The form of the Allais paradox known as the common ratio effect (CRE) is a violation of deterministic expected utility theory that has been widely replicated with monetary outcomes. Its robustness has stimulated the development of numerous alternative models of risky choice. However, much less is known about the prevalence of the CRE in decisions involving non-monetary outcomes. We conduct a controlled laboratory comparison of the CRE for money versus consumer goods. The CRE is very strong with money, but largely disappears for goods, primarily as a result of differences in risk attitudes between goods and money. We caution against assuming that findings from experiments involving monetary lotteries will reliably generalise to other types of consequences.