Policy Inducement Effects in Energy Efficiency Technologies. An Empirical Analysis of the Residential Sector - Green Energy and Efficiency: An Economic Perspective - page 215

þ b 2 MarketSy s i ; t 1

¼ þb 0 þ b 1 Innov Sys i ; t 1

þ b 4 Energy Sys i ; t 1

þ b 3 EE Policy i ; t 1 ð 4 Þ

þ e i ; t

þ b 5 EðÞ Controls i ; t

We use a log-log xed effects speci cation to take into account country-speci c

unobservable heterogeneity; Hausman xed

effects. 7 The maximum likelihood method is used to estimate the model parameters.
All variables referring to the systems investigated are modelled with a one year lag
in order to reduce potential endogeneity bias while preserving the standard
inducement effect framework. In this sense, when the resilience of the innovation
process is accounted for, it is commonplace to expect policies or market inducement
effects to present a time lag from the time when the phenomenon occurs and the
reaction in terms of innovations by

' s test con rms our choice of using

rms. As a standard method of addressing this
issue, a one year lag reduces endogeneity and enables resilience to be accounted for,
but a minimal number of observations is lost.
Different model speci

cations are estimated to test the contributions of the
different systems affecting the dynamics of invention of EE technologies. The
policy variables are maintained in all the speci cations, while different variables for

measuring the contribution of other innovation drivers are tested. Moreover, further
estimations show the impacts of each policy type by disaggregating the policy
dataset according to Table 2 .
Table 3 tests a general policy inducement effect together with the contribution of
two different proxies of the innovation system. More speci cally, estimations (1 - 4)

include the stock of GERD, while in estimations (5 - 8) innovation capacity is

measured by the speci c stock of R&D in EE. Broadly speaking, the contribution to

invention of the national innovation system is positive and signi cant both when

the effect is tested on the total number of patents and when patents are divided into
the three sub-domains. Unfortunately, our dataset suffers from a large number of
missing data for speci c R&D in EE, which translates into several missing

observations. Therefore, in the estimations below we keep only the GERD variable
for measuring the contribution of national innovation systems.
The price-inducement effect, represented by the price-tax bundle, also positively
and signi

cantly affects our dependent variable, although the statistical robustness
is lower than for R&D. Since we measure prices at end-use level, it can be inferred
that producers pay more attention to price changes, probably in a demand-driven
effect in which consumers are highly sensitive to energy consumption and prefer-
ably choose high-ef

ciency goods to counterbalance increases in energy prices.
In the case of electrical appliances and lighting, two sectors characterised by
intensive energy use but prompt responsiveness for energy saving, this effect is
particularly strong, for at rst, the lifecycle of lamps and

appliances is shorter than that of buildings, and the reactivity of consumer

least two reasons:

' s choices

7

Test results are available upon request.