Time discounting is the process by which the value assigned to an outcome changes with the delay before it occurs. In economics, it usually refers to placing less weight on a benefit or cost received in the future than on an otherwise equivalent outcome received sooner. It is central to decisions about consumption, saving, borrowing, investment, and public policy. Related research in psychology examines how people compare outcomes available at different times, often through choices between a smaller, sooner reward and a larger, later reward.
Time discounting is not a single mechanism. A delayed payment may be worth less because funds could earn a return elsewhere, because future consumption is expected to be more abundant, because delivery is uncertain, or because a person intrinsically prefers earlier enjoyment. These explanations have different implications and should not be treated as interchangeable. In particular, an observed preference for immediate money does not by itself establish a strong psychological preference for the present.
The principal theoretical distinction is between exponential discounting, which assigns a constant proportional discount rate across delays, and nonexponential models, which allow the rate to vary with the time horizon. This distinction matters because some nonexponential preferences generate reversals between plans and later choices. At the collective level, discounting determines how present expenditures are compared with benefits extending decades or centuries into the future.
Intellectual origins and development
The analysis of time discounting developed from debates about capital, interest, and the distribution of consumption over time. In The Theory of Political Economy (1871), William Stanley Jevons discussed the influence of the anticipated duration and intensity of pleasure on present decisions. Eugen von Böhm-Bawerk’s Positive Theory of Capital (1889) offered an influential account of why present goods could command a premium over future goods. His explanations included differences between present and future provision, imperfect anticipation of future wants, and the productive opportunities associated with using resources earlier.
Irving Fisher gave these questions a systematic treatment in The Rate of Interest (1907) and The Theory of Interest (1930). He distinguished impatience about consumption from investment opportunities that transform present resources into future output. In this framework, a market interest rate emerges from the interaction of preferences, resources, and productive possibilities, rather than representing a direct reading of impatience alone. This distinction remains important: the rate at which someone discounts future utility need not equal the return available on financial assets.
Frank Ramsey’s 1928 article “A Mathematical Theory of Saving” examined optimal accumulation over an extended horizon. Ramsey objected to discounting future enjoyment solely because it was more distant, although later models frequently introduced positive pure time preference. Paul Samuelson’s 1937 article “A Note on Measurement of Utility” provided the compact discounted-utility formulation that became standard in economics. Its attraction was analytical simplicity: an entire sequence of experiences could be evaluated using a utility function and a set of time weights.
Subsequent work challenged the descriptive universality of that representation. Robert Strotz’s 1955–1956 article “Myopia and Inconsistency in Dynamic Utility Maximization” analyzed people whose future choices would depart from their earlier plans. Richard Thaler’s 1981 research documented systematic variation in elicited discount rates across delays, amounts, and gains versus losses. George Ainslie developed accounts of preference reversal based on sharply declining short-horizon discount rates, while David Laibson’s 1997 article “Golden Eggs and Hyperbolic Discounting” helped establish a tractable model of present-focused preferences in macroeconomic analysis.
These developments produced two related but distinct research programs. One uses discounting to construct internally coherent models of intertemporal choice and financial valuation. The other studies how actual choices depart from those models and what explains the departures. Modern behavioral economics draws heavily on the second program, without implying that every observed deviation from a simple model is a mistake.
Discount functions, utility, and rates
A standard discounted-utility representation evaluates a consumption sequence as
where is consumption at date , is the utility associated with that consumption, and is its discount factor as viewed from date zero. The conventional normalization is . Under positive time discounting, generally falls with delay. The utility function describes how outcomes are valued at a given date, whereas the discount function describes how timing changes their weight.
The distinction between a discount factor and a discount rate is essential. A factor is a multiplier applied to a future value. A rate describes the proportional reduction associated with moving an outcome farther into the future. In continuous time, the instantaneous discount rate corresponding to a differentiable, positive discount function is
A declining discount factor therefore does not necessarily imply a declining discount rate. Exponential discounting has a declining factor but a constant rate. Nonexponential functions can produce rates that decrease, increase, or vary irregularly with delay.
