Sunteți pe pagina 1din 24

Paper to be presented at the DRUID 2012

on
June 19 to June 21
at
CBS, Copenhagen, Denmark,

AN INVESTIGATION INTO THE DETERMINANTS OF DIFFUSION OF WIND


POWER
Konstantinos Delaportas
UCL - University College London
SSEES - School of Slavonic and East European Studies
konstantinos.delaportas@uclmail.net

Abstract
My research seeks to investigate the factors that influence the diffusion of wind power in a total of 130 countries over the
time period 1990-2009. The paper treats electricity from wind power as an ecoinnovation, and tries to add to the
literature that examines the barriers of diffusion of ecoinnovations, by drawing upon the theory of diffusion of
innovations. In particular, it tries to unify a ?neoclassical economic approach? with a sociological perspective, and then
uses hazard models to examine the factors that can explain differences in the speed of diffusion of wind power across
countries.
To model the aforementioned theoretical framework, the paper uses hazard models in an attempt to identify the
reasons why a certain event has occurred within a given period of time. In this paper two major questions are examined:
first, what are the factors explaining whether a country has integrated wind energy into its electricity generation system,
and second, what factors can explain the speed of diffusion of wind technologies into a country. The results mainly
confirmed the neoclassical economic approach, suggesting the various elements of the market as the most important
determinants of profitability these included the profitability of adopters, of suppliers, the availability of wind, as well as the
carbon and market lock in. Moreover, the existence of a change agent in the form of a green party and the country?s
trade with the pioneers of wind energy were also significant determinants, while all other institutional determinants were
found insignificant.
Jelcodes:O33,-

AN INVESTIGATION INTO THE DETERMINANTS OF DIFFUSION


OF WIND POWER

INTRODUCTION
The aim of this paper is to examine the determinants of first adoption of wind energy.
Although the issue of innovation and the issue of invention has been the focus of a plethora
of research works, the issue of adoption of an innovation has not received the appropriate
attention. This paper aims to deal with this issue by investigating the factors that can explain
a countrys decision to adopt wind turbines in order to generate electricity.
The structure of the paper is the following: section two investigates some of the main
literature on diffusion of innovation, and examines the particularities of the energy sector
with respect to diffusion. Section three gives an overview of the data and the methodology
that is used to examine adoption, section four presents and discusses the results, while
section five concludes.

LITERATURE REVIEW
The determinants of innovation and its diffusion, although yet not clearly understood, have
been the focus of a wide deal of research; nevertheless, relatively less emphasis has been
placed on the determinants of first adoption of the innovation. In other words, I expect to be
a different set of factors that influence the decision of an agent to substitute an existing
technology. These factors vary according to the extent that the agent decides to substitute
the existing technology.
The rest of this section summarizes two key theoretical approaches in the diffusion of
innovation literature, continues with examining the particularities of innovation diffusion in
the energy sector.
To understand what factors influence the diffusion of innovation, first we need to
understand what is the mechanism of diffusion. Then according to each stage, the various
determinants will be proposed.

DIFFUSION PROCESS
Burt (1973) summarizes the literature on the diffusion process by arguing that the process
begins when an agent becomes aware of the innovation and ends by the individuals
decision to adopt or reject it.
The literature suggests the existence of three stages between the moment of information on
the existence of the innovation and the decision towards its adoption. In particular, Burt
(1973) argues that this process of adoption can be broken into three successive stages. The
process is initiated by the potential adopter becoming aware of an innovations availability.
Following awareness of the innovation, the potential adopter proceeds to gather information
1

and/or advice relevant to the innovation from both personal and formal sources. During the
course of this communication activity, the potential adopter reaches a psychological decision
regarding adoption of the innovation. This second phase of adoption can thus be termed a
decision period. The final phase of the adoption is the process behavioral adoption.
In this research I call these three stages as the information period, the decision-shaping
period and the decision-making period. Schematically this process can be illustrated in the
following diagram:
Information period

