Methodological reflections

82

website, were consumers can indicate how many items less they can imagine buying in a set time period. After providing feedback for the energy and water saving potential for this number of items, consumers should be consecutively asked to commit to these items as saving goal and obtain suggestions that can help them reach their goal.

Lastly, the results of the three-month follow up and their practical implications need reflection.

Participants of all intervention groups reduced their clothing consumption at the three-month follow up. We therefore cannot eliminate the possibility that simply counting the items purchased can be a successful strategy for reducing consumption. Future studies with a different set up for the control group are needed to further examine this possibility.

83

therefore recall intervals, the behaviour measurement used in this thesis can be distorted, due to both unconscious recall error and to social desirability. One possible example of a imaginable distortion can be found in Study 3, where on average all participants reported having bought significantly less in the past three months at the follow-up (M = 4.14) as compared to the same three-month period the previous year (M = 7.76) and another three-month period before the intervention (M = 8.88). Any explanation for this can only be of speculative nature. Before indicating how many items they have bought the past three months, participants were asked whether they had bought any items in the past month. The number of items they bought in the past month most likely is smaller than what they bought in the previous three months, therefore the question regarding the last month potentially set an anchor leading to an underestimation of the items bought in the previous three months. Without an objective measurement for the actual number of items purchased we cannot exclude e.g. such anchoring or other effects (Jacowitz &

Kahneman, 1995; Mussweiler, Strack, & Pfeif, 1991).

Equally, with the measurement of behaviour at the level of past-month purchases at different time points throughout the year, seasonal effects might be possible, therefore limiting the temporal validity of the results. This is particularly pronounced in the diary study, as the timeframe for data collection was only two weeks. Purchases in such short time periods might not be representative for the average level of clothing purchase, as in previous qualitative research consumers reported often buying many items at a time. An under- or over-estimation of the number of items purchased on average can therefore not be excluded. Across all studies, we tried to reduce the possible influence of this effect by relying on different time frames for behaviour measurement. In Study 3, e.g., we compare past-month and past-three months behaviour, both at different time points of the year and for the same time period across years.

Lastly, it should be mentioned that we did not measure other things that could have influenced reduced consumption, e.g. financial circumstances and whether they changed during the study period. Another example are higher-order motives (i.e. motives that influence multiple behaviours) other than morality, e.g. frugality or otherwise egoistic, self-enhancement motives (Gatersleben et al., 2017) Frugality is a voluntary forgoing of the acquisition and using of economic goods and services, and it is related to environmental behaviour through values of self-transcendence like unity with nature on the one hand and self-enhancement like ambition on the other. Morality does not necessarily motivate frugality. Possibly, the strong role of personal

84

norms we found would have been diminished if we had included a broader variety of further behaviour determinants.

4.4.2. Sample and procedure

With regard to the sample and therefore generalizability of the results, it is important to make clear the distinctions between Studies 1, 2 and 3. The sample of Study 1 was close to being representative with regard to age, sex, education and region. Limitations to representativeness are due to the fact that the survey was split into two parts, and participants decided themselves whether to return for part two or not. While the first part was completely representative, the procedure led to a self-selection bias and e.g. an underrepresentation of men in Poland or an overrepresentation of older participants in Germany. For Studies 2 and 3, which were both conducted in the United Kingdom, we did not aim for representativeness. The samples for both studies are characterized by an overrepresentation of female participants and lower income classes. Both biases might have influenced the results, about which we can only speculate. On average, we find that women consume more clothing items than men do. The overrepresentation of women therefore could have led to a higher average number of items purchased. Equally, we find a small but significant tendency that more income leads to a higher number of items consumed. This limits the generalizability of the results across the entire population of British citizens.

A further consideration with regard to the procedure is the mixed nature of the intervention conditions in the intervention. Each treatment condition combined multiple strategies, which is preferable if the goal is to achieve behaviour change (Abrahamse et al., 2007; Lokhorst et al., 2013). At the same time, it a disadvantage for further theory refinement, as we are not able to identify the exact mechanisms that brought about change in the intervention conditions.

Lastly, it should be noted that participants in all studies received monetary rewards for their participation, either in the form of vouchers or as cash compensation. This potentially limits especially the applicability of the intervention design in a real life setting. When not provided with an incentive, consumers might not be motivated to follow such an extensive intervention comprising multiple input units and requiring commitment to change one’s behaviour on a voluntary basis.

85