A Baseline DSGE Model

Jesús Fernández-Villaverde

Duke University, NBER, and CEPR

Juan F. Rubio-Ramírez

Duke University and Federal Reserve Bank of Atlanta

October 10, 2006

1
1. Introduction

In these notes, we present a baseline sticky prices-sticky wages model. The basic structure of
the economy is as follows. A representative household consumes, saves, holds money, supplies
labor, and sets its own wages subject to a demand curve and Calvo’s pricing. The final output
is manufactured by a final good producer, which uses as inputs a continuum of intermediate
goods manufactured by monopolistic competitors. The intermediate good producers rent
capital and labor to manufacture their good. Also, these intermediate good producers face
the constraint that they can only change prices following a Calvo’s rule. Finally, there is
a monetary authority that fixes the one-period nominal interest rate through open market
operations with public debt.

1.1. Households

There is a continuum of households in the economy indexed by j. The households maximizes

the following lifetime utility function, which is separable in consumption, cjt , real money
balances, mjt /pt (where pt is the price level), and hours worked, ljt :
( µ ¶ )
X
∞
mjt
1+γ
ljt
E0 β t dt log (cjt − hcjt−1 ) + υ log − ϕt ψ
t=0
pt 1+γ

where β is the discount factor, h is the parameter that controls habit persistence, γ is the
inverse of Frisch labor supply elasticity, dt is an intertemporal preference shock with law of
motion:
log dt = ρd log dt−1 + σ d εd,t where εd,t ∼ N (0, 1),

and ϕt is a labor supply shock with law of motion:

log ϕt = ρϕ log ϕt−1 + σ ϕ εϕ,t where εϕ,t ∼ N (0, 1).

Note that the preference shifters are common for all households. Also, we have selected a
utility function (log utility in consumption) whose marginal relation of substitution between
consumption and leisure is linear in consumption to ensure the presence of a balanced growth
path with constant hours.
Households can trade on the whole set of possible Arrow-Debreu commodities, indexed
both by the household j (since the household faces idiosyncractic wage-adjustment risk that
we will describe below) and by time (to capture aggregate risk). Our notation ajt+1 indicates
the amount of those securities that pay one unit of consumption in event ω j,t+1,t purchased

2
by household j at time t at (real) price qjt+1,t . To save on notation, we drop the explicit
dependence of qjt+1,t and ajt+1 on the event when no ambiguity arises. Summing over different
individual assets we can price securities contingent only on aggregate states. Households also
hold an amount bjt of government bonds that pay a nominal gross interest rate of Rt .
Then, the j − th household’s budget constraint is given by:
Z
mjt bjt+1
cjt + xjt + + + qjt+1,t ajt+1 dω j,t+1,t
pt pt
¡ ¢ mjt−1 bjt
= wjt ljt + rt ujt − μ−1
t a [ujt ] kjt−1 + + Rt−1 + ajt + Tt + zt
pt pt

where wjt is the real wage, rt the real rental price of capital, ujt > 0 the intensity of use of
capital, μ−1
t a [ujt ] is the physical cost of use of capital in resource terms, μt is an investment-
specific technological shock to be described momentarily, , Tt is a lump-sum transfer, and zt
are the profits of the firms in the economy. We assume that a [1] = 0, a0 and a00 > 0.
Investment xjt induces a law of motion for capital
µ ∙ ¸¶
xjt
kjt = (1 − δ) kjt−1 + μt 1 − S xjt
xjt−1

where δ is the depreciation rate and S [·] is an adjustment cost function such that S [Λx ] = 0,
S 0 [Λx ] = 0, and S 00 [·] > 0 where Λx is the growth rate of investment along the balance
growth path. We will determine that growth rate below. Note our capital timing: we index
capital by the time its level is decided. The investment-specific technological shock follows
an autoregressive process:

μt = μt−1 exp (Λμ + zμ,t ) where zμ,t = σ μ εμ,t and εμ,t ∼ N (0, 1)

The value of μt is also the inverse of the relative price of new capital in consumption terms.
Given our description of the household’s problem, the lagrangian function associated with
it is:
⎡ ½ ³ ´ 1+γ
¾ ⎤
mjt ljt
⎢ dt log (cjt − hcjt−1 ) + υ log pt − ϕt ψ 1+γ ⎥
X ⎢
∞ ⎢ ( R ) ⎥
mjt bjt
cjt + xjt + pt + pt + qjt+1,t ajt+1 dω j,t+1,t ⎥
E0 βt ⎢
⎢ −λjt ¡ ¢
⎥
⎥
mjt−1 bjt−1
t=0 ⎢ −w jt ljt − rt u jt − μ−1
a [ujt ] kjt−1 − − R t−1 − a jt − Tt − zt ⎥
⎣ n t ³pt h i´
pt o ⎦
xjt
−Qjt kjt − (1 − δ) kjt−1 − μt 1 − S xjt−1 xjt

where they maximize over cjt , bjt , ujt , kjt , xjt , wjt , ljt and ajt+1 (maximization with respect to

3
money holdings comes from the budget constraint), λjt is the lagrangian multiplier associated
with the budget constraint and Qjt the lagrangian multiplier associated with installed capital.
The first order conditions with respect to cjt , bjt , ujt , kjt , and xjt are:

dt (cjt − hcjt−1 )−1 − hβEt dt+1 (cjt+1 − hcjt )−1 = λjt

Rt
λjt = βEt {λjt+1 }
Πt+1
rt = μ−1 0
t a [ujt ]
© ¡ ¢ª
Qjt = βEt (1 − δ) Qjt+1 + λjt+1 rt+1 ujt+1 − μ−1 t+1 a [ujt+1 ]
µ ∙ ¸ ∙ ¸ ¶ ∙ ¸µ ¶2
xjt 0 xjt xjt 0 xjt+1 xjt+1
−λjt + Qjt μt 1 − S −S + βEt Qjt+1 μt+1 S = 0.
xjt−1 xjt−1 xjt−1 xjt xjt

We do not take first order conditions with respect to Arrow-Debreau securities since, in our
environment with complete markets and separable utility in labor, their equilibrium price
will be such that their demand ensures that consumption does not depend on idiosyncractic
shocks (see Erceg et. al., 2000).
Q
If we define the (marginal) Tobin’s Q as qjt = λjtjt , (the ratio of the two lagrangian
multipliers, or more loosely the value of installed capital in terms of its replacement cost) we
get:

dt (cjt − hcjt−1 )−1 − hβEt dt+1 (cjt+1 − hcjt )−1 = λjt

Rt
λjt = βEt {λjt+1 }
Πt+1
rt = μ−1 0
t a [ujt ]
½ ¾
λjt+1 ¡ −1
¢
qjt = βEt (1 − δ) qjt+1 + rt+1 ujt+1 − μt+1 a [ujt+1 ]
λjt
µ ∙ ¸ ∙ ¸ ¶ ∙ ¸µ ¶2
xjt 0 xjt xjt λjt+1 0 xjt+1 xjt+1
1 = qjt μt 1 − S −S + βEt qjt+1 μt+1 S .
xjt−1 xjt−1 xjt−1 λjt xjt xjt

The last equation is important. If S [·] = 0 (i.e., there are no adjustment costs), we get:

1
qjt =
μt

i.e., the marginal Tobin’s Q is equal to the replacement cost of capital (the relative price of
capital). Furthermore, if μt = 1, as in the standard neoclassical growth model, qjt = 1.
The first order condition with respect to labor and wages is more involved. The labor
used by intermediate good producers to be described below is supplied by a representative,

4
competitive firm that hires the labor supplied by each household j. The labor supplier
aggregates the differentiated labor of households with the following production function:
µZ 1 η−1
¶ η−1
η

ltd = η
ljt dj (1)
0

where 0 ≤ η < ∞ is the elasticity of substitution among different types of labor and ltd is the
aggregate labor demand.1
The labor “packer” maximizes profits subject to the production function (1), taking as
given all differentiated labor wages wjt and the wage wt . Consequently, its maximization
problem is: Z 1
max wt ltd − wjt ljt dj
ljt 0

whose first order conditions are:

µZ 1 ¶ η−1
η
−1
η η−1
η η − 1 η−1 −1
wt ljt dj ljtη − wjt = 0 ∀j
η−1 0 η

Dividing the first order conditions for two types of labor i and j, we get:
µ ¶− η1
wit lit
=
wjt ljt

or: µ ¶ η1
lit
wjt = wit
ljt
Hence: η−1
1
wjt ljt = wit litη ljtη

and integrating out:

Z 1 1
Z 1 η−1 1 ¡ ¢ η−1
wjt ljt dj = wit lit η
ljtη dj = wit litη ltd η
0 0

R1
Now, by the zero profits condition implied by perfect competition wt ltd = 0
wjt ljt dj, we get:
1 ¡ ¢ η−1 1 ¡ ¢ 1
−
wt ltd = wit litη ltd η ⇒ wt = wit litη ltd η

1
Often, papers write θ = η−1 1 η
η and θ = η−1 . For that reparametrization, −∞ < θ ≤ 1, where as θ → −∞
(i.e., as η → 0), we go to a Leontieff production function, θ = 0 (i.e., η = 1) , we have a Cobb-Douglas, and
θ = 1 (i.e., η → ∞), a linear production function.

5
and, consequently, the input demand functions associated with this problem are:
µ ¶−η
wjt
ljt = ltd ∀j (2)
wt

This functional form shows the effect of elasticity η, on the demand for j − th type of labor.
R1
To find the aggregate wage, we use again the zero profit condition wt ltd = 0 wjt ljt dj and
plug-in the input demand functions:
Z 1 µ ¶−η Z 1
wjt
wt ltd = wjt ltd dj ⇒ wt1−η = 1−η
wjt dj
0 wt 0

to deliver:
µZ 1 ¶ 1−η
1

1−η
wt = wjt dj .
0

Idiosyncratic risks come about because households set their wages following a Calvo’s
setting. In each period, a fraction 1 − θw of households can change their wages. All other
households can only partially index their wages by past inflation. Indexation is controlled by
the parameter χw ∈ [0, 1]. This implies that if the household cannot change her wage for τ
Yτ χw
Πt+s−1
periods her normalized wage after τ periods is Πt+s
wjt .
s=1
Therefore, the relevant part of the lagrangian for the household is then:
( )
X
∞ 1+γ
ljt+τ Yτ χw
Πt+s−1
max Et (βθw )τ −dt ϕt ψ + λjt+τ wjt ljt+τ
wjt
τ =0
1+γ s=1
Πt+s

subject to à τ !−η
Y Πχt+s−1
w
wjt d
ljt+τ = lt+τ ∀j
s=1
Πt+s wt+τ

or, substituting the demand function (2), we get:

⎧ Ã τ !−η ⎫
⎪
⎪ Yτ χw Y Πχw w ⎪
⎪
⎪
⎪ λ
Πt+s−1
wjt t+s−1 jt d
lt+τ ⎪
⎪
X∞ ⎪
⎨ jt+τ Πt+s Πt+s wt+τ ⎪
⎬
s=1 s=1
max Et (βθw )τ ⎛ τ
Y Πχw w ⎞−η(1+γ)
wjt ⎪
⎪ ⎪
⎪
τ =0 ⎪
⎪ ⎝ t+s−1 jt ⎠
⎪
⎪
⎪
⎩ −d ϕ ψ s=1
Πt+s wt+τ
¡ d ¢1+γ ⎪
⎭
t+τ t+τ 1+γ
lt+τ

6
which simplifies to
⎧ Ã τ !1−η ⎫
⎪
⎪ Y Πχw ⎪
⎪
⎪
⎪ λjt+τ t+s−1 wjt d
wt+τ lt+τ ⎪
⎪
X
∞ ⎪
⎨ Πt+s wt+τ ⎪
⎬
s=1
max Et (βθw )τ ⎛ τ
Y Πχw w ⎞−η(1+γ)
wjt ⎪
⎪ ⎪
⎪
τ =0 ⎪
⎪ ⎝ t+s−1 jt ⎠
⎪
⎪
⎪
⎩ −d ϕ ψ s=1
Πt+s wt+τ
¡ d ¢1+γ ⎪
⎭
t+τ t+τ 1+γ
lt+τ

