Money in the Utility Function

Macroeconomı́a II
Maestrı́a en Economı́a
UTDT

Francisco J. Ciocchini

[email protected]

2023

1/43
Plan
I In this lecture

I General discussion about money.

I Set up an RBC model with Money.

I Introduce Money in the Utility Function (MIUF) to derive a

money-demand function.

I Obtain the equilibrium conditions of the model.

I Log-linearize the equilibrium conditions around the nonstochastic

steady sate.

I Write the log-linearized model in matrix form.

I Calibrate the model and compute the solution using Uhlig’s method.

I Display impulse response functions, obtain second moments, and

analyze results.
2/43
RBC Model with MIUF
I Discrete time, infinite horizon: t 2 {0, 1, 2, ...}.
I One-sector model.
I Closed economy.
I Government: issues money through lump-sum transfers.
I No long-run growth in productivity: g = 0.
I No population growth: n = 0.
I Notation similar to Cooley & Hansen (1989).
I Aggregate production function:

Yt = Zt Kt✓ Ht1 ✓
✓ 2 (0, 1)

I TFP:
ln Zt+1 = (1 ) ln Z + ln Zt + "t+1

where 2 (0, 1), Z > 0, and {"t+1 }1 t=0 is a sequence of zero-mean i.i.d.
random variables with variance 2" .

3/43
RBC Model with MIUF
I Government
Vt = Mt Mt 1

= µt M t 1 Mt 1

= (µt 1)Mt 1

where Vt are lump-sum transfers to the private sector during period t, Mt is the
money supply at the end of t, the initial stock M 1 > 0 is given, and µt is the
gross growth rate of the money supply between t 1 and t.

I We assume:
ln µt+1 = (1 ↵) ln µ + ↵ ln µt + ⌘ t+1

with ↵ 2 (0, 1), µ > 0, and {⌘ t+1 }1

t=0 ⇠ IID(0,
2 ).
⌘

I Money supply shocks (⌘ t ) and TFP shocks ("t ) are independent from
each other.
4/43
RBC Model with MIUF

I Representative Household
I Expected Utility:
P1 ⇣ ⌘
E0 t Mt
t=0 u Ct , H t , Pt

where ⇣ ⌘ ⇣ ⌘
Mt Mt
u Ct , H t , pt
= ln Ct BHt + a ln Pt

with 2 (0, 1), B > 0 and a > 0. We assume < µ.

I Budget Constraint:

Bt Mt
Ct + Kt+1 + Pt
+ Pt
= wt Ht + Rt Kt + ⇧t + (1 )Kt
(1+it 1 )Bt 1 Mt 1 Vt
+ Pt
+ Pt
+ Pt

where ⇧t = dividends, and M 1 > 0, B 1 = 0, K0 > 0 are given.

5/43
Profit Maximization

I Profit maximization:

⇧t ⌘ maxKt ,Ht Zt Kt✓ Ht1 ✓

w t Ht R t Kt

I FOC: ⇣ ⌘ (1 ✓)
Kt Yt
Rt = ✓Zt Ht
= ✓K t

⇣ ⌘✓
Kt Yt
wt = (1 ✓) Zt Ht
= (1 ✓) Ht

I Maximized profits (= dividends):

⇧t = 0

6/43
Utility Maximization
I Lagrangian:

8 P1 t 9
>
> t=0 u (Ct , Ht , Mt /Pt ) + >
>
>
< >
=
L = E0 " (1+it )Bt Mt #
>
> P1 t wt Ht + Rt Kt + ⇧t + (1 )Kt + 1
Pt
1
+ Pt
1
>
>
>
: t=0 t V Bt Mt >
;
+ Pt Ct Kt+1 Pt Pt
t

I FOC
Ct : uC (t) = t

Ht : uH (t) = t wt

Mt : u M (t) P1t 1
tP
t
+ Et { t+1 P
1
t+1
}=0
P

Kt+1 : t = Et { t+1 (Rt+1 +1 )}

1+it
Bt : 1
tP
t
= Et { t+1 P
t+1
}

plus the budget constraint and appropriate transversality conditions.

