The problems and solutions of predicting participation in energy efficiency programs

Highlights: • Energy efficiency pilot studies suffer from severe volunteer bias. • We formulate an approach for accommodating volunteer bias. • A short questionnaire and classification trees can control for the bias. - Abstract: This paper discusses volunteer bias in residential energy efficiency studies. We briefly evaluate the bias in existing studies. We then show how volunteer bias can be corrected when not avoidable, using an on-line study of intentions to enroll in an in-home display trial as an example. We found that the best predictor of intentions to enroll was expected benefit from the in-home display. Constraints on participation, such as time in the home and trust in scientists, were also associated with enrollment intentions. Using Breiman's classification tree algorithm we found that the best model of intentions to enroll contained only five variables: expected enjoyment of the program, presence in the home during morning hours, trust (in friends and in scientists), and perceived ability to handle unexpected problems. These results suggest that a short questionnaire, that takes at most 1 min to complete, would allow better control of volunteer bias than a more extensive questionnaire. This paper should allow researchers who employ field studies involving human behavior to be better equipped to address volunteer bias