Inward investment, transactions linkages, and productivity spillovers.

Transcription

1 Inward investment, transactions inkages, and productivity spiovers. Nige Driffied, Max Munday* Annette Roberts Abstract The paper examines the extent to which foreign manufacturing firms in the UK promote productivity growth in the domesticay-owned manufacturing sector through their buying and suppying reationships. The paper reveas evidence for intra- and inter-regiona externaities from the presence of foreign manufacturing, and intra- and inter-industry effects. Externaities in the domestic sector are most noticeabe where foreign manufacturing ses to domestic manufacturing. These externaities are, however, not whoy robust to different specifications of spatia dependence. The findings are positioned in a debate which has tended to view backward (as opposed to forward) inkages from mutinationas to domesticay-owned suppy bases as a critica driver of indirect economic benefits. Keywords: Productivity spiovers; Foreign manufacturing; Input-Output tabes; Spatia dependence. JEL: C67, L14, O33, R58 Nige Driffied, Birmingham Business Schoo, Birmingham University, Edgbaston, Birmingham. B15 2TT. Te Fax Emai: N.L.Driffied@bham.ac.uk *Address for correspondence Max Munday*& Annette Roberts Wesh Economy Research Unit & ESRC Centre for Business Reationships Accountabiity Sustainabiity and Society, Cardiff Business Schoo, Coum Drive, Cardiff CF10 3EU, United Kingdom. Te Fax Emai MundayMC@cf.ac.uk RobertsA1@cf.ac.uk June 2004 Acknowedgements: The Authors woud ike to thank three anonymous referees and the journa editor for their hepfu comments on earier versions of this paper. We woud aso ike to thank Professor John Dewhurst and Dr Macom Beynon for comments on earier drafts.

2 1. Introduction There has been much debate in recent years over the scae and scope of productivity spiovers that are generated by foreign manufacturing investment. One transmission mechanism for such spiovers is expected to be the extent of the buyer-suppier reationships between foreign manufacturers and their domesticay-owned counterparts. This paper expores the extent to which transactions inkages between foreign and domestic firms impact on productivity growth in domestic manufacturing. This is potentiay the most beneficia effect foreign capita on industria and regiona deveopment within the host country. Extensive poicy resources in the UK and esewhere have been expended on seeking to foster inkages between foreign manufacturers and oca (within-region) suppy and services bases. Such inkages are seen as vita for generating the indirect mutipier impacts expected from foreign firms in their host regions, as we as potentiay the most ikey vehice for technoogy transfer. There have, however, been few empirica attempts to assess the intra- and inter-industry productivity impications of these transaction inkages. Moreover, and a centra contribution of this paper, is the fact that previousy there has been ony imited attention paid to the productivity externaities generated by foreign firms saes to domestic industries (i.e. in forwardy inked sectors). This paper then assesses whether the productivity impacts of foreign manufacturing are greater where domesticay-owned firms buy from, or se to foreign manufacturing firms. The paper aso adopts an inter-regiona focus, examining whether production externaities from the foreign sector are restricted to host regions, or whether there is equa evidence of externaities in regions beyond the host region. The second section of the paper examines the iterature on productivity spiovers resuting from the presence of, what is assumed to be, reativey productive foreign 2

3 manufacturers. The review adopts a UK focus, and goes on to demonstrate the potentia roe of buyer-suppier inkages as a transmission mechanism for spiovers. The third section deveops some hypotheses on how transactions inkages (whether backwards or forwards) from the mutinationa enterprise might affect the productivity of domestic manufacturers. The fourth section describes how productivity externaities from the foreign sector might be assessed and measured, outines the data sources used in the anaysis, discusses the construction of variabes and a mode to test the hypotheses. The fifth section describes mode estimation, and provide detais of the main resuts. The fina section concudes. 2. Foreign Manufacturing, Buyer-Suppier Linkages, and Productivity in the Domestic Sector Connected to issues surrounding the contribution to domestic productivity growth made by manufacturing capita, is work both appied and theoretica, seeking to examine the roe of mutinationas in the host country s productivity growth, and roe in improving aocative efficiency (Barre and Pain, 1997). Theoretica perspectives on mutinationa enterprise generay suggest that foreign firms are potentia agents of technoogica diffusion, having a series of ownership advantages over domestic firms (Dunning 1993). The mutinationa firms origina stock of advantages may spiover or be appropriated by indigenous competing or suppier firms (Caves 1996; Markusen and Venabes 1999). This type of appropriation is consistent with endogenous growth theory where non-internaised technoogica change and deveopment from one industry can become an externaity which is captured by other industries and individuas (Fingeton 2001). 3

