log zt = ρlogzt–1+εt.

Simulating a Simple Real

Then, consider the problem of choosing a ‘plan’ {(ct , lt , k t )}∞t =0
Business Cycle Model that maximises the expected value (evaluated at the end of
using Excel period 0) of
∞

Toru Hokari, Masaki Iimura, Seiji Murakoshi ∑ β [log c

t =0
t
t + γ log(1 − lt )]
and Yoshiko Onuma
University of Tsukuba, Japan subject to

+ (1 − δ )k t for t = 0, 1, 2,
α 1−α
c t + k t +1 ≤ z t k t lt

Abstract
The first order conditions to this problem are given by
Simulating the real business cycle models is a popular topic ⎡αz k α −1l 1−α + 1 − δ ⎤
1
in first-year graduate courses on macroeconomics. Usually, (1) = β E t ⎢ t +1 t +1 t +1 ⎥,
ct ⎢⎣ ct +1 ⎥⎦
Maple and Matlab are used for this purpose, mainly because
they can be used both for solving and for simulating the
+ (1 − δ )k t ,
α 1−α
(2) c t + k t +1 = z t k t lt
models. Strulik (2004) demonstrates that Excel can be used
both for solving and for simulating a standard RBC model. (1 − α )zt k t lt α −α
γ
In this paper, we propose a more elementary approach that (3) = ,
1 − lt ct
might be suitable for undergraduate courses. We illustrate
(i) how to solve a simple RBC model by hand and (ii) how
(4) log z t +1 = ρ log z t + ε t +1 .
to use Excel to simulate the solution.
where Et [X] represents the expected value of the random
Introduction variable X evaluated at the end of period t.

Simulating the real business cycle models (Kydland and We linearise (1)–(4) around the ‘deterministic steady-state’
Prescott, 1982; Long and Plosser, 1983) is a popular topic in (c, l, k, z) defined by
first-year graduate courses on macroeconomics. Usually,
1 1 − δ + αzk α −1l 1−α ,
Maple and Matlab are used for this purpose, mainly because =β⋅
c c
they can be used both for solving and for simulating the
models. Strulik (2004) demonstrates that Excel can be used c + k = (1 − δ )k + zk α l 1−α ,
both for solving and for simulating a standard RBC model.

In this paper, we propose a more elementary approach that γ

=
(1 − α )zk α l −α ,
might be suitable for undergraduate courses. We illustrate 1− l c
(i) how to solve a simple RBC model by hand and (ii) how
log z = ρ log z .
to use Excel to simulate the solution.
From these equations, we get
1
A simple real business cycle model ⎛ ⎞ 1−α
⎜ ⎟
Let α, β, γ, δ, ρ∈ (0,1) and z0, k0, σ∈ℜ+ be given. Let k ⎜ α ⎟ α
= ⎛k⎞
{ε t }t∞=1 and {zt }t∞=1 be stochastic processes such that for each l ⎜ 1 ⎟ (1 − α )⎜ ⎟
⎜ −1+ δ ⎟ l = , ⎝l⎠
t, εt is an i.i.d. drawn (at the beginning of period t) from the ⎝β ⎠ ⎛ ⎛ k ⎞α ⎞ α
⎜ ⎜ ⎟ − δ ⎜ ⎟ ⎟γ + (1 − α )⎛⎜ k ⎞⎟
⎛ k ⎞
normal distribution N(0,σ2), and zt is a random variable ⎜⎝ l ⎠ ⎟ ,
⎝ ⎝ l ⎠⎠ ⎝l⎠
defined by