All households set the same wage because complete markets allow them to hedge the risk
of the timing of wage change. Hence, we can drop the jth from the choice of wages and λjt .
The first order condition of this problem is:
⎧ Ã τ !1−η ⎫
⎪ Y Πχw ³ ´−η ⎪
⎪
⎪ t+s−1 wt∗ d ⎪
⎪
⎪
⎨ (1 − η) λt+τ lt+τ ⎪
⎬
X
∞ Πt+s wt+τ
Et (βθw )τ Ã s=1 !−η(1+γ) =0
⎪
⎪ Y τ χw ¡ d ¢1+γ ⎪
⎪
τ =0 ⎪
⎪ η Π ∗
t+s−1 wt ⎪
⎪
⎩ + wt∗ dt+τ ϕt+τ ψ Πt+s wt+τ
lt+τ ⎭
s=1

or
à τ !1−η µ ¶−η
η−1 ∗ X Y Πχt+s−1
∞
τ
w
wt∗ d
wt Et (βθw ) λt+τ lt+τ =
η τ =0 s=1
Π t+s w t+τ
⎛ Ã τ !−η(1+γ) ⎞
X∞ Y χw
Πt+s−1 wt ∗ ¡ ¢ 1+γ
Et (βθw )τ ⎝dt+τ ϕt+τ ψ d
lt+τ ⎠
τ =0 s=1
Πt+s wt+τ

Now, if we define:
à τ !1−η µ ¶η
η−1 ∗ X Y Πχt+s−1
∞ w
wt+τ
ft1 = wt Et (βθw )τ λt+τ d
lt+τ
η τ =0 s=1
Πt+s wt∗

and à τ !−η(1+γ) µ
X
∞ Y Πχt+s ¶η(1+γ)
τ
w
wt+τ ¡ d ¢1+γ
ft2 = Et (βθw ) dt+τ ϕt+τ ψ lt+τ
τ =0 s=1
Πt+s−1 wt∗

we have that the equality ft1 = ft2 is just the previous first order condition. Note that for
those sums to be well defined (and, more generally for the maximization problem to have a
à τ !1−η
Y χw
solution), we need to assume that (βθw )τ λt+τ goes to zero faster than Πt+s /Πt+s−1
s=1
goes to infinity in expectation.

7
 One can express ft1 and ft2 recursively as:
µ χ ¶1−η µ ∗
¶η−1
η − 1 ∗ 1−η Πt w wt+1
ft1 = (wt ) λt wtη ltd + βθw Et 1
ft+1
η Πt+1 wt∗

and: µ ¶η(1+γ) µ ¶−η(1+γ) µ ¶η(1+γ)

χ
wt ¡ d ¢1+γ Πt w ∗
wt+1
ft2 = ψdt ϕt lt + βθw Et 2
ft+1 .
wt∗ Πt+1 wt∗
Now, since ft1 = ft2 , we can define ft = ft1 = ft2 , such that:
µ χ ¶1−η µ ∗
¶η−1
η − 1 ∗ 1−η Πt w wt+1
ft = (wt ) λt wtη ltd + βθw Et ft+1
η Πt+1 wt∗

and: µ ¶η(1+γ) µ ¶−η(1+γ) µ ¶η(1+γ)

χ
wt ¡ d ¢1+γ Πt w ∗
wt+1
ft = ψdt ϕt lt + βθw Et ft+1 .
wt∗ Πt+1 wt∗
Since we assume complete markets and separable utility in labor (see Erceg et. al., 2000),
we consider a symmetric equilibrium where cjt = ct , ujt = ut , kjt−1 = kt , xjt = xt , λjt = λt ,
∗
qjt = qt , and wjt = wt∗ . Therefore, the first order conditions associated to the consumer’s
problems are:

dt (ct − hct−1 )−1 − hβEt dt+1 (ct − hct )−1 = λt

Rt
λt = βEt {λt+1 }
Πt+1
rt = μ−1 0
t a [ut ]
½ ¾
λt+1 £ ¡ −1
¢¤
qt = βEt (1 − δ) qt+1 + rt+1 ut+1 − μt+1 a [ut+1 ]
λt
µ ∙ ¸ ∙ ¸ ¶ ∙ ¸µ ¶2
xt 0 xt xt λt+1 0 xt+1 xt+1
1 = qt μt 1 − S −S + βEt qt μt+1 S .
xt−1 xt−1 xt−1 λt xt xt

the budget constraint:

Z
mjt bjt+1
cjt + xjt + + + qjt+1,t ajt+1 dω j,t+1,t
pt pt
¡ ¢ mjt−1 bjt
= wjt ljt + rt ujt − μ−1
t a [ujt ] kjt−1 + + Rt−1 + ajt + Tt + zt
pt pt

and the laws of motion for ft :

µ χ ¶1−η µ ∗
¶η−1
η − 1 ∗ 1−η Πt w wt+1
ft = (wt ) λt wtη ltd + βθw Et ft+1
η Πt+1 wt∗

8
and: µ ¶η(1+γ) µ ¶−η(1+γ) µ ¶η(1+γ)
χ
wt ¡ d ¢1+γ Πt w ∗
wt+1
ft = ψdt ϕt lt + βθw Et ft+1 .
wt∗ Πt+1 wt∗
We need both laws of motion to be able, later, to solve for all the relevant endogenous
variables.
R1
Note that using the zero profits condition for the labor supplier, wt ltd = 0 wjt ljt dj and
the net zero supply of all securities, we have that the aggregate budget constraint can be
written as:
R1 R1
0
mjt dj bjt+1 dj
ct + xt + + 0 =
pt pt
R1 R1
¡ ¢ mjt−1 dj bjt dj
wt ltd + rt ut − μ−1
t a [ut ] kt−1 +
0
+ Rt−1 0 + Tt + zt
pt pt

Thus, in a symmetric equilibrium, in every period, a fraction 1 − θw of households set wt∗

as their wage, while the remaining fraction θw partially index their price by past inflation.
Consequently, the real wage index evolves:
µ χ ¶1−η
Πt−1
w

wt1−η = θw 1−η
wt−1 + (1 − θw ) wt∗1−η .
Πt

i.e., as a geometric average of past real wage and the new optimal wage. This structure is a
direct consequence of the memoryless characteristic of Calvo pricing.

1.2. The Final Good Producer

There is one final good is produced using intermediate goods with the following production
function: µZ 1 ¶ ε−1
ε
ε−1
ytd = yit ε di . (3)
0

where ε is the elasticity of substitution.

Final good producers are perfectly competitive and maximize profits subject to the pro-
duction function (3), taking as given all intermediate goods prices pti and the final good price
pt . As a consequence their maximization problem is:
Z 1
max pt ytd − pit yit di
yit 0

Following the same steps than for the wages, we find the input demand functions associ-

9
ated with this problem are: µ ¶−ε
pit
yit = ytd ∀i,
pt
R1
where ytd is the aggregate demand and the zero profit condition pt ytd = 0
pit yit di to deliver:

µZ 1 ¶ 1−ε
1

pt = p1−ε
it di .
0

1.3. Intermediate Good Producers

There is a continuum of intermediate goods producers. Each intermediate good producer i

has access to a technology represented by a production function

α
¡ d ¢1−α
yit = At kit−1 lit − φzt

where kit−1 is the capital rented by the firm, litd is the amount of the “packed” labor input
rented by the firm, and where At follows the following process:

At = At−1 exp (ΛA + zA,t ) where zA,t = σ A εA,t and εA,t ∼ N (0, 1)
1 α
The parameter φ, which corresponds to the fixed cost of production, and zt = At1−α μt1−α
guarantee that economic profits are roughly equal to zero in the steady state. We rule out
the entry and exit of intermediate good producers.
1 α
Since zt = At1−α μt1−α , we have that

zA,t + αzμ,t ΛA + αΛμ

zt = zt−1 exp (Λz + zz,t ) where zz,t = and Λz = .
1−α 1−α

Intermediate goods producers solve a two-stages problem. In the first stage, taken the
input prices wt and rt as given, firms rent litd and kit−1 in perfectly competitive factor markets
in order to minimize real cost:
min wt litd + rt kit−1
d ,k
lit it−1

subject to their supply curve:

( ¡ d ¢1−α ¡ d ¢1−α
α α
At kit−1 lit − φzt if At kit−1 lit ≥ φzt
yit =
0 otherwise

10
Assuming an interior solution, the first order conditions for this problem are:

α
¡ d ¢−α
wt = % (1 − α) At kit−1 lit
¡
α−1 d
¢1−α
rt = %αAt kit−1 lit

where % is the Lagrangian multiplier or:

α wt d
kit−1 = l
1 − α rt it

The real cost is then:

α
wt litd + wt litd
1−α
or: µ ¶
1
wt litd
1−α
Given that the firm has constant returns to scale, we can find the real marginal cost mct
by setting the level of labor and capital equal to the requirements of producing one unit of
α
¡ d ¢1−α
good At kit−1 lit = 1 or:
µ ¶α µ ¶α
α
¡ d ¢1−α α wt d ¡ d ¢1−α α wt
At kit−1 lit = At l lit = At litd = 1
1 − α rt it 1 − α rt

that implies that: ³ ´−α

α wt
1−α rt
litd =
At
Then: ³ ´−α
µ ¶ α wt
1 1−α rt
mct = wt
1−α At
that simplifies to: µ ¶1−α µ ¶α 1−α α
1 1 wt rt
mct =
1−α α At
Note that the marginal cost does not depend on i: all firms receive the same technology
shocks and all firms rent inputs at the same price.
In the second stage, intermediate good producers choose the price that maximizes dis-
counted real profits. To do so, they consider that are under the same pricing scheme than
households. In each period, a fraction 1 − θp of firms can change their prices. All other
firms can only index their prices by past inflation. Indexation is controlled by the parameter
χ ∈ [0, 1], where χ = 0 is no indexation and χ = 1 is total indexation.

11
 The problem of the firms is then:
(Ã τ ! )
X
∞
λt+τ Y pit
max Et (βθp )τ Πχt+s−1 − mct+τ yit+τ
pit
τ =0
λt s=1
pt+τ

subject to à τ !−ε
Y pit
yit+τ = Πχt+s−1 d
yt+τ ,
s=1
pt+τ
where the marginal value of a dollar to the household, is treated as exogenous by the firm.
Since we have complete markets in securities and utility separable in consumption, this mar-
ginal value is constant across households and, consequently, λt+τ /λt is the correct valuation
on future profits.
Substituting the demand curve in the objective function and the previous expression, we
get:
⎧⎛Ã !1−ε Ã τ !−ε ⎞ ⎫
X
∞
λt+τ ⎨ Y τ
pit Y pit ⎬
max Et (βθp )τ ⎝ Πχt+s−1 − Πχt+s−1 mct+τ ⎠ yt+τ
d
pit
τ =0
λt ⎩ s=1
pt+τ s=1
pt+τ ⎭

or
⎧⎛Ã !1−ε Ã τ !−ε ⎞ ⎫
X
∞
λt+τ ⎨ Y τ χ
Πt+s−1 pit Y χ
Πt+s−1 pit ⎬
max Et (βθp )τ ⎝ − mct+τ ⎠ yt+τ
d
pit
τ =0
λt ⎩ s=1
Πt+s pt s=1
Πt+s pt ⎭

whose solution p∗it implies the first order condition:

⎧⎛ Ã τ !1−ε Ã τ !−ε ⎞ ⎫
X
∞
λt+τ ⎨ Y Πt+s−1 p
χ ∗ Y Πt+s−1 p
χ ∗ ⎬
Et (βθp )τ ⎝(1 − ε) it ∗−1
pit + ε it ∗−1 ⎠ d
pit mct+τ yt+τ = 0
τ =0
λt ⎩ s=1
Πt+s pt s=1
Πt+s pt ⎭

or
⎧⎛ Ã τ !1−ε Ã τ !−ε ⎞ ⎫
X
∞ ⎨ Y Πt+s−1
χ ∗
pit Y Πt+s−1
χ ⎬
Et τ
(βθp ) λt+τ ⎝(1 − ε) +ε ⎠ d
mct+τ yt+τ = 0
τ =0
⎩ s=1
Πt+s pt s=1
Πt+s ⎭

where, in the second step, we have dropped irrelevant constants and we have used the fact
that we are in a symmetric equilibrium. Note how this expression nests the usual result in
the fully flexible prices case θp = 0:

ε
p∗it = pt mct+τ
ε−1

12
i.e., the price is equal to a mark-up over the nominal marginal cost.
Since we only consider a symmetric equilibrium, we can write that p∗it = p∗t and:
⎧⎛ Ã τ !1−ε Ã τ !−ε ⎞ ⎫
X
∞ ⎨ Y χ
Πt+s−1 ∗
pt Y χ
Πt+s−1 ⎬
Et (βθp )τ λt+τ ⎝(1 − ε) +ε mct+τ ⎠ yt+τ
d
=0
τ =0
⎩ s=1
Πt+s pt s=1
Πt+s ⎭

To express the previous first order condition recursively, we define:

à τ !−ε
X
∞
τ
Y Πχt+s−1
gt1 = Et (βθp ) λt+τ d
mct+τ yt+τ
τ =0 s=1
Πt+s

and à τ !1−ε
X
∞ Y Πχt+s−1
τ p∗t d
gt2 = Et (βθp ) λt+τ y
τ =0 s=1
Πt+s pt t+τ

and then the first order condition is εgt1 = (ε − 1)gt2 .

