Productivity Levels Between Canadian and U.S. Industries

Transcription

1 Productvty Levels Between Canadan and U.S. Industres Someshwar Rao, Janmn Tang and Wemn Wang Mcro-Economc Polcy Analyss Branch Industry Canada 235 Queen Street, Room 1002B Ottawa, Ontaro K1A 0H5, Canada We wsh to thank Andrew Sharpe for comments and suggestons and the staff of the Income and Expendture Accounts Dvson of Statstcs Canada for performng the detaled level calculatons and aggregaton requred to derve the ndustry purchasng power partes (PPPs) for ths study. We are also grateful to Ronald Roux of the Input and Output Dvson of Statstcs Canada for provdng commodty margn data and Joanne Flemng for edtoral assstance. All opnons expressed n ths paper are entrely our own and should n no way be attrbuted to Industry Canada. 1

2 Abstract Ths study, usng the methodology gven n Jorgenson and Kuroda (1995) and Lee and Tang (2001) and the new 1999 benchmark data on expendture-based blateral commodty purchasng power partes (PPPs) jontly developed by Statstcs Canada and the U.S. Bureau of Labor Statstcs, estmates Canada-U.S. blateral PPPs for gross output, ntermedate nput, two types of captal stock [structures and machnery and equpment (M&E)] and value added n 31 ndustres. These PPPs n turn are used to estmate the Canada-U.S. labour productvty and multfactor productvty (MFP) level gaps for all these ndustres. These benchmark gaps n conjuncton wth the ndustry productvty growth rates n the two countres are then used to develop tme seres data on the labour productvty level gaps. The new estmates show that Canada s aggregate labour productvty was more than 20 percent below the U.S. level n 2003 and wdened sgnfcantly snce In addton, they show that Canada s more productve than the Unted States n most of the resource-based manufacturng ndustres, but less productve n the two machnery ndustres and n all servce ndustres. The new results n general are consstent wth the exstng lterature on the subject, but there are two key dfferences. Frst, the Canada-U.S. manufacturng labour productvty gap n ths study s sgnfcantly lower than prevous estmates. Ths dfference s manly due to the sgnfcant downward revsons to the productvty gaps n the two machnery ndustres. Second, ths paper shows that the dfferences n captal ntensty levels explan about 30 percent of the aggregate labour productvty gap between the two countres n 2001, and more than 50 percent of the Canada-U.S. manufacturng labour productvty gap. These estmates are sgnfcantly hgher than those n Rao, Tang and Wang (2003). 2

3 1. Introducton Canadans enjoy one of the hghest standards of lvng and qualty of lfe n the world. Moreover, both productvty and real per-capta ncome n Canada have ncreased at a sgnfcantly faster pace n the second half of the 1990s and n the 2000s. But, GDP per-capta n Canada, measured n purchasng power party (PPP) exchange rate, was stll on average 15 percent below the level n the Unted States (U.S.) n the past decade. About 85 percent of the ncome gap was attrbuted to the Canada-U.S. productvty gap, and the gap was generally pervasve across ndustres and provnces (Rao and Tang, 2004b). A relable estmate for the Canada-U.S. productvty gap s crucal for two mportant reasons. Frst, Canada's ablty to compete effectvely n nternatonal markets fundamentally depends on Canada's productvty relatve to ts tradng partners, especally the U.S., ts largest tradng partner. Second, Canada's strength n attractng and retanng physcal and human captal and nnovaton actvtes, fundamental drvers of economc growth, largely depends on Canada s ablty to pay hghly compettve returns to all the factors of producton, whch s fundamentally determned by ts relatve productvty. Hence, large and wdenng productvty gaps would erode Canada's future economc performance, creatng a vcous cycle of low economc growth and a wdenng of productvty and real ncome gaps. But, estmatng real ncome and productvty level gaps n varous ndustres across countres s tedous and tme consumng, and requres large amounts of hgh qualty data on prces, output and nputs. To compare productvty levels n dfferent ndustres between Canada and the Unted States, accurate data on blateral PPP exchange rates by ndustry s necessary, because use of the market exchange rate wll lead to unrelable and msleadng estmates for productvty gaps. Lee and Tang (2001), usng the 1993 benchmark data on expendture-based Canada-U.S. blateral commodty PPPs, developed estmates for gross output based labour productvty and multfactor productvty (MFP) gaps for about 30 ndustres. Rao, Tang and Wang (2003), usng the 1987 value-added based blateral PPPs from Plat (1996), estmated the labour productvty level gaps for Canadan ndustres. There have been a number of sgnfcant developments snce these prevous studes of PPPs. Frst, software purchases are now treated as nvestments, as fnal demand rather than as ntermedate nputs n the prevous natonal account system. The new treatment sgnfcantly ncreases output and captal stock n terms of both levels and growth rates. Second, composton of ndustry output has changed sgnfcantly snce 1993, especally durng the economc resurgence n the second half of the 1990s. Informaton and communcaton technologes (ICTs) have been playng an ncreasngly mportant role n both ICT-producng and -usng ndustres n the two countres. 1 Fnally, there s a new benchmark survey on expendture-based blateral commodty PPPs between Canada and the Unted States, jontly developed by Statstcs Canada 1 A varety of studes fnd that the productvty growth revval n the second half of 1990s n the U.S. s due to producton and use of ICTs (Jorgenson, Ho and Stroh, 2002; Plat and Lee, 2001). Some studes also show that the productvty growth pck-up n the second half of 1990s n Canada s manly due to the use of ICTs (e.g. Rao and Tang, 2001). 3

