Productivity, Taxes, and Hours Worked in Spain:

Transcription

1 Federal Reserve Bank of Minneapolis Research Deparmen Saff Repor 550 July 2017 Produciviy, Taxes, and Hours Worked in Spain: Juan C. Conesa Sony Brook Universiy Timohy J. Kehoe Universiy of Minnesoa, Federal Reserve Bank of Minneapolis, and Naional Bureau of Economic Research ABSTRACT In he early 1970s, hours worked per working-age person in Spain were higher han in he Unied Saes. Saring in 1975, however, hours worked in Spain fell by 40 percen. We find ha 80 percen of he decline in hours worked can be accouned for by he evoluion of axes in an oherwise sandard neoclassical growh model. Alhough axes play a crucial role, we canno argue ha axes drive all of he movemens in hours worked. In paricular, he model underpredics he large decrease in hours in and he large increase in hours in The lack of produciviy growh in Spain during has lile impac on he model s predicion for hours worked. Keywords: dynamic general equilibrium; hours worked; disorionary axes; oal facor produciviy JEL Codes: C68, E13, E24, H31 * This work has benefied from ousanding research assisance by Parisa Kamali. The auhors can be reached via a juan.conesa@sonybrook.edu and kehoe@umn.edu. The daa and he explanaion of all he consruced daa used in his projec can be accessed online a hps://sies.google.com/sie/jcconesa/research and a hp://users.econ.umn.edu/~kehoe/. Earlier versions of his paper were circulaed beginning in Augus The views expressed herein are hose of he auhors and no necessarily hose of he Federal Reserve Bank of Minneapolis or he Federal Reserve Sysem.

2 1. Inroducion Wha forces have driven oupu growh and flucuaions in Spain over he las hree decades? Wha has been he impac of he evoluion of axes and oal facor produciviy (TFP) on aggregae hours worked and oupu? We sudy hese quesions hrough he lens of growh accouning and he neoclassical growh model. Our resuls show ha he evoluion of aggregae hours worked in Spain has been consisen wih he evoluion of axes, whereas lack of TFP growh has had a minor impac on hours worked. A shorcoming of our model is ha i fails o accoun for abou 20 percen of he movemens in hours worked in booms and recessions. The model, however, does accoun for he declining rend in hours worked over he period The mehodology used is ha inroduced in Kehoe and Presco (2002), following he mehodology proposed in Cole and Ohanian (1999) in heir sudy of he U.S. Grea Depression. See Conesa e al. (2007) for an exposiion of his mehodology and an explanaion of how o exend i o differen model environmens. As he firs sep in sudying growh and hours worked in Spain over he period , we use growh accouning o quanify he conribuion of TFP, capial deepening, and aggregae hours worked for he dynamics of oupu per working-age person. Nex, we consruc a sandard neoclassical growh model in which a sand-in household chooses hours worked, consumpion, and capial holdings, aking as given he deerminisic evoluion of working-age populaion, TFP, and ax raes. This mehodology provides us wih a quaniaive ool for idenifying he relevan margins for poenial candidae explanaions for changes in such variables as hours worked. A sriking feaure of he Spanish growh experience is he lack of TFP growh since Our exercise is silen abou he reasons behind his observaion. Diaz and Franjo (2016) argue ha excessive invesmen in srucures raher han in capial equipmen accouns for much of he sagnaion of TFP. Garcia-Sanana e al. (2016) argue ha misallocaion of resources because of crony capialism is responsible for his feaure. Regardless of he reason for his sagnaion, however, our quaniaive exercise indicaes ha he lack of TFP growh has had only a minor impac on he evoluion of aggregae hours worked. Prior o 1975, TFP and oupu per working-age person moved ogeher in Spain. (Because of daa availabiliy, we define working age o be ages 15 64; our resuls are no 1

3 sensiive o minor changes in his definiion.) Afer 1975, however, his has no been he case. The reason is ha 1975 marked he beginning of a rend of decreasing aggregae hours worked. This rend of decreasing hours worked sharply conrass wih he U.S. experience, where hours worked per working-age person have been roughly consan. Over he period , hours worked per working-age person in Spain, 23.7 hours per week, were higher han hose in he Unied Saes, 23.4 hours per week. Spain provides an exreme example of a general rend in European labor marke dynamics. In France, for example, hours worked per working-age person have been sysemaically falling since he 1960s, alhough he decline in France has no been as seep as in Spain. The differences in he labor marke experiences in he Unied Saes and Europe have been exensively sudied. Mos of he lieraure in his research area has focused on he impac of differences in labor marke insiuions. Benolila and Berola (1990), Blanchard and Summers (1986), and Alesina e al. (2005), among ohers, focus on he role of insiuions and labor marke resricions. Ljungqvis and Sargen (1998) focus on he ineracion beween shocks and insiuions. Presco (2004), however, argues ha differenial axaion alone can accoun for he differences in he curren level of aggregae hours worked beween Europe and he Unied Saes. Ohanian e al. (2008) provide a comparison across counries ha are members of he Organisaion for Economic Co-operaion and Developmen (OECD) and focus mosly on he correlaion beween disorionary wedges on labor supply and hours worked. While we work wih a heoreical framework ha is similar o ha of Ohanian e al. (2008), we pay more aenion o he deails of he aggregae growh process for he Spanish economy. In paricular, our calibraion is specific o he Spanish case (insead of maching some OECD average), and we compare model oucomes o daa along all he relevan dimensions a every poin in ime o idenify he specific episodes in which here are deparures beween daa and model predicions. Besides he fac ha we work wih a longer ime horizon, i is hard o compare our resuls o hose in Ohanian e al. (2008) since hey do no repor a specific comparison of model oucomes and daa for he case of Spain. Our analysis shows ha he evoluion of he axaion of consumpion and facor earnings can accoun for he secular rend decrease in hours worked observed in Spain. Of course, our exercise is silen abou he disribuion of aggregae hours worked wihin he 2

