5D Library › Choice Architecture, Cumulative Prospect Theory and Computational Behavioural Economics of Pricing Interfaces · 中文

Chapter 6 Time: Quasi-Hyperbolic Discounting, Procrastination and the Friction of Forms

Why a few minutes’ work is postponed again and again

Someone wants to subscribe to a service, claim a refund, take up a benefit or cancel a membership she no longer needs. It would take a few minutes and the benefit is clear, yet she keeps putting it off until the offer expires, the refund window closes or another month is charged. The author’s framework asks to compare Laibson’s quasi-hyperbolic (β\beta–δ\delta) discounting with exponential discounting and to compute the friction cost of multi-step forms and the “procrastination–default” threshold. This chapter first checks the source and properties of the β\beta–δ\delta model, then follows O’Donoghue and Rabin in distinguishing naive from sophisticated procrastination, and finally gives an author’s model of multi-step forms.

Three discount functions

Exponential discounting D(t)=δtD(t) = \delta^t applies the same discount to each extra day of waiting, so today’s preference between “next Monday or next Tuesday” is the same as on Monday itself (time consistency). Empirically, people are often willing to wait when the choice is distant and unwilling when it is close. Hyperbolic discounting D(t)=1/(1+kt)D(t) = 1/(1 + kt) is one function describing this; quasi-hyperbolic discounting is a convenient approximation to it.

Evidence

The quasi-hyperbolic discount function D(0)=1D(0) = 1, D(t)=βδtD(t) = \beta\delta^t (t≥1t \ge 1, 0<β<10 < \beta < 1) was first proposed by Phelps and Pollak (1968) to describe how each generation weighs its own consumption against that of later generations, in “game-equilibrium growth” between generations (Phelps and Pollak 1968). Laibson (1997) applied the same function to one person’s “selves” at different dates, analysing consumption, saving and how illiquid assets (“golden eggs”) act as commitment devices (Laibson 1997). Frederick, Loewenstein and O’Donoghue’s 2002 review discusses the wide spread of estimated discount rates and the evidence for different discounting models (Frederick et al. 2002). Augenblick, Niederle and Sprenger (2015) found clear present bias in real-effort tasks and almost none in intertemporal choices over money (Augenblick et al. 2015).

The framework’s “Laibson quasi-hyperbolic discounting (β–δ)” should therefore note that the functional form originates with Phelps and Pollak. Augenblick and co-authors’ result matters especially here: filling in a form is an effort, not an intertemporal choice over money, and present bias is more pronounced in exactly this kind of task.

Calculation

# Exponential versus quasi-hyperbolic (beta-delta) discounting. Phelps & Pollak (1968)
# introduced the beta-delta form for successive generations; Laibson (1997) applied it to
# one person's successive selves: D(0) = 1, D(t) = beta * delta^t for t >= 1.
# Exponential: D(t) = delta^t. Hyperbola (Mazur form) for comparison: D(t) = 1/(1 + k t).
# Time unit: one day. Illustrative parameters: beta 0.7, delta 0.9995, k 0.05.
BETA, DELTA, K = 0.7, 0.9995, 0.05

def d_exp(t):  return DELTA ** t
def d_bd(t):   return 1.0 if t == 0 else BETA * DELTA ** t
def d_hyp(t):  return 1 / (1 + K * t)

