Transcript ppt - Institute for Fiscal Studies

Productivity and Government Policy Towards
R&D
Rupert Harrison
Institute for Fiscal Studies
Plan of the lecture
• Motivation
• The UK ‘Productivity Gap’
• UK R&D performance
• Private and social returns to R&D
• Theory
• Evidence
• R&D tax credits
• The basic idea
• Do R&D tax credits work?
• The UK R&D tax credits
• Conclusions
The productivity gap: labour productivity
Index, UK=100
140
120
100
UK
Germany
France
USA
80
60
40
20
0
Value added
per worker
Value added
per hour
worked
Value added
per person of
working age
Source: HM Treasury, Pre-Budget Report 2004
Service sectors account for an increasingly large part of
the gap
Share of UK-US Gap, 1990, 1995 and 2001
Mining/quarrying
Electricity, gas, water
Transport and storage
Construction
Post and telecom
Hotels and restaurants
Manufacturing (excl. mach+equip)
Business services
Machinery and equipment
Financial intermediation
Wholesale and retail
Distribution of employment
-5%
0
5%
10% 15% 20% 25%
1990
1995
2001
Source: authors' calculations using OECD data, US measured at f actor costs
Source: Griffith, Harrison, Haskel and Sako (2003)
Determinants of productivity
• R&D and innovation
• creation of new knowledge and technologies
• diffusion and adoption of existing technologies
• Human capital
• direct effect on labour productivity
• indirect effect as skills and technological progress
may be complementary
• Investment climate
• Competition, regulatory regime
• Infrastructure
R&D as a % of GDP
UK R&D performance, 1981-2002
3.00
UK
2.50
USA
France
2.00
Germany
1.50
1981 1984 1987 1990 1993 1996 1999 2002
Year
Japan
Who performs R&D in UK? (2002)
£m
%
13,100
67%
Higher Education R&D
(HERD)
4,400
22%
Government R&D
(GOVERD)
1,800
9%
Business Enterprise R&D
(BERD)
Decline in BERD intensity, 1981-2002
BERD as a % of GDP
2.5
UK
USA
2.0
France
1.5
Germany
1.0
1981 1984 1987 1990 1993 1996 1999 2002
Year
Japan
Increase in overseas R&D…?
R&D level, rebased
200
150
R&D done
by UK
firms
R&D done
in the UK
(BERD)
100
50
0
1992
1994
1996
Year
1998
2000
Why should government support R&D?
• The ‘policy-makers argument’
• “support innovation…”
• “improve competitiveness…”
• The economist’s response
• Where’s the market failure?
• Does the market create sufficient incentives for individuals
and firms to engage in the socially optimal amount of
innovation and technology transfer?
• If not, can government intervention effectively provide the
appropriate incentives at sufficiently low administrative and
compliance cost, and without creating further distortions?
Why should government support R&D?
• ‘Spillovers’ justification
• In the absence of perfect intellectual property rights,
knowledge is partially non-excludable
• Total benefits of new knowledge may not be captured by the
innovator
• Private returns to innovation are lower than social returns
• The market will not provide the socially optimal level of
innovation
• Role of R&D and spillovers in models of endogenous growth
(e.g. Romer 1990, Aghion and Howitt 1992)
• Non-rival nature of knowledge
Private and social returns to innovation
• What evidence is there that SROR > PROR ?
• Intuition, case-studies…
• Econometric evidence
• 3 broad types of econometric evidence:
• Cross-country studies at economy level
• Cross-industry studies (often across countries as well)
• Plant- and firm-level studies (usually for 1 country)
Augmented production function approach (see
Griliches, 1998)
Yit 
L K R
Ait Lit K it Rit
ln Yit   L ln Lit   K ln K it   R ln Rit  eit
R
Yit
R
Rit
is the elasticity of output w.r.t. the firm’s R&D stock
is the (private) rate of return to the firm’s R&D stock
Estimate external r.o.r. from production function
ln Yit   L ln Lit   K ln Kit   R ln Rit
  E ln Rit  eit
E
is the elasticity of output w.r.t. others’ R&D stock
Empirical evidence: firm and industry-level studies
• Griliches (1998) concludes from the literature that
• “R&D spillovers are present, their magnitude may be quite large,
and social rates of return remain significantly above private rates”
• Estimates at firm level
• Private rate of return: 15% to 30%
• Social rate of return: 30% to 50%
• Estimates at industry level
• Social rate of return (only within-industry spillovers): 20% to 40%
• Social rate of return (incl inter-industry spillovers): 50% to 100%
R&D – imitation as well as innovation?
• R&D does not just generate new knowledge, ie push
out the technology ‘frontier’
• It may also allow firms behind the frontier to imitate
those at the frontier by increasing their ‘absorptive
capacity’
• Implications:
• 1. Imitation may be costly
• 2. Doing R&D may allow a firm / economy to catch
up with high productivity firms / economies, raising
its growth rate in the short run until it catches up
The two faces of R&D (Griffith et al, 2001a)
 Ai 
 R
 ln Ait  1     ln AFt  1 ln  
 Y it 1
 AF t 1
technology transfer
 Ai 
 R
  2   ln    X it 1  uit
 Y it 1  AF t 1
absorptive capacity
Estimates of the social rate of return
Griffith et al (2001a)
• Innovation effect (1)  40%
• This is the rate of return to R&D for the country at the
frontier in a given industry
• But UK TFP was only 63% of US TFP (1974-90)
• So R&D may also enable us to catch up with the frontier,
boosting the social return to R&D if 2<0
• Total effect (innovation + imitation/technology transfer)
  1 - 2ln(Ai/AF)t-1  90%
(3)
Implications for government policy
• Evidence supports some kind of subsidy to R&D as externalities
appear to be substantial
• Gap between private and social rate of return implies that subsidy
should be quite large
• In practice no government offers this much subsidy
• Direct subsidy vs R&D tax credit
• Gradual move away from discretionary support schemes
• Tax-based schemes allow firms to choose R&D projects
• Tax-credits directly address the externality by bringing the
marginal private return closer to the social return
Alternative credit designs
• Volume-based credit
• Payable on all R&D
• Incremental credit
• Payable on all R&D above a rolling base
• Fixed base credit
• Payable on all R&D above a fixed base (eg 50% of level in
base year)
Key criteria for R&D credit design
• Cost-effectiveness
• additional R&D (value added) generated per pound of
exchequer cost
• Simplicity
• low compliance and administrative costs
• Certainty for companies
• how much credit will they receive and when?
Cost-effectiveness 1:additional R&D
• Additional R&D generated depends on:
• amount of R&D eligible for the credit
• effect of tax credit on the ‘price’ of the last pound of R&D
(marginal effective tax credit)
• responsiveness of R&D to the lower ‘price’ of R&D
• Is additional R&D of the same quality as existing R&D?
• Marginal projects
• Re-labelling of other activities as R&D?
Cost-effectiveness 2: exchequer cost
• Exchequer cost of tax credit depends on:
• credit rate (and statutory rate of corporation tax)
• amount of ‘existing’ R&D that receives the credit (‘deadweight’)
• amount of new R&D generated
• ‘Deadweight’ cost is by far the largest component of cost in most
designs (often >95%)
Pros and cons of each design
• Volume–based credit
• simple to understand and predict but high ‘deadweight’
• Incremental credit
• lowest ‘deadweight’ but frequent uprating of base reduces
effectiveness (METC < credit rate)
• Fixed base credit
• intermediate deadweight but uncertainty over future uprating
of base may reduce effectiveness
• Complex rules necessary for incremental and fixed base
designs
R&D tax credits in the UK
SMEs
Large firms
<250
250+
<40m euros
40m+ euros
Number of firms
4,500
1,000
Amount of R&D
c. £0.5 bn
c. £10 bn
50%
25%
Corporation tax rate (B)
19% - 30%
30%
Marginal effective tax credit
(A*B)
9.5% - 15%
7.5%
Yes (@24%)
No
Employment
Turnover
Rate of credit (A)
Repayable?
How much new R&D might this generate?
• Additional R&D will be equal to:
Eligible R&D
* %Δprice
* price-elasticity of demand for R&D
Change in price of R&D
(see Griffith et al, 2001b)
User cost of R&D after
credit =
Where Ad is NPV of capital
allowances before credit
Ac is METC for R&D
1  A d  Ac
(   )
1 t
 is the firm’s real discount rate
 is the rate of economic
depreciation of R&D
t is rate of corporation tax
 1  Ad 
1  0.287


