The European Commission s science and knowledge service Joint Research Centre ICT use, innovation and employment growth in Gazelles 7 th IRIMA Workshop Innovation, Employment, Firm Growth and Job Creation Federico Biagi and Martin Falk, Bruxelles, 28 June 2016 1
Motivation Aim: explore whether ICT use has a positive impact on employment growth for Gazelles (various indicators are used). We also explore the relationship between employment growth and process/product innovation. Contribution Use of internationally comparable data based on linked Structural Business Survey (SBS) and ICT Use by Enterprises Survey (ICT Survey) and Community Innovation Survey (CIS) for 10 European countries for 2002-2010 Outline: Data Stylized facts on Gazelles Determinants of employment growth of Gazelles Role of ICT and innovation 2
Motivation Firm level is the preferred aggregation level to study the impacts of ICT use, however: a) limitations because of cross-section nature of ICT Survey and CIS Survey (rotating design); b) access to firm level data in several EU countries is restricted. Advantages of micro aggregated data (ESSLait Micro Moments Database) Panel data Linked with other surveys can be easily conducted Large number of innovation and ICT/E-commerce indicators (unlike in the EU Klems/WIOD database) 3
Data http://ec.europa.eu/eurostat/web/microdata/micro-moments-dataset 4
Data Distributed Microdata (DMD) protocol with software Common Code accesses, links and aggregates otherwise not accessible microdata, Bartelsman et al (2004), Eurostat (2008, 2012, 2013). 1 2 3 CIS ESSLait Micro Moments Database (MMD) 5 5
Data ESSLAIT micro moments (MMD) database (Bartelsman et al. 2014) Disaggregated data for 14 European countries for the period 2000-2010 (AT, DE, DK, FI, FR, IE, IT, LU, NL, NO, PL, SE, SI and UK), but only 2 countries available at the Safe centre (EUROSTAT). Structural Business Statistics, Business register (age), CIS and ICT/Ecommerce survey Data available for several disaggregated groups Broad industry groups and gazelles/broad industry groups and high growth firms/two digit industry data Data from 2008 onwards converted from NACE 2 to NACE 1.1 Restricted to firms with 10 or more employees No innovation data for DE; no two digit industry level data for LU. We use DK, FI, FR, IE, IT, NL, NO, SE, SI, UK 6 6
Stylized Facts Percentage of Gazelle firms (10 percent growth per year on average over 3yrs and less than 5 yrs old in year 3) is close to about 5 percent in the EU-10 countries Highest in Business Services and Personal Services and lowest in Manufacturing Percentage of Gazelles is relatively stable over time 7
Share of Gazelles 12 Market services: FinBU, Pers., Distr. 10 8 6 4 2 percentage of gazelle firms NL UK DK FI EU - 10 SE IT FR 0 2005 2006 2007 2008 2009 2010 Notes: EU-10 includes DK, FI, FR, IE, IT, NL, NO, SE, SI, UK. Source ESSLAIT micromoments database 8
Share of Gazelles 8 Manufacturing: ConsG, IntmdG, InvesG, Other: excl. Elecom 7 6 5 4 3 2 1 IT FI NL UK DK EU - 10 SE FR 0 2005 2006 2007 2008 2009 2010 Source ESSLAIT micro moments database 9
Share of Gazelles 12 10 8 6 4 2 0 Elecom: Electrical machinery, post and communication DK 2005 2006 2007 2008 2009 2010 UK NL EU - 10 SE FI IT FR Source ESSLAIT micro moments database 10
Percentage and Employment Share of Gazelles 16 14 12 14.8 Importance of gazelles in EU-10 in 2010 percentage of gazelles 12.3 employment share of gazelle firms 10 8 6 4 8.1 6.0 5.5 7.1 7.4 4.8 4.5 6.1 7.3 4.6 3.9 3.8 2 0 Source ESSLAIT micro moments database 11
Empirical model log linear labour demand equation (Van Reenen 1997) applied to gazelle firms ln L ic t = + b ic K + b ic t WP + 1 2 ln ic t 3 b b l + e i=industry, c=country, t=time L employment (headcounts or full time equivalents) K capital stock in constant prices WP real wage l set of variables capturing ICT use/diffusion and innovation ß ic fixed (group) effect Estimation in first differences Lict = ic + b1 ln Kict + b2 lnwpict + 1 ICTict + 1INNO ict + b3 Estimation by robust regression method ic t ~ ~ ~ ln DYR + u ict 12
Empirical model Variables used to capture ICT use: 1) Share of broadband enabled workers (infrastructure but also enabling process and organizational innovation) 2) Share of firms sharing electronic data internally and with providers and distributors (enabling process and organizational innovation) 3) Share of firms selling through the internet (ecommerce variable) 4) Share of firms purchasing through the internet (epurchasing variable) 13
