Intertemporal Choice

Intertemporal choice is the broad research domain studying decisions whose consequences unfold across different points in time. The umbrella covers saving versus spending, health behavior with delayed payoffs, investment decisions, academic effort allocation, addiction and self-control, and retirement planning. The domain spans economics (since Samuelson 1937), operant psychology (Rachlin 1974 onward), and behavioral economics (where the integration of the two traditions has produced the contemporary literature).

The defining structural feature of an intertemporal choice is that at least one option's consequences are realized at a different time than the decision is made. This separates intertemporal choice from choice under certainty (where consequences are immediate) and from choice under risk (where consequences are uncertain but contemporaneous). Inter-temporal choices often involve both delay and risk, but the construct is principally about the temporal dimension.

Within the domain, several distinguishable phenomena have their own research literatures. Delay discounting — the decrease in subjective value of a reward with delay — is the most-studied measurable behavior. The hyperbolic functional form of that discounting (hyperbolic discounting) is the most-supported mathematical model. Future self continuity is one of the psychological mediators. Present bias, formalized by Laibson (1997) as the β-δ quasi-hyperbolic model, captures the asymmetric over-weighting of the immediate present relative to all future periods.

In economic policy and finance, intertemporal choice is the foundation for retirement-policy design, climate-policy discounting debates, optimal consumption-saving models, and consumer-credit regulation. The shift from classical DU-based policy analysis to behavioral-economics-informed policy analysis since the 1990s has been driven largely by accumulated evidence of intertemporal-choice anomalies that the DU model cannot accommodate. Default-enrollment retirement policies (Thaler & Benartzi 2004) and pre-commitment savings devices are direct applications of present-bias research.

In clinical and addiction research, inter-temporal choice paradigms produce some of the most-replicated transdiagnostic markers in behavioral psychiatry. Steep delay discounting correlates with addiction, obesity, smoking, problem gambling, and ADHD with moderate effect sizes (d ~ 0.4-0.6 across substance categories). The "reinforcer pathology" framework proposes that these conditions share a behavioral signature in inter-temporal choice. Pre-commitment interventions (Disulfiram, Contingency Management) are direct applications of intertemporal-choice theory to clinical practice.

In everyday decision-making, intertemporal choices structure most consequential life decisions: saving for retirement, choosing a college major, exercising, eating well, maintaining relationships through inconvenience, completing long projects. The contemporary research picture: humans systematically deviate from the DU model in identifiable ways, and understanding the specific patterns of deviation enables both better personal decision-making and better policy design. The interventions catalogued in the nudge theory literature are predominantly applications of intertemporal-choice research to consumer welfare contexts.

Economic treatment of inter-temporal choice traces to Eugen von Böhm-Bawerk in the 1880s, whose Capital and Interest (1889) introduced the concept of "time preference" as a primitive of economic behavior. Irving Fisher's Theory of Interest (1930) developed the foundational two-period intertemporal-choice diagram still used in textbooks. The discounted-utility framework that became the modern reference model was introduced by Paul A. Samuelson in his 1937 Review of Economic Studies paper "A Note on Measurement of Utility" — significantly, as a brief technical aside in a paper primarily about utility measurement. Samuelson himself noted that the DU formulation was "completely arbitrary" and made no claim to descriptive realism, but the model nevertheless became the dominant normative and descriptive framework in economics for the next four decades.

Operant psychology developed a parallel literature largely independent of economics. Howard Rachlin and colleagues at Stony Brook in the 1970s applied matching-law principles to inter-temporal choice in pigeons and rats. George Ainslie (1975, Psychological Bulletin) extended the framework to humans and developed the theoretical case that hyperbolic discounting predicts preference reversals (the "picoeconomics" of self-control) that the dominant economic model could not explain. James E. Mazur's 1987 chapter formalized the hyperbolic model in its canonical form.

Contemporary behavioral economics integrated these two traditions starting in the late 1980s. Richard Thaler's 1981 Economics Letters paper "Some empirical evidence on time inconsistency" documented several anomalies in inter-temporal preferences using economic-style elicitation. David Laibson's (1997) Quarterly Journal of Economics paper "Golden eggs and hyperbolic discounting" introduced the quasi-hyperbolic (β-δ) model and the formal theory of self-control through commitment devices — the foundation for the economics-of-self-control literature that has dominated the field since.

