Do How-to-votes Matter? Evidence from the 2022 SA Election

Too often I hear criticism that preferential voting is corrupted by the majority of voters simply copying how-to-vote (HTV) preference recommendations. The same critics tend to decry that preference “deals” determine elections.

Talk of preference “deals” is a hangover from the now abolished Senate group voting ticket (GVT) system. In the GVT era, 90-95% of ballot papers in the Senate were single ‘1’ above the line votes with preferences determined by the party’s registered preference ticket. Preferences became a tradeable commodity with parties delivering 90-95% of preferences by arranged deals.

Those arranged “deals” never applied to lower house elections and have not applied for Senate elections since the abolition of GVTs in 2016. So-called “deals” on lower house preferences are about what parties and candidates distribute as how-to-vote recommendations. Most so-called “deals” are more expressions of shared or common interest between parties.

HTVs can have an impact, but much less than opponents of preferential voting suggest. All available research on ballot paper completion shows that the majority of voters do not copy HTV preference recommendations.

The Victorian and South Australian Electoral Commissions regularly analyse ballot papers and have determined that around 40% of major party voters complete their ballot paper with their preferred party’s HTV preference sequence. The proportion following HTVs is even lower for minor parties and independents. The difference in follow rate is largely related to the proportion of voters who receive the HTV for a particular party or candidate. Obviously the probability of a voter copying a HTV recommendation is higher if they receive a HTV than if they don’t.

Most countries ban campaigning at polling places, but Australia permits campaigning and the distribution of how-to-vote material, if done politely.

Several years ago I wrote a long post on why candidates and parties distribute how-to-votes. Overwhelmingly the point of HTVs is to influence first preferences and to encourage the casting of formal votes. For minor parties being able to influence preferences may also be important, but the follow rate for minor party HTVs is low if they do not actively campaign.

The weakness in trying to assess the impact of HTVs is that not every voter receives their preferred party or candidate’s HTV. If we knew who received a how-to-vote, and how they completed their ballot paper, we would have a better measure of HTV influence. But the secrecy of the ballot makes it impossible to link a ballot paper to a voter.

Instead we are stuck with follow rates based on all ballot papers. But do we get a better measure of HTV influence if we are confident that most voters saw all candidate’s HTVs.

There is one Australian jurisdiction where we know that many more voters see HTV recommendations.

That is South Australia where there is a unique provision that allows HTVs for all candidates in a district to be displayed in front of a voter in a voting booth.

Let me explain the South Australian system and then go on to compare preference flow data from the 2022 South Australian election with the HTVs displayed in voting booths.

For those who want the data behind this post, you can find the full data by district, and summary tables by party, at the following link.

The preferences document is part of a longer publication I’ve prepared on the results of the 2022 South Australian election. You can find it here or read about it here in a post about the publication yesterday.

You can find my 2022 South Australian Election results publication at this link.

I also published a similar analysis of 2018 South Australian preference flows and HTVs at this link.

What is Displayed in South Australian Voting Booths

Four candidates nominated for the seat of Bragg at the 2022 South Australian election. Below is the sample HTVs displayed in all voting booths in all polling places in Bragg.

Source: Election Commission of South Australia.

Note that each party adopts a particular style for their displayed how-to-votes. How-to-votes distributed outside polling places will usually be the A4 size we are all familiar with including candidate photos, larger party names and more colour. The HTVs in voting booths are for information, those distributed outside for campaigning.

As the above examples show, candidates and parties can choose whether to include all party affiliations, as the Greens have done, or include only their own candidate’s party affiliation. As party names are printed on ballot papers it makes hiding them on HTVs seem a bit pointless. But the view of some parties is that not showing all affiliations encourages voters to follow the how-to-vote.

Candidates can lodge “just vote 1” options that indicate ‘1’ for them then a message to number all the other squares. A party can issue a split ticket where two alternative preferences recommendations are shown. And a candidate who does not lodge a HTV has an outlined box which includes the text “NO HOW TO VOTE CARD LODGED BY” then the name and affiliation of the candidate.

The picture below shows how the HTV display appears in voting booths.

Source: Election Commission of South Australia.

A final note on another unique SA feature, voting tickets. Candidates can lodge a voting ticket, a registered set of preferences, which can be used as a savings provision for ballot papers with incomplete numbering. If a ballot paper has a valid first preference, or in some cases several valid preferences, but does not have a complete set of numbers, the ballot paper can be ‘saved’ from informality by the ticket vote being used to complete the ballot paper’s preferences.

I do not have the figures for 2022, but in 2018 3.8% of the total votes were ballot papers saved by this provision. It is why informal voting is always lower at state elections in South Australia compared to Federal elections in the state.

In this post I have not taken account of the small numbers of ballot papers saved by voting tickets as there number is not large enough to change the overall analysis.

What Data on Preferences is Available

Like the Australian Electoral Commission, the Electoral Commission of SA (ECSA) conducts two types of preference counts. The first is the required formal distribution of preferences done to determine the winning candidate. The successive exclusions of lowest polling candidates sees ballot papers examined and transferred to continuing candidates based on the numbers written by voters on their ballot papers.

