Price level convergence within the euro area: How Europe caught up with the US and lost terrain again

Transcription

1 No. 497 / January 2016 Price level convergence wihin he euro area: How Europe caugh up wih he US and los errain again Marco Hoeberichs and Ad Sokman

2 Price level convergence wihin he euro area: How Europe caugh up wih he US and los errain again Marco Hoeberichs and Ad Sokman * * Views expressed are hose of he auhors and do no necessarily reflec official posiions of De Nederlandsche Bank. Working Paper No. 497 January 2016 De Nederlandsche Bank NV P.O. Box AB AMSTERDAM The Neherlands

3 Price level convergence wihin he euro area: how Europe caugh up wih he US and los errain again * Marco Hoeberichs a and Ad Sokman a a De Nederlandsche Bank, Economics & Research Division, PO Box 98, 1000 AB Amserdam, he Neherlands, m.m.hoeberichs@dnb.nl and a.c.j.sokman@dnb.nl 18 January 2016 Absrac Persisen price differences across euro area counries are an indicaion of incomplee economic inegraion. We analyze long and shor run developmens of price level dispersion in he euro area and compare he resuls wih he siuaion in he US. We find ha moneary and economic inegraion in Europe has been successful in esablishing a major downward rend in price level differences across counries since In 2007, price level dispersion in he euro area was a he same level as in he US. Afer he financial crisis, dispersion firs coninued is downward rend before diverging economic condiions across euro area counries conribued o a widening of price level differences again. Shor-run dynamics show ha price dispersion in Europe deviaes more from he long-erm equilibrium han in he US, alhough deviaions have become smaller since EMU. Keywords: economic inegraion, price-level convergence, law of one price, EMU, US. JEL classificaions: E31, E37, F15. * Acknowledgemens: The paper has benefied from useful commens from Peer van Els, Jakob de Haan and Johan Verbruggen. The views expressed in his paper are hose of he auhors and do no necessarily reflec hose of De Nederlandsche Bank. Any remaining errors are he auhors' own responsibiliy.

4 1. Inroducion Persisen price differences across counries are ofen seen as an indicaion of incomplee economic inegraion of he counries. Engel and Rogers (1996) show ha boh borders and sicky prices cum volaile exchange raes are key elemens in explaining price differences across counries. Their approximaion of he US-Canadian border effec by a disance equivalen of 75,000 miles has challenged many o invesigae he sources of inernaional price dispersion. Gorodnichenko and Tesar (2009), e.g., demonsrae ha commonly used dummies o capure border effecs canno separae border fricions from economic facors ha give rise o differen disribuions of price volailiy wihin counries. Ohers invesigaed price differenials across regions wihin a single counry and hereby circumven he problems associaed wih measuring counry border effecs. See for a recen example Crucini, Shinani and Takayuki (2010). Exchange-rae volailiy and shipping coss are examples of ime varying segmenaion effecs ha are found o be imporan deerminans of price dispersion (see e.g. Parsley and Wei, 2001). Nex o exchange rae variabiliy, Faber and Sokman (2009) consider long erm rade developmens among European counries as a measure of diminishing border effecs. In his paper, we analyse long erm developmens in price level dispersion wihin Europe and wihin he US from 1960 up o and including This enables us o idenify he imevarying componen of border effecs ha are hard o disinguish for shor episodes. In he pas 50 years, price level dispersion wihin he euro area more han halved. Price level convergence was paricularly srong in he 1960s, early 1970s and early 1990s. There have been wo episodes in which price level convergence sagnaed. Tha was from mid 1970s up o mid 1980s and in recen years from 2010 onwards. Wihin he US, conrary o Europe, price level dispersion remained by and large sable unil 1983, well below he European level. Halfway he 1980s, he US price dispersion rae sared o rise. In he year 2007, he gap beween he euro area and US price level dispersion was closed. In his paper we invesigae how Europe 1