Change in price = ln 
 ln 
 0.11
d
c

1  0.287  0.075 
1  A  A 
Effect on R&D expenditure
• Bloom et al (2002) use data on tax treatment of R&D
in a panel of OECD countries to estimate the priceelasticity of demand for R&D as:
• 0.12 in the short run
• 0.86 in the long run
• Implies a change in R/Y (R&D intensity) of around
• 0.11 * 0.12 = 1.3% in the short run
• 0.11 * 0.86 = 9.5% in the long run
What might this do to productivity growth in
manufacturing? (Griffith et al 2001b)
•
•
•
•
Given eligible R&D in manufacturing  £8bn
Manufacturing value added
 £150bn
Hence R/Y
 5.3%
Effect on growth of TFP
= (R/Y) * %(R/Y) * elasticity of TFP wrt (R/Y)
= (R/Y) * %(R/Y) * [1 - 2ln(Ai/AF)t-1 ]
= 0.04% in short run or 0.30% in long run
assuming 1 = 0.433, 2 = 1, and Ai/AF = 85% of
frontier productivity
Conclusions
• Theory and empirical evidence suggests that social
return > private return to R&D due to spillovers
• R&D tax credits go some way to internalise externality
at sufficiently low admin and compliance cost
• R&D tax credits are likely to cost effective, at least in
the long run
• Too early to evaluate the impact of the UK R&D tax
credits – early results possible in maybe 1 or 2 years
• Not going to narrow the ‘productivity gap’ on their own
Caveats
• If the supply of scientists and engineers is inelastic in
the short run then initial impact of R&D tax credits
may be mainly to raise their wages
• There is some evidence that this happened in the US
• UK firms are doing more R&D overseas, especially in
the US
• Some evidence that this R&D is more productive and
provides access to cutting edge technologies
• Should we encourage them to do more R&D here or
are they better off doing it in the US (frontier)?