SOLUTIONS TO CHAPTER 12

Problem 12.1
(a) From mt – pt = c – b(Et [pt+1] – pt ), collecting the terms in pt yields
(1) pt (1 + b) = mt – c + bEt [pt+1],
and so pt is given by
 b   1 
(2) p t   E t [p t 1 ]    m t  c .
1 b  1 b 

(b) Equation (2) holds in all periods so that we can write pt+1 as
 b   1 
(3) p t 1   E t 1[p t  2 ]    m t 1  c  .
1 b  1 b 
Taking the expected value, as of time t, of both sides of equation (3) yields
 b   1 
(4) E t [p t 1 ]   E t [p t  2 ]    E t [m t 1 ]  c ,
1 b  1 b 
where we have used the law of iterated projections, which states that Et [Et+1 [pt+2]] = Et [pt+2]. If this did
not hold, individuals would be expecting to revise their estimate of pt+2 either up or down, which would
imply that their original estimate was not rational.

(c) Substituting equation (4) into equation (2) yields

2
 b   1   b  
(5) p t    E t [p t  2 ]   m t  c   E t [m t 1 ]  c  .
1 b   1  b  1 b  
Again using the fact that equation (2) holds in all periods, we can write pt+2 as
 b   1 
(6) p t  2   E t  2 [p t  3 ]   m t  2  c .
1 b  1 b 
Taking the expected value, as of time t, of both sides of equation (6) gives us
 b   1 
(7) E t [p t  2 ]   E t [p t  3 ]   E t [m t  2 ]  c ,
1 b  1 b 
where we have again used the law of iterated projections so that Et [Et+2 [pt+3]] = Et [pt+3]. Substituting
equation (7) into equation (5) leaves us with
 1  
3 2
 b   b   b 
(8) p t    E t [p t  3 ]   m t  c   E t [m t 1 ]  c     E t [m t  2 ]  c  .
1 b   1  b  1 b  1 b  
The pattern should now be clear. We can write pt as the following infinite sum:
 1  
2 3
 b   b   b 
(9) p t   m t  c   E t [m t 1 ]  c     E t [m t  2 ]  c    E t [m t  3 ]  c  
 1  b  1 b  1 b  1 b  

(d) With output and the real interest rate constant, the price level must adjust to clear the money market.
If mt+i is higher, pt+i will need to be higher to clear the money market. Thus if individuals expect, in
period mt+i-1 , that mt+i will be higher they will also expect pt+i to be higher. Thus in period t + i – 1,
expected inflation will be higher. This reduces real money demand in period t + i – 1. For a given value
of mt+i-1 , this means that pt+i-1 will need to rise to clear the money market. Now go back one more period.
Suppose that individuals expect, in period t + i – 2, that mt+i will be higher. Then they expect, through
the reasoning above, that pt+i-1 will be higher. Thus expected inflation in t + i – 2 will be higher, real
money demand will be lower and thus pt+i-2 will be need to be higher to clear the money market.
12-2 Solutions to Chapter 12

Reasoning backward, as soon as people expect the nominal money supply to rise in some future period,
the price level will rise in the current period.

(e) Equation (9) can be written using summation notation as

1  b 
i
(10) p t    E t [m t  i ]  c  .
1  b i  0 1  b 
Substituting the assumption that Et [mt+i] = mt + gi into equation (10) yields
1   b i 1    b i   b i 
(11) p t    ( mt  gi  c)   ( mt  c)    g i  .
1  b i0 1  b  1 b  i0 1  b  i0  1  b  
Now we can use the facts that
  b i  b   b 2 1
(12)    1       1 b,
i0 1  b  1  b  1  b  1  [ b / (1  b)]
and
  b i b / (1  b) b / (1  b)
(13)  i      b(1  b) .
i0  1  b  {1  [ b / (1  b)]}2 1 / (1  b) 2
Equation (13) uses the result that
 x
(14)  ix i  .
i0 (1  x) 2
A (not entirely rigorous) way to see why (14) and thus (13) hold is to note that with x < 1, we have
1
(15) 1  x  x 2  x 3   .
1 x
Differentiating both sides of equation (15) with respect to x (which means differentiating term by term on
the left-hand side) gives us
1
(16) 1  2 x  3x 2  .
(1  x) 2
Multiplying both sides of equation (16) by x yields
x
(17) x  2 x 2  3x 3  .
(1  x) 2
Note that (17) and (14) are equivalent; the left-hand side of equation (14) is simply the left-hand side of
(17) written in summation notation.

Substituting equations (12) and (13) into equation (11) yields

1
(18) p t 
1 b
 ( m t  c)(1  b)  gb(1  b) .
Thus the price level is given by
(19) pt = (mt – c) + bg.

To see how the price level changes when money growth changes, use equation (19) to take the derivative
of pt with respect to g:
 pt
(20)  b  0.
g
Thus a rise in money growth, even without a rise in the level of the current period's money supply, causes
an upward jump in the current price level.
 Solutions to Chapter 12 12-3

Problem 12.2
(a) Substituting the normalized, flexible-price level of output, y0 = 0, into the IS equation, y0 = c – ar0 ,
gives us 0 = c – ar0 . Solving for the real interest rate in period 0 yields
(1) r0 = c/a .
Since the nominal money stock is expected to be constant, the price level is expected to be constant and
thus expected inflation from period 0 to period 1 is
(2) E0 [p1 ] – p0 = 0.
The nominal interest rate in period 0, i0 = r0 + [E0 [p1 ] – p0 ], is simply equal to the real interest rate:
(3) i0 = c/a.
Finally, substituting the assumptions that m0 = 0 and y0 = 0 as well as equation (3) into the condition for
equilibrium in the money market, m0 – p0 = b + hy0 – ki0 , yields – p0 = b – (ck/a) or simply
(4) p0 = (ck/a) – b.

(b) In period 2, the economy is once again at the flexible-price equilibrium level of output, which is 0.
Substituting this fact into the IS equation allows us to solve for the real interest rate in period 2:
(5) r2 = c/a.
Since expected inflation from period 2 to period 3 is equal to g – the price level is expected to rise by the
same amount as the nominal money supply each period – the nominal interest rate in period 2 is given by
(6) i2 = (c/a) + g.
Since m was equal to 0 in period 0 and then increases by g in each following period, the nominal money
supply in period 2 is m2 = 2g. Substituting this fact as well as y2 = 0 and i2 = (c/a) + g into the condition
for equilibrium in the money market leaves us with
(7) 2g - p2 = b – (ck/a) – kg.
Solving for p2 gives us
(8) p2 = – b + (ck/a) + (2 + k)g.

(c) The price level is completely unresponsive to unanticipated monetary shocks for one period. Thus
the price level in period 1 does not change from its period 0 value and hence
(9) p1 = (ck/a) – b.
The expectation of inflation from period 1 to period 2, E1 [p2 ] – p1 , is therefore
(10) E1 [p2 ] – p1 = – b + (ck/a) + (2 + k)g – (ck/a) + b = (2 + k)g,
where we have used equations (8) and (9) to substitute for p2 and p1 .

Now substitute the IS equation, y1 = c – ar1 , into the money-market-equilibrium condition,

m1 – p1 = b + hy1 – ki1 , to obtain
(11) m1 – p1 = b + hc – ahr1 – ki1 .
By assumption, the nominal money supply in period 1 is g. In addition, i1 = r1 + [E1 [p2 ] – p1 ], which,
using equation (10), is equivalent to i1 = r1 + (2 + k)g. Substituting these facts as well as equation (9) for
the price level into equation (11) gives us
(12) g – (ck/a) + b = b + hc – ahr1 – kr1 – (2 + k)kg.
Simplifying and collecting the terms in r1 yields
(13) r1 [ah + k] = hc + (ck/a) – g – (2 + k)kg.
Thus the real interest rate in period 1 is given by
hc  (ck a )  g  (2  k ) kg
(14) r1  .
ah  k
Finally, substituting equations (10) and (14) into i1 = r1 + [E1 [p2 ] – p1 ] gives us
hc  (ck a )  g  (2  k ) kg hc  (ck a )  g  (2  k ) kg  (2  k )ahg  (2  k ) kg
(15) i1   (2  k )g  .
ah  k ah  k
Thus the nominal interest rate in period 1 is given by
12-4 Solutions to Chapter 12

hc  (ck a )  g  (2  k )ahg
(16) i1  .
ah  k

(d) Using equations (16) and (3), the change in the nominal interest rate from period 0 to period 1 is
hc  (ck a )  g  (2  k )ahg hc  (ck a )  g  (2  k )ahg  hc  (ck a )
(17) i1  i 0   (c a )  .
ah  k ah  k
Simplifying yields
(2  k )ahg  g
(18) i1  i 0  .
ah  k
We can determine the condition required of the parameters for the nominal interest rate to fall from
period 0 to period 1; that is, for i1 – i0 < 0. From equation (18), this condition is
g  (2  k )ah  1
(19)  0,
ah  k
or simply
(20) (2 + k)ah < 1.
The smaller is a (the elasticity of output with respect to changes in the real interest rate), the smaller is h
(the income elasticity of real money demand) and the smaller is k (the interest semi-elasticity of real
money demand), the more likely it is for the condition in (20) to be satisfied and thus the more likely it is
for the nominal interest rate to fall in response to the monetary expansion.

For the nominal interest rate, i = r + e, to fall, we need the liquidity effect to outweigh the expected
inflation effect. That is, we need the real interest rate to fall by more than expected inflation rises. With
the price level fixed by assumption in period 1, y and i must adjust to ensure money market equilibrium.
If k is small, changes in i will not affect real money demand very much. We need y to rise to increase
real money demand and get it equal to the new higher real money stock. If h is small, we need y to rise a
lot to accomplish this. If y is to rise a lot, we need – from the IS equation – the real interest rate to fall a
lot. If furthermore, a is small, we need r to fall a lot just to generate an increase in output. Thus small
values of k, h, and a all work to make the drop in r larger and thus make it more likely that i will fall.

Problem 12.3
(a) Any shock to the nominal money supply in period t + 1 is fully reflected in the price level by period
t + 2. That is, the only reason the price level will change from period t + 1 to period t + 2 is if there is a
non-zero realization of u in period t + 1. From the law of iterated projections, we have
(1) Et [Et+1 [pt+2 ] – pt+1 ] = Et [pt+2 - pt+1 ].
Since the expected value, as of period t, of ut+1 is zero, the price level is not expected to change from
period t + 1 to period t + 2. Thus
(2) Et [Et+1 [pt+2 ] – pt+1 ] = 0.

Since the condition for money market equilibrium must hold each period, we can write
(3) mt+1 – pt+1 = b + hyt+1 – krt+1 – k(Et+1 [pt+2 ] – pt+1 ),
where we have substituted in for it+1 = rt+1 + (Et+1 [pt+2 ] – pt+1 ). Taking the expected value of both sides
of equation (3) yields
(4) E t [m t 1 ]  E t [p t 1 ]  b  hyn  kr n ,
where we have used the result from equation (2) that Et [Et+1 [pt+2 ] – pt+1 ] = 0. In addition, since yt+1 and
rt+1 will only depend on the ut+1 shock, which is expected to be zero, they are expected to be equal to their
average values.
 Solutions to Chapter 12 12-5

(b) Rearranging equation (4), we have

(5) E t [p t 1 ]  E t [m t 1 ]  b  hyn  kr n .
Since mt+1 = mt + ut+1 , Et [mt+1]= mt . Using this fact and subtracting pt from both sides of equation (5)
yields
(6) E t [p t 1 ]  p t  (m t  p t )  b  hyn  kr n .
As explained in part (a), expected inflation is equal to ut and thus we can write
(7) u t  (m t  p t )  b  hyn  krn .
Substituting mt = mt-1 + ut into equation (7) and rearranging to solve for pt yields
(8) p t  m t 1  b  hyn  krn .

