SOLUTIONS TO CHAPTER 5

Problem 5.3
(a) The equations describing the evolution of technology are given by
~
(1) ln A t  A  gt  A t ,
and
~ ~
(2) A t  A A t 1   A ,t ,
where -1 < A < 1. From equation (1) and letting lnA0 denote the value of lnA in period 0, we have
~ ~
ln A 0  A  g(0)  A 0 . Rearranging to solve for A 0 gives us
~
(3) A 0  ln A 0  A .
Using equations (1) and (2) we can write the following for period 1:
~
(4) ln A1  A  g  A1 ,
and
~ ~
(5) A1  A A 0   A ,1.
Substituting equation (3) into equation (5) yields
(6) A1  A  ln A 0  A    A,1.
~

Finally, substituting equation (6) into equation (4) gives us

(7) ln A1  A  g  A  ln A 0  A    A,1.

Using equations (1) and (2), we can write the following for period 2:
~
(8) ln A 2  A  2g  A 2 ,
and
~ ~
(9) A 2  A A1   A ,2 . Substituting equation (6) into equation (9) yields
 
(10) A 2  A A  ln A 0  A    A,1   A,2  A 2  ln A 0  A   A  A,1   A,2 .
~

Finally, substituting equation (10) into equation (8) gives us

(11) ln A 2  A  2g  A 2  ln A 0  A   A  A,1   A,2 .

Again using equations (1) and (2), we can write the following for period 3:
~
(12) ln A 3  A  3g  A 3,
and
~ ~
(13) A 3  A A 2   A ,3.
Substituting equation (10) into equation (13) yields
 
(14) A 3  A A 2  ln A 0  A   A  A,1   A,2   A,3  A 3  ln A 0  A   A 2 A,1  A  A,2   A,3 .
~

Finally, substituting equation (14) into equation (12) gives us

(15) ln A 3  A  3g  A 3  ln A 0  A   A 2 A,1  A  A,2   A,3 .

(b) Using equation (7) to find the expected value of lnA1 yields
(16) E[ln A1 ]  A  g  A  ln A 0  A  ,
since E[A,1 ] = 0.
Using equation (11) to find the expected value of lnA2 yields
(17) E[ln A 2 ]  A  2g  A 2  ln A 0  A  ,
since E[AA,1 ] = AE[A,1 ] = 0, E[A,2 ] = 0.
5-2 Solutions to Chapter 5

Using equation (15) to find the expected value of lnA3 yields

(18) E[ln A 3 ]  A  3g  A 3  ln A 0  A  ,
since E[A2A,1 ] = A2 E[A,1 ] = 0, E[AA,2 ] = A E[A,2 ] = 0, E[A,3 ] = 0.

Problem 5.4
(a) We need to solve the household's one period problem assuming no initial wealth and normalizing the
size of the household to one. Thus the problem is given by
(1) max lnc + b(1 - l )1- /(1 - ),
c,l
subject to the budget constraint,
(2) c = wl.

Set up the Lagrangian:

(3) L = lnc + b(1 - l )1- /(1 - ) + [wl - c].
The first-order conditions are
(4) L/c = (1/c) -  = 0,
and
(5) L/l = -b(1 - l )- + w = 0.
Substituting the budget constraint into equation (4) yields
(6)  = 1/c = 1/(wl ).
Substituting equation (6) into equation (5) yields
(7) -b(1 - l )- + w/(wl ) = 0,
and simplifying gives us
(8) 1/l = b/(1 - l ).
Although labor supply, l, is only implicitly defined by equation (8), we can see that it will not depend
upon the real wage.

(b) We want a formula for relative leisure in the two periods, (1 - l1)/(1 - l2). Assume that the household
lives for two periods, has no initial wealth, has size Nt /H = 1 for both periods and finally that there is no
uncertainty. Thus the problem can be formalized as
(1  l1 )1 (1  l 2 )1
(9) max ln c1  b  e  ln c 2  e  b ,
1  1 
subject to the intertemporal budget constraint given by
c w l
(10) c1  2  w1l 1 2 2 .
1 r 1 r

Set up the Lagrangian:

(1  l1 )1 (1  l 2 )1  w l c 
(11) L = ln c1  b  e  ln c 2  e  b    w1l 1 2 2  c1  2  .
1  1   1 r 1 r
The first-order conditions are given by the following four equations:
(12) L/c1 = (1/c1 ) -  = 0, and (13) L/c2 = (e-/c2 ) - [/(1 + r)] = 0,
(14) L/l1 = -b(1 - l1 )- + l1 = 0, and (15) L/l2 = -e-b(1 - l2 )- + [l2 /(1 + r)] = 0.
Rearranging equation (14) yields one expression for  :  = b(1 - l1 )- /l1.
 Solutions to Chapter 5 5-3

Rearranging (15) yields another expression for  :  = [e-b(1 - l2 )- (1 + r)]/l2. Equating these two
expressions for  yields
e  b(1  r ) b (1  l 1)  1 w2
(16)    ,
   
(1  l 2 ) w 2 (1  l 1) w1 (1  l 2 ) e (1  r ) w1
and thus
1
(1  l 1)  1 w2 
(17)   .
(1  l 2 )  e  (1  r ) w 1 
If w2 /w1 rises, then (1 - l 1 )/(1 - l 2 ) rises. That is, suppose the real wage in the second period rises
relative to the real wage in the first period. Then the individual increases first-period leisure relative to
second-period leisure, or reduces first-period labor supply relative to second-period labor supply. We
can calculate the elasticity, denoting – for ease of notation – (1 - l 1 )/(1 - l 2 )  l * and w2 / w1  w*:

(18)
 1

l * w * 1 1 e  (1  r ) w *(1  )1 w *
 .
w * l *  l*
Substitute in the denominator for l*  (1 - l 1 )/(1 - l 2 ) from equation (17) to yield
 
l * w * 1 1 e  (1  r ) 1  w *1  1
(19) 

 w * l *  1 e   (1  r ) w * 1   
 .

Thus the smaller is  – or the bigger is 1/ – the more the individual will adjust relative labor supply in
response to a change in relative real wages.

From equation (17), we can also see that if r rises then (1 - l 1)/(1 - l 2) falls. That is, suppose that there is
a rise in the real interest rate. Then the individual reduces first-period leisure relative to second-period
leisure, or increases first-period labor supply relative to second-period labor supply. It is straightforward
to show that
(1  l 1) (1  l 2 ) 1  r  1
(20)  .
1  r  (1  l1 ) (1  l 2 ) 
Thus the smaller is  – or the bigger is 1/ – the more the individual will respond to a change in the real
interest rate. Note that with log utility, where  = 1, this elasticity is equal to one.

Intuitively, a low value of  means that utility is not very sharply curved in l. This means that l responds
a lot to changes in wages and the interest rate.

Problem 5.5
(a) The problem is to maximize utility as given by
(1) lnc1 + bln(1 - l 1 ) + e-[lnc2 + bln(1 - l 2 )],
subject to the following lifetime budget constraint:
1 1
(2) c1  c 2  w1l1  w 2l2
1 r 1 r

Set up the Lagrangian:

 
(3) L  ln c1  b ln(1  l1 )  e  ln c 2  b ln(1  l 2 )   w1l1 
1 1
w 2 l 2  c1  c2  .
 1 r 1 r 
The four first-order conditions are given by
5-4 Solutions to Chapter 5

L 1
(4)  0;
c1 c1
L e   
(5)   0 ;
c 2 c2 1  r
L b
(6)    w1  0 ; and
 l1 (1  l1 )
L  e   b  w 2
(7)   0 .
l 2 (1  l 2 ) 1  r
Equation (4) implies that
1
(8) c1  .

