Industry Canada Research Publications Program FOREIGN DIRECT INVESTMENT AND PRODUCTIVITY GROWTH: THE CANADIAN HOST-COUNTRY EXPERIENCE

Transcription

1 Industry Canada Research Publcatons Program FOREIGN DIRECT INVESTMENT AND PRODUCTIVITY GROWTH: THE CANADIAN HOST-COUNTRY EXPERIENCE Workng Paper Number 30 Aprl 1999

2 Industry Canada Research Publcatons Program The Industry Canada Research Publcatons Program provdes a forum for the analyss of key mcro-economc challenges n the Canadan economy and contrbutes to an nformed publc debate on these ssues. Under the drecton of the Mcro-Economc Polcy Analyss Branch, the Program s research paper seres features peer-revewed analytcal workng papers or polcy-related dscusson papers wrtten by specalsts on mcro-economc ssues of broad mportance. The vews expressed n these papers do not necessarly reflect the vews of Industry Canada or of the federal government.

3 Industry Canada Research Publcatons Program FOREIGN DIRECT INVESTMENT AND PRODUCTIVITY GROWTH: THE CANADIAN HOST-COUNTRY EXPERIENCE By Surendra Gera, Wulong Gu and Frank C. Lee Industry Canada Workng Paper Number 30 Aprl 1999 Auss dsponble en franças

4 Canadan Catalogung n Publcaton Data Gera, Surendra Foregn drect nvestment and productvty growth : the Canadan host-country experence (Workng paper) Text n Englsh and French on nverted pages. Ttle on added t.p.: Investssement étranger drect et crossance de la productvté. Includes bblographcal references. ISBN Cat. no. C21-24/ Investments, Foregn Canada. 2. Captal productvty Canada. 3. Labor productvty Canada. 4. Technologcal nnovatons Canada. I. Gu, Wulong, II. Lee, Frank C. (Frank Chung) III. Canada. Industry Canada. IV. Seres: Workng paper (Canada. Industry Canada) V. Ttle. HG4538.G '3'0971 C E The lst of ttles avalable n the Research Publcatons Program and detals on how to obtan copes can be found at the end of ths document. Abstracts of research volumes and papers publshed n Industry Canada s varous seres, and the full text of our quarterly newsletter, MICRO, are avalable on STRATEGIS, the Department's onlne busness nformaton ste, at Comments should be addressed to: Someshwar Rao Drector Strategc Investment Analyss Mcro-Economc Polcy Analyss Industry Canada 5th Floor, West Tower 235 Queen Street Ottawa, Ontaro K1A 0H5 Tel.: (613) Fax: (613) E-mal: rao.someshwar@c.gc.ca

5 ACKNOWLEDGEMENTS An earler verson of ths paper was presented to the Polcy Research: Creatng Lnkages Conference, Ottawa, October 1-2, 1998 and to the Canadan Economc Assocaton Meetngs, May 28-31, 1998, Ottawa. We would lke to thank Shamka Srmanne, Tm Sargent and two external revewers for ther comments.

6

7 TABLE OF CONTENTS ABSTRACT... I. INTRODUCTION...1 Prevous Emprcal Fndngs: FDI and Productvty n the Host-Country...1 Objectve of Our Study...2 II. EMPIRICAL FRAMEWORK...3 III. DATA AND TRENDS...5 IV. EMPIRICAL RESULTS...15 Parameter Estmates...15 Estmates of Factor-Prce Elastctes...15 The Effects of Inward FDI...17 Effects of Domestc R&D Spllovers...20 Effects of Internatonal R&D Spllovers...21 V. SUMMARY AND CONCLUSIONS...23 APPENDIX A: DATA CONSTRUCTION...25 APPENDIX B: TOTAL FACTOR PRODUCTIVITY DECOMPOSITION...27 NOTES...29 BIBLIOGRAPHY...31 INDUSTRY CANADA RESEARCH PUBLICATIONS...33

8

9 ABSTRACT In ths paper, we analyze the mpact of technology transfers and spllovers from nward foregn drect nvestment (FDI) on the producton cost and structure of Canadan ndustres. Specfcally, we: (1) estmate the effects of nward FDI on the cost of producton; and (2) examne the mpact of FDI on the structure of producton,.e., the effects on demand for factors such as captal, labour, ntermedate goods and R&D captal. In dong so, we control for domestc and nternatonal R&D spllovers. We fnd that nward FDI lowers producton cost and ncreases productvty n most Canadan ndustres, and that t alters the structure of producton as ndustres adjust ther demand for factor nputs. Our estmates show that nward FDI s based aganst the use of captal, labour and ntermedate goods and somewhat based toward the use of domestc R&D. We also fnd evdence of sgnfcant postve nternatonal R&D spllovers through trade flows. The results ndcate that nternatonal R&D spllovers are based aganst the use of physcal captal, labour and ntermedate goods, and based toward the use of domestc R&D captal. The relatonshp between domestc R&D and nternatonal R&D spllovers suggests that domestc frms must nvest n R&D to capture the benefts of R&D spllovers from abroad.

