No 2004 02 Technology dfferences, nsuons and economc growh : a condonal condonal convergence Hervé Boulhol
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Hervé Boulhol No 2004 02 February
Technology dfferences, nsuons and economc growh : a condonal condonal convergence TABLE OF CONTENTS SUMMARY...4 ABSTRACT...5 RÉSUMÉ...6 RÉSUMÉ COURT...7 1. INTRODUCTION...8 2. TECHNOLOGY DIFFERENCES AND DIFFUSION...10 3. GROWTH EQUATIONS AND INSTITUTIONS...11 4. DATA...13 5. FROM THE INSTITUTIONAL DATABASE TO THE ECONOMETRIC SPECIFICATION...15 6. RESULTS...17 7. CONCLUSION...24 BIBLIOGRAPHY...25 APPENDIX 1...38 APPENDIX 2...41 APPENDIX 3...45 APPENDIX 4...46 LIST OF WORKING PAPERS RELEASED BY CEPII...58 3
CEPII, Workng Paper No 2004-02 TECHNOLOGY DIFFERENCES, INSTITUTIONS AND ECONOMIC GROWTH : A CONDITIONAL CONDITIONAL CONVERGENCE SUMMARY The augmened Solow model by Mankw, Romer and Wel (1992) exhbed he role of human capal for long erm growh pah and led s auhors o accep eher he assumpon of dencal echnology across counres or he reamen of echnology dfferences as resduals n he growh equaon. However, Hall and Jones (1996, 1999) and Klenow and Rodrguez-Clare (1997) have shown ha producvy levels and oupu per worker are hghly correlaed, whch cass doubs on he condonal convergence scenaro. Ye he cross secon leraure has no drawn he necessary mplcaons. Acknowledgng he mporance of akng no accoun producvy dfferences, we break down producvy no wo componens: a pure echnologcal par and s complemen called effcency. From a smple model of echnology dffuson, we focus on he neracons beween nsuons and echnology dfferences and denfy hree complemenary channels hrough whch nsuons mpac growh: effcency n he use of echnology, long erm TFPgrowh and echnology dffuson. To shed lgh on how growh and nsuons nerplay, our framework s esed from a new and dealed daabase on nsuons developed by he French Mnsry of Economy, Fnance and Indusry (MINEFI). Daa was colleced hrough a quesonnare by he Economc Mssons of he MINEFI n 51 counres represenng 80% of world GDP. The daabase consss of 330 ems on nsuons n a broad sense, each recevng a rankng from 0 o 4 for each counry. The robusness of he daabase has been esablshed by Berheler, Desdogs and Ould-Aouda (2003) hrough a comparave sudy wh oher nsuonal daabases used n varous economc sudes. We fnd ha echnology dffuson subsanally mpacs economc performance and ha cachng-up s condonal o he qualy of he approprae nsuon a mx of R&D, nnovaon and capal-rsk suppor -, wh he annual rae of convergence o he echnologcal froner varyng from 0% o 12.4% dependng on he counry. Insuons also maer for echnologcal effcency, as our non-corrupon varable, for nsance, conrbues as much as he sock of human capal o he producvy level heerogeney. Moreover, long run TFP-growh dfferences are sgnfcanly deermned by such nsuons lke he ones reflecng a compeve produc marke or a favourable nnovaon envronmen. Havng conrolled for nsuons, nernaonal rade measured, by he openness rae, s nsgnfcan o explan neher TFP-growh dfferences nor echnology dffuson. However, when we ake he manufacurng share n expors no accoun, we fnd a sgnfcan mpac of rade on TFP-growh, comng solely from he rcher counres n he sample, whch clearly pons o a non-lneary. Ths suggess ha he conrbuon of rade s posve only f he specalsaon s approprae and he developmen farly advanced. Ths las resul s enave because of poenal endogeney bases. Includng human capal flows as a deermnan of he seady-sae reveals ha MRW s approach and ours are 4
Technology dfferences, nsuons and economc growh : a condonal condonal convergence complemens raher han subsues. Condonal convergence here s also condonal o sharng he same echnology and qualy of nsuons, whch renders recen observed dvergence well accouned for. ABSTRACT Hghlghng ha echnology s only a componen of producvy, hs sudy focuses on he neracons beween nsuons and echnology dfferences o explan cross-counry growh paern. Three complemenary channels hrough whch nsuons mpac growh are denfed: effcency n he use of echnology, long erm TFP-growh and echnology dffuson. From a new and dealed daabase on nsuons developed by he French Mnsry of Economy, Fnance and Indusry, poor nsuonal qualy, beyond human capal, s esmaed o be he source of an annual growh-rae loss of beween 2.4 and 6.1 percenage pon for half of he counres. Technology dffuson speed s nsuonally relaed and he dsance o he echnology froner s reduced from 0% o 12.4% annually dependng on he counry. Trade has a non lnear nfluence on growh, beng sgnfcan only for counres already advanced n he developmen phase. Condonal convergence here s also condonal o sharng he same echnology and qualy of nsuons, renderng recen observed dvergence well accouned for. J.E.L. classfcaon: Keywords: O11; O33; O47 Technology dffuson; Insuons; Producvy; Growh 5
CEPII, Workng Paper No 2004-02 ECARTS TECHNOLOGIQUES, INSTITUTIONS ET CROISSANCE ÉCONOMIQUE: UNE CONVERGENCE CONDITIONNELLE CONDITIONNELLE RÉSUMÉ L exenson du modèle de Solow par Mankw, Romer e Wel (1992) a ms en évdence le rôle du capal human pour le sener de crossance à long erme e a condu ses aueurs à acceper l hypohèse d une echnologe denque pour l ensemble des pays ou le raemen des dfférences echnologques comme résdus des équaons de crossance. Hall e Jones (1996, 1999) e Klenow e Rodrguez-Clare (1997) on cependan monré que les nveaux de producvé e de producon par êe éaen foremen corrélés, remean en cause le scenaro de convergence condonnelle, sans que la léraure en a ré oues les mplcaons qu s mposen. Reconnassan l enjeu de la prse en compe des dfférences de producvé, nous décomposons la producvé en deux élémens : une composane puremen echnologque e son complémen que nous appelons «effcacé». A parr d un modèle smple de dffuson echnologque, nous nous concenrons sur les neracons enre les nsuons e les dfférences de echnologes pour explquer l évoluon comparée de la crossance enre pays. Tros canaux complémenares par lesquels les nsuons on une nfluence sur la crossance son denfés: l effcacé dans l ulsaon des echnologes, la crossance de la producvé à long erme e la dffuson des echnologes. Pour clarfer les neracons enre les nsuons e la crossance économque, nore modèle es esé à parr d une nouvelle base de données nsuonnelles déallée, développée par le MINEFI. Les données on éé collecées par les Mssons Economques du MINEFI dans 51 pays représenan 80% du PIB mondal. La base de donnée es composée de 330 ems sur les nsuons, noon prse dans son sens large, chacun d enre eux recevan une noe enre 0 e 4 pour chaque pays. Berheler, Desdogs e Ould-Aouda (2003) on éabl la robusesse de la base en la rapprochan d aures bases de données nsuonnelles ulsées dans les ravaux économques. Nous rouvons que la dffuson echnologque a un mpac subsanel sur la performance économque e que le rarapage es condonnelle à la qualé de l nsuon adéquae un mélange de R&D, d nnovaon e de suppor au capal-rsque -, avec une vesse annuelle de convergence vers la fronère echnologque varan de 0% à 12,4% selon les pays. Les nsuons mporen auss pour l effcacé dans l ulsaon des echnologes, e nore mesure de la non-corrupon, par exemple, conrbue auan que le sock de capal human à la dsperson de la producvé. De plus, les dfférences de crossance de la producvé oale des faceurs (PTF) à long erme son sgnfcavemen déermnées par des nsuons elles que celles refléan un marché des produs concurrenel ou un envronnemen favorable à l nnovaon. En conrôlan les dfférences nsuonnelles, le commerce nernaonal, mesuré par le aux d ouverure, n es sgnfcaf n pour explquer les dfférences de crossance de la PTF n pour la dffuson des echnologes. Cependan, lorsque l on prend en compe la par des bens manufacurés dans les exporaons, alors nous rouvons un mpac sgnfcaf du commerce sur la crossance de la PTF, provenan 6
Technology dfferences, nsuons and economc growh : a condonal condonal convergence seulemen des pays les plus rches dans l échanllon, ce qu ndque une non-lnéaré. Cela suggère que la conrbuon du commerce es posve seulemen s la spécalsaon es approprée e le développemen déjà avancé. Ce derner résula es fragle en rason d évenuels bas d endogénéé. De plus, la prse en compe de l mpac du flux de capal human pour l éa réguler révèle que l approche de MRW e la nore son des complémens pluô que des subsus. La convergence condonnelle es c condonnelle auss au fa de dsposer des mêmes echnologes e des mêmes nsuons, e rend compe de la dvergence récemmen observée. RÉSUMÉ COURT Inssan sur la dsncon enre producvé e echnologe, cee éude se concenre sur les neracons enre les nsuons e les écars de echnologques pour explquer l évoluon comparée de la crossance enre pays. Tros canaux complémenares par lesquels les nsuons on une nfluence sur la crossance son denfés: l effcacé dans l ulsaon des echnologes, la crossance de la producvé à long erme e la dffuson des echnologes. A parr d une nouvelle base de données nsuonnelles déallée, développée par le MINEFI, nous esmons que l nsuffsane qualé des nsuons, au-delà du capal human, es la cause d un défc de aux de crossance annuelle enre 2,4 e 6,1 pon de pourcenage pour la moé des pays. La vesse annuelle de dffuson des echnologes vare de 0% à 12,4% selon les pays en foncon de leur nveau nsuonnel. Le commerce a un mpac non lnéare sur la crossance pusqu l es sgnfcaf seulemen pour les pays déjà avancés dans la phase de développemen. La convergence condonnelle es c condonnelle auss au fa de dsposer des mêmes echnologes e de la même qualé des nsuons, e rend compe de la dvergence récemmen observée. J.E.L. classfcaon: Mos-clés: O11 ; O33 ; O47 Dffuson des echnologes ; Insuons ; Producvé ; Crossance 7
CEPII, Workng Paper No 2004-02 TECHNOLOGY DIFFERENCES, INSTITUTIONS AND ECONOMIC GROWTH : A CONDITIONAL CONDITIONAL CONVERGENCE (*) Hervé Boulhol 1. INTRODUCTION The semnal arcle by Mankw, Romer and Wel (1992), subsequenly denoed MRW, shed lgh on he conrbuon of human capal n reconclng he measured low speed of condonal convergence beween counres wh a physcal capal share of around onehrd. Ther man concluson s ha, when human capal s added, he hen augmened Solow model s well sued o analyse growh across counres. I follows ha, conrary o endogenous growh heory, he growh process s solely drven by facor accumulaon (ncludng human capal), s conssen wh a rae of convergence of around 2% a year (raher han he 4%-5% expeced from he Solow exbook model) and valdaes underlyng assumpons of consan reurns o scale and dencal echnology. The MRW framework has been crcsed on dfferen grounds. Frs, he embedded human capal heory reas human capal jus as anoher accumulang facor. Ths mples ha human capal should ener no growh equaon hrough s growh rae, bu here s confuson n wheher he sock of human capal maers raher han he flow as n MRW. Benhabb and Spegel (1994) convncngly suppor he vew ha he sock of human capal s a deermnan of he magnude of a counry s Solow resdual. Second, as mos of he cross-counry leraure, he MRW approach s subjec o he bas comng from he dencal echnology assumpon. The condonal convergence predcons has been more and more dffcul o reconcle wh he facs ponng o global dvergence as oulned by Prche (1997). Promped by J.Temple s fourh queson abou he causes of ncome dfferences (Temple, 1999, p.113), we sar from he nference ha f poor counres are poor no only because of a lack of npus ha wll accumulae faser foserng convergence, mus be ha hey are poor also because of overall effcency and echnology dfferenals - whaever hese mean - ha may perss or aggravae over me especally f hey are due o nsuonal dfferences. 1 The role of nsuons as a deermnan of long erm growh s ncreasngly recognsed, ye more research s needed o dsenangle whch nsuons maer for economc performance and above all o ncorporae hem properly n economc heory. Even hough Rodrk (2002, 2003) s convncng n argung ha he same nsuons may have dfferen economc (*): I would lke o express my graude o Lonel Fonagné, Jacques Ould-Aouda and Parck Arus for havng made hs sudy possble. I would parcularly lke o hank Agnès Bénassy-Quéré, Gullaume Gauler, Roman Rancère and he parcpans of he Cep semnar for her valuable commens. Fnally I very much apprecaed he help I receved from Mayls Coupe and Davd Galvn, as well as he warm welcome I receved from Cep employees. 1 (Q4) Are poor counres poor manly because hey lack npus, or because of echnology dfferences? 8
