The sources of total factor productivity growth: Evidence from Canadian data

Transcription

1 The sources of oal facor produciviy growh: Evidence from Canadian daa By Kenneh Carlaw Universiy of Canerbury Sephen Kosempel Universiy of Guelph Deparmen of Economics and Finance Universiy of Guelph Discussion Paper This is an Acceped Manuscrip of an aricle published by Taylor & Francis in Economics of Innovaion and New Technology, 2004, Vol. 13(4), June, pp The final publicaion of his aricle is available a DOI: hp://dx.doi.org/ /

2 The Sources of Toal Facor Produciviy Growh: Evidence from Canadian Daa Final Version June 16, 2003 Kenneh Carlaw Deparmen of Economics Universiy of Canerbury Privae Bag 4800, Chrischurch, New Zealand Sephen Kosempel * Deparmen of Economics Universiy of Guelph Guelph, Onario, N1G 2W1, Canada * Corresponding auhor: Sephen Kosempel, Deparmen of Economics, Universiy of Guelph, Guelph, Onario, N1G 2W1, Canada; Fax: (519) ; Kosempel@uoguelph.ca

3 ABSTRACT A dynamic general equilibrium model is consruced and used o idenify sources of oal facor produciviy growh in Canada and o quanify heir imporance. The model also provides procedures for consrucing measures of echnological progress. We find ha periods of low produciviy growh correspond o periods of high growh in invesmen-specific echnology (IST) or high raes of echnology embodimen. For example, he growh rae of IST was relaively high beween 1974 and The higher growh rae of IST during his period should have increased he rae of produciviy growh by an esimaed 0.29 percenage poins, ceeris paribus. Ye, produciviy growh slowed. Why? Keywords: Invesmen-specific echnological change; oal facor produciviy; economic growh. JEL classificaion: D24; O33; O47 2

4 1. INTRODUCTION Like many oher economies, Canada experienced a produciviy slowdown saring in he early 1970s and persising hrough o he lae 1990 s. This presens an empirical puzzle, because here is significan empirical and anecdoal evidence ha echnological change seems o have increased in he same period ha he produciviy slowdown is recorded. This is especially rue when looking a he innovaions ha have occurred o informaion and communicaions echnology, roboics and microchip echnology. In his paper we will aemp o make progress owards resolving his puzzle. The analysis o be underaken involves isolaing he conribuions of invesmen-specific echnology (IST) and residual-neural echnology (RNT) o overall produciviy, measured as oal facor produciviy (TFP). IST, calculaed as he raio of he price of consumer goods o he price of machinery and equipmen, accouns for improvemens in he qualiy of capial. IST growh direcly capures he embodimen of echnological advancemens in new machinery and equipmen, such as innovaions in informaion and communicaions echnologies. In comparison, RNT growh capures changes in produciviy associaed wih he organizaion of capial and labor. I will also capure improvemens in he qualiy of labor, o he exen ha hose improvemens are no accouned for in he measure of he labor inpu. The reason ha he RNT-IST spli is ineresing, relaive o oher poenial produciviy composiion disaggregaions, is ha IST gives an indicaion of invesmen qualiy which is a proximae measure of echnological diffusion. Imporanly, IST increases were found o be negaively correlaed wih TFP and RNT. So his paricular spli reveals somehing abou he relaionship beween echnological diffusion and measured TFP. This has direc implicaions for 3

