=
1
e

1
1+
_
Hashmi (National U of Singapore) RCK Model August 22, 2014 6 / 48
Assumptions
The instantaneous utility function is
u[C(t)] =
C(t)
1
1
1
, > 0 and n (1 )g > 0. (2.2)
This utility function is called CRRA (constant relative risk aversion)
Risk aversion is related to the curvature of the utility function
ArrowPratt coecient of Absolute Risk Aversion: u
//
/u
/
Coecient of Relative Risk Aversion:
cu
//
/u
/
= c () c
1
/c
=
Hashmi (National U of Singapore) RCK Model August 22, 2014 7 / 48
Assumptions
is the inverse of the intertemporal elasticity of substitution
If is small (approaching zero), large swings in consumption are
acceptable to the consumers
If 1, u = ln C
Proof:
lim
1
C
1
1
1
= lim
1
C
1
1
(1)
= lim
1
C
1
ln C
1
= ln C. Q.E.D.
The rst equality follows from LHopitals rule.
n (1 )g > 0 is needed to ensure that lifetime (which is
innite here) utility does not diverge
Hashmi (National U of Singapore) RCK Model August 22, 2014 8 / 48
Behavior of the rm
Firms problem is static so we suppress the time subscripts
The rm maximizes prots taking factor prices as given
max
L,K
F(K, AL) WL rK.
We have normalized the price of output to 1
The rst order conditions for this problem are
F
L
= W and
F
K
= r .
Hashmi (National U of Singapore) RCK Model August 22, 2014 9 / 48
Behavior of the rm
Since F is CRS,
Y = F(K, AL) = ALf (k),
where k =
K
AL
.
==
F
K
= f
/
(k) and
F
L
= A[f (k) f
/
(k)k].
The rst order conditions become
A[f (k) f
/
(k)k] = W (2.4) , and
f
/
(k) = r . (2.3)
The wage rate per eective unit of labor is
W
A
= w = f (k) f
/
(k) k (2.5)
Hashmi (National U of Singapore) RCK Model August 22, 2014 10 / 48
Behavior of the household
The household chooses a path of consumption to maximize its utility
taking timepaths of r and W as given
Dene R(t) =
t
_
=0
r ()d
This allows for time varying interest rate
If the interest rate is constant, say at r , then R(t) = r t
Hashmi (National U of Singapore) RCK Model August 22, 2014 11 / 48
Behavior of the household
The household budget constraint is
_
t=0
e
R(t)
C(t)
L(t)
H
dt _
K(0)
H
+
_
t=0
e
R(t)
W(t)
L(t)
H
dt (2.6)
==
K(0)
H
+
_
t=0
e
R(t)
[W(t) C(t)]
L(t)
H
dt _ 0 (2.7)
== lim
s
_
K(0)
H
+
s
_
t=0
e
R(t)
[W(t) C(t)]
L(t)
H
dt
_
_ 0 (2.8)
Hashmi (National U of Singapore) RCK Model August 22, 2014 12 / 48
Behavior of the household
The expression in square brackets in (2.8) is the present value of
capital holdings of the household at time s.
To see this, note that the capital holdings of the household at time s
are
K(s)
H
=
K(0)
H
e
R(s)
+
s
_
t=0
e
R(s)R(t)
[W(t) C(t)]
L(t)
H
dt. (2.9)
Taking the present value we get
K(s)
H
e
R(s)
=
K(0)
H
+
s
_
t=0
e
R(t)
[W(t) C(t)]
L(t)
H
dt.
The righthand side of the last equation is the same as the bracketed
expression in (2.8)
Hashmi (National U of Singapore) RCK Model August 22, 2014 13 / 48
Behavior of the household
Substituting the last equation in (2.8) we get
lim
s
_
K(s)
H
e
R(s)
_
_ 0, (2.10)
which says that the present value of the household assets cannot be
negative in the limit
Using the household budget constraint in (2.6), we have derived a
boundary condition in (2.10).
Hashmi (National U of Singapore) RCK Model August 22, 2014 14 / 48
Behavior of the household
Before solving the household maximization problem it is instructive to
write the objective function and constraints in terms of variables per
eective unit of labor
Recall the utility function
U =
_
t=0
e
t
C(t)
1
1
L(t)
H
dt.
Dene c(t) =
C(t)
A(t)
==C(t) = c(t)A(t) = c(t)A(0)e
gt
.
Also recall L(t) = L(0)e
nt
. The utility can then be written as
U =
_
t=0
e
t
[c(t)A(0)e
gt
]
1
1
L(0)e
nt
H
dt.
Hashmi (National U of Singapore) RCK Model August 22, 2014 15 / 48
Behavior of the household
U =
_
t=0
e
t
[c(t)A(0)e
gt
]
1
1
L(0)e
nt
H
dt
=
[A(0)]
1
L(0)
H
_
t=0
e
[n(1)g ]t
c(t)
1
1
dt
= B
_
t=0
e
t
c(t)
1
1
dt, (2.12)
.
where B =
[A(0)]
1
L(0)
H
and = n (1 )g > 0.