F = (("exponential", d_exp), ("beta-delta", d_bd), ("hyperbolic", d_hyp))
print("discount factor     t = 0     1     7    30   365")
for lab, f in F:
    print(f"   {lab:14s} " + "".join(f"{f(t):6.3f}" for t in (0, 1, 7, 30, 365)))
print("per-day rate 1 - D(t+1)/D(t)   t = 0       1      30     365")
for lab, f in F:
    rates = [1 - f(t + 1) / f(t) for t in (0, 1, 30, 365)]
    print(f"   {lab:26s} " + "".join(f"{r:8.4f}" for r in rates))
print("100 on day s or 120 on day s + 7, as judged today (day 0):")
for s in (0, 1, 7, 30, 90):
    pick = ["120 later" if 120 * f(s + 7) > 100 * f(s) else "100 sooner" for _, f in F]
    print(f"   s = {s:2d}: exponential {pick[0]:10s} beta-delta {pick[1]:10s}"
          f" hyperbolic {pick[2]}")
print("beta-delta: 100 now iff 1.2 beta delta^7 < 1"
      f" (beta < {1 / (1.2 * DELTA ** 7):.4f});"
      " for s >= 1 beta cancels")
s_star = next(s for s in range(1000) if 120 * d_hyp(s + 7) > 100 * d_hyp(s))
print(f"hyperbolic: sooner for s < {s_star}, later from s = {s_star}"
      f" (later iff s > 35 - 1/k = {35 - 1 / K:.0f})")
# discount factor     t = 0     1     7    30   365
#    exponential     1.000 1.000 0.997 0.985 0.833
#    beta-delta      1.000 0.700 0.698 0.690 0.583
#    hyperbolic      1.000 0.952 0.741 0.400 0.052
# per-day rate 1 - D(t+1)/D(t)   t = 0       1      30     365
#    exponential                  0.0005  0.0005  0.0005  0.0005
#    beta-delta                   0.3004  0.0005  0.0005  0.0005
#    hyperbolic                   0.0476  0.0455  0.0196  0.0026
# 100 on day s or 120 on day s + 7, as judged today (day 0):
#    s =  0: exponential 120 later  beta-delta 100 sooner hyperbolic 100 sooner
#    s =  1: exponential 120 later  beta-delta 120 later  hyperbolic 100 sooner
#    s =  7: exponential 120 later  beta-delta 120 later  hyperbolic 100 sooner
#    s = 30: exponential 120 later  beta-delta 120 later  hyperbolic 120 later
#    s = 90: exponential 120 later  beta-delta 120 later  hyperbolic 120 later
# beta-delta: 100 now iff 1.2 beta delta^7 < 1 (beta < 0.8363); for s >= 1 beta cancels
# hyperbolic: sooner for s < 16, later from s = 16 (later iff s > 35 - 1/k = 15)

Illustrative parameters are β=0.7\beta = 0.7, δ=0.9995\delta = 0.9995 per day and k=0.05k = 0.05. The quasi-hyperbolic function applies a 30 % discount once between day 0 and day 1 and only 0.05 % per day thereafter, like the exponential; the hyperbolic daily rate falls gradually from 4.76 % to 0.26 %. In the choice “100 on day ss or 120 on day s+7s+7”, the exponential discounter always takes 120; the quasi-hyperbolic discounter takes 100 only when s=0s = 0 (100 is available today), provided 1.2βδ7<11.2\beta\delta^7 < 1, that is β<0.8363\beta < 0.8363; the hyperbolic discounter takes 100 for all s<16s < 16, the reversal being set by s>35−1/k=15s > 35 – 1/k = 15. The quasi-hyperbolic function separates “now” from “later” sharply, the hyperbolic lets preferences change continuously with distance; both produce preference reversals, the exponential does not. The left panel of Figure 5 shows the three curves.

Left: exponential, quasi-hyperbolic and hyperbolic discount functions (\beta = 0.7, \delta = 0.9995, k = 0.05, illustrative). Right: in the author’s multi-step form model, the probability that each type completes the form within the 14-day window, against the number of steps.

Naive and sophisticated

Evidence

O’Donoghue and Rabin (1999) analysed when present-biased people choose to do things. A naive person believes her future selves will not be present-biased and will follow today’s plan; a sophisticated person knows her future selves are just as present-biased and reasons backwards. For tasks whose costs come now and benefits later (writing a report, filling in a form), a naive person may procrastinate indefinitely; a sophisticated person, foreseeing her own delay, sometimes acts earlier, but may also delay comfortably because she foresees that a future self will act. For activities whose rewards come now and costs later (indulgences), the directions reverse (O’Donoghue and Rabin 1999). Ariely and Wertenbroch (2002) found that students voluntarily set themselves costly deadlines, but that self-imposed deadlines were less effective than evenly spaced external ones (Ariely and Wertenbroch 2002).

The author’s model: multi-step forms and deadlines

What follows is an author’s model showing how the “friction of multi-step forms” interacts with present bias. It borrows O’Donoghue and Rabin’s naive/sophisticated distinction, but the specific set-up is this book’s, not a published result.

An online form has mm steps, each costing effort ee; completing it costs C=meC = me on the day it is done. The benefit VV arrives later, so on that day it is multiplied by β\beta (δ≈1\delta \approx 1 within two weeks). The form closes after day TT; each day of delay, the person “drops out” with probability qq (forgets, closes the tab, is absorbed by something else). The self on day tt compares “do it today”, −C+βV-C + \beta V, with “wait for the next self who will do it”, β(1−q)k(V−C)\beta(1-q)^k(V – C), where kk is how many days she expects that to be. The exponential discounter (β=1\beta = 1) forecasts correctly; the naive person thinks tomorrow’s self will certainly act (k=1k = 1); the sophisticated person reasons back from day TT and knows on which day the next self who really acts will be. Comparing only “today” with “tomorrow”, today is better if

C≤βqV1−β(1−q).C \le \frac{\beta q V}{1 – \beta(1-q)}.