Discounting utility is also different from discounting money. Suppose utility has diminishing increments, so that another unit of consumption contributes less when consumption is already high. An individual expecting substantially greater future income may prefer an additional unit of consumption today even with no pure preference for earlier utility. This result reflects marginal utility, not necessarily impatience. Conversely, someone expecting a period of low consumption may place a particularly high value on resources delivered during that period.
A simple monetary comparison illustrates the identification problem. Indifference between receiving 100 units of currency today and 110 in one year implies a one-year monetary discount factor of , or approximately 0.909, if the amounts are valued linearly and the later payment is certain. It corresponds to a 10 percent annual effective rate. With nonlinear utility, transaction costs, borrowing constraints, or doubts about payment, that same choice does not identify a pure utility discount rate.
The basic additive formula also imposes substantive assumptions. It treats utility at each date as separable from experiences at other dates. Habits, anticipation, adaptation, and preferences over improving or deteriorating sequences can violate that assumption. Negative time discounting is mathematically possible as well: a person might prefer an otherwise equivalent enjoyable experience later, perhaps because anticipation itself is valuable. Such preferences require careful interpretation rather than automatic classification as irrational.
Exponential discounting and financial valuation
Under exponential discounting, a future outcome receives the weight
in continuous time, where is a constant discount rate. In discrete time, the corresponding expression is , with . If is an effective interest rate per period, the monetary discount factor is . The continuous rate equivalent to that effective rate is , so numerical rates depend on the compounding convention.
The defining property is multiplicativity:
Discounting an outcome over two successive intervals gives the same result as discounting it over their combined length. Under suitable regularity conditions, this property characterizes the exponential form. It also supports stationarity: moving both outcomes in a comparison forward by the same amount of time leaves their relative discount weights unchanged. A preference between consumption in one year and two years therefore has the same timing structure as a preference between consumption today and in one year.
This structure underlies much present-value analysis. With a constant per-period rate , a certain monetary payment at date has present value . At a 5 percent annual effective rate, 1,000 currency units payable in ten years have a present value of approximately 614. Under discounted cash flow analysis, values are added across dates, and an investment’s net present value subtracts its initial and subsequent costs from its discounted benefits.
Financial discounting is closely connected to opportunity cost and arbitrage. If otherwise identical certain payments could be bought and sold at inconsistent prices, investors might construct offsetting transactions that generate a profit without a net initial investment. Market prices therefore constrain the discount factors used to value tradable cash flows. Those constraints do not establish that every investor has the same psychological discount rate, and risky payments cannot generally be valued with the risk-free rate without an appropriate adjustment.
A constant rate is not necessary for coherent financial valuation. The yield curve can imply different discount rates for different maturities. A deterministic, date-specific sequence of one-period rates produces a consistent collection of discount factors even though no single constant rate describes every maturity. Likewise, changing information can alter valuations without constituting a preference reversal. Exponential discounting is the benchmark for stationary preferences; it is not a requirement that all market yields or all forecasts remain constant.
Declining impatience and preference reversal
Hyperbolic discounting assigns relatively strong weight to differences between near-term delays and weaker weight to equivalent differences in the distant future. A commonly used form is
where . Its instantaneous discount rate is , which declines with delay. Waiting an additional month can therefore matter much more when a reward is imminent than when it is several years away. Generalized hyperbolic functions allow greater flexibility in the shape and speed of this decline.
Declining impatience can generate preference reversal. Consider a person choosing between 100 units available at an earlier date and 120 units one year later. With and delays measured in years, a comparison made ten years before the earlier payment assigns the alternatives values of about 16.67 and 18.46, respectively. The later, larger payment is preferred. When the earlier payment becomes immediately available, the same discount function assigns the alternatives values of 100 and 80. The earlier, smaller payment is then preferred, even though neither payment has changed.