The canonical synthesis is Frederick, Loewenstein and O'Donoghue's (2002) Journal of Economic Literature critical review "Time discounting and time preference: A critical review." The review summarized three decades of empirical work, documented the substantial heterogeneity in measured discount rates across paradigms, and argued that "time preference" in the classical economic sense is not a unitary psychological construct but a label for a family of partially-independent influences on inter-temporal choice. Subsequent work has elaborated this thesis through systematic investigation of mediating processes, including future self continuity (Hershfield 2011), episodic future thinking (Peters & Büchel 2010), and identity-based mechanisms (Bartels & Urminsky 2011).

The discounted-utility (DU) model is the classical economic framework. In its standard form, an agent evaluates a stream of consumption (c₀, c₁, ..., c_T) according to U = ∑_t=0..T δ^t u(c_t), where u(·) is an instantaneous utility function and δ is a constant per-period discount factor. The model has several strong assumptions: time-separable utility, exponential discounting (a constant per-period rate produces time consistency), and a single discount rate that applies across all goods, contexts, and stakes.

Empirical research has documented systematic anomalies. Frederick, Loewenstein and O'Donoghue (2002) catalogued the main ones: hyperbolic discounting (per-period discount rate decreases as the delay grows); the magnitude effect (people discount large amounts less steeply than small amounts); the sign effect (discounting is steeper for gains than losses); the delay-speedup asymmetry (compensation demanded to delay a reward exceeds compensation accepted to speed it up); preference for spread (people prefer income that increases over time even when net present value would prefer a flat or decreasing stream); and preference reversals over time. Each is well-replicated and not specific to a single paradigm or population.

Quasi-hyperbolic discounting, introduced by Laibson (1997), is the most-cited modification of DU to accommodate present bias. The β-δ model evaluates U = u(c₀) + β ∑_t=1..T δ^t u(c_t), where β < 1 captures a special weight on the immediate present and δ is the long-run exponential factor. The model produces dynamic inconsistency: today's plan for tomorrow differs from tomorrow's choice. This generates a motive for commitment devices and an economic theory of self-control, with documented applications to retirement saving, addiction, procrastination, and credit-card borrowing.

Other refinements include Loewenstein and Prelec's (1992) hyperbolic model with a generalized hyperbola, multiple-self models from Schelling (1984) and Strotz (1956), and dual-system models from Fudenberg and Levine (2006). No single functional form is dominant across all paradigms; the most honest reading of the literature is that different models capture different aspects of empirical heterogeneity.

Intertemporal choice is operationally measured through a family of paradigms targeting different aspects of the construct. The most widely used are delay-discounting tasks: adjusting-amount procedures (Mazur 1987) and the Monetary Choice Questionnaire (Kirby, Petry & Bickel 1999) discussed in their own entry. These yield individual estimates of the discount rate k under a hyperbolic functional form.

Beyond delay discounting, the empirical inter-temporal-choice literature uses several complementary paradigms. Front-end-delay manipulations compare choice between an immediate reward and a delayed one against choice between two delayed rewards at the same relative delay; they test for present bias and identify the β parameter in quasi-hyperbolic models. Choice-bracketing experiments manipulate whether multiple inter-temporal choices are presented sequentially or as a single bundle. Real-effort task paradigms (Augenblick & Rabin 2019) measure inter-temporal choice over effort rather than money, where the gain-loss framing is clearer.

Each measurement paradigm produces different point estimates of the same underlying preferences. Frederick et al. (2002) documented implicit annual discount rates ranging from negative values to several thousand percent across published studies — variation that reflects real differences in what different paradigms measure, not just measurement error. Researchers in the field have increasingly favored measuring multiple paradigms and reporting them separately rather than treating any single paradigm as the canonical measurement.

