Lower exhaustion rates amid a more concentrated upper house vote

One of the things that stood out in the NSW Legislative Council results was that Animal Justice increased their vote from the level that they polled in 2015 and 2019, but they fell short of winning. I wanted to try and understand what happened. What changed this year that raised that bar?

In this post I’m exploring some different data points about how the smaller field of serious contenders for the NSW Legislative Council has played into a higher proportion of votes contributing to a successful candidate, and a higher bar for winning a seat.

First up, this shows the lowest vote (both primary vote and vote at the end of the count). Where the candidate was from a party that won more than one seat, I’ve subtracted the whole quotas to make a fair comparison.

The level of support required to win a seat is the highest it has been since the current system started in 2003.

In 2007, the Nationals' Trevor Khan won the final seat with 2.4% of the primary vote, but this time the seventh Liberal Rachel Merton required 2.7%.

The two AJP wins in 2015 and 2019 were the wins with the lowest primary vote on record, and the AJP's first win in 2015 was from a remarkably low vote even at the end of the count - just 2.1%.

This time around, the AJP managed 2.3% of the primary vote and 2.9% at the end of the count. Such a vote would have won a seat at any of the last five elections, but not this year.

I've previously noted that the pool of preferences available for the final rounds of the count is less than it was in 2019. One possible explanation is the reduction in the number of groups with an above-the-line box. Thanks to six groups failing to register the required 15 candidates, the number of above-the-line boxes reached the equal lowest number under the post-GVT system.

It's not just the vote for the final candidates that has increased - across the board, the last-elected candidates polled better than at previous elections.

Every election under the current system has followed the same pattern - 17 candidates elected with a full quota (or close to it in case of the Greens' Amanda Cohn, who reached a quota before the final rounds of the count after polling close to a quota), and the final four elected without a quota after the remaining candidates are excluded (although Fred Nile did reach a quota at the very end of the 2007 count). This leaves one more candidate as the last candidate excluded whose votes are not passed on. Thus every vote either ends up with one of the 21 elected candidates or the last defeated candidate, or is exhausted.

The total share of the vote that was cast for those last 4 candidates, and the share that ended up with them by the end of the count, was notably higher this year than at the last five elections, and particularly higher than in 2019.

If the figure reached 18%, that would suggest that all four ended up polling a full quota, and no votes exhausted. It's not quite there, and I don't think it would ever reach that level, but this year that figure reached 15%.

This next chart is the one that started me on this topic. I calculated the total number of exhausted votes as a share of the formal vote. This figure bounced around between 7% and 8% at the last five elections, but has now dropped to just 5.6%.

Now this could mean a number of things - it could suggest that voters are filling out more preferences. But it may also mean that less votes were available to be exhausted, and that is definitely part of the story. We know there were less votes that were available as preferences in this count, and we know that more of the vote was already sitting with the parties that stayed in the count to the end.

We will need to wait for the full preference dataset to be able to say definitively if preference rates went up - I've noticed some evidence that this has happened in the lower house, so I wouldn't be surprised.

In order to isolate the effects of a greater concentration of the vote on exhaustion rates, for this next chart I have simply looked at votes that passed between groups. So for the big parties, votes that stayed in the group and were not passed on as a surplus are not included. Nor are the primary votes for the five candidates left standing at the end of the count. For the sake of simplicity I have assumed all primary votes for a group ended up staying with the last standing candidate in that group.

That remaining pool of votes is then effectively the 'preference pool'. It was between 8.4% and 9.3% of the total formal vote from 2003 until 2015, and then climbed to 10.7% in 2019 before dropping to just 7.7% in 2023.

Of these votes, the share that ended up exhausting hovered in the low 80s from 2003 to 2015, but dropped to the low 70s in 2019, when there was an uptick in preferencing behaviour.

The analysis for 2023 suggests the proportion ending up flowing to one of the last five candidates increased slightly from the previous high, with 27.7% of votes in the preference pool not exhausting.

Overall there is evidence here for two main trends:

• Preferencing looks set to roughly maintain the improved rate we saw in 2019.
• A smaller field of parties has seen a concentration in votes and thus requires a higher minor party vote to win one of the final seats.

Liked it? Take a second to support the Tally Room on Patreon!