Economic Integration and the Non-tradable Sector: the European Experience

Transcription

1 Economic Inegraion and he Non-radable Secor: he European Experience Sophie Pion Work in progress. Preliminary version as of May 21, Absrac: Since he inroducion of he Euro, macroeconomic imbalances widened among Member Saes. This divergence ook he form of srong differences in he dynamics of he non-radable secors beween he core and he periphery. Adaping a model of srucural change for a small open economy wih a radable and a non-radable secor, his aricle shows ha economic inegraion resuls in a relaive expansion of he non-radable secor in oal employmen. To do so, he model revisis he radiional Balassa-Samuelson and Baumol s effecs o incorporae he effec of financial inegraion a collapse in he ineres rae on he dynamics of he non-radable secor. Using a novel daa se for 12 counries of he Euro area, his aricle hen documens he expansion of he non-radable secor over in he Euro area periphery (+4.8p.p.) significan even when excluding he housing secor from he sample (+2.8p.p.). This expansion happened simulaneoulsy o (i) faser produciviy growh in he radable secor han in he non-radable secor (ii) declining long-erm nominal ineres raes. A quanificaion of hese effecs shows ha, in Porugal over , economic inegraion can explain up o 90% of he expansion of he non-radable secor. This broad effec of economic inegraion accouns for much more han he sole Balassa-Samuelson effec (abou 13%). Key words: srucural change, non-radable secor, unbalanced growh, macroeconomic imbalances, Euro area. Paris School of Economics Universiy Paris 1 Panhéon Sorbonne & CEPII. sophie.pion@ps .eu. I am exremely graeful o my advisor Agnès Bénassy-Quéré for her coninuous guidance and suppor. I am indebed o Michel Agliea, Jean Imbs, Lise Paureau, Richard Pores, Ricardo Reis, Fabien Tripier and paricipans in he Dynamics, Economic Growh and Inernaional Trade DEGIT XXI conference, he Royal Economic Sociey junior symposium 2015, he Spring 2015 CESifo-Delphi Conference, he 2015 ECB Forum on Cenral Banking and he 2015 EEA Congress, as well as seminars a Paris School of Economics and he OECD for heir useful discussions and commens.

2 1. Inroducion Greece, Ireland, Porugal and Spain he so-called "periphery" have accumulaed large curren accoun deficis since he Euro s incepion. Firs inerpreed as good imbalances, curren accoun deficis were supposed o reflec a cach-up and convergence process of he poores counries of he area. 1 The single currency was expeced o make balance of paymens irrelevan beween he member saes. 2 This view was called ino quesion in he afermah of he recession and he idea ha curren accouns deficis refleced a convergence process was challenged by boh economiss and policymakers. Debaes emerged o reassess he mechanisms behind he accumulaion of curren accoun deficis in he Euro area periphery. They focused on he observaion ha counries which accumulaed he larges deficis were counries wih low aggregae TFP growh. This aricle focuses on he naure of hese imbalances and heir origins. More specifically, i asks wheher economic inegraion hrough he single marke bu also hrough moneary and financial inegraion could have fosered uneven growh raes across differen secors depending on heir exposure o inernaional rade. Economic inegraion has had wo main dimensions (Blanchard and Giavazzi, 2002): financial and moneary inegraion, and radable marke inegraion (he single European marke). Tradable marke inegraion led o fas produciviy growh in he radable secor of he periphery. Financial and moneary inegraion resuled in he convergence of nominal ineres raes among he Eurozone counries, hence o a seep decline in he risk premia of counries in he periphery. Exending he baseline muli-secor model of Ngai and Pissarides (2007) o a small open economy composed of a radable secor and a non-radable secor, I show ha marke and financial inegraion conribues grealy o he expansion of he share of he nonradable secor in oal employmen. This reallocaion of ressources ino he non-radable secor he secor wih he lowes TFP growh reduces aggregae TFP growh. Two mechanisms are a play o explain he effec of economic inegraion on he share of he non-radable secor in oal employmen: a relaive (non-radable o radable) price effec revisied Balassa-Samuelson effec, and he fac ha consumpion grows faser han oupu unbalanced growh effec. Faser produciviy growh in he radable han in he non-radable secor leads o a relaive price increase (Balassa-Samuelson effec). Similarly, a collapse in he cos of capial leads o a relaive price increase as i benefis less he labor-inensive non-radable secor (Acemoglu and Guerrieri, 2008). As long as here is a small (below one) elasiciy of subsiuion beween raded and non-raded goods, boh effecs lead o he expansion of he share of employmen in he non-radable secor (Baumol s effec). 3 Financial inegraion also fosers a demand boom, i.e. emporary unbalanced growh. Tradable goods can be impored, bu non-radable goods mus be produced domesically: i resuls in an increase in he share of employmen of he non-radable secor, and an accumulaion of curren accoun deficis. Using a novel daa se for 12 counries of he Euro area, I hen documen he dynamics of he non-radable secor over The share of he non-radable secor in employmen rose seeply in he periphery of he Euro area over (+4.8p.p.). During he same period, his share remained sable in he so-called core counries. 4 The increase in peripheral counries is significan even if he housing secor is 1 In heir seminal aricle of 2002, Blanchard and Giavazzi showed ha financial inegraion and lower ineres raes along wih goods markes inegraion would lead boh o a decrease in saving and an increase in invesmen in poorer counries, and so, o large curren accoun deficis. Deficis would be reduced as counries would converge. 2 Ingram poined ou in 1973 ha "he radiional concep of a defici or a surplus in a member naion s balance of paymens becomes blurred " (Ingram, 1973, p.15). 3 Baumol (1967) suggess ha fas produciviy growh in manufacuring aciviies fuels an increase in wages. This cos increase canno be offse in services aciviies since i faces slower produciviy growh. I hus leads o a relaive (service o manufacuring) price increase. As long as he relaive oupu of service and manufacuring aciviies are mainained, an ever increasing proporion of he labor force mus be channeled ino hese aciviies and he rae of growh of he economy mus be slowed correspondingly. 4 The periphery includes he four counries of he EA12 (counries which adoped he euro in 2001 and before) wih he lowes GDP per capia (a purchasing power sandards) in I includes: Greece, Ireland, Porugal, Spain. Core counries are: Ausria, Belgium, Finland, France, Germany, Ialy, Luxembourg, he Neherlands. Discussion on he composiion of he radable 2