4 Productivity spiovers from foreign-owned to domestic firms can arise in a number of ways. Bomstrom and Kokku (1996) demonstrate that mutinationa enterprises might improve aocative efficiency as they enter sectors characterised by high entry barriers, and that they can then break down domestic monopoies and improve competition. In a simiar vein, a significant reationship between comparative advantage of UK manufacturing industries and new foreign direct investment into those industries was found by Driffied and Munday (2000). This, and simiar studies, have demonstrated the more impicit, dynamic effects of foreign direct investment spiovers into the indigenous manufacturing sector of the UK economy. A range of studies have suggested that buyer-suppier partnerships invoving foreign firms are a mechanism for productivity spiovers, technoogy diffusion (Morris et a. 1993; Gorg and Ruane 1998), and more fundamenta vaue chain restructuring (O huaachain and Wasserman 1999). In a wider review, Crone and Roper (1999) examined the specific iterature on knowedge transfers from mutinationas, and concuded that the suppy chain is the main route through which knowedge is transferred from mutinationa pants to indigenous firms, and that such transfers ead to important improvements in suppier performance. The more system-wide efficiency effects of growth in I were aso demonstrated by Giespie et a., (2000) for Scotand. Despite its importance, examining the significance of production externaities generated by I in the suppy chain has hitherto been probematic, argey due to data constraints. Nevertheess, there is some case evidence of the importance of such effects (see Oiver and Wikinson 1992; Morris et a. 1993). Moreover, there is an apparent consensus that ow eves of input-output inkages between the foreign and domestic sectors are an impediment to custer deveopment, a theme which is 4

5 increasing in importance in severa UK regiona deveopment agency strategy documents. Indeed there is more genera evidence purporting to demonstrate that those foreign investors with the owest rates of oca inkages contribute east to regiona growth prospects and competitiveness (Crone and Roper 1999; see aso Brand et a. 2000). There is then an underying assumption that higher eves of transactions inkages between foreign and domestic firms are beneficia to the domestic sector, with an impicit recognition that the intensity of input-output inkages encourage knowedge and technoogy spiovers to indigenous sectors. The focus of previous anaysis has been on the eve of backward inkage from mutinationas to the indigenous suppy base, on the assumption that those firms with the highest backward inkage contribute most to economic deveopment prospects (Hirschman 1958, see aso Scott 1982). This perspective can be inked to growth poe theory which focused attention on technoogica input-output inkages as a key generator of regiona growth, particuary as a resut of expansion in a reativey productive ead firm (see Erickson 1974). Backward inkages then have the potentia to generate greater indirect empoyment impacts than forward inkages in the regiona economy. There is aso some expectation that backward inkages are more important than forward inkages in creating productivity spiovers into the indigenous sector (Munday and Roberts 2001). However, as discussed beow, there is some concern over who appropriates the derived gain. In this paper forward inkage effects (i.e. foreign industries seing to domestic industries) are investigated with some expectation that these connections aso impact on domestic firms in terms of production externaities. For exampe, Erickson (1974) demonstrated the potentia for externay generated economies in forwardy inked sectors. These resuted from ead firms generating cost reductions for customer groups 5

6 because of their greater efficiency, and with other benefits deriving from the production of technoogicay superior intermediate goods. However, it is ikey that technoogy-reated spiovers wi resut from within-region technica change, and/or resut from extra regiona technica change (see Fingeton 2001). For exampe, in the atter case an adjacent region with a high concentration of reativey productive foreign firms coud, through purchasing inks, create spiover effects in domesticayowned suppier industries in another region. In this paper the presence of both interand intra-regiona effects from the presence of foreign manufacturing investment are investigated. 3. Input-Output Linkages with Foreign Manufacturers and Domestic Industry Performance The potentia for domesticay-owned suppiers to benefit from inkages with foreign manufacturers may be seen to have increased in the context of modern manufacturing environments. Fexibe production systems form a key component of cutures of continuous manufacturing improvement. Centra to the new manufacturing ethos has been the need to redefine reationships with suppiers and subcontractors. At one extreme, O huachain and Wasserman (1999) show in the Braziian case, how the demands paced by foreign car assembers on their first tier suppiers to produce more compex sub-systems and partake in design, ed through to some of these same first tier suppiers (often foreign-owned themseves) internaising activities that were previousy undertaken by a very arge number of their smaer subcontractors. Basicay tier one suppiers with greater responsibiities had vaid concerns about the quaity of production in the second tier and beow. This combined with norma contractua hazards ed to greater vertica integration in the vaue chain. In the UK, foreign manufacturers, particuary the Japanese, have ed the way in a rather different 6

7 redefinitiona process. Within this redefinition has been a trend away from short-term arms-ength contractua reationships emphasising exit in the case of probems, towards co-makership, and the estabishment of onger term reationships based on voice 1 (Heper 1991). More predominant voice reationships which are at the heart of continuous improvement manufacturing systems provide the potentia for a wider range of production externaities between buyer and suppier, particuary where the buyer is a mutinationa enterprise in possession of a set of ownership advantages embodied in physica and intangibe assets, technoogy and knowedge. 3.1 Backward inkages from mutinationa firms and spiovers. Case evidence reveas that in the above circumstances, foreign manufacturers might improve the capabiities of their suppiers in a number of ways. For exampe, coser communication, perhaps aided by Eectronic Data Interchange (EDI) may provide the basis for joint probem soving, exchange of ideas and technica information. Morris et a. (1993) demonstrate how Japanese mutinationas within coser buyer-suppier partnerships provide technoogica assistance, in terms of design, purchasing, marketing information, tooing, and in the promotion and reward of productivity improvements. Moreover, buyer-suppier partnerships and cose inspection of foreign firm activities has sometimes prompted indigenous suppier firms to adopt methods of personne management and work organisation found in the foreign sector (Oiver and Wikinson 1992). A growing body of empirica evidence (Barre and Pain 1997; Giespie et a. 2000), suggests that the UK economy improves technoogicay in the aggregate as a resut of such production externaities. Crone and Roper (1999) demonstrate the 1 Heper cassifies suppier reationships by methods with which the parties invoved resove probems. In an exit reationship, the customer finds a new suppier, whie in a voice reationship the customer works with the suppier to resove probems. 7

