Fiscal Fo esigh and he E ec s o Go e nmen Spending∗
Ma io Fo ni†
Uni e si `a di Modena e Reggio Emilia
CEPR and RECen
Luca Gambe i‡
Uni e si a Au onoma de Ba celona
and RECen
May 6, 2010
Abs ac
We s udy he e ec s o go e nmen spending by using a s uc u al, la ge dimensional,
dynamic ac o model. We ind ha he go e nmen spending shock is non- undamen al
o he a iables commonly used in he s uc u al VAR li e a u e, so ha i s impulse
esponse unc ions canno be consis en ly es ima ed by means o a VAR. Go e nmen
spending aises bo h consump ion and in es men , wi h no e idence o c owding ou .
The impac mul iplie is 1.7 and he long un mul iplie is 0.6.
JEL classi ica ion: C32, E32, E62.
Keywo ds: s uc u al ac o model, sign es ic ions, iscal policy, go e nmen spending
shock, undamen alness, non- undamen alness.
∗We a e g a e ul o Vale ie Ramey o p o iding us wi h he da ase o he pape “Iden i ying
go e nmen spending shocks: I ’s all in he iming”.
†Con ac : Dipa imen o di Economia Poli ica, ia Be enga io 51, 41100, Modena, I aly. Tel. +39
0592056851; e-mail: [email p o ec ed]
‡The inancial suppo om he Spanish Minis y o Science and Inno a ion h ough g an ECO2009-
09847 and he Ba celona G adua e School Resea ch Ne wo k is g a e ully acknowledged. Con ac :
O ice B3.1130 Depa amen d’Economia i His o ia Economica, Edi ici B, Uni e si a Au onoma de
Ba celona, Bella e a 08193, Ba celona, Spain. Tel (+34) 935814569; e-mail: luca.gamb[email p o ec ed]
1
1 In oduc ion
Unde s anding he e ec s o disc e iona y iscal policy ac ions is key o assessing com-
pe ing heo ies o he business cycle and p o iding guidance o policymake s. Recen
de elopmen s in he conduc o iscal policy in he US and o he indus ialized coun ies
ha e spa ked a enewed in e es in he opic. Li le consensus howe e has eme ged o e
he las yea s: economis s disag ee abou he sign o he esponse o p i a e agg ega e
demand componen s, in pa icula consump ion, and, as a consequence, he magni ude
o he go e nmen spending mul iplie . In hei seminal pape , Blancha d and Pe o i
(2002) use a VAR model and iden i y a go e nmen spending shock by imposing ha
go e nmen spending is no a ec ed on impac by he o he shocks. The main inding is
ha go e nmen spending leads o a la ge inc ease in consump ion. Simila esul s a e
ob ained by Fa as and Miho (2001), Gali, Lopez Salido, and Valles (2007), Moun o d
and Uhlig (2002), and Pe o i (2002, 2007), which may be included in he so-called
“go e nmen spending inno a ion app oach”. On he con a y, Ramey and Shapi o
(1998), using a dummy a iables iden i ica ion app oach, ind ha consump ion alls,
implying a e y small alue o he go e nmen spending mul iplie . Bu nside, Eichen-
baum, and Fishe (2004), Ca allo (2005), Edelbe g, Eichenbaum, and Fishe (1999)
and Eichenbaum and Fishe (2005) ind simila esul s.
Recen ly, a ew wo ks ha e con incingly a gued ha one o he in insic cha ac e -
is ic o iscal policy ac ions is ha hey a e an icipa ed (see e.g. Yang, 2007, Leepe ,
Walke and Yang, 2008, Me ens and Ra n, 2009). Tha is, p i a e agen s ecei e
signals abou u u e changes in axes and go e nmen spending be o e hese changes
ac ually ake place. The eason is he exis ence o legisla i e and implemen a ion lags:
i akes ime o a policy ac ion o be passed and implemen ed. The phenomenon is
called “ iscal o esigh ”; empi ical es ima es o he lag ange om a ew mon hs o a
couple o yea s.
Leepe , Walke and Yang (2008) show ha iscal o esigh poses a o midable chal-
lenge o he econome ician. The au ho s conside a simple neoclassical g ow h model
wi h wo shocks, a echnology and an an icipa ed ax shock. They show ha he MA
ep esen a ion o any pai o a iables among capi al, axes and echnology, is non-
undamen al; ha is, he de e minan o he MA ma ix has oo s smalle han one
in modulus. The implica ion is ha he a iables do no ha e a VAR ep esen a ion
in he s uc u al shocks, so ha he ue iscal policy shock and he ela ed impulse
esponse unc ions canno be ound by es ima ing a VAR.
The p oblem can be e o mula ed in e ms o in o ma ion se s. Typically, economic
agen s can see he s uc u al shocks. By con as , he econome ician can only obse e
he economic a iables. Ob iously, such a iables con ey in o ma ion abou he shocks,
bu i he impac e ec s a e small and he delayed e ec s a e la ge, such in o ma ion
2
is no enough o eco e he shocks (Lippi and Reichlin, 1993).
Pe suasi e e idence ha he in o ma ion se used in he VAR iscal policy li e a u e
is indeed oo poo is p o ided in Ramey (2009). Ramey shows ha he iscal policy
shock ob ained by using a VAR simila o he one in Pe o i (2007) is no an inno a ion
wi h espec o a ailable mac oeconomic in o ma ion, being G ange -caused by he
o ecas o go e nmen spending om he Su ey o P o essional Fo ecas e s.
In ecen yea s a ew wo ks ha e ied o o e come he p oblem posed by iscal
o esigh . Two di e en s a egies ha e been adop ed. On he one hand, Me ens and
Ra n (2009) es ima e he e ec s o go e nmen spending shocks using he me hodology
p oposed by Lippi and Reichlin (1994), based on Blaschke ma ices. On he o he
hand, some au ho s augmen he VAR wi h a iables p esumably con eying be e
in o ma ion abou disc e iona y iscal policy ac ions. Ramey (2009) cons uc s wo
se ies o exogenous go e nmen spending shocks: one is based on na a i e e idence o
de ense spending, he second is based on he Su ey o P o essional Fo ecas e s. Fishe
and Pe e s (2009) iden i y go e nmen spending shocks wi h s a is ical inno a ions o
he accumula ed excess e u ns o la ge US mili a y con ac o s. Bo h app oaches ha e
sho comings. The o me equi es many es ic ions, some o hem elying on he
co ec speci ica ion o he heo e ical model; mo eo e , he s uc u al shocks canno
be es ima ed consis en ly. As o he la e , i is ha d o judge whe he he addi ional
a iables included in he VAR a e ully success ul in cap u ing he ele an in o ma ion.