The next step is to solve for output in period t. Rearranging the condition for money market equilibrium
to solve for it yields
(9) i t  [ b  hy t  ( mt  p t )] / k .
From equation (7), we have
(10) (m t  p t )  u t  b  hyn  krn .
Substituting equation (10) into equation (9) gives us
b  hyt  u t  b  hyn  kr n h ( y t  y n )  kr n  u t
(11) i t   .
k k
Substituting equation (11) for it and using the fact that te = ut , the IS equation becomes
 h ( y t  y n )  kr n  u t 
(12) y t  c  a    au t .
 k 
Collecting the terms in yt , we have
 k  ah  ahyn  akrn  au t
(13)   y t  c   au t ,
 k  k
which implies
kc  ahyn  akrn  au t  kaut
(14) y t  ,
k  ah
and thus output in period t is given by
kc  a[hyn  kr n  (1  k )u t ]
(15) y t  .
k  ah
To determine the real interest rate, rearrange the IS equation to obtain
(16) rt = (c/a) - (yt /a).
Substituting equation (15) into equation (16) yields
c kc  a[hyn  kr n  (1  k )u t ]
(17) rt   ,
a a (k  ah)
which implies
ck  cah  kc  a[hyn  kr n  (1  k )u t ] ch  [hyn  kr n  (1  k )u t ]
(18) rt   .
a (k  ah) k  ah
Thus the real interest rate in period t is
h (c  y n )  kr n  (1  k )u t
(19) rt  .
k  ah
12-6 Solutions to Chapter 12

The nominal interest rate is it = rt + te , where te = ut :

(20) it = rt + ut .
Substituting equation (19) into equation (20) gives us
h (c  y n )  kr n  (1  k )u t  (k  ah)u t h (c  y n )  kr n  (ah  1)u t
(21) i t   .
k  ah k  ah
(c) From equation (21), with te = ut , we have
h (c  y n )  kr n ah  1 e
(22) i t   t .
k  ah k  ah
From equation (22), we can see that changes in expected inflation are not reflected one-for-one in the
nominal rate. This is because prices are completely unresponsive to the monetary disturbance for one
period. This means that, in general, output and the nominal interest rate will adjust to clear the money
market. For output to change, the real interest rate must change and therefore, in general, the nominal
interest rate will not move one-for-one with inflation.

Problem 12.4
(a) Under rational expectations,
(1) t+1 = Et [t+1]+ t+1 ,
where t+1 is a disturbance that is uncorrelated with anything known at t. Now consider the regression:
(2) it = a + bt+1 + et .
Using the hint in the question, the OLS estimator of b is given by
cov(i t ,  t 1 )
(3) b  .
var( t 1 )
Using it = rt + Et [t+1] and equation (1), we can write the covariance in the numerator as
(4) cov(it , t+1 ) = cov(rt + Et [t+1], Et [t+1] + t+1 ).
Since rt and Et [t+1] are uncorrelated and t+1 is uncorrelated with anything known at t, this implies
(5) cov(it , t+1 ) = var(Et [t+1]).
Again using equation (1), the variance in the denominator of equation (3) can be written as
(6) var(t+1 ) = var(Et [t+1] + t+1 ) = var(Et [t+1]) + var(t+1 ),
where we have used the fact that cov(Et [t+1], t+1 ) = 0. Substituting equations (5) and (6) into equation
(3) allows us to write the OLS estimator as
var(E t [ t 1 ])
(7) b̂   1.
var(E t [ t 1 ])  var( t 1 )
The hypothesis that the real interest rate is constant, so that changes in expected inflation cause one-for-
one movements in the nominal interest rate, only predicts that the coefficient on t+1 should be positive
and less than one, not that it will take on any specific value.

(b) Now consider a regression of the form

(8) t+1 = a' + b' it + et'.
The OLS estimator of b' is of the form
cov(i t ,  t 1 )
(9) b   .
var(i t )
The covariance in the numerator of equation (9) is still given by equation (5). Since it = rt + Et [t+1], we
can write the denominator of equation (9) as
(10) var(it ) = var(r) + var(Et [t+1]),
where we have used the fact that cov(r, Et [t+1]) = 0. Substituting equations (5) and (10) into equation
(9) gives us the following OLS estimator:
 Solutions to Chapter 12 12-7

var(E t [ t 1 ])
(11) b̂  .
var(r )  var(E t [ t 1 ])
The hypothesis that the real interest rate is constant, so that var(r) = 0, predicts a coefficient of one on i t .

(c) Consider the following regression:

(12) it = a + b0 t + b1 t-1 + ... + bn t-n + t .
So, for example, the coefficient b0 represents the direct impact on it of a change in t , holding the other
's constant.

Now suppose that the behavior of actual inflation is given by

(13) t = t-1 + et .
If it = r + Et [t+1], with r constant, changes in expected inflation should cause one-for-one movements in
it. Thus since t+1 = t + et+1 , a change in t of t will cause Et [t+1], and thus it , to change by t .
So we would expect b0 =  in the above regression.

But now, controlling for t , the other variables – t-1 , ..., t-n – provide no new information about t+1 .
Any effect that t-1 , say, has on t+1 is already captured indirectly by t-1's impact on t . Thus we would
expect b1 = ... = bn = 0 in the above regression. Thus the claim is incorrect since we would have
b0 + b1 + ... + bn = , not b0 + b1 + ... + bn = 1.

Problem 12.5
(a) We have t = pt – pt-1 and te = pte – pt-1 . Thus t – te = (pt – pt-1 ) – (pte – pt-1 ) = pt – pte . We can
therefore write the Lucas supply function as
(1) yt = yn + b(pt – pte ).
Setting aggregate supply equal to aggregate demand (which is given by yt = mt – pt ) gives us
(2) mt – pt = yn + b(pt – pte ).
Solving equation (2) for pt yields
1 b e 1 n
(3) p t  mt  pt  y .
1 b 1 b 1 b
With rational expectations, the expected value of both sides of equation (3) must be equal. Hence
(4) p et 
1
m t 1  a   b pet  1 y n ,
1 b 1 b 1 b
where we have used the fact that the expected value of mt = mt-1 + a + t is equal to mt-1 + a since  is
white noise. Subtracting equation (4) from equation (3) yields
1 1 1
(5) p t  p et 
1 b
mt 
1 b
 m t 1  a  
1 b
 m t  m t1  a .
Substituting equation (5) into equation (1) gives us
(6) y t  y n 
b
m t  m t 1  a  .
1 b

(b) From equation (6), we can see that we also need to know a, as well as mt and mt-1 , in order to
determine the current level of output. Intuitively, equation (6) says that only unexpected money affects
output since the difference between mt and (mt-1 + a) is the random shock, t . However, if we don't know
a, we cannot determine how much of the change in the nominal money stock from period t – 1 to period t
was due to a (and thus was expected) and how much was due to  (and thus was unexpected).
12-8 Solutions to Chapter 12

(c) Again, it must be true that with rational expectations, the expected value of both sides of equation (3)
must be equal. However, the expected value of mt is now mt-1 + (0) + (1 – )a = mt-1 + (1 – )a since
private agents believe that the probability that a = 0 is . Thus
(7) p et 
1
m t 1  (1  )a   b pet  1 y n .
1 b 1 b 1 b
Subtracting equation (7) from equation (3) yields
1
(8) p t  p et 
1 b
 m t  m t1  (1  )a  .
Substituting equation (8) into equation (1) gives us
(9) y t  y n 
b
m t  m t 1  (1  )a  .
1 b

(d) Equation (6) holds in any period in which there is no regime shift. Thus if there is no regime shift in
period t – 1, we can write
(10) y t 1  y n 
b
m t 1  m t  2  a  .
1 b
Subtracting equation (10) from equation (6) yields
b
(11) y t  y t 1 
1 b
 
 m t  m t1    m t1  m t2  .
Defining yt  yt – yt-1 and mt  mt – mt-1 , we have
  m t   m t1 .
b
(12)  y t 
1 b
Equation (12) states that in the absence of regime shifts, output growth is determined by the change in
money growth.

If there is a regime shift in period t, equation (9) holds. Subtracting equation (10) from equation (9)
yields
b b
(13) y t  y t 1 
1 b
  
m t  m t 1    m t 1  m t 2  
1 b
 a  (1  )a  ,
or simply
ab
  m t   m t1 .
b
(14)  y t  
1 b 1 b
Under the null hypothesis of no credibility of the announcement of the regime shift,  = 0, the first term
on the right-hand side of equation (14) is equal to zero. Thus if the announcement is not believed,
equations (14) and (12) are identical. Thus we can run a regression of yt on [mt – mt-1 ] and a dummy
variable that equals one in the period of a regime shift. The coefficient on that dummy variable will
reflect the amount of credibility of the policymaker's announcement. In fact, since we will have an
estimate of b/(1 + b) and can determine a (the average change in the money stock before the regime
shift), we can calculate an estimate of  from the coefficient on the dummy variable.

Problem 12.6
(a) (i) The one-period nominal interest rate is given by it1 = Et [t+1] since the real interest rate is
assumed constant at zero. Since t+1 = mt+1 , we have
(1) it1 = Et [mt+1].
Since money growth is given by
(2) mt = kmt-1 + t ,
and since equation (2) holds in all periods, we can write
 Solutions to Chapter 12 12-9

(3) mt+1 = kmt + t+1 .

Substituting equation (3) into equation (1), we have
(4) it1 = Et [kmt + t+1 ] = kmt ,
where we have used the fact that mt is known as of time t and Et [t+1 ] = 0.

(a) (ii) The expectation, as of time t, of the nominal interest rate from period t + 1 to t + 2 is
(5) Et [it+11] = Et [t+2] = Et [mt+2].
Since equation (2) holds every period, we can write
(6) mt+2 = kmt+1 + t+2 .
Substituting equation (3) into equation (6) gives us mt+2 as a function of mt :
(7) mt+2 = k2 mt + kt+1 + t+2 .
Substituting equation (7) into equation (5) gives us
(8) Et [it+11] = Et [k2 mt + kt+1 + t+2 ] = k2 mt ,
where we have used the fact that mt is known at t and the 's are mean-zero disturbances.

(a) (iii) Under the rational-expectations theory of the term structure, the two-period interest rate is
(9) it2 = (i1 + Et [it+11])/2.
Substituting equation (8) into equation (9), we have
(10) it2 = (it1 + k2 mt )/2.
From equation (4), kmt = it1 and so equation (10) can be rewritten as
(11) it2 = (it1 + kit1 )/2 = it1 (1 + k)/2.

(a) (iv) From equation (11), a rise in k will increase the two-period interest rate, it2 , for any given one-
period rate. For a given level of inflation in period t, expected inflation for period t + 1 will now be
higher. Thus for a given one-period interest rate in t, the one-period rate in t + 1 is expected to be higher.
Therefore it2 , which is the average of the one-period rate in t and the expected one-period rate in t + 1,
will now be higher for a given it1 .

Note that as k goes to one, so that money growth and thus inflation approach a random walk, the two-
period interest rate becomes equal to the one-period interest rate. That is because with inflation a
random walk, next period's inflation (and thus next period's one-period nominal rate) is expected to be
equal to this period's inflation (and thus this period's one-period nominal rate).

