Invemen-pecific Technology Shock, Neural Technology Shock and he Dunlop-Tarhi Obervaion: Theory and Evidence Moren O. Ravn, European Univeriy Iniue and he CEPR Saverio Simonelli, European Univeriy Iniue and Univeriy of Naple
INTRODUCTION The Dunlop-Tarhi obervaion i key for buine cycle reearch near orhogonaliy beween hour worked and real wage, and beween hour worked and aggregae labor produciviy a he buine cycle frequencie picure ofen een a a limu e of reaonable buine cycle heorie appear conrary o much of he echnology driven buine cycle lieraure
INTRODUCTION I i a puzzle for RBC yle heorie and Keyneian yle heorie becaue: RBC yle heorie rely on an aggregae produciviy hock: hif labor demand and herefore give rie o poiive hour-produciviy (- real wage) comovemen Keyneian yle heorie: conradic he imple icky wage ory which would imply negaive real wage comovemen
INTRODUCTION I ha been he concern of much reearch in he recen buine cycle lieraure mix of labor demand and labor upply hock (Chriiano and Eichenbaum, 992, Braun, 994, McGraan, 994): Fical impule indiviible labor (Hanen, 986): High labor upply elaiciy home-producion heorie (Benhabib, Rogeron and Wrigh, 992): modified labor upply repone o echnology hock
THIS PAPER Three aim:. Look a condiional correlaion rucure: How do hour and produciviy comove condiional on hock? Neural permanen echnology hock Invemen-pecific echnology hock 2. Look a ecoral apec: Conumpion ecor v. invemen ecor 3. Conra condiional and ecoral reul wih economic heory
EMPIRICS We inveigae US quarerly daa 960-2003 ample We ue a rucural VAR approach o idenify wo ype of echnology hock: neural permanen echnology hock invemen-pecific permanen echnology hock We hen examine he impac of hee idenified hock on aggregae and ecor level variable
SVAR We eimae he following VAR: X X ( L) = k + B X + e [ i n n n n = Δp, Δa, h, c y, i y ]' p i : he log of he invemen o conumpion price a : he log of aggregae labor produciviy h : he log of hour worked c n -y n : he log of nominal conumpion expendiure o nominal oupu i n -y n : he log of nominal invemen o nominal oupu k: conan and rend
SVAR The wo hock are idenified auming. Only permanen invemen-pecific echnology hock can affec long-run level of relaive invemen price 2. Only permanen invemen-pecific echnology hock and permanen neural echnology hock can affec long-run level of aggregae labor produciviy β X = κ + 0 βi X i + ε P i= Eimaed uing Shapiro-Waon 2SLS + riangular 2SLS eimaion procedure
SVAR Having eimaed he wo hock, we hen eimae heir impac on ecoral variable from: ~ h ~ α γ h + μ P P h = n + βi y i + i= i= h i i h denoe derended hour worked in ecor conumpion ecor (non-durable) invemen ecor (durable) y denoe he vecor of idenified hock
RESULTS
RESULTS
THE HOURS-PRODUCTIVITY RELATIONSHIP Aggregae Conumpion Invemen in quaniie Invemen wih relaive price adjumen
MOMENTS Secor Aggregae: level: Conumpion Large negaive ecor: correlaion Negaive comovemen condiional upon invemen-pecific hock. Invemen ecor: Large poiive correlaion condiional upon quaniie: poiive comovemen neural hock wih price adjumen: a in he aggregae
EVIDENCE In ummary: The Dunlop-Tarhi obervaion hold uncondiionally in he aggregae bu: i doe no hold uncondiionally a he ecor level i doe no hold condiionally on neural and invemen-pecific echnology hock yemaic relaionhip in he aggregae yemaic relaionhip a he ecor level Implicaion: Theory hould no decouple hour and produciviy
THEORY We examine buine cycle verion of Greenwood, Hercowiz, and Kruell (997) wo-ecor economy: conumpion good producing ecor good are non-orable invemen good ecor good canno be conumed boh ecor are compeiive neural and invemen pecific echnology hock co of adjumen relaed o variaion in capial ock and in hour worked variable capaciy uilizaion
