Knowledge Spillovers from Multinationals to Local Firms: International and Irish Evidence

Transcription

1 Knowledge Spillovers from Multinationals to Local Firms: International and Irish Evidence DATE 12 April 2018 VENUE ESRI, Whitaker Square, Sir John Rogerson s Quay, Dublin 2 AUTHORS Iulia Siedschlag Mattia Di Ubaldo

2 Recent International and Irish Evidence Di Ubaldo, M., M. Lawless, I. Siedschlag (2018). Productivity spillovers from multinational activity to indigenous firms in Ireland, ESRI Working Paper No. 587 Siedschlag, I., M. Di Ubaldo, M. Tong Koecklin (2018). Comparative Performance of Indigenous and Multinational Firms Operating in Ireland, European Commission. Jude, C. (2016). Technology spillovers from FDI. Evidence on the intensity of different spillover channels, The World Economy, Newman, C., J. Rand, T. Talbot, F. Tarp (2015). Technology transfers, foreign investment and productivity spillovers, European Economic Review 76: Haller, S. (2014). "Do domestic firms benefit from foreign presence and import competition in the Irish services sectors? The World Economy, 37(2): Havranek, T., Z. Irsova (2011). Estimating vertical spillovers from FDI: Why results vary and what the true effect is, Journal of International Economics, Vol.85, pages Barrios. S., H. Görg, E. Strobl (2011). Spillovers through backward linkages from multinationals: Measurement matters!, European Economic Review, 55: Meyer, K., E. Sinani (2009). When and where does foreign direct investment generate positive spillovers? A meta- analysis, Journal of International Business Studies, 40: Ruane, F., A. Uğur (2004). Foreign direct investment and productivity spillovers in Irish manufacturing industry: Evidence from plant level panel data, International Journal of the Economics of Business, 11(3):

3 Outline Research and Policy Context Knowledge Spillovers from Multinationals to Local Firms Main Channels and Measurement Productivity Spillovers International and Irish Evidence Trade Spillovers International and Irish Evidence Policy Implications 3

4 Research and Policy Context Multinationals are associated with new technologies and advanced management know-how International evidence indicates that multinationals are more productive than local firms Indirect positive effects on the performance of local firms via knowledge spillovers Knowledge diffusion from multinationals to local firms could increase aggregate productivity International evidence suggests that the productivity slowdown in recent years is linked to a slowdown of knowledge and innovation diffusion from top performance firms to laggard firms Ireland one of the most globalised economies in the world Multinational firms contribute substantially to Ireland s economic performance and competitiveness Foreign-owned firms - large scale, capital-intensive, concentrated in high-productivity sectors, concentrated geographically, highly globalised Irish-owned firms - much smaller, more labour-intensive, more-embedded in local economies, less internationalised 4

5 Knowledge Spillovers from Multinationals to Local Firms Main Channels Demonstration effects: local firms learn about new technologies and markets from the activities of multinationals Competition effects: local firms innovate more Supply chain linkages: knowledge transfer between multinationals and local firms in upstream and downstream sectors Labour mobility: tacit knowledge transfer (embedded in human capital/skills) 5

6 Knowledge Spillovers from Multinationals to Local Firms Measurement Horizontal Spillovers linked to the presence of foreign-owned firms in the local firm s industry or region Intra-industry spillovers HOR j,t : The share of foreign affiliates in the total output or employment of industry j at time t Intra- region spillovers HOR r,t : The share of foreign affiliates in the total output or employment of region r at time t Vertical Spillovers via supply chain linkages Inter-industry forward linkages - supplies by foreign affiliates to local firms FOR j,t : Measure capturing the intermediate inputs available from foreign affiliates in upstream industries to domestic firms in industry j. Inter-industry backward linkages supplies by local firms to foreign affiliates BAC j,t : Measure capturing the local firms output in industry j available to foreign affiliates in downstream industries. 6

7 AUTHORS Mattia Di Ubaldo Martina Lawless Iulia Siedschlag Productivity Spillovers from Multinational Activity to Domestic Firms: Evidence from Ireland

