Macroeconometria PDF
Macroeconometria PDF
Macroeconometria PDF
a1 yt 1 + εt si yt 2 0
yt =
a2 yt 1 + εt si yt 2 <0
a1 yt 1 + εt si yt 2 0
yt =
a2 yt 1 + εt si yt 2 <0
a1 yt 1 + εt si yt 2 0
yt =
a2 yt 1 + εt si yt 2 <0
yt = aφ(yt 2 )yt 1 + εt
a1 yt 1 + εt si yt 2 0
yt =
a2 yt 1 + εt si yt 2 <0
yt = aφ(yt 2 )yt 1 + εt
yt = a1 yt 1 + ... + ap yt p + εt
yt = a1 yt 1 + ... + ap yt p + εt
donde
aj = φj + π j exp( αyt2 1 )
yt = a1 yt 1 + ... + ap yt p + εt
donde
aj = φj + π j exp( αyt2 1 )
yt = a1 yt 1 + ... + ap yt p + εt
donde
aj = φj + π j exp( αyt2 1 )
yt = a1 yt 1 + ... + ap yt p + εt
donde
aj = φj + π j exp( αyt2 1 )
yt = εt + bε2t 1
yt = a1 yt 1 + ... + ap yt p + εt
donde
aj = φj + π j exp( αyt2 1 )
yt = εt + bε2t 1
o en general NLMA(q)
yt = g (εt , εt 1 , ..., εt q )
Modelos bilineales
yt = εt + bεt 1 yt 2
Modelos bilineales
yt = εt + bεt 1 yt 2
o
yt = ayt 1 + εt + bεt 2 yt 2
Modelos bilineales
yt = εt + bεt 1 yt 2
o
yt = ayt 1 + εt + bεt 2 yt 2
yt = βt yt 1 + εt
βt = m + αβt 1 + η t
Modelos bilineales
yt = εt + bεt 1 yt 2
o
yt = ayt 1 + εt + bεt 2 yt 2
yt = βt yt 1 + εt
βt = m + αβt 1 + η t
Modelos bilineales
yt = εt + bεt 1 yt 2
o
yt = ayt 1 + εt + bεt 2 yt 2
yt = βt yt 1 + εt
βt = m + αβt 1 + η t
yt = g (zt 1 ) + h (zt 1 ) εt
yt = g (zt 1 ) + h (zt 1 ) εt
donde εt es MDS
yt = g (zt 1 ) + h (zt 1 ) εt
donde εt es MDS
en este caso
g (zt 1) = E (yt j zt 1)
yt = g (zt 1 ) + h (zt 1 ) εt
donde εt es MDS
en este caso
g (zt 1) = E (yt j zt 1)
h2 (zt 1) = V (yt j zt 1)
yt = µ + ∑ bi ,j εi ,t j + ∑ ∑ bi ,j ,kl εi ,t j εk ,t l + ...