Calculation

# THE AUTHOR'S MODEL of a multi-step form, in the setting of O'Donoghue & Rabin (1999).
# Completing it costs C = m * e on the day it is done (m steps, effort e each); the
# benefit V comes later, so on that day it carries the factor beta (delta = 1 within the
# window). The offer closes after day T; each day a person who has not acted drops out
# (forgets, loses the tab) with probability q. Self t acts iff acting beats her forecast:
#   act now: -C + beta V;  if the next self to act does so k days later:
#   wait:    beta (1 - q)^k (V - C).
# Exponential (beta = 1) forecasts correctly; a naive person forecasts that tomorrow's
# self will act (k = 1) whenever C <= V; a sophisticated person solves backwards.
V, E, BETA, Q, T = 10.0, 0.5, 0.7, 0.05, 14

def plan(C, beta, kind):
    """Return the first day on which the form is done (None = never)."""
    act = [False] * (T + 1)
    act[T] = -C + beta * V >= 0
    for t in range(T - 1, -1, -1):
        if kind == "naive":                       # believes she will do it tomorrow
            k = 1 if C <= V else None
        else:                                     # knows the next self who acts
            k = next((j - t for j in range(t + 1, T + 1) if act[j]), None)
        wait = beta * (1 - Q) ** k * (V - C) if k else 0.0
        act[t] = -C + beta * V >= max(wait, 0.0) and -C + beta * V >= 0
    return next((t for t in range(T + 1) if act[t]), None)

def done(m, kind):
    day = plan(m * E, 1.0 if kind == "exponential" else BETA, kind)
    return day, (0.0 if day is None else (1 - Q) ** day)

c_s = BETA * Q * V / (1 - BETA * (1 - Q))
print(f"V = {V}, e = {E} per step, beta = {BETA}, q = {Q} per day, window T = {T} days")
print("acting today beats acting tomorrow iff C <= beta q V / (1 - beta (1 - q))"
      f" = {c_s:.3f}")
print("   steps   cost    exponential    naive          sophisticated")
print("                   (first day on which the form is done, P(done))")
for m in (1, 2, 3, 4, 6, 10, 13, 14, 21):
    cells = []
    for kind in ("exponential", "naive", "sophisticated"):
        day, p = done(m, kind)
        cells.append(f"{'never' if day is None else 'day ' + str(day):>6s} {p:5.3f}")
    print(f"   {m:5d}  {m * E:5.1f}    " + "    ".join(cells))
print("sophisticated, first day by steps 1..15: "
      + " ".join("-" if d is None else str(d) for d in
                 (done(m, "sophisticated")[0] for m in range(1, 16))))
print("enrolment after the window, m = 6 steps, 30 % of people would rather not join:")
for kind in ("exponential", "naive", "sophisticated"):
    p = done(6, kind)[1]
    print(f"   {kind:13s}: opt-in form {0.7 * p:.3f};"
          f"  opt-out form {0.7 + 0.3 * (1 - p):.3f}"
          f" (target 0.700)")
# V = 10.0, e = 0.5 per step, beta = 0.7, q = 0.05 per day, window T = 14 days
# acting today beats acting tomorrow iff C <= beta q V / (1 - beta (1 - q)) = 1.045
#    steps   cost    exponential    naive          sophisticated
#                    (first day on which the form is done, P(done))
#        1    0.5     day 0 1.000     day 0 1.000     day 0 1.000
#        2    1.0     day 0 1.000     day 0 1.000     day 0 1.000
#        3    1.5     day 0 1.000    day 14 0.488     day 0 1.000
#        4    2.0     day 0 1.000    day 14 0.488     day 2 0.902
#        6    3.0     day 0 1.000    day 14 0.488     day 2 0.902
#       10    5.0     day 0 1.000    day 14 0.488     day 3 0.857
#       13    6.5     day 0 1.000    day 14 0.488    day 14 0.488
#       14    7.0     day 0 1.000    day 14 0.488    day 14 0.488
#       21   10.5     never 0.000     never 0.000     never 0.000
# sophisticated, first day by steps 1..15: 0 0 0 2 2 2 2 0 5 3 14 14 14 14 -
# enrolment after the window, m = 6 steps, 30 % of people would rather not join:
#    exponential  : opt-in form 0.700;  opt-out form 0.700 (target 0.700)
#    naive        : opt-in form 0.341;  opt-out form 0.854 (target 0.700)
#    sophisticated: opt-in form 0.632;  opt-out form 0.729 (target 0.700)

