Multi-factor productivity and capital deepening in OECD countries: the role of regulation and tax structure

Transcription

1 Preliminary draf please do no cie Muli-facor produciviy and capial deepening in OECD counries: he role of regulaion and ax srucure Balázs Éger * 1 Absrac In his paper, we seek o undersand he drivers of counry-level MFP and capial deepening wih a special emphasis on produc and labour marke policies and ax srucure. For a panel of OECD counries, we find ha more sringen produc and labour marke regulaions and policies are associaed wih lower levels of MFP and capial sock. Cross-counry MFP variaions can be also explained by cross-counry variaion in labour marke regulaions, barriers o rade and invesmen and insiuions. Capuring he correlaion beween ax srucure and oucome variables is difficul. There is an overall negaive relaion beween MFP and he overall level of axaion. I is however difficul o esablish a firm relaionship beween differen ypes of axes and MFP. The posiive resul beween he capial sock and overall axaion is difficul o explain and needs furher invesigaion. JEL classificaion codes: C2, J2, O4 Keywords: muli-facor produciviy, invesmen, capial sock, produc marke regulaion, labour marke regulaion, ax srucure 1 Inroducion Undersanding he deerminans of muli-facor produciviy (MFP) and capial deepening is essenial o enable policy makers o adjus policy wih a view o improve oucomes. Empirical research on he drivers of muli-facor produciviy and capial deepening is abundan. A large fracion of empirical evidence for MFP is based on firm-level and secoral observaions. Evidence based on firm-level daa suggess ha labour and produc marke regulaions and FDI resricions penalise allocaive efficiency (Andrews and Cingalo, 2014). An increasing number of sudies also idenify an imporan correlaion beween he sringency of regulaion and muli-facor produciviy performance a he indusry level (OECD, 2003; Nicolei and Scarpea, 2003; Arnold e al., 2008; Bourlès e al., 2013; Cee e al., 2013a,b and 2014). As * OECD, Economics Deparmen. balazs.eger@oecd.org 1 The views expressed in his paper do no reflec he views of he OECD. 1

2 for capial deepening, only a few sudies analyse he relaionship beween produc and labour marke policies on he one hand and invesmen on he oher. A large majoriy of he papers looking ino he relaionship beween produc marke regulaion and invesmen use secoral and firm-level daa (see e.g. Alesina e al., 2005 and Varia, 2008 for secor-level daa and Araujo, 2011 and Cambini and Rondi, 2011 for firm-level daa). We only know of one paper (Kedrain e al. 2010) ha invesigaes he exen o which invesmen is associaed wih labour marke regulaions. I is agains his background ha his paper seeks o undersand he drivers of counry-level MFP and capial deepening. I exends on Éger (2016a) and Éger (2016b) by including ax srucure on op of produc and labour marke policies as deerminans of MFP and he capial sock. We esimae he longerm drivers of MFP and capial deepening in a large sample of 32 OECD member counries 2 for a ime period covering almos 30 years. In doing so, MFP is regressed on a se of long-erm drivers, including variables such as produc and labour marke regulaions, rade openness and measures of innovaion inensiy; and various measures of ax srucure. The esimaion resuls confirm ha anicompeiive produc marke regulaions are associaed wih lower MFP and capial deepening. We also show ha crosscounry MFP variaions can be explained by cross-counry variaion in labour marke regulaions, barriers o rade and invesmen and insiuions. Capuring he correlaion beween ax srucure and oucome variables is difficul. There is an overall negaive relaion beween MFP and he overall level of axaion. I is however difficul o esablish a firm relaionship beween differen ypes of axes and MFP. The posiive resul beween he capial sock and overall axaion is difficul o explain and needs furher invesigaion. The res of he paper is organised as follows. Secion 2 provides a discussion on modelling and daa issues. Secion 3 deals wih esimaion issues. Secion 5 repors he esimaion resuls. Secion 6 provides some concluding remarks. 2 Modelling and daa issues 2.1 Measuring MFP Mismeasuring he level and dynamics of MFP has imporan implicaions. I deermines growh a he counry level. I also affecs he fronier level o which convergence is supposed o ake place in he long run. We calculae MFP as he residual of oupu once all inpus (capial and labour) are accouned for. Cross-counry comparabiliy of MPF levels is ensured by using he (consan 2005) purchasing power pariy exchange rae (he raio of absolue price levels in he domesic and reference counry), which accouns for differen levels of economic developmen. We compue a number of alernaive aggregae MFP measures and find ha wheher human capial is included or excluded from MFP maers he mos. The oher parameers, including he ype of PPP rae o make MFP levels comparable across counries and alernaive measures of capial and labour inpu maer o a lesser exen. MFP series, which exclude human capial (oupu minus human and physical capial and labour) are implausible: a decline over decades in counries close o he fronier, and he USA being far from he fronier. Therefore, an MFP measure including human capial will be used in he empirical analysis (Éger, 2016a). This is consisen wih ypical MFP measures used in firm- and indusry-level sudies (Gal, 2013). 2 Chile and Mexico are excluded: daa on corporae axes are no available for hese wo counries. 2