Decision-shaping period

Decision-making period

DETERMINANTS OF DIFFUSION
NEOCLASSICAL ECONOMICS PERSPECTIVE
Neoclassical economists believe that the only determinant of diffusion of an innovation is
related to its cost; i.e. that technology choices respond to changes in the price; the cheaper
the product, or the higher the expected profitability the higher should the rate of its
adoption be. However, Griliches (1957) research on hybrid corn illustrated that only part of
the increased adoption rate of hybrid corn in the US can be attributed to profitability.
However, he argued that if economic profits were clear-cut, then this facilitated the
diffusion process.
Although investigation in the topic of hybrid seed might not sound an attractive research
subject, this research is of particular interest because of the importance of hybrid seeds to
the farmers at that stage. Seeds are the main input in the production process of farmers,
which was at that time the major source of income. Thus, when we investigate the diffusion
rates of certain technologies we need to take into account the importance of these
technologies into the agents economic activities.
Griliches, analyses the diffusion process by suggesting three categories of factors that
examine diffusion at each of the stage. One group of factors can explain the decision to
adopt, another explains the speed of diffusion once the adoption decision has been made,
and the last one group examines the ceiling, i.e. the maximum share of the market that
the innovation can capture (Griliches 1957, pp. 505-506).
Griliches argues that the decision to adopt is first and foremost dependent upon the
availability of the innovation in the region in question (Griliches 1957, p. 507); the
availability is a proxy for the supply of the innovation; the producers decision to supply
depends on the perceived profitability of the regions from the suppliers of the innovation. In
turn, this depends on the size of the market, the marketing costs, the cost of innovation in
that area, and the expected rate of acceptance by the consumers. What he concluded was
that the expected pay-off of the suppliers was the major factor influencing their decision to
introduce the product to the market.
When attempting to examine the diffusion factors, i.e. the reasons that lead consumers to
purchase/adopt the innovation, Griliches argues that the expected profitability of the new
innovation to the agents is the principle reason for adoption, since the higher the stimulus
the faster is the probability of adoption (Griliches 1957, p. 516). His empirical findings

suggested that profitability can explain a substantial variation (around 50-60%) in the
differences in the diffusion speeds across regions. It is obvious that this approach assumes
that all farmers have a similar understanding of the benefits of the innovation, partly
implying a perfect functioning of the market, a standard approach of the neoclassical
economists. However, how realistic is it that all agents understand fully and can predict the
monetary benefits of adopting an innovation? It is undeniable that one of the main reasons
for adopting an innovation is the potential benefits, but what determines their perception of
these benefits? What factors shape their beliefs? Moreover, arent there some institutional
factors e.g. adopting because of the competitor adopted that could help explaining their
decisions? The issue of the importance of profitability is an interesting one particularly when
examining the RETS, where micro agents (investors) might be solely motivated by the
profitability, but the profitability is dependent on the price, which is set by government. So it
could be quite interesting to examine the interplay. This is a way of artificially increasing the
benefits of an innovation in order to attract new agents is captured by the literature on
incentives1.
On the issue of the importance of profitability, Rogers summarizes the literature on
agriculture that illustrates that profitability is important, but not the sole important factor in
the diffusion process (Rogers 1988, p. 215).

ROGERS DIFFUSION FRAMEWORK


There are 2 ways to understand the determinants of diffusion. The first way looks at the
characteristics of an agent that decides whether to adopt an innovation. The second way
looks at the innovation and what are its characteristics that make it attractive to be adopted.
Rogers approach follows the latter, and suggests that 49 to 87 percent of the variance in the
rate of diffusion can be predicted based on an innovations 5 characteristics: Relative
advantage, Compatibility, Complexity, Trialability, Observability (Rogers 1988). The
remaining variation can be attributed by a wide spectrum of factors such as the type of
innovation-decision, the nature of communication channels diffusing the innovation at
various stages in the innovation-decision process, the nature of the social system, and the
extent of change agents' promotion efforts in diffusing the innovation (Rogers 1988, p.
232).
The following table is a graphical representation of his diffusion framework:

