Using Expenditure PPPs for Sectoral Output and Productivity Comparisons

Transcription

1 Internatonal Comparson Program Chapter 24 Usng Expendture PPPs for Setoral Output and Produtvty Comparsons Robert Inklaar and Marel P. Tmmer Measurng the Sze of the World Eonomy ICP Book

2 Contents Methodology for Produtvty Comparsons... 5 Output and Expendture Pres wthn a Supply-Use Table Framework... 9 PPPs n the GGDC Produtvty evel Database Output and Value Added PPPs abor and Captal PPPs Produtvty evel Comparsons for the OECD Produtvty evels Senstvty of PPPs and Produtvty evels Conluson Annexes Annex A. Senstvty to Choe of PPPs, Annex B: Country Codes Referenes

3 Purhasng power partes (PPPs) have a wde range of analytal and poly applatons. 1 Tradtonally, PPPs have been used for nternatonal omparsons of nome, expendture, and output. Most well known are the omparsons of levels of gross domest produt (GDP) per apta as publshed by the World Bank (e.g., World Bank 2008). Dfferenes n GDP per apta aross ountres are manly determned by levels of labor produtvty at the aggregate eonomy level. Today, omparatve seres of output per worker and per hour worked are beng produed routnely by varous statstal organzatons and n aadema, suh as the Penn World Tables and the seres produed by the Organsaton for Eonom Co-operaton and Development (OECD), Conferene Board, and Gronngen Growth and Development Centre (GGDC) at the Unversty of Gronngen. Varous useful analytal applatons of produtvty levels, however, are also found at the more detaled level of ndvdual ndustres. Beause of the large dfferenes n eonom struture aross ountres, nternatonal omparsons of output and produtvty at the setor level (agrulture, manufaturng, and serves) provde useful omplements to omparsons of GDP or by expendture ategores. Prevous researh has shown that a low overall level of produtvty s not neessarly ndatve of large gaps n all setors. Generally, t s assumed that produtvty gaps n manufaturng an be large, and that gaps n serves produtvty aross ountres are muh smaller. Ths well-known fndng s at the heart of the Harrod-Balassa-Samuelson effet and often nvoked to explan the lower relatve pres of serves n low-nome ountres ompared wth hgh-nome ountres. Also, studes of onvergene and dvergene n the world eonomy are nreasngly made at the ndustry level. Tests of nternatonal trade theores and endogenous growth models requre measures relatve to the world produtvty fronter by setor. By defnton, a major part of the researh n these areas requres PPPs from the produton sde, rather than from the expendture sde that s, PPPs should reflet the dfferenes n the output pres of goods and serves rather than the expendture pres. In the remander of ths hapter, we all these produton PPPs. Comparsons of produtvty, loosely defned as output per unt of nputs, also requre the relatve pres of aptal nputs, labor nputs, and ntermedate nputs. Ths so-alled ndustry of orgn approah was poneered by Page and Bombah (1959) n a omparson of the Unted ngdom and the Unted States. The earler work was then onvenently summarzed by ravs (1976). Over the last two deades, ths method was further developed and used n the Internatonal Comparsons of Output and Produtvty (ICOP) projet 1 The authors would lke to thank Bart van Ark and D. S. Prasada Rao for useful omments and suggestons on an earler draft of ths hapter. The researh for ths hapter was part of the EU EMS projet on Growth and Produtvty n the European Unon. Ths projet was supported by the Researh Dretorate General of the European Commsson, as part of the 6th Framework Programme, Prorty 8, Poly Support and Antpatng Sentf and Tehnologal Needs. 3

4 at the Unversty of Gronngen and at the Natonal Insttute of Eonom and Soal Researh n ondon (O Mahony 1999). Van Ark and Maddson (1994) and Maddson and van Ark (2002) provde an overvew of the progress made n the early phase of ths projet through the researh nput of a dozen sholars and work on about 30 ountres. More reently, setoral output and produtvty trends for a large set of ountres were provded by the EU EMS database (avalable at and dsussed by Tmmer et al. 2010). However, t appears that produton PPPs were sare and emprally dffult to obtan, whh manly related to the lak of readly avalable produer pre surveys. An alternatve to the ndustry of orgn approah s to use data from nternatonally oordnated surveys on expendture pres suh as those olleted for the Internatonal Comparson Program (ICP) under the auspes of the Unted Natons and the World Bank (ravs, Heston, and Summers 1982; Summers and Heston 1991). Sne the early 1980s, the Organsaton for Eonom Co-operaton and Development (OECD) has regularly publshed estmates of expendture PPPs derved from ts jont program wth Eurostat. Expendture-based PPP omparsons are based on purhasers pres of fnal goods and serves wth a detaled produt spefaton. Hene, to apply them to output and produtvty omparsons by ndustry, the PPPs have to be mapped from expendture ategores to ndustry groups. The expendture approah to setoral PPPs was poneered by Jorgenson, uroda, and Nshmzu (1987) and most reently appled by van Besebroek (2009) and Sørensen and Shjernng (2008). In general, PPPs based on expendture pre surveys suffer less from qualty problems than unt values beause produt omparsons are based on detaled spefatons. However, the approah also has some drawbaks for omparsons of output and produtvty at the ndustry level beause t requres detaled adjustments for margns, taxes, and nternatonal trade. Furthermore, by defnton these PPPs only over pres for fnal expendture and do not reflet the relatve pres of ntermedate goods. Followng Plat (1996), van Ark and Tmmer (2009) argue that a mxture of PPPs derved from the expendture and ndustry of orgn approahes should be used for produtvty omparsons at the ndustry level. Buldng on ths dea, Inklaar and Tmmer (2009) onstruted the GGDC Produtvty evel database. Ths database provdes omparsons of output, nputs, and produtvty at a detaled ndustry level for a set of 30 OECD ountres for 1997, and t s publly avalable at The PPPs used for ths database were a ombnaton of produton- and expendture-sde PPPs. Ths study also provded pre omparsons of aptal, labor, and ntermedate nputs, alongsde output pre omparsons so that large-sale omparsons of multfator produtvty at the setor level beame feasble. Ths hapter begns by outlnng what types of PPPs are needed for setoral, sngle-, and multfator produtvty omparsons by ndustry. We argue n the frst seton that, gven the lmted avalablty of produton PPPs, expendture PPPs from the ICP are ndspensable n faltatng these omparsons. However, untl now the lterature has not learly desrbed what adjustments are needed to onvert PPPs by expendture ategory nto output PPPs by ndustry and 4