3 2.2 Modelling MFP The deerminans of MFP The produciviy fronier Secor- and firm-level sudies (Cee e al. 2013a,b; Nicolei and Scarpea, 2003; Griffih e al., 2004) ypically include he absolue produciviy fronier in he esimaed equaions. The coefficien esimae on he produciviy fronier gives he exen of convergence o he fronier. If he coefficien equals 1, here is perfec convergence in he long-run o he produciviy fronier. 3 If i is smaller han 1, convergence is only parial. If, on op of he produciviy fronier, oher variables such as produc and labour marke regulaion are also included ino he long-run relaionship, convergence can ake place o he absolue fronier correced for regulaions in he counry considered: regulaions can permanenly lower he level of he MFP fronier a counry converges o in he long run. Using he absolue produciviy fronier a he macroeconomic level is rickier for wo reasons. Firs, a he firm or indusry level, produciviy is compared for relaively homogenous firms and indusries. A he aggregae level, however, composiion effecs may play an imporan role: a counry specialised in agriculure may never be able o exhibi MFP levels observed in a highly indusrialised counry by increasing MFP in agriculure. 4 Second, if he counry a he produciviy fronier is small, i is unreasonable o expec ha large counries would converge o i. In our daase, Luxembourg and Norway are ofen found o be a he absolue MFP fronier. To alleviae he second problem, he MFP level of he USA will be used as he MFP fronier in our empirical analysis. An alernaive and more sandard approach in erms of coinegraion would be o model MFP levels wihou convergence o a condiional fronier derived from he absolue MFP fronier. In such a seup, one would seek o esablish he drivers of MFP levels and convergence would ake place owards he esimaed, counry-specific long-run relaionship in an error correcion framework. In his sense, his is also convergence, which akes place owards a counry-specific (esimaed) MFP fronier. This approach is used for insance in Wesmore (2013) and Andrews and Wesmore (2014). They deermine he counry-specific produciviy fronier - o which convergence akes place hrough an error correcion mechanism - as a coinegraing relaionship beween observed MFP and produc marke regulaion. However, year fixed effecs inroduced ino he coinegraion analysis are akin o he use of an MFP fronier. Bu he fronier given by year fixed effecs is more flexible han using a counry or group of counries as he MFP fronier: i acually capures he common rend for all counries in a panel seing ha is allowed o change over ime. The covariaes used in such regressions explain he gap beween he common rend and counryspecific MFP developmens, i.e. by how much counries diverge from his common rend due o he covariaes. 3 In pracice, many sudies use he gap o he fronier. This indeed is anamoun o resricing he coefficien on he fronier o uniy. 4 In he very long run, convergence can ake place hrough a secoral shif in he economy. 3

4 Innovaion inensiy The level of MFP can be hough o be driven by a number of facors. I is widely acceped ha MFP depends on he creaion, ransmission and absorpion of knowledge (Isaksson, 2007). The wo variables capuring hese phenomena are innovaion inensiy and rade openness. Innovaion inensiy can be measured hrough spending on research & developmen, as a share of GDP (R&D inensiy) or, alernaively, hrough paening aciviy. Higher innovaion inensiy is expeced o boos MFP levels. Trade openness Trade openness is usually considered as enhancing echnology diffusion and adopion hrough rade (and foreign direc invesmen), resuling in a higher MFP level. Trade openness can be calculaed as expors and impors of goods and services over GDP divided by wo. One poenial problem wih rade openness is ha i migh be auomaically greaer for smaller counries and lower for larger counries. For insance, individual federal saes in Germany are comparably open o counries similar in size (e.g. Slovakia or he Czech Republic). Neverheless, Germany as a counry is much less open, as federal saes rade a lo wih each oher. One way o accoun for his size-bias is o regress openness on counry size, measured for insance by he log of oal populaion. The regression residuals are he par of openness, which canno be explained by counry size. Therefore, hey can be considered as he size-adjused openness raio. This paper uses such a size-adjused openness raio. Produc, labour marke regulaions and ax policies More sringen regulaion in various pars of he economy, including produc and labour markes, can impede he efficien allocaion of capial and labour wihin and across firms and indusries. This can impac negaively on MFP. Produc marke regulaion could be capured by he OECD s Produc Marke Regulaion (PMR) indicaor or he World Bank s Doing Business indicaor. The drawback of he PMR indicaor is ha i is available every five years (1998, 2003, 2008 and 2013). 5 The Doing Business indicaors are available a annual frequency. However, hey only cover he period from 2002 o The OECD s elecriciy, ranspor and communicaions regulaion (ETCR) indicaor, a subse of he OECD s Produc Marke Regulaion (PMR) indicaor, covers a longer period as i sars in 1975 and ends in I also has annual observaions. For hese wo reasons, his paper uses he ETCR indicaor, which measures he degree of produc marke regulaion on a scale of 0 o 6. Low numbers indicae less regulaion, higher numbers refer o more sringen regulaion. The raional for using his indicaor is ha more regulaion would be associaed wih a lower MFP level. We will also use he wo componens of ETCR: barriers o enry and public ownership. In addiion o produc marke regulaion, labour marke regulaion can also bear an impac on MFP hrough he direc effecs of he allocaion of labour resources and he indirec impac on capial reallocaion. Therefore, we use wo indicaors capuring labour marke regulaion: i.) spending on acive labour marke policies (ALMP), and ii.) he employmen proecion legislaion (EPL) indicaor (for permanen conracs), 5 Wesmore (2013) and Andrews and Wesmore (2014) use he PMR indicaor by filling in he gaps beween he observaions in 1998, 2003, 2008 and 2013 via linear inerpolaion. 4