3 excluded from he sample. This expansion happened simulaneoulsy o (i) faser produciviy growh in he radable secor han in he non-radable secor, (ii) declining long-erm ineres raes. Finally, an illusraive calibraion of he model is underaken o invesigae wheher he dynamics generaed by he model are broadly consisen wih he paerns in European daa. A simple growh accouning exercise conludes ha, in Porugal for example, he sandard Balassa-Samuelson effec accouns for 13% of he increase in he share of he non-radable secor in oal employmen over However, when reassessing is broader effecs boh he revisied Balassa-Samuelson and unbalanced growh effecs, economic inegraion can explain up o 90% of he expansion of he non-radable secor. The conribuion of his aricle if hreefold. Firs, i proposes a heoreical analysis of he effecs of a collapse in he ineres rae and differen TFP growh raes across secors on he dynamics of he nonradable secor. Second, i builds a new daabase o analyze he dynamics of he non-radable secor and he differen dimensions of economic inegraion for 12 counries of he Euro area. Third, i quanifies he conribuions of economic inegraion on he dynamics of he non-radable secor for he 12 counries of he Euro area over This aricles relaes o previous analyses of defecive growh paerns. Paerns of defecive growh have already been examined for he US by Hlashwayo and Spence (2014). The auhors argue ha hree main elemens explain he American defecive growh paern prior o he 2007 financial crisis: an ousized domesic demand, accommodaive capial inflows, and he lack of srucural flexibiliy and resource mobiliy o accompany echnological and srucural change. This resuled in a shif of facors of producion o he non-radable secor of he economy, crowding ou he US radable secor, dampening is scope and compeiiveness. The impac of EMU on caching-up economies can relae o hese hree elemens: he compression of bond spreads in he Euro area periphery following moneary inegraion fosered demand booms, moneary and financial inegraion fueled subsanial increase in privae leverage in he peripheral counries, he single marke inegraion aced as a caalys for srucural change. In Europe also i resuled in a relaive expansion of he non-radable secor (Giavazzi and Spavena, 2011). These paerns are all he more defecive ha hey are good predicors of financial crises. Gourinchas and Obsfeld (2012) show ha domesic credi expansion and real currency appreciaion are he mos robus and significan predicors of financial crises, regardless of wheher a counry is emerging or advanced. Kalanzis (2015) shows how in a small open economy he deepening of financial openness resuling in capial inflows, followed by an expansion in he relaive size of he nonradable secor, increases he financial fragiliy of he economy. They are wo differen approaches of he mechanisms hrough which economic inegraion can affec he dynamics of he non-radable secor. The firs one exends he sandard Balassa-Samuelson effec. This effec is a play o explain he shif of facors of producion o he non-radable secor (Gregorio e al., 1994). However, Esrada e al. (2013) sugges ha produciviy growh in he radable secor canno be he sole explanaion of he dynamics of he relaive price in he periphery. The Balassa-Samuelson framework have been exended o include differences in labor and produc-marke regulaions (in he non-radable secors paricularly 5 ) across counries. Differences in regulaions could also have conribued o mainain persisen inflaion differenials across counries (Bénassy-Quéré and Coulibaly, 2014), and could have been a driver of he iner-secoral misallocaion of facors of producion (Epifani and Gancia, 2011). These exensions of he Balassa-Samulson effec documen an increasing relaive price of non-radables, and hereby could explain he relaive expansion of he non-radables secor. In a sandard model of he echnological explanaion of srucural change, ressources reallocae o he secor wih he fases growing relaive price (Baumol, 1967; Ngai and Pissarides, 2007). and non-radable secor is presened in Secion 3. The radable secor includes he manufacuring, mining and agriculural aciviies, as well as six service secors for which a large par of he oupu is inernaionally raded. 5 In a Speech given a he Annual Hyman P. Minsky Conference on April 10, 2014, Peer Prae Member of he Execuive Board of he ECB already saed ha he incomplee marke inegraion in goods and services, and a general lack of compeiive processes in he non-radable secor, allowed some firms in so-called caching-up economies o exrac excessive rens and disor capial allocaion. 3

