Analysing preference flows under a Hare-Clark system like that used in Tasmania is a lot more complex than a single-member electorate system. It isn’t simply a matter of a series of candidates being knocked out until two remain – different parties benefit from preferences at points in time before later giving away preferences of their own, and parties have multiple candidates with preferences that don’t always stay within the party ticket.
Inspired by a commenter, Cameron Scott, who did a little bit of this sort of analysis in the comments, for this post I’m looking at how much of the vote it generally takes to win a certain number of seats, how this number isn’t the same for every party, and the rates at which preferences leave each party group and how many preferences a party group picks up through the distribution of preferences.
For this post I’m analysing as far back as 2006. This is partly because my data repository has recently been expanded to include the results of the 2006 election (although the actual distribution of preferences is not included). The 2021 results are free to everyone, but the 2006-2018 results are available for anyone who pays $5 or more per month via Patreon. These datasets include primary vote results at the electorate and booth level, candidate lists (including who was elected), lists of polling places with latitude and longitude, and data on enrolment and quotas.
To start simple, this first chart shows the relationship between the total number of quotas polled by each of the three main parties at the last five elections, and how that relates to the number of seats won.
There’s a reasonably tight correlation, with the Greens understandably occupying the bottom left of the chart.
You can see a bit of evidence of something that will become clearer later on. You can see that Labor maanged to win seven seats and ten seats with smaller votes than the vote needed for the Liberal Party to win the same number of seats.
This next chart shows the same data another way. It shows how many quotas each party polled in excess of their seat victories. Of course there are more quotas than seats – there would be almost 30 quotas across the five electorates, but just 25 seats.
Labor regularly had the smallest gap, generally winning less votes per seat. The only exception is in 2014, when the Liberal Party had a very efficient distribution of votes per seat.
Indeed in 2018 and 2021 the ALP managed to win more seats than its quota share.
So now I’m going to dive down to the level of the electorate. This next chart shows all 75 outcomes for the three parties at the five elections since 2006 – how many primary vote quotas did each party poll, and how many seats did they win?
Some things become clear when looking at this chart – parties polling under 0.7 quotas usually don’t win, and those polling over that do win. There isn’t much overlap.
But there is much more overlap as to whether parties win one or two seats when they poll between 1.5 and 2 quotas.
The Liberal Party polled 1.88 quotas in Franklin in 2006, but only won one seat. Meanwhile the ALP polled 1.56 quotas in Bass in 2021 but won two.
I won’t try and dive into the full story about why these divergences occur here. Some of the story would be about the relative strength of other parties, and the Ginninderra effect of how an even distribution of votes within a party can make one more seat competitive.
But I suspect part of the story is about how preferences flow. If you look at all three of these charts, the ALP generally seems to perform better than either the Greens or the Liberal Party (although it is harder to compare the Greens to the others since their total vote level is usually lower).
My hypothesis is that the ALP tends to do better because they are more likely to get preferences from the left than the Liberal Party, and more likely to get preferences from the right than the Greens. A position in the centre of the spectrum will help you, on average.
To dive a bit deeper into this, I looked at the distribution of preferences for every contest since 2006. I was specifically interested in when a candidate is excluded and their preference flows to a party other than their own party. So when a Liberal candidate is excluded, how many votes are lost from the Liberal group (leakage) and how many votes are gained by each of the other groups (gains).
Labor has generally done better, particularly since 2014. The fact that they’ve done better since then may suggest that their performance has improved due to their relatively weaker primary vote position, rather than something intrinsic to the ALP, since the figures in 2006 and 2010 were much closer.
Finally, this chart compares the number of votes gained to the number of votes lost (otherwise known as leakage). Understandably, if a party has candidates who stay in the count longer they will have more potential to gain votes and less potential to leak votes. I have also not counted the final exclusion of a party’s last candidate as “leakage”.
There is an inverse relationship between the two metrics, but the party balance is again obvious. The vast majority of cases where a party gained over 0.2 quotas have been for the ALP.
I don’t have a final conclusion here. There is a lot more to do in terms of looking at potential factors, and perhaps there are better metrics that remove some potential messiness. I am not convinced that the ALP’s better performance is a permanent feature.
These issues may also change in 2024, both because of the larger candidate fields and the lower quota.
But this is useful when we think about how those primary votes that will be reported on election night might translate into a final result.