Eden-Monaro By-election – Preferences and Commentary on the Result

The final data dump of results from the Eden-Monaro by-election has been published, including the distribution of preferences and data on preference flows by party. You can find all the data at the AEC website.

In this post I want to look at final flows of preferences, were there differences caused by when people voted, how did the count unfold on election night, and was it a good result for Labor?

How the Preferences Flowed

An earlier release of the distribution of preferences showed that the final exclusion of the Shooters, Fishers and Farmers (SFF) candidate Matthew Stadtmiller pushed Labor from second place into the lead. This seemed to back claims that the SFF were responsible for Labor winning the by-election.

That Stadtmiller appeared in first place on the ballot paper played a part in him leap frogging the National and Green candidates during the distribution. On the exclusion of every candidate, even very low polling candidates, there was a drift of preferences to Statdmiller. This allowed Stadtmiller to pass the higher polling Green and National candidates.

While there is a lot of talk about the advantage of the donkey vote for a candidate at the top of the ballot paper, in a 14 candidate field, a lot of the advantage is created simply by voters numbering the balance of preferences top to bottom down the list. The top to bottom drift of preferences played a part in Labor’s victory, as I discuss below.

Complicating analysis based on the preference distribution is that Stadtmiller’s vote rose from 5,066 votes on first preferences to 9,434 on his exclusion at the final step of the count. There were 4,368 votes that drifted through Stadtmiller before making a choice between the Labor and Liberal candidates.

The AEC has now released preference flow data that gives a more nuanced picture of the preference flows than is provided by the distribution of preferences. Preference flow data tallies the final destination preference of ballot papers. In Eden-Monaro this is a tally of the final choice between Labor and Liberal for all votes with a first preference for one of the other 12 candidates.

If you take the preference flow data, which removes the intermediary steps of the preference distribution, than before the SFF preferences were distributed, Labor’s Kristy McBain already led 44,999 votes to 44,870 for Liberal Fiona Kotvojs. That 129 vote lead would have been overturned without a flow of SFF preferences, but Labor gained a net 606 SFF preferences to achieve its 735 vote winning margin. Labor still needed SFF preferences to achieve a majority of the vote, but it didn’t start from behind when preferences of SFF votes were distributed.

The preference flow data is shown in Table 1, the source first preference candidate shown in column 1, the tally of destination preferences for Labor and Liberal shown along the line. The table is in descending first preference vote order, and the destination party that received the majority of a candidate’s preferences is shown in bold.

Table 1 – Preference Flows By Party – 2020 Eden-Monaro By-election

First to Labor to Liberal
Candidate (Party) Pref % Votes Pct Votes Pct
Hicks (NAT) 6.4 1,348 22.3 4,704 77.7
Griff (GRN) 5.7 4,895 90.9 490 9.1
Stadtmiller (SFF) 5.3 2,836 56.0 2,230 44.0
Balderston (HEMP) 2.3 1,449 67.3 705 32.7
Porter (IND) 1.3 759 62.3 459 37.7
Jansson (SCP) 1.1 809 75.5 262 24.5
Angel (SAP) 1.0 596 63.1 348 36.9
McCrae (LDP) 0.7 242 37.2 409 62.8
Holgate (IND) 0.7 427 67.1 209 32.9
Storey (CDP) 0.6 119 19.4 495 80.6
Bosi (IND) 0.5 203 39.6 310 60.4
Potter (FED) 0.2 79 46.5 91 53.5
Total Preferences 25.8 13,762 56.2 10,712 43.8

Source: Calculations by Antony Green sourced from AEC website.

Table 2 summarises the preference flows into four party categories and compares the by-election flows with Eden-Monaro at last year’s Federal election.

Table 2 – Preference Flows By Party – Eden-Monaro 2019 and 2020

% of Vote % Prefs to Labor
Party 2019 2020 2019 2020
National 7.0 6.4 12.8 22.3
Greens 8.8 5.7 87.0 91.7
SFF 0.0 5.3 0.0 56.0
Others 8.1 8.4 38.9 56.0
Total 23.8 25.8 49.0 56.2

Source: Calculations by Antony Green sourced from AEC website.

The National vote fell slightly. and the flow of National preferences to Labor rose from 12.8% to 22.3%, or conversely, the flow to the Liberal Party fell from 87.2% to 77.7%. The Green vote fell, but the flow of preferences to Labor rose from 87.0% to an extremely high 91.7%. It is unusual for any party’s preference flows to pass 90%.

