OBJECTIVES: Much recent debate has focused on the merits of differential discounting of costs and health effects. Yet relatively little attention has been paid to the merits of using discount rates that are time-dependent (i.e. non-constant). Recent theory shows that appropriate discount rates depend upon the real rate of borrowing. Since this is determined by the bond market, and since real yields differ on bonds of different maturity, this implies that discount rates ought to be non-constant. Recent research has also demonstrated that conventional objections to non-constant discounting, such as the risk of time-inconsistency, may no longer hold. Our objective is to shed light on this under-researched topic and to ignite a debate about the merits of non-constant discounting. METHODS: We review the theoretical and empirical literature around the use of non-constant discounting, consider whether non-constant discounting is appropriate for social decision making, and (if so) propose how decision makers can incorporate non-constant discounting in a way that is time-consistent and which accounts for intergenerational equity and other social value considerations. We also consider, and propose solutions to, technical hurdles associated with adopting non-constant discounting. RESULTS: The conventional approach to discounting (using a constant rate) is appropriate only in special cases where specific assumptions are adopted. In general, non-constant discounting is preferable. Recent work has overcome both the theoretical and technical hurdles to the adoption of non-constant discounting CONCLUSIONS: Decision makers should reconsider their existing discounting methodology to ensure that it is compatible with their perspective on social choice, any budget constraints faced, and other considerations. Where non-constant discounting is found to be appropriate, it should be embraced by decision makers.