What affects MFP in the long-run? Evidence from Canadian industries Danny Leung and Yi Zheng Bank of Canada, Research Department Structural Studies May 2008 The views expressed in this presentation are those of the authors. No responsibility should be attributed to the Bank of Canada.
March Motivation Changes in labour productivity growth driven by changes in MFP Recent interest in ICT as a determinant of MFP, results mixed Sensitive to assumptions on timing Difficulty in accounting for other variables Two above issues are not exclusive to ICT
March Contribution Evaluate possible determinants of MFP using Canadian industry-level data Evaluate possible determinants of MFP using Canadian industry-level data Adopt an econometric methodology that allows the estimation of the long-run effects of the determinants of MFP Adopt an econometric methodology that allows the estimation of the long-run effects of the determinants of MFP Examine the main determinants of MFP in the same framework Examine the main determinants of MFP in the same framework
March Preview of Results In the long-run, ICT capital services seem to have a small positive effect on MFP In the long-run, ICT capital services seem to have a small positive effect on MFP In recent years, the impact of ICT has been stronger, but not instantaneous In recent years, the impact of ICT has been stronger, but not instantaneous Outsourcing and trade openness also influence MFP positively Outsourcing and trade openness also influence MFP positively The impact of competition is unexpectedly negative The impact of competition is unexpectedly negative
March Variables Considered ICT – general purpose technology, facilitates organizational change that raises MFP – innovation, imitation, spillover; generally not fully accounted for in measured MFP R&D – innovation, imitation, spillover; generally not fully accounted for in measured MFP FDI – efficient conduit of technological transfer FDI – efficient conduit of technological transfer Trade – tech. transfer, competition, economies of scale through expanded markets Trade – tech. transfer, competition, economies of scale through expanded markets Outsourcing & offshoring – specialization, economies of scale, firm turnover, restructuring Outsourcing & offshoring – specialization, economies of scale, firm turnover, restructuring Competition – its relation with innovation is ambiguous Competition – its relation with innovation is ambiguous Public infrastructure – “omitted” capital Public infrastructure – “omitted” capital
March The Model (1) ARDL ARDL Reparameterized to ECM Reparameterized to ECM
March The Model (2) The Pooled Mean Group Estimator The Pooled Mean Group Estimator Pesaran et al (1999) Pesaran et al (1999) Constrains long-run parameters to be identical, but allows short-run coefficients and error variances to differ across groups Constrains long-run parameters to be identical, but allows short-run coefficients and error variances to differ across groups Traditional dynamic fixed effect model restricts all coefficients to be the same Traditional dynamic fixed effect model restricts all coefficients to be the same Estimating each equation separately may not be feasible given sample size Estimating each equation separately may not be feasible given sample size
March The Data (1) Twelve 2-digit NAICS industries (covering all of business sector), Twelve 2-digit NAICS industries (covering all of business sector), MFP – from Canadian productivity accounts (CPA) MFP – from Canadian productivity accounts (CPA) ICT capital services, CPA ICT capital services, CPA R&D measures ( ), various sources R&D measures ( ), various sources Own-industry R&D stock (RD) Own-industry R&D stock (RD) R&D spill-over; weighted R&D stock of suppliers (SRD) R&D spill-over; weighted R&D stock of suppliers (SRD) R&D intensity (RDI) R&D intensity (RDI)
March The Data (2) Public infrastructure (infra_g) or mass infrastructure (infra_m) - Stock of engineering capital owned by gov’t or infra_g plus engineering capital of transportation and utilities sectors, CANSIM Public infrastructure (infra_g) or mass infrastructure (infra_m) - Stock of engineering capital owned by gov’t or infra_g plus engineering capital of transportation and utilities sectors, CANSIM Outsourcing – intermediate inputs costs over nominal gross output, CPA Outsourcing – intermediate inputs costs over nominal gross output, CPA Global trade openness – nominal world imports and exports over world GDP, IMF Global trade openness – nominal world imports and exports over world GDP, IMF Markup Markup price over average variable cost –nominal output over labour, energy and materials costs, CPA price over average variable cost –nominal output over labour, energy and materials costs, CPA price over average cost – from production function estimation with unobserved component, Leung (2008) price over average cost – from production function estimation with unobserved component, Leung (2008)