5 under whh ondtons expendture PPPs would provde a good proxy for produton PPPs. The man ontrbuton of ths hapter s to derve these adjustments and ondtons based on a system of supply-use tables (SUTs). SUTs are a major buldng blok of the natonal aounts n many ountres. They provde a systemat framework of the flows of goods and serves from produton and mports to fnal expendture. The seond seton of ths hapter shows the usefulness of the dervaton and onfrontaton of varous types of pres (bas and purhasers pres) n a SUT system. For example, ths system provdes lear gudane on how the pres of onsumpton goods and ntermedate nputs are related to the output pres of ndustres. These relatonshps an be used to generate ross-heks n ase dfferent data soures (from the produton and expendture sdes) exst or to llustrate the adjustments to be made to the expendture PPPs to be used n setoral produtvty omparsons. Usng a supply-use framework, we show that fnal expendture pres need to be adjusted for trade and transportaton margns, for taxes and subsdes, for the pres of exports and mports, and fnally for the pres of ntermedate use n order to provde a good proxy for output pres. Beause of these adjustments, the usefulness of expendture PPPs n setoral produtvty omparsons dffers aross setors. Ths dfferene s llustrated wth some results from the GGDC Produtvty evel database for OECD ountres n the thrd seton of ths hapter. The fourth seton presents some produtvty omparsons, and the fnal seton some onludng remarks. In those remarks, we stress the mutual dependene and potental spllovers from setoral produtvty omparsons and expendture results from the ICP. Methodology for Produtvty Comparsons In ths seton, we present the methodology for omparng levels of output, nput, and produtvty aross ountres that we used n onstrutng the GGDC Produtvty evel database (Inklaar and Tmmer 2009). Beause we were tryng to onstrut a omparable set of produtvty measures for a large number of ountres and ndustres at the same tme, varous hoes had to be made, not only about the use of partular ndex number formulas, but also about ther atual mplementaton. Ths seton lays out the bas methodology (for a more detaled dsusson of methodology and emprs, see Inklaar and Tmmer 2009). The man am of the GGDC Produtvty evel database ( s to ompare produtvty between ountres. The aounts provde so-alled bnary omparsons that s, omparsons between a ountry and a base ountry that s the same n all omparsons. Beause the greatest nterest les n omparng the performane of ountres to the world produtvty and tehnology leader, t s natural that we 5

6 hoose the Unted States as our base ountry n the produtvty omparsons. 2 The most ommonly used sngle produtvty measure for nternatonal omparsons of levels s labor produtvty. Ths term s generally defned as an output measure dvded by a labor nput measure. The labor nput measure an be the number of persons employed, number of employees, or number of hours worked. The output measure an be ether the volume of gross output or the volume of value added. If Q VA s value added and H s hours worked, then value added based labor produtvty (P_VA) s gven by VA Q / H (24.1) P_ VA. VA Q / H US US Alternatvely, more than one nput an be aounted for the so-alled multfator produtvty (MFP) measures. MFP measures are well rooted n eonom theory, but beause of ther heavy data requrements they are used muh less than sngle produtvty measures suh as labor produtvty. MFP and labor produtvty measures are not ndependent of eah other. Multfator produtvty measures an be used to explan sngle-fator produtvty dfferenes. For example, dfferenes n labor produtvty levels an be explaned by dfferenes n the rato of aptal to labor and dfferenes n multfator produtvty. These and other lnks have been establshed wth the help of the eonom theory of produton. The GGDC Produtvty evel aounts provde estmates for both value added based MFP, takng nto aount both labor and aptal serves, and gross output based MFP, takng nto aount labor, aptal, and ntermedate nputs. In ths hapter, we outlne only the methodology for value added based measures. Followng Jorgenson and Nshmzu (1978), we defne the translog quantty ndex of dfferene n multfator produtvty based on value added (MFP_VA). Ths ndex s defned as VA Q Q Q (24.2) ln MFP _ VA ln ˆ ln ˆ w w ln, VA Q Q Q US US US where Q s the quantty of aptal serves, Q s the quantty of labor serves, and ŵ s the share of aptal serves n value added averaged over the two ountres that s, 1 wˆ w w, where w 2 US V V V, wth V the nomnal value of aptal serves ompensaton n ountry n natonal urreny) and smlarly for labor so that w ˆ wˆ 1. Under the standard neolassal assumptons, ths measure ndates the dfferene n the level of tehnology between the two ountres (see Jorgenson and Nshmzu 1978). 2 Usng the Unted States as our base ountry does not mply the use of a star system of omparsons. As dsussed later, we wll use base-nvarant produtvty measures. 6

7 The formulas (24.1) and (24.2) ndate that omparable volume measures of output and nput for the two ountres are needed. When a sngle output s beng ompared, physal measures suh as numbers of ars are possble. However, when omparsons are made at the ndustry or aggregate level where output s not represented by a sngle produt, output s gven n terms of real values. In that ase, a orreton for dfferenes n relatve pre levels between ountres s needed. Ths s usually done wth a purhasng power party (PPP) that ndates the rato of the pre of output n one ountry relatve to that n another ountry, eah gven n loal urrenes. Volume ndes are alulated mpltly by the rato of the nomnal values and the relevant pre ndes. For example, the aggregate value added quantty n ountry s gven by VA VA V (24.3) Q. VA PPP For labor nput, one an use number of workers or total hours worked as a volume measure. However, for multfator produtvty omparsons one would also lke to nlude the omposton of labor n terms of varous labor types wth dfferent produtvtes for example, low- and hgh-sklled labor. Ths an be done by hoosng an approprate PPP based on relatve wages so that V (24.4) Q, PPP wherev s the nomnal value of labor ompensaton n ountry (.e., n natonal urreny), and PPP s the relatve pre of labor serves n ountry. Smlarly, for aggregate aptal nput n ountry (24.5) ~ V Q PPP, where V ~ s the nomnal value of ex ante aptal ompensaton n ountry, and relatve pre of aptal serves n ountry. 3 PPP s the One of the man applatons n produtvty omparsons s the so-alled level aountng. evel aounts provde a deomposton of dfferenes n value added per hour worked nto dfferenes n aptal per hour worked (aptal ntensty), n labor omposton (skll ntensty), and n MFP. Ths deomposton s arred out as 3 Note that beause of our approah to aptal measurement, aptal ompensaton n ths formula s based on ex ante measures of rates of return and wll dffer from the ex post measure of aptal ompensaton used as the weght n equaton (24.2). 7