The LBL Future Self Continuity Index does not measure inter-temporal choice directly; it measures one of the psychological mediators (future self continuity) that the empirical literature most-strongly links to individual differences in measured inter-temporal preferences.

vs. delay discounting — intertemporal choice is the research domain; delay discounting is the canonical measurable behavior within it. The hierarchical relationship: delay discounting is to inter-temporal choice what reaction time is to cognitive processing — a focused operationalization of a broader phenomenon. Many intertemporal-choice studies use delay-discounting paradigms as their primary measurement; many do not.

vs. hyperbolic discounting — hyperbolic discounting is the specific mathematical model of delay discounting that fits empirical data better than exponential discounting. It is a model within the inter-temporal-choice domain, not an alternative to it. The distinction matters: the empirical phenomenon (subjective value declines with delay) is robust across species and paradigms; the specific functional form is one of several competing models.

vs. future self continuity — future self continuity is one of the psychological mediators theorized to explain individual differences in inter-temporal choice. The relationship is mediator-domain: FSC is the psychological state, inter-temporal choice is the behavioral and decision-theoretic domain. FSC predicts shallower delay discounting and patient inter-temporal preferences; it does not explain all variance in inter-temporal choice.

vs. risk preference — risk preference is the domain studying choices under uncertainty about outcome occurrence. Inter-temporal choice involves delay; risk preference involves uncertainty. The two domains are partially separable but interact: delayed outcomes are often uncertain partly because of the delay. Andersen et al. (2008) and later work has developed joint elicitation of time and risk preferences.

vs. prospect theory — prospect theory is the descriptive model of decision-making under risk. The sign and magnitude effects in inter-temporal choice parallel patterns predicted by prospect-theoretic framing, and recent integrations (Loewenstein & Prelec 1992; Scholten & Read 2010) treat inter-temporal choice as another domain where prospect-theoretic mechanisms operate.

A graduate-school decision

A 24-year-old considers a five-year doctoral program in a field they are interested in. The program requires accepting a stipend substantially below their current entry-level salary, which means five years of constrained spending. The benefits accrue later: a credential that opens specific career paths, completion satisfaction, and longer-term earning potential in a research career.

The intertemporal-choice reading: this is a high-stakes decision involving (a) an immediate consumption cost (five years of reduced income), (b) a delayed identity benefit (the PhD credential), (c) an uncertain long-horizon earnings outcome, and (d) substantial irreversibility once started. The classical DU model would compute expected utility across all periods and pick the higher value. Empirical research suggests respondents typically combine several distinct mechanisms: delay discounting of the future earnings stream, future self continuity with the PhD-completing person, prospect-theoretic loss aversion regarding the income drop, and uncertainty-based discounting of long-horizon outcomes. The pattern of which mechanism dominates predicts not just whether the person enrolls but how they respond to future setbacks.

A health-behavior decision

A 40-year-old whose doctor has recommended weight loss for cardiovascular reasons decides whether to maintain a daily 30-minute walk through a winter season. Each day the choice is between the small immediate cost (cold, time, effort) and the diffuse delayed benefit (cumulative cardiovascular protection over years).

The intertemporal-choice reading: this diffuse-benefit structure is the kind that exponential discounting predicts to be straightforward but that hyperbolic-discounting and quasi-hyperbolic models predict will be systematically under-pursued. The daily cost is immediate and salient; the daily benefit accumulates invisibly. Laibson's (1997) quasi-hyperbolic model predicts the present-self plans to walk tomorrow but the tomorrow-self, facing the immediate cost, often does not. Pre-commitment devices (scheduled appointments, gym buddies, public goals) work by shifting the decision to a moment when the cost is also delayed. Empirical commitment-device literature in health behavior (Volpp et al. 2008; Halpern et al. 2015) confirms this prediction.

Intertemporal choice is one of the most-studied research domains in behavioral economics. The contemporary literature has refined several earlier claims in ways that matter for both research and application.

"Time preference" is not a unitary construct. Frederick et al. (2002) argued, and subsequent work has confirmed, that the family of psychological mechanisms producing inter-temporal-choice patterns includes pure time preference, anticipatory utility, uncertainty about delivery, future-self continuity, choice-bracketing effects, prospect-theoretic framing, and habit formation. Single-paradigm measurements that report "a discount rate" implicitly aggregate over these mechanisms, with weights that vary by paradigm. This is the most-important limitation of the field's standard measurement vocabulary.