i ,j i ,j k ,l
E uro P ound
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
0 4 -J a n -2 0 0 4 2 0 -M a r -2 0 0 5 0 4 -J u n -2 0 0 6 1 9 -A u g -2 0 0 7 0 4 -J a n -2 0 0 4 2 0 -M a r -2 0 0 5 0 4 -J u n -2 0 0 6 1 9 -A u g -2 0 0 7
2 3
C an Y en
2
1
1
0 0
-1
-1
-2
-2 -3
0 4 -J a n -2 0 0 4 2 0 -M a r -2 0 0 5 0 4 -J u n -2 0 0 6 1 9 -A u g -2 0 0 7 0 4 -J a n -2 0 0 4 2 0 -M a r -2 0 0 5 0 4 -J u n -2 0 0 6 1 9 -A u g -2 0 0 7
2 2
0 0
-2 -2
-4 -4
1 4 -J a n -2 0 0 0 2 6 -J u l-2 0 0 2 0 4 -F e b -2 0 0 5 1 7 -A u g -2 0 0 7 1 4 -J a n -2 0 0 0 2 6 -J u l-2 0 0 2 0 4 -F e b -2 0 0 5 1 7 -A u g -2 0 0 7
3 4
C an Y en
2
2
1
0 0
-1
-2
-2
-3 -4
1 4 -J a n -2 0 0 0 2 6 -J u l-2 0 0 2 0 4 -F e b -2 0 0 5 1 7 -A u g -2 0 0 7 1 4 -J a n -2 0 0 0 2 6 -J u l-2 0 0 2 0 4 -F e b -2 0 0 5 1 7 -A u g -2 0 0 7
Piensa que
Et J h (yt ) = Et J ...Et 2 Et 1 h (yt )
Piensa que
Et J h (yt ) = Et J ...Et 2 Et 1 h (yt )
y al hacer que J ! ∞
yt = εt σ t
σ2t = γ + αyt2 1
yt = εt σ t
σ2t = γ + αyt2 1
yt = εt σ t
σ2t = γ + αyt2 1
yt = εt σ t
σ2t = γ + αyt2 1
E (yt j zt 1) =0
2 2
V (yt j zt 1 ) = σt = γ + αyt 1
yt = εt σ t
σ2t = γ + αyt2 1
E (yt j zt 1) =0
2 2
V (yt j zt 1 ) = σt = γ + αyt 1
yt j zt 1 N (0, γ + αyt2 1 )
yt j zt 1 N (0, γ + αyt2 1 )
grá…camente
y 0.8
0.6
0.4
0.2
-5 -4 -3 -2 -1 0 1 2 3 4 5
x
yt j zt 1 N (0, γ + αyt2 1 )
grá…camente
y 0.8
0.6
0.4
0.2
-5 -4 -3 -2 -1 0 1 2 3 4 5
x
Eyt = ...Et 2 Et 1 yt =0
Eyt = ...Et 2 Et 1 yt =0
Eyt = ...Et 2 Et 1 yt =0
2 2 2
= ...Et 3 Et 2 σ t Et 1 ε t = ...Et 3 Et 2 σ t = ...Et 3 Et 2 ( γ + αyt2 1 )
Eyt = ...Et 2 Et 1 yt =0
2 2 2
= ...Et 3 Et 2 σ t Et 1 ε t = ...Et 3 Et 2 σ t = ...Et 3 Et 2 ( γ + αyt2 1 )
2 2 2
= ...Et 4 Et 3 ( γ + αEt 2 yt 1 ) = ...Et 4 Et 3 ( γ + αEt 2 εt 1 σ t 1 )
Eyt = ...Et 2 Et 1 yt =0
2 2 2
= ...Et 3 Et 2 σ t Et 1 ε t = ...Et 3 Et 2 σ t = ...Et 3 Et 2 ( γ + αyt2 1 )
2 2 2
= ...Et 4 Et 3 ( γ + αEt 2 yt 1 ) = ...Et 4 Et 3 ( γ + αEt 2 εt 1 σ t 1 )
= ...Et 4 Et 3 ( γ + ασ2t 1 Et 2
2 εt 1 ) = ..Et 4 Et 3 ( γ + ασ2t 1 )
2 2
= ...Et 4 (γ + α γ + αEt 3 εt 2 σ t 2 )
2
= ...Et 4 ( γ + α γ + αEt 3 ( γ + αyt 3 ) )
2 2
= ...Et 4 (γ + α γ + αEt 3 εt 2 σ t 2 )
2
= ...Et 4 ( γ + α γ + αEt 3 ( γ + αyt 3 ) )
γ
= γ + αγ + α2 γ + α3 γ + ... =
1 α
2 2
= ...Et 4 (γ + α γ + αEt 3 εt 2 σ t 2 )
2
= ...Et 4 ( γ + α γ + αEt 3 ( γ + αyt 3 ) )
γ
= γ + αγ + α2 γ + α3 γ + ... =
1 α
asumiendo que α < 1
Momentos incondicionales
Momentos incondicionales
Kurtosis: (ver tarea) sí hay, de hecho
3(1 α2 )
κ= >3
(1 3α2 )
Momentos incondicionales
Kurtosis: (ver tarea) sí hay, de hecho
3(1 α2 )
κ= >3
(1 3α2 )
Distribución condicional
Distribución condicional
yt es MDS
Distribución condicional
yt es MDS
Distribución condicional
yt es MDS
Distribución condicional
yt es MDS
Distribución condicional
yt es MDS
Distribución condicional
yt es MDS
Distribución condicional
yt es MDS
yt2 = γ + αyt2 1 + vt
Distribución condicional
yt es MDS
yt2 = γ + αyt2 1 + vt
o sea yt2 sigue un AR(1)
Lobato (ITAM) (Institute) MacEcApli/EcFin. 30 / 63
Modelo ARCH(1)
Como yt es MDS su predictor óptimo es 0
yt = εt σ t
σ2t = γ + α1 yt2 1 + α2 yt2 2 + ... + αp yt2 p
yt = εt σ t
σ2t = γ + α1 yt2 1 + α2 yt2 2 + ... + αp yt2 p
α (9 i )
αi = , para i = 1, ..., 8.