8 Productivity Spillovers International Evidence Recent reviews and quantitative meta-analyses of international evidence Havranek and Irsova (2011); Meyer and Sinani (2009); Görg and Strobl (2001) Productivity spillovers vary across firms and host countries Mixed evidence on productivity spillovers via horizontal linkages weak or negative spillovers More robust evidence on productivity spillovers via supply chain linkages - in particular spillovers to local suppliers Ability to benefit from knowledge spillovers varies across firms large firms, firms more intensive in skills and R&D are more successful in absorbing and translating knowledge into innovation and productivity Productivity spillovers are greater in countries open to international trade having stronger incentives to innovate and related non-linearly (U-shaped form) to the level of development in host country human capital, enabling institutional framework 8

9 Previous Research on Ireland Within-industry spillovers (horizontal) with manufacturing and service sectors examined separately (e.g. Ruane and Ugur, 2005; Haller, 2014) Positive/negative but statistically insignificant effects Within and inter-industry (vertical) spillovers (Barrios, Görg, and Strobl, 2011) allow for the input sourcing behaviour of multinationals to be different from that of domestic firms Positive and statistically significant spillovers via backward linkages Negative but statistically insignificant horizontal spillovers and via forward linkages

10 Research Objectives and Novelties Research Objectives Examine productivity spillovers within sectors and within regions Examine vertical spillovers via forward and backward linkages Examine whether spillovers differ across sectors Distinguish spillovers from affiliates owned by the EU and by non-eu multinationals Examine the role of absorptive capacity of domestic firms Novelties Estimate and compare sector specific productivity spillovers manufacturing and services Explore the sensitivity of productivity spillovers to the origin of investors investors based in EU versus those based in non-eu countries Explore the sensitivity of productivity spillovers to a range of measures capturing supply chain linkages 10

11 Key Findings Only limited evidence of productivity spillovers linked to the presence of foreign-owned firms in the same industry or in the same region Positive intra-industry spillovers in services mainly linked to EU based MNEs Crowding out effects in manufacturing linked to both EU and non-eu based MNEs Stronger evidence of both negative and positive spillovers through supply chain linkages Positive spillovers on local buyers from EU based MNEs in services Negative effects on local buyers from non-eu MNEs in manufacturing Positive spillovers appear to be conditioned by the absorptive capacity of local firms their ability to learn and translate new knowledge into innovation and productivity 11

12 Data Unbalanced panel of annual firm-level data, provided by the Central Statistics Office (CSO) Census of Industrial Production (CIP) Annual Services Inquiry (ASI) Business Register Multifactor productivity (MFP) estimates at firm level obtained as part of the MultiProd OECD project provided by the Department of Finance Inter-country Input Output Tables (WIOD),

13 Foreign ownership premia Source: Authors estimates based on data from the CIP and ASI. Estimates of MFP provided by the Department of Finance obtained as part of the MultiProd project. All regressions include industry, region and time fixed effects. 13

14 Empirical Methodology Following Javorcik (2004); Barrios, Görg, and Strobl (2011) d ln TFP ijrt = β 0 + β 1 HOR j,t 1 + β 2 HOR r,t 1 + β 3 FOR j,t 1 + β 4 BAC j,t 1 + β 5 SALES jt + β 6 HHI jt + β 7 Z d ijrt + μ i + γ j + ρ r + τ t + ε ijrt d lntfp ijrt The productivity of domestic firm i in sector j, in county r and in year t. HOR HOR Y Y f j, t 1 j, t 1 / j, t 1 Y Y f r, t 1 r, t 1 / r, t 1 The share of foreign affiliates employment in total employment in industry j at time t-1. The share of foreign affiliates employment in total employment in region r at time t-1 FOR, j, t 1 ljhorl, t 1 l j l ; lj : the share of inputs of industry j purchased from industry l BAC, j, t 1 ljhorl, t 1 l j l ; lj : the share of output of industry j supplied to industry l Z d ijrt: domestic firm characteristics age, skill intensity, R&D intensity, international experience, size April 2018