10

11 I. INTRODUCTION The nternatonalzaton of producton has been a major phenomenon n the modern day global economy. Ths s partly a result of the removal of many barrers to nternatonal captal movements and partly a result of the wdespread use of the nformaton technology whch has made t possble for frms to ntegrate ther worldwde producton and nvestment actvtes. Most sgnfcant has been the dramatc growth n foregn drect nvestment (FDI) along wth other transactons that support nternatonal producton, ncludng subcontractng, lcensng, franchsng and allances. Between 1970 and 1990, drect nvestment flows among OECD countres more than doubled as a share of GDP (from 0.5 to 1.2 percent). Smlar trends are observed n Canada. Over the past ten years, the stock of nward FDI more than doubled n Canada, reachng 22.6 percent of GDP n The ncreased mportance of FDI n the economy has renewed the debate on such polcy ssues as whether nward FDI affects employment, producton and export growth n the host country, and whether outward FDI substtutes for exports and reduces captal nvestment and employment n the home country. 1 In ths paper, we focus on the relatonshp between nward FDI and productvty n the host country. The exstng lterature on FDI ponts to two major channels through whch FDI mprove the effcency of producton n the host country: technology transfers and spllover benefts to domestc frms (Blomstr`m and Kokko, 1994; and Blomstr`m, 1991). Technology transfers from foregn parent frms can take place through the addton of more productve captal stock and the adopton of product and process nnovatons, and through the R&D and nnovatve actvtes of these frms n the host economy. The nflow of new technology and workng practces of foregn frms could create sgnfcant spllover benefts to domestc frms n the host country. These spllovers can occur n a number of ways. Frst, they can occur smply because of the publc good characterstcs of knowledge. Lke all other types of knowledge, the frm-specfc assets of multnatonal enterprses (MNEs) may escape from ther owner s control and leak nto competng domestc frms. Second, knowledge and technology could spllover from foregn frms to domestc frms through the tranng of labour and management, whch wll subsequently beneft local frms. Thrd, MNEs can stmulate mprovements n qualty and relablty of nputs by local supplers. Fourth, domestc frms can learn by watchng (Balasubramanyam et al., 1996). Fnally, nward FDI can also lead to ncreased competton and force less effcent domestc frms to nnovate or to ext. Prevous Emprcal Fndngs: FDI and Productvty n the Host-Country A number of emprcal studes have shown that nward FDI mproves the productvty performance of host countres through technology transfer and spllover benefts. Borenszten et al. (1994) found that FDI outflows from the OECD countres are an mportant vehcle of technology transfer for developng countres and they appear to have contrbuted postvely to GDP growth. Smlarly, Caves (1996) and Dunnng (1993) observed that FDI has contrbuted to productvty convergence among countres. Mansfeld and Romeo (1980) found that technologes transferred to afflates were consstently of later vntage than those sold to outsders. A Canadan study by McFetrdge (1987) also concluded that technology transfer lags tend to be shorter for ntra-frm transfers than for other types of transfers. Evdence also shows that there are sgnfcant spllover benefts from MNEs for competng local frms. Blomstr`m and Wolff (1994) found that productvty of domestc frms n Mexco ncreased more rapdly, and the productvty gap from foregn subsdares narrowed, the hgher the foregn subsdares ntal share and the larger the ntal productvty gap. Based on a panel of 1,270 German manufacturng frms over the perod , Bertschek (1995) found that nward FDI had a postve and sgnfcant

12 2 Introducton effect on product and process nnovatons of domestc frms. Smlarly, Mansfeld and Romeo (1980) found that nnovatve efforts by domestc frms n the Unted Kngdom were hastened n response to technology transfers from the U.S. multnatonals to ther subsdares n the Unted Kngdom. The Canadan evdence on spllover benefts of nward FDI s somewhat mxed. Globerman (1979) found that labour productvty across Canadan-owned plants was postvely related to the share of foregn ownershp of an ndustry n However, studes by Saunders (1980) and Bernhardt (1981) questoned the exstence of spllover benefts of FDI for domestc producers. Saunders (1980) found a negatve relatonshp between the labour productvty of Canadan ndustres relatve to U.S. ndustres and the share of foregn ownershp for a sample of 3-dgt ndustres n In contrast, Bernhardt (1981) found no relatonshp between the two varables n 1966 and a negatve correlaton n In a recent study, Barrell and Pan (1997) analyze the mpact of technology transfers and spllovers from foregn-owned frms on productvty growth. Ther estmates for West Germany show that a 1 percent ncrease n the real nward FDI stock rases techncal progress by 0.27 percent. For the Unted Kngdom the results provde an nterestng sectoral contrast. In the manufacturng sector, nward FDI s found to have a sgnfcant effect on techncal progress a 1 percent rse n the FDI stock rases techncal progress by 0.26 percent. FDI accounted for about 30 percent of U.K. manufacturng productvty growth snce By contrast, FDI s not found to contrbute sgnfcantly to productvty growth n the non-manufacturng sector where some two-thrds of nward FDI has taken place. For the authors, these fndngs suggest that the benefts of nward FDI seem to le n sectors (such as manufacturng) where domestc producers are at a comparatve dsadvantage and relatvely less productve. Objectve of Our Study Followng Barrell and Pan (1997), we analyze the mpact of technology transfers and spllovers from nward FDI on producton costs and the structure of Canadan ndustres. More specfcally: (1) we estmate the effects of nward FDI on the cost of producton and total factor productvty; and (2) we examne the mpact of FDI on the structure of producton,.e. the effects on demand for factor nputs such as captal, labour, ntermedate goods and R&D captal. Our study makes two emprcal contrbutons to the exstng lterature. Frst, n our analyss we apply a cost functon approach a rgorous framework of factor demand analyss. Ths allows us to trace the response of factor demands to FDI n the producton process whle nvestgatng ts mpact on producton costs. Second, our analyss controls for domestc and nternatonal R&D spllovers. These factors have been found to affect productvty growth n small open economes such as Canada (Coe and Helpman, 1995; Bernsten, 1994; and Gera, Gu and Lee, 1998). The rest of the paper s organzed as follows. In secton II, we descrbe the emprcal model used for our estmatons. In secton III, we present an overvew of the data and general trends. Secton IV, contans our estmaton results. Fnally, n secton V, we offer some conclusons.