(b) (i) Equation (4) holds in all periods and thus the actual one-period interest rate in t + 1 is
(12) it+11 = kmt+1 .
Substituting equation (3) into equation (12) yields
(13) it+11 = k2 mt + kt+1 .
Thus
(14) it+11 – it1 = k2mt + kt+1 – kmt = k(k – 1)mt + kt+1 .
From equation (11), we can write
(15) it2 – it1 = [it1 (1 + k)/2] – it1 = [it1 (1 + k – 2)/2],
or substituting in for it1 = kmt , we have
k ( k  1)m t
(16) i 2t  i1t  .
2

Using the hint in the question, the OLS estimator of b in the following regression:
(17) it+11 – it1 = a + b[it2 – it1 ] + et+1 ,
is given by
12-10 Solutions to Chapter 12

 cov[(i1t 1  i1t ), (i 2t  i1t )]

(18) b  .
var(i 2t  i1t )
Using equations (14) and (16), the covariance in the numerator of (18) can be written as
 k ( k  1) m t 
(19) cov[(i1t 1  i1t ), (i 2t  i1t )]  cov  k ( k  1) m t  k t 1 ,  .
 2
Since  is white noise and var(mt ) = 2, we have
k 2 ( k  1) 2
(20) cov[(i1t 1  i1t ), (i 2t  i1t )] 
 2 .
2
Using equation (16), the variance in the denominator of equation (18) can be written as
2 1 k 2 ( k  1) 2 2
(21) var(i t  i t )   .
4
Substituting equations (20) and (21) into equation (18) gives us
k 2 ( k  1) 2 2

 2
(22) b  2  2.
k ( k  1) 2 2

4

(b) (ii) With the time-varying term premium, equation (16) becomes
k ( k  1)m t
(23) i 2t  i1t   t .
2
Using equations (14) and (23), the covariance in the numerator of equation (18) is now given by
 k ( k  1) m t 
(24) cov[(i1t 1  i1t ), (i 2t  i1t )]  cov  k ( k  1) m t  k t 1 ,  t  .
 2 
Since  and  are white noise, this is simply
k 2 ( k  1) 2 2
(25) cov[(i1t 1  i1t ), (i 2t  i1t )]   .
2
This covariance is the same as it was without the time-varying term premium. The variance of
(it2 – it1 ) will change, however. It is now given by
2 1 k 2 ( k  1) 2 2
(26) var(i t  i t )    2 ,
4
where we have used the fact that the covariance between  and  is zero.

Substituting equations (25) and (26) into equation (18) gives us

k 2 ( k  1) 2 2

2 2
(27) b   .
 k 2 ( k  1) 2   4 2 

  2    2 1   2 2

 4   k ( k  1) 

(b) (iii) Since k2 ( k – 1)2 reaches a maximum at k = 1/2, the OLS estimator is highest when k = 1/2. For
k > 1/2, an increase in k (more persistent money growth and inflation), reduces the value of the OLS
estimator. As k approaches one, so that money growth, inflation and thus the one-period nominal interest
rate all approach random walks, the OLS estimator goes to zero.
 Solutions to Chapter 12 12-11

Problem 12.7
The new aggregate supply equation can be written as
y t 1   t .
(1)  t   t 1  ~
Since the change to the model described in the problem affects just the AS equation, we must only
consider how the change affects E[π2]. Substituting equation (12.23), ~y t  q t   IS Y
t   t , into our
new AS equation lagged forward one period yields

(2)  t 1  (1  q) t   IS Y
t   t   t 1 .


t ,  t , and  t 1 are all uncorrelated
Next, we want to take the expectation of the square of (1). Since  IS Y

with each other and with πt and have a mean of zero, all their cross terms become zero. Therefore,
(3) E[ 2t 1 ]  (1  q) 2 E[ 2t ]   2  2IS   2  2Y   2 ,
where  2 is the variance of   .

Since in this model E[2t 1]  E[2t ] in the long run, we can substitute and solve for E[π2], giving us
 2 (2IS  2Y )  2  2 (2IS  2Y )  2
(4) E[2 ]   .
1  (1  q) 2 q(2  q)
Having found E[π2] in terms of q, we turn our attention to finding E[(y – y*)2]. Since the problem
changed only the aggregate supply equation, the equation for E[(y – y*)2] given by (12.26) is still valid.
From (4) and (12.26), we can obtain q* by taking the first order condition of E[(y – y*)2] + λE[π2] with
respect to q. That is, we have
(5) E[(y – y*)2] + λE[π2] = 2  q 2 E[ 2 ]   2Y   2IS  E[ 2 ]  (q 2  )E[ 2 ]  2   2Y   2IS .
The first-order condition for q is thus given by
 E[ 2 ]
(6) (q 2  )  2qE[ 2 ]  0 ,
q
or equivalently
E[ 2 ]
(7) (q 2  )  2q 0.
 E[ 2 ]  q
Defining K   2 ( 2IS   2Y )   2 , we have
K
(8) E[ 2 ]  ,
q(2  q)
and
 E[ 2 ]  K (2  2 2 q)
(9)  .
q [q(2  q)]2
Thus, the ratio is given by
K
E[ ]2
q(2  q) q(2  q)
(10)   .
2 2
 E[ ]  q  K (2  2 q) (2  2 2 q)
[q(2  q)]2
Substituting equation (10) into the first-order condition, equation (7), gives us
12-12 Solutions to Chapter 12

q(2  q)
(11) (q 2  )  2q 0.
(2  2 2 q)
We can rewrite equation (11) as
(12) (q 2  )(2  2 2 q)  2q[q(2  q)]  0 .
Dividing both sides by 2α results in
(13) (q 2  )(1  q)  q 2 (2  q)  0 .
Expanding equation (13) yields
(14) q 2    q 3   q  2q 2  q 3  0 .
Thus, we have the following polynomial in q:
(15)  q 2   q    0 .
Using the quadratic formula, the solution for q* is given by
    2 2  4
(16) q*  .
2
We find that q* is unchanged from the one described by (12.27). Note that once again, we throw out the
negative root since a negative q* causes the variances of y and π to be infinite.

Problem 12.8
(a) When ϕπ = 1, the matrix A simplifies to
1 0 
(1) A   .
  
The characteristic equation of A is defined by det(A – tI) = 0, where I is the identity matrix and the
solutions of the equation, t, are the eigenvalues. Recall that the determinant of a 2 x 2 matrix, such as
a b 
(2) B   ,
c d 
is given by
(3) B  ad  bc.
Therefore, the characteristic equation of A is given by
1 t 0
(4)  (1  t )(  t )  0
 t
The characteristic equation implies that the eigenvalues are t 1 = β and t2 = 1.

To analyze the self-fulfilling movements, we first make the observations that

(5) E t [~
y t 1 ]  ~
yt ,
and
(6) Et [πt+1] = λπt .
Substituting equations (1), (5), and (6) into equation (12.41) from the text and performing the
multiplication yields
~
y t  1 0  ~ yt   ~
yt 
(7)        ~ .
 t     t  y t   t 
From equation (7), we obtain two equalities. The first is given by
(8) ~
yt  ~yt .
There are two possibilities that satisfy equation (8): ~
yt  0 ; or λ = 1. The second equality that we obtain
from equation (7) is
 Solutions to Chapter 12 12-13

(9)  t  ~
y t   t .

Consider the first possible solution for equation (8). Substituting ~

yt  0 into (9), we obtain
(10)  t   t ,
or simply
(11) λ = 1/β.
Since β < 1, equation (11) implies λ > 1, which contradicts the assumption of the problem. Therefore we
eliminate the possibility that ~yt  0 and conclude λ = 1. We can now substitute λ = 1, πt = λt Z, and
ỹt = cλtZ into equation (9) to obtain
(12) Z = κcZ + βZ.
Equation (12) implies that
(13) c = (1 – β)/κ.
Thus the self-fulfilling movements of π and ỹ result in both variables remaining constant at values Z and
[(1 – β)/κ]Z, respectively.

(b) In part (a), we found that when ϕπ = 1, one of the eigenvalues is 1 and the other is strictly inside the
unit circle (since β < 1). According to equation (12.42), the eigenvalues are continuous functions of ϕ π.
Consequently, the derivative of equation (12.42) evaluated at ϕπ = 1 will determine whether the
eigenvalues will increase or decrease for values of ϕπ that are infinitesimally close to 1. Before we take
the derivative, note that we need only focus on the eigenvalue given by
1      (1    ) 2  4
(14)   ,
2
since this is the solution that equals 1 when ϕπ = 1. Taking the derivative of both sides of equation (14)
with respect to ϕπ gives us
d 1  1 
(15)   [(1    ) 2  4] 1 / 2 2(1    )  .
d 2  4 
Since α ≡ κ(1 – ϕπ)/θ, we have
 
(16)  .
  
Substituting equation (16) into (15) and simplifying leaves us with
d  1 
(17)   [(1    ) 2  4] 1 / 2(1    )
d 2 2 
We want to evaluate the derivative in (17) at ϕπ = 1. Substituting the fact that ϕπ = 1 implies that α = 0
into equation (17) yields
d  1 
(18)    [(1  ) 2  4] 1 / 2(1  ) .
d  1 2 2 

Since (1  ) 2  4  (1  ) 2 , equation (18) simplifies to

d  1 (1  ) 
(19)   .
d  1 2 2 (1  ) 
Since both terms on the right-hand side of equation (19) are negative, we conclude that d / d is
negative when ϕπ = 1. Consequently, for values of ϕπ infinitesimally larger than 1, both eigenvalues are
inside the unit circle whereas for values of ϕπ infinitesimally smaller than 1, one of the eigenvalues is
outside the unit circle.
12-14 Solutions to Chapter 12

(c) Using the given information, the matrix A simplifies to

  2(1  ) 
1
(20) A    .
 
    2(1  )
To find the eigenvalues, we must solve for the values of t in the characteristic equation of A, which is
det(A – tI) = 0, or
 2(1  )
1 t
(21)   0.
   2(1  )  t
Calculating the determinant gives us
(22) (1  t )[  2(1  )  t ]  [2(1  ) / ]  0 .
Simplifying and expanding the terms on the left-hand-side of equation (22) yields
(23) (1  t )  2(1  )  2t (1  )  t (1  t )  2(1  )  0 .
Simplifying gives us
(24)   t  2t  2t  t  t 2  0 .
Collecting terms leaves us with the following quadratic equation in t:
(25) t 2  (1  )t    0 .
Using the quadratic formula to find the roots of equation (25) gives us
 (1  )  (1  ) 2  4
(26) t 
2
2
Since (1  )  4  (1  ) 2 , this simplifies to
 (1  )  (1  )
(27) t  .
2
Thus the eigenvalues are t1 = –1 and t2 = –β.

We can determine the values of λ and c that satisfy equation (12.41) in the text by following the same
steps as in the solution to part (a). Substituting our known values into (12.41) yields
 2(1  )   t 1   2t 1Z(1  ) 
y t  ct Z 1
~  c Z  ct 1
Z  
(28)          .
 t  t Z      2(1  ) t 1Z   t 1   
  t 1
 c Z   Z  2(1  ) Z
t 1

The second row of the system described by (28) implies that

(29) t Z  ct 1Z  t 1Z  2(1  )t 1Z .
Dividing both sides of equation (29) by λtZ and simplifying gives us
(30) c  1    2(1  ) .
Equation (30) simplifies to
1  (2  )
(31) c  .

The first row of the system described by (28) implies that
2t 1Z(1  )
(32) ct Z  ct 1Z  .

Dividing both sides of equation (32) by λtZ gives us
2(1  )
(33) c  c  .

 Solutions to Chapter 12 12-15

Substituting equation (31) into equation (33) yields

1  (2  ) 1  (2  ) 2(1  )
(34)   .
  
