two samples selected will identify those social factors that determine participation. Second, the causes of any energy savings can be determined with maximum validity, by se lecting from the sample of nonparticipants a comparison group that includes households as similar as possible to those in the sample of participants.
The lessons we can draw from the existing evaluations of energy conservation programs enable us to conclude that any studies that are limited to technical/ energy measure ments are not particularly useful. The major drawback is that they do not conceive of the results of a program as the outcome of a social process, in which both facilitating and constraining factors interact and overlap. An under standing of these factors is obviously crucial to improving the experimental apparatus.
Our recommendations emphasize the necessity of adopt ing a two-fold approach, which integrates and articulates this interaction between socio-technical results, program characteristics, sociodemographic and other contextual factors, and household participation.
Finally, although we recommend that researchers adopt these methodological guidelines, they also have to take a pragmatic approach, adapted to the type of program evalu ated, and the context into which it has been implemented. The initial hypotheses, will obviously vary. For example they will be different in the case of nonsubsidized energy conservation work by retired couples, as compared with mass communications about wasted lighting addressed to all sociocultural groups. For this reason some degree of flexibility is vital in the specification of the conceptual model, and in the choice of the most appropriate method.
Berry L., 1983. "Residential conservation program im pacts--Methods of reducing self-selection bias." Eval uation Review, 7(6).
Department of Energy, Mines and Resources, 1984. "Eval uation of the Canadian home insulation program." Canada.