When and How to Support Renewables?—Letting the Data Speak - Green Energy and Efficiency: An Economic Perspective - page 290

function nd that the (semi)elasticity of

patents ranges between 0.4 and 0.7 suggesting decreasing return to scales. 7 As the
current year accounts for over 60 % of total R&D elasticity, they conclude that
R&D impacts patenting at an early stage of the R&D sequence.
Public RD&D spending on particular technologies is also deemed to create
innovations. 8

' based on R&D and lagged R&D. They

nd that public RD&D expenditure stimulates
innovation in renewable energy technologies.

For example, [ 3 ]

1.2.3 A Combination of Deployment and RD&D is Driving Innovation

Based on earlier literature Wiesenthal et al. [ 17 ] present a two-factor learning curve
model that disentangles two of the most important learning factors: learning by
doing and learning by researching. The latter describes the relationship between the
accumulated knowledge stock and production costs. For a given technology t and
time period y , the curve can be described as follows:

C t ; y ¼ aQ a

t ; y KS b

t ; y

where:
C /W)

Q Cumulative Production (W)
KS Knowledge stock (here: approximated through R&D investments,

Costs of unit production ( €

€ )

ʱ Elasticity of learning by doing

ʲ Elasticity of learning by researching

a Normalisation parameter with respect to initial conditions

Soederholm and Sandqvist [ 13 ] use a two-variable model using deployment and
R&D to estimate the effectiveness of different subsidy schemes. They show that
learning rates depend crucially on the speci

cation used. Quantifying effects

remains dif cult and the authors stress that simultaneity can lead to possible biases

as for example reduced costs can lead to higher deployment.
Lindman and S ö derholm ( 2012 ) review 35 studies on learning rates for wind
power and warn that results are econometrically spurious in most empirical esti-
mates. They argue that more attention should be paid on “ learning and knowledge

spillovers in the renewable energy eld, as well as to the interaction between

technology learning and R&D efforts .

Koseoglu et al. [ 8 ] discuss the allocation of subsidies to either R&D or market
application. Their conclusion is that R&D is underused compared to market
application subsidies. A possible reason could be that short term effects of
deployment are more visible than R&D and therefore favoured by policy makers.

”

7

Similar to Hall et al. ( 1986 ) who analysed data set from the seventies with similar models.

8

This is despite crowding-out effects of private RD&D spending. See for example [ 10 ].