8 importance of transactions inkages from the foreign sector to indigenous suppiers as a critica driver of such positive production externaities. However, this production externaity might not be refected in the growth of vaue added in suppiers serving the foreign sector. Utimatey, foreign manufacturers by their size may have sufficient market/purchasing power to squeeze margins in suppying sectors. These market power effects are even more ikey to occur in the context of modern buyer-suppier partnerships, where restructuring has often invoved increased attention on suppier seection or even take-over (see O huachain and Wasserman 1999), a rationaisation of the suppy base i.e. a movement away from mutipe to singe sourcing, and the deveopment of onger term contractua reations, with contracts being awarded across mode ives. Moreover, a manufacturing preference for just-in-time deivery of components can potentiay increase the stock hoding costs of smaer suppiers which are unabe insist on just-in-time deivery from their own suppier set (Oiver and Wikinson 1992). Linked to the above review is previous research that has showed how coser buyersuppier partnerships can invove the suppier providing detais of financia costs to the manufacturer. In part the free exchange of cost information can assist in processes of vaue anaysis and vaue engineering, eading to systematic cost reductions in products (see for exampe Hiromoto 1988), abeit with Munday (1992) demonstrating that assembers tended to use cost data as a means of putting pressure on suppier prices. Indeed, some suppiers facing increased pressures within a more monopsonistic situation may actuay try to diversify their cient bases (Morris et a. 1993) and reduce dependence. Whie indigenous suppiers might be we paced to appropriate productivity externaities from the foreign sector, these gains might not resut in consistent improvements in their vaue added. The stronger bargaining 8

9 position of arger foreign manufacturing firms, couped with more stringent price, quaity and deivery standards, coud effectivey reduce margins verticay up the suppy chain. Through its market power, particuary purchasing bargaining power, the mutinationa is better paced to appropriate the gains from coser suppier inkages. 3.2 Forward inkages and spiovers In cases where indigenous firms buy from the foreign sector they potentiay benefit from the greater scae and scope efficiencies, competency, innovative capacity and technoogy of the mutinationa. In some cases, foreign mutinationas may provide direct assistance to customer groups. For exampe, Dunning (1993) found that US affiiates in the UK were more ikey to provide training for cients, than their domestic customers. Moreover, reviewing what scarce evidence there is, Dunning suggests that foreign firms, through inward investments bringing new management techniques and production processes to host nations, have had the effect of raising standards in downstream sectors, this inking to the fact that by improving the quaity of the output of their industria customers, they create new advantages for themseves. Then domestic customers of foreign manufacturers coud be in a stronger bargaining and earning position than domestic suppiers, such that the proprietary knowedge embodied in the product and technoogy of the mutinationa wi spi over more easiy into domestic firms who are essentiay the customers in the agreement. Further, Caves (1996) shows that entry by foreign mutinationas can increase competition in domestic markets, whist Geroski (1995), in a review, demonstrates that high price cost margins may serve to induce entry by foreign firms. In either case foreign entry coud break through domestic industry entry barriers, reduce prices, and 9

10 be a partia soution to market faiure (see aso Driffied and Munday, 1998). Ceary, market power approaches to the theory of the mutinationa woud suggest that the objective of overseas operations is simpy to increase monopoy rents, and as such foreign entry simpy eads to greater industry concentration. However, Driffied (2001) in a study of foreign entry into the UK, reveaed that entry by foreign manufacturing reduced concentration, and increased the speed at which industry sectors move to equiibrium. The above review indicates that there are expected to be production spiovers from the foreign sector arising through their direct buyer-suppier inkages. However, the review aso suggests that these effects may be pronounced in cases where foreign firms se to the domestic sector, rather than in the case of domestic firms who se to the foreign sector. In the atter case the foreign firms may be better abe to appropriate the gains from coser inkages. Norma transaction reationships woud aso suggest that evidence of externaities wi be found within industry, for exampe, where foreign firms transact with industries in their own broad sector, and aso where they transact with industries in other sectors. Moreover, given the intra-regiona suppy and demand constraints that inevitaby mitigate the attempts of foreign manufacturing pants to purchase/se extensivey in their immediate ocaities, then it is expected that evidence for externaities wi be found both within and outside the host region. 4. Assessing Production Externaities: Method, Data and Variabes Method The genera iterature on the growth of productivity is connected to that concerning the measurement of production externaities through production function approaches 10