Page 16 CHEER Volume 19

 α
⎛k⎞ ⎡ cˆt ⎤ ⎡cˆt +1 ⎤ ⎡ ε t +1 ⎤
(1 − α )⎜ ⎟ (1 − l ) ⎢kˆ ⎥ = M ⎢kˆ ⎥ + M ⎢ E [cˆ ]− cˆ ⎥
c= ⎝l⎠ ,
(10) ⎢⎢ zˆ ⎥⎥
t 0 ⎢ t +1 ⎥ 1⎢ t t +1 t +1 ⎥
γ ⎣ t⎦ ⎢⎣ zˆ t +1 ⎥⎦ ⎢⎣ Et [zˆ t +1 ]− zˆt +1 ⎥⎦,
α
⎛k ⎞
l⎜ ⎟ − c where
l
k= ⎝ ⎠ , −1
⎡ a11 0 0 ⎤ ⎡b11 b12 b13 ⎤
δ
M 0 ≡ ⎢ a 21 a 22 a 23 ⎥ ⎢0 b22 0 ⎥,
⎢ ⎥ ⎢ ⎥
z =1.
⎣⎢ 0 0 a33 ⎦⎥ ⎢⎣ 0 0 b33 ⎦⎥
−1
By linearising (1)–(4) around (c, l, k, z), we get ⎡ a11 0 0 ⎤ ⎡0 b11 b13 ⎤
M 1 ≡ ⎢⎢ a 21 a 22 a 23 ⎥⎥ ⎢0 0 0 ⎥⎥.
[ ]
α −1 1−α α −1 1−α
1 β (1 − δ + αzk l ) β (1 − α )αzk l ⎢
(5) − ⋅ cˆt = − ⋅ Et [cˆt +1 ]+ ⋅ Et lˆt +1
c c c ⎢⎣ 0 0 a33 ⎥⎦ ⎢⎣1 0 b33 ⎥⎦

