R&D as a Value Creating Asset
Emma Edworthy
Gavin Wallis
Methodological Overview: Linking Tables
• UK Business R&D linking Tables:
BERD
MINUS
Capital expenditure on land and buildings and plant and machinery
PLUS
Acquisition of R&D to be used as input in R&D production
PLUS
Capital Services
Methodological overview: Freely available
R&D
• Canberra II group made the following recommendation:
‘In principle, freely available R&D should not be included as Capital
formation, but in practice it may not be possible to exclude it. The
assumption is that including freely available R&D would not lead to
significant error’.
• On the basis that it is too difficult to separate it out, the UK are going to
include all freely disseminated R&D.
Methodological overview: Double
Counting
• Computer software: software development outside the computer
industry is likely to be picked up in the BERD and own account software
numbers
• GFCF
• 1. Estimates from surveys collecting data on GFCF (CAPEX, ABI) will
include some intellectual property
• 2. Not all expenditure by companies in the R&D industry will result in
intellectual property. They will also invest in furniture and fittings etc.
Current Price GFCF
Total expenditure on
R&D (Frascati Manual
based)
=
Current expenditure on
R&D (labour costs etc.)
Current expenditure on
R&D (labour costs etc.)
+
+
R&D related
expenditure on other
assets (Plant and
machinery, purchased
software etc.)
R&D GFCF
Derived estimate of
capital service flows
from other asset
classes*
addition of R&D GFCF
R&D related
expenditure on other
assets (Plant and
machinery, purchased
software etc.)
+
estimation
=
=
National Accounts
GFCF
estimation
National Account
capital stock
Capital consumption
non-R&D related
expenditure on assets
(Plant and machinery,
software etc.)
GDP
* Can either be derived as consumption of fixed capital COFC (capital consumption) plus a normal return on capital used or direct capital
services estimates
Current Price GFCF
•
We use three different methods for calculating the capital service flows from
other asset classes:
•
Method 1: Consumption of Fixed capital (COFC) plus an assumed return


GFCFt   Ct   I at    I at   COFCat   Rat
a
a
a

 a
•
Method 2: Capital services estimated using rentals


GFCFt   Ct   I at    I at   CSat
a
a

 a
•
COFCat  K at   a
CSat  K at  rat
Method 3: Capital services estimated using capital services growth rates


GFCFt   Ct   I at    I at   CSat
a
a

 a
CSa 0  K a 0  ra 0
CSat 1  CSat  g at
t  1, 2,
Current Price GFCF
• Method one is a proxy estimate of Capital services
• In methods two and three the capital service flow from the asset used
in R&D is measured directly.
• Capital services growth rate is a much more common output of
statistical offices than estimated rental rates.
• UK currently publishes capital services growth rates annually, but not
rentals.
• Methods two and three are preferred on theoretical grounds, as they
directly measure capital services flows. Which is best?
Constant Price GFCF
Data sources used:
R&D component
Proxied by
Source
Wages and salaries
•Index of earnings of
ASHE
science and technology
Professionals
•Index of average earnings
of technicians
•Index of average earnings
of Administrative
occupations
Other current (materials
etc)
PPI (input) materials and
fuels purchased by
manufacturing excluding
FBTP
PPI
Capital
Separate index for plant
and machinery and land
and buildings
National Accounts
capital stock deflators
Constant Price GFCF
120
115
110
105
GDP deflator
R&D deflator
100
95
90
1998 1999 2000 2001 2002 2003
year
Depreciation rate
In calculating an R&D stock, we use the Perpetual Inventory model.