11 (see for exampe, Griiches 1992; Griiches and Lichtenberg 1984). Cabaero and Lyons (1990, 1992) examine industry-eve spiovers from output growth and use a production function of the foowing type: nq irt = a + s n m m ny 1 m, irt + X c c irt 1 + u irt.(1) In Equation 1, Q is a measure of output, Y represents the vector of s factors of production, and X represents the vector of potentia externaities. This mode incorporates industry (i), region (r), and time (t) dimensions. There are a series of methodoogica considerations with this type of mode. Firsty, econometric studies of productivity growth indicate the importance of earning by doing and the cumuative effects of continuous production (Irwin and Kenow 1994). This suggests that accumuated experience is expected to be an important determinant of current productivity eves (see for exampe, Isam 1995). Therefore there is a need for a dynamic specification in which accumuated experience is proxied by a agged output. This is shown in Equation (2). Equation (2) captures the importance of past eves of inputs in the current production process. s n nq irt = a + 0 nq irt-1 + n m irt + X 1 c irt + u irt.(2) m m Y 1, In addition however, as is we understood, there is the possibiity for spatia dependence with these type of data. In modes such as this, it is ikey that the fixed effects mode wi not sufficienty contro for the spatia component in productivity growth, particuary where spiovers between regions occur. For exampe, Case (1991) examined the possibiity of spatia dependence where a region s output growth is in part dependent on output growth in a neighbouring region. This is we understood in the regiona science iterature (see for exampe, Ansein and Keejian c 11

12 1997). The crucia distinction here is between a spatia error mode, and a spatia ag mode. The spatia ag mode generay takes the form: nq irt = a + 0 nq irt-1 +W r Q irt + s n m m ny 1 m, irt + X c c irt 1 + u irt.(3) Where W r is the spatia contiguity matrix for region r. In the above case, output of the industry in other regions may impact on the output in the region r. The econometric issues associated with such a specification are discussed within a standard cross-sectiona framework in Ansein and Forax (1995), and LeSage (1999). The econometric treatment of spatia dependence is an important issue when seeking to identify sources of productivity spiovers, particuary when seeking to distinguish between genuine productivity effects. In the industria economics iterature, there is considerabe debate on the degree to which productivity or technoogy spiovers can be correcty identified. For exampe, there is the possibiity that observed increases in productivity represent simpy a demand or scae effect, where output increases simpy due to aggregate demand shocks, rather than spiovers that impact on productivity (Outon 1996; Basu and Fernad 1995). Spatia dependence is an important issue here. If productivity spiovers can sti be detected in the presence of spatia dependence, then one can be more confident in the interpretation. However, whie this ceary represents an improvement on Equation (2), this nevertheess impies a restriction that the size of the inter-regiona dependency does not vary across regions other than by variations in the contiguity matrix across coumns. Within a cross-sectiona framework this restriction is difficut to avoid, due to degrees of freedom constraints. Using pane data, where there is an industry as we as regiona component to the data, then the foowing reaxation of this restriction can be empoyed: 12

13 nq irt = a + 0 nq irt-1 + r W r Q irt + s n m m ny 1 m, irt + X c c irt 1 + u irt.(4) Where u irt = i + t + r + irt.. This therefore aows the spatia dependence term to vary across regions. It is aso possibe to aow to vary across industries and time, though not of course jointy due to degrees of freedom constraints. There is, however, no evidence of variation across industries or time, whie the region-specific measures are presented in Tabe 3. A standard way of estimating a spatia dependence mode is to empoy a maximum ikeihood approach (LeSage 1999, see aso Fingeton 2001). However, this approach becomes rather cumbersome, particuary with arge data sets (see Keejian and Prucha (1999) for further discussion of this). Further aternatives are discussed in Ehorst (2003), who highights the probems of deaing with arge pane data modes where N is arge reative to t, and aso the probem of deaing with endogeneity of regressors within any of the standard maximum ikeihood approaches to this type of probem. A simiar comparison is made by Be and Bockstae (2000). This probem becomes particuary pertinent in modes that incude a agged dependent variabe which by construction is correated with the fixed effects specified in Equation (4). Converting the data to first differences removes this probem, but the agged dependent variabe becomes endogenous and must therefore be instrumented. This is done by empoying further ags. Equay, the Y terms may aso be endogenous, and so again can be instrumented with further ags within this framework. nq irt = nq irt-1 + r W r Q irt + s m m n Y 1 m, irt + 6 c4 c X irt 1 +e irt.(5) 13

14 The Genera Method of Moments Instrumenta Variabes (GMM) estimator suggested by Areano and Bond (1988, 1991) can be appied to estimate (5), and this generates heteroscedastic-consistent estimates. Irrespective of the chosen specification between (3) and (4), there are further considerations with this estimation, as it is potentiay inefficient in the presence of spatia autocorreation. The appropriate estimation of space-time modes with endogenous variabes in the presence of spatia autocorreation and spatia dependency is however the subject of on going work. There remains a distinction however between modes that present the probem of spatia error, requiring spatia autocorreation to be aowed for in the estimation process, and spatia ag modes where the cause of the autocorreation can be identified. The resuts presented here are in the spirit of the atter. The most efficient test for spatia autocorreation, foowing Ansein and Rey (1991) is based on Moran s I statistic (Moran 1950). Once the spatia ag mode was empoyed in this case, there was no evidence of any spatia autocorreation in the errors. Equay, there is no evidence of inter-industry autocorreation, or seria correation in the errors Variabes and data Industry data for output, capita and abour in 20 manufacturing sectors were based on two digit industry (SIC 80) cassifications and derived from the UK Census of Production from the UK Office for Nationa Statistics (ONS). Data were avaiabe for for 10 UK regions (for descriptive statistics see Appendix 1). The dependent variabe (Q irt ) is gross vaue added in the domestic industry in the region. The independent variabe is the capita stock (K irt ) of the domestic industry. The change in this is given by net capita investment in the UK owned sector. Data on the 14