In his pape we depa om he VAR app oach and use ins ead a la ge s uc u al
ac o model. The mo i a ion is ha , as a gued in Fo ni, Giannone, Lippi and Reich-
lin (2009), la ge ac o models a e no a ec ed by he non- undamen alness p oblem.
The basic in ui ion is ha hese models ypically use mos o he a ailable mac oeco-
nomic in o ma ion and his helps in closing he gap be ween he in o ma ion se o he
econome ician and ha o economic agen s.
To be e unde s and how non- undamen alness a ises and how he ac o model can
a oid he p oblem, le us s a om a ec o MA ep esen a ion, ob ained om a DSGE
model. Typically he numbe o a iables is la ge han he numbe o shocks, so ha we
ha e a ec angula , “ all” MA sys em. As we shall show, o such sys ems, obse ing he
a iables is equi alen o obse ing he shocks, and he non- undamen alness p oblem
is no he e. In he model o Leepe , Walke and Yang (2008), o ins ance, he all
sys em made up by he h ee s a e a iables and he wo shocks is undamen al (see
Sec ion 2).
Un o una ely, a ec angula sys em canno be es ima ed by using s anda d VAR
echniques. This is because obse ed se ies do no ha e educed dynamic ank: by
es ima ing a VAR wi h n a iables, we end up wi h nlinea ly independen esiduals
and ind oo many s uc u al shocks. In o de o es ima e a VAR we ha e o igno e
3
some a iables and “cu ” he all sys em o ge a squa e one. Bu in such a way we
open he doo o non- undamen alness.
The ac o model ollows an al e na i e s a egy o handle he educed ank p ob-
lem. I e ains all o he a iables and adds measu emen e o s. Since he numbe
o a iables is e y la ge, and he e o s a e poo ly co ela ed ac oss sec ion, we can
ge id o hem by aking sui able linea combina ions o he a iables ( he p incipal
componen s). In such a way we end up wi h a undamen al, ec angula sys em which
can be es ima ed consis en ly by means o a educed ank VAR echnique.
Le us now summa ize ou main indings.
To begin, we ind ha he go e nmen spending shock is non- undamen al o he
a iables commonly used in he s uc u al VAR li e a u e. P ecisely, we selec a ew
squa e sub-ma ices o ou all impulse- esponse ma ix, co esponding o s anda d
VAR speci ica ions. Then we compu e he smalles oo o he de e minan and ind
ha in mos cases i is smalle han one in modulus.
Then we iden i y a go e nmen spending shock by using sign es ic ions (Uhlig,
2005). Mo e speci ically, an expansiona y go e nmen spending shock is de ined as a
shock ha ing a posi i e e ec on go e nmen expendi u e, ou pu , p ices, he p ime
a e, he go e nmen p ima y de ici and ax eceip s ( he las inequali y is imposed
o dis inguish he go e nmen spending shock om a ax shock). All es ic ions a e
imposed only on esponses delayed by six mon hs ( he hi d coe icien o he impulse
esponse unc ions), so ha he impac e ec on all a iables, and in pa icula go -
e nmen expendi u e, is le un es ic ed o a oid he iscal- o esigh c i icism.
The main esul s a e he ollowing. Fi s , ou es ima ed shock, unlike he VAR
shock, passes Ramey’s G ange -causa ion es , i.e. i is no caused by he o ecas o
go e nmen spending om he Su ey o P o essional Fo ecas e s. Second, he shape
o he impulse esponse unc ions sugges s ha ac ually he e is a g ea deal o an ic-
ipa ion. A e an immedia e and signi ican inc ease, go e nmen spending g adually
ises and eaches i s maximum, which is abou wo imes la ge han he ini ial e ec ,
a e a couple o yea s. By con as , he e ec on consump ion is ansi o y and eaches
i s maximum on impac . Finally, he e is no e idence o c owding-ou . Consump ion
eac s posi i ely o he iscal shock. Mo e su p isingly, he eac ion o o al in es men
is posi i e and signi ican on impac and becomes nega i e only in he long- un. Ou
es ima ed mul iplie is 1.7 on impac , eaches i s maximum, 2.2, a e 3 qua e s and
hen declines owa ds i s long- un alue, abou 0.6.
The emainde o he pape is o ganized as ollows: Sec ion 2 discusses non- undamen alness;
Sec ion 3 p esen s he ac o model; Sec ion 4 shows esul s; Sec ion 5 concludes.
4
2 Fundamen alness, s uc u al VARs and iscal o esigh
2.1 Fundamen alness in squa e and all sys ems
Le us conside he s a is ical MA ep esen a ion
χ =B(L)u ,(1)
whe e χ = (χ1 · · · χn )0is an n- ec o o weakly s a iona y a iables, B(L)isa(n×q)
ma ix o a ional unc ions in he lag ope a o L, wi h n≥q, and u = (u1 · · · uq )0is
aq-dimensional whi e-noise no malized o ha e iden i y a iance-co a iance ma ix.
By equa ion (1), χ lies in he space spanned by p esen and pas alues o u , i.e.
χ ∈Hu
= span(u,j= 1, . . . , q, τ ≤ ). Howe e , he con e se does no necessa ily
hold. I i does, i.e. u ∈Hχ
, we say ha ep esen a ion (1) is undamen al and u is
undamen al o χ . In such a case, obse ing χ is equi alen o obse ing u , in he
sense ha Hu
=Hχ
. Mo eo e , by he uniqueness o he o hogonal decomposi ion,
(B(L)−B(0))u is he p ojec ion o χ on o i s own pas Hχ
−1and B(0)u is he
esidual, i.e. he inno a ion o he in o ma ion se Hχ
.1A undamen al whi e noise
is no unique, bu i is easily seen ha i is also undamen al, hen i is a linea
ans o ma ion o u . By con as , non- undamen al whi e-noise ec o s can be ob ained
om u by applying linea il e s ha in ol e he u u e o u and he so-called Blaschke
ma ices (see e.g. Lippi and Reichlin, 1994).
I B(z) is a ional, as assumed abo e, we can cha ac e ize undamen alness in e ms
o i s ank: ep esen a ion (1) is undamen al i , and only i , he ank o B(z) is q o
all zsuch ha |z|<1 (see e.g. Rozano , 1967, Ch. 1, Sec ion 10, and Ch. 2, p. 76).