See Rogers pp. 217-223

novations

Attributes of Innovations and Their Rate of Adoption

233

ion are
ed and
cult to
vability
em, is

ch are
action
volved
nents:
es, the
aspect
ited in
e (the
of a
nnovavabil-

ion is
as the
d. So
adop-

pe of
is its
of the
lative
ility).
other
comn the
, and
he in-

III. Communication Channels (e.g., mass


media or interpersonal)
IV. Nature of the Social System
(e.g., its norms, degree of interconnectedness, etc
V. Extent of Change Agents' Promotion Efforts

Figure 6-1. A paradigm of variables determining the rate of adoption of


Relative
advantage is defined as the degree to which an innovation is perceived as being
innovations.
better than the idea it supersedes (Rogers 1988, p. 213). This is the most straightforward
measurable
diffusion
parameter, andiscan
be expressed
in eitherrate
monetary or social
The type
of innovation-decision
related
to an innovation's
advantages,
depending
on the nature
the innovation.
Therequiring
higher thean
innovations relative
of adoption.
We generally
expectof that
innovations
advantage,
the higher innovation-decision
its rate of diffusion. He
identifiesmore
two categories
individual-optional
willthen
be adopted
rapidly that reflect this
than when anprofitability,
innovationand
is adopted
by an organization
(Chapter
10).especially related to
characteristic:
status. Rogers
argues that status,
a trait
The more
involved
in making
an innovation-decision,
the process, but its
highly
visiblepersons
innovations,
is important
at the
beginning of the diffusion
slower
the
rate
of
adoption.
If
so,
one
route
to
speeding
the
rate
ofits status begins to
importance decreases with time as more and more people adopt it and
adoption
is
to
attempt
to
alter,
the
unit
of
decision
so
that
fewer
indecline.

dividuals are involved. For instance, it has been found in the United

States that when


thedegree
decisiontotowhich
adoptan
fluoridation
Compatibility
is the
innovationofismunicipal
perceivedwater
as consistent with the
supplies
is
made
by
a
mayor
or
city
manager,
the
rate
of
adoption
is
existing values, past experiences, and needs of potential adopters (Rogers
1988, p. 223).
quicker
than
when
the
decision
is
made
collectively
by
a
public
The higher the compatibility of the innovation with the already established practices, the
referendum.
lower
is the uncertainty and thus the higher the potential for diffusion. Moreover, this is an
The communication channels used to diffuse an innovation also
important criterion as the previously introduced ideas constitute the basis of comparison.
may have an influence on the innovation's rate of adoption (Figure
The
compatibility
criteria
relate to the
sociocultural
values
beliefs, the previously
6-1).
For example,
if interpersonal
channels
must be
used and
to create
introduced
ideas, and the as
client
needs for
innovations.
Rogers
suggests
awareness-knowledge,
frequently
occurs
among later
adopters,
thethat it is positively
related
to
the
rate
of
diffusion,
but
that
the
statistical
evidence
does
not illustrate it as a
rate of adoption will be slowed.
majorThe
diffusion
determinant
(Rogers
1988, p. 226).channels
This criterion
relationship
between
communication
and can
ratebeoffurther analysed by
adoption
even more
complicated
than
6-1and
suggests.
Thepositioning,
atlooking
into are
its various
characteristics
such
asFigure
its name,
its market
tributes of the innovation and the communication channels probably
In an attempt to reap the full benefits of compatibility, it might be tempting to introduce an
innovation with a high degree of similarity with existing practices; however, such an action
entails two crucial risks. Firstly, there is some kind of trade-off between compatibility and
relative advantage. The more similar the innovation is to the one it is replacing, maybe the
lower is its potential to improve existing practices, and thus the lower its perceived
advantage. Secondly, there is always the issue of innovation negativism, whereby a failure of
an incremental innovation may hinder the further stages of an innovation process (Rogers
1988). If you decide to gradually diffuse an innovation and thus split its adoption process
into various steps, you might lower the chances of each step being rejected as it is less
radical than the established practices, but at the same time you increase the number of
stages and thus lowering the probability of adopting the eventual innovation.