7/43
Utility Maximization

I From the FOC w.r.t. Ct and Ht we get:

uH (t)
= wt
uC (t)

I From the FOC w.r.t. Ct and Kt+1 we get:

uC (t) = Et {uC (t + 1)(Rt+1 + 1 )}

I From the FOC w.r.t. Ct and Bt we get:

n o
uC (t) = Et uC (t + 1)(1 + it ) PPt+1
t

n o
uC (t) = (1 + it ) Et uC (t + 1) PPt+1
t

8/43
Utility Maximization
I From the FOC w.r.t. Ct and Mt we get:

n o
uC (t) = u M (t) + Et uC (t + 1) PPt+1
t
P

I Combining the previous two expressions we get:

uC (t)
uC (t) = u M (t) +
P 1 + it

I Then:
u M (t) it
P
=
uC (t) 1 + it

9/43
Utility Maximization
I For u (C, H, M/P ) = ln C BH + a ln (M/P ) we have:
1
uC = C

uH = B

a
uM = M
P P

I Define:
Mt
mt ⌘ Pt

Pt+1
⇡ t+1 ⌘ Pt

I Then, uH (t)
uC (t)
= wt )
BCt = wt

10/43
Utility Maximization
I From uC (t) = Et {uC (t + 1)(Rt+1 + 1 )} we get:

n o
Ct
1
= Et Ct+1
(Rt+1 +1 )

n o
I From uC (t) = u M (t) + Et uC (t + 1) PPt we get:
P t+1

n o
1
Ct
= a
mt
+ Et 1 1
Ct+1 ⇡ t+1

u M (t)
I From P
uC (t)
= it
1+it
we obtain the money-demand function:

mt = a 1+i
it
t
Ct

11/43
Market Clearing

I For each t, market clearing requires:

Yts = Ytd

Kts = Ktd

Hts = Htd

Mtd = Mts

Btd = 0

where Ytd ⌘ Ctd + Itd .

12/43
Walras Law
I From the budget constraint of the representative household we get:
d Mtd
d d Bd Md s s (1+it )B
t V
Ct + It + Pt + P t = wt Ht + Rt Kt + ⇧t + + Pt
1 1 1
Pt
+ Pt
t t t

where we have used It = Kt+1 (1 )Kt .

I Using ⇧t = Yts wt Htd Rt Ktd and Vt = Mts Mts 1 :

d
Bt Mtd
d d s s s d d
Ct + It + Pt
+ Pt
= w t H t + Rt K t + Y t wt H t Rt K t

(1+it
d
)B Mtd Mts Mts
1 t 1 1 1
+ Pt
+ Pt
+ Pt

I Then:
!
Bd Mtd Mts
(Ctd + Itd Yts ) + wt (Htd Hts ) + Rt (Ktd Kts ) + Pt + Pt Pt
t

d
!
(1+it 1
)B
t 1
Mtd 1
Mts 1
= Pt
+ Pt Pt

I Using Ytd ⌘ Ctd + Itd and assuming market clearing in t 1 we get:

✓ ◆
d
Bt Mtd Mts
(Ytd Yts ) + wt (Htd Hts ) + Rt (Ktd Kts ) + Pt + Pt Pt =0

13/43
Equilibrium
I In a competitive equilibrium, market clearing must hold in every period. Then:

Yts = Ytd = Yt

Kts = Ktd = Kt

Hts = Htd = Ht

Mtd = Mts = Mt

and Btd = 0 is implied by Walras Law.

I Also:
Mt
mdt = mst = mt =
Pt

I In addition, from Mt = µt Mt 1 we get: Mt

Pt
= µt
Mt
Pt
1
. Then:
Mt P t 1 Mt 1
Pt
= µt P Pt
)
t 1
µt
mt = mt 1
⇡t

14/43
Equilibrium
I We can summarize the equilibrium conditions as follows (12 equations, 12 variables):

Yt = Zt Kt✓ Ht1 ✓

Yt
wt = (1 ✓) Ht

Y
Rt = ✓ Kt
t

It = Kt+1 (1 )Kt

Y t = Ct + It

1+it
mt = a it Ct

BCt = wt

µt
mt = ⇡t mt 1

n o
= Et
1 Ct
Ct+1 (Rt+1 + 1 )

n o
1
Ct = a
mt + Et 1 1
Ct+1 ⇡ t+1

ln Zt+1 = (1 ) ln Z + ln Zt + "t+1

ln µt+1 = (1 ↵) ln µ + ↵ ln µt + ⌘ t+1
15/43
Nonstochastic Steady State
I We use upper bars to denote the nonstochastic steady state (nss).