4 and the U.S. Bureau of Labor Statstcs. Ths paper wll address these developments and n partcularly, t wll sngle out both ICT producton and use. The objectve of ths study s to develop new estmates for the Canada-U.S. productvty gaps for 31 ndustres, takng nto account these new developments. Usng the methodology n Jorgenson and Kuroda (1995) and Lee and Tang (2001) and the 1999 benchmark data on expendture-based blateral commodty PPPs between Canada and the Unted States, we calculate blateral PPPs for gross output, ntermedate nput, two types of captal [(structures and machnery and equpment (M&E)] and value added for 31 ndustres. These n turn are used to develop estmates for valueadded based labour productvty and MFP gaps n all the ndustres. A gross output PPP s defned as the amount of Canadan dollars receved by a Canadan producer from sellng the same quantty of goods and servces that a U.S. producer sells for one U.S. dollar. Smlarly, an nput PPP s defned as the amount of Canadan dollars requred for a Canadan producer to purchase the same quantty of nput that a U.S. producer purchases for one U.S. dollar. We use Canada-U.S. blateral commodty PPPs and convert them nto ndustry PPPs by usng gross output, nput and nvestment structures of each ndustry that are estmated from nputoutput tables of the two countres. We group the busness sector nto 31 ndustres. The ndustres are not exact NAICS (North Amercan Industry Classfcaton System) ndustres. Instead, they are grouped n such a way so that we can establsh a concordance between NAICS and SIC (Standard Industral Classfcaton). 2 The ndustry lst and the correspondng NAICS codes are reported n Table 1. They are roughly at the 2-dgt level for prmary ndustres, the 3- dgt level for manufacturng ndustres and the 2-dgt level for servces ndustres. Note, however, that computer and communcaton equpments are separated from electronc equpments due to the ncreasng mportance of the two ICT ndustres. The fndngs of ths study n general are consstent wth the conclusons of earler studes (Rao, Tang and Wang, 2003; Lafortune and Lee, 2003). However, two fndngs of the current study dffer from prevous studes n two mportant aspects. Frst, the manufacturng labour productvty gap n ths study s sgnfcantly lower than those n the earler studes. For nstance, the estmated gap n 1999 n ths study was 18 percent, compared to 31 percent n Rao, Tang and Wang (2003). Nevertheless, both studes show a consderable wdenng of the gap snce The productvty gaps n the two machnery ndustres n ths paper are consderably lower than those n the prevous study, largely due to dfferences n the PPPs. Second, ths study suggests that dfferences n captal ntensty n the two countres can explan about 30 percent of the aggregate Canada-U.S. labour productvty gap n In the manufacturng sector, more than 50 percent of the gap can be attrbuted to the captal ntensty gap. These are sgnfcantly larger than those n Rao, Tang and Wang (2003). Our paper s organzed n the followng way. Secton 2 outlnes the conceptual framework for developng blateral PPPs between Canada and the Unted States, and dscusses the new blateral PPP estmates. Secton 3 provdes new estmates for labour productvty and MFP level gaps. It also provdes estmates for Canada's prce compettve poston n all the ndustres. The last 2 The U.S. ndustry data for productvty analyss n ths paper s based on SIC. 4

5 secton, secton 4, summarzes the key fndngs of the study and explores ther mplcaton for future research and publc polcy. 2. Estmatng Blateral Purchasng Power Partes for Output and Inputs In ths secton, we explan how to adjust expendture-based commodty PPPs for margns, whch s requred for estmatng gross output PPPs at producers prces. We then present the methods for convertng commodty PPPs nto ndustry PPPs, followed by a bref descrpton of data requred for constructng ndustry PPPs. Fnally, we dscuss the PPP estmates PPPs at Producers Prces The avalable raw PPPs, whch are the essental data for ths study, are expendture-based blateral commodty PPPs between Canada and the Unted States, evaluated at purchasers prces. To estmate ndustry PPPs for gross output, we need to have producton-based commodty PPPs at producers prces. To obtan producton-based commodty PPPs at producers prces, the expendture-based commodty PPPs must be adjusted for margns (whch nclude ndrect net commodty taxes and dstrbuton margns). 3 The purchasers prce of a commodty s equal to ts producers prce plus all the related margns. Assumng that the margn value of purchasng a unt of the commodty can be expressed as a percentage share of ts producers prce, we have: E P (1) P = P 1+ δ ), s s ( s where P, and E j s P, are expendture prces of commodty j n country s (expressed as the P j s natonal currency of country s ) at purchasers and producers prces, respectvely, and δ s s the margn rate of commodty j n country s. The Canada-U.S. expendture-based blateral commodty PPP of a commodty s the rato of natonal prces of the commodty n the two countres,.e., (2) PPP P j P E P CA P CA (1 + δ CA ) 1+ δ US E = = = PPPj, P P (1 + δ ) 1+ δ P US E US US CA 3 Ideally, they should also be adjusted for nternatonal trade (exports and mports), as the one suggested by Hooper and Vrankovch (1995). Exports are domestcally produced but not domestcally consumed, whle mports are domestcally consumed but not domestcally produced. An adjustment for exports and mports s desrable to convert expendture-based PPPs to producton-based PPPs. However, due to data lmtaton, ths paper, lke all prevous studes, wll not make the adjustment. 5

6 E P where PPP j and PPP j are Canada-U.S. expendture-based blateral commodty PPPs for commodty j at purchasers and producers prces, respectvely. The margn rates for each commodty can be estmated from the nput-output accounts of the two countres Industry PPPs Industry PPPs can be derved from producton-based and expendture-based commodty PPPs and nput-output structures of the two countres. Ths study follows the methodologes used n Jorgenson and Kuroda (1995) and Lee and Tang (2001) PPPs for Gross Output In ths subsecton we descrbe the method for convertng commodty PPPs at producers prces to gross output PPPs at the ndustry level. An ndustry gross output s defned as the value of the ndustry s producton at producers prces. For most ndustres, t s the sum of all commodtes produced at producers prces. However, for margn ndustres such as transportaton, retal and wholesale trade ndustres, gross outputs are manly margns receved, not the value of goods transported or traded. Thus, gross output PPPs for these ndustres cannot be derved from expendture-based commodty PPPs. In ths study, gross output PPPs for these ndustres are assumed to be 1.18, the same as the Statstcs Canada 1999 GDP PPP. In the remander of ths subsecton, we construct gross output PPPs for all ndustres except transportaton, retal and wholesale trade. As defned prevously, the Canada-U.S. blateral gross output PPPs are the amount of Canadan dollars receved by Canadan producers from sellng the same quantty of the output that a U.S. producer sells for one U.S. dollar. For ndustry, the gross output PPP s estmated by aggregatng the ndustry s expendture-based commodtes PPPs at producers prces over all the commodtes produced by ndustry n a trans-log form, usng nomnal shares n the commodty mx as weghts for the ndustry. That s, Y Y Y P Y (3) ln ( PPP ) = ( w, CA + w, US ) ln( PPP ) j s = j s j, wth w,, Y,, where Y j, s j 1 2 j Y, s,, s the nomnal gross output of commodty j at producers prce n ndustry n country s, obtaned from the make matrces of the nput-output tables, and share of commodty j n ndustry n country s PPPs for Intermedate Input Y w j, s, s the nomnal Unlke ndustry gross output PPPs, ndustry PPPs for ntermedate nputs should be calculated from the purchasers standpont, whch mples that they need to be converted from expendturebased commodty PPPs at purchasers prce. That s, 6

