Balassa-Samuelson: Still haven t found what we re looking for?

Transcription

1 Balassa-Samuelson: Sill haven found wha we re looking for? rond-arne Borgersen Øsfold Universiy College Faculy of Economics, Social Sciences and Languages Remmen N Halden Roswiha M King Øsfold Universiy College Faculy of Economics, Social Sciences and Languages Remmen N Halden roswihamking@hiofno Absrac Despie is srong heoreical posiion when i comes o explaining inflaion in ransiion economies, he empirical meris of he Balassa-Samuelson (B-S) effec are weak his paper poins o hree heory-based conribuing facors o hese weak findings and offers an alernaive approach Firs, he shor-erm focus makes B-S prone o underesimaing he magniude of he produciviy growh differenial Second, he convenional demand side CPI based secoral composiion reduces he exen o which he produciviy growh differenial is passed hrough o inflaion hird, by ignoring he dependence beween he wo main B-S componens, a second downward bias o he produciviy growh pass hrough comes abou he key o our proposed alernaive cenres around an endogenous relaion beween he produciviy growh differenial and secor sizes he long-run supply-side approach allows us o capure inflaion drivers ha convenional B-S fails o incorporae he paper provides a blueprin for fuure empirical analysis JEL classificaions: P2, O14 Keywords: Balassa-Samuelson, srucural inflaion, Scandinavian Model of Inflaion, reallocaion, resrucuring Acknowledgemens We hank he paricipans of he 23 rd Research Seminar of he Managing Economic ransiion nework a he Universiy of Brighon on May 8, 2009 and of he inernaional conference a he allinn School of Economics and Business Adminisraion Economies of Cenral and Easern Europe: Convergence, Opporuniies and Challenges, June 14-16, 2009, for helpful commens on an earlier draf of his paper 1

2 1 Inroducion Despie is srong heoreical posiion when i comes o explaining inflaion and real exchange rae developmens in ransiion economies, he empirical meris of he Balassa-Samuelson (B-S) effec are weak 1 In fac, over he las years several papers have claimed he B-S effec irrelevan for he difficulies ransiion economies in easern-europe have encounered in complying wih he Maasrich inflaion crierion 2 Irrespecive of weak empirical esimaes, he role of srucural inflaion should no be aken lighly Counries ransiioning from cenrally planned economies o marke economies experience large srucural changes in compressed ime Common sense suggess ha srucural shifs of such magniude will be accompanied by inflaion originaing in srucural causes If here is anywhere a showcase for srucural inflaion i should be found in his ype of economy For decades economiss have been araced by hese feaures, rying o highligh causal relaions beween he epochal srucural shifs and inflaion in paricular along he lines of B-S 3 However, quesions regarding how he B-S effec ypically is analyzed seem o remain he purpose of his paper is firs of all o idenify some of he heory-based reasons for why he empirical suppor for srucural inflaion, as manifesed by he B-S effec, has been so disappoining Second, he paper proposes an alernaive concepual approach he paper essenially argues ha he reason why we sill have no found wha we are looking for is ha we have been looking in he wrong places he lieraure on he B-S effec, as described by for example Éger (2002a), concern iself wih how produciviy growh in he radable secor oupaces ha of he non-radable secor Under he assumpion ha wage growh is roughly equal beween secors, wages increase in boh secors As a consequence he relaive price of non-radable goods rises, and inflaion picks up his way he heory esablishes a causal relaion beween secoral produciviy growh and srucural inflaion he B-S analysis has gone hrough a number of variaions and exensions he original supply side orienaion by Balassa and Samuelson 4 has been exended o include demand side variables, such as per capia income and governmen expendiures 5 Recen applicaions also disinguish beween a so-called inernal (or domesic) B-S effec and an exernal (or inernaional) one Boh versions of he B-S effec are 1 For example, for a selecion of Eas European economies Mihaljek and Klau (2008) show, on average, he shor run domesic B-S effec o be 013%, he long run domesic B-S effec 11%, he shor run inernaional B-S effec 01%, and he long run inernaional B-S effec 12% Mihaljek and Klau (2004) repor ha he B-S effec, in counries of Cenral and Easern Europe, on average only explains beween 02 % and 2 % of he annual inflaion differenials relaive o he euro area Small effecs are also repored, for example, Cipriani (2000), Coricelli & Jazbec (2001), Éger (2002a), Éger (2002b), Éger e al (2003), Flek e al (2003), and Kovács (2002) 2 For example, Éger (2005) concludes ha he B-S effec plays o 2 For example, for a selecion of Eas European economies Mihaljek and Klau (2008) show, on average, he shor run domesic B-S effec o be 013%, he long run domesic B-S effec 11%, he shor run inernaional B-S effec 01%, and he long run inernaional B-S effec 12% Mihaljek and Klau (2004) repor ha he B-S effec, in counries of Cenral and Easern Europe, on average only explains beween 02 % and 2 % of he annual inflaion differenials relaive o he euro area Small effecs are also repored, for example, Cipriani (2000), Coricelli & Jazbec (2001), Éger (2002a), Éger (2002b), Éger e al (2003), Flek e al (2003), and Kovács (2002) nly a very limied role in he deerminaion of inflaion and real exchange rae movemens Breuss (2003) concludes ha he B-S effec is no obsacle for he eas European euro area aspirans Furhermore, he hypohesis ha srucural facors relaed o changes in produciviy are NO significan inflaion drivers in ransiion economies has promped a search for oher causal relaions germain o he ransiion experience For example Éger and Podpiera (2008) advance an alernaive explanaion in heir qualiy bias hypohesis 3 For a review of he lieraure on he B-S effec, see for example, Éger (2003) or Éger e al (2006) 4 See Balassa (1964) and Samuelson (1964) See also Rogoff (1996) or Obsfeld & Rogoff (1996) for a descripion of he original supply side perspecive of B-S analysis 5 For B-S analysis exended o include he demand side, see Bergsrand (1991) or De Gregorio e al (1994) 2