4 A second view focuses raher on he impac of moneary inegraion on he expansion of he non-radable secor. Financial inegraion is modeled hrough a real ineres rae decrease 6 and a subsequen capial inflow in he European periphery. This lower cos of capial fueled a demand boom and he subsequen expansion of he non-radable secor (Fagan and Gaspar, 2007; Benigno and Fornaro, 2014), and more specifically an increase in house prices (Ferrero, 2015), causing a degradaion of curren accoun deficis (Geerolf and Grjebine, 2013). Financial fricion could also explain he disored allocaion of capial inflows following financial inegraion, in favor of he non-radable secor (Reis, 2013; Gopinah e al., 2015), or once again more specifically in favor of he housing secor (Adam e al., 2012). This aricle depars from previous analyses in hree ways: i synhesizes he effecs of he differen dimensions of economic inegraion on resource allocaion in a model of srucural change; i documens paerns of defecive growh and he differen dimensions of economic inegraion for 12 counries of he Euro area; i quanifies he conribuion of economic inegraion o he building-up of imbalances. The remainder of he aricle is organized as follows. Secion 2 develops he heoreical framework ha is able o invesigae he impac of economic inegraion on he dynamics of he non-radable secor in a small open economy. Secion 3 presens novel daa on he dynamics of non-radable secors and he differen dimensions of economic inegraion in he Euro area since Secion 3 quanifies he conribuion of economic inegraion on he dynamics of he share of he non-radable secor in oal employmen. Secion 4 concludes. 2. A wo-secor small open economy model This secion presens a model o invesigae he impac of economic inegraion on he dynamics of he non-radable secor in a small open economy. I is assumed ha his economy is par of a group of counries rading goods and asses among hemselves. For convenience, his group of counries is referred o as he world. Appendix 1 conains proofs of he main conclusions Se-up The model exends he baseline muli-secor model of Ngai and Pissarides (2007) o a small open economy. The wo secors considered in he economy here are no he manufacuring and services secors, bu he radable secor (T) and he non-radable secor (N). Two reasons moivae his choice. Firs, analyzing he secoral dynamics in erms of radable versus non-radable secors allows o derive implicaions for he dynamics of expors and impors and hereby for he curren accoun. As is oulined in Blanchard (2007), he dynamics of large imbalances imply significan iner-secoral shifs in economic aciviy: during a defici phase, he non-radable secor expands and he radable secor shrinks in relaive erms; conversely, curren accoun rebalancing requires a relaive conracion of he nonradable secor and he expansion of he radable secor. Second, analyzing he radable versus non-radable secor allows us o disinguish secors depending on heir exposure o inernaional compeiion. A small economy is a price-aker in he radable secor. On he opposie, non-radable aciviies face only domesic compeiion. Tradiionally, economiss use he shorcu of labeling he indusry as radable and services as non-radable. Analyzing he dynamics of he radable versus non-radable secors would hen be equivalen o analyzing indusry versus service secors. However, he share of services in oal world rade is increasing seeply, and especially in he Euro area. In Greece, services represened more han 50% of he value of oal expors in Moreover, recen sudies have shown he recen serviizaion of he economies, i.e. he fac ha he divide beween manufacuring and service aciviies is becoming more and more blurry (Bernard and For, 2015). 6 See Hale and Obsfeld (2016) for a discussion on he effec of moneary inegraion he suppression of bond yields in he European periphery up o

5 By analogy o Ngai and Pissarides (2007), srucural change hereafer hus refers o a change in he share in oal employmen of he non-radable secor. We assume ha non-radable goods can only be consumed domesically, whereas radable goods can be consumed, invesed or raded. The radable good is used as he numeraire. There are wo inpus for producion: labor and capial. Boh are perfecly mobile across secors. Labor is no mobile across counries: he labor force is exogenous and grows a he rae ν. Conversly, capial is mobile and he counry can borrow or lend unlimied amouns on he inernaional capial marke. As in Blanchard and Giavazzi (2002), he nominal rae of ineres is given exogenously and depends on he world ineres rae r and a wedge x : R = (1 + r)(1 + x ). This wedge x could reflec a spread semming from he counry s borrowing cos premium due o he currency risk or oher ypes of uncerainy (uncerainy regarding financial regulaions, or credi risk for example). This wedge falls as economies inegrae. Toal financial wealh is composed of domesic capial K minus he level of foreign deb F. The represenaive household The economy is inhabied by a represenaive household who derives uiliy V a ime from he discouned sum of fuure consumpion: V = [β(1 + ν)] s ln(c s ) s= where β ]0, 1[ is he discoun facor, and c s 0 is consumpion per capia a ime s. This represenaive household works, borrows on foreign markes and owns domesic firms. The budge consrain, expressed in erms of radables and per uni of labor, is: p c = ω + d + f +1 (R ν)f (1) where c is aggregae consumpion per capia and p he consumer price index in erms of he radable good. We have p c = c T + p N c N wih c T he consumpion of radables and c N of non-radables, p N is he relaive price of non-radables. The represenaive household receives he wage ω and dividends from he firms he owns d (for simpliciy he represenaive household owns all firms in he domesic economy and here is no foreign direc invesmen in he model 7 ). Borrowing and lending ake place via one-period foreign bonds. Le f be he per capia value of he bonds borrowed a he end of he period 1 a he exogenous ineres rae R (a negaive f means a posiive asse holding). R f mus be reimbursed a he end of period, possibly by borrowing f +1. Aggregae consumpion is a CES funcion of he consumpion of boh goods: c = [γ 1 T θ c θ 1 θ + (1 γ) 1 N θ c θ 1 θ ] θ θ 1 Wih γ [0, 1] he share of he non-radable good, and θ he elasiciy of subsiuion beween he wo goods. The consumpion price index p is a funcion of he relaive price of he non-raded goods p N : p = [γ + (1 γ)(p N ) (1 θ) ] 1 1 θ (2) Sandard firs order condiions yield he inra-emporal allocaion of real consumpion: c T c N = γ 1 γ (pn ) θ (3) 7 For simpliciy, here is no FDI in he model. Blanchard and Giavazzi (2002) show, however, ha invesmen ouflows o oher EU counries amoun o only 15 percen of oal ouflows. 5