Note that
β (α − 1)αzk α −1l 1−α ˆ βα zk α −1l 1−α
㸠 ⋅ k t +1 + ⋅ E t [zˆt +1 ] ,
c c ⎡ 1 ⎤
⎢ 0 0 ⎥
⎢ a11
−1
⎡ a11 0 0 ⎤ ⎥
k ⎢a a 1 a 23 ⎥
(6) ccˆt + kkˆt +1 = (1 − α ) zk α l 1−α ⋅ lˆt + ⋅ kˆt + zk α l 1−α ⋅ zˆ t , ⎢ 21 a 22 a 23 ⎥⎥ = ⎢− 21 −
β ⎢ a11 a 22 a 22 a 22 a33 ⎥
⎢⎣ 0 0 a 33 ⎥⎦ ⎢
⎢ 0 1 ⎥⎥
0
γl (1 − α ) zk α l −α α (1 − α ) zk α l −α ˆ ⎢⎣ a33 ⎥⎦
(7) − ⋅ lˆt = − ⋅ cˆt + ⋅ kt
(1 − l ) 2
c c
Thus, −1
⎡ a11 0 0 ⎤ ⎡b11 b12 b13 ⎤
α (1 − α ) zk α l −α ˆ (1 − α ) zk α l −α
−
c
⋅ lt +
c
⋅ zˆt , M 0 = ⎢ a 21 a 22 a 23 ⎥ ⎢0 b22 0⎥
⎢ ⎥ ⎢ ⎥
⎢⎣ 0 0 a33 ⎥⎦ ⎢⎣ 0 0 b33 ⎥⎦
(8) zˆt +1 = ρzˆt + ε t +1 ,
⎡ 1 ⎤
c − c ˆ lt − l z −z ⎢ 0 0⎥
k −k
where cˆt ≡ t
c
, lt ≡ , kˆt ≡ t , and zˆt ≡ t
z
. ⎢ a11 ⎥ ⎡b11 b12 b13 ⎤
l k a 1 a 23 ⎥ ⎢
From (7), we get = ⎢− 21 − 0 b22 0 ⎥⎥
⎢ a11 a 22 a 22 a 22 a33 ⎥ ⎢
⎢ ⎥ ⎢0
⎣ 0 b33 ⎦⎥
(1 − α ) zk α l −α α (1 − α ) zk α l −α ˆ (1 − α ) zk α l −α ⎢ 0 1 ⎥
− ⋅ cˆt + ⋅ kt + ⋅ zˆ t 0
c c c ⎢⎣ a33 ⎥⎦
lˆt = α −α
α (1 − α ) zk l γl .
− ⎡ b11 b12 b13 ⎤
c (1 − l ) 2
⎢ ⎥
⎢ a11 a11 a11 ⎥ ⎡ m11 m12 m13 ⎤
Thus, we can use it to eliminate lˆt and lˆt–1 from (5) and (6). a b a b b a 21b13 a 23b33 ⎥ ⎢
= ⎢− 21 11 − 21 12 + 22 − − ≡ m 21 m22 m 23 ⎥⎥.
Then, (5), (6) and (8) can be written in a matrix form as
⎢ a a a11 a 22 a 22 a11 a 22 a 22 a 33 ⎥ ⎢
⎢ 11 22 ⎥ ⎢⎣ 0 0 m33 ⎥⎦
0 ⎤ ⎡ cˆt ⎤ ⎡b11 b13 ⎤ ⎡ cˆt +1 ⎤ ⎡0 b11 ⎢ 0 b33 ⎥
⎡ a11 0 b12 b13 ⎤ ⎡ ε t +1 ⎤ 0
⎢a a 22 a 23 ⎥ ⎢kˆt ⎥ = ⎢ 0 b22 0 ⎥ ⎢kˆt +1 ⎥ + ⎢0 0 0 ⎥ ⎢ E t [cˆt +1 ]− cˆt +1 ⎥ ⎣⎢ a33 ⎦⎥
(9) ⎢ 21 ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥,
⎢⎣ 0 0 a33 ⎥⎦ ⎢⎣ zˆ t ⎥⎦ ⎢⎣ 0 0 b33 ⎥⎦ ⎢⎣ zˆ t +1 ⎥⎦ ⎢⎣1 0 b33 ⎥⎦ ⎢⎣ Et [zˆt +1 ]− zˆt +1 ⎥⎦
Note that
where
⎡m11 − λ m12 m13 ⎤
1 ⎢ ⎥
a11 ≡ − , M 0 − λ I = ⎢ m21 m22 − λ m23 ⎥
c
⎢ 0 0 m33 − λ ⎥⎦
β (1 − α )αzk α −1l 1−α ⎣
β (1 − δ + αzk α −1l 1−α ) c (1 − α ) zk α l −α
b11 ≡ − + ⋅ , and
c α (1 − α ) zk α l −α γl c
− 2
c (1 − l ) M 0 − λ I = (m11 − λ )( m22 − λ )(m33 − λ ) − m12 m21 (m33 − λ )
β (α − 1)αzk α −1l 1−α
β (α − 1)αzk α −1l 1−α c α (1 − α ) zk α l −α
b12 ≡ + α −α
⋅ , = (m33 − λ )(m11 m 22 − m12 m 21 − (m11 + m22 )λ + λ2 ).
c α (1 − α ) zk l γl c
−
c (1 − l ) 2 Thus, the eigen values of M0 and a corresponding matrix Q
β (1 − α )αzk α −1l 1−α of eigen vectors are given by
βα zk α −1l 1−α c (1 − α ) zk α l −α
b13 ≡ + α −α
⋅ , m11 + m22 − (m11 + m 22 ) 2 − 4(m11 m 22 − m12 m 21 )
c α (1 − α ) zk l γl c λ1 ≡ ,
− 2
c (1 − l ) 2
(1 − α )αzk α l 1−α (1 − α ) zk α l −α m11 + m 22 + (m11 + m22 ) 2 − 4(m11 m22 − m12 m 21 )
a 21 ≡ c + ⋅ , λ2 ≡ ,
α −α
α (1 − α ) zk l γl c 2
−
c (1 − l ) 2 λ3 ≡ m33 ,
k (1 − α )αzk α l 1−α α (1 − α ) zk α l −α
a 22 ≡− − ⋅ , and
β α (1 − α ) zk α l −α γl c
−
c (1 − l ) 2 ⎡ m12 m23 − m13 (m 22 − λ3 ) ⎤
b22 ≡ − k , ⎢ m12 m12
(m11 − λ 3 )( m21 − λ3 ) − m12 m21 ⎥ ⎡ q
⎢ ⎥ 11 q12 q13 ⎤
a33 ≡ ρ , m21 m13 − m23 (m11 − λ3 ) ⎥ ⎢
Q = ⎢λ1 − m11 λ 2 − m11 ≡ ⎢q 21 q 22 q 23 ⎥⎥.
⎢ (m11 − λ 3 )( m21 − λ3 ) − m12 m21 ⎥
b33 ≡ −1. ⎢ 0 ⎥ ⎢⎣ 0 0 1 ⎥⎦
0 1
⎢ ⎥
In what follows, we use Farmer’s (1999) method to solve ⎢⎣ ⎦⎥
the linearised system (9). From (9), we get
Then,