8 VA H Q H Q H (24.6) ln wˆ ln wˆ ln ln MFP _ VA, VA H Q H Q H US US US US US US where ŵ and ŵ are defned as n equaton (24.2). The PPPs for outputs and nputs requred n (24.3) (24.5) are derved on the bass of detaled sets of output and nput pres. 4 Pres are aggregated usng the multlateral translog pre ndes (CCD ndex) ntrodued by Caves, Chrstensen, and Dewert (1982). Basally, n ths methodology an artfal ountry s reated by averagng over all ountres n the data set. Ths onstruted ountry s then used as a brdge when makng bnary omparsons between two ountres. Ths method reates so-alled transtve PPPs that are base ountry ndependent (see hapters 1, 4, and 5 n ths volume for further dsusson). As wth our MFP ndes, the PPPs are normalzed wth the Unted States = 1. abor and aptal volume ndes grounded n produton theory should take nto aount the omposton of eah fator nput, suh as dfferent levels of sklls or types of aptal goods, n partular nformaton and ommunaton tehnology (ICT) assets versus non-ict assets. For labor, ths an be aheved by deflaton wth an approprate PPP ( PPP ) based on the relatve wages of eah labor type l as follows: (24.7) PPP wl ln PPPl ln PPPl ln,, l where the bar n the last term ndates a geometr average over all ountres ndexed by runnng from 1 to N, and N s the number of ountres. It follows that ln PPP l 1 N ln PPPl, and w l 2 [ l, l, w l s the average weght of labor type l defned as 1 w ( w / N )] wth w, the share of labor type l n total labor ompensaton n ountry : w l Vl, l, V. The PPP for eah labor type s derved on the bass of relatve wages. A smlar proedure s appled for the dervaton of PPPs for aptal ntermedate nputs X PPP. PPP, output Y PPP and For the deflaton of value added, a double deflaton proedure s used based on separate PPPs for gross output and ntermedate nputs as requred (Jorgenson, uroda, and Nshmzu 1987). We follow a CCD-lke approah by takng a geometr mean of all possble bnary Törnqvst ndes for a partular ountry. Frst, we alulate the bnary value added PPP for eah ountry par (,d) as follows: 4 We aggregate over pres rather than over quanttes beause varaton n pres aross ountres s muh less than varaton n quanttes (also see Allen and Dewert 1981). 8

9 (24.8) ln PPP Z ln PPP 1 1 w Y X Z d Y ln PPP Y Y X ln PPP w ln PPP X ln PPP d X d. The weght ountres. Y wx s the share of ntermedate nputs n gross output, averaged over the two Seond, a Gn-ES proedure s appled to multlateralze the set of value added PPP bnares gven n (24.8), as n Caves, Chrstensen, and Dewert (1982). Together, these equatons provde the system used to derve MFP measures onsstent wth neolassal produton theory. Output and Expendture Pres wthn a Supply-Use Table Framework The theoretally most approprate approah for nternatonal omparsons of output and produtvty levels s to apply PPPs that are based on the ndustry of orgn approah. Varous alternatve ways an be used to obtan PPPs for gross output, partly dependng on the data avalablty for ndvdual ndustres. One way s to make use of produer pres for spefed produts, but these are sare beause large-sale nternatonal surveys of produton pres are not onduted. The most wdely used approah to obtan produton PPPs s the unt value rato (UVR) method. Ths method makes use of produton statsts suh as ensuses or busness statsts surveys that reord the output values and quanttes for produt tems. By dvdng the output value by the orrespondng quanttes, one obtans unt values, whh an then be used for alulatng unt value ratos for mathed tems between ountres. Beause of lak of data, ths approah an be used only for a lmted set of ndustres and ountres. In addton, unt value ratos an suffer from qualty adjustment problems n nternatonal omparsons. Detaled produt haratersts are dffult to observe dretly from produton statsts beause those statsts report quantty and values for produt groups rather than for spefed produts, and produt desrptons are often bref. 5 An alternatve to the ndustry of orgn approah s to use data from nternatonally oordnated surveys on expendture pres suh as n the Internatonal Comparson Program (ICP). Expendture-based PPP omparsons are based on the purhasers pres of fnal goods and serves wth a detaled produt spefaton. Hene, to apply them to output and produtvty omparsons by ndustry, bas headng PPPs need to be mapped from expendture ategores to ndustry groups. However, so far no one has learly outlned under whh ondtons expendture PPPs provde a good proxy for the PPPs of gross output and what knd of adjustments are needed. In ths seton, we use the bas balane equaton between supply and use of goods n the supply- 5 See van Ark and Tmmer (2009) for a reent survey. 9