Technology dfferences, nsuons and economc growh : a condonal condonal convergence mpacs based on a counry s dosyncrases, we sugges ha he qualy of some nsuons does nfluence economc performance overall. In hs sudy, we show ha echnology dfferences play an mporan role n he cross-counry growh paern and ha nsuons maer for oal facor producvy (TFP) level and growh raes, and for echnology dffuson. Many reasons may explan why echnology dffers sgnfcanly beween counres. Paen proecon, learnng by dong, knowledge dfferences derac echnology from s publcgood preenson (Casell, Esquvel and Lefor, 1996). Moreover, akng a broader vew of producvy, nsuons lnked o he socal, polcal or legal aspecs of effcency conrbue o producvy dfferenals. Surprsngly, he growh leraure does no pay enough aenon o heerogeney n echnology. When does, hs heerogeney s reaed n panel esmaes and as a fxed effec, he deals of whch are rarely avalable, makng an assessmen of wheher hey do represen wha hey should dffcul. Excepons are Hall and Jones (1996, 1999), Klenow and Rodrguez-Clare (1997) who precsely esmae producvy dfferences and Bloom, Cannng and Sevlla (2002) whose concerns are close o ours. We hope our conrbuon o be heorecal and emprcal. Theorecally we denfy hree channels hrough whch nsuons may mpac producvy. Frs, a sac conrbuon hrough effcency n he use of echnology, second a perssen dynamc one hrough long run TFP-growh raes and hrd a emporary dynamc one hrough echnology dffuson. Emprcally we es our framework usng a new and dealed daabase on nsuons ha was developed by he French Mnsry of Economy, Fnance and Indusry (MINEFI). Daa from a quesonnare was colleced by he Economc Mssons of he MINEFI n 51 counres represenng 80% of world GDP. The daabase consss of 330 ems on nsuons n a broad sense, each recevng a rankng from 0 o 4 for each counry. We fnd ha echnology dffuson subsanally mpacs economc performance and ha cachng-up s condonal o he qualy of he approprae nsuon a mx of R&D, nnovaon and capal-rsk suppor -, wh he annual rae of convergence o he echnologcal froner varyng from 0% o 12.4% dependng on he counry. Insuons also maer for echnologcal effcency, as our non-corrupon varable, for nsance, conrbues as much as he sock of human capal o he producvy level heerogeney. Moreover, long run TFP-growh dfferences are sgnfcanly deermned by such nsuons lke he ones reflecng a compeve produc marke or a favourable nnovaon envronmen. In addon, ncludng human capal flows as a deermnan of he seadysae reveals ha MRW s approach and ours are complemens raher han subsues. Havng conrolled for nsuons, nernaonal rade measured, by he openness rae, has a non-lnear conrbuon o TFP-growh, beng sgnfcan and posve only f he specalsaon s approprae and he developmen farly advanced. Ths las resul s enave because of poenal endogeney bases. The sudy s organsed as follows. Secon 2 hghlghs he mporance of akng no accoun echnology dfferences and nroduces our approach regardng he echnologcal process. Secon 3 deals how nsuons and growh nerplay n he model. Secon 4 descrbes he daa and he selecon of he nsuonal varables, leadng o he economerc 9
CEPII, Workng Paper No 2004-02 specfcaon n Secon 5. Resuls are presened n Secon 6 where economerc ssues are also dscussed. Fnally, Secon 7 concludes. 2. TECHNOLOGY DIFFERENCES AND DIFFUSION The mos dspuable crcal assumpon n he cross-secon growh leraure suggess ha eher all counres share he same level of echnology and echnologcal progress or ha he dfferences n hese levels are reaed as resduals, mplyng ha hey are beng consdered ndependen from oher explanaory varables. Ths s an exreme conjecure snce means ha echnologcal change spreads nsananeously and compleely o every counry, whaever he level of openness or nsuonal profle, and leads o havng only he dfferences of capal per un of labour o explan dfferences of oupu per capa. Assumng a Cobb-Douglas producon funcon and wh sandard noaons, Y = K a b 1 a b ). H ( A L (1) where Y s oupu, K and H are socks of physcal and human capal respecvely, A s he producvy level and L he number of workers. Oupu per worker y for he counry s herefore gven by 1 a b a b 1 a b.( K / L ).( H / L ) A KH y Y / L = A.( Z / L ) a (2) wh Z KH b/ a = K.( H / L) defnng a capal aggregae bul from he physcal capal sock and he human capal sock per capa. If we assume ha he producvy level A s ndependen of he counry ( = A, ), hen he rao of oupu per capa n 1980 A beween he USA and Uganda of 48 o 1, beng he wo exremes n our daa, ranslaes no KH a hghly unrealsc capal aggregae, Z, per capa rao of 110,700 o 1 usng a physcal capal share of one-hrd. Moreover, recognsng he producvy dfferences, s apparen from equaon (2), vald a each me, ha boh he nal producvy level and he nal oupu per capa are closely lnked, whch renders growh equaon esmaes assumng dencal producvy serously based. Whle hs nconvenence s wellacknowledged, he growh leraure has no drawn ye all he necessary mplcaons. Over he las ffeen ears, economc research has made some noable advances n he undersandng of wha producvy s. However, he essence of s conens remans unknown, and he parameer A s ofen ndsncly desgnaed as eher he producvy or he echnology level. Ths creaes confuson n denfyng he role of echnology and herefore, adopng a dfferen posure, we nss here on he dsncon beween he wo noons and call he complemen of echnology n producvy: effcency. Inspred by Bassann and Scarpea (2001), we break down he oal level of producvy A no wo componens: a pure echnology level B and he degree of effcency n usng hs echnology X so ha A equals B. X. As he benchmark, he counry wh he hghes GDP per capa n 1980, he USA, has been chosen. We denoe b = B / B, an nverse 10 ben
Technology dfferences, nsuons and economc growh : a condonal condonal convergence ndcaor of he dsance o he froner, x = X / X, he rao of he relave effcency o he benchmark, and a = A / Aben = b. x, he relave producvy level. Insuonal qualy s consdered as mpacng he echnologcal effcency X and possbly he echnology dffuson whch process s mos smply governed by: 2 ben b& ( ) = v.(1 b ( )) (3) where sands for me: n he long run echnologes converge o he froner a a pace represened by v, whch wll be esed as beng consan across counres or nsuonallyrelaed. Noe s only he pure echnology componen ha s assumed o converge (or dverge f v s negave), and oal producvy dscrepances may perss as a resul of dfferences n nsuonally-relaed effcency X. 3. GROWTH EQUATION AND INSTITUTIONS We suppose ha nsuons ener no growh equaons hrough hree dfferen channels: he level of echnologcal effcency, he progress of echnologcal effcency and possbly he speed of echnology dffuson. Nocng ha he producvy level A can be wren as A. x. b, we can spl TFP-growh no hree componens: ben A & A = A & A ben ben x& x b& b (4) The frs erm on he rgh s smply he benchmark TFP-growh, denoed by g, he second, denoed c, s he long run TFP-growh defc o he benchmark and he hrd s he echnologcal cach-up componen derved from resolvng he dfferenal equaon (3). A & A = g c e v.(1 b (0)) v. (1 b (0)) (5) Insuons wll have an mpac hrough X (0), hrough c, and poenally hrough v. We now need o negrae equaon (5) no he growh equaon. The growh model we hen develop s he augmened Solow model enrched o ake no accoun he heerogeney of echnologes and he conrbuon of nsuons. Wh n denong he populaon growh, d K H he physcal and human capal deprecaon rae, s and s he fracon of oal ncome nvesed n physcal and human capal respecvely, Appendx 1 esablshes he followng growh equaon: 2 In a recen paper, Benhabb and Spegel (2003) refers o hs dffuson process, orgnang n he Nelson- Phelps model, as he confned exponenal dffuson process. 11
CEPII, Workng Paper No 2004-02 Log y ( ) Log y (0) 1 e = β. g c f y (0) 1 e. Log a (0). A (0) v, ( b (0)) ben β. K a H b ( s ).( s ). Log a ( n d g c ) b 1/(1 a b) where β = 1 a b).( n d g c ) s he usual speed of condonal convergence and he las erm ( v. 1 1 (1 b (0)). e 1 f = v, ( b (0)). Log 1. v (7) b (0) b (0) s he conrbuon of echnology dffuson o growh: s posvely relaed o he speed v and o he dsance o he echnologcal froner. Equaon (6) s o be compared o he augmened-solow growh equaon whch s exacly he same as f we assume ha nal oal producvy level A (0) s dencal across counres, ha every counry s a he froner ( b = 1) and ha here s no long erm TFP-growh dfferences ( c = 0). The growh process s herefore he resul of four dsnc forces: he adjused absolue convergence hs source of convergence s lessened here because of overall producvy level dfferences, herefrom he adjecve adjused -, a second convergence componen comng from echnologcal cach-up, he usual non-convergence semmng from dfferences n long erm pahs due o dfferen nvesmen raes and an addonal dvergence force comng from long erm TFP-growh dfferences. The man reason for consderng ha all counres share he same echnology les n he dffculy o observe relave echnology levels. Wha s only needed here, as shown by equaon (6), s he relave level of nal producvy and here s one pece of nformaon from whch we can esmae. Indeed equaon (2) llusraes ha nal oupu per worker s ceranly lnked o nal producvy. We wll assume ha a par η of nal ncome raos can be explaned by nal producvy raos, he complemen 1 η beng explaned by capal dfferences, and herefore we formally wre: (6) a (0) A (0) A ben (0) y = y ben (0) (0) η (8a) Hence η s characersed by: Cov η = ( Log a (0), Log y (0)) Var ( Log y (0)) (8b) Klenow and Rodrguez-Clare (1997) use esmaes of socks of physcal and human capal o nfer producvy levels and assess ha producvy dfferences accoun for half or more of level dfferences n 1985 GDP per worker (p.75). For nsance accordng o 12
Technology dfferences, nsuons and economc growh : a condonal condonal convergence equaon (2), wh η = 0. 75 and a = b = 1/ 3, he oupu per capa rao beween he USA and Uganda of 48 n 1980 s explaned by a conrbuon from he producvy level rao KH of 2.7 and a capal aggregae, Z, per capa rao of 6,100 nsead of 110,700. For sure, he exreme smplfcaon n equaon (8a), whch has hough he mer of hghlghng he poenal correlaon beween nal producvy and nal oupu per worker, and of avodng he man pfall of he cross-secon approach, s a srong ad hoc assumpon, bu akng η = 0, as s done n mos of he cross-secon leraure, s as srong and ceranly a far more naccurae ad hoc assumpon. 4. DATA 4.1. General daa The me frame s he perod from 1980 o 2000. For daa enerng he radonal Solow model, we use Penn World Table 6.1. 3 The per capa oupu s he real GDP per capa a Purchasng Power Pary, chan seres. For he nvesmen rae K s, we use he average over he perod of he nvesmen share of real GDP (varable k n he daabase). As n MRW, he proxy for he rae of human capal accumulaon s he percenage of he workng-age populaon n second-level educaon. Ths percenage s consruced by mulplyng he gross secondary enrollmen rae (World Developmen Indcaors) by he percenage of he workng-aged populaon aged 15 o 19. The average of hs varable over he years 1980, 1990 and 2000 s named HCFLOW. 4.2. Insuons daabase We use an orgnal daabase on nsuons developed by he MINEFI, well descrbed and analysed n Berheler, Desdogs and Ould-Aouda (2003). Ths daabase focuses prmarly on emergng counres as 44 of he 51 counres are developng counres and also ncludes a conrol group of developed counres. Daa was gahered from a very dealed quesonnare: for each counry, 330 ems aggregaed n 115 ndcaors were made avalable. As an example he ndcaor relaed o he effcency of publc polcy lnked o he qualy of he ax sysem s bul from four ems assessng he mporance of he black marke, he mporance of fraud n he formal economy, n cusoms and he capacy of he Admnsraon o mplemen ax measures. Imporanly, Berheler e al have esablshed he robusness of he daabase by hghlghng s convergence wh varous nsuonal daabases (World Bank, Fraser Insue, Economs Inellgence Un, Polcal Economc Rsk Consulancy, IMF, Transparency Inernaonal among ohers) all coverng 30% of he sock varables from he quesonnare. The lmaons of he daabase are wofold: a farly small number of counres and a me perod lmed o a sngle pon n or close o year 2000. We wll consder ha he nsuonal profle for a gven counry s sable over he 3 Alan Heson, Rober Summers and Bena Aen, Penn World Table Verson 6.1, Cener for Inernaonal Comparsons a he Unversy of Pennsylvana (CICUP), Ocober 2002. 13
CEPII, Workng Paper No 2004-02 perod of he growh analyss. Ths rases posonous quesons of endogeney snce a counry experencng a favourable economc developmen ncreases s chances of developng beer nsuons and he causaly beween growh and nsuons mgh be reversed. Therefore, hose ndcaors ha are oo suspcous n hs respec are excluded and hs endogeney ssue s economercally addressed n secon 6.4. For he purpose of he sudy, we cluser he remanng 94 ndcaors no fve groups represenng dsnc aspecs of he nsuonal profle. These fve nsuonal domans cover produc marke, labour marke, fnancal sysem, nnovaon and a general headng for all oher ndcaors. Each doman s analysed hrough a facor analyss whch aggregaes he orgnal ndcaors and provdes robus nsuonal varables synhessng mos of he nformaon n he daabase. 4.3. Facor Analyss The chosen approach s very close o ha of Ncole, Scarpea and Boylaud (1999) developed o buld produc marke regulaon ndcaors. I dffers only n he facors aggregaon mehodology and deals are found n Appendx 2. The dea s smple: for each doman, we run a facor analyss and selec he number of relevan facors accordng o usual ess. The axes are hen roaed n order o enhance nerpreaon of facors and an aggregaed ndex s bul for each doman. Unsurprsngly hese aggregaed ndces are exremely correlaed wh each oher so ha valuable dealed nformaon s los n he aggregaon process. As a consequence we preferred o use, as explanaory varables for he role of nsuons on growh, he facors ha conaned enough nformaon and were easly denfed. Therefore we wll focus on varables whch rean more han 25% of her doman varance. Table 1 summarses he varables whch passed hese ess. 4 In addon o he aggregaed ndces, fve oher varables are seleced: CORRUPTION, an ndcaor of he non-corrupon level, CONTRACT, a varable referrng o he conracual mporance n he labour marke, BANKRULES, assessng he qualy of bank regulaon and prudenal rules, R&D-CAPRISK, an ndcaor of R&D/nnovaon effor and favourable capal rsk sysem and INTPROP, a varable lnked o he proecon of nellecual propery rghs (IPRs). 5 If we lower our hreshold o 20% of he varance, hree of he sx hen added varables are of parcular neres snce hey represen a very dsnc aspec of he produc marke nsuons: TRADECOMPET, an ndcaor of he nernaonal and domesc procompeve envronmen, LARGECO, he share of large companes n he dsrbuon secor, whch may be an ndcaor of effcen scale, and NEWENTRY, represenng low barrers o new enry. Chars 1 and 2 represen he counres n a wo-dmensonal plan for he nsuonal varables ha wll be of parcular neres below: CORRUPTION, R&D- CAPRISK and PRODINDEX. Char 1, for nsance, ndcaes a srong posve correlaon 4 As always wh facor analyss, he advanage of such a mehodology s ha he fnal ndcaors are compued as objecvely as possble based on he daa. The man nconvenence s almos dencal. Indcaors are daa based whch means ha hey depend upon he specfc sample and f we add new counres, ndcaors for counres n he orgnal sample wll be alered. 5 For he labour marke and fnancal sysem domans, because he general ndex s farly nangble, he facors only wll be used (see Appendx 2, $ 3.3 and 3.4). 14