4 working-age populaion. The quaniaive exercise ha we perform allows us o idenify years in which daa deviae from heory in a quaniaively imporan way. We wan o idenify hese episodes because hey sugges avenues for fuure research. In paricular, he comparison beween he model oucomes and he daa reveals ha in periods of rapid changes in aggregae hours worked, he model sysemaically underesimaes he magniude of such changes. In paricular, he model fails o accoun for much of he large decrease in hours worked during and for much of he large increase in hours worked during Our model also predics ha hours worked and oupu sar o fall afer 2009, whereas he recession had already sared in Hours fall in he model because boh consumpion axes and labor income axes increase. The discrepancy beween model oucomes and he daa is consisen wih he criique of our model ha i predics smooher movemens in hours han hose observed in he daa. The observaion ha Spanish labor markes reac more o shocks han in oher counries is well esablished in he lieraure. Benolila e al. (2012) aribue he differenial labor marke response beween Spain and France during he recession o he dualiy of Spanish labor marke insiuions. In paricular, a large fracion of employmen in Spain is covered by emporary conracs wih very low hiring and firing coss. For a more deailed reamen of he naure and consequences of he Grea Recession in Spain, see Jimeno and Sanos (2014) and he oher papers in he special issue of SERIEs on he Grea Recession in Spain. 2. Growh accouning exercise Cole and Ohanian (1999) and Kehoe and Presco (2002) have developed he grea depressions mehodology o deec and analyze large deviaions from balanced growh behavior. They use economic heory in he form of he neoclassical growh model o guide heir view of economic daa. The model feaures an aggregae producion funcion of he Cobb-Douglas form, Y AK L, (1) where Y is oupu, A is oal facor produciviy, K is capial, and L is labor inpu. If he populaion grows a a consan rae, N N 0, and TFP grows a a consan rae, 3

5 A 1 A, hen he economy has a balanced growh pah in which all quaniies per 0 working-age person grow by he facor, excep hours worked per working-age person, which are consan. Kehoe and Presco (2002) rewrie he producion funcion (1) as /(1 ) 1/(1 ) A Y K L. (2) N Y N This decomposiion is useful because, along a balanced growh pah, where A grows a a consan rae, he capial-oupu raio, K / Y, and hours worked per working-age person, L / N, are consan. 225 Figure 1: Growh accouning for Spain oupu per working age person produciviy facor index 1970 = capial facor 75 labor facor Figure 1 depics he decomposiion of he growh of oupu per working-age person Y / N in Spain over he period ino hree facors: he produciviy facor , he capial facor K / Y /(1 ) 1/(1 ) A, and he labor facor L / N. Noice ha afer 1975, Spain 4

6 has experienced large and persisen deviaions from a balanced growh pah since he growh in GDP per working-age person differs from ha in he produciviy facor, and he capial and labor facors are no consan. To consruc Figure 1, we assemble daa on annual GDP and invesmen from naional accouns. Real invesmen is curren-price invesmen deflaed by he GDP deflaor. Conesa e al. (2007) explain why using curren-price invesmen deflaed by he invesmen deflaor is only appropriae in a wo-secor model in which he relaive price of invesmen o consumpion varies. We calculae he sock of capial by cumulaing invesmen: K 1 (1 ) K I, (3) where he depreciaion rae is calibraed so ha he average value of he raio of capial consumpion o GDP, K / Y, maches is average value in he daa. We need o choose an iniial value for he capial sock in equaion (3). A simple way o do his is o equae he capial-oupu raio in he iniial period o he average for he firs decade of daa, K Y K. (4) Y Noice ha, if we have many years of daa before he period in which we are ineresed, he choice of he iniial value of capial would make lile difference because of he compounding of depreciaion in he cumulaion of invesmen (3). For he growh accouning repored in Figure 1, we have daa on invesmen and GDP saring in We calibrae We also use naional income accouns o calibrae he capial share and obain See he online daa appendix for deails. Saring in 1975, he Spanish economy diverges sharply from a balanced growh pah. Oupu per working-age person Y / N grows more slowly (if a all) han does he produciviy facor 1/(1 ) A because hours worked per working-age person L / N fall sharply even hough here is a sharp increase in he capial facor K / Y /(1 ). Beween 1974 and 1986, hours per working-age person fall by abou 40 percen. Over he nex 10 years, hours flucuae, and in 1994 hey begin o rise unil hey sar o fall again in