Parameters are V=10V = 10, e=0.5e = 0.5, β=0.7\beta = 0.7, q=0.05q = 0.05, T=14T = 14 (illustrative). The threshold for today beating tomorrow is C≤1.045C \le 1.045, at most 2 steps. The results:

  • The exponential discounter completes on day 0 whenever C≤VC \le V (up to 20 steps).
  • The naive person completes on the day with 1 or 2 steps; with 3 or more she thinks every day that she will do it tomorrow and delays to the last day, by which time she is still “around” with probability only 0.9514=0.4880.95^{14} = 0.488. Beyond 14 steps (C>βV=7C > \beta V = 7) she does not act even on the last day.
  • The sophisticated person still completes on the day with 3 steps (she knows that if she does not, the next self to act is far away); with 4 to 7 steps she completes on day 2, with probability 0.902; with 11 to 14 steps she, like the naive person, waits for the last day. Between 8 and 10 steps her day of action is not monotone (days 0, 5 and 3): the “next self who acts” found by backward induction jumps with the cost, a known feature of O’Donoghue and Rabin’s sophisticated model rather than a program error.

There is no single “procrastination–default threshold” here but two: when the number of steps crosses the today/tomorrow threshold (2 steps here), the naive completion rate falls from 1 to 0.488; when cost crosses βV\beta V, every present-biased person gives up. The sophisticated completion rate falls overall as the number of steps grows, but not monotonically, and lies broadly between the two (Figure 5, right).

The last lines apply the model to defaults. Suppose 70 % of people really want to join and 30 % do not. If joining requires a 6-step form (opt-in), the share who end up enrolled is 0.700 for exponential discounters, 0.341 for naive and 0.632 for sophisticated people; if enrolment is the default and leaving requires the same form (opt-out), it is 0.700, 0.854 and 0.729. For exponential discounters the default changes nothing; for naive people it changes the outcome for about half (0.341 against 0.854, a gap of 51.3 points); for sophisticated people for about a tenth (0.632 against 0.729, 9.7 points). This is one mechanism behind default effects, consistent with Madrian and Shea’s (2001) finding that automatic enrolment substantially raised 401(k) participation and that many employees stayed at the default contribution rate and investment option (Madrian and Shea 2001); they attribute this to inertia and to treating the default as advice, and the model captures only the procrastination channel.

Evidence

In a field experiment with the US Internal Revenue Service, Bhargava and Manoli (2015) sent eligible non-claimants of the Earned Income Tax Credit different versions of reminder notices; simplification and emphasising the benefit raised take-up, and they attribute part of the incomplete take-up to lack of information, confusion and “psychological frictions” (Bhargava and Manoli 2015). The meta-analysis by Jachimowicz and co-authors mentioned in Chapter 1 finds an average default effect of d=0.68d = 0.68 (Jachimowicz et al. 2019).

Note

The ethics of this section. The same model can be used to lower the barrier to something good for the user, or to raise the barrier to cancelling a subscription or claiming a refund. The latter is the kind of “dark pattern” some regulators target. The Chapter 8 checklist flags “asymmetric steps to join and to leave” and “pre-ticked add-ons”; this book offers no method for using procrastination to stop users from leaving.

Hypothesis · not established · The step threshold matters more for the naive than for the sophisticated

The author’s hypothesis (not established): for an online application with a deadline, cutting the form from 6 steps to 2 raises completion by more among people who describe themselves afterwards as “often thinking I’ll do it tomorrow and then not doing it” than among those who describe themselves as “knowing I’ll delay and so planning ahead”. What would show it wrong: in a pre-registered experiment randomising 2-step and 6-step forms and measuring self-reported procrastination type in advance, the confidence interval for the interaction between number of steps and procrastination type includes zero or has the opposite sign.