Hashmi (National U of Singapore) RCK Model August 22, 2014 16 / 48
Behavior of the household
To rewrite the budget constraint, note that C(t)
L(t)
H
= c(t)
A(t)L(t)
H
,
W(t)
L(t)
H
= w(t)
A(t)L(t)
H
and
K(0)
H
= k(0)
A(0)L(0)
H
.
Substituting these in (2.6) we get
_
t=0
e
R(t)
c(t)
A(t)L(t)
H
dt _ k(0)
A(0)L(0)
H
+
_
t=0
e
R(t)
w(t)
A(t)L(t)
H
dt.
Using A(t)L(t) = A(0)L(0)e
(n+g )t
and dividing by
A(0)L(0)
H
we have
_
t=0
e
R(t)
c(t)e
(n+g )t
dt _ k(0) +
_
t=0
e
R(t)
w(t)e
(n+g )t
dt. (2.14)
Hashmi (National U of Singapore) RCK Model August 22, 2014 17 / 48
Behavior of the household
To rewrite (2.10), note that K(s) = k(s)A(s)L(s). Substituting this
in (2.10)
lim
s
_
k(s)
A(s)L(s)
H
e
R(s)
_
_ 0.
==
lim
s
_
A(0)L(0)
H
e
R(s)
e
(n+g )s
k(s)
_
_ 0.
Dividing by
A(0)L(0)
H
we have
lim
s
_
e
R(s)
e
(n+g )s
k(s)
_
_ 0. (2.15)
Hashmi (National U of Singapore) RCK Model August 22, 2014 18 / 48
Behavior of the household
The objective of the household is to choose c(t) to maximize (2.12)
subject to (2.14).
The Lagrangian for the problem is
L = B
_
t=0
e
t
c(t)
1
1
dt+
_
k(0) +
_
t=0
e
R(t)
w(t)e
(n+g )t
dt
_
t=0
e
R(t)
c(t)e
(n+g )t
dt
_
. (2.16)
The rst order condition with respect to c(t) is
Be
t
c(t)
= e
R(t)
e
(n+g )t
. (2.17)
Hashmi (National U of Singapore) RCK Model August 22, 2014 19 / 48
Behavior of the household
Take logs and use R(t) =
t
_
=0
r ()d
ln B t ln c(t) = ln
t
_
=0
r ()d + (n +g)t. (2.18)
Dierentiate with respect to time
c(t)
c(t)
= r (t) + (n +g). (2.19)
Substitute n (1 )g for and solve for
c(t)
c(t)
c(t)
c(t)
=
r (t) g
. (2.20)
Hashmi (National U of Singapore) RCK Model August 22, 2014 20 / 48
Behavior of the household
Recall C(t) = c(t)A(t). Take logs and dierentiate with respect to
time
C(t)
C(t)
=
A(t)
A(t)
+
c(t)
c(t)
Substitute (2.20) and g =
A(t)
A(t)
into the last equation
C(t)
C(t)
= g +
r (t) g
which simplies to
C(t)
C(t)
=
r (t)
. (2.21)
Hashmi (National U of Singapore) RCK Model August 22, 2014 21 / 48
Behavior of the household
If r (t) > , consumption is rising because people consume less (and
save more) now and consume more in future
If r (t) < , consumption is falling because people are impatient and
they want to consume more now
If is low, intertemporal elasticity of substitution is high. People
allow greater variation in consumption for a given dierence between
r (t) and
If , intertemporal elasticty of substitution approaches zero.
People want perfect consumption smoothing.
Hashmi (National U of Singapore) RCK Model August 22, 2014 22 / 48
Dynamics of the Economy
Recall
c(t)
c(t)
=
r (t)g
(2.20).
Since r (t) = f
/
[k(t)] we can write (2.20) as
c(t)
c(t)
=
f
/
[k(t)] g
. (2.24)
c(t) = 0 if f
/
[k(t)] = + g. Let us call the unique k(t) that
satises this equation k
+
.
c(t) > 0 if f
/
[k(t)] > + g i.e. k(t) < k
+
.
c(t) < 0 if f
/
[k(t)] < + g i.e. k(t) > k
+
.
Hashmi (National U of Singapore) RCK Model August 22, 2014 23 / 48
Dynamics of the Economy
Hashmi (National U of Singapore) RCK Model August 22, 2014 24 / 48
Dynamics of the Economy
For k note that
. (2.24)
_
t=0
e
R(t)
c(t)e
(n+g )t
dt
_ k(0) +
_
t=0
e
R(t)
[w(t) G(t)]e
(n+g )t
dt. (2.41)
Hashmi (National U of Singapore) RCK Model August 22, 2014 37 / 48
Comparative Dynamics (2)
Hashmi (National U of Singapore) RCK Model August 22, 2014 38 / 48
Comparative Dynamics (2)
Hashmi (National U of Singapore) RCK Model August 22, 2014 39 / 48
Slope of the Saddle Path
We shall focus on the Saddle Path in the neighborhood of the
steadystate
To do so, we shall linearize the system around the steadystate
Recall the rstorder Taylor approximation formula in case of one
function
f (x) f (x
+
) + f
/
(x
+
)[x x
+
].
