Vote by Type at the 1999 Republic Referendum PLUS how Referendum Night will Unfold.

Several times in the last fortnight I’ve been asked how I think Postal and Pre-poll votes will split at the referendum.

As a general rule at Australian elections, postal votes significantly favour the Coalition compared to polling day votes. Pre-poll votes slightly favour the Coalition though by how much varies from election to election. The smaller categories of Absent and Provisional votes tend to favour Labor.

At referendums, it is fair to say a No vote is for the status quo and a Yes vote for change. On that basis you would expect postal voting to display the same pattern as at a general election, favouring the conservative position. Pre-poll voting could also have a small lean to the status quo.

This observation on postal voting is backed by the chart below that shows the Yes/No percentages by vote type for the 1999 Republic referendum. (I’ve co-opted the common colours being used by Yes and No in 2023.)

As the column showing percentage votes in each category makes clear, the past two decades have seen a massive shift away from voting on election day. So will the same trends be evident in 2023? Here are a few important points.

• In 1999 83.8% of all votes were cast as ordinary votes, the overwhelming majority on polling day. For the 2022 Federal election the corresponding figure for ordinary votes was only 48.7%.
• Postal votes made up only 3.2% of all votes in 1999 compared to 14.3% in 2022. The figure rose from 8.3% in 2019 to 14.3% in 2022 in the aftermath of Covid. In 2023 the rate of postal voting may decline from its 2022 level, but will still be three to four times the rate seen in 1999.
• In 1999 the postal vote was 11% more conservative than ordinary votes. The gap between postal votes and ordinary votes has declined in recent years as the number of postal votes has grown. However, a narrower gap on a larger number of votes can shift the final result away from the election night final figures. The inclusion of postal votes will lower the Yes % reported on election night.
• Pre-poll votes have increased massively from 5.4% in 1999 to 32.8% in 2022. In 1999 the pre-poll vote for Yes was 1.2% higher than for ordinary votes, but you can't draw conclusions from a category that has grown six-fold since 1999. What can be said is that pre-polls are very important given they now make up one-third of the total vote.
• In 1999 Pre-Poll votes were declaration votes and not counted on election night. Since 2010 all within district pre-polls are cast as ordinary votes and are counted on election night.
• In 1999 Absent votes made up 5.4% of all votes, but only 3.2% at the 2022 election. Absent votes are votes cast by a voter on election day in a polling place outside their home division. The number of Absent votes has declined with the greater availability of pre-poll voting, along with the increased use of joint polling places.
• In 1999 only polling day ordinary votes were counted on election night. All other votes were declaration votes and counted in the days after polling.
• In 2023 all polling day ordinary votes will be counted on election night, along with the vast majority of pre-poll votes. On election night this may cause the growth of votes counted to slow after about two hours of counting in each state. It is increasingly a feature of Australian election coverages that election night counting of polling places is completed more quickly than in the past, but there is then a delay until the giant pre-poll centres complete counting and report.

Will we know the result on referendum night?

The Australian Electoral Commissioner, Tom Rogers, has stated that he won't be able to declare any state until postal votes have been counted. The return date for postal votes is two weeks after polling day, and this has been mis-reported as being we won't know the result for two weeks.

Unless the result is very close, we will know the result on election night. And if the polling is correct, we may know the result very quickly.

The Australian Electoral Commission (AEC) cannot declare the result until after the close off date for return of postal votes, which is the Friday two weeks after polling day. In law the AEC has to count all the votes before returning the writ and declaring whether the referendum has passed or failed.

But that doesn't mean we won't know the result on the night. If Yes or No has a significant lead on election night when four-fifths of votes are counted, it is highly unlikely that the outstanding one-fifth of the vote can overturn an election night lead. Except in the case of a very narrow Yes lead.

Commissioner Rogers' statement about a two week delay with the postal votes has set off the conspiracy theories, reviving all the stories of the 2020 US election being 'stolen'.

Fevered imaginations have "them" stealing the election for "Yes" with a postal vote stack. This is garbage for two reasons. One, as mentioned above, the Postal Vote will favour the No side and will not help deliver a Yes victory. Second, the only postal votes that can be included in the count are those for voters who registered to be sent a postal vote before the election. You can't invent postal votes for people who didn't apply, and why would someone who wants to vote Yes not vote and instead wait to use a postal vote once the election night result was known. It is a theory that makes no sense.

As usual, the election night count is only preliminary. Scrutineers for Yes and No are permitted to watch the count in each polling place. All ballot papers are transferred to the office of the Returning Officer after polling day, and all results are confirmed by a check count in the days after the election. The check count is conducted by different counting staff to election night, and again can be observed by scrutineers representing Yes and No.