5 he inerpreaion of TFP. For example, Hulen (2000) and many ohers say, TFP is no a measure of echnological change. Our analysis is a way of confirming ha view. A recen IMF repor on Canada (see Dunaway e. al., 2000) revealed ha he average annual growh rae of IST was higher, and he growh rae of RNT lower, afer he produciviy slowdown began. We use TFP, measured as oupu per uni of combined labor and capial, o rack Canada s produciviy performance. Table 1 summarizes he measures of TFP, IST and RNT over seleced periods. Beween 1961 and 1996, produciviy grew a an average annual rae of 0.71 percen. However, beween 1961 and 1973 TFP grew a an average annual rae of 1.75 percen and hen slowed o an annual rae of 0.17 percen over he period. Alhough he rae of produciviy growh slowed in 1974, he 1970s also signified he beginning of an era of high IST growh. The average annual growh rae of IST increased from 1.29 percen over he period o 5.12 percen over he period. In comparison, he average annual growh rae of RNT declined from 1.80 percen over he period o percen over he period. 1 These observaions lead he reader o correcly conclude ha he increase in he growh rae of IST was no sufficienly large o compensae for he reducion in he growh rae of RNT, and herefore he rae of produciviy growh slowed. However, he daa displayed in Table 1 and in he IMF repor, do no provide precise indicaions of he imporance of IST and RNT as sources of produciviy growh in Canada. The curren paper improves our undersanding of he various sources of produciviy growh; by quanifying he individual conribuions of IST and RNT o he rae of TFP growh in Canada for seleced periods. 1 On he surface i appears ha we have experienced negaive echnology growh, somehing ha is implausible in a maximizing framework. Our index of inpu neural efficiency improvemen has changed negaively, bu his could reflec a rade off beween new embodied echnology oday and inpu neural efficiency growh in he fuure. 4

6 Table 1 Average Annual Growh Raes: Produciviy and Technology Toal Facor Produciviy 1.75% 0.17% 0.71% Invesmen-Specific Technology a 1.29% 5.12% 3.81% Residual-Neural Technology 1.80% -0.07% 0.57% a The ime pah of IST was racked using movemens in he relaive price of equipmen. Beween 1961 and 1996 he price of new equipmen relaive o he price of consumer non-durables and services declined by approximaely 4.0 per year, while he rae of invesmen in equipmen rose by 4.0 percen per year. This negaive co-movemen has been inerpreed in he lieraure as evidence of IST (see Greenwood e al. 1997, 2000 and Dunaway e. al. 2000). Time series for TFP and RNT were calculaed from Canadian daa using parameer values derived from a Real Business Cycle (RBC) model ha was calibraed o he Canadian economy. The effecs ha IST has on TFP growh are no ransparen, and his may explain why hose effecs have no been compleely esablished in he exising lieraure. For example, he relevan real business cycle (RBC) lieraure (see Greenwood, Hercowiz and Krusell 1997, 2000) has focused on measuring he effecs of IST and RNT on labor produciviy growh. The curren paper adds o his lieraure, by sudying IST, RNT and heir conribuions o an alernaive measure of produciviy growh, TFP. The papers by Greenwood e al. do no disinguish beween RNT and TFP. The conribuions of IST (or embodied echnological change) o TFP growh has, however, been sudied in he growh accouning lieraure. Hulen (1992) and Jorgenson (1966) have focused on he measuremen of he efficiency of he capial sock and he effecs of measuremen errors on produciviy esimaes. The auhors of hose papers argued ha qualiy change (or IST growh) is difficul o observe, and herefore may no be measured accuraely in he Naional Income and Produc Accouns (NIPA). Furhermore, he failure o adjus capial for qualiy 5

7 change has he effec of suppressing he qualiy erms ino he TFP residual. In oher words, due o he qualiy measuremen problem, some changes in he capial sock are overlooked, and herefore he effecs of hose changes on oupu are incorrecly assigned as an improvemen in TFP. In order o obain an esimae of he size of error associae wih he official capial sock esimaes, Hulen used qualiy-correced daa from Gordon (1990). Gordon found ha he official deflaors for producer durable equipmen oversae qualiy-correced inflaion in capial goods, and herefore undersae increases in capial inpu. There are a couple reasons why i would be inappropriae o use Gordon s daa for he curren sudy. Firs, Gordon s daa is for he U.S. economy and reflecs he composiion of U.S. capial invesmen expendiures. Second, and mos imporanly, he daa gahering agencies in he U.S. and Canada have improved heir procedures for esimaing capial goods prices and have revised heir old esimaes. The revised official esimaes of capial goods prices in Canada do no exhibi subsanial differences from Gordon s esimaes, a leas no over he period ha his sudy is focused. In he curren paper we assume ha capial invesmen is measured correcly in he NIPA. Noneheless, we will sill demonsrae ha IST made imporan conribuions o TFP during he period. In addiion o he errors in measuremen ha have already been discussed in he lieraure, IST has more direc effecs on TFP. Firs, IST growh improves produc qualiy, and herefore i enables more effecive unis of oupu o be produced per uni of capial and labor employed. Second, IST growh reduces he price of new capial relaive o he price of personal consumpion goods, and herefore i provides an incenive for economic agens o reallocae heir expendiures away from consumpion goods and owards capial goods. In oher words, 6

