Final Report
Total Page:16
File Type:pdf, Size:1020Kb
FINAL REPORT Information Society: ICT impact assessment by linking data from different sources Grant Agreement Number – 49102.2005.017-2006.128 August 2008 Authors Aarno Airaksinen Statistics Finland Andrea de Panizza ISTAT, Italy and European Commission Eric Bartelsman (Professor) Vrije Universiteit Amsterdam, Tinbergen Institute and IZA Eva Hagsten Statistics Sweden George van Leeuwen Statistics Netherlands Mark Franklin Office for National Statistics, UK Mika Maliranta Statistics Finland & The Research Institute of the Finnish Economy (ETLA) Patricia Kotnik University of Ljubljana, Faculty of Economics Peter Stam Office for National Statistics, UK Petri Rouvinen The Research Institute of the Finnish Economy (ETLA) Shikeb Farooqui Office for National Statistics, UK Simon Quantin INSEE, France Stefan Svanberg Statistics Sweden Tony Clayton Office for National Statistics, UK Yoann Barbesol INSEE, France With thanks and acknowledgement to Allessandra Nurra ISTAT, Italy Brian Ring CSO Ireland Chiara Criscuolo London School of Economics, Centre for Economic Performance Eugene van der Pijll Statistics Netherlands Gerolamo Giungato ISTAT, Italy Hans-Olof Hagen Statistics Sweden Jonathan Haskel (Professor) Queen Mary University of London Joseph Robjohns Office for National Statistics, UK Martin Lundo Statistics Denmark Martin Mana Czech Statistical Office Natalia Cherevichenko Statistics Denmark Nina Djahangiri Statistics Austria Ole-Petter Kordahl Statistics Norway Oliver Bauer Federal Statistics Office, Germany Ritchie McMahon CSO Ireland Stefan Bender Institute for Employment Research Vaclav Kosina Czech Statistical Office We would like to thank all those who have supported and contributed to this project, not least the Information Society team at Eurostat. However, any errors or omissions are the responsibility of the authors. Disclaimer All data in this report comply with statistical disclosure measures and standards throughout all project member countries. Furthermore, data refer to experimental and provisional research datasets. They should not be treated as - or compared to - any official national statistics. Eurostat Agreement No. 49102.2005.017-2006.128 Contents Chapter Page 1. Recommendations, Summary Results and Conclusions 1 Tony Clayton 2. Introduction and Background 20 Tony Clayton 3. Methods and Data Sources 29 Mark Franklin 4. Describing the Data 40 Peter Stam 5. Properties of Linked Data Evidence from the ICT Impacts Project 68 Eric Bartelsman 6. Productivity and Core ICT Metrics at Firm Level 94 Mark Franklin Shikeb Farooqui 7. ICT Characteristics of Fast Growing Firms 120 Simon Quantin Yoann Barbesol 8. Employment, Skills and Information Technology 134 Eva Hagsten Patricia Kotnik 9. ICT Business Integration 149 Mark Franklin Tony Clayton 10. ICT Investment and Productivity 163 George van Leeuwen Shikeb Farooqui 11. Offshoring 190 Andrea de Panizza Eva Hagsten Patricia Kotnik Simon Quantin Stefan Svanberg Yoann Barbesol 12. ICT, Innovation and Productivity 222 George van Leeuwen Shikeb Farooqui 13. IT Outsourcing in Finnish Business 240 Aarno Airaksinen Mika Maliranta Petri Rouvinen 14. From Micro to Macro 255 Eric Bartelsman References 272 Appendix I – Project User guide 280 Peter Stam Eurostat Agreement No. 49102.2005.017-2006.128 Chapter 1 Chapter 1 Summary Results, Conclusions and Recommendations Tony Clayton UK Office for National Statistics This chapter summarises the main analytical results of the project, uses them to draw conclusions from the new findings on Information and Communications Technology (ICT) impacts, alongside prior research, and sets out recommendations for action by Eurostat and NSIs. Where possible the full analytical evidence in subsequent chapters is referenced. Section 1.1 Summary of Results This section brings together the analytical findings from the project. The range of work covers all the milestones set for the original contract (see Appendix A), and the development of new methodology. The results here derive from the following different types of analytical work: • at the most basic level, one-off studies of firm level productivity impact where only one country has the data to perform a specific piece of analysis (for example work from Finland on ICT outsourcing). • several examples of groups of countries collaborating on micro data analysis for topics where all have similar (but not necessarily identical) firm level data which enable a common analytical framework to be used and compared (eg Netherlands and UK on ICT investment, Sweden, France and Italy on offshoring, Sweden, Netherlands and UK on innovation). • an encouraging range of firm level analysis using common metrics and common analytical code with identical data sources, either