Empirical model Variables used to capture innovation activities 1) Share of firms introducing a product innovation new to the market (higher innovation) 2) Share of firms introducing an improved product or a product that is new to the firm but not to the market (lower innovation) 3) Share of firms introducing process innovation 14
Results First difference specification estimated by robust regression method. E-commerce activities and broadband internet enabled employees are significantly related to employment growth of Gazelles. So are variables related to ecommerce and, especially epurchasing. Coefficient of 0.09 for share of broadband enabled employees =>10 percent increase in the latter leads to an increase in the employment growth rate of Gazelles by 1 approximately percentage point (careful on causality!) Relative wages have the expected negative sign and are highly significant. 15
Results Capital accumulation highly relevant for Gazelles too (stands in contrast to our own earlier work for all firms: evidence of embodied technological change?). Product innovations are significant (more so when we use the share of Gazelles introducing products that are new to the market, relative to when we use the share of Gazelles introducing improved version of already existing products or products that are new to the firm but not to the market). Process innovation are not significantly related to employment growth. ICT and ecommerce variables appear more relevant than classical innovation variables (but ICT variables are often enablers of innovation as well). 16
Results Estimation results of the labour demand equation Dep. var: employment growth of gazelles Robust regression method Coef. t Coef. t lnwp -0.76 *** -11.36-0.81 *** -11.89 lnk 0.12 *** 2.72 0.09 * 1.79 # broadband enabled workers t-1 0.09 ** 2.12 # firms sharing Electronic Data t-1 0.13 ** 2.28 constant 0.09 *** 3.23 0.06 1.09 year effects yes yes # obs 186 128,** and *** are statistically significant at the 10%, 5% and 1% level, respectively. OLS estimates based on heteroscedasticity consistent standard errors. The sample consists of data for 10 European countries (DK, FI, FR, IE, IT, NL, NO, SE, SI and UK) for the period 2006-2010 17
Results Estimation results of the labour demand equation Dep. var: employment growth of gazelles Robust regression method Coef. t Coef. t lnwp -0.77 *** -11.57-0.75 *** -11.42 lnk 0.11 ** 2.51 0.11 *** 2.71 # firms selling via the internet t-1 0.12 * 1.95 e-purchases: share of orders through internet t-1 0.31 *** 3.09 constant -0.03-1.80 0.12 *** 5.79 year effects yes yes # obs 184 186,** and *** are statistically significant at the 10%, 5% and 1% level, respectively. OLS estimates based on heteroscedasticity consistent standard errors. The sample consists of data for 10 European countries (DK, FI, FR, IE, IT, NL, NO, SE, SI and UK) for the period 2006-2010 18
Results Estimation results of the labour demand equation Dep. var: employment growth of gazelles Robust regression method Coef. t Coef. t lnwp -0.77*** -11.24-0.77*** -11.26 lnk 0.14 *** 3.00 0.14 *** 2.97 # firms with new market products t-1 0.13 * 1.75 # firms with new & improved products t-1 0.08 1.34 constant -0.06** -2.39-0.06** -2.11 year effects yes yes # obs 185,** and *** are statistically significant at the 10%, 5% and 1% level, respectively. OLS estimates based on heteroscedasticity consistent standard errors. The sample consists of data for 10 European countries (DK, FI, FR, IE, IT, the NL, NO, SE, SI and UK) for the period 2006-2010 19
Conclusions First empirical evidence on the employment effects of ICT diffusion and innovations. Novel (panel) data set of internationally comparable industry level data for 10 European countries. Positive employment growth estimates for ICT and ecommerce activities Positive employment growth estimates also for the share of firms introducing new-to-the-market products. For Gazelles ICT and ecommerce variables appear more relevant than classical technological innovations Future work IV variable method to account for the endogeneity of ICT Estimating the demand for workers with higher education (tertiary degree) for the group of Gazelles 20