Technology dfferences, nsuons and economc growh : a condonal condonal convergence beween he non-corrupon varable and he nal GDP per capa, bu also shows ha he corrupon ndex adds valuable nformaon, as a gven GDP per capa level may be assocaed wh a wde range of corrupon levels. < Table I, Chars 1 and 2 > 5. FROM THE INSTITUTIONAL DATABASE TO THE ECONOMETRIC SPECIFICATION 5.1. Insuonal varables The ndcaors n Table 1 are he canddaes for he compuaon of our nsuonal varables. These ndcaors are ndfferen o a lnear ransformaon and we lnearly normalse hem so ha hey mee he consrans embedded n he model presened n c X (0) dff secons 2 and 3 as follows. We defne I, I, I he nsuonal ndcaors ha are respecvely lnked o long erm TFP-growh, nal effcency n he use of echnology and echnology dffuson. As by defnon c ben equals 0, he long erm TFP-growh defc o he benchmark, c, s defned by: c c ben c = c.( I I ) (INST 1) where c s a parameer o esmae and whch measures he mpac of hs parcular c nsuon, I, on long erm TFP-growh. As regards he nal use of echnology, equaon (8a) descrbes how we nfer he relave levels of nal producvy a (0) whch s broken down no b 0). x (0). So, by makng he furher assumpon ha he lowes effcency ( level n he sample, x (0), equals half (n logarhm) he mnmum producvy level (0) mn mn a (n logarhm), we deduce x (0) for each counry and hen b (0). 6 7 Formally, b (0) = a X(0) ben (0) / x (0) X(0) x (0) = 1 ζ.( I I ), wh ζ such ha x mn (0) = a mn (0) 1/ 2 (INST 2) Fnally o es wheher nsuons mpac he speed of echnology dffuson, we wre: v dff dff = v.( I I mn ) w (INST 3) 6 For a varable z, we defne zmn = mn z. 7 Ex pos, we assess ha he bes f s obaned beween 50% and 60% 15
CEPII, Workng Paper No 2004-02 where v s a parameer o esmae and measures he mpac of hs nsuon, dffuson speed, and w s a consan. 5.2. Economerc specfcaon dff I, on he The specfcaon s drecly derved from equaon (6) by lnearsng n c. Wh he ~ addonal noaons for he speed of convergence β = (1 a b).( n d g) and ~ β ρ = 1/.(1 e ), we oban: Log y ( ) Log y (0) y (0) = ρ. Log a (0). A c. Z f v, ben ( b (0)) ρ. Log (0) u K a H ( s ).( s ) ( n d g) b a b 1/(1 a b) g (9) where K a H y (0) a b ρ ( s ).( s ) Z = 1 (1 a b).(1 ρ. ). Log. (1 ρ. ). Log a (0). A (0) 1 a b n d g ( n d g) ben b a b and u s he resdual. The reamen of demographc growh s somemes confusng. Ceran auhors have been nconssen, reang ρ as a consan bu keepng, he demographc varable, n counry-dependen elsewhere n he equaon. We chose o run he esmaes eher by conssenly keepng n counry dependen everywhere or by consderng ( n d g) a consan everywhere because of he smplfcaons enals bu a he expense of neglecng any demographc mpac. Takng equaon (INST 1) no equaon (9) leads o: Log y ( ) Log y(0) y (0) = ρ. Log a (0). A c.( I c ben I c ). Z ben K a ( s ).( s ρ. Log (0) ( n d g) f v, ( b (0)) u H b ) a b 1/(1 a b) g (10) wh b (0) and v gven by (INST2) and (INST3). We recall ha nsuons nervene hrough x (0) from whch we deduc b (0), hrough c and poenally hrough v. The consan ( d g) s fxed o a realsc 0.06 and resuls are no much mpaced f s n he (0.05,0.08) range. When consdered as a consan, ( n d g) wll be fxed a 0.08. 16
Technology dfferences, nsuons and economc growh : a condonal condonal convergence 6. RESULTS Ths secon sars wh he esmaes of he radonal Solow model and s MRW exenson (6.1). In order o faclae he undersandng of he conrbuons of he nsuons whn he specfcaon of equaon (10) and o denfy he role of educaon separaely, he resuls are successvely presened whou human capal (6.2) and ncludng human capal (6.3). Then, addresses economerc ssues (6.4), provdes more quanfcaon of he role of nsuons (6.5) and fnally dscusses he mpac of nernaonal rade (6.6). 6.1. Sarng pon esmaes Because of mssng daa, our sample s lmed o 44 counres. As a sarng pon, applyng MRW approach o our daa ( a ( 0) = b (0) = 1, c = 0 n equaon (10) ), we es he followng specfcaon: Log y( ) Log y(0) a s b s = ρ. Log y (0). ρ. Log. ρ. Log ce u 1 a b n g d 1 a b n g d K H An esmaed b sgnfcanly dfferen from zero dsngushes he augmened Solow model from he exbook verson. The speed of convergence mpled by he nal ncome ~ coeffcen ρ s here β = (1 a b).( n d g), lower han n he Solow model. Table II presens he resuls for he Solow model n he frs wo columns and for he augmened verson n he las wo. Columns (2) and (4) dffer from columns (1) and (3) respecvely by negang he populaon growh dfferences across counres. Dsapponngly, boh specfcaons have poor explanaory power when we nclude each counry s demographc evoluon because of he resrcon mposed lnkng he speed of convergence o ndvdual demographc growh. 8 If we lm ourselves o columns (2) and (4), we fnd agan he man resuls of MRW, hanks economercally o he posve correlaon beween nal oupu and he human capal varable: he esmaed physcal capal share s closer o he expeced (0.3-0.4) range, human capal accumulaon plays a sgnfcan role, he common emprcally esmaed 2% speed of convergence s conssen wh he model. < Table II > 8 Durlauf and Johnson (1995, Table II) showed ha MRW sample ressed o he approprae resrcon. 17
CEPII, Workng Paper No 2004-02 6.2. Model esmaes whou human capal To ncorporae he role of nsuons, we sar wh he nal effcency level x (0) whch 18 X(0) we mos smply derve from he general doman of nsuons, usng as I eher he aggregaed ndex GENINDEX or he frs facor CORRUPTION. We hen nfer he nal producvy level and he pure echnology componen followng equaons (8) and (INST2). Table III gves hese esmaed levels, usng η = 0. 75. The mpled nal dsance o he echnologcal froner s very close wheher we use one ndcaor or he oher, confrmng ha mos of he nformaon n hs doman s ncluded n he non-corrupon varable. Because of he sraghforward reason why corrupon may nduce weak effcency, we wll lm ourselves o I X (0) = CORRUPTION from now on. < Table III > Insuons ha mos lkely explan long erm producvy dfferenals have now o be c chosen. The core equaon (10) s esed wh I beng deermned accordng o (INST1) by he global ndex of he produc and nnovaon domans, and he man ndcaor n he labour marke and fnancal sysem domans successvely, keepng he rae of echnology dffuson v consan across counres. We assess he qualy of he resuls, summarsed n Table IV, accordng o hree crera: sgnfcance of he parameer esmaes, explanaory power, physcal capal share esmae closer o heorecal predcon. Along hese lnes, he resuls are very close o one anoher excep wh BANKRULES where he esmaes are less precse (remember ha varable defnons are found n Table I). The labour marke varable CONTRACT s relaed o boh lmed chld labour and small nformal economy. Because of he very lkely endogeney of hs varable, wll be dropped for furher analyss. The man prelmnary nferences are: frs, sgnfcan esmaes excep for he echnologcal cach-up annual speed whch s esmaed a around 5% bu s weakly sgnfcan, casng doub on uncondonal echnology dffuson; second, nsuons maer for long erm TFP-growh, n parcular he global qualy of he compeve produc marke and he nnovaon-frendly envronmens have a sgnfcan mpac on long erm growh enalng poenal dvergence; hrd, he explanaory power s very encouragng especally compared o prevous resuls shown n Table II; fourh, despe havng aken no accoun echnology heerogeney, capal share esmaes are oo hgh rasng he lkelhood of msspecfcaon and bases; sxh, and lnked o he ffh, here effecvely s condonal convergence (here s also condonal o sharng he same echnology and effcency) a a speed of around 3%, bu oo low o be n lne wh he underlyng Solow model. < Table IV > As suggesed n sub-secon 5.1, one channel hrough whch nsuons may nfluence growh s by promong or hnderng echnology dffuson. To es wheher he dffuson s dff condonal o nsuons, we use equaon (INST3) akng as I he nsuons mos lkely o do he ask n he nnovaon doman, R&D-CAPRISK and INTPROP, and resrcng ourselves o usng PRODINDEX or INNINDEX for he nsuon mpacng
Technology dfferences, nsuons and economc growh : a condonal condonal convergence long erm TFP-growh (Table V). The resuls are very sensve o he nsuon o whch we condon he cach-up. Proecon of IPRs for nsance (columns (2) and (4)) does no speed up echnologcal cach-up: s mpac s nsgnfcan and f anyhng negave, harmng convergence o he froner of counres whch ry o proec IPRs. On he conrary, nsuons whch favour R&D, nnovaon and capal-rsk do have a sgnfcan and srongly posve nfluence on cach-up speed whch hghlghs he condonal naure of echnology convergence. Columns (1) and (3) also exhb a capal share esmae of around 0.50, whch mples a Solow-ype condonal convergence speed slghly above 4%, and dmnshed he rsks of based esmaes. Unforunaely, a hgher speed of convergence does no necessarly mean ha a poor counry converges faser han wha s usually esmaed, bu ha would, had he same overall producvy (A) as he benchmark counry. Moreover, specfcaon n columns (1) and (3) have very sasfyng explanaory power as assessed by he adjused R-square above 0.66. As he consan par of he dffuson speed w s nsgnfcan, we dscard and re-esmae he specfcaon of column (1) o reach column (5), whch ends up beng he base equaon of hs sudy whou human capal, and column (6) by furher relaxng he consrans on he demographc growh varables n. Resuls n hese las wo columns are very smlar, excep ha akng each counry s demographc growh no accoun leads o a less precse esmae of he long erm TFP-growh heerogeney parameer c bu o a capal share even closer o he one-hrd sandard. Focusng on column (5), he cenral esmae of 0.0064 for v means ha he annual dffuson speed ranges from 0% for Zmbabwe o 12.4% for Tawan. Usng a dfferen mehodology, Bloom e al. (2002) found an uncondonal echnologcal convergence of 2% a year. Wh he esmaed value of c n column (5) and (3), he wors esmaed long erm TFP-growh performance n he sample s Syra wh an annual spread of respecvely 3.3% and 4.0% compared o he USA. We develop he quanfcaon of he mpac nsuons have on economc performance furher n subsecon 6.5. < Table V > The general ndex PRODINDEX comes from he aggregaon of facor scores n he produc marke doman, he hree mos prevalen beng TRADECOMPET, LARGECO and NEWENTRY. Based on he core esmae of column (5), Table V, whch of hese hree aspecs of produc marke compeon sgnfcanly conrbues o long run TFP-growh? The frs wo varables are no found sgnfcan whereas NEWENTRY leads o smlar esmaes: eher low barrers o enry alone or s combnaon wh domesc and nernaonal compeon (TRADECOMPET) and large sze of frms (LARGECO) explan he nfluence of a more compeve produc marke on TFP-growh. The esmaons have been conduced so far by consderng ha hree quarers of he dfferences n he nal oupu per capa (n logarhm) could be explaned by dfferences n he oal level of nal producvy, meanng n oher words ha η was fxed a 0.75. We now wan o es how our resuls are sensve o he choce of a gven value for η (he lower he η he closer he nal producvy levels beween counres) and Table VI provdes some comparave esmaes,. Based on he hree crera defned above o assess 19
CEPII, Workng Paper No 2004-02 he qualy of our esmaes sgnfcance, explanaory power, conssen capal share -, he conclusons are clear-cu: allowng for producvy heerogeney defnely mproves he resuls as he esmaes wh he lower η s, by far, less good whaever he crera. Acually, columns (3 o 5) srongly suppor our approach, by suggesng ha he share of dfferences n nal ncome per capa due o he dfferences n oal nal producvy levels s ceranly greaer han one-half, whch s conssen wh Klenow and Rodrguez- Clare s analyss. Moreover, hese resuls mply ha nsuons maer for he effcency n he use of echnology, approxmaed by a non corrupon ndex, and omng hs mpac by consderng dencal effcency and echnology s msplaced and leads o based esmaes. 6.3. Human capal < Table VI > Resuls of he prevous sub-secon are subjec o poenal bases due o omng poenally mporan varables lke human capal. There are many ways n whch he educaon level can nfluence growh. Educaonal achevemen may have an mpac on he effcency n he use of a gven echnology (our X varable), on long erm TFP-growh or on he rae of human capal accumulaon. In he frs wo cases, he sock of human capal, HCSTOCK, s probably he varable of neres and we use he average years of second-level schoolng for he populaon aged 25 and over n 1980 from he Barro-Lee daabase. In he hrd case, we wll use he human capal flow varable à la MRW called HCFLOW. To es wheher educaon has an nfluence over he effcency level n our framework, X equaon (10) s smply esmaed by usng I = HCSTOCK. The resuls ndeed sugges ha he sock of human capal a he begnnng of he perod s sgnfcan n explanng nal effcency (Table VII, column (2)). In fac, boh he nsuonal varable CORRUPTION, a proxy for he non-corrupon level, and he sock of educaon HCSTOCK probably maer for effcency. These wo varables exhb a posve lnear correlaon coeffcen of 0.55 and by dchoomy we show n column (3) ha he opmal combnaon s acheved wh I X (0) (0) = CORRUPTION 4. HCSTOCK. MRW assers ha akng no accoun human capal as anoher accumulang facor enables one o valdae he Solow assumpon of dencal producvy across counres. 9 We now show ha he role of human capal hey pu forward does no conflc wh our framework bu raher complemens. Table VIII reproduces he resuls where we nfer ha human capal accumulaon adds valuable nformaon n he growh process whou much alerng he basc parameers esmae of secon 6.1. The physcal capal share esmae s now very close o he common sense value, he jon effec of emphassng he role of human capal accumulaon and producvy heerogeney. 9 Or a leas uncorrelaed wh he regressors. 20