5 succeeded in closing he gap and why price dispersion sared rising again afer We also invesigae he dynamics around he long-erm equilibrium and find remarkable differences beween he euro area and he US. The paper is organized as follows. In Secion 2 we presen our model. Secion 3 discusses our daase. Secion 4 shows he resuls of he esimaion and some robusness checks and Secion 5 concludes. 2. Model We define price level dispersion across a se of counries (or across regions wihin he US). is he absolue price level of a baske of producs in counry j (or region j) a ime. Price level dispersion a ime is measured by he cross-counry (cross-region) sandard deviaion of ln ( P j ): P j n n 1 1 ( p ) (ln ( P ) ) [ln ( P ) ln ( P )] j j i n j 1 n i 1 2 (1) Nex, we derive he relaionship beween price level dispersion and is main deerminans. Following Crucini e al. (2005), producion in a counry is described by a Cobb-Douglas echnology wih a raded and a non-raded inpu facor. The disincion beween raded and nonraded inpus is crucial for analyzing price differences beween counries, as radable goods open he possibiliy of arbirage. In all counries under consideraion, he bundle of goods is produced by he same Cobb Douglas echnology wih consan reurns o scale. Furhermore, we assume ha here is perfec compeiion. This gives P W Q (2) 1 j j j 2

6 is he share of non-raded inpus required for producing one uni of he baske of goods. is he price of he non-raded inpu and W j Q j he price of he raded inpu in counry j a ime. From Eq. 2 we derive an approximaion for he long-run relaionship beween he level of price dispersion and is deerminans, firs by aking he naural log on he righ and lef hand side of Eq. 2 (p=ln(p), w=ln(w)) and q=ln(q)). p w (1 ) q (3) j j j A each poin in ime, we calculae he variance of he price level across counries (or regions): 2 2 ( j ) [ ( j )] ( j ) (1 ) ( j ) 2 (1 ) ( j) ( j) ( j, j) 1 Var p p w q w q Cor w q (4) For sufficienly high correlaion beween w and q he second erm may be ignored (see Faber and Sokman, 2009). This gives he following expression for ( pj ), shorly p : p w (1 ) q (5) The dispersion of price levels is a weighed average of he dispersion of non-raded inpu coss w and he dispersion of raded inpu coss q. The dispersion of raded inpu coss is expeced o be higher if arbirage coss are higher (see e.g. Rogoff, 1996). Arbirage coss depend on exchange rae volailiy ( vol ) as a measure of cross border uncerainy semming from he presence of naional currencies. Arbirage coss also depend on ransporaion coss. The crude oil price p oil is aken as a proxy of he ime dependen par of disance. Finally, he share of goods rade among he counries (regions) under consideraion relaive o GDP ( open ) summarizes he developmen over ime of all oher rade coss like (non-) ariffs barriers and informaion coss. q f ( open ( ), vol ( ), p ( )) open vol ln p (6) oil, oil, 3

7 Subsiuing Eq. 6 ino Eq. 5 gives he following relaionship for price level dispersion p w (1 )[ open vol ln p ] (7) oil, Eq. 7 provides us wih an analyical ool o idenify he long-erm facors driving price level dispersion. As he variables in Eq. 7 have a uni roo we will perform an inegraioncoinegraion analysis. Eq. 7 is also our saring poin o invesigae possible regulariies in price dispersion dynamics. p p w open vol ln p ecm (8) oil, 1 1 ecm is he residual erm in he esimaed Eq Daa Saring in 1995, EUROSTAT publishes absolue HICP aggregae price levels for EU member saes relaive o he absolue price level for he EU as a whole. The choice of he denominaor does no affec he size of our price dispersion measure. HICP indices are available back o To consruc absolue levels of HICP for years preceding 1995, we apply a similar mehodology as used by Chen and Devereux (2003) for US ciy CPIs. Firsly, all counry HICP indices are convered ino a common currency using yearly averages of marke exchange raes. Nex, he HICP indices up o and including 1995 are 1 Source: OECD Economic Oulook (Number 75, June 2004). 4