13 II. EMPIRICAL FRAMEWORK The dualty theory of producton has establshed that there s a unque correspondence between the producton functon and the cost functon. In order to examne the underlyng producton technology, one can ether employ a producton functon or ts assocated cost functon. In ths paper, we use the cost functon approach 2. Our choce of the cost functon approach s manly motvated by our desre to understand the followng behavoral response: to what extent nward FDI affects the demand for factors of producton among Canadan ndustres. We assume that each ndustry produces one output usng four nputs: physcal captal, labour, ntermedate goods and R&D captal. The cost of producton s affected by nward FDI, domestc R&D spllovers and nternatonal R&D spllovers. Gven that a multnatonal s decson to nvest abroad s nfluenced by many factors ncludng the economc clmate of host and competng countres, we assume that nward FDI, along wth domestc and nternatonal R&D spllovers are determned exogenously. 3 Assumng also that the ndustry mnmzes ts total cost of producton, we can wrte a general form of the ndustry s cost functon as: (1) C( w, Y; t, Z), where w = { w }s a vector of factor prces (the rental prce of physcal captal, wage of labour nput, prce of ntermedate goods and rental prce of R&D captal); = K, L, M, R ndexes physcal captal, labour nput, ntermedate goods and R&D captal; Y s the ndustry output; t s an ndex of tme representng pure techncal change; Z = { Z k } s a vector of exogenous varables (FDI, domestc R&D spllovers, and nternatonal R&D spllovers) that affect the cost of producton of an ndustry; and k = FDI, DRD, FRD ndexes nward FDI captal stock, domestc R&D spllovers and nternatonal R&D spllovers. C = w x s the total cost of producton of the ndustry, equal to the sum of the rental prce of captal stock, the wage blls, the cost of ntermedate goods, and the rental prce of R&D captal stock; and x { x } s a vector of the quanttes of factor nputs comprsed of physcal captal stock, labour nput, ntermedate goods and R&D captal stock. Accordng to Shepherd s lemma, the dervatve of the cost functon wth respect to an nput prce yelds the derved demand for the nput: C w = x ( w, Y; t, Z) = K, L, M, R. In our emprcal formulaton, FDI affects ndustry producton n two ways. FDI shfts the cost of producton curve and t affects the structure of producton as the ndustry adjusts ts demand for factor nputs such as captal, labour, ntermedate goods and R&D captal stock. For our emprcal nvestgaton, we have chosen the transcendental logarthmc (translog) cost functon modfed to ncorporate the effects of exogenous varables Z on the total cost of producton. The translog cost functon can be vewed as a second order logarthmc approxmaton to an arbtrary

14 4 Emprcal Framework twce contnuously dfferentable cost functon (Dewert and Wales, 1987). Due to possble multcollnearty n our data, we assume constant returns to scale as n Nadr and Km (1996). 4 translog cost functon wth constant returns to scale technology can be wrtten as: The (3) ln C( w, Y; t, Z) lny = α + α ln w + α t + β ln Z o t k k k γ j ln w ln w j + γ t ln w t + θ k ln w ln Z j α t + β ln Z ln Z + β ln Z t, 2 tt 2 k l kl k l kt k k where, j = K, L, M, R ndex captal, labour, ntermedate goods and R&D captal stock and k, l = FDI, DRD, FRD ndex nward FDI captal stock, domestc R&D spllovers and nternatonal R&D spllovers. Partally dfferentatng the translog average cost functon (3) wth respect to the logarthm of factor prces and usng Shepherd s lemma (2), we obtan the followng set of cost-share equatons: k k (4) s lnc = = α + γ j ln w j + γ tt + θ k ln Z k, = K, L, M, R, ln w j k where s = w x C s the share of cost accounted for by factor. There are several parametrc restrctons mposed on the cost functon. The cost functon s lnearly homogeneous n factor prces. Ths mples the followng set of restrctons: (5) α = 1, γ j = 0, γ t = 0, θ k = 0. j Note that the same set of restrctons are mposed by the constrant that the cost shares of physcal captal, labour, ntermedate goods and R&D captal add up to one. The set of restrctons (5) mply that only three out of the four share equatons are lnearly ndependent. As a result, the share equaton for R&D captal stock s dropped from the system of estmaton equatons. The cross partal dervatves of the cost functon must be equal from Young s theorem n calculus. Ths mples the followng set of symmetry condtons. 5 (6) γ = γ, β = β. j j kl lk The fnal system of estmaton equatons conssts of the cost functon (3) and three share equatons (4) for captal, labour and ntermedate goods wth parameter restrctons (6). An error term s added to each equaton partly to reflect optmzaton errors. The share equaton for R&D captal stock s dropped from the system of estmaton equaton. To ensure that our parameter estmates are nvarant to whch share equaton s dropped, the teratve seemngly unrelated regresson (SUR) method s used. To estmate the cost functon and share equatons, we pooled together cross-ndustry tme-seres data. The pooled cross-sectonal tme-seres data generate more varaton, whch make t more lkely to dentfy the parameter estmates than f we were usng the tme seres data for a sngle ndustry. When