In he pa icula case n=q, such condi ion educes o he equi emen ha de B(z)
does no anish wi hin he uni ci cle in he complex plane.2
Ou main poin he e is ha , as a gued in Fo ni, Giannone, Lippi and Reichlin
(2009), he e is a subs an ial di e ence be ween he case n=q, on one hand, and
n>q, on he o he hand. In he o me case, he de e minan is a a ional unc ion,
which gene ally anishes somewhe e and may well anish wi hin he uni ci cle. In
he la e case, B(z) is a “ all”, ec angula ma ix; i s ank is less han q o some z
only i all o he (q×q) sub-ma ices o B(z) a e singula . Hence in gene al B(z) is
“ze oless”, i.e. has ank q o all z, and non- undamen alness is e y unlikely. Mo e
p ecisely, le ing pbe he l- ec o whose en ies a e he pa ame e s o B(L) and Π ∈Rl
he se o all possible p, in he case n > q undamen alness holds gene ically (i.e. he
subse o Π whe e undamen alness does no hold is meag e), whe eas in he case n=q
undamen alness is no gene ic.
1Con e sely, i B(0)u is he inno a ion o Hχ
,u is undamen al o χ .
2Obse e ha in e ibili y implies undamen alness, bu he con e se does no hold, because i he
ank alls o some uni modulus z, we do no ha e in e ibili y.
5

Conside o ins ance he simple case n= 2, q= 1, χ1 =u +b1u −1,χ2 =
u +b2u −1. Now conside he squa e subsys em made up by he i s equa ion: u
is non- undamen al o χ1 i and only i |b1|>1. In his case, he undamen al
ep esen a ion is χ1 =η +b−1
1η −1,η =(1 + b1L)/(b1+L−1)u .3Simila ly u is
non- undamen al o χ2 i and only i |b2|>1. Howe e , he all sys em made up by
bo h equa ion is non- undamen al i and only i b1=b2and |b1|>1. Obse e ha u
is gene ally undamen al o χ e en i i is non- undamen al o bo h χ1 and χ2 .
2.2 Fundamen alness and VAR models
Now le us assume ha (1) is de i ed as he solu ion o a DSGE model, so ha he
a iables in χ a e he mac oeconomic a iables o in e es , he en ies o u a e s uc-
u al shocks and B(L) is a ma ix o impulse- esponse unc ions (whose coe icien s a e
unc ions o he deep pa ame e s o he model). The numbe o a iables nis ypically
la ge han he numbe o shocks q, so ha B(L) is a all ma ix and χ is dynamically
singula (i.e. i s spec al densi y ma ix has educed ank q).
u is he inno a ion o he in o ma ion se o economic agen s. This is qui e eason-
able e en i u is no di ec ly obse able, because, as no ed abo e, u will be undamen al
o χ , excep o negligible cases, implying ha B(0)u is he esidual o he p ojec ion
o χ , which we assume obse able, on o i s own pas Hχ
−1.
A his s age he economis passes on he ba on o he econome ician. The aim is
o es ima e B(L) and u , s a ing om he in o ma ion in Hχ
. Un o una ely, mac oe-
conomic se ies a e no dynamically singula , pe haps because χ is obse ed wi h e o .
By es ima ing an n-dimensional VAR we would end up wi h nlinea ly independen
shocks, in con lic wi h he heo y. The s anda d s a egy is hen he ollowing: (i)
selec ing a squa e, q-dimensional subsys em, say
χ∗
=B∗(L)u ; (2)
(ii) es ima ing he VAR A(L)χ1 = o ind ou he inno a ions  =B∗(0)u and
he MA il e A(L)−1=B∗(L)B∗(0)−1; (iii) iden i ying B∗(0), and he e o e ep esen-
a ion (2), by imposing he no maliza ion B∗(0)B∗(0)−1= Σalong wi h iden i ying
es ic ions de i ed om heo e ical conside a ions.
Howe e , as a gued abo e, he squa e subsys em could be non- undamen al, o ,
equi alen ly, he educed in o ma ion space used by he econome ician, Hχ∗
, could be
smalle han he one o he agen s, Hu
. In such a case, u is no a linea ans o ma ion
o  , so ha s ep (ii) is w ong and he VAR canno p oduce he co ec esul , wha e e
be he iden i ica ion scheme adop ed in (iii). Ob iously, he choice o he subsys em in
3Obse e ha he Blaschke ac o b(L) = (1 + b1L)/(b1+L−1)is such ha b(z)b(z−1) = 1, so
ha he spec al densi y o η is cons an and η is whi e noise.
6
s ep (i) may be ele an , bu , as shown in he example below, a undamen al subsys em
does no necessa ily exis .
2.3 A iscal o esigh example
Leepe , Walke and Yang (2008) show ha iscal shock non- undamen alness in VAR
models na u ally a ises in an economy wi h iscal o esigh .4S a ing wi h a s anda d
g ow h model wi h log p e e ences and inelas ic labo supply, he au ho s ob ain he
equilib ium capi al accumula ion equa ion
k =λ1k −1−λ−1
2
∞
X
i=0
θiE (ν0a +i+1 −ν1a +i+ψτ +i+1) (3)
whe e k ,a and τ deno e he log o capi al, he log o echnology and he ax a e,
espec i ely, in de ia ion om he s eady s a e. The pa ame e s appea ing in he abo e
equa ion a e unc ions o he deep pa ame e s o he model; om he heo y we know
ha |θ|<1 (θis a discoun a e). Technology and axes ollow he exogenous law o
mo ions
a =uA,
τ =uτ, −2
whe e uτ, and uA, a e i.i.d. shocks ha economic agen s can obse e. The second
equa ion says ha he e ec o iscal policy on axes is delayed by wo pe iods.
Sol ing o k we ge 5



a
k
τ


=


0 1
−λ−1
2ψ(L+θ)
1−λ1L
λ−1
2ν1
1−λ1L
L20


 uτ,
uA, !=B(L)u
Le us conside he squa e subsys em gi en by he i s wo ows ( echnology and cap-
i al): he de e minan λ−1
2ψ(z+θ)
1−λ1z anishes o z=−θ, which is less han 1 in modulus.
Simila ly, he de e minan o he subma ix gi en by he i s and he las ows o B(z)
( echnology and axes) is z2, which anishes o z= 0. Finally, he de e minan o he
subsys em o med by he second and he las ow (capi al and axes) also anishes o
z= 0. In conclusion, u = (uτ, uA, )0is non- undamen al o any pai o a iables on
he le -hand side, implying ha s anda d VAR echniques a e unable o co ec ly es-
ima e he iscal shock. To be e app ecia e he ole o an icipa ion, obse e ha wi h
4Simple examples o non- undamen alness in economic models can also be ound in Lippi and Re-
ichlin (1993) and Fe n´andez-Villa e de, Rubio-Rami ez, Sa gen and Wa son (2006).