Equation (5) implies that
e  (1  r )
(9) c 2  .

Equation (6) implies that
b
(10) l1  1  .
w1
Equation (7) implies that
(1  r )e  b
(11) l2 1  .
w 2
Now substitute equations (8) - (11) into the lifetime budget constraint, equation (2), to obtain
1 e   (1  r )  b  w 2  (1  r ) e   b 
(12)   w 11   1  .
  (1  r )   w 1  1  r   w 2 
Multiplying both sides of equation (12) by  gives us
  w 1  b   w 2   w 2  (1  r ) e   b 
  .
(13) 1  e   w 1 
  w 1  1  r   w2 
Simplifying further allows us to obtain
 w2
(14) 1  e     w 1  b   e  b.
1 r
Now use the fact that 1 + e- + b + e-b = (1 + e-)(1 + b) to solve for :
(1  e   )(1  b)
(15)   .
 w1  w 2 (1  r )
To obtain an expression for first-period labor supply, substitute equation (15) into equation (10) to obtain
b bw 1  w 2 (1  r )
(16) l1  1  1 .

(1  e )(1  b) (1  e  )(1  b)w 1
w
w1  w 2 (1  r) 1
Dividing the top and bottom of the second term on the right-hand side of (16) by w1 yields
b1  w 2 w1 1 (1  r ) 
(17) l1  1  .
(1  e  )(1  b)
 Solutions to Chapter 5 5-5

Note that l 1 is a function of the relative wage, w2 /w1 . Thus any change in w1 and w2 that leaves w2 /w1
unchanged will leave l 1 unchanged.

To obtain an expression for second-period labor supply, substitute equation (15) into equation (11) to
obtain
(1  r )e  b (1  r )e  bw 1  w 2 (1  r )
(18) l2 1  1 .
(1  e  )(1  b) (1  e  )(1  b)w 2
w
w1  w 2 (1  r) 2
Dividing the top and bottom of the second term on the right-hand side of (18) by w2 yields
(1  r )e  bw1 w 2   1 (1  r )
(19) l2  1  .
(1  e  )(1  b)
Again, note that l 2 depends only on the relative wage, w1 /w2 . Thus any change in w1 and w2 that leaves
w1 /w2 unchanged will leave l 2 unchanged.

(b) (i) The fact that the household has initial wealth of Z > 0 will not affect equation (5.23) in the text –
the Euler equation – that relates consumption in one period to expectations of consumption the following
period. The fact that the household has initial wealth does not change the marginal utility lost from
reducing current consumption by a small amount today nor does it change the expected marginal utility
gained by using the resulting greater wealth to increase consumption next period above what it otherwise
would have been. That is, it does not affect the experiment by which we informally derived the Euler
equation. The budget constraint, just as in the Ramsey model, only becomes important when determining
the level of consumption each period.

(b) (ii) The result in part (a) will not continue to hold if the household has initial wealth. The new
lifetime budget constraint is given by
1 1
(20) c1  c 2  Z  w1l1  w 2l2 .
1 r 1 r
The addition of a constant to lifetime wealth does not affect the four first-order conditions. Take those
first-order conditions, equations (8) through (11), and substitute them into this new budget constraint:
1 e   (1  r )  b  w 2  (1  r ) e   b 
(21)   Z  w 11   1  .
  (1  r )   w 1  1  r   w 2 
Following the same algebra steps as in part (a) now yields
(1  e  )(1  b)
(22)   .
Z  w1  w 2 (1  r)
To obtain an expression for first-period labor supply, substitute equation (22) into equation (10) to obtain
b bZ  w 1  w 2 (1  r )
(23) l1  1  1 .

(1  e )(1  b) (1  e  )(1  b)w 1
w
Z  w1  w 2 (1  r) 1
Dividing the top and bottom of the second term on the right-hand side of (23) by w1 yields
bZ w1   1  w 2 w1 1 (1  r ) 
(24) l1  1  .
(1  e  )(1  b)
5-6 Solutions to Chapter 5

Taking the derivative of l 1 with respect to w1 – imposing the condition that w2 /w1 remains constant –
yields
 l1 bZ w 12
(25)   0.
 w 1 (1  e  )(1  b)
Thus a change in w1 , even if it is accompanied by a change in w2 such that relative wages remain
constant, does affect first-period labor supply. In fact, a rise in the first-period wage increases first-
period labor supply.

To obtain an expression for second-period labor supply, substitute equation (22) into equation (11) to
obtain
(1  r )e  b (1  r )e  bZ  w 1  w 2 (1  r )
(26) l2 1  1 .
(1  e  )(1  b) (1  e  )(1  b)w 2
w
Z  w1  w 2 (1  r) 2
Dividing the top and bottom of the second term on the right-hand side of (26) by w2 yields
(1  r )e  bZ w 2   w1 w 2   1 (1  r )
(27) l2 1  .
(1  e  )(1  b)
Taking the derivative of l 2 with respect to w2 – imposing the condition that w1 /w2 remains constant –
yields
l 2 (1  r )e  bZ w 2 2
(28)   0.
w 2 (1  e  )(1  b)
Thus a change in w2 , even if it is accompanied by a change in w1 such that relative wages remain
constant, does affect second-period labor supply. In fact, a rise in the second-period wage increases
second-period labor supply.

Problem 5.6
(a) The real interest rate is potentially random, so let r = E[r] +  where  is a mean-zero random error.
The individual maximizes expected utility as given by
(1) U  ln C1  E[ln C 2 ] .
Substituting in for C2 yields
(2) U  ln C1  E[ln((1  E[r]  )(Y1  C1 ))] .
Set the derivative of equation (2) with respect to C1 equal to zero to obtain the first-order condition:
(3) U C1  1 C1  E(1)(1  E[r]  ) (1  E[r]  )(Y1  C1 )  0 ,
or simplifying
 
(4) 1 C1  E 1  Y1  C1   0 .
Since 1/(Y1 - C1 ) is not random, E[1/(Y1 - C1 )] = 1/(Y1 - C1 ) and thus we can write
(5) C1 = Y1 /2.
In this case, the choice of C1 is not affected by whether r is certain or not. Even if r is random, the
individual simply consumes half of first-period income and saves the rest.

(b) Now the individual does not receive any first-period income but receives income Y2 in period 2. So
the individual's problem is to maximize expected utility as given by equation (1), subject to
(6) C1 = B1 ,
and
 Solutions to Chapter 5 5-7

(7) C 2  Y2  (1  E[r]  )B1  Y2  (1  E[r]  )C1 ,

where B1 represents the amount of borrowing the individual does in the first period. Substituting (7) into
the expected utility function (1) yields
(8) U  ln C1  E[ln(Y2  (1  E[r]  )C1 )] .
Set the derivative of equation (8) with respect to C1 equal to zero to find the first-order condition:
(9) U C1  1 C1  E(1  E[r]  ) C 2   0 .
Use the formula for the expected value of the product of 2 random variables – E[XY] = E[X]E[Y] +
cov(X,Y) – to obtain
(10) 1 C1  (1  E[r])E[1 C 2 ]  cov1  E[r]  ,1 / C 2  .
The covariance term is positive. Intuitively, a higher  means the individual has to pay more interest on
her borrowing which forces her to have lower C2 and thus higher 1/C2 .