There was no SFF candidate in 2019, and at the by-election SFF preferences flowed to Labor at 56.0%. The vote for other candidates was about the same as last year, but where in 2019 Labor received an unhelpful flow of only 38.9% other party preferences, mainly UAP,  at the by-election Labor received 56.0%.

Labor’s first preference vote was down 3.3 percentage points at the by-election, and down 0.5 percentage points on two-party preferred. Overall, Labor’s loss of first preference vote was compensated by receiving stronger flows of preferences.

In a future post I will look at the SFF preference flows and will assess how many voters were influenced by receiving an SFF how-to-vote.

Did How and When People Voted Affect the Result?

As I tracked in a post before the election, there was an increase in both pre-poll and postal voting. It is certain that some of this increase was driven by concerns over Covid-19. In the case of postal voting, Labor’s active campaign at the by-election to encourage postal voting always made it likely that Labor would do better with postal voting than it had at last year’s Federal election.

That turned out to be the case. Table 3 breaks down the results into four categories, polling day votes, pre-poll votes within district, postal votes, and an “Other” category that includes absent, pre-poll absent, provisional and special hospital votes. This final category was significant at last year’s Federal election, but barely present at the by-election.

Table 3 – 2-Party Result by Vote Type – Eden-Monaro 2019-2020

As % of Vote Labor 2-Party %
Vote Type 2019 2020 2019 2020 Swing
Polling Day 50.6 43.6 52.8 53.0 +0.2
Pre-poll 35.5 42.9 48.9 48.8 -0.1
Postal 5.6 13.1 42.9 46.7 +3.8
Others 8.3 0.4 53.0 52.4 -0.6

Turnout in Eden-Monaro was down from 93.3% last year to 89.1% at the by-election. There was little change in informal voting, 6.8% in 2019 and 6.7% at the by-election. There were 8 candidates in 2019 and 14 at the by-election, but any increase in numbering errors was cancelled out by there being no confusion with the Senate ballot paper. When all the major parties contest a by-election, there tends to be a decline in informal voting compared to the previous election.

Polling day votes as a proportion of all votes was down 7.0 percentage points and Pre-polls up 7.4 points. The Labor two-party preferred vote was up 0.2 points on polling day and down 0.1 points with pre-poll votes. It is tempting to view the change in polling day and pre-poll votes as compensating changes.

As expected from its postal vote campaign, Labor did better with postal votes. What is hard to assess is whether this just swept up voters who would have used Absent voting. The fall in Absent voting was in total greater than the fall in turnout, so presumably many of last year’s Absent voters used other voting channels at the by-election.

Progress of the Count

The count for Eden-Monaro was always going to have a rush of early figures followed by a long drawn out wait for pre-poll centres to report. All polling day votes were reported by 9pm and represented 38% of  enrolment. Around the same number of votes were cast as pre-polls, and after an hour’s delay, pre-poll results started to arrive after 10pm. Some of the larger pre-poll centres did not report their two-party preferred counts until after midnight.

(I’ve previously written a blog post on whether the AEC should be allowed to start counting pre-poll votes before 6pm as a measure to speed up the count.)

The first graph below plots the progressive two-party preferred percentage against the percentage of the vote counted. The green line shows the Labor 2PP as a percentage of the total votes. The red line is a projected Labor 2PP% based on comparing results counting centre by counting centre.

As the graph shows, the last time Labor trailed during the count was at 1% on the projected 2PP% and at 4% on the simple 2PP%.

The first graph makes calling an election look a little easier than it really is. The prediction graph levels out at around 10%, but waiting for 10% of the vote to be counted takes time.

The graph below gives a more realistic plot of what the result looks like in real time. Getting to 10% counted took until about 7:30pm in the coverage, putting pressure on anyone trying for an early call.

The projected 2PP% line never had the Liberal Party ahead after 7pm, and the raw 2PP% never had the Liberal Party ahead after 7:10pm. There were some odd calls late in the evening suggesting the Liberal Party were winning, but these cannot have been based on the count tallies released by the Australian Electoral Commission.

The swing against Labor was larger on first preferences than it was on two-party preferred. It took some time to get two-party preferred counts from the large pre-poll counting centres, the final 2PPs not coming through until after midnight.