10 use table framework to derve an exat relatonshp between expendture and ndustry output pres. Ths relatonshp s used to ndate whh adjustments need to be made to expendture pres so they are a reasonable proxy for bas output pres. Alternatvely, the adjustments an be used to assess the usefulness of expendture PPPs n output omparsons. Supply-use tables provde a onvenent way of summarzng and presentng a oherent set of eonom transatons wthn a ountry. Whereas a supply table ndates for eah produt where t omes from (domestally produed or mported), the use table provdes nformaton on produt use as ntermedate or fnal onsumpton. In the SUT framework, the valuaton of domest output s at bas pres, whle the use of produts s reorded at purhasers pres. 6 Ths dstnton s rual n the lnk between expendture and produton PPPs (Commsson of the European Communtes et al. 1993, se. VI.J): Purhasers pre = bas pre of the produt reeved by the produer + taxes on the produt 7 subsdes on the produt + trade and transport margns n delverng the produt to the purhaser. Ths seton provdes an exposton of the full struture of the SUT framework. The followng notaton s used, whh nludes ommodtes ( = 1,..., m) and ndustres j (j = 1,..., n); quanttes are ndated wth aptals: S = quantty of total supply of produt U = quantty of total use of produt M = mported quantty of produt F = quantty of produt for fnal domest demand E = quantty of produt exported Y j = quantty of ommodty produed by ndustry j X j = quantty of ommodty used as ntermedate nput by ndustry j j = value of labor serves used n ndustry j j = value of aptal serves used n ndustry j. owerase p s are used for the orrespondng pres: 6 Exports are valued at free on board (f.o.b.) pres and the mports at ost, nsurane, and freght (..f.) pres. The export f.o.b. pre s essentally a purhaser s pre, nludng net taxes and trade and transport margns up to the border of the exportng ountry. The mport..f. pre s essentally a bas pre, exludng net taxes leved after rossng the border and trade and transport margns wthn the ountry. 7 Taxes nlude any taxes on produts at the sales pont suh as a sales or a value added tax. 10

11 Y p j = bas pre reeved by ndustry j for sellng ommodty X p j = purhaser s pre pad by ndustry j for ntermedate onsumpton of ommodty M p = bas (..f.) pre of mported ommodty F p = purhaser s pre for fnal domest demand of ommodty E p = purhaser s (f.o.b.) pre of exported ommodty. Margns and taxes are defned as follows: T = total taxes net of subsdes on produts R = total trade and transport margns S r = trade and transport margn rate on suppled produt S t = net tax rate on suppled produt. Fnally, a aptal V n front of a quantty symbol s used to ndate ts value through multplaton by the orrespondng pre. 11

12 Table 24.1 Outlne of a Supply-Use Table USE table at purhasers' pres 1 1. j. n : : : : : : Commodtes. P X jx j. Total ntermedate nput at purhasers' pres Gross value added at bas pres Gross output at bas pres m Industres VX P F F P E E VX +VF + VE : : : : : : j. VX VF VE VX + VF + VE. j + j.. VY j.. VX Total ntermedate use Fnal domest Exports demand f.o.b. Total use at purhasers' pres SUPPY table at bas pres Commodtes. P Y jy j. Total at bas pres 1 Total Industres domest Import 1. j. n supply..f. : : : : : : : : VY P M M t VS r VS (1+ +r t )VS : : : : : : : : m. VY Total supply at bas pres VS = VY + VM Taxes mnus subsdes Trade and transport margns Total supply at purhasers' pres j. VY VM VS= VY + VM T R VS + R + T

13 Table 24.1 s a smplfed outlne of a supply-use table. 8 In both the supply and use tables, ommodtes are n the rows and ndustres n the olumns. The use table ndates the usage for eah produt : ntermedate, fnal domest demand, or export. The last olumn ndates total use. The entres are at the purhasers pres. In addton, the use table ontans a so-alled value added blok. For eah ndustry j, total ntermedate nput at purhasers pres plus value added at the bas pre adds up to gross output at bas pres (last row). The supply table ndates the orgn of eah produt, whether domest produton or mport. The fourth olumn reords total supply at bas pres. The other olumns provde nformaton on taxes and subsdes on produts and trade and transport margns. These are needed to arrve at the total supply at purhasers pres, whh an be set aganst total use at purhasers pres from the use table. The output of all produts produed n ndustry j valued at bas pres sums to gross output at bas pres n ths ndustry (last row n the supply table). The supply and use tables are lnked by two bas denttes: row and olumn. The olumn dentty requres dentty for eah ndustry between the sum of gross output over all produts produed n an ndustry, on the one hand, and value added plus ntermedate onsumpton, on the other: VX V V VY (24.9) j j j j The row dentty requres balane between use and supply for eah produt and lnks the expendture and produton approahes at the produt level. The dentty should hold both n terms of quanttes and values. In quantty terms, F E S (24.10) X Ths dentty states that the quantty of supply of produt must be equal to ts use, onsstng of ntermedate use, fnal domest demand, and exports. In value terms, at purhasers pres the dentty s (24.11) VX VF VE VS T R. 8 See the Eurostat manual on supply-use tables for more nformaton (Eurostat 2008). 13

14 The value of total ntermedate use of ( VX ) s equal to the sum of values of ntermedate use of by all produers, and the total value of supply ( VS ) s equal to the value of supply by all produers and mports. By rewrtng values as the produt of pres and quanttes, (24.11) an be stated as (24.12) j p X j (1 t X S j p r S F ) F j p p Y j Y E j E p M M By rewrtng equaton (24.12), the relatonshp between purhasers pres ( the one hand, and bas output pres ( X p j F p E p ), on Y p j ), on the other, an be derved. Ths dentty provdes the bas relatonshp between the fnal domest demand pre and the output pre at the produt level, whh we are seekng. To brng out ths relatonshp more learly, we assume, wthout loss of generalty, that there s only one bas pre n the system for an ndvdual produt that s, the bas output pre of a produt s ndependent from ts ndustry of orgn: (24.13) p Y j p Y By rearrangng equaton (24.12), substtutng (24.13), and usng dentty (24.10), the followng bas result an be derved (omttng ndex for larty). Under the assumpton gven n (24.13), the general relatonshp between bas output pres and fnal domest demand pres an be wrtten as (24.14) p Y 1 S (1 t r S ) p F A E, M A X Ths s a key result for our purpose. It ndates that three types of adjustments are needed to derve an output pre from a fnal domest demand pre: an adjustment for margns and taxes, an adjustment for nternatonal trade (A E,M ), and an adjustment for ntermedate onsumpton (A X ). The latter two are gven by 1 S (1 t r E, M E F S S M F (24.15a) A P P (1 t r ) P P and S ) E Y M Y 14

