Most currently employed Integrated Assessment Models are of a dynastic nature, commonly assuming a fixed relation between pure time preference, economic growth and interest rate. This rigid relation has led to much debate on which level of discounting to adopt. Especially the quantitative results of Integrated Assessment Models have been subject to controversy because of their strong sensitivity on future discounting. Many economic analysts advocate employing a descriptive time preference, based on historic data, which usually represents an approximate efficient use of environmental resources. Others encourage assuming a prescriptive time discounting, allowing them - by taking low discount values - to model a sustainable use of environmental services. This paper argues that, although a fixed time preference relation might be convenient for economic analysis, such a supposition can be misleading. By avoiding a rigid discounting relation, the descriptive-prescriptive controversy can be avoided. It is concluded that dynastic Integrated Assessment Models are in many respects inappropriate for providing policy makers with quantitative figures about the costs of carbon dioxide emissions, and their desirable reduction levels. In contrast, Overlapping Generations models, allowing the use of a flexible discounting relation, are suitable for designing a broad class of policy instruments. With the Integrated Assessment Model ALICE 2.0, it is shown how various assumptions on demographic change and public institutions can affect the interest rate, thereby influencing the desired optimal greenhouse gas emission reductions. It is recommended that economic modelling intensify attempting to establish both efficient and sustainable resource use. The discount rate should be treated as an endogenous variable, rather than an exogenous parameter characterising a central planner.