5 These daa series are borrowed from Gal and Theising (2015), who provide deails on daa sources and definiions. Finally, a few measures of ax srucure are also used: he share of direc axes as a share of GDP and average labour ax wedge.in our baseline specificaion, he level of MFP is regressed on innovaion inensiy, rade openness, adjused for counry size and a measure of produc marke regulaion. Tesable equaions In our baseline specificaion, he level of MFP is regressed on innovaion inensiy, rade openness, adjused for counry size and a measure of produc marke regulaion (equaions 1). Equaion (2) is augmened using a measure of labour marke regulaion/policy: MFP f OPEN, INNOVATION, PMR,) (1) j, ( j, j, j, MFP f OPEN, INNOVATION, PMR, LMR ) (2) j, ( j, j, j, j, Modelling capial deepening The acceleraor model An acceleraor model is a simple and commonly used way o model business invesmen. 6 I assumes ha invesmen (I ) can be spli ino ne invesmen (I N ) and replacemen invesmen (I R ). Replacemen invesmen in period equals he depreciaion of he capial sock in -1: I R = δk 1 (3) where δ is he depreciaion rae. Ne invesmen is assumed o equal o changes in he desired capial sock: n I N = i=0 β i K i (4) In urn, he desired sock of capial (K ) is considered o be a linear funcion of oupu (Y). Hence, real invesmen can be wrien as curren and pas real GDP growh and lagged capial sock: 2 n I = i=0 β i Y i + δk 1 (5) Equaion (5) is usually esimaed in consan domesic prices (Oliner e al., 1995; Lee and Rabanal, 2010; Barkbu e al., 2015). I can be easily ransformed ino a ne invesmen equaion: I δk 1 = n β i Y i i=0. 6 A sric definiion of business invesmen is aggregae invesmen excluding housing invesmen and public invesmen. This ype of invesmen can be called privae business invesmen. Business invesmen according o OECD erminology excludes housing invesmen bu includes public invesmen. 5

6 2.3.2 The user cos of capial Obviously, invesmen can depend on more han jus oupu growh and lagged capial sock. According o he neoclassical model, he desired sock of capial is no only a posiive funcion of oupu bu i also depends negaively on he user cos of capial (UCC) (Chirinko, 1993; Oliner e al., 1995): I = n i=0 β i (Y i UCC σ i ) + δk 1 (6) where σ is a consan elasiciy of subsiuion beween capial and labour in he producion funcion (Chirinko, 1993). Tevlin and Whelan (2003) argue ha he capial sock is a non-saionary variable and propose a saionarised varian of equaion (6) in growh raes: I n = + K i=0 β i y i + γ i ucc i 1 n i=0 (7) Fundamenally, here exiss a long-run relaionship linking he desired sock of capial o he level of oupu and he user cos of capial, which can be wrien as follows: K = αy UCC σ (8) Log-linearisaion and he error correcion represenaion give he following equaion: 3 k = γk 1 + βy 1 + δucc 1 + θ y + π ucc (9) Invesmen and srucural policies Invesmen and capial deepening can depend on counry characerisics including he ease wih which funding is available o business. The business environmen is also an imporan facor. If he direc and indirec coss of saring a business is low, he number of business sar-ups will increase (World Bank, 2014). This in urn can ranslae ino more invesmen. Similarly, pro-compeiive produc marke regulaion is likely o push firms o inves more o say ahead of compeiors or allow he enry of new compeiors willing o inves. Less relaxed labour marke regulaions allow for a less cosly reallocaion of capial and labour wihin and across firms and could, by lowering adjusmen coss, encourage invesmen. To accoun for he impac of srucural policies on he long-run level of invesmen, equaion (6) could be augmened by produc and labour marke regulaions, as in equaion (8): K = f(y, UCC σ, PMR, LMR) (10) where PMR and LMR refer o produc and labour marke regulaions. Pelgrin e al (2002) use a similar line of reasoning o plug in financial developmen as a driver of long-erm business invesmen ino he neoclassical invesmen model. 7 7 Pelgrin e al. (2002) use real invesmen raher han he capial sock as he dependen variable in equaion (8). They argue ha in he seady sae, invesmen can be wrien as a consan share of he capial sock if one assumes a consan growh rae of he capial sock: I = (δ + g)k where δ is he depreciaion rae and g is oupu growh in he seady-sae. In pracice, he capial sock can be replaced by invesmen if he hisorical invesmen o capial sock raio is a saionary process. This is, however, no always he case in pracice. 6