Multiplying both sides of equation (34) by κ gives us
1  (2  )
(35)  1  (2  )  2(1  ) .

Multiplying both sides of equation (35) by λ and simplifying gives us
(36) 1  (2  )    22  2  22  22 .
Collecting terms yields
(37) 1  (2  )    2 ,
or simply
(38) 2  (1  )  1  0 .
Using the quadratic formula to solve equation (38) leaves us with
 (1  )  (1  ) 2  4

(39) 2
 (1  )  (1  )
 .
2
The two possible solutions are λ1 = –1 and λ2 = –1/β. Since –1/β < –1, this answer violates the condition
that   1 and so we can eliminate it. We can therefore substitute λ1 = –1 into equation (31) to obtain
1  (2  )
(40) c   ,

or simply
1 
(41) c  .


Using the values of λ and c that we have solved for, we can now conclude that the self-fulfilling
movements take the form
(42)  t  (1) t Z ,
and
1  
(43) ~
y t  (1) t  Z.
  
Therefore π and ỹ oscillate between the values ± Z and ± (1 + β)Z/κ, respectively.

Problem 12.9
(a) When the policymaker fixes i, the money-market equilibrium condition is irrelevant. Equilibrium
output is determined by the IS curve and the fixed nominal interest rate, i̇̅ . Substituting i̇̅ into the IS curve
yields
(1) y = c – a i̇̅ + 1.
The variance of y is simply
(2) var(y) = var(1 ) = 12.

(b) When the policymaker fixes m, the equilibrium level of output is determined by the intersection of
the IS and money-market equilibrium equations. Rearranging the IS curve to solve for i gives us
(3) i = (c + 1 – y)/a .
Substituting equation (3) and the assumption that m = ṁ̅ into the money-market equilibrium equation,
12-16 Solutions to Chapter 12

m – p = hy – ki + 2 , gives us
(4) ṁ̅ – p = hy – [k(c + 1 – y)/a] + 2 = [h + (k/a)]y – (kc/a) – (k/a)1 + 2 .
Solving for y yields
m  p  (kc a )  (k a )1   2 a ( m  p)  kc  k1  a 2
(5) y   .
h  (k a ) ah  k
The variance of y is
2 2
 k  2  a  2
(6) var(y)    1    2 .
 ah  k   ah  k 

(c) If 12 = 0 – if there are only monetary shocks – then from equations (2) and (6):
(7) var( y)  0,
i i
and
2
 a  2
(8) var(y) mm    2  0 .
 ah  k 
Thus interest-rate targeting leads to a lower variance of output than money-stock targeting. In fact,
output is constant under interest targeting.

(d) If 22 = 0 – if there are only IS shocks – then from equations (2) and (6):
(9) var(y) ii  12 ,
and
2
 k  2 2
(10) var(y) m m    1  1 .
 ah  k 
Thus money-stock targeting leads to a lower variance of output than interest-rate targeting.

(e) Consider the situation in part (c) in which there are only monetary shocks. If the policymaker targets
the nominal interest rate, it ensures that i remains constant in the face of any monetary shock. Since i is
not allowed to change, planned expenditure does not change and thus the level of output that equates
planned and actual expenditure does not change. If the policymaker targets the nominal money stock, the
monetary shocks do require a change in the interest rate to restore money-market equilibrium. The
change in the interest rate then changes planned expenditure and thus the level of output that equates
planned and actual expenditure.

Consider the situation in part (d) in which there are only IS shocks. A positive IS shock shifts the IS
curve to the right. If the policymaker keeps m fixed, then as Y rises to equate planned and actual
expenditure, i rises as well for the money market to remain in equilibrium. This rise in i reduces planned
expenditure and thus mitigates some of the positive shock. If, instead, the policymaker targets the
nominal interest rate, equilibrium output changes by the full extent of the shock to the IS curve.

(f) If there are only IS shocks, it is possible to keep y constant at some target level y . By rearranging
the money-market equilibrium equation with y set to y , the nominal money supply must be such that
(11) m = p + h y - ki.

The policymaker knows the fixed p, has picked y herself and can observe i. Thus when i changes – and
since there are only IS shocks, we know this must be due to a shift of the IS curve – the policymaker must
 Solutions to Chapter 12 12-17

change m accordingly. As i rises, for example, the policymaker must reduce m. The policymaker can
stop reducing m when m and i are such that equation (11) is satisfied.

Problem 12.10
(a) Using the fact that for a random variable X, var(X) = E[X2 ] – (E[X])2 or E[X2 ] = var(X) + (E[X])2,
we have
(1) E[(y – y*)2 ] = var(y – y*) + (E[y – y*])2.
Substituting the expression for output, y = x + (k + k )z + u, into var(y – y*) and simplifying yields
(2) var(y – y*) = var(x + kz + k z + u – y*) = z2 k2 + u2.
Substituting for output in (E[y – y*])2 and simplifying yields
(3) (E[y – y*])2 = (E[x + kz + k z + u – y*])2 = (x + kz – y*)2,
where we have used the fact that k and u both have mean zero. Substituting equations (2) and (3) into
equation (1) gives us
(4) E[(y – y*)2] = z2 k2 + u2 + (x + kz – y*)2.

(b) The policymaker chooses z in order to minimize E[(y – y*)2 ]. Using equation (4), the first-order
condition is
(5) (E[(y – y*)2])/z = 2zk2 + 2k(x+ kz – y*) = 0.
Collecting the terms in z yields
(6) z(k2 + k2 ) = (y* – x)k,
and thus the optimal choice of z is
( y *  x) k
(7) z  .
k 2  k 2

(c) To see the way in which policy should respond to shocks (i.e. changes in x), to take the derivative of
z, as given by equation (7), with respect to x:
z k
(8)   0.
x k 2  k 2
The fact that the derivative in (8) is negative implies that higher values of x should be offset with lower
values of z to keep output from varying as much.

Note that z/x does not depend on u2, the parameter that represents uncertainty about the state of the
economy. Thus in this model, the optimal degree of "fine-tuning" does not depend on the amount of
uncertainty about the state of the economy.

(d) In contrast, z/x does depend on k2, the parameter that represents uncertainty about the effects of
the policy instrument. In fact, we have
  z x k
(9)   0.
 k 2
k 2  k 22

Since z/x is negative to begin with, a rise in k2 makes it less negative. That is, higher values of k2 –
more uncertainty about the effects of the policy instrument – reduces the amount that policy should
respond to shocks or in other words, reduces the amount of "fine-tuning" that should be done.
12-18 Solutions to Chapter 12

Problem 12.11
(a) From the assumption of the problem, we first conclude
(1) ~y(0)  b[i(0)  (0)]  0 .
The inequality expressed by (1) also implies
(2)  (0)  ~
y(0)  0
since λ is positive. Therefore, π is decreasing at time zero. We also know that π(0) is a positive quantity
since we are told to assume
(3) 0 < bπ(0), b > 0.
As time moves forward, inflation decreases since its derivative is negative. Because inflation decreases,
output also decreases since
(4) ~y(t )  b[i(t )  (t )]  b(t )  i(0) .
~
As y( t ) decreases,  ( t ) gets more negative, causing π(t) to decrease at a faster rate. Because the interest
rate is constant, this cycle continues indefinitely, causing ~y( t ) and π(t) to diverge to – ∞ as t goes to ∞,
as reflected in the following sketches:

(b) Our graphs will look identical to the sketches in part (a) until π(t) reaches below the x-axis at a time
we will call time T. At time T, ~y( t ) will immediately increase to a less negative value since
(5) ~y(T)  b[i(T)  (T)]  b(T)  b(T)  i(0) .
This implies that the derivative of inflation also increases,
(6)  (T)  ~
y(T) .
But since the derivative is still negative, π(t) continues to decrease after time T, causing ~y( t ) to also
decrease after time T. This behavior is described by the following sketches:

(c) Since i(0) = 0, we conclude

(7) ~ y(0)  b[i(0)  (0)]  b(0)  0 .
Since ~ y(0)  0 , we also deduce
(8)  (t )  ~
y( t )  0 ,
since λ > 0. Thus, π(t) initially increases since its time derivative is positive. As inflation increases,
~y( t ) also increases, causing  ( t ) to increase even further. This cycle continues and causes both ~y( t )
and π(t) to diverge to ∞ as t goes to ∞, as reflected in the following sketches:
 Solutions to Chapter 12 12-19

(d) According to this model, waiting to reduce the interest rate until the economy experiences deflation
can result in a brief bump in output followed by a relapse back into recession. On the other hand, acting
quickly can immediately boost output and avoid a recession, while also causing inflation in the near
future. Having avoided the deflationary collapse, the central bank presumably could raise i at some point
to prevent inflation from exploding.

Problem 12.12
(a) We can begin with equation (12.56) from the text, which states that household optimization requires
the following condition to hold for periods t and t + 1:
1 at (1  i t ) a t 1
(1) dX  dX .
Pt C t Pt 1 C t 1
With the period-1 price level fixed at Ṗ̅ and the assumption that the economy is in steady state in period
2, equation (1) for periods 1 and 2 becomes
1 a1 (1  i1 ) a *
(2)  .
P C1 P* C*
Since consumption and output must be equal in each period, we can rewrite equation (2) as
1 a1 (1  i1 ) a *
(3)  . ii MM
P y1 P* y
Solving for the interest rate in period 1 gives us
a P*y
(4) i1  1  1.
a * Py1
Equation (4) is analogous to equation (12.60) in the text.
We will refer to it as the CC curve. It is downward-
sloping in (y1,i1) space. See the figure at right. 0
CC
The cash-in-advance constraint, equation (12.55) in the
text, is given by y1
 Mt
 P if i t  0
 t
(5) C t 
 M t if i  0.
 Pt t

Since in equilibrium consumption equals output and because we are assuming the price level in period 1
is fixed at Ṗ̅, it follows that in period 1 we have
12-20 Solutions to Chapter 12

 M1
 P if i1  0
(6) y1 
 M1 if i  0.
 P 1

Equation (6) is analogous to equation (12.61) in the text. The set of points satisfying (6) is shown as the
MM curve in the figure above. It is horizontal at i1 = 0 until y1 = M1/Ṗ̅, and then vertical. In the case
depicted in the figure, the CC curve intersects the vertical portion of the MM curve and thus the nominal
interest rate is positive and output equals M1/Ṗ̅.

(b) For i1 to equal 0, the CC curve must intersect the horizontal portion of the MM curve. Thus the level
of Y implied by CC at i1 = 0 must be less than M1/Ṗ̅. From equation (4), when i1 = 0, CC implies
a1P * y
(7)  1.
a * Py1
Solving equation (7) for y1 yields
1  a   P *
(8) y1   1   y .
  a *  P 
In steady state, we know that P* = M*/ y. Thus we can write equation (8) as
1  a  M * 
(9) y1   1   .
  a *  P 
Thus for i1 = 0 we require
1  a   M *  M1
(10)  1    ,
  a *  P  P
or equivalently
1 a 
(11) M1   1  M * .
  a *
Intuitively, expanding the money supply “too much” so that M1 is high, will drive the nominal interest
rate to 0.

From equation (8), we can see that for y1 to be less than y, we require
1  a   P *
(12)  1     1,
  a *  P 
or equivalently
1 a 
(13) P   1  P * .
  a *

(c) When i1 = 0, the CC curve intersects the horizontal portion of the MM curve. From equation (4), we
can see that M1 does not appear in the CC curve and thus an
increase in M1 has no effect on CC. i1 CC MM MMNEW

An increase in M1 extends the horizontal portion of the MM

curve out to the new value of M1NEW/Ṗ̅ and shifts the vertical
portion of MM to the right. See the figure. Just as in the case
with flexible prices, an increase in the first-period money
supply, holding M* fixed, has no effect on demand. Thus, 0

y1 M1/Ṗ̅ M1NEW/Ṗ̅ y1
 Solutions to Chapter 12 12-21

here, it has no effect on first-period output, which remains at y1 in the figure.

