It is commonly assumed that people make intertemporal choices by "discounting" the value of delayed outcomes, assigning discounted values independently to all options, and comparing the discounted values. We identify a class of anomalies to this assumption of alternative-based discounting, which collectively shows that options are not treated independently but rather comparatively: The time difference, or interval, between the options sometimes counts more and sometimes counts less if it is taken as a whole than if it is divided into shorter subintervals (superadditivity and subadditivity, respectively), and whether the interval counts more or less depends on the money difference, or compensation, involved (inseparability). We develop a model that replaces alternative-based discounting with attribute-based tradeoffs. In our model, people make intertemporal choices by weighing how much more they will receive or pay if they wait longer against how much longer the wait will be, or, conversely, how much less they will receive or pay if they do not wait longer against how much shorter the wait will be. This model, called the tradeoff model, accommodates, in a psychologically plausible way, all anomalies that the discounting approach can and cannot address.

According to most models of intertemporal choice, an agent's discount rate is a function of how far the outcomes are removed from the present, and nothing else. This view has been challenged by recent studies, which show that discount rates tend to be higher the closer the outcomes are to one another (subadditive discounting) and that this can give rise to intransitive intertemporal choice. We develop and test a generalized model of intertemporal choice, the Discounting By Intervals (DBI) model, according to which the discount rate is a function of both how far outcomes are removed from the present and how far the outcomes are removed from one another. The model addresses past challenges to other models, most of which it includes as special cases, as well as the new challenges presented in this paper: Our studies show that when the interval between outcomes is very short, discount rate tends to increase with interval length (superadditive discounting). In the discussion we place our model and evidence in a broader theoretical context.

People prefer to receive good outcomes immediately rather than wait, and they must be compensated for waiting. But what influences their decision about how much compensation is required for a given wait? To give a partial answer to this question, we develop the DRIFT model, a heuristic description of how framing influences intertemporal choice. We describe 4 experiments showing the implications of this model. In the experiments, we vary how the difference between a smaller sooner outcome and a larger later outcome is framed-either as total interest earned, as an interest rate, or as total amount earned (the conventional frame in studies of intertemporal choice)-and whether the larger later outcome is described as resulting from the investment of the smaller sooner one. These alternate frames have several effects. First, the investment language increases patience. Second, the explicit provision of the (otherwise implicit) experimental interest rate sharply reduces the magnitude effect. Correspondingly, we find that interest frames increase patience when the rewards are small, but they decrease patience when they are large. Third, the interest-rate frame induces somewhat greater discounting for longer time periods and, thus, reverses the common finding of "hyperbolic" discounting. Thus, many of the "stylized facts" implied by studies involving choices between a smaller sooner and a larger later amount are eliminated or reverse under alternate outcome frames.