R&D and ICT R&D in Rhomolo

Transcription

1 R&D and ICT R&D in Rhomolo Ben Gardiner and Wojtek Szewczyk Workshop on: Modelling the economic impact of EU ICT R&D Expenditures Seville, 16th April 2012 The views expressed are purely those of the authors and may not in any circumstances be regarded as stating an official position of the European Commission.

2 Outline Context and background Model description R&D scenarios Plans relating to (ICT) R&D Conclusions

3 Context and background Previous modelling exercises of impact of EU R&D policies partial equilibrium studies extensive use of HERMIN and QUEST structural models Court of Auditors (2006) recommendations it is important to strengthen the capacity of the Commission to oversee the evaluation process, including the application of complex macro-econometric models if a macro-economic model is used it must take proper account of the characteristics of the economies under review Barca report (2009) refocus on core priorities re-emphasise place-based policy (vs spatially blind debate)

4 Context and background Demand for sub-national analysis implications of recent theoretical developments and their empirical applications (ie New Economic Geography) consideration of spatial heterogeneity in impact of EC policies (+ spillovers) Aims of this communication: to show regional heterogeneity in the impact of alternative EC policies (even within countries) to assess the size of spillovers across regions of investments to show that RHOMOLO can account for these features, and thus that it complements evidence from other macro-models

5 Model description (brief) RHOMOLO Computable General Equilibrium Model - A complex Holistic Model with different dimensions addressed simultaneously - Aimed at studying the impact of Cohesion Policies on regional economic growth. - Grounded on: New economic geography (Krugman 1991) Endogenous growth theory (Romer 1990) General equilibrium theory (Arrow & Debreu 1954) - 2 versions (so far) of the Model: -Prototype (5 Countries, 23 sectors); Version 1 (EU27, 6 sectors)

6 RHOMOLO vs alternative models Integrated Modelling Theory vs Data-Driven 100 (Internal consistency of the model) Regional Disaggregation (Degree of coverage of economic, social and environmental indicators) (100% = full NUTS2 coverage) Sectoral Disaggregation (100% = full NACE 2-digit coverage, ie 60 sectors) QUEST HERMIN RHOMOLO (prototype) RHOMOLO (v1 EU27)

7 Treatment of R&D Prototype Model R&D and TFP: gtfp = a + a LnH + a RD + a LnH TFP + e where: rs, rs, 0 1 r 2 r 3 r * rs, TFP rs, H (Human Capital) average years of schooling RD (R&D expenditures/gva) R&D intensity. Total intramural R&D expenditure (GERD) by sectors of performance and region Interaction term H*TFPGAP, proxies for technology flows absorbed through region-specific human capital R&D sector (NACE 72.) produces output using L, K, and I RDr endogenous to the model (driven by growth of R&D sector)

8 Treatment of R&D Version 1 R&D and TFP: gtfp differences: r TFP β + ε r = 0 + β1rdr + β2 * TFPr Human Capital effect removed Sectoral dimension removed Interaction term replaced by pure catch-up (leader vs follower) term r No R&D sector any longer R&D enters the model exogenously

9 Impact assessment: R&D scenarios (using Prototype (5-country) model) Scenario 1: Full shock to ALL region simultaneously so as to reach the R&D intensity target by Scenario 2: Shock to specific leader region in each country. Scenario 3: a) Shock to R&D intensive (RDI) regions only b) Shock to R&D non-intensive (RDNI) regions only

10 Summary of Impact assessment Scenario 1: Positive impact of increasing R&D intensity on real GDP growth, although size of impact differs across countries The dispersion of regional GDP growth varies across countries, with Germany showing the lowest dispersion Results show an asymmetric response on unemployment rates. Scenario 2: The positive shock to the leader spills over the other regions. The channels are TFP catch-up, Trade, and Migration Stronger spillover effects in economies with stronger interactions across regions (Germany vs Poland), and that are prone to adopt innovations (eg human capital composition). Scenario 3: Shocking the R&D non-intensive regions seems to lead to stronger spillover effects than when the intensive R&D regions are shocked The overall country effect is also stronger when the shock is applied to the less intensive R&D regions

11 Summary so far... RHOMOLO is a new regional CGE model for the analysis of EC policies (principally Cohesion Policy) The model structure allows spatial effects, interactions and spillovers to be identified across the EU27 R&D plays a role in the TFP equation, and for now remains both exogenous and not separated into an ICT vs non-ict component