15 X 1 X F (24.15b) A Pj P S S (1 t r ) j X Y j. The frst term on the rght-hand sde of (24.14) s the fnal expendture pre, 9 adjusted for average net taxes and margns on total supply of the produt. The seond adjustment s for nternatonal trade, gven n (24.15a). Ths s espeally mportant for omparsons nvolvng small, open eonomes. The sze of the adjustment depends on the dfferenes between the fnal expendture pres and the export and mport pres, and on the ratos of export and mport quanttes to total domest output. The thrd adjustment n (24.15b) depends on the sze of the dfferenes between the fnal expendture pre and the ntermedate onsumpton pre for a partular tem, and on the rato of ntermedate onsumpton to total domest output for that tem. Ths bas result suggests that f export, mport, and domest pres dffer, for produts that are haraterzed by larger shares of mports and exports and/or ntermedate onsumpton n total output, expendture pres wll be poorer proxes for output pres. But for ndustres that are manly produng for fnal onsumpton and whose produts are hardly nternatonally traded, an adjusted fnal expendture pre mght be a reasonable proxy for the ndustry output pre. Ths type of nformaton an be derved from nput-output tables. Usng ths result, we an also state the adjustments needed for expendture PPPs to properly reflet output PPPs: (24.16) PPP Y 1 S (1 t r 1 S (1 t r US S S US p ) ) p F F US A A E, M E, M US A X A X US. 9 We use the terms fnal expendture pre and fnal domest demand nterhangeably. 15

16 If the adjustments for nternatonal trade and ntermedate onsumpton are assumed to be zero n both ountres, (24.17) s smplfed to S S ( 1 t r ) F S S (1 t r ) Y US US (24.17) PPP PPP. In ths ase, the expendture PPP, adjusted for average net taxes and margns on total supply of the produt, equals the output PPP. In fat, ths adjusted expendture PPP has been used as a proxy for output pres by, for example, Jorgenson, uroda, and Nshmzu (1987), ee and Tang (2000), and most reently van Besebroek (2009) and Sørensen and Shjernng (2008). However, (24.17) shows that two further adjustments may be needed. In a poneerng attempt, Hooper (1996) tred to adjust expendture PPPs for nternatonal trade pres, but hs methodology was ad ho, and the adjustments were based on very aggregate data. As a result, ths approah was not pursued. Also, adjustments for ntermedate onsumpton have not yet been tred, and there s lttle hope n vew of the pauty of data on pres for ntermedate onsumpton. The onluson s that only the frst adjustment for margns and taxes s feasble n prate. The expeted sze of the other adjustments that need to be made but annot provdes nformaton about the possble bas assoated wth the use of expendture PPPs as a proxy for output PPPs. Based on nformaton from nput-output tables (see van Ark and Tmmer 2009), t s possble to make some general statements about the possble bases, n partular n omparsons nvolvng small open eonomes wth large export and mport shares. Ths bas wll dffer for eah setor. For ndustres that manly produe goods for ntermedate demand suh as agrulture, mnng, transport, bas goods manufaturng, and busness serves, expendture PPPs (E-PPPs) are not useful as proxes for output PPPs (O-PPPs). 10 On the other hand, E-PPPs appear aeptable for relatve pre dfferenes n food manufaturng, utltes, and ommunatons after adjustng for taxes and margns. E-PPPs are also useful for onstruton, hotels, and real estate atvtes beause fnal expendture shares are very hgh for these setors and mports are neglgble. 10 In addton, subsdes on ertan transport serve ategores are very dffult to peel off from the fnal expendture pre (e.g., bus, subway, and ral transport pres). 16

17 PPPs n the GGDC Produtvty evel Database Ths seton dsusses the soures and methodology of the PPPs for output, ntermedate nput, labor, and aptal used n the GGDC Produtvty evel database. Ths set of PPPs s based on a ombnaton of expendture and produton PPPs. Output and Value Added PPPs The output PPPs used n the GGDC Produtvty evel database rely heavly on a study by Tmmer, Ypma, and van Ark (2007). They presented a new and omprehensve data set of blateral output PPPs for a set of 30 OECD ountres, wth the Unted States as the base ountry at the ndustry level for the year Ths data set s based on a ombnaton of several data soures, both from the expendture and ndustry of orgn approahes. Expendture PPPs by expendture ategory were taken from the OECD omparson alloated to ndustres. For example, the expendture pres of bread and sugar were alloated to food manufaturng. In a seond step, the expendture PPPs were adjusted to a bas pre onept by peelng off trade and transport margns and taxes net of subsdes as outlned n the prevous seton. Produton PPPs were based manly on unt value ratos. The hoe for PPPs derved from ether the expendture or ndustry of orgn approah was n part dtated by the avalablty of pre data for a partular (three-dgt) ndustry and a areful assessment of the usefulness of eah n ase both alternatves were avalable. Ths approah s outlned n van Ark and Tmmer (2009) and Tmmer, Ypma, and van Ark (2007). Table 24.2 gves the type of PPP used for output for major setors. In addton, ts qualty s assessed. PPPs are ranked from 1 (very poor) to 5 (very good) on the bass of the followng rtera. The qualty of E-PPPs at the ndustry level depends on the share of fnal expendture n total use and the share of mport n total supply, as desrbed n the prevous seton. Ths qualty an be evaluated on the bases of nput-output tables, also as desrbed earler. The rtera for assessng the qualty of O-PPPs are dfferent from those for E-PPPs. Coneptually O-PPPs refer to the pres of domestally produed produts. But as dsussed earler, the man weaknesses of O-PPPs are the produt mx and qualty problems. Espeally for hgh-teh goods, or heterogeneous serves, O-PPPs an be affeted. In addton, for many serves no data are avalable on unt values beause of a lak of approprate value data and the dffulty n defnng quanttes. O-PPPs are therefore partularly useful for ndustres for whh produts are relatvely homogeneous and for whh dfferenes n produt qualty problems are small. 17

