Cross-classified Multilevel Logistic Regression Model

A Randomised Control Trial (RCT) asked whether the use of tailored feedback could be used to encourage participation in a kerbside food waste recycling scheme in Oldham, Greater Manchester. The rationale for the trial was that most people underestimate the extent to which their peers engage in pro-social behaviour and compare their own behaviour to these under-estimates. We randomly assigned 318 streets to a treatment and control group. Households in the treatment group were sent two postcards that provided feedback on how their street performed on food waste recycling compared to the average for their neighbourhood. Participation in the food waste scheme was measured for all households on three occasions：at baseline, and after the receipt of the first and second feedback cards. We estimated the effect of our treatment using cross-classified multilevel logistic regression models, controlling for baseline, street size and the interaction of treatment and baseline. Normative feedback can have a positive impact on participation but the impact is small. The impact was greater on streets that were small (with between 8 and 30 households) and where the baseline was low. We also estimated that with the feedback postcard nudge, we can reduce food waste which can contribute 5,332 pounds worth less of food waste in the area. From the trials, feedback had the positive cumulative effect on the participation rate.