More than a survey

Some of the backers and potential backers of this project have asked me whether this project is a survey.

The short answer is that it is, but that its more than that.

We will be surveying about 500 people about things like their voting intention, and subjecting them to a quick numeracy test.

However we'll be using inferential statistical techniques (primarily a regression model) to look deeply into the results of each individual respondent to the survey. This means trying to show correlations between different groups within the sample, and drawing conclusions about the real world.

This allows us to conduct a true experiment where if we ask people to consider the meaning of non-political data, do they answer differently when presented data about politically controversial topics? We can also look at whether there statistically significant differences between highly numerate Greens voters and highly numerate One Nation voters - or do they react the same way when presented with data that confirms or challenges their pre-existing biases?

This allows us to mathematically show whether or not the motivated numeracy exists in Australia in the topic of our study (closing down coal-fired power stations).

We'll then be able to build a monte carlo simulation to show the probabilities of people who support different political parties getting the correct answer at different numeracy levels. These will be along the lines of the graphs below, produced by Professor Kahan last year.

 

Motivated numeracy results from Kahan et al (2017)

So while a survey is certainly part of the exercise (and certainly the most expensive!), we'll be doing a bit more with the data than what you might see with a typical survey you might read in your local paper.

If you'd like to back this project just click here.

References

Kahan, Dan M. and Peters, Ellen, Rumors of the 'Nonreplication' of the 'Motivated Numeracy Effect' Are Greatly Exaggerated (August 26, 2017). Yale Law & Economics Research Paper No. 584. Available at SSRN: https://ssrn.com/abstract=3026941 or http://dx.doi.org/10.2139/ssrn.3026941