3 hypoheses abou srucural causes of inflaion he inernaional version focuses on he endency for inflaion in caching-up economies o exceed ha of he economies hey are converging o In he domesic version he B-S effec represens he endency for he domesic prices of non-radable goods o rise faser han he prices of radables (Mihaljek and Klau, 2008) In his paper we argue ha B-S analysis is bese by hree unforunae feaures ha preven i from capuring ha which i is seeking hese feaures reside on he heoreical side Firs, alhough usually hough of as a supply side model, closer examinaion of is ypical formulaion reveals a clear demand side channel Second, and more imporanly, srucural changes evolve over he longer erm However, by consrucion, B-S has a shor-erm orienaion hird, he B-S effec reas as independen wha in realiy are srongly relaed elemens: he produciviy growh differenial beween secors and he secors relaive magniude aking hese shorcomings ino accoun suggess a blueprin for an alernaive approach when esimaing srucural inflaion and he B-S effec his enails a re-focusing on he supply side, he longer erm and on he inerdependence beween he wo main facors deermining how srucural shifs impac on inflaion his inerdependence allows conemporaneous-, as well as leading- and lagging values of he produciviy growh differenial o impac on secor sizes, and, evenually srucural inflaion In he absence of leading- or lagging relaions beween he produciviy growh differenial and secor sizes, our approach reduces o he convenional B-S approach he long-run perspecive implies ha we absrac from (ofen erraic) year-on-year changes of srucural inflaion drivers, in favor of cumulaive changes and heir annual averages, in line wih he longerm processes underlying ransiion Srucural inflaion can now vary boh over ime and beween counries a differen sages of ransiion for wo reasons: changes in he produciviy growh differenial, and changes in how he differenial is passed hrough o inflaion In paricular, by endogenising relaive secor sizes, we can assess how he impac of he produciviy growh pass-hrough on srucural inflaion changes over ime In his conex i is useful o regard he B-S effec as a componen of a more general varian of he srucural inflaion hypohesis - decidedly anchored in he supply side and decidedly long-erm - ha has become known as he Scandinavian Model of Inflaion 6 Srucural inflaion is he conribuion o inflaion from he real side of he economy As he real facors are relaed o he srucural shifs a he hear of he ransiion process hey hold a promise of fuure growh and higher sandard of living Srucural inflaion should hus be less worrisome o policy makers han he nominal (moneary) conribuion o inflaion Ye, he Maasrich Convergence Crierion on inflaion, which has been a barrier o euro zone enry for a number of Eas European EU member saes, does no disinguish beween real and nominal conribuions o inflaion Considering ha he Eas European new EU member saes sill have a long way o go o be a par wih he sandard of living of Wes European economies and heir economic srucures, i is o be expeced ha srucural inflaion will persis ino he foreseeable fuure If srucural inflaion accouns for a significan porion of overall inflaion, and is expeced o persis, i pus a quesion mark over he appropriaeness of he Maasrich convergence crieria for ransiion economies aspiring o join he euro area In his conex he magniude of he srucural componen o overall inflaion is significan for policymakers So far, radiional B-S analysis wih is weak empirical suppor has backed up Wes European policy-makers opinions ha special consideraions for ransiion economies are no warraned 7 In his paper we presen analyical foundaions for a poenial reconsideraion of his opinion 6 See Lindbeck (1979) for an exposiion of he Scandinavian Model of Inflaion 7 he official sance of he European Commission and he European Cenral bank, expressed in 2001 as follows, has essenially no changed: nowihsanding is imporance, he Balassa-Samuelson effec should no be oversaed when explaining curren inflaion raes in he accession counries he Maasrich inflaion crierion, which will no be revised o ake ino accoun any possible Balassa-Samuelson effec, should no be seen as an immediae requiremen for hese counries, bu raher as a medium-erm objecive for cenral banks (Chrisian Noyer, 2001) 3