7 1 2 Φ Φ Φ E Φ (4) ln ( PPP ) = ( w, CA + w, US ) ln( PPPj ), wth w, s = Φ, s Φ, s, j j where Φ s the value of commodty j as an ntermedate nput n ndustry n country s,, s Φ w j, s obtaned from the use matrces of the nput-output tables, and, s the share of commodty j n total value of ntermedate nputs used by ndustry n country s PPPs for Investment Industry PPPs for new nvestment should also be calculated from the purchasers standpont. These PPPs are defned as the amount of Canadan dollars requred for a Canadan producer to purchase the same quantty of new nvestment goods that a U.S. producer purchases for one U.S. dollar. We group new nvestment goods nto M&E and structures. The PPP for new nvestment type I (M&E, structures, or both (total nvestments)) n ndustry s aggregated over commodtes n a trans-log form,.e., 1 ln, CA,, US, s =, s, 2 j j I I I E I I I (5) ( PPP ) = ( w + w ) ln( PPP ), wth w K K, j j s I K j, s where, s the value of commodty j as nvestment type I n ndustry n country s, obtaned from the fnal demand matrx for Canada and the captal flow matrx for the Unted I States, and w, j, s s the share of commodty j n total value of nvestment type I purchased by ndustry n country s PPPs for Value Added PPPs for value added are derved from ts relatonshp wth gross output and ntermedate nput. Value added s calculated as a resdual of gross output mnus ntermedate nput. Followng Jorgenson (1995), a trans-log form of Canada-U.S. relatve value added can be defned as 1 1 ν ln CA, US, CA, US, CA, ν ν (6) ( V V ) = ln( Y Y ) ln( Φ Φ ),, US where V, s s value added for ndustry n country s ; Y, s s gross output value for ndustry n country s ; Φ, s s value of ntermedate nputs for ndustry n country s ; and ν s the average share of nomnal value added n gross output of the two countres for ndustry,.e., (7) 1V, CA V, US ν = +. 2 Y, CA Y, US 7

8 Note that equaton (6) s vald n natonal currences or PPP adjusted values, so the ndustry PPPs for value added can be expressed as (8) ln PPP 1 1 ν ν Y Φ = ln PPP ln PPP ν ν Thus, the derved PPPs from equaton (8) s the mplct PPPs for value added for ndustry Data for the Estmaton of PPPs To mplement the methodology descrbed above, we requre data on PPPs for commodtes between Canada and the Unted States, and the nput-output tables wth dfferent margns for both Canada and the Unted States.. The basc data on commodty PPPs are Canada-U.S. expendture-based blateral PPPs for 221 basc headngs n 1999 and they are aggregated over more than 2,000 commodtes. 4 They are developed jontly by Statstcs Canada and the U.S. Bureau of Labor Statstcs. However, the commodty PPPs are manly for commodtes that are produced for fnal consumptons. The dataset does not have PPPs for most commodtes that are prmarly used as ntermedate nputs. To fll the gap, we frst use 1999 Canada-U.S. market exchange rate as a proxy PPP for commodtes (such as grans, wheat, copper, steel, and precous metals) that are heavly traded n the North Amercan or the world markets. Other commodtes that are used as ntermedate nputs n the manufacturng sector, such as chemcal, rubber and plastc, non metallc mnerals, and metal products, are proxed by unt value ratos (UVRs) 5 for 1997 from van Ark, Inklaar and Tmmer (2000), and updated to 1999, usng nformaton from the KLEMS database for both Canada and the Unted States (Ho, Rao and Tang, 2004). The second major data requred for the exercse are nput and output tables expressed n current dollar for both Canada and the Unted States. For Canada, we obtaned the make and use tables at the publc aggregate level (713 commodtes by 283 NAICS-based ndustres) and the fnal demand table at the publc aggregate level (713 commodtes by 170 expendtures) for 1999 from Statstcs Canada. The fnal demand table contans nformaton on captal expendtures by asset types for detaled ndustres. These tables are manly used to measure the commodty composton for each ndustry. For Canada, we also obtaned nput and fnal demand margns for 1999 (transportaton, wholesale, retal and net ndrect commodty tax) of 713 commodtes from Statstcs Canada. They are used to derve commodty PPPs at producers prces by peelng off the margns from expendture commodty PPPs. 4 Commodty PPPs are updated every three years. The 1999 data were the most current at the tme of ths study. Statstcs Canada performed the detaled level calculatons and aggregaton of these PPPs. 5 UVR and PPP are alternatves used n the lterature to convert dfferent currences nto a common currency n nternatonal productvty comparson. The major dfference between the two approaches s that the former s based on product unt values derved from value and quantty nformaton for product group and the latter s based on fnal expendture nformaton. For the dscusson of the strengths and weaknesses of the two approaches, see van Ark, Inklaar and Tmmer (2000). 8