6 and he iner-emporal Euler equaion: c +1 c = β(1 + r)(1 + x +1 ) p p +1 (4) Proposiion 1 : he growh rae of consumpion is a posiive funcion of he wedge x +1. The higher he wedge, he more impaien is he counry. The counry reaches he world seady sae only when x has converged o zero. Firms In each secor, here is a represenaive firm indexed by j = T, N. Firms use homogeneous capial K and labor L, and we have: n T + n N = 1; k T n T + k N n N = k (5) where n j is he share of secor j in oal employmen, k he aggregae capial-o-labor raio, and k j he capial-labor raio in secor j. Producion funcions are Cobb-Douglas: = A j (K) j αj (L j ) (1 αj ) wih α j ]0, 1[ he capial inensiy of secor j, and A j he secor-specific TFP. This producion funcion can be wrien in unis per labor: y j = A j n(k j ) j αj. Y j Firms are equiy-financed and seek o maximize he presen discouned value of dividends. Dividend (expressed in erms of radables) in each period equals revenues ne of wages and capial expendiures: D j = py j j ω L j q I j where q is he price of invesmen goods and I j represens gross invesmen. If he firm has marke power, hen price p j depends on is choice of oupu: p(y j j ). 8 Wih perfec foresigh, he firms programme a ime is: max p j s= R 1,s (p j sy j s ω s L j s q s I j s) where R,s = (1 + r) s s τ= (1 + x τ) (1 + x ) R,s is he discoun facor. 9 The firm s programme is subjec o iniial capial K j 0, he producion funcion, and he consrain ha capial inpu depends on invesmen and depreciaion δ. 10 The user cos of capial a ime (he same in boh secors, U ) is a funcion of he price of invesmen goods, he ineres rae and he depreciaion rae: U = q 1 (1 + r)(1 + x ) q (1 δ) = q 1 [(R 1) + δ (1 + ˆq 1 ) ˆq 1 ] (6) Wih ẑ he growh rae of variable z. Since he radable price is he numeraire, firs order condiions in he radable secor yield he equaion for he wage: ω = [ A T ] 1 U αt µ T (1 α T ) 1 αt (α T ) αt 1 α T (7) 8 This assumpion depars from he basic Balassa-Samuelson se-up since firms in each secor have marke power o fix heir prices. In he radiional Balassa-Samuelson framework, he radable price follows he law of one price. One would need his assumpion o compare price levels across counries. Here he focus is raher on differences in price and employmen dynamics, and hence here is no need o make any assumpion on he level of he radable price. 9 We have R T = 1 and R+1 T = (1 + r)(1 + x +1) = R +1. If x = x is consan, hen R,s reduces o R s = [(1+r)(1+x)] s. 10 We have K j +1 = Ij + (1 δ)kj where Ij is oal invesmen in secor j a he end of period, and Kj is capial inpu a he begining of ime. 6

7 Wages are a decreasing funcion of he user cos of capial U (and hereby a decreasing funcion of he spread x ), an increasing funcion of radable produciviy A T and a decreasing funcion of a markup µ T. 11 The equaion for he relaive price of he non-radable good, which depends only on echnological condiions, is: p N = (AT /µ T ) 1 αn 1 α T (A N /µ N ) α N α T [(1 α T ) 1 αt (α T ) αt ] 1 αn 1 α U T 1 α T (8) (1 α N ) 1 αn (α N ) αn 2.2. Economic inegraion and he dynamics of he non-radable secor This secion sudies implicaions of rade and financial oppenness on srucural change. I assume ha he non-radable secor is more labor-inensive han he radable secor: α N < α T. This assumpion will be discussed in he empirical secion (Secion 3). Proposiion 2 : The relaive price of non-radable goods increases (ˆp N > 0 ) if : (1) Balassa-Samuelson effec: produciviy (real facor paymens) grows faser in he radable han in he non-radable secor; (2) Financial inegraion: he user cos of capial decreases. Proof: Rewriing equaions 8, we ge he growh rae of p N : ( ) ( 1 α ˆp N N α = 1 α T â T â N N α T + 1 α }{{} T Balassa-Samuelson effec ) Û } {{ } effec of financial inegraion where a j = Aj, wih j = N, T, is produciviy, or, in he case where firms have marke power wih µ j µ j 1, real facor paymens 12. Given ha 0 < α N < α T < 1, we ge a posiive impac of ( ) â T â N and a negaive impac of Û on ˆp N. Changes in he relaive price reflecs he ypical Balassa-Samuelson effec, i.e. a posiive link beween faser produciviy growh in he radable secor and he relaive price of he non-radable good. This effec sems from he fac ha produciviy growh in he radable secor leads o a wage increase, which ensures ha he marginal cos of radables remains consan. However, i increases he marginal cos, and hence he relaive price of he non-radable good he more so ha he non-radable secor is labor-inensive. In urn, he impac of a fall in he user cos of capial on he relaive price of non-radables depends on he capial inensiy of he non-radable relaively o he radable secor (α N α T ). Indeed, a fall in he ineres rae is mached by a wage increase ensuring ha he marginal cos of radables remains consan. If he non-radable secor is relaively more labor inensive, his rise in wages will increase is marginal cos, and hence he relaive price, of he non-radable good: because he non-radable secor is relaively more labor inensive, his rise in wages will no be compensaed by he fall in he ineres rae in his secor. Considering ha rade inegraion involves upward convergence in he produciviy of he radable secor and ha financial and moneary inegraion involves a downward convergence of he ineres rae (fall in he wedge x ), he relaive price of he non-radable good increases hrough he wo channels menionned in Proposiion 2. ( 11 Wih he markup µ j (1 = + ) ( p j /Y j )) 1. This markup derives from he case where firms have a marke power, p j / Y j hen firms se heir price as a markup over maginal coss. We hen ge, as in Fernald and Neiman (2011), ha value added in each secor can be decomposed ino he labor and capial shares in cos, and a profi share. In ha case, measures of TFP can diverge from rue echnology growh A j if hey do no accoun for he profi share. See model Appendix for a discussion of his bias. 12 In he case where firms make profis (µ j 1), and hese profis evolve over ime, we have Âj ˆµj = (1 αj )(ˆω ˆp j ) + α j (Û ˆp j ). If here are no profi, hen produciviy equals real facor paymens. 7

