Extra subsidy implies extra R

1
Extra subsidy implies extra RD?
Maarten Cornet (m.f.cornet_at_cpb.nl) Björn
Vroomen (b.l.k.vroomen_at_cpb.nl) CPB Netherlands
Bureau of Economic Policy Analysis Cardiff, 8
September 2005 (Work in Progress)
2
Introduction

• Innovation important driver of economic growth
• Effectiveness of innovation support programmes in
  general unclear
• causality or correlation?
• Analysis of (change in) policy with social
  experiment framework
• We present two examples where causal relation can
  and can not be identified

3
The WBSO

• WBSO Dutch RD tax credit programme
• Credit based on RD-employment (total RD wage
  expenditures)
• 40 tax credit for the first 68k, 13 for the
  remaining
• 2001 two additions to the programme
• additional support for new RD-firms
  (technostarter)
• extension of the first tax-credit bracket (
  68k? 90k)

4
Data

• Data provided by
• Dutch Tax and Customs Administration
• SenterNovem (Ministry of Economic Affairs,
  programme administrator)
• Data provided on participants in the
  WBSO-programme
• total yearly RD wages
• firm characteristics (branch, turnover, total
  wages)
• period 1994-2003 (2000-2001 analysis)
• 28000 records

5
Technostarter programme

• RD wages expenditures till 90k 60 tax credit
  instead of standard 40
• Firms have to be younger than 5 years
• Firms should not have obtained more than 2 years
  of WBSO funding
• ? max of three years of technostarter funding

6
Technostarter identification
7
Ideal case?

• Nearly ideal
• Difference in age may influence reaction of firms
• Difference in experience with the WBSO program
  (that is, experience with RD) may influence
  reaction of firms
• Include control variables
• age and WBSO experience
• firm characteristics

8
Increase RD wage expenditures?

• Technostarter vs near-technostarter
• difference-in-difference estimator 18 (p0.00)
• first-difference estimator 14 (p0.00)
• Technostarter vs hidden technostarter
• FD-estimator 11 (p0.01)
• Effect of policy on RD wage expenditures is
  estimated between 10 and 20

9
Value for money?

• Bang-for-the-Buck (BFTB) Ratio between extra RD
  and extra public funding
• Effect 10-20 BFTB 40-60
• 1 extra public funding implies 0.50 extra RD
  wage costs, given an effect of 15
• Are the social benefits higher than the social
  costs?
• in general they are

10
Extension first tax-credit bracket

• Increase of 68k to 90k
• Category 1 (control group), firms below 68k
• no change, credit remains 40
• Category 2 (experimental group), firms between
  68k and 90k
• increase of 13 to 40
• Category 3 (Control group), firms above 90k
• no change, remains 40 13, however lump sum
  subsidy

11
Methodological issues

• Firms in control group can also be influenced
  by treatment
• difference in effect is estimated
• Choice of group sizes may influence results
• Group membership is not constant over time
• ? No convincing causal relation!
• Lets ignore problems for now

12
Increase RD wage expenditures? (ignoring
methodological issues)

• Compare 50k-68k with 68k-90k
• DD-estimator 2.6 increase (p0.01)
• FD-estimator ?
• Compare 90k-120k with 68k-90k
• DD-estimator 2.9 increase (p0.01)
• FD-estimator ?
• DD , FD ? ? group composition effect?

13
Value for money?

• Extra funding firms above 90k is quite costly
• lump sum ( 6000) ? minimal effect expected
• in 2000 1932 firms near 12 million
• 1 extra public funding implies 0.20 extra RD
  wage expenditures, given an effect of 3 (BFTB
  20)
• Are the social benefits higher than the social
  costs?
• in general they are not

14
Conclusions

• Effectiveness
• Technostarter program successful, clear causal
  relation (10-20 increase)
• Extension first tax-credit bracket unclear,
  methodological issues that hinder identification
  causal relation (? 3 increase)
• Bang-for-the-Buck, literature on average 100,
  Technostarter 50, Extension bracket ? (20)
• This study is one of a kind or other studies to
  high?

15
Discussion
16
Setting

• Identifying a causal relation between instrument
  and outcome is far from trivial
• the causal relation between many innovation
  policy instruments and innovation is unclear
• Social experiment framework
• evaluation of (change in) policy instruments
• Two cases where causal relation can and can not
  be identified
• firm-level data on RD employment

17
Social experiment

• Identify two groups (experimental control) of
  similar participants, in that manner that the
  difference in outcome can only be explained by
  the treatment
• participants are not randomly assigned and aware
  of being treated
• characteristics of experiment (policy) can help
  to identify both groups
• Two methods of estimation effect
• Difference-in-Difference estimator (DD) compare
  group averages over time
• First-Difference estimator (FD) compare
  growth/change of individual participants over
  time

18
Possible reaction curve