7 3. Theoreical framework We follow Cole and Ohanian (1999) and Kehoe and Presco (2002) in using a simple general equilibrium growh model o mach heory wih daa and o idenify episodes where model predicions deviae from he daa. The model has a sand-in household ha has uiliy for consumpion and leisure and ha makes decisions on he rade-off beween labor and leisure and ha beween consumpion and invesmen. There is a one-secor producion echnology given by he producion funcion (1), and he economy is closed o foreign rade. The governmen ransfers ax revenue (in excess of governmen consumpion) o he household as a lump sum so as o balance he budge every period. The household has perfec foresigh over he evoluion of all he exogenous variables: populaion, axes, governmen consumpion, lump-sum ransfers, and TFP. The household s maximizaion problem is max [ log C (1 )log( N h L )] 1970 s.. (1 ) C K K (1 ) wl (1 )( r ) K T c k 1 (5) K K Here N denoes he number of people in he working-age populaion in he economy, and h denoes he yearly disposable ime endowmen of each individual. The fiscal variables ha he household secor akes as given are he axes on consumpion, on labor income, and on c k ne capial income,,, and he ransfers T. The choice variables are sequences of aggregae consumpion levels worked L. C, aggregae capial socks In each period, he resource consrain is The governmen budge consrain is 1 K, and levels of aggregae hours C K 1 (1 ) K G AK L. (6) c k C wl ( r ) K G T. (7) The familiar condiions for cos minimizaion and zero profis are 6

8 w (1 ) AK L (8) and r AK L. (9) 1 1 Because he axes on consumpion bu no hose on labor or capial are indirec axes ha ener ino he marke prices ha he household pays for consumpion, we mus modify our definiion of GDP in he model, measuring i as saisical agencies do: Y AK L C. (10) 1 c This modificaion implies ha he produciviy erm in he model, A in equaion (10), differs from oal facor produciviy in he growh accouning exercise in he previous secion. In paricular, we use a measure of TFP of GDP a facor prices: A Y C, (11) c 1 K L whereas he TFP measure in he growh accouning exercise is TFP of GDP a marke prices. I is worh noing ha he concep of GDP a facor prices ha we use in he model, Y C, is no precisely he measure of GDP a facor prices, or GDP a facor cos, found c in Spain s naional income saisics. The reasons are wofold. Firs, our measure fails o subrac ou of GDP a marke prices all indirec axes, in paricular ariffs on impors and indirec axes on purchases of inermediae goods ha were prevalen before Spain s 1986 value added ax reform. Second, since we are using a closed economy model, our measure of consumpion includes he rade balance. To be precise, we could refer o he produciviy erm A in equaion (11) as TFP of GDP a facor prices, bu we refer o i simply as TFP in he res of his paper wih he warning ha i differs from he TFP in he growh accouning secion, which is TFP of GDP a marke prices. Our benchmark specificaion implies ha all ax proceeds in excess of governmen consumpion are rebaed o he household in lump sum. There are wo reasons for doing so: firs, because we are measuring marginal ax raes insead of average ax raes, a lump-sum rebae makes our axes progressive, and second, because a very large fracion of he 7

9 governmen s budge is used o finance ransfers direcly. In fac, looking a he composiion of governmen expendiure, we find ha he sum of pensions, healh care, unemploymen insurance, and educaion in Spain amouns o more han half of he governmen oulays. Our benchmark specificaion implies ha all of governmen consumpion, as opposed o governmen ransfers, is a wase, or equivalenly i finances some public good ha eners uiliy in a separable form. Noice ha his specificaion is no neural in deermining he implicaion of axes for hours worked. Assuming insead ha all governmen revenues are rebaed o he household ha is, making G 0 in he governmen budge consrain would imply ha here is no income effec of axaion, making hours worked fall even more when households are faced wih higher axes. I is worh poining ou ha we can ignore governmen deb because he model exhibis Ricardian equivalence wih respec o he iming of ransfers. In paricular, he exisence of governmen deb would only change he iming of ransfers, leaving all oher variables unaffeced. Of course, he model does no exhibi complee Ricardian equivalence since axes are disorionary, bu his only means ha he iming of axes maers, no he iming of borrowing, lending, and ransfers. 4. Evoluion of fiscal variables We obain esimaes of effecive marginal ax raes using a mehodology similar o ha of Mendoza e al. (1994). Our mehodology differs from heirs in wo respecs: Firs, we aribue a fracion of households nonwage income o labor income o be consisen wih our esimae of he capial share. Second, we accoun for ax progressiviy by adjusing average income axes by he raio of he marginal income ax o he average income ax from an esimaed ax funcion (following Gouveia and Srauss (1994)). The esimaed ax for he Spanish economy in he 1990s suggess ha he marginal income ax is 1.8 imes he average income ax, and his is he adjusmen facor ha we use. See he online daa appendix for deails on he use of his adjusmen facor. Our esimaes show a subsanial increase in axaion of boh he income of facors of producion and consumpion. Regarding he axaion of capial income, our esimaes generae a very large spike in his axaion. The reason for his large spike is ha during he boom prior o he recession, ax revenues from corporae income axes grew faser han ne 8

