Quantitative Macroeconomics
Q&A Session
Seda Basihos Andreas Tischbirek
Department of Economics
HEC Lausanne
14 December 2020
Log-Linearization
To solve nonlinear dynamic models, one needs to approximate the nonlinear
equations characterizing the equilibrium with log-linear ones.
Let xt a strictly positive variable, x̂t is logaritmic deviation from its
steady-state x:
xt
x̂t ≡ ln xt − ln x = ln
x
x xt − x
= ln +
x x
= ln(1 + % change ) ≈ % change
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization
Now suppose that we have a nonlinear function of xt such that f (xt ). Let
us define the Taylor approximation of logarithm of this function:
0
f (x)
ln f (xt ) ≈ ln f (x) + (xt − x)
f (x)
0
f (x) xt − x
≈ ln f (x) + x
f (x) x
0
f (x)x
≈ ln f (x) + x̂t
f (x)
UNIL-HEC Q&A Session 14 December 2020
Steady State
Recall that the system of nonlinear equations governs the dynamic
behaviour of RBC model:
At ktα (1 − It )1−α = ct + kt+1 − kt + δkt (1)
ln At = ρ ln At−1 + εt (2)
ct α−1
(1 − It+1 )1−α + 1 − δ = 1
Et β αAt+1 kt+1 (3)
ct+1
1
(1 − α)At ktα (1 − It )−α = χ (4)
ct
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Euler Equation (Eq. 3)
ct α−1 1−α
Et β αAt+1 kt+1 ht+1 + 1 − δ = 1
ct+1
where hours worked defined by ht ≡ 1 − lt
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Euler Equation (Eq. 3)
ct α−1 1−α
Et β αAt+1 kt+1 ht+1 + 1 − δ = 1
ct+1
Taking logarithm:
α−1 1−α
ln β + ln ct − ln Et ct+1 + ln Et αAt+1 kt+1 ht+1 + 1 − δ = 0
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Euler Equation (Eq. 3)
Recall the log-linearization formula
f1 (x1 , x2 , x3 )x1
ln f (x1,t , x2,t , x3,t , ...) ≈ ln f (x1 , x2 , x3 ) + x̂1,t
f (x1 , x2 , x3 )
f2 (x1 , x2 , x3 )x2
+ x̂2,t
f (x1 , x2 , x3 )
f3 (x1 , x2 , x3 )x2
+ x̂3,t
f (x1 , x2 , x3 )
+ ...
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Euler Equation (Eq. 3)
Applying this formula:
ln β + ln c + ĉt − Et (ln c + ĉt+1 ) + ln αAk α−1 h1−α + 1 − δ
αk α−1 h1−α A
+ Et Ât+1
αAk α−1 h1−α + 1 − δ
α(α − 1)Ak α−2 h1−α k
+ Et k̂t+1
αAk α−1 h1−α + 1 − δ
α(1 − α)Ak α−1 h−α h
+ Et ĥt+1 ≈ 0
αAk α−1 h1−α + 1 − δ
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Euler Equation (Eq. 3)
Note that in SS
ln β + ln c − ln c + ln αAk α−1 h1−α + 1 − δ = 0
Then we can re-write:
αk α−1 h1−α A
ĉt − Et ĉt+1 + Et Ât+1
αAk α−1 h1−α + 1 − δ
α(α − 1)Ak α−2 h1−α k
+ Et k̂t+1
αAk α−1 h1−α + 1 − δ
α(1 − α)Ak α−1 h−α h
+ Et ĥt+1 ≈ 0
αAk α−1 h1−α + 1 − δ
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Euler Equation (Eq. 3)
Le us make it more clear
αAk α−1 h1−α
ĉt − Et ĉt+1 + Et Ât+1
αAk α−1 h1−α + 1 − δ
α(α − 1)Ak α−1 h1−α
+ Et k̂t+1
αAk α−1 h1−α + 1 − δ
α(1 − α)Ak α−1 h1−α