4 he paper is srucured as follows: In Secion 2 we documen he disincion beween shor- and long-erm measures of B-S and srucural inflaion, and pinpoin he shorcomings of he former A long-erm supply side focused approach for srucural inflaion based on he Scandinavian Model of Inflaion is derived in Secion 3 he model endogenises he relaionship beween secor sizes and he produciviy growh differenial, he wo main componens of he B-S effec, and highlighs he disincion beween shor- and long-run measures he implicaions for ransiion are developed in Secion 4 he relaionship beween he produciviy growh pass-hrough and is underlying causes is discussed in Secion 5 he las par concludes 2 Concepual consideraions: Srucural inflaion and Balassa-Samuelson When he empirical findings of he B-S effec are declared o be weak wo ypes of aiudes may be conveyed One perspecive is ha he B-S effec is weak because srucural inflaion drivers are weak, and ha - like i or no - his is simply wha he daa ells us (Éger and Podpiera, 2008) he oher perspecive is ha he daa harbors srong evidence of srucural inflaion drivers, bu he convenional B-S mehodology fails o capure i I is his laer perspecive ha we pursue in his secion he core of our argumen is ha long-run supply side ransiion dynamics come in wo layers, which in he following is referred o as reallocaion and resrucuring, erms which are o be defined laer he paper argues ha he convenional B- S effec capures he former, bu no he laer In he shor-erm he relaive sizes of he radable and he non-radable secor (key elemens in B-S analysis) are fixed Profi-seeking behavior is confined o finding improvemens in facor proporions, o increase produciviy, produc qualiy, improve work organizaion and managerial pracices, and oher aciviies compaible wih fixed relaive secor sizes Economic expansion or decline would affec all secors simulaneously he economy may grow or shrink, bu relaive secor shares remain unchanged Blanchard (1997), for example, poins ou ha he sark shrinkage of oupu in he very beginning of ransiion was disribued evenly across all secors his would fi wihin our frame of reference of fixed relaive secor shares he accompanying impac on inflaion is a shor-erm effec However, Blanchard (1997) coninues wih over ime, here has been increasing differeniaion across secors, and here now appears o be slow convergence of secoral composiion o ha observed in OECD counries 8 his describes wha is mean by long-erm or large scale dynamics: Changes in relaive secor sizes hrough reallocaion or resrucuring processes We will refer o he effecs of changing relaive secor sizes on inflaion as long-erm effecs In his frame of reference we define ransiion as a shif oward a larger share of he non-radable secor - essenially in agreemen wih Blanchard (1997) Viewed from his perspecive he srucural shifs a he very hear of ransiion essenially remain ouside he convenional B-S analysis his may be an obvious reason for is weak empirical resuls Anoher poenial cause is ha relaive secor sizes are ypically defined as shares in he CPI raher han in GDP We will address hese poenial reasons for he weak B-S effec one by one, and in more deail below Le us ake as a poin of deparure he definiion of srucural inflaion ( P SR ) as suggesed by he Scandinavian Model of Inflaion, which is in congruence wih B-S mehodology 9 1) P = (( 1 )[ ]) SR where (1-) denoes he relaive size of he non-radable secor and [ q ] N q N he produciviy growh differenial beween he radable and non-radable secor he original srucural inflaion hypohesis does 8 Blanchard (1997), pp 5 9 See again Lindbeck (1979) for he Scandinanvian Model of Inflaion, and Mihaljek and Klau (2008) for he convenional B-S framework 4

5 no concern iself wih iner-emporal consideraions, and is hus, in principle, subjec o a variey of exensions and inerpreaions he wo dominan varians are: (i) Fixed relaive size of he non-radable secor pre-muliplying a sum of ime-indexed produciviy growh differenials 10 2a) (1 ) j= ( N ) j (ii) Alernaively, he producs of ime-indexed produciviy growh differenials and imeindexed relaive size of he non-radable secor summed over he relevan ime horizon Noe ha here conemporaneous non-radable secor size and conemporaneous produciviy growh differenials are he facors of he producs, ie ( 1 ) and ( N ) have always he same ime index 11 2b) ( 1 ) j ( N ) j j= Expressions (2a) and (2b) are inerpreed as srucural inflaion over a period of ime, ie he par of inflaion ha has is origin in he real side of he economy, as opposed o he moneary side o obain average annual values of srucural inflaion expression (2a) and/or (2b) are divided by he number of years under observaion hree commens are in order: (i) As menioned above, analysis of he B-S effec ypically compues he relaive size of he non-radable secor, (1 ), by aking he secor s share in he CPI raher han is share in GDP his, while analyically no incorrec, has been shown o underesimae he secor s size, and, by implicaion, underesimae he B-S effec his is paricularly so for he ransiion economies of Easern Europe For example, as Mihaljek and Klau (2008) poin ou, marke-based non-radables are responsible for only around 20%-30% of he Balic Saes CPI baske, while ypically conribuing more han 60% o valued added his implies ha swiching from CPI o GDP when calculaing he size of he non-radable secor will augmen he B-S effec Even so, CPI-based (ie demand side) secoral share compuaions are doubful when explaining srucural inflaion (ii) Sandard B-S analysis akes he size of he non-radable secor (1 ) and he produciviy growh differences beween radable and non-radable secors ( N ) as independen of each oher A dependency relaion is however argued by Borgersen and King (2009) In paricular, relaive secor sizes are argued o depend on he produciviy growh differenial beween secors So, if for example produciviy growh in he radable secor is sronger han in he non-radable secor he radable secor will expand relaive o he non-radable secor (iii) Acceping a dependence relaion, as described in (ii), and defining relaive secor size as a funcion of produciviy growh differences, de faco rules ou expression (2a) since here he secor size (1 ) is consan, as indicaed by he absence of a ime index In conras o (2a), expression (2b) accommodaes a dependence of he relaive secor size (1 ) on he produciviy growh difference ( N ) However, a new quesion arises: Why should secor size a ime depend only on produciviy growh differenial a ime? Would i no be more plausible ha secor size a ime also depends on produciviy growh in earlier years? Afer all, i will ake ime o change secor sizes in response o produciviy growh signals And furher, is no plausible ha secor sizes depend 10 he majoriy of papers dealing wih he B-S effec are of his kind, as a review of he survey papers Éger, Balázs (2003), and/or Éger e al (2006) reveals 11 For example Mihaljek and Klau (2008) ake his approach 5

