Observed saving behavior
Application to
the theory of consumption
Lectures in Behavioral economics
Spring 2016, Part 3
12.11.2015
G.B. Asheim, ECON4260, #3
1
Relevance of behavioral economics
Acc. to Vernon Smith (”Rational choice: The contrast between economics and psychology”, JPE 99, 1991, 877-89)
much beh. economic critique stems from the premises that
Two-sided auction: buyers announce bids, sellers announce asking prices.
1.
... rationality in the economy … derives from the rationContracts on acceptance. Some information but incomplete information
alitytoofeach.
theExperiments
individual decision
makers
in theconverges
economy.
available
show that such
a market
to price
and volume
represented
by theoretical
intersection intensive,
of supply and
demand
2.
... individual
rationality
is a cognitively
calculating
curves--competitive
equilibrium. "This
is some
kind of magic!" What it shows
process of maximization
of the
self-interest.
is that people are quite good at discovering prices and quantities that clear a
3.
... even
economic
theory
be tested
by testing
directly the
market
if they have
onlycan
incomplete
market
information.
economic rationality of the individuals isolated from
interactive experience in social and economic environments.
Instead, V. Smith thinks that beh. economic critique is important
if it predicts different economic behavior at an aggregate level.

12.11.2015
G.B. Asheim, ECON4260, #3
2

Present-biased pref. can explain these phenomena.

Illiquid assets combined with credit card debt
Illiquid assets are commitments.
Credit cards provide consumption smoothing,
but facilitate tempting shopping splurges.

Co-movement between income and consumption
Consequence of self-imposed liquidity constraints.
Becomes more binding at the time of retirement.

Laibson: A model that explains empirical evidence.
12.11.2015
3
G.B. Asheim, ECON4260, #3
Outline

Cons.-saving model w/illiquid asset
Laibson (1997).
Application of the multi-self approach, where
sophisticates use illiquid asset as commitment.
Yields explanation of observed saving behavior.

A theoretical result in the cons.-saving model
Laibson (1998).

Empirical test of the cons.-saving model
Angeletos et al. (2001).
Show that present-biased pref. fit data better.
12.11.2015
G.B. Asheim, ECON4260, #3
4
1
Laibson’s (1997) consumption-saving model
Consumer makes decisions in periods 1, , T

max  ln(c1 )   ln(c2 )   2 ln(c3 )
One liquid asset x
One illiquid asset z
Exogenous initial asset holdings x0 , z0  0
subject to c1 
In period t :  earn labor inco me yt
 earn asset inco me Rt ( xt 1  zt 1 )
 choose consumptio n ct and a new
c10 
asset allocation xt and zt such that
ct  xt  zt  yt  Rt ( xt 1  zt 1 )
ct  yt  Rt xt 1
zt 1 cannot be con - 
xt , zt  0
sumed at time t .

U  Et u (ct )    t 1 
t
T
 t
u (c )

subject to c1 
7
c2 c3
y
y
 2  y1  2  32  W
R R
R R
Solution :
c2 c3
y
y
 2  y1  2  32  W
R R
R R
c11 
W
RW
 (R ) 2 W
1
1
,
c

,
c

2
3
1     2
1     2
1     2
Numerical illustrati on (optimal period - 1)
0) :
Numerical illustrati on :
c110  $$422
369.30
00,, cc1220  $$334
365..46
31, c3130  $$331
361..11
66
  0.8,   0.9, R  1.1, and W  $1000.
G.B. Asheim, ECON4260, #3
G.B. Asheim, ECON4260, #3
max ln(c1 )   ln(c2 )   2 ln(c3 )
Instantane ous utility : u (c)  ln c
12.11.2015
12.11.2015
Optimal period-1 behavior
3-period consumption-saving example
Budget constraint : c1 
W
RW
(R ) 2 W
0
0
,
c

,
c

2
3
1    2
1    2
1    2
c10  $369.00, c20  $365.31, c30  $361.66
Laibson’s (1997) consumption-saving model (2)
Period - t intertemp oral preference s :
c2 c3
y
y
 2  y1  2  32  W
R R
R R
Numerical illustrati on :
5
G.B. Asheim, ECON4260, #3

Solution :
The illiquid asset
12.11.2015
Optimal period-0 behavior
6
12.11.2015
G.B. Asheim, ECON4260, #3
8
2
Naive period-2 behavior
Sophisticated period-1 behavior
max ln(c1 )   ln(c 2 (c1 ))   2 ln(c 3 (c1 ))
max ln(c2 )   ln(c3 )
subject to c2 
c3
 R (1   )W
 W2N  R (W  c11 ) 
R
1     2
  R 2 (W  c1 ) 
 R (W  c1 ) 

   2 ln 
max ln c1   ln 
1  
 1   


Solution :
c2N 
Solution :
W
RW
, c3N 
1  
1  
N
2
N
2
c1S  1 W  2 ,
R (1 )W
R R (1 )W
c2S  c 2 (c1 )  11  
, c3S  c 3 (c1 )  1
   1    2
1    2
Numerical illustrati on (naivete)
(optimal :period - 0) :
Numerical illustrati on (sophistic
(naivete)
(optimal :period
ation) :- 0) :
c  $$422
369.30
00,, cc  $$365
369..31
46,, cc $$361
292.66
.61
c110S  $$422
369
422..30
00
30,, cc2N202S  $$365
369..31
46,, cc303NS $$361
292.66
.61
10
1
N
0
22
12.11.2015
0N
33
G.B. Asheim, ECON4260, #3
9
Sophisticated period-2 behavior
c3
 W2  R (W  c1 )
R
(which yields $334.46 in period 2)
W2
R (W  c1 )