8 convered ino absolue price levels hrough synchronizaion wih he counry HICP absolue price level for he year Aggregae price levels for 20 US ciies from 1960 onwards are consruced similarly. 3 By doing so, we obain a consisen daase covering he period. Chen and Devereux (2003) and Faber and Sokman (2009) show ha his approximaion of he underlying absolue values of HICP is reliable. The se of parly consruced and parly colleced absolue price levels allows us o calculae price level dispersion for he euro area and for he US (see Fig. 1A below). Following he model specificaion in Eq. 7 and Eq. 8, addiional daa are required for non-raded inpu coss, openness, exchange rae volailiy and ransporaion coss. The level of income dispersion wihin he euro area, which we use as a proxy for dispersion in he cos of he non-raded inpu in Eq. 7, is calculaed on he basis of he per capia gross domesic produc a facor coss in each of he 11 euro area counries convered o common unis using PPP measures. 4 For he US he consrucion of income dispersion is based on annual daa of GDP per head in US saes. Figure 1B depics he developmen of GDP per head dispersion in he euro area and he US. For openness wihin he euro area we ake he average share in GDP of goods expors from all 11 euro area counries o oher EU counries (members 2003). For he US no ime series on rade among US saes or regions are available. From he Commodiy Flow Survey, which offers he mos comprehensive naionwide source of freigh daa, he value of goods raded beween US regions was calculaed for 1977, 1993, 1997, 2002 and See Fig 1C. 2 Source: Eurosa Chronos. 3 Source: Bureau of Labor Saisics and Cecchei, Mark, and Sonora (2002) for ciy CPIs. 4 Source: OECD Economic Oulook and addiional daa from he World Developmen Indicaors daabase. (Europe) / Bureau of Economic Analysis (US). A correcion is made for he German reunificaion. 5 Source: Bureau of Transporaion Saisics, Commodiy Flow Surveys, US Dep. of Transporaion. 5

9 Fig. 1 The daa A Price level dispersion wihin Europe and he US 1D Nominal exchange rae volailiy euro area US euro area wihin US.35 1B Income dispersion wihin Europe and he US 1E Price crude oil per barrel (deflaed by CPI 2010=1) euro area US euro area US.36 1C Openness inernal marke Europe and he US euro area US 6

10 Long-erm European exchange rae volailiy vol is measured by he sandard deviaion of all monhly changes in he exchange rae of a counry agains he German DM in a given year, averaged over all counries in he group and over eigh years. 6 We use he nominal bilaeral exchange rae wih Germany because Germany was de faco he anchor counry in he European exchange rae sysem. Since we analyze unweighed price dispersion across counries, we do no use an effecive exchange rae based on rade weighs. Eigh years is abou he average lengh of he European business cycle during he pas 50 years. See Fig. 1D. Finally, ransporaion coss are represened by he crude oil price per barrel deflaed by he CPI (in $ for he US, DM/euro for Europe). See Fig 1E. 4. Empirical findings In his secion, we presen our empirical resuls. As price level dispersion in he euro area and he US are non-saionary variables, we apply a co-inegraion analysis. This pus us in a posiion o disenangle he long-erm equilibrium behaviour from he shor-erm dynamics. The long-erm equilibrium is specified in Eq. 7 and he dynamic par in Eq. 8. Long run Firs, we ake a closer look a he oucomes for he long-run relaionship of price level dispersion. In he euro area, price level dispersion, income dispersion, exchange rae volailiy, he naural logarihm of he oil price and openness all have a uni roo of order one for he full sample (Table 1). Price level dispersion wihin he US is non-saionary; income dispersion wihin he US is saionary (Table 2). Time series daa for rade wihin he US are lacking. The 6 Source: IMF IFS and Reinhar and Rogoff (2004). 7