15 Productivity spillovers EU vs Non-EU multinationals (1) (2) (3) Dependent variable Ln(MFP_Solow) t Ln(MFP_Solow) t Ln(MFP_Solow) t Sample All firms Services Manufacturing Intra-industry EU (t-1) ** * (0.0784) (0.0845) (0.132) Intra-industry non-eu (t-1) ** *** (0.0620) (0.100) (0.0755) Intra-region EU (t-1) (0.0636) (0.0649) (0.159) Intra-region non-eu (t-1) (0.0407) (0.0401) (0.0979) Forward link EU (t-1) ** 1.622* (0.657) (0.901) (4.207) Forward link non-eu (t-1) *** (1.005) (1.565) (1.976) Backward link EU (t-1) ** (1.044) (1.038) (5.530) Backward link non-eu (t-1) (0.894) (0.984) (3.789) N 19,215 14,138 5,077 Source: Authors estimates using data from Ireland s Central Statistics Office (CSO). Notes: Estimates of TFP based on the Solow Index obtained as part of the MultiProd project. Standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01. Forward and backward spillover measures are computed excluding imported inputs. The control variables included in all regressions not reported above are: industry growth, industry competition, age, skill intensity, engagement in exporting, engagement in importing, engagement in R&D, and dummy variables for firm size. All regressions include time, firm, industry, and region fixed effects. 15

16 EU vs Non-EU MNEs - The Role of Absorptive Capacity Horizontal Spillovers (1) (2) (3) Dependent variable Ln(MFP_Solow) t Ln(MFP_Solow) t Ln(MFP_Solow) t Sample All firms Services Manufacturing Intra-industry_EU (t-1) ** (0.0794) (0.0826) (0.136) Intra-industry_nonEU (t-1) ** *** (0.0619) (0.0986) (0.0767) Intra-region_EU (t-1) (0.0634) (0.0651) (0.153) Intra-region_nonEU (t-1) (0.0409) (0.0402) (0.0953) Intra-industry_EU* RD/Emp (t-1) (0.102) (0.118) (0.658) Intra-industry_nonEU* RD/Emp (t-1) * (0.0285) (0.0247) (0.394) Intra-region_EU* RD/Emp (t-1) *** ** (0.0408) (0.0418) (0.522) Intra-region_nonEU* RD/Emp (t-1) (0.0696) (0.0707) (0.629) N 19,215 14,138 5,077 Source: Authors estimates using data from Ireland s Central Statistics Office (CSO). Notes: Estimates of TFP based on the Solow Index obtained as part of the MultiProd project. Standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01. Forward and backward spillover measures are computed excluding imported inputs. The control variables included in all regressions not reported above are: industry growth, industry competition, age, skill intensity, engagement in exporting, engagement in importing, R&D intensity, and dummy variables for firm size. All regressions include time, firm, industry, and region fixed effects. 16

17 EU vs Non-EU MNEs - The Role of Absorptive Capacity Vertical Spillovers (1) (2) (3) Dependent variable Ln(MFP_Solow) t Ln(MFP_Solow) t Ln(MFP_Solow) t Sample All firms Services Manufacturing Forward link EU (t-1) ** 1.618* (0.662) (0.911) (4.286) Forward link non EU (t-1) ** (0.994) (1.547) (1.966) Backward link EU (t-1) ** (1.038) (1.034) (5.596) Backward link non EU (t-1) (0.891) (0.977) (3.779) Forward link EU* RD/Emp (t-1) (1.235) (1.272) (11.64) Forward link non EU* RD/Emp (t-1) ** (1.481) (1.499) (8.459) Backward link EU* RD/Emp (t-1) 0.775* 1.052*** *** (0.406) (0.394) (7.774) Backward link non EU* RD/Emp (t-1) ** ** 23.43*** (0.297) (0.300) (4.515) N 19,215 14,138 5,077 Source: Authors estimates using data from Ireland s Central Statistics Office (CSO). Notes: Estimates of TFP based on the Solow Index obtained as part of the MultiProd project. Standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01. Forward and backward spillover measures are computed excluding imported inputs. The control variables included in all regressions not reported above are: industry growth, industry competition, age, skill intensity, engagement in exporting, engagement in importing, R&D intensity, and dummy variables for firm size. All regressions include time, firm, industry, and region fixed effects. 17