If r is not random – so that r = E[r] with certainty – we can rewrite equation (10) as
(11) 1 C1  (1  E[r])(1 C 2 )  (1  E[r]) Y2  (1  E[r])C1  ,
which implies that
(12) Y2  (1  E[r])C1  (1  E[r])C1 .
Solving for C1 yields
(13) C1  Y2 2(1  E[r]) .
In the case where r is random, equation (10) can be written as
(14) 1 C1  E[1  E[r]  ] E[1 C 2 ]  cov1  E[r]  ,1 / C 2  .
Since 1/C2 is a convex function of C2 , then by Jensen's inequality we have E[1 C2 ]  1 / E [C2 ] . In
addition, because the covariance term is positive, we can write
(15) 1 C1  (1  E[r])E[1 C 2 ]  cov1  E[r]  ,1 C 2   (1  E[r])1 E[C 2 ] .
Substituting into this inequality the fact that E[C2 ] = Y2 - (1 + E[r])C1 yields
(16) 1 C1  (1  E[r]) Y2  (1  E[r])C1  ,
which implies that
(17) Y2  (1  E[r])C1  (1  E[r])C1 .
The inequality in (17) can be rewritten as
(18) 2(1  E[r])C1  Y2
or simply
(19) C1  Y2 2(1  E[r]) .
Note from equation (13) that the right-hand side of (19) is the optimal choice of C1 under certainty. Thus
we have shown that if r becomes random with no change in the expected value of r, the optimal choice of
C1 becomes smaller. Essentially, if there is some uncertainty about how much interest the individual will
have to pay in the second period, she is more cautious in her decision as to how much to borrow and
consume in the first period.

Problem 5.7
(a) Imagine the household increasing its labor supply per member in period t by a small amount l.
Suppose it then uses the resulting greater wealth to allow less labor supply per member in the next period
and allowing for consumption per member to be the same in both periods as it otherwise would have
been. If the household is behaving optimally, a marginal change of this type must leave expected lifetime
utility unchanged.

Household utility and the instantaneous utility function of the representative member of the household
are given by
5-8 Solutions to Chapter 5

t 
(1) U   e t u(c t ,1  l t ) N t H,
t 0
and
(2) ut = lnct + bln(1 - l t ).
From equations (1) and (2), the marginal disutility of working in period t is given by
(3) -U/l t = e-t(Nt /H)[b/(1 - l t )].
Thus increasing labor supply per member by l has the following utility cost for the household:
(4) Utility Cost = e-t(Nt /H)[b/(1 - l t )]l.
This change in labor supply raises income per member in period t by wtl. Note that the household has
en times as many members in period t + 1 as in period t. Thus the increase in wealth per member in
period t + 1 is e-n[(1 + rt+1 )wtl ].

We need to determine how much this change in labor supply in period t allows labor supply per member
in period t + 1 to fall, if the path of consumption is to be unaffected. In period t + 1, giving up one unit
of labor per member costs wt+1 in lost income per member. Thus giving up 1/wt+1 units of labor per
member implies lost income of one per member. Or, giving up [e-n (1 + rt+1 )wtl ]/wt+1 units of labor
results in lost income per member of e-n (1 + rt+1 )wtl . That is exactly equal to the extra wealth per
member the household has from working more last period. Thus we have determined that labor supply
per member can fall below what it otherwise would have been by the amount [e-n (1 + rt+1 )wtl ]/wt+1
while still allowing consumption to be the same as it otherwise would have been. This allowable drop in
labor supply per member results in the following expected utility benefit as of period t:
 N b e  n (1  rt 1 ) w t l 
(5) Expected Utility Benefit = E t e ( t 1) t 1 .
 H (1  l t 1 ) w t 1 

Equating the cost and expected benefit yields

N b  N b e n (1  rt 1 ) w t l 
(6) e t t l  E t e ( t 1) t 1 .
H (1  l t )  H (1  l t 1 ) w t 1 
Since e-(t+1)(Nt+1 /H)e-n is not uncertain and since Nt+1 = Nt en, this simplifies to
b  b(1  rt 1 ) w t 
(7)  e  E t  .
(1  l t )  (1  l t 1 ) w t 1 

(b) Consider the household in period t. Suppose it reduces its current consumption per member by a
small amount c and then uses the resulting greater wealth to increase consumption per member in the
next period above what it otherwise would have been. The following equation, (5.23) in the text, gives
the condition this experiment implies, assuming the household is behaving optimally:
1  1 
(5.23)  e  E t  (1  rt 1 )  .
ct  c t 1 

Now imagine the household increasing its labor supply per member in period t by a small amount l and
using the resulting income to increase its consumption in that period. The following equation, (5.26) in
the text, gives the condition that this experiment implies, assuming that the household is behaving
optimally:
 Solutions to Chapter 5 5-9

ct w
(5.26)  t .
1  lt b
Solving for 1/ct gives us
1 b
(8)  .
c t (1  l t ) w t
Note that equations (5.26) and (8) hold in every period. Thus for period t + 1, we can write
1 b
(9)  .
c t 1 (1  l t 1 ) w t 1
Substituting equations (8) and (9) into equation (5.23) yields
b  b(1  rt 1 ) 
(10)  e  E t  .
(1  l t ) w t  (1  l t 1 ) w t 1 
Multiplying both sides of equation (10) by wt , and using the fact that Et [wt ] = wt gives us
b  b(1  rt 1 ) w t 
(11)  e  E t  .
(1  l t )  (1  l t 1 ) w t 1 
This is the same condition obtained from the experiment in part (a).

Problem 5.8
(a) To obtain the first-order condition or Euler equation, we can use the informal perturbation method.
The experiment is to suppose the individual reduces period-t consumption by C. She then uses the
resulting greater wealth in period t + 1 to increase consumption above what it otherwise would have
been. The utility cost in period t of doing so is given by the following expression:
(1) Utility Cost = 1 (1  ) u  C t  C  1 (1  ) 1  2 C t  C ,
t t

where we have used the instantaneous utility function, u(Ct ) = Ct - Ct2 , to calculate u ' (Ct ).

The following expression gives the expected utility gain in period t + 1 from the above experiment:
 t 1

(2) Exp. Utility Gain = E t 1 (1  ) u C t 1 1  A  C = 1 (1  ) E t 1  2C t 1 1  A C ,
t 1

where A is the real interest rate. Equation (2) simplifies to

(3) Exp. Utility Gain = 1 (1  )t 11  2E t [C t 1 ]1  A C .

If the individual is optimizing, the utility cost from this perturbation must equal the expected utility gain:
t 1
(4) 1 (1  ) 1  2 C t  C  1 (1  ) 1  2E t [C t 1 ]1  A C ,
t

or simply
(5) 1  2 C t  1 (1  )1  A 1  2E t [C t 1 ] .
Using the fact that  = A and simplifying yields
(6) Ct = Et [Ct+1 ].
Consumption follows a random walk. The expected value of consumption next period is simply equal to
today's actual realization of consumption.

(b) We will guess that consumption takes the form

(7) C t    K t   e t .
Substitute equation (7) and the production function, Yt = AKt + et , into the capital-accumulation
equation, Kt+1 = Kt + Yt - Ct , to obtain
(8) K t1  K t  AK t  e t     K t   e t ,
or simply
5-10 Solutions to Chapter 5

(9) K t 1     1  A    K t  (1   )e t .