8 produciviy increases in par because of beer producion echnologies in he capial goods producing secor, and in par because more resources ge allocaed o he more producive secor. To preview he resuls, IST growh was found o accoun for approximaely 20 percen of he growh in TFP over he enire period, zero percen over he pre-1974 period and 100 percen over he pos-1974 period. RNT growh accouned for he remainder in each period. Given ha he average annual growh rae of IST was posiive during he period, i was surprising o find ha hese echnological advancemens had a negligible impac on produciviy growh. This finding will be explained laer in he paper. The resuls also sugges ha he increase in he growh rae of IST over he pos-1974 period produced an increase in he rae of produciviy growh of 0.29 percenage poins, ceeris paribus. However, his gain was more han offse by he reducion in he growh rae of RNT, which produced a decline in he rae of produciviy growh of 1.87 percenage poins, ceeris paribus. The ne effec was a fall in he rae of produciviy growh by 1.58 percenage poins. The remainder of he paper is organized as follows: The model o be sudied is consruced in Secion 2. The resuls are presened and discussed in Secion 3. Some problems associaed wih he use of TFP as a measure of echnological change are idenified in Secion 4. Concluding remarks are provided in Secion THE ECONOMIC ENVIRONMENT In his secion a model of he Canadian economy is described, which will help o idenify appropriae mehods for calculaing ime series esimaes of echnology and produciviy and for isolaing he conribuions of RNT and IST o TFP. The model described is a version of he RBC model consruced by Greenwood e al. (1997), and used by Greenwood e al. (1997, 2000) for 7

9 measuring he impac of IST and RNT on he rae of labor produciviy growh in he U.S. economy. This paricular version of he RBC model allows for new equipmen o become increasingly more efficien wih he passage of ime Households The represenaive household has preferences defined over sreams of consumpion (c ) and leisure (l ), given by E 0 U ( c, l ), 0 < β < 1. = 0 β (1) The household budge consrain is given by, c + i ) s, + pqie, ( 1 τ n ) w n + (1 τ k )( rs, s + re, e + τ. (2) Here w is he real wage rae, r s, is he real renal rae of srucures and r e, is he real renal rae of equipmen. Labor income is axed a a rae of τ n, and capial income a a rae of τ k. The household receives a lump-sum ransfer of τ from governmen. The household s ime endowmen is normalized o one, and herefore n denoes he fracion of ime he household devoes o work. The capial sock is composed of srucures, s, and equipmen, e. Invesmen expendiures are given by i s, for srucures and p q i e, for equipmen, where p measures he purchase price of new equipmen relaive o he price of he consumpion good and q indicaes he qualiy of he newes vinage of equipmen. 8

