Hyperbolic Discounting

Also called Hyperbolic Time Discounting, Hyperbolic Discount Function

Quick answer

Hyperbolic discounting is the way people value a reward less as it gets later, with the biggest drop in value over the first short delays. It makes a person take a smaller reward now over a larger one later, yet choose the larger one when both are far away. Plans made in advance therefore get broken when the date arrives.

~20 sec read

012 min

The problem Hyperbolic Discounting names

Say a freelance designer is offered a choice. They can have $100 today or $110 in one week. Many people take the $100 today, because a week of waiting feels long.

Now change the question. The designer can have $100 in one year or $110 in one year and one week. The gap is the same: one more week of waiting for $10 more. Many people now take the $110, because one more week at that distance feels short.

Each answer is reasonable alone. Together they are a problem. The designer who picks the $110 in the second question is planning to wait one week for $10. A year later, the first payment is one day away and the choice is the first question again. The designer takes the $100. Nothing new was learned. Only the distance changed. (The $100 and $110 here are made-up numbers.)

The same reversal costs real money and time. Picture a team that plans to fix a slow checkout page after the launch. When the launch is over, the team plans to fix it after the next launch. Or picture a founder who plans to save part of next month's income. Next month arrives and the plan moves again. Every plan looks sensible when it is made, and it is never carried out.

Economists describe this gap between plan and action as a failure of self-control. Hyperbolic discounting is one model of why it happens.

023 min

Understanding Hyperbolic Discounting

The designer's reversal makes sense only if the cost of waiting changes with distance. Money today is worth more than the same money later, because it can be spent or invested now. This is the time value of money. The open question is how fast the value drops as the wait gets longer. Discounting is the name for that drop.

The standard model in economics is exponential discounting. It takes away the same percentage of value for each extra period of waiting, such as 1 percent a day. Under that rule, a person's preferences are time-consistent, which means that the choices they make later confirm the plans they made earlier. Two rewards one week apart keep the same ranking whether they are one week away or one year away. A person who prefers $110 in 53 weeks to $100 in 52 weeks also prefers $110 in one week to $100 today. The designer's reversal cannot happen under this rule.

Hyperbolic discounting changes one thing: the percentage lost is not constant. It is large for the first days of waiting and small for days far away. One common form is:

V = A / (1 + k × D)

V is what the reward is worth today. A is its amount, D is the days of delay and k is a number for how fast this person's valuation drops. A bigger k means a faster drop.

The table uses k = 0.05 per day. That value is made up to show the effect and is not a measurement of anyone.

ChoiceDelay in daysWorth today
$100 now0$100.00
$110 in one week7$81.48
$100 in one year364$5.21
$110 in one year and one week371$5.63

Today the $100 wins, because 100.00 is more than 81.48. A year ahead the $110 wins, because 5.63 is more than 5.21. The far values are small because this made-up k is large, so only the ranking matters. One extra week of waiting costs a lot when it comes right after now. It costs very little when it comes after a year. That difference is the whole effect.

The day the choice reverses

With the same made-up numbers, the reversal has a date. Set the two worth-today values equal and the sooner reward is 50 days away. If the $100 is more than 50 days away, the $110 wins. If it is closer than 50 days, the $100 wins. So the designer's plan to wait holds for weeks and then ends. The date of the reversal can be predicted from k, the two amounts and the gap between them.

A simpler form for models: β and δ

Hyperbolic formulas are hard to fit into long economic models. In 1997 David Laibson used a simpler two-number form that Phelps and Pollak had proposed in 1968 for a different problem. Under it, a reward now counts in full. Every later reward is multiplied once by β, a number between 0 and 1, and then by δ for each period of waiting. The single drop at the start makes the present count extra, and present bias is the name for β below 1. This form reproduces the reversal between now and later. It has a steady rate after the first period, so it only approximates the hyperbolic shape.

031 min

Where Hyperbolic Discounting comes from

The hyperbolic shape was proposed in the 1970s and 1980s by researchers studying self-control. In 1975 George Ainslie suggested that a reward's value falls in proportion to one divided by its delay. Richard Herrnstein (1981) and James Mazur (1987) then proposed the form shown above, in which the amount is divided by one plus a constant times the delay.

The word hyperbolic comes from the shape. A reward's value, plotted against its delay, follows a hyperbola, a curve that falls fast at first and then flattens.

041 min

What experiments show about Hyperbolic Discounting

Proposing a shape is easy, so the next question was whether people's answers follow it. The first widely cited numbers came from a survey. Richard Thaler (1981) asked people how much money in one month, one year or ten years would make them as happy as $15 today. The middle answers were $20, $50 and $100. Written as yearly interest rates, that is 345 percent for the one-month wait, 120 percent for one year and 19 percent for ten years.