Rogers defines complexity as the degree to which an innovation is perceived as relatively


difficult to understand and use (Rogers 1988, p. 230), and argues that its negatively related
to the rate of adoption, an argument that has yet to find rigid empirical support. In my
opinion, this is very user specific criterion; for example, you cannot expect a recent
mathematics graduate to have the same ease in understanding on how a new computer
software operates as a historian emeritus.
Trialability is the degree to which an innovation may be experimented with on a limited
basis (Rogers 1988, p. 231); the higher its degree of trialability, the lower the uncertainty
that surrounds the innovation and thus the likelier its adoption. Moreover, the importance
of this characteristic decreases with the number of adopters, and its assumed to be more
important at the early adoption stages. But this assumes that riskiness is negatively
associated to adoption. Is that always the case??
Observability is degree to which the results of an innovation are visible to others (Rogers
1988, p. 232). He argues that the more visible a technology is to members of a social group,
the higher its rate of diffusion.
The type of innovation decision suggests that emphasis needs to be paid on the number of
agents involved in the adoption-decision making. The more agents involved the slower is the
adoption process, suggesting that an adoption decision by an organization is taken with
more of a difficulty than the decision of a single agent. The communication channels and the
way they operate in a given social system have a definite role in the rate of diffusion; yet,
Rogers does not specify the nature of this interaction as well as its impact on the diffusion
rate. Lastly, the efforts of the change agent have a certain influence on the diffusion rate,
but its importance again depends on the diffusion rate and is a topic still under-researched.
Rogers recognizes the dynamic nature of all the above determinants, and suggests the
diffusion effect, which is defined as the cumulatively increasing degree of influence upon
an individual to adopt or reject an innovation, resulting from the activation of peer networks
about an innovation in a social system (Rogers 1988, p. 232). In other words, there is a
domino effect of adoption implying that the diffusion rate accelerates with its number of
adopters. This effect relates to the availability of information on the innovation as well as its
communication systems, and suggests that there is a minimum level of information
necessary for an agent to adopt an innovation. But is this level of information the same to all
agents, or it varies? Maybe more risk loving agents may require less information than riskloving individuals.

PARTICULARITIES OF INNOVATIONS IN THE ENERGY SECTOR


Innovations in the energy sector are in many ways particular in their analysis, so some
authors have gone as far as providing a new concept that of eco-innovation. An ecoinnovation can be viewed as The production, assimilation or exploitation of a product,
production process, service or management or business methods that is novel to the
organization (developing or adopting it) and which results, throughout its life cycle, in a
reduction of environmental risk, pollution and other negative impacts of resources use
(including energy use) compared to relevant alternatives (Oltra 2008, p.2).
Oltra (2008) and Popp (2010, p. 2) argue that ecological innovations are similar to all other
innovation types, in the sense of that their analysis depends on factors which are difficult to

evaluate. But most importantly, the main particularity related to innovation in this sector is
the fact that there is a double externality problem, where the existing market failures of
the innovation process, are accentuated in the environmental markets, because of the
failure of the market to price pollution, and thus the producers have no incentive to reduce
its production in the absence of policy/regulations.
Yet some authors argue that the negative externalities of this market failure are somewhat
alleviated if reductions in pollution are viewed as inefficiencies, and thus firms have
incentives to improve their efficiency and thus reduce their negative environmental impact.
Moreover, they argue that because of their positive environmental impact, these
innovations are always socially desirable, and thus justifying the need for state involvement
to establish equilibrium in the market. However, there are some RETs, which despite their
evident environmental benefits, fail to gather unanimous public support. A typical example
is the case of wind turbines, whose installation faces significant opposition by various social
groups despite their clear environmental benefits, for reasons related to appearance. 2
Moreover, RETs have various physical/technical differences when compared to the
traditional energy generating facilities that need to be taken into account. Conventional
electricity generating facilities such as coal/oil fired and nuclear require extensive and
expensive physical facilities. On the contrary, some RETs are small scale and of a modular
nature, allowing for factory-based automatic production, and much less on-site
construction. Therefore, renewable energy technologies are more similar to massproduction technologies than to conventional power plants (Neij 1997, p. 1100). This
suggests that RETs do not share all the particularities of existing analyses of the energy
sector. At the same time, RETs because of they have lower energy densities than traditional
energy producing means, these installations occupy larger physical space and are thus more
likely to create influence a larger number of stakeholders (Wstenhagen et al 2007, p. 2684).
These differences have illustrated the importance of examining other stakeholders who
were earlier considered of minor importance when investigating the diffusion process.
Another issue that needs to be taken into consideration when analyzing the renewable
energy innovation has to do with the complicated infrastructures to which energy
technologies are bounded to (Wstenhagen 2007, p. 2685). This technological
complicatedness makes the introduction of any radical innovation more difficult than the
diffusion of other independent products. Moreover, although their introduction is small
scale their investment and siting decision still affects a multitude of other stakeholders, and
consequently becomes a political rather than a pure economical decision (Wstenhagen
2007, p. 2686). Therefore, their diffusion analysis simply on an economic basis might not be
ideal.
RETs, and wind technologies in particular, belong to a sector that has some distinct
characteristics when compared to other mainstream industries, something that needs to be
accounted for when investigating it. In particular, Jacobsson (2000) recognizes 3
particularities. The first one is related to the size of the market, which is enormous and thus
the amount of time needed for any substantial transformation to take place is quite
extensive. The second one relates to the subsidization of the incumbent/traditional energy
sources, either in the form of R&D incentives or the non-reflection of the environmental