9 Smlarly, we obtaned the correspondng 1999 nput and output tables for the Unted States from the U.S. Bureau of Economc Analyss. However, the U.S. make and use tables (margns ncluded) are at a very aggregated level (97 commodtes by 97 SIC-based ndustres). So we supplemented the 1999 tables wth nformaton from the 1997 make and use tables (margns ncluded) that are at the detaled level (511 commodtes by 511 NAICS-based ndustres). In other words, we used the commodty composton structure from the 1997 tables, controllng the totals from the 1999 tables. In addton, we obtaned the 1997 captal flow table (180 commodtes by 123 NAICS-based ndustres) for the Unted States, whch was used to measure the commodty composton of nvestment n each ndustry. The data assocated wth the nput and output tables for Canada and the Unted States have dfferent commodty codes and they are at a dfferent ndustry level. In addton, the commodtes wth PPPs are more aggregated than those n the nput and output tables. As a fnal step n the data preparaton, we establsh a common commodty level wth 227 commodty headngs for both Canada and the Unted States. Both the Canadan and U.S. tables are converted nto new tables wth the 227 commodtes and the 31 ndustres (Table 1). These new tables are then used to calculate the margn rates and ndustry shares for each headng n both countres. The margn rates wll be used to convert expendture-based commodty PPPs at purchasers prce to producton-based commodty PPPs at producers prce, and the ndustry shares of each commodty, measurng the commodty composton structure for each ndustry, wll be used to estmate ndustry PPPs Dscusson of the PPP Estmates In ths secton, we brefly dscuss the ndustry PPP estmates and compare them to offcal estmates at the aggregate level PPPs for Gross Output, Intermedate Input and Value Added PPP estmates for gross output, ntermedate nputs and value added are reported n Table 2. Gross output PPPs vary sgnfcantly across ndustres. Industres wth ther outputs heavly traded n the North Amercan or the world market have hgh gross output PPPs and are close to the market exchange rate. These ndustres are agrculture, textle and clothng, wood, rubber and plastc, fabrcated metal, machnery, computer, communcaton equpment, electrcal equpment, and furnture. On the other hand, for ndustres that are not heavly traded nternatonally such as servces and constructon, the gross output PPPs are well below the market exchange rate, and are closer to the party. PPPs for ntermedate nputs are less varable across ndustres than gross output PPPs. 6 There s no ndustry that has ntermedate nput PPP hgher than the market exchange rate. Lke gross output PPPs, the ntermedate PPPs for servce ndustres are on the low sde, and those for most manufacturng ndustres are on the hgh sde. 6 Ths partly reflects the fact that we used the proxy data (the market exchange rate) for some commodty PPPs used for ntermedate nputs. 9

10 The PPP for value added n the busness sector s 1.14, whch s smlar to the offcal estmates for the total economy (1.18). The small dfference s partly because the value added n ths paper s at producers prce whle the offcal estmate s at purchasers prce. It may also reflect the dfference n coverage (ths study does not cover the non-busness sector). At the ndustry level, value added PPPs are much more varable across ndustres than both gross output and ntermedate nput PPPs. Ths may be partly due to the fact that they are estmated mplctly. The value added PPPs are hgher than the market exchange rate n 13 ndustres whle they are below one n three ndustres PPP for Investment PPP estmates for total new nvestment as well as nvestments n M&E and structures are reported n Table 3. Our estmate of PPP for total nvestment for the busness sector n 1999 s 1.20, whch s smlar to the offcal Statstcs Canada estmate for gross fxed captal formaton (1.17). Agan, the slght dfference between the two estmates may reflect the dfference n coverage. The PPP for M&E nvestment for the busness sector s 1.37 n 1999, whch ncludes expendtures on not only M&E but also software products. The offcal Statstcs Canada estmate s 1.47 for M&E excludng software, whch more or less s equal to the market exchange rate, and 0.79 for software products. The PPP for nvestment n structures for the busness sector s 1.03 n 1999, compared to 1.02 of the offcal estmate by Statstcs Canada. It s dentcal for all ndustres but one, because there s only one type of nvestment n structures (constructon) n all these ndustres. The PPP estmates for nvestment n M&E beng sgnfcantly hgher than those for nvestment n structures reflect the fact that much of the M&E s mported. 3. Productvty Levels and Prce Compettveness of Canadan and U.S. Industres Ths secton fulfls the second objectve of ths study, that s, to compare productvty levels and nternatonal prce compettveness of Canadan and U.S. ndustres, based on the PPPs estmates above. To carry out such comparsons, we frst set up the framework The Framework for Productvty Level Comparsons As n Jorgenson and Nshmzu (1978), our theoretcal framework for productvty and nternatonal prce compettveness comparsons between Canada and the Unted States s based on a trans-log producton functon, orgnally ntroduced by Chrstensen, Jorgenson and Lau (1971, 1973). Here, output (value added) 7 s a trans-log functon of captal and labour, as well as a dummy varable equal to one for Canada and zero for the Unted States, and tme as an ndex of technology for each ndustry. 7 For the analyss at the ndustry level, Jorgenson and hs assocates often rely on the gross output concept for productvty analyss, that s, nputs also nclude ntermedate nputs. Unfortunately, we do not have the data for conductng a smlar analyss. 10

