evaluation of the norwegian r&d tax credits 160919... · 2020. 1. 13. · norwegian tax credits...
TRANSCRIPT
Evaluation of the Norwegian R&D tax credits
Marina RybalkaStatistics Norway
Joint workshop on firms, industries and productivity
Lisboa (INE) | September 16, 2019
Norwegian tax credits SkatteFUNN
24.09.2019 | Slide 2
• Launched in 2002
• 18 percent (20 for SMEs) tax deduction of certain amount of firm’s R&D expenses
– up to EUR 0.5 million in the start– up to EUR 2.5 million today– works as a grant in case if the tax credit exceeds the tax payable by the firm or if the firm is not in
a tax position
• Simple application and reporting procedures • No discrimination between types of projects or industries• Favouring collaboration with R&D institutions
– project cup is doubled in case of joint project with an approved R&D institution
Main changes in the scheme2002-2017
24.09.2019 | Slide 3
Timeline for SkatteFUNN
Tax credit SMEs: 20 pct. Intramural R&D: NOK 4 mill Purchased R&D: NOK 8 mill
Total: NOK 8 mill Introduction of SkatteFUNN
2002
2003
Tax credit SMEs: 20 pct. Tax credit others: 18 pct.
2007
Cap on hourly wage: NOK 500
Cap on yearly hours: 1 850
2009
Intramural R&D: NOK 5.5 mill Purchased R&D: NOK 11 mill
Total: NOK 11 mill
2011
Changes in definition of R&D and SMEs and in calculation
of hourly wage. Cap on hourly wage: NOK 530
2014
Intramural R&D: NOK 8 mill Purchased R&D: NOK 22 mill
Total: NOK 22 mill Cap on hourly wage: NOK 600
Intramural R&D: NOK 15 mill Purchased R&D: NOK 33 mill
Total: NOK 33 mill
2015
2016
Intramural R&D: NOK 20 mill Purchased R&D: NOK 40 mill
Total: NOK 40 mill
Intramural R&D: NOK 25 mill Purchased R&D: NOK 50 mill
Total: NOK 50 mill
2017
Thresholds for tax-deductible R&D expendituresNOK million. 2002-2018
0
5
10
15
20
25
30
35
40
45
50
0
5
10
15
20
25
30
35
40
45
50
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
Intramural Purchased (and total)
24.09.2019
Changes in thresholds
| Slide 4
Changes in firms exploiting SkatteFUNNNumber of firms by firm subsidy size. NOK 1000. 2002-2015
24.09.2019
Use of SkatteFUNN
| Slide 5
0
500
1000
1500
2000
2500
3000
3500
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
>1600[1440,1600](1100;1440)[990,1100](800;990)[720,800]<720
Changes in firms exploiting SkatteFUNNShare of firms by firm subsidy size. NOK 1000. 2002-2015
24.09.2019
Use of SkatteFUNN
| Slide 6
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
>1600[1440,1600](1100;1440)[990,1100](800;990)[720,800]<720
Input additionality• A key point in the evaluation
– Important criterion in itself, cf. policy goals– A ”necessary condition for success”
• Counterfactual question: What would R&D investments have been without SkatteFUNN?