10 All variables in he household s budge consrain are measured in unis of he consumpion good. Thus, he household mus sacrifice one uni of consumpion per uni ha i invess in srucures, and p unis of consumpion per uni ha i invess in equipmen. In he daa, he relaive price series displays a downward rend, indicaing ha invesmen in equipmen is becoming relaively less expensive as ime passes. Essenially, efficiency gains also lead o cos savings for invesors. For his reason, he reciprocal of he relaive price provides a measure of he efficiency, or he qualiy (q ), of he newes vinage of equipmen. Greenwood e al. (1997, 2000) have also inerpreed movemens in he relaive price as evidence of IST. 2 The household s capial socks evolve according o: i = + ) s, and (3) s, s 1 (1 δ s q = + ) e. (4) ie, e 1 (1 δ e Here δ s and δ e denoe he raes of depreciaion of srucures and equipmen, respecively Firms Producion can be hough of as occurring in wo sages. Consumpion goods are produced in Sage 1, and capial goods in Sage 2. Firms produce consumpion goods by combining labor, srucures and equipmen according o a Cobb-Douglas producion funcion, 2 In his analysis all movemens in he relaive price are perceived as IST. We feel ha IST is he mos likely explanaion for he downward rend observed in his ime series. However, i is no inconceivable ha oher economic forces have some bearing on relaive price movemens. For example, since Canada is a ne imporer of capial goods and a ne exporer of raw maerials, erms of rade shocks may also affec he relaive price. 9

11 F αn α s αe ( N, S, E ; A ) = A N S E wih α α α + + = 1, (5) n s e where A denoes he sae of residual-neural echnology. Firms are compeiive and seek o maximize profis, max π = F( N, S, E : A ) w N r S r E. (6) { N, S, E } s, e, Consumpion goods are sold o households and used in several capaciies: (i) hey may be consumed; or (ii) hey are used in Sage 2 of he producion process, o produce srucures and equipmen. Each consumpion good can be used o produce one uni of srucures or 1/p unis of equipmen Governmen The governmen axes labor and capial income and hen rebaes he proceeds o households subjec o he following budge consrain: τ = τ w n + τ r s + r e ). (7) n k ( s, e, 2.4. Compeiive equilibrium c for he ypical, i,, q i, n, s, e A compeiive equilibrium consis of an allocaion { } s e, =0 household, an allocaion { }, S, E =0 such ha: w,, r,, r N for he ypical firm and a se of prices { } s e, =0 10

12 1. he household s allocaion solves he household s problem of maximizing (1) subjec o (2)-(4), aking prices as given; 2. he firm s allocaion solves he firm s problem of maximizing (6), aking prices as given; 3. he governmen s budge consrain (7) is saisfied; and 4. all markes clear: n = N, s = S, e = E, F( n, s, e ; A ) = c + is, + ie,. (8) 2.5. Measuremen issues The resource consrain (8) indicaes ha a household ha purchases qi e unis of new equipmen is only required o sacrifice i e unis of curren consumpion. Despie is appearance, he qualiy componen embodied in equipmen invesmen is no free. Households pay a price for each efficiency uni of equipmen (qi e ) ha hey purchase. However, when oupu is measured in unis of he consumpion good, qualiy improvemens lower he uni price ( p = 1 q) of new equipmen. A lower uni price offses he charge for higher qualiy. As a resul, qualiy improvemens do no appear in he oal cos of an invesmen in equipmen. When all ypes of expendiure and sources of income have been divided by he price of he consumpion good, so ha each iem has a consan uni of measuremen, he measure of oupu is as shown by equaion (8). In he daa, however, invesmen expendiures are no measured in erms of forgone consumpion. Insead a separae deflaor is applied o each caegory of expendiure, so ha each iem is measured in consan dollars, bu no necessarily in a common uni. In he daa, nominal oupu (p y y) is given by p y, y pc, c + ps, is, + pe, qie,, (9) 11

13 and real oupu (y) by y c + is, + qie,. (10) Here p x denoes he implici price deflaor for iem x y, c, i s, i ). Noice he esimae of oupu ( e obained from he daa (10) does no coincide wih he esimae obained via (8). Qualiy improvemens add o oal oupu (y), bu hey do no add o oal expendiures (F). The sandard measure of he rae of oal facor produciviy growh (ΔTFP) is aken from oupu as measured by equaion (10), so ha: ΔTFP = y! N! S! E! α n + α s + α e. (11) y N S E Our measure of he rae of residual-neural echnological change (ΔRNT) is calculaed from oupu as measured by equaion (8), so ha: ΔRNT = F! F N! S! E! α n + α s + α e. (12) N S E In boh equaions (11) and (12) he measure of inpus o producion are idenical. I is only he measure of oupu ha differs wih equaion (12) having he invesmen-specific improvemens in qualiy removed. Since he values of F and y coincide for an economy ha experiences no IST growh, here mus be a one-o-one relaionship beween changes in RNT and changes in TFP. 12