15 Emprcal Framework 5 usng tme seres data for a sngle ndustry, varables often move together generatng a serous multcollnearty problem. In order to control for nterndustry dfferences n the producton process, we ntroduce ndustry fxed effects n the share equatons and the cost equaton. That s, we allow the parameters α o and α ( = K, L, M, R ) to vary across ndustres. The own-prce and cross-prce elastctes of factor demands, ε j, can be computed drectly from the translog cost functon (3): (7) ε = γ s + s 1 ε = γ s + s, for j,, j = K, L, M, R. j j j The cost functon (3) allows us to test addtonal restrctons regardng the underlyng producton technology: neutralty of pure techncal change and untary partal elastctes of substtuton among all nputs (for a further dscusson, see Nadr and Schankerman, 1981). Techncal change s sad to be th-factor neutral (th-factor savng, or th-factor usng) f the cost share of the th factor s unchanged (lowered, or rased) n response to techncal change and s represented by γ t = 0 (γ t < 0 or γ t > 0 ). The neutralty of the techncal change can be tested by mposng the restrctons: (8) γ t = 0, = K, L, M, R. The cost functon exhbts untary partal elastctes of substtuton among all nputs f (9) γ j = 0, j and, j = K, L, M, R. Usng the cost functon (3), we can estmate the effects of FDI, domestc and nternatonal R&D spllovers on total cost. They are measured by the cost elastcty computed by takng the frst order dervatve of the cost functon (3): lnc (10) η ck = = β k + θ k ln w + β kl ln Zl + β ktt, k = FDI, DRD, FRD. ln Zk l The cost elastcty defned n (10) represents the percentage change n total cost resultng from a one percent ncrease n nward FDI stock, domestc or nternatonal R&D spllovers. The effects of FDI, domestc and nternatonal R&D spllovers on the structure of producton can be measured by the elastcty of factor demand: (11) η k ln x θ = = η ck + ln Z s k k, = K, L, M, R, k = FDI, DRD, FRD. A postve elastcty of nput demand wth respect to varable k ndcates that varable k s based toward the use of nput. A negatve elastcty suggests that varable k s based aganst the use of nput. 6

16

17 III. DATA AND TRENDS The data requred for our emprcal analyss nclude: quanttes of output; quanttes and prces of captal, labour, ntermedate goods and R&D captal; nward FDI captal stock; domestc R&D spllovers; and nternatonal R&D spllovers. Data on nward FDI stock are avalable for 13 SIC-C one-dgt ndustres (Standard Industral Classfcaton for Companes and Enterprses 1980). All remanng data are aggregated nto these 13 SIC-C one-dgt ndustres. 7 The perod of analyss s from 1973 to The quantty of output s gross output n 1986 constant dollars. The quantty of labour s measured by the total number of hours worked. Intermedate goods consst of energy, materals and servces. They are all obtaned from Statstcs Canada s KLEMS database that provdes ndustry data on total output (Y), physcal captal (K), labour (L), energy (E), materals (M) and servces (S). The captal stock s the net physcal captal stock based on geometrc deprecaton, obtaned from the Fxed Captal Stock and Flows data from Statstcs Canada. We use the begnnng-of-year net captal stock snce nvestment s assumed to be productve the followng year. The R&D stock s calculated usng the perpetual nventory method. Rental prces of physcal and R&D captal are calculated usng the formula n Appendx A. Data on the book value of the nward FDI stock are obtaned from Statstcs Canada. They are avalable for 13 SIC-C (1980) one-dgt ndustres from 1983 onwards. Before 1983, the data on the book value of nward FDI are based on an earler SIC (1970) classfcaton. They are converted to the SIC-C (1980) ndustry classfcaton usng the converson matrx n Statstcs Canada Catalogue no The real value of the nward FDI stock for an ndustry s calculated by dvdng the book value of the nward FDI stock by the nvestment deflator of that ndustry. Domestc and nternatonal R&D spllovers are calculated as R&D captal stock emboded n goods purchased domestcally and nternatonally. Essentally, we are measurng emboded R&D spllovers as opposed to dsemboded R&D spllovers. 8 More specfcally, domestc R&D spllovers receved by ndustry are constructed as: (16) DRD X x = j, j v where x R j j j R s the R&D captal stock of ndustry j, v j s the real value-added of ndustry j, X j s the amount of domestc ntermedate goods ndustry purchased from ndustry j. Domestc R&D spllovers defned n equaton (16) measure R&D emboded n goods purchased by an ndustry from all other ndustres. It s commonly referred to as domestc nterndustry emboded R&D spllovers. Internatonal R&D spllovers are assumed to be from other G-7 countres. Analogous to domestc R&D spllovers, nternatonal R&D spllovers receved by Canadan ndustry are measured as foregn R&D captal stock emboded n the ndustry mports and s constructed as: 9 (17) FRDS = M j jk j k jk x R α. jk v

18 8 Data and Trends In equaton (17), M j represents total mport of good j by ndustry. α jk s the share of total mports of good j accounted for by country k. x R jk s the R&D captal stock of ndustry j n country k. 10 v jk s the real value-added of ndustry j n country k. 11 Internatonal R&D spllovers, as measured n equaton (17), represent nternatonal nterndustry emboded R&D spllovers. We exclude R&D spllovers emboded n ntra-ndustry trade from our measurement of nternatonal R&D spllovers due to two consderatons. Frst, ntra-ndustry trade s manly comprsed of ntrafrm trade of the MNEs. Thus, nward FDI should capture nternatonal R&D spllovers emboded n ntrafrm trade or ntra-ndustry trade. Second, we measure domestc R&D spllovers as R&D emboded n domestc nterndustry trade. To be consstent wth our measure of domestc R&D spllovers, we have chosen to measure nternatonal R&D spllovers as R&D emboded n nterndustry mports. We now turn to the basc data for our regresson analyss (Tables 1 to 5). Tables 1 and 2 present the mean values and annual growth rates of some key varables for the regresson analyss ncludng: total cost, gross output, physcal captal stock, labour nput, ntermedate goods, R&D captal stock, nward FDI stock, and domestc and nternatonal R&D spllovers. These varables show consderable varaton across ndustres. Compared wth the non-servce ndustres, the servce ndustres n general had larger annual growth rates n gross output, physcal captal stock, employment, ntermedate goods and R&D captal stock over the perod. The stock of nward FDI ncreased n all ndustres except energy wth annualzed growth rates rangng from a low of percent n the energy ndustry to a hgh of 6.44 percent n the fnance and nsurance ndustry. The mean values of cost shares accounted for by captal, labour, ntermedate goods, and R&D captal are shown n Table 3. Intermedate goods account for the largest share of total cost n almost all ndustres, rangng from 31.8 percent n other servce ndustres to 73.5 percent n the transportaton equpment ndustry. Not surprsngly, the servce ndustres show a very dfferent pattern of cost structure than the non-servce ndustres. Compared wth the non-servce ndustres, the servce ndustres have relatvely large cost shares of labour and relatvely small shares of ntermedate goods. In all ndustres, the cost share of R&D captal s rather small; the electrcal and electronc products manufacturng ndustry has the hghest cost share of R&D captal (6.6 percent). Gven our focus on nward FDI, the followng two tables (Tables 4 and 5) examne closely the dstrbuton and relatve mportance of nward FDI among Canadan ndustres. Inward FDI s found to be concentrated n the energy ndustry and n the fnance and nsurance ndustry. However, the rato of FDI to gross output s the hghest n machnery and equpment (64 percent), followed by chemcals and chemcal products (38 percent).