There is also some evidence that wind turbines have a negative impact on birds
and their surroundings.
2

costs in the prices of energy produced from them. Moreover, the energy sector, because of
its strategic importance for the government and is thus always under strict monitoring and
regulation. Thus, any analysis of major technological change in this sector needs to
incorporate the role of the government as a major stakeholder. Lastly, Jacobsson, but also
other authors like Nakicenovic (Nakicenovic 2002) and Unruh (2000) underline that the
resistance of the established players in any kind of change is quite significant. This resistance
gives rise to dangers of potential technological lock-in to fossil fuel technologies, a
development which could lead to a reversal of the world economys trend towards
decarbonization.
THE DIFFUSION PROCESS OF RETS
Grubb (1990) constitutes one of the earliest attempts to examine the reasons for which RETs
had failed to achieve a deep entry into the energy sector, despite the common agreement
on their medium and long term advantages. He argues that RETs constitute a Cinderella
option implying that their use and potential is neglected by current policy makers (those in
the 1980s) and proposes four potential explanations for this neglect. The first concerns a
lack of data on the actual energy value of renewable resources, which leads to uncertainty
and underinvestment. The second is related to the excessive conservatism exhibited by
international organizations and governments based on the pessimist intellectual legacy of
the 1970s, with these studies offering pessimistic projections on technologies and costs,
thus discouraging government support funds to flow in the renewable sectors. Thirdly, there
seems to be a lack of the necessary informative institutions that would efficiently
disseminate the necessary renewables information to the policy community and allow it to
create measures promoting the innovation. Lastly, he suggests a lack of vision i.e. skepticism
and adverse attitude of policy makers towards potential of renewable technologies.
Studies supporting the importance of regulation for diffusion of environmental technologies
have also been the topic for most empirical diffusion research. The main rationale stems
from the double externality problem, and claims that if the introduction of these
technologies does not bring efficiency gains or in other words provides sufficient profitability
to justify its adoption by the investors, then policy is required to promote diffusion. For
example, Popp argues that even if the all the advantages of RETs are taken into
consideration, these benefits are largely external to the individual producer; the producer
will have an incentive to adopt more costly clean technologies only if they provide additional
costs savings, thus justifying incentives for government intervention (Popp et al 2010, p. 24).
Other research comes from Gray and Shadbegian (1998), who find a strong relationship
between regulation and diffusion. Their research was focused on the paper and pulp
industry, and aimed at examining the factors that influence the technology adoption choice
of the plants. Similarly,
Moreover, Kerr and Newell (2003) investigate the diffusion of lead reduction technologies in
US refineries. They develop and test a model that suggests that diffusion will take place by
firms gradually as the cost falls and thus the benefits increase, while simultaneously
regulatory stringency increases the value of adoption; firms with lower benefits or higher
costs will adopt more slowly. The importance of regulations is such that the authors argue is
the main explanatory variable behind the diffusion of lead reducing technologies comes
from increased regulatory stringency.
However, not all types of regulation have an immediate effect on the adoption of
ecoinnovations. Snyder et al. (2003) examines the diffusion of membrane-cell technology in