As it was the case for f ’s that we found in the household problem, we need (βθp )τ λt+τ
to go to zero sufficiently fast in relation with the rate of inflation for gt1 and gt2 to be well
defined and stationary.
Then, we can write the g’s recursively:
µ ¶−ε
Πχt
gt1 = λt mct ytd + βθp Et 1
gt+1
Πt+1

and µ ¶1−ε µ ¶
Πχt Π∗t
gt2 = λt Π∗t ytd + βθp Et 2
gt+1
Πt+1 Π∗t+1
where:
p∗t
Π∗t = .
pt
Given Calvo’s pricing, the price index evolves:
¡ ¢1−ε 1−ε
p1−ε
t = θp Πχt−1 pt−1 + (1 − θp ) p∗1−ε
t

or, dividing by p1−ε

t ,
µ ¶1−ε
Πχt−1
1 = θp + (1 − θp ) Π∗1−ε
t
Πt

13
1.4. The Government Problem

The government sets the nominal interest rates according to the Taylor rule:
⎛ ⎛ yd ⎞γ y ⎞1−γ R
µ ¶γ R µ ¶γ Π t
Rt Rt−1 ⎝ Πt
d
⎝ yt−1 ⎠ ⎠
= exp (mt )
R R Π Λyd

through open market operations that are financed through lump-sum transfers Tt . Those
transfers insure that the deficit are equal to zero:
R1 R1 R1 R1
0
mjt dj 0
mjt−1 dj 0
bjt+1 dj 0
bjt dj
Tt = − + − Rt−1
pt pt pt pt

The variables Π represents the target level of inflation (equal to inflation in the steady state),
R steady state nominal gross return of capital, and Λyd the steady state gross growth rate of
ytd . The term mt is a random shock to monetary policy that follows mt = σ m εmt where εmt is
distributed according to N (0, 1). The presence of the previous period interest rate, Rt−1 , is
justified because we want to match the smooth profile of the interest rate over time observed
in U.S. data. Note that R is beyond the control of the monetary authority, since it is equal
to the steady state real gross returns of capital plus the target level of inflation.
Applying the definition of transfers above, the aggregated budget constraint of households
is equal to:
¡ ¢
ct + xt = wt ltd + rt ut − μ−1
t a [ut ] kt−1 + zt .

1.5. Aggregation

First, we derive an expression for aggregate demand:

ytd = ct + xt + μ−1
t a [ut ] kt−1

With this value, the demand for each intermediate good producer is
µ ¶
¡ −1
¢ pit −ε
yit = ct + xt + μt a [ut ] kt−1 ∀i,
pt

and using the production function is:

µ ¶
α
¡ d ¢1−α ¡ −1
¢ pit −ε
At kit−1 lit − φzt = ct + xt + μt a [ut ] kt−1
pt

14
Since all the firms have the same optimal capital-labor ratio:

kit−1 α wt
d
=
lit 1 − α rt

and by market clearing Z 1

litd di = ltd
0

and Z 1
kit−1 di = ut kt−1 .
0

it must be the case that:

kit−1 ut kt−1
d
= .
lit ltd
Then: µ ¶α µ ¶α
α
¡ d ¢1−α kit−1 ¡ d¢ ut kt−1
At kit−1 lit = At lit = At litd
litd ltd
Integrating out
Z µ ¶α µ ¶α Z
1
ut kt−1 ut kt−1 1 ¡ ¢1−α
At litd di = At litd di = At (ut kt−1 )α ltd
0 ltd ltd 0

and we have
Z 1µ ¶−ε
¡ ¢1−α α ¡ ¢ pit
At (ut kt−1 ) ltd − φzt = ct + xt + μ−1
t a [ut ] kt−1 di
0 pt

R 1 ³ pit ´−ε
Define vtp = 0 pt
di. By the properties of the index under Calvo’s pricing

µ ¶−ε
Πχt−1
vtp = θp p
vt−1 + (1 − θp ) Π∗−ε
t .
Πt

we get: ¡ ¢1−α
At (ut kt−1 )α ltd − φzt
ct + xt + μ−1
t a [ut ] kt−1 = p
vt
Now, we derive an expression for aggregate labor demand. We know that
µ ¶−η
wjt
ljt = ltd
wt

15
If we integrate over all households j, we get
Z 1 Z 1µ ¶−η
wjt
ljt dj = lt = djltd
0 0 wt

where lt is the aggregate labor supply of households.

Define Z 1µ ¶−η
w wjt
vt = dj.
0 wt
Hence
1
ltd = lt .
vtw
Also, as before:
µ χ ¶−η
wt−1 Πt−1
w
−η
vtw = θw w
vt−1 + (1 − θw ) (Πw∗
t ) .
wt Πt

2. Equilibrium

A definition of equilibrium in this economy is standard and the symmetric equilibrium policy
functions are determined by the following equations:

• The first order conditions of the household

dt (ct − hct−1 )−1 − hβEt dt+1 (ct+1 − hct )−1 = λt

Rt
λt = βEt {λt+1 }
Πt+1
rt = μ−1 0
t a [ut ]
½ ¾
λt+1 ¡ −1
¢
qt = βEt (1 − δ) qt+1 + rt+1 ut+1 − μt+1 a [ut+1 ]
λt
µ ∙ ¸ ∙ ¸ ¶ ∙ ¸µ ¶2
xt 0 xt xt λt+1 0 xt+1 xt+1
1 = qt μt 1 − S −S + βEt qt+1 μt+1 S
xt−1 xt−1 xt−1 λt xt xt
µ χw ¶1−η µ ∗ ¶η−1
η − 1 ∗ 1−η Πt wt+1
ft = (wt ) λt wtη ltd + βθw Et ft+1
η Πt+1 wt∗
µ χw ¶−η(1+γ) µ ∗ ¶η(1+γ)
¡ ¢
∗w −η(1+γ) d 1+γ Πt wt+1
ft = ψdt ϕt (Πt ) lt + βθw Et ft+1
Πt+1 wt∗

16
• The firms that can change prices set them to satisfy:
¶−ε µ
Πχt
gt1
= λt mct ytd
+ βθp Et 1
gt+1
Πt+1
µ χ ¶1−ε µ ∗ ¶
2 ∗ d Πt Πt 2
gt = λt Πt yt + βθp Et ∗
gt+1
Πt+1 Πt+1
εgt1 = (ε − 1)gt2

where they rent inputs to satisfy their static minimization problem:

ut kt−1 α wt
d
=
lt 1 − α rt
µ ¶1−α µ ¶α 1−α α
1 1 wt rt
mct =
1−α α At

• The wages evolve as:

µ χ ¶1−η µ ¶1−η
Πt−1
w
wt−1 1−η
1 = θw + (1 − θw ) (Π∗w
t )
Πt wt

and the price level evolves:

µ ¶1−ε
Πχt−1
1 = θp + (1 − θp ) Π∗1−ε
t
Πt

• Government follow its Taylor rule

⎛ ⎛ yd ⎞γ y ⎞1−γ R
µ ¶γ R µ ¶γ Π t
Rt Rt−1 ⎝ Πt ⎝
d
yt−1
⎠ ⎠
= exp (mt )
R R Π Λyd

• Markets clear:
¡ ¢1−α
At (ut kt−1 )α ltd − φzt
ytd = p
vt
ytd −1
= ct + xt + μt a [ut ] kt−1

17
 where

lt = vtw ltd
µ χ ¶−ε
p Πt−1 p
vt = θp vt−1 + (1 − θp ) Π∗−ε
t
Πt
µ χw ¶−η
w wt−1 Πt−1 w −η
vt = θw vt−1 + (1 − θw ) (Π∗w
t )
wt Πt

and µ ∙ ¸¶
xt
kt − (1 − δ) kt−1 − μt 1 − S xt = 0.
xt−1

3. Stationary Equilibrium

Since we have growth in this model induced by technological change, most of the variables
are growing in average. To solve the model, we need to make variables stationary.

3.1. Manipulating Equilibrium Conditions

First, we work on the first order conditions of the household

µ ¶−1 µ ¶−1
ct ct−1 zt−1 ct+1 zt+1 ct
dt −h − hβEt dt+1 −h = λt zt
zt zt−1 zt zt+1 zt zt
zt Rt
λt zt = βEt {λt+1 zt+1 }
zt+1 Πt+1
μt rt = a0 [ut ]
½ ¾
λt+1 zt+1 zt μt £ ¤
qt μt = βEt (1 − δ) qt+1 μt+1 + μt+1 rt+1 ut+1 − a (ut+1 )
λt zt zt+1 μt+1
à " x # " x # x !
t t t
zt zt z z t z zt
1 = qt μt 1 − S xt−1 − S 0 xt−1t t
xt−1
zt−1
z t−1 zt−1
zt−1 z
zt−1 t−1
" xt+1 # Ã xt+1 !2
λt+1 0 zt+1 zt+1 zt+1 zt+1
+βEt qt+1 μt+1 S xt xt
λt zt
zt zt
zt
⎛ ∗ ⎞η−1
µ ∗ ¶1−η µ ¶η µ χw ¶1−η wt+1
η − 1 wt wt Πt z t+1
ft = λt zt ltd + βθw Et ⎝ zt+1∗ ⎠ ft+1
η zt zt Πt+1 wt
zt t
z
⎛ ∗ ⎞η(1+γ)
µ χw ¶−η(1+γ) wt+1
−η(1+γ) ¡ d ¢1+γ Πt z t+1
ft = ψdt ϕt (Π∗w
t ) lt + βθw Et ⎝ zt+1∗ ⎠ ft+1
Πt+1 w
zt
t
zt

18
 The firms that can change prices set them to satisfy:
µ χ ¶−ε
ytd Πt
gt1
= λt zt mct + βθp Et 1
gt+1
zt Πt+1
d
µ χ ¶1−ε µ ∗ ¶
2 ∗ yt Πt Πt 2
gt = λt zt Πt + βθp Et ∗
gt+1
zt Πt+1 Πt+1
εgt1 = (ε − 1)gt2

where they rent inputs to satisfy their static minimization problem:

ut kt−1 α wt 1 zt μt
d z
=
lt t−1 μt−1 1 − α zt rt μt zt−1 μt−1
³ ´1−α
µ ¶1−α µ ¶α wt zt1−α
1 1 zt
(rt μt )α μα
t
mct =
1−α α At

The wages evolve as:

µ χ ¶1−η Ã wt−1 !1−η

Πt−1
w
zt−1 zt−1 1−η
1 = θw wt + (1 − θw ) (Π∗w
t )
Πt zt
zt

and the price level evolves:

µ ¶1−ε
Πχt−1
1 = θp + (1 − θp ) Π∗1−ε
t
Πt

Government follow its Taylor rule

⎛ ⎛ ytd
⎞γ y ⎞1−γ R
zt zt
µ ¶γ R ⎜µ ¶γ Π ⎜ d
yt−1 zt−1 ⎟ ⎟
Rt Rt−1 ⎜ Πt ⎜ ⎟ ⎟
= ⎜ ⎜ zt−1
⎟ ⎟ exp (mt )
R R ⎜ Π ⎜ Λyd ⎟ ⎟
⎝ ⎝ ⎠ ⎠