5S ic ly speaking he sys em is jus a block o he model since o simplici y we abs ac om
consump ion. Howe e he implica ions discussed la e emain unchanged.
7
no implemen a ion delay (τ =uτ, ), u would be undamen al o all o he subsys ems,
whe eas, wi h a one-pe iod delay (τ =uτ, −1), undamen alness would s ill hold o he
subsys em wi h axes and capi al. In ui i ely, he a iables con ey in o ma ion abou
he cu en alues o he iscal shock, as long as hey a e con empo aneously a ec ed by
such shock. In p esence o implemen a ion delay, axes a e no a ec ed on impac , and
he e o e do no p o ide use ul in o ma ion. Capi al is mo e help ul; howe e , i he
delay is la ge han one pe iod, i s con ibu ion is no su icien o eco e he shock.
Le us now conside he whole sys em. u is undamen al o χ , since B(z) is
ze oless, i.e. has ank 2 e e ywhe e in he complex plane. In ac , i is easily seen ha
χ has he educed ank VAR ep esen a ion6




1 0 0
λ−1
2ψL 1−λ1L0
ν1(L−θ)L2
ψθ2
(L−θ)(1−λ1L)L2
λ−1
2ψθ21−L2
θ2






a
k
ˆτ


=


0 1
−λ−1
2ψθ λ−1
2ν1
0 0


 uτ,
uA, !.
Pu di e en ly, p esen and pas alues o he h ee a iables, capi al, axes and echnol-
ogy, a e su icien o es ima e he wo shocks. Un o una ely, s anda d VAR echniques
canno be used o his end. In he nex sec ion we p esen a s uc u al ac o model,
whose co e is a all sys em like he one abo e. Such a sys em can be consis en ly
es ima ed h ough app op ia e p ocedu es.
3 The la ge ac o model
3.1 Rep esen a ion
In he p esen sec ion we p o ide a p esen a ion o ou model and es ima ion p ocedu e.
Fo addi ional de ails see Fo ni, Giannone, Lippi and Reichlin (2009), FGLR om now
on.
A con enien assump ion, which is s anda d in he la ge ac o model li e a u e, is
ha he e a e in ini ely many a iables xi ,i∈N. The econome ician obse es he
i s no hem, and consis ency esul s a e ob ained o bo h nand T( he numbe o
ime obse a ion) going o in ini y.
Each mac oeconomic a iable is he sum o wo mu ually o hogonal unobse able
componen s, he common componen χi and he idiosync a ic componen ξi :
xi =χi +ξi .(4)
The idiosync a ic componen s a e poo ly co ela ed in he c oss-sec ional dimension
(see FGLR, Assump ion 5 o a p ecise s a emen ). They a ise om shocks o sou ces
6No ice ha he VAR ep esen a ion has ini e o de . Exis ence o a ini e VAR ep esen a ion is a
gene al p ope y o ze oless all a ional sys ems (Ande son and Deis le 2008).
8
o a ia ion which conside ably a ec only a single a iable o a small g oup o a iables;
in his sense, we could say ha hey a e no “mac oeconomic” shocks. Fo a iables
ela ed o pa icula sec o s, like indus ial p oduc ion indexes o p oduc ion p ices,
he idiosync a ic componen may e lec sec o -speci ic a ia ions (wi h a sligh abuse
o language we could say “mic oeconomic” luc ua ions); o s ic ly mac oeconomic
a iables, like GDP, in es men o consump ion, he idiosync a ic componen mus be
in e p e ed essen ially as a measu emen e o . Wi h equa ion (1) in mind i is easily
seen ha he ac o model can be in e p e ed as he log-linea solu ion o a DSGE
model augmen ed wi h a measu emen e o .7
The common componen s a e esponsible o he main bulk o he co-mo emen s be-
ween mac oeconomic a iables, being linea combina ions o a ela i ely small numbe
o ac o s 1 , 2 ,· · · , , no depending on i:
χi =a1i 1 +a2i 2 +· · · +a i =ai .(5)
Such ac o s can be in e p e ed as he s a e a iables o he economic sys em.
The dynamic ela ions be ween he mac oeconomic a iables a ise om he ac
ha he ec o ollows he ela ion
=N(L)u ,(6)
whe e N(L) is a ×qma ix o a ional unc ions in he lag ope a o Land u =
(u1 u2 · · · uq )0is a q-dimensional ec o o o hono mal whi e noises, wi h q < .
Such whi e noises a e he s uc u al mac oeconomic shocks.8
Since N(L) is all, he discussion in he p e ious sec ion mo i a e he assump ion
ha N(z) is ze oless, i.e. ankN(z) = q o any z, which implies undamen alness. This
ensu e ha has he ini e o de VAR ep esen a ion (Ande son and Deis le , 2008)
D(L) = =Ru ,(7)
whe e D(L) is a × ma ix o polynomials such ha D(L)−1R=N(L) and R=N(0).
F om equa ions (4) o (7) i is seen ha he model can be w i en in he dynamic
o m
xi =bi(L)u +ξi ,(8)
7See also Al ug, 1989, Sa gen , 1989, and I eland 2004 o he link be ween ac o models and DSGE
models.
8In he la ge dynamic ac o model li e a u e hey a e some imes called he “common” o “p imi i e”
shocks o “dynamic ac o s” (whe eas he en ies o a e he “s a ic ac o s”). Equa ions (4) o (6)
need u he quali ica ion o ensu e ha all o he ac o s a e loaded, so o speak, by enough a iables
wi h la ge enough loadings (see FGLR, Assump ion 4); his “pe asi eness” condi ion is necessa y o
ha e uniqueness o he common and he idiosync a ic componen s, as well as he numbe o s a ic
ac o s and dynamic ac o s q.
9
p ices (CPI and he GDP de la o ), he p ime a e, he go e nmen p ima y de ici
and ax eceip s. The posi i e e ec on ou pu and p ices is imposed o dis inguish
he shock om a sys ema ic spending eac ion o a ecessiona y shock s emming om
he p i a e sec o . An inc easing de ici is imposed o exclude expendi u es en i ely
inanced wi h addi ional eceip s. The las inequali y is imposed o dis inguish he
go e nmen spending shock om a ax shock. All o he es ic ions a e imposed only
on he esponses delayed by six mon hs ( he hi d coe icien o he impulse esponse
unc ions), so ha he impac e ec on all a iables, and in pa icula go e nmen
spending, is le un es ic ed o a oid he iscal- o esigh c i icism.
