and Labor votes are greater than Green votes

  • The sum of Labor and Green votes can be greater than the coalition votes.

  • All Labor votes preference Greens and vice-versa, which is not necessarily the case. As such this modelling can be taken to be the least extreme case. If Labor preferences flow to the coalition, the results will be more likely to favour the coalition.

  • 20000 simulations were carried out for each rate preferencing tested

  • Preference rates of between 5% and 100% were tested in increments of 5% to determine the flow between parties. These preference rates represents the rate at which individuals mark above the line successfully. That is, they don't vote with a single one above the line and don’t have their vote saved through the saving mechanism.

“Upset” definition:

The model inspects cases in which the Liberal Party has two senators elected before preference flows and the Greens and labour have 3 senators. It determines if the Greens and Labour were expected to get the final seat under the current preferencing system but instead the Liberal party gets that seat instead.

Results:

The model results show that an “upset” is possible at any level of incomplete above-the-line preferences. Interestingly, a failure to complete the preference rate of just 20% results in 10% of election results being “upsets”.

 

LIbsGraph.png

 

In fairness, it should be noted that in order for there to be an upset, the Liberal party vote needs to be approaching that of the minimum necessary to reach 3 senators on quota alone (42.85%).

Proportion of people preferencing

Number of sims

Number of upsets

% upsets

Mean estimate of votes (in all simulations)

Estimate of votes (in simulations with upsets)

       

Liberal

Labor

Greens

Liberal

Labor

Greens

5%

555

329

59.30%

40.30%

35.60%

24.00%

41.30%

36.50%

22.20%

10%

550

289

52.50%

40.20%

35.80%

24.00%

41.40%

36.50%

22.20%

15%

545

285

52.30%

40.30%

35.60%

24.10%

41.60%

36.30%

22.20%

20%

522

233

44.60%

40.30%

35.80%

23.90%

41.70%

36.50%

21.80%

25%

500

213

42.60%

40.20%

35.50%

24.20%

41.70%

35.90%

22.40%

30%

492

200

40.70%

40.30%

35.80%

23.90%

41.70%

36.50%

21.80%

35%

530

169

31.90%

40.10%

35.90%

24.00%

41.80%

36.10%

22.10%

40%

534

176

33.00%

40.30%

35.80%

23.90%

41.90%

36.10%

21.90%

45%

494

135

27.30%

40.20%

35.70%

24.10%

42.10%

35.70%

22.20%

50%

514

140

27.20%

40.30%

35.60%

24.20%

42.20%

35.70%

22.10%

55%

543

110

20.30%

40.20%

35.40%

24.30%

42.30%

35.60%

22.10%

60%

484

94

19.40%

40.20%

35.90%

23.90%

42.30%

36.50%

21.20%

65%

528

81

15.30%

40.20%

35.70%

24.10%

42.40%

36.40%

21.20%

70%

531

63

11.90%

40.10%

35.60%

24.30%

42.50%

35.30%

22.20%

75%

508

63

12.40%

40.40%

35.60%

24.00%

42.50%

35.40%

22.00%

80%

517

54

10.40%

40.30%

35.80%

23.90%

42.60%

36.40%

21.00%

85%

531

38

7.20%

40.20%

35.70%

24.10%

42.70%

36.20%

21.10%

90%

520

21

4.00%

40.20%

35.70%

24.10%

42.70%

36.70%

20.60%

95%

522

14

2.70%

40.20%

35.90%

24.00%

42.80%

37.10%

20.10%

100%

580

0

0.00%

40.20%

35.70%

24.00%

     


However, this model underestimates the impact of vote exhaustion as it only considers 3 players. Given many more parties, even with most votes filled out formally but still exhausting at 6, the result could be much worse.

 

Conclusion

The implementation of an optional preferential vote above-the-line system can cause upsets at elections where smaller parties have overlapping demographics and hence split the vote.

Discussion

If this system is implemented, a core way reduce the likelihood of a conservative controlled Senate is that those who have progressive tendencies should not run against the Labor Party. It is highly confusing why the Greens are choosing to go down this path.

Resources

The spreadsheet for this model (with less rows to reduce upload size) can be downloaded from here:

https://drive.google.com/file/d/0B7eQg17zrm7RNlpCYmtVUzdhaGc/view?usp=sharing

 

Created with NationBuilder