18 Table 24.2 Soure and Gradng for Industry PPPs n GGDC Produtvty evel Database Industry ISIC rev. 3 Grade PPP type used Remark A. Output PPPs Agrulture ode O-PPP Homogeneous goods Mnng and quarryng O-PPP Homogeneous goods Manufaturng E-PPP/O-PPP Food, drnk, and tobao 15, 16 4 Manly O-PPP Homogeneous goods Bas goods 17, 20, 21, Manly O-PPP Homogeneous goods Nondurable 18,19, 22, 36, 37 4 Manly O-PPP Homogeneous goods Durable Manly E-PPP Qualty, mport, and overage problem Eletrty, gas, and water supply 40, 41 4 E-PPP/O-PPP Homogeneous goods Construton 45 4 E-PPP Hgh expendture share Trade O-PPP/E-PPP Qualty problem Hotels and aterng 55 4 E-PPP Hgh expendture share Transport O-PPP Qualty problem Communatons 64 3 O-PPP Qualty problem Fnane E-PPP Referene PPP Real estate atvtes 70 4 E-PPP Hgh expendture share Busness serves E-PPP Small expendture share Publ admnstraton and defense 75 1 Manly wages Based on nput PPPs Eduaton and health 80, 85 1 Manly wages Based on nput PPPs Other serves E-PPP Df. produt mx B. Intermedate nput PPPs All ndustres 1-4 E-PPP/O-PPP C. Captal nput PPPs, all ndustres Construton 3 E-PPP PPP for nvestments Mahnery and equpment 4 E-PPP PPP for nvestments Based on output PPPs, so grade depends on mx of nput produts D. abor nput, all ndustres 4 Wages Relatve wages for varous labor types Soure: The authors. Note: Rankng: 0 = not avalable; 1 = very poor; 2 = poor; 3 = aeptable; 4 = useful; 5 = very useful. E- PPP refers to expendture PPPs for the OECD from the 1999 round, and O-PPP refers to produton PPPs for 1997 from Tmmer, Ypma, and van Ark (2007). Agrultural output onssts almost exlusvely of the produts used for ntermedate nput by other frms, not for fnal onsumpton. Therefore, expendture PPPs annot be used as a proxy for agrultural output PPPs. Instead, the agrultural PPPs for ths study are developed along the same lnes as earler ICOP work on agrulture (Prasada Rao 1993). We rely exlusvely on produton PPPs based on produer pres from the FAOSTAT database of the Food and Agrulture Organzaton (FAO). Ths database ontans a very extensve set of quanttes and farm pre values of up to 146 agrultural produts. Smlarly, for mnng and most manufaturng ndustres unt value ratos are used. Unt values for European ountres were derved from PRODCOM, whh has a harmonzed set of produt data for European Unon member states. The PRODCOM database nludes quanttes and sales values by produt, lnked to the NACE lassfaton, for up to 7,000 produt tems. 11 Ths database s essentally based on the orgnal natonal produton ensuses and ndustry surveys, but uses a harmonzed produt 11 NACE s the aronym used for the General Industral Classfaton of Eonom Atvtes wthn the European Communtes. 18

19 odng system. PRODCOM greatly enhanes the number of produt mathes on the bass of whh unt value ratos were onstruted. For non-european ountres, t uses omparable data from natonal produton surveys. In addton, t apples hedon UVRs for ars. Muh of the output n manufaturng ndustres suh as textles, pulp and paper, bas metals, nonmetall mnerals, and hemals onssts of relatvely homogeneous bas goods, and hene the qualty of the O-PPPs s relatvely hgh. Ths s not true for ndustres produng more sophstated spealzed goods suh as n eletral and noneletral equpment, transport equpment, and nstrument manufaturng. For these ndustres, the data set manly makes use of adjusted omponent expendture PPPs from the ICP. Although ther qualty s hgher than that of unt value ratos, qualty problems also plague purhaser s pre omparsons of hgh-teh goods. E-PPPs annot be dretly used for the dstrbuton setor beause the output n ths setor s measured as the margn of sales over purhases and separate deflaton of ntermedate purhases s neessary. But wth an adjustment for the margn to sales rato for eah tem, E-PPPs an be used for the retal setor. For retal trade, expendture PPPs for ndvdual expendture ategores were dretly appled to sales output. The PPP at the margn level s derved as a weghted average of the sales PPPs of all goods, orreted for dfferenes n the margn to sales ratos between two ountres. In the ase of wholesale trade, only the unt values of goods purhased by the wholesale setor are observable. Margn PPPs are derved by adjustng for dfferenes n margn to ost ratos between two ountres. Informaton on retal and wholesale sales, purhases, and margns were obtaned from natonal trade ensus and survey data. Although ths approah s superor to usng unadjusted E-PPPs, t only partally orrets for dfferenes n the qualty of the trade serve provded see Tmmer and Ypma (2006) for addtonal dsusson on ths approah. For transport and ommunaton serves, UVRs were used, based on value and quantty data from a wde varety of nternatonal soures. Beause of the hgh level of ntermedate use, E- PPPs are only poor proxes. But gven the relatvely broad desrptons of the produts used (up to nne produt groups are dstngushed), the qualty of the UVRs s lower than that for manufaturng ndustres. For other setors (e.g., onstruton), produton PPPs are very poor or nonexstent, and adjusted expendture PPPs are used extensvely. Coneptually, expendture PPPs are a very good proxy for the PPPs for the gross output of onstruton, hotel and restaurant, and real estate atvtes 12 beause almost all the output of these ndustres s for fnal expendture, wth very lttle export and mport atvty, so that adjustments need only be mnmal. However, the qualty of the ICP expendture PPPs themselves s not always partularly hgh (e.g., for onstruton). For other ndustres, expendture PPPs are poor proxes beause a szable share of these serves s used for ntermedate, not fnal, onsumpton. And the mx of serves used by produers wll 12 Ths PPP an be based on the expendture PPP for rents. 19