PREFERENCES Houehold are aumed o be infiniely lived, have raional expecaion, and heir preference are given a: ( ) ( ) = + + + = 0,, 0 0 / i c n n C E V κ σ σ κ φ σ β /σ i he ineremporal elaiciy of ubiuion /κ i he invere of he aggregae labor upply elaiciy i a growh facor ha i included o guaranee he exience of a balanced growh pah
TECHNOLOGIES The producion echnologie are given a: C I = = A z A z 2 ( ) α ( ) c u K h x c, ( ) α ( ) i i u K h i, z : Neural echnology hock ha affec boh ecor imulaneouly c, x : Invemen-pecific echnology hock ha affec he invemen ecor only Invemen good canno be conumed and conumpion good canno be inveed: I = Ic, + Ii, i, c, i, α c α
ADJUSTMENT COSTS We aume ha i i coly o vary capial and labor inpu: ( ) ( ) ( ) ( ) ( ) K K I G I K u K n n n F h,,,,,,,,,,, / / + Λ = = + δ Where F and G are aumed o be uch ha here are no adjumen co along he balanced growh pah Thee co are needed o limi he exen o which facor of producion can inananeouly be reallocaed acro ecor he model would be counerfacual wihou uch co of adjumen
OUTPUT AND TECHNOLOGY GROWTH Aggregae oupu and he echnology procee are: ( ) ( ) ( ) ( ) x x z z x x z z x x x x z z z z PI C Y ε γ ε γ ρ ρ ρ ρ exp / exp / 2 2 = = + = Growh in echnology lead o growh in: oupu and conumpion invemen and capial ock relaive invemen price
MODEL VS. DATA In order o aure ha only invemen-pecific hock have permanen effec on relaive invemen price we aume: α c = α and i ε c ε i However: We ill do no know I he model conien wih he dynamic impac of neural and invemen-pecific echnology hock on aggregae variable? I he model conien wih he dynamic effec of echnology hock on ecor level variable?
STRUCTURAL ESTIMATION In order o evaluae he model, we need o paramerize i: Θ : Parameer ha we calibrae Θ 2 : Parameer ha we eimae The eimaion i done by limied informaion approach: Θ ( daa heory( )) daa heory IR IR Θ Θ ' W IR IR ( Θ ) ( ) 2 = minθ 2 2 2 arg 2 Θ IR daa : The empirical eimae of he impac of echnology hock IR heory : The impac of he hock in he model given he parameer W: A weighing marix 2
PARAMETERS
Indiviible labor High adjumen co in invemen ecor Higher perience of neural hock Invemenpecific hock much more volaile
THE MODEL VS. DATA
MODEL VS. DATA
MODEL VS. AGGREGATE DATA The model doe a grea job of accouning for mo of he aggregae dynamic: very precie eimae of he impac of he wo echnology hock on oupu conumpion invemen hour worked lighly wore in erm of neural echnology hock on labor produciviy
MODEL VS. DATA Invemen-pecific hock Wih price adjumen Precie eimae of he impac of invemen-pecific hock
MODEL VS. DATA Neural echnology hock Wih price adjumen Here he fi i wore in erm of impac on conumpion ecor
DUNLOP-TARSHIS OBSERVATION
WHAT S MISSING? The model provide a moive for reallocaion of labor bu: doe no inroduce furher ecor pecificiie uch a kill difference acro ecor durable wage around 5-20 % higher han conumpion ecor (and ecor premium i procyclical) in boom: killed labor flow from conumpion o invemen ecor lef for fuure reearch
CONCLUSIONS We have hown:. While hour and produciviy are nearly orhogonal a he buine cycle frequencie, he condiional correlaion rucure doe no confirm near orhogonaliy Neural hock: Poiive comovemen Invemen-pecific hock: Negaive comovemen 2. Syemaic difference acro ecor Poiive produciviy-hour comovemen in invemen ecor Negaive comovemen in conumpion ecor
CONCLUSIONS 3. Economic heory can accoun for aggregae evidence very well. 4. Sill work o do in erm of accouning fully for he ecoral evidence bu heory doe beer han expeced!
Dunlop-Tarhi
Dunlop-Tarhi RETURN