K at   (1   a ,t  )  I a ,t 
 0
•
We made the following assumption about the net capital stock in the
initial year, assuming a steady state.
Ka0  I a0 /  a
Depreciation rate
We estimated a depreciation rate for the whole economy using
econometric methods.
•
Looked at the impact past R&D had on productivity (GVA at market
prices)
•
We estimated the following:
GVAt 
 G
s 1... t
s
t s
Depreciation rate
Our preliminary results were run for the period 19982003:
Gva_diff
coefficient
Standard error
t
L1
-0.770473
7.516772
-0.01
L2
-18.89507
10.71141
-1.76
L3
42.80761
10.53161
4.06
L4
-26.19576
10.39732
-2.52
L5
2.838254
8.741311
0.32
Depreciation Rate:
•
Suggest a life length mean of 5 years
•
Depreciation is calculated as follows:
  R /T
Where R is the declining balance rate (equal to 2) and T is the life
length mean (5)
•
Implies a depreciation rate of 40%
•
We recognise that this is a very crude method. It is just an early
investigation in to a possible approach.
•
The approach taken could provide sensible estimates of depreciation
following more development
Results: Current price Business GFCF
•
The results from the three methods are quite similar
Year
MA14: Total
R&D
expenditure
Method 1
Method 2
Method 3
2003
13.7
15.1
15.1
14.6
2002
13.1
14.9
15
14.5
2001
12.3
13.5
13.4
13.1
2000
11.5
12.4
12.5
12.1
1999
11.3
12.5
12.7
12.3
1998
10.1
10.9
11.1
10.8
1997
9.5
11.3
10.4
11.2
Results: Business R&D Capital stock
15% Depreciation
Year
Method Method Method
1
2
3
64.2
64.5
62.7
2003
57.8
58
56.7
2002
50.4
50.1
49.6
2001
43.5
43.9
42.9
2000
36.7
37
36.3
1999
28.4
28.7
28.2
1998
20.5
20.6
20.5
1997
12
12.1
12.1
1996
40% Depreciation
Year
Method Method Method
1
2
3
34.9
35.1
33.9
2003
33.1
33.3
32.3
2002
30.4
30.4
29.8
2001
28.3
28.5
27.9
2000
26.5
26.7
26.1
1999
23.2
23.5
23.1
1998
20.5
20.7
20.5
1997
17
17.1
17.1
1996
Preliminary Productivity Analysis: Firm
Level
yit  a  1 nit   2 kT ,it   3 k R ,it   4 D  eit
Where:
y: value added
n: labour
kT ,it : tangible capital
a: impact of external knowledge on the firm’s productivity
k R ,it : R&D
e: error term
D: Service industry dummy
Preliminary Productivity Analysis:
Employment
0.71*** (40)
Tangible Capital
0.25*** (19)
R&D Capital stock (RD)
0.08*** (11)
Services dummy
0.17*** (4.4)
Services*RD
0.01*** (3.2)
Patent Dummy
0.08*** (4.0)
Domestic Dummy
0.05** (2.5)
Patent dummy*RD
0.0 (0.05)
UK dummy*RD
0.005** (2.01)
US dummy*RD
0.007** (2.23)
Japan dummy*RD
0.01 (1.38)
EU dummy*RD
0.00 (-1.82)
*** significant at 1% level
** significant at 5% level
* Significant at 10% level
Preliminary Productivity Analysis:
•
Our analysis considers the link between the level of value added and
the level of R&D
•
By estimating the Cobb Douglas production function we obtained a
coefficient on R&D that represents the elasticity of R&D with respect
to TFP i.e. the % change in productivity for a % change in R&D
•
The estimates of R&D elasticity is 0.08. This implies that a 10%
increase in BERD is associated with a productivity increase of 0.8%.
•
Average difference between the impact of services and
manufacturing on productivity. Services firms are on average more
productive.
Preliminary Productivity Analysis:
•
Interacting services and the R&D capital stock suggested
that an increase in R&D capital stock leads to a bigger
increase in productivity for Services than manufacturing.
•
Sector breakdown: An increase in the R&D capital stock
in services and primary industries leads to a larger
increase in productivity than manufacturing.
•
Whereas construction and energy have a negative impact
compared to manufacturing.
Preliminary Productivity Analysis:
•
The addition of dummies for patented industries and foreign ownership
showed that patented firms and UK firms add more to productivity.
•
R&D coefficient on R&D capital stock still remains significant after the
addition of the patent dummy.
•
UK owned, manufacturing firms in the patent industry add most to
productivity.
•
However, interacting the patent and foreign owned dummies with the R&D
capital stock showed that there was no additional affect from being a UK firm
or in a patent industry.
•
UK owned firms and US firms have an additional effect from an increase in
the R&D capital stock on productivity over and above the rest of the world.