Abstract
Entrepreneurship education and training experiments are usually encouragement designs, where training is offered to a randomly assigned treatment group. In encouragement designs, some subjects assigned to the treatment condition do not actually take the treatment, and some subjects assigned to the control condition manage to receive the treatment. A new modeling approach is warranted in order to account for this non-compliance. Field experiments in entrepreneurship must anticipate four subject types: Compliers, NeverTakers, AlwaysTakers, and Defiers. By anticipating two-sided non-compliance and using these terms, entrepreneurship researchers can estimate the effect of interventions on entrepreneurs who actually comply with the training as well as the efficacy of the overall program. Integrating this straightforward methodological innovation using instrumental variables regression can add precision and credibility to entrepreneurship education and training experiments.