• Challenges:– Non-experimental setting – SkatteFUNN is a general scheme– no formal control group
• Strategies:– Econometric evaluation (exploring changes in the scheme)– Surveys and interviews
24.09.2019 | Slide 7
Construction of R&D expenditures variable
1. Reported R&D expenditures from R&D survey
2. Reported R&D expenditures in 3 years period in SkatteFUNN application
3. Information on direct subsidies from other public agencies and programs– Research Council of Norway– EU (FP7, H2020)– Innovation Norway– Regional research funds
24.09.2019 | Slide 8
R&D data from Tax authorities compared with other data sources. In thousand NOK
0
5000
10000
15000
20000
25000
30000
35000
0
5000
10000
15000
20000
25000
30000
35000
R02-03 R04-06 R07-08 R09-10 R11-13 R14-15
<720 [720,800] (800;990) [990,1100] (1100;1440) [1440,1600] >1600
24.09.2019
Firms report to the Tax authorities only part of R&D that is eligible for the tax credit
| Slide 9
Input additionality
24.09.2019
«Discontinuity approach» (jf. Hægeland og Møen, 2007)
| Slide 10
0.000 %
0.005 %
0.010 %
0.015 %
0.020 %
0.025 %
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000
Estimated distribution of intramural R&D as reported in application to SkatteFUNN. NOK 1 000. 2003-2013
Thresholds for tax-deductible R&D expendituresNOK million. 2002-2018
0
5
10
15
20
25
30
35
40
45
50
0
5
10
15
20
25
30
35
40
45
50
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
Intramural Purchased (and total)
24.09.2019
Changes in thresholds
| Slide 11
Input additionality
Only firms with R&D expenditures under the project cost cap are stimulated to increase their R&D efforts
Probability to start R&D (extensive margin) Positive effect after introduction of the scheme, no significant effect of the
increase in 2009
Probability to continue R&D Positive effect in both studies
BFTB: 1.01 (0.96/0.73) vs. 2 in Hægeland og Møen, 2007 relevant only for firms with past R&D>0 (past R&D > 1 NOK million/past R&D > 2
NOK million) post-crisis period
24.09.2019
«Discontinuity approach» (jf. Hægeland og Møen, 2007)
| Slide 12
Input additionality
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 13
Sources: Statistics Norway and Samfunnsøkonomisk analyse
0
500
1000
1500
2000
2500
3000
3500
0
500
1000
1500
2000
2500
3000
3500
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
New users of Skattefunn by tax credit size
>1600
[1440,1600]
(1100;1440)
[990,1100]
(800;990)
[720,800]
(600;720)
<=600
Total active
Input additionality
• Explore all changes in the scheme
• Main idea: Follow different SkatteFUNN-user «generations» over different policy regimes– allow firms that begin using the policy early differ from
firms that start using the policy later • probably encouraged by the specific policy change
• Period: 1999-2015– 1999-2001 is a period before SkatteFUNN
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 14
Input additionality
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 15
Sources: Statistics Norway and Samfunnsøkonomisk analyse
0
500
1000
1500
2000
2500
3000
3500
0
500
1000
1500
2000
2500
3000
3500
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
New users of Skattefunn by tax credit size
>1600
[1440,1600]
(1100;1440)
[990,1100]
(800;990)
[720,800]
(600;720)
<=600
Total active
1
2
34
5
6
Input additionality
• Main challenge: selection into the scheme– a direct comparison of SkatteFUNN-users and non-users will
give a biased result
• Matching procedure: – pick control-firms among non-users of SkatteFUNN that are as
similar as possible to SkatteFUNN-users prior to introduction of the policy
– Make these two groups more comparable– BUT loose some information
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 16
Input additionality