6 on expecaions abou fuure produciviy growh? Such expecaions could, for example be raised by announcemens of economic policy ha favors a paricular secor his would arac enrepreneurial aciviy and invesmen ino his secor even before such policy is enaced, in order o secure a firs move advanage In oher words, should we no consider boh leads and lags in he relaionship beween he produciviy growh differenial and secor sizes? his also fis nicely wih he long-erm orienaion of he Scandinavian Model of Inflaion and is srong anchoring in he supply side he mahemaical expression analogous o (2b), bu in conras o (2b), capuring leads and lags in he above sense is given by 12 3) ( 1 ) ( ) j= j i= N i We presen expression (3) as an alernaive formulaion of srucural inflaion, and refer o i as cumulaive srucural inflaion As before, average annual values of cumulaive srucural inflaion are obained by dividing he expression by he number of years under observaion In he following we refer o expression 3) as he long-run B-S effec As shown in Borgersen and King (2009) for he case of Lavia, in a simple accouning framework, an analysis along he lines of cumulaive srucural inflaion assigns a srong role o srucural facors impacing on inflaion In sandard B-S analysis erraic year-on-year jumps beween advances in produciviy of he radable secor alernaing wih advances in produciviy in he non-radable secor make hese manifesaions of shor- erm supply side facors effecively cancelling each oher ou, giving he misleading impression ha srucural facors exer only a weak influence on inflaion Defining srucural inflaion according o equaion (3) miigaes hese muually cancelling effecs, and conribues o a sronger impac of srucural inflaion drivers Furhermore, expression (3) incorporaes leads and lags in he relaion beween he relaive size of he secors and he produciviy growh differenial ( N ) Srucural shifs depend on an array of pas, curren and expeced fuure produciviy growh differenials, acing as signals for profi seeking invesors In addiion o he curren produciviy growh differenial, invesmens are moivaed by expecaions of upcoming produciviy growh in a secor (leads), and pas produciviy growh differenials which migh have lagging effecs on invesmens and secor sizes So, using expression (3) serves a dual purpose: Firs, i miigaes he cancelling ou effec of yearon-year reversals in produciviy growh differenials beween radable and non-radable secor Second, and more significanly, equaion (3) capures somehing ha is missed by sandard B-S analysis: he longer erm srucural shifs a he hear of ransiion ha bring abou changes in secoral composiion For illusraive purposes we expand expression (3), and arrange is erms in marix form 12 Noe ha while expression (2b) is a sum of producs, expression (3) is a produc of sums 6

7 able 1: Produciviy growh differenials and secor size (1-) ( q ) N (1-) ( q ) N +1 (1-) ( q N ) +2 (1-) ( q N ) -1 (1-) ( q N ) (1- ) +1 ( q ) N (1-) ( +1 q ) N +1 (1-) ( +1 q N ) +2 (1-) ( +1 q N ) -1 (1-) ( +1 q N ) (1- ) +2 ( q ) N (1-) ( +2 q ) N +1 (1-) ( +2 q N ) +2 (1-) ( +2 q N ) -1 (1-) ( +2 q N ) (1- ) +3 ( q ) N (1-) ( +3 q ) N +1 (1-) ( +3 ) N +2 (1-) +3 ( q ) N -1 (1-) ( +3 q N ) (1- ) -1 ( q ) N (1-) ( -1 q ) N +1 (1-) ( -1 q ) N +2 (1-) ( -1 q ) N -1 (1-) ( -1 q N ) (1- ) ( q ) N (1-) ( q ) N +1 (1-) ( q ) N +2 (1-) ( q ) N -1 (1-) ( q N ) Noe ha expression (2a) represens one fixed bu arbirary row of he marix, while expression (2b) represens is principal diagonal (indicaed by he shaded cells in he marix) By conras, expression (3) represens he enire marix he disincion beween couning he diagonal, or one of he rows, and he full marix illusraes he disincion beween he shor- and he long-erm B-S approach able 1 will be useful when in Secion 5 we analyze how srucural inflaion drivers evolve In paricular, for visualizing how leading and lagging produciviy growh differenials migh impac on srucural inflaion 13 he erms o he lef of he (shaded) principal diagonal lend hemselves o he inerpreaion of lags in he above described sense hese are he erms of a sub-sum of (3), given by 4a) j= ( 1 ) ( ), j i= N i while he erms o he righ of he principal diagonal are inerpreed as leads, and are represened by a subsum of (3) given by 1 j i= + 1 4b) ( 1 ) ( ) j= N i 3 A long-run supply side approach o srucural Inflaion he Scandinavian Model of Inflaion incorporaes he Balassa-Samuelson effec ino a comprehensive longrun framework, where wage growh is equalized beween secors allowing he produciviy growh 13 In an economeric analysis he weighs assigned o he various marix elemens in compuing he value of expression (3) for a given daase are esimaed, ie would be endogenous o he model 7