(c) Substitute equation (7) and equation (7) lagged forward one period into the first-order condition,
equation (1):
(10)   K t  e t  E t   K t1  e t1  .
Substituting equation (9) into equation (10) yields
(11)   K t   e t     E t    (1  A  )K t  (1   )e t    E t  e t1  .
Noting that E t  e t1   E t e t   t1   e t , we can collect terms to obtain
(12)    K t   e t  (1  )  (1  A  )K t      (  )e t .

In order for equation (12) to hold, we need the coefficients on K t and et , as well as the constant term, to
be the same on both sides. Equating the coefficients on Kt gives us
(13)   (1  A  ) ,
or simply
(14)   A .
Equating the coefficients on et gives us
(15)      (  ) .
Using equation (14) and simplifying yields
(16)  (1   A)  A ,
or simply
A
(17)   .
1   A
Finally, equating the constant terms yields
(18)   (1  ) .
Unless   A  1, this requires
(19)   0 .
Note that we are also ignoring the case in which   0 and   0 with no restriction on .

(d) Substituting equations (14), (17), and (19) into the guess for consumption, equation (7), and the
capital-accumulation equation, equation (9), yields
 A 
(20) C t  AK t   et ,
1    A 
and
 1  
(21) K t1  K t   et .
1    A 

To keep the analysis simple, and without loss of generality, we can assume that  , and thus e, both equal
0 until some period t. In period t, there is a one-time, positive realization of  t  1    A . From period
t + 1 forward,   0 again. In what follows, the change in a variable refers to the difference between its
actual value and the value it would have had in the absence of the one-time shock (i.e. if  and e had
remained at 0 forever).

In period t, Kt is unaffected. From equation (21), we can see that Kt is determined by last period’s capital
stock and last period’s realization of e. From the production function, Yt = AKt + et , we have
(22) Yt  AKt  et  0  (1    A) .
 Solutions to Chapter 5 5-11

Thus output in the period of the shock is higher by (1   A). From equation (20), the change in
consumption is given by
 A   A 
(23) C t  AK t   e t  0   1    A   A .
1    A  1    A 
Thus consumption in the period of the shock is higher by A.

In period t + 1, even though  t1 is assumed to be 0 again, et+1 is different than it would have been in the
absence of the one-time shock due to the autoregressive form of the e terms. More precisely
(24) et 1  et  (1    A).
From equation (21), the change in the capital stock is given by
 1    1  
(25) K t 1  K t   e t  0   1    A   (1  ) .
1    A  1    A 
Intuitively, last period, output rose by (1   A) but consumption rose only by A. The rest of the
increase in output, (1 ), was devoted to investment and hence the rise in this period's capital stock by
an equal amount (we are assuming no depreciation). From the production function, Yt+1 = AKt+1 + et+1 ,
the change in output is
(26) Yt1  AK t1  e t1  A(1  )  (1  )  A  A  A    A   2  A  (1  ) .
From equation (20), the change in consumption is given by
 A   A 
(27) C t1  A K t1   e t1  A(1  )   (1    A )  A  A  A  A .
1    A  1    A 
Thus there are no further dynamics for consumption. It remains higher than it would have been in the
absence of the shock by the amount A.

Similarly, we can calculate these changes for period t+2:

(28) e t2  e t1   2 (1    A);
 1    (1  )(1    A )  2
(29) K t2  K t1   e t1  (1  )     (1  )  (1  )  1   ;
1    A   1   A 
(30) Yt2  AK t2  e t2  A(1  )  A(1  )   2 (1    A)  A   2 (1  ) ; and
 A  2  A  2 2 2
(31) C t2  A K t 2   e t2  A(1  )    (1    A )  A   A   A  A .
1    A  1    A 

The pattern can now be inferred. Suppose there is a one-time shock of  t  1    A . In the period of the
shock, consumption rises by A and permanently stays at that new level with no further dynamics. In
addition, n periods after the shock, the change in output is
(32) Yt  n  A   n (1  ) ,
and the change in the capital stock is
(33) K tn  1   n .

The nature of the dynamics of Y and K depends on the value of  . In the special case in which  is equal
to 0, so that there is no persistence in the technology shock, there are no further dynamics after period
t + 1. The period after the shock, and in all those thereafter, capital is higher by one and output is higher
by A.
5-12 Solutions to Chapter 5

For the case of 0    1, the capital stock rises by (1 ) the period after the shock. It then increases
more each period until it asymptotically approaches its new long-run level that is one higher than it
would have been in the absence of the shock. Output rises by (1   A) the period of the shock. It then
decreases each period until it asymptotically approaches its new long-run level that is A higher than it
would have been in the absence of the shock.

For the case of 1    0 , capital and output oscillate – alternating above and below their new long-run
levels in successive periods – and gradually settle down to be one and A higher, respectively.

Problem 5.9
(a) To obtain the first-order condition or Euler equation, we can use the informal perturbation method.
The experiment is to suppose the individual reduces period-t consumption by C. She then uses the
resulting greater wealth in period t + 1 to increase consumption above what it otherwise would have
been. The utility cost in period t of doing so is given by the following expression:

(1) Utility Cost = 1 (1  ) u C t   C  1 (1  ) 1  2  C t  t   C ,
t t

where we have used the instantaneous utility function, u(Ct ) = Ct - (Ct + t )2 , to calculate u ' (Ct ).

The expected utility gain in period t + 1 from the above experiment is given by the following:
(2) Expected Utility Gain = E t  1 (1  ) u C t 1   1  A  C
 t 1 

= 1 (1  )
t 1
 
E t 1  2  C t 1  t 1   1  A  C ,
where A is the real interest rate. Since  is white noise, Et[t+1 ] = 0, and thus
(3) Expected Utility Gain = 1 (1  )
t 1
1  2E t [C t1 ]1  A C .
If the individual is optimizing, the utility cost from this perturbation must equal the expected utility gain:
 
(4) 1 (1  ) 1  2  C t  t   C  1 (1  ) 1  2E t [C t 1 ]  1  A  C ,
t t 1

or simply
(5) 1  2  C t  t   1 (1  )  1  A 1  2E t [C t 1 ] .
Using the fact that  = A and simplifying yields
(6) C t  t  E t [C t 1 ].

(b) We will guess that consumption takes the form

(7) C t    K t   t .
Substitute equation (7) and the production function, Yt = AKt , into the capital-accumulation equation,
Kt+1 = Kt + Yt - Ct , to obtain
(8) K t 1  K t  AK t     K t   t ,
or simply
(9) K t 1    1  A    K t   t .

(c) Substitute equation (7) and equation (7) lagged forward one period into the first-order condition,
equation (6):
(10)    K t   t  t  E t [   K t1   t 1 ] .
Noting that Et[Vt+1 ] = 0, we have
(11)    K t  (   1)t     E t [ K t 1 ] .
 Solutions to Chapter 5 5-13

Substitute equation (9) into equation (11). Since Kt+1 is a function of Kt and t which are both known at
time t, we have
 
(12)    K t  (   1)t       1  A    K t  t .
Simplifying yields
(13)    K t  (   1)t   1      1  A    K t  t .
In order for equation (13) to hold, we need the coefficients on K t , t ,and the constant term to be the
same on both sides. That is, we need
(14)   1  A    ,
which implies that
(15)  = A.
In addition, we require that
(16)   1   ,
or simply
(17)    1 (1  ) .
Substituting equation (15) into equation (17) gives us
(18)   1 (1  A) .
Finally, we require that
(19) 1      .
If  does not equal 0, this implies that
(20)  = 0.
There is another set of parameter values that satisfies equation (13):  = 0 and  = -1 with no restriction
on . This second solution is economically unappealing, however, since  = 0 implies that consumption
does not depend on the capital stock. This is not realistic since consumption depends on output which in
turn is determined by the capital stock. Thus we can, on economic grounds, ignore this second solution.