The formal distribution for Bragg is shown below.

Source: Election Commission of South Australia.

ECSA also conducts a second type of preference count for information purposes. This count accumulates preference flow data for all excluded candidates and tallies how many of a candidate’s preferences reached each of the two final candidates in the contest. Below is the preference flow data for Bragg which removes the intermediary steps in the count. Note that in these tables I have shown the party direction of HTVs and indicated the impact of a donkey vote where appropriate.

Source: Calculations and table by Antony Green based on data released by the Election Commission of South Australia.

This data is different from the formal distribution of preferences. For instance, the Family First candidate was the first excluded in Bragg, and 290 of his 1175 votes had second preferences for the Greens. The above table for Bragg ignores these intermediary flows and simply looks at the final destination of a ballot paper’s preference, ignoring the route by which ballot papers reached one of the final two candidates. (A note to make on voters following HTVs is that the Family First HTV in Bragg had the Greens’ candidate last but 25% of voters gave the Green second preference.)

If you combine the data in the two tables, you can work out that of the 290 Family First ballot papers with second preference for the Greens, 88 had a third preferences for Liberal and 202 third preference for Labor. Unpicking the data like this is not possible for contests with more than four candidates.

For those surprised by Family First’s preference recommendation in Bragg, I’ll discuss those and the impact of the party varying its preference recommendations in the section on Family First further down the post.

Limitations of the Data

The preference flow does not tell us whether ballot papers copied a HTV preference. Flow data tells us the percentage of a candidate’s ballot papers that ended up with the same final choice candidate as a HTV, not whether the voters used the same sequence of preferences.

For instance, the data on Greens ballot papers below shows that 81.2% Green votes reached Labor candidates after preferences ahead of Liberal candidates. But the flow data does not tell us the intermediary numbers completed, telling us the destination not the route.

In that sense the preference flow data doesn’t tell us whether voters followed a how-to-vote, just whether ballot papers finished with the recommended candidate in the final pairing after preferences.

But if, as some people who don’t look at the available data argue, 90% of people just follow the HTV, then you would expect to see 90% rates of preference flow. The Greens get close to that due to most Greens voters feeling some ideological affinity with Labor. But the rates are nothing like that for other parties.

And as the example above for Bragg shows, the flow of Green preferences to Labor was 82.0%, Family First 56.9%. Two HTVs in the same voting booth but two very different rates of preference flow. Voters don’t just follow a HTV, a lot of them clearly make up their own minds.

State-wide Preference Flows

Minor Party Preference Flows – 2022 South Australian Election

	Candi-	Pct First	% Preferencess	% Preferences
Party	dates	Preferences	to Labor	to Liberal
The Greens	43	9.1	81.2	18.8
Independents	20	7.3	47.1	52.9
Family First	34	3.7	47.4	52.6
One Nation	19	2.6	46.8	53.2
Animal Justice	10	0.6	69.8	30.2
The Nationals	8	0.5	33.9	66.1
Aust Family Party	6	0.3	45.9	54.1
SA Best	1	0.2	48.3	51.7
Real Change SA	4	0.1	52.1	47.9
Liberal Democrats	1	0.1	35.5	64.5
Total	146	24.4	60.1	39.9

Source: Calculations by Antony Green based on raw data released by the Electoral Commission of South Australia.

At the state-wide level, this data shows clearly that a lot of voters make up their own minds on preferences. Only the Greens and Animal Justice produce flows above 65%. But as I discuss below, Family First’s variation in preference recommendation between seats varies the flow from seat to seat.

This is an even greater factor for Independents with relatively strong flows in different seats that cancel each other out in state-wide figures.

There were six seats that did not finish as two-party preferred contests – Finniss, Flinders, Kavel, Mount Gambier, Narungga and Stuart. All finished as Liberal-Independent contests. I’ll discuss these contests separately at the end of the post.

Green Preferences

The chart below plotting district level Green preference flows to Labor against % Greens first preference vote looks as expected. The trend line shows a default 75% flow to Labor with the flow rate rising as the Green primary vote increases. Essentially every candidate in a contest attracts a small number of “random” first preference votes, and these decline as a proportion of overall vote for a party as its vote rises.

As an example, the three lowest preference flows to Labor were Flinders (10.2% vote, 69.8% flow), King (5.6%, 67.9%) and Playford (9.5%, 63.7%). In all three seats the Greens candidate drew top spot on the ballot paper, and a ‘donkey’ vote numbering straight down the ballot paper favoured the Liberal Party. That seems to explain why the flow of preferences to Labor was lower in these three seats.

Family First Preferences

Family First contested 34 seats. In 14 the HTV displayed in the polling place recommended preferences to Labor and the flows were Labor 53.6% and Liberal 46.4%. In the 20 seats recommending preferences to the Liberal Party, flows were only 43.6% to Labor 56.4% to Liberal. If voters just copied HTV recommendations, you would see much stronger preference flows than these.