A particularly influential alternative is quasi-hyperbolic discounting, also called the beta–delta model:
where and . The parameter governs trade-offs among future periods, while gives all future utility an additional reduction relative to current utility. When , the model reduces to exponential discounting. Edmund Phelps and Robert Pollak used a related mathematical structure in a 1968 model of intergenerational saving; later work applied it to conflict among successive decisions by the same person.
The behavioral interpretation is present bias: immediate outcomes receive disproportionate weight. This is more specific than simply having a high discount rate. An exponential discounter can be extremely impatient while still making plans that remain consistent as time passes. Present bias instead introduces a discontinuity, or an especially steep decline, near the present.
The resulting problem concerns time consistency. A person may prefer today that a costly task be completed next week, but postpone it again when next week arrives. A sophisticated decision-maker anticipates such reversals; a naive decision-maker does not fully anticipate them. These classifications describe beliefs about future behavior, not intelligence or general competence. Nonexponential discounting also does not invariably produce an observable reversal: the available alternatives, constraints, and actual outcome sequence determine whether a conflict appears.
Measuring discounting in experiments
A common measurement task offers repeated choices between smaller, sooner payments and larger, later payments. Researchers vary the amount, delay, or both, and use switching points to estimate indifference. An adjusting-amount procedure changes one reward until choices approximately balance. A multiple-price list presents an ordered series of alternatives, such as a fixed immediate payment against progressively larger delayed payments. These methods can estimate a discount function, but the estimates depend on assumptions about utility and how subjects understand the task.
One major complication is that receiving money and consuming goods are different events. Participants can save an immediate payment, borrow against expected income, or use payments to reduce debt. Consequently, monetary choices partly reveal access to financial markets and the value of liquidity. An apparently high discount rate may reflect an urgent payment obligation or an expensive borrowing option. A model that assumes linear utility and unconstrained financial access can misattribute these circumstances to impatience.
Payment procedures create additional confounds. Immediate payments may be delivered in person, whereas delayed payments require a bank transfer, another visit, or confidence that the experimenter will remain available. Such differences involve transaction costs and perceived reliability. Researchers often introduce a common front-end delay, comparing payments tomorrow and next month rather than today and next month. This helps equalize delivery procedures, but it can also remove the precise contrast needed to identify an immediacy effect.
Nonlinear utility is another identification problem. If larger payments have diminishing marginal value, the amount required to compensate for delay cannot be interpreted through simple monetary ratios. Joint estimation of time and risk preferences is one response, although the result then depends on the assumed connection between choices under risk and choices over time. Andreoni and Sprenger’s convex time-budget experiments, published in 2012, allowed subjects to allocate payments across dates along experimental budget lines, helping separate utility curvature from discounting under their maintained model.
Several empirical patterns recur, but none supplies a universal numerical discount rate. Larger rewards often produce lower estimated rates than smaller rewards, an effect commonly called the magnitude effect. Gains and losses can yield different implied rates, and preferences over food, money, and other outcomes need not coincide. Hypothetical choices and choices with real payments may also differ. A meta-analysis by Shane Frederick, George Loewenstein, and Ted O’Donoghue in 2002 emphasized the wide dispersion of estimates and the difficulty of separating pure time preference from other influences.
Psychological mechanisms and contextual influences
Time discounting summarizes choices, but it does not by itself identify the psychological process producing them. One explanation emphasizes the representation of future events. Immediate experiences are often concrete and vivid, whereas remote events are described more abstractly. A delayed benefit that is difficult to imagine may receive less attention or weaker emotional weight. Experiments involving detailed imagination of future experiences have sometimes reduced measured delay discounting, although effects depend on the task, population, and intervention.