36
α (9 i )
αi = , para i = 1, ..., 8.
36
yt = εt σ t
σ2t = γ + α1 yt2 1 + ... + αp yt2 p + β1 σ2t 1 + ... + βq σ2t q
yt = εt σ t
σ2t = γ + α1 yt2 1 + ... + αp yt2 p + β1 σ2t 1 + ... + βq σ2t q
yt = εt σ t
σ2t = γ + α1 yt2 1 + ... + αp yt2 p + β1 σ2t 1 + ... + βq σ2t q
yt = εt σ t
σ2t = γ + α1 yt2 1 + ... + αp yt2 p + β1 σ2t 1 + ... + βq σ2t q
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
es MDS
yt2 = σ2t + (yt2 σ2t ) = σ2t + vt
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
es MDS
yt2 = σ2t + (yt2 σ2t ) = σ2t + vt
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
es MDS
yt2 = σ2t + (yt2 σ2t ) = σ2t + vt
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
es MDS
yt2 = σ2t + (yt2 σ2t ) = σ2t + vt
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
es MDS
yt2 = σ2t + (yt2 σ2t ) = σ2t + vt
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
es MDS
yt2 = σ2t + (yt2 σ2t ) = σ2t + vt
yt = εt σ t
σ2t = γ + α1 yt2 1 + β1 σ2t 1
es MDS
yt2 = σ2t + (yt2 σ2t ) = σ2t + vt
2 2
= ...Et = ...Et 3 Et 2 σ2t =
3 Et 2 σ t Et 1 ε t
2 2
...Et 3 Et 2 ( γ + α1 yt 1 + β1 σt 1 )
2 2
= ...Et = ...Et 3 Et 2 σ2t =
3 Et 2 σ t Et 1 ε t
2 2
...Et 3 Et 2 ( γ + α1 yt 1 + β1 σt 1 )
2 2
= ...Et 4 Et 3 ( γ + α 1 Et 2 yt 1 + β1 Et 2 σt 1 )
2 2 2
= ...Et 4 Et 3 ( γ + α 1 E t 2 ε t 1 σ t 1 + β 1 Et 2 σ t 1 )
2 2
= ...Et = ...Et 3 Et 2 σ2t =
3 Et 2 σ t Et 1 ε t
2 2
...Et 3 Et 2 ( γ + α1 yt 1 + β1 σt 1 )
2 2
= ...Et 4 Et 3 ( γ + α 1 Et 2 yt 1 + β1 Et 2 σt 1 )
2 2 2
= ...Et 4 Et 3 ( γ + α 1 E t 2 ε t 1 σ t 1 + β 1 Et 2 σ t 1 )
2
= ...Et 4 Et 3 ( γ + ( α 1 + β 1 ) Et 2 σt 1 )
2 2
= ...Et = ...Et 3 Et 2 σ2t =
3 Et 2 σ t Et 1 ε t
2 2
...Et 3 Et 2 ( γ + α1 yt 1 + β1 σt 1 )
2 2
= ...Et 4 Et 3 ( γ + α 1 Et 2 yt 1 + β1 Et 2 σt 1 )
2 2 2
= ...Et 4 Et 3 ( γ + α 1 E t 2 ε t 1 σ t 1 + β 1 Et 2 σ t 1 )
2
= ...Et 4 Et 3 ( γ + ( α 1 + β 1 ) Et 2 σt 1 )
p q q
yt2 = γ+ ∑ αj yt2 j ∑ βj yt2 j + ∑ βj σ2t j + vt
j =1 j =1 j =1
p q q
yt2 = γ+ ∑ αj yt2 j ∑ βj yt2 j + ∑ βj σ2t j + vt
j =1 j =1 j =1
p q q
yt2 = γ + ∑ αj yt2 j + ∑ βj yt2 j + ∑ βj (σ2t j yt2 j ) + vt