Ha ing de ined he ele an sign es ic ions, we p oceeded as explained a he end
o Sec ion 3.2 o ge a se o admissible impulse esponse unc ions (sa is ying he
es ic ions) and a se o co esponding iscal shock se ies. We ob ained 350 admissible
shock se ies ou o 20,000 d awings o he o a ion pa ame e s. We ook he simple
a e age o such se ies as ou es ima e o he iscal shock.
Finally we pe o med he boo s apping p ocedu e explained a he end o Sec ion
3.3 o ge a pos e io densi y dis ibu ion o he impulse esponse unc ions. We
gene a ed 300 a i icial samples X∗and o each one o hem we d ew 1,000 o a ion
ec o s H1. We e ained 1,029 admissible se s o impulse esponse unc ions. In he
pic u es below we show he a e age along wi h he 16 h and he 84 h pe cen iles o he
ela ed dis ibu ion.
4.4 G ange causa ion and an icipa ion
Ha ing ob ained ou es ima e o he go e nmen spending shock we e i y whe he
such shock passes Ramey’s G ange -causa ion es . As al eady no ed, Ramey (2008)
shows ha he go e nmen spending shock ob ained wi h a VAR simila o ha o
Pe o i (2007) is G ange -caused by he go e nmen spending o ecas om he Su ey
o P o essional Fo ecas e s. He e we pe o m a simila exe cise using ou es ima ed
shock. Speci ically, we eg ess he go e nmen spending shock on ou lags o he shock
i sel and ou lags o he go e nmen spending o ecas . Table 4 shows he esul s.
None o he pa ame e s is s a is ically signi ican . The F-s a is ic ob ained unde he
null hypo hesis ha he pa ame e s o he lags o he o ecas a iable a e join ly ze o
is 1.862, which is e y much smalle han he 10% c i ical alue. In conclusion, ou
go e nmen spending shock is no G ange -caused by he go e nmen spending o ecas
se ies.
Le us now go deepe in o he analysis o go e nmen spending an icipa ion by exam-
ining he es ima ed impulse esponse unc ions. Figu es 1-3 depic he eac ion p o ile
o se e al a iables o in e es o a go e nmen spending shock which aises go e nmen
spending by one pe cen o GDP as he maximum e ec (ho izon 8). Consis en ly wi h
16

he exis ence o implemen a ion lags, go e nmen spending inc eases slowly, eaching
he maximal le el a e wo yea s. Abou one hal o he o al spending akes place
immedia ely, hal is delayed by one qua e o mo e. By con as , consump ion and
in es men each hei maximal le el ei he on impac (consump ion), o 1-2 qua -
e s a e he shock (GDP and in es men ). Hence, he spending is sp ead o e ime,
whe eas economic agen s eac immedia ely. This seems o i he s o y ha agen s
ecei e signals abou changes in axes and go e nmen spending, and eac o hem,
be o e hese changes a e ully in place.
4.5 C owding-ou and he mul iplie
Now le us look a he eac ion o GDP and i s componen s. GDP eac s immedia ely,
inc easing by 2%. The esponse s ays a ha le el o abou one yea and s a s
dec easing a e wa d, he e ec s being no longe signi ican a e 6 qua e s. Gi en
he no maliza ion imposed, he impulse esponse unc ion ep esen s he go e nmen
spending mul iplie . The es ima ed mul iplie is 1.7 on impac , eaches i s maximum,
2.2, a e 3 qua e s and hen declines owa ds i s long- un alue, abou 0.6. Con idence
bands show ha he mul iplie is signi ican ly abo e one a ho izon 2, while is no
di e en om ze o in he long un.
The size and shape o he mul iplie can be explained by looking a he esponse o
p i a e consump ion and in es men . Bo h consump ion and in es men immedia ely
and signi ican ly inc ease by abou 1% and 3% espec i ely. The esponse o consump-
ion is e y sho -li ed, declining and becoming no signi ican a e he second qua e .
On he con a y, he esponse o in es men appea s o be mo e pe sis en , he e ec
anishing only a e abou six qua e s. In he long un, he poin es ima e o he
esponse o bo h a iables is nega i e, al hough no signi ican .
By inspec ing he disagg ega ed componen s (Fig. 2), i is clea ha he esponse
o agg ega e in es men is mainly d i en by non- esiden ial in es men while esiden ial
in es men is c owded ou a e he nea ly ze o impac e ec . As a as consump ion is
conce ned, he h ee componen s eac posi i ely and signi ican ly on impac . Go e n-
men spending c owds-in, in he sho un, p i a e componen s o agg ega e demand.
Resul s o consump ion s and in sha p con as wi h he s anda d p edic ion o
RBC models, whe e go e nmen spending shock gene a e nega i e weal h e ec ha
dep esses consump ion. They a e consis en wi h he e idence in Blancha d and Pe o i
(2002), wi h he ema kable di e ence ha he e he esponse is empo a y and sho -
li ed, wi h he maximal e ec s obse ed on impac .
The esponse o in es men con adic s he s anda d ex book c owding-ou e -
ec igge ed by he inc ease in he in e es a e. A posi i e esponse o in es men ,
howe e , is he ou come p edic ed in Bax e and King (1993) a e a pe manen go -
17
e nmen spending shock. The e, he inc ease in in es men is caused by an inc ease in
he ma ginal p oduc i i y o capi al ollowing a sha p inc ease in employmen which is
also ound he e (see Fig. 3).17
4.6 Va iance decomposi ion
Table 2 shows he a iance decomposi ion o se e al a iables o in e es . Columns
2-5 epo he pe cen age o o ecas e o a iance o he a iables lis ed in column 1,
accoun ed o by he shock a a ious ho izons. Column 6 epo s he pe cen age o
he a iance o he se ies, ans o med o each s a iona i y (e.g. in la ion ins ead o
p ices), accoun ed o by he shock. The shock accoun s o abou 25% o he a iance
o go e nmen spending (bo h ede al and o al) and abou 10% o he a iance o
de ici and axes. A a i s sigh , hese numbe s could seem small bu ecall ha (i) we
a e uling ou ax shocks; (ii) we a e uling ou spending no inc easing cu en de ici ;
and (iii) his is disc e iona y policy, in ha i excludes sys ema ic eac ions o shocks
s emming om he p i a e sec o .
