Inducing risk preferences in multi-stage multi-agent laboratory experiments

Though there has been some debate over the practical efficacy of using binary lotteries for controlling risk preferences in experimental environments, the question of its theoretical validity within the contexts it is often used, namely multi-stage multi-agent settings, has not been addressed. Whilst the original proof of its validity featured a single-agent single-stage context, its practical use has seen a wide range of implementations. Practitioners have implicitly assumed that whenever the setting and form of implementation they have chosen deviates from the original single-agent single-period proof, it remains theoretically valid. There has been virtually no debate in the practitioner literature on the theoretical validity of binary lotteries in a more general context, or on whether the form of implementation matters. The current article addresses these questions, establishes limitations on validity and suggests some design principles for future implementation of binary lotteries for the purpose of controlling risk preferences.