8 differenial o simulaneously impac he radable and he non-radable secor he formaion of wages is relaed o he produciviy differenial beween secors, and hence he B-S effec he relaionship beween wages and he produciviy differenial comes abou o mainain fixed income shares for boh capial and labour In he model, he demand side is implicily suppressed, highlighing he supply side deerminans of srucural inflaion Consider a small open economy, spli ino a radable and a non-radable secor along he lines of Lindbeck (1979), where labour and capial are used as inpus in boh secors of producion he wo secors differ when i comes o pricing, as purchasing power pariy is assumed o govern pricing of radable goods In he non-radable secor here is mark-up pricing, as prices are se in relaion o uni labour coss Labour is homogenous and facor income shares are fixed his in basic consrains he model o he long run Expressing all variables as raes of change, he srucural approach o inflaion reads 5) P = p + e + ( 1 )[ ] W where (P) equals he domesic price index, ( pw ) world marke prices and (e) he exchange rae In addiion, produciviy growh in he radable ( q ) and in he non-radable secor ( q N ) impac inflaion, where he produciviy growh differenial is scaled down by he size of he non-radable secor, ( 1 ), measuring he share of aggregae oupu of he non-radable secor In equaion 5) he las erm represens P = ( 1 ), as already inroduced in equaion 1 srucural inflaion, ie [ ] SR N his represenaion of srucural inflaion is common o boh convenional B-S analysis and he Scandinavian Model of Inflaion For highlighing he disincion beween convenional B-S analysis, and our exended version of he Scandinavian Model of Inflaion, le us consider he impac on srucural inflaion of an increase in he produciviy growh differenial As i is only he difference in produciviy growh beween secors ha maers, we simplify he exposiion by assuming ha produciviy growh in he non-radable secor is zero ( N = 0) ) his makes he produciviy growh differenial equal o produciviy growh in he radable secor Le us also assume ha secor sizes are fixed in he shor-run, moivaed by he discussion in secion 2 Now, he shor-run produciviy growh pass-hrough equals dp SR 7) = ( 1 ) > 0 d he shor-run B-S effec is posiive when he produciviy growh differenial increases he exen of produciviy growh pass-hrough is deermined by he size of he non-radable secor However, as discussed above, he shor-erm analysis mus be exended when analysing long-erm shifs In paricular, he fixed secor size assumpion mus be relaxed his is wha our exension of he Scandinavian Model of Inflaion inroduces According o he B-S effec, srucural inflaion has wo componens: he produciviy growh differenial and how his produciviy growh differenial is passed hrough o inflaion Variaions in srucural inflaion can now come from wo sources: direcly via changes in he produciviy growh differenial, and indirecly hrough variaions in how he produciviy growh differenial is passed-hrough o inflaion While equaion 7) akes ino accoun he former of hese wo, he laer is ignored By endogenising secor sizes, variaions in produciviy growh pass-hrough are also accouned for hinking of secor sizes as a consequence of produciviy growh differenials moivaes our exension of he Scandinavian Model of Inflaion: o define he relaive size of he radable secor as an increasing funcion of he produciviy growh differenial 14 N 14 I is assumed ha = ( ) is coninuous and differeniable 8

9 ' 8) ( q ) where ( q ) > 0 = he posiive relaionship beween he produciviy growh differenial and he radable secor implies a negaive relaionship beween he produciviy growh differenial and he non-radable secor his defines an inverse relaionship beween he produciviy growh differenial and he produciviy growh passhrough Hence, even if he produciviy growh differenial decreases, srucural inflaion can sill increase his as long as he secor size response is subsanial, and he increase in produciviy growh pass-hrough more han ouweighs he reducion in he produciviy growh differenial Hence, he long-run approach adds complexiy o he analysis of srucural inflaion and he B-S effec he combined effec on srucural inflaion when secor sizes are endogenous is analysed by Borgersen and King (2009), disinguishing beween a produciviy effec (PE) and a size effec (SE) Whereas he former incorporaes he direc effec on srucural inflaion from he produciviy growh differenial, he laer capures he indirec effec following he secor size response and he changing produciviy growh pass-hrough 15 he combined effec on srucural inflaion is: 9) dp d SR = ( 1 ) q ( q ) '( q ) q PE SE Figure 1 gives a schemaic relaionship beween a number of differen produciviy measures in he radable secor and he secor size he size of he radable secor is measured on he horizonal axis, while he verical axis measures radable secor produciviy he graph shows he relaion beween he wo, and illusraes he relaionship beween he size of he radable secor and he produciviy growh informaion se (highlighed below he firs axes) Figure 1: he relaionship beween he size of he radable secor and is produciviy measures q * I II III IV > 0 > 0 < 0 < 0 > 0 < 0 < 0 > 0 15 See Borgersen and King (2009) for comparaive saics 9

10 he inerpreaion of figure 1 can briefly be described as follows: When produciviy is a is peak, he radable secor is close o full resource uilizaion When he radable secor is smaller han is opimal size (for insance in regime ii), he posiive produciviy growh will make he secor expand When radable secor size is above is opimal value, he negaive produciviy growh promps a conracion In he regimes II and III, where he acceleraion of radable secor produciviy growh decreases, he size of he radable secor adjus owards full capial uilisaion In regime II, where he radable secor increases (and he nonradable secor decreases), he increase in he produciviy growh differenial simulaing srucural inflaion, is couneraced by he reducion in produciviy growh pass- hrough In regime iii on he oher hand, where he radable secor decreases, he reducion in he produciviy growh differenial is accompanied by an increase in he non-radable secor, ie in he produciviy growh pass-hrough he endogenous secor size allows us o address he combined impac on srucural inflaion following an increased produciviy growh differenial, which can be eiher posiive or negaive By dp SR manipulaing equaion 9) he condiions for when > 0 can be expressed as: d 1 10a) q > 1 10b) q < El when ( < 0) El when ( > 0) '( ) where El = is he secor size elasiciy, measuring how he disribuion of value added beween secors is affeced by a one percen increase in he produciviy growh differenial he condiions for when an increased produciviy growh differenial increases srucural inflaion, depends on he relaionship beween he secor size elasiciy and he prevailing disribuion of value added, bu varies according o acceleraion of he produciviy growh differenial In he cenre of figure 1, srucural inflaion will increase if he secor size elasiciy exceeds he exising disribuion of value added 16 In he wings, he secor size elasiciy has o be smaller han he prevailing disribuion of value added for srucural inflaion o increase Saed differenly, in he cenre a higher produciviy growh differenial will increase srucural inflaion when he size effec exceeds he produciviy effec When he secor size elasiciy exceeds he exising disribuion of value added, he higher produciviy growh pass-hrough will compensae for he reduced produciviy growh differenial In he wings on he oher hand, where he secor size elasiciy is smaller han he prevailing disribuion of value added, he reduced pass-hrough has o be accompanied by an increased produciviy growh differenial for srucural inflaion o increase When comparing he shor-run B-S analysis o our long-run B-S approach, some ineresing observaions come abou: While in he shor-run framework he condiions for inflaionary pressure is only exered in regimes I and II of figure 1 (see equaion 7), i can be associaed wih all four regimes in he longrun (see equaions 10a and 10b) his illusraes why he shor-run B-S approach migh undersae he B-S effec, being suscepible o erraic year-on-year produciviy growh differenials beween secors he endogenous produciviy growh pass-hrough gives a more differeniaed undersanding of how produciviy growh differenials impac on srucural inflaion ha is, alhough he shor-run B-S effec is negaive he long-run B-S effec can sill be posiive 16 he condiions for when srucural inflaion decreases are given by condiions (10a) and (10b), bu now he inequaliies are reversed 10