j =1 j =1 j =1
p q q
yt2 = γ+ ∑ αj yt2 j ∑ βj yt2 j + ∑ βj σ2t j + vt
j =1 j =1 j =1
p q q
yt2 = γ + ∑ αj yt2 j + ∑ βj yt2 j + ∑ βj (σ2t j yt2 j ) + vt
j =1 j =1 j =1
p q
yt2 = γ + ∑ φj yt2 j + ∑ βj vt j + vt
j =1 j =1
p q q
yt2 = γ+ ∑ αj yt2 j ∑ βj yt2 j + ∑ βj σ2t j + vt
j =1 j =1 j =1
p q q
yt2 = γ + ∑ αj yt2 j + ∑ βj yt2 j + ∑ βj (σ2t j yt2 j ) + vt
j =1 j =1 j =1
p q
yt2 = γ + ∑ φj yt2 j + ∑ βj vt j + vt
j =1 j =1
en donde φi = αi + βi y p = max(p, q )
p q q
yt2 = γ+ ∑ αj yt2 j ∑ βj yt2 j + ∑ βj σ2t j + vt
j =1 j =1 j =1
p q q
yt2 = γ + ∑ αj yt2 j + ∑ βj yt2 j + ∑ βj (σ2t j yt2 j ) + vt
j =1 j =1 j =1
p q
yt2 = γ + ∑ φj yt2 j + ∑ βj vt j + vt
j =1 j =1
en donde φi = αi + βi y p = max(p, q )
por tanto, yt2 sigue un ARMA(p , q)
o sea
γ + (1 + θ )yt2 1
σ2t =
(1 + θL)
o sea
γ + (1 + θ )yt2 1
σ2t =
(1 + θL)
por tanto
γ + (1 + θ )yt2 1
yt2 = σ2t + vt = + vt
(1 + θL)
o sea
γ + (1 + θ )yt2 1
σ2t =
(1 + θL)
por tanto
γ + (1 + θ )yt2 1
yt2 = σ2t + vt = + vt
(1 + θL)
o sea
(1 + θL)yt2 = γ + (1 + θ )yt2 1 + (1 + θL)vt
∆yt2 = γ + vt + θvt 1
o sea
γ + (1 + θ )yt2 1
σ2t =
(1 + θL)
por tanto
γ + (1 + θ )yt2 1
yt2 = σ2t + vt = + vt
(1 + θL)
o sea
(1 + θL)yt2 = γ + (1 + θ )yt2 1 + (1 + θL)vt
∆yt2 = γ + vt + θvt 1
con θ < 0
Lobato (ITAM) (Institute) MacEcApli/EcFin. 41 / 63
Modelo IGARCH
yeT +l jT = E (yT +l j zT ) = 0
yeT +l jT = E (yT +l j zT ) = 0
yeT +l jT = E (yT +l j zT ) = 0
ej ARCH(1)
ej ARCH(1)
se pueden estimar α y γ por MCO en la ecuación
yt2 = γ + αyt2 1 + vt
ej ARCH(1)
se pueden estimar α y γ por MCO en la ecuación
yt2 = γ + αyt2 1 + vt
ej ARCH(1)
se pueden estimar α y γ por MCO en la ecuación
yt2 = γ + αyt2 1 + vt
yt j zt 1 N (0, γ + αyt2 1 )
f (yT , yT 1 , ..., y2 , y1 )
= f (yT j yT 1 , ..., y1 ) f (yT 1 j yT 2 , ..., y2 , y1 ) ...f (y2 j y1 ) f (y1 )
f (yT , yT 1 , ..., y2 , y1 )
= f (yT j yT 1 , ..., y1 ) f (yT 1 j yT 2 , ..., y2 , y1 ) ...f (y2 j y1 ) f (y1 )
entonces
T
log L = log 2π
2
1 T 1 T yt2
2 t∑ ∑
log(γ + αyt2 1 ) + log f (y1 )
=2 2 t =2 γ + αyt2 1