14 As such, he conribuion of IST growh o TFP growh can be calculaed as he difference beween (11) and (12), y! F! Δ TFP Δ RNT =. (13) y F 3. THE RESULTS A ime series for TFP growh and RNT growh were obained by inpuing Canadian daa hrough equaions (11) and (12), respecively. 3 A ime series for IST growh was calculaed as he rae of decline in he relaive price series ( Δ IST = P! / P). The conribuion of IST growh o TFP growh was isolaed using (13). Addiional deails on hese calculaions can be found in he Appendix. The individual conribuions of RNT growh and IST growh o he average annual rae of produciviy growh are displayed in Table 2. Based on he resuls displayed in Table 2, i is eviden ha here has been a considerable shif in he sources of produciviy growh. Beween 1961 and 1973, no produciviy growh could be explained by improvemens in invesmenspecific echnology. However, since he mid 1970s he conribuion of IST growh o produciviy growh increased significanly. A he end of our daa se, IST growh accouns for almos all of Canada s produciviy gains. 3 These calculaions require esimaes of he share parameers in he producion funcion. The values of he share parameers were deermined from a calibraed version of he model (see Carlaw and Kosempel, 2001). Labor s share parameer (α n ) is se o , and maches he raio of labor income o oal income in Canada. Capial s share parameers are se so ha in he seady sae he purchases of srucures accouns for 5.90 percen of oal expendiures and he purchases of equipmen 7.49 percen: i s F( ) = and F( ) = These values were obained from he daa, and imply = α and = s α. e i e 13

15 Table 2 Conribuions o he Rae of TFP Growh Amoun Conribued o he Rae of TFP Growh: (A) ΔTFP-ΔRNT (IST s conribuion) -0.05% 0.24% 0.14% (B) ΔRNT (RNT s conribuion) 1.80% -0.07% 0.57% (C) ΔTFP (A+B) 1.75% 0.17% 0.71% Percenage Conribuion o he Rae of TFP Growh: (D) Invesmen-Specific (A/C)*100% 0% 100% 20% (E) Residual-Neural (B/C)*100% 100% 0% 80% The resuls in Table 2 also sugges ha he increase in he growh rae of IST over he pos period produced an increase in he rae of produciviy growh of 0.29 percenage poins, ceeris paribus. However, his gain was more han offse by he reducion in he growh rae of RNT, which produced a decline in he rae of produciviy growh of 1.87 percenage poins, ceeris paribus. The ne effec was a fall in he rae of produciviy growh by 1.58 percenage poins. A comparison of Tables 1 and 2 reveal ha he conribuion of IST growh o TFP growh is small relaive o he acual growh rae of IST. This is because invesmen in equipmen accouns for only a small fracion of oal oupu. Noice also ha IST growh had a negligible impac on he growh rae of produciviy during he pre-1974 period, despie he fac ha he growh rae of IST was posiive during his period. This resul can be explained by sudying equaion (13) in more deail. Recall, 14

