Variability of appliance using behavior, i.e., the way it distributes across a day, is linked to the context structure of a behavior. Because it is not free to distribute just “anywhere” throughout the day, it can be assumed that certain times are more suitable for appliance using behavior to be shifted to than other times.

1. 1.

   The indicator FIAD has been further developed in Waseem, Sajjad, Martirano and Manganelli (2017) to theModified flexibility index of aggregate demand (MFIAD), but the conceptual idea remains the same.

2. 2.

   The research design was approved by the Technische Universität Braunschweig, Institute of Psychology Ethics Committee. Project approval Number: D–2018–01.

3. 3.

   A date, not sample size was employed as criteria for ending the survey after four weeks. The survey period was extended once for two weeks. Return rates dropped after about two and a half weeks into the survey, so another call for participation was sent out. See Appendix H for return numbers.

4. 4.

   Case number 1049.

5. 5.

   Case number 1049 chose weekend behavioral pattern 4.

6. 6.

   Case number 1049 falls into the category 71 to 80%.

7. 7.

   Case 1049 falls into the category of 30 years or younger and female. Additional descriptive characteristics (distribution of age, income and distribution of answers to the question about time spend per week on occupation) are in Appendix J.

8. 8.

   The x-axis displaying time of day is only precise to one hour but TUD limits are exact to ten minutes. The limits are put in the middle of the category label when falling exactly to a full hour and in the other cases before or after the full hour category regardless of the 10-minute interval.

9. 9.

   Looking at the answering pattern of this subject, one sees that for all appliances all possible usual using times (24) are selected. Checking the comment section points towards the participant not answering the question, but checking all possible time boxes because “I decide when my washing machine or dryer runs, when I drink coffee or stream a hardcore strip”. The answers from this subject are reported, but cannot be interpreted as usual times of using an appliance.

10. 10.

    Six participants (VP_26 / case 541; VP_31 / case 559; VP_49 / case 795; VP_51 / case 866; VP_78 / case 1037; VP_90 / case 1074) answered in a way describable by a linear function with a y-axis intercept of 0. According to the task this is interpreted as no behavioral effort required for shifting this appliance using behavior in time under current context structure. Looking at the comment section, this interpretation is problematic for VP_78 (shortened and translated): “I can use my washing machine either at 8 a.m. (before work) or from 6 p.m. onwards (after work), […]. If someone would ask me to run it later I would do so without wanting money for it. Since our washing machine runs approx. 3 hours, we start it the latest at 7 p.m. […]. If there was a person to come to our home to turn on the washing machine, I would be fine with it running also at other times.” This shows that the participant does have difficulty changing the using time due to context restrictions, but instead of answering the task in such a fashion, the statement is made that a change in behavior would not require monetary compensation, but the possibility to do so. The case will be reported, but should not be interpreted as a linear curve type.

11. 11.

    Upper limits are imposed by the provided scale ranging up to 10 €. For each scale participants could choose to not answer for the specific time shift. This no answer option often lies around provided BAC values and some comments suggest the upper limits sometimes being too low. So, instead of not including these time shift points in the curve description, not answering in those cases is included for the curve fitting as a BAC value of 11 € changing the upper limit.

12. 12.

    VP_78 with curve type linear is excluded from the loglinear model fitting. The reason is given in footnote 39.

13. 13.

    The counts in the cells of the I x J table are assumed to be independent events from a Poisson random component,\(n_{ij} \sim Poisson\left( {\mu_{ij} } \right)\) and the cell counts are linked to the explanatory terms using the log link (Agresti, 1996).

14. 14.

    In the glm() function in R the following specification is made: family = poisson(link = log).

15. 15.

    This is an aspect which is also discussed for the categorization approach of activities in TUD and since the BAC study design followed the same approach, it is very likely that some of the appliance using behaviors treated as one operant include also other behaviors which use the same appliance, but are a different behavior. This is another source of error.

Authors and Affiliations

1. Braunschweig, Germany

   Farina Wille

Corresponding author

Correspondence toFarina Wille.

Reprints and permissions

© 2021 The Author(s), under exclusive license to Springer Fachmedien Wiesbaden GmbH, part of Springer Nature

Policies and ethics