People often pick $100 today over $120 next week. Push both dates out by two months and many pick the $120 instead. That flip is hyperbolic discounting. It explains why a small reward now can beat a bigger reward later, and why many loyalty programs pair the two.
What is hyperbolic discounting?
Hyperbolic discounting is a pattern in which the discount rate falls as the delay grows. People mark a reward down steeply over the next few days and weeks. Over months and years they mark it down much less. A 2020 review by Cohen and three co-authors puts it simply. With a hyperbolic discount function, the rate of decline falls as the horizon increases (Cohen et al., 2020).
The result is a split between today’s self and tomorrow’s self. A plan made for a distant reward does not survive when the reward gets close. David Laibson wrote that this pattern makes preferences inconsistent over time. It gives people a reason to bind their own future choices (Laibson, 1997).
The idea has a long history. The same review traces non-exponential discounting to Strotz in 1955 and to George Ainslie’s work from 1975. It lists present-biased preferences, from Laibson in 1997 and O’Donoghue and Rabin in 1999, as a close relative (Cohen et al.). The two names point at one problem: the near term looms too large.
How is it different from exponential discounting?
Standard economics assumes exponential discounting. Each extra day costs the same share of value. Hyperbolic discounting front-loads the penalty. The Cohen review plots three functions, and the table sums them up (Cohen et al.).
| Model | Discount function | Rate as delay grows | Choices flip over time |
|---|---|---|---|
| Exponential | Value falls by the same factor each period | Constant | No |
| Hyperbolic | (1 + αt) to the power −γ/α, from Loewenstein and Prelec | Falls | Yes |
| Quasi-hyperbolic | 1 now; β times δ to the power t later, with β below 1 | Drops once, then constant | Yes |
For illustration, suppose a member values a reward at its amount divided by (1 + 0.1 x days of delay). Now against a week out, $100 beats $120: $100 against $120 / 1.7, which is $70.60. Move both dates out. Eight weeks out, $100 is worth $15.15. Nine weeks out, $120 is worth $16.44. The member now waits for the bigger reward. The same one-week wait is refused when near and accepted when far. An exponential discounter picks the same side both times.
Is hyperbolic discounting real?
It is real, and it is stronger for rewards people can use right away. The Cohen review finds that money tasks show weak effects of adding a short front-end delay and tasks with real rewards show strong ones. In one test with juice for thirsty people, present bias was substantial when the juice was immediate and vanished when every option was shifted ten minutes out. A work-effort task gave a present-bias factor of 0.90, against 0.97 for money (Cohen et al.). Lower numbers mean more present bias.
Field data point the same way. DellaVigna and Malmendier studied 7,752 health club members. Those on a flat fee of over $70 a month went 4.3 times a month. They paid over $17 a visit, though a 10-visit pass cost $10 a visit (DellaVigna and Malmendier, 2006). Their leading explanations are overconfidence about future self-control or effort.
The evidence has limits. In one study of low- and middle-income people, 36% showed present bias and 9% showed the reverse (Cohen et al.). Wikipedia notes that many people learn to choose consistently. It adds that a 2014 paper faulted studies for leaning on students (Wikipedia).
How does hyperbolic discounting show up in loyalty programs?
Programs mix instant and deferred rewards. The 2026 Monopoly Game at McDonald’s, run through its app, has two tracks. Some game pieces win a food prize at once. The rules give overall odds of about 1 in 5.23 for such prizes. Others are Collect to Win pieces that must be combined into a winning set (McDonald’s official rules). The first track serves the present self and the second serves the saver.
A points ladder works the same way. Starbucks pays Green Level members 1 Star per dollar. Its lowest rung is 25 Stars for $1 off a beverage add-on, which is $25 of spend at the base rate. The 100-Star rung needs $100 and the 400-Star rung needs $400 (Starbucks Rewards Terms). The low rungs give a quick first reward. The high rungs give a goal worth saving for.
Present bias also affects redemption. Ted O’Donoghue and Matthew Rabin showed that people with this bias do immediate-reward activities too soon (O’Donoghue and Rabin, 1999). A reward that comes with an experience gets used early. Programs can expect quick burn on small rewards, and a reward that posts late acts as a delayed reward.
Paid plans show the commitment side. DellaVigna and Malmendier found that monthly-contract members were 17% more likely to stay past a year than members who signed for a year (DellaVigna and Malmendier). The authors link their results to overconfidence about future self-control. A program should price a fee on real use and make canceling easy.
How should a program design around it?
The research points to four design rules. Each is an application and not a tested loyalty result.
- Give an early first reward, so a new member sees a payoff in the first few visits.
- Keep a larger goal in view, so the member has something to save for.
- Post small rewards promptly, since a late reward acts as a delayed one.
- Offer a way to commit, such as a goal with a reward at the end, because people value ways to bind future choices.
The pairing is the point. Instant rewards keep today’s self engaged, and the deferred goal gives tomorrow’s self a reason to stay. A program with only the first pays out small rewards and loses the long game. A program with only the second asks members to wait for a payoff they discount.
How can a loyalty team test it?
Test timing against value, and keep the two rewards equal in dollar value so timing is the only change. For illustration, a program gives 10,000 members $1 off applied at once on the next visit, and gives another 10,000 points worth $1 that post after 7 days. If 31% of the first group return within 14 days against 27% of the second, timing alone added 4 points, or 400 visits per 10,000 members.
Then repeat the test with a consumable reward and with a cash-like one. The Cohen review suggests the gap will be larger for the consumable one. Also chart the days from earn to redemption for each group. Present-biased members redeem sooner. Read the results by segment, since the 36% and 9% shares show that members differ.
How GRAVTY handles hyperbolic discounting
Reward timing is a setting a team controls. The technology page lists average latency under 50 ms. Programs on GRAVTY run instant-win games, as our gamification guide says. Visual Rules score each action as it arrives. Our glossary entry on instant win has more.
Frequently asked questions
What is hyperbolic discounting in simple terms?
It is the tendency to value a reward much less as it moves away from the present, with the drop slowing at longer delays. The discount rate falls as the horizon increases (Cohen et al., 2020). That is why $100 today can beat $120 next week while $120 in nine weeks beats $100 in eight.
What is the difference between hyperbolic and exponential discounting?
Exponential discounting cuts value by the same share each day. Hyperbolic discounting cuts it less as the delay grows. Only the hyperbolic form lets choices flip as a reward nears (the review). Laibson called this inconsistency over time (Laibson, 1997).
What is quasi-hyperbolic discounting?
Quasi-hyperbolic discounting is a simple version. A reward gets full weight now and beta times delta to the power of the delay later, with beta below 1. Laibson used it in 1997 (Laibson, 1997). It keeps the fast first drop of hyperbolic discounting and stays easy to model (Cohen et al.).
What is an example of hyperbolic discounting?
Health club contracts are a field example. Members on a flat monthly fee of over $70 attended 4.3 times a month and paid more than $17 a visit, though a 10-visit pass cost $10 a visit (DellaVigna and Malmendier, 2006). People overestimated future use.
How does hyperbolic discounting affect loyalty programs?
It makes instant rewards pull harder than bigger rewards that sit far away. McDonald’s Monopoly pairs instant food prizes with Collect to Win prizes (McDonald’s official rules), and the Starbucks ladder starts at 25 Stars for a $1 reward (Starbucks Rewards Terms). Programs run both short and long reward paths.