11 4 Reallocaion and resrucuring While he shor-erm B-S effec is associaed wih fixed secor sizes, he long-run approach in Secion 3 allows for changes in secor sizes, bu keeps he opimal disribuion of value added fixed According o he reasoning in Secion 2, abou he differen version of srucural inflaion applied in he B-S analysis, he shor-run B-S effec where secor sizes are fixed is represened by expression (2a) Likewise, expression (2b), allowing for changes in secor sizes, can be associaed wih our long-run B-S effec above Sill, he reasoning regarding srucural inflaion should also allow for poenial shifs in he opimal disribuion of value added ransiion is, as highlighed by Blanchard (1997)and described earlier, associaed wih changes in he opimal disribuion of value added beween secors over ime More specific, ransiion is defined as a relaive decrease (increase) in he radable (non-radable) secors opimal share of valued added as he economy maures Hence, in addiion o he reallocaion effec capured by equaion (2b), we also allow for a more fundamenal srucural shif, referred o as resrucuring hese laer shifs are associaed wih changes in opimal secor sizes While reallocaion implies movemens along one paricular curve in figure 2 below, resrucuring encompasses shifs o anoher of he family of curves Boh reallocaion and resrucuring are relaed o changes in he produciviy growh differenial ha is, increased produciviy in a secor increases profis and rens, which ac as a signal for shifing of resources ino he secor, hrough changes in relaive prices However, while reallocaion is characerised by expansions - or conracions - of resources highly sensiive o he curren produciviy growh differenial, he shifing of resources accompanying resrucuring requires somehing more han jus insananeous simulus I depends on he underlying momenum in he produciviy growh differenial, incorporaing boh curren-, pas-, and expeced fuure produciviy growh differenials Hence, resrucuring is relaed o he measure of srucural inflaion given by expression 3 Figure 2 shows wo of a family of curves, which in heir enirey race ou a counry s ransiion rajecory, including boh reallocaion and resrucuring As ransiion is defined as changes in he disribuion of value added owards he non radable secor, he closer o he origin a curve is locaed, he more maure he economy he dynamics of resrucuring are linked o reallocaion Consider an economy iniially a poin B (equivalen wih regime iii in figure 1), where resources are no opimised, given he * Assuming ha condiion 10a) holds srucural inflaion increases opimal disribuion of value added ( ) 2 11

12 Figure 2: Reallocaion and resrucuring he produciviy informaion se is relaed o ( 2 ) q * 1 * 2 A B C 1 2 > 0 > 0 < 0 < 0 > 0 < 0 < 0 > 0 A poin B he deflaionary impac on srucural inflaion accompanying he negaive produciviy growh differenial is dominaed by he increased produciviy growh pass-hrough When he economy moves from poin B owards poin A, srucural inflaion increases As he reallocaion process moves he economy owards poin A, he ( 2 ) curve flaens ou Once he secor size elasiciy approaches he prevailing disribuion of value added, he produciviy growh differenial becomes incapable of inducing furher volume responses Addiional reallocaions can now no longer ranslae ino produciviy gains he reallocaion process is characerised by a negaive produciviy growh differenial, which simulaes he nonradable secor Conrary o he falling produciviy growh differenial, he rising produciviy growh passhrough simulaes inflaion A poin A he produciviy growh differenial approaches zero, and he produciviy growh pass-hrough dominaes he impac on srucural inflaion he increased impac of he produciviy growh pass-hrough builds up cumulaive srucural inflaion When srucural inflaion increases, and (evenually) reaches a criical value K, i challenges he exising opimal secor srucure Now * * resrucuring changes he opimal secor sizes, ie here is a shif from ( 2 ) o ( 1 ), and he economy jumps from poin A o poin C, where a new process of reallocaion sars over he condiion for shifing beween curves, illusraing resrucuring, can, where ( q C ) * cumulaive average produciviy growh differenial, and ( ) as follows: ˆ denoes a fixed bu arbirary ( ) represens he *, be denoed 11) * dp SR ˆ 1 iff < K C dq * * * = where ( ˆ * dp < ˆ 1) and ( K 0) SR ˆ K iff C dq When he cumulaive average produciviy growh differenial makes srucural inflaion exceed a given hreshold (K), he opimal disribuion of value added changes Resrucuring comes abou when he 12