Technology dfferences, nsuons and economc growh : a condonal condonal convergence < Tables VII VIII > Fnally, deparng from he reamen of human capal as jus anoher facor and neglecng complex ssues of accumulaon, we smply wonder f he sock of human capal has an nfluence on long erm TFP-growh and herefore we nclude he sock measure HCSTOCK n addon o he nsuonal varable PRODINDEX n our c varable. The economercs do show a posve conrbuon of human capal sock o producvy-growh and weaken somewha he sgnfcance of our nsuonal varable. However, hese resuls are doubful because he esmaed physcal capal share s rased back o around 70%. To summarse our resuls on he role of educaon, we fnd ha he sock of human capal plays an mporan par n explanng effcency, quanavely equvalen o he one comng from oher nsuonal varables lke non-corrupon (see below 6.5), and ha he flow of human capal has explanaory power n he deermnaon of long-erm growh pah whch complemens he analyss we have run so far. However he lnk beween he sock of human capal and he long erm TFP-growh s confusng. 6.4. Economerc ssues To reach he above resuls we made wo srong assumpons. The frs s ha he nsuonal qualy ha we measure has been deemed o be sable over he perod under sudy. There s no much we can do here as he nformaon s only avalable for around year 2000, oher han beng on he robus srucural dmenson of hese nsuons. Obvously, hs s an approxmaon as he nsuonal envronmen mgh have changed sgnfcanly for some counres beween 1980 and 2000. The second ssue refers o he poenal endogeney of he nsuons, and reverse causaly from growh o nsuons ha may ensue. To llusrae why hese concerns may no be oo problemac, he R&D- CAPRISK varable, for nsance, can be expeced o capure an exogenous srucural aspec of nnovaon faclang echnology ransfers. Indeed, s manly deermned by he conrbuon of fve ems: suppor o R&D and nnovaon from he Mnsry of Research, from he publc or prvae research cenres and from he echnopoles (hree ems), he qualy of he relaonshps beween unverses, research cenres and companes, and fnally he mporance of a fnancal scheme favourng capal rsk. We address hese pons and check he robusness of our resuls n hree seps. Frsly, smple descrpve sascs reveal ha our nsuonal varables capure deep and lasng srucural componens of he counres characerscs. The eases check s our non-corrupon varable snce oher comparable measures exs. CORRUPTION exhbs a Pearson lnear correlaon coeffcen of 82% wh Knack and Keefer s measure for he 1980-1989 perod, called here KK80-89. 10 Along hose lnes and wh he dea of nsrumenng he nsuonal varables ha ener our core specfcaon, CORRUPTION, PRODINDEX and R&D-CAPRISK, we regress hese hree varables wh poenal nsrumens: he Knack and Keefer s non-corrupon measure (KK80-89), he logarhm of oupu per capa n 1980 (LOGY80), he annual 10 The source s Easerly and Levne daabase: hp://www.worldbank.org/research/growh/ddeale.hm. 21
CEPII, Workng Paper No 2004-02 average growh rae of oupu per capa beween 1970 and 1980 (LAGDLOGY), he ferly rae n 1980 (World Developmen Indcaors, FERT80) and n addon he K logarhm of he nvesmen rae s (LOGSK) for he PRODINDEX regresson and he R&D compose ndcaor for 1980-1983 (World Developmen Indcaors, R&D80-83) for he R&D-CAPRISK regresson. Despe hs lmed se of regressors, he hgh levels of 2 adjused- R obaned (76%, 62%, 70% respecvely for CORRUPTION, PRODINDEX and R&D-CAPRISK) sugges ha consderng hese nsuonal varables exogenous n he equaon explanng growh beween 1980 and 2000 should no be problemac. Secondly, we formally assess he endogeney of he nsuonal varables for our wo core equaons, wh and whou human capal, hrough a Hausman es. For he specfcaon whou human capal, as we are here concerned wh he endogeney of our varables and no wh he bas comng from omed varables (human capal), one precauon has o be aken n he choce of he nsrumens: human capal measures are no vald nsrumens as hey are mos lkely correlaed wh he resduals. Table IX presens he resuls wh columns (1) and (4) beng our base OLS esmaes whou and wh human capal respecvely columns (1) and (4) of Table VIII. Unsurprsngly, gven he economerc relaons denfed above, nsrumenal varable esmaes (IV) and OLS do no dffer sgnfcanly, even when nsrumenng he nvesmen rae varables. Hausman es sascs mply ha we should lm ourselves o he OLS. In oher words, our nsuonal varables reflec robus srucural componens of he counres n 1980. Obvously, one can hnk of cases, lke Chna or Ireland, where rejecng endogeney of nsuons qualy for growh esmaes over he las 20 years seem dubous. However, he ess show ha, n he sample overall, he deep and lasng characerscs of he nsuons seleced oversep endogeney concerns. Fnally, anoher way o es he robusness of our assumpons s o compare our esmaes wh esmaes usng a dfferen me frame, he dea beng ha f nsuons have grealy evolved over he perod, hs wll nduce very dfferen esmaes for a longer perod. The resuls, no presened here, ndcae ha esmaes for he 1970-2000 perod prove o be very close o our base resuls for he 1980-2000 perod, whch renforces he prospecs of our nsuonal varables capurng srucural parameers, bu non-endogeney s no rejeced as clearly as before snce he Hausman es χ -square probably s only 0.16 whou human capal. < Table IX> The presence of oulers has also been esed. Ther excluson does no aler he man resuls and deals wll be provded upon requess. 6.5. Quanfyng he mpac of nsuons furher Ths s beyond he scope of hs sudy o sele he debae concernng he conrbuon of human capal o growh as jus anoher accumulang facor. Therefore, he followng resuls are presened wh he specfcaon eher whou or wh human capal. From he specfcaon whou human capal, he esmaed annual growh can be broken down no four componens as dealed n Appendx 3: adjused (o accoun for nal producvy level dfferences) absolue convergence, nvesmen rae conrbuon, long-erm TFP- 22
Technology dfferences, nsuons and economc growh : a condonal condonal convergence growh and echnology dffuson. 11 When one akes he dfferences o he mean n he sample, Table X s obaned: for nsance, Chna s annual growh-rae over he perod was 4.2% above he average, he esmaon of he model s 4.6%, broken down no 1.0% comng from adjused absolue convergence due o decreasng capal reurns, 0.5% due o dfferences n nvesmen rae, -0.8% semmng from long-erm producvy-growh dfferences explaned here by dfferences n produc marke effcency dermenal o Chna and 4.0% due o echnology dffuson, he man drver of Chna s over-performance. 12 < Table X > Furhermore, n order o measure he conrbuon of each componen globally, s sandard devaon s compued and repored n Table XI. In ha sense, physcal capal nvesmen rae and echnology dffuson conrbue mos o he dfferences n annual growh beween counres havng a dsperson wce as grea as he one of he adjused absolue convergence componen and almos hree mes he one of long erm TFP-growh. When human capal s ncluded, he conrbuon of s accumulaon rae comes hrd n explanng growh dsperson across counres n beween physcal capal nvesmen rae and echnology dffuson on he one hand and long erm TFP-growh and adjused absolue convergence on he oher. Based on our esmaes, we can calculae wha coss a counry n erms of annual growh rae no o have he bes nsuons n he sample. Table XII ndcaes ha Ngera loses he mos from he poor qualy of nsuons wh a counerfacual growh loss of 6.1% annually, broken down no 2.3% due o poor effcency (CORRUPTION), 2.7% because of a weak echnology dffuson (R&D-CAPRISK) and 1.1% n long erm TFP-growh loss (PRODINDEX). For a comparson, Arad and Sala--Marn (2003) denfy egh deermnans of Afrca s economc underperformance o OECD counres, whch oals up o an annual growh defc of around 8%. In erms of effcency, he loss comng from a weak sock of human capal compared o he USA averages 1.3% annually, o be compared o 0.9% for he non-corrupon varable. However, he dspersons are comparable and hs s due o he over-performance of he USA n erms of educaon: he counry ha comes n second, Japan, loses 0.6% annually o he USA because of human capal sock dfferences. The medan annual loss over he sample due o he hree major nsuonal aspecs denfed oals up o 2.4%. Because he sample s very heerogeneous wh boh counres among he rches and among he poores, he nsuons measures, resulng from he daa based facor analyss, canno dfferenae much beween he mos developed counres, and 11 The producvy-growh componen s no jus c as calculaed a he end of secon 6.1 - -3.3% for Syra bu raher as apparen from equaon (10) c. Z. 12 For he USA, 1.7 percenage pon of annual growh s no accouned for by he model whou human capal. Wh he base esmaon wh human capal, USA s growh s correcly esmaed, bu he USA do no appear a he echnologcal froner n 1980, (where Argenna and Germany sand): s nal oupu per capa s explaned by he srong conrbuon of he sock of human capal o he effcency raher han by s hen echnologcal performance. USA s economc growh over-performance comes hen from he conrbuon of human capal o he seady-sae and from echnologcal cach-up. 23
CEPII, Workng Paper No 2004-02 he counerfacual loss s calculaed o be under 0.3% for he rches counres. Japan s an excepon wh a loss of 1.4% manly due o a weak produc marke rankng. Schemacally, Lan Amercan counres suffer a loss n he 1.5%-2.5% range, Norh-Afrcan n he 2.5%- 4%, Sub-Saharan above 4% and Asan and Easern European are more wdespread ou. The addve average annual growh losses, over he counres n he sample, due o poor overall effcency (as measured by CORRUPTION), o non-compeve produc marke (PRODINDEX) and o unfavourable envronmen for ransferrng echnology (R&D- CAPRISK) are respecvely 0.9%, 0.7% and 0.8%. Fnally, dreamng of a world n 1980 where each counry would have enjoyed a qualy of s nsuons a he maxmum over he sample, we buld Table XIII where we show dfferen measures of nequaly. By comparng he las wo rows, we can see for nsance ha he world would be 52% rcher and he unweghed Gn ndcaor would have fallen from 0.42 o 0.32 durng he perod, nsead of he acual rse o 0.48. 6.6. Inernaonal rade < Tables XI - XII - XIII > Appendx 4 shows ha, conrollng for nsuons, rade does no conrbue sgnfcanly o long erm TFP-growh. Trade conrbuon becomes sgnfcan only f expors are mosly manufacures and he counry suffcenly developed, suggesng ha he mpac of rade s non-lnear and depends on he specalsaon of expors. As s beyond he scope of hs sudy o overcome he ssue rased by he endogeney of rade, hs las resul s enave. 7. CONCLUSION The facor analyss of he nsuonal daabase has pu forward varables ha poenally mpac growh. Once, he role of producvy dfferences n he growh analyss has been hghlghed, nsuonal qualy s shown o maer grealy n explanng growh pahs, where convergence appears o be no only condonal on facor accumulaon raes bu also on sharng he same echnology and qualy of nsuons. In fac, we denfed ha he growh oucome s he resul of wo convergen forces, decreasng reurns o capal and echnology dffuson, one dvergen force, long run TFP-growh dfferences due o nsuonal heerogeney, n addon o he non-convergence semmng from he dfferences n facor accumulaon raes. The greaes mpacs of nsuons channel hrough he effcency level n he use of echnology, well accouned for by our non-corrupon varable, and hrough he dffuson of echnology. Uncondonal dffuson appears o be weakly sgnfcan. However, dffuson s sgnfcanly fosered by nsuons favourng R&D, nnovaon and capal-rsk fnancng, whereas IPRs proecon, for nsance, f anyhng, hampers dffuson. In oher respecs, a pro-compeve produc marke and an nnovaon-frendly envronmen add a sgnfcan conrbuon o long run TFP-growh dfferences bu of a lesser magnude. If hese resuls are correc, hey enal ha developmen polcy for he leas developng counres s necessary and should prmarly arge echnology dffuson. Inernaonal 24
Technology dfferences, nsuons and economc growh : a condonal condonal convergence suppors should focus on counres bes prepared for growh, for nsance hose havng a low corrupon ndex, and am naurally a developng human capal hrough he healh and school sysems, and also a brngng experse o expede he dffuson of echnology and knowledge. Oher measures are secondary n mporance and should evenually appear a a laer sage n he developmen process. BIBLIOGRAPHY Acemoglu D., P.Aghon and F.Zlbo (2002), Dsance o froner, selecon and economc growh, NBER Workng Paper 9066. Arad E. and X.Sala--Marn (2003), The economc ragedy of he XXh cenury: growh n Afrca, NBER Workng Paper 9865. Bassann A. and S.Scarpea (2001), Does human capal maer for growh n OECD counres? Evdence from PMG esmaes, OECD Workng Paper 282. Benhabb J. and M.M. Spegel (2003), Human capal and echnology dffuson, FRBSF Workng Paper 2003-02. Benhabb J. and M.M. Spegel (1994), The role of human capal n economc developmen Evdence from aggregae cross-counry daa, Journal of Moneary Economcs 34(2): 143-173. Berheler P., A.Desdogs and J.Ould Aouda (2003), Insuonal profles Presenaon and analyss of an orgnal daabase of he nsuonal characerscs of developng, n ranson and developed counres, Drecon de la Prévson, Workng Paper. Bloom D.E., D.Cannng and J.Sevlla (2002), Technologcal dffuson, condonal convergence and economc growh, NBER Workng Paper 8713. Casell F., G.Esquvel and F.Lefor (1996), Reopenng he convergence debae: a new look a cross counry growh emprcs, Journal of Economc Growh 1: 363-389. Durlauf S.N. and P.A.Johnson (1995), Mulple regmes and cross-counry growh behavour, Journal of Appled Economercs 10(4): 365-384. Easerly W. and R.Levne (2002), I s no facor accumulaon: sylzed facs and growh models, Cenral Bank of Chle, Workng Paper 164. Hall R.E. and C.I.Jones (1996), The producvy of naons, NBER Workng Paper 5812. 25
CEPII, Workng Paper No 2004-02 Hall R.E. and C.I.Jones (1999), Why do some counres produce so much more oupu per worker han ohers?, Quarerly Journal of Economcs, February: 83-116. Klenow P.J. and A.Rodrguez-Clare (1997), The neoclasscal revval n growh economcs: has gone oo far?, NBER Macroeconomcs Annual 12: 73-103. Mankw G.N, D.Romer and D.Wel (1992), A conrbuon o he emprcs of economc growh, Quarerly Journal of Economcs 107: 407-437. Ncole G., S.Scarpea and O.Boylaud (1999), Summary ndcaors of produc marke regulaon wh an exenson o employmen proecon legslaon, OECD Workng Paper 226. Prche L. (1997), Dvergence, bg me, Journal of Economc Perspecves 11(3): 3-17. Rodrk (2003), Growh sraeges, NBER Workng Paper 10050. Rodrk (2002), Feasble globalzaons, NBER Workng Paper 9129. Rodrk D., A.Subramanan and F.Trebb (2002), Insuons rule: he prmacy of nsuons over geography and negraon n economc developmen, NBER Workng Paper 9305. Temple J. (1999), The new growh evdence, Journal of Economc Leraure 37(1): 112-156. 26