10 capial income in he naional accouns, while here was a large drop in corporae ax revenues wih he recession. Corporae ax revenues as a share of ne capial income grew from 7 percen in 1994 o 29 percen in Figure 2: Evoluion of effecive marginal ax raes 0.5 ax on capial 0.4 ax on labor ax on consumpion Esimaing effecive marginal ax raes requires ha we ake ino accoun he progressiviy of income axes. Taxaion of labor earnings can be decomposed ino income axaion and payroll axes (which are mosly social securiy conribuions ha are roughly proporional). Since he relevan daa for our model in erms of he disorionary implicaions are he marginal, and no he average, ax raes, we scale up by 83 percen he axaion of households income in order o compue effecive marginal ax raes. This adjusmen is consan over ime and only affecs he levels of ax raes, no is evoluion over ime, which is wha maers in our model. Figure 2 repors our esimaes of marginal ax raes. The axes in our model do no accoun for all of he axes in he Spanish economy. In addiion o income axes on labor income and capial income and indirec axes on consumpion, here are ariffs on impors and indirec axes ne of subsidies on firms. A 9

11 simple way in which we could accoun for hese addiional axes would be o include hem in he axes on consumpion: Y Y. (12) F c C Here F Y is GDP a facor prices from he naional accouns, which nes ou of GDP a marke prices no only indirec axes on consumpion bu also ariffs on impors and indirec axes paid by firms. In his case, o compue TFP of GDP a facor prices as in equaion (11), we replace Y C wih c F Y. Recalibraing he model in his manner has very lile impac on our resuls since more han 80 percen of all indirec axes in Spain over he period were axes on consumpion, and he evoluion of all indirec axes followed ha of consumpion axes. 25 Figure 3: Evoluion of governmen consumpion 20 Percen GDP

12 Figure 3 repors he evoluion of governmen consumpion, showing how i has doubled is size as a share of GDP, wih mos of he increase occurring beween 1974 and Calibraion and numerical experimens In he numerical experimens ha we perform, our model deermines he evoluion of he endogenous variables, given he iniial capial sock in 1970 as measured in he daa. The endogenous variables reac o he evoluion of he exogenous variables, which are he evoluion of TFP of GDP a facor prices, he evoluion of he working-age populaion, and he evoluion of ax raes as measured in he daa. To run our numerical experimens, we need o make assumpions abou he evoluion of exogenous variables afer We assume ha TFP grows a he rae a which i grew in he years immediaely prior o he recession, Figure 1 shows ha TFP of GDP a marke prices did no grow from 1995 unil 2015, and he behavior of TFP a facor prices in equaion (11) is almos idenical. Consequenly, we are assuming no TFP growh afer We assume ha he working-age populaion is consan afer 2015, even hough projecions by he Unied Naions indicae ha he working-age populaion in Spain will decline. Finally, we assume ha ax raes say consan a heir 2015 level. None of our resuls seem o depend criically on hese assumpions, since mos of he model predicions are driven by he saic labor supply condiion. Given our assumpions on he evoluion of he exogenous variables, we calculae he equilibrium up o he erminal dae of 2044 assuming ha he equilibrium has converged o a seady sae by ha dae. Conesa e al. (2007) provide a deailed explanaion of he compuaional procedure and a MATLAB program for implemening i. Nex, we need o assign values o all he parameers in he model. To deermine he value of he disposable ime endowmen of individuals, h, we assume ha each adul has a ime endowmen of 100 hours a week. We choose he depreciaion rae so ha he raio of capial consumpion o GDP coincides wih he average value observed in he daa over he period Calculaing 2015 K 0.139, (13) Y we obain We esimae he capial share in Spain o be using he same 11

13 naional accouning daa as hose used for he growh accouning exercise and for he esimaion of he marginal ax raes. To calibrae he preference parameers, we use he firs-order condiions from he household problem (5) and daa observaions for he period Deriving he firsorder condiions wih respec o preference parameers as funcions of daa observaions: C and L and rearranging, we can wrie he values of he (1 ) C (1 ) (1 (1 )( )) c 1 1 c k C r (14) c (1 ) C. (15) c (1 ) C (1 ) w ( N h L ) Using condiions (14) and (15), we can compue annual observaions for he parameers and. The parameers we assign o our economy are he averages over he period , and Noice ha we have calibraed he echnology parameers he depreciaion rae and he labor share using daa from he enire period , whereas we have calibraed he behavioral parameers he discoun facor and he consumpion share using only daa for Changing he period used for calibraing he echnology parameers has negligible effecs on our resuls, whereas changing he period for calibraing he behavioral parameers has significan effecs on he resuls. Our benchmark calibraion follows Bergoeing e al. (2002) in choosing behavioral parameers o mach household behavior in a period in which we are no ineresed, in his case , and we es he model by seeing how well i maches behavior in he period in which we are ineresed, in his case We find ha our resuls for real GDP per working-age person and he capial-oupu raio improve significanly when we recalibrae he behavioral parameers o mach behavior over bu ha our resuls on hours worked per working-age person do no change in any significan way. 6. Resuls for he benchmark economy and he counerfacual wih consan axes Given ha our goal is o quanify he implicaions of he evoluion of axes on hours worked, our principal exercise is o compare our benchmark model resuls wih a counerfacual in 12

