What social media ads and polling tells us about the Voice to Parliament referendum campaign strategies so far

If you live in one of Australia’s smallest states, you might soon be inundated with ads about the Voice to Parliament referendum.

Campaign strategies are starting to take shape, and on one side at least, the path to victory appears to sidestep Australia’s biggest cities.

Facebook ad data shows that the Fair Australia and Yes23 pages are the biggest referendum-related spenders on the social network so far.

Both camps spent about $110,000 on Facebook ads between mid-March and mid-June, but they’re choosing to spend that money in very different ways.

While the Yes campaign is advertising roughly evenly across the states, the No campaign is clearly targeting its money more strategically.

In Queensland, the Fair Australia page is spending more than double the Yes23 page, while in Victoria, the reverse is true.

Why would the No campaign pay less attention to the two most populous states?

While in an election it’s the marginal seat voters that get the most attention, in a referendum we’re better off talking about the marginal states.

For the referendum to pass, a majority of voters in four states — the ACT and Northern Territory don’t count for this condition — must write “yes” on their ballot paper.

If three states vote “no”, it’s all over, even if most Australians voted “yes”.

Indeed, if Victoria and New South Wales vote in favour of the Voice, that alone could be enough to see the nation record a majority, even with three or four states against it.

Warren Mundine, a leading advocate for the No campaign, confirmed to the ABC’s RN Breakfast program in May that was precisely his side’s strategy.

“For us, it’s sensible, because they have to have the double whammy of a majority plus a majority of states,” he said.

“All we’ve got to do is pick up three states and that’s the end of the referendum.

“We will be targeting four states mainly … we’re feeling very confident at this stage, but there’s a hell of a long way to go yet.”

The four states being targeted

The four states to focus on are clear when you look at the limited public opinion polling that’s been published.

We don’t tend to report on individual polls at ABC News, but when taken all together, they can give us useful context on public opinion.

Here are three of the larger ones, with big enough samples to produce state-by-state estimates of support.

The biggest poll published so far was conducted in March by YouGov for the Yes campaign, with a sample size of more than 15,000 people.

Don’t focus too much on the numbers, what’s informing the campaign strategy here is the order of the states.

In all three of these polls, Victoria and New South Wales were in the top half of the states.

Queensland is consistently the least likely state to back the Voice, while Western Australia typically comes next.

Tasmania and South Australia have bounced around the state rankings more, but that’s unlikely to be because voters there are constantly changing their minds.

Rather, those are the states in which pollsters would be finding the smallest samples, and therefore the figures in those states have larger margins of error.

Across each of those polls, there was at least a 9 percentage point difference between the state with the most support and the least support.

That tells us quite clearly that it’s possible for most Australians to support the Voice, but still see the referendum fail.

And that’s the easiest way to defeat the referendum.

Where to from here?

More recent polls from different companies have shown tighter national figures than we saw in March.

The most extreme of these is a Resolve Strategic poll published in the Nine newspapers, which found 49 per cent support for the Voice, and a majority against the Voice in three states.

When considered in aggregate, the polls have been tightening all year, according to a polling average model devised by Professor Simon Jackman from the University of Sydney.

In the first few months of the year, support for Yes fell by about a quarter of a percentage point a week, but that has accelerated to almost half a percentage point a week in the last month or so.

The averaging model uses what we know about poll sample sizes and margins of error to also calculate a margin of error for the average.

The result is not a prediction of the referendum outcome, but an effort to interpret the existing published polling.