The shock accoun s o abou 16%, 13% and 16% o he a iance o GDP, in es men
and consump ion, espec i ely. In e es ingly, he shock is mo e impo an in he e y
sho un (on impac i explains 21% 14% and 20% o he h ee a iables, espec i ely)
han a longe ho izons (a ho izon 20 pe cen ages a e 9%, 8% and 9%, espec i ely.
4.7 Robus ness
This subsec ion s udies he obus ness o he esul s o changes in model speci ica ion.
Fi s le us compa e he esul s o ou benchma k speci ica ion ( = 13, q= 6) wi h
i e al e na i e speci ica ions: 1) = 10, q= 6; 2) = 16, q= 6; 3) = 10, q= 4;
4) = 13, q= 4; 5) = 16, q= 4. Figu e 4 displays he impulse esponse unc ions
o consump ion and in es men o he six di e en speci ica ions. The i s column
depic s he esponses o he 4 dynamic shock speci ica ion, he second hose o he
6 dynamic shock speci ica ion. O e all he esul s a e ema kably simila bo h om a
quali a i e and om a quan i a i e poin o iew. The only mino di e ence is ha he
e ec s end o be sligh ly la ge in he 10 s a ic ac o speci ica ion and sligh ly smalle
in he 16 ac o speci ica ion han in ou benchma k.
We also made se e al o he checks lis ed below.
1) We used he ede al unds a e and he 10 yea bond a e ins ead o he p ime a e
o iden i y he shock.
2) We used ede al go e nmen spending ins ead o and oge he wi h o al go e nmen
17Howe e , unlike Bax e and King (1993), he e he pe sis en inc ease in labo canno be caused by
a nega i e weal h e ec , gi en ha consump ion inc eases.
18
spending o iden i y he shock.
3) We did no es ic he in e es a e.
4) We used wo ins ead o h ee lags in he VAR o he ac o s.
5) We imposed he iden i ying es ic ion o pe iods 4 o /and 5.
6) We used he second (ins ead o i s ) di e ences o he log o p ices and o he nominal
a iables.
7) We used he es ima ion p ocedu e p oposed by Fo ni and Lippi (2010).
In all hese expe imen s we ound he same esul s ob ained in he benchma k model.
O e all esul s seem o be obus o changes in model speci ica ion.
5 Conclusions
This pape s udied he e ec s o go e nmen spending shocks in he US using a s uc-
u al, la ge dimensional, dynamic ac o model. The main mo i a ion is ha in his
model, unlike in VARs, he shocks a e undamen al e en in p esence o iscal o esigh .
We ind ha he go e nmen spending shock is non- undamen al o he a iables com-
monly used in he s uc u al VAR li e a u e, so ha i s impulse esponse unc ions
canno be consis en ly es ima ed by means o a VAR. Go e nmen spending aises
bo h consump ion and in es men , wi h no e idence o c owding ou . The impac
mul iplie is 1.7 and he long un mul iplie is 0.6.
19
Appendix: Da a
T ans o ma ions: 1=le els, 2= i s di e ences o he o iginal se ies, 5= i s di e ences
o logs o he o iginal se ies, 5= second di e ences o logs o he o iginal se ies.
no.se ies T ans . Mnemonic Long Label
1 5 GDPC1 Real G oss Domes ic P oduc , 1 Decimal
2 5 GNPC96 Real G oss Na ional P oduc
3 5 NICUR/GDPDEF Na ional Income/GDPDEF
4 5 DPIC96 Real Disposable Pe sonal Income
5 5 OUTNFB Non a m Business Sec o : Ou pu
6 5 FINSLC1 Real Final Sales o Domes ic P oduc , 1 Decimal
7 5 FPIC1 Real P i a e Fixed In es men , 1 Decimal
8 5 PRFIC1 Real P i a e Residen ial Fixed In es men , 1 Decimal
9 5 PNFIC1 Real P i a e Non esiden ial Fixed In es men , 1 Decimal
10 5 GPDIC1 Real G oss P i a e Domes ic In es men , 1 Decimal
11 5 PCECC96 Real Pe sonal Consump ion Expendi u es
12 5 PCNDGC96 Real Pe sonal Consump ion Expendi u es: Nondu able Goods
13 5 PCDGCC96 Real Pe sonal Consump ion Expendi u es: Du able Goods
14 5 PCESVC96 Real Pe sonal Consump ion Expendi u es: Se ices
15 5 GPSAVE/GDPDEF G oss P i a e Sa ing/GDP De la o
16 5 FGCEC1 Real Fede al Consump ion Expendi u es & G oss In es men , 1 Decimal
17 5 FGEXPND/GDPDEF Fede al Go e nmen : Cu en Expendi u es/ GDP de la o
18 5 FGRECPT/GDPDEF Fede al Go e nmen Cu en Receip s/ GDP de la o
19 2 FGDEF Fede al Real Expend-Real Receip s
20 1 CBIC1 Real Change in P i a e In en o ies, 1 Decimal
21 5 EXPGSC1 Real Expo s o Goods & Se ices, 1 Decimal
22 5 IMPGSC1 Real Impo s o Goods & Se ices, 1 Decimal
23 5 CP/GDPDEF Co po a e P o i s A e Tax/GDP de la o
24 5 NFCPATAX/GDPDEF Non inancial Co po a e Business: P o i s A e Tax/GDP de la o
25 5 CNCF/GDPDEF Co po a e Ne Cash Flow/GDP de la o
26 5 DIVIDEND/GDPDEF Ne Co po a e Di idends/GDP de la o
27 5 HOANBS Non a m Business Sec o : Hou s o All Pe sons
28 5 OPHNFB Non a m Business Sec o : Ou pu Pe Hou o All Pe sons
29 5 UNLPNBS Non a m Business Sec o : Uni Nonlabo Paymen s
30 5 ULCNFB Non a m Business Sec o : Uni Labo Cos
31 5 WASCUR/CPI Compensa ion o Employees: Wages & Sala y Acc uals/CPI
32 5 COMPNFB Non a m Business Sec o : Compensa ion Pe Hou
33 5 COMPRNFB Non a m Business Sec o : Real Compensa ion Pe Hou
20
no.se ies T ans . Mnemonic Long Label
34 5 GDPCTPI G oss Domes ic P oduc : Chain- ype P ice Index