• Results of matching procedure for firms with information on previous R&D• Stratification by firms’ industry, region and cohort, and use of additional
public support
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 17
Before match After match
SKF-firmsnot
SKF-firms %bias SKF-firmsnot SKF-
firms %biasNo. of employees 21.98 2.64 36.1*** 10.71 10.07 2.2Total assets 77882 14261 5.8*** 14089 12304 2.9Organisational age 3.35 0.45 44.4*** 3.08 3.06 0.2Share of high-skilled 0.36 0.13 68.5*** 0.31 0.30 2.7Financial liqudity rate 36.62 81.42 -20.3*** 42.49 44.86 -1.2Dummies:R&D experience 0.15 0.002 58.0*** 0.04 0.03 9.4***Other types of support 0.30 0.004 90.2*** 0.02 0.02 0No. firms 7060 227934 3089 11199F-test 3418.44*** 1.2
Input additionality
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 18
Graphical test of common trend assumption, entire population
Input additionality
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 19
Graphical test of common trend assumption, after matching4
56
78
9M
ean
log
R&D
expe
ncie
s
199920002001200220032004200520062007200820092010201120122013
Control G02-03G04-06 G07-08G09-10 G11-13G14-15
Input additionality
R&D expenditures for early users generation against the benchmark of no use
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 20
0
200
400
600
800
1000
1200
1400
SkatteFUNN-user generetion 2002-2003
Not SKF-firms G02-03: no SKF G02-03: with SKF
Input additionality
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 21
0%
20%
40%
60%
80%
100%
R02-03 R04-06 R07-08 R09-10 R11-13 R14-15
Effects of SKF partisipation on R&D expenditures
G02-03 G04-06 G07-08 G09-10 G11-13 G14-15
Input additionality
• “Bang for the buck”: main model
24.09.2019
«Generalized diff-in-diff approach» (jf. Mohnen et al.,2017)
| Slide 22
RegimeAll generations G02-03 G04-06 G07-08 G09-10 G11-13 G14-15
R02-03 2.22 2.22R04-06 2.68 2.86 2.41R07-08 2.34 2.26 2.83 1.63R09-10 1.95 2.21 2.38 0.74 2.15R11-13 1.60 3.07 1.85 0.65 1.64 0.90R14-15 1.66 2.34 2.47 1.42 2.17 1.45 1.15Total 2.07 2.55 2.42 1.06 1.91 1.12 1.15For firms with past R&D>0 2.04 2.47 2.33 0.94 1.99 1.29 1.17
Input additionality• We find positive and significant input additionality
– but effects vary depending on the type of change and the type of user-generation
• For all generations input additionality is strongest in the start of use of SkatteFUNN and is declining thereafter – until most recent policy change with significant increase of project cost
cap
• Not significant difference in BFTB between large and small firms, but between large and small projects
• Much higher BFTB for firms with zero past R&D
24.09.2019
Main conclusions
| Slide 23
Innovation output analysis
24.09.2019 | Slide 24
• Probability of different types of innovations (probit model)– new product, new process, new product to the marked
• Innovative sales (ordered probit model)
• Probability of different types of protection (probit model)– Patent– Trademark– Design– Copyright
• Main data sources: CIS 2001-2014– merged with the matched data from the previous stage
Innovation output analysis
24.09.2019 | Slide 25
• Main model– Split the observed R&D intensity into two parts
• Additional R&D intensity generated by a tax credit, Δr– the treatment effect on the treated predicted for each firm from the input
additionality analysis
• Counterfactual R&D intensity that each company would have done in the absence of a tax credit, rc
– obtained as the difference between r and Δr
• Inspired by Czarnitzki and Hussinger (2004), Cerulli and Poti(2012) and Freitas et al. (2017)
Innovation output analysis
24.09.2019 | Slide 26
• Main model: types of innovationNew or improved
productNew or improved
processNew product for the
markedShare of turnover
from new products(1) (2) (1) (2) (1) (2) (1) (2)