I Setting "t+1 = 0 and ⌘ t+1 = 0 8t we get:

Z=Z

µ=µ

I From mt = µt
⇡t
mt 1 we get: m = µ
⇡
m. Then: ⇡ = µ )

⇡=µ

n o
I From 1
= Et Ct
Ct+1
(Rt+1 +1 ) , Ct = Ct+1 = C, and Rt+1 = R we
1
get: R = (1 ))
R=⇢+
1
where we have used ⌘ 1+⇢
.
16/43
Nonstochastic Steady State
⇣ ⌘ (1 ✓) ⇣ ⌘ (1 ✓)
I From Rt = ✓Zt Kt
Ht
we get R = ✓Z K
. Then:
H

✓ ◆ 1
K ✓Z 1 ✓
=
H R

⇣ ⌘✓ ⇣ ⌘✓
I From wt = (1 ✓) Zt Kt
Ht
we get: w = (1 ✓) Z K
)
H

⇣ ⌘ ✓
✓Z 1 ✓
w = (1 ✓) Z R

I From BCt = wt we get:

w
C= B

17/43
Nonstochastic Steady State
n o
I From 1
Ct
= mat + Et 1 1
Ct+1 ⇡ t+1
we get 1
= a
m
+ 1 1
. Then, using
C C ⇡
⇡ = µ, we get:

m = a µµ C

I Recall we have assumed µ > . Then, m > 0.

I From mt = a 1+i
it
t
Ct we get m = a 1+i C. Using the previous expression
i
to eliminate m we get: a µ µ C = a 1+i
i
C. Then: µ
µ
= 1+i
i
. Then:

µ
i=

I Since we have assumed µ > , we get i > 0.

18/43
Nonstochastic Steady State
I From the previous expression we get 1 + i = 1 µ. Using µ = ⇡ :

1
1+i= ⇡

I Notice that, in the nss, the gross real interest rate is 1

, and the net real
1
interest rate is 1 = ⇢. This is consistent with the expression we found
earlier for the rental rate of capital: R = ⇢.

I Now we write the rest of the variables in terms of K and then solve for K.

I From It = Kt+1 (1 )Kt :

I= K
⇣ ⌘ 1
I From K
= ✓Z 1 ✓
:
H R
⇣ ⌘ 1
R 1 ✓
H= ✓Z
K

19/43
Nonstochastic Steady State

I From Yt = Zt Kt✓ H⇣t1 ✓⌘we get Y = ZK ✓ H 1 ✓

. Using the expression
✓ R 1 ✓
above: Y = ZK ✓Z
K . Then:

Y = ✓1 RK

I From Yt = Ct + It we get Y = C + I. Using the previous expressions:

1 w
✓
RK = B
+ K. Then:

w
K=
(R
✓
)B

where K > 0 because R = ⇢ + > and ✓ 2 (0, 1).

I Having found K we can go back and recover I, H and Y .

I Remark: in the nss, the only real variable that depends on µ is m (because
it depends on the nominal interest rate i, which increases with ⇡ = µ).