16 Δ TFP Δ RNT = y! F! y F c c! i s i! s qie i! e q! c c! is i! s ie i! e = y c y is y ie q F c F is F ie 1 1 c$ i$ $ s q 1 ie qie q$ = c is ie y F + c i + + s y F. (13ʹ ) ie y q #% %% "%%%! #%"%! % #"! ( i) ( ii) ( iii ) For convenience ime subscrips have been dropped. All daa obained for he variables in (13ʹ ) were convered o 1992 dollars (see he daa appendix). By selecing 1992 as he base year for prices, we are also implicily assigning a value of one o he level of qualiy in ha year. As a resul, q < 1 and y < F over he enire period. Therefore, during he pre-1974 period, erm (i) in equaion (14) is posiive and erm (ii) negaive. Term (iii) is always posiive. Wheher or no invesmen-specific echnology lowers or raises TFP growh depends on he size of erms (i) and (iii) relaive o he size of erm (ii). Along he model s balanced growh pah i is easy o show ha erms (i) and (ii) cancel ou leaving only erm (iii). Thus, along he balanced growh pah IST affecs TFP growh posiively. However, during he period, boh he level and growh rae of invesmen-specific echnology were below heir balanced growh levels. As a resul, erm (ii) was negaive and large (in absolue value) and erm (iii) posiive bu small. The ne effec was ha IST had virually no impac on he rae of TFP growh during his period. The resuls above are sensiive o he base year ha is chosen for prices. Each ime Saisics Canada adjuss he base year for is price deflaors, i also changes he base year for qualiy. As a resul, he curren level of qualiy (or he efficiency of capial) will never deviae far from one. This is rue despie he fac ha new capial goods become more efficien as ime passes. Essenially, insead of raising he qualiy of new capial, he curren procedure is o lower he 15

17 qualiy of old capial. Furhermore, each ime he level of qualiy is scaled down, he weighs aached o he various growh raes in (14) change, and herefore he conribuion of IST o TFP growh also changes. Since Saisics Canada is currenly providing is real values in 1992 dollars, 1992 was seleced as he base year for his sudy. Thus, he resuls in his sudy are inerpreed as esimaes of he conribuions of RNT and IST o he rae of TFP growh when values are measured in 1992 equivalen unis. 4. MEASURES OF TECHNOLOGICAL CHANGE Figure 1 reveals some ineresing correlaions beween he growh raes of TFP, residualneural echnology and invesmen-specific echnology. In periods of TFP slowdown, he growh rae of IST is increasing and vice versa, while he growh rae of RNT racks he growh rae of TFP posiively. Possibly here is a relaionship beween IST and RNT ha has no been considered in his paper. For example, if i akes ime for workers o learn how o operae new high echnology capial goods efficienly, hen following he inroducion of advanced echnologies he growh raes of RNT and TFP will be low. {Inser Figure 1 here} Is TFP is a good indicaor of he effecs of echnological progress? 4 The produciviy slowdown puzzle is only a mysery if one expecs ha echnological change o resul in immediae economic pay off in erms of more oupu per uni inpu. However, if echnological 4 Noe ha Hulen (2000) and Carlaw and Lipsey (2003) argue ha TFP is no a measure of echnological change. So discovering a divergence beween TFP growh and changes in invesmen qualiy is no surprising from heir view. 16

18 change involves a process of embodimen in erms of insalling new equipmen, wriing off obsolee equipmen, reorganizing producion and invesing in new skills eiher direcly or hrough learning by doing and using processes; hen we should expec ha oupu per uni inpu may no have a posiive relaionship wih echnological change. In fac, he cos of echnological change should be all up fron wih he economic benefis spread ou in o he fuure and produciviy gains realized only afer all of he complemenary invesmens made. Thus, here may no be a produciviy puzzle. The resul of he modeling and measuremen exercise sugges ha if one wishes o observe wheher echnological change is occurring, one ough o seek o measure echnological change. No curren period oupu per uni inpu. By measuring improvemens in he qualiy of invesmen goods relaive o non-durable consumpion goods direcly, we find evidence ha new and higher qualiy echnology is being pu ino use in he period where he anecdoal evidence suggess rapid echnological change is occurring. No doub IST growh is no a complee measure of echnological change wihin he Canadian economy, bu our findings do show ha where an independen measure of echnological change can be employed we observe a negaive correlaion beween i and TFP. I was discovered ha periods of low produciviy growh or produciviy slowdown correspond o periods of high IST growh or high raes of echnology embodimen. 5 These resuls provide some suppor for he view ha, wha ever else i measures, oal facor produciviy is no a measure of echnological change. 6 5 This resul is consisen wih he findings of Paul David (1991), Aghion and Howi (1998), Lipsey, Bekar and Carlaw (1998), and Bekar (1999), who argue ha in periods of rapid echnological developmen, such as occur wih he inroducion of new general purpose echnologies, invesmen in new echnologies is all fron end loaded and he payoff o such invesmen occurs in he fuure as he echnology maures and learning by doing occurs. 6 Griliches (1994) suggess ha TFP is a measure of ignorance. Jorgenson (1995) argues ha TFP only measures spillovers from knowledge and no echnological change direcly. 17