7 Tesable equaions Our baseline equaion includes he real sock of capial as he dependen variable. Real oupu, he user cos of capial and produc marke regulaion are used as regressors: K f Y, UCC, PMR ) (11) j, ( j, j, j, Equaion (9) implies ha he capial sock depends on he level of real oupu, user cos and he level of produc marke regulaion. The user cos of capial is decomposed ino i.) long-erm real ineres rae, ii.) corporae axes, and iii.) relaive invesmen prices. The level of he sringency of produc marke regulaion is capured by he OECD s Energy Transpor and Communicaions Regulaion (ETCR) indicaor. Tigher regulaion can be expeced o resul in less invesmen and a lower capial sock in he long run. Equaion (9) is augmened by including labour marke regulaion (LMP): K f Y, UCC, PMR, LMR ) (12) j, ( j, j, j, j, Labour marke regulaions are measured by four indicaors drawn from OECD daa sources: he ax wedge, he Employmen Proecion Legislaion (EPL) indicaor (for permanen conracs), he unemploymen benefi replacemen rae (UBRR) and spending on acive labour marke policies (ALMP). Higher values of he firs hree variables indicae more sric regulaion. Higher labour marke regulaion is expeced o be correlaed wih a lower capial sock. The rise in he las indicaor, hrough faciliaing labour reallocaion, is hough o be posiively correlaed wih he capial sock. 2.4 The role of ax srucure Taxes have disorive effecs on economic oucomes. Taxes in many OECD counries are high as a share of GDP and have been increasing over he las 30 years or so. Arnold e al. (2011) showed ha he srucure of axes maer. In oher words, disorions may be dampened for a given level of ax revenues-o-gdp raio if ax revenues are colleced from more growh-friendly axes. In paricular, Arnold e al. (2011) repor ha income ax on business are he mos harmful for long-erm growh and ha indirec axes, especially hose levied on immovable propery and negaive environmenal exernaliies (so-called green axes) are he leas disorive. We se ou o es empirically he exen o which differen ypes of axes impac on MFP and capial deepening. We do his in wo sages. Firs, we add differen caegories of axes (as a share of GDP) o equaions (1) and (2) for MFP and equaions (11) and (12) for capial deepening. Neverheless, when doing so, we need o conrol for overall axes. Therefore, in a second sep, various ax caegories are added o he equaions ogeher wih oal ax revenues over GDP. The following ypes of axes are considered in he empirical analysis. 7

8 These daa were drawn from he OECD s Revenue Saisic Personal income ax (Taxes on income, profis and capial gains of individuals) PIT & social securiy conribuions Corporae income ax (Taxes on income, profis and capial gains of corporaes) Propery axes o Recurren axes on propery o Recurren axes on wealh o Esae, inheriance and gif axes o Taxes on financial and capial ransacions o Non-recurren axes on propery Taxes on goods and services 3 Esimaion mehods Given he non-saionary naure of he daa 8, coinegraion echniques are needed o esimae he level relaionships linking muli-facor produciviy wih is long-erm drivers. If he variables are no relaed hrough a coinegraing vecor, he esimaed level equaions may be spurious. The long-erm coefficiens are esimaed on he basis of he Dynamic OLS (DOLS) esimaor: Over he sandard OLS esimaor, i has he advanage ha i correcs for he possible endogeneiy of he regressors and auocorrelaion in he residuals by incorporaing leads and lags of he regressors in firs differences (Sock and Wason, 1993): Y j, 0 n X n j, i, i 1 i 1 l k n k 2 X 1 i, l j, i, l (13) where Y is muli-facor produciviy and X is he vecor of MFP drivers. j sands for individual counries, i for he regressors, and k1 and k2 represen respecively leads and lags. Equaion (12) can be esimaed using counry and ime fixed effecs. Wheher or no he variables of ineres are coinegraed can be esed in wo ways. Firs, he residuals obained from he long-erm relaionship (ε ) can be used o esimae he error correcion model in he second sage. Weak evidence for he presence of coinegraion is if he error correcion erm in his second sage is saisically significan and has a negaive sign. This implies an error correcion mechanism o be in place. A second and more formal es of coinegraion is when he esimaed residuals from he long-erm relaionship are esed for he presence of a uni roo. The rejecion of he null hypohesis of a uni roo can be inerpreed in favour of coinegraion, in he spiri of he Engle and Granger residual-based coinegraion approach. This paper uses Kao s residual-based panel coinegraion ess (Kao 1999), which, along equaion (13), allow for counry-specific inerceps bu imposes homogenous coefficiens. 8 Éger (2016a,b) provides empirical evidence for his. 8