An increase in M* causes P* to rise proportionately. Since P* appears in the CC curve given by equation
(4), we can see that it causes the CC curve to shift out to the right (i1 would now be higher for any given
y1). Since M1 is assumed to be unchanged when M* rises, the
MM curve given by equation (6) is unaffected. See the figure i1 MM
NEW
at right. CC CC

As drawn, the new CC curve still intersects the horizontal

portion of the MM curve. Thus the economy remains in a
liquidity trap with i1 = 0. But output rises from y1 to y1NEW. If
the shift in CC were larger, the new curve could intersect the
vertical portion of the MM curve and both output and the 0
nominal interest rate would rise. Intuitively, the increase in
M* raises expected future price levels and thus raises
expected inflation with the first-period price level assumed y1 y1NEW y1
to be fixed. If i1 remains unchanged, the increase in expected
inflation lowers the real interest rate. This increases the demand for goods. In this sticky-price model,
output increases in response to the increased demand.

Problem 12.13
The CC curve for period 1 is given by
1  a  P * 
(1) i1   1     1.
  a *   P1 
Since P* = M*/y*, a fall in y* implies an increase in the steady-state price level, P*. From equation (1),
we can see that this means the CC curve would shift out to the right (i1 would be higher for any given
level of P1).

The period-1 MM curve is given by

 M1
 y if i1  0
 1
(2) P1 
 M1 if i  0.
 y1 1

Since y1 does not change – we are told that y1 need not i1 CC CCNEW MM
equal y* – and M1 does not change, the MM curve is
unaffected.

From the figure at right, we can see that this permanent

adverse supply shock is expansionary. It causes the price
level in period 1 to rise from P1 to P1NEW. As drawn, the
interest rate in period 1 remains at 0. If the shift in the CC 0
curve had been larger, the nominal interest rate would
also rise.
P1 P1NEW P1
Intuitively, the fall in y* raises expectations of future
price levels. If neither P1 nor i1 changed, expected
12-22 Solutions to Chapter 12

inflation would be higher and so the real interest rate would be lower. The demand for goods would
exceed the supply of goods. Since this cannot happen, the result is that the price level is bid up in period
1.

Problem 12.14
(a) (i) The period-2 CC curve is given by i2 MM MMNEW
1  a  P * 
(1) i 2   2     1.
  a *   P2  CC
The period-2 MM curve is given by
 M2
 y if i 2  0
 0 E
(2) P2 
 M 2 if i  0.
 y 2
P2
Since we begin with i2 = 0, we know that the CC curve
intersects the horizontal portion of the MM curve. See the
figure at right. The increase in M2 has no effect on the CC curve. It extends the horizontal portion of the
MM curve and shifts the vertical portion to the right. Thus the increase in M2, holding M1 and M* fixed,
has no effect on the price level or nominal interest rate in period 2. See the figure at right.

In period 1, the CC curve is defined by

1  a  P 
(3) i1   1   2   1 .
  a 2   P1 
Since P2 is unaffected by the increase in M2, there is no impact on the period-1 CC curve.

The MM curve in period 1 is given by

 M1
 y if i1  0

(4) P1 
 M1 if i  0.
 y 1

The change in M2 has no effect on the MM curve in period 1 either. Thus the increase in M2, with M1 and
M* held fixed, has no impact on the price level in period 1.

(a) (ii) An increase in the steady-state money supply, M*, will raise the steady-state price level since
P* = M*/y. From equation (1), we can see that this increase in P* will shift the period-2 CC curve out to
the right (i2 would be higher for any given P2).

Since M2 is held constant, we can see from equation (2) i2 MM

that there is no effect on the MM curve. See the figure at
right. The increase in M* causes the price level in period 2 CC CCNEW
to rise from P2 to P2NEW. As drawn, the nominal interest
rate remains at 0. If the CC curve had shifted more, both
the price level and the nominal interest rate would have
risen. Intuitively, the increase in expected future money 0
supply raises expected future price levels. If neither P2 nor
i2 changed, expected inflation would be higher and the real
interest rate would be lower. Thus demand for goods
P2 P2NEW P2
 Solutions to Chapter 12 12-23

would exceed the supply. Since this cannot happen, the price level and/or the nominal interest rate must
rise.

The increase in the price level in period 2 now has an i1

effect in period 1. From equation (3), we can see that the MM
rise in P2 causes the period-1 CC curve to shift out to the CC CCNEW
right. Since M1 is held constant, we can see from
equation (4) that the period-1 MM curve is unaffected.
See the figure at right. The increase in M* causes the
price level in period 1 to rise from P1 to P1NEW. As drawn,
the nominal interest rate remains at 0. If the CC curve 0
had shifted more, both the price level and the nominal
interest rate would have risen. The intuition is the same
as for period 2. The increase in expected future price P1 P1NEW P1
levels means that the price level must rise today.

(a) (iii) The increase in M* increases P* proportionately. From equation (1), we can see that this increase
in P* will shift the period-2 CC curve out to the right (i2
would be higher for any given P2) and have no effect on i2
the period-2 MM curve. The increase in the money MM MMNEW
supply in period 2, M2, has no effect on the CC curve in
that period but it extends the horizontal portion of the CC
MM curve and shifts the vertical portion to the right, in
proportion to the increase in M2. See the figure at right.
The price level in period 2 rises from P2 to P2NEW, and the 0
increase will be in proportion to the increase in M2 and CCNEW
M*. That is, unlike the case in the solution to part (a) (ii),
we know that the nominal interest rate will remain at 0 in P2 P2NEW P2
period 2.

We can see from equations (3) and (4), that the effect in i1
period 1 will be similar. The increase in P2 shifts the MM MMNEW
period-1 CC curve out to the right in proportion to the CC
increase in P2 and thus M2. The increase in M1
extends the horizontal portion of the MM curve and
shifts the vertical portion to the right in proportion to the
increase in M1. The result is that the price level in period 0 CCNEW
1 rises in proportion to the increase in M1, M2, and M*.
See the figure at right in which the price level in period 1
rises from P1 to P1NEW. P1 P1NEW P1
12-24 Solutions to Chapter 12

(b) For i2 to be greater than > 0, the CC curve must

intersect the vertical portion of the MM curve. See the i2
figure at right. From equation (2), the increase in M2 MM MMNEW
extends the horizontal portion of the MM curve and CC
shifts the vertical portion to the right. The new MM
curve would be vertical at P2 = M2NEW/y. From equation i2
(1), the increase in M2 has no effect on the CC curve. As
we can see from the figure, the period-2 price level rises i2NEW
=0
to P2NEW. Note that it does not rise in proportion to the
increase in M2 because i2 falls to 0. That is, the increase
in M2 is expansionary in period 2 but it does not have the
P2=M2/y P2NEW P2
full conventional effect on aggregate demand since the
economy hits the zero lower bound on the nominal
interest rate.

Equation (3) tells us that the rise in P2 shifts out the

period-1 CC curve. It has no effect on the MM curve in i1 MM
period 1. Thus the price level in period 1 rises to P1NEW.
See the figure at right. Intuitively, the monetary CC CCNEW
expansion in period 2 raises the expected price level and
thus raises expected inflation in period 1. If neither P1
nor i1 changed, this would lower the real interest rate and
demand for goods would exceed supply. Since this 0
cannot happen, prices are bid up. Through this
mechanism, the monetary expansion in period 2 – even
though it means the economy will be in a liquidity trap
for two periods rather than one – is expansionary in P1 P1NEW P1
period 1.

Problem 12.15
As described in the text, in equilibrium, output equals yn and inflation equals * + (b/a)(y* – yn).
Substituting these values into the loss function given by equation (12.64) in the text, which is given by
L = (1/2)(y – y*)2 + (1/2)a( – *)2, yields the following value of the loss function in equilibrium:

       
2
1 n 2 1 b 1 2 1 b2 2
(1) LEQ  y  y *  a  y * y n  y * yn  y * y n ,
2 2 a 2 2 a
or simply

(2) LEQ 
1
 2
y *  y n 1 
b2 
 .
2  a 
Output equals yn in equilibrium, regardless of the value of a. Thus to see how the equilibrium loss varies
with a, use equation (2) to take the derivative of LEQ with respect to a:

(3)
LEQ  b 2
a
 
2a 2
 2
y * y n  0 .

Equation (3) states that a fall in a increases LEQ. That is, a reduction in the cost of inflation increases the
loss to society. It is true that any given deviation in inflation from its optimal level, *, has a lower cost
to society. However, the problem is that this causes the equilibrium level of inflation itself to be higher.
Intuitively, at a given e, the marginal cost of additional inflation is now lower for the policymaker. For a
given e, it then becomes optimal to set a higher inflation rate. But the public knows this and thus the
 Solutions to Chapter 12 12-25

level of  for which e =  is now higher. It turns out that the fact that EQ exceeds * by more than it
used to, outweighs the fact that any given deviation in EQ from * has a lower cost to society.

Problem 12.16
(a) Let S be the amount of social welfare from a given policy. Thus, we have
 a   a 
(1) S   y1  12    y 2  22  .
 2   2 
Substituting equation (12.63) for output into equation (1), we get
a a
(2) S  ( y n  b1  12  b1e )  ( y n  b2  22  be2 ) .
2 2
Taking the derivative of (2) with respect to π2 and setting the result equal to zero gives the solution
π2 = b/a.

(b) Since the type 1 policymaker never chooses π1 = 0, there is no doubt in the second period that the
policymaker is of type 1, since a type 2 policymaker would have picked π1 = 0. Therefore, people will
expect the policymaker to maximize social welfare in the second period. Consequently, e2  b / a .

Taking the derivative of (2) with respect to π1 and setting the result equal to zero, we find that the
policymaker selects π1 = b/a to maximize social welfare. Setting 1  2  e2  b / a in equation (2), we
get
(3) S  2y n  b1e .

(c) Since the public expects the policymaker to select π2 = b/a with probability p and π2 = 0 with
probability (1 – p), we have
(4) e2  p(b / a )  (1  p)0  p(b / a ) .
Therefore, substituting π1 = 0, 2  b / a , and e2  p(b / a ) into equation (2), we get
b2  1 
(5) S  2 y n  b1e    p .
a 2 

(d) Since 0 < p < 1/2, the value of S in part (c), equation (5), is larger than the value of S obtained in part
(b), equation (3), because the term (b2/a)[(1/2) – p] is positive. Thus, a type 1 policymaker would select
π1 = 0. Notice also that as p gets smaller, S gets larger. This implies that a strong reputation as a
policymaker who is tough on inflation can allow a type 1 policymaker to achieve higher social welfare by
first selecting π1 = 0.

(e) For 1/2 < p < 1, if the public believes the type 1 policymaker will pick π1 = 0, then e2  p(b / a ) as in
part (c). However, now (1/2) – p is negative and consequently the term (b2/a)[(1/2) – p] reduces the
value of S in equation (5). Thus, if the public believes that a type 1 policymaker will pick π1 = 0, he or
she will choose π1 = b/a.