15 capita stock are not avaiabe at this eve of aggregation, so the sum of net investment over the previous five years is used as a proxy. A standard depreciation rate of 10% was used. Foowing Haske and Heden (1999) it is expected that types of abour affect industry output differenty i.e. abour is heterogenous. The Census of Production provides information on manua (ML irt : empoyment of operatives in domestic owned industry at time t in the region) and non-manua abour (NML irt empoyment of nonoperatives in domestic owned industry at time t in the region) and the mode estimated then distinguishes between abour type. Finay, time dummies were incuded to proxy exogenous technoogica progress. More compex is the process of designing a variabe that combines estimates of the inkage intensity of domestic firms with foreign firms (i.e. aowing exporation of inter/intra industry/region effects, both backwardy from the mutinationa to its domestic suppiers, and forwards from the mutinationa to its customers), with an estimate of new foreign manufacturing investment in those industries connected to domestic industries. The approach taken is to specify how the idea variabes might be estimated and then to show how the measure was constructed with imited information. Construction of the inkage component of the variabes woud require inter-regiona input-output tabes (for an exampe of this see Mier and Bair 1985) or sets of bi-regiona tabes (see for exampe, Oosterhaven et a. 2001). Within these tabes it woud be necessary to separate out the foreign and domesticay-owned eements of individua industries. More formay, Z can be defined as a matrix of intermediate inter-regiona transactions. For ease of description in the forthcoming discussion in reation to construction of idea variabes, the regions are denoted in superscripts. Hence an 15

16 eement in Z is defined Z, where 1 h, k R (R is the number of regions considered), and each Z describes transactions between a number S h of seing industries in the h th region and a number P k of purchasing industries in the k th region. In addition, in this discussion individua seing industries are denoted by i, and purchasing industries by j, 1 i S h, and 1 j P k. Within this framework (but not shown here) woud be associated matrices of fina demands and of outputs. Z = Z Z Z R1 Z Z Z R2 Z Z Z 1R 2R RR. Matrix Z coud then be expanded by separating the industries into foreign-owned (F) and domesticay-owned (D) components. Then for each Z eement in the matrix Z, a sub-matrix of the form Z = z z DD z z DF FF can be specified. The first (top eft) eement, z DD, represents transactions between domestic seing industries in region h and domestic purchasing industries in region k. The top right transaction, z DF, has the same interpretation, except that seing industries are domesticay owned and purchasing industries are foreign owned. By identifying the industry transactions, each singe eement Z can be further expanded into the form Z = z z DD z z DF FF z11 z11 = zsh1 zsh1 DD DD z z z 11DF 11FF z Sh1DF S 1FF h z z z 1Pk DD 1P z k ShPk DD S P h k z z z 1Pk DF 1P FF z k ShPk DF S P FF h k. 16

17 To construct the idea inkage variabe described earier, certain coefficients woud be derived (describing the intensities of both purchases and saes between industries). The purchasing coefficients woud be constructed for each industry by dividing each transaction ( z etc.) by the sum of intermediate purchases (i.e. the coumn totas ijdd from matrix Z), rather than the more usua gross output denominators. The objective in constructing this variabe is to measure the strength of inter/intra industry inkages. A gross output denominator incudes wages, imports, and other vaue added which coud serve to obscure the scae of inter-sectora transactions, and it is the atter which are hypothesised to infuence spiover effects. The coumn totas for domestic purchasing industries in matrix Z above in a R Sh particuar region k are given by, zijdd zij h1 i1 R Sh industries, the coumn totas woud be : zijdf zijff ( z h1 i1, whist for foreign purchasing Hence, without oss of generaity, in reation to the case of a particuar transaction ijdd ) between domestic industry i ocated in region h, and domestic industry j ocated in region k, a purchasing coefficient, here denoted by. ijdd, is derived by dividing this transaction by the coumn tota for the domestic purchasing industry in region k, ij DD R Sh z z qjdd qj h1 q1 z ij DD. (To avoid confusion, the index q is used for seing industries in pace of i, which represents a specific industry in this expression.) 17

18 Foowing the same descriptive process set out in reation to the eements in matrix T, these purchasing coefficients can anaogousy be represented in a matrix, here defined L, and given by L = L L R L L L L R2 L L L 1R 2R RR, which contains eements L, which again as described in reation to Z, can be expanded to specify industry trade reationships. L = DD DF FF = Sh1 Sh1 DD DD 11DF 11FF Sh1DF S 1FF h 1Pk DD 1P k ShPk DD S P h k 1Pk DF 1P FF k ShPk DF S P FF h k. Ideay the components of sub-matrix L (i.e. ijdd, ij etc) become the basis for estimating the inkage variabes representing the purchasing patterns/intensities of domestic industries. In a simiar way, and using eements within matrix Z, saes coefficients, representing the saes of domestic i industries to other industries coud be derived. In this case, each specific transaction within Z is divided by the corresponding row tota (i.e. the sum of intermediate saes). In reation to matrix L, and the construction of the idea variabes described earier, there woud be a requirement for a different matrix representing inkages in each of 18