35 5 GNPCTPI G oss Na ional P oduc : Chain- ype P ice Index
36 5 GDPDEF G oss Domes ic P oduc : Implici P ice De la o
37 5 GNPDEF G oss Na ional P oduc : Implici P ice De la o
38 5 INDPRO Indus ial P oduc ion Index
39 5 IPBUSEQ Indus ial P oduc ion: Business Equipmen
40 5 IPCONGD Indus ial P oduc ion: Consume Goods
41 5 IPDCONGD Indus ial P oduc ion: Du able Consume Goods
42 5 IPFINAL Indus ial P oduc ion: Final P oduc s (Ma ke G oup)
43 5 IPMAT Indus ial P oduc ion: Ma e ials
44 5 IPNCONGD Indus ial P oduc ion: Nondu able Consume Goods
45 2 AWHMAN A e age Weekly Hou s: Manu ac u ing
46 2 AWOTMAN A e age Weekly Hou s: O e ime: Manu ac u ing
47 2 CIVPART Ci ilian Pa icipa ion Ra e
48 5 CLF16OV Ci ilian Labo Fo ce
49 5 CE16OV Ci ilian Employmen
50 5 USPRIV All Employees: To al P i a e Indus ies
51 5 USGOOD All Employees: Goods-P oducing Indus ies
52 5 SRVPRD All Employees: Se ice-P o iding Indus ies
53 5 UNEMPLOY Unemployed
54 5 UEMPMEAN A e age (Mean) Du a ion o Unemploymen
55 2 UNRATE Ci ilian Unemploymen Ra e
56 5 HOUST Housing S a s: To al: New P i a ely Owned Housing Uni s S a ed
57 2 FEDFUNDS E ec i e Fede al Funds Ra e
58 2 TB3MS 3-Mon h T easu y Bill: Seconda y Ma ke Ra e
59 2 GS1 1-Yea T easu y Cons an Ma u i y Ra e
60 2 GS10 10-Yea T easu y Cons an Ma u i y Ra e
61 2 AAA Moody’s Seasoned Aaa Co po a e Bond Yield
62 2 BAA Moody’s Seasoned Baa Co po a e Bond Yield
63 2 MPRIME Bank P ime Loan Ra e
64 5 BOGNONBR Non-Bo owed Rese es o Deposi o y Ins i u ions
65 5 TRARR Boa d o Go e no s To al Rese es, Adjus ed o Changes in Rese e
66 5 BOGAMBSL Boa d o Go e no s Mone a y Base, Adjus ed o Changes in Rese e
67 5 M1SL M1 Money S ock
68 5 M2MSL M2 Minus
69 5 M2SL M2 Money S ock
21

no.se ies T ans . Mnemonic Long Label
70 5 BUSLOANS Comme cial and Indus ial Loans a All Comme cial Banks
71 5 CONSUMER Consume (Indi idual) Loans a All Comme cial Banks
72 5 LOANINV To al Loans and In es men s a All Comme cial Banks
73 5 REALLN Real Es a e Loans a All Comme cial Banks
74 5 TOTALSL To al Consume C edi Ou s anding
75 5 CPIAUCSL Consume P ice Index Fo All U ban Consume s: All I ems
76 5 CPIULFSL Consume P ice Index o All U ban Consume s: All I ems Less Food
77 5 CPILEGSL Consume P ice Index o All U ban Consume s: All I ems Less Ene gy
78 5 CPILFESL Consume P ice Index o All U ban Consume s: All I ems Less Food & Ene gy
79 5 CPIENGSL Consume P ice Index o All U ban Consume s: Ene gy
80 5 CPIUFDSL Consume P ice Index o All U ban Consume s: Food
81 5 PPICPE P oduce P ice Index Finished Goods: Capi al Equipmen
82 5 PPICRM P oduce P ice Index: C ude Ma e ials o Fu he P ocessing
83 5 PPIFCG P oduce P ice Index: Finished Consume Goods
84 5 PPIFGS P oduce P ice Index: Finished Goods
85 5 OILPRICE Spo Oil P ice: Wes Texas In e media e
86 5 USSHRPRCF US Dow Jones Indus ials Sha e P ice Index (EP) NADJ
87 5 US500STK US S anda d & Poo ’s Index i 500 Common S ocks
88 5 USI62...F US Sha e P ice Index NADJ
89 5 USNOIDN.D US Manu ac u e s New O de s o Non De ense Capi al Goods (BCI 27)
90 5 USCNORCGD US New O de s o Consume Goods & Ma e ials (BCI 8) CONA
91 1 USNAPMNO US ISM Manu ac u e s Su ey: New O de s Index SADJ
92 5 USVACTOTO US Index o Help Wan ed Ad e ising VOLA
93 5 USCYLEAD US The Con e ence Boa d Leading Economic Indica o s Index SADJ
94 5 USECRIWLH US Economic Cycle Resea ch Ins i u e Weekly Leading Index
95 2 GS10-FEDFUNDS
96 2 GS1-FEDFUNDS
97 2 BAA-FEDFUNDS
98 5 GEXPND/GDPDEF Go e nmen Cu en Expendi u es/ GDP de la o
99 5 GRECPT/GDPDEF Go e nmen Cu en Receip s/ GDP de la o
100 2 GDEF Go e nnen Real Expend-Real Receip s
101 5 GCEC1 Real Go e nmen Cons. Expendi u es & G oss In es men , 1 Decimal
102 5 Real Fede al Cons. Expendi u es & G oss In es men Na ional De ense
103 2 Fede al p ima y de ici
104 5 Real Fede al Cu en Tax Re enues
105 5 Real Go e nmen Cu en Tax Re enues
106 2 Go e nmen p ima y de ici
22
Re e ences
[1] Al ug, S. (1989). Time- o-Build and Agg ega e Fluc ua ions: Some New E idence,
In e na ional Economic Re iew 30, 889-920.
[2] Amengual, D. and M.W. Wa son (2007). Consis en Es ima ion o he Numbe o
Dynamic Fac o s in a La ge N and T Panel, Jou nal o Business and Economic
S a is ics 25, 91-96.
[3] Bai, J. (2003). In e en ial Theo y o Fac o Models o La ge Dimensions, Econo-
me ica 71, 135-171.
[4] Bai, J., and S. Ng (2002). De e mining he numbe o ac o s in app oxima e ac o
models, Econome ica 70, 191-221.
[5] Bai, J., and S. Ng (2007). De e mining he Numbe o P imi i e Shocks in Fac o
Models, Jou nal o Business and Economic S a is ics 25, 52-60.
[6] Bax e , M. and R. King (1993). Fiscal Policy in Gene al Equilib ium. Ame ican
Economic Re iew 83(3), 315-34.
[7] Be nanke, B. S., J. Boi in and P. Eliasz (2005). Measu ing Mone a y Policy: A
Fac o Augmen ed Au o eg essi e (FAVAR) App oach, The Qua e ly Jou nal o
Economics 120, 387-422.