Suppose instead the public believes the type 1 policymaker will pick π1 = b/a. If he or she picked π1 =
b/a, the public would know that the policymaker was of type 1 and would adjust their expectation so that
e2  (b / a ) . Substituting 1  2  e2  b / a into equation (2) gives us
(6) S  2y n  b1e .
12-26 Solutions to Chapter 12

If the policymaker chose π1 = 0, the public would believe for certain that she is a type 2 policymaker
(because the public believes that type 1 policymakers always choose π1 = b/a), and so e2  0 . Choosing
π2 = b/a would then achieve a level of social welfare given by
b2
(7) S  2 y n  b1e 
2a
Since (7) yields a higher level of social welfare than (6), the policymaker of type 1 will choose π1 = 0 if
the public believes that a type 1 policymaker will pick π1 = b/a.

Therefore, neither the public always believing a type 1 policymaker will pick π1 = 0 nor the public always
believing a type 1 policymaker will pick π1 = b/a can be an equilibrium. The equilibrium will instead
consist of the type 1 policymaker playing a mixed strategy for period 1 in which he or she selects π1 = 0
with some probability and π1 = b/a with some probability.

Problem 12.17
(a) The policymaker chooses inflation to maximize her objective function, which is given by
W = cy – (a2 /2), subject to the constraint that output is given by the Lucas Supply function,
y = yn + b( – e ). Thus the policymaker's problem is
(1) max W = c[ y + b( – e )] – (a2 /2).


The first-order condition is

(2) W/ = bc – a = 0.
Thus the policymaker's choice of  is
(3)  = bc/a.

(b) The public knows the policymaker sets inflation according to equation (3). Thus with rational
expectations, expected inflation must equal the expectation of the right-hand side of equation (3):
(4) e = E[bc/a] = bcE[]/a = bc  /a.

(c) The true social welfare function is given by WSOC = y – (a2 /2). Taking the expectation of both
sides of this expression with respect to the public's information set, so that  is random, gives us
(5) E[WSOC ] = E[(yn + b( – e )) – (a2 /2)],
where we have substituted for y = yn + b( – e ). Now substitute the policymaker's choice of , equation
(3), and the public's expectation of inflation, equation (4), into equation (5):
   bc bc  ab c  
2 2 2
(6) E[ WSOC ]  E   y n  b     .
   a a  2a 2 
Simplifying yields
b 2cE[  2 ] b 2cE[ ] b 2c 2 E[  2 ]
(7) E[ WSOC ]  y n E[ ]    .
a a 2a
Since E[] =  , equation (7) becomes

(8) E[ WSOC ]  y n  
b 2c
a

E[  2 ]   2 b 2c 2 E[  2 ]
2a
.
Now use the facts that for a random variable X:
(9) var(X) = E[X2 ] – (E[X])2,
and
(10) E[X2 ] = var(X) + (E[X])2.
 Solutions to Chapter 12 12-27

Here, this means that we can write

(11) 2 = E[2 ] –  2,
and
(12) E[2 ] = 2 +  2.
Substituting equations (11) and (12) into equation (8) gives us the following expected value of the true
social welfare function:

(13) E[ WSOC ]  y n  
a
 
2a

b 2c 2 b 2c 2 2
   2 . 
(d) To find the first-order condition for the maximization, use equation (13) to set the derivative of the
expected value of the social welfare function with respect to c equal to zero:
 E[WSOC ] b 2 2 b 2 c
(14)
c

a
 
a
 
 2   2  0 .

Solving for c yields

 2
(15) c  .
 2   2
There is a tradeoff here. From equation (3), we can see that choosing a more "conservative" policymaker
– one with a low value of c – produces a better performance in terms of average inflation. Such a
policymaker would not respond well to the shocks, however. Thus there is some optimal level of
"conservatism" that balances these two forces.

The value of c that maximizes the expected value of true social welfare is decreasing in the mean of .
Since we know that e equals  on average (since  equals  on average), output equals full-employment
output on average, regardless of the values of c or  . From equation (3), we can see that if  is higher on
average, inflation will also be higher on average, for a given c. Thus it will be welfare-improving to
offset this higher average  and keep inflation lower on average by having a policymaker with a lower c;
that is, having a more "conservative" policymaker.

However, the value of c that maximizes expected social welfare is increasing in the variance of the 
shock. The more variable is the shock, the less "conservative" the central banker should be. Since the
policymaker can act after  is realized, she can choose to offset any deviation in  from its expected
value, which raises welfare. The policymaker will do this only to the extent that she cares about the
shock's effect. Thus the more that  varies, the better it is to have a policymaker who cares about the
shock's effect and will act to offset it.

Problem 12.18
(a) Social welfare is higher when the policymaker turns out to be a Type-1, the type that shares the
public's preferences concerning output and inflation. The choice of setting  = 0 in both periods – as the
Type-2 policymaker does – is a choice available to the Type-1 policymaker. She chooses not to do this;
to maximize social welfare, she decides to choose another pair of inflation rates. Since she is attempting
to maximize social welfare, welfare must be higher under the choices made by the Type-1 policymaker.
For example, as explained in the text, if  < 1/2, it is optimal for the Type-1 policymaker to choose 1 =
b/a and 2 = b/a. That must be because it achieves higher welfare than choosing 1 = 0, 2 = 0.

(b) Expected inflation, e, is determined by the public's beliefs. So both the " a' " policymaker and the
"a" policymaker face the same e, since in either case, the public believes it is facing an " a' "
12-28 Solutions to Chapter 12

policymaker. Thus both policymakers have the same choice set. The "a" policymaker makes her choice
to maximize true social welfare, whereas the " a' " policymaker makes her choice to maximize something
else. Thus social welfare must be higher with the "a" policymaker.

Problem 12.19
(a) Suppose that  differs from  in some period t0 . Then e = b/a for all periods after t0 . Substituting
this expression for expected inflation into the Lucas supply function, yt = yn + b(t – te ), gives us output
in each subsequent period:
(1) yt = yn + b(t – b/a) for all t > t0 .
With expected inflation now constant into the future, the equilibrium in each period is independent of the
policymaker's action in the previous period. Thus we can concentrate on a representative period, t; the
equilibrium in all periods will be the same. Substituting equation (1) into the policymaker's objective
function for period t, wt = yt – (a2 /2), yields
(2) wt = yn + b(t – b/a) – at2 /2 for all t > t0 .
The first-order condition for the choice of inflation is
(3) wt /t = b – at = 0,
and thus the policymaker chooses
(4) t = b/a for all t > t0 .
Since t = t = b/a, then from the Lucas supply function we have
e

(5) yt = yn for all t > t0 .

(b) To keep things simple, we can assume that the monetary authority chooses to depart from  =  in
period 0. This does not alter the message. Since  has always been equal to  , 0e =  . Substituting
this into the Lucas supply function gives us
(6) y0 = yn + b(0 – b/a).
Given the fact that the policymaker is choosing to depart from  =  , the choice of 0 does not affect e
and thus the equilibrium in future periods. Thus only the current period's objective function matters to
the policymaker. She will choose 0 to maximize
(7) w0 = yn + b(0 –  ) – (a02 /2).
The first-order condition for the choice of 0 is
(8) w0 /0 = b – a0 = 0,
and thus the policymaker chooses
(9) 0 = b/a.

With this choice of inflation, using the Lucas supply function, output in period 0 is given by
(10) y0 = yn + b[(b/a) –  ].
Substituting equations (9) and (10) into the policymaker's objective function, w0 = y0 – (a02 /2), yields
(11) w0 = yn + (b2 /a) – b  – (b2 /2a),
or simply
(12) w0 = yn + (b2 /2a) – b  .

As shown in part (a), in all subsequent periods after the policymaker has deviated, t = b/a and yt = yn.
Substituting these values of output and inflation into the objective function, wt = yt – (a2 /2), gives us
(13) wt = yn – (b2 /2a) for all t > 0.
Thus the policymaker's lifetime objective function if she deviates is given by

t yn  (b2 / 2a).


D n 2
(14) W  y  (b / 2a )  bˆ 
t 1
 Solutions to Chapter 12 12-29

Pulling the [yn – (b2 /2a)] out of the summation sign and using the fact that, since  < 1, we have


t 1
2 3 2

(15)            1        (1  ) ,
t

we can write the lifetime objective function as
b2     n b2     n    b2
(16) W D  y n   bˆ     y    1  y  bˆ  1   ,
2a  1     2a   1     1    2a
or simply
 1  n  1  2  b 2
(17) W D   y  bˆ    .
1     1    2a

If the policymaker chooses  =  every period, output will be equal to yn every period. The
policymaker's objective function in each period is therefore given by
(18) wt = yn – (a  2 /2).
Thus the policymaker's lifetime objective function if she does not deviate is given by

t yn  (aˆ 2 / 2).


(19) W ND 
t 0
Pulling the [yn – (a  2 /2)] out of the summation sign and using the fact that, since  < 1, we have

(20)    1       1 (1  ) ,
t 2
t 0
we can write the lifetime objective function as
ND  1  n a
 ˆ2 
(21) W   
 1    y   .
  2 

(c) One way of solving the problem is to calculate the benefit and cost of deviating as a function of 
and the other parameters. We can then examine the range of  's over which the cost exceeds the benefit
and thus the range of  's over which the policymaker will choose not to deviate from  =  .
The benefit of departing from  =  in some period t0 is that welfare in period t0 is yn + (b2 /2a) – b 
[see equation (12)] rather than yn – (a  2 /2) [see equation (18)]. Thus the benefit of deviating, B, is
(22) B = yn + (b2 /2a) – b  – yn + (a  2 /2),
or simply
(23) B = (b2 /2a) + (a  2 /2) – b  .

The cost of deviating is that in all periods subsequent to t0 , welfare will be yn – (b2 /2a) [see equation
(13)] rather than yn – (a  2 /2). Thus the cost of deviating in each future period is
yn – (a  2 /2) – yn + (b2 /2a) or simply (b2 /2a) – (a  2 /2). The total cost of deviating, discounted to time
t0 is

(24) C 
t  t o 1
    
  t  t 0 ( b 2 / 2a )  (a 2 / 2)  ( b 2 / 2a )  (a 2 / 2)    2   3  .

Substituting the result in equation (15) into equation (24) gives us the following cost of deviating:
    b 2 a 2 
(25) C     .
 1     2a 2 
12-30 Solutions to Chapter 12

We can plot the benefit and cost from deviating as a function of  . First, we will deal with the benefit
from deviating. From equation (23), we have
(26) B/  = a  – b,
and
(27) 2 B/  2 = a > 0.
Thus B is a parabola that reaches a minimum at  = b/a. From equation (23), at  = 0, B = b2 /2a.
Finally, at its minimum at  = b/a, B = (b2 /2a) + (b2 /2a) – (b2 /a) = 0. B, the benefit from deviating as a
function of  , is plotted in both figures below.

Now dealing with the cost of deviation, we have from equation (25)
(28) C/  = – [/(1 – )]a  ,
and
(29) 2 C/  2 = – [/(1 – )]a < 0.
Thus C is an inverted parabola that reaches a maximum at  = 0. From equation (25), the value of the
cost of deviating at  = 0 is given by C = [/(1 – )](b2 /2a). In addition, at  = b/a, C = 0.

B,C B,C
 b2
1   2a
Cost
b2 /2a Benefit

Cost
Benefit
b2 /2a
 b2
1   2a

b(1  2) b b(1  2) b

 2   1    2   1  

a a
a a

The case of  < ½, so that /(1 – ) < 1, is depicted in the left-hand figure. The case of  > ½, so that
/(1 – ) > 1, is depicted in the right-hand figure. We need to solve for the values of  where the benefit
of deviating equals the cost of deviating. Setting the right-hand sides of equations (23) and (25) equal
yields
b 2 aˆ 2   b 2 aˆ 2 
(30)   bˆ    ,
2a 2 1    2a 2 
which implies that
   aˆ 2   b2
(31) 1    bˆ  1   0,
 1  2  1    2a
or simply
 1  a 2  1  2  b 2
(32)    b     0.
1    2  1    2a
Multiplying both sides of equation (32) by (1 – )2a gives us an equivalent condition for B = C :
 Solutions to Chapter 12 12-31

(33) a2  2 – 2a(1 – )b  + (1 – 2)b2 = 0.