19 the nine years incuded within the anaysis. Hence in the forthcoming matrix a further sub-script t (time) coud be added to each eement. However this has not been incuded for presentationa purposes. In addition, to create the variabes ony particuar eements of L (and its associated sub-matrices) woud be required. For exampe, to examine potentia spiovers to domestic industries as a resut of their purchases from foreign industries, ony the coefficients woud be used. Hence a sub-matrix of L coud be defined (L ) which woud ony incude coefficients of the genera form ij regiona trade, is therefore given by,. This L sub-matrix, expanded to define specific industria and L = S11 R1 11 R1 21 R1 S R S 2 R1 12 R R1 S 2 R 11 1P1 11 2P 11 S P R1 1P1 R1 2P R1 S P R 1 1R 11 1R 21 1R S 1 RR 11 RR 21 1 RR S 1 R 1R 12 1R 22 1R S 2 RR 12 RR 22 1 RR S 2 R 1R 1PR 1R 2P 1R S P 1 R RR 1PR RR 2P RR S P R R R R. The L sub-matrix above shows the intensities of the purchasing reationships between domestic and foreign industries. Using the notation given previousy for eements within L (i.e. ij ) four sets of inkage reationships coud be defined for each domestic industry j ocated in region k. To assess inkages between domestic and foreign firms in the same industry and in the same region, the coefficients in L where i = j and h = k woud be taken from L. To show domestic firm inkages to foreign firms in the same industry but in a different region, the coefficients where i = j, h k woud be used. For inkages with different foreign industries within the same 19

20 region a i j, h = k coefficients woud be required, and finay a coefficients where i j, h k woud represent connections to different foreign industries in different regions. In the same way, but not iustrated here, a sub-matrix of saes coefficients coud be derived to show domestic industries seing to foreign industries. In this case, the submatrix woud contain ony DF coefficients, and a further four sets of inkage reationships coud be defined. Finay, each of the inkage coefficients (or sums of groups of coefficients) in each time period woud be combined with a measure of new foreign direct investment in those industries inked to given domestic industries. Corresponding with the sets of inkage reationships described earier (i.e. ij ), the I measures (i.e. h I i ) in each time period are: the stock of foreign capita in the same industry and region; the stock of foreign capita in the same industry across a other regions; the stock of foreign capita in other industries in the region, and the stock of foreign capita in other industries across a other regions. This process woud resut in eight measures of I intensity in inter-inked industry sectors. For any given domestic industry j in region k there woud be four FORWARD inkage variabes: FWD INTRA-INDREG woud measure the concentration of new foreign manufacturing in industry i and in region h which domestic firms in industry j and region k buys from, where i = j and h = k, (i.e. same industry and same region). It h hh foows that FWD INTRA-INDREG woud be given by I. FWD INTER-INDREG woud measure the concentration of new foreign manufacturing in a seing industries which domestic industry j buys from, where i i ii 20

21 j, but h = k (i.e. different industries within the same region). Therefore FWD INTER- R Sh h hh INDREG coud be defined as I i ij h1 i1 i j FWD INTRA-INDOREG woud measure the concentration of new foreign manufacturing in industry i which domestic firms in industry j buys from, where i = j and h k, (i.e. same industry in different regions). FWD INTRA-INDOREG woud R Sh h then be defined as I i ii h1 i1. FWD INTERIND-OREG is a measure of the concentration of new foreign manufacturing in a seing industries which domestic industry j buys from, but where i j and h k (i.e. different industries in different regions). This variabe woud be R Sh h specified by I i ij h1 i1 i j. BACKWARD inkage variabes woud be derived in a simiar fashion but woud be in terms of the concentration of new I in those industries to which domestic industries se (hence the inkage coefficient woud have a DF subscript and the I measure woud be k I j. The inter-regiona ownership disaggregated trade data to enabe construction of the above defined variabes is, however, not currenty avaiabe in the UK, athough attempts have been made to examine ownership disaggregated transactions at the singe region eve (see Munday and Roberts 2001; see aso Giespie et a. 2000). The inkage measures were therefore constructed using the UK input-output data for 1990 from the Office of Nationa Statistics. Using the Make and Use matrices it was possibe to construct a symmetrica industry by industry transactions matrix for. 21

22 the UK, with the manufacturing industries aggregated to the defined 2 digit industry groups. With this imited input-output information, the anaysis of the impacts of inward I on domestic productivity growth is based on two input-output matrices, (representing the intensity of saes from each industry to other industries) and (representing the strength of purchasing reationships between industries) for the UK. In this framework, for exampe, the matrix can then be written as: n n2 1n 2n nn where n is the number of industries. The eading diagona refers to intra-industry inkages, and the other terms to inter-industry inkages. The inkage coefficients were derived in the same manner as described earier in reation to the idea data i.e. to derive purchasing coefficients for each specific industry, each transaction was divided by the sum of intermediate purchases, whereas the saes coefficients have intermediate saes as a denominator. Hence in reation to the FORWARD inkage variabes defined earier, the respective ij eements are approximated by (and repaced with) a nationa inkage coefficient, defined here as v ij. Simiary, the BACKWARD inkage variabes aso contain coefficients derived from the nationa input-output matrix. The h and k superscripts have been maintained for the construction of the variabes, however, there is no variation in these industry inkage coefficients by region. By combining data from the two input-output matrices of nationa inkage coefficients with measures of foreign direct investment (by industry, region and time) 22