the chlorine manufacturing industry, and argues that direct regulations to chlorine
manufacturing facilities did not have a significant impact on the diffusion of new technology.
Rather indirect measures to end-products influenced consumer demand, which in turn
influenced the decision of producers to adopt new, cleaner technologies. Others (Popp
2006) have proposed optimal policy mixes, but the extent of this optimality is very much
country specific, something not sufficiently taken into account in these papers.
Linked to the issue of regulations is the strength, effect and mechanism of political economy
that permeates the system, as it affects demand for regulation (Lovely, & Popp 2008).
Moreover, Lovely and Popp (2008) suggest that openness facilitates technology adoption
and diffusion, and also that developing countries adapt regulations in earlier stages of
economic development than the developed countries.
Micro-level analysis finds that the more sophisticated a plant, the easier/smoother/faster
the diffusion of the new technology is (Kerr, & Newell 2003). A similar argument could me
made for diffusion of RETs, and test the hypothesis that the more sophisticated and
technological advance a country is the faster is the diffusion of a new technology. Or if we
want to narrow the hypothesis even more, we could argue that the more technologically
advanced and diversified the energy sector of a country, the faster the diffusion of RETs. To
support this claim, we could look back in the literature of diffusion and match it with a
theory.
The principal-agent problem can also be used to explain why some environmental friendly
technologies can face difficulties with their adoption. The main argument is that if the
person deciding for the adoption of the new technology is not the one that reaps the
rewards then the process of diffusion is impeded. The typical example is that of the adoption
of energy savings technologies for a rented house. The owner has no incentive to invest in
more expensive energy efficient measures, if these costs are not passed on to the tenant. At
the same time the tenant could benefit by paying lower utility bills, but these costs might
take longer to be paid back, and assuming he is interested in a short-let, he might choose an
alternative home with a lower rent (Popp et al 2009). This conflict of interest might lead to
what is known as the energy paradox, whereby although some environmental technologies
provide cost savings to their adopters, they are still not diffused (Jaffe, & Stavins 1994;
Newell et al 2004). As a corollary, policy aimed at improving the financial benefits of these
technologies is ineffective (Popp et al 2010, p. 24).
Another explanation for this energy paradox comes from Shama (1983), who observed the
low speed of diffusion of energy conservation technologies, which bring positive economic
costs to adopters. He argues that if this phenomenon is examined purely from the
economics and engineering perspective, then it seems as an irrational and paradoxical
behavior of consumers. However, if the behavioral aspect is included in the analysis, then
this behavior can be considered as perfectly rational. In more details, he uses Rogers
diffusion framework to examine this energy paradox, and concludes by providing various
policy recommendations.
For the adoption of an innovation it is necessary to examine the preferences of the agents
involved in the process. Such an attempt is made by Masini and Menichetti (2010) who
investigate the decision making process behind RET investments. They use behavioural
finance and build a model that examines the structural and behavioral characteristics of
investors as factors determining diffusion. The main shortcoming of this paper is the fact
that it treats RETs as one, and does not distinguish between different types of technologies
8