8 To recover he share of he non-radable secor in oal employmen, we combine he firs-order condiions in he radable and non-radable secor and he consrain ha all non-radable oupu mus be consumed in each period. Le us denoe by n N he share of he non-radable secor in oal employmen, and by ñ N he following funcion of n N : n N /s L,N where s L,j = 1 αj µ j ñ N = n N /s L,N + n T /s L,T j {T, N} is he secoral share of labor in income, ñ N is a posiive funcion of n N. The expression for he share of he non-radable secor in oal employmen is hen: (9) ( ) p ñ N N 1 θ = (1 γ) p c (10) p p y where y is he aggregae oupu per capia in erms of radables. The wo firs erms on he righ side represen he employmen needed o saisfy he consumpion demand for he non-radable good. The hird produc is he consumpion rae. Differeniaing equaion 10, we ge he dynamics of ñ N which saisfies: ˆñ N = (1 θ) (ˆp N ˆp ) + ˆχ ˆñ N = (1 θ)(1 ψ )ˆp N + ˆχ [( ) ( 1 α ˆñ N N α = (1 θ)(1 ψ ) 1 α T â T â N N α T ) ] + 1 α }{{} T Û + ˆχ }{{} }{{} effec of Balassa-Samuelson effec effec of financial inegraion financial inegraion (11) ( ) p c p N 1 θ, where χ = p y and ψ = (1 γ) p ψ ]0, 1[ is he share of non-radables in aggregae nominal consumpion. The properies of srucural change follow immediaely from equaion 11. There are hree drivers of srucural change. The firs is differences in observed secoral TFP growh raes (i.e., â T â N ). If produciviy grows faser in he radable secor han in he non-radable secor, hen he relaive price increases (Balassa-Samuelson effec). Wih θ < 1, consumpion demands are oo inelasic o mach all he oupu change due o TFP growh, so employmen has o move ino he slow-growing non-radable secor (Baumol s effec). Only if θ = 1, hen he employmen share is consan while he relaive price changes. Wih consan employmen shares, he faser-growing radable secor produces relaively more oupu over ime. The aggregae price changes in his case are such ha consumpion demands exacly mach all he oupu changes due o he differen TFP growh raes. The second driver is he effec of financial inegraion on relaive prices. Financial inegraion fosers a relaive price increase, and if θ < 1 i leads o an expansion of he non-radable secor (see proposiion 2). In his laer case, consumpion demands are oo inelasic o mach all he oupu change due o he cheaper capial cos benefiing he capial-inensive radable secor, so employmen has o move ino he labor-inensive non-radable secor. Finally, he hird driver is he effec of financial inegraion on he consumpion rae p c /p y : if his raio emporarily increases, he non-radable secor expands. An increase in his raio means ha he invesmen rae is falling or ha he counry accumulaes a curren accoun defici. Labor moves ou of he radable secor and ino he non-radable secor. This is he case when he counry is impaien enough (he counry is impaien if β(1 + r)(1 + x +1 ) > 1, see Appendix for a discussion on his effec). An anicipaed fall in he wedge x +1 fuels consumpion growh in he curren period, increasing he demand for boh he 8

9 non-radable and radable goods. However, non-radable goods mus be produced domesically, whereas radable goods can be impored: he share of he non-radable secor increases, and he curren accoun balance deerioraes. Proposiion 3: Wih differences in TFP and capial inensiies across secors, here are 3 drivers of srucural change: (1) he Balassa-Samuelson effec (i.e. â T > â N ) if θ 1. This effec leads o an expansion of he nonradable secor if θ < 1; (2) financial inegraion, hrough is effec on he relaive price (i.e. Û < 0 wih θ 1). This effec is a play even if he economy is on a balanced growh pah (i.e., ĉ = ŷ ). Financial inegraion leads o an expansion of he non-radable secor if θ < 1 and α N < α T ; (3) financial inegraion, by fueling a emporary demand boom wih ĉ > ŷ. Then he share of he nonradable secor expands and he curren accoun deerioraes. This effec is a play even if θ 1. If here are no differences in capial inensiies across secors and no markups, equaion 10 becomes: ( ) p n N N 1 θ = (1 γ) p c p p y The expression of srucural change hen reduces o he expression found in Ngai and Pissarides (2007): ˆn N = (1 θ)(1 ψ )(Â T Â N ) + }{{} ˆχ }{{} Balassa-Samuelson effec effec of financial inegraion (12) Proposiion 4: Absen differences in capial inensiies across secors and wih no markups, here are only 2 drivers of srucural change: (1) he Balassa-Samuelson effec (i.e. Â T > Â N ) if θ 1. This effec leads o an expansion of he nonradable secor if θ < 1; (2) financial inegraion, by fueling a emporary demand boom wih ĉ > ŷ. Economic inegraion affecs boh emporarily and permanenly he dynamics of he non-radable secor. In his secion was firs incorporaed a Balassa-Samuelson effec in a model of srucural change of a small open borrowing economy. I resuls ha he Balassa-Samuelson effec, by inducing a relaive price increase in he long-run, leads as long as TFP grows faser in he radable secor o a reallocaion of labor ino he slow-growing non-radable secor. This effec holds as long as here is a low (below one) elasiciy of subsiuion beween radable and non-radable goods. A similar effec arises if here is financial inegraion and he non-radable secor is labor inensive: financial inegraion, by lowering he user cos of capial, benefis he capial-inensive secor. However, if consumpion demands are oo inelasic o mach all he oupu change due o he cheaper capial cos, employmen has o move ino he labor-inensive non-radable secor. On op of hese wo long-run effecs, financial inegraion can also fuel a ransiory expansion of he non-radable secor: financial inegraion fuels foreign capial inflows ino he caching-up economy, and fuels a emporary demand-boom. Non-radable goods mus be produced domesically, whereas radable goods can be impored: he share of he non-radable secor increases, and he curren accoun balance deerioraes. 3. Empirical Evidence This secion presens a novel daabase ha documens he dynamics of he radable and non-radable secors and he main dimensions of economic inegraion in Europe. The daabase uses naional accouns daa a he indusry-level as well as daa on rade in goods and services o build a series of indicaors of growh and produciviy accouns for he radable and non-radable secor of European counries. Daa are available for up o 24 counries and covers up o he years , bu he coverage differs widely across counries. This aricle focuses on a subse of 12 Euro area counries over