Markets clear: ³ ´α ¡ ¢
α At kt−1 1−α
ytd μαt−1 zt−1 zt
ut zt−1 μ ltd −φ
t−1
=
zt vtp
zt−1
but since μαt−1 zt−1
α
= At−1
, we have
³ ´α ¡ ¢
zt−1 At kt−1 1−α
ytd At−1 zt
ut zt−1 μ ltd −φ
t−1
=
zt vtp

19
 ytd ct xt zt−1 μt−1 kt−1
= + + a [ut ]
zt zt zt zt μt zt−1 μt−1
where

lt = vtw ltd
µ χ ¶−ε
p Πt−1 p
vt = θp vt−1 + (1 − θp ) Π∗−ε
t
Πt
à wt−1 χw
!−η
w
z
zt−1 t−1 Πt−1 w −η
vt = θw wt vt−1 + (1 − θw ) (Π∗w
t )
z
zt t
Πt

and à " #!
xt
kt zt μt kt−1 μ zt z
zt t xt
− (1 − δ) − t 1−S xt−1 = 0.
zt μt zt−1 μt−1 zt−1 μt−1 μt−1 zt−1 z
zt−1 t−1
zt

3.2. Change of Variables

We now redefine that variables to obtain a system on stationary variables that we can easily
manipulate. Hence, we define e et = λt zt , ret = rt μt , qet = qt μt , x
ct = zctt , λ et = xztt , w
et = wztt ,
∗ d
e∗ = wt , e
wt zt
kt = kt , and yed = yt .
zt μt t zt
Then, the set of equilibrium conditions are:

• The first order conditions of the household:

µ ¶−1 µ ¶−1
zt−1 zt+1 et
dt e
ct − he
ct−1 − hβEt dt+1 e ct+1 − he
ct =λ
zt zt
et+1 zt Rt }
et = βEt {λ
λ
zt+1 Πt+1
ret = a0 [ut ]
( )
et+1 zt μt
λ
qet = βEt ((1 − δ) qet+1 + ret+1 ut+1 − a (ut+1 ))
et zt+1 μt+1
λ
µ ∙ ¸ ∙ ¸ ¶
et zt
x 0 et zt
x et zt
x
1 = qet 1 − S −S
et−1 zt−1
x et−1 zt−1 x
x et−1 zt−1
et+1 zt ∙ ¸ µ ¶2
λ 0 xet+1 zt+1 et+1 zt+1
x
+βEt qet+1 S
et zt+1
λ et zt
x et zt
x
µ χw ¶1−η µ ∗ ¶
η − 1 ∗ 1−η e η d Πt et+1 zt+1 η−1
w
ft = (wet ) λt (w et ) lt + βθw Et ft+1
η Πt+1 et∗ zt
w
µ χw ¶−η(1+γ) µ ∗ ¶
¡ ¢
∗w −η(1+γ) d 1+γ Πt et+1 zt+1 η(1+γ)
w
ft = ψdt ϕt (Πt ) lt + βθw Et ft+1
Πt+1 et∗ zt
w

20
• The firms that can change prices set them to satisfy:
µ χ ¶−ε
e Πt
gt1 d
= λt mct yet + βθp Et 1
gt+1
Πt+1
µ χ ¶1−ε µ ∗ ¶
2 et Π ye + βθp Et Πt
gt = λ ∗ d Πt 2
gt+1
t t ∗
Πt+1 Πt+1
εgt1 = (ε − 1)gt2

where they rent inputs to satisfy their static minimization problem:

ut e
kt−1 α w et zt μt
d
=
lt 1 − α ret zt−1 μt−1
µ ¶1−α µ ¶α
1 1
mct = et )1−α retα
(w
1−α α

• The wages evolve as:

µ χ ¶1−η µ ¶1−η
Πt−1
w
et−1 zt−1
w 1−η
1 = θw + (1 − θw ) (Π∗w
t )
Πt et zt
w

and the price level evolves:

µ ¶1−ε
Πχt−1
1 = θp + (1 − θp ) Π∗1−ε
t
Πt

• Government follow its Taylor rule:

⎛ ⎛ yed ⎞γ y ⎞1−γ R
µ ¶γ R µ ¶γ Π t zt
Rt Rt−1 ⎝ Πt
d z
⎝ yet−1 t−1 ⎠ ⎠
= exp (mt )
R R Π Λyd

• Markets clear:

zt−1 μt−1
yetd = e
ct + xet + a [ut ] e
kt−1
zt μt
³ ´α ¡ ¢
At zt−1 e 1−α
A t−1 zt
u t kt−1 ltd −φ
yetd =
vtp

21
 where

lt = vtw ltd
µ χ ¶−ε
p Πt−1 p
vt = θp vt−1 + (1 − θp ) Π∗−ε
t
Πt
µ χw ¶−η
w et−1 zt−1 Πt−1
w w −η
vt = θw vt−1 + (1 − θw ) (Π∗w
t )
et zt Πt
w

and µ ∙ ¸¶
e zt μt zt μt et zt
x
kt − (1 − δ) e
kt−1 − 1−S et = 0.
x
zt−1 μt−1 zt−1 μt−1 et−1 zt−1
x

4. Solving the Model

We will solve the model by loglinearizing the equilibrium conditions and applying standard
techniques. Before loglinearizing, we need to find the steady-state of the model. We will solve
the normalized model defined in the last section. Note that later, when we bring the model
to the data, we will need to undo the normalization.

4.1. The Steady-State

Now, we will find the deterministic steady-state of the model. First, let ze = exp (Λz ),
μ e = exp (ΛA ). Also, given the definition of e
e = exp (Λμ ), and A et , w
c, x et∗ , and yetd , we have
et , w
that Λc = Λx = Λw = Λw∗ = Λyd = Λz .
Then, in steady-state, the first order conditions of the household can be written as:

1 1 e
d h
− hβd =λ
e
c − ze e
c ecze − he c
1R
1=β
ze Π
0
re = a [1]
1
qe = β ((1 − δ) qe + reu − a [1])
zeμ
e
qe
1 = qe (1 − S [e z ] − S 0 [ez ] ze) + β S 0 [e
z ] ze2
ze
µ χ ¶1−η
η − 1 ∗ 1−η e η d Π w
f= e ) λw
(w e l + βθw zeη−1 f
η Π
µ χ ¶−η(1+γ)
¡ w∗ ¢−η(1+γ) ¡ d ¢1+γ Π w
f = ψdϕ Π l + βθw zeη(1+γ) f
Π

22
the first order conditions of the firm as:
¶
χ −ε
µ
e y + βθp
1 Π
d
g = λmce g1
Π
µ χ ¶1−ε
2 e ∗ d Π
g = λΠ ye + βθp g2
Π
εg1 = (ε − 1)g 2
uek α w e
= zeμ
e
ld 1 − α re
µ ¶1−α µ ¶α
1 1
mc = e1−α reα
w
1−α α

the law of motion for wages and prices as:

µ ¶1−η
Πχw ¡ ∗ ¢1−η
1 = θw ze−(1−η) + (1 − θw ) Πw
Π
µ χ ¶1−ε
Π
1 = θp + (1 − θp ) Π∗1−ε
Π

and the market clearing conditions as:

e
c+x e = yed
A
v p yed = (ue k)α (ld )1−α − φ
z
l = v w ld
µ χ ¶−ε
p Π
v = θp vp + (1 − θp ) Π∗−ε
Π
µ χ ¶−η
w Π w ∗
v = θw zeη v w + (1 − θw ) (Πw )−η
Π
e
ke
zμe − (1 − δ) ek − zeμ e (1 − S [z]) x
e = 0.

To find the steady-state, we need to choose functional forms for a [·] and S [·]. For a [u] we
pick: a [u] = γ 1 (u − 1)+ γ22 (u−1)2 . Since in the steady state weh havei u = 1, then re = a0 [1] =
γ 1 and a [1] = 0. The investment adjustment cost function is S xxt−1 t
= κ2 ( xxt−1
t
− Λx )2 . Then,
along the balanced growth path, S [Λx ] = S 0 [Λx ] = 0. Using this two expressions, we can

23
rearrange the system of equations that determine the steady-state as:

1 1 e
(1 − hβe
z) =λ
1 − hze e
c
Πe
z
R=
β
re = γ 1
µ ¶
β β
re = 1 − (1 − δ) /
zeμ
e zeμ
e
¡ ¢ η − 1 ∗ e w∗ −η d
1 − βθw ze(η−1) Π−(1−χw )(1−η) f = w λ(Π ) l
η
¡ ¢ ¡ ∗ ¢−η(1+γ) ¡ d ¢1+γ
1 − βθw zeη(1+γ) Πη(1−χw )(1+γ) f = ψ Πw l
¡ ¢
1 − βθp Π(1−χ)ε g1 = λmcee yd
¡ ¢
1 − βθp Π−(1−χ)(1−ε) g2 = λΠ e ∗ yed

εg1 = (ε − 1)g 2
e
k α w e
= zeμ
e
l d 1 − α re
µ ¶1−α µ ¶α
1 1
mc = e1−α reα
w
1−α α
1 − θw Π−(1−χw )(1−η) ze−(1−η) ¡ w∗ ¢1−η
= Π
1 − θw
1 − θp Π−(1−χ)(1−ε)
= Π∗1−ε
1 − θp
e
c+x e = yed
A e
vp yed = (e k)α (ld )1−α − φ
ze
l = v w ld
1 − θp Π(1−χ)ε p
v = Π∗−ε
1 − θp
¡ ¢
1 − θw zeη Π(1−χw )η w ∗
v = (Πw )−η
1 − θw
e zeμe
k= e.
x
zeμ
e − (1 − δ)

First, notice that there is some restrictions on γ 1

β
1− zeμ
e
(1 − δ)
r̃ = β
= γ1
zeμ
e

24
and that the nominal interest rate is:
Πe
z
R=
β
The relationship between inflation and optimal relative prices is:
µ ¶ 1−ε
1

∗ 1 − θp Π−(1−ε)(1−χ)
Π =
1 − θp

expression from which we obtain the following two results:

1. If there is zero price inflation (Π = 1), then Π∗ = 1.

2. If there is full price indexation (χ = 1), then Π∗ = 1.

From the optimal price setting equations, we get that the marginal cost is:

ε − 1 1 − βθp Π(1−χ)ε
mc = Π∗
ε 1 − βθp Π−(1−χ)(1−ε)

The relationship between inflation and optimal relative wage is:

µ ¶ 1−η
1

w∗ 1 − θw Π−(1−χw )(1−η) ze−(1−η)

Π = .
1 − θw

Then we find that: ³ ³ α ´α ´ 1−α

1

e = (1 − α) mc
w
re
and the optimal wage evolves:
∗
e∗ = wΠ
w e w

Again, we can note the following:

∗
z = 1), then Πw = 1.
1. If there is zero price inflation (Π = 1) and zero growth rate (e
∗
z = 1), then Πw = 1.
2. If there is full wage indexation (χw = 1) and zero growth rate (e

We have the two following equations for the wage household decision:

¡ ¢ η − 1 ∗ ¡ w∗ ¢−η e d
1 − βθw zeη−1 Π−(1−χw )(1−η) f = w Π λl
η

and
¡ ¢ ¡ ∗ ¢−η(1+γ) ¡ d ¢1+γ
1 − βθw zeη(1+γ) Πη(1−χw )(1+γ) f = ψ Πw l .

25
Dividing the second by the first delivers:
¡ ∗ ¢−ηγ ¡ d ¢γ
1 − βθw zeη(1+γ) Πη(1−χw )(1+γ) ψ Πw l
= η−1
1 − βθw ze(η−1) Π−(1−χw )(1−η) η
w∗ λ̃

e Now, we will look for another relationship between ld and

which defines ld as a function of λ.
e to have two equations with two unknowns.
λ
Before doing so, we highlight that, again, there are two interesting cases:

1. If there is zero price inflation (Π = 1) and zero growth rate (e

z = 1), then we get the
η
static condition that real wages are a markup η−1 over the marginal rate of substitution
between consumption and leisure.