19 Industry (SIC-C) Total cost Gross Output Table 1 Mean values of man varables by ndustry, (mllons of 1986 dollars) Physcal captal stock Labour (mllons of hours) Intermedate goods R&D captal stock Inward FDI stock Foregn R&D spllovers Domestc R&D spllovers 1. Food, beverage & 53, , , , , , , , tobacco 2. Wood & paper 31, , , , , , Energy 60, , , , , , , , Chemcals & chemcal 26, , , , , , products 5. Metallc mnerals & 37, , , , , , , , metal products 6. Machnery & 6, , , , , equpment 7. Transportaton 32, , , , , , , , equpment 8. Electrcal & electronc 12, , , , , , products 9. Constructon 64, , , , , , , , Transportaton & 51, , , , , , , , communcaton 11. Fnance & nsurance 47, , , , , Other ndustres 1 51, , , , , , , , Consumer goods & servces 73, , , , , , Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng.

20 Industry (SIC-C) Total cost Table 2 Annual growth rates of man varables by ndustry, (percentage) Gross Output Physcal captal stock Labour (mllons of hours) Intermedate goods R&D captal stock Inward FDI stock Foregn R&D spllovers 1. Food, beverage & tobacco 2. Wood & paper Energy Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment 7. Transportaton equpment 8. Electrcal & electronc products 9. Constructon Transportaton & communcaton 11. Fnance & nsurance Other ndustres Domestc R&D spllovers 13. Consumer goods & servces Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng.

21 Data and Trends 11 Industry (SIC-C) 1. Food, beverage & tobacco Table 3 Average cost shares of nputs by ndustry, (percentage) Physcal Labour Intermedate R&D captal goods captal Wood & paper Energy Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment Transportaton equpment 8. Electrcal & electronc products 9. Constructon Transportaton & communcaton 11. Fnance & nsurance Other ndustres Consumer goods & servces Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng.

22 12 Data and Trends Table 4 Dstrbuton of nward FDI by ndustry, 1973 and 1992 (percentage) Industry (SIC-C) Change 1. Food, beverage & tobacco 2. Wood & paper Energy Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment 7. Transportaton equpment 8. Electrcal & electronc products Constructon Transportaton & communcaton 11. Fnance & nsurance Other ndustres Consumer goods & servces Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng.

23 Data and Trends 13 Table 5 Sgnfcance of FDI by ndustry, 1973 and 1992 ( percentage of gross output) Industry (SIC-C) Change 1. Food, beverage & tobacco 2. Wood & paper Energy Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment 7. Transportaton equpment Electrcal & electronc products 9. Constructon Transportaton & communcaton Fnance & nsurance 12. Other ndustres Consumer goods & servces Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng.

24

25 IV. EMPIRICAL RESULTS As dscussed earler, the estmaton model conssts of the cost equaton (3) and three share equatons (4) for captal, labour and ntermedate goods. In the estmaton process, we also mpose parameter restrctons (5). 12 Regressons are performed on pooled cross-secton tme seres data for 13 ndustres over the perod Parameter Estmates Table 6 presents the parameter estmates of the model and ther asymptotc t-statstcs. The hgh R-square value suggests that the model fts the data qute well. The ndustry dummy varables are generally found to be statstcally sgnfcant. In our emprcal formulaton, the factor demands and the cost of producton are affected by pure techncal change. Our estmates show that the based techncal change coeffcents on captal, labour and ntermedate goods ( γ, γ, γ ) are not statstcally dfferent from zero. However, the coeffcent on Kt Lt Mt ( γ Kt + γ Lt + γ Mt R&D captal ( γ Rt = ) ) s postve and statstcally sgnfcant at the 5 percent level. These results mply that pure techncal change s essentally Hcks-neutral wth respect to physcal captal, labour and ntermedate goods, and s R&D captal-usng across Canadan ndustres. Furthermore, the estmated rate of pure techncal change ( ( α + t) ) s found to be negatve. 13 The jont hypothess of t α tt untary partal elastctes of substtuton among all nputs s strongly rejected. Thus, the Cobb-Douglas functonal form does not adequately represent the producton technology of Canadan ndustres a common fndng n the lterature. Estmates of Factor-Prce Elastctes To gve a concse descrpton of the producton structure, we calculated the own- and cross-prce elastctes of factor demand usng equaton (7). As an llustraton, we present the mean values of the prce elastctes over the perod for the electrcal and electronc products ndustry (Table 7). The estmated own-prce elastctes of demand (along the man dagonal) have the expected negatve sgn and are statstcally sgnfcant at the 5 percent level, wth the excepton of the own-prce elastcty of captal. The cross-prce elastcty estmates ndcate that ( ) both captal and labour are substtutes for ntermedate goods; () captal and labour are weak substtutes; and () R&D captal s a complement of physcal captal but a substtute for labour and ntermedate goods. The estmates of own- and cross-prce elastctes of factor demands for all other ndustres are smlar to those for the electrcal and electronc products ndustry. 14 An excepton s the sgn of own-prce elastcty of demand for R&D captal. The own-prce elastcty of demand for R&D captal for the remanng 12 ndustres s found to be postve. Ths surprsng result may be due to the fact that R&D captal only accounts for less than 2 percent of total cost n these ndustres. It can be argued that the rental prce of R&D captal s unlkely to be a major factor n determnng the frm s choce of R&D expendtures when R&D accounts for such a small share of total cost. 15