11 From the producton functon, Jorgenson, Kuroda and Nshmzu (1995) and Chrstensen, Cummngs and Jorgenson (1995) show that dfferences n the logarthms of the mult-factor productvty (MFP) levels between Canada and the Unted States, for the th ndustry, can be expressed as the value of the dfference between the logarthms of the value added, less a weghted average of the dfferences between the logarthms of nputs: MFP ( ) ( ) ( ) (9) ln ln ln Can LP Can ) K k Can = v, MFP ( US) LP ( US) k ( US) where MFP (S) s the MFP level for the th ndustry n country S; LP (S) s the PPP-based labour productvty level, defned as value added per hour worked, for the th ndustry n country S; k (S) s the PPP-based captal ntensty level, defned as captal stock per hour worked, for the K 1 K K th ndustry n country S; and vˆ [ v ( Can) v ( US) ] = +, the average compensaton share of 2 captal n Canada and the Unted States for the th ndustry. Thus, the relatve MFP between Canada and the Unted States s equal to the relatve labour productvty net of the relatve contrbutons attrbuted to the captal ntensty. Note, however, that the MFP measure s sgnfcantly dfferent from Lee and Tang (2001). Followng the Jorgensonan framework closely, they also control for the qualty of captal and labour nputs. In other words, they recognze that the margnal products of dfferent captal and labour are dfferent. So they treat M&E and structures, and unversty educaton and non-unversty educaton dfferently by assgnng dfferent weghts to them. 8 In short, ther MFP measure s net of qualty mprovements n captal and labour. In contrast, the MFP estmate n the current study ncludes the qualty mprovements The Framework for Prce Compettveness Comparsons Followng Jorgenson and Kuroda (1995), we measure prce compettveness by relatve value added prces, defned as value added PPPs dvded by the market exchange rate ($CDN per $US). Accordng to Lee and Tang (2001), the value added prce for a Canadan ndustry relatve to ts U.S. counterpart s postvely related to the relatve nput costs and negatvely related to the relatve MFP. A relatve nput cost s defned as the PPP of the nput dvded by the market exchange rate. The relatve MFP s the rato of MFP n Canada to that n the Unted States. If the relatve value added prce for the th ndustry s below one, then the ndustry s sad to be more compettve n Canada than n the Unted States, and vce versa Data for Productvty and Prce Compettveness Comparsons To mplement the above framework, we requre data on output, captal and labour by ndustry. In addton, we need data on factor compensaton. For Canada, the nomnal GDP at the basc prce s from CANSIM table , whch s converted to factor cost by peelng off the net 8 The weghts are factor compensaton, whch essentally assumes that the margnal product of a unt of factor s equal to the factor compensaton for the unt. 11

12 ndrect taxes on producton. Total hours worked s from CANSIM table Nonresdental captal stock, geometrc (nfnte) end-year net stock s extracted from CANSIM table And, fnally, the factor compensaton data are derved from the nput and output tables. All Canadan data are NAICS-based and converted nto the 31 ndustres n Table 1. For the Unted States, GDP at factor cost and factor compensaton data are from the U.S. Bureau of Economc Analyss (GPO87SIC.xls). Data on total hours worked for from Ho, Rao and Tang (2004), s extended to 2001 usng growth rate of full-tme equvalent employment from the U.S. Bureau of Economc Analyss (GPO87SIC.xls). Data on non-resdental captal stock are also from the U.S. Bureau of Economc Analyss. All U.S. data are SIC-based and converted nto the 31 ndustres n Table Productvty Levels n Canada Relatve to Those n the Unted States In ths secton, we dscuss the results on productvty level comparsons between Canadan and U.S. ndustres. We frst deal wth labour productvty level comparsons Labour Productvty Levels n Canada Relatve to Those n the Unted States We frst calculate labour productvty levels n Canada relatve to those n the Unted States for 26 9 ndustres n The results are presented n Table 4, and they are generally consstent wth the fndngs n our earler study (Rao, Tang, and Wang, 2003), wth a few exceptons. In 1999, Canada s labour productvty level n the busness sector was 82 percent of the U.S. level, an 18 percent gap between the two countres. The gap s smlar to the prevous estmate of 16 percent for the total economy n 1999 (Table 5). However, the new estmate of the labour productvty gap of 18 percent n the manufacturng sector s consderably smaller than the gap of 31 percent n Rao, Tang and Wang (2003). There are at least three man possble reasons for the dscrepancy between the two estmates. 10 Frst, there s no relable value added PPP for the total manufacturng sector n the earler study. Usng the labour productvty growth rates n the manufacturng sector n Canada and the Unted States, the earler study derved the labour productvty level gap for 1999 by extrapolatng the benchmark value for 1987 (de Jong, 1996). The benchmark estmate was calculated on the bass of the unt value ratos (UVRs) approach developed by the Internatonal Comparson of Output and Productvty (ICOP) project at the 9 Due to data lmtatons, computer and machnery are combned nto one ndustry, and communcaton equpment, other electronc product and electrcal product are combned nto another ndustry. We also drop mscellaneous manufacturng (23) and other servces (31) due to data qualty concerns. 10 There are other two possble reasons for causng the dscrepancy. Labour productvty n ths paper s defned as value added per hour worked, whle t s defned as value added per worker n the earler study. As we know, on average, Amercans work more hours than Canadans. For nstance, the average hours worked per employee n 1999 for the total manufacturng sector was 1,982 hours n Canada (Source: Statstcs Canada) and 2,055 hours n the Unted States (Sources: Ho, Rao and Tang, 2004 and the U.S. Bureau of Economc Analyss). As a result, the Canada-U.S. labour productvty level gap based on value added per hour worked s smaller. In addton, the base year n the current paper s 1999 whle our earler estmate s based on Ths may be another reason for the dfference snce the relatve labour productvty level s senstve to the base year. Note that the output n the manufacturng sector s aggregated over many dfferent commodtes/ndustres, whch s nfluenced by both quanttes and prces of the commodtes/ndustres n the manufacturng sector (Tang and Wang, 2004). 12