f (yT , yT 1 , ..., y2 , y1 )
= f (yT j yT 1 , ..., y1 ) f (yT 1 j yT 2 , ..., y2 , y1 ) ...f (y2 j y1 ) f (y1 )
entonces
T
log L = log 2π
2
1 T 1 T yt2
2 t∑ ∑
log(γ + αyt2 1 ) + log f (y1 )
=2 2 t =2 γ + αyt2 1
yt = xt β + εt
Var (εt j zt 1 ) = γ + αε2t 1
yt = xt β + εt
Var (εt j zt 1 ) = γ + αε2t 1
yt = xt β + δσt + εt
yt = xt β + δσt + εt
yt = σt εt
yt = σt εt
σ2t = exp(ht )
yt = σt εt
σ2t = exp(ht )
ht = γ + φht 1 + ηt
yt = σt εt
σ2t = exp(ht )
ht = γ + φht 1 + ηt
donde
εt IIDN (0, 1)
yt = σt εt
σ2t = exp(ht )
ht = γ + φht 1 + ηt
donde
εt IIDN (0, 1)
yt = σt εt
σ2t = exp(ht )
ht = γ + φht 1 + ηt
donde
εt IIDN (0, 1)
la primera ecuación
yt = σt εt
la primera ecuación
yt = σt εt
la primera ecuación
yt = σt εt
σ2t = exp(ht )
la primera ecuación
yt = σt εt
σ2t = exp(ht )
en donde
ht = γ + φht 1 + ηt
la primera ecuación
yt = σt εt
σ2t = exp(ht )
en donde
ht = γ + φht 1 + ηt
la primera ecuación
yt = σt εt
σ2t = exp(ht )
en donde
ht = γ + φht 1 + ηt
E (yt j zt 1) = E (σt j zt 1 )E ( εt j zt 1)
E (yt j zt 1) = E (σt j zt 1 )E ( εt j zt 1)
= E (σt j zt 1) 0=0
E (yt j zt 1) = E (σt j zt 1 )E ( εt j zt 1)
= E (σt j zt 1) 0=0
usando que
E (εt j zt 1) = E ( εt ) = 0
V (yt j zt 1 ) = E (yt2 j zt 1 )
= E (σ2t j zt 1 )E (ε2t j zt 1) = E (σ2t j zt 1)
V (yt j zt 1 ) = E (yt2 j zt 1 )
= E (σ2t j zt 1 )E (ε2t j zt 1) = E (σ2t j zt 1)
usando que
E (ε2t j zt 1) = E (ε2t ) = 1
V (yt j zt 1 ) = E (yt2 j zt 1 )
= E (σ2t j zt 1 )E (ε2t j zt 1) = E (σ2t j zt 1)
usando que
E (ε2t j zt 1) = E (ε2t ) = 1
V (yt j zt 1 ) = E (yt2 j zt 1 )
= E (σ2t j zt 1 )E (ε2t j zt 1) = E (σ2t j zt 1)
usando que
E (ε2t j zt 1) = E (ε2t ) = 1
V (yt j zt 1 ) = E (yt2 j zt 1 )
= E (σ2t j zt 1 )E (ε2t j zt 1) = E (σ2t j zt 1)
usando que
E (ε2t j zt 1) = E (ε2t ) = 1
nota que
γ σ2η
ht N( , )
1 φ 1 φ2
nota que
γ σ2η
ht N( , )
1 φ 1 φ2
o sea σ2t sigue una log-normal
nota que
γ σ2η
ht N( , )
1 φ 1 φ2
o sea σ2t sigue una log-normal
cuyos momentos se derivan usan la función generadora de momentos
(FGM) de la normal
nota que
γ σ2η
ht N( , )
1 φ 1 φ2