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table I: Seleced nsuonal ndcaors Doman Varable Inerpreaon %Varance n he doman General CORRUPTION Non-corrupon 36% General GENINDEX Aggregaed ndex 72% Produc marke PRODINDEX Aggregaed ndex 67% Labour marke CONTRACT Lmed chld labour and small nformal 26% economy Fnancal sysem BANKRULES Bank conrol, rules, ransparency 25% Innovaon R&D-CAPRISK R&D effor and capal rsk sysem 37% Innovaon INTPROP Inellecual propery rghs 28% Innovaon INNINDEX Aggregaed ndex 77% General FREEDOM Freedom 21% Produc marke TRADECOMPET Lmed barrers o rade and compeon 21% enforcemen Produc marke LARGECO Large companes n he dsrbuon secor 25% Produc marke NEWENTRY Faclaon of new enry 21% Labour marke UNIONFREED Trade-unon rghs 20% Table II: Sarng pons Solow Augmened Solow n n=0.02 n n=0.02 (1) (2) (3) (4) a (phys.capal) 0.546 (0.048)*** 0.647 (0.029)*** 0.399 (0.147)*** 0.427 (0.100)*** b (human capal) - - 0.163 (0.152) 0.288 (0.106)*** ce 0.210 (0.018)*** 0.173 (0.013)*** 0.213 (0.018)*** 0.162 (0.014)*** 1 Impled β 0.035 0.028 0.034 0.023 2 R -0.05 0.31-0.05 0.36 Observaons 44 44 44 44 1 when akng no accoun he ndvdual demographc growh, convergence speed s calculaed wh average populaon growh across counres. (***) asympoc sgnfcance a 99% level (**) asympoc sgnfcance a 95% level (*) asympoc sgnfcance a 90% level 27
CEPII, Workng Paper No 2004-02 Table III (*): Inferred relave producvy and echnology levels n 1980 GENINDEX a (0) b (0) CORRUPTION a (0) b (0) Algera -4.9 0.32 0.71-4.2 0.32 0.63 Argenna -0.8 0.59 0.92-1.3 0.59 0.91 Brazl 1.7 0.40 0.52 0.3 0.40 0.55 Cameroon -5.1 0.17 0.39-7.3 0.17 0.47 Chle 3.2 0.35 0.42 7.4 0.35 0.33 Chna -6.7 0.10 0.28-1.0 0.10 0.15 Colomba -0.3 0.29 0.44-2.1 0.29 0.48 Coe d'ivore -2.0 0.19 0.33-6.4 0.19 0.48 Egyp -8.2 0.19 0.66-5.2 0.19 0.41 France 8.6 0.81 0.75 6.7 0.81 0.79 Germany 10.6 0.80 0.68 8.7 0.80 0.72 Ghana -3.3 0.11 0.21-0.7 0.11 0.16 Greece 6.6 0.64 0.64 3.3 0.64 0.74 Hong Kong 5.7 0.67 0.70 6.1 0.67 0.67 Hungary 5.9 0.48 0.50 5.3 0.48 0.50 Inda -1.1 0.11 0.17-7.9 0.11 0.32 Indonesa -3.4 0.16 0.30-6.0 0.16 0.37 Iran -7.5 0.28 0.88-3.8 0.28 0.53 Ireland 9.7 0.56 0.49 9.0 0.56 0.49 Israel 5.8 0.62 0.65 5.8 0.62 0.63 Japan 4.0 0.79 0.91 4.4 0.79 0.86 Korea, Rep. 1.5 0.32 0.43 0.3 0.32 0.44 Malaysa 0.0 0.32 0.48 3.3 0.32 0.37 Mexco -0.6 0.45 0.70-3.1 0.45 0.82 Morocco -3.7 0.22 0.44-0.9 0.22 0.33 Ngera -4.6 0.11 0.24-10.2 0.11 0.49 Norway 9.5 0.83 0.74 9.1 0.83 0.73 Paksan -7.3 0.11 0.33-6.6 0.11 0.27 Peru -0.4 0.32 0.49-2.2 0.32 0.54 Phlppnes 0.2 0.24 0.35-4.8 0.24 0.50 Poland 5.9 0.43 0.45 1.9 0.43 0.54 Porugal 2.0 0.52 0.67 4.3 0.52 0.57 Romana -0.3 0.17 0.26-3.2 0.17 0.31 Sngapore 2.8 0.62 0.76 8.6 0.62 0.56 Souh Afrca 1.8 0.47 0.61 0.7 0.47 0.64 Syra -9.5 0.22 0.99-2.9 0.22 0.39 Tawan 1.8 0.37 0.49 3.0 0.37 0.44 Thaland -2.7 0.21 0.38-0.5 0.21 0.30 Tunsa -4.2 0.30 0.62 2.7 0.30 0.36 Turkey -0.1 0.29 0.43-1.4 0.29 0.46 Uganda -0.4 0.05 0.08-4.9 0.05 0.11 Uned Saes 6.8 1.00 1.00 6.3 1.00 1.00 Venezuela, RB -1.4 0.47 0.77-4.5 0.47 0.95 Zmbabwe -3.9 0.20 0.41-6.6 0.20 0.51 (*) calculaons wh η = 0. 75 28
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table IV 1 : Insuons maer for long erm producvy-growh c I PRODINDEX CONTRACT BANKRULES INNINDEX a (phys.capal) 0.624 (0.088)*** 0.531 (0.092)*** 0.619 (0.075)*** 0.629 (0.099)*** c ( I c ) 0.0023 0.0025 0.0017 0.0026 (0.0012)* (0.0008)*** (0.0011) (0013)** w (dffuson speed) 2 2 0.053 0.040 0.049 0.060 (0.033) (0.021)* (0.029)* (0.039)* LogAben (0) 7.78 8.57 7.83 7.80 (0.79)*** (0.56)*** (0.66)*** (0.88)*** Impled β 0.030 0.032 0.021 0.030 2 R 0.56 0.59 0.54 0.57 Observaons 44 44 44 44 1 2 I X (0) = v = w (v=0) CORRUPTION g = 0.02 n = 0.02 (***) asympoc sgnfcance a 99% level (**) asympoc sgnfcance a 95% level (*) asympoc sgnfcance a 90% level Table V 1 - Condonal echnology dffuson c I PRODINDEX INNINDEX PRODIND. PRODIND. dff I R&D- INTPROP R&D- INTPROP R&D- R&D- CAPRISK CAPRISK CAPRISK CAPRISK (1) (2) (3) (4) (5) (6) a (phys.capal) 0.484 (0.076)*** 0.609 (0.094)*** 0.502 (0.083)*** 0.587 (0.103)*** 0.483 (0.061)*** 0.407 (0.048)*** c ( I c ) 0.0015 0.0027 0.0021 0.0031 0.0015 0.0010 (0.0007)** (0.0012)** (0.0007)*** (0.0011)*** (0.0006)*** (0.0006)* v ( dff 2 I ) 0.0064-0.0023 0.0099-0.0034 0.0064 0.0050 (0.0029)** (0.0021) (0.0050)** (0.0024)* (0.0028)** (0.0020)*** w 0.0002 0.068-0.0024 0.081 - - (0.0092) (0.044)* (0.0089) (0.050)* LogAben (0) 8.54 8.07 8.53 8.42 8.54 8.73 (0.38)*** (0.76)*** (0.41)*** (0.72)*** (0.31)*** (0.22)*** Impled β 0.041 0.031 0.040 0.033 0.041 0.046 2 R 0.66 0.58 0.68 0.61 0.67 0.66 Observ. 44 44 44 44 44 44 I X (0) = 2 CORRUPTION dff dff v = v.( I I mn ) w g = 0.02 n = 0.02 excep column (6) where n s unconsraned and β s calculaed wh average populaon (*** asympoc sgnfcance a 99% level (**) asympoc sgnfcance a 95% level (*) asympoc sgnfcance a 90% level 29
CEPII, Workng Paper No 2004-02 Table VI 1 : Producvy dfferences η = 0.05 η = 0.5 η = 0.75 η = 1 η = 1.25 (1) (2) (3) (4) (5) a (phys.capal) 0.642 0.590 0.483 0.366 0.233 (0.065)*** (0.072)*** (0.061)*** (0.064)*** (0.080)*** c ( I c ) 0.0021 0.0021 0.0015 0.0009 0.0003 (0.0006)*** (0.0007)*** (0.0006)*** (0.0006)* (0.0006) v ( I dff ) 0.0205 0.0134 0.0064 0.0041 0.0032 (0.1492) (0.0113) (0.0028)** (0.0012)*** (0.0008)*** LogAben (0) 7.68 7.97 8.54 9.02 9.40 (0.56)*** (0.51)*** (0.31)*** (0.23)*** (0.21)*** 2 R 0.42 0.63 0.67 0.68 0.64 Observ. 44 44 44 44 44 1 I X (0) = CORRUPTION I c = PRODINDEX I dff = R & D CAPRISK g = 0.02 n = 0.02 (***) asympoc sgnfcance a 99% level (**) asympoc sgnfcance a 95% level (*) asympoc sgnfcance a 90% level KH cov( Log ( Z / Y ), Log Y ) (1 a) (1 a b).η I can be shown ha = and herefore f, as he Var( Log Y ) a nuon suggess, hs covarance s posve, here s an upper bound o η whch s ( 1 a) /(1 a b). Table VII: Educaon and nal effcency I X(0) CORRUPTION HCSTOCK CORRUPTION 4.HCSTOCK (1) (2) (3) a (phys.capal) 0.483 0.452 0.528 (0.061)*** (0.059)*** (0.065)*** c ( I c ) 0.0015 0.0021 0.0019 (0.0006)*** (0.0007)*** (0.0006)*** v ( I dff ) 0.0064 0.0042 0.0073 (0.0028)** (0.0017)*** (0.0032)** LogAben (0) 8.54 8.63 8.50 (0.31)*** (0.29)*** (0.35)*** 2 R 0.67 0.66 0.70 Observ. 44 44 44 30
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table VIII: Educaon and human capal accumulaon X(0) I CORRUPTION CORRUPTION CORRUPTION CORRUPTION 4.HCSTOCK 4.HCSTOCK (1) (2) (3) (4) a (phys.capal) 0.483 (0.061)*** 0.302 (0.091)*** 0.528 (0.065)*** 0.361 (0.093)*** b (human capal) - 0.293 (0.099)*** - 0.271 (0.098)*** c ( I c ) 0.0015 0.0017 0.0019 0.0022 (0.0006)*** (0.0008)** (0.0006)*** (0.0009)*** v ( I dff ) 0.0064 0.0093 0.0073 0.0096 (0.0028)** (0.0046)** (0.0032)** (0.0046)** LogAben (0) 8.54 8.65 8.50 8.62 (0.31)*** (0.39)*** (0.35)*** (0.45)*** 2 R 0.67 0.72 0.70 0.74 Observ. 44 44 44 44 31
CEPII, Workng Paper No 2004-02 Table IX: Endogeney ess whou human capal wh human capal OLS (1) IV (2) IV (3) OLS (4) IV (5) IV (6) a (phys.capal) 0.483 (0.061)*** 0.476 (0.100)*** 0.462 (0.159)*** 0.361 (0.093)*** 0.350 (0.108)*** 0.314 (0.272) b (human capal) - - - 0.271 (0.098)*** 0.250 (0.110)** 0.236 (0.250) c ( I c ) 0.0015 0.0024 0.0025 0.0022 0.0025 0.0029 (0.0006)*** (0.0011)** (0.0014)* (0.0009)*** (0.0013)* (0.0014)** v ( I dff ) 0.0064 0.0077 0.0078 0.0096 0.0084 0.0081 (0.0028)** (0.0067) (0.0068) (0.0046)** (0.0062) (0.0059) LogAben (0) 8.54 8.69 8.76 8.62 8.84 9.12 (0.31)*** (0.49)*** (0.49)*** (0.45)*** (0.51)*** (0.76)*** Observ. 44 44 44 44 44 44 2 R 0.67 0.65 0.64 0.74 0.73 0.70 Insrumens KK80-89 LOGY80 LAGDLO GY FERT80 RD80-83 LOGSK KK80-89 LOGY80 LAGDLO GY FERT80 RD80-83 KK80-89 LOGY80 LAGDLO GY FERT80 RD80-83 LOGSK LOGSH HCSTOCK KK80-89 LOGY80 LAGDLO GY FERT80 RD80-83 LOGSH80 HCSTOCK Hausman es Sasc Pr > ChSq Heeroscedascy Whe Sasc Pr > ChSq 12.0 0.60 2.25 0.69 1.25 0.87 24.9 0.21-2.76 1.00-2.77 1.00 Breusch-Pagan S. Pr > ChSq 2.1 0.35 3.8 0.15 32
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table X: Growh componen conrbuons (dfferences o mean) (*) annual growh esmaed annual growh adjused absolue convergence (3) phys. capal nvesmen rae long erm TFP-growh (5) echn. dffuson (1) (2) (4) (6) Algera -1.9% -1.3% 0.0% 0.1% -0.5% -0.9% Argenna -1.9% -2.3% -0.5% 0.0% 0.0% -1.7% Brazl -1.5% 0.1% -0.2% 0.3% -0.1% 0.1% Cameroon -2.2% -2.0% 0.5% -1.9% -0.1% -0.6% Chle 1.0% -0.2% -0.1% 0.2% 0.4% -0.6% Chna 4.2% 4.6% 1.0% 0.5% -0.8% 4.0% Colomba -0.9% -0.8% 0.1% -0.8% -0.1% 0.0% Coe d Iv. -3.5% -2.9% 0.4% -2.7% 0.2% -0.7% Egyp 0.7% -1.9% 0.4% -2.0% -0.1% -0.1% France -0.4% -0.3% -0.8% 1.0% 0.5% -1.0% Germany -0.2% 0.2% -0.8% 0.9% 0.7% -0.6% Ghana -1.5% -1.4% 0.9% -2.6% -0.1% 0.4% Greece -1.0% -0.2% -0.6% 0.7% 0.6% -0.9% Hong Kong 1.7% 0.4% -0.6% 1.2% 0.4% -0.5% Hungary -0.8% 0.9% -0.4% 0.3% 0.6% 0.3% Inda 1.8% 0.9% 0.9% -0.8% -0.3% 1.1% Indonesa 1.2% 0.6% 0.6% 0.1% -0.4% 0.3% Iran -0.1% -1.3% 0.1% 0.5% -1.3% -0.7% Ireland 2.8% 1.3% -0.5% 0.5% 0.7% 0.5% Israel -0.1% 0.7% -0.6% 1.1% 0.4% -0.2% Japan 0.2% -0.7% -0.8% 1.8% -0.2% -1.4% Korea, Rep. 3.9% 2.7% 0.0% 2.0% -0.3% 1.0% Malaysa 1.5% 2.2% 0.0% 1.1% -0.3% 1.5% Mexco -1.4% -1.7% -0.3% 0.3% 0.1% -1.7% Morocco -1.0% 0.7% 0.3% -0.6% -0.1% 1.1% Ngera -4.7% -2.5% 0.9% -1.7% -0.3% -1.3% Norway 0.3% 0.5% -0.8% 1.6% 0.5% -0.7% Paksan 0.7% 1.6% 0.9% -0.9% -0.1% 1.7% Peru -2.3% -0.7% 0.0% 0.3% 0.1% -1.1% Phlppnes -1.8% -0.3% 0.2% -0.2% -0.2% -0.1% Poland -0.7% 0.7% -0.3% 0.5% 0.2% 0.2% Porugal 0.8% 0.7% -0.4% 0.8% 0.5% -0.2% Romana 1.4% 2.2% 0.5% 0.8% -0.2% 1.1% Sngapore 2.8% 2.8% -0.6% 2.6% 0.5% 0.3% Souh Afr. -2.2% -2.1% -0.3% -1.4% 0.2% -0.6% Syra -0.5% -1.2% 0.3% -0.9% -1.0% 0.4% Tawan 3.9% 1.7% -0.1% 0.5% 0.1% 1.3% Thaland 2.6% 2.9% 0.3% 1.7% -0.5% 1.4% Tunsa 0.2% 1.3% 0.0% -0.3% 0.0% 1.6% Turkey 0.3% 0.3% 0.1% 0.1% 0.0% 0.1% Uganda 1.7% 0.2% 1.5% -4.6% 0.3% 3.0% Uned Saes 0.2% -1.5% -1.0% 0.6% 0.7% -1.9% Venezuela -3.0% -2.7% -0.3% -0.3% -0.2% -1.8% Zmbabwe -2.2% -2.3% 0.4% -0.5% -0.3% -1.9% (*) The esmaed annual growh dfference o he mean n column (2) s he sum of he four conrbuons n columns (3 o 6). 33
CEPII, Workng Paper No 2004-02 Table XI: Growh componen dsperson (see Appendx 3, for he deals) Whou human capal Cenered Varables Sandard Devaon Mnmum Maxmum dlog y 0.0200-0.0466 0.0424 Resdual 0.0108-0.0221 0.0259 Absolue convergence 0.0058-0.0102 0.0158 Phys. cap. nves. rae 0.0134-0.0451 0.0269 Long erm TFP-growh 0.0045-0.0124 0.0076 Technology dffuson 0.0126-0.0182 0.0398 Wh human capal Cenered Varables Sandard Devaon Mnmum Maxmum dlog y 0.0200-0.0466 0.0424 Resdual 0.0095-0.0250 0.0210 Absolue convergence 0.0045-0.0073 0.0116 Seady sae conrb.: - physcal capal 0.0111-0.0368 0.0221 - human capal 0.0070-0.0237 0.0102 Long erm TFP-growh 0.0047-0.0137 0.0073 Technology dffuson 0.0147-0.0186 0.0444 34
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table XII: Defc n annual growh rae due o poor nsuonal qualy 1980-2000 (model wh human capal) Toal CORRUPTION R&D-CAPRISK PRODINDEX To compare: Sock of human capal Ngera 6.1% 2.3% 2.7% 1.1% 2.3% Zmbabwe 5.4% 1.7% 2.4% 1.3% 1.9% Ghana 5.2% 1.0% 3.4% 0.8% 1.6% Iran, Islamc Rep. 4.3% 1.0% 0.9% 2.4% 1.3% Inda 4.3% 1.9% 1.1% 1.3% 1.9% Indonesa 4.1% 1.6% 1.2% 1.3% 1.8% Cameroon 4.0% 1.8% 1.5% 0.7% 2.0% Syran Arab Republc 4.0% 1.0% 0.9% 2.1% 1.3% Uganda 3.9% 2.0% 1.7% 0.2% 2.5% Egyp, Arab Rep. 3.9% 1.4% 1.4% 1.1% 1.6% Coe d'ivore 3.5% 1.7% 1.4% 0.4% 1.9% Paksan 3.5% 1.8% 0.9% 0.8% 1.9% Romana 3.4% 1.1% 1.1% 1.2% 1.4% Algera 3.2% 1.2% 0.5% 1.5% 1.6% Phlppnes 3.2% 1.2% 0.8% 1.2% 1.3% Peru 3.1% 0.9% 1.4% 0.8% 1.2% Chna 3.1% 1.0% 0.2% 1.9% 1.5% Thaland 2.9% 0.9% 0.7% 1.3% 1.5% Chle 2.8% 0.1% 2.3% 0.4% 0.9% Morocco 2.5% 1.0% 0.6% 0.9% 1.6% Turkey 2.5% 1.0% 0.7% 0.8% 1.4% Colomba 2.4% 0.9% 0.6% 0.9% 1.3% Mexco 2.3% 1.1% 0.4% 0.8% 1.4% Venezuela, RB 2.2% 1.1% 0.2% 0.9% 1.2% Korea, Rep. 2.0% 0.6% 0.1% 1.3% 0.8% Malaysa 1.9% 0.4% 0.2% 1.3% 1.0% Brazl 1.8% 0.7% 0.2% 0.9% 1.3% Tunsa 1.6% 0.5% 0.2% 0.9% 1.3% Argenna 1.5% 0.8% 0.0% 0.7% 1.1% Souh Afrca 1.5% 0.7% 0.2% 0.6% 1.2% Japan 1.4% 0.3% 0.0% 1.1% 0.6% Tawan 1.3% 0.5% 0.0% 0.8% 0.9% Poland 1.2% 0.5% 0.2% 0.5% 1.0% Porugal 0.9% 0.4% 0.2% 0.3% 1.1% Hungary 0.7% 0.3% 0.2% 0.2% 1.1% Hong Kong, Chna 0.7% 0.2% 0.1% 0.4% 0.6% Israel 0.6% 0.2% 0.0% 0.4% 0.7% Greece 0.5% 0.4% 0.0% 0.1% 1.0% France 0.5% 0.2% 0.0% 0.3% 0.8% Sngapore 0.3% 0.0% 0.0% 0.3% 1.0% Norway 0.3% 0.0% 0.0% 0.3% 0.7% Uned Saes 0.2% 0.1% 0.0% 0.1% 0.0% Ireland 0.2% 0.0% 0.2% 0.0% 0.7% Germany 0.1% 0.0% 0.0% 0.1% 1.0% mean 2.4% 0.9% 0.7% 0.8% 1.3% sd 1.6% 0.6% 0.8% 0.5% 0.5% 35
CEPII, Workng Paper No 2004-02 Table XIII: A brave new world: Inequaly of ncome per capa ($ PPP) Tme Insuons Mean Sandard Dev. Varaon Coeff. Gn 1980 acual 6500 5100 0.78 0.42 2000 acual 9750 8600 0.88 0.48 2000 bes 14850 7950 0.54 0.32 36
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Char 1: GDP per capa and (non-)corrupon ndex NOR IRESIN GER CHL FRA HKO USA ISR HUN POR JAP MAL GRE TUN TAI POL KOR BRA SOA GHA THA -3-2 CHN -1 MOR TUR 0 ARG 1 Log (GDP/capa2 COL PER n 1980, PPP) ROM SYR MEX IRA ALG UGA PHI VEN EGY INO PAK COT CAM IND ZIM NIG CORRUPTION Char 2: R&D-CAPRISK and PRODINDEX CHL PRODINDEX IRE USA GRE GER HUN POR NOR FRA SIN HKO ISR POL TAI MEX PER ARG ROM TUR SOA COT PAK BRA COL JAP TUN KOR -10-5 UGA PHI MOR 0 5 10 15 MAL R&D-CAPRISK GHA CAM INO ZIM IND THA VEN EGY NIG ALG CHN IRA SYR 37
