Table 2 Mixed effect models with added covariates, political leaning, and egoistic values

Outcome variable	Best model	LRT	FE (covariate)	FE (treatment)	FE (interaction)
Heating carbon footprint	Best model (see Table 1) + egoistic values fixed effect	5.33, p = 0.021^a	−4.680 (−8.78, −0.78)	−3.563 (−13.30, 7.29) 0.444 (0.40, 0.48) −0.098 (−0.15, −0.04)	NA
Food carbon footprint	Best model (see Table 1) + political leaning fixed effect	8.342, p = 0.004	0.632 (0.20, 1.08)	1.411 (−0.90, 3.72) 0.029 (0.01, 0.05) −0.031 (−0.06, 0.001)	NA
Food carbon footprint	Next best model (see Table SI 2.2, M3) + treatment/egoistic values interaction FE	5.057, p = 0.024^a	0.218 (−0.72, 1.27)	−7.300 (−14.3, −0.50)	2.333 (0.38, 4.41)
Car carbon footprint	Best model (see Table 1) + political leaning fixed effect	7.085, p = 0.008^a	0.913 (0.28, 1.59)	4.424 (0.09, 9.00)	NA
Car carbon footprint	Best model (see Table 1) + treatment/egoistic values interaction FE	6.108, p = 0.040^a	−0.673 (−2.23, 0.87)	−12.557 (−27.70, 1.76)	5.066 (0.89, 9.50)
Civic climate positivity score	Best model (see Table 1) + treatment/political leaning interaction FE	6.35, p = 0.042^a	−0.094 (−0.18, −0.01)	−0.187 (−0.80, 0.42)	0.118 (−0.001, 0.24)
Civic climate positivity score	Best model (see Table 1) + treatment/egoistic values interaction FE	7.43, p = 0.024^a	−0.232 (−0.43, −0.04)	−0.855 (−1.73, 0.02)	0.362 (0.11, 0.62)

Note that analyses with domain-specific outcomes and with egoistic values are explorative. Full details of all models in Supplementary Tables 3.1, 3.2, 3.3, 3.4, 3.5, 3.6, 3.7.
LRT likelihood ratio test, showed for the test against the respective best model in Table 1.
FE fixed effect; for food carbon footprint and heating carbon footprint, the interaction FE between treatment and time is shown (first estimate for treatment, second for time, third for interaction).
95% Confidence interval in brackets.
Bonferroni-corrected alpha value for confirmatory (pre-registered) tests is 0.002, 0.006 for exploratory (non-pre-registered) tests with egoistic values as covariate (across all outcome variables) and 0.008 for political leaning as covariate across domain-specific carbon footprints as outcome variables. ^a Models that do not pass the Bonferroni-corrected alpha value.
N = 7615 (156 participants, 56 days).

Search