1 Availability refers to whether or not respondents will have the capacity to understand these frames; accessibility refers to whether or not they have ‘regular or recent exposure’ to the consideration; applicability refers to whether or not, in a competitive context, these frames could fit the phenomenon (Chong and Druckman Citation2007a).

2 We checked our own inductive analysis having two independent readers perform a framing analysis independently, based on a random sample of the articles appearing on the Royal Commission and banking scandals between June, 2017 and November, 2018. The independent readers agreed on four frames: responsibility for regulatory/government failure; victims of banking misconduct; banking system failure; and role of banks in Australia/impact on markets. The first two frames are those we used in our media framing experiment. For further analysis of the media frames used during the Australian banking scandal, see Nicholls and Culpepper (Citation2020).

3 Each article also featured a picture related to the emphasis of the article. We included pictures because they increase verisimilitude – most press coverage that people read of the Australian banking scandal, whether online or in paper, featured accompanying pictures reinforcing a key aspect of the article. The methodological drawback of including pictures is that they may include information beyond that included in the text, which is why we devoted considerable time to identifying pictures that reinforced the core message of the article. The victim-framed article features an editorial cartoon from the Australian Financial Review, showing grave robbers from Commonwealth Bank digging in a graveyard; it also reproduced a tweet that featured in an Australian press article about an individual victim. The regulatory failure article showed a sign held by protestors outside the commission hearings, featuring the slogan ‘Jail-in for bankers not bail-in for deposits.’ The non-partisan article on regulatory capture featured an editorial cartoon from the Australian Financial Review, which showed the head of ASIC, James Shipton surrounded by pigs that looked like piggy banks, representing the banks (see Appendix 1).

4 Breaking up banks in the Australian context refers to separating retail banking from wealth management services, including pension investments.

5 In our survey of 1200 Australians, respondents were asked, ‘Generally speaking, do you usually think of yourself as Liberal, Labor, National, or what?’ Roughly 31 percent (unweighted) opted for Labor, 24 percent for Liberal, 5 percent for National. About 9 percent volunteered their affiliation with the Greens and an additional 6 percent mentioned some other political party. Exactly 25 percent gave no party affiliation; that quarter of respondents are analysed here as non-partisans.

6 There is a second mode of reasoning by which non-partisans might find the third article more appealing. This interpretation draws on the work of Zaller (Citation1992). Zaller argues that individuals who are moderately aware are likeliest to be responsive to new information and to change their policy preferences. If the reader did not know that it was governments that had allowed the growth of ever larger financial institutions – then it would be the case that this article contains new information to which non-partisans are more likely than partisans to be responsive. We are not able to adjudicate between these two interpretations, since they are observationally equivalent.

7 Extracted factor scores were rotated using oblique rotation (promax option) based on our priors that these economic policy factors might not be independent and might be correlated. The components that load heavily on factor 2 are close to zero on factor 1. Most of factor 1 variables sit close to zero on factor 2 and the highest load (bailing out banks) is only 0.2.

8 This ‘baseline’ model in effect captures what Kinder (Citation1998) referred to as the ‘primary ingredients of opinion.’

9 Our statistical analysis of average treatment effects for this systemic capture frame is unchanged when we control for this difference in educational background.

10 In Appendix 3 we present a causal mediation analysis to consider the potential linkage between the affective response and changes in policy preferences of the non-partisans.

11 We thank an anonymous reviewer for raising the important point that the field dates for our data collection overlap with the summer holiday season. The field dates were chosen as a result of wanting to get the survey into the field as soon as possible after the final hearings of the Banking Royal Commission, when we judged the issue would have maximum salience. While proximity to the summer holidays may have affected recruitment and completion rates for the online panellists, there are no priors we can think of on how the field dates would bias how respondents answer our questions or react to our media treatments.