carried out by local researchers across most countries direct from local datasets, or using the data created for the project and centrally written code to run identical regression analysis, for all countries in the project except Denmark and Slovenia. • construction of an extensive ‘metadata warehouse’, which is used to weight and aggregate ICT use, structural business and business register data from surveys in all 13 countries in as comparable a way as is possible, producing distributed microdata datasets (DMD); the aggregation process operates to produce estimates of complex indicators (constructed from more than one variable from a survey) as well as indicators which depend on intersections between surveys; this metadata system can also be used to generate datasets on a highly comparable basis for firm level regression analysis within countries. • industry / country level analysis of ICT impacts, using the large (and still under-explored) dataset produced by the distributed microdata (DMD) analysis system, where we have a very high level of confidence in the comparability of indicators, and on the ability to draw reliable comparisons between industries /countries and over time. The summary below presents results in terms of the key topics tackled in the study, and draws on evidence from both firm level and DMD analysis as appropriate. 1 Eurostat Agreement No. 49102.2005.017-2006.128 Chapter 1 1.1.1 Results from firm level analysis 1.1.1.1 Coverage (Chapters 4, 5 and 6) All 13 countries participating in the study (including Denmark which had to withdraw in the late stages due to lack of personnel to undertake analysis and check results) have succeeded in producing regression and / or correlation results from firm level data, either individually or using the DMD analysis methodology developed in the project – and in most cases both. Project participants have completed analysis to show results relating ICT use at firm level to labour productivity from 11 of the 13 participating countries, on common metrics using an exactly comparable method, and using the common metadata to define and link variables. We know we can get results from the remaining countries with minor additional resources; the missing micro-data analysis is due to analytical resource constraints, or to limitations under which access to data was available for this particular project. Analysis of the properties of linked datasets in the project, using methodologies developed in earlier studies, shows that sample reweighting, using metadata and methods included in the project, is capable of dealing with most issues of ‘representativeness’ of data. This breaks down in cases where overlap between datasets is inadequate, and we have not advocated modelling in such cases. Linking of datasets in many countries, using sampling designs currently in use, leaves the overlap between ICT surveys and firm performance surveys heavily biased towards larger firms. This affects both firm level analysis and the DMD analysis. For impact conclusions adequately to reflect small firms, sampling strategies would need to change. 1.1.1.2 Common firm level analysis across all NSIs (Chapter 6) The core ICT use metrics used in the project (computer use, e-sales, e-purchases, fast internet enabled or using employees) show reasonably consistent, positive, labour productivity effects at firm level across manufacturing industries in all countries in the project, beyond the six which have been covered by earlier studies. This suggests that productivity impacts related to use of ICT in manufacturing are now relatively well established and transferable across countries within the EU. The same core ICT use metrics have much more varied relationships with labour productivity across services at firm level in different countries; for the UK, France, Nordic countries and Netherlands, positive correlations seen in prior studies, and reported in early work from this project are confirmed, in other countries participating in the project, productivity effects are insignificant or even, in one or two cases, negative. There seems to be at least some correlation between the countries (Nordic states, Netherlands, UK, France) where ICT use by firms is relatively more intensive and communications infrastructure is strong, or where there is greater market flexibility / dynamism, and the strength of the statistical relationship between ICT use and firm level productivity in services. These differences in impact for services could be explained by a number of factors, including: • differences in competitive conditions in national services markets, and / or • productivity gains requiring ‘critical mass’ in networks and ICT use, and / or • measurement difficulties in services which are better tackled in some states. The common analysis