Experimentalists are increasingly using allocation problems to make inferences about subjects’ preferences – the reason being that allocation problems appear more informative than other types of problems ‒ such as pairwise choices, Holt-Laury price lists and the Becker-DeGroot-Marschak mechanism. At the same time some experimentalists are broadening the type of allocation problem, moving from allocations over just two events to allocations over more than two, again to get more information from experiments. Even with just two allocations, the issue of the error process is already interesting; going to allocations over more than two increases the interest as well as the complexity of the problem. This paper examines some of the various possibilities and solutions, and also carries out a simulation exercise to investigate the problems caused by using the wrong stochastic specification. Finally we report on an experiment and on the results of the econometric analysis of the experimental data.