13 Unversty of Gronngen (van Ark, Inklaar and Tmmer, 2000). The smple extenson of the relatve level n 1987 to 1999 may not truly reflect the changes n the commodty composton over ths perod. Second, both the U.S. Bureau of Labor Statstcs and Statstcs Canada recently made major revsons to labour productvty data for the manufacturng sector. Fnally, PPPs derved from the expendture-based commodty PPPs, used n the current study, can be sgnfcantly dfferent from unt value ratos (UVRs) used by de Jong (1996), hence a sgnfcant dfference n the estmate of the labour productvty gap between the two countres. We beleve that the estmates n the current paper are more up-to-date and more relable. In 1999, Canada was less productve than the Unted States n 17 of the 26 ndustres, ncludng all the ndustres assocated wth the nformaton and communcaton technologes ndustres (machnery and computers, electronc and electrcal equpment, and nformaton and cultural ndustres), as shown n Table 4. The productvty gap s sgnfcantly larger n petroleum and coal products, fabrcated metals, textle and clothng, and FIRE (fnance, nsurance, real estate and rental and leasng). On the other hand, Canada s more productve n nne ndustres. Many of these are resource-based ndustres. The productvty advantage s greater n constructon, prmary metals, paper, and non-metallc mnerals. In machnery and computer, and electronc and electrcal equpment, Canada s labour productvty levels were 63 and 70 percent of the U.S. levels, respectvely. The estmates are sgnfcantly larger than the earler estmates of only about 30 percent (Table 5). We beleve the major dfference between the two studes s due to the fact that the value-added PPPs for the two ndustres n the current study are sgnfcantly lower than the PPPs used n Rao, Tang and Wang (2003). The change n the benchmark year as well as the revsons to labour productvty data n the two countres mght have also contrbuted to the dfference. We extended the level estmates n 1999 backward to 1997 and forward to 2001 usng the labour productvty growth rates n Canada and the Unted States for Both the level and growth estmates are reported n Table 6. Bascally, the table shows that n the last fve years, Canada has mproved ts labour productvty relatve to the Unted States n prmary ndustres and constructon, but lagged the Unted States n the manufacturng sector. Canada s poor performance n the manufacturng sector s manly due to the two ndustres assocated wth computers and electroncs. The machnery and computer ndustry n Canada saw ts productvty level relatve to the Unted States fall from 85 percent n 1997 to 63 n Smlarly, for electronc and electrcal equpment, the relatve productvty level fell from 66 to 44 percent. In addton, we extended the estmates n 1999 to for the aggregate manufacturng sector and the busness sector (Fgure 1). Clearly, the Canada-U.S. labour productvty gap wdened dramatcally snce the mddle of the 1990s, especally n the manufacturng sector. In 2003, Canada s labour productvty level n the manufacturng sector was 69 percent of the U.S. level, compared to 88 percent n Smlarly, durng ths perod, n the busness sector, the productvty gap ncreased by 8 percentage ponts MFP Levels n Canada Relatve to those n the Unted States. 11 Real GDP data are based on CANSIM table for Canada and the U.S. Bureau of Economc Analyss (GPO87SIC.xls) for the U.S. We cover only the perod of because the data for Canada only starts at 1997 (due to the adopton of the NAICS) and the data for U.S. are only avalable up to

14 Based on equaton (9), we calculated relatve MFP levels n 1999 n Canada and the Unted States for the 26 ndustres. The estmated relatve MFP levels by ndustry are also reported n Table 4. In 1999, Canada was less productve than the Unted States n the busness sector, and the gap was 14 percent. At the ndustry level, Canada was less productve than the Unted States n 18 of the 26 ndustres. In partcular, Canada was much less productve n petroleum and coal products, utlty, and fabrcated metal. On the other hand, n 1999, Canada was sgnfcantly more productve than the Unted States n other transportaton equpment, prmary metal, prntng, non-metallc, and constructon. On average, Canada s dong much better n MFP than n labour productvty vs-à-vs the Unted States. The MFP gap n 1999 was lower than the labour productvty gap n the two major sectors: manufacturng and servces. The dfference n the two gaps was partcularly sgnfcant n the manufacturng sector (18 percent for the labour productvty gap compared to 10 percent for the MFP gap). In 17 of the 26 ndustres, the MFP gap was lower than the labour productvty gap. The dfference n the two gaps s manly attrbutable to the lower captal ntensty 12, especally the M&E captal ntensty (Table 4). 13 The correlaton between the relatve labour productvty level and the relatve captal ntensty level was 0.55 n 1999, whch s hghly sgnfcant. The captal ntensty gap between the two countres explans about 25 percent of the labour productvty gap n the busness sector and more than 50 percent of the manufacturng sector labour productvty gap. 14 These are sgnfcantly hgher than the fndngs for 2000 n Rao, Tang and Wang (2003). As shown n Rao, Tang and Wang (2003), the M&E captal ntensty s lower n Canada than that n the Unted States because of a relatvely hgher cost of M&E nvestment n Canada due to the heavy relance on mported captal and the weak Canadan dollar. In 1999, the M&E captal ntensty n Canada was only 55 percent of the U.S. level, compared to 89 percent for the total captal ntensty Prce Compettveness between Canadan and U.S. Industres Ths secton assesses dfferences n prce compettveness between Canadan and U.S. manufacturng ndustres and relates them to ther MFP levels. The relatve prces for value added, captal and labour, and relatve MFP levels for 1999 are reported n Table 7. In 18 of the 26 ndustres, Canadan ndustres had a lower relatve prce for value added and hence were more prce compettve than ther U.S. counterparts. Recall that relatve value-added prce s postvely related to relatve nput prces and negatvely related to relatve MFP level. 12 Canada s captal ntensty s based on captal stock estmates from the Investment and Captal Stock Dvson of Statstcs Canada. The captal stocks are estmated usng the same declnng-balance rates as those n the U.S. 13 As shown n Rao, Tang and Wang (2003), M&E captal s lower because of the relatvely hgher cost of M&E nvestment whch reles on mports. In 1999, the captal ntensty (PPP based) n Canada was only 55 percent of the U.S. level for M&E compared to 89 percent of the level for total captal. 14 The contrbuton rates are calculated when the productvty gaps are n log dfference. 14