2. If there is full wage indexation (χw = 1) and zero growth rate (e

z = 1), then we get the
η
static condition that real wages are a markup η−1 over the marginal rate of substitution
between consumption and leisure.

The expression for the dispersion of prices is given by:

1 − θp
vp = Π∗−ε
1 − θp Π(1−χ)ε

where no price inflation or full price indexation delivers no price dispersion in steady-state.
Using the expression for Πw∗ we find that the wage dispersion in steady-state is:

1 − θw ∗
vw = (Πw )−η
1 − θw Π(1−χ )η
w z eη

where no price inflation or full wage indexation combine with zero growth delivers no wage
dispersion in in steady-state.
The relationship between labor demand and labor supply is:

l = v w ld .

Note now that

Ae e α d 1−α
d ze
(k) (l ) −φ
ye = .
vp
But, since in steady-state e
k= zeμ
zeμ
e
e −(1−δ)
e,
x it is the case that:

Ae e α d 1−α
zeμ
e − (1 − δ) e ze
(k) (l ) −φ
e
c+ k = yed =
zeμ
e vp

26
that allows us to find e
k as function of ld :

e
k α w e
=Ω= e⇒e
zeμ k = Ωld .
l d 1 − α re

Then:
Ae α d
ze
Ω l −φ zeμ
e − (1 − δ) d
e
c = − Ωl
vp zeμ
e
à !
Ae e
z e
μ − (1 − δ)
= (vp )−1 Ωα − Ω ld − (v p )−1 φ
ze zeμ
e

Now, we can express the marginal utility of consumption in terms of hours, and get another
e
relationship between ld as a function of λ:

µ ¶−1 ÃÃ e ! !−1
h A p −1 α zeμ e − (1 − δ) e
(1 − hβ)de
z 1− (v ) Ω − Ω ld − (v p )−1 φ =λ
ze ze zeμ
e

Using both relationship we can solve for ld and get:

1 − βθw zeη(1+γ) Πη(1−χw )(1+γ)

=
1 − βθw ze(η−1) Π−(1−χw )(1−η)
¡ ∗ ¢−ηγ ¡ d ¢γ
ψ Πw l
¡ ¢ ³³ ´ ´−1 .
η−1 ∗ h −1 Ae p )−1 Ωα − zeμ
e −(1−δ) d − (v p )−1 φ
η
w (1 − hβ)de
z 1 − ze ze
(v zeμ
e
Ω l

Note that this is nonlinear equation. Therefore we will use a root finder to find ld .
Once we have ld , we can solve for capital, investment, output, and consumption as follows:

e
k = Ωld
zeμe − (1 − δ) e
e =
x k
zeμ
e
Ae e α d 1−α
d ze
(k) (l ) −φ
ye =
à vp !
Ae e
z e
μ − (1 − δ)
e
c = (v p )−1 Ωα − Ω ld − (vp )−1 φ
ze zeμ
e

4.2. Loglinear approximations

For each variable vart , we define vd

art = log vart − log var, where var is the steady-state value
for the variable vart . Then, we can write vart = var expvd art
.

27
 We start by log linearizing the marginal utility of consumption:
µ ¶−1 µ ¶−1
zt−1 zt+1 et .
dt e
ct − he
ct−1 − hβEt dt+1 e
ct+1 − he
ct =λ (4)
zt zt
³ ´−1
zt−1
It is helpful to define the auxiliary variable auxt = dt e
ct − he
ct−1 zt . Then, we have that:

et
auxt − hβEt zet+1 auxt+1 = λ

where zet+1 = zt+1

zt
.
Then, (4) can be written as:

b b
e
dt
aux expaux d t+1 +zet+1
−hβzauxEt expaux e λt
= λe

which can be loglinearized as:

³ ´
b
aux ad auxt+1 + b
uxt − hβzEt (d eλ
zet+1 ) = λ et . (5)

Using the following two steady state relationship that aux(1 − hβe e and that Etb
z) = λ zet+1 = 0,
we can write:
b
et .
ad
uxt − hβzEt ad uxt+1 = (1 − hβz) λ

Next, we loglinearize the auxiliary variable:

µ ¶−1
dt
aux dbt be
ct h ct−1 −b
be zet
auxe = d exp e
c exp − e c exp .
ze

The loglinear approximation is:

µ ¶−1 µ ¶−2
h h
auxd
auxt c− e
= d e c dbt − d e c− e c cb
ee
ct
ze ze
µ ¶−2
h he c ³b ´
+d ec− e c ct−1 − b
e zet .
ze ze
¡ ¢−1
c − hze e
Making use of the fact that aux = d e c , we find:
à µ ¶−1 µ ¶−1 !
h h hec ³ ´
auxt = aux dbt − e
auxd c− e c cb
ee
ct + e
c− e c b
ct−1 − b
e zet ,
ze ze ze

28
that simplifies to:
õ ¶−1 µ ¶−1 ³ !
h h h ´
auxt = aux dbt − 1 −
auxd b
e
ct + 1 − ct−1 − b
b
e zet
ze ze ze

µ ¶−1 Ã b b
!
h b he
c h e
z
uxt = dbt − 1 −
t−1 t
ad e
ct − + . (6)
ze ze ze

Putting the (5) and (6) together:

b
et =
(1 − hβz)λ
µ ¶−1 Ã ! ( µ ¶−1 Ã !)
h b hb
e
ct−1 hbzet h b h b
e
c hb
ze
dbt − 1 − − hβzEt dbt+1 − 1 −
t t+1
e
ct − + e
ct+1 − + .
ze ze ze ze ze ze

After some algebra, we arrive to the final expression:

2
et = dbt − hβzEt dbt+1 − 1¡ + h β¢ b
b
(1 − hβz)λ e
ct + ¡
h
¢ b
e
ct−1 + ¡
βhz
¢ Et
be
ct+1 − ¡
h
¢b
zet
1 − hze ze 1 − hze 1 − hze ze 1 − hze

This expression helps to understand the role of the habit persistence parameter h. If we set
h = 0 (i.e., no habit), we would get:

b
et = dbt − b
λ e
ct

where the lags and forward terms drop.

Now, we loglinearize the Euler equation:

et+1 zt Rt }.
et = βEt {λ
λ
zt+1 Πt+1

To do so, we write the expression as

Rt b
e λet+1 1 Re }
b b
e λe = βEt {λe
λe
zeezz,t ΠeΠb t+1

By using the fact that

Πe
z
R=
β
we simplify to:
b
b
e
λ
b
e t+1
λ 1 eRt
e = Et {e }
ezz,t eΠb t+1

29
Now, it is easy to show that:

b b
et = Et {λ
et+1 + R
bt − Π
b t+1 }.
λ (9)

Let us now consider:

ret = a0 [ut ] .

First, we write:
b £ ¤
reeret = a0 u expubt ,

where the loglinear approximation delivers:

reb
ret = a00 [u] ub
ut .

Since re = a0 [u], then:

b a00 [u] u
ret = 0 bt ,
u
a [u]
or
b γ
ret = 2 ût . (10)
γ1
The next equation to consider relates the shadow price of capital to the return on invest-
ment: ( )
et+1 1 1
λ
qet = βEt ((1 − δ) qet+1 + ret+1 ut+1 − a [ut+1 ])
et zet+1 μ
λ et+1
μt+1
et+1 =
where μ μt
. We can we write this expression as
( b
)
b β b
e b b (1 − δ) qeeqet+1 +
qeeqet = Et exp∆λt+1 −zet+1 −μe t+1 b £ ¤ .
zeμ
e reu expret+1 +but+1 −a u expubt+1

Loglinearization delivers:
µ ¶
β b
qeb
qet = e b
Et 4λt+1 − zet+1 − μ b
et+1 ((1 − δ) qe + reu − a [u])
zeμ
e
β ³ ³ ´ ´
+ qet+1 + reu b
Et (1 − δ) qeb ret+1 + u
bt+1 − a [u] ubut+1 .
zeμ
e

Making use of the following steady-state relationships: u = 1, a [u] = 0, re = a0 [u], qe = 1

30
 ³ ´
and 1 = β
zeμ
e
re + zeβμe (1 − δ), and Et −b b
et+1 = 0, the previous expression simplifies to:
zet+1 − μ

b β et+1 + β (1 − δ) Etb
b
qet = ((1 − δ) + re) Et ∆λ
β
qet+1 + reuEtb
ret+1 ,
zeμ
e zeμ
e zeμ
e

that implies: µ ¶
b b β (1 − δ) b β(1 − δ)
e
qet = Et ∆λt+1 + Et qet+1 + 1 − Etb
ret+1 . (11)
zeμ
e zeμ
e
The next equation to loglinearize is:
µ ∙ ¸ ∙ ¸ ¶ et+1 zt ∙ ¸µ ¶2
et zt
x 0 et zt
x et zt
x λ 0 xet+1 zt+1 et+1 zt+1
x
1 = qet 1 − S −S +βEt qet+1 S
et−1 zt−1
x et−1 zt−1 x
x et−1 zt−1 et zt+1
λ et zt
x et zt
x

which can be rearranged as:

³ h i h i ´
b
qet b
∆xet +b
zet 0 ∆xb
et +b
zet b
et +b
∆x zet
1 = qe exp 1 − S ze exp − S ze exp ze exp +
q̃ b h i b b
b b e b b
+β Et expqet −zet+1 +4λt+1 S ze exp∆xet+1 +zet+1 ze2 exp2(∆xet+1 +zet+1 ) .
00

ze

Taking the loglinear approximation (and using the fact that qe = 1) we get:
³ ´ β 00 ³ ´
b 00
0 = qet − S [e 2
z ] ze ∆x b b
et + zet + S [e 3 b b
et+1 + zet+1 .
z ] ze Et ∆x
z

Reorganizing:
³ ´
z 2 ∆b
κe x zet = b
et + b z 2 Et ∆b
qet + βκe et+1
x (12)

where κ comes from the adjustment cost function.

We move now into loglinearizing the equations that describe the law of motion of ft
µ χ ¶1−η µ ∗
¶η−1
η − 1 ∗ 1−η e Πt w et+1
w zt+1
ft = (w et )η ltd + βθw Et
et ) λt (w ft+1
η Πt+1 et∗ zt
w

and
µ χ ¶−η(1+γ) µ ¶η(1+γ)
−η(1+γ) ¡ d ¢1+γ Πt w w ∗
et+1 zt+1
ft = ψdt ϕt (Π∗w
t ) lt + βθw Et ft+1 .
Πt+1 et∗ zt
w

31
 The first equation can be written as:

b η − 1 ∗ 1−η e η d b∗ b e b bd
f expft = e ) λw
(w e l exp(1−η)wet +λt +ηwet +lt +
η
b b b b∗ b
βθ Π−(1−η)(1−χw ) zeη−1 f E expft −(1−η)(Πt+1 −χw Πt +∆wet+1 +zet+1 ) .
w t

Since Etb
zet+1 = 0, we can loglinearize the previous expression as:
µ ¶
η − 1 ∗ b
be + λ bet + b
f fbt = (w ∗ 1−η
e) λ ew η d
e l (1 − η) w t
et + η w d
lt +
η
³ ³ ∗
´´
βθw Π −(1−η)(1−χw ) η−1 b b b b
ze f Et ft+1 − (1 − η) (Πt+1 − χw Πt ) + ∆w
et+1 .