10 3.1. Daa The daa are consuced in wo seps: firs I build indicaors o documen secor dynamics a he mos disagregaed level available; hen I classify each secor as radable or non-radable and aggregae he daa in hese wo secors. The consrucion of he daabase is deailled in Appendix 2. In he firs sep, using Eurosa Naional Accouns daa, a se of secor-level indicaors describing secor dynamics is buil for 24 European counries 13 for up o in 19 secors of he Nace revision 2 classificaion. Growh accouning indicaors are consruced using EU-KLEMS mehodology (O Mahony and Timmer, 2009). This daabase covers a wider se of counries han EU KLEMS in is 2016 updae bu wih less informaion on employmen srucure 14. This daase differs also from EU KLEMS since i allows for he exisence of profis o disinguish he share of labor, capial and profis in gross value added. The exisence of profis if no accouned for in he measure of inpus and heir revenue shares can biais he measure of TFP (Fernald and Neiman, 2011). Conrary o EU KLEMS, I do no make he assumpion ha labor and capial compensaions sum exacly o he value added, herefore I canno deduce capial compensaions from gross value added minus labor compensaions bu raher need o esimae capial compensaions. To esimae capial compensaions, informaion on he user cos of capial and capial sock are needed. User coss of capial are consruced using daa on invesmen prices and depreciaion rae (boh secor and asse specific), and a proxy of renal raes: he long-erm nominal ineres raes (benchmark cenral governmen bonds of 10 years, idenical across secors). 15 Capial compensaions are he produc of user coss of capial and capial socks a he counry-year-secor-asse level. The profi share is ulimaely deduced as he residual of he labor share and he capial share. A radabiliy indicaor is hen buil o classify each secor as radable or non-radable. To do so, I use daa on producion provided in Eurosa naional accouns. Daa on rade in services come from Eurosa balance of paymens for each European counries in he BPM5 classificaion over and in he BPM6 classificaion over (daa for 2015 are no declared for all counries and iems). Finally, daa on rade in goods come from BACI, CEPII s daabase based on COMTRADE which provides a harmonized world rade marix for values a he 6-digi level of he Harmonized Sysem of 1992 (5 699 producs) for 253 counries over 1989 o All daabases are convered ino he NACE revision 2 classificaion. The radabiliy of each secor depends on is openness raio he raio of oal rade (impors + expors) o oal producion. A secor is considered as radable if is openness raio is greaer han 10%, on average for he oal area (24 counries) and over Table 1 repors he openness raio by secor on average for he 24 counries. Unsurprisingly, mining and quarrying, manufacuring and agriculure aciviies are found radable. Concerning services, six indusries are considered radable. The non-radable secor accouns for 43% of oal producion, 52% of GVA (Gross Value Added, a curren prices) and 51% of employmen on average for he area over On average over , he share of he non-radable secor is he larges in Denmark (57% of oal employmen, 56% of GVA, 47% of producion) and smalles in Slovenia (40% of oal employmen, 49% of GVA, 41% of producion). Ineviably, he hreshold of 10% is arbirary. One possibiliy could be o apply differen radabiliy crieria for differen counries, bu applying he same crierion for all counries leads o more clearcu resuls The 24 counries are counries of he EU28 excluding Bulgaria, Croaia, Cyprus, Romania, Mala due o poor daa qualiy bu including also Norway. Counries are hus: AT: Ausria; BE: Belgium; CZ: Czech Republic; DE: Germany; DK: Denmark; EE: Esonia; EL: Greece; ES: Spain; FI: Finland; FR: France; HU: Hungary; IE: Ireland; IT: Ialy; LT: Lihuania; LU: Luxembourg; LV: Lavia; NL: Neherlands; NO: Norway; PL: Poland; PT: Porugal; SE: Sweden; SI: Slovenia; SK: Slovakia; UK: Unied Kingdom. 14 EU KLEMS uses various micro-daa sources o ge informaion on employmen srucure of he workforce, and use his informaion o build indicaors of labor services used as labor inpu for he measure of TFP. Here I raher use an indicaor of he volume of hours worked as labor inpu for he measure of TFP. 15 Since EU KLEMS ulimaely deduces capial compensaions from subsracing labor compensaions from gross value added, heir renal rae is endogenous and do no correspond o he nominal ineres rae as i incorporaes also he dynamics of profis. 16 A counry-level, radabiliy could be affeced by marke regulaions or marke srucure, which should no maer for he 10

11 Secor Tradable secor Table 1 Openness raio by secor, on average for he 24 counries 1995 Openness raio (%) , change in p.p , average B Mining and quarrying C Manufacuring I Accommodaion and food service aciviies A Agriculure, foresry and fishing H Transporaion and sorage N Adminisraive and suppor service aciviies M Professional, scienific and echnical aciviies J Informaion and communicaion K Financial and insurance aciviies Non-radable secor D Elecriciy, gas, seam and air condiioning supply R Ars, enerainmen and recreaion G Wholesale and reail rade O Public adminisraion and defence F Consrucion S Oher service aciviies E Waer supply and wase managemen P Educaion Q Human healh and social work aciviies L Real esae aciviies Source: auhor s calculaions using Eurosa and BACI. Noe: he openness raio is he raio of oal rade (impors+expors) o oal producion. Grey cells are non service aciviies. 11

