Macroeconomics, Mcgraw Hill Education, Chapter 3 and 5

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 42

Theneoclassicalgrowthframeworkandstylizedfactsaboutconvergenceand

divergence

ReferencesontheSolowmodel:
P.B.Srensen,H.J.Whitta.Jacobsen(2010),IntroducingAdvanced
Macroeconomics,McGrawHillEducation,chapter3and5

D.Romer,AdvancedMacroeconomics,McGrawHill,chapter2.

D.Acemoglu(2009),IntroductiontoModernEconomicGrowth,Princeton
UniversityPress,chapter2;verydetailed,useselectively.

ReferencesonSolowandFacts1and2:
D.Acemoglu(2009),IntroductiontoModernEconomicGrowth,Princeton
UniversityPress,chapter3.
P.B.Srensen,H.J.Whitta.Jacobsen(2010),IntroducingAdvanced
Macroeconomics,McGrawHillEducation,chapter5and6


1
st
approach: GrowthAccounting(Solow1957)

theproximatecauseofincomegrowthisfactoraccumulation

ProximatecausesofeconomicgrowthintheSolowmodel

Assume:competitivemarkets+equilibriumaxiomnodemandlimitstogrowth
Theproximatecausesofincomegrowtharethoseshapingthedinamicsoffactor
supply:
AccumulationofcapitalstockK(t):fallsinthedomainofeconomics
AccumulationoflabourforceL(t):fallsmainlyinthedomainofdemography
AccumulationoftechnologyA(t): fallsinthedomainofscienceandengineering
Howfarcaneconomicsreachintheexplanationoftheproximatecausesofgrowth?


DifferentiateY(t)=F(K(t),L(t),A(t))withrespecttotime

F
A
A/Y=elasticityofYwithrespecttoA
F
K
K/Y=elasticityofYwithrespecttoK
F
L
L/Y=elasticityofYwithrespecttoL

K
=capitalshareinY=F
K
K/Y=elasticityofYwithrespecttoK

L
=labourshareinY=F
L
L/Y=elasticityofYwithrespecttoL


x=contributionoftechnologicalprogress(growthofTotalFactorProductivityTFP)

Labouraugmentingtechnologicalprogress

elacityofYwithrespecttoA=elacityofYwithrespecttoL
F
A
A/Y=F
L
L/Y=
L
easytomeasure!!
A
L
A A A
F
x
A Y A

/
L
A
x
A


Mainproblemswiththisexercise
deviaonsfromperfectcompeon
mismeasurement

Reasonsformismeasurementarerelatedtotheheterogeneityoflaborandcapital
inputsandtochangesininputqualitythroughtime:


Earlygrowthaccountingresults:


2
nd
approach:
takeSolowmodelseriouslyandderivepredictionstobetestedagainstdata
Solowmodelpredictsconvergenceofoutputpercapitay(t)tosteadysatey*(t)
define: k=K/AL y=Y/L
A
g
A

y(t)=y*(t)+[y(t)y*(t)]
y*(t)=steadystate,orpermanent,componentofgrowthpath
[y(t)y*(t)]=convergence,ortemporary,component

Decompositionbetweentemporaryandpersistent
componentingrowthpathsisnottrivial:y*(t)isunknown



ExampleGDPpercapitainEU1519602000:boldlineEU15GDPpercapita

Toanalysecrosscountrydifferencesinthesteadystatecomponenty*(t)

here: y=percapitaincomeY/L
f(k)=F(K,AL)/AL=Y/AL
k=K/AL


NowassumeF()isCobbDouglas(forsimplicity)
Y=K

(AL)
(1)

K
*=isconstant!!

devidingbothsidesbyAL=(AL)

(AL)
(1)

GDPperunitofefficiency = Y(t)/[A(t)L(t)]=k(t)


GDPpercapita = y(t)=A(t)k(t)


Insteadystate: grosssaving=capitaldepreciation
sf(k*)=sk*

=(g+n+)k*
s=(g+n+)k*
(1)
(6.1)
k*=[s/(n+g+)]
1/(1)
(6.2)
(6.2)expressesthesteadystatek*asafunctionofthemodelparameters
Noticethatk*isconstant

y*(t)=A(t)(k*)

=A(t)[s/(n+g+)]
/(1)

logy*(t)=logA(t)+[

/(1)]log[s/(n+g+)]

logy*(t)=logA(t)+[

/(1)]log[s/(n+g+)]
ifneoclassicalassumptionshold =KshareinGDP0.3
ifgandareuniformacrosscountries

crosscountrydifferencesinsteadystate(longterm)percapitaGDPare
explainedby:
initialtechnologylevelA(t)
populationgrowthn
propensitytosaves
Nowconsiderthetemporaryandpersistentgrowthrates
componentstogether:
( ) ( )(1 )
y
t g n g
y

[logy(t)logy*(t)] (7)
(seeendfootnotes)

growthrateofpercapitaoutputy(t)onatransitionpathexplainedby:
1. persistentgrowthcomponentduetotechnologicalprogressg
2. temporaryortransitionalgrowthcomponentcausedby
convergence:thenarrowingoftheinitialdeviation
[logy(t)logy*(t)]fromsteadystatepath

Noticethatthehigher,theloweristhegrowthcomponent
explainedbyy(t)<y*(t) theslowerisconvergence


Nowconsidergrowthtrajectoriesratherthangrowthrates
( ) ( )(1 )
y
t g n g
y

[logy(t)logy*(t)] (7)

Whenconsideringtrajectories,thepositionofsteadystatepercapita
incomepathy*(t)matters