11 Table 1 Adf uni roo es saisics euro area Level p (0.14) (0.00) w (0.17) (0.00) open (0.51) (0.00) l vol (0.52) (0.03) ln p dm (0.53) ( 0.00) oil, Noe p-values beween parenheses Table 2 Adf uni roo es saisics US Level Level p (0.53) (0.00) (0.12) (0.00) w (0.03) (0.00) (0.05) (0.02) ln p (0.66) (0.00) (0.39) (0.00) oil, Noe p-values beween parenheses available years reveal no clear upward or downward rend. We rea his series in he following as a saionary variable. The log oil price in US dollars is a non-saionary variable. Concerning price level dispersion in he euro area, we adop he Johansen co-inegraion procedure o deermine he underlying long-erm relaionship. The model is esimaed for he full sample The co-inegraion rank ess for he euro area indicae he presence of one co-inegraing relaion a he 1% significance level (Table 3). The oil price is no significan and is herefore lef ou in he long-erm equilibrium equaion (Table 3, column 1 and 2). This does no mean ha ransporaion coss are no relevan, bu ha hese likely affec price dispersion hrough oher channels like openness. The remaining hree esimaed parameers are all significan and have he expeced sign: openness lowers price-level dispersion, exchange rae volailiy increases price-level dispersion and income dispersion is posiively relaed o price-level dispersion. Oil prices are significan in explaining he long-erm equilibrium for 8

12 Table 3 Esimaes euro area long-erm equilibrium parameers coefficien of (1) (2) (3) consan w 0.35 (0.07) 0.01 (0.17) 0.38 (0.02) (0.07) 0.37 (0.19) open (0.16) (0.07) (0.17) l vol (0.006) (0.007) (0.006) ln p dm oil, (0.006) (0.007) - p Johansen es -race es -eigenvalue es 1 co-in relaion * 1 co-in relaion * 1 co-in relaion * 1 co-in relaion * 1 co-in relaion * 1 co-in relaion * Inercep (no rend) in CE Sandard errors beween parenheses * = 1% significance level price dispersion beween ciies in he US (Table 4). The augmened Dickey-Fuller es, however, indicaes ha co-inegraion mus be rejeced for he full sample. One explanaion for his is ha he energy inensiy of he US economy has come down subsanially, especially in he period from he firs oil crisis in he early 1970s o he second oil crisis in he early 1980s. Saring from 1984 onwards, he presence of a co-inegraing relaionship beween he US price level dispersion and he oil price canno be rejeced. As expeced, higher oil prices are associaed wih more price dispersion. The coefficien for income dispersion is no significan. Bergin and Glick (2007), for example, also find a srong relaionship beween inernaional price dispersion and oil price movemens. Their period of invesigaion runs from 1990 o Table 4 Esimaes US long-erm equilibrium parameers p coefficien of (1) (2) (3) (4) consan 0.10 (0.01) 0.10 (0.04) 0.12 (0.02) 0.11 (0.003) w (0.12) (0.20) - ln p (0.004) (0.003) (0.004) (0.004) oil, Aug DF es sa * * * * Noe: Sandard errors beween parenheses *Asympoical criical -value a 1% level is (Davidson, Mc Kinnon, 1993) 9

13 Shor run The dynamic par is esimaed by means of an error-correcion specificaion. The errorcorrecion erm for he euro area is he residual from he long-run equaion presened in Table 3 (parameer esimaes in column 3), and for he US he residual from he equaion in Table 4 (column 4). Firs we ake a closer look a he esimaion resuls for Europe (Table 5). Column 1 shows he specificaion wih he lagged changes in price level dispersion and changes in he dispersion of income, openness, exchange rae volailiy and he oil price. The esimaed errorcorrecion parameer is highly significan. Wih a value of 0.33, he half life of reurning o he underlying equilibrium level of price dispersion in Europe is abou wo years. Furhermore, a significan par of lagged changes in price dispersion are passed hrough o he curren year. In oher words, price dispersion movemens end o persis in he shor run. From he facors ha may affec shor-erm dynamics, only he lagged change in exchange rae volailiy is significan. Changes in openness do no significanly affec price dispersion movemens in he Table 5 Esimaes euro area error-correcion equaion p excl (1) (2) (3) ecm (0.087) (0.089) (0.096) 1 consan (0.002) (0.002) (0.002) p (0.125) (0.131) (0.139) w (0.193) (0.202) (0.255) open (0.204) (0.210) (0.265) 1 l vol (0.017) (0.018) (0.018) p (-0.007) (0.008) (0.008) ln dm oil, 1 _ R SR plm(1) * plm(2) * plm(4) * pnormaliy * pheeroskedasiciy * Noe: sandard errors beween parenheses * p-values

