Evidence of affiliation of values in a repeated trial auction experiment

Most auction experiments designed to determine subjects' values use repeated trials. This is primarily to give subjects a chance to understand the auction and to learn from market feedback. However, concerns exist that repetition could lead to affiliation of values with subjects following final prices over trials. This could be especially true for items consumers are unfamiliar with. To test these hypotheses, the current research employed repeated auction trials for conventional, organic, and non-GM chocolate bars, with each respectively considered less familiar to subjects. The auction results were first analysed to determine whether the premiums, calculated as the differences in bids between bar types, were statistically significant over each trial. A tobit regression analysis was then conducted to determine which factors statistically influenced premiums over each trial. Analysis showed that the premium for the non-GM bar over the conventional and most of the factors explaining the premiums in the first trial became insignificant after repetition. This suggested strong evidence of affiliation and demonstrated how it can lead to loss of information about subjects' initial formulation of values. It is recommended that, depending on the focus of the study, single trials should be used to avoid these problems.