Since in the steady-state we have that

η−1 −(1−η)(1−χw )
η−1
η
(w∗ )1−η λwη ld
1 − βθw ze Π = ,
f

we can write that:

µ ¶
¡ ¢ ∗ b
fbt = η−1 −(1−η)(1−χw )
1 − βθw ze Π b e
et + λt + η w
(1 − η) w b bd
et + lt +
³ ³ ´´
βθw Π−(1−η)(1−χw ) zeη−1 Et fbt+1 − (1 − η) Π b t+1 − χw Π
b t + ∆wbe∗ (13)
t+1

Let us know consider the second equation describing the behavior of f . First we can write it
as:

b b b b∗ bd
f expft = ψdϕ(ld )1+γ expdt +bϕt +η(1+γ)(wet −wet )+(1+γ)lt +
b b b b∗ b
+βθ zeη(1+γ) Πη(1+γ)(1−χw ) f E expft+1 +η(1+γ)(Πt+1 −χw Πt +∆wet+1 +zet+1 ) .
w t

It can be shown that in steady-state:

η(1+γ) η(1+γ)(1−χw ) ψdϕ(ld )1+γ

1 − βθw ze Π =
f

and since Etb

zet+1 = 0, that implies
¡ ¢³ ³ ´ ´
fbt = 1 − βθw zeη(1+γ) Πη(1+γ)(1−χw ) dbt + ϕb t + η (1 + γ) w be∗ + (1 + γ)b
bet − w l d
t t
³ ³ ´´
+βθw zeη(1+γ) Πη(1+γ)(1−χw ) Et fbt+1 + η (1 + γ) Π b t+1 − χw Π be∗
b t + ∆w . (14)
t+1

32
 Let us loglinearize the law of motion for gt1 and gt2 . First consider
µ ¶−ε
et mct yed + βθp Et Πχt
gt1 =λ t
1
gt+1
Πt+1

that can be rewritten as:

b
e bd b b
1
e y d expλt +mc
g1 expbgt = λmce c t +yet
+βθp g1 Πε(1−χ) Et expε(Πt+1 −χΠt )+bgt+1 .
1

If we loglinearize that last expression, we get:

µ ¶ ³ ´
e y λ b
et + mc b d
b t+1 − χΠ
b t) + b
g1 b
gt1 = λmce d
c t + yet + βθp g 1 Πε(1−χ) Et ε(Π 1
gt+1 .

It can shown that in steady state

e yd
λmce
1 − βθp Πε(1−χ) =
g1

therefore,
µ ¶ ³ ´
¡ ¢ b d
gt1
b = 1 − βθp Πε(1−χ)
λet + mc b ε(1−χ)
c t + yet + βθp Πt b t+1 − χΠ
Et ε(Π b t ) + gbt+1
1
. (15)

Let us now consider:

µ ¶1−ε µ ¶
et Π∗ yed + βθp Et Πχt Π∗t
gt2 =λ t t
2
gt+1 ,
Πt+1 Π∗t+1

that can rewritten as:

b
e b ∗ bd b b b∗ b∗
2
e ∗ yed expλt +Πt +yet +βθp Π−(1−ε)(1−χ) g 2 Et exp−(1−ε)(Πt+1 −χΠt )−(Πt+1 −Πt )+bgt+1 .
g2 expgbt = λΠ
2

If we loglinearize that last expression, we get:

µ ¶
e ye λ b bt + b
et + Π d
g2b
gt2 = λΠ ∗ d ∗
yet
³ ³ ´ ³ ´ ´
+βθp Π−(1−ε)(1−χ) 2 b b b ∗ b ∗ 2
g Et −(1 − ε) Πt+1 − χΠt − Πt+1 − Πt + gbt+1 .

It can be shown that:

e ∗ yed
λΠ
1 − βθp Π−(1−ε)(1−χ) = ,
g2

33
therefore:
µ ¶
¡ ¢ b d
gt2
b = 1 − βθp Π−(1−ε)(1−χ) e b ∗ b
λt + Πt + yet
³ ³ ´ ³ ´ ´
+βθp Π−(1−ε)(1−χ) Et −(1 − ε) Π b t+1 − χΠ
bt − Πb ∗t+1 − Π
b ∗t + b2
gt+1 . (16)

Note that is easy to show that εgt1 = (ε − 1)gt2 loglinearizes to:

gbt1 = b
gt2 . (17)

Now, let us loglinearize the relationship between the capital-labor ratio and the real wage-
real interest rate
ut e
kt−1 α w et zt μt
d
= .
lt 1 − α ret zt−1 μt−1
It is easy to show that
b bet − b
u k t−1 − b
bt + e ltd = w ret + b b
zet + μ
et . (18)

Let us loglinearize the marginal cost

µ ¶1−α µ ¶α
1 1
mct = et )1−α retα
(w
1−α α

to get
mc bet + αb
c t = (1 − α)w ret . (19)

Let us now concentrate on the aggregate wage law of motion:

µ χ ¶1−η µ ¶1−η
Πt−1
w
w̃t−1 zt−1 ¡ ∗ ¢1−η
1 = θw + (1 − θw ) Πw
t
Πt w̃t zt

the above expression can be rewritten as:

b b bw ∗ b w∗
1 = θw Π−(1−χw )(1−η) ze−(1−η) exp−(1−η)(Πt −χw Πt−1 +Πt +ezt ) + (1 − θw ) (Πw )1−η exp(1−η)Πt ,

and loglinearized to:

θw Π−(1−χw )(1−η) ze−(1−η) (Π b t−1 + Π

b t − χw Π bw b w∗
et ) = (1 − θw ) (Πw )1−η Π
t +z t ,
∗

34
that we can write as:

θw Π−(1−χw )(1−η) ze−(1−η) b bw

b t−1 + Π be∗ − w
bet .
w ∗ 1−η (Πt − χw Π et ) = w
t +z t (20)
(1 − θw ) (Π )

Now we concentrate on the aggregate price law of motion:

µ ¶1−ε
Πχt−1
1 = θp + (1 − θp ) Π∗1−ε
t
Πt

that can be rewritten as:

b b b∗
1 = θp Π−(1−ε)(1−χ) exp−(1−ε)(Πt −χΠt−1 ) + (1 − θp ) (Π∗ )(1−ε) exp(1−ε)Πt

that loglinearizes to:

θp Π−(1−ε)(1−χ) b b t−1 ) = Π
b ∗t .
(Πt − χΠ (21)
(1 − θp ) (Π∗ )(1−ε)
The Taylor rule loglinnearizes to:
³ d
´
b b b b b
Rt = γ R Rt−1 + (1 − γ R ) γ Π Πt + γ y (4yet + zet ) + m
b t. (22)

The market clearing conditions:

lt = vtw ltd ,
a [ut ] e
kt−1
e et +
ct + x = yetd ,
et zet
μ

and
Aet ³ ´α ¡ ¢1−α
vtp yetd = e
ut kt−1 ltd −φ
zet
can be written as:
b w bd
l explt = vw ld expvbt +lt ,
h i b
b
et e
u e
kt−1
be b
a ue exp k exp b d
ct et
x d yet
e
c exp +e x exp + = e
y exp ,
e expμeb t +bzet
zeμ
and µ ¶
d Ae ³ ´α ³ ´1−α A
be b
et +α
t −z u
b
kt−1 +(1−α)e
bt +e ltd
btp +b
v p yed expv yet
= ue
k e
ld exp −φ
ze
et+1 =
respectively, where A At+1
.
At

35
 Loglinearizing we get:
b vtw + b
lt = b ltd , (23)
γ ek d
ct + x̃b
c̃b
e et + 1 u
x bt = ỹ db
yet , (24)
zeμ
e
and
³ d
´ Ae ³ ´α ³ ´1−α µ b µ
b
¶ ¶
d p p b
(ỹ v ) vbt + yet = e
uk e
l d e b
At − zet + α u e ed
bt + kt−1 + (1 − α) lt . (25)
ze

Let us consider now

µ ¶−ε
Πχt−1
vtp = θp p
vt−1 + (1 − θp ) Π∗−ε
t
Πt

that can be written as

p b b p b∗
v p expvbt = θp Πε(1−χ) v p expε(Πt −χΠt−1 )+bvt−1 + (1 − θp ) Π∗−ε expεΠt .

Using
¡ ¢
1 − θp Πε(1−χ) vp = (1 − θp ) Π∗−ε

we get: ³ ´ ¡ ¢ ∗
vbtp = θp Π ε(1−χ) b b p bt .
ε(Πt − χΠt−1 ) + vbt−1 − 1 − θp Πε(1−χ) εΠ (26)

Let us know consider the law of motion of the wage dispersion:

µ χ ¶−η
Πt−1
w
1 w̃t−1 −η
vtw = θw w
vt−1 + (1 − θw ) (Πw∗
t ) .
Πt zet w̃t

that can be written

w b b b b b w b∗ b
v w evbt = θw Πη(1−χw ) zeη v w eη(Πt −χw Πt−1 +wet −wet−1 +zet )+bvt−1 + (1 − θw ) (Πw∗ )−η e−η(wet −wet ) .

Using
¡ ¢
1 − θw Πη(1−χw ) zeη vw = (1 − θw ) (Πw∗ )−η

we get:
³ ³ ´ ´ ¡ ¢
vbtw = θw Πη(1−χw ) zeη η Πb t − χw Π bet − w
b t−1 + w bet−1 + b w
zet + vbt−1 be∗ −w
− 1 − θw Πη(1−χw ) zeη η(w bet ).
t
(27)
Finally, let us loglinearize the law of motion of capital

36
 µ ∙ ¸¶
x̃t
k̃t zet μ
et = (1 − δ) k̃t−1 + μ
et zet 1 − S zet x̃t .
x̃t−1
If we rearrange terms, we get:
b̃ b̃
³ h i´
kt +b b
zet +μ
et b b
et
zet +μ b b
et
zet +4x b
k̃ee exp
zμ = (1 − δ) k̃ exp kt−1
+ee exp
zμ 1 − S ze x̃ expxet .

Loglinearizing:
³ ´
k̃ee b̃
zμ kt + b et = (1 − δ) k̃b̃
b
zet + μ ex̃(b
kt−1 + zeμ b
et + b
zet + μ et ).
x

zeμ
e
Note that, using k̃ = zeμ
e −(1−δ)
x̃, we can rearrange the previous expression to get:

b̃ (1 − δ) b̃ zeμ
e − (1 − δ) b b
kt + b b
zet + μ
et = kt−1 + et + b
(zet + μ et )
x
zeμ
e zeμ
e

or
b̃ (1 − δ) b̃ zeμ
e − (1 − δ) b 1 − δ ³b b ´
kt = kt−1 + et −
x zet + μ
et . (28)
zeμ
e zeμ
e zeμ
e

4.3. System of Linear Stochastic Difference Equations

We now present the equations in the system as ordered in Uhlig algorithm model2fun.m

Equation 1 The first equation is

θw Π−(1−χw )(1−η) z̃ −(1−η) b bw

b t−1 + Π be∗ − w
bet
(Π t − χw Π t + e
zt ) = w t
(1 − θw ) (Πw∗ )1−η

We make use of the fact that Π bw bet − w

bet−1 , and define the auxiliary parameter a1 =
t = w
θw Π−(1−χw )(1−η) z̃ −(1−η) −(1−χw )(1−η) z̃ −(1−η)
1−η . In the Uhlig code this expression appears as a1 = θw Π 1−η ,
(1−θw )(Πw∗ ) (1−θw ){exp[log(Πw∗ )]}
∗
because when we solve for the steady state of Πw , we express it in log-levels.
Then, the equation boils down to:

b t − χw a1 Π
a1 Π bet − a1 w
b t−1 + a1 w be∗ − w
bet−1 + a1 zet = w bet
t

and rearranging:

b t − χw a1 Π
a1 Π bet − a1 w
b t−1 + (1 + a1 )w be∗ = 0
bet−1 + a1 zet − w (29)
t

37
Equation 2 The second equation is:

θp Π−(1−ε)(1−χ) b b t−1 ) = Π
b ∗t
(Πt − χΠ
(1 − θp ) (Π∗ )(1−ε)

θp Π−(1−ε)(1−χ)
In order to make notation for compact, define a2 = (1−θp )(Π∗ )(1−ε)
. Note that in the file this
Π−(1−ε)(1−χ)
expression appears as a2 = (1−θ θ){exp[log(Π
p
∗ )]}(1−ε)
, because when we solve for the steady state
p
value Π∗ , we express it in log-levels. Substituting for a2 :

b t − χΠ
a2 (Π b t−1 ) = Π
b ∗t

Rearranging:
a2 Π b t−1 − Π
b t − a2 χΠ b ∗t = 0 (30)

Equation 3 The third equation is

b
ret = φu ût ,

where φu = γ 2 /γ 1 . Then,
−b
ret + φu ût = 0 (31)

Equation 4 The fourth equation is

gbt1 = gbt2

Rearranging:
gbt1 − b
gt2 = 0 (32)