Given the above tracks of two-party preferred results through the night, I suspect some of the calls of a Liberal victory were based on first preference swings in pre-poll centres. I made my final call of a Labor victory on election night after waiting for the final 2PP counts to be reported.

A Good or Bad Result for Labor?

The general consensus has been that the Eden-Monaro result was politically neutral. Eden-Monaro was a Labor-held seat, and Labor retained it. In the binary world of political perceptions, a win is a win.

Yet if just a few hundred votes had fallen differently, Labor would have lost and there would have been a completely different picture painted. The “Morrison Miracle” of victory at last year’s Federal election would have been followed by the once-in-a-century event of a government taking a seat from the opposition at a by-election.

The fine line between attracting “nothing happened” reviews from a narrow win, and bad headlines from a narrow defeat, may have come down to something as simple as the random draw for ballot position in Eden-Monaro.

Labor candidate Kristy McBain drew position 8 while Liberal candidate Fiona Kotvojs drew the last spot in a field of 14. The National candidate drew above Labor on the ballot paper, always likely to produce a slight leak of National preferences to Labor. The Shooters Fishers and Farmers (SFF) candidate drew top spot on the ballot paper, which probably inflated the flow of preferences to Labor on top of what flowed from the SFF how-to-vote preference recommendation for Labor.

Labor won by 735 votes out of  out of 94,935 formal votes. Anyone who has ever examined ballot papers will know there is always a preference drift down the list of candidates. After voting for the candidates they know, a small percentage of voters resort to numbering everyone else top to bottom. If the Labor and Liberal candidate’s ballot positions had been reversed, 368 voters using top to bottom preferencing would have been enough to swap the result and deliver victory to the Liberal candidate. 368 voters is less that 0.4% of the vote, within most measures of the so-called donkey vote.

As explained earlier, the constant drift of preferences to the SFF candidate in top spot on the ballot paper backs this preference drift view of how Labor won.

Beyond the “win’s a win” analysis, the overall result for Labor wasn’t great measured against history. But we don’t have a lot of examples of by-elections held in a pandemic at a time when there is general goodwill towards a government trying co-operate with the states in the national interest. It is a political climate that makes it hard for an opposition to be heard.

I’ve always argued against using average by-elections swings as a reliable performance measure. In the sample space of by-election swings, the average is only one measure of a very varied set of data. The size of a by-election swing is caused by the circumstances of the by-election, so merging all these into an average for use in measuring the Eden-Monaro by-election fought under different circumstances may be meaningless.

For example, the average swing against government at by-elections since 1949 has been 3.8%. In that time, the average swing in seats held by government has been 5.0%, but the average in the smaller number of seats held by oppositions has been a much lower 1.6%.

In Eden-Monaro, the loss of the sitting member vote, and anger at a resignation causing the by-election, were anti-Labor rather than anti-government factors. With a small 0.4% swing towards the government, Labor did worse than the average by-election swing in opposition held seats.

Labor might have had the advantage of the government’s maladroit handling of last summer’s bushfire crisis, but the government had the advantage of the public seeing its handling of the pandemic in much better light.

In my view, the ballot draw was a little bit of luck that helped Labor over the line. Sometimes luck plays a role in determining political success or failure. In Eden-Monaro, a little bit of luck at a bad moment in the political cycle allowed Labor, and its Leader Anthony Albanese,  to avoid coming under the significant political pressure that would have flowed from defeat.

I’m not saying it was only luck that allowed Labor to win. Clearly the choice of the right candidate was important. Clearly the messy process to choose Liberal and National candidates was important. The SFF how-to-vote was important. The weaker flow of National preferences was important. But after all those factors, a little bit of luck mattered as well.


1 thought on “Eden-Monaro By-election – Preferences and Commentary on the Result”

  1. Antony, what’s your opinion on rotating ballot order on preferential ballots, like the ACT & Tasmania do with Hare-Clark? It seems like it would dilute the effect of top to bottom preferencing, which might have an effect in close elections.

    COMMENT: Ballot paper rotation under Hare-Clark is partly about randomising the donkey vote, but its introduction has mainly been about breaking the power of a party to control who gets elected from the party’s ticket. The rotation also does not affect the order parties appear in columns, merely the order candidate’s appear in each party’s column.

Leave a Reply