Reminders, deadlines and a counter-intuitive property of the model

The author’s model has another property worth stating. The today/tomorrow threshold βqV/(1−β(1−q))\beta qV/(1 – \beta(1-q)) is the naive person’s decision rule (she always believes the next self to act is tomorrow’s, k=1k = 1); its derivative in qq is βV(1−β)/(1−β+βq)2>0\beta V(1-\beta)/(1-\beta+\beta q)^2 > 0: the smaller the drop-out risk qq, the lower the threshold and the more readily the naive person delays. The sophisticated person’s kk comes from backward induction and she uses this threshold only when the next self to act is tomorrow’s; in general a smaller qq also makes waiting cheaper, (1−q)k(1-q)^k, pushing in the same direction by an amount that depends on kk. So a measure that lowers the risk of “forgetting” (a reliable reminder service, for instance) may in the model reduce the number who complete on the day while raising the share who eventually complete — for the naive person, completion on the last day, (1−q)T(1-q)^T, rises as qq falls. Judging a reminder by “same-day conversion” alone may give the opposite verdict from “eventual completion”. The inference comes only from the author’s model and needs testing.

Ariely and Wertenbroch supply evidence on the deadline side: in their classroom and proofreading studies, students were willing to set themselves binding deadlines, and such self-imposed deadlines improved performance, but less than evenly spaced deadlines imposed from outside (Ariely and Wertenbroch 2002). In this chapter’s model an intermediate deadline cuts one long window into several short ones, and the last day of each forces the naive person to act once; this suggests why the number and spacing of deadlines may matter more than having a deadline at all. The correspondence is an analogy; the model has no learning or fatigue.

Chapter summary

The quasi-hyperbolic discount function originates in Phelps and Pollak’s intergenerational model; Laibson applied it to one person’s dates; it and the hyperbolic function both produce preference reversals, the exponential does not. Present bias is clearer in effort tasks than in choices over money. In the author’s form model the threshold for today beating tomorrow is 2 steps; with 3 to 14 steps the naive person waits for the last day and completes with probability 0.488, while the sophisticated person completes on day 2 with probability 0.902 for 4 to 7 steps, her day of action becoming non-monotone beyond that; when cost exceeds βV\beta V both give up. Defaults make no difference for exponential discounters, change the outcome for about half of naive people and for about a tenth of sophisticated ones. The framework’s “procrastination–default threshold” is, in this model, two thresholds, and it is the author’s model, not a published result.

References

Ariely, Dan, and Klaus Wertenbroch. 2002. “Procrastination, Deadlines, and Performance: Self-Control by Precommitment.” Psychological Science 13 (3): 219–24. https://doi.org/10.1111/1467-9280.00441.
Augenblick, Ned, Muriel Niederle, and Charles Sprenger. 2015. “Working over Time: Dynamic Inconsistency in Real Effort Tasks.” Quarterly Journal of Economics 130 (3): 1067–115.
Bhargava, Saurabh, and Dayanand Manoli. 2015. “Psychological Frictions and the Incomplete Take-up of Social Benefits: Evidence from an IRS Field Experiment.” American Economic Review 105 (11): 3489–529.
Frederick, Shane, George Loewenstein, and Ted O’Donoghue. 2002. “Time Discounting and Time Preference: A Critical Review.” Journal of Economic Literature 40 (2): 351–401.
Jachimowicz, Jon M., Shannon Duncan, Elke U. Weber, and Eric J. Johnson. 2019. “When and Why Defaults Influence Decisions: A Meta-Analysis of Default Effects.” Behavioural Public Policy 3 (2): 159–86.
Laibson, David. 1997. “Golden Eggs and Hyperbolic Discounting.” Quarterly Journal of Economics 112 (2): 443–77. https://doi.org/10.1162/003355397555253.
Madrian, Brigitte C., and Dennis F. Shea. 2001. “The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior.” Quarterly Journal of Economics 116 (4): 1149–87.
O’Donoghue, Ted, and Matthew Rabin. 1999. “Doing It Now or Later.” American Economic Review 89 (1): 103–24. https://doi.org/10.1257/aer.89.1.103.
Phelps, Edmund S., and Robert A. Pollak. 1968. “On Second-Best National Saving and Game-Equilibrium Growth.” Review of Economic Studies 35 (2): 185–99.

Fan, Y. (2026). Choice Architecture, Cumulative Prospect Theory and Computational Behavioural Economics of Pricing Interfaces: From Loss-Aversion Value Functions and Asymmetric Dominance to Hyperbolic Discounting, Read Beside the Qingzhong Chapters of the Guanzi [选择架构、累积前景理论与销售界面价格锚定的计算行为学] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836260 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).