23 an index is provided for each of the manufacturing groups of the significance of new foreign manufacturing investment in industries that domesticay based industries buy from/se to both in the same region, and in other regions. Hence whist there is no variation in industry inkage coefficients by region or by time, variation in these independent variabes is attained through the I measures which are stratified by industry, region and time. There are a number of issues reated to variabe construction. First, the inkage coefficients were derived for It was not possibe to derive inter- or intraindustry inkage coefficients for each of the years corresponding to the sampe of industry data (i.e ) as the required input-output tabes are ony produced at discrete intervas. For this exercise inkage coefficients for 1990 are assumed to be a reasonabe proxy for those occurring in the period. Second, the foreign-owned capita stock is used as the measure of foreign direct investment in inked sectors. As the foreign-owned capita stock is cacuated using a perpetua inventory method, and depreciated at 10% p.a., more weight is naturay given to new investment. As the estimation of the mode is carried out on firstdifferenced data, we are effectivey reating the change in I stocks to changes in tota factor productivity. Third, the measure introduces restrictions on the mode in that the inkage coefficients derived ony proxy for the transactions reationships between domestic and foreign industries, such as those defined earier in matrix L. With imited information, the coefficients refect the activities of a industries (foreign-owned and domestic), with some expectation that these reationships wi differ between domestic and foreign industries (Brand et a., 2000). However, the purpose of this anaysis was to generate a genera measure of inkages between industry groups. Moreover, this 23

24 process, which reies on nationa rather than inter-regiona trade information, assumes that firms in industries in one region, have simiar purchasing propensities to the same industries in another region. As highighted earier, in order to derive more accurate inkage coefficients, an ownership disaggregated inter-regiona input-output tabe, with each industry separated into foreign and domestic components woud be required. Nevertheess, the methodoogy presented here marks a significant improvement on previous work seeking to investigate econometricay the impacts of I on tota factor productivity growth. This anaysis presents a first attempt to reate I across the fu popuation of industries and regions, to impacts in both upstream and downstream industries. Constructing eight variabes in this fashion enabes an examination of the extent to which there are different effects on domestic industries reated to whether there is new I in the same industry or different industries to which they are inked. For exampe, it might be the case that where a domestic industry ses/buys to a foreign firm in the same industry the degree of externaity may be mediated because firms have better knowedge of one another and the potentia for opportunistic behaviour is imited in both parties to the transaction. As outined in the previous section the expectation is that where industries with concentrations of new foreign manufacturing se to domestic industries (i.e. the FORWARD variabes) then there are positive output externaities. However, where industries with concentrations of new foreign manufacturing buy from domestic industries (i.e. the BACKWARD variabes) then effects are expected to be negative. 5. Resuts. 24

25 There are potentia coinearity probems in estimating Equation 5 with eight externaity variabes. The process undertaken here was to estimate the mode with the forward and backward inkage variabes separatey to evauate the different sets of effects. The fu mode is reported in Appendix 2. Additiona diagnostics, with regard to mode specification and other tests are detaied in the footnote to Tabe 1. Equation 3 was estimated, initiay to confirm the existence of spatia dependence within these data, and secondy to determine its nature, that is, whether r = b for b r across regions. The existence of inter-industry dependence was aso investigated, but its existence was rejected. Time series ags, as we as spatia ags in spatia dependence were aso investigated. Both a rook contiguity, and a standard first order contiguity matrix were empoyed here, athough given the number of the regions, there is itte difference between the two, and the resuts are not sensitive to matrix specification. The resuts presented here in Tabes 1 and 2 are based on the standard first order matrix. The three coumns in Tabes 1 and 2 reate to the various assumptions concerning the existence (and nature) of spatia dependence discussed above. The first coumn in each tabe presents the resuts based on the estimation of Equation (2), that is, incuding a agged dependent variabe, but excuding any terms aowing for spatia dependence in tota factor productivity growth. The second coumn in the tabes iustrate the resuts from the estimation of (3), that is with the degree of spatia dependence constant across regions, with contiguity inferred from a standard first order contiguity matrix. The third coumn in each case refers to the estimation of (4), that is aowing the magnitude of spatia dependence to vary across region, with the mean vaues of presented in Tabes 1 and 2, with the regiona specific vaues of r (spatia dependence) presented in Tabe 3. However, whie there is some variation in the regiona dependence effects across regions, and the 25