19 5. CONCLUSIONS The Canadian NIPA daa appear o presen an empirical puzzle, a produciviy slowdown in a period of rapid echnological change. We have idenified a possible answer o his puzzle ha is consisen wih argumens posied by oher researchers. We have exended he findings in he recen real business cycle lieraure and research done by he IMF ha idenifies residual-neural and invesmen-specific echnological change as possible sources of labor produciviy growh. The curren paper has conribued o his lieraure by quanifying he conribuions of hose echnologies o he rae of TFP growh in Canada over seleced periods. The resuls are as follows: IST growh was found o accoun for approximaely 20 percen of he growh in TFP over he enire period, zero percen over he pre-1974 period and 100 percen over he pos-1974 period. RNT growh accouned for he remainder in each period. The increase observed in he growh rae of IST over he pos-1974 period produced an increase in he rae of produciviy growh of 0.29 percenage poins, ceeris paribus. However, his gain was more han offse by he reducion observed in he growh rae of RNT, which produced a decline in he rae of produciviy growh of 1.87 percenage poins, ceeris paribus. The ne effec was a fall in he rae of produciviy growh by 1.58 percenage poins. We noe ha alhough he average annual rae of IST increased by 3.83 percenage poins afer 1974, his change had only a sligh impac on he rae of produciviy growh. This is because invesmen in equipmen represens only a small fracion of oal oupu. However, he rae of invesmen in equipmen was found o display a posiive rend. If Canadians coninue o acquire relaively more of he goods ha embody advanced echnologies, hen in he fuure produciviy growh will become more sensiive o IST, and herefore he rae of produciviy growh should increase. 18

20 The rae of produciviy growh may also improve as people learn how o operae high echnology capial goods more efficienly. The process of learning by doing should increase he growh rae of RNT, and subsequenly affec TFP. A poenial avenue for furher research lies in idenifying he relaionship beween IST and RNT. Alhough TFP growh is commonly held o be a measure of he rae of echnological progress in an economy, he daa and our analysis seem o sugges ha TFP may no be he appropriae measure of echnological progress. Our esimaes show IST increasing in he period when rapid echnological change occurs and ye TFP is declining. Perhaps IST is a beer way o rack echnological progress wihin an economy. 19

21 APPENDIX: DATA DEFFINITIONS AND SOURCES A descripion of he daa and macroeconomic variables used in his sudy is provided below. All expendiures and facor inpus were deflaed by he 16+ populaion. All dollar values are based on 1992 prices. Daa is from he Canadian Socio-economic Informaion and Managemen Sysem Daabase (CANSIM). Oupu (y): real gross domesic produc (D22467) ne of gross housing produc (measured by gross impued rens (D22923) and gross paid rens (D22924)). Consumpion (c): real expendiures on consumer semi-durables (D22439), non-durables (D22440) and services (D22441). Invesmen in Srucures (i s ): real business gross fixed capial formaion in srucures (D22449). Invesmen in Equipmen (qi e ): real business gross fixed capial formaion in machinery and equipmen (D22450). Toal Expendiures (F): calculaed as oupu (y) less invesmen in equipmen (qi e ) plus expendiures on equipmen (i e ). Here expendiures on equipmen are measured in erms of forgone consumpion, ha is, hey have been deflaed by he implici price deflaor for consumer non-durables, semi-durables and services. Hours Worked (n): oal hours a work in non-agriculural indusries. Calculaed from oal hours 20