Otherthingsequal,positionofy*(t)pathisdeteminedbysavings
ratesandinitialtechnologylevelA(0)


PredictionsforcountriesA,Bstructurallyidentical:y
A
*(t)=y
B
*(t)
ast y
A
(t)y
A
*(t) y
B
(t)y
B
*(t)

fromy
A
*(t)=y
B
*(t): y
A
(t)>y
B
(t) g
y
A
(t)<g
y
B
(t)

thepersistentgrowthcomponentsareidenticalinAandB

thetemporarygrowthcomponentislowerintherichcountry
AthaninthepoorcountryB

growthisfasterinBthaninA=unconditionalconvergence

UnconditionalGDPpercapitaconvergenceinthe6originaryEU
countries(exception:Luxemburgafter1980)


GDPpercapitaconvergenceinotherEUlatecomers
lackofconvergenceinGDPpercapitainotherEUcountries


SlowconvergenceintheEastEUcountries?


theabovefiguressuggest:
1. thereareimportantexceptiontoconvergenceinGDPpercapita
evenincountriesatasimilarstageofdevelopment

2. Whereconvergencetakesplace,itisrelativelyslow..


( ) ( )(1 )
y
t g n g
y

[logy(t)logy*(t)] (7)

Speedofconvergenceismeasuredby = (1)(n + + g)
speedofconvergenceisdeterminedbyhowfastthe
marginalproductofcapitalfallswhenkincreases.
Convergenceisfasterifisclosertozeroandslowerifis
closerto1.

If=1weareintheY=AKmodel
MPK=A=constant
noconvergenceatall.
theeconomyisalwaysonabalancedgrowthpath

Calibrationyieldscounterfactualpredictions:


Barroregressions

b
0
=g+(n+g+)(1)logy*(t1)

Ifb
0
isnotcountryspecific,butisuniformacrosscountries

equaon(8)testsforunconditionalconvergence(absoluteconvergence)

Unconditional convergence:


BarroResults:averagepercapitaGDPgrowth19601985positivelycorrelated
withGDPpercapita1960,onacrosssectiondataof91countries


Conditionalconvergence


BacktotheSolowmodel:
( ) ( )(1 )
y
t g n g
y

[logy(t)logy*(t)] (7)
useadiscretetimeapproximation
g
t,t1
g(n+g+)(1)[logy(t1)logy*(t1)]
foruniformn,g,,thegrowthrateincreaseswiththedistance:
[logy*(t1)logy(t1)]
ifweallowforthefactthatcountriessteadystatesmaydiffer
therichcountrymaygrowfasterthanthepoor!

Modellingsteadystatedifferences
logy*(t1)=logA(t1)+[

/(1)]log[s/(n+g+)]
foruniformn,g,,steadystatesmaydifferacrosscountriesbecause:
thesavingsbehaviordiffers
thetechnologyleveldiffers

g
t,t1
g(n+g+)(1)[logy(t1)logy*(t1)]
g
t,t1
=b
0
i
b
1
logy
t1
+
i,t

b
0
i
capturestheeffectofallthevaraiblesaffectingthesteady
state,inparticularinitialtechnologyA(t1)ands


GDPpercapitagrowth19601985netofallexplanatoryvariablesofsteady
state1960versusGDPpercapita


Problem1.Persistentandunpersistenteffects


Problem2.InatheoreticalframeworkbroaderthantheSolowmodel:
somevariablesinX
i,t
mayaffectlongrungrowth,notjustconvergence
X
i,t
maybeendogenous:jointlydeterminedwithg
i,t,t1

Infact,regressionssimilarto(10)abovehavebeenusedtoestimate
thedeterminantsofeconomicgrowth,notjustconvergence

Estimatesofcoefficientvectorwouldthenprovideinformationabout
causaleffectsongrowthofthevariablesinvectorX

ButifvariablesinvectorXareeconometricallyendogenous
Causalinterpretationofisnotappropriate
Estimateofconvergencecoefficientb
1
isinconsistent,thatis,as
thedatapointsinthesampleincrease,theprobabilitythatthe
estimateofb
1
iscorrectdoesnotconvergeto1.


Problem3:
ConditionalconvergenceobtainsbutisfarslowerthanSolowsuggests.


Problem4: SomevariablesinX
i,t
arehardtomeasure

Byintroducingadhoclargedifferencesinlogy*
j,0
,theSolowmodel
canexplainhighlydivergingdynamicbehavior.

logy*
j,0
=logA
j,0
+[

/(1)]log[s
j,
/(n
j
+g+)]
s
j
andn
j
areeasytomeasure,buttechnolgyA
j
isnot!

Footnote(optional!!)
Step1. y*(t)=A(t)f(k*) andtakelogs

Step2.
considerkfunctionoflogk: k=e
logk
k/(logk)=k
TakeTaylorexpansionof logyasfunctionoflogkaroundlogk*:

(6.2)
logk(t)logk*[f(k*)]
1
[logy(t)logy*(t)] (6.2.bis)

Step3.
(k/t)=sf(k)(n++g)k
fromfirstorderexpansionoflogkaroundk*: logklogk*+(1/k)(kk*)

(6.3)
Note:tosubstituteforsuse(6.1)


Step4. Use(6.2)togetherwithlogy*(t)=logA(t)+logf(k*)towrite:

f(k*)[logk(t)logk*]

1. toderivethepreviousresult,substiuteforthegrowthrateofkfrom6.3
2. nowsubstituteforlogk(t)logk*from6.2.bis
[logy(t)logy*(t)]

You might also like