This paper orders in preference various scheduling policies comprised of dispatching and regeneration rules in a multi-level assembly production system. Traditional expected-value statistical and second-degree stochastic dominance preference ordering rules are used to identify the most efficient-scheduling policies for risk-averse managers using various measures of performance. The results indicate that selection of an efficient regeneration rule is contingent upon the selection of a dispatching rule and these rules must be selected jointly to develop efficient scheduling policies. For this study, simple intuitively appealing scheduling policies were found to be efficient.