14 which we hold ax raes and governmen consumpion consan a heir iniial levels. We require ha governmen consumpion grow by a facor ha is a produc of he growh facor of populaion and he growh facor of produciviy, so ha a balanced growh pah would be possible. In paricular, in he counerfacual we hold he ax raes consan a heir average raes over and governmen consumpion consan a is average level over he same period. Alernaively, we could have held governmen consumpion as a fracion of GDP consan wihou changing he resuls in any imporan way. Figure 4 shows he evoluion of GDP per working-age person in he daa, in he benchmark model, and in he counerfacual model wih consan axes and consan governmen consumpion. Noice ha he benchmark model does reasonably well unil he lae 1980s, bu i fails o predic he increase in oupu aferward. The model wih consan axes does beer in predicing he increase in oupu over he whole period bu overpredics he increase in he 1980s and 1990s. Figure 4: GDP per working-age person in benchmark model 40,000 35,000 daa 30,000 consan ax model 2010 euros 25,000 20,000 benchmark model 15,000 10, Figures 5 and 6 show he evoluion of he capial-oupu raio and hours worked per working-age person. In Figure 5, we see ha he benchmark model does a poor job in 13

15 capuring he increase in he capial-oupu raio. 4.5 Figure 5: Capial-oupu raio in benchmark model 4.0 daa 3.5 raio consan ax model benchmark model In Figure 6, he daa on hours worked show larger flucuaions relaive o he predicions of he benchmark model, boh when hours reach heir minima in 1985 and 1994 and when hey reach heir maximum in Noice, however, ha he benchmark model does a far beer job in capuring he decline in hours worked from 1974 o 2015 han he model wih consan axes: In he daa, hours worked per working-age person fall by 21.0 percen over his period, whereas in he benchmark model, hey fall by 27.7 percen. In conras, in he model wih consan axes, hours worked per working-age person fall by only 11.7 percen. 14

16 Figure 6: Hours worked per working-age person in benchmark model consan ax model hours per week benchmark model 16 daa We see ha wo reasons explain why he model wih consan axes does a beer job in capuring he average growh in GDP per working-age person over he period han does he benchmark model: Firs, he model wih consan axes does a beer job in capuring he capial deepening ha sars in he 1980s. Second, he model wih consan axes accouns for more oupu growh because i does no accoun for he drop in hours worked. The firs reason in he model wih consan axes is compaible wih he daa bu leaves unanswered he quesion of why he capial-oupu raio increased as capial income axes were increasing. Overall, he benchmark model he model wih increasing axes and governmen consumpion does a beer job in accouning for he movemens in hours worked in Spain han does he model wih consan axes and governmen consumpion. I fails o accoun for economic growh because i fails o accoun for he capial deepening ha occurred in Spain saring in he 1980s. 15

17 7. Sensiiviy o recalibraion of behavioral parameers Our mehodology allows us o idenify he reasons for he failure of he benchmark model o accoun for is poor performance in racking he growh of real GDP per working-age person in Spain in Figure 4. We use equaions (14) and (15) o recalibrae he behavioral parameers and using daa for he enire period We obain a discoun facor of which implies subsanially less discouning han he discoun facor calibraed o daa, and a consumpion share of which implies a shif in uiliy o favoring more leisure compared o he consumpion share calibraed o daa, Figure 7: GDP per working-age person wih recalibraed parameers 40,000 35,000 daa 2010 euros 30,000 25,000 20,000 model wih recalibraed parameers benchmark model 15,000 10, When we redo he numerical experimens in he previous secion wih he behavioral parameers recalibraed o daa, we find ha he model wih changing axes and governmen consumpion performs far beer han does he model calibraed o daa. Figure 7 shows he improved performance of he model in capuring he growh of real GDP per working-age person. Noice, however, ha he model fails o capure he boom in 16

18 This is worh more sudy. Figure 8 shows he improved performance of he model in capuring he capial deepening ha occurred in Spain from he 1980s onward. Figure 9 shows ha he model s resuls for hours worked change very lile. Figure 8: Capial-oupu raio wih recalibraed parameers raio model wih recalibraed parameers daa 2.5 benchmark model

19 Figure 9: Hours worked per working-age person wih recalibraed parameers model wih recalibraed parameers 22 daa hours per week benchmark model Kehoe and Presco (2002) regard he behavioral parameers and as capuring no only he preferences of households bu also, in a reduced form way, insiuions and policies in he Spanish economy ha affec invesmen markes in he case of and labor markes in he case of. Following heir approach, we inerpre he large change in as incorporaing changes in insiuions and policies affecing invesmen markes in Spain. The resuls in Figures 4 and 5 sugges ha a large par of hese changes in insiuions and policies occurred in he 1980s. Obvious candidaes for hese changes are he 1986 inegraion of Spain ino wha was hen he European Communiy and he accompanying financial liberalizaion. Kehoe and Presco s (2002) mehodology like he wedges mehodologies of Mulligan (2005) and Chari e al. (2007) allow us o idenify where and when imporan policy changes have occurred and poin o direcions in which more general equilibrium modeling is needed. The large change in suggess he need for an open economy model ha is capable of modeling he large capial inflows ha followed Spain s inegraion ino he European Communiy. One such modeling exercise is ha of Fernandez de Cordoba and Kehoe (2000). I is worh poining ou ha, as Figure 1 shows, capial deepening in Spain acually 18