CHEER Volume 19 Page 17

 ⎡ q 22 − q12 − q 22 q13 + q12 q 23 ⎤
Simulating the model using Excel
1 ⎢− q
Q −1
= q11 q 21 q13 − q11 q 23 ⎥ Step 1. (Figure 1) Choose the values of parameters, α, β, γ,
q11 q 22 − q12 q 21 ⎢ ⎥
21

⎣⎢ 0 0 q11 q 22 − q12 q 21 ⎦⎥ , δ, ρ, and σ. Here, we use the same values as in Table 2 of

King and Rebelo (1999: 955).
and
⎡λ1 0 0⎤
M 0 = Q⎢ 0 λ2 0 ⎥ Q −1 .
⎢ ⎥
⎣⎢ 0 0 λ3 ⎦⎥

By pre-multiplying both sides of (10) by Q–1, and by putting

⎡ xt1 ⎤ ⎡ cˆt ⎤
⎢ 2⎥ −1 ⎢ ˆ ⎥ ,
x
⎢ t⎥ ≡ Q k
⎢ t⎥
⎢ x t3 ⎥ ⎢
⎣ zˆt ⎦⎥
⎣ ⎦
we obtain
⎡ xt1 ⎤ ⎡λ1 0 0 ⎤ ⎡ xt1+1 ⎤ ⎡ ε t +1 ⎤
⎢ 2⎥ ⎢ ⎢ ⎥
⎢ xt ⎥ = ⎢ 0 λ2 0 ⎥ ⎢ xt2+1 ⎥ + Q −1 M 1 ⎢ Et [cˆt +1 ]− cˆt +1 ⎥ .
⎥ ⎢ ⎥
⎢ x t3 ⎥ ⎢⎣ 0
⎣ ⎦ 0 λ3 ⎥⎦ ⎢⎣ xt3+1 ⎥⎦ ⎢⎣ Et [zˆt +1 ]− zˆt +1 ⎥⎦

Taking the expected value of both sides evaluated at the

end of period t, Figure 1
⎡x 1
t
⎤ ⎡λ1 0 0 ⎤ ⎡ Et x [ ]⎤ 1
t +1
Step 2. (Figure 2) Calculate the deterministic steady-state (c,
⎢ ⎥ ⎢ ⎢
⎢x
2
t ⎥ =⎢0 λ2 0 ⎥ ⎢ Et x
⎥
[ ]⎥⎥ 2
t +1
.
l, k, z). Since
⎢x
⎣
3
t
⎥ ⎣⎢ 0
⎦ 0 [ ]⎥⎦
λ3 ⎦⎥ ⎢⎣ E t x 3
t +1
1
⎛ ⎞ 1−α
⎜ ⎟
It follows that for each T>0, x 1t = λ1 Et [xt1+T ]. Also, it can be T
k ⎜
=
α ⎟ ,
shown that λ1 <1. Thus, l ⎜1 ⎟
⎜β − 1 + δ ⎟
T
[ ]
x 1t = lim λ1 Et xt1+T = 0 ,
T →∞
⎝ ⎠
enter
so that =($B$3/(1/$B$4-1+$B$6))^(1/(1-$B$3))
q 22 cˆt − q12 kˆt + (q12 q 23 − q 22 q13 ) zˆt = 0 . in cell D3. Similarly, enter the formulas for l, c, k, and z in
cells D4–D7.
Thus, the solution to the linearised system (10) is given by
q12 ˆ ⎛ q q ⎞
(11) cˆt = ⋅ k t + ⎜⎜ q13 − 12 23 ⎟⎟ ⋅ zˆ t ,
q 22 ⎝ q 22 ⎠
ˆ
a cˆ + a 22 k t + a 23 zˆt
(12) kˆt +1 = 21 t ,
b22

(13) zˆt +1 = ρzˆt + ε t +1 .