Using the quadratic formula, we have

2ab(1  )  4a 2 b 2 (1  ) 2  4a 2 b 2 (1  2)  
2ab(1  )  4a 2 b 2 1  2   2  1  2
(34)    .
2a 2 2a 2
Some further algebra yields
2ab(1  )  2ab b(1  )  b
(35)    ,
2a 2 a
and thus finally
b(1  )  b b
(36)  1   ,
a a
and
b(1  )  b b(1  2)
(37)  2   .
a a
These two values of  for which the benefit of deviating just equals the cost of deviating are depicted in
the figures above. Note that for the case of  > 1/2 – the figure on the right –  2 is negative and is thus
not relevant. We can now interpret the figures.

For the case of  > 1/2 – depicted in the figure on the right – the cost of deviating exceeds the benefit of
deviating for any  such that 0   < b/a. With these values of the parameters, the policymaker will
choose not to deviate from  =  . Right at  = b/a, the policymaker is indifferent and in fact at  = b/a,
deviating is the same as producing  =  . Finally, for any value of  > b/a, the benefit of deviating
exceeds the cost of deviating and hence the policymaker will in fact deviate from  =  .

For the case of  < 1/2 – depicted in the figure on the left – the cost of deviating exceeds the benefit of
deviating for any value of  such that [b(1 – 2)]/a <  < b/a. With these values of the parameters, the
policymaker will choose not to deviate from  =  . Right at  = b/a and  = [b(1 – 2)]/a, the
policymaker is indifferent. Finally, for any value of  < [b(1 – 2)]/a or  > b/a, the benefit of
deviating exceeds the cost of deviating and hence the policymaker will in fact deviate from  =  .

For the policymaker to set  = 0 if  = 0, we would need the cost of deviating to exceed the benefit of
deviating, evaluated at  = 0. From our earlier discussion, we know this will be true if  > 1/2. Thus
regardless of the values of a and b, the policymaker will choose to set inflation to zero if
 = 0 as long as the discount rate is greater than 1/2.

Problem 12.20
(a) We can use the same technique as in part (c) of the solution to Problem 12.19. We can examine the
range of  's over which the cost of deviating from setting  =  exceeds the benefit of deviating. This
gives the range of  's over which the policymaker chooses  =  each period. The benefit from
deviating, B, is the same as it was in Problem 12.19. Thus we have
(1) B = (b2 /2a) + (a  2 /2) – b  .
The cost of deviating in some period is that in the following period, e = b/a rather than e =  . As
shown in part (a) of the solution to Problem 12.19, when e = b/a, the policymaker chooses  = b/a. Thus
output is equal to yn in the following period. Since the policymaker chooses  = e in the period after
deviating, expected inflation reverts to e =  in all subsequent periods. Thus there is only a one-period
cost to deviating. Specifically, the cost is that in the period after deviating, the value of the policymaker's
objective function is yn – (b2 /2a) rather than yn – (a  2 /2). Discounting that back to the period in which
the deviation occurs yields the following cost, C:
12-32 Solutions to Chapter 12

(2) C = [ yn – (a  2 /2) – yn + (b2 /2a)] = [(b2 /2a) – (a  2 /2)].

We can now plot the benefit and cost of deviating as a function of  . The benefit from deviating is the
same as in Problem 12.19 and so we can concentrate on the cost. From equation (2):
(3) C/  = –a  ,
and
(4) 2 C/  2 = –a < 0.
Thus C is an inverted parabola that reaches a maximum at  = 0. From equation (2), the value of the
cost of deviating at  = 0 is given by C = (b2 /2a) < (b2 /2a) since  < 1. The next step is to solve for
the values of  where the benefit of deviating equals the cost of deviating. Setting the right-hand sides
of equations (1) and (2) equal yields
b 2 aˆ 2 b 2 aˆ 2
(5)   bˆ   ,
2a 2 2a 2
which simplifies to
(1  )a 2 (1  )b 2
(6) ˆ  bˆ   0.
2 2a
Multiplying both sides of equation (6) by 2a gives us the following equivalent condition for B = C:
(7) (1 + )a2  2 – 2ab  + (1 – )b2 = 0.
Using the quadratic formula gives us

2ab  4a 2 b 2  4a 2 b 2 (1  )(1  ) 
2ab  4a 2 b 2 1  1   2 
(8)    .
2a 2 (1  ) 2a 2 (1  )
Some further algebra yields
2ab  2ab b(1  )
(9)   2  ,
2a (1  ) a (1  )
and thus finally
b(1  ) b b(1  )
(10)  1   , and (11)  2  .
a (1  ) a a (1  )

B,C From the figure at left, we can see that the cost of
deviating from  =  exceeds the benefit from
deviating for any  such that
(12) b(1 – )/a(1 + ) <  < b/a.
b2 /2a Benefit
With these values of the parameters, the
policymaker will choose not to deviate. For any
Cost value of  greater than b/a or less than
b(1 – )/a(1 + ), the benefit from deviating
b2 /2a exceeds the cost of deviating and hence the
policymaker will in fact deviate from  =  .

b(1  ) b
 2   1  
a (1  ) a
 Solutions to Chapter 12 12-33

(b) Again, we will employ the same technique. The benefit from deviating remains the same; it is given
by equation (1), B  ( b 2 / 2a)  (a 2 / 2)  b .

We need to determine the cost of deviating for the policymaker. Suppose the policymaker deviates in
some period t. Then in period t + 1, t+1e = 0 > b/a. We can also write this as t+1e = b/a + x, x > 0. The
variable x captures the extent of the punishment for deviating. When the policymaker takes expected
inflation as given, she chooses to set inflation equal to b/a. Thus, using the Lucas supply function, output
in period t + 1, the period after deviating, is
(13) y t 1  y n  b[(b / a )  (b / a )  x]  y n  bx .
Thus output is below the natural rate the period after deviating. The value of the policymaker's objective
function in period t + 1 is
(14) w t 1  y t 1  (a2 / 2)  y n  bx  (b 2 / 2a ) .
Thus the cost of deviating in period t + 1 is that welfare is given by (14) rather than y n  (aˆ 2 / 2) .
Discounting this back to period t, we have the cost in period t + 1:
(15) Ct 1  [ y n  (aˆ 2 / 2)  y n  bx  (b 2 / 2a )] ,
or simply
(16) C t1   [ bx  (a 2 / 2)  ( b 2 / 2a)] .

Now consider the situation in period t + 2, two periods after a deviation. Expected inflation equals b/a.
Taking expected inflation as given, the policymaker chooses to set inflation equal to b/a. Thus output is
at the natural rate. The value of the policymaker's objective function in t + 2 is
(17) w t  2  y t  2  (a2 / 2)  y n  (b 2 / 2a ) .
Thus the cost of deviating in period t + 2 is that welfare is equal to y n  (b 2 / 2a ) rather than
y n  (aˆ 2 / 2) . Discounting this back to period t, we have the cost in period t + 2:
(18) C t  2  2[ y n  (aˆ 2 / 2)  y n  (b 2 / 2a )]  2 [(b 2 / 2a )  (aˆ 2 / 2)] .

In period t + 3, since actual inflation last period was equal to expected inflation last period, expected
inflation reverts to  and there is no further cost to the deviation in period t.

Thus the total cost of the deviation is

(19) C   [ bx  (a 2 / 2)  ( b 2 / 2a)]   2 [( b 2 / 2a)  (a 2 / 2)],
or simply
(20) C  bx  (1  )[( b 2 / 2a)  (a 2 / 2)] .

From equation (20),

(21)  C   (1  )[2a / 2]  (1  )a ,
and
2 2
(22)  C   (1  )a  0 .
Thus C is an inverted parabola that reaches a maximum at   0 . The value of the cost of deviating at
  0 is given by bx + (1 + )(b2/2a). From earlier analysis, we know that the benefit of deviating at
  0 is b2/2a. Thus if the value of x, the excess punishment, is high enough, the cost of deviating at
  0 will exceed the benefit and there can be an equilibrium with zero inflation. Specifically, we need
the following condition to hold:
(23) bx + (1 + )(b2/2a) > b2/2a,
12-34 Solutions to Chapter 12

or
(24) bx > (b2/2a)[1 – (1 + )],
or simply
(25) x > (b2/2a)[1 – (1 + )]/b.

We can determine the value of  at which C = 0. From equation (20), C = 0 when

(26) (1  )[(a 2 / 2)  ( b 2 / 2a)]  bx ,
which implies
(27) (a 2 / 2)  ( b 2 / 2a)  bx / (1  ) ,
and thus
(28)  2  [2bx / a(1  )]  ( b 2 / a 2 ) .
Therefore C = 0 when
2 bx b2 b
(29)    2  .
a (1  ) a a
Thus the cost of deviating is equal to zero at a value of  greater than the one for which the benefit of
deviating is equal to zero (which is   b / a ). The values of  for which it is an equilibrium for the
policymaker not to deviate are those – just as in part (a) – where the cost of deviating exceeds the benefit.
The basic idea is that higher values of x lead to a wider range of  's for which the cost exceeds the
benefit and thus a wider range of  's for which the policymaker does not deviate.

(c) As we have shown previously, if the policymaker takes expected inflation as given, she chooses
inflation equal to b/a. Thus if e = b/a, the policymaker chooses  = b/a, so that the public's expectation
is fulfilled and output is at the natural rate. There is no incentive for the policymaker to choose a
different inflation rate and there is no incentive for the public to change its expectation of inflation and
thus  = e = b/a will be an equilibrium for any a > 0, b > 0.

Problem 12.21
The key is to reason backward from the last period, which we will denote T. In that last period, the
policymaker's choice of  has no effect on next period's expected inflation; there is no next period. Thus
the policymaker's problem in the final period is to take expected inflation as given and choose  in order
to maximize the period T objective function. From previous analysis in the solution to Problem 12.19,
we know that the policymaker's choice of inflation in this type of situation is T = b/a. Since the public
knows how the policymaker behaves, expected inflation also equals b/a and thus output equals yn.

Now consider the situation in period T – 1. The important point is that the policymaker knows her
choice of T-1 will have no bearing on what happens the next and final period. Regardless of the level of
 she chooses in period T – 1, expected inflation next period will be b/a, as described above. Since the
policymaker's problem has no impact on the future, she chooses , taking e as given, to maximize the
period T – 1 objective function. Again, the optimal choice is T-1 = b/a. The public knows this and so
T-1e = b/a and thus output in period T – 1 equals yn.