20 sared in I is possible ha some of he increase in can be ascribed o changes in insiuions and policies ha followed he change in poliical regime iniiaed wih he deah of he dicaor Francisco Franco. We could also look for changes in insiuions and policies ha affec labor markes o accoun for he drop in he consumpion share, bu he change in does no have a major impac on he resuls in Figures 7 9. Given he success of he model wih recalibraed parameers in capuring he increase in oupu and he capial deepening in Spain, we employ he sraegy of calibraing he behavioral parameers and using daa for he enire period In erms of he mehodology of Kehoe and Presco (2002), his means ha we are implicily incorporaing changes in insiuions and policies ha affeced invesmen markes and, o a lesser exen, labor markes, over he period Sensiiviy o he choice of labor supply elasiciy Labor economiss ypically measure he responsiveness of labor supply o variaion in aferax wages using he Frisch elasiciy, he elasiciy of labor wih respec o he afer-ax wage keeping he marginal uiliy of income consan. In our benchmark economy, wih a uiliy funcion ha is separable in consumpion and leisure in he sand-in household s problem (5), his elasiciy is u Lu Nh L L. (16) In he Spanish daa ha we sudy, he average number of hours worked per working-age person is 18.6 hours per week, which leaves 81.4 hours per week for leisure. This implies ha he Frisch elasiciy of labor supply in our model is abou 4.3, which is a large value compared o many esimaes in he lieraure. Considerable conroversy surrounds he issue of wha is he bes specificaion of preferences for leisure and wha is he mos plausible labor supply elasiciy. For example, Alesina e al. (2005) argue ha Presco s (2004) resuls depend on an implausibly high elasiciy of labor supply relaive o microeconomeric esimaes based on he variaion of working hours wih respec o he wages of prime-age employed males. Their criicism could also be applied o our exercise. 19

21 Noice, however, ha in boh Presco s work and ours, he sand-in household makes decisions for he aggregae economy. Therefore, he soluion o he household s problem should embed he paricipaion decisions of individuals. For example, young workers decide when o ener he labor marke, workers decide when o move ino and ou of he labor marke in response o changes in household s circumsances, old workers decide when o reire. Incorporaing his sor of decision which deermines wha is usually referred o as labor supply a he exensive margin ino he behavior of he sand-in household requires behavior ha need no replicae ha of he average prime-age employed male. Heckman (1993) argues ha he elasiciy of paricipaion decisions is large, and, in fac, mos of he movemen in aggregae hours worked is due o variaions on he exensive margin. In fac, if we wan o model flucuaions of hours worked in he aggregae economy, he elasiciy of labor supply should be esimaed using aggregae daa and no individual daa. To he exen ha hours worked in he model are less volaile ha hose in he daa, his suggess ha our implied labor supply elasiciy may be smaller han i would be if we esimaed his elasiciy using he Spanish aggregae daa ha we are considering for he ime period of ineres. (This may no be he case, however, if workers are no able o deermine heir own hours worked because of involunary unemploymen or oher sors of rigidiies.) Even wihin he conex of esimaion using microeconomic daa, several papers show how he esimaes of labor supply elasiciies based on he variaion of he working hours of prime-age employed males are biased downward. I is well known ha women have higher labor supply elasiciy han men. Moreover, Imai and Keane (2004) argue ha in a framework wih endogenous human capial accumulaion, in paricular wih learning by doing, he labor supply elasiciy is Domeij and Flodén (2006) have shown boh heoreically and empirically ha he presence of borrowing consrains biases he esimaed individual labor supply elasiciies downward. Theoreical work focusing on indivisible labor, such as Hansen (1985) and Rogerson (1988), has shown ha small (in heir papers, zero) individual elasiciies can aggregae o a big number (in heir papers, infinie) depending on he pariculars of he heoreical environmen and he risk-sharing arrangemens available o people. Esimaes of he labor supply elasiciy ha ake ino accoun he enire populaion and boh he inensive and exensive margin generae numbers more in line wih he implied elasiciies using aggregae 20