14 shor run, bu are imporan drivers of price dispersion in he longer run. We also esimae he model for he full sample excluding he years 2007, 2008 and 2009, hereby eliminaing he poenially disorionary influence of he financial crisis. We obain very similar resuls (see column 2). Neiher does cuing off he esimaion period in 2007 (column 3) aler hese findings. The esimaed parameers for he shor-run equaion for price dispersion across US ciies are shown in Table 6. The sample period for he US is The error-correcion parameer is highly significan (column 1). Wih a half life of 5 o 6 years, a reurn o he longerm equilibrium akes subsanially longer in he US han in he euro area. Wha may explain his? Firs of all, measured in squared kilomeres, he US area is four imes bigger han he euro area. Larger disances are accompanied wih rade coss. These rade coss are an imporan reason for he presence of price differences beween locaions. Secondly, since he esablishmen of he European Economic Union, price dispersion developmens in he euro area have o a subsanial degree been deermined by a process of moneary and economic Table 6 Esimaes US error-correcion equaion p excl (1) (2) (3) ecm (0.027) (0.029) (0.032) 1 consan (0.0004) (0.0004) (0.0005) (0.140) (0.152) (0.163) (0.066) (0.070) (0.074) p 1 w 1 ln p oil, (0.001) (0.002) (0.002) _ R SR plm(1) * plm(2) * plm(4) * pnormaliy * pheeroskedasiciy * Noe: sandard errors beween parenheses * p-values 11

15 inegraion. Borders have become much less significan in Europe. This conribued o shrinking price level differences beween counries. Again as wih he euro area, leaving ou he financial crisis years or cuing of he sample in he year does no affecs he findings for he US (see Table 6, column 2 and 3). Figure 2 shows he combined long and shor erm performance of he model by means of a dynamic simulaion. In he dynamic simulaion, we sar wih realisaions for he dependen variable in he firs year and calibrae he level of price dispersion in year 2. In year 3, price level dispersion is deermined by he simulaed level of price level dispersion in year 2 and he values of he exogenous variables in ha year, and so on. We ake he equaions for he period up o 2007 as a saring poin. The dynamic simulaion up o and including 2007 is an in-sample exercise. From 2008 up o 2014, he oucomes are ou-of-sample esimaes. Overall he descripion given by he model simulaion is saisfacory. In he euro area, he larges deparure from realised price level dispersion occurs in he second half of he 1980s and early 1990s. Rogers (2007) noed ha he unprecedened drop in price level dispersion during his period coincided wih he compleion of he European Single Marke in Fig. 2 Dynamic simulaion price level dispersion euro area measured euro area dynamic simulaion ( ou-of-sample) US measured US dynamic simulaion ( ou-of-sample) 12