CEPII, Workng Paper No 2004-02 38 APPENDIX 1: The growh equaon The producvy growh of counry a me, ) ( g s gven by equaon (5): (0)) (1 (0)).(1 ) ( ) (. v b e b v c g A A g = & (A1) and under he assumpon ha, n he long run, here s echnologcal convergence o he froner ( 0 v ) : * ) ( lm g c g g = and we noe (0)) (1 (0)).(1 ) ( ) (. * v b e b v g g = γ Usng noaon of equaon (7) n he man ex, equaon (A1) has he followng soluon: f g v e A A ). (, * (0). ) ( = (A2) I s common o work wh he quany of capal and oupu per un of effecve labour, L A., ). /( ~ ). /( ~ ). /( ~ L A Y y L A H h L A K k (A3) From he dynamc of physcal capal, we derve he dynamc of k ~ : K y s k d g n k ~. ~ ). ( ~ = & (A4a) and equvalenly for human capal. H y s h d g n h ~. ~ ). ( ~ = & (A4b) Therefore, one can show as n MRW ha he seady-sae levels are: ) 1/(1 * * ) 1/(1 * 1 * ) 1/(1 * 1 * ) ) /( ).( ) (( ~ )) /( ).( ) (( ~ )) /( ).( ) (( ~ b a b a b H a K b a a H a K b a b H b K d g n s s y d g n s s h d g n s s k = = = (A5) I follows from (A4a-b): ) ( ). ( )) ( ~ ~..( )) ( ~ ~..( ~ ~. ~ ~. ~ ~ * * b a d g n h y s b d g n k y s a h h b k k a y y H K γ = = & & &
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Log-lnearsng around he seady-sae, wh he rae of * convergence β = (1 a b).( n g d), and usng he expresson n A(5) leads o: ~ ~ ~ ~ * dlog y y& ~ * y y * y = ~ = a.( n g d ).( Log ~ Log ~ ) b.( n g d ).( Log * ~ d y k k h.( ~ ~ * = β Log y Log y ) ( a b). γ ( ) ~ * y Log ~ ) ( a b). γ ( ) * (A6) h overesmae he growh of boh physcal and human capal per un of effecve labour. The soluon o he dfferenal equaon (A6) s: Log y ~ Log y ~ * = e β..( Log ~ y (0) Log ~ y * ) ( a b). e β.. 0 e β. x. γ ( x). dx By usng he seady-sae expresson gven by (A5), and nong β. 1 β. x γ v, ( a b). e.. e. γ ( x). dx, (A6) becomes he growh equaon : 0 K a H Log ~ y ( ) Log y ~ (0) s s Log ~ ( ).( ) = ρ. y (0) ρ. Log ( n d g c b a b ) 1/(1 a b) γ v, (A6) β wh ρ = 1/.(1 e ) Oupu per worker s easly derved from oupu per effecve labour by usng equaon (A2): K a H Log y( ) Log y (0) y (0) ( s ).( s ) = ρ. Log ρ. Log A (0) ( n d g c b a b ) 1 /(1 a b) The specfcaon of equaon (6) s hen esablshed by neglecng g c f v, γ v, γ v, compared o v Based on smulaons, he followng Table gves an order of he magnude of hs approxmaon. The lower he physcal and human capal share he more legmae he approxmaon s. Acually, he mporan hng o noce s ha boh f v, and h v, ncrease wh he rae of echnologcal dffuson. Gven capal shares, as he rao of he approxmaon (las column) s sable over he counres (dfferen b (0), see he frs hree lnes), specfcaon of equaon (6), whch neglecs, smply means ha he rae of dffuson s somewha underesmaed, he more so ha capal shares are hgh. γ v, f,. 39
CEPII, Workng Paper No 2004-02 v ab b (0 ) f v, γ v, f v, γ v, 5% 0.35 0.25 5.35% 0.98% 0.82 5% 0.35 0.50 2.45% 0.47% 0.81 5% 0.35 0.75 0.96% 0.19% 0.80 5% 0.60 0.50 2.45% 1.01% 0.59 1% 0.35 0.50 0.83% 0.18% 0.79 10% 0.35 0.50 3.12% 0.55% 0.82 f v, 40
Technology dfferences, nsuons and economc growh : a condonal condonal convergence APPENDIX 2: Facor analyss 1. Defnng fve domans and redressng wo values The 94 ndcaors (see secon 4.2) were clusered no 5 domans: produc marke, labour marke, fnancal marke, nnovaon and a general headng for all oher ems, each represenng a specfc aspec of a counry s nsuons. Table A1 gves he breakdown of he ndcaors. Each ndcaor s defned n a (0;4) range. A descrpve sascal analyss leads o redress wo ndcaor values for he USA. In he general doman, he ndcaor B300 ( Colluson beween he Sae and frms ) akes he value of 1.5 for he USA whereas, for nsance for OECD counres, France and Germany ake a 4, Japan a 2.5. We decded o redress he USA noaon o a conservave 3. In he produc marke doman, ndcaor B704 ( Pressure on compeon from employer organsaons ) shows: USA 1, Greece 4, France and Germany 3. We redressed he value for he USA o 3. 2. Facor analyss mehodology For each doman, we run a facor analyss n covarance whch provdes a more nformave sascal summary. Whn a gven doman, he underlyng facors are consruced as a lnear combnaon of he ndcaors n he doman, n such a way ha hese facors rean he mos nformaon as measured by he cross-counry varance. We am a denfyng he man facors wh sub-domans whch would be meanngful n economc erms. The mehodology s very close o Ncole, Scarpea and Boylaud (1999) who calculae ndcaors for Produc Marke Regulaon n OECD counres. For each doman, we valdae our approach by a ch-squared sgnfcance es (Barle es) and we selec he number of facors based on Akake s nformaon creron and Schwarz s bayesan creron. Once he opmal dmenson of he projecon space has been deduced, we use he varmax mehod o roae he axes n order o mnmse he number of varables wh hgh absolue loadngs on each facor, so ha facor nerpreaon s faclaed. Each counry has hen a score on each roaed facor. From facor scores, an aggregaed ndex s calculaed (for each doman). Here he aggregaon procedure dffers somewha from Ncole e al. s. The aggregaon deals are gven below. We use marx algebra. For a gven doman, X = [ x(, j)] s he orgnal daa wh ndexng he counres from 1 o I, and j ndexng he prmary ndcaors n he doman from 1 o J. The facor scores marx s Z = [ z(, k)] where k ndexes he roaed facors from 1 o K, K beng he number seleced facors. By denong V = [ v( j, k)] he marx of he K normalsed egen vecors, he facor analyss algebra s wren: Z = X. V Var Z = L 41
CEPII, Workng Paper No 2004-02 where L s he (K,K) dagonal marx wh he k-h coordnae beng he egen value l k. Before calculang he aggregaed ndex, we projec orhogonally he varables X j n he facors space Z and oban he esmaed Xˆ as follows: ˆ ' 1 ' 1 ' ' 1 ' X = Z.( Z Z). Z X = X. V. L. V X. X = X. V. L. LV. = Z. V ' As we chose o roae he axes, wha makes sense for he aggregaed ndex s only he overall nformaon n he facor space no on each facor ndvdually. A varable X j whch s a 100% n he facor space wll be lef nac by he ransformaon from X j o Xˆ j whereas a varable X ' orhogonal o he facor space wll dsappear. We hen defne j our aggregaed ndex as he average of he ransformed varables. INDEX K Xˆ ' = 1/.1 = 1/ K Z. V. 1 J J where 1 J s he (J,1) vecor of coordnae j equals o 1. Developng he marx formula above leads o: INDEX K z k = 1 = k v j K jk K v jk. v jk. x k= 1 j j = K 2 v jk k= 1 j j Ths aggregaed measure s appealng because s drecly based on he facor scores,.e. he sub-doman ndcaors, each beng weghed usng he coordnaes of he correspondng egen vecor. I s close o Ncole e al s aggregaed ndcaor whch s compued as follows: K lk = = 2 INDEX NSB. v jk. x k 1 lk j k j The neresng propery of our aggregaed ndcaor s he conssency wh he choce of roang he axes. Indeed, s easly shown ha hs ndcaor s nvaran o an orhogonal roaon of he facor space. If T s one of hese orhogonal ransformaon, characersed by ' T. T = I K, he K-dmenson deny marx, hen he egen vecors become V ~ = V. T and he ndex s lef unchanged: INDEX ~ ~ ~ ' = 1/ K X. V. V.1 = 1/ K X. V. V '. 1 J J = INDEX 42
Technology dfferences, nsuons and economc growh : a condonal condonal convergence 3. Resuls Resuls are presened for each doman. Tables A show he ess and he egen values breakdown. Tables B refer o he facor nerpreaon,.e. sub-doman denfcaon, hghlghng he varables whch conrbue mos o he varance. Tables C conan he subdoman ndcaors (.e. he facor scores) and he aggregaed ndex. All hese ndcaors are normalsed so ha hey are cenered and her sandard devaon equals 5. 3.1 General doman (GEN) The ess descrbed above lead us o selec four facors for he general doman, bu he frs wo ogeher conrbue o 57% of he varance, whereas he las wo o only 16%. Moreover, he nerpreaon of facors 3 and 4 are no obvous. The frs facor, CORRUPTION, s clearly an ndcaor of non-corrupon; ranges from 9.1 for Norway down o 10.2 for Ngera. The second facor, FREEDOM, s deermned by a se of prmary ndcaors referrng o dfferen aspecs of freedom. The aggregaed ndex, GENINDEX, based on hese four facor-scores ranges from 10.6 for Germany o 9.5 for Syra. 3.2 Produc marke doman (PROD) The produc marke doman can be lmed o hree sub-domans ha explan 67% of he varance. Each sub-doman s easly denfed and conrbues o roughly a hrd of he cross-counry nformaon whn he doman. These are: TRADECOMPET, an ndcaor of he lmed resrcons o mpors, of he enforcemen of nernaonal rade rules, for nsance hrough membershp o he WTO, and of compeon regulaon o avod colluson and abuse of marke power; LARGECO, he share of large companes n he dsrbuon secor whch s possbly a broader ndcaor of frm sze and effcen scale, NEWENTRY, barrers o new enry. 3.3 Labour marke doman (LABR) Four facors are necessary n he labour marke doman. The frs one, CONTRACT, only conrbues o more han 20% of he varance. I refers o he weak conrbuons of chld labour and nformal economy, whch s ndeed an mporan characersc as he sample focuses prmarly on developng counres. The second, UNIONFREED, an ndcaor of rade-unon freedom, accouns for roughly 20% of he varance. The las wo have a conrbuon of only 11% and are represenave of labour flexbly and unemploymen among he sklled young. The general ndex LABRINDEX s a mxure of componens ha we generally end o oppose o each-oher, lke srong unons and labour flexbly. Therefore, s no easly readable and wll be dropped from he analyss. 43
CEPII, Workng Paper No 2004-02 3.4 Bank and Fnancal sysem (FIN) Ths doman s he more dffcul o nerpre. Alhough only wo facors conrbue o 20% or more, ess requre fve facors. The frs one, BANRULES, has a clear-cu nerpreaon, referrng o bankng regulaon and prudenal rules, and conrbues o around 25% of he overall varance n he doman. The second roaed facor, DEPGUARANT, s mosly explaned by em C602 whch s a dummy varable denfyng counres where here s a law by whch he Sae guaranees deposs. I s farly dubous. Three oher domans are denfed bu hey each represen 10% or less of oal varance. These are domesc and nernaonal compeon n he fnancal sysem, cred access, and, more enavely, debor proecon. Here also, he general ndex s farly nangble and herefore we wll lm ourselves o BANKRULES. 3.5 Innovaon doman (INN) Facor analyss n hs doman clearly denfes hree dsnc aspecs lnked o he nnovaon envronmen. The frs sub-doman, R&D-CAPRISK, has n fac a double componen: on he one hand, effors o promoe R&D and nnovaon, and on he oher, a sysem favourng capal rsk fnancng. I conrbues 37% of he varance. The second subdoman, INTPROP, refers o he proecon of nellecual propery rghs and, o a lesser exen, o a developed envronmen of nsurance companes and penson funds. I also has a srong conrbuon of 28%. The hrd sub-doman, PROP, s more broadly lnked o he proecon of propery rghs bu conrbues o only 12.5% of he doman varance. 3.6 Correlaon The compuaon of lnear Pearson coeffcens beween he aggregaed ndces of each doman reveals a srong correlaon. Table A2 shows ha lnear correlaon coeffcens range from 0.67 o 0.88. Ths enals ha f we wan o capure a specfc aspec of nsuons, we should lean owards focusng on less aggregaed daa as much as possble, lke he denfed facors hemselves. In oher words, we have o carefully wegh he radeoff beween greaer varance and more dealed nformaon. 44
Technology dfferences, nsuons and economc growh : a condonal condonal convergence APPENDIX 3: Componens of he growh equaon The specfcaon whou human capal s aken as an example. I s easly exended when human capal s ncluded. From equaon (10) n he man ex, we derve four componens of esmaed growh. The frs s he adjused absolue convergence erm. The ypcal erm s lessened here because of dfferences n nal producvy level. The second s he nvesmen rae conrbuon hrough s nfluence on he seady-sae. The hrd refers o he X(0) dynamc effcency loss due o he nsuon I. I does no capure he whole long erm TFP-growh erm g c because of he ransonal dynamcs. To undersand hs, keep n mnd ha he usual frs wo erms above ake no accoun he ndvdual TFP-growh. I s he oal effec of havng a lower perssen TFP-growh ha s refleced n he varable Z and aken no accoun n hs hrd erm. The las componen of growh s a ransonal conrbuon comng from echnologcal dffuson. Formally, usng equaon (10): y (0) " adjused" absolueconvergence = ρ. Log a (0). A (0) nvesmen rae a =. ρ. Log 1 a n long erm TFP growh = g c.( I c ben 1 1 (1 b (0)). e echnolog cal dffuson =. Log b (0) s g d c I ). Z To calculae wha coss a counry no o have perfec nsuons as s repored n Table XII, we assess counerfacually wha would be s esmaed growh wh x ( 0) = 1, dff dff max c c ben I = I and I = I. As here s a cross erm n he Z expresson, we spl hem equally beween he wo nsuonal varables. Formally, mpac( I mpac( I mpac( I X (0) dff c ) = ( ρ 1 1 (1 b (0)). e ) =. Log 1 (1 b (0)). e ) = c. Z v. c.(1 a).(1 ρ. / 2)). Log( x dff dff v.( I max I mn ). dff dff v.( I I mn ). ben c.(1 a).(1 ρ. / 2). Log( x (0)) (0)) 45