1  
1  
 save $273.64 in the illiquid asset
(which yields $331.11 in period 3)
RW2 R 2 (W  c1 )
c 3 (c1 ) 

1  
1  
12.11.2015
11
If   0.8,   0.9, R  1.1, Y1  1000, and Y2  Y3  0, then
In period 1 :  consume $422.30
 save $304.05 in the liquid asset
Solution :
c 2 (c1 ) 
G.B. Asheim, ECON4260, #3
Sophisticates can implement optimal period-1
behavior by using the illiquid asset
max ln(c2 )   ln(c3 )
subject to c2 
12.11.2015
G.B. Asheim, ECON4260, #3
10

Sophisticates strictly prefer the use of illiquid assets.

Naifs do not recognize the commitment value of
illiquid assets.
12.11.2015
G.B. Asheim, ECON4260, #3
12
3
Proof of a theoretical result in the cons.-saving model
The illiquid asset is not a perfect commitm. techn.
 cash on hand : xt  yt  R ( xt 1  ct 1 )
dU t 1
 choose consumptio n : ct  xt
dxt 1
dU t  2
dU t  2 u (ct 1 )


u (ct 1 )   R
dxt  2
dxt  2
 R
3-period consumption-saving example
  0.5,   1, R  1, and W  $1200.
You cannot prevent yourself
from consuming current income.

● If
u (ct )   R
Optimal period-1

Income
Consumption
Liquid asset
  0.8,   0.9, R  1.1, Y1  500, Y2  550,
 dc  dU t  2
dU t 1 dct 1

u (ct 1 )  R1  t 1 
dxt 1 dxt 1
 dxt 1  dxt  2
and Y3  0, then the illiquid asset does not help at all.
An illiquid asset does not work as a commitment
device if you can borrow against its future payoff.

 dc
 dc  1 
  t 1  1  t 1  u (ct 1 )
 dxt 1  dxt 1   
  dc 
 dc  
u (ct )  R  t 1    1  t 1  u (ct 1 )
 dxt 1  
  dxt 1 
● Credit cards (and other liquidity enhancing intruments)
may undermine the commitment value of illiquid assets.
12.11.2015
13
G.B. Asheim, ECON4260, #3
A theoretical result in the consum.-saving model

(Laibson, 1998)
Assume no illiquid asset
Period - t intertemp oral preference s :

U  Et u (ct )    t 1 
t
T
Optimality condition if   1 :
 t

u (c )

s
 cash on hand : xt  yt  R( xt 1  ct 1 )
dU t 1
 choose consumptio n : ct  xt
dxt 1
dU t  2
dU t  2 u (ct 1 )


u (ct 1 )   R
dxt  2
dxt  2
 R

 dc
 dc  1 
  t 1  1  t 1  u(ct 1 )
 dxt 1  dxt 1   
  dc 
 dc  
u (ct )  R  t 1    1  t 1  u(ct 1 )
 dxt 1  
  dxt 1 

 dc 
 dc  
u (ct )  Et R  t 1    1  t 1  u (ct 1 )
 dxt 1  
 dxt 1 
A source of ”under-saving”.
12.11.2015
G.B. Asheim, ECON4260, #3
14
15
Compare data with simulations Angeletos et al. (2001)
under exponential disc. and hyperbolic disc. (sophisticates)
Households with liquid
assets > 1 month’s income:

Exponent. simulation: 73%
Hyperbolic simulation: 40%
Data:
43%
Generalize d opt. condition for sophistica tes with   1 :


u (ct )   R
 dc  dU t  2
dU t 1 dct 1

u(ct 1 )  R1  t 1 
dxt 1 dxt 1
 dxt 1  dxt  2
G.B. Asheim, ECON4260, #3
Empirical test of the consumption-saving model
Proof of a theoretical result in the cons.-saving model
u (ct )  Et Ru (ct 1 )  Et R  u (ct  s )
s
12.11.2015

Households with positive
credit-card borrowing:
Exponent. simulation: 19%
Hyperbolic simulation: 51%
Data:
70%
12.11.2015
Mean credit-card borrowing
(all households):
Exponent. simulation: $900
Hyperbolic simulation: $3400
Data:
>$5000

Cons.-income comovement
(income coefficent):
Exponent. simulation: 0.032
Hyperbolic simulation: 0.166
Data:
 0.2
G.B. Asheim, ECON4260, #3
16
4
12.11.2015
G.B. Asheim, ECON4260, #3
17
12.11.2015
G.B. Asheim, ECON4260, #3
19
12.11.2015
G.B. Asheim, ECON4260, #3
18
12.11.2015
G.B. Asheim, ECON4260, #3
20
5
Additional references



Choi, J., D. Laibson, B. Madrian & A. Metrick (2002),
Defined Contribution Pensions: Plan Rules, Participant
Decisions, and the Path of Least Resistance, Tax Policy
and the Economy 16, 67–114
DellaVigna, S. & U. Malmendier (2006), Paying Not to
Go to the Gym, American Economic Review 96, 694–719
Shapiro, J.M. (2005), Is there a daily discount rate?
Evidence from the food stamp nutrition cycle, Journal of
Public Economics 89, 303–325
12 November 2015
G.B. Asheim, ECON4260, #4
21
6