16 Price dispersion across US regions displays a genly upward sloping rend, wih a slighly lower dispersion in he firs half of he 1980s. This is racked by he dynamic simulaion wih he overall model for he US. 5. Conclusion We have analysed he long-erm rend and shor-erm dynamics of price level dispersion in Europe and in he US. Our resuls for he long-run rend show ha he price levels across he euro area counries converged beween 1960 and The long-run level of price dispersion came down o he level of he US in 2007, fell furher unil 2010 and sared rising again in he years afer Imporan drivers behind he converging rend are lower exchange rae volailiy, smaller income dispersion and more openness of European economies. These drivers can a leas parly be aribued o he success of economic and moneary inegraion in Europe. The more recen rising rend in he years is mosly due o rising income dispersion as a resul of he European deb crisis. The long-run developmen of price dispersion across US regions is much more sable and unil 2007 a a lower level han in Europe. The oil price is he only imporan driver behind he long-run rend in price dispersion in he US. The dynamic simulaions reveal some imporan differences beween he euro area and he US. Deviaions from he simulaed pah have become smaller since EMU, bu are sill considerably larger han in he US. Alhough moneary and economic inegraion in Europe has been successful in esablishing a downward rend in differences beween price levels across counries, large deviaions are an indicaion ha idiosyncraic shocks in Europe are ransmied ino price dispersion, more so han in he US. This may be due o more limied risk sharing arrangemens across European counries. Consisen wih his possible explanaion of incomplee European economic inegraion is a final observaion abou price dispersion during he years afer he financial crisis. In Europe, 13

17 he dynamic simulaion shows an upward rend since 2010 (see Fig. 2). This developmen is driven by larger income dispersion across European counries (see Fig. 1b), which was a resul of he recession and he deb crisis following he financial crisis. In he US, however, income dispersion did no increase afer he crisis. 14

18 References Bergin, P.R. and R. Glick, (2007), Global price dispersion: Are prices converging or diverging?, Journal of Inernaional Money and Finance, 26, Cecchei, S.G., Nelson, C. M. and R.J. Sonora, (2002), Price index convergence among Unied Saes ciies, Inernaional Economic Review, 43, Chen, L and J. Devereux, (2003), Wha can US ciy price daa ell us abou purchasing power pariy?, Journal of Inernaional Money and Finance, 22, Crucini, M.J., Telmer, C.I. and M. Zachariadir, (2005), Undersanding European real exchange raes, American Economic Review, 95(3), Crucini, M.J., Shinani M. and T. Takayuki, (2010), The law of one price wihou he border: he role of disance versus sicky prices, The Economic Journal, 120, Engel, C. and J.H. Rogers, (1996), How wide is he border?, American Economic Review, 86(5), Engel, C., and J.H. Rogers, (2001), Deviaions from purchasing power pariy: causes and welfare coss, Journal of Inernaional Economics, 55(1), Faber, R.P., and A.C.J. Sokman, (2009), A shor hisory of price level convergence in Europe, Journal of Money Credi and Banking, 41, 2-3, Gorodnichenko, Y. and L.L. Tesar, (2009), Border effec or counry effec? Seale may no be so far from Vancouver afer all, American Economic Journal; Macroeconomics, 1, OECD Economic Oulook (Number 75, June 2004). Parsley, C. and S. Wei, (2001), Explaining he border effec: he role of exchange rae variabiliy, shipping coss, and geography, Journal of Inernaional Economics, 55, Reinhar, C.M. and K.S. Rogoff, (2004), The modern hisory of exchange rae arrangemens: a reinerpreaion, Quarerly Journal of Economics, 119, Rogers, J.H., (2007), Moneary union, price level convergence, and inflaion: How close is Europe o he USA?, Journal of Moneary Economics, 54, 3, Rogoff, K.S., (1996), The purchasing power pariy puzzle, Journal of Economic Lieraure, 34,

19 Previous DNB Working Papers in 2016 No. 493 Jacob Bikker, Dirk Gerrisen and Seffie Schwillens, Compeing for savings: how imporan is crediworhiness during he crisis? No. 494 Jon Danielsson and Chen Zhou, Why risk is so hard o measure No. 495 Gabriele Galai, Irma Hindrayano, Siem Jan Koopman and Marene Vlekke, Measuring financial cycles wih a model-based filer: Empirical evidence for he Unied Saes and he euro area No. 496 Dimiris Chriselis, Dimiris Georgarakos, Tullio Jappelli and Maaren van Rooij, Consumpion uncerainy and precauionary saving

20 De Nederlandsche Bank N.V. Posbus 98, 1000 AB Amserdam dnb.nl