22 daa. For example, Erosa e al. (2016) esimae a value of In a life-cycle conex, Rogerson and Wallenius (2009) find ha axes have large effecs on aggregae labor supply, even if he labor supply elasiciy a he individual level is as low as i is in microeconomeric sudies. As a sensiiviy analysis, we explore separable nonhomoheic uiliy o deermine how sensible our resuls are o changes in he Frisch labor supply elasiciy. Consider he uiliy funcion 1970 c C 1 Nh L 1 1 c N N. (17) This sor of uiliy funcion has been used for microeconomeric esimaion using individual daa, as in Heckman (1993) and Browning e al. (1999), and for wealh disribuion analysis, as in Casañeda e al. (2003). Our specificaion of uiliy in he sand-in household s problem (5) is he limiing case of he uiliy funcion (17) where boh c 0 and 0 By varying, we can vary he Frisch elasiciy:. u Lu 1 Nh L. (18) 1 L By seing 4.768, we obain a Frisch elasiciy of labor supply of abou 0.75, which is he sor of value favored by Chey e al. (2011). The value of c conrols wheher hours worked rise or fall when income is increasing. We choose a value of c 0.5, which implies ha hours worked fall as income increases. King e al. (1988) show ha, unless c 0, he uiliy funcion (17) is no consisen wih he exisence of a balanced growh pah. Since our specificaion has no produciviy growh afer 2015, however, we can ignore his poenial problem. Numerical experimens under his alernaive specificaion of he uiliy funcion show ha he evoluion of axes sill maers. We calibrae he behavioral parameers of he model o daa as in he previous secion and obain and Noice ha, since we have changed he funcional form of he sand-in household s uiliy funcion, hese parameers are no easily comparable o hose used in he previous secions. 21

23 Figures repor our resuls for he case wih low labor supply elasiciy for boh he case wih axes evolving as in Figure 2 and wih governmen consumpion evolving as in Figure 3, and he counerfacual wih consan axes and governmen consumpion. Figure 10: GDP per working-age person wih low labor supply elasiciy 40,000 35,000 consan ax model 30,000 ax model 2010 euros 25,000 20,000 daa 15,000 10,

24 5.0 Figure 11: Capial-oupu raio wih low labor supply elasiciy consan ax model raio 3.5 ax model 3.0 daa Figure 12: Hours worked in he model wih low labor supply elasiciy consan ax model hours per week daa ax model

25 The resuls in Figure 12 show ha he model wih increasing axes performs beer in explaining hours worked han he model wih consan axes. In paricular, while hours worked per working-age person fall by 21.0 percen from 1974 o 2015 in he daa, in he model wih increasing axes, hey fall by 17.5 percen. In conras, in he model wih consan axes, hours worked per working-age person fall by only 6.8 percen. Alhough he performance of he model wih low labor supply elasiciy in capuring he fall in hours worked in Spain in Figure 12 is no as impressive as he resuls for he model wih a higher labor supply elasiciy, repored in Figures 6 and 9, he resuls sill show he imporance of increasing axes in accouning for Spanish labor marke oucomes. 9. The role of he lack of produciviy growh A sriking feaure of he growh accouning for Spain repored in Figure 1 is ha produciviy has been fla since To undersand he impac of he lack of produciviy growh on labor marke oucomes, we ask he model wha would have happened if TFP afer 1994 had grown a he same average rae as i had grown from 1970 o 1994: 1.22 percen per year. Again, we calibrae he behavioral parameers of he model o he daa from he enire period , as in Secion 7. We refer o he model wih TFP calibraed o he daa as he calibraed model raher han as he benchmark model because he benchmark model in Secion 6 has behavioral parameers calibraed o daa from 1970 o The resuls repored in Figures show ha he lack of produciviy growh over he period is very imporan for undersanding he evoluion of GDP per workingage person and he capial-oupu raio, bu ha is impac on hours worked is small. 24

26 Figure 13: GDP per working-age person wih high TFP growh 40,000 35,000 high produciviy growh model 30,000 calibraed model 2010 euros 25,000 20,000 daa 15,000 10, Figure 14: Capial-oupu raio wih high TFP growh raio calibraed model daa 2.5 high produciviy growh model

27 Figure 15: Hours worked in he model wih high TFP growh hours per week daa calibraed model high produciviy growh model The resuls in Figure 15 sugges ha here was lile relaion beween he changes in produciviy and changes in hours worked in Spain over our period of ineres, Noneheless, he iming of changes in produciviy and changes in hours suggess he possibiliy of an endogenous relaion beween he wo. In paricular, noice ha, in he growh accouning in Figure 1, produciviy sagnaes when hours worked boom during he period I is possible ha he labor marke reforms ha allowed greaer use of emporary conracs and encouraged his boom in hours also encouraged hiring less producive workers, which led o he drop in produciviy. This possibiliy is worh exploring furher. 10. Concluding remarks In his paper, we have argued ha he rend of aggregae hours worked in Spain is consisen wih neoclassical growh heory given he observed rends of axes and governmen consumpion. The impac of increases in axes on hours worked is dampened if he labor supply elasiciy is subsanially lower han ha ypically used in macroeconomic research bu is no eliminaed. 26

28 Our model underpredics he magniude of large flucuaions in hours worked. Recen research suggess ha he insiuional arrangemens of he Spanish labor marke imply a larger reacion of hours worked or employmen o shocks. See Benolila e al. (2012) for an analysis of he differenial response of France and Spain o he recen recession, and Jimeno and Sanos (2014) for a comprehensive analysis of he Spanish experience during he las recession in In conras o he evoluion of axes, he lack of TFP growh in Spain since 1994 does no generae a subsanial impac on he evoluion of hours worked in our model. 27

