This paper presents a technique which uses statistical behavioral models to estimate the level of free
riders in DSM programs. These models area variation of the discrete-choice participation models which
have been used extensively in DSM impact evaluations. These behavioral models relate the decision to
undertake the conservation action promoted by the utility (whether or not it was done through the
program) as a function of demographic and attitudinal variables as well as the program incentive
the rebate level or the loan buydown).. The level of free-ridership is then determined by finding the
probability of undertaking the conservation action given that there is no program incentive.

In addition to estimating the level of free riders, these models can also be used to simulate the level of
free riders when the program incentive is changed or different segments of the customer population are
targeted for participation. Therefore, these models can be used for both program design and program
evaluation.

These models represent a potentially powerful tool for DSM evaluation, since they quantify the level of
free-ridership and produce testable results. In addition, unlike other survey-based methods, these models
do not require a subjective interpretation by the researchers and are not prone to hypothetical bias and
cognitive dissonance. However, they may be influenced by the presence of free drivers.

The paper reviews the theory behind these models and then presents estimated results from a recent
impact evaluation of a residential audit program.