(d) Substituting equations (15), (18) and (20) into the guess for consumption, equation (7), and the
capital-accumulation equation, equation (9), yields
(21) C t  AK t  1 (1  A) t ,
and
(22) K t 1  K t  1 (1  A) t .
Without loss of generality, we can assume that  = 0 until some period t when there is a one-time positive
realization of t . To keep the analysis simple, assume that t = (1 + A). From period t + 1 forward,
 = 0 again.

In period t, Kt is unaffected. It is determined by last period's capital stock and last period's saving. From
the production function, Yt = AKt , Yt is unaffected since Kt is unaffected. From equation (21), we can
see that consumption in period t, Ct , is lower by 1 (1  A) t  1 (1  A)  1  A  1.

In period t + 1, we can see from equation (22) that Kt+1 is higher by 1 (1  A) t  1 (1  A)  1  A  1.
Intuitively, last period's drop in consumption by one, with unchanged output, meant an increase in saving
of one. This in turn means an increase in this period's capital stock by one. Through the production
function, since Kt+1 is higher by one, output is higher by A. Finally, with t+1 assumed to be 0, then since
Kt+1 is higher by one, Ct+1 must be higher than it was in period t - 1 (before the shock) by A. This last
fact can be seen from equation (21).
5-14 Solutions to Chapter 5

From period t + 2 forward, assuming  = 0 forever, there will be no further dynamics. K stays at its new
higher level: one higher than in period t - 1. Y stays at its new higher level: A higher than in period t - 1.
C stays at its new higher level: A higher than in period t - 1. All of this is depicted in the figure below.

C Y K
• • •

• • • • • •
• • •
• • • • •

t-1 t t+1 t+2 t+3 t-1 t t+1 t+2 t+3 t-1 t t+1 t+2 t+3

Problem 5.10
(a) From the Solow, Ramsey, and Diamond models it is clear that on the balanced growth path without
shocks, the growth rates of Y, K and C are all equal to n + g. In addition, the growth rate of w is g, the
growth rate of L is n, and the growth rates of l and r are zero. Given the logarithmic structure here,
"growth rate" means the change in the logarithm of the variable; the fact that the growth rate of K is n + g
means that ln(Kt+1 ) - ln(Kt ) = n + g.

Dividing both sides of the production function, Yt = Kt[AtLt ]1- , by AtLt yields
(1) Yt /AtLt = Kt[AtLt ]- = [Kt /AtLt ].
Since y* and k* are the balanced-growth-path values of Y/AL and K/AL respectively, we have
(2) y* = k*.
Similarly, dividing both sides of the capital-accumulation equation, Kt+1 = Kt + Yt - Ct - Gt - Kt , by AtLt
gives us
K t 1 Kt Yt Ct Gt K t
(3)      .
A t Lt A t Lt A t Lt A t Lt A t Lt A t Lt
Using the notation given in the question and the fact that Kt+1 = enegKt on the balanced growth path and
thus that Kt+1 /AtLt = enegKt /AtLt yields
(4) enegk* = k* + y* - c* - G* - k*.
Dividing both sides of the equation giving the real wage, wt = (1 - )[Kt /AtLt ]At , by At yields
(5) wt /At = (1 - )[Kt /AtLt ].
Denoting the value of w/A on the balanced growth path as w* gives us
(6) w* = (1 - )k*.
From equation (5.4) in the text giving the real interest rate, we have on a balanced growth path
(7) r* = k*-(1-) - .

We need to transform textbook equation (5.26), which relates the trade-off between current consumption
and current labor supply, into an expression concerning the balanced growth path without shocks. Note
that in equation (5.26) , ct /(1 - l t ) = wt /b , c is consumption per person, C/N. We are interested in c*
which is consumption per unit of effective labor, C/AL. Since C/N = (C/AL)(L/N)A, on the balanced
growth path it is true that c = c*l *A. Using this fact and dividing both sides of equation (5.26) by A, we
obtain
 Solutions to Chapter 5 5-15

c*l *A A wA
(8)  .
(1  l *) b
Since w* = W/A, we have
c*l * w*
(9)  .
(1  l *) b
Finally, we need to transform textbook equation (5.23), which relates the tradeoff between current and
future consumption. First, since there is no uncertainty without any shocks, we can write equation (5.23),
1/ct = e-Et [(1 + rt+1 )/ct+1 ], as
(10) 1/ct = e- (1 + rt+1 )/ct+1 .
Then multiply both sides of equation (10) by ct+1 :
(11) ct+1 /ct = e-(1 + rt+1 ).
On the balanced growth path, consumption per person grows at rate g and thus ct+1 = ct eg or
ct+1 /ct = eg . Thus we have
(12) 1 + r* = e+g.
We now have six equations in the following six variables: y*, k*, c*, w*, l *, and r*.

(b) We need to assume the following parameter values:  = 1/3, g = 0.005, n = 0.0025,  = 0.025,
r* = 0.015, and l * = 1/3. Note that these are quarterly values for n, g and r*.

From equation (7), we can obtain an expression for capital per unit of effective labor on the balanced
growth path, k*:
(13) k* = [/(r* + )]1/(1-).
Substituting the given parameter values yields
(14) k* = [(1/3)/(0.015 + 0.025)]1/(1-1/3) = 24.0563.
Substituting this value for k* into equation (2) gives us a value for quarterly output per unit of effective
labor on the balanced growth path:
(15) y* = k* = (24.0563)1/3 = 2.8868.
We are told that the ratio of government purchases to output on the balanced growth path is (G/Y)* = 0.2.
This means that:
(16) [G/AL]/[Y/AL] = 0.2
and substituting for the value of y* calculated above gives us
(17) G/AL  G* = (0.2)(2.8868) = 0.5774.
From equation (4), we can solve for consumption per unit of effective labor on the balanced growth path,
c*:
(18) c* = k* + y* - G* - k* - enegk* = (1 -  - eneg )k* + y* - G*.
Substituting for the values we know yields
(19) c* = (1 - 0.025 - e0.0025e0.005 )(24.0563) + 2.8868 - 0.5774 = 1.5269.