13 benefis from reallocaion have been exhaused he condiion for resrucuring and ransiion can alernaively, when applying equaion 10a), be expressed as: dp SR 12) K C d ha is when El K 1 ( K 0) ha is, when he secor size elasiciy exceeds a hreshold defined by he exising disribuion of value added, adjused for he K-componen, resrucuring comes abou he analysis poins o wo inerlinked phases of ransiion, boh relaed o produciviy growh developmens in he radable and in he non-radable secor his lends iself o he inerpreaion ha reallocaion is associaed wih medium-erm processes, resrucuring describes longer-erm developmens When he reallocaion gains have been exhaused he energies from he produciviy growh differenial emerge as an acceleraor of srucural inflaion When srucural inflaion reaches an inensiy level K, he reallocaion process evenually simulaes he resrucuring dynamics 5 Srucural inflaion drivers Given he discussion in secion 2 we allow for an inerpreaion of K in order o discuss how srucural inflaion drivers migh vary over ime and beween counries a differen sages of ransiion Le us sar ou by endogenising K, and assume ha i can be represened as a funcion (g) of he long run B-S effec, ha is he cumulaive produciviy growh differenial 13) K = g j j= i= ( 1 ) ( ) N i he condiions for long-erm ransiion is hen relaed o he difference beween he secor size elasiciy, he exising disribuion of value added beween secors, and K, measuring he underlying rend in he produciviy growh differenial For considering he exreme cases regarding he relaionship beween he produciviy growh differenial and ransiion le us assume ha he g-funcion akes he following form: 1 4) ( 1 ) ( ) (1 ( q q ) K = j N ) i j= i= j= j N j ha is he full marix minus he principal diagonal, or saed differenly he cumulaive srucural inflaion less he year-on-year derived srucural inflaion When excluding K<0, hree possible cases emerge: Case 1: Firs, in he absence of lead- and lag effecs beween srucural inflaion and he produciviy growh differenial K=0, and he condiion for ransiion will equal condiion (8a) In his case our long run B-S measure of srucural inflaion and ha of he convenional B-S analysis (shor-erm B-S) coincide Case 2: If on he oher hand K>0, he condiions for long-erm ransiion are harder o saisfy ha is, no under all circumsances is srucural inflaion associaed wih long-erm ransiion If we consider an esablished ransiion economy, one where ransiion from a planned o a marke led economy is well under way, and he produciviy growh differenial for a long period of ime has been favourable for ransiion, hen he lag effec beween pas produciviy growh differenials and srucural inflaion is srong If expecaions of coninued fuure ransiion suddenly vanish he K funcion would only consis of he erms o 13

14 he lef of he principal diagonal Srucural inflaion is now compleely driven by he lag-effec, and he K- funcion equals 15) 1 ( 1 ) ( ) (1 ) ( q q K = j N j N ) i j= + 1 i= j= j Case 3: If K>0 and we consider an economy in is iniial phases of ransiion where he governmen announces a policy shif, emphasising ransiion and boosing expecaions abou fuure produciviy growh developmens, K would only consis of he erms o he righ of he principal diagonal, ha is 16) K ( 1 ) ( ) ( 1 ) ( q q ) 1 = j N i j= i= + 1 j= j N j All in all, he condiion for when srucural inflaion will be high enough o generae ransiion varies boh beween counries and over periods he hree cases oulined above are exremes, in he sense ha hey display pure lead- or pure lag effecs beween he produciviy growh differenial and srucural inflaion, or complee absence of such effecs In realiy hese exreme cases are unlikely o be observed, as a mix of boh lead and lags can be expeced Applying economeric ools o counry daa could assign weighs o he cells in he marix and allow us o classify counries according o heir sage on he ransiion rajecory Anyhow, conex specific analyses are necessary o predic which rae of srucural inflaion ha will induce long-erm ransiion 6 Summary and discussion his paper quesions he convenional framework for analyzing srucural inflaion and he B-S effec, and suggess an alernaive based on an exension o he Scandinavian Model of Inflaion he heoreical foundaions of he B-S effec are addressed, leaving empirical applicaions for fuure research In comparison o convenional B-S analysis, he Scandinavian Model of Inflaion is more decidedly anchored in he supply side, and wih a more firm long-erm focus We exend his framework for srucural inflaion by endogenising he relaionship beween secor sizes and he produciviy growh differenial, he wo main componens of he B-S model All in all, his allows us o address hree main shorcomings of convenional B-S analysis: he demand side channel, he shor-erm focus, and he lack of dependence beween he wo main facors impacing on srucural inflaion While ypical B-S analysis compues secoral shares based on CPI (a demand side saisic), a supply side emphasis will ypically make GDP a preferred choice In ransiion economies he CPI based measures have been shown o weaken he B-S effec Focusing on he demand side and downplaying he supply side, seems a quesionable sraegy for analysing srucural inflaion Raher, he par of he economy where srucural facors reside should be he key objec of analysis A supply side emphasis immediaely leads he way for a long-run perspecive, as srucural shifs residing on he supply side are long-erm processes his perspecive implies ha no only conemporaneous values of he produciviy growh differenial should be aken ino accoun when analyzing srucural inflaion We inroduce he concep of cumulaive produciviy growh differenial as a long-run driver of srucural inflaion and he B-S effec In our erminology, couning he enire daa-marix, insead of jus is principal diagonal, or one of is rows, comes abou as a way of accouning for how produciviy growh differences beween secors impac on inflaion his allows raher arbirary values of he year-on-year- 14