18 AUTHORS Mattia Di Ubaldo Iulia Siedschlag Trade Spillovers from Multinationals to Local Firms

19 Research and Policy Context Activity of MNEs, both on domestic and foreign markets, can affect the trade performance of local firms. Competition effects and knowledge spillovers are the main channels, similarly to what affects firm productivity. More specifically for trade: Lower cost of trade and increased exporting of local firms. Reduced informational barriers to trade, learning about export opportunities. Necessity to target foreign markets, as domestic market shrinks. Either by exporting more products or reaching more destinations. Weaker local firms are driven out of exporting. Increased production capabilities and product quality upgrading can foster both export success and imports of new/better inputs. 19

20 International Evidence Aitken et al (1997); Greenaway et. al, (2004): higher export probability for Mexican and UK firms, from horizontal spillovers. Swenson (2008): new export connections (product and destinations) for Chinese exporters information spillovers (horizontal). Swenson and Chen (2014): higher prices and higher survival rates for new export transactions from horizontal spillovers in Chinese firms. Harding and Javorcik (2012): export quality upgrade in developing countries. Industry data on FDI targeting and export unit-values. Bajgar and Javorcik (2017): product quality upgrading through vertical spillovers, custom data from Romania, negative effect from horizontal spillovers. Ciani and Imbruno (2017): quality upgrading from forward spillovers, quality downgrading from horizontal spillovers; data from Bulgaria. 20

21 Research Objectives and Novelties Examine how MNEs affect the trading activity of local firms in terms of diversity of transactions and density of network. Number of products, destinations and export/import intensity. Analyse both exports and imports Long time span: Examine the effect of both horizontal (i.e. same sector or region) and vertical spillovers (i.e. supply chain channel). Various spillover measures Examine the role of absorptive capacity of domestic firms, identified by their R&D activity. 21

22 Key Findings for Ireland Trade of local firms is impacted positively by MNEs activity. Exports: Foreign presence in the same sector leads to exporting to more destinations Competition-innovation; imitation; learning; network. Supplying inputs to MNEs leads to: Imports: More products shipped innovation, productivity. More destinations reached network, productivity. Purchasing inputs from MNEs leads to: More products imported complexity, productivity. More origins for imports network, productivity. Both effects more prevalent for R&D active firms. 22

23 Key Findings for Ireland Some negative effect arises from MNEs through regional spillovers: Lower number of products exported and imported. Lower number of import origins. Both effects are likely due to crowding out of local firms. Absorptive capacity of local firms appears to reinforce some effect, mostly forward spillovers: When buying inputs from MNEs, R&D active firms: Expand the range of imported products. Expand the range of export destinations. Expand the range of import destinations. 23

24 Evidence from Ireland - Data We exploit confidential firm and product level data made available by the Central Statistics Office (CSO). Data are from 1996 to Trade data: o Merchandise exports and imports o CN-8 digit product level. o Disaggregated by destinations (exports) and origins (imports). Firm level data: o o Balance sheet information from the Census of Industrial Production (CIP). Information on employment, revenue, materials, wages, R&D expenditure etc. 24

25 Empirical Methodology ln(y) d 5 ijrt = β 1 Hor j,t 1 +β 2 Hor r,t 1 + β 3 For j,t 1 +β 4 Bac j,t 1 + q=1 β Z q ijrt 1 + μ i + ρ t + ε ijrt d ln(y) ijrt The trade outcome of domestic firm i in sector j, in county r and in year t. no. products; no. destinations/origins; export- or import- intensity. Hor j,t 1 = Emp f jt 1 Emp jt 1 Hor r,t 1 = Emp f rt 1 Emp rt 1 The employment share of foreign firms in industry j at time t-1. The employment share of foreign firms in region r at time t-1 For j,t 1 = l δ lj Hor j,t 1 where δ lj is the share of inputs of industry j purchased from industry l Bac j,t 1 = l γ lj Hor j,t 1 where γ lj is the share of output of industry j supplied to industry l 25

