Tine Jeppesen FIW Research Conference Vienna December 10 th 2010
2 Research Question and Motivation Contributions of paper Theory & Literature Data & Empirical methodology Results Conclusion
3 Research Question Does FDI affect the export participation and performance of domestic firms? What are the channels of export spillovers? ◦ Horizontal / Vertical linkages Research Question & Motivation
4 Why study export spillovers? Promoting exports is common policy goal Correlation between growth of real exports and real output (Greenaway 2004 European Journal of Political Economy) Exporting causes productivity increases? De Locker (2007) JIE Export spillovers may be an additional benefit from hosting FDI Research Question & Motivation
5 Evidence from multiple countries Region where FDI is increasing rapidly Firm-level information on vertical linkages Contributions
6 Theoretical Framework Findings in empirical literature on export behaviour ◦ Sunk costs (Roberts & Tybout, 1997) ◦ Exporter are more productive than non-exporters (Bernard & Jensen ; Melitz, 2003) MNEs can affect export performance of domestic firms by ◦ Lowering sunk costs ◦ Increasing domestic productivity Theory & Literature
7 Foreign market information (Aitken et al JIE) Knowledge/ technology spillovers (labour movements, imitation) Competition (Greenaway et al European J. of political Economy) Channels of export spillovers: Horizontal spillovers Theory & Literature
8 Greater scope for spillovers? (Javorcik, 2004 AER) Evidence of productivity spillovers to upstream industries in Eastern Europe (Javorcik, 2004 AER ; Gorodnichenko et al IZA; Javorcik and Spatareanu, 2009) Information, reputational effects Theory & Literature
9 Data Enterprise Surveys, The World Bank Group Firm-level data Survey rounds: 2002, 2005, 2008/09 Survey universe ‘commercial, service or industrial business establishments with at least five fulltime employees’ 2002 and 2005: Quota Sampling 2008/09: Stratified random sampling Data and empirical methodology
10 23 countries Albania, Armenia, Azerbaijan, Belarus, Bulgaria, Croatia, Czech Rep. Estonia, Georgia, Hungary, Kazakhstan, Kyrgyz Rep. Latvia, Lithuania, Moldova, Poland, Romania, Russia, Slovakia, Slovenia, Tajikistan, Turkey, Ukraine. Manufacturing industries ISIC Rev.3.1: Small, medium and large firms Data and empirical methodology
11 Multinationals are bigger, more productive, use a higher share of foreign inputs, supply a higher share of inputs to other multinationals and are more export oriented than domestic firms. The same differences are found between domestic exporters and non-exporters. Data and empirical methodology
12 Heckman selection model Exclusion restriction: International quality certification (ISO 9000, ISO 1400) Model is estimated only on domestic firms. Exclude government owned enterprises. Small sample size: Only include industries with more than 5 firms Empirical Strategy Data and empirical methodology
13 Spillover Measures Horizontal ◦ Hor_emp jkt ◦ Hor_exp jkt Vertical ◦ Share of output sold to MNE within the country (2002, 2005) ◦ Share of inputs of foreign origin Data and empirical methodology
14 Control variables Industry level controls ◦ Industry share of total country exports ◦ Industry share of total country employment Standard firm-level controls ◦ Labour productivity ◦ Employment ◦ Age Country, year, industry dummies Data and empirical methodology
15 Variable Mean non- exporters Mean exportersDifference Log of employment *** Log of labour productivity *** Age *** MNE Sales *** Foreign Input *** Quality certificate *** *** p<0.01, ** p<0.05, * p<0.1
16 Industry level Data and empirical methodology
17 Excluding vertical linkagesIncluding vertical linkages ParticipationPerformanceParticipationPerformance (1) (2)(3) (4) Hor_emp *** * Hor_exp *** *** Mnesales0.617***0.423* Foreign input0.518***0.337* Expshare_ind0.979***0.724** * Empshare_ind ***0.277 log employment0.516***1.404***0.416***1.495*** log employment *** ** log labour prod *** *** log labour prod ** Age *** Quality certificate0.304***0.321*** Number obs Uncensored obs Log Likelihood ρ 0.215*** 0.317*** Results: All Firms *** p<0.01, ** p<0.05, * p<0.1 Clustered adjusted standard errors in parentheses all regressions include country, year and industry dummies. Wald test rejects independence between participation and performance equation Constant not shown
18 Results: Small Firms *** p<0.01, ** p<0.05, * p<0.1 Clustered adjusted standard errors in parentheses all regressions include country, year and industry dummies. Wald test rejects independence between participation and performance equation Constant not shown Excluding vertical linkagesIncluding vertical linkages ParticipationPerformanceParticipationPerformance (1) (2)(3) (4) Hor_emp Hor_exp Mnesales0.724***1.220*** Foreign input0.400***0.642*** Expshare_ind0.798**1.456** Empshare_ind ***1.789 log employment *** ** log employment log labour prod ***-1.658***1.577*** log labour prod *** Age ** ** Quality certificate0.298***0.469*** Number obs Uncensored obs Log Likelihood ρ 0.327** 0.453*
19 Excluding vertical linkagesIncluding vertical linkages ParticipationPerformanceParticipationPerformance Hor_emp *** Hor_exp-0.393**-0.940***-0.790***-1.024*** Mnesales0.550*0.234 Foreign input0.697***0.29 Expshare_ind1.132*** Empshare_ind *1.08 log employment0.816***1.481***0.987***1.930*** log employment ** * log labour prod.0.355**0.844** log labour prod Age *** Quality certificate0.327***0.343*** Number obs Uncensored obs Log Likelihood ρ 0.153*** 0.372*** Results: Larger Firms *** p<0.01, ** p<0.05, * p<0.1 Clustered adjusted standard errors in parentheses all regressions include country, year and industry dummies. Wald test rejects independence between participation and performance equation Constant not shown
20 Conclusion Effect of MNEs on domestic export performance is complex Different effects of horizontal and vertical spillovers Vertical linkages positively affects both export participation and intensity. Underlines importance of local linkages. Conclusion