Attention and framing can alter the same objective trade-off. “Receive 100 today or 120 in a year” highlights monetary amounts and waiting. Expressing the options as calendar dates can change how the interval is perceived. Explicitly stating both the payment and the absence of a payment at the other date can also shift choices: “100 today and nothing next year” is descriptively fuller than “100 today.” These effects show why a discount parameter estimated in one format may not transfer unchanged to another.
Emotions and reward-specific motivation provide another mechanism. Hunger can change the attractiveness of food available immediately relative to food available later, while leaving other trade-offs less affected. Anticipation itself can contribute utility: waiting for an enjoyable event may be pleasant, whereas waiting for an unpleasant event may involve dread. In such cases, preferring later enjoyment or earlier discomfort need not reflect a reversed preference over the underlying experience. It can reflect the additional experiences generated during the waiting period.
Neuroscientific research connects intertemporal decisions with systems involved in valuation, memory, prospective thinking, and cognitive control. McClure and colleagues’ 2004 study reported different patterns of brain activity for choices involving immediate rewards and choices involving only delayed rewards. Kable and Glimcher’s 2007 work linked subjective discounted value to activity in valuation-related regions. These findings support investigation of how timing enters valuation, but they do not establish a simple anatomical division between an irrational present-oriented system and a rational future-oriented system.
Economic circumstances also shape measured patience. An unreliable environment can make waiting unattractive because promised rewards may not arrive. Scarcity can direct attention toward urgent needs, while low access to insurance and credit can make current resources especially valuable. Associations between discounting, income, education, and behavior therefore do not by themselves establish the direction of causation. Preferences may influence circumstances, circumstances may influence choices, and measurement may capture both. Treating a fitted discount parameter as a fixed personal trait can obscure these distinctions.
Saving, borrowing, and dynamic choice
Time discounting is a central component of models of saving and consumption. A household allocates resources across dates subject to income, wealth, returns, and a budget constraint. Earlier consumption provides utility sooner, but saving can finance greater consumption later. The resulting choice depends jointly on time preference, expected resources, utility curvature, and financial opportunities. Discounting alone does not determine whether a household borrows or saves.
For an unconstrained household maximizing expected discounted utility, an interior optimum often satisfies the consumption Euler equation:
where is the real return between dates and . The equation balances the marginal utility sacrificed by saving one additional unit today against its expected discounted marginal benefit tomorrow. It relies on assumptions including suitable preferences, an available marginal saving opportunity, and no binding constraint preventing the relevant adjustment.
With constant relative risk aversion utility and a certain constant return, the relationship becomes especially transparent. If , then
A lower discount factor tends to reduce optimal consumption growth, holding other conditions fixed. But a higher interest rate has both substitution and income effects: it makes future consumption cheaper relative to current consumption while also changing the resources available to creditors and debtors. Consequently, aggregate saving need not respond mechanically to the return.
Present-biased preferences introduce a conflict between current and planned behavior. Someone may intend to accumulate retirement wealth but repeatedly favor immediate spending. The same logic can apply to undertaking education, completing administrative work, or making payments before a deadline. Yet an observed delay does not prove present bias: uncertain returns, fixed costs, limited attention, or rational waiting can produce similar behavior.
Commitment devices restrict later choices in order to support an earlier plan. Examples include restricted-access savings accounts and contracts that impose a cost for abandoning a target. Their theoretical value depends on whether the person anticipates future reversals and whether the restriction addresses the relevant conflict. Automatic enrollment and reminders work differently: they can change defaults or attention without fully restricting choice. Assessing these arrangements within welfare economics raises a difficult question when preferences conflict across time—whether evaluation should privilege the earlier plan, the later choice, or some independently justified criterion.
Social discounting and intergenerational decisions
Social discounting is used in cost–benefit analysis to compare public costs and benefits occurring at different dates. It is especially consequential for long-lived infrastructure, environmental protection, and climate change. At a constant 3 percent annual rate, a benefit of one million currency units received in 100 years has a present value of about 52,000. At 7 percent, its present value is about 1,150. These are arithmetic illustrations, not claims that either rate is appropriate for every project.