20/43
Log-linearization
I For any variable X define:

et ⌘ ln Xt
X ln X

I Using Uhlig’s method to log-linearize the equilibrium conditions we get:

Yet = Z
et + ✓K
e t + (1 et
✓)H

et = Yet
w et
H

et = Yet
R et
K

Iet = K
e t+1 (1 et
)K

Y Yet = C C
et + I Iet

et
et = C
m eit
µ

21/43
Log-linearization

I Log-linearization (cont.):

et = w
C et

et = m
m et 1 et
+µ et
⇡
n o
et+1
0 = Et C et
C et+1
RR
n o
0 = Et et+1 + ⇡
C et+1 et + (µ
µC et
)m

et+1 = Z
Z et + "t+1 with "t+1 ⇠ IID 0, 2
"

2
et+1 = ↵e
µ µt + ⌘ t+1 with ⌘ t+1 ⇠ IID 0, ⌘

22/43
Price Level
I From ⇡ t+1 ⌘ Pt+1
Pt
we get Pt+1 = ⇡ t+1 Pt )

ln Pt+1 = ln ⇡ t+1 + ln Pt

I From ⇡
et+1 = ln ⇡ t+1 ln ⇡ and ⇡ = µ we get:

et+1
ln ⇡ t+1 = ln µ + ⇡

I Combining the expressions above:

et+1 + ln Pt
ln Pt+1 = ln µ + ⇡

I At t = 0, P0 is determined from the money-market equilibrium condition:

m0 = a 1+i
i0
0
C0 ) M0
P0
= a 1+i
i0
0
C0 )
1 i0
P0 = aC0 1+i0
M0

I From the solution of the model, we know C e0 and ei0 .

I e
Therefore, we know C0 = CeC0 and i0 = iei0 .
e

I Hence, given M0 , the expression above determines P0 .

I Knowing P0 and {e ⇡ t+1 }1
t=0 (that we obtained from the solution), we can
1
compute {Pt+1 }t=0 from ln Pt+1 = ln µ + ⇡ et+1 + ln Pt .
23/43
Matrix Form
I Define:
2 3
Yet
6 C et 7
6 7
6 7
 6 Iet 7  
e t+1 6 et 7 et
K 6 H 7 Z "t
xt ⌘ yt ⌘ 6 7 zt ⌘ ✏t ⌘
et
m 6 e
Rt 7 et
µ ⌘t
6 7
6 wet 7
6 7
4 eit 5
et
⇡

I Rewrite our log-linearized system in matrix form, as follows:

0 = Axt + Bxt 1 + Cyt + Dzt

0 = Et {F xt+1 + Gxt + Hxt 1 + Jyt+1 + Kyt + Lzt+1 + M zt }

zt+1 = N zt + ✏t+1

24/43
Matrix Form

I Matrices A, B, C, D are the following:

2 0 0 3 2 ✓ 0 3
6 0 0 7 6 0 0 7
6 7 6 7
6 0 0 7 6 1 0 7
6 7 6 7
6 1 0 7 6 (1 ) 0 7
A=6 7 B=6 7
6 0 0 7 6 0 0 7
6 7 6 7
6 0 1 7 6 0 0 7
4 0 0 5 4 0 0 5
0 1 0 1
2 3 2 3
1 0 0 (1 ✓) 0 0 0 0 1 0
6 1 0 0 1 0 1 0 0 7 6 0 0 7
6 7 6 7
6 1 0 0 0 1 0 0 0 7 6 0 0 7
6 7 6 7
6 0 0 0 0 0 0 0 7 6 0 0 7
C=6
6 Y C I 0 0 0 0 0
7
7 D=6 7
6 7 6 0 0 7
6 7 6 7
6 0 1 0 0 0 0 µ
0 7 6 0 0 7
4 5 4 0 0 5
0 1 0 0 0 1 0 0
0 0 0 0 0 0 0 1 0 1

25/43
Matrix Form
I Matrices F, G, H, J, K, L, M, N are the following:

  
0 0 0 0 0 0
F = G= H=
0 0 0 µ 0 0


0 1 0 0 R 0 0 0
J =
0 0 0 0 0 0


0 1 0 0 0 0 0 0 0
K =
0 µ 0 0 0 0 0 0 0

 
0 0 0 0
L = M=
0 0 0 0


0
N =
0 ↵

26/43
Calibration
I For quarterly data:

Z = 1
✓ = 0.36
= 0.99
B = 2.86
= 0.025
= 0.95
↵ = 0.48
a = 0.01
" = 0.00721
⌘ = 0.009

Similar to the Cash-in-Advance (CIA) model in Cooley & Hansen (1989). The value of a is
chosen so that, when µ = 1, the value of m is similar to the one implied by Cooley &
Hansen (1989). This is the value chosen by McCandless (2008), Chapter 9.