and the investors attitudes towards them, and also its exclusive focus on the European
market.
Another factor that has most likely an impact on the diffusion of RETs is prices. Some earlier
research on ecoinnovations has illustrated energy prices as significant determinants of
technology adoption (Boyd, & Karlson 1993). Thus, it could be argued that prices are also
significant for the diffusion of RETs. The only paper that investigated this phenomenon
explicitly was Rehfelds (2007), who stressed the importance of getting the prices
economically right, in order to achieve maximum diffusion.
Similarly, literature suggests the importance of factor costs; firms with higher factor costs
will tend to adopt a cost cutting technology faster than those with lower costs. FisherVanden et al. (2006) use a panel of 22,000 Chinese large and medium enterprises and
attempt to determine the factors that explain Chinas improvements in energy efficiency
over the period 1995-2001. Their findings suggest that energy efficiency improvements
stemming from the diffusion of cleaner technology are mainly attributed to rising energy
prices. This finding which underplays the eminent role of regulation of all previous studies
might help to shed light on the particularities of transition/emerging economies, since this is
one of the very few studies that examine diffusion in such economic environment.
Therefore, a significant determinant of technological diffusion is its impact on the firms
profitability.
Neij (1997) uses experience curves to study the diffusion of wind and solar, and argues that
the most important factor for their diffusion is how fast their prices (measured as the cost of
generating electricity) will fall when compared to traditional electricity producing factories.
She argues that the potential for cost reduction is higher for RETs than for traditional
technologies; her research also illustrates the importance of R&D for decreases in the total
cost, and only if these two facts are combined one could expect successful diffusion of RETs.
Similarly, Nakicenovic (2002) illustrates the importance of learning by doing in the diffusion
of new technologies, and makes a case study of the decarbonization of energy. He proposes
a theoretical model that shows that although it is more costly to invest in clean technologies
now, the learning effects from the diffusion of these technologies will allow for a prompt
payoff of the investments and thus facilitate even more the diffusion.
General literature on diffusion stipulates a positive relationship between firm size and
adoption propensity (Karshenas, & Stoneman 1993; Geroski 2000; Levin et al 1987; Saloner,
& Shepard 1992). The number of studies supporting a negative relationship is very limited,
with Oster and her study of steel firms (Oster 1982). The lower the number of adopters, the
higher the speed of diffusion as less agents will have to accept the new innovation. Hence, it
could be argued that more oligopolistic/higher concentrated industries might accept the
innovation more easily. Moreover, the larger the size of a firm the easier and less costly and
risky it is to adopt new technologies, as it is more diversified.
In the energy sector, Rose and Joskow (1988) investigate the impact of size and ownership
on technology diffusion in the US electricity market in a sample of 144 electric utilities over
the 1950 through 1980 period. Their results indicate that diffusion is positively related to
firm size, but at the same time argue for the existence of a maximum optimal size after
which further size increases leads to lower diffusion. Moreover, their results underline the
importance of ownership structure for diffusion, with investor/privately-owned companies
faster adopters than foreign one, and find weak support for the influence fuels prices as a
diffusion determinant. Kerr and Newells research in petroleum refineries also support this
9

finding in the energy sector, with larger and more sophisticated refineries more likely to
adopt new technologies (Kerr, & Newell 2003). In this case however the size was a proxy for
sophistication, which was in turn thought of as positively related to diffusion. This finding is
noteworthy because it could be argued that the more sophisticated the plant, the more
difficult it would be to introduce radical changes, i.e. the more locked-in it would be to
established technologies. But at the same time, empirical evidence suggested the opposite.
Looking at the cross-country adoption of RETs, the argument of size could be translated into
a proxy for the size of the country, or more appropriately the wealth and/or entrepreneurial
climate and/or the innovativeness of the country. The more wealthy and entrepreneurial
and innovative the countrys mentality is, the more likely it is to take risks and adopt new
technologies. Similarly, it could be argued that countries with more
concentrated/monopolistic energy sectors, the diffusion could be higher. However, this
hypothesis might not be validated, as it might seem that concentrated energy sectors are
state dominated, and thus less prone to innovation. But again, it is different to talk about
innovation and different about its diffusion. In state-owned companies it might be easier for
governments to impose the regulations, and thus in this case diffuse an innovation.
Lanjouw and Mody (1996) construct a patent data set from 1972 to 1986 for the US, Japan
and Germany in order to study the creation and diffusion of environmental technologies.
Moreover, they examine international technology transfer from these 3 industrialized
economies to 14 lower and middle income countries. This period is important because it was
a period of a period of rapidly increasing public awareness and concern about environmental
damage, similar to todays situation. Environmental concerns came to the forefront in the
early 1970s, triggered in some in- stances by specific accidents, as in Japan. They found that
innovation is related to pollution abatement expenditures, which are in turn related to the
level and stringency of environmental regulations. To study diffusion they investigate trade
in environmental technology and domestic patenting by foreigners in the sphere of
environmental technology, and find that most patents in developing countries comes from
technologies sited for developed countries rather than technologies adapted for developing
countries, suggesting the importance of technology transfer from developed to developing
countries.
Dechezleprtre et al. ( 2010) looks at patents and the diffusion of RETs in emerging markets,
a total of 76 countries, including emerging markets. They find that R&D and innovations
mainly focused on industrialized countries, particularly Japan, Germany and the USA.
However, some 16% of total patents comes from emerging markets, especially China, Russia
and South Korea. On the issue of international diffusion of these technologies, the evidence
suggests that there is not much transfer activity across countries; most of the transfer takes
place among developed countries, but there seems to be no evidence indicating diffusion
among emerging markets. Thus, what is an important area of research for diffusion is how
this technology flows across international boarders, esp. to emerging economies that are
now developing at extremely high rates and are thus the prime polluters.
Lovely and Popp (2008) focus on the adoption of environmental regulation as the first step
in the international diffusion of environmental technologies (Popp et al 2010, p. 28). In
particular, they examine how the existing technological stock, mostly originating in
developed countries, induces regulation in developing countries. They focus their research
on the adoption of coal-fired power plants to adopt pollution control regulations in a
mixture of 45 developed and developing countries.