Equation 5 The fifth equation is:

b bet − b
bt + e
u kt−1 − b
ltd = w ret + b b
zet + μ
et

Rearranging:
b
bt + b
u ret + e
kt−1 − b bet − b
ltd − w b
zet − μ
et (33)

Equation 6 The sixth equation is:

mc bet + αb
c t = (1 − α)w ret

Rearranging:
bet + αb
(1 − α)w ret − mc
ct = 0 (34)

38
Equation 7 The seventh equation is:
³ d
´
bt = γ R R
R b t + γ y (4b
bt−1 + (1 − γ R ) γ Π Π yet + b
zet ) + m
bt

Rearranging:
d d
bt + γ R R
−R b t + (1 − γ R )γ y b
bt−1 + (1 − γ R )γ Π Π zet + (1 − γ R )γ y b
yet − (1 − γ R )γ y b b t = 0 (35)
yet−1 + m

Equation 8 The eighth equation is:

γ ek d
ct + x̃b
c̃b
e et + 1 u
x bt = ỹ db
yet
zeμ
e

Note that in the code, because we have solved for the log-steady h ³ ´i state, constants enter
£ ¡ ¢¤
as follows: c̃ = exp [log (c̃)], x̃ = exp [log (x̃)] , k̃ = exp log k̃ , ỹ d = exp log ỹ d .
Rearranging:
γ ek d
ct + x̃b
c̃b
e et + 1 u
x bt − ỹ db
yet = 0 (36)
zeμ
e

Equation 9 The ninth equation is:

³ d
´ Ae ³ ´α ³ ´1−α µ b µ
b
¶ ¶
d p p b
(ỹ v ) vbt + yet = e
uk e
l d e b
At − zet + α u e ed
bt + k t−1 + (1 − α) lt
ze
³ ´α ³ ´1−α
e
Define the parameter produc = Aze uek e
ld . Note that in terms of Uhlig notation, this
n h ioα n h io1−α
Ae e ed
is defined as produc = ze exp [log(u)] exp log(k) exp log(l ) . Also note that in
d
£ ¡ d ¢¤ p p
terms of the code, we have that ỹ = exp log ỹ , v = exp [log (v )]. Substituting:

³ ´ µ µ ¶ ¶
d b b
(ỹ v ) vbt + b
d p p
yet = produc Aet − b
zet + α u k t−1 + (1 − α) e
bt + e d
lt

Rearranging:

d be
vtp + (ỹ d vp )b
(ỹ d vp )b yet − (produc)A b
et − (α)(produc)b
t + (produc)z ut
b
−(α)(produc)e kt−1 − (1 − α) (produc)eltd = 0 (37)

39
Equation 10 The tenth equation is:
³ ´ ¡ ¢ ∗
b t − χΠ
vbtp = θp Πε(1−χ) ε(Π b t−1 ) + vbt−1
p bt
− 1 − θp Πε(1−χ) εΠ

Define a3 = βθp Πε(1−χ) . Then, θp Πε(1−χ) = aβ3 . This type of parameter definition will become
clearer when we analyze the price setting equations. Then, substituting:
µ ¶
a3 b a3 b a3 p a3 b ∗t
vbtp = εΠt − χε Πt−1 + vbt−1 − 1 − εΠ
β β β β

And rearranging:
µ ¶
a3 ε b a3 εχ b a3 p a3 b ∗t − vbtp = 0
Πt − Πt−1 + vbt−1 − 1 − εΠ (38)
β β β β

Equation 11 The eleventh equation is:

³ ³ ´ ´ ¡ ¢
vbtw = θw Πη(1−χw ) zeη η Π b t−1 + w
b t − χw Π bet − w
bet−1 + b w
zet + vbt−1 be∗ −w
− 1 − θw Πη(1−χw ) zeη η(w bet )
t

First, let’s define a4 = θw Πη(1−χw ) zeη . Substituting:

³ ³ ´ ´
vbtw = a4 η Πb t − χw Π bet − w
b t−1 + w bet−1 + b w
zet + vbt−1 be∗ − w
− (1 − a4 ) η(w bet )
t

Rearranging:
∗
b t − χw ηa4 Π
vbtw = a4 η Π bet − a4 η w
b t−1 + a4 η w bet−1 + a4 ηb w
zet + a4 vbt−1 be + (1 − a4 ) η w
− (1 − a4 ) η w bet
t

Then,

b t − χw ηa4 Π
a4 η Π bet − a4 η w
b t−1 + η w bet−1 + a4 ηb w
zet + a4 vbt−1 be∗ − vbw = 0
− (1 − a4 ) η w (39)
t t

Equation 12 The twelfth equation is:

lt = vbtw + b
b ltd

Which we rearrange in the code to appear as:

b vtw − b
lt − b ltd = 0 (40)

40
Equation 13 The thirteenth equation is:

b̃ (1 − δ) b̃ zeμ
e − (1 − δ) b 1 − δ ³b b ´
kt = kt−1 + et −
x et
zet + μ
zeμ
e zeμ
e zeμ
e

which we rearrange to:

∙ ¸
(1 − δ) b̃ (1 − δ) b 1 − δ ³b b ´ b̃
k t−1 + 1 − et −
x zet + μ
et − k t = 0 (41)
zeμ
e zeμ
e zeμ
e

Equation 14 The fourteenth equation is:

be b
b A et
t + αμ
zet =
1−α

which we rearrange to:

1 be α b b
At + e − zet = 0
μ (42)
1−α 1−α t

Equation 15 The fifteenth equation is:

b 1 + b2 β b b βbz̃ b
(1 − bβe et = dbt − bβe
z )λ z Et dbt+1 − ¡ b
ct + ¡
¢e b
¢b
ct−1 + ¡
e b
¢ Etb
ct+1 − ¡
e ¢b
zet
1 − ze ze 1 − ze 1 − ze ze 1 − zeb

Rearranging:

1 + b2 β b b b βbz̃ b b b
dbt −bβe
z Et dbt+1 − ¡ b
¢ e
ct + ¡ b
¢ e
ct−1 + ¡ b
¢ Et e
ct+1 − ¡ b
¢b
zet −(1−bβe et = 0 (43)
z )λ
1 − ze ze 1 − ze 1 − ze ze 1 − ze

Equation 16 The sixteenth equation is:

b b
et+1 + R
et = Et {λ bt − Π
b t+1 }
λ

which we rearrange to:

b b
et+1 − λ
et + R
bt − Π
b t+1 } = 0
Et {λ (44)

Equation 17 The seventeenth equation is:

µ ¶
b b β (1 − δ) b β(1 − δ)
e
qet = Et ∆λt+1 + Et qet+1 + 1 − Etb
ret+1
zeμ
e zeμ
e

41
which we rearrange to:
µ ¶
b b β (1 − δ) β(1 − δ)
et+1 − λ
Et λ et + Etb
qet+1 + 1 − ret+1 − b
Etb qet = 0 (45)
zeμ
e zeμ
e

Equation 18 The eighteenth equation is:

³ ´
z 2 ∆b
κe x zet = b
et + b z 2 Et ∆b
qet + βκe et+1
x

which we rearrange by undoing the first-difference operator:

b z 2 Et b
qet + βκe x z2b
et+1 − (1 + β)κe z2b
et + κe
x z 2b
et−1 − κe
x zet = 0 (46)

Equation 19 The nineteenth equation is:

µ ¶
¡ ¢ ∗ b
fbt = η−1 −(1−η)(1−χw )
1 − βθw ze Π b e
et + λt + η w
(1 − η) w b bd
et + lt +
³ ³ ´´
βθw Π−(1−η)(1−χw ) zeη−1 Et fbt+1 − (1 − η) Π b t + ∆w
b t+1 − χw Π be∗
t+1

define a5 = βθw zeη−1 Π−(1−η)(1−χw ) . Substituting:

∙ ¸ h ³ ´i
b b ∗ b
e b bd b b b b ∗
et + λt + η w
ft = (1 − a5 ) (1 − η) w et + lt + a5 Et ft+1 − (1 − η) Πt+1 − χw Πt + ∆w
et+1

Rearranging:

b
be∗ + (1 − a5 ) λ bet + (1 − a5 ) b
et + (1 − a5 ) η w
(1 − η) w t ltd + a5 Et fbt+1
b t+1 − (η − 1)a5 χw Π
+(η − 1)a5 Et Π be∗ − fbt = 0
b t − (1 − η)a5 Et w (47)
t+1

Equation 20 The twentieth equation is:

¡ ¢³ ³ ´ ´
fbt = 1 − βθw zeη(1+γ) Πη(1+γ)(1−χw ) dbt + ϕb t + η (1 + γ) w be∗ + (1 + γ)b
bet − w l d
t t
³ ³ ∗
´´
+βθw zeη(1+γ) Πη(1+γ)(1−χw ) Et fbt+1 + η (1 + γ) Π b t+1 − χw Π be
b t + ∆w t+1

Define a6 = βθw zeη(1+γ) Πη(1+γ)(1−χw ) . Then,

³ ³ ´ ´
fbt = (1 − a6 ) dbt + ϕ
b t + η (1 + γ) w be∗ + (1 + γ)b
bet − w l d
t t
³ ³ ∗
´´
+a6 Et fbt+1 + η (1 + γ) Π b t+1 − χw Π be
b t + ∆w t+1

42
Rearranging,

(1 − a6 ) dbt + (1 − a6 ) ϕ
b t + η (1 + γ) (1 − a6 ) w be∗ + (1 + γ) (1 − a6 ) b
bet − η (1 + γ) w ltd − fbt
t
∗
+a6 Et fbt+1 + a6 η (1 + γ) Et Π
b t+1 − a6 η (1 + γ) χw Π be = 0
b t + a6 η (1 + γ) Et w (48)
t+1

Equation 21 The twenty first equation is

µ ¶ ³ ´
¡ ¢ b d
gt1
b = 1 − βθp Πε(1−χ) e b ε(1−χ)
c t + yet + βθp Πt
λt + mc Et ε(Π b t ) + gbt+1
b t+1 − χΠ 1

As before, define a3 = βθp Πε(1−χ) . Then:

b d
(1 − a3 ) λ c t + (1 − a3 ) b
et + (1 − a3 ) mc b t+1 − χεa3 Π
yet + εa3 Et Π b t + a3 Et gbt+1
1
gt1 = 0
−b (49)

Equation 22 The twenty second equation is

µ ¶
¡ ¢ b d
gbt2 = 1 − βθp Π−(1−ε)(1−χ)
λ bt + b
et + Π∗
yet
³ ³ ´ ³ ´ ´
+βθp Π−(1−ε)(1−χ) Et −(1 − ε) Π b t+1 − χΠ
bt − Πb ∗t+1 − Π
b ∗t + b2
gt+1

Define: a7 = βθp Π−(1−ε)(1−χ) . Substituting:

b b ∗t + (1 − a7 ) b
et + Π d
b t+1 − χ(ε − 1)a7 Π
b t − a7 Et Π
b ∗t+1 + a7 Et gbt+1
2
(1 − a7 ) λ yet + (ε − 1)a7 Et Π gt2 = 0
−b

Shocks The preference shocks has the following structure:

dbt = ρd dbt−1 + εd,t

b t = ρϕ ϕ
ϕ b t−1 + εϕ,t

Note that Uhlig does not allow to write something like dbt = ρdbt−1 + σ d εd,t . We declare the
variance-covariance matrix later. The following shocks are not defined because of their i.i.d.
nature.

μb
et = zμ,t
be
A t = zA,t

mt = σ m εm,t

43
4.4. Solving the Model

Now, let µ ¶0
bet , b b bt , b d b b
statet = Πb t, w 1
gt , e
gt , b2
kt, R yet , b
e vt , vbt , b
ct , bp w
qet , fet , b et , b
et , λ
x zet ,
³ ´0
nstatet = b
ret , u b ∗t , b
bt , Π c t, b
ltd , mc be∗ ,
lt , w t
³ ´0
exot = zμ,t , dbt , ϕ
b t , zA,t , mt ,

and
εt = (εμ,t , εd,t , εϕ,t , εA,t , εm,t )0 .