29 References Alesina, A. F., E. L. Glaeser, and B. Sacerdoe (2005), Work and leisure in he U.S. and Europe: Why so differen? in M. Gerler and K. Rogoff (eds.), NBER Macroeconomics Annual 2005, Vol. 20 (Cambridge, MIT Press), Benolila, S., and G. Berola (1990), Firing coss and labour demand: How bad is Eurosclerosis, Review of Economic Sudies, 57, Benolila, S., P. Cahuc, J. J. Dolado, and T. Le Barbanchon (2012), Two-ier labour markes in he Grea Recession: France versus Spain, Economic Journal, 122, F155 F187. Bergoeing, R., P. J. Kehoe, T. J. Kehoe, and R. Soo (2002), A decade los and found: Mexico and Chile in he 1980s, Review of Economic Dynamics, 5 (2002), Blanchard, O. J., and L. H. Summers (1986), Hyseresis and he European unemploymen problem, in S. Fischer (ed.), NBER Macroeconomics Annual 1986, Vol. 1 (Cambridge, MIT Press), Browning, M., L. P. Hansen, and J. J. Heckman (1999), Micro daa and general equilibrium models, in J. B. Taylor and M. Woodford (eds.), Handbook of Macroeconomics, Vol. 1, Par A (Amserdam: Elsevier), Casañeda, A., J. Díaz-Giménez, and J. V. Ríos-Rull (2003), Accouning for he U.S. earnings and wealh inequaliy, Journal of Poliical Economy, 111, Chari, V. V., P. J. Kehoe, and E. R. McGraan (2007), Business cycle accouning, Economerica, 75, Chey, R., A. Guren, D. Manoli, and A. Weber (2011), Are micro and macro labor supply elasiciies consisen? A review of evidence on he inensive and exensive margins, American Economic Review, 101, Cole, H. L., and L. E. Ohanian (1999), The Grea Depression in he Unied Saes from a neoclassical perspecive, Federal Reserve Bank of Minneapolis Quarerly Review, 23, Conesa, J. C., T. J. Kehoe, and K. J. Ruhl (2007), Modeling grea depressions: The depression in Finland in he 1990s, Federal Reserve Bank of Minneapolis Quarerly Review, 31, Diaz, A., and L. Franjo (2016), Capial goods, measured TFP and growh: The case of Spain, European Economic Review, 83, Domeij, D., and M. Flodén (2006), The labor-supply elasiciy and borrowing consrains: Why esimaes are biased, Review of Economic Dynamics, 9, Erosa, A., L. Fuser, and G. Kambourov (2016), Towards a micro-founded heory of aggregae labor supply, Review of Economic Sudies, 83, Fernandez de Cordoba, G., and T. J. Kehoe (2000), Capial flows and real exchange rae flucuaions following Spain's enry ino he European Communiy, Journal of Inernaional Economics, 51, Garcia-Sanana, M., E. Moral-Benio, J. Pijoan-Mas, and R. Ramos (2016), Growing like Spain: , CEPR Discussion Paper

30 Gouveia, M., and R. P. Srauss (1994), Effecive federal individual income ax funcions: An exploraory empirical analysis, Naional Tax Journal, 47, Hansen, G. D. (1985), Indivisible labor and he business cycle, Journal of Moneary Economics, 16, Heckman, J. J. (1993), Wha has been learned abou labor supply in he pas weny years? American Economic Review, 83, Imai, S., and M. P. Keane (2004), Ineremporal labor supply and human capial accumulaion, Inernaional Economic Review, 45, Jimeno, J. F., and T. Sanos (2014), The crisis of he Spanish economy, SERIEs-Journal of he Spanish Economic Associaion, 5, Kehoe, T. J., and E. C. Presco (2002), Grea depressions of he 20h cenury, Review of Economic Dynamics, 5, King, R. G., C. I. Plosser, and S. T. Rebelo (1988), Producion, growh and business cycles: I. The basic neoclassical model, Journal of Moneary Economics, 21, Ljungqvis, L., and T. J. Sargen (1998), The European unemploymen dilemma, Journal of Poliical Economy, 106, Mendoza, E. G., A. Razin, and L. L. Tesar (1994), Effecive ax raes in macroeconomics: Cross-counry esimaes of ax raes on facor incomes and consumpion, Journal of Moneary Economics, 34, Mulligan, C. B. (2005), Public policies as specificaion errors, Review of Economic Dynamics, 8, Ohanian, L., A. Raffo, and R. Rogerson (2008), Long-erm changes in labor supply and axes: Evidence from OECD counries, , Journal of Moneary Economics, 55, Presco, E. C. (2004), Why do Americans work so much more han Europeans? Federal Reserve Bank of Minneapolis Quarerly Review, 28, Rogerson, R. (1988), Indivisible labor, loeries and equilibrium, Journal of Moneary Economics, 21, Rogerson, R., and J. Wallenius (2009), Micro and macro elasiciies in a life cycle model wih axes, Journal of Economic Theory, 144,