26 16 Emprcal Results Table 6 Parameter estmates (13 ndustres over the perod) Parameters Estmates t-statstcs Parameters Estmates t-statstcs α o α K α K α K α L α K α M α K α t α K γ KK α K γ LL α L γ MM α L γ KL α L γ KM α L γ LM α L θ KFDI α L θ KDRD α L θ KFRD α L θ LFDI α L θ LDRD α L θ LFRD α L θ XFDI α L θ XDRD α M θ XFRD α M γ Kt α M γ Lt α M γ Mt α M β DRD α M β FRD α M β FDI α M β FDIFDI α M β DRDDRD α M β FRDFRD α M β FDIDRD α M β FDIFRD α o β DRDFRD α o α tt α o β FDIt α o β DRDt α o β FRDt α o α K α o

27 Emprcal Results 17 Table 6 (cont d) Parameters Estmates t-statstcs Parameters Estmates t-statstcs α K α o α K α o α K α o α K α o α K α o Equaton R 2 Std. error Total cost Captal share Labour share Intermedate goods share Log lkelhood 3011 Table 7 Own- and cross-prce elastctes of factor demand n the electrcal and electronc products ndustry (mean values for the perod) 1 Captal Labour Intermedate goods Captal (-0.03) (-1.42) (1.83) Labour (-1.42) (-12.09) (13.10) Intermedate goods (1.83) R&D captal (-1.69) (13.10) (7.03) (-14.62) (13.90) R&D captal (-1.69) (7.03) (13.90) (-20.04) 1. Numbers n parentheses are t-statstcs. The Effects of Inward FDI Table 8 presents the mean values of estmated cost and factor demand elastctes of FDI for the perod These elastctes measure the mpact of FDI on the cost and structure of producton across Canadan ndustres. The estmated elastctes of captal, labour and ntermedate goods are generally negatve and statstcally sgnfcant. It shows that nward FDI s based aganst the use of captal, labour and ntermedate nputs n almost all Canadan ndustres. In other words, nward FDI lowers the ndustry s demand for captal, labour and ntermedate goods for a gven level of output. However, f we take nto account the output expanson (scale) effects of FDI, the overall employment of captal, labour and ntermedate nputs may well ncrease.

28 18 Emprcal Results The relatonshp between FDI and R&D captal s not found to be sgnfcant for most manufacturng ndustres. However, FDI appears to be based toward the use of R&D n most servces ndustres. For all ndustres, we fnd that nward FDI has a negatve and statstcally sgnfcant nfluence on the cost of producton. As one would expect, there s consderable varaton n the estmated cost elastctes across ndustres. The maxmum gans from nward FDI arse n the energy ndustry each 1 percent ncrease n the FDI stock reduces the total cost of producton by 0.5 percent n the long-run. By contrast, there does not appear to be any perceptble sgnfcant effect from nward FDI n ndustres such as machnery and equpment, electrcal and electronc products, and transportaton and communcaton. Fnally, n fnance and nsurance, whch has attracted most recent nward FDI flows, a 1 percent rse n the FDI stock s estmated to reduce the total cost of producton by 0.16 percent. Table 8 Elastctes of foregn drect nvestment 1 (mean values for the perod) Industry (SIC-C) Total cost Captal Labour Intermedate goods 1. Food, beverage & tobacco (-5.59) (-2.88) (-6.46) (-5.80) 2. Wood & paper (-4.42) (-1.80) (-5.15) (-4.68) 3. Energy (-6.67) 4. Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment 7. Transportaton equpment 8. Electrcal & electronc products (-4.86) (-6.32) (-1.53) (-6.64) (-0.18) 9. Constructon (-4.10) 10. Transportaton & communcaton (-1.08) 11. Fnance & nsurance (-2.44) 12. Other ndustres (-3.07) 13. Consumer goods & servces (-2.76) (-6.14) (-2.65) (-3.42) (0.26) (-1.64) (0.65) (-0.44) (-0.48) (-1.71) (-2.20) (-0.49) (-6.93) (-5.55) (-6.87) (-2.16) (-7.14) (-0.78) (-4.63) (-1.65) (-2.87) (-3.38) (-3.28) (-6.75) (-5.01) (-6.63) (-1.51) (-6.91) (-0.09) (-4.21) (-0.98) (-2.40) (-3.01) (-2.82) R&D (1.55) (1.31) (-0.93) (-0.74) (0.17) (0.82) (-1.95) (0.33) (1.87) (1.87) (1.81) (1.27) (1.88) 1. Numbers n parentheses are t-statstcs. 2. Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng.