9 5 Esimaion resuls 5.1. MFP Baseline specificaion The baseline specificaion in equaion (3) is esimaed using he DOLS esimaor including counry and ime fixed effecs. A number of ineresing resuls emerge from he regression analysis. Firs, here is a srong posiive correlaion beween overall R&D expendiures and openness on he one hand and MFP on he oher. There is a srong negaive relaionship beween produc marke regulaion, capured by he ETCR indicaor and MFP. A negaive correlaion beween produc marke regulaion and MFP suggess ha more sringen regulaion is associaed wih lower levels of MFP and ha more compeiion friendly regulaion spurs MFP. Neverheless, his finding needs qualificaion. The srong negaive relaionship beween aggregae ETCR and MFP holds only if counry fixed effecs are used. If boh counry and ime fixed effecs are included ino he regressions, only he coefficien esimae on public ownership is found o be saisically significan, overall ETCR and barriers o enry have large sandard errors. Beyond produc marke regulaion, he way how labour markes are regulaed can also influence aggregae MFP direcly hrough he allocaion of labour wihin and across firms and secors, and indirecly hrough invesmen decisions. To invesigae his maer, he baseline specificaion is augmened by spending on acive labour marke policies (ALMP). The sign of he coefficien esimaes on ALMP is in line wih expecaions: more spending on ALMP is associaed wih higher MFP, indicaing ha increased spending on ALMP helps labour reallocaion probably via reducing skill mismaches. Tax srucure In addiion o produc and labour marke regulaions, ax policies could influence MFP by changing incenives o inves, innovae, o sar a new business or he way how labour can be reallocaed wihin and across firms and secors. Arnold e al. (2011) found ha higher direc axes are harmful for economic growh and produciviy. The esimaion resuls show ha oal ax revenues have a saisically significan negaive relaionship wih MFP. An increase in axes is indeed correlaed wih a decrease in MFP. When looking a differen axes, personal income axes; and recurren and non-recurren axes also exhibi a negaive link o MFP. The oher axes do no have any saisically significan relaion o MFP. An excepion is corporae income axes: i has a posiive sign. This appears counerinuiive a firs sigh. Bu such a posiive sign could be explained as follows: businesses could reac o a rising ax burden by becoming more efficien and increasing produciviy (Table 1). These resuls do no change if we conrol for he overall level of axaion (Table 2). They are also robus o differen sample sizes (excluding CEE counries and oher less developed OECD economies) or o he use of he share of he differen ax caegories in oal axes raher han in GDP. Adding ALMP spending o he regressions upse he resuls for propery axes: he coefficien on recurren axes is esimaed wih large sandard errors. A he same ime, axes on ne wealh have a srong posiive sign and esae, inheriance and gif axes have a negaive sign. 9

10 Level effecs of policies and ax srucure on MFP The relaionships beween policies and MFP have hus far been idenified on he basis of he wihin dimension: he coefficiens were esimaed using he average ime variaion in our sample. In oher words, he coefficien esimaes ell us how changes in policies are relaed o changes in MFP on average in our panel. Neverheless, hey do no ell us why he level of MFP may differ across counries and wheher differences in he level of axes may influence differences in MFP levels. The enire cross-counry variaion in our daa is absorbed by he counry fixed effecs. And he MFP series have subsanial crosscounry variaion. One simple way o figure ou he MFP drivers influence cross-counry dispersion is o replace counry fixed effecs by some variables, which vehicle economic informaion. We augmen our baseline specificaion (in which MFP is regressed on he public ownership componen of ETCR, privae spending on R&D and rade openness adjused for counry size) by four variables, which only have cross-secion variaion. They are he counry averages of wo labour marke indicaors (EPL and ALMP), barriers o rade and invesmen from he OECD s PMR indicaor and a measure of insiuional qualiy. Insiuional qualiy is capured by an indicaor of he rule of law. The ax variables are added o he regressions. These regressions are esimaed wihou counry fixed effecs bu hey include year fixed effecs. The esimaion resuls are based no only on he wihin dimension bu on he wihin and beween dimensions. They show ha he new variables capuring cross-secion variaion in MFP are saisically significan a he sandard 5% level and ha heir sign makes economic sense. More sringen EPL and more resricive barriers o rade and invesmen are relaed o lower MFP levels. A he same ime, higher spending on ALMP and beer insiuions (rule of law) are associaed wih higher levels of MFP. A he same ime, he variables used in earlier esimaions are fairly robus o he replacemen of counry fixed effecs: openness is precisely esimaed wih a posiive sign. Neverheless, he R&D variable becomes saisically non-significan. The public ownership par of he ETCR indicaor is negaively linked o MFP. Looking now a he ax variables, some of he resuls observed in models including counry fixed effecs become unsable. For insance, he overall ax revenue variable ends o be negaive bu i is ofen no precisely esimaed. Propery axes are eiher saisically no significan or have a posiive sign. By conras, personal income axes and axes on goods and services are found o have a negaive link o MPF, jus as earlier. The coefficien on corporae income axes shows a negaive connecion o MFP (Table 4). Tenaively, we can make o ell he following sory: counries wih higher corporae income axes end o have lower MFP levels. A he same ime, average increases in hese axes may lead o improved produciviy oucomes as businesses will ry o compensae for he higher ax burden. Indeed, he crosscounry (beween) variaion in corporae income axes is considerably larger han he average of counryspecific (wihin) variaion. This could make i plausible ha business manage o improve efficiency a a small scale bu large differences in CIT will resul is lower MFP. The quesion one may now ask is how his new se of esimaes compares wih earlier resuls conaining counry fixed effecs. The answer o his quesion is ha hey fare reasonably well. Models wihou counry fixed effecs show a high goodness of fi, no very far from hose seen wih counry fixed effecs. 10