CEPII, Workng Paper No 2004-02 APPENDIX 4: Trade conrbuon I s beyond he scope of hs sudy o overcome he ssue rased by he endogeney of rade n s conrbuon o growh. Recognsng hs lmaon, we wll neverheless show some resuls hopng hey shed some lgh on he complex mechansm a work. Sac gans from rade spread over a ranson perod or dynamc gans from spllovers should resul, a leas emporarly, n ncreased oal producvy. If we use as a rade varable he sum of expors plus mpors dvded by GDP (OPENC n he Penn World Table), ake he average over he perod and add no (INST1) o es he mpac of openness over TFP-growh usng c c c.( ben max = c I I ) d.( Trade Trade ) (INST1 ) he sensvy of growh o rade, as measured by he parameer d, s found o be nsgnfcan. More rade does no seem o expede growh. However, we found wo sgnfcan dfferen conrbuons of rade. The frs s obaned when crossng he openness level wh he percenage of manufacures expors n oal expors (World Developmen Indcaors). The nuon refers o he dea ha rade n agrculure or raw maerals may no be benefcal a all n erms of TFP-growh. Table A4 shows, sarng n column (1) from our base esmae from Table V column (5), our new esmaes when we nclude rade as openness rao n column (2) and when we cross wh he manufacurng expor share n column (3). 13 Wh he average esmae of 0.0212 for d, a counry lke Hungary wh 64.5% of s expors beng manufacured producs s expeced o add 0.27% annually o long erm growh-rae when s expors and mpors each ncrease by 10% of GDP whereas mos of Sub-Saharan counres wll add vrually nohng. We obvously dd no go furher no secor deals, bu hs resul suggess eher ha specalsaon maers for he mpac of rade on growh or ha, f we ake he manufacurng share n expors as a developmen ndcaor, rade s mporan for effcency laer on n he developmen phase. The second sgnfcan conrbuon of rade we obaned renforces hese nferences. Gong one sep furher and crossng he rade varable of column (3) wh dummy varables whch separae he sample no wo, wh he break pon a average nal ncome per capa, we conclude ha exporng manufacures (as a share of GDP) ncreases growh for relavely rch counres only (column 4), he effec beng nsgnfcan or f anyhng negave for poor counres (column 5). We deduce, echong Acemoglu, Aghon and Zlbo (2002) ha he posve effec of (specalsed) rade on growh s non lnear wh lberalsng o rade no helpng much n he early sage of developmen. The magnude of he conrbuon measured by he dsperson as n Tables XI pons o a sandard devaon of he rade conrbuon of 0.25%, 0.35% and 0.50% for specfcaon n columns (2), (3) and (4) respecvely, equvalen o he one denfed for produc marke compeon level whch by he way, becomes less sgnfcan as rade does. We also esed wheher ncreased rade acceleraes he speed of echnology dffuson hrough an ncreased speed and found no sgnfcan mpac. Fnally, esng he mpac of rade hrough he more horough 13 Our openness measures s he OPENC average of years 1980, 1990 and 1995 - chosen because of mssng daa for 2000 - capped a 100% o avod some awkward effecs comng for example from exreme case lke Sngapore wh openness level over 400%. No cappng ends o flaen ou he mpac and probably leads o less relevan resuls. 46
Technology dfferences, nsuons and economc growh : a condonal condonal convergence specfcaon ncludng human capal leads o smlar esmaes. Alhough we acknowledge he lmaons descrbed above due o he endogeney ssue, hs resul conflcs wh hose (more rgorous) of Easerly and Levne (2002) and Rodrk, Subramanan and Trebb (2002). < Table A4 > Table A4: Trade conrbuon o growh TRADE VARIABLE - Openness Openness * Manufacures expor share Openness * Manufacures expor share * Openness * Manufacures expor share * Dummy rch Dummy poor (1) (2) (3) (4) (5) a (phys.capal) 0.484 0.483 0.442 0.466 0.462 (0.076)*** (0.063)*** (0.075)*** (0.079)*** (0.054)*** c ( I c ) 0.0015 0.0015 0.0008 0.0006 0.0016 (0.0007)** (0.0007)** (0.0006)* (0.0006) (0.0007)** v ( I dff ) 0.0064 0.0063 0.0048 0.0071 0.0079 (0.0029)** (0.0029)** (0.0021)** (0.0031)** (0.0038)** d (Trade) - 0.003 0.021 0.027-0.033 (0.013) (0.012)** (0.011)*** (0.030) LogAben (0) 8.54 8.57 8.98 8.89 8.23 (0.38)*** (0.34)*** (0.40)*** (0.38)*** (0.37)*** 2 R 0.67 0.66 0.68 0.70 0.67 Observ. 44 44 44 44 44 Table A2: Pearson Correlaon Coeffcens, N = 51 genndex prodndex labrndex fnndex nnndex genndex 1.000 0.884 0.819 0.832 0.881 prodndex 0.884 1.000 0.748 0.846 0.875 labrndex 0.819 0.748 1.000 0.667 0.693 fnndex 0.832 0.846 0.667 1.000 0.845 nnndex 0.881 0.875 0.693 0.845 1.000 47
CEPII, Workng Paper No 2004-02 Table A: GEN Sgnfcance Tess Based on 51 Observaons Tes DF Ch-Square Prob > ChSq H0 : No common facors HA : A leas one common facor H0 : 4 Facors are suffcen HA : More facors are needed 231 931.80 <.0001 149 163.56 0.1960 Ch-Sqare whou Barle s Correcon 208.80 Akake s Informaon Creron -89.20 Schwarz s Bayesan Creron -377.04 Tucker and Lews s Relably Coeffcen 0.968 Egen Values Conrbuon roaon NbFACTOR Facor1 0.501 0.362 4 Facor2 0.090 0.206 4 Facor3 0.073 0.078 4 Facor4 0.061 0.079 4 sum 0.725 0.725. Table B: GEN Roaed Facors Imporan Varables Inerpreaon Varable Names Commens Facor 1 a303 a307 Non corrupon CORRUPTION Facor 2 a103 a104 a800 a100 Freedom FREEDOM Facor 3 b300 a801 a106? GEN3 Facor 4 a606 Inernaonal Arbrage of dsagreemens INTARB 48
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table C: GEN CORRUPTION FREEDOM GEN3 INTARB GENINDEX Algera -4.2-5.6 2.7 4.9-4.9 Argenna -1.3 0.3-3.4 3.9-0.8 Bulgara 0.3 4.2-3.4-1.0 1.7 Brazl 1.0 2.4-6.4-3.1 0.1 Cameroon -7.3 1.1 3.0-3.8-5.1 Chle 7.4 0.0-3.4-9.0 3.2 Chna -1.0-10.5-4.2 4.2-6.7 Colomba -2.1 0.9-1.4 5.7-0.3 Ivory Coas -6.4 2.9 1.2 4.8-2.0 Czech Republc 1.4 4.1-1.7-3.7 2.3 Egyp -5.2-8.6 2.1 1.0-8.2 France 6.7 2.0 6.2 4.8 8.6 Germany 8.7 2.2 8.5 3.7 10.6 Ghana -0.7 0.7-2.2-13.3-3.3 Greece 3.3 4.0 9.4-1.3 6.6 Hong Kong 6.1-0.3-2.1 8.2 5.7 Hungary 5.3 2.2-2.6 5.9 5.9 Inda -7.9 5.9 6.3 1.6-1.1 Indonesa -6.0 3.4 1.6-5.3-3.4 Iran -3.8-6.9-0.2-3.2-7.5 Ireland 9.0 2.6 6.8-0.7 9.7 Israel 5.8-1.4 3.4 6.8 5.8 Japan 4.4 3.8-4.9-2.6 4.0 Korea, Rep. 0.3 4.9-9.1 2.3 1.5 Lhuana 4.5 5.9-14.2 7.4 5.2 Malaysa 3.3-6.1 2.6 1.4 0.0 Mexco -3.1 3.2 0.6-0.4-0.6 Morocco -0.9-2.8-4.1-2.7-3.7 Ngera -10.2 4.3 3.5 0.6-4.6 Norway 9.1 2.1 3.1 3.1 9.5 Paksan -6.6-2.6 0.1-4.2-7.3 Peru -2.2 0.8 1.2 3.0-0.4 Phlppnes -4.8 4.7 1.2 4.5 0.2 Poland 1.9 4.9 3.2 4.9 5.9 Porugal 4.3 4.8-9.0-10.6 2.0 Romana -3.2 2.7-1.1 4.2-0.3 Russa -2.5 0.8-0.2-1.1-1.7 Saud Araba 2.9-15.9 3.4-0.4-6.1 Sngapore 8.6-7.9 0.4 2.7 2.8 Souh Afrca 0.7 1.3 7.4-5.3 1.8 Syra -2.9-10.7-3.8-2.0-9.5 Tawan 3.0 3.6 0.3-13.0 1.8 Thaland -0.5-2.0-9.4 4.4-2.7 Tunsa 2.7-9.2-0.2-5.3-4.2 Turkey -1.4 0.0 4.2 0.3-0.1 Ukrane -4.9 4.6 4.3-0.4-0.4 Uganda -4.6 0.6-8.1 1.6-4.6 Uned Saes 6.3 3.0 6.4-6.0 6.8 Venezuela, RB -4.5 2.8-2.2 4.6-1.4 Venam -2.3-6.7-2.2-2.8-6.5 Zmbabwe -6.6-0.8 6.7 0.8-3.9 49
CEPII, Workng Paper No 2004-02 Table A: PROD Sgnfcance Tess Based on 51 Observaons Tes DF Ch-Square Prob > ChSq H0 : No common facors HA : A leas one common facor 91 383.90 <.0001 H0 : 3 Facors are suffcen HA : More facors are needed 52 55.13 0.3573 Ch-Sqare whou Barle s Correcon 64.85 Akake s Informaon Creron -39.14 Schwarz s Bayesan Creron -139.60 Tucker and Lews s Relably Coeefcen 0.9813 Egen Values Conrbuon Roaon NbFACTOR Facor1 0.499 0.214 3 Facor2 0.092 0.252 3 Facor3 0.083 0.207 3 sum 0.673 0.673. Table B: PROD Roaed Facors Facor 1 Imporan Varables B8020 a701 b800 b801 Inerpreaon Varable Names Commens Few barrers o mpors, nernaonal rade rules (WTO) and compeon enforcemen Facor 2 B702 Large companes n he dsrbuon secor Facor 3 B301 b700 Enhancemen of new frms TRADECOMPET Srong negave coordnae of b702 whch conrbues grealy o facor2 LARGECO NEWENTRY To a lesser exen: b601, qualy norms 50
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table C: PROD TRADECOMPET LARGECO NEWENTRY PRODINDEX Algera -0.9-5.6-3.1-5.0 Argenna -2.5 3.0 1.2 0.3 Bulgara -2.2 3.9-1.9-0.3 Brazl -4.7 9.1-5.7-1.1 Cameroon 2.6-0.6-8.9-2.3 Chle 1.7 2.2 5.0 4.7 Chna -5.7 3.9-9.2-6.1 Colomba 0.9 2.5-7.0-1.2 Ivory Coas -1.3 0.0 1.8-0.1 Czech Republc 5.6-0.8 0.1 3.7 Egyp -3.0 1.6-5.6-3.8 France 9.1-2.3 2.5 6.4 Germany 7.4 1.6 4.3 8.1 Ghana 1.3-9.7 3.2-2.8 Greece 9.7 0.3 1.4 7.8 Hong Kong 1.5 2.6 7.0 5.5 Hungary 1.2 5.9 7.1 7.2 Inda 4.2-5.2-6.9-2.8 Indonesa -3.2-2.2 1.9-2.6 Iran -12.6-12.5 7.2-12.6 Ireland 5.8 2.6 6.7 8.5 Israel -0.4 6.2 4.3 4.9 Japan 0.0 2.0-4.0-0.7 Korea, Rep. 0.9-0.2-2.5-0.6 Lhuana -0.8 10.0-1.7 4.0 Malaysa -6.8 3.9 2.0-1.9 Mexco -1.1 4.3-0.4 1.3 Morocco 3.9-2.6-7.0-1.7 Ngera 4.4-5.6-11.0-4.7 Norway 2.2 3.9 6.7 6.6 Paksan 8.0-8.8-3.3-0.4 Peru 1.4 0.1 0.9 1.4 Phlppnes -4.2 1.9 0.3-1.9 Poland -3.4 7.8 4.1 3.5 Porugal 5.9 0.9 3.6 6.3 Romana 2.6-1.5-2.8-0.2 Russa -3.7-0.3-8.0-6.4 Saud Araba -3.0-7.9 3.6-4.8 Sngapore 4.7-0.4 7.3 6.4 Souh Afrca -3.1 7.0-1.8 0.7 Syra -14.7-5.7 1.2-13.1 Tawan 0.2 1.0 2.9 2.0 Thaland -0.4 2.5-7.3-2.2 Tunsa 5.4-8.2 0.3-0.3 Turkey -0.9 4.3-1.6 1.0 Ukrane 3.3-4.7-0.6-0.4 Uganda -1.4-4.2-4.1-5.0 Uned Saes 4.4 3.1 7.8 8.3 Venezuela, RB -6.1 2.1 0.0-3.3 Venam -7.2-8.3 5.6-7.1 Zmbabwe -5.2-2.7 4.4-3.3 51
CEPII, Workng Paper No 2004-02 Table A: LABR Sgnfcance Tess Based on 51 Observaons Tes DF Ch-Square Prob > ChSq H0 : No common facors HA : A leas one common facor H0 : 4 Facors are suffcen HA : More facors are needed 105 342.55 <.0001 51 60.47 0.1710 Ch-Sqare whou Barle s Correcon 72.85 Akake s Informaon Creron -29.15 Schwarz s Bayesan Creron -127.67 Tucker and Lews s Relably Coeefcen 0.9179 Egen Values Conrbuon roaon NbFACTOR Facor1 0.349 0.256 4 Facor2 0.142 0.198 4 Facor3 0.100 0.111 4 Facor4 0.086 0.111 4 sum 0.677 0.677. Table B: LABR Roaed Facors Imporan Varables Facor 1 D904 d403 Lmed chld labour and small nformal economy Inerpreaon Varable Names Commens CONTRACT Facor 2 d100 d101 Trade-unon rghs UNIONFREED Facor 3 d701 d700 -d600 d601 Decenralsed wage negoaons. Lle consrans from mnmum wages and layoff procedures FLEX Facor 4 d903 Sklled young unemploymen EDUC 52
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table C: LABR CONTRACT UNIONFREED FLEX EDUC LABRINDEX Algera -3.5-0.2-2.6-2.5-3.9 Argenna -3.1-0.1-2.1 4.5-2.7 Bulgara -5.8 1.7-6.9 4.6-5.1 Brazl -0.2 1.9 8.3 3.3 2.8 Cameroon -8.9 4.9 0.0-5.6-5.8 Chle 1.2 2.3 7.3 3.2 3.9 Chna -2.7-11.3 6.4 7.1-5.6 Colomba -2.6-1.9-1.5 2.9-3.2 Ivory Coas -6.1 4.0 0.9-4.6-3.5 Czech Republc 2.5 3.8 2.1 4.7 4.8 Egyp -4.8-7.6 0.0-7.1-8.4 France 7.5 2.8-2.8 0.5 7.1 Germany 8.8 4.1-6.3 0.7 8.2 Ghana -5.7 0.2-3.8 4.7-5.1 Greece -0.7 6.1-0.1-3.2 1.9 Hong Kong 5.9 5.1 3.9-10.8 7.2 Hungary 6.0 4.9-0.6 0.3 7.3 Inda -9.5 3.9 1.2-5.2-6.5 Indonesa -3.0 5.6 5.7-5.0 0.8 Iran 2.3-13.4-3.5-4.8-5.6 Ireland 6.8 3.4-8.8 2.0 5.7 Israel 5.7 1.8-7.5 1.7 4.3 Japan 2.6 0.7-8.7-2.3 0.5 Korea, Rep. -0.1-2.0 0.3-3.5-1.3 Lhuana 6.8 5.3 4.4 0.8 9.3 Malaysa 5.8-3.1 4.6-5.8 3.9 Mexco -0.9-2.8-1.3 2.7-2.1 Morocco -3.5 0.6 4.8-6.2-2.3 Ngera -11.6 0.1-4.8 7.6-10.1 Norway 8.7 4.2-5.6 0.6 8.3 Paksan -4.9 2.9 5.1-8.5-2.5 Peru -1.5 3.4 3.1 3.2 1.3 Phlppnes -1.7-1.2 6.0 4.2-0.4 Poland 6.3 3.6 1.3 0.2 7.4 Porugal 1.7 4.9-0.2 4.0 4.1 Romana 1.3 2.0-4.2 2.8 1.4 Russa 3.4-0.7 2.3 4.4 3.5 Saud Araba 7.5-14.8 6.4-8.0 0.0 Sngapore 6.9-9.5 2.2 5.3 2.3 Souh Afrca -1.0-0.2-1.8 4.6-0.9 Syra -2.3-9.1-13.5-5.3-9.6 Tawan 2.6 1.5 0.4 5.7 3.5 Thaland -5.2-2.6 3.0 3.0-4.7 Tunsa 0.1-6.5-7.4-3.1-4.9 Turkey -1.4-0.8 0.1-6.2-2.1 Ukrane -0.3 3.7 2.9-9.0 1.3 Uganda -0.4 0.2 7.4 5.6 1.8 Uned Saes 3.3 0.5 7.6 5.1 5.1 Venezuela, RB -3.6 4.1-4.7-5.0-2.6 Venam -2.5-6.2 4.6 5.6-3.5 Zmbabwe -6.3 0.1-3.8 6.3-5.6 53
CEPII, Workng Paper No 2004-02 Table A: FIN Sgnfcance Tess Based on 51 Observaons Tes DF Ch-Square Prob > ChSq H0 : No common facors HA : A leas one common facor H0 : 5 Facors are suffcen HA : More facors are needed 231 630.59 <.0001 131 139.74 0.2846 Ch-Sqare whou Barle s Correcon 181.48 Akake s Informaon Creron -80.52 Schwarz s Bayesan Creron -333.59 Tucker and Lews s Relably Coeefcen 0.9614 Egen Values Conrbuon roaon NbFACTOR Facor1 0.335 0.246 5 Facor2 0.157 0.088 5 Facor3 0.086 0.106 5 Facor4 0.073 0.197 5 Facor5 0.067 0.080 5 sum 0.717 0.717. Table B: FIN Roaed Facors Imporan Varables Inerpreaon Varable Names Commens Facor 1 C601 c703 c704 Bank conrol, rules (Cook BANKRULES c705 rao, normalsed accoun. sysem), ransparency Facor 2 C801 c701 c700 c400 Domesc and nernaonal compeon COMPET Facor 3 C901 c600 c900 -c402 Cred access and nonndependence of he cenral bank CREDIT Facor 4 C602 Sae guarany on deposs DEPGUARAN Facor 5 C702 Promong compeon and Fc702 debors proecon? 54
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table C: FIN BANKRULES COMPET CREDIT DEPGUARANT Fc702 FININDEX Algera -3.3-6.7 3.8 4.0 1.5-4.1 Argenna 2.2 4.0-4.0 3.4-6.0 3.2 Bulgara 2.1-1.5-11.1 3.4 0.9-0.2 Brazl -3.9-0.2 7.8 4.1 0.2-0.5 Cameroon -5.2-0.3 2.8-5.2-5.7-5.8 Chle 4.8 3.2 2.0 2.7-7.1 5.6 Chna -5.2-6.1 4.1-4.9 4.5-7.2 Colomba -2.6 5.6-1.1 5.4-6.3 1.4 Ivory Coas 2.2 6.7 6.6-5.7-7.7 4.0 Czech Republc -4.7 1.3 6.7 5.0 2.0 0.0 Egyp -5.9-0.9-6.1 3.7-1.1-5.3 France 3.4 4.2 6.7-5.3 7.9 5.7 Germany 7.3-1.5 1.8 4.4 7.6 7.4 Ghana 1.7 0.1-4.7-5.4-6.5-1.9 Greece 6.9-4.1 7.2 3.8 7.3 6.5 Hong Kong 8.6 2.4-2.3-7.7-2.8 5.0 Hungary 2.3 6.8 4.9 3.7 0.4 7.3 Inda 0.5 0.0-7.2-6.4 6.8-1.9 Indonesa -5.7 9.6 0.1-4.6-0.9-0.7 Iran -7.2-11.4-3.8-4.7 1.5-13.4 Ireland 9.7-7.8 1.1 5.1-8.4 3.9 Israel 6.9-10.9 3.8 3.3-5.8 0.5 Japan -1.6 4.3 0.1 2.9 9.5 3.0 Korea, Rep. -1.4 4.9-4.3 4.3 5.6 2.6 Lhuana -3.8-0.5 4.0 5.6 0.4-0.8 Malaysa 3.3-0.6-4.2-5.6-1.6-0.3 Mexco -3.5 7.1 1.9 4.1-6.1 1.8 Morocco -0.4 5.0-7.2-5.9 7.1 0.2 Ngera -5.2 7.4-2.6-5.9-0.9-2.3 Norway 7.3-2.4 0.7 3.8 7.1 6.4 Paksan -9.1 2.0-1.5 4.8-0.9-5.0 Peru -1.1 2.5-4.1 5.4-5.6 0.4 Phlppnes 1.0 4.9-0.1 4.0-5.1 3.8 Poland 2.3 7.6 7.6-5.4 8.5 6.9 Porugal 2.6-2.8-1.3 4.3 2.6 1.8 Romana -4.9-2.8 5.2 4.9 0.5-2.9 Russa -2.6 1.5 7.7-5.3-3.1-1.6 Saud Araba 2.4-3.7 6.9-6.0 1.4-0.2 Sngapore 10.4-5.8-3.6-6.7-1.6 2.3 Souh Afrca 3.9-1.3 2.0-5.7 2.1 1.4 Syra -10.4-11.5-0.8-4.7 0.1-15.5 Tawan -1.1 2.2-9.3 4.5 7.1 0.6 Thaland 1.5 1.0-0.2-6.5-3.3-0.6 Tunsa -2.2-2.9-10.3 3.6 3.1-3.8 Turkey -3.2 0.5 3.9 6.2-3.3-0.2 Ukrane 2.1-4.2 2.8 5.2-1.1 1.3 Uganda -3.6-1.5 3.1-5.4-1.9-4.7 Uned Saes 6.5 4.4-2.9 2.9 6.5 8.4 Venezuela, RB 1.0-1.6-5.8 4.7-3.0-0.3 Venam -8.4-6.5-0.7-4.9 1.9-11.2 Zmbabwe 3.2 0.2-6.0-5.6-8.1-1.1 55
CEPII, Workng Paper No 2004-02 Table A: INN Sgnfcance Tess Based on 51 Observaons Tes DF Ch-Square Prob > ChSq H0 : No common facors HA : A leas one common facor H0 : 3 Facors are suffcen HA : More facors are needed 78 489.10 <.0001 42 41.71 0.4837 Ch-Sqare whou Barle s Correcon 48.68 Akake s Informaon Creron -35.31 Schwarz s Bayesan Creron -116.45 Tucker and Lews s Relably Coeefcen -1.00 Egen Values Conrbuon Roaon NbFACTOR Facor1 0.587 0.368 3 Facor2 0.109 0.280 3 Facor3 0.077 0.125 3 sum 0.773 0.773. Table B: INN Roaed Facors Facor 1 Facor 2 Imporan Varables C501 A501 B603 c502 c500 Inerpreaon Varable Names Commens R&D effor and capal rsk srucure Inellecual propery rghs proecon and o a lesser exen nsurance companes and penson funds R&D-CAPRISK INTPROP Facor 3 A600 Propery rghs PROP There s a weak correlaon beween he propery rghs em (A600) and he nellecual propery rghs em (B603) 56