It is then straightforward to use these values for c* and y* to solve for the share of output devoted to
consumption on the balanced growth path:
(20) C/Y = [C/AL]/[Y/AL]  c*/y* = 1.5269/2.8868 = 0.5289.
Thus consumption's share in output is approximately 53 percent. Since output is devoted to
consumption, investment, or government purchases, we know that
(21) I/Y = 1 - C/Y - G/Y = 1 - 0.5289 - 0.2 = 0.2711,
and thus investment's share in output is roughly 27%. Compared to actual figures for the U.S. this is
giving slightly too much weight to investment and slightly too little weight to consumption. Finally, the
implied ratio of capital to annual output on the balanced growth path is
(22) K/4Y = [K/AL]/[4Y/AL]  k*/4y* = 24.0563/[(4)(2.8868)] = 2.083.
5-16 Solutions to Chapter 5

Problem 5.11
Before invoking the simplifying assumptions, the model here is the RBC model with no government and
100-percent depreciation, given by
~
(1) Yt = Kt [At Lt ]1- , (2) Kt+1 = Yt - Ct , (3) ln A t  A  gt  A t ,
~ ~
(4) A t  A A t 1   A ,t , (5) ln N t  N  nt , and (6) ut = lnct + bln(1 - l t ).
In this question, we are simplifying by assuming n  g  A  N  0 . This results in the following
adjustments to the model. Population is given by
(7) ln Nt = 0,
which implies that
(8) Nt = 1.
Since we have normalized the population to one, labor supply per person, l t , will be the same as total
labor supply, Lt. Thus we can rewrite the production function as
(9) Yt = Kt [At l t ]1-.
~
With respect to technology, since g and A are equal to 0, we have ln A t  A t . Using equation (4) to
rewrite this yields
(10) lnAt = A lnAt-1 + A,t .

(a) Define the value function at time t as

 
(11) Vt  max E t  e (s t ) ln C s  b ln(1  l s ) .

s t 
Since we are solving the social planner's problem, the maximization is subject to the constraints given by
the production function, equation (9), the capital-accumulation equation (2), and the technology equation
(10). Thus the value function at time t is the expected present value of lifetime utility, from time t
forward, evaluated at all the optimal choices of consumption and labor supply. The technique we will
use here allows us to reduce what looks like a complicated multiperiod problem down to a two-period
problem. That is because the value function must satisfy Bellman's Equation given by
 
(12) Vt (K t , A t )  max [ln C t  b ln(1  l t )]  e  E t Vt 1 (K t 1 , A t 1 ) .
C,l
Equation (12) says that the value function at time t is equal to utility at time t, evaluated at the optimal C t
and lt , plus the discounted expected value as of time t of next period's value function. That is, the
expected value of maximized lifetime utility is maximized lifetime utility "today" plus "today's"
expectation of maximized lifetime utility from "tomorrow" on, appropriately discounted.

(b) We will guess that the value function is of the form

(13) Vt (Kt , At ) =  0 +  K lnKt +  A lnAt .
Substituting this guess into equation (12), the Bellman equation, yields
 
(14) Vt (K t , A t )  max [ln C t  b ln(1  l t )]  e  E t  0   K ln K t 1   A ln A t 1  .
C,l
Taking logs and then expectations of both sides of equation (2), the capital-accumulation equation, yields
(15) Et[lnKt+1 ] = Et[ln(Yt - Ct )] = ln(Yt - Ct ),
since Yt and Ct are both known as of time t. Taking the expected value of both sides of equation (10)
yields
(16) Et[lnAt+1 ] = A lnAt ,
since the  shocks have mean zero. Substituting equations (15) and (16) into equation (12) yields
 
(17) Vt (K t , A t )  max [ln C t  b ln(1  l t )]  e   0   K ln(Yt  C t )   A  A ln A t  .
C,l
 Solutions to Chapter 5 5-17

The first-order condition for Ct is

(18) 0  1 C t  e  K (1) (Yt  C t ) ,
which implies that
(19) 1 C t  e  K Yt  C t  .
Equation (19) can be rewritten as
(20) Yt  C t  e  K C t ,
or simply
(21) C t (1  e  K )  Yt .
Thus consumption is given by
(22) C t  [1 / (1  e  K )]Yt .
The ratio of consumption to output is given by
(23) Ct /Yt = 1/(1 + e-  K ).
Equation (23) shows that the ratio of consumption to output does not depend on Kt or At.

(c) The first-order condition for l t (noting that Lt = l t ) is

(24) 0 = -b/(1 - l t ) + [e-  K /(Yt - Ct )](1 - )Kt At1- l t -.
Simplifying yields
(25) b/(1 - l t ) = [e-  K /(Yt - Ct )](1 - )(Yt / l t ).
Substituting equation (20) into equation (25) yields
(26) b/(1 - l t ) = [e-  K /e-  K Ct )](1 - )(Yt / l t ) = (Yt /Ct )[(1 - )/ l t ].
Substituting equation (23) into equation (26) yields
(27) b/(1 - l t ) = (1 - )(1 + e-  K )/ l t .
Multiplying both sides of equation (27) by (1 – l t )l t gives us
(28) b l t = (1 – l t )(1 - )(1 + e-  K ).
Further simplification allows us to obtain
(29) l t [(1 - )(1 + e-  K ) + b] = (1 - )(1 + e-  K ),
and thus
(1  )
(30) l t  .
(1  )  [b /(1  e  K )]
Thus l t , labor supply per person, does not depend on Kt or At either. In addition, with some simple
algebra, it is possible to solve for optimal leisure, an expression which will be useful later on:
(31) (1  l t )  b [(1  )(1  e  K )  b] .

(d) Now take these optimal choices of consumption and leisure, as well as the production function, and
substitute them all into the value function. It will turn out that the original guess that the value function
is loglinear in capital and technology is valid.

Formally, substitute equations (20), (22) and (31) into equation (17) to obtain

Vt (K t , A t )  ln[Yt (1  e  K )]  b ln b [(1  )(1  e  K )  b]  
 
(32)
e   0   K ln[e  K Yt (1  e  K )]   A  A ln A t .
Substituting the production function, equation (9), into equation (32) and expanding some of the
logarithms yields
5-18 Solutions to Chapter 5

 
Vt (K t , A t )   ln K t  (1  ) ln A t  (1  ) ln l t  ln(1  e  K )  b ln b [(1  )(1  e  K )  b]

(33)  e  0  e  K ln(e  K )  ln(1  e  K )   ln K t  (1  ) ln A t  (1  ) ln l t 
 e  A  A ln A t .
There is no need to substitute in for l t since we already know that it does not depend on Kt or At and it is
really the coefficients on lnKt and lnAt that we are interested in. It is possible to rewrite equation (33) as
(34) Vt (Kt , At ) =  0' +  K' lnKt +  A' lnAt ,
where  K'  (1 + e-  K ),  A'  (1 - )(1 + e-  K ) + e-  A A, and  0'  the rest of the terms that do not
depend upon Kt or At .

(e) In order for our original guess to be correct, we need the coefficient on lnKt in equation (34) to be
equal to  K. That is, we need  K = (1 + e-  K ). Solving for  K yields
(35)  K = /(1 - e- ).
We also need the coefficient on lnAt in equation (34) to be equal to  A. That is, we need
(36)  A = (1 - )(1 + e-  K ) + e-  A A .
Substituting the expression for  K , equation (35), into equation (36) yields
(37)  A = (1 - ){1 + [e- /(1 - e- )]} + e-  A A .
Collecting the terms in  A and simplifying yields
(38)  A (1 - e- A ) = (1 - )/(1 - e- ),
and thus finally,  A is given by
(39)  A = (1 - )/[(1 - e- )(1 - A e- )].

(f) Substitute the value of  K that was derived above into the earlier solutions for Yt /Ct and l t. That is,
substitute equation (35) into equation (23):
Ct 1 1
 
     
(40) ,
Yt 1   e   1   e   1   e   e 1   e
or simply
(41) Ct /Yt = 1 - e- .
This is the same ratio of consumption to output that was obtained by deriving the competitive solution to
this model, with the additional assumption of n = 0 incorporated.

