Game Theory Slides Chapter 6x6timeinconsistentpreferences

Time Inconsistent Preferences
Felix Munoz-Garcia
Advanced Microeconomics II - Washington State University

Time Inconsistent Preferences - Tadelis 8.3.4
Previously, we assumed that players maximize their discounted

sum of payo¤s.
This is typically done using exponential discounting, where
every future period is multiplied by δt .
We can show that using exponential discounting, an
individual’s consumption choices are consistent across time
periods.
Consider a player with u (x ) = ln(x ) who needs to allocate a

…xed budget K across three periods. Assume that prices all
equal 1. His maximization problem is
max ln(K x2 x3 ) + δ ln(x2 ) + δ2 ln(x3 )

x2 ,x3 | {z }
x1
with …rst-order conditions

1 δ
+ = 0
K x2 x3 x2
1 δ2
+ = 0
K x2 x3 x3
Solving, we …nd
1
x1 = K
1 + δ + δ2
δ
x2 = K
1 + δ + δ2
δ2
x2 = K
1 + δ + δ2
But what if the player made his choices sequentially? Would
he reach the same equilibrium? Let’s try backward induction.
In period 2, the player solves
max ln(x2 ) + δ ln(K2 x2 )

x2 | {z }
x3
where
1 δ + δ2
K2 = K x1 = K K = K
1 + δ + δ2 1 + δ + δ2
Solving, we …nd
K2 δ
x2 = =K
1+δ 1 + δ + δ2
Thus, regardless of his previous actions, the player will choose

the same consumption amounts sequentially as he will
simultaneously. Thus, this player has time consistent
preferences.
This is a useful property of exponential discounting.
What about other types of discounting, namely hyperbolic
discounting.
In this case, a player uses the discount rate δ as seen in
exponential discounting, but he uses an additional discount
factor β 2 (0, 1) to discount all of the future compared to
present consumption. Our previous example would look as
follows under hyperbolic discounting:
max ln(K x2 x3 ) + βδ ln(x2 ) + βδ2 ln(x3 )

x2 ,x3 | {z }
x1
Intuitively, if the player is looking toward the future, the

discount factor he uses between periods t = 1 and t = 2 is
stronger than the one he uses between periods t = 2 and
t = 3.
Hyperbolic discounting will cause problems with self-control,
in that a player will plan to do one thing but later choose to
revise his plan.
This is studied frequently in behavior economics. In laboratory
setting, it is frequently found that individuals do not consume
"rationally." Robby Rosenman’s module in Applied
Microeconomics touches on this very topic, as well as other
similar ones.
Let δ = 1 for simplicity. We know that a rational person using
exponential discounting will equalize his consumption across
the three periods, i.e., xi = K3 i = 1, 2, 3. Let’s look at the
case where a player uses hyperbolic discounting, and for
simplicity, we’ll assume that β = 21 .
Our simultaneous problem becomes
1 1
max ln(K x2 x3 ) + ln(x2 ) + ln(x3 )
x2 ,x3 | {z } 2 2
x1
with solution
1
x1 = K
2
1
x2 = K
4
1
x3 = K
4
Checking to see if the player’s sequential solution is the same,

we use backward induction. Our period 2 problem becomes
1
max ln(x2 ) + ln(K x1 x2 )
x2 2 | {z }
x3
This yields best-response functions of

2
x2 (x1 ) = (K x1 )
3
1
x3 (x1 ) = (K x1 )
3
We can substitute these best-response functions back into our

period 1 problem to obtain
1 2 1 1
max ln(x1 ) + ln (K x1 ) + ln (K x1 )
x1 2 3 2 3
with solution
1
x1 = K
2
1
x2 = K
3
1
x3 = K
6
Note: This di¤ers from Tadelis’solution. He has an error in his

FOCs. (This is why we check them!)
This results in a very di¤erent solution than our simultaneous

benchmark, x1 = 21 K , x2 = 14 K , x3 = 14 K .
In this situation, the player consumes the same amount as he
would normally in period 1, but in period 2, he becomes
impatient once again, and overconsumes what he would have
wanted to in period 1. We refer to this as the player having
time inconsistent preferences.
Comparing time consistent with time inconsistent preferences

in the simultaneous game (δ = 1):
Time Inconsistent Preferences, β = ½

½
Time Consistent Preferences, β = 1

⅓
¼
Period 1 Period 2 Period 3

Comparing time consistent with time inconsistent preferences

in the sequential game (δ = 1):
Time Inconsistent Preferences, β = ½

½
Time Consistent Preferences, β = 1

⅓
Period 1 Period 2 Period 3

Game Theory Slides Chapter 6x6timeinconsistentpreferences

Uploaded by

Copyright:

Available Formats

Game Theory Slides Chapter 6x6timeinconsistentpreferences

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Game Theory Slides Chapter 6x6timeinconsistentpreferences

Uploaded by

Copyright:

Available Formats

Time Inconsistent Preferences

Advanced Microeconomics II - Washington State University

Previously, we assumed that players maximize their discounted

Consider a player with u (x ) = ln(x ) who needs to allocate a

max ln(K x2 x3 ) + δ ln(x2 ) + δ2 ln(x3 )

with …rst-order conditions

In period 2, the player solves

max ln(x2 ) + δ ln(K2 x2 )

Thus, regardless of his previous actions, the player will choose

max ln(K x2 x3 ) + βδ ln(x2 ) + βδ2 ln(x3 )

Intuitively, if the player is looking toward the future, the

Checking to see if the player’s sequential solution is the same,

This yields best-response functions of

We can substitute these best-response functions back into our

Note: This di¤ers from Tadelis’solution. He has an error in his

This results in a very di¤erent solution than our simultaneous

Comparing time consistent with time inconsistent preferences

Time Inconsistent Preferences, β = ½

Time Consistent Preferences, β = 1

Period 1 Period 2 Period 3

Comparing time consistent with time inconsistent preferences

Time Inconsistent Preferences, β = ½

Time Consistent Preferences, β = 1

Period 1 Period 2 Period 3

You might also like