15 Wth respect to captal, Canada had a lower or equal captal prce for total nvestment n all ndustres. 15 The lower cost of total nvestment was manly because Canada had a substantally lower cost n structures nvestment. On the other hand, Canada had a slghtly hgher M&E nvestment prce n 11 ndustres, all n the manufacturng sector. In Table 7, we also report the relatve wage rate between Canadan and U.S. ndustres. 16 Unlke nvestment prce, all Canadan manufacturng ndustres had a sgnfcant advantage over ther U.S. counterparts n labour compensaton, on average about 66 percent of the U.S. level. The varaton n relatve wage rate across ndustres was very small. As a result of the dfferences n nput prces, the nput structures are dfferent n the two countres. Canadan ndustres are generally more labour-ntensve, whle ther U.S. counterparts tend to be more captal-ntensve. These fndngs are consstent wth the conclusons of Rao, Tang and Wang (2003). The correlaton coeffcent between relatve value-added prces and relatve MFP levels across all ndustres n 1999 s whle the correlaton coeffcents wth captal and labour prces are 0.36 and 0.15, respectvely. These correlatons ndcate that nter-ndustry varaton n relatve value-added prces across ndustres s more strongly nfluenced by nter-ndustry varaton n relatve MFP levels. We summarze the relatonshp between value-added prces and MFP levels by plottng relatve value-added prces aganst relatve MFP levels for 1999 across ndustres, as shown n Fgure 2. To better llustrate the relatonshp between prce compettveness and relatve MFP levels, we dvde the fgure nto four quadrants. Quandrants I and II show Canadan ndustres whch are less prce compettve than ther U.S. counterparts, whle quadrants III and IV show Canadan ndustres that are more prce compettve than ther U.S. equvalents. At the same tme, Canadan ndustres n quadrants II and III are more productve than ther U.S. counterparts, whle relatvely less productve ndustres n Canada are located n quadrants I and IV. In 1999, eght Canadan ndustres were less prce compettve and less productve than U.S. ndustres (quadrant I). They are agrculture, textle and clothng, wood products, plastc and rubber products, fabrcated metal, machnery and computers, electronc and electrcal equpments, and furnture and related products. As shown n Table 7, lower productvty s the reason for those ndustres beng less prce compettve. No ndustry was less prce compettve but more productve than ts U.S. counterpart (quadrant II). An examnaton of quadrant III reveals that 10 Canadan ndustres were more prce compettve and more productve than U.S. ndustres. Nne of these ndustres had relatvely lower nput prces as well as hgher MFP level than ther U.S. counterparts (Table 7). The last quadrant IV shows eght ndustres where Canada was more prce compettve but less productve than the Unted States (mnng, paper, petroleum and coal products, chemcals, utlty, retal trade, transportaton, nformaton and 15 Note that nvestment prces are very dfferent from the user cost of captal snce the latter s also nfluenced by other factors ncludng taxes (see Jorgenson and Yun, 2001). 16 Relatve wage rate s calculated as the rato of wage rate n Canada n Canadan dollars to the wage rate n the U.S. n U.S. dollars, dvded by the market exchange rate. In ths paper, the wage rate n each country s smply calculated as total labour compensaton dvded by total hours worked. Thus, the relatve wage rate s dfferent from the relatve prce for labour nput n Lee and Tang (2001) who control for qualty of hours worked. 15

16 cultural ndustres, and fnance, nsurance and real estate). Canada s compettve poston n these ndustres stemmed entrely from lower nput prces. 4. Conclusons The man objectve of ths study has been to develop new and up-to-date estmates for the Canada-U.S. labour productvty and MFP gaps at the ndustry level. Usng the methodology n Jorgenson and Kuroda (1995) and Lee and Tang (2001) and the 1999 expendture-based blateral commodty PPPs, we calculated blateral PPPs for gross output, ntermedate nput, two types of captal stocks and value added for 31 ndustres. These PPP estmates n turn were used to develop estmates for labour productvty and MFP gaps n all the ndustres. The benchmark gaps n conjuncton wth ndustry labour productvty growth rates n the two countres were used to develop a tme seres for the gaps durng the perod Ths paper shows that the aggregate Canada-U.S. labour productvty gap n 1999 was around 18 percent and the MFP gap was about 14 percent. In the manufacturng sector, the labour productvty gap was 18 percent and the MFP gap was about 10 percent. The labour productvty estmates n the manufacturng sector are sgnfcantly lower than the prevous estmates (Rao, Tang and Wang, 2003). The dfference n the aggregate manufacturng labour productvty gap between the current study and the earler study can be manly attrbuted to the dfferences n the estmates for PPPs for the two machnery ndustres. The new PPPs are sgnfcantly lower than the prevous estmates. Consequently, the labour productvty gaps n these two ndustres n ths study are consderably lower than the prevous estmates. More mportantly, ths study shows that dfferences n captal ntensty n the two countres explan about 30 percent of the aggregate Canada-U.S. labour productvty gap and more than 50 percent of the manufacturng sector labour productvty gap n 2001, whch are sgnfcantly hgher than those reported n Rao, Tang and Wang (2003). Ths paper shows that n 17 of the 26 ndustres, Canadan ndustres had lower labour productvty and 18 had lower MFP levels compared to ther U.S. counterparts n It also fnds that n 1999, 18 of the 26 Canadan ndustres had a lower relatve value-added prce and thus were more prce compettve than ther U.S. counterparts. MFP level has been an mportant determnant of Canada s nternatonal prce compettveness. Nne of the Canadan ndustres wth prce compettveness advantage relatve to the Unted States n 1999 were entrely due to lower nput prces n Canada, especally labour compensaton. On average, the wage rate n Canada was only about 66 percent of the U.S. level. These fndngs ndcate the mportance of mprovng Canada s relatve productvty for mprovng Canada s prce compettveness, and also the busness clmate n Canada. Canada needs to mantan hghly compettve market framework polces for retanng and attractng busness nvestment and nnovaton actvtes to Canada. A recent OECD (2003) study shows that by brngng down ts foregn drect nvestment (FDI) regulatons to U.K. levels whch are the lowest among the OECD countres, and reducng the product market regulatons to the OECD average, Canada could double ts FDI stock. Such an ncrease n FDI stock wll ncrease captal ntensty, expand trade, ncrease research and development (R&D) and ultmately mprove productvty and lvng standards n Canada (Rao and Tang, 2004a). The recent 16

17 apprecaton of the Canadan dollar n late 2003 and early 2004 would reduce the prce of captal relatve to labour, especally for M&E captal, whch n turn would stmulate nvestment and mprove productvty n Canada. Our study suggests that future research should do an n-depth analyss of the determnants of MFP and captal ntensty n Canada and the Unted States. In partcular, the role of taxes, nvestment ncentves, labour compensaton, avalablty of sklled labour, labour market and FDI regulatons need to be explored. Canada-U.S. border rsk n busness nvestment decsons n Canadan and U.S. ndustres needs also to be studed. 17