Hypothetical-versus-real-reward variance. Most published discount-rate estimates use hypothetical rewards. The empirical literature generally finds similar rates across hypothetical and real-reward paradigms (Madden et al. 2003; Johnson & Bickel 2002), but the magnitude effect and other parameters can be sensitive to whether respondents believe rewards are real. This matters particularly for studies recruiting from low-income or clinical populations.

WEIRD-sample dominance. The intertemporal-choice literature has been criticized for predominantly Western, English-speaking, college-educated samples. Cross-cultural replications exist but are less extensive than the literature's confidence often suggests. Discount rates vary with cultural framings of time and saving, with scarcity exposure, and with socioeconomic background.

Single-paradigm overgeneralization. A single-paradigm discount-rate estimate is often treated in popular treatments as a universal individual constant. The actual literature treats discount rates as paradigm-specific, stake-specific, and domain-specific. Comparisons across studies require careful attention to whether paradigms are commensurable.

The integration with risk preference is incomplete. The classical DU model treats delay and uncertainty as separable; empirical evidence suggests they interact. Joint elicitation of time and risk preferences (Andersen et al. 2008) produces different estimates than separate elicitation. The integration of inter-temporal choice with risk preference and with prospect theory remains an active area of theoretical and empirical work.

The empirical inter-temporal-choice literature uses behavioral tasks (delay-discounting paradigms) rather than self-report inventories as its primary measurement. The LBL Future Self Continuity Index measures one of the most-cited psychological mediators of inter-temporal-choice patterns: future self continuity. The tool measures FSC across three dimensions (similarity, vividness, positivity) at two time horizons (1 year and 10 years), producing a 4-quadrant archetype assignment and evidence-based pathway recommendations. People with strong FSC tend to show shallower delay discounting and more-patient inter-temporal preferences in lab paradigms.

Intertemporal choice is the research domain studying decisions whose consequences unfold over time. The classical model is Paul A. Samuelson's discounted utility (DU) framework, introduced in his 1937 Review of Economic Studies paper. Empirical research has documented systematic anomalies: hyperbolic rather than exponential discounting (Mazur 1987; Ainslie 1975), present bias (Laibson 1997), the magnitude effect, the sign effect, the delay-speedup asymmetry, and preference reversals. Frederick, Loewenstein and O'Donoghue's (2002) Journal of Economic Literature critical review remains the canonical synthesis: "time preference" is not a unitary psychological construct but a family of partially-independent influences on inter-temporal decisions. The most-cited modification of DU is the quasi-hyperbolic (β-δ) model (Laibson 1997), which formalizes present bias and generates an economic theory of self-control through commitment devices. The contemporary measurement picture uses delay-discounting paradigms (Mazur 1987; Kirby et al. 1996, 1999) supplemented by front-end-delay manipulations and choice-bracketing experiments. Psychological mediators include future self continuity (Hershfield 2011), episodic future thinking (Peters & Büchel 2010), and identity-stability effects (Bartels & Urminsky 2011). Honest limitations include paradigm-dependence of discount-rate estimates, hypothetical-vs-real-reward sensitivity, WEIRD-sample dominance, and incomplete integration with risk preference.

This entry is intended as a citable scholarly reference. Choose the format that matches your context. The retrieval date should reflect when you accessed the page, which may differ from the entry's last-reviewed date shown above.

LifeByLogic. (2026). Bounded Rationality: Simon, Satisficing, Heuristics. https://lifebylogic.com/glossary/bounded-rationality/

LifeByLogic. "Intertemporal Choice: Samuelson, DU, and Anomalies." LifeByLogic, 18 May 2026, https://lifebylogic.com/glossary/intertemporal-choice/.

LifeByLogic. 2026. "Intertemporal Choice: Samuelson, DU, and Anomalies." May 18. https://lifebylogic.com/glossary/intertemporal-choice/.

@misc{lblboundedrationality2026,
  author = {{LifeByLogic}},
  title = {Bounded Rationality: Simon, Satisficing, Heuristics},
  year = {2026},
  month = {may},
  publisher = {LifeByLogic},
  url = {https://lifebylogic.com/glossary/bounded-rationality/},
  note = {Accessed: 2026-05-14}
}