The methodology for ACM's survey on how Australians plan to vote in the Voice to Parliament referendum

In an ACM survey of 10,131 people, across 10 days in June, 38 per cent of respondents said they would vote "yes" for the Indigenous advisory body, while 55 per cent say they will vote "no", leaving 7 per cent undecided.

In January, a smaller-scale ACM survey of 1060 Australians found 63 per cent of respondents would vote "yes" on the Voice.

The survey used a dual sample collection approach to reach respondents, ACM's Chi Squared research director Alex Mihalovich explained.

About 1000 of the responses came from the CrackerJack panel, an opt-in panel of 9286 regional Australians, which ACM distributed the survey to.

But the bulk of respondents were reached through existing ACM publications, and polled between June 16 and June 26.

The data collected was then reweighted in line with demographical data from the Australian Bureau of Statistics.

"We recognize that a lot of our readers, the skew or the bias, if you like, is towards regional Australia as opposed to metropolitan Australia," Mr Mihalovich said.

"It's also skewed older, a lot of our readers are older, not younger. And a lot of our readers are skewed towards male and not female.

"So what we recognize is that intrinsically in in that sample, there is a skew or what we call a bias."

The researchers compared their sample with the ABS demographic data on metro/rural splits, age demographic splits and male/female splits.

READ MORE:

"Then what we do is we reweight the data," Mr Mihalovich said.

"So for example, let's say for example in our sample, we have 60 per cent male and 40 per cent female, but in the Australian population, it's 50-50.

"What we do is we up-weight, the 40 per cent of females to make it 50 per cent, and we down-weight the 60 per cent of males to make it 50 per cent, so that it accurately reflects the true Australian population.

"And we do that that re-balancing for age, gender and geographic region."

Mr Mihalovich added the large sample size and the reweighting of data left researchers satisfied with the survey results.