18 References Chrstensen, Laurts R., Danne Cummngs, and Dale W. Jorgenson, Relatve Productvty Levels, : An Internatonal Comparson, n Dale W. Jorgenson, Productvty: Internatonal Comparsons of Economc Growth, Vol. 2, Cambrdge: The MIT Press, pp Chrstensen, Laurts R., Dale W. Jorgenson, and Lawrence J. Lau, Conjugate Dualty and the Transcendental Logarthmc Producton Functon, Econometrca 39 (4), , Transcendental Logarthmc Producton Fronters, Revew of Economcs and Statstcs 55 (1), de Jong, Gjalt, Canada s postwar manufacturng performance: a comparson wth the Unted States, Research Memorandum GD-32, Gronngen Growth and Development Centre, Unversty of Gronngen. Ho, Mun S., Someshwar Rao, and Janmn Tang, Sources of Output Growth n Canadan and U.S. Industres n the Informaton Age, n Economc Growth n Canada and the Unted States n the Informaton Age, ed. Dale W. Jorgenson (Harvard Unversty), Industry Canada Research Monograph. Hooper, Peter and Elzabeth Vrankovch, Internatonal Comparsons of the Level of Unt Labour Costs n Manufacturng, Internatonal Fnance Dscusson Paper, no. 527, Board of Governors of the Federal Reserve System. Jorgenson, Dale W., Productvty and Economc Growth, n Dale W. Jorgenson, Productvty: Internatonal Comparsons of Economc Growth, Vol. 2, Cambrdge: The MIT Press, pp Jorgenson, D.W., Mun S. Ho, and Kevn Stroh, Growth of U.S. Industres and Investment n Informaton Technology and Hgher Educaton, mmeo, Harvard Unversty. Jorgenson, Dale W. and Masahro Kuroda, Productvty and Internatonal Compettveness n Japan and the Unted States, , n Dale W. Jorgenson, Productvty: Internatonal Comparsons of Economc Growth, Vol. 2, Cambrdge: The MIT Press, pp Jorgenson, Dale W., Masahro Kuroda, and Meko Nshmzu, Japan-U.S. Industry-Level Productvty Comparsons, , n Dale W. Jorgenson, Productvty: Internatonal Comparsons of Economc Growth, Vol. 2, Cambrdge: The MIT Press, pp Jorgenson, Dale W. and Meko Nshmzu, U.S. and Japanese Economc Growth, : An Internatonal Comparson, Economc Journal 88,

19 Jorgenson, Dale W. and Kun-Young Yun, Tax Reform and the Cost of Captal, Clarendon Press, Oxford. Lafortune, Jeanne and Frank C. Lee, The Canada-US Productvty Gap, Mmeo, Fnance Canada, Ottawa, Canada. Lee, Frank C. and Janmn Tang, Productvty Levels and Internatonal Compettveness Between Canada and the Unted States, n Industry-Level Productvty and Internatonal Compettveness Between Canada and the Unted States, ed. Dale W. Jorgenson and Frank Lee, Industry Canada Research Monograph. OECD, OECD Economc Outlook: Specal Focus on Foregn Drect Investment 2003/1, No. 73, June, (Pars). Plat, Drk, Labour Productvty Levels n OECD Countres: Estmates for Manufacturng and Selected Servce Sectors, Economcs Department Workng Papers, No. 169, OECD, Pars. Plat, Drk and Frank C. Lee, Productvty Growth n ICT-Producng and ICT-Usng Industres: A Source of Growth Dfferentals n the OECD, OECD STI Workng Papers, no. 2001/4. Rao, Someshwar and Janmn Tang, The Contrbuton of ICT to Productvty Growth n Canada and the Unted States n the 1990s, Internatonal Productvty Montor 3 (Fall 2001): Rao, Someshwar and Janmn Tang, 2004a. Foregn Ownershp and Total Factor Productvty, n Multnatonal and Publc Polcy, ed., Lorrane Eden and Wendy Dobson, Northampton: Edward Elgar (forthcomng). Rao, Someshwar and Janmn Tang, 2004b. Compettveness Challenges Facng Canadan Industres, Mmeo, Industry Canada. Rao, Someshwar, Janmn Tang and Wemn Wang, Canada s Recent Productvty Record and Captal Accumulaton, Internatonal Productvty Montor, No. 7, pp Tang, Janmn and Wemn Wang, Sources of Aggregate Labour Productvty n Canada and the Unted States, Canadan Journal of Economcs, 37(2), pp van Ark, Bart, Robert Inklaar and Marcel Tmmer, The Canada-U.S. Manufacturng Productvty Gap Revsted: New ICOP Results, Gronngen Growth and Development Centre, Unversty of Gronngen. 19

20 Table 1: Industry Classfcaton Industres NAICS Codes Prmary Industres 1 Agrculture (ncludng forestry, fshng and huntng) 11 2 Mnng 21 Constructon 3 Constructon 23 Manufacturng Industres 4 Food, beverage and tobacco Textle and clothng Wood product Paper Prntng and publshng* 323, 511, Petroleum and coal products Chemcals Plastc and rubber products Non-metallc mneral product Prmary metal Fabrcated metal Machnery Computer Communcatons equpment Other electronc equpments Electrcal equpments Motor vehcle Other transportaton equpment Furnture and related product Mscellaneous manufacturng 339 Servces Industres 24 Utlty Wholesale trade Retal trade Transportaton Informaton and cultural ndustres** 512, FIRE (Fnance, nsurance, real estate and rental and leasng) Busness servces Other servces * We do not have data to separate nformaton servces and data processng servces ndustry (NAICS 514) from publshng ndustry (NAICS 511) for Canada. So both ndustres are ncluded n prntng and publshng ndustry. ** It only ncludes Moton pcture and sound recordng ndustres (NAICS 512) and Broadcastng and telecommuncatons (NAICS 513) 20