In case of two functions and two variables it extends to
_
f (x, y)
g(x, y)
_
_
f (x
+
, y
+
) + f
1
(x
+
, y
+
)[x x
+
] +f
2
(x
+
, y
+
)[y y
+
]
g(x
+
, y
+
) + g
1
(x
+
, y
+
)[x x
+
] +g
2
(x
+
, y
+
)[y y
+
]
_
Hashmi (National U of Singapore) RCK Model August 22, 2014 40 / 48
Slope of the Saddle Path
The dynamic system that we want to linearize consists of
c(t) =
f
/
[k(t)]g
c(t). (2.24)
k
_
c=c
+
k=k
+
+
_
k
c
_
c=c
+
k=k
+
[c c
+
] +
_
k
k
_
c=c
+
k=k
+
[k k
+
]. (2.27)
Hashmi (National U of Singapore) RCK Model August 22, 2014 41 / 48
Slope of the Saddle Path
To simplify the notation, note that in the steadystate, c =
k = 0.
Dene c = c c
+
,
k = k k
+
,
c = c and
k =
k
c
_
c
c
_
c=c
+
k=k
+
c +
_
c
k
_
c=c
+
k=k
+
k (2.28)
k
_
k
c
_
c=c
+
k=k
+
c +
_
k
k
_
c=c
+
k=k
+
k (2.29)
There are four partial derivatives that we need to nd and evaluate at
the steady state values of c and k
Hashmi (National U of Singapore) RCK Model August 22, 2014 42 / 48
Slope of the Saddle Path
Recall
c(t) =
f
/
[k(t)] g
c(t) (2.24).
_
c
c
_
c=c
+
k=k
+
=
_
f
/
[k(t)] g
_
c=c
+
k=k
+
= 0,
because f
/
(k
+
) = + g.
_
c
k
_
c=c
+
k=k
+
=
_
c (t)
f
//
[k(t)]
_
c=c
+
k=k
+
= c
+
f
//
[k
+
]
.
Hashmi (National U of Singapore) RCK Model August 22, 2014 43 / 48
Slope of the Saddle Path
Recall
k
c
_
c=c
+
k=k
+
= 1.
_
k
k
_
c=c
+
k=k
+
= [f
/
[k(t)] (n +g)]
c=c
+
k=k
+
= f
/
[k
+
] (n +g)
= + g (n +g)
=
Hashmi (National U of Singapore) RCK Model August 22, 2014 44 / 48
Slope of the Saddle Path
Substituting these derivatives in (2.28) and (2.29)
c
c
+
f
//
[k
+
]
k (2.30)
k c (2.31)
Dividing (2.30) by c and (2.31) by
k
c
c
c
+
f
//
[k
+
]
k
c
(2.32)
k
c
k
(2.33)
Hashmi (National U of Singapore) RCK Model August 22, 2014 45 / 48
Slope of the Saddle Path
For simplicity, we assume
c
c
=
k
= . (2.32) becomes
=
c
+
f
//
[k
+
]
k
c
.
==
c
k
=
c
+
f
//
[k
+
]
. (2.34)
Substitute (2.34) in (2.33)
c
+
f
//
[k
+
]
. (2.35)
==
2
+
c
+
f
//
[k
+
]
= 0. (2.36)
Hashmi (National U of Singapore) RCK Model August 22, 2014 46 / 48
Slope of the Saddle Path
The two roots of (2.36) are
1
,
2
=
_
2
4
c
+
f
//
[k
+
]
2
.
Since
c
+
f
//
[k
+
]
2
_
is less than zero (the other is
greater than zero and implies a divergent path). Let us call this root
1
.
To be on the linearized saddle path, at time zero c must be equal to
c
+
+
c
+
f
//
[k
+
]
1
(k k
+
).
. .
(cc
+
) from (2.34)
From there,
k(t) = k
+
+e
1
t
[k(0) k
+
] = k
+
+e
1
[k(t 1) k
+
]
c(t) = c
+
+e
1
t
[c(0) c
+
] = c
+
+e
1
[c(t 1) c
+
]
Hashmi (National U of Singapore) RCK Model August 22, 2014 47 / 48
Speed of Adjustment
Assume f (k) = k
==f
//
(k) = ( 1)k
+(2)
From (2.24) k
+
=
_
+g
_ 1
1
From (2.25) c
+
= k
+
(n +g)k
+
Parameters: =
1
3
, = 0.04, n = 0.02, g = 0.01 and = 1
Using the above functional form and parameter values we get
1
= 5.4%
It means each period both c and k cover 5.4% of their current
distance from c
+
and k
+
Hashmi (National U of Singapore) RCK Model August 22, 2014 48 / 48