Similarly for labor supply, substitute equation (35) into equation (30):
(1  )
(42) l t  .
(1  )  b [1  (e  (1  e  ))]
We have already worked on an expression like the one in the denominator just above and thus
(43) l t = (1 - )/[(1- ) + b(1 - e- )].
This expression for labor supply is the same as the one that was obtained when deriving the competitive
solution to the model, with the additional assumption of n = 0 incorporated.

Problem 5.12
The derivation of a constant saving rate, s t  s , and a constant labor supply per person, l t  lˆ , does not
depend on the behavior of technology. As the text points out, it is the combination of logarithmic utility,
Cobb-Douglas production, and 100 percent depreciation that causes movements in both technology and
capital to have offsetting income and substitution effects on saving. These assumptions allow us to
derive an expression for the saving rate that does not depend on technology. See equation (5.31) in the
 Solutions to Chapter 5 5-19

text. Once a constant saving rate is established, technology plays no role in the derivation that labor
supply is also constant. That relies on the form of the utility function and on Cobb-Douglas production.
The latter is necessary so that labor's share in income is a constant.

Problem 5.13
(a) Imagine the household increasing its labor supply per member in period t by a small amount l and
using the resulting income to increase its consumption in that period. Household utility and the
instantaneous utility function are given by
t 
(1) U   e t u(c t ,1  l t ) N t H ,
t 0
and
(2) ut = lnct + b(1 - l t )1- /(1 - ).
From equations (1) and (2), the marginal disutility of working in period t is given by
(3) -U/l t = e-t(Nt /H)b(1 - l t )-.
Thus increasing labor supply per member by l has the following utility cost for the household:
(4) Utility Cost = e-t(Nt /H)b(1 - l t )-l.

Since the change raises consumption by wtl, it has the following utility benefit for the household:
(5) Utility Benefit = e-t(Nt /H)(1/ct )wtl.

If the household is behaving optimally, a marginal change of this type must leave expected lifetime utility
unchanged. Thus the utility cost must equal the utility benefit; equating these two expressions gives us
N b N 1
(6) e t t l  e t t w t l ,
H (1  l t )  H ct
or simply
ct w
(7)  t .

(1  l t ) b
Equation (7) relates current leisure and consumption given the wage.

(b) With this change to the model, the saving rate is still constant. The derivation of a constant saving
rate begins from the condition relating current consumption to expectations of future consumption,
1/ct = e-Et [(1 + rt+1 )/ct+1 ]. That relation is not affected by this change to the instantaneous utility
function. The rest of the derivation depends on Cobb-Douglas production and 100 percent depreciation,
but not on the way that utility is affected by leisure. Thus equation (5.33) in the text, s  = en- ,
continues to hold.

(c) Leisure per person is still constant as well. Note that ct in equation (7) is consumption per person. It
can be written as ct  Ct /Nt = (1 - s  )Yt /Nt , where s is the constant saving rate. Taking the natural
logarithm of both sides of equation (7) and substituting for ct yields
 )Yt /Nt ] - ln(1 - l t ) = lnwt - lnb.
(8) ln[(1 - s

Since the production function is Cobb-Douglas, labor's share of output is (1 - ) and thus
5-20 Solutions to Chapter 5

wt l t Nt = (1 - )Yt . Note that we have used Lt  l tNt ; the total amount of labor, Lt , is equal to labor
supply per person, l t , multiplied by the number of people, Nt . Rearranging, we have wt = (1 - )Yt / l t
Nt. Substituting this fact into equation (8) yields
 ) + lnYt - lnNt - ln(1 - l t ) = ln(1 - ) + lnYt - ln l t - lnNt - lnb.
(9) ln(1 - s

Canceling terms and rearranging gives us

(10) ln l t - ln(1 - l t ) = ln(1 - ) - ln(1 - s
 ) - lnb.
Taking the exponential function of both sides of equation (10) yields
lt (1  )
(11)  .
 b(1  ŝ)
(1  l t )
Equation (11) implicitly defines leisure per person as a function of the constants , , b, and s
 . Thus
leisure per person is also a constant.

Problem 5.14
(a) Taking logs of the production function, Yt = Kt(AtLt )1-, gives us
(1) lnYt = lnKt + (1 - )(lnAt + lnLt ).
In the model of Section 5.5 it was shown that labor supply and the saving rate were constant so that
Lt = lˆl Nt and K t  sY
 t 1. Thus we can write
(2) lnYt = ln s + lnYt-1 + (1 - )(lnAt + ln lˆ + lnNt ).
~
Finally we can use the equations for the evolution of technology and population, ln A t  A  gt  A t and
ln N t  N  nt , to obtain
(3) ln Yt   ln ŝ   ln Yt 1  (1  )(A  gt )  (1  )A t  (1  )[ln lˆ  N  nt] .
~

We need to solve for the path that log output would settle down to if there were no technology shocks.
Start by subtracting (n + g)t from both sides of equation (3):
(4) ln Yt  (n  g)t   ln ŝ   ln Yt 1  (n  g)t  (1  )[A  ln lˆ  N  A t ] .
~

Add and subtract (n + g) to the right-hand side of equation (4) to yield
(5) ln Yt  (n  g)t   ln ŝ  (1  )[A  ln lˆ  N]  (n  g)  [ln Yt 1  (n  g)(t  1)]  (1  )A t .
~

Now define Q   ln ŝ  (1  )[A  ln lˆ  N]  (n  g) and use this to rewrite equation (5) as
~
(6) ln Yt  ( n  g) t  Q  [ln Yt 1  ( n  g)( t  1)]  (1  )A t .
~
On a balanced growth path with no shocks to technology, the A s are uniformly 0. In addition, we know
that output will simply grow at rate n + g. With this logarithmic structure that means lnYt - lnYt-1 = n + g
or lnYt-1 = lnYt - (n + g). Substituting these facts into equation (6) yields
(7) lnYt - (n + g)t = Q + [lnYt - (n + g) - (n + g)(t - 1)] = Q + [lnYt - (n + g)t].
Further simplification yields
(8) [lnYt - (n + g)t](1 - ) = Q,
or simply
(9) lnYt* = Q/(1 - ) + (n + g)t.
Equation (9) gives an expression for lnYt*, the path that log output would settle down to if there were
never any technology shocks.
 Solutions to Chapter 5 5-21

~ ~
(b) By definition, Yt  ln Yt  ln Yt * , where lnYt* is the path found in part (a). Thus Yt gives us the
difference between what log output is in any period and what it would have been in the complete absence
of any technology shocks. Substituting for lnYt* from equation (9) yields
~
(10) Yt = lnYt - Q/(1 - ) - (n + g)t.
Note that equation (10) holds every period and so we can write
~
(11) Yt1 = lnYt-1 - Q/(1 - ) - (n + g)(t - 1).
Multiplying both sides of equation (11) by  and solving for lnYt-1 yields
~
(12) lnYt-1 =  Yt1 + [/(1 - )]Q + (n + g)(t - 1).
Substituting equation (12) into equation (3) and then substituting the resulting expression into equation
(10) yields
Yt   ln ŝ  Yt 1   (1  )Q  (n  g)( t  1)
~ ~
(13)
 
 (1  ) A  gt  A t  ln lˆ  N  nt  Q (1  )  (n  g) t.
~

Simplification yields
~ ~ ~
(14) Yt = (n + g)t + (1 - )(n + g)t - (n + g)t +  Yt1 + (1 - ) A t ,
and thus finally
~ ~ ~
(15) Yt =  Yt1 + (1 - ) A t .
Equation (15) is identical to equation (5.40) in the text.