Then, we need to write the system defined above in Uhlig’s format, i.e.:

0 = AA ∗ statet + BB ∗ statet−1 + CC ∗ nstatet + DD ∗ exot ,

à !
F F ∗ statet+1 + GG ∗ statet + HH ∗ statet−1
0 = Et ,
+JJ ∗ nstatet+1 + KK ∗ nstatet + LL ∗ exot+1 + MM ∗ exot
and
exot+1 = NN ∗ exot + Σ1/2 ∗ εt+1 with Et εt+1 = 0.

4.4.1. Writing the Model in Uhlig’s Form

First, note that from this section on, and in the codes, we define the variables in terms of
their loglinear deviation from steady-state. The matrices in Uhlig’s notation are as following

44
 ⎛ ⎞
a1 1 + a1 0 0 0 0 0 0 0 0 0 0 0 0 a1
⎜ ⎟
⎜ a2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 1 −1 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 −1 0 0 0 0 0 0 0 0 0 0 0 0 −1 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 1−α 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ (1 − γ R )γ Π 0 0 0 0 −1 (1 − γ R )γ y 0 0 0 0 0 0 0 (1 − γ R )γ y ⎟
⎜ ⎟
AA = ⎜ ⎟
⎜ 0 0 0 0 0 0 −ỹ d c̃ 0 0 0 0 x̃ 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 ỹ d vp 0 ỹ d vp 0 0 0 0 0 produc ⎟
⎜ ⎟
⎜ ⎟
⎜ a3 ε/β 0 0 0 0 0 0 0 −1 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ηa4 η 0 0 0 0 0 0 0 −1 0 0 0 0 ηa4 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 −1 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 −1 0 0 0 0 0 0 0 1 − (1−δ) 0 (1−δ)
− z̃μ̃ ⎟
⎝ z̃μ̃ ⎠
0 0 0 0 0 0 0 0 0 0 0 0 0 0 −1
⎛ ⎞
−χw a1 −a1 0 0 0 0 0 0 0 0 0 0 0 0 0
⎜ ⎟
⎜ −χa2 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
⎜ 0 0 0 0 1 0 0 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
⎜ 0 0 0 0 0 γ R −(1 − γ R )γ y 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
BB = ⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
⎜ 0 0 0 0 −α(produc) 0 0 0 0 0 0 0 0 0 ⎟ 0
⎜ ⎟
⎜ −a3 εχ ⎟
⎜ 0 0 0 0 0 0 0 a3
0 0 0 0 0 ⎟ 0
⎜ β β ⎟
⎜ −χ ηa −ηa 0 0 0 0 0 0 0 a4 0 0 0 0 0 ⎟
⎜ w 4 4 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 1−δ ⎟
⎝ 0 0 0 0 z̃μ̃
0 0 0 0 0 0 0 0 0 0 ⎠
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

45
 ⎛ ⎞
0 0 0 0 0 0 −1
⎜ ⎟
⎜ 0 0 −1 0 0 0 0 ⎟
⎜ ⎟
⎜ −1 φu 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 1 1 0 −1 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ α 0 0 0 −1 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
⎜ ⎟
CC = ⎜ γ 1 k̃ ⎟
⎜ 0 0 0 0 0 0 ⎟
⎜ z̃μ̃ ⎟
⎜ 0 −α(produc) 0 −(1 − α)(produc) 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 −(1 − a3
)ε 0 0 0 0 ⎟
⎜ β ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 −η(1 − a4 ) ⎟
⎜ ⎟
⎜ 0 0 0 −1 0 1 0 ⎟
⎜ ⎟
⎜ ⎟
⎝ 0 0 0 0 0 0 0 ⎠
0 0 0 0 0 0 0
⎛ ⎞
0 0 0 0 0
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ −1 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 1 ⎟
⎜ ⎟
DD = ⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 −produc 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ − 1−δ 0 0 0 0 ⎟
⎝ z̃μ̃ ⎠
α 1
1−α
0 0 1−α
0

46
 ⎛ βbz̃
⎞
0 0 0 0 0 0 0 1− z̃b
0 0 0 0 0 0 0
⎜ ⎟
⎜ −1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 ⎟
⎜ ⎟
⎜ β(1−δ) ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 z̃μ̃
0 0 1 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 βκz̃ 2 0 0 ⎟
FF = ⎜
⎜
⎟
⎟
⎜ a5 (η − 1) 0 0 0 0 0 0 0 0 0 0 a5 0 0 0 ⎟
⎜ ⎟
⎜ a6 η(1 + γ) 0 0 0 0 0 0 0 0 0 0 a6 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎝ a3 ε 0 a3 0 0 0 0 0 0 0 0 0 0 0 0 ⎠
a7 (ε − 1) 0 0 a7 0 0 0 0 0 0 0 0 0 0 0

⎛ ⎞
0 0 0 0 0 0 0 GG1,8 0 0 0 0 0 −(1 − bβ z̃) GG1,15
⎜ ⎟
⎜ 0 0 0 0 0 1 0 0 0 0 0 0 0 −1 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 −1 0 0 −1 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 1 0 GG4,13 0 −κz̃ 2 ⎟
GG = ⎜
⎜
⎟
⎟
⎜ −a5 χw (η − 1) (1 − a5 )η 0 0 0 0 0 0 0 0 0 −1 0 1 − a5 0 ⎟
⎜ ⎟
⎜ −a6 χw η(1 + γ) GG6,2 0 0 0 0 0 0 0 0 0 −1 0 0 0 ⎟
⎜ ⎟
⎜ −a3 εχ 0 −1 0 0 0 1 − a3 0 0 0 0 0 0 1 − a3 0 ⎟
⎝ ⎠
−a7 (ε − 1)χ 0 0 −1 0 0 1 − a7 0 0 0 0 0 0 1 − a7 0

where

1 + βb2
GG1,8 = −
1 − z̃b
b
GG1,15 = −
z̃(1 − z̃b )
GG6,2 = (1 − a6 )η(1 + γ)
GG4,13 = −(1 + β)κz̃ 2
⎛ b
⎞
0 0 0 0 0 0 0 z̃(1− z̃b )
0 0 0 0 0 0 0
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 κz̃ 2 0 0 ⎟
HH = ⎜
⎜
⎟
⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎝ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ⎠
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

47
 ⎛ ⎞
0 0 0 0 0 0 0
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 1− β(1−δ)
0 0 0 0 0 0 ⎟
⎜ z̃μ̃ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
JJ = ⎜
⎜
⎟
⎜ 0 0 0 0 0 0 −a5 (1 − η) ⎟
⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 a6 η(1 + γ) ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
⎝ ⎠
0 0 −a7 0 0 0 0
⎛ ⎞
0 0 0 0 0 0 0
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 0 0 ⎟
KK = ⎜
⎜
⎟
⎟
⎜ 0 0 0 1 − a5 0 0 1−η ⎟
⎜ ⎟
⎜ 0 0 0 (1 − a6 )(1 + γ) 0 0 −η(1 + γ) ⎟
⎜ ⎟
⎜ 0 0 0 0 1 − a3 0 0 ⎟
⎝ ⎠
0 0 1 0 0 0 0
⎛ ⎞
0 −βbz̃ 0 0 0
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
LL = ⎜
⎜ 0
⎟
⎜ 0 0 0 0 ⎟ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎝ ⎠
0 0 0 0 0
⎛ ⎞
0 1 0 0 0
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
MM = ⎜
⎜ 0
⎟
⎟
⎜ 0 0 0 0 ⎟
⎜ ⎟
⎜ 0 1 − a6 1 − a6 0 0 ⎟
⎜ ⎟
⎜ 0 0 0 0 0 ⎟
⎝ ⎠
0 0 0 0 0

48
 ⎛ ⎞
0 0 0 0 0
⎜ ⎟
⎜ 0 ρd 0 0 0 ⎟
⎜ ⎟
NN = ⎜
⎜ 0 0 ρϕ 0 0 ⎟
⎟
⎜ ⎟
⎝ 0 0 0 0 0 ⎠
0 0 0 0 0
⎛ ⎞
σ 2μ 0 0 0 0
⎜ 2 ⎟
⎜ 0 σd 0 0 0 ⎟
⎜ ⎟
Σ=⎜
⎜ 0 0 σ 2ϕ 0 0 ⎟
⎟
⎜ 2 ⎟
⎝ 0 0 0 σA 0 ⎠
0 0 0 0 σ 2m

4.4.2. Writing the Likelihood function

Uhlig’s solution method gives us the following solution:

statet = P P ∗ statet−1 + QQ ∗ exot

and
nstatet = RR ∗ statet−1 + SS ∗ exot .

We observe obst = (log Πt , log Rt , 4 log wt , 4 log yt )0 . Therefore, to write the likelihood
function, we need to write the model in the following state space form:

St = A ∗ St−1 + B ∗ εt
obst = C ∗ St−1 + D ∗ εt

where St−1 = (1, statet , statet−1 , exot−1 ).

4.4.3. Building the A and B matrices

Hence, ⎡ ⎤
1 0 0 0
⎢ ⎥
⎢ 0 PP 0 QQ ∗ NN ⎥
A=⎢
⎢
⎥
⎥
⎣ 0 I 0 0 ⎦
0 0 0 NN

49
and ⎡ ⎤
0
⎢ ⎥
⎢ QQ ⎥
B=⎢
⎢
⎥ ∗ Σ1/2 .
⎥
⎣ 0 ⎦
I

4.4.4. Building the C and D matrices

Define the observables vector:

obst = (log Πt , log Rt , 4 log wt , 4 log yt )0 .

Note that:

b t = P P (1, :) ∗ statet−1 + QQ (1, :) ∗ exot =

Π
P P (1, :) ∗ statet−1 + QQ (1, :) ∗ NN ∗ exot−1 + QQ (1, :) ∗ Σ1/2 ∗ εt ,

bt = P P (6, :) ∗ statet−1 + QQ (6, :) ∗ exot =

R
P P (6, :) ∗ statet−1 + QQ (6, :) ∗ NN ∗ exot−1 + QQ (6, :) ∗ Σ1/2 ∗ εt ,

bet − w
w bet−1 = P P (2, :) ∗ statet−1 + QQ (2, :) ∗ exot − P P (2, :) ∗ statet−2 − QQ (2, :) ∗ exot−1 =

P P (2, :) ∗ statet−1 − P P (2, :) ∗ statet−2

+QQ (2, :) ∗ NN ∗ exot−1 − QQ (2, :) ∗ exot−1 + QQ (2, :) ∗ Σ1/2 ∗ εt ,

and

yet − b
b yet−1 = P P (7, :) ∗ statet−1 + QQ (7, :) ∗ exot − P P (7, :) ∗ statet−2 − QQ (7, :) ∗ exot−1 =
P P (7, :) ∗ statet−1 − P P (7, :) ∗ statet−2
+QQ (7, :) ∗ N N ∗ exot−1 − QQ (7, :) ∗ exot−1 + QQ (7, :) ∗ Σ1/2 ∗ εt ,

Also, remember that Π b t = log Πt − log Πss , R bt = log Rt − log Rss , wbet − w
bet−1 = 4 log wet =
4 log wt − 4 log zt , and byet − b αΛ +Λ
yet−1 = 4 log yet = 4 log yt − 4 log zt . Since 4 log zt = 1−α A +
μ

αzμ,t +zA,t
1−α
, we have:
obst = C ∗ St−1 + D ∗ εt .

50
where ⎡ ⎤
log Πss P P (1, :) 0 QQ (1, :) ∗ NN
⎢ ⎥
⎢ log Rss P P (6, :) 0 QQ (6, :) ∗ NN ⎥
C=⎢
⎢ αΛμ +ΛA P P (2, :) −P P (2, :) QQ (2, :) ∗ (NN − I)
⎥
⎥
⎣ 1−α ⎦
αΛμ +ΛA
1−α
P P (7, :) −P P (7, :) QQ (7, :) ∗ (NN − I)
and ⎡ ⎤
QQ (1, :)
⎢ ⎥
⎢ QQ (6, :) ⎥
⎢ µ ¶ ⎥
⎢ ⎥
D=⎢ α 0 0 1 0 ⎥ ∗ Σ1/2
⎢ QQ (2, :) + ⎥
⎢ µ 1−α ¶ ⎥
⎣ α 0 0 1 0 ⎦
QQ (7, :) + 1−α

51