A person who discounted at one steady rate would give the same yearly rate at every distance. These people did not. The rate was very high for short waits and much lower for long ones, which is the hyperbolic pattern.

Later studies tested the reversal itself. They found that a choice between two delayed rewards can reverse toward the nearer one as both dates approach. Reversals of this kind have been seen in people and also in pigeons.

051 min

How Hyperbolic Discounting plays out in a health club

The clearest test outside a lab is a health club. A visitor can buy a pass for ten visits at $10 a visit, or pay a flat fee every month. Economists Stefano DellaVigna and Ulrike Malmendier followed 7,752 members of three U.S. health clubs over three years.

Members who chose a flat monthly fee above $70 went to the club 4.3 times a month on average and paid more than $17 for each visit they could expect to make. Over a whole membership they gave up about $600 compared with paying per visit.

This case follows the pattern described above. A member signs up when the visits are weeks away and plans to go often. Each visit then has to be chosen again on the day, when the effort is immediate and the benefit comes later.

monthly feeflat feeevery monthten-visit pass$10 a visit
Look at what each choice charges: one fee that repeats every month, or a price for each visit.

061 min

A second case: a month of food

A health club shows people choosing a contract once, so the next case looks at daily choices with no contract at all. Jesse Shapiro studied food stamp recipients, who receive their food benefit on one day each month. He used a 1989–91 food intake survey run by the U.S. Department of Agriculture.

Recipients ate 10 to 15 percent fewer calories by the end of the 30 days after the benefit arrived. Shapiro's estimates of daily patience did not fit any reasonable exponential model. He also tested other explanations, such as food losing value as it is stored, and his data rejected them.

One variable differs from the health club. Members chose a contract once, in advance, while these households chose what to eat again every day. The contrast shows that the pattern appears both in sign-up decisions and in basic daily consumption.

071 min

What Hyperbolic Discounting means for your work

These two cases matter to anyone who plans work or sets up choices for other people. Picture a product manager building a roadmap in January. Every quarter has a cleanup item for slow pages and old code, and every quarter the item moves to the next one. The manager chooses again each quarter, when the cost of cleanup is this quarter's work and the benefit is far away. Under the hyperbolic pattern, the item is ranked below this quarter's work every time. Before the year began, when every option was far away, the same manager ranked the cleanup first. This is procrastination at the level of a team. The decision to change is when to decide: put the cleanup in the plan while the options are distant, and make it costly to remove.

A designer sees the same pattern on the user's side. A new user decides whether to finish setup when the effort is now and the benefit comes later. Moving the first visible benefit earlier in the flow shortens the wait that costs the most. A founder faces it with company money. Cash reserves, savings and invoicing all have costs now and benefits later. A rule set in advance, such as moving a fixed share of revenue to a reserve on the day it arrives, removes the daily choice.

082 min

How to apply Hyperbolic Discounting

In every case above, the problem arises at the moment of choice, when the near reward is close and the larger one is not. The aim is to stop depending on that moment.

  • Decide while both options are far away. The reversal appears when the smaller reward gets near, so a decision made earlier is the one the person still agrees with later.
  • Bind the decision. Precommitment means arranging things now so that the later choice is hard to change, such as an automatic transfer on payday or a deadline that someone else enforces.
  • Plan for the right kind of person. A naive person expects future preferences to match current ones. A sophisticated person predicts the change and plans around it. A rule that works for the second can fail for the first.
  • Watch one signal. A task that moves to the next period twice in a row shows this problem.

Save More Tomorrow is a real plan built on these rules. Richard Thaler and Shlomo Benartzi designed it and published it in 2004. An employee signs a form today that raises their savings rate only when the next pay raise arrives. At signing, both the cost and the benefit are in the future, so the choice is made between two distant options. Over 28 months, participants raised their average savings rate from 3.5 to 11.6 percent, more than three times as much.

28 months3.5 percent11.6 percent
The form is signed now, and the change comes later. Compare the two bars on one scale.

091 min

When Hyperbolic Discounting breaks

The pattern also has limits that are easy to miss.

Hyperbolic is not the same as impatient

Hyperbolic discounting is often described as impatience, and that reading is wrong. A person can be very impatient and still consistent. An exponential discounter with a high rate wants rewards soon, but keeps the same plan when the date arrives. Hyperbolic discounting is about the shape of the drop, and the shape is what changes the plan. Two people can both take $100 today. Only the hyperbolic discounter also plans to wait a year and then changes their mind.

It does not say which choice is right

The model predicts that choices will reverse. It does not say that the earlier plan was the better one. A person who takes the smaller reward today may have a good reason, such as a bill due this week. The pattern becomes a problem only when the person later wishes they had kept the plan.