March ICT
March Own R&D capital
March Spill-over R&D capital
March Public infrastructure
March Mass infrastructure
March Outsourcing
March Trade openness
March Markup (P/AVC)
March Unit Root Test (1) Im, Pesaran and Shin (2003), ADF Im, Pesaran and Shin (2003), ADF H0: series from each industry contain unit root H0: series from each industry contain unit root Ha: fraction of series are stationary with heterogenous coefficients Ha: fraction of series are stationary with heterogenous coefficients Hadri (2000) Hadri (2000) H0: series are stationary H0: series are stationary Pesaran (2003), CADF Pesaran (2003), CADF IPS test that relaxes assumption of cross- sectional independence between industries IPS test that relaxes assumption of cross- sectional independence between industries
March Unit Root Test (2)
March Unit Root Test (3)
March Cointegration (1) Pedroni (1999, 2004), residual-based tests with null of no cointegration Pedroni (1999, 2004), residual-based tests with null of no cointegration four “panel” tests with alternative that residuals have homogeneous autoregressive coefficient four “panel” tests with alternative that residuals have homogeneous autoregressive coefficient three “group” tests with alternative without homogeneity assumption three “group” tests with alternative without homogeneity assumption General-to-specific approach General-to-specific approach Variables dropped one at a time from relationship to see if null is still rejected Variables dropped one at a time from relationship to see if null is still rejected Most parsimonious specification includes MFP, ICT capital, outsourcing, trade openness, and markup Most parsimonious specification includes MFP, ICT capital, outsourcing, trade openness, and markup
March Cointegration (2)
March PMG Results – Long Run (1)
March PMG Results – Long Run (2) Impact of ICT is small Impact of ICT is small Effect of outsourcing is positive Effect of outsourcing is positive Effect of trade openness is positive Effect of trade openness is positive Higher competition (lower markup) has a negative effect on MFP Higher competition (lower markup) has a negative effect on MFP Inverted U-shaped relation between competition and innovation Inverted U-shaped relation between competition and innovation
March PMG Results – Long Run (3) Resonable fit Resonable fit No evidence of non-normal errors, Jarque- Bera test No evidence of non-normal errors, Jarque- Bera test Two of twelve industries exhibit heteroskedasticity, Breush-Pagan test Two of twelve industries exhibit heteroskedasticity, Breush-Pagan test Breush-Godfrey test of serial correlation and RESET test indicate a possibility of missing variables for certain industries Breush-Godfrey test of serial correlation and RESET test indicate a possibility of missing variables for certain industries
March PMG Results – Short Run Speed of adjustment is -0.4 on average Speed of adjustment is -0.4 on average 95% of a deviation from long-run equilibrium corrected in 5 years 95% of a deviation from long-run equilibrium corrected in 5 years Average coefficients on current and lagged growth rates of ICT are negative, but statistically insignificant Average coefficients on current and lagged growth rates of ICT are negative, but statistically insignificant Sum of short-run ICT coefficients is negative and significant in retail trade and manufacturing Sum of short-run ICT coefficients is negative and significant in retail trade and manufacturing
March Dynamic Decomposition (1) It is possible to decompose MFP growth into contributions by factor It is possible to decompose MFP growth into contributions by factor assuming speed adjustment is the same for all factors and assuming equilibrium in the initial period assuming speed adjustment is the same for all factors and assuming equilibrium in the initial period
March Dynamic Decomposition (2)
March Concluding Remarks The long-run impact of ICT is small, but its contribution to recent MFP growth in some industries is large The long-run impact of ICT is small, but its contribution to recent MFP growth in some industries is large Outsourcing and global trade openness have a positive effect on MFP in the long- run Outsourcing and global trade openness have a positive effect on MFP in the long- run The effect of competition in MFP is negative The effect of competition in MFP is negative