All declaration votes, mainly Postal and Absent votes, will have their envelopes processed in the days after the election. If the envelopes pass the check of details, which scrutineers for Yes and No are permitted to observe, the ballot papers are extracted, counted once, check counted, and added to the result totals.

Referendum counts are much simpler than general election counts. There is only one voting square on the ballot paper and there is no need to check for a valid sequence of preferences.

How will the Media Report Results on Election Night?

How results are reported on referendum night is the same as at a general election. Counts are conducted BY MANUAL COUNTING, and once complete, reported by polling place and by pre-poll centre and entered into the AEC's results computer. On its website, the AEC will report results nationally, by state and territory, and by electoral division. Results down to polling place will be available from the Sunday after polling day.

All results are transmitted to the media by a results feed. The feed has some technical difference by reporting Yes/No results rather than candidate results. Media and other interested parties can access the data which includes results down to polling place level. But unlike at general elections, the AEC feed does not include any historic data.

At general elections, how quickly polling place results are reported depends on -

1. How many votes are taken. Staff at a polling place with fewer ballot papers can count and report results quicker than a polling place taking more votes.
2. How many candidates there are. The more candidates on the ballot paper, the slower the sort of ballot papers before counting starts.
3. The number of votes cast for candidates other than the top two. The more candidates there are, the more votes for other candidates, the longer it takes to conduct an indicative preference count.

At a referendum, candidate numbers and preferences are irrelevant. Every polling place and pre-poll centre in every electorate across the country has essentially the same ballot paper - a two-horse race between candidates called Yes and No.

Even more than at general elections, the first results will come from smaller rural polling places, with results from larger towns and the cities reporting later. The even larger pre-poll centres will report later still.

At Australian elections, there is a significant positive correlation between polling place size and the percentage of Labor two-party preferred vote. Small rural booths generally have a lower Labor vote, so the early cumulative count has a distinct anti-Labor lean. In technical terms, early counts exhibit statistical 'bias', the cumulative percentages are a biased estimator of the final result.

As the count progresses, the bias in the early figures is cancelled out by the arrival of larger polling places from urban areas with more mixed results.

It is a feature of Australian elections that the Labor two-party preferred vote starts low and rises. In recent decades this has not been evident in television election coverages where swing or change by polling place is used to correct for early statistical bias. Australian election coverages tend to display projected results rather than raw numbers, giving a clearer picture of the result than can be gleaned from the raw numbers.

This is in contrast to United States elections where raw numbers and percentages are broadcast. Viewers will be familiar with analysts pointing at numbers and maps to explain why a number will rise or fall. Most American networks use county based methods of comparing results using swing, but it is less reliable under voluntary voting. Above all, US networks broadcast raw numbers and percentages, not the projected percentages used in Australian election coverages.

There may be a feel of US election coverages on referendum night. The numbers being broadcast will be raw numbers and percentages.

Given Labor has few seats in rural areas, and given the National Party and many Liberals have advocated for a No vote, it is certain that the election night Yes % will follow a similar trend to the raw Labor % at general elections. It is unlikely that the cumulative Yes % will be the same as a cumulative Labor 2-party %, but initially at least it should follow a similar trajectory in starting low and rising.

On election night 14 October, the Yes percentage in every state will start low and rise. The question is, at what percentage counted does the curve of cumulative Yes % flatten? When the curve flattens, you can predict the result in each state.

But based on analysis of the 1999 Republic referendum, there is a very high likelihood that results in metropolitan seats will not conform with traditional Labor-Coalition voting patterns. In country seats at the 1999 Republic referendum, there was some correlation between the Yes % and the Labor 2-Party % at the 1998 Federal election. But that correlation evaporated in the cities.

At the Republic referendum, many affluent Liberal metropolitan seats voted Yes at rates well above the level of Labor vote at the previous year's Federal election. And many metropolitan Labor seats voted No, their Yes % well below the Labor two-party preferred vote the previous year.

Which means calling the result will have to wait for some results to arrive from each state's capital.

For the Yes case to win a state, the Yes % will need to approximate the Labor two-party preferred in urban Labor seats at the 2022 Federal election.

At the Republic referendum, some Labor seats recorded Yes percentages 20-30 percentage points below the Labor vote the previous year. A few Liberal and 'teal' Independent seats voting Yes in 2023 will struggle to overcome a possible No vote from the more numerous metropolitan Labor seats.