12 Moreover, he use of a hreshold has he virues of being based on he sample daa and is easily subjecable o sensiiviy checks. Using a hreshold of 15% would exclude financial and insurance aciviies and informaion and communicaion from he radable secor. Using a hreshold of 20% would also exclude professional, scienific and echnical aciviies from he radable secor. Appendix 2 discusses furher he choice of he indicaor and he choice of he 10% hreshold Sylized facs The dynamics of he non-radable secor Figure 1 displays he share of he non-radable secor in oal hours worked over 1995 o 2014 in Euro area counries: core counries (Ausria, Belgium, Germany, Finland, France, Ialy, Luxembourg, Neherlands) and he periphery (Greece, Spain, Ireland, Porugal). 17 The share of he non-radable secor (Figure 1a) rose seeply in he periphery over (+4.8p.p.), while i declined slighly in core counries (-0.3p.p.). These shares sared declining afer he 2008 global financial crisis in he periphery bu no in core counries. The increase in he non-radable share before 2008 in he periphery is significan even when excluding he consrucion and real esae secors from he sample (see Figure 1b). The share of he non-radable secor in hours worked increased mos in Ireland and Greece, while i decreased in Germany (see Figure 2). The housing bubbles conribued grealy o he dynamics of he non-radable secors as he consrucion secor was he fases growing secor in mos caching-up counries over (excep for Porugal). However, he housing secor (consrucion and real esae) does no explain he bulk of he non-radable secor (excep for Spain), and oher secors played an imporan role (wholesale and reail rade more paricularly is one of he mos dynamic secors over he period in he periphery). Produciviy growh The heoreical model shows ha labor reallocaes in secors where produciviy grows relaively slowly. Figure 3a shows he change in (unbiased) TFP in he radable relaive o he nonradable secor for each group of counries: core counries (black) and he periphery (green). TFP increased faser in radable han non-radable secors in all groups of counries over The increase was seeper for he periphery (+20.9% vs. 13.9% in core counries). This effec should hus play in favor of a faser expansion of he non-radable secor in periphery han in core counries. Long-erm ineres rae The heoreical framework also shows ha financial and moneary inegraion conribues o he expansion of he non-radable secor, as long as he non-radable secor is more laborinensive. The daase shows ha labor compensaions represen on average 77% of GVA in he non-radable secor (excluding consrucion and real esae aciviies), while he share is 67% in he radable secor. The evidence is robus when correcing facor shares by he profi share (shares are hen resp. of 80% and 74%). Financial and moneary inegraion led o a convergence of nominal ineres raes among Euro area counries o abou 4% around he mid-2000s. This downward convergence induced a srong decline in he risk premia of peripheral economies, wih ineres raes declining by 7.6 p.p. on average over , while ineres raes declined by only 3.9 p.p. on average in core counries. Ineres rae increased again afer he 2008 global financial crisis and more paricularly he 2011 Euro area crisis. These dynamics are largely refleced in long-erm ineres raes deflaed by he price of radables (deflaor of GVA in he radable secor, see Figure 3a). EA11 since radable secors are well inegraed in Europe (Esrada e al., 2013). 17 The 12 core and peripheral counries of he Euro area all adoped he Euro in 1999 or 2001 for Greece. These 12 counries were classified as periphery if hey were in heir GDP per capia, in pucharsing power sandard, was in he boom hird in 1995, hey are else considered as core counries. 12

13 Figure 1 Share of he non-radable secor in oal hours worked, by counry group, , in % (a) Toal economy (b) Excluding consrucion and real esae Source: auhor s calculaions using Eurosa and BACI. Noe: a hreshold of 10% is used for he measure of radabiliy. Averages over counries weighed by he number of hours worked. The periphery includes he four counries of he EA12 (counries which adoped he euro in 2001 and before) wih he lowes GDP per capia (a purchasing power sandards) in The res of he EA12 are considered as core counries. The periphery includes: EL; ES; IE; PT. The core counries are: AT; BE; DE; FI; FR; IT; LU; NL. Figure 2 Change in he share of he non-radable secor in oal hours worked (p.p.) (a) (b) Source: auhor s calculaions using Eurosa and BACI. *For Belgium and Ireland, daa sar only in Noe: a hreshold of 10% is used for he measure of radabiliy. 13

14 Figure 3 Change in he share of he non-radable secor in oal employmen, relaive (T/N) TFP and nominal long-erm ineres raes, oal economy (dos: excl. consrucion & real esae) (a) (b) Source: auhor s calculaions using Eurosa and BACI. Noe: The measure of TFP is unbiased. Iniial year for he periphery: A hreshold of 10% is used for he measure of radabiliy. The periphery includes: EL; ES; IE; PT. The core counries are: AT; BE; DE; FI; FR; IT; LU; NL. In oal, he rising share of he non-radable secor in peripheral counries before he crisis is concomian o he wo following sylized facs (Figure 3a): a seep rise in he TFP in he radable secor relaive o he non-radable secor, a collapse in he long-erm ineres raes. 4. Quanificaion This secion assesses he conribuion of financial and marke inegraion on he dynamics of he nonradable secor over for he 11 core and periphery counries of he Euro area using a growh accouning exercise. Secion I emphasized ha boh he change in relaive (T/N) TFP and he fall in long erm ineres raes are drivers of he share of he non-radable secor in oal employmen. These wo effecs fas radable produciviy and fall in he ineres rae have already been shown o be imporan drivers of relaive prices in he long-run. Indeed, Pion (2016) shows ha he impac of a -1% differenial in he real ineres rae increases he non-radable price by 0.86% o 1.52% relaive o he Euro area. In Greece, he fall in he real ineres rae over could explain almos half of he non-radable price increase relaive o he EA average, and ogeher wih he Balassa-Samuelson (BS) effec, accoun up o 80% of is variaions. We here focus on he dynamics of he share of non-radable employmen raher han on relaive prices. To confron he daa wih he model, an illusraive calibraion is underaken o invesigae wheher he dynamics generaed by he model are broadly consisen wih he paerns in European daa. Wheher we focus on equaion (11) or equaion (12), he firs imporan parameer for our calibraion is he elasiciy of subsiuion beween he wo secors. The model suggess a way of evaluaing he elasiciy. In paricular, i provides a relaionship beween prices and quaniies: ψ = pn c N ( ) p N 1 θ = (1 γ) p c p 14