11 5.2. Capial deepening Baseline specificaion The sandard invesmen model works fine: only he real ineres rae is found weakly relaed o he capial sock. We firs esimae an invesmen model which links he level of capial sock o he level of oupu and he componens of he user cos (long-erm real ineres raes, corporae axes and relaive invesmen prices). All regressions include he oupu gap o conrol for cyclical flucuaions in he dependen and independen variables. Oupu has a posiive and almos uniy correlaion wih he capial sock. Long-erm real ineres raes, he corporae axes-o-gdp raio and he relaive invesmen price variable bear he expeced negaive sign. Produc marke regulaion, measured by he OECD s ETCR indicaor, shows a fairly robus negaive relaionship wih he capial sock series. The ETCR indicaor has a srong negaive correlaion wih he capial sock. As for labour marke regulaions and policies, he employmen proecion legislaion (EPL) indicaor has a srong and quaniaively imporan negaive relaionship o he capial sock. The coefficien esimae on he EPL is precisely esimaed for boh he level and log-linear specificaions Tax srucure As menioned above, he corporae axes-o-gdp raio is linked in a negaive fashion o capial deepening. Bu he oher findings are raher puzzling. The effec of overall axaion is posiive and in mos of he regressions. Recurren propery axes and axes on financial and capial ransacions are also found o be posiively correlaed wih he capial sock. These surprising findings need furher research and invesigaion. 9 6 Conclusions This paper invesigaed he deerminans of MFP and capial deepening a he macroeconomic level wih a special emphasis on produc and labour marke regulaions and ax srucure. We esablished ha anicompeiive produc marke regulaions are associaed wih lower MFP and capial deepening. We also showed ha cross-counry MFP variaions can be explained by cross-counry variaion in labour marke regulaions, barriers o rade and invesmen and insiuions. Capuring he correlaion beween ax srucure and oucome variables is difficul. There is an overall negaive relaion beween MFP and he overall level of axaion. I is however difficul o esablish a firm relaionship beween differen ypes of axes and MFP. The posiive resul beween he capial sock and overall axaion is difficul o explain and needs furher invesigaion 9 Noe ha he looking a cross-counry differences in he capial sock variable does no make sense: capial sock is a cumulaed index wih no cross-counry meaning (larger counries will have larger capial socks). 11

12 Table 1. MFP and various ax caegories (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) c ** ** 10.47** ** ** ** ** ** ** ** ** rade openness (size adjused) 0.007** 0.008** 0.008** 0.008** 0.008** 0.008** 0.008** 0.008** 0.008** 0.008** 0.007** business expendiures on R&D 0.054** 0.048** 0.048** 0.037** 0.037** 0.052** 0.044** 0.045** 0.056** 0.045** 0.041** produc marke regulaion (ETCR pub.own) -0.02** ** ** ** -0.02** ** ** ** ** ** ** labour marke policy (log ALMP) oal ax revenues ** personal income ax (PIT) ** PIT & SSC corporae income ax 0.021** propery axes ** recurren propery axes ** recurren axes on ne wealh esae, inheriance and gif axes axes on financial and capial ransacions non-recurren propery axes * axes on goods and services ** Kao coinegraion es: null of no coinegraion (p-value) error correcion erm ** ** ** ** ** ** ** ** -0.04** ** ** adjused R-squared No. of observaions No. of counries Noe: * and ** denoe saisical significance a he 10% and 5% levels, based on robus sandard errors. Kao is he Kao panel coinegraion es. The regressions include a measure of human capial and a measure of he cycle (oupu gap) o ake ou cyclical flucuaions from he daa. The regressions include counry and year fixed effecs. 12

13 Table 2. MFP and various ax caegories conrolling for oal ax revenues (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) c ** ** ** ** ** ** ** ** ** ** rade openness (size adjused) 0.007** 0.007** 0.007** 0.007** 0.007** 0.007** 0.007** 0.007** 0.007** 0.007** business expendiures on R&D 0.053** 0.049** 0.048** 0.046** 0.057** 0.052** 0.054** 0.063** 0.054** 0.046** produc marke regulaion (ETCR pub.own) ** ** ** ** ** -0.02** ** -0.02** -0.02** ** labour marke policy (log ALMP) oal ax revenues * ** ** ** * ** ** ** ** personal income ax (PIT) PIT & SSC corporae income ax 0.025** propery axes ** recurren propery axes -0.07** recurren axes on ne wealh esae, inheriance and gif axes axes on financial and capial ransacions non-recurren propery axes -0.03* axes on goods and services ** Kao coinegraion es: null of no coinegraion (p-value) error correcion erm ** ** ** ** ** -0.04** ** ** ** ** adjused R-squared No. of observaions No. of counries Noe: * and ** denoe saisical significance a he 10% and 5% levels, based on robus sandard errors. The regressions include a measure of human capial and a measure of he cycle (oupu gap) o ake ou cyclical flucuaions from he daa. The regressions include counry and year fixed effecs. 13