12 Extending to ICT How Rhomolo can be used to simulate impact of ICT R&D policies, such as Digital Agenda for Europe with target of doubling expenditure on ICT R&D by 2020? 1. Extend / modify the model s structure so it would have explicit representation of instruments considered (e.g. GBAORD) 2. Undertake calculations outside Rhomolo to create intermediate link between the policy instruments and the existing Rhomolo structure no modifications to the model. page 12

13 Overview of transmission mechanism ICT GBAORD r historical data ICT GBAORD r baseline 8% by 2020 POLICY r 2x GBAORD x pub-priv elast. ICT BERD by reg/sect until 2020 ICT BERD by reg/sect until 2020 Reallocation only EU2020 (GERD=3%GDP by 2020) BERD r,s until 2020 Rhomolo TFP r,s =f(r&d r,s )

14 GBAORD projections DAE specifies only the EU target for doubling public ICT R&D. There is no relevant country specific targets. Hence, the treatment: ICT ( EU 20 ) φdistr grdr = RDr S + r α grd ICT -growth rate in intensity of ICT GAORD RD EU20 -growth rate in public R&D expenditure as determined by EU2020 targets DIST r S r -distance of country r from ICT R&D leader (calculated as difference in ICT R&D intensities) -is slope of the fitted linear growth for country r for years Alpha and theta estimated to achieve the policy targets

15

16 BERD projections lnrd = c + β + α lnrd + α lngrd + α lntg + μ ICT noict ICT rt, r 1 rt, 2 rt, 3 rt, 2 rt, RD ICT BERD on ICT RD noict BERD on non-ict GRD ICT public spending on ICT R&D TG availability of ICT skilled graduates (enrollments into ISC460 and ISC48 (Mathematics, Statistics and Computing), lagged by 7 years Calibrated on data

17 Some estimates and results (policy)

18 year itrdp, country == MT itrdp, country == CY itrdp, country == SI itrdp, country == LV itrdp, country == SK itrdp, country == HU year itrdp, country == BG itrdp, country == PL itrdp, country == EL itrdp, country == EE itrdp, country == LT itrdp, country == CZ year itrdp, country == BE itrdp, country == DK itrdp, country == AT itrdp, country == RO itrdp, country == PT itrdp, country == IE year itrdp, country == FI itrdp, country == IT itrdp, country == UK itrdp, country == DE itrdp, country == SE itrdp, country == ES itrdp, country == FR

19 Sectoral allocation to fit Rhomolo structure

20 Sectoral allocation For the ICT sectors: Baseline: Policy: RE RD = RD, t = (2011, 2020) ICT ICT s, r,2010 ICT sict,, r t ICT r, t REr,2010 ICT ICT ICT ICT REsICT, r GVAr RD ' s,,,, ( ',, ) ICT r t = RDsICT r t + RD r t RDr t RE sict, r GVA sict, r sict sict For the non-ict sectors: Policy:, ( ' ) RD ' = RD RD RD RE ICT ICT ICT ICT s r snict, r, t snict, r, t r, t r, t REsr, s

21 Regionalization With region/sector GVA shares for 2005, and region GVA shares for , regional R&D intensity is: RD ' = RD jt, s,2005 s w GVA js, GVA jw, The region-specific factors are treated as fixed effects, while Europe 2020/ICT induced changes are assumed to affect sectors across regions in the same way. The RD j,t series is then adjusted to fit historical / starting data for This method allows a set of forecast R&D intensities to be calculated which are consistent with the agreed targets, and which are also consistent with the latest available regional R&D page historical 21 data

22

23 Further potential for integrating ICT (R&D) into Rhomolo The presented approach allows for simulation of ICT R&D policy effects in model which does not specifically account for ICT R&D. Further advancement would be to modify the TFP function in Rhomolo, so the ICT-type of R&D enters it directly. As ICT R&D has more widespread effects than other types of R&D, additional technology diffusion/spillover mechanism would be complementary.

24 Conclusions RHOMOLO is a new model, designed to be capable of tracing the origin of regional heterogeneity in the impact of alternative policies, and on the role played by the different channels of externalities In particular, the design brief is to capture such types of phenomena and then to help in the assessment of the impact of innovation policies at the regional level. RHOMOLO v1 is ready to provide results for the EU27-NUTS2, with further developments expected in the near future (v2 expected in mid-2012). There is potential for analysis of ICT and ICT R&D related policies with Rhomolo