29 Emprcal Results 19 Table 9 Contrbuton of FDI to TFP growth, (annualzed growth rates) Industry (SIC-C) TFP growth FDI contrbuton Cost elastcty of FDI Growth of FDI 1. Food, beverage & tobacco 2. Wood & paper Energy Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment Transportaton equpment 8. Electrcal & electronc products 9. Constructon Transportaton & communcaton 11. Fnance & nsurance Other ndustres Consumer goods & servces Average The contrbuton of FDI to TFP growth equals mnus the cost elastcty of FDI multpled by the growth of FDI. 2. Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng. Next, we dentfy the contrbuton of nward FDI to TFP growth across Canadan ndustres. The contrbuton s calculated by multplyng the cost elastcty estmates by the change n FDI stock. The results are presented n Table 9. The results show that nward FDI contrbuted to TFP growth across Canadan ndustres by an average of 0.5 percent per year over the perod. In the manufacturng sector, the effect of FDI on TFP growth was found to be the strongest n food, beverage and tobacco (1.2 percent per year), chemcals and chemcal products (0.8 percent), and transportaton equpment (1.1 percent). FDI also contrbuted sgnfcantly to TFP growth n fnance and nsurance (1 percent per year). Despte the sgnfcant contrbuton of nward FDI to TFP growth, TFP grew, on average, by a meagre 0.11 percent per year n Canadan ndustres over the perod. The factors behnd the slow productvty growth stll reman an engma. 16

30 20 Emprcal Results Effects of Domestc R&D Spllovers In our emprcal model, we also take nto account R&D spllovers (domestc and nternatonal) across Canadan ndustres. Ths secton dscusses the effects of domestc R&D spllovers on the total cost of producton and the demand for factor nputs. The results are presented n Table 10. Table 10 shows that the estmated elastctes of domestc R&D spllovers are postve but not statstcally sgnfcant at the 5 percent level n 7 out of 13 ndustres. Whle these results appear to be counter-ntutve, other researchers have found smlar results for Canadan ndustres. For example, Bernsten (1994) found that domestc R&D spllovers ncreased producton cost n a number of ndustres such as chemcal products, fabrcated metals, non-electrcal machnery, and non-metallc mneral products. 17 Table 10 Elastctes of domestc R&D spllovers 1 (mean values for the perod) Industry (SIC-C) Total cost Captal Labour Intermedate goods 1. Food, beverage & tobacco (2.58) (2.81) (0.05) (3.05) 2. Wood & paper (1.71) (2.32) (-0.23) (2.25) 3. Energy (2.30) 4. Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment 7. Transportaton equpment 8. Electrcal & electronc products (0.36) (2.16) (0.13) (2.30) (-1.58) 9. Constructon (2.72) 10. Transportaton & communcaton 11. Fnance & nsurance (2.43) (-1.14) 12. Other ndustres (1.51) 13. Consumer goods & servces (1.04) (2.56) (1.51) (2.59) (1.94) (2.61) (1.10) (2.91) (2.80) (0.12) (2.14) (2.06) (-0.78) (-1.83) (0.15) (-1.25) (-0.06) (-2.81) (1.86) (1.44) (-2.30) (0.77) (0.08) (2.78) (0.75) (2.69) (0.47) (2.71) (-1.37) (3.02) (2.93) (-0.72) (2.06) (1.85) R&D (-1.47) (-1.42) (-0.85) (-1.52) (-1.11) (-1.54) (-0.38) (-2.13) (-1.58) (-1.44) (-1.69) (-1.39) (-1.62) 1. Numbers n parentheses are t-statstcs. 2. Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng.

31 Emprcal Results 21 Table 11 Elastctes of nternatonal R&D spllovers 1 (mean values for the perod) Industry (SIC-C) Total cost Captal Labour Intermedate goods 1. Food, beverage & tobacco (-3.50) (-3.21) (-4.05) (-3.06) 2. Wood & paper (-2.70) (-2.73) (-3.17) (-2.09) 3. Energy (-3.99) 4. Chemcals & chemcal products 5. Metallc mnerals & metal products 6. Machnery & equpment 7. Transportaton equpment 8. Electrcal & electronc products (-3.05) (-3.86) (-1.84) (-4.04) (-0.65) 9. Constructon (-2.33) 10. Transportaton & communcaton (-1.35) 11. Fnance & nsurance (-1.14) 12. Other ndustres (-2.78) 13. Consumer goods & servces (-2.02) (-4.05) (-2.99) (-3.45) (-2.43) (-3.05) (-2.05) (-2.59) (-1.95) (-1.82) (-3.09) (-2.41) (-3.99) (-3.52) (-4.20) (-2.28) (-4.33) (-1.10) (-2.70) (-1.73) (-1.52) (-3.02) (-2.41) (-3.36) (-2.56) (-3.43) (-1.26) (-3.77) (-0.04) (-1.83) (-0.37) (-0.20) (-1.88) (-0.82) R&D (3.64) (3.61) (2.73) (2.71) (3.09) (2.94) (2.02) (0.67) (3.77) (3.77) (3.78) (3.42) (3.77) 1. Numbers n parentheses are t-statstcs. 2. Other ndustres nclude servces to busness; educaton, health & socal servces; accommodaton, restaurants & recreaton; and food retalng. Effects of Internatonal R&D Spllovers The spllover effects of nternatonal R&D on the producton cost and nput demand across Canadan ndustres are presented n Table 11. Internatonal R&D spllovers lower producton costs and hence ncrease TFP n all Canadan ndustres. These results confrm the prevous fndng that nternatonal R&D spllovers have a postve and sgnfcant effect on productvty growth n small open economes lke Canada (Bernsten, 1994; Coe and Helpman, 1995; Gera, Gu and Lee, 1998). The maxmum gans from nternatonal R&D spllovers arse n energy each 1 percent ncrease n nternatonal R&D spllovers reduces the total cost of producton by 0.3 percent. The elastcty estmates ndcate that nternatonal R&D spllovers are based aganst the use of physcal captal, labour, and ntermedate nputs across Canadan ndustres. However, t s nterestng to note that nternatonal R&D spllovers are based toward the use of domestc R&D captal n Canadan

32 22 Emprcal Results ndustres. 18 Ths result suggests that nternatonal technology transfers lead to hgher domestc R&D n Canadan ndustres. We nterpret ths result to mply that domestc ndustres must nvest n R&D to beneft from R&D orgnatng from foregn sources.