o sea σ2t sigue una log-normal
cuyos momentos se derivan usan la función generadora de momentos
(FGM) de la normal
recuerda si x N (µ, σ2 ) entonces
2 σ2
E (e tx ) = e t µ+0.5t
Por tanto
Por tanto
E σ2t = E exp(ht )
Por tanto
E σ2t = E exp(ht )
γ 1 σ2η
E (yt2 ) = E σ2t = E exp(ht ) = exp( + )
1 φ 2 1 φ2
Por tanto
E σ2t = E exp(ht )
γ 1 σ2η
E (yt2 ) = E σ2t = E exp(ht ) = exp( + )
1 φ 2 1 φ2
Por tanto
E σ2t = E exp(ht )
γ 1 σ2η
E (yt2 ) = E σ2t = E exp(ht ) = exp( + )
1 φ 2 1 φ2
2γ 2σ2η
E (yt4 ) = E (ε4t )E σ4t = 3E exp(2ht ) = 3 exp( + )
1 φ 1 φ2
yt es RB (como ya sabíamos)
2γ σ2η
=? exp( + )
1 φ 1 φ2
2γ σ2η
=? exp( + )
1 φ 1 φ2
ahora
E (yt2 yt2 s ) = E (σ2t ε2t σ2t s ε2t s ) = E (ε2t ε2t s )E (σ2t σ2t s )
2γ σ2η
=? exp( + )
1 φ 1 φ2
ahora
E (yt2 yt2 s ) = E (σ2t ε2t σ2t s ε2t s ) = E (ε2t ε2t s )E (σ2t σ2t s )
Por tanto
2γ σ2η σ2η
E (yt2 yt2 s ) = E exp(ht + ht s) = exp( + +φ s
)
1 φ 1 φ2 1 φ2
Por tanto
2γ σ2η σ2η
E (yt2 yt2 s ) = E exp(ht + ht s) = exp( + +φ s
)
1 φ 1 φ2 1 φ2
y
cov (yt2 , yt2 s ) = E (yt2 yt2 s ) E (yt2 )2
Por tanto
2γ σ2η σ2η
E (yt2 yt2 s ) = E exp(ht + ht s) = exp( + +φ s
)
1 φ 1 φ2 1 φ2
y
cov (yt2 , yt2 s ) = E (yt2 yt2 s ) E (yt2 )2
Por tanto
2γ σ2η σ2η
E (yt2 yt2 s ) = E exp(ht + ht s) = exp( + +φ s
)
1 φ 1 φ2 1 φ2
y
cov (yt2 , yt2 s ) = E (yt2 yt2 s ) E (yt2 )2
! !
2γ σ2η s
σ2η
= exp( + ) exp(φ ) 1
1 φ 1 φ2 1 φ2
nota que
σ2η
exp(φs ) 1>0
1 φ2
nota que
σ2η
exp(φs ) 1>0
1 φ2
y que
σ2η
lim exp(φs ) 1=0
s !∞ 1 φ2
nota que
σ2η
exp(φs ) 1>0
1 φ2
y que
σ2η
lim exp(φs ) 1=0
s !∞ 1 φ2
por tanto la FACV de yt2 decae hacia cero a medida que s aumenta
Nota que
log yt2 = ht + log ε2t
nota que si εt es IID entonces log ε2t es también IID, además
Nota que
log yt2 = ht + log ε2t
nota que si εt es IID entonces log ε2t es también IID, además
ht = γ + φht 1 + ηt
ht = γ + φht 1 + ηt
tenemos
γ + ηt
zt = c + + εt
1 φL
ht = γ + φht 1 + ηt
tenemos
γ + ηt
zt = c + + εt
1 φL
o sea
zt = c + φzt 1 + γ + η t + εt φεt 1
ht = γ + φht 1 + ηt
tenemos
γ + ηt
zt = c + + εt
1 φL
o sea
zt = c + φzt 1 + γ + η t + εt φεt 1