A sample of 500 Australian adults will be recruited to participate by an online polling company which will administer the questions by their online testing platform.  250 participants will be randomly allocated to each treatment condition. Participants were financially compensated by Galaxy Research.


There is a risk that the sample will not be sufficiently polarised by the condition question. To mitigate this risk we have designed the question to be based on a government forcibly closing coal fired power stations  (instead of the gun control question used by Kahan). This question has politically polarised a small trial sample earlier conducted late last year.

Pre Analysis Plan

Specifically this research will seek to conceptually replicate Kahan et al. (2017)’s third hypothesis in Australian participants:

H0: That ideological polarization in the treatment conditions will be the same among participants with low numeracy and those with high numeracy.

The results will be assessed by using a binary logit logistic regression model which is appropriate for analyzing a binary output variable. This will allow for inferences to be drawn about which explanatory variables influenced the participants’ ability to choose the correct answer or not (numeracy or political affiliation).


Browse the protocols that are part of the experimental methods.