33 V. SUMMARY AND CONCLUSIONS Ths paper nvestgates the effects of FDI on the cost of producton and total factor productvty of Canadan ndustres. In addton, the paper examnes the mpact of FDI on the structure of producton,.e., the effects on demand for factor nputs such as captal, labour, ntermedate goods and R&D captal. Our major fndngs are as follows: Frst, nward FDI lowers producton cost and hence ncreases productvty n most Canadan ndustres. Second, nward FDI alters the structure of producton as ndustres adjust ther demand for factor nputs. Our results show that nward FDI s based aganst the use of captal, labour and ntermedate goods. In contrast, nward FDI s somewhat based toward the use of domestc R&D. Thrd, nternatonal R&D spllovers have a sgnfcant negatve mpact on producton costs across Canadan ndustres. Our results ndcate that nternatonal R&D spllovers through trade are based aganst the use of physcal captal, labour and ntermedate goods, and based toward the use of domestc R&D captal across Canadan ndustres. The relatonshp between domestc R&D and nternatonal R&D spllovers suggests that domestc frms must nvest n R&D to capture the benefts of R&D spllovers from abroad. In vew of our fndngs, nward FDI appears to act as an mportant channel for the dffuson of deas and nnovatons. Whle Canada s nward FDI stock has been ncreasng over the past decade, ts share of FDI n North Amerca and worldwde has been declnng snce the Free Trade Agreement (FTA). A key challenge for polcymakers s how to attract more FDI nto Canada.

34

35 APPENDIX A DATA CONSTRUCTION Rental Prce of Captal Stock As n Mohnen and Dagenas (1997), the rental prce of captal stock s measured as: (A.1) w k = pi ( γ + δ I )( 1 tci )( 1 ucz), ( 1 u ) c where p I s the nvestment deflator. γ s the annual nterest rate on prme busness loans from the Bank of Canada (CANSIM 2560, B14020). The captal deprecaton rateδ I s computed as a resdual from the data on gross fxed captal formaton and geometrc net captal stock: δ I ( t) = 1 ( K( t) I ( t)) K( t 1) where K( t) and K( t 1) are the end-of-year net captal stock n years t and t 1 and I( t) s the gross fxed captal formaton n year t. The nvestment tax credt tc I and the corporate ncome tax rate u c are taken from the varous ssues of Statstcs Canada Catalogue no (Corporaton Taxaton Statstcs) and the Canadan Master Tax Gude. The present value of captal cost allowance s computed as z = α ( α + γ ), where the geometrc deprecaton rateα s assumed to be 5 percent for non-resdental structures and 20 percent for machnery and equpment. R&D Captal Stock R&D captal stock s defned as the begnnng-of-year R&D stock and s calculated from real R&D expendtures usng the perpetual nventory formula: (A.2) x ( t) = ( 1 δ ) x ( t 1) + RD( t 1), R R R where the R&D captal deprecaton rateδ R s assumed to be 10 percent, a rate commonly used n the lterature (Mohnen and Dagenas, 1997); x R s the begnnng-of-year R&D captal stock; and RD s real R&D expendtures. The R&D captal stock n the base year s equal to: (A.3) x R ( 0) = RD( 0) g + δ R where g s the average growth rate of real R&D expendtures over the entre perod. The rental prce of the R&D captal stock s constructed as follows: (A.4) w R = [ ] pr ( γ + δ R) ( 1 uc )( 1 tcr ) ucd, 1 u c

36 26 Appendx A where p R s the R&D nvestment deflator. γ and u c are nterest rates and corporate ncome tax rates. The R&D nvestment tax credt tc R was frst ntroduced n 1977 and s taken from the varous ssues of Statstcs Canada Catalogue no (Corporaton Taxaton Statstcs) and the Canadan Master Tax Gude. An ncremental R&D nvestment tax credt was ntroduced n the early 1980s (1983, 1984, and 1985). The ncremental R&D credt allowed corporatons to transfer to ther nvestors a tax credt equal to 50 percent of desgnated R&D expendtures. For those years, the present value of ncremental R&D nvestment allowances s calculated as: (A.5) d = a R 1 3 = 1 1 t + r, 3( 1 ) where a R s the ncremental nvestment allowance rate and s equal to 50 percent (Bernsten, 1996). For all other years, d = 0.

37 APPENDIX B TOTAL FACTOR PRODUCTIVITY DECOMPOSITION The effects of FDI, domestc and nternatonal R&D spllovers on productvty growth can be analyzed usng the estmated cost functon. The Dvsa ndex of total factor productvty (TFP) growth s defned as: (B.1) T FP & = Y& s x&, where the dot ( ) ndcates the rate of change and s = w x C s the cost share of th factor nput. The cost functon s defned as: (B.2) C( w, Y; t, Z) = w x. Takng the logarthm of Equaton (B.2) and then dfferentatng wth respect to tme t yelds: (B.3) ln C ln C ln C ln C w& + Y& + + Z& k = s x& + ln w ln Y t k ln Z k s w&. equaton: Combnng (B.1) and (B.3) to substtute out s & x, we obtan the followng decomposton C (B.4) TFP & ln = ( 1 ηcy ) Y& ηck Z& k, t k where η = ( ln C ln Y ) s the cost elastcty of output, and η = ( ln C ln Z ) s the cost cy elastcty of FDI, domestc or nternatonal R&D spllovers. Accordng to equaton (B.4), TFP growth can be decomposed nto three components: () a scale effect gven by the frst term; () contrbutons of FDI, domestc and nternatonal R&D spllovers gven by the second term; and () contrbuton of pure technologcal change. Wth constant returns to scale producton as assumed n ths paper, the cost elastcty of output η cy equals 1 and the scale effect equals zero. k k