14 Table 3. MFP and various ax caegories conrolling for oal ax revenues and labour marke policies (ALMP) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) c ** ** ** ** ** ** ** ** ** 11.76** rade openness (size adjused) 0.006** 0.005** 0.007** 0.006** 0.006** 0.006** 0.006** 0.006** 0.006** 0.006** business expendiures on R&D * 0.033* * 0.03* produc marke regulaion (ETCR pub.own) ** ** ** ** ** ** ** ** ** ** labour marke policy (log ALMP) 0.035** 0.031** ** 0.039** 0.048** 0.049** 0.044** 0.042** 0.032** oal ax revenues ** ** ** -0.01** ** ** ** ** -0.01** * personal income ax (PIT) PIT & SSC corporae income ax 0.026** propery axes recurren propery axes recurren axes on ne wealh 0.132** esae, inheriance and gif axes ** axes on financial and capial ransacions non-recurren propery axes axes on goods and services ** Kao coinegraion es: null of no coinegraion (p-value) error correcion erm ** ** -0.06** ** ** ** ** ** ** ** adjused R-squared No. of observaions No. of counries Noe: * and ** denoe saisical significance a he 10% and 5% levels, based on robus sandard errors. The regressions include a measure of human capial and a measure of he cycle (oupu gap) o ake ou cyclical flucuaions from he daa. The regressions include counry and year fixed effecs. 14

15 Table 4. MFP and various ax caegories allowing for cross-counry variaion as well in he explanaory variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) c 10.77** 10.76** ** ** 10.39** ** ** ** ** 10.7** produc marke regulaion (ETCR pub.own) ** ** ** ** ** ** rade openness (size adjused) 0.003** 0.002** 0.003** 0.004** 0.004** 0.003** 0.003** 0.003** 0.003** 0.003** business expendiures on R&D by indusry * ** employmen proecion legislaion (EPL) ** ** ** ** ** ** ** ** ** ** labour marke policy (ALMP) 0.003** 0.003** 0.003** 0.002** 0.003** 0.003** 0.003** 0.002** 0.002** 0.003** rule of law (cross counry) 0.417** 0.434** 0.405** 0.408** 0.393** 0.385** 0.391** 0.41** 0.397** 0.433** PMR barriers o rade and invesmen (cross counry) -0.19** -0.17** ** -0.2** ** ** ** ** ** ** oal ax revenues ** * * * 0.008** personal income ax (PIT) ** PIT & SSC 0.02** corporae income ax ** propery axes 0.071** recurren propery axes 0.043** recurren axes on ne wealh esae, inheriance and gif axes axes on financial and capial ransacions 0.068** non-recurren propery axes 0.326** axes on goods and services ** adjused R-squared No. of observaions No. of counries Noe: * and ** denoe saisical significance a he 10% and 5% levels, based on robus sandard errors. The regressions include year fixed effecs. 15

16 Table 5. Capial deepening and ax srucure (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) c 1.691* 1.762* * ** ** 1.662* 1.907* log real oupu 0.961** 0.96** 1.003** 0.96** 0.969** 0.953** 0.969** 0.933** 0.961** 0.954** log relaive invesmen prices -0.27** ** ** ** ** ** ** ** ** ** long-erm real ineres rae ** ** ** ** ** ** ** ** ** ** produc marke regulaion (ETCR pub.own) ** ** ** ** ** ** ** ** ** ** employmen proecion legislaion (EPL) oal ax revenues 0.011** ** 0.012** 0.012** 0.012** 0.013** 0.012** 0.013** 0.013** personal income ax (PIT) PIT & SSC 0.011** corporae income ax ** propery axes 0.014** recurren propery axes 0.01 recurren axes on ne wealh 0.07** esae, inheriance and gif axes ** axes on financial and capial ransacions 0.037** non-recurren propery axes axes on goods and services Kao coinegraion es: null of no coinegraion (p-value) error correcion erm ** ** ** ** -0.1** ** ** ** -0.1** ** adjused R-squared No. of observaions No. of counries Noe: * and ** denoe saisical significance a he 10% and 5% levels, based on robus sandard errors. The regressions include a measure of he cycle (oupu gap) o ake ou cyclical flucuaions from he daa. The regressions include counry and year fixed effecs. 16

17 Table 6. Capial deepening and ax srucure wih labour marke regulaion (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) c 7.284** 7.583** 5.992** 7.304** 7.181** 7.577** 7.047** 8.689** 7.183** 7.448** log real oupu 0.774** 0.765** 0.818** 0.773** 0.777** 0.764** 0.782** 0.723** 0.777** 0.769** log relaive invesmen prices ** ** ** -0.42** ** ** ** ** ** ** long-erm real ineres rae ** ** ** ** ** ** ** ** ** ** produc marke regulaion (ETCR pub.own) * * * employmen proecion legislaion (EPL) ** ** ** ** ** ** ** ** ** ** oal ax revenues 0.006** ** 0.007** 0.008** 0.007** 0.008** 0.006** 0.008** 0.008** personal income ax (PIT) PIT & SSC 0.012** corporae income ax -0.01** propery axes 0.02** recurren propery axes recurren axes on ne wealh 0.068** esae, inheriance and gif axes ** axes on financial and capial ransacions 0.065** non-recurren propery axes axes on goods and services Kao coinegraion es: null of no coinegraion (p-value) error correcion erm ** -0.09** ** ** ** ** -0.09** ** ** ** adjused R-squared No. of observaions No. of counries Noe: * and ** denoe saisical significance a he 10% and 5% levels, based on robus sandard errors. The regressions include a measure of he cycle (oupu gap) o ake ou cyclical flucuaions from he daa. The regressions include counry and year fixed effecs. 17