26 26 Results Exports: Products

27 27 Results Exports: Destinations

28 28 Results Exports: Intensity

29 Summary of findings - Exports Horizontal spillovers (sectors and regions): Products: negative regional spillovers crowding out. Destinations: positive sectoral spillovers competition, learning, network. Vertical spillovers (value-chain): Products: positive backward spillovers. Destinations: positive backward spillovers. Positive forward spillovers for R&D active firms. 29

30 30 Results Imports: Products

31 31 2 February 2018 Results Imports: Origins

32 32 2 February 2018 Results Imports: Intensity

33 Summary of Findings - Imports Horizontal spillovers (sectors and regions): Products: negative regional spillovers crowding out Partially offset by firms doing R&D. Origins: negative regional spillovers crowding out Vertical spillovers (value-chain): Products: positive forward spillovers Due to firms doing R&D Origins: positive forward spillovers Stronger in R&D active firms 33

34 Policy Implications Attracting foreign direct investment is not sufficient to generate benefits to local firms via knowledge spillovers Since knowledge spillovers are not automatic, enhancing the absorptive capacity of local firms is key in order to ensure they can benefit from advanced knowledge and technologies Since most knowledge spillovers appear to come about through supply chain linkages, enabling production linkages between local and multinational firms could be beneficial for aggregate productivity and trade performance 34

35 Ireland: Enterprises with innovation expenditures (%) All expenditure on innovation In-house R&D Purchase of external R&D Acquisition of machinery, equipment and software Acquisition of other external knowledge Irish-owned Foreign-owned Source: Own elaboration based on data from the Community Innovation Survey 2016, CSO 35

36 Ireland: Enterprises engaged in co-operation for technological innovation (%) Any type of cooperation Cooperation with other enterprises within own enterprise group Cooperation with universities and or third level institutions Cooperation with suppliers of equipment, materials, components or software Cooperation with Cooperation with competitors and or consultants and or other enterprises commercial in same sector laboratories or private research and development institutes Cooperation with government or public research institutes Irish owned Foreign owned Source: Own elaboration based on data from the Community Innovation Survey 2016, CSO 36

37 Ireland: Enterprises engaged in joint research projects (%) Engaged in any collaboration Other firms in Ireland Other firms outside Ireland Higher education or other institutes in Ireland Higher education or other institutes outside Ireland Irish owned Foreign owned Source: Own elaboration based on data from the Business Expenditure on Research and Development, , CSO 37

38 EU countries: Enterprises providing ICT training (% of all enterprises) AT BE FI IE DE LU DK UK SI SE ES MT PT EU CZ HR CY NL FR SK HU EL EE IT LV PL LT BG RO Source: Own elaboration based on the data from the European Innovation Scoreboard

39 EU countries: Population aged having completed tertiary education (%) CY LT IE LU SE UK DK NL BE FR PL SI LV EE EL ES FI AT EU PT MT SK HR BG CZ DE HU IT RO Source: Own elaboration based on the data from the European Innovation Scoreboard 2017

40 EU countries: Population aged involved in lifelong learning (%) SE DK FI FR NL LU EE AT UK SI EU PT ES CZ DE IT MT LV BE CY IE HU LT EL PL HR SK BG RO Source: Own elaboration based on the data from the European Innovation Scoreboard

41 Economy wide restrictiveness of product market regulations in EU countries NL UK AT DK DE EE FI IT PT SK HU BE CZ EU ES IE LU FR SE LT BG MT LV PL CY RO SI EL HR Source: Own calculations based on most recent OECD data 41

42 EU countries: Regulatory barriers to entrepreneurship SK NL IT DK AT PT UK FI EE LT PL DE FR EU HU BG SE LU BE SI CZ EL IE HR LV RO CY ES MT Source: Own calculations based on most recent data from the OECD 42