Time-of-use prices and electricity demand: Allowing for selection bias in experimental data

We address self-selection in time-of-use experiments. Our methodology is especially appropriate when (i) theory does not provide an exclusion restriction between the participation and consumption equations or (ii) the demand system contains a large number of parameters estimated from a difficult objective function. We find that correcting for selection bias is important. Generally, small commercial establishments are not very responsive to time-of-use pricing. However, for some subgroups (such as those with neither electric heating nor air conditioning), significant responsiveness occurs given a sufficiently short peak period and a sufficiently large peak/off-peak price differential.