A standard consumption-based framework derives the social discount rate from a Ramsey-style relationship:
where is the pure rate of time preference, is the elasticity of marginal utility with respect to consumption, and is the growth rate of per-person consumption. Under the assumptions supporting this expression, future consumption receives less weight both because of pure time preference and because richer future people obtain less marginal utility from an additional unit of consumption. The formula concerns a marginal consumption discount rate; applying it to actual projects requires attention to risk, resource costs, and the nature of their benefits.
The parameters contain different kinds of judgments and evidence. Expected consumption growth is an empirical projection. Pure time preference concerns how welfare at different dates is weighted. Utility curvature can represent willingness to substitute consumption across time and, in suitable social welfare formulations, concern about inequality. These roles are related but not identical. A single numerical parameter may therefore carry several substantive assumptions, especially when a model aggregates people within generations as well as welfare across generations.
An alternative emphasis is the opportunity cost of capital. Public investment can displace private investment, current consumption, or both; the appropriate treatment depends on the financing and economic setting. In distorted markets, the return earned on private capital need not equal the consumption discount rate. Shadow-pricing methods can account for displaced investment rather than simply applying the same market rate to every future effect. Taxes, externalities, and incomplete markets complicate the translation from observed yields to social values.
Intergenerational analysis also raises distributional issues that a single rate cannot resolve. Future populations may differ in size, income, and exposure to environmental damage. A global average growth rate can conceal worsening conditions for particular groups. The Stern Review in 2006 and subsequent debates, including those involving William Nordhaus, illustrated how assumptions about time preference, growth, damages, and risk can produce substantially different evaluations of climate policy. Discounting cannot replace explicit treatment of who receives the benefits and who bears the costs.
Uncertainty, risk adjustment, and long horizons
Uncertainty changes discounting in several distinct ways. A promised payment may fail to arrive, future consumption may differ from expectations, and the returns available for reinvestment may change. These are not equivalent to a pure preference for earlier utility. For example, a payment conditional on surviving to a future date may have a reduced expected personal benefit, even if utility conditional on being alive is not discounted. Likewise, distrust of the payer introduces delivery risk rather than necessarily greater impatience.
In asset pricing, a general valuation relationship is
where is a future payoff and is a stochastic discount factor. In a representative-agent consumption model, it is commonly expressed as
The payoff’s value depends on its covariance with marginal utility, not only on its expected amount and date. A payment delivered when consumption is scarce can be especially valuable. A payment delivered mainly in prosperous states may be worth less despite having the same expected monetary amount.
This explains why a single risk-adjusted discount rate can be misleading for projects with changing exposure over time. Risk may be treated through state-contingent valuation or certainty-equivalent cash flows; alternatively, a suitable rate adjustment can sometimes represent it. Reducing expected cash flows for a particular risk and then adding a premium for the same risk to the discount rate can double-count the adjustment. The direction of the appropriate adjustment also depends on whether a project provides insurance against adverse states or performs poorly precisely when resources are most valuable.
Uncertain future discount rates can themselves generate declining effective rates. Suppose a constant continuous rate will be either 1 percent or 5 percent, each with probability one-half. The expected discount factor is
It is not equal to , the factor calculated from the mean rate. Over long horizons, the low-rate component increasingly dominates the expected factor. Martin Weitzman’s 1998 analysis of the far-distant future and his 2001 work on gamma discounting helped establish the importance of this mechanism in long-horizon policy evaluation.
The precise implication depends on the persistence and structure of uncertainty. A rate fixed forever but initially unknown differs from a short-term rate that fluctuates independently each year. Updating information can also change the appropriate schedule. For long-horizon projects, another distinction concerns relative prices: if environmental amenities or other nonmarket goods become scarcer relative to consumption, their value in consumption units may rise. That change belongs in the valuation of benefits and should not automatically be interpreted as a change in pure time preference.