20 dffer onsderably from the serves used by fnal onsumers, suh as for busness and other serves. Beause we lak alternatves, we use adjusted expendture PPPs, but note n table 24-2 that they are a poor proxy for an output PPP. The expendture PPP for fnane s a referene PPP that s based on the overall expendture PPP rather than on the relatve pres of fnanal serves. Beause of the way n whh fnanal output s urrently deflated n most natonal aounts, ths prate s perhaps defendable, but t s learly unsatsfatory, and more researh s needed to measure both fnanal output and pres. For publ admnstraton, eduaton and health expendture PPPs have not been used. In almost all ountres, the output n these setors s measured by means of nputs. There s a reent tendeny n some ountres to ome up wth genune output measures. However, by and large our assumpton that output s measured by nputs holds true, n partular for our benhmark year, By mplaton, produtvty levels should be the same aross all ountres. Put otherwse, output PPPs should be a weghted sum of the nput PPPs, wth weghts ndatng the share of eah nput n total output. However, when we ompared our nput PPPs wth the expendture PPPs gven by the OECD for these ndustres, large dfferenes were found. In partular, the labor PPPs used by the OECD are rather dfferent from our labor PPPs. Further srutny of the ICP PPPs for ths setor s warranted (see hapter 11 of ths book). Therefore, we deded to defne the output PPPs for nonmarket serves (ndustres, M, N, and P) as a weghted sum of our nput PPPs. Consequently, the omparatve multfator produtvty levels n these ndustres are all equal to one by defnton. Intermedate nput PPPs should reflet the osts of aqurng ntermedate delveres and math the pre onept used n the nput-output tables, hene at bas pres plus net taxes. The data problems assoated wth obtanng nput PPPs for ndvdual ndustres are larger than those assoated wth output. There s often no nput pre parallel to the output PPPs. Busness statsts surveys and produton ensuses provde lttle or no nformaton on quanttes and values of nputs n manufaturng, and for nonmanufaturng ndustres the nformaton s largely absent. Moreover, by defnton, PPPs from the expendture sde do not reflet the pres of ntermedate nputs beause they over only the fnal expendture ategores. In ths study, we use output PPPs as a proxy for relatve ntermedate nput pres under the assumpton that the bas pre of a good s ndependent of ts use that s, we use the same gross output PPP of an ndustry to deflate all ntermedate delveres from ths ndustry to other ndustres. The aggregate ntermedate nput PPP for a partular ndustry an be derved by weghtng ntermedate nputs at the output PPP from the delverng ndustres. Imported goods are dentfed separately, and exhange rates are used as onverson fators for mports. Ideally, one would lke to have separate estmates of mport PPPs based on trade data beause there s lttle evdene that the law of one pre holds for all goods even when nternatonally traded. However, so far these data are not readly avalable. 20

21 PPPs for output and ntermedate nput an be ombned to alulate PPPs for value added as desrbed earler. Table 24.3 presents value added PPPs for the market eonomy and three man setors of the eonomy manufaturng, other goods, and market serves for In addton, t provdes PPPs for aggregate GDP and exhange rates as avalable from the OECD; both have been used as alternatves n prevous studes. As noted n the lterature, ths approah gnores the dfferenes n pres aross varous ndustres as well as the dfferenes n the pres of ntermedate nputs and outputs generally seen as a major weakness (Sørensen 2001). Instead, the value added PPPs have been derved by separate deflaton of output and ntermedate nputs as n equaton (24.9). As shown n table 24.3, the rato of setoral value added to GDP PPPs an vary between 75 perent and more than 200 perent. The PPPs for the market eonomy are generally hgher than the GDP PPPs, manly beause the latter nludes nonmarket serves, whh, aordng to the OECD PPP results, are expensve n the Unted States ompared wth other ountres. Importantly, the table shows large dfferenes n relatve pres aross setors, onfrmng the fndngs by van Besebroek (2009) and Sørenson and Shjernng (2008). For example, the PPP for other goods n Japan s muh hgher than the PPP for manufaturng goods. Ths dfferene s manly due to the hgh output pres n the agrultural setor, whh s famous for ts weak ompettveness and strong mport proteton (van Ark and Plat 1993). The use of an overall GDP PPP would greatly overestmate produtvty levels n ths setor. On balane, the value added PPPs for manufaturng dffer by about 16 perent from the GDP PPP aross our set of ountres (absolute log dfferenes). Ths dretly translates nto a 16 perent dfferene n measures of produtvty levels. For market serves, the dfferene s omparable (15 perent), whle for other goods t s even bgger (32 perent). 21

22 Table 24.3 Varous Alternatve Value Added PPP Measures: Seleted Countres, 1997 Value added PPPs GDP (OECD) Market eonomy Manufaturng Other goods Market serves Exhange rate Australa Austra Belgum Czeh Republ Denmark Fnland Frane Germany Hungary Ireland Italy Japan uxembourg Netherlands Portugal Slovena Span Sweden Unted ngdom Unted States Soures: Value added PPPs based on the GGDC Produtvty evel database (Inklaar and Tmmer 2009); GDP PPPs and exhange rates from OECD (2002). The GDP PPP s extrapolated from 1999; see OECD (2002). Note: All entres n are natonal urreny per U.S. dollar. For ountres that adopted the euro n 1999, the 1999 onverson rate was used on the pre-euro urrenes. abor and Captal PPPs Comparsons that use a homogeneous (or raw ) labor onept n the denomnator of the produtvty equaton, suh as number of workers or total hours worked, do not need urreny onverters for labor nput, beause the omparson s already gven n terms of volume. In the ase of a heterogeneous labor onept for example, workers of dfferent skll types labor nput PPPs are needed to orret total labor ompensaton for dfferenes n the relatve pres of dfferent ategores of workers. Ideally, ths labor nput PPP should be based on labor osts, nludng all osts nurred by the produers n the employment of labor suh as taxes leved, health ost payments, other types of nsurane and ontrbutons to retrement pad by the employer, fnanal benefts suh as stok optons, and the value of payments n knd and allowanes (suh as housng and rent). 22