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Table C: INN R&D-CAPRISK INTPROP PROP INNINDEX Algera -3.3-4.2-2.4-5.8 Argenna -5.1 6.6 2.3 3.2 Bulgara 2.6 0.8-10.8-1.6 Brazl -1.7-3.5 2.0-3.0 Cameroon -4.9-6.0-0.4-7.4 Chle -5.9 9.4 6.2 6.2 Chna 4.8-10.8-2.2-6.7 Colomba -1.4 1.7 4.4 2.0 Ivory Coas -5.1 2.1-1.7-1.6 Czech Republc 1.7 2.1-3.2 1.5 Egyp -4.0-2.9-1.1-4.7 France 8.2 4.1-12.2 3.6 Germany 10.9 3.6-12.9 4.4 Ghana -5.9 0.2 1.2-2.6 Greece 2.0 6.0-2.9 4.8 Hong Kong 3.7 7.7-11.7 4.1 Hungary 2.3 7.5 4.9 8.7 Inda -1.8-1.3 0.9-1.7 Indonesa -3.3-1.6-4.0-4.3 Iran -4.1-5.1-6.1-8.1 Ireland 4.5 8.0 4.3 10.1 Israel 7.9 0.9 5.7 6.8 Japan 1.7 4.1 3.3 5.2 Korea, Rep. 5.7-5.2 4.4 0.4 Lhuana 1.8 0.0 5.0 2.6 Malaysa 5.2-3.6 3.9 1.2 Mexco -6.6 4.3-0.6-0.3 Morocco -0.7-5.6 2.2-4.1 Ngera -6.6-0.4 0.0-3.9 Norway 8.3 2.7 3.4 7.6 Paksan -1.0-5.0-0.7-4.7 Peru -5.4 2.4 1.8-0.4 Phlppnes -2.0 1.1-2.2-0.9 Poland 2.7 2.1 6.4 5.2 Porugal 0.8 2.7 4.8 4.1 Romana -2.0-3.4-3.5-4.9 Russa -2.9-3.2-6.9-6.3 Saud Araba -4.1-3.3 6.6-2.6 Sngapore 9.3 2.0 2.8 7.4 Souh Afrca 0.1 3.4 1.7 3.2 Syra -2.4-9.4-2.9-9.6 Tawan 11.7-7.9 4.8 1.5 Thaland -0.8-5.3 7.0-2.3 Tunsa 4.0-7.5-0.5-3.9 Turkey -2.0 1.1-0.9-0.5 Ukrane -2.8 0.8 6.6 1.3 Uganda -3.8-4.0-2.6-6.0 Uned Saes 6.9 5.8 4.4 9.7 Venezuela, RB -6.0 5.3-2.9 0.0 Venam -3.1-6.9-2.4-7.8 Zmbabwe -7.7 8.2-3.6 1.1 57
CEPII, Workng Paper No 2004-02 LIST OF WORKING PAPERS RELEASED BY CEPII 14 No Tle Auhors 2004-01 Crossance e régmes d nvesssemen P. Vlla 2003-22 A New Look a he Feldsen-Horoka Puzzle usng an Inegraed Panel 2003-21 Trade Lnkages and Exchange Raes n Asa : The Role of Chna 2003-20 Economc Implcaons of Trade Lberalzaon Under he Doha Round 2003-19 Mehodologcal Tools for SIA Repor of he CEPII Worshop held on 7-8 November 2002 n Brussels 2003-18 Order Flows, Dela Hedgng and Exchange Rae Dynamcs A.Banerjee P. Zangher A. Bénassy-Quéré & Amnal Lahrèche-Révl J. Francos, H. Van Mejl, F. Van Tongeren N. Kousnezoff B. Rzepkowsk 2003-17 Tax compeon and Foregn Drec Invesmen A. Bénassy-Quéré, L. Fonagné & A. Lahrèche-Révl 2003-16 Commerce e ransfer de echnologes : les cas comparés de la Turque, de l Inde e de la Chne F. Lemone & D. Ünal-Kesenc 2003-15 The Emprcs of Agglomeraon and Trade K. Head & T. Mayer 2003-14 Noonal Defned Conrbuon: A Comparson of he French and German Pon Sysems 2003-13 How Dfferen s Easern Europe? Srucure and Deermnans of Locaon Choces by French Frms n Easern and Wesern Europe 2003-12 Marke Access Lberalsaon n he Doha Round: Scenaros and Assessmen 2003-11 On he Adequacy of Moneary Arrangemens n Sub- Saharan Afrca F. Legros C. Dsder & T. Mayer L. Fonagné, J.L. Guérn & S. Jean A. Bénassy-Quéré & Mayls Coupe 14 Workng papers are crculaed free of charge as far as socks are avalable; hank you o send your reques o CEPII, Sylve Huron, 9, rue Georges-Pard, 75015 Pars, or by fax : (33) 01 53 68 55 04 or by e-mal Huron@cep.fr. Also avalable on: \\www.cep.fr. Workng papers wh * are ou of prn. They can neverheless be consuled and downloaded from hs webse. 14 Les documens de raval son dffusés grauemen sur demande dans la mesure des socks dsponbles. Merc d adresser vore demande au CEPII, Sylve Huron, 9, rue Georges-Pard, 75015 Pars, ou par fax : (33) 01 53 68 55 04 ou par e-mal Huron@cep.fr. Egalemen dsponbles sur : \\www.cep.fr. Les documens de raval comporan * son épusés. Ils son ouefos consulable sur le web CEPII. 58
Technology dfferences, nsuons and economc growh : a condonal condonal convergence 2003-10 The Impac of EU Enlargemen on Member Saes : a CGE Approach 2003-09 Inda and he World Economy : Tradonal Specalsaons and Technology Nches 2003-08 Imnaon Amongs Exchange-Rae Forecasers : Evdence from Survey Daa 2003-07 Le Currency Board à ravers l expérence de l Argenne H. Bchr, L. Fonagné & P. Zangher S. Chauvn & F. Lemone M. Bene, A. Bénassy-Quéré & H. Colas S. Chauvn & P. Vlla 2003-06 Trade and Convergence : Revsng Ben-Davd G. Gauler 2003-05 Esmang he Fundamenal Equlbrum Exchange- Rae of Cenral and Easern European Counres he EMU Enlargemen Perspecve 2003-04 Sklls, Technology and Growh s ICT he Key o Success 2003-03 L nvesssemen en TIC aux Eas-Uns e dans quelques pays européens 2003-02 Can Busness and Socal Neworks Explan he Border Effec Puzzle? 2003-01 Hypernflaon and he Reconsrucon of a Naonal Money: Argenna and Brazl, 1990-2002 2002-18 Programme de raval du CEPII pour 2003 2002-17 MIRAGE, a Compuable General Equlbrum Model for Trade Polcy Analyss 2002-16 Evoluons démographques e marché du raval : des lens complexes e parfos conradcores 2002-15 Exchange Rae Regmes and Susanable Pares for CEECs n he Run-up o EMU Membershp B Eger & A. Lahrèche-Revl J. Melka, L. Nayman, S. Sgnano & N. Mulder G. Cee & P.A. Noual P.P. Combes, M. Lafourcade & T. Mayer J. Sgard M.H. Bchr, Y. Decreux, J.L. Guérn & S. Jean L. Cadou, J. Gene & J.L. Guérn V. Couder & C. Couharde 2002-14 When are Srucural Defcs Good Polces? J. Chaeau 2002-13 Projecons démographques de quelques pays de l Unon Européenne (Allemagne, France, Iale, Royaume-Un, Pays-Bas, Suède) R. Sleman 2002-12 Regonal Trade Inegraon n Souhern Afrca S. Chauvn & G. Gauler 2002-11 Demographc Evoluons and Unemploymen: an J. Châeau, J.L. Guérn Analyss of French Labour Marke wh Workers & F. Legros 59
CEPII, Workng Paper No 2004-02 Generaons 2002-10 Lqudé e passage de la valeur P. Vlla 2002-09 Le concep de coû d usage Puy-Clay des bens durables 60 M.G. Foggea & P. Vlla 2002-08 Mondalsaon e régonalsaon : le cas des ndusres du exle e de l habllemen M. Fouqun, P. Morand R. Avsse G. Mnvelle & P. Dumon 2002-07 The Survval of Inermedae Exchange Rae Regmes A. Bénassy-Quéré & 2002-06 Pensons and Savngs n a Moneary Unon : An Analyss of Capal Flow 2002-05 Brazl and Mexco s Manufacurng Performance n Inernaonal Perspecve, 1970-1999 2002-04 The Impac of Cenral Bank Inervenon on Exchange-Rae Forecas Heerogeney 2002-04 The Impac of Cenral Bank Inervenon on Forecas Heerogeney 2002-03 Impacs économques e socaux de l élargssemen pour l Unon européenne e la France 2002-02 Chna n he Inernaonal Segmenaon of Producon Processes 2002-01 Illusory Border Effecs: Dsance Msmeasuremen Inflaes Esmaes of Home Bas n Trade 2001-22 Programme de raval du CEPII pour 2002 2001-21 Crossance économque mondale : un scénaro de référence à l horzon 2030 2001-20 The Fscal Sablzaon Polcy under EMU An Emprcal Assessmen 2001-19 Drec Foregn Invesmens and Producvy Growh n Hungaran Frms, 1992-1999 2001-18 Marke Access Maps: A Blaeral and Dsaggregaed Measure of Marke Access 2001-17 Macroeconomc Consequences of Penson Reforms n Europe: An Invesgaon wh he INGENUE World B. Coeuré A. Jousen & F. Legros N. Mulder, S. Monou & L. Peres Lopes M. Bene, A. Benassy-Quéré, E. Dauchy & R. MacDonald M. Bene, A. Benassy-Quéré, E. Dauch & R. MacDonald M.H. Bchr & M. Maurel F. Lemone & D. Ünal-Kesenc K Head & T. Mayer N. Kousnezoff A. Kadareja J. Sgard A. Bouë, L. Fonagné, M. Mmoun & X. Pcho Equpe Ingénue
Technology dfferences, nsuons and economc growh : a condonal condonal convergence Model 2001-16* La producvé des ndusres méderranéennes A. Chevaller & D. Ünal-Kesenc 2001-15 Marmoe: A Mulnaonal Model L. Cadou, S. Dees, S. Guchard, A. Kadareja, J.P. Laffargue & B. Rzepkowsk 2001-14 The French-German Producvy Comparson Revsed: Ten Years Afer he German Unfcaon 2001-13* The Naure of Specalzaon Maers for Growh: An Emprcal Invesgaon 2001-12 Forum Economque Franco-Allemand - Deusch- Französsches Wrschafspolsches Forum, Polcal Economy of he Nce Treay: Rebalancng he EU Councl and he Fuure of European Agrculural Polces, 9 h meeng, Pars, June 26 h 2001 L. Nayman & D. Ünal-Kesenc I. Bensdoun, G. Gauler & D. Ünal-Kesenc 2001-11 Secor Sensvy o Exchange Rae Flucuaons M. Fouqun, K. Sekka, J. Malek Mansour, N. Mulder & L. Nayman 2001-10* A Frs Assessmen of Envronmen-Relaed Trade Barrers 2001-09 Inernaonal Trade and Rend Sharng n Developed and Developng Counres L. Fonagné, F. von Krchbach & M. Mmoun L. Fonagné & D. Mrza 2001-08 Econome de la ranson : le dosser G. Wld 2001-07 Ex Opons for Argenna wh a Specal Focus on Ther Impac on Exernal Trade 2001-06 Effe fronère, négraon économque e 'Foreresse Europe' 2001-05 Forum Économque Franco-Allemand Deusch- Französsches Wrschafspolsches Forum, The Impac of Easern Enlargemen on EU-Labour Markes and Pensons Reforms beween Economc and Polcal Problems, 8 h meeng, Pars, January 16 2001 S. Chauvn T. Mayer 61
CEPII, Workng Paper No 2004-02 2001-04 Dscrmnaon commercale : une mesure à parr des flux blaéraux 2001-03* Heerogeneous Expecaons, Currency Opons and he Euro/Dollar Exchange Rae G. Gauler B. Rzepkowsk 2001-02 Defnng Consumpon Behavor n a Mul-Counry Model 2001-01 Pouvor prédcf de la volalé mplce dans le prx des opons de change 2000-22 Forum Economque Franco-Allemand - Deusch- Französsches Wrschafspolsches Forum, Trade Rules and Global Governance: A long Term Agenda and The Fuure of Bankng n Europe, 7 h meeng, Pars, July 3-4 2000 2000-21 The Wage Curve: he Lessons of an Esmaon Over a Panel of Counres 2000-20 A Compuaonal General Equlbrum Model wh Vnage Capal 2000-19 Consumpon Hab and Equy Premum n he G7 Counres 2000-18 Capal Sock and Producvy n French Transpor: An Inernaonal Comparson 2000-17 Programme de raval 2001 2000-16 La geson des crses de lqudé nernaonale : logque de falle, prêeur en derner ressor e condonnalé O. Allas, L. Cadou & S. Dées B. Rzepkowsk S. Guchard & J.P. Laffargue L. Cadou, S. Dées & J.P. Laffargue O. Allas, L. Cadou & S. Dées B. Chane Kune & N. Mulder J. Sgard 2000-15 La mesure des proecons commercales naonales A. Bouë 2000-14 The Convergence of Auomoble Prces n he European Unon: An Emprcal Analyss for he Perod 1993-1999 2000-13* Inernaonal Trade and Frms Heerogeney Under Monopolsc Compeon 2000-12 Syndrome, mracle, modèle polder e aures spécfcés néerlandases : quels ensegnemens pour l emplo en France? G. Gauler & S. Haller S. Jean S. Jean 2000-11 FDI and he Openng Up of Chna s Economy F. Lemone 2000-10 Bg and Small Currences: The Regonal Connecon A. Bénassy-Quéré & B. Coeuré 62
Technology dfferences, nsuons and economc growh : a condonal condonal convergence 2000-09* Srucural Changes n Asa And Growh Prospecs Afer he Crss 2000-08 The Inernaonal Moneary Fund and he Inernaonal Fnancal Archecure 2000-07 The Effec of Inernaonal Trade on Labour-Demand Elasces: Inersecoral Maers 2000-06 Foregn Drec Invesmen and he Prospecs for Tax Co-Ordnaon n Europe 2000-05 Forum Economque Franco-Allemand - Deusch- Französsches Wrschafspolsches Forum, Economc Growh n Europe Enerng a New Area?/The Frs Year of EMU, 6 h meeng, Bonn, January 17-18, 2000 2000-04* The Expecaons of Hong Kong Dollar Devaluaon and her Deermnans 2000-03 Wha Drove Relave Wages n France? Srucural Decomposon Analyss n a General Equlbrum Framework, 1970-1992 2000-02 Le passage des reraes de la réparon à la capalsaon oblgaore : des smulaons à l ade d une maquee 2000-01* Rappor d acvé 1999 1999-16 Exchange Rae Sraeges n he Compeon for Aracng FDI 1999-15 Groupe d échanges e de réflexon sur la Caspenne. Recuel des compes-rendus de réunon (déc. 97- oc. 98)" 1999-14 The Impac of Foregn Exchange Inervenons: New Evdence from FIGARCH Esmaons 1999-13 Forum Economque Franco-Allemand - Deusch- Französsches Wrschafspolsches Forum, Reducon of Workng Tme/Easward Enlargmen of he European Unon, 5 h meeng, Pars, July 6-7 1999 1999-11* La dversé des marchés du raval en Europe : Quelles conséquences pour l Unon Monéare ; 63 J.C. Berhélemy & S. Chauvn M. Aglea S. Jean A. Bénéssy-Quéré, L. Fonagné & A. Lahrèche-Révl B. Rzepkowsk S. Jean & O. Bonou O. Rougue & P. Vlla A. Bénassy-Quéré, L. Fonagné & A. Lahrèche-Révl D. Panell & G. Sokoloff M. Bene, A. Bénassy-Quéré & C. Lecour 1999-12* A Lender of Las Resor for Europe M. Aglea L. Cadou, S. Guchard
CEPII, Workng Paper No 2004-02 Deuxème pare : Les mplcaons macroéconomques de la dversé des marchés du raval 1999-10* La dversé des marchés du raval en Europe : Quelles conséquences pour l Unon Monéare ; Premère pare : La dversé des marchés du raval dans les pays de l Unon Européenne 1999-09 The Role of Exernal Varables n he Chnese Economy; Smulaons from a macroeconomerc model of Chna 1999-08 Haue echnologe e échelles de qualé : de fores asyméres en Europe 1999-07 The Role of Capal Accumulon, Adjusmen and Srucural Change for Economc Take-Off: Emprcal Evdence from Afrcan Growh Epsodes 1999-06 Enerprse Adjusmen and he Role of Bank Cred n Russa: Evdence from a 420 Frm s Qualave Survey 1999-05 Cenral and Easern European Counres n he Inernaonal Dvson of Labour n Europe 1999-04 Forum Economque Franco-Allemand Economc Polcy Coordnaon 4 h meeng, Bonn, January 11-12 1999 1999-03 Models of Exchange Rae Expecaons: Heerogeneous Evdence From Panel Daa 1999-02 Forum Economque Franco-Allemand Labour Marke & Tax Polcy n he EMU 1999-01 Programme de raval 1999 & M. Maurel L. Cadou & S. Guchard S. Dees L. Fonagné, M. Freudenberg & D. Ünal-Kesenc J.C. Berhélemy & L. Söderlng S. Brana, M. Maurel & J. Sgard M. Freudenberg & F. Lemone A. Bénassy-Quéré, S. Larrbeau & R. MacDonald 64
CEPII, Workng Paper No 2003 -?? CEPII DOCUMENTS DE TRAVAIL / WORKING PAPERS S vous souhaez recevor des Documens de raval, merc de remplr le coupon-réponse c-jon e de le reourner à : Should you wsh o receve copes of he CEPII s Workng papers, jus fll he reply card and reurn o: Sylve HURION Publcaons CEPII 9, rue Georges-Pard 75740 Pars Fax : (33) 1.53.68.55.04 M./Mme / Mr./Mrs... Nom-Prénom / Name-Frs name... Tre / Tle... Servce / Deparmen... Organsme / Organsaon... Adresse / Address... Vlle & CP / Cy & pos code... Pays / Counry... Tél.... Désre recevor les Documen de raval du CEPII n : Wsh o receve he CEPII s Workng Papers No:........................ Souhae êre placé sur la lse de dffuson permanene (pour les bblohèques) Wsh o be placed on he sandng malng ls (for Lbrares).