Working backward, the same thing happens each period. The policymaker knows that expected inflation
the following period will be b/a regardless of what she does this period. Thus she acts to maximize the
one-period objective function and chooses  = b/a, which results in output equal to the natural rate.
Therefore the unique equilibrium for all periods is te = t = b/a and yt = yn.
 Solutions to Chapter 12 12-35

Problem 12.22
The politician faces the following problem, where E is defined as the probability of being reelected:
(1) max E  Pr[ 2  u 2  K] ,
u1, u 2
subject to
(2)  t   t 1  (u t  u n )  st ,
and
(3) uL  ut  uH ,
for t = 1, 2. Substituting equation (2) evaluated at t = 2 into the probability that the politician is reelected
yields
(4) E  Pr[1  (u 2  u n )  s2  u 2  K] .
We can rewrite equation (4) as
(5) E  Pr[s2  K  1  u n  (  )u 2 ] .
The probability on the right-hand side of equation (5) is simply the cumulative distribution function of
2S evaluated at K  1  u n  (  )u 2 and so we can write the probability of being reelected as
(6) E  F(K  1  u n  (  )u 2 ) .
Note that the choice of u2 does not depend on 1, which in turn is a function of u1. To see the way in
which the probability of being reelected varies with the choice of u1, take the derivative of E with respect
to u1:
E 
(7)  f (K  1  u n  (  )u 2 )( 1 ) ,
 u1  u1
where f(•) is the probability density function of 2S. Since 1  0  (u1  u n )  1s , where 0 is given,
 1  u1   . Thus we can write
E
(8)  f (K  1  u n  (  )u 2 )
 u1
Since f(•)  0, the derivative given in equation (8) is greater than or equal to zero. Thus picking a higher
value for first-period unemployment can never reduce the probability of being reelected and might
increase it. Thus it is optimal to pick the highest feasible level of unemployment in the first period, uH.

Intuitively, since only second-period inflation and unemployment determine the probability of being
reelected, the politician wants to face the best possible inflation-unemployment tradeoff in period 2.
From equation (2), we can see that is accomplished by having the lowest possible inflation rate in the
previous period, period 1. That, in turn, is accomplished by having the highest possible unemployment
rate in period 1.

Problem 12.23
(a) Rearranging the relationship between output and inflation given by
(1) y t  y n  b( t  E t 1[ t ]) ,
to solve for t gives us
(2)  t  (1/ b)(y t  y n )  E t 1[ t ] .

The liberal leader chooses yt to maximize aLyt – t2/2 subject to inflation being determined by equation
(2). The first-order condition is
12-36 Solutions to Chapter 12

 t
(3) a L   t 0.
yt
From equation (2), taking Et-1[t] as given, the derivative of t with respect to yt is (1/b). Substituting that
fact, as well as equation (2), into the first-order condition gives us
(4) a L  [(1/ b)(y t  y n )  E t 1[ t ]](1/ b)  0 .
Equation (4) can be rewritten as
(5) (1 / b)(y t  y n )  E t 1[ t ]  ba L .
Solving equation (5) for yt gives us
(6) (1/ b) y t  (1/ b) y n  ba L  E t 1[ t ] ,
and multiplying both sides of equation (6) by b yields
(7) y L n 2
t  y  b a L  bE t 1[ t ] .
Equation (7) gives the value of output that a liberal leader will choose.

An analogous exercise would allow us to derive the following expression for the level of output chosen
by a conservative leader:
(8) y C n 2
t  y  b a C  bE t 1[ t ] .

(b) Substitute the liberal's choice for output into equation (2) to obtain
(9)  t  (1/ b)(y n  b 2a L  bE t 1[ t ]  y n )  E t 1[ t ]
Thus the inflation rate with a liberal leader is given by
(10)  Lt  ba L .
Similarly, the inflation rate with a conservative leader is
(11)  Ct  ba C .
Individuals know this is how the leaders will behave in period 1. Since the public knows that the
probability of a liberal leader is p, and the probability of a conservative leader is (1 – p), the expected
value of inflation in period 1 is a weighted average of the two possible inflation rates given by (10) and
(11), where the probabilities serve as weights. That is,
(12) E 0 [1 ]  pba L  (1  p)ba C  b[pa L  (1  p)a C ] .

To determine output in period 1 under a liberal leader, substitute equation (12) into equation (7):
(13) y1L  y n  b 2a L  bE 0 [1 ]  y n  b 2a L  b 2 [pa L  (1  p)a C ] .
Equation (13) can be rearranged to obtain
(14) y1L  y n  b 2a L (1  p)  b 2 (1  p)a C ,
or simply
(15) y1L  y n  b 2 (1  p)[a L  a C ] .

To determine output in period 1 under a conservative leader, substitute equation (12) into equation (8):
(16) y1C  y n  b 2a C  bE 0 [1 ]  y n  b 2a C  b 2 [pa L  (1  p)a C ] .
Equation (16) can be rearranged to obtain
(17) y1C  y n  b 2 [pa L  paC ] ,
or simply
(18) y1C  y n  b 2 p[a L  a C ] .
 Solutions to Chapter 12 12-37

Since aL > aC > 0, output and inflation are higher in period 1 under a liberal leader than they are under a
conservative leader.

(c) In period 1, the public knows for certain who the leader will be in period 2. Thus if a liberal is in
power, the public expects inflation in period 2 to equal baL; similarly, if a conservative is in power, the
public expects inflation in period 2 to equal baC.

The leaders continue to maximize their objective function, taking expected inflation as given. Thus,
from equation (7), the liberal leader's choice of y2 is given by
(19) y L n 2
2  y  b a L  bE1[ 2 ] .
Substituting the fact that E1[2] = baL if there is a liberal leader gives us
(20) y L n 2 2
2  y  b aL  b aL .
And thus period-2 output under a liberal leader is simply
(21) y L
2 y .
n

Similarly, from equation (8), the conservative leader's choice of y2 is given by

(22) y C n 2
2  y  b a C  bE1[ 2 ] .
Substituting the fact that E1[2] = baC if there is a conservative leader gives us
(23) yC n 2
2  y  b aC  b aC .
2

And thus period-2 output under a conservative leader is also given by

(24) y C
2 y .
n

Without the uncertainty about who the leader will be, output will not deviate from potential.

Problem 12.24
We can focus on a situation in which gM , , i, and r are constant and in which e = . Although not
technically correct – since Y and thus M/P are growing – we will refer to such a situation as a steady
state in what follows. Under these assumptions, it is therefore reasonable to assume that output, and the
real interest rate are unaffected by the rate of money growth and that actual and expected inflation are
equal. Taking the exponential function of both sides of the money demand function, which is given by
ln(M(t)/P(t)) = a – bi + lnY(t), yields
(1) M(t)/P(t) = e a bi Y( t ) .
The nominal interest rate is given by i = r + e. In steady state, e and r are constant and thus so is the
nominal interest rate. Thus in steady state, the quantity of real balances must grow at the same rate as
Y(t). In other words, M  ( t ) M( t )  P ( t ) P( t )  g . Solving for inflation yields
Y
(2)  = gM – gY ,
where gM is the growth rate of the nominal money stock. This means that the nominal interest rate in
steady state is given by
(3) i  r    r  g M  g Y ,
where we have used the fact that actual and expected inflation are equal. Substituting equation (3) into
equation (1) gives steady-state real balances:
 b ( r  g M  gY )
(4) M( t ) P( t )  e a e Y( t ) .

Seignorage is given by
12-38 Solutions to Chapter 12

 (t)
M  ( t ) M( t )
M M( t )
(5) S(t) =  .  gM
P( t ) M ( t ) P( t ) P( t )
Substituting equation (4) into equation (5) gives steady-state seignorage:
 b ( r  g M  gY )  bgM
(6) S( t )  g M e a e Y( t )  Cg M e Y( t ) ,
a  b( r gY )
where C  e e . We need to find the choice of nominal money growth, gM , that maximizes
steady-state seignorage. Again, we are assuming that output is unaffected by money growth. The first-
order condition is
 bg  bg
(7)  S( t )  g M  Ce M Y( t )  bCg M e M Y( t )  0 ,
which simplifies to
(8) 1 – bgM = 0.
Thus seignorage is maximized when money growth is given by
(9) gM = 1/b.
From equation (2), we know that  = gM – gY and thus the rate of inflation that maximizes seignorage is
(10)  = (1/b) – gY .
Equation (10) implies that the higher is the growth rate of real output, the lower is the rate of inflation
that maximizes steady-state seignorage.

Problem 12.25
(a) Desired real money holdings are given by
(1) m( t )  Ce  b ( t ) .
e

The assumption is that expected inflation adjusts gradually toward actual inflation. Specifically, our
assumption is
(2)  e ( t ) = [(t) – e (t)].
As usual, seignorage is given by M  (t)/M(t)][M(t)/P(t)]. Assuming that the
 (t)/P(t) or equivalently [ M
nominal money supply is growing at rate gM (t), we can write seignorage as
(3) S(t) = gM (t)m(t).

To see the dynamics of inflation and money holdings formally, note that the growth rate of real money,
m (t)/m(t), equals the growth rate of nominal money, gM (t), minus the rate of inflation, (t). Rewriting
this as an equation for inflation gives us
(4) (t) = gM (t) – [ m
 (t)/m(t)].
Define G as the amount of real purchases that the government needs to finance with seignorage. Thus
from equation (3), we have
(5) gM (t) = G/m(t).
Taking the time derivative of both sides of equation (1) yields
 ( t )   b e ( t )Ce  b ( t ) .
e
(6) m
Dividing both sides of equation (6) by m(t) gives us
(7) m ( t ) m( t )   b e ( t ) .
Substituting equations (5) and (7) into equation (4) yields
G
(8) ( t )   b e ( t ) .
m( t )
Substituting equation (8) into equation (2) gives us
 G 
(9)  e ( t )     b e ( t )   e ( t )  .
 m( t ) 
Collecting the terms in  e ( t ) yields
 Solutions to Chapter 12 12-39

 G 
(10)  e ( t ) 1  b      e (t) ,
 m( t ) 
and thus
  G   e ( t ) m( t ) 
(11)  e ( t )   .
1  b  m( t ) 

(b) The assumption that G > S* (where S* represents the maximum steady-state value of seignorage) is
equivalent to G > e m for all possible values of e. Thus since b < 1, the right-hand side of equation
(11) is everywhere positive: regardless of where it starts, expected inflation grows without bound. To
e
examine the nature of the phase diagram, substitute m( t )  Ce  b ( t ) into equation (11):
G 
(12)  e ( t )    e (t) .
(1  b) Ce  b ( t ) 1  b
e

The following derivatives will be useful:

e
d e ( t ) bGe b ( t ) 
(13)   ,
d e ( t ) (1  b) C 1  b
and
e
d 2  e ( t ) b 2 Ge b ( t )
(14)   0.
d e ( t ) 2 (1   b) C
By setting the right-hand side of equation (13)
equal to zero, it is straightforward to show  e ( t )
that  e ( t ) reaches a minimum at e (t) =
[ln(C/bG)]/b. Thus the phase diagram has the
shape depicted in the figure at right. From
equation (1), and since e rises without
bound, the real money stock is continually
falling. If m(t) is continually falling, then
from equation (3), it must be the case that the
growth rate of the nominal money supply,
gM (t), is continually rising if the government
is to obtain G in seignorage. e (t)

(c) Now consider the case of G < S*. The left-hand figure below reproduces Figure 12.12 from the text.
It depicts the amount of seignorage the government can obtain in steady state as a function of the growth
rate of the nominal money supply. In the case of G < S*, there are two possible growth rates of the
nominal money supply, labeled g1 and g2 in the figure, consistent with raising the amount G in
seignorage. Recall that in a steady state, expected inflation equals actual inflation which in turn equals
the constant growth rate of the nominal money supply. Thus, by assumption, e (t)m(t) = G at e (t) = g1
and e (t) = g2. From equation (11) then,  e ( t ) = 0 at e (t) = g1 and e (t) = g2 . From the figure on the
left, when g1 < e (t) < g2 , we have e (t)m(t) > G and thus  e ( t ) < 0. Otherwise, e (t)m(t) < G and thus
 e ( t ) < 0. Putting all this information together gives us the phase diagram depicted on the right. The
low-inflation steady state with e (t) = (t) = g1 is stable and the high-inflation steady state with e (t) =
(t) = g2 is unstable.
12-40 Solutions to Chapter 12

S(t)  e ( t )

S*

g1 g2

e (t)

g1 g2 gM (t)