23 The PPP for labor represents the relatve pre of one unt of labor between two ountres. For eah type of labor, relatve wages an be alulated. The EU EMS Growth Aounts dstngush between 18 dfferent labor types: two gender ategores, three age ategores, and three eduatonal attanment ategores. The eduatonal attanment ategores are low sklled (preprmary, prmary, and lower seondary eduaton Internatonal Standard Classfaton of Eduaton [ISCED] 0 2), medum sklled (upper seondary eduaton, ISCED 3), and hgh sklled (total tertary eduaton, ISCED 5 7). However, n partular for level omparsons ths lassfaton s rough and mght be msleadng beause eduatonal systems n Europe and the Unted States are very dfferent. In partular, the dfferent roles of voatonal shoolng systems ause problems of omparablty aross ountres. For example, n Germany voatonal tranng s mportant to enterng many oupatons, but ths s less prevalent n the Unted States. Based on the work by the Brtan s Natonal Insttute of Eonom and Soal Researh (see Mason, O eary, and Veh 2009), we made a more detaled omparson and further deomposed the medum-skll level nto three ategores for a total of fve. Further researh on the omparablty of shoolng qualfatons aross ountres s needed, however. To onvert aptal nput measured n natonal pres nto ommon pres, aptal nput PPPs must be developed. Captal PPPs gve the relatve pre of the use of a unt of aptal n two ountres from the purhaser s perspetve. The alulaton of the aptal nput PPP s less straghtforward than for output, ntermedate nput, or labor nput PPPs beause of the oneptualzaton of aptal nput as aptal serves rather than aptal stoks. To obtan the relatve pres for aptal nput, we follow Jorgenson and Nshmzu (1978). Under the assumpton that the relatve effeny of new aptal goods s the same n both ountres, PPP k ountry C and the base ountry, the Unted States, s alulated as, the relatve rental pre of asset k between I pk, C p I k, C (24.18) PPPk PPPk. I p p k, US k, US Ths defnton ndates that the relatve rental pre of a unt of aptal between two ountres depends on the relatve purhaser s pre of a new aptal good of asset k between ountry C and the I Unted States ( PPP, ) and the relatve ost of aptal nput, wth k t nvestment pre. One an thnk of the user ost rato p k p I k pk the user ost and I p k the as the relatve pre of hrng a smlar aptal good for one year n both ountres. In the absene of taxaton, the famlar ost of aptal equaton for asset type k an be wrtten as (Jorgenson and Grlhes 1967) Ths formula s an approxmaton beause we exlude seond-order effets. Country subsrpts are dropped to avod lutterng. 23

24 pk (24.19) r ˆ I k pk. I p k Ths formula shows that the user ost s determned by the nomnal rate of return (r), the rate of eonom depreaton (δ), and the asset-spef aptal gans measured as the hange n nvestment pre ( I pˆ k ). Investment PPPs are olleted n the OECD ICP program and are avalable for 35 aptal assets from the OECD (2002) for Beause they are already at purhasers pres, no adjustment s needed to arrve at the nput pre onept. The PPPs for the 35 assets are aggregated to the eght assets n ths study usng a CCD aggregaton proedure. Investment deflators by asset and ndustry from the EU EMS database are used to move these PPPs to the benhmark year, The rates of depreaton are geometr rates that vary aross assets, but are assumed to be dental aross ountres as n the EU EMS growth aounts (orgnally based on Fraumen 1997). Beause we use the ex ante approah to aptal measurement, the nomnal rate of return s the 10-year government bond yeld for 1997 taken from the Internatonal Monetary Fund s database Internatonal Fnanal Statsts. It ndates the opportunty ost of usng the nvestment fund n a rsk-free alternatve. The asset revaluaton term an be derved from nvestment pre ndes. To mnmze the mpat of sometmes volatle hanges, annual averages over the perod are used. Fnally, we multply the asset- and ndustry-spef rental pres wth the aptal stok taken from the aptal nput fles from the EU EMS database (Marh 2008 release) to derve the ex ante aptal ompensaton. Ths wll typally dffer from the (ex post) aptal ompensaton as gven n the natonal aounts. We normalze the ompensaton by asset to the natonal aounts fgure by proportonalty. 14 Table 24.4 s an overvew of the labor and aptal PPPs for the market eonomy n For referene, we also nlude the value added PPP for eah ountry. It appears that aptal serves are relatvely expensve n most ountres ompared wth the Unted States the aptal nput PPP s often muh hgher than the value added or the labor PPP. Espeally n Eastern Europe and Ireland, the use of aptal n produton s expensve. abor-nput PPPs, on the other hand, vary wdely aross the set of ountres. In ountres suh as Belgum, Denmark, Germany, and the Netherlands, wages are relatvely hgh, whereas n the lower-nome ountres suh as Portugal, Span, and those n Eastern Europe, the labor nput s heap. But that s not beause the share of low-sklled workers s hgher n the latter set of ountres; we ompare the wages of groups of workers wth equvalent levels of eduatonal attanment n all ountres See Oulton (2007) for ths hybrd approah to aptal measurement. 15 See the aveats n the prevous seton. 24

25 Table 24.4 abor and Captal PPPs: Seleted Countres, 1997 Value added PPP abor nput PPP Captal nput PPP Australa Austra Belgum Czeh Republ Denmark Fnland Frane Germany Hungary Ireland Italy Japan uxembourg Netherlands Portugal Slovena Span Sweden Unted ngdom Unted States Soures: PPPs are based on the GGDC Produtvty evel database (Inklaar and Tmmer 2009). Note: All entres are n natonal urreny per U.S. dollar. For ountres that adopted the euro n 1999, the 1999 onverson rate was used on the pre-euro urrenes. Produtvty evel Comparsons for the OECD The GGDC Produtvty evel database provdes a wde range of results on omparatve pres, nput and output quanttes, and produtvty at the ndustry level. In ths seton, we provde only a seleton of the data avalable n the GGDC Produtvty evel database; detaled results an be found at We also dsuss the senstvty of the results. Produtvty evels Basally, our produtvty level omparsons are based on deflatng nomnal nputs and output as gven n natonal nput-output tables by a set of relatve pres. The nomnal values of nputs and outputs are based on natonal ndustry by ndustry nput-output tables (IOTs), wth separate nformaton on domest and mported supples of ommodtes. IOTs are not avalable for all ountres n a ommon benhmark year, and so we used supply-use tables to onstrut omparable IOTs. The startng pont of our analyss s the natonal supply-use table for eah ountry, valued n natonal urreny for For Canada, the Unted States, and Australa, these tables were obtaned from the natonal statstal offes. Eurostat makes these tables avalable for the European ountres 25