27/43
Calibration
I We solve the model for two di↵erent values of µ, 1.015 and 1.15. We get:

µ = 1.015 µ = 1.15

R 0.0351 0.0351

w 2.3706 2.3706

C 0.8288 0.8288

K 11.4324 11.4324

H 0.3009 0.3009

Y 1.1147 1.1147

I 0.2858 0.2858

m 0.3365 0.0596

i 0.0253 0.1616

⇡ 1.0150 1.1500
28/43
Solution

I We use Uhlig’s method to find a recursive solution of the form:

xt = P xt 1 + Q zt
yt = R xt 1 + S zt

I For µ = 1.015 we get:

 
0.9418 0 0.1552 0
P = Q=
0.5316 0 0.4703 0.8803
2 3 2 3
0.0550 0 1.9417 0
6 0.5316 0 7 6 0.4703 0 7
6 7 6 7
6 1.3273 0 7 6 6.2091 0 7
6 7 6 7
6 0.4766 0 7 6 1.4715 0 7
R=6
6
7
7 S=6
6
7
7
6 0.9450 0 7 6 1.9417 0 7
6 0.5316 0 7 6 0.4703 0 7
6 7 6 7
4 0.0000 0 5 4 0.0000 0.9026 5
0.5316 1 0.4703 1.8803

29/43
Solution
I Then:
e t+1 = 0.9418K
K e t + 0.1552Z
et

m e t + 0.4703Z
e t = 0.5316K et 0.8803e
µt

Yet = 0.0550K
e t + 1.9417Z
et

et = 0.5316K
C e t + 0.4703Z
et

Iet = e t + 6.2091Z
1.3273K et

et =
H e t + 1.4715Z
0.4766K et

et =
R e t + 1.9417Z
0.9450K et

e t + 0.4703Z
et = 0.5316K
w et

eit = 0.9026e
µt

et =
⇡ et + m
0.5316K et 1
et + 1.8803e
0.4703Z µt

30/43
Solution
I Notice that the solution for K
e t+1 , Yet , C
et , Iet , H
et, R
e t and w
et coincides
with the solution of the real model in Hansen (1985). Upon reflection this
is not surprising since the system of equations that characterizes the
equilibrium can be broken down in two blocks.

I Block 1
Yet = Z
et + ✓ K
e t + (1 et
✓)H

et = Yet
w et
H

et = Yet
R et
K

Iet = K
e t+1 (1 et
)K

Y Yet = C C
et + I Iet

et = w
C et
n o
0 = Et Cet+1 et
C et+1
RR

et+1 = Z
Z et + "t+1

31/43
Solution
I Block 2
et
et = C
m ei
µ t

et = m
m et 1 et
+µ et
⇡
n o
0 = Et et+1 + ⇡
C et+1 et + (µ
µC et
)m

et+1 = ↵e
µ µt + ⌘ t+1

I Block 1 is exactly the same as the one in Hansen’s model. This block of 8 equations can be
e t+1 , Yet , C
solved to obtain the paths of K et , Iet , H
et , R
et , w et (8 variables). Given
et and Z
et obtained from Block 1, we can solve Block 2 (4 equations) to obtain the
the solution for C
e t , eit , ⇡
paths of m e t and µ
et (4 variables).
I Separability of real balances in the utility function is important for this result.
I The only real variable that is a↵ected by money growth is mt (real balances). When this
happens we say that money is superneutral. It is natural to expect that m is a↵ected by
money growth since µ a↵ects the nominal interest rate, which is a determinant of the
demand for real balances.

32/43
Simulations
I Percent Standard Deviations
Real Balances 0.9445 0.54

Output 1.7619 1.00

Consumption 0.5069 0.29

Investment 5.6316 3.20

Labor 1.3415 0.76

Rental Rate 1.7945 1.02

Real Wage 0.5069 0.29

Nominal Interest Rate 0.8201 0.47

Gross Inflation 1.6533 0.94

Results based on 100 simulations. Each simulation consists of 115 periods, which is the
sample size in Cooley & Hansen (1989). Data was detrended using the HP filter with
= 1600. Second column: standard deviations relative to output.