Additional R&D intensity, Δr 0.030*** 0.047*** 0.023*** 0.050*** 0.029*** 0.041*** 0.012* 0.031***Counterfactual R&D intensity, rc 0.070*** 0.053*** 0.045*** 0.033*** 0.051*** 0.045*** 0.068*** 0.064***Share of high-skilled 0.094** 0.086 -0.078** -0.083 0.076** 0.107** 0.046 -0.066Cooperation: in the group 0.065** 0.065** 0.030 0.030 0.038* 0.036 0.01 -0.011Cooperation: another firm 0.197*** 0.167*** 0.155*** 0.133*** 0.126*** 0.091*** 0.109*** 0.070**Cooperation: university -0.009 -0.011 0.063** 0.053* 0.022 0.027 0.011 -0.002Yt-1 - 0.203*** - 0.148*** - 0.163*** - 0.156***Number of observations 4401 2158 4391 2151 4377 2158 3934 1839Pseudo R2 0.29 0.34 0.18 0.20 0.26 0.29 0.14 0.15
Innovation output analysis
24.09.2019 | Slide 27
• Main model: types of innovationNew or improved
productNew or improved
processNew product for the
markedShare of turnover
from new products(1) (2) (1) (2) (1) (2) (1) (2)
Additional R&D intensity, Δr 0.030*** 0.047*** 0.023*** 0.050*** 0.029*** 0.041*** 0.012* 0.031***Counterfactual R&D intensity, rc 0.070*** 0.053*** 0.045*** 0.033*** 0.051*** 0.045*** 0.068*** 0.064***Share of high-skilled 0.094** 0.086 -0.078** -0.083 0.076** 0.107** 0.046 -0.066Cooperation: in the group 0.065** 0.065** 0.030 0.030 0.038* 0.036 0.01 -0.011Cooperation: another firm 0.197*** 0.167*** 0.155*** 0.133*** 0.126*** 0.091*** 0.109*** 0.070**Cooperation: university -0.009 -0.011 0.063** 0.053* 0.022 0.027 0.011 -0.002Yt-1 - 0.203*** - 0.148*** - 0.163*** - 0.156***Number of observations 4401 2158 4391 2151 4377 2158 3934 1839Pseudo R2 0.29 0.34 0.18 0.20 0.26 0.29 0.14 0.15
Innovation output analysis
24.09.2019 | Slide 28
• Main model: types of innovation protection
Patent Trademark Design Copyright(1) (2) (1) (2) (1) (2) (1) (2)
Additional R&D intensity, Δr 0.011** 0.022*** -0.003 0.003 0.002 0.010 -0.006 -0.011Counterfactual R&D intensity, rc 0.025*** 0.018*** 0.029*** 0.024*** 0.016*** 0.016** 0.014*** 0.018***Share of high-skilled 0.092*** 0.083* 0.021 0.001 0.005 0.050 0.079*** 0.111**Cooperation: in the group 0.027 -0.009 0.048** -0.006 0.034** 0.001 0.001 -0.023Cooperation: another firm 0.033** 0.045** 0.024 0.083*** 0.027* 0.080*** 0.022 0.037Cooperation: university 0.022 0.004 0.040* 0.030 -0.009 -0.026 0.015 0.026Yt-1 - 0.197*** - 0.169*** - 0.152*** - 0.145***Number of observations 4382 2158 4398 2158 3988 1807 3988 1821Pseudo R2 0.19 0.29 0.14 0.21 0.17 0.25 0.14 0.19
Innovation output analysis• We find positive and significant results for different
innovation types– But effects vary depending on the type of innovation
• We find positive and significant result for patents– But not for other types of innovation protection
• The results hold even when we control for persistency of innovation
• Skills of employees and cooperation with other firms are important for innovation
24.09.2019
Main conclusions
| Slide 29
Other results
• The estimated effect of R&D capital on labour productivity is positive and significant
• The external effects of R&D are difficult to measure quantitively
• We find no any significant increase in cooperation after expansions of the scheme
– However, it was an increase in cooperation after the introduction of the scheme
24.09.2019
Main conclusions
| Slide 30
Misuse• 5 risk criteria defined by the Tax authorities
– high-risk group when the firm meets 2 of 5 criteria
• 150 randomized controls in high-risk group and 50 in low-risk group
• Misuse of SkatteFUNN does occur– For about 10 percent of the controlled firms it was revealed discrepancies
between hours reported to the government and in their SkatteFUNN-claim– About 13 percent of the firms reported ordinary costs as R&D investments
• However, the overall BFTB > 1 even after adjustments for the misuse
24.09.2019
Main conclusions
| Slide 31
Thank you!Marina Rybalka