[8] Blancha d, O.J. and R. Pe o i (2002). An Empi ical Cha ac e iza ion o he Dy-
namic E ec s o Changes in Go e nmen Spending and Taxes on Ou pu , The
Qua e ly Jou nal o Economics: 1329-1368.
[9] Bu nside, C., M. Eichenbaum, and J. D. Fishe (2004). Fiscal Shocks and Thei
Consequences. Jou nal o Economic Theo y 115, 89-117.
[10] Ca allo, M. (2005). Go e nmen Employmen Expendi u e and he E ec s o Fis-
cal Policy Shocks. Fede al Rese e Bank o Chicago Wo king Pape 2005-16.
[11] Chambe lain, G. (1983). Funds, ac o s, and di e si ica ion in a bi age p icing
models, Econome ica 51, 1281-1304.
[12] Chambe lain, G., and M. Ro hschild (1983). A bi age, ac o s uc u e and mean
a iance analysis in la ge asse ma ke s, Econome ica 51, 1305-1324.
[13] Conno , G., Ko ajczyk, R.A., 1988. Risk and e u n in an equilib ium APT. Ap-
plica ion o a new es me hodology. Jou nal o Financial Economics 21, 255-89.
23
[14] Edelbe g, W., M. Eichenbaum, and J. D. Fishe (1999). Unde s anding he E ec s
o a Shock o Go e nmen Pu chases. Re iew o Economic Dynamics 2 (1), 166-
206.
[15] Eichenbaum, M. and J. D. Fishe (2005). Fiscal Policy in he A e ma h o 9/11.
Jou nal o Money, C edi and Banking 37 (1), 1-22.
[16] Fa as, A. and I. Miho (2001). The E ec s o Fiscal Policy on Consump ion and
Employmen : Theo y and E idence. CEPR Discussion Pape No. 2760.
[17] Fe nndez-Villa e de J., J.F. Rubio-Ram ez, T.J. Sa gen and M.W. Wa son
(2007). ABCs (and Ds) o Unde s anding VARs. Ame ican Economic Re iew,
Ame ican Economic Associa ion, 97(3):1021-1026.
[18] Fo ni, M., L. Gambe i (2010) The dynamic e ec s o mone a y policy: A s uc-
u al ac o model app oach, Jou nal o Mone a y Economics, o hcoming.
[19] Fo ni, M., D. Giannone, M. Lippi and L. Reichlin (2009). Opening he Black
Box: S uc u al Fac o Models wi h La ge C oss-Sec ions, Econome ic Theo y
25, 1319-1347.
[20] Fo ni, M., M. Hallin, M. Lippi and L. Reichlin (2000). The gene alized dynamic
ac o model: iden i ica ion and es ima ion, The Re iew o Economics and S a is-
ics 82, 540-554.
[21] Fo ni, M., M. Hallin, M. Lippi and L. Reichlin (2005). The gene alized ac o
model: one-sided es ima ion and o ecas ing. Jou nal o he Ame ican S a is ical
Associa ion 100, 830-840.
[22] Fo ni, M. and M. Lippi (2001). The gene alized dynamic ac o model: ep esen-
a ion heo y, Econome ic Theo y 17, 1113-1141.
[23] Fo ni, M., Lippi, M. (2010). The un es ic ed dynamic ac o model. Rep esen a-
ion esul s. Fo hcoming in Jou nal o Econome ics.
[24] Fo ni, M. and L. Reichlin (1998). Le ’s ge eal: a ac o analy ical app oach o
disagg ega ed business cycle dynamics, Re iew o Economic S udies 65, 453-473.
[25] Gali, J., D. Lopez Salido, and J. Valles (2007). Unde s anding he E ec s o
Go e n- men Spending on Consump ion. Jou nal o he Eu opean Economic As-
socia ion 5, 227-270.
[26] Geweke, J. (1977). The dynamic ac o analysis o economic ime se ies, in D.J.
Aigne and A.S. Goldbe ge , Eds., La en Va iables in Socio-Economic Models,
No h Holland, Ams e dam.
24
[27] Hallin M. and R. Liska (2007). De e mining he numbe o ac o s in he gene al
dynamic ac o model, Jou nal o he Ame ican S a is ical Associa ion 102, 603-
617.
[28] I eland, P.N. (2004). A me hod o aking models o he da a, Jou nal o Economic
Dynamics and Con ol 28, 1205-1226.
[29] Leepe , E.M., Walke , T.B. and S.S. Yang (2008). Fiscal Fo esigh : Analy ics and
Econome ics, NBER Wo king Pape No. 14028.
[30] Leepe , E.M., Walke , T.B. (2009). In o ma ion Flows and News D i en Business
Cycles, mimeo Indiana Uni e si y
[31] Lippi, M. and L. Reichlin (1993). The Dynamic E ec s o Agg ega e Demand and
Supply Dis u bances: Commen , Ame ican Economic Re iew 83, 644-652.
[32] Lippi, M. and L. Reichlin (1994). VAR analysis, non undamen al ep esen a ion,
Blaschke ma ices, Jou nal o Econome ics 63, 307-325.
[33] Moun o d, A. and H. Uhlig (2002). Wha A e he E ec s o Fiscal Policy Shocks.
CEPR Discussion Pape No. 3380.
[34] Ona ski, A. (2009). Tes ing Hypo heses Abou he Numbe o Fac o s in La ge
Fac o Models, Econome ica 77, 1447-1479.
[35] Pappa, E. (2009). The e ec s o iscal shock on employmen and he eal wage,
In e na ional Economic Re iew 50:217-244.
[36] Pe o i, R. (2002). Es ima ing he E ec s o Fiscal Policy in OECD Coun ies.
CEPR Discussion Pape No. 3380.
[37] Pe o i, R. (2007). In Sea ch o he T ansmission Mechanism o Fiscal Policy.
NBER Mac oeconomics Annual.
[38] Ramey, V.A. (2009). Iden i ying Go e nmen Spending Shocks: I ’s All in he
Timing, NBER Wo king Pape s no. 15464.
[39] Ramey, V.A. and M. Shapi o (1998). Cos ly Capi al Realloca ion and he E ec s
o Go e nmen Spending. Ca negie Roches e Con e ence on Public Policy 48,
145-194.
[40] Ra n, M. O. and K. Me ens (2009). An icipa ion o Fiscal policy. Uni e si y o
Sou hamp on manusc ip .
25
Figu e 2: Impulse esponse unc ions.
32

Figu e 3: Impulse esponse unc ions.
33
Figu e 4: Robus ness: 13 ac o s (solid line), 10 ac o s (do ed line), 16 ac o s
(dashed line).
34