Reverse Preference Reversals for Short FrontEnd Delays

1
Reverse Preference Reversals for Short Front-End
Delays

• Ayse Öncüler
• INSEAD, France
• (joint work with Serdar Sayman, Koc U., Turkey)

2
Time Inconsistency

• Discounted Utility Model (Koopmans, 1960)
• Stationarity in Discounting
• If (x, t)(y, t), then (x, tj)(y, tj)
• A set of experimental evidence on (regular) time
  inconsistency, starting with Thaler (1981),
  Benzion et al (1989) Individuals prefer the
  later-larger reward (LL) at the beginning but
  closer to the date, switch to the sooner-smaller
  one (SS).

3
Time Inconsistency

• (Regular) Time Inconsistency

Prefer LL
Prefer SS
4
Time Inconsistency

• Discounted Utility Model (Koopmans, 1960)
• Stationarity in Discounting
• If (x, t)(y, t), then (x, tj)(y, tj)
• A set of experimental evidence on (regular) time
  inconsistency, starting with Thaler (1981),
  Benzion et al (1989) Individuals prefer the
  later-larger reward (LL) at the beginning but
  closer to the date, switch to the sooner-smaller
  one (SS).
• Current study provides evidence on reverse time
  inconsistency Individuals aim for the
  sooner-smaller reward and over time, they switch
  to the later-larger one
• Experimental evidence and a possible explanation

5
Time Inconsistency

• (Regular) Time Inconsistency

• (Reverse) Time Inconsistency

present value
present value
LL
SS
SS
LL
time
SS
LL
time
SS
LL
Example Vacation plans of workaholics
Example Dieting, Christmas clubs
6
Experimental Design
Hypothetical
Real Payoffs
Loyalty Programs
Study 1
Study 3
Monetary Payoffs
Study 4
Study 2
7
Study 1 Hypothetical Loyalty Program

• Aim To check for regular/reverse time
  inconsistency
• INSEAD, Singapore (n97) and Koc U., Turkey
  (n100)
• Hypothetical scenarios on retailer loyalty
  programs
• Local supermarket offers a 10 check if the buyer
  makes 100 purchases. Reward is 25 when the
  cumulative purchase is 200
• Subjects favorite movie theater offers a free
  ticket after 3 movie attendances or 2 free
  tickets after 5 attendances
• 2 conditions beginning of program, midway
  through

8
Study 1 Results - INSEAD
Movie
Supermarket
RTI
TI
RTI
TI
Beginning A 10 after 100 B 25 after
200 Midway A 10 today (100
completed) B 25 after 100 more (200 total)
Beginning A 1 ticket after 3 B 2 tickets
after 5 Midway A 1 ticket now (3
completed) B 2 tickets after 2 more (5 total)
9
Study 1 Results Koc Uni.
Movie
Supermarket
RTI
TI
RTI
Beginning A 10 after 100 B 25 after
200 Midway A 10 today (100
completed) B 25 after 100 more (200 total)
Beginning A 1 ticket after 3 B 2 tickets
after 5 Midway A 1 ticket now (3
completed) B 2 tickets after 2 more (5 total)
10
Study 1- Conclusion

• Overall, no significant time inconsistency in
  either direction
• Some women exhibited regular TI- becoming more
  impatient
• for imminent reward
• Men exhibited reverse TI- deferring an imminent
  reward
• Possible pitfall Effort required for payoff
• Next Choice task between two monetary payoffs

11
Study 2 Hypothetical Monetary Outcomes

• Aim to check time inconsistency for choice
  task between two monetary outcomes to be received
  at different time periods
• 2 pairs of outcomes (7-10 and 20-25)
• 4 pairs of time periods

12
Study 2 - Results
(Koc U., n1 70, n2 60, btw ss)
preferring to wait
RTI
RTI
RTI
RTI
13
Study 2 - Results cont.
(Koc U., n1 70, n2 60, btw ss)
preferring to wait
RTI
RTI
RTI
RTI
14
Study 2- Conclusion

• No gender difference
• We observe RTI in almost all cases
• subjects preferring to wait decreases with
  delay
• Replicated at INSEAD (n67)
• Possible pitfall Not longitudinal
• Next Real payoffs

15
Study 3 Real Payoffs, Loyalty Program

• Aim To check for regular/reverse time
  inconsistency in an incentive-compatible design
• INSEAD, Fontainebleau (n47)
• During the experiment, the campus café offered a
  two-level reward program students and staff
  could receive 1 croissant after buying 10
  croissants, or could receive 2 croissants after
  15 croissants. Customers are given a patronage
  card which bears an ID number. This card is shown
  when they make purchases, and the purchases are
  recorded on the card by means of a stamp.
• Intended choices are elicited at the beginning
• Purchase and promotion redemption is recorded for
  6 weeks

16
Study 3 Results (n47, within-ss)

• preferring to wait (B)
• Male Female TOTAL
• A 1 free crois. after 10
• B 2 free crois. after 15 70 25 51
• A 1 free crois after 10 purchased
• B 2 free crois. after 5 more 56 95 72

RTI
RTI
TI
17
Study 3- Conclusion

• At the beginning, most men reported preferring
  later/larger reward but some switched to
  sooner/smaller reward during the experiment
  (Regular Time Inconsistency)
• At the beginning, most women reported preferring
  sooner/smaller reward but they switched to
  later/larger reward during the experiment
  (Reverse Time Inconsistency)
• Overall, there seems to be some evidence on RTI
  and TI
• 40 RTI
• 19 TI
• 41 consistent with earlier preference

18
Study 4 Real Payoffs, Monetary Outcomes

• Replication of Study 2 with real payoffs, n38,
  France
• 2 pairs of outcomes, 3 pairs of time periods

19
Study 4 - Results
(INSEAD, n38, within ss)
preferring to wait
RTI
RTI
RTI
20
Study 4 Results cont.
(INSEAD, n38, within ss)
preferring to wait
RTI
TI
21
Study 4 - Results

• for 7 vs. 10 both females and males exhibit
  RTI
• for 20 vs. 25 for small time periods RTI,
• for larger periods females exhibit TI

22
Main Result
Evidence for reverse time inconsistency For
loyalty programs Men exhibit RTI for
hypothetical rewards women exhibit RTI for real
rewards For neutral payoffs It depends on the
relative size of SS and LL and the time interval
between them Main conclusion Intertemporal
preferences are richer than previously recognized
23
Discussion

• What can explain reverse time-inconsistency?
• Utility from delayed consumption (savoring,
  Loewenstein 1987)
• Effort Uncertainty (sunk cost of time, Soman
  2003)
• Affect
• Form of discounting (elicitation study shows some
  subjects exhibit a modified discounting
  behaviour)

24
(No Transcript)
25
Conclusion

• An explicit identification of reverse
  time-inconsistency
• Also observed in some recent experiments
• Sholten and Read (2005), Airoldi et al (2005),
  Attema et al (2006)
• Implications in many domains Loyalty programs,
  bank loan schedules, etc