The value of using microdata and microdata linking to investigate innovation impacts

Transcription

1 The value of using microdata and microdata linking to investigate innovation impacts Dirk Czarnitzki, K.U.Leuven, Belgium Joint NESTI-TIP Workshop on Innovation Indicators for Policy Making and Impact Assessment Paris, June 12th, 2007

2 Keywords collected from text on workshop s background Does policy ask the right questions? Indicators and analysis on contribution of R&D/innovation to economic growth Indicators for effectiveness of innovation policies Links between innovation performance and policies Commercialization of public research Scope and extent of open/collaborative innovation Cross-country studies

3 Indicators Building comparable indicators across countries is important BUT indicators are never evidence just pointers. Economic modelling is needed to understand the underlying processes (Mr. Evans, Keynote this morning) aim of linking micro-databases Indicators are just input for multivariate analyzes identifying causal relationships Main drawbacks: - Comparability over countries, data availability - No easy construction of indicators

4 Structure of Talk Why linking databases? Linking innovation and R&D surveys over time Examples of linked European innovation databases - Steunpunt O&O Indicatoren, K.U.Leuven - Mannheim Innovation Panel, ZEW Germany - Examples: What has been done? Comment on cross-country studies Example on tracing technology transfer from science to industry Limitations of current research

5 Why linking databases? European CIS/R&D surveys collect detailed data about innovation and R&D, but typical weakness of surveys: - surveys often have no indicators on capital or access to credit markets in general; other business strategies (e.g. marketing) - cross-sectional structure of surveys gives rise to simultaneity problems in (multivariate) analyzes (direction of causalities?) - cross-sectional analysis does not allow to control for unobserved heterogeneity, such as risk preferences, managerial structure, etc.

6 Panel databases: Linking survey data over time

7 Long-term development of R&D in Flemish business sector R&D employment R&D personnel R&D expenses (constant prices of 1995) R&D expenses in Mio. EUR

8 Linking survey data over time Extrapolated data give a macroeconomic number, but are estimates! Publication of only point estimates is misleading; standard errors should be calculated Making use of panel structure: - Test at the firm level: H 0 = R&D(i,t) / R&D(i,t-1) = 1? - Can be split by firm size classes or sectors to see where growth/decline originates. - More controls can be added

9 Examples of linked European innovation databases

10 ZEW Innovation Surveys and their links PROFI database from German Federal Ministry for Education and Research (BMBF) - all federally funded civilian R&D/innovation projects since 1980 German and EPO patent database - Link by patent assignee for all years since Links are updated on irregular basis Mannheim Innovation Panel - CIS survey annually since Designed as panel database - link is updated annually Creditreform database FDI data (German Central Bank) - FDI since No regular update - largest credit rating agency in Germany - covers (relevant) population - credit rating index and many other infos (foundation year, capital structure, ownership, etc.) - Updated twice per year

11 IWT ICAROS database - all publicly funded R&D/innovation projects since 1992 by Flemish Government - link is updated regularly EPO patent database - Link by patent assignee for all years since Links are updated regularly Data at Steunpunt O&O Indicatoren CIS Surveys - Designed as panel since 2001 BELFIRST database (Bureau van Dijk) - Balance sheet data - covers (relevant) population - Updated annually R&D surveys - panel as they covery inventory of R&D performers Note: Population databases can be used to calculate further controls, e.g. regional indicators (competition, agglomeration/network effects etc.), or technological indicators from patents etc. Under construction: capture national benefits via patent applications (group structure, assignee vs. inventor); and maybe R&D tax credits

12 What has been done? Some examples Use Panel structure to analyze persistence of innovation B. Peters, ZEW; Mohnen/Raymond et al., MERIT Do subsidies trigger additional R&D investment or does crowding-out happen? - Use linked database of subsidies and CIS Almus/Czarnitzki (2003); Hussinger (2007) Why do we observe different levels of R&D (intensities) across countries, regions, sectors, firm sizes? - Maybe due to financial constraints? Subsidy levels? - Interaction of credit market restrictions and subsidies - Use linked data on credit rating, subsidies and CIS Czarnitzki (2006), Binz/Czarnitzki (2007) Export, FDI and Productivity Growth - CIS and FDI data of German Central Bank Arnold and Hussinger (2006a, 2006b)

13 What has been done? Some examples Is additionally stimulated R&D productive or does it end up in crowding-out private investment or just higher wages for R&D personnel only? Use link of CIS with patent database and subsidies - Step 1: Treatment effects analysis on R&D investment - Step 2: Disentangle R&D into purely privately financed and publicly stimulated component - Step 3: Use patents as output measure both R&D components have positive effect on output See Czarnitzki/Hussinger (2004); Czarnitzki/Licht (2006)

14 Comment on cross-country studies

15 Example: R&D and the response to collaboration and subsidies in Germany and Finland Apply heterogeneous treatment effects analysis to (Western) Germany and Finland (CIS III) Treatments: Subsidy receipt and engagement in collaboration Question: Do treatments increase R&D investment and output?

16 Example: R&D and its response to collaboration and subsidies in Germany and Finland What would a firm have done if it had not been treated? - Both subsidies and collaboration increase R&D investment Could R&D and innovation output be further extended through subsidies and collaboration? - Interesting result: In Germany higher potential than in Finland - but why? - Funding and collaboration patterns differ in Finland and in Germany? - Or level of funding and collaboration are different in countries (decreasing returns to scale?) - Source: Czarnitzki/Ebersberger/Fier (2007)

17 Example on tracing technology transfer from science to industry

18 US SBIR database: Academic Entrepreneurship (Toole/Czarnitzki) CRISP database (National Institutes of Health) - Research Grants Biomedical Research - Amount of Grant, Type, Field - (Public) Institution - Principal Investigator Venture Xpert - Venture Capital at firm level SBA database - All SBIR projects since Firm name, address - Amount of Grant, Type USPTO database - Patents since Assignee - Inventor Further data: Publications per researcher, Ph.D. year and field, quality of institution, etc.

19 Limitations of current research

20 Limitations Difficult to assess why country differences may be found In policy evaluation, quantitative analyzes have so far not been able to differentiate among programs and to advice policy on best design of programs Output often only patents, but narrow measure. What about products, processes? (see Canadian Innovation Surveys => but cross-sectional structure) - Problem: different time lags from invention to market (e.g. pharma vs. IT service)