33/43
Simulations
I Contemporaneous Correlations with Output

Real Balances 0.45 (0.10)

Output 1.00 (0.00)

Consumption 0.88 (0.02)

Investment 0.99 (0.00)

Labor 0.98 (0.00)

Rental Rate 0.96 (0.01)

Wage 0.88 (0.02)

Nominal Interest Rate 0.01 (0.13)

Gross Inflation 0.11 (0.10)

Results based on 100 simulations. Each simulation consists of 115 periods, which is the
sample size in Cooley & Hansen (1989). Data was detrended using the HP filter with
= 1600. Second column: small sample standard errors (simulation based).

34/43
Impulse-Response Functions
I TFP shock

35/43
Impulse-Response Functions
I Shock to money growth

36/43
Modified version of the model
I Government:
Mt Mt 1
Gt = Pt

Mt Mt 1
where Gt = government purchases of goods. and Pt
⌘ St is seigniorage.

I Eliminate lump-sum transfers (Vt ) from the budget constraint of the

household.

I We assume Gt is exogenous and Mt is endogenous (fiscal dominance).

I In particular, assume:

ln Gt+1 = (1 ) ln G + ln Gt + ⇠ t+1

I Now the resource constraint is:

Y t = C t + It + G t

37/43
Modified version of the model
I It’s not difficult to show that, in the nonstochastic steady state,
seigniorage is:

µ 1 µ 1 aC
S= µ
m = µ 1
µ

µ 1
S= µ
aC

I The steady-state value of Gt is:

G=G

I The growth rate of money is now endogenous. Its equilibrium value, µ,

solves G = S. Then µ is determined by:

µ 1
G= µ aC

I Remark: in this model, C is independent of µ and G.

38/43
Modified version of the model
I Determination of µ

39/43
Modified version of the model
I Increase in G

40/43
Appendix: MIUF when the nominal interest rate is zero
I From the FOC w.r.t. Ct and Mt we get:
n o
uC (t) = u M (t) + Et uC (t + 1) PPt+1
t
(1)
P

I From the FOC w.r.t. Ct and Bt we get:

n o
uC (t) = (1 + it ) Et uC (t + 1) PPt+1
t
(2)

I Suppose it = 0. Then, (2) becomes:

n o
uC (t) = Et uC (t + 1) PPt+1
t
(3)

I Substituting (1) into (3) we get:

n o n o
u M (t) + Et uC (t + 1) PPt+1
t
= Et uC (t + 1) PPt+1
t
P

41/43
Appendix: MIUF when the nominal interest rate is zero
I Then
u M (t) = 0
P

I The expression above implies that, when it = 0, the household would

demand real balances until satiation (zero marginal utility).

I Notice, however, that u M (t) = 0 cannot hold if the utility function always
P
displays u M > 0, like in our logarithmic case.
P

I What’s happening?

I Well, with it = 0 and u M > 0, the household could borrow like crazy
P
(Bt /Pt ! 1) and use the proceeds to increase real balances
(Mt /Pt ! +1). This operation would always increase utility (since
u M > 0). Hence, the maximization problem of the household would have
P
no solution.

42/43
Bibliography

I Cooley, T. and Hansen, G. (1989). “The Inflation Tax in a Real Business

Cycle Model,” American Economic Review , Vol. 79, No. 4 (September),
pp. 733-748.

I McCandless, George (2008). The ABCs of RBCs: An Introduction to

Dynamic Macroeconomic Models, Harvard Univ. Press. Chapters 8 & 9.

I Uhlig, Harald (1999). “A Toolkit for Analysing Nonlinear Dynamic

Stochastic Models Easily,” in Ramon Marimon and Andrew Scott (eds.),
Computational Methods for the Study of Dynamic Economies, Oxford
University Press, pp. 30-61.
https://www.wiwi.hu-berlin.de/de/professuren/vwl/wipo/research/MATLAB Toolkit

I Walsh, Carl (2017). Monetary Theory and Policy , 4th edition, MIT Press.
Chapters 2 & 3.

43/43