+ Et ĥt+1 ≈ 0
αAk α−1 h1−α + 1 − δ
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Euler Equation (Eq. 3)
Note that αAk α−1 h1−α + 1 − δ is marginal product of capital and it equals
1
β in SS:
1
αAk α−1 h1−α + 1 − δ =
β
Then, we have
1 1 − (1 − δ)β
αAk α−1 h1−α = − (1 − δ) =
β β
Plugging these yields log-linearized form of Euler eq.:
ĉt − Et ĉt+1 + 1 − (1 − δ)β Et Ât+1
+ (α − 1) 1 − (1 − δ)β Et k̂t+1
+ (1 − α) 1 − (1 − δ)β Et ĥt+1 ≈ 0
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
At ktα (ht )1−α = ct + kt+1 − kt + δkt
where hours worked defined by ht ≡ 1 − lt
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
At ktα (ht)1−α = ct + kt+1 − kt + δkt
Recall the log-linearization formula
0
f (x)x
ln f (xt ) ≈ ln f (x) + x̂t
f (x)
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
At ktα (ht)1−α = ct + kt+1 − kt + δkt
Recall the log-linearization formula
0
f (x)x
ln f (xt ) ≈ ln f (x) + x̂t
f (x)
Let us first go for left-hand side:
1
⇒ ln At ≈ ln A + AÂt
A
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
At ktα (ht)1−α = ct + kt+1 − kt + δkt
Recall the log-linearization formula
0
f (x)x
ln f (xt ) ≈ ln f (x) + x̂t
f (x)
Let us first go for left-hand side:
1
⇒ ln At ≈ ln A + AÂt
A
1
⇒ α ln kt ≈ α(ln k + k k̂t )
k
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
At ktα (ht)1−α = ct + kt+1 − kt + δkt
Recall the log-linearization formula
0
f (x)x
ln f (xt ) ≈ ln f (x) + x̂t
f (x)
Let us first go for left-hand side:
1
⇒ ln At ≈ ln A + AÂt
A
1
⇒ α ln kt ≈ α(ln k + k k̂t )
k
1
⇒ (1 − α) ln ht ≈ (1 − α)(ln h + hĥt )
h
Then, we have:
ln A + α ln k + (1 − α) ln h + Ât + αk̂t + (1 − α)ĥt (5)
UNIL-HEC Q&A Session 14 December 2020
Apply the same to right-hand side:
So, what we have actually something like
f1 (x1 , x2 , x3 )x1
ln f (x1,t , x2,t , x3,t , ...) ≈ ln f (x1 , x2 , x3 ) + x̂1,t
f (x1 , x2 , x3 )
f2 (x1 , x2 , x3 )x2
+ x̂2,t
f (x1 , x2 , x3 )
+ ...
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
Apply the same to right-hand side:
c
ln ct + kt+1 − (1 − δ)kt ≈ ln c + k − (1 − δ)k + ĉt
c + k − (1 − δ)k
k (1 − δ)k
+ k̂t+1 − k̂t
c + k − (1 − δ)k c + k − (1 − δ)k
(6)
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
Using (5) and (6), we get:
c k
Ât + αk̂t + (1 − α)ĥt ≈ ĉt + k̂t+1
c + k − (1 − δ)k c + k − (1 − δ)k
(1 − δ)k
− k̂t
c + k − (1 − δ)k
Note that
ln A + α ln k + (1 − α) ln h = ln c + k − (1 − δ)k
Ak α h1−α = c + k − (1 − δ)k
UNIL-HEC Q&A Session 14 December 2020
Log-Linearization of Resource Constraint (Eq. 1)
We can write log-linearized resource constraint such:
(1 − δ)k c k
Ât + α + k̂t + (1 − α)ĥt ≈ ĉt + k̂t+1
c + δk c + δk c + δk
UNIL-HEC Q&A Session 14 December 2020