26 coefficients are significanty different from zero, they are not significanty different from one another. A inputs into the production function, and the agged dependent variabe are treated as theoreticay endogenous, and are therefore instrumented with a avaiabe ags. The Sargan test for the vaidity of the instruments confirms the suitabiity of this approach. The resuts shown in Tabes 1 and 2 revea significant evidence of spatia dependence, that is that productivity spiovers (not from I, just generay) do occur across the mainand UK regions (or simpy that macro, or industry eve shocks impact on industries across regions). There is no sign of an equivaent inter-industry dependence. The concusions regarding the beneficia effect of I in sectors which domestic industries buy from are sensitive to the specification of spatia dependence. However, in the case of the negative effects occurring where domestic firms se to industries with the foreign presence then the resuts are not sensitive to the specification for spatia dependence. The resuts then indicate that the modeing framework picks up the impacts of foreign infuence in the vaue chain on domestic producers, that is, the framework is not simpy describing demand effects. Importanty, the beneficia effects on the domestic sector (Tabe 1) are found to diminish as restrictions concerning spatia dependence are reaxed, suggesting that inter-regiona productivity gains from I may simpy be demand effects. This is an interesting resut, with reference to the more genera work on spiovers and mode specification, which suggests that many reported spiover effects, are simpy demand effects, rather than genuine productivity gains (see for exampe, Outon 1996). Tabes 1 & 2 about here 26

27 The resuts in Tabe 1 indicate that domestic industries coud be benefiting from purchasing inkages with industries with strong foreign invovement. Significant resuts were found where domestic industries purchase from the same industry group with high eves of foreign invovement, but in a different region, and where domestic industries purchase from different industry groups with high eves of foreign invovement in the same region. The remaining externaity variabes were not significant but have the anticipated sign. Then there is some evidence that domestic firms may be benefiting from the avaiabiity of better quaity components, better technoogy and ower prices in inked industries with a strong foreign component. However, as highighted the resuts are sensitive to the specification of the mode. As reveaed in the review in section 3, this finding offers some support for the contention that foreign manufacturers do bring innovative manageria techniques and production processes to host areas which have the effect of improving standards in downstream sectors. It is we understood that the scae and scope of spiovers is party dependent on the actions of the domestic firm, and their abiity to assimiate the externaity (Bomstrom et a. 2001). Thus, one woud expect that the greater the technoogica simiarities between the foreign and domestic firms, the greater the potentia for productivity growth in the domestic sector. Positive spiovers occurring forward through transaction inkages might utimatey be connectabe to onger term improvements in the competitiveness of segments of domestic industry. The presence of these inter/intra-industry and inter/intra regiona effects certainy reates cosey to the wider nationa and regiona objectives underying the attraction of foreign capita to the UK, and means that genera UK evidence on the more dynamic positive impacts of foreign manufacturing across UK industry cannot be dismissed (see for exampe, Driffied and Munday 2000). 27

28 Tabe 2 demonstrates that in cases where domestic firms se to industries with high eves of foreign manufacturing investment the nature of inkage externaities is ess cear. Consequenty, where domestic industries se to the same or different industry with higher eves of foreign investment, but in other regions, then there is evidence of a negative impact on gross vaue added. This provides some evidence that foreign manufacturers in the same industry coud have sufficient market power to appropriate the gains from buyer-suppier partnerships. Moreover, where foreign firms purchase from domestic firms in the same industry sector, then they coud have more expert knowedge of suppier conditions which reduces scope for suppy side opportunistic behaviour. This resut does not equate with no efficiency improvements within the indigenous suppier, but that foreign firms are in a stronger position to appropriate the gains. This situation occurs where the inward investor sources products from outside its immediate ocaity, for one of severa reasons. Firms esewhere are superior to oca firms, either due to economies of scae, or other technoogica advantages, or it may simpy be the case that there are no potentia suppiers to be found ocay. In either case, it is unikey that the foreign firm wi exhibit the same bargaining advantage over such firms, as they do over oca suppiers, and so the foreign firm is unabe to assimiate the gain. 6. Concusions The paper describes a methodoogy for assessing the significance of production externaities occurring from foreign enterprise to domestic enterprise through transactions reationships. The paper provides evidence that productivity growth in the domestic sector is affected by the nature of transactions inkages with foreign-owned 28

29 manufacturing firms. Gains to the domestic sector appear to be greater where domestic firms purchase from foreign firms, athough effects weaken with aternative specifications. One contention of the review section of this paper is that regiona poicy anaysis (in the UK at east) has tended to focus upon the economic significance of backward inkages from the mutinationa firm into the host region economy. Whist this focus might be justified in terms of assessing the indirect empoyment, output and jobs supported by foreign investment, this paper reveas that the nature of production spiovers backwards to the domestic suppy sector is far from cear. Whist the encouragement of forward inkages has assumed a esser importance in terms of regiona poicy strategy, this paper has demonstrated that this is ikey to be the strongest channe for positive production spiovers to the domestic sector, when considered in terms of generating rea productivity growth. The resuts then suggest that there is merit in poicy terms of investigating the issue of who domestic firms buy from. This is probematic argey because the resuts here tend to suggest that the dispacement of purchases from domestic firms, by purchases from foreign-owned ones may actuay ink through to productivity advantages, and then potentiay through to comparative advantage as reveaed in manufacturing export performance. The concusions from the paper are tentative argey because of the difficuties invoved in assessing the eves of foreign manufacturing invovement in different domestic industry transaction chains. However, the paper makes steps towards a method of assessing such externaities incorporating information from input-output tabes. In further research there may be merit in investigation on an industry by industry basis, such that case materia can be used to assist in the interpretation of empirica findings. Case materia might aso be examined to support the anaysis of 29