15 Expressing all variables in heir logarihm, we obain he following relaionship: [ ( )] p N log (ψ ) = log(1 γ) + (1 θ) log p (13) To esimae he parameer θ, once again he share of non-radable consumpion in oal consumpion ψ = pn cn p c is needed. Eurosa does no provide daa on oal final expendiure on consumpion per indusry bu only for he oal economy. To ge a proxy of non-radable consumpion, I use he assumpion made in he model ha all non-radable producion mus be consumed in each period. A srong limiaion wih his assumpion is ha he non-radable secor includes he real esae and consrucion aciviies, which are largely used for invesmen and no only for consumpion. I exclude his secor in he following. 18. Wih hese assumpions, radable consumpion can be deduced by rerenching non-radable gross value added from oal final expendiure ne of axes less subsidies on producs. Tradable consumpion should also be equal o gross value added minus oal invesmen and minus he curren accoun in he radable secor. 19 These wo approaches of radable consumpion give very similar measures (hey differ by +/- 5%). Finally, non-radable consumpion represens 48% of oal consumpion on average for he 12 EA counries over The elasiciy of subsiuion θ can now be esimaed using equaion (13). I assume his parameer o be he same for he 12 counries in he sample. Then he esimaing relaionship will no only include an idiosyncraic error erm bu also counry fixed effecs (assuming ha way ha he parameer γ differs across counries). Since he focus of he relaive price effec is on medium-run frequencies (raher han business cycle flucuaions), I use he Hodrick-Presco filer o smooh boh he independen and he dependen variables and use smoohed variables for he esimaion. This simple regression yields an esimae of θ 0.81 and a wo sandard error confidence inerval of [0.66; 0.97]. I chose θ = 0.81 for he benchmark esimaion. This esimae is close o he one used in Acemoglu and Guerrieri (2008): hey find an elasiciy of subsiuion of 0.76 beween capial-inensive and labor-inensive goods. Equipped wih an esimae of θ, I can measure he conribuion of financial and marke inegraion on he dynamics of he non-radable secor. To do so, I firs decompose he change in he share of he non-radable secor in oal employmen ino a he effec of he consrucion and real esae secor (H), and he effec of he change of he share of he non-radable secor in oal employmen excluding consrucion and real esae from he sample (N H): ˆ (n N + n H ) = n N (ˆn N H + ˆ (1 n H )) + n H ˆn H }{{} Consrucion and real esae wih n N + n H + n T = 1 and n N H + n T = 1 Then, using equaion (12), I decompose furher he change of he share of he non-radable secor in oal employmen excluding consrucion and real esae (ˆn N H ), ino a radiional Balassa-Samuelson effec, he addiional impac of he revisied Balassa-Samuelson effec, and he effec of unbalanced growh. The revisied Balassa-Samuelson effec includes boh he effec of he declining ineres rae on relaive prices, and he effec of correcing biased TFP. Finally, he unbalanced growh effec relies in he fac ha he 18 Invesmen in dwellings and oher buildings and srucures (asses N111 and N112 in he AN F 6 classificaion) is an imporan share of oal invesmen. When measuring he raio of his laer invesmen o GVA in he consrucion and real esae secors, invesmen represens a lile more han 90% of oal GVA on average over for he 12 counries. I hus make he srong assumpion ha all producion in hese wo secors are used for invesmen only, and do no rerench housing consumpion from final expendiure ne of axes less subsidies on producs. 19 Rewriing he budge consrain in level raher han in per capia, and replacing dividends by is expression given in he firms secion, we ge ha: p C = p Y q I + F +1 R F. Since all non-radable producion is consumed in each period, we easily ge: C T = Y T q I + F +1 R F, so radable consumpion should equalize radable gross value added minus oal invesmen (gross fixed capial formaion excluding invesmen in dwellings and oher buildings and srucures) and minus he curren accoun (wih he curren accoun proxied by he rade balance, CA = F +1 R F X M ). 15

16 economy is no on a balanced growh pah so ˆχ 0. Rewriing equaion (12), we ge a decomposiion of ˆn N H : ˆn N H = [( 1 n N H 1 ñ N H ) ] ˆñ N H + (1 n N H )(ŝ L,N H ŝ L,T ) = (1 θ)(1 ψ)( T Fˆ P T T Fˆ P N H ) + ˆχ }{{}}{{} Sandard Balassa-Samuelson effec Unbalanced growh effec ( ) ( [( ) ( 1 n N H 1 α N H α + 1 ñ N H (1 θ)(1 ψ) 1 α T â T â N H N α T ) ] ) + 1 α T Û + ˆχ [ ] (1 θ)(1 ψ)( T Fˆ P T T Fˆ P N H ) + ˆχ + (1 n N H )(ŝ L,N H ŝ L,T ) }{{} Addiional conribuion of he he Balassa-Samuelson effec revisied I compue hese effecs for wo sub-periods: (since growh raes are needed, here is no daa for 1995) and For variable wih no ime subscrips, I use heir average over he period. I use he change over he period in he sandard primal measure of TFP (biased as i does no accoun for he dynamics of profis, see Appendix for a discussion on he measures of TFP) o measure he sandard Balassa-Samuelson effec. The conribuions of each effec for each counry over he wo sub-periods are summarized in Figure 4. Unsurprisingly, beween 1996 and 2007, mos of he increase in he non-radable secor in Spain is due o he housing bubble. In Ireland he housing secor also played an imporan role: 65% of he increase in he share of he non-radable secor is due o he housing bubble and more han 25% is due o economic inegraion. In Germany, employmen grew 5% less in he non-radable secor han in he oal economy, and his is enirely due o he dynamics of he housing secor and he declining consumpion rae. In Greece and Porugal, economic inegraion can explain up o resp. 80% and 90% of he change in he share of he nonradable secor in oal employmen. Economic inegraion had a much larger impac han he sole sandard Balassa-Samuelson effec (which explain resp. 11% and 13%). Surprisingly, in Greece, he unbalanced growh effec did no play any role, and even aced slighly negaively on he share of he non-radable secor. Since he 2008 global financial crisis, he non-radable secor shrinked in every peripheral counries (Greece, Spain, Ireland, Porugal). Bu his readjusmen is mosly due o he fall in he consumpion rae and he collapse of he housing secor raher han TFP dynamics. 16

17 Figure 4 Conribuion (in p.p.) of he Balassa-Samuelson, Balassa-Samuelson revisied and unbalanced growh effecs o he change (in %) in he share of he non-radable secor in oal employmen (a) (b) Source: auhor s calculaions using Eurosa and BACI. *: daa sar in 1998 for Ireland, 1999 for Belgium, 2001 for he Neherlands and Luxembourg. Noe: a hreshold of 10% is used for he measure of radabiliy. 17