18 References Alesina, A., S. Ardagna, G. Nicolei, F. Schianarelly, 2005, Regulaion and invesmen, Journal of he European Economic Associaion, 3(4), Andrews, D. and B. Wesmore (2014), Managerial capial and business R&D as enablers of produciviy convergence, OECD Economics Deparmen Working Papers, No.1137, OECD publishing. Andrews, D. and F. Cingano (2014), Public policy and resource allocaion: evidence from firms in OECD counries, Economic Policy, 29(78), Araújo, S., 2011, Has deregulaion increased invesmen in infrasrucure?: Firm-level evidence from OECD counries, OECD Economics Deparmen Working Paper No Arnold, J., B. Brys, C. Heady, A. Johansson, C. Schwellnus and L. Varia (2011), Tax policy for economic recovery and growh, Economic Journal, 121(550), F59-F80. Arnold, J., G. Nicolei and S. Scarpea (2008), Regulaion, allocaive efficiency and produciviy in OECD counries: Indusry and firm-level evidence, OECD Economics Deparmen Working Paper No Barkbu, B., S. P. Berkmen, P. Lukyansau, S. Saksonovs and H. Schoelermann (2015), Invesmen in he euro area: why has i been weak?, IMF Working Papers, No. WP/815/832. Bourlès, R., G. Cee, J. Lopez, J. Mairesse, G. Nicolei (2013), Do produc marke regulaions in upsream secros curb produciviy growh? Panel daa evidence for OECD counries, The Review of Economics and Saisics, 95(5), Cambini, C and L. Rondi (2011), Regulaory independence, invesmen and poliical inference: Evidence from he European Union, RSCAC Working Paper 2011/42, European Universiy Insiue. Cee, G., J. Lopez and J. Mairesse (2013), Produc and labour marke regulaions, producion prices, wages and produciviy, NBER Working Paper No Cee, G., J. Lopez, J. Mairesse (2013), Upsream produc marke regulaions, ICT, R&D and produciviy, NBER Working Paper No Cee, G., J. Lopez, J. Mairesse (2014), Produc and labour marke regulaions, producion prices, wages and produciviy, NBER Working Paper No Chirinko, R. (1993), Business fixed invesmen spending: modelling sraegies, empirical resuls, and policy implicaions, Journal of Economic Lieraure, Vol. 31(4), pp Éger, B. (2016a), Regulaion, insiuions and produciviy : new macroeconomic evidence from OECD counries, American Economic Review, Papers & Proceedings, 106(5), forhcoming, hps:// Éger, B. (2016b), Regulaion, insiuions and aggregae invesmen : new evidence from OECD counries, OECD Economics Deparmen working paper (forhcoming) Gal, P. (2013), Measuring oal facor produciviy a he firm level using OECD-ORBIS, OECD Economics Deparmen working paper No

19 Gal, P. and A. Theising (2015), The macroeconomic impac of policies on labour marke oucomes in OECD counries: a reassessmen, OECD Economics Deparmen Working Paper No Griffih, R., S. Redding, J. Van Reenen (2004), Mapping he wo faces of R&D: produciviy growh in a panel of OECD indusries, The Review of Economics and Saisics, 86(4), Isaksson, A. (2007), Deerminans of oal facor produciviy: a lieraure review, Unied Naionals Indusrial Developmen Organizaion, Research and Saisics Branch Saff Working Paper 02/2007. Kao, C. (1999), Spurious regression and residual-based ess for coinegraion in panel daa, Journal of Economerics, 90(1), Kedrain, C., I. Koske, I. Wanner (2010), The impac of srucural policies on saving, invesmen and curren accouns, OECD Economics Deparmen Working Papers No Lee, J. and P. Rabanal (2010), Forecasing U.S. invesmen, IMF Working Papers, No. WP/10/246. Nicolei, G. and S. Scarpea (2003), Regulaion, produciviy and growh: OECD evidence, Economic Policy, 18(36), Oliner, S., G. Rudebusch and D. Sichel (1995), New and old models of business invesmen: a comparison of forecasing performance, Journal of Money, Credi and Banking, Vol. 27(3), pp Pelgrin, F., S. Schich and A. de Serres (2002), Increase in business invesmen raes in OECD counries in he 1990s: How much can be explained by he fundamenals? OECD Economics Deparmen Working Paper No Sock, J. and M. Wason (1993), A simple esimaor of coinegraing vecors in higher order inegraed sysems, Economerica, 61 (4), Tevlin, S. and K. Whelan (2003), Explaining he invesmen boom in he 1990s, Journal of Money, Credi and Banking, Vol. 35(1), pp Varia, L. (2008), How do axes affec invesmen and produciviy? An indusry-level analysis of OECD counries, OECD Economics Deparmen Working Paper No Wesmore, B. (2013), R&D, Paening and Growh: The Role of Public Policy, OECD Economics Deparmen Working Papers, No World Bank (2014), Research on he effecs of business regulaion. hp:// repors/~/media/giawb/doing%20business/documens/annual-repors/english/db14- Chapers/DB14-Research-on-he-effecs-of-business-regulaions.pdf 19