Recall that from (7), we have

(14) lˆt = d1cˆt + d 2 kˆt + d 3 zˆt ,
1 (1 − α ) zk α l −α
d1 ≡ −
where α −α
⋅ ,
α (1 − α ) zk l γl c
−
c (1 − l ) 2 Figure 2
1 α (1 − α ) zk α l −α
d2 ≡ α −α
⋅ ,
α (1 − α ) zk l γl c
− Step 3. (Figure 3) Enter the formulas for the components of
c (1 − l ) 2
matrices A, B, and M0, the eigen values of m0, and the
1 (1 − α ) zk α l −α
d3 ≡ ⋅ . components of matrix Q. For instance, since
α (1 − α ) zk α l −α γl c
−
c (1 − l ) 2 zˆ1 = ρzˆ 0 + zε 1 ,
a cˆ + a 22 kˆ0 + a 23 zˆ 0
Given the values of z0, k0, and {ε t }t∞=1 , we can use (11)–(14) kˆ1 = 21 0 ,
b22
to simulate the evolution of ct, lt, kt, and zt.
enter
= -$B$17*$B$15/($B$13*$B$18)+$B$20/$B$18
in cell D17.

Step 4. (Figure 4) Enter 0 in the cells from A30 to J30.

Step 5. Enter =A30+1 in cell A31.

Step 6. Enter =NORMINV(RAND(),0,$B$8^2) in cell B31.

Step 7. (Figure 5) Enter the formulas for zˆ1, k̂1, cˆ1, and lˆ1 in
cells C31–F31. For instance, since

Page 18 CHEER Volume 19

 zˆ1 = ρzˆ 0 + zε 1 ,
a cˆ + a 22 kˆ0 + a 23 zˆ 0
kˆ1 = 21 0 ,
b22
enter
=$B$7*C30+$D$7*B31|
in cell C31, and
=($B$17*G30+$B$18*D30+$B$19*C30)/$B$20
in cell D31.

Step 8. Copy the cells A31–F31 into subsequent rows. Then

we obtain a simulated time series of z1, k̂1, cˆ1, and lˆ1.

Let
Figure 3 it ≡ k t +1 − (1 − δ )k t ,
y t ≡ ct + it ,
i ≡ k − (1 − δ )k = δk ,
y ≡ c + k − (1 − δ )k = c + δk ,
y − y ccˆt + kkˆt +1 − (1 − δ )kkˆt
yˆ t ≡ t = ,
y y
i − i kkˆt +1 − (1 − δ )kkˆt
iˆt ≡ t = ,
i i
Figure 4 rt ≡ αzt k t lt − δ = α ( zzˆt + z )(kkˆt + k )α −1 (llˆt + l )1−α − δ ,
α −1 1−α

w ≡ (1 − α ) z k l = (1 − α )( zzˆ + z )( kkˆ + k )α (llˆ + l ) −α ,

α −α
t t t t t t t

w ≡ (1 − α ) zk α l −α ,
wt − w (1 − α )( zzˆt + z )(kkˆt + k ) α (llˆt + l ) −α − w
wˆ t ≡ = .
w w

By considering these variables together with z1, k̂1, cˆ1, and

Figure 5
lˆ1, one can create a table and a figure (Figures 6 and 7)
similar to Table 3 and Figure 10 of King and Rebelo (1999).1

Figure 6
CHEER Volume 19 Page 19
 Notes
1 Sample Excel files are available at
http://member.social.tsukuba.ac.jp/hokari/

References
Farmer, R.E.A. (1999) Macroeconomics of Self-fulfilling
Prophecies,
2nd ed, MIT Press.

King, R.G., and S.T. Rebelo (1999) ‘Resuscitating real business

cycles’ in Handbook of Macroeconomics, volume 1B, by J.B.
Taylor and M. Woodford (Eds), Elsevier, 927–1007.

Kydland, F. and E. Prescott (1982) ‘Time to build and aggregate

fluctuations’, Econometrica 50, 1345–1371.

Long, Jr., J.B. and C.I. Plosser (1983) ‘Real business cycles’,
Journal of Political Economy 91, 39–69.

Strulik, H. (2004) ‘Solving rational expectations models using

Excel’, Journal of Economic Education 35, 269–283.

Contact details
Toru Hokari
Associate Professor
Graduate School of Humanities and Social Sciences
University of Tsukuba
1-1-1 Ten’nodai
Tsukuba
Figure 7 Ibaraki, 305-8571
Japan
Email: [email protected]
Fax: +81-298-53-6611

Page 20 CHEER Volume 19