15 produciviy growh differenial- complicaing convenional B-S esimaes - o be replaced by cumulaive differenials, beer suied for analyzing long-run supply side driven srucural shifs he long-run focus also provides addiional insighs on he B-S effec from he endogenous relaionship beween secor sizes and he produciviy growh differenial he lack of dependence beween he wo is shown o hamper shor-erm B-S analysis In our long-run B-S approach a reduced produciviy growh differenial can be compensaed by an increased produciviy growh pass-hrough, or vice versa Hence, he long-run B-S approach shows ha he link beween produciviy growh and srucural inflaion is more complex han wha he shor-erm approach leads us o believe Focusing on he cumulaive produciviy growh differenial implies ha we allow srucural inflaion and he B-S effec o be influenced by boh leads- and lags in he produciviy growh differenial While he lagging effecs can be argued srong in maure ransiion economies, he leads migh be dominan among newcomers he relaive srengh of he differen inflaion drivers can be used o classify counries along he ransiion rajecory Augmening he shor-erm B-S approach even allows us o separae beween differen ypes of srucural shifs Boh reallocaion and resrucuring are long-run srucural shifs driven by he produciviy growh differenial beween secors he relaion beween he wo illusraes he complexiy of he ransiion process Resrucuring comes abou as he reallocaion process looses seam, and srucural inflaion escalaes In urn, he shifs brough abou by resrucuring creae new opporuniies for reallocaion o resume References Balassa, B 1964he Purchasing Power Pariy Docrine: A Reappraisal - JPE, 72, 6, Bergsrand, JH 1991 Srucural Deerminans of Real Exchange Raes and Naional Price Levels: Some Empirical Findings - he American Economic Review, 81, 1, Blanchard, O 1997 he Economics of Pos-Communis ransiion Clarendon Lecures in Economics Oxford Universiy Press Borgersen, A and R M King 2009 Inflaion in Lavia: How real is i? - Arbeidsrappor 2009:2 Øsfold Universiy College, Halden, Norway, hp://wwwhiofno/indexphp?id=4035&lang=nor Breuss, F 2003 Balassa-Samuelson Effecs in he CEEC: Are hey Obsacles for Joining he EMU? - Insiue for European Affairs (IEF) working paper No 53Wirschafsuniversiä Wien May 2003 hp://epubwu-wienaca Cipriani, M 2000 he Balassa-Samuelson Effec in ransiion Economies - Mimeo hp://homegwuedu/~mciprian/cipriani-balassaransiionpdf Coricelli, F and B Jazbec 2001 Real exchange rae dynamics in ransiion economies - Srucural Change and Economic Dynamics, 15, De Gregorio, J, A Giovanni and HC Wolf 1994 Inernaional evidence on radables and nonradables inflaion - European Economic Review, 38, 6, Éger, B 2005 Balassa-Samuelson mees Souh-Easern Europe, he CIS, and urkey: a close encouner of he hird kind? - European Journal of Comparaive Economics, 2,2, Éger, B 2003 Assessing equilibrium exchange raes in CEE acceding counries: can we have DEER wih BEER wihou FEER? A criical survey of he lieraure - Focus on ransiion, 2,

16 Éger, B 2002a Esimaing he Impac of he Balassa-Samuelson Effec on Inflaion and he Real Exchange Rae During he ransiion - Economic Sysems, 26,1, 1-26 Éger, B 2002b Invesigaing he Balassa-Samuelson Hypohesis in he ransiion: Do We Undersand Wha We See? A Panel Sudy - Economics of ransiion, 10,2, Éger, B and J Podpiera 2008 Srucural Inflaion and Real Exchange Rae Appreciaion in Visegrad-4 Counries: Balassa-Samuelson or Somehing Else? - Policy Insigh, 20, Cenre for Economic Policy Research (CEPR) Éger, B, L Halpern and R MacDonald 2006 Equilibrium exchange raes in ransiion economies: aking sock of he issues - Journal of Economic Surveys, 20,2, Éger, B, I Drine, K Lommazsch and C Raul 2003 he Balassa-Samuelson Effec in Cenral and Easern Europe: Myh or Realiy? - Journal of Comparaive Economics, 31,3, Flek, V, L Marková and J Podpiera 2003 Secoral produciviy and real exchange rae appreciaion: much ado abou nohing?, - Czech Journal of Economics and Finance, 3-4 Kovács, M A ed 2002 On he esimaed size of he Balassa-Samuelson effec in five cenral and easern European counries - Working paper 5/2002 Naional Bank of Hungary Lindbeck, A 1979 Impored and Srucural Inflaion and Aggregae demand: he Scandinavian Model Reconsruced In: Assar Lindbeck, ed Inflaion and Employmen in Open Economies Oxford Universiy Press Mihaljek, D and M Klau 2008 Caching-up and inflaion in ransiion economies: he Balassa-Samuelson effec revisied - BIS Working Paper No he Balassa-Samuelson effec in cenral Europe: a disaggregaed analysis - Comparaive Economic Sudies, 46, Noyer, C 2001 Challenges Ahead: he Accession Process - European Cenral Bank, Foreign and Commonwealh Office, London, November 12, 2001, hp://wwwbisorg/review/r011120bpdf Samuelson, P A 1964 heoreical Noes on rade Problems - he Review of Economics and Saisics, XLVI, 2, Obsfeld, M and K Rogoff 1996 Foundaions of Inernaional Macroeconomics he MI Press Rogoff, K 1996 he Purchasing Power Pariy Puzzle - Journal of Economic Lieraure, XXXIV, 2,