102 min

Where the evidence on Hyperbolic Discounting is contested

The pattern is well documented, but its details are disputed. Four points matter to anyone who uses it.

Delay brings doubt

A reward that comes later may not come at all. Lab studies tell people to assume that the delayed reward is certain, but delay normally brings some doubt. A steep drop over short waits may therefore partly measure doubt, not impatience. Frederick, Loewenstein and O'Donoghue point out that it is hard to tell how much of the shape comes from time preference and how much from falling confidence that the reward will arrive.

Lab reversals are tested in one sitting

The lab studies cited for reversals show them at a single point in time, with people ranking both pairs of rewards together. The reversal that matters in life is the one that happens later, when the date gets close. Frederick, Loewenstein and O'Donoghue note an implicit belief that the same people would reverse again when the dates arrive. That follows only if a person's discounting stays the same over time, which is an assumption.

The drop may stop after about a year

In the same review, discount rates fell as the horizon grew. After the studies with horizons of one year or less were removed, the link between the length of the wait and the share of value kept was almost exactly zero. The fast fall in value may be mostly a feature of short waits.

Money choices and effort choices differ

Ned Augenblick, Muriel Niederle and Charles Sprenger (2015) ran two versions of one study. People chose between money at different dates, and they chose how much work to do on different dates. In the money version, choices were very consistent over time. In the work version, people showed considerably more present bias. A test that uses only money can therefore miss the pattern.

For a practitioner, the safe reading is narrow. Expect plans to slip when the cost lands now and the benefit lands later, especially for effort. Do not treat k as a fixed trait of a person.

111 min

Hyperbolic Discounting vs. nearby concepts

Four nearby ideas get mixed up with hyperbolic discounting, and each differs on one point.

ConceptWhat it describesWhat separates it
This pageHyperbolic DiscountingIt predicts that a choice will reverse as the date nearsWhat it describes: A discount rate that falls as the delay grows
PsychologyPresent BiasIt is a tendency. Hyperbolic discounting is one model of itWhat it describes: Extra weight on what happens now
BusinessTime Value of MoneyIt follows from interest and needs no psychologyWhat it describes: A dollar now is worth more than a dollar later
PsychologyProcrastinationIt is a behaviour. Hyperbolic discounting is one explanation of itWhat it describes: Putting off a task

Exponential discounting, the standard model, is the sharpest contrast. It can be as impatient as you like and still never reverse a plan.

?4 questions

Questions people ask

How do researchers measure someone's k?

They offer choices between a smaller sooner reward and a larger later reward at several delays, find where the person switches, and fit the formula to those points. The fitted number is the estimate of k.

Does hyperbolic discounting mean people are irrational?

Not by itself. The pattern describes how choices change with distance, and it does not say which choice is right. It becomes a problem when the person later wishes they had kept the earlier plan.

Can a company use hyperbolic discounting to sell more?

Yes. Offers that give a benefit now and delay the cost, such as a free trial or a monthly fee, work with the pattern. Helping customers keep a plan is different from hiding a cost from them.

Is procrastination caused by hyperbolic discounting?

It is one explanation. When a task's cost comes now and its benefit comes later, a hyperbolic discounter keeps moving the start date. Other causes of procrastination also exist.

§6 sources

Sources

Show all 6 sources
  1. David Laibson, "Golden Eggs and Hyperbolic Discounting", Quarterly Journal of Economics 112(2) (1997). doi.org

  2. Shane Frederick, George Loewenstein and Ted O'Donoghue, "Time Discounting and Time Preference: A Critical Review", Journal of Economic Literature 40(2) (2002). This review also reports the studies by Thaler, Ainslie, Herrnstein and Mazur named above. blogs.cornell.edu, PDF copy

  3. Stefano DellaVigna and Ulrike Malmendier, "Paying Not to Go to the Gym", American Economic Review 96(3) (2006). eml.berkeley.edu

  4. Jesse M. Shapiro, "Is There a Daily Discount Rate? Evidence from the Food Stamp Nutrition Cycle", Journal of Public Economics 89 (2005), working paper version of 2003. econwpa.ub.uni-muenchen.de

  5. Ned Augenblick, Muriel Niederle and Charles Sprenger, "Working over Time: Dynamic Inconsistency in Real Effort Tasks", Quarterly Journal of Economics 130(3) (2015), abstract page. resolver.caltech.edu

  6. Chicago Booth Review, "Save More Tomorrow" (2004), a summary of Richard Thaler and Shlomo Benartzi's plan. chicagobooth.edu

Keep reading

More from Psychology

All of Psychology
All of Psychology

Missing a concept?

Tell us what to write next. We’ll email you when it’s live.

Several? Separate them with commas.

Only to tell you when it’s live.

How we use your email: privacy.