22 a work (I190301) less hours worked in agriculure (I190302). Capial Inpus (s and e): Time series for he capial socks were esimaed by ieraing on heir respecive laws of moion. In hese calculaions he raes of capial depreciaion were se equal o he values used in he relevan lieraure): δ and δ = (see Greenwood e al. 1997, s = e 2000 or Dunaway e al. 2000). The iniial values for he capial socks were se a heir balanced growh levels. This implies ha he flows and socks of capial experienced he same average growh raes over he enire period. Relaive Price (p): he raio of he implici price deflaor for real business gross fixed capial formaion in machinery and equipmen o he implici price deflaor for consumer semi-durables, non-durables and services. Invesmen-Specific Technology (q): he reciprocal of he relaive price (base = 1992). The Rae of Residual-Neural Technological Change: calculaed as he growh rae of oal expendiures (F) less he share weighed growh raes of he labor and capial inpus. The Rae of Toal Facor Produciviy Growh: calculaed as he growh rae of oupu (y) less he share weighed growh raes of he labor and capial inpus. Recall, oupu is measured in effecive unis, and herefore i includes qualiy componens. Expendiures are measured in erms of forgone consumpion, and herefore hey exclude qualiy componens. Hence, here is a difference beween RNT and TFP. 21

23 16+ Populaion: Toal populaion (D for pre-1971 and C for pos-1971) less populaion under age 15 (D for pre-1971 and C for pos-1971). REFERENCES Aghion, P., and P. Howi. (1998) On he Macroeconomic Effecs of Major Technological Change. In E. Helpman (Ed.), General Purpose Technologies and Economic Growh. Cambridge, Mass.: MIT Press, Bekar, C. (1999) Two Produciviy Puzzles in Briish Economic Hisory. Ph.D. Thesis, Simon Fraser Universiy. Carlaw, K.I., and R.G. Lipsey. (2003) Produciviy, Technology and Economic Growh: Wha is he relaionship? (forhcoming) Journal of Economic Surveys, 17(3). Carlaw, K., and S. Kosempel. (2001) Accouning for Canada's Economic Growh, Discussion Paper No , Universiy of Guelph. David, P. (1991) Compuer and Dynamo: The Modern Produciviy Paradox in a No Too Disan Mirror. In Technology and Produciviy: The Challenge for Economic Policy. Paris: OECD. Dunaway, S., e al. (2000) Canada: Seleced Issues. IMF Saff Counry Repor No. 00/34, Gordon, R. (1990). The Measuremen of Durable Goods Prices. Chicago: Universiy of Chicago Press. Greenwood, J., Z. Hercowiz and P. Krusell. (1997) Long Run Implicaions of Invesmen- Specific Technological Change. American Economic Review, 87(3), (2000) The Role of Invesmen-Specific Technological Change in he Business Cycle. European Economic Review, 44(1),

24 Griliches, Z. (1994) Produciviy, R&D, and he Daa Consrain. American Economic Review, 84(1), Hulen, C.R. (2000) Toal Facor Produciviy: A Shor Biography. NBER working paper Hulen, C.R. (1992) Growh Accouning When Technological Change is Embodied in Capial. American Economic Review, 82(4), Jorgenson, D.W. (1966) The Embodimen Hypohesis. Journal of Poliical Economy, 74(1), (1995) Produciviy, Volume 1: Poswar U.S. Economic Growh. Cambridge, Mass.: MIT Press. Lispey, R.G., C. Bekar and K.I. Carlaw. (1998) The Consequences of Changes in GPTs. In E. Helpman (Ed.), General Purpose Technologies and Economic Growh. Cambridge, Mass.: MIT Press,

25 0.2 TFP growh 0.15 RNT growh TFP growh - RNT growh 0.1 IST growh FIGURE 1. Produciviy Growh Raes in Canada,