Problem 5.15
(a) (i) The equation of motion for capital is given by
(1) Kt+1 = Kt + Yt - Ct - Gt - Kt ,
or substituting the production function into equation (1), we have
(2) Kt+1 = Kt + Kt(At Lt )1- - Ct - Gt - Kt .
Using equation (1), lnKt+1 /lnKt (holding At , Lt , Ct , and Gt fixed) is
 ln K t 1  K t 1 K t   Yt  Kt
(3)   1    .
 ln K t  K t K t 1   K t  K t 1
Since factors are paid their marginal products, the real interest rate is r t = Yt /Kt -  and thus
 ln K t 1 Kt
(4)  (1  rt ) .
 ln K t K t 1

(a) (ii) On the balanced growth path without shocks, capital grows at rate n + g, so that Kt+1 = en+gKt . In
addition, using r* to denote the balanced-growth-path value of the real interest rate, equation (4) can be
rewritten as
 ln K t 1 Kt 1 r *
(5)  (1  r*) ng  n g .
 ln K t e Kt e
bgp

(b) Using equation (2), lnKt+1 /lnAt (holding Kt , Ct , Gt and Lt fixed) is

 ln K t 1  K t 1 A t  A t  (1  ) K t  (A t L t ) 1
    1
 (1  ) K t A t L t 
(6)  .
 ln A t  A t K t 1  K t 1  K t 1
Using Yt = Kt (At Lt )1-, equation (6) becomes
 ln K t 1 (1  ) Yt
(7)  .
 ln A t K t 1
5-22 Solutions to Chapter 5

On the balanced growth path without shocks, Kt+1 = en+g Kt . Since the production function is Cobb-
Douglas, the amount of income going to capital – which is the marginal product of capital multiplied by
the amount of capital, (r* + )Kt – is equal to Yt . Thus Yt = (r* + )Kt /. Substituting these two facts
into equation (7) yields
 ln K t 1 (1  )( r * ) K t (1  )( r * )
(8)  n g
 n g
.
 ln A t  e K t  e
bgp

Using equation (2), lnKt+1 /lnLt (holding Kt , Ct , Gt and At fixed) is

 ln K t 1  K t 1 L t  L t  (1  ) K t  (A t L t ) 1
  1 
 (1  ) K t A t L t 
(9)  .
 ln L t  L t K t 1  K t 1  K t 1
Comparing equation (9) to equation (6), we can see that lnKt+1 /lnLt = lnKt+1 /lnAt . Thus by
performing the same manipulations as above, we can write
 ln K t 1 (1  )( r * )
(10)  .
 ln L t
bgp e ng

Using equation (2), lnKt+1 /lnGt (holding Kt , Ct , Lt and At fixed) is

 ln K t 1  K t 1 G t Gt
(11)   .
 ln G t  G t K t 1 K t 1
Multiplying and dividing the right-hand side of equation (11) by Yt gives us
 ln K t 1 Yt (G t Yt )
(12)  .
 ln G t K t 1
As explained above, on the balanced growth path without shocks, Kt+1 = en+g Kt and Yt = (r* + )Kt /.
Substituting these two facts into equation (12) and using (G*/Y) to denote the ratio of G to Y on the
balanced growth path yields
 ln K t 1 ( r * ) K t (G * Y) ( r * )(G * Y)
(13)  
 .
 ln G t
bgp e n g
Kt e ng

Using equation (2), lnKt+1 /lnCt (holding Kt , Gt , Lt and At fixed) is

 ln K t 1  K t 1 C t Ct
(14)   .
 ln C t  C t K t 1 K t 1
Using hint (2), we can write this derivative evaluated at the balanced growth path as

(15)
 ln K t 1

 Yt  G * K t  (e ng  1) K t 
.
 ln C t K t 1
bgp
Defining 1  (1 + r*)/en+g, 2  (1 - )(r* + )/(en+g ) and 3  -(r* + )(G*/Y)/(en+g ), we need to
show that lnKt+1 /lnCt evaluated at the balanced growth path is equal to 1 - 1 - 2 - 3 . By definition,
1  r * (1  )( r * ) ( r * )(G * Y)
(16) 1  1   2   3  1  ng   ,
e e ng e ng
or simply
1  r * ( r * ) (G * Y)  (1  )
(17) 1  1  2  3  1  ng  .
e e ng
Note that we can write (r* + ) as
(18) (r* + ) = Yt /Kt = Kt-1 (At Lt )1- = Yt /Kt .
 Solutions to Chapter 5 5-23

Substituting equation (18) into equation (17) gives us

1  r * Yt  (G * Yt )  (1  ) 1  r * G * (1  )Yt
(19) 1  1  2  3  1  ng  n g  1  n g 
e K t e e e n g K t
Obtaining a common denominator and using the fact that on the balanced growth path without shocks,
Kt+1 = en+g Kt , we have
e ng K t  (1  r*) K t  G * Yt  Yt
(20) 1  1   2   3  .
K t 1
From equation (18), Yt = (r* + )Kt . Substituting this into equation (20) yields
e ng K t  K t  r * K t  G * Yt  r * K t  K t
(21) 1  1   2   3  .
K t 1
Collecting terms yields

(22) 1  1   2   3 
 Yt  G * K t  (e ng  1) K t 
.
K t 1
Comparing equations (15) and (22), we have shown that
 ln K t 1
(23)  1  1   2   3 .
 ln C t
bgp

The log linearization is of the form

~  ln K ~  ln K ~  ln K ~
t 1 t 1 t 1
K t 1   K t   A t   L t
 ln K t bgp   ln A t bgp   ln L t bgp 
     
(24) .
 ln K ~  ln K ~
t 1 t 1
 G t   C t
 ln G t bgp   ln C t bgp 
   
Using equations (4), (8), (10), (13), and (23) as well as the definitions of 1 , 2 and 3 to substitute for
the derivatives, we have
~ ~ ~ ~ ~ ~
(25) K t 1  1 K t  2 (A t  L t )  3 G t  (1  1  2  3 ) C t .

~ ~ ~ ~
(c) Substituting equation (5.43), C t  a CK K t  a CA A t  a CG G t , and equation (5.44),
~ ~ ~ ~
L t  a LK K t  a LA A t  a LG G t , into equation (25) yields
~ ~ ~ ~ ~ ~ ~
K t 1  1 K t  2 A t  2 (a LK K t  a LA A t  a LG G t )  3 G t
(26) ~ ~ ~
 (1  1  2  3 ) (a CK K t  a CA A t  a CG G t ).
Collecting terms gives us
K t 1   1  2 a LK  (1  1  2  3 )a CK  K t   2  2 a LA  (1  1  2  3 )a CA  A t
~ ~ ~
(27)
  2 a LG  3  (1  1  2  3 )a CG  G t .
~

Defining bKK  1 + 2 aLK + (1 - 1 - 2 - 3 )aCK , bKA  2 (1 + aLA ) + (1 - 1 - 2 - 3 )aCA and

bKG  2 aLG + 3 + (1 - 1 - 2 - 3 )aCG , equation (27) can be rewritten as
~ ~ ~ ~
(28) K t 1  b KK K t  b KA A t  b KG G t .
Equation (28) is identical to equation (5.53) in the text.