To begin the analysis we set out to profile the five types of households (as shown in Table 1). First we looked at the proportion of each in the sample. Of the total sample of 2,278 households, 1,084 (47.6%) had no champion present (NCP), 173 (7.6%) had a champion who was not the respondent (CNR), 530 (23.3%) had an energy-efficiency champion who was the respondent (EEC), 96 (4.2%) had a climate-change champion who was the respondent (CCC), and 395 (17.3%) had a combined energy efficiency and climate-change champion who was the respondent (EEC/CCC).
Second, we considered how the five household types differed in their attitudes (Table 1). The NCP and EEC household types had the lowest acceptance of anthropogenic climate change (63–75%), with the remaining household types having acceptance levels between 7–30% higher (p < 0.001). A similar pattern of outcomes was found for worry about climate change, issue importance, and issue involvement (p < 0.001). For efficacy, the mean response was highest for the EEE/CCC group which was 44.5% higher than for NCP households, with the mean for the remaining household types 21.7 to 29.9% higher than NCP households.
Third in the profile was an examination of the households across the four behaviours (Table 1) examined in this study. This revealed that behaviours were higher in households with champions. The mean for the CCC and EEC/CCC households was 58.9 to 65.3% higher respectively than the NCP households. Means for EEC only households and CNR households were 23.4% and 29.0% higher than NCP households. Overall, this demonstrated that when a champion was present in the household, climate-change and energy-efficiency behaviours were significantly and meaningfully different/higher.
Fourth was a sociodemographic profile of the five household types. Significant differences were found for education, with CCC and CCC/EEC household types having the highest level of education. Significant differences were also found for age (EEC highest age, CCC lowest), and also for employment status, with EEC households the most likely to be retired and CCC the least likely. Significant differences were also found for gender (female higher for CCC and CNR), while no significant differences were identified for income.
Finally, we examined differences in political persuasion and religious behaviours between the five household types (Table 1). For political persuasion, negligible to moderate differences were observed in the proportion of Labor voters across household types (from 0.4–9.6%), with more substantive differences observed for Liberal voters, with the EEC/CCC and CCC types having the lowest proportion of Liberal voters (20.2 to 24.1%) and the EEC types having the highest (39.5%). Unsurprisingly, the EEC/CCC and CCC types had the highest proportion of Greens voters. Significant differences were also identified across religious behaviours: NCP households had the highest percentage that were neither devout nor religious (72.7%), with EEC having a marginally lower proportion. The lowest percentage for neither was CCC households (58.3%) followed by EEC/CCC households (64.6%).
This profile of household types revealed significant differences between households with and without champions, in terms of attitudes, beliefs, intentions and behaviours, and sociodemographics. The variation in behaviour provided evidence of the impact of a champion on households and indicates that different approaches may be required for households with and without champions.
To strengthen this claim, the next step was to assess the effect of a champion being present in the household independently of the effect of differences in sociodemographic profile, political persuasion and religiosity. To do this we estimated five ordinal regression equations for :1) involvement, 2) issue importance, 3) efficacy, 4) past behaviour and 5) behavioural intentions using no champion present as the reference category (Table 2) (see Methods for more details). In each of these regressions, every champion type was found to be significantly different from no champion present and had positive coefficients. That is, households with champions had greater involvement, felt the issue was of greater importance, and had higher efficacy, had higher behavioural intentions and engaged in more mitigation behaviours than households without champions.
To further understand these results, we compared the regression results across champion types. To do this we then used Wald tests to identify significant differences (Supplementary Material). The degree to which the champions spoke to their family and friends about the issue (involvement) and issue importance significantly differed across all champion types except for between champion present but not the respondent and energy-efficiency champion. Households with super champions – that is both efficiency champions and climate-change champions (EEC/CCC) – had the (significantly) highest coefficient for household type in all of these regressions.
After the EEC/CCC households, the second highest coefficients for involvement and efficacy were for the CCC households, while for importance it was for the EEC households. The super-champion households also had the highest coefficient for behaviour and behavioural intentions, and this was significantly different from all other household types apart from CCC households. For both behaviour and behavioural intentions, the magnitude of the coefficients for CCC households was higher than for EEC households, but this contrast was not significantly different. Overall, these findings suggest that the presence of an EEC and CCC or EEC/CCC have unique effects on household behaviour and behavioural intentions, again pointing to the need for differentiated communication and/or policy approaches.
Returning to Table 2, in addition to champion types, we also included sociodemographic variables, religiosity and political persuasion in the ordinal regressions. Significant and positive associations were identified for gender (being female) (importance, efficacy and behavioural intentions) and degree-level education (involvement, behaviour and behavioural intentions). Significant and negative associations were found for age in three of the regressions (importance, efficacy and behaviour), which is consistent with previous Australian research16, showing that older people tend to believe that action by individuals would make little difference to climate change, that climate change is less important to them personally, and are less likely to engage in energy-efficient behaviour.
For the three religiosity variables (Table 2), being devout (those who take their religion seriously) has a positive significant coefficient for involvement and efficacy. In contrast, being regular (attending a religious service regularly, but not devout) has a negative significant coefficient for efficacy, while being devout and regular has a positive significant coefficient for behaviour.
Perhaps the clearest set of results in Table 2 were for political persuasion. Here voters supporting the Green Party (left-wing party) have positive and significant coefficients for all five dependent variables. In contrast, the conservative voters (Australian Liberal Party and Australian National Party) have negative and significant coefficients.
In terms of magnitude of effect across all types of variables, the largest effects are from having champions present, especially climate-change champions or both energy-efficiency and climate-change champions. This is followed by political persuasion, religion, age and education.