The difference in flows shows up clearly in the chart below plotting % vote and % flows to Labor. The points are colour coded by HTV recommendation and the seats where preferences were directed to Labor have minimal overlap with the seats where Liberal preferences were recommended. The differences caused by the recommendation are about 10%, which suggests HTV recommendations impact flows at the margin rather than wildly shift preference flows.

Using an average across seats can remove weighting caused by seats with a higher first preference vote. This created little difference to the analysis with the average flows to Labor being 53.0% where directed to Labor opposed to 43.2% where directed to Liberal. The difference in flows created by changing the HTV recommendation remains at around 10%. At the 2025 Federal election in South Australia, Family First preferences flowed 64.8% to Liberal .

For those interstate who are bemused by Family First recommending preferences to Labor in some seats, it is an issue to do with the right faction in South Australia, associated with Senator Don Farrell, having a significant presence of moral conservative MPS. A number of the seats where Family First recommended preferences to Labor are held by right faction moral conservatives. Several other were Liberals seats with particularly strong small-l Liberal MPs, such as former Attorney-General Vickie Chapman in Bragg.

One Nation and Others

Overall One Nation preferences flowed 46.1% to Labor and 53.9% to Liberal. Preferences were directed to Labor in four seats and flowed 50.0% to Labor as against 46.8% in seats with a Liberal preference recommendation. Of the nine seats where preferences to Labor were above 50%, three were in seats with Labor recommendation, and three in seats with a Liberal recommendation but the donkey vote favouring Labor. Clearly voters did not pay much attention to HTV recommendations. Nor did there appear to be strong anti-Labor feeling amongst One Nation voters. This compares to the 2025 Federal election when One Nation preferences were 71% to Liberal in South Australia.

The flows for other parties are shown in the state-wide preferences table earlier in the post.

Independents

The table below shows the preference flows for the 11 Independents who polled a significant vote at the election. The table shows district, candidate name, HTV recommendation, % vote and two-party preference flows.

Selected Independent Preference Flows – SA 2022 Election

District	Candidate	HTV	% Vote	% to Labor	% to Liberal
Finniss	Lou Nicholson	1-only	19.6	57.3	42.7
Flinders	Liz Habermann	Liberal	27.2	33.8	66.2
Frome	Kate Hunter	Split	16.6	60.3	39.7
Hammond	Arlie Keen	Labor	15.7	69.1	30.9
Kavel	Dan Cregan	Liberal	50.5	42.2	57.8
Mt Gambier	Troy Bell	Liberal	45.7	29.7	70.3
Narungga	Fraser Ellis	Liberal	32.5	22.8	77.2
Newland	Frances Bedford	1-only	12.3	63.3	36.7
Stuart	Geoff Brock	1-only	48.5	64.2	35.8
Waite	Sam Duluk	Liberal	19.7	23.8	76.2
Waite	Heather Holmes-Ross	1-only	14.6	75.9	24.1

The table shows clearly that preferences favoured the Liberal Party in the five seats where the Independent recommended preferences to the Liberal Party. The one Independent recommendation for Labor also produced a strong flow to Labor in Hammond.

But the five seats where Independents had 1-only or split preference recommendations, all produced strong preference flows to Labor. This repeats a phenomena we have seen at all recent Federal elections. Successful as well as strong polling defeated Independents all produce strong flows of preferences to Labor, even though the candidates do not make preference recommendations. Some of this can be explained by Labor and Green supporters voting strategically for the Independent and then preferencing back to their real preferred party.

That still doesn’t explain all the preference flows. It seems that there is a slice of former Liberal voters who, having made the decision not to vote for their former party, also choose not to give them further preferences. It is a problem to which the Liberal Party is struggling to find a solution.

Liberal-Independent Contests

There were six districts that finished as Liberal versus Independent contests. Labor contested all six seats with the Greens, One Nation and Family First contesting four.

All six demonstrated the problem the Liberal Party is dogged by across the country. When challenged, the Liberal Party always does badly on preferences from all other parties in races against Independents.

Overall preferences flows to the Independent in each contest were 79.7% to Lou Nicholson in Finniss, 73.9% to Liz Habermann in Flinders, 87.1% to Dan Cregan in Kavel, 70.9% to Troy Bell in Mount Gambier, 66.8% to Fraser Ellis in Narungga and 80.5% in Geoff Brock in Stuart.

Labor recommended preferences to the Independent in all six contests, producing total preference flows of 78.5% to Independent ranging from 70.5% in Narungga to 87.0% in Kavel. In Kavel it seemed all Dan Cregan’s opponents much preferred him to Liberal candidate Rowan Mumford

The Greens recommended preferences to the Independents in the four seats it contested producing flows of 84.8% to the Independents, ranging from 74.1% in Stuart (only 3.1% of the primary vote) to 90.5% in Kavel.

Family first also recommended preferences to Independents in four seats producing flows of 63.5%. One Nation produced flows of 76.6% to Independents across four seats, including 74.2% in Stuart where the party’s HTV favoured the Liberal candidate but Independent Geoff Brock benefited from the donkey vote.