10

One of the most interesting arguments of the paper is that when we investigate diffusion of
clean technology, the starting point should be the introduction of the regulation rather than
the introduction of the technology itself. Thus, they argue that the first step for
understanding international diffusion of clean technologies is the understanding of the
determinants of regulation.
Popp (2004) investigates the innovation and diffusion process of air pollution control
equipment; he focuses on USA, Japan, and Germany and uses patent data to examine how
technological innovations in the field of air pollution occur and diffuse between these three
countries. His findings justify the hypothesis that stricter regulation induces innovation, but
the interesting point of this paper is his attempt to examine the diffusion of these
technologies across borders. In more details, he measures innovation in terms of patents
and views the diffusion as taking place in two ways: either as the direct adoption of a
technology from a different country, or as an input in the creation of a new technology (i.e. a
knowledge spillover). If, for example, a patent of an innovation created in Japan is cited by a
patent made in the USA, then this suggests the existence of a knowledge spillover from
Japan to the USA. The patent evidence suggests that technology transfer is more indirect,
and takes mostly the form of knowledge spillovers rather than direct adoption of foreign
innovations.
However, I believe that the main problem of this paper is the selection of countries. By
selecting only these 3 countries they make the implicit assumption that this technology is
pioneered and exists only in these 3 countries. As a non-expert in the field of this
technology, I cannot claim that this assumption is not valid. However, it would be surprising
if no other OECD country developed any similar technology within this time span, something
which if true can cause considerable omitted variable bias in the analysis.

11

METHODOLOGY
Modeling the time to technology diffusion/adoption leads naturally to the use of statistical
methods developed for analyzing duration data; these methods are commonly known as
duration or hazard models. Hazard models are focused on the occurrence of a particular
event, usually known as failure, which occurs after a certain period of time has passed. The
interest is not solely whether or not the event will occur, but also the timing of the event.
The hazard rate is the probability of the country failing in the time interval of our analysis t,
given that it has survived up until time t. Examples of failures are the failure of a component
in a machine, the death of a patient, or the movement into unemployment of a worker. In
this analysis, although counterintuitive, failure denotes the fact that a country has started to
exploit its wind resources in order to generate electricity.
The survivor function is the probability that no event has occurred before time t. The
cumulative distribution function serves as the complement to the survivor function and
illustrates what is the probability of an event occurring before time t.
A common issue in survival analysis is the issue of censoring. Censoring occurs, or better, an
observation is censored if it does not fail within the time period of the analysis.
There are three main approaches that could be used to model time to an event: nonparametric, semi-parametric, and parametric. The non-parametric method lets the data to
talk by themselves, i.e. does not attempt to estimate either the baseline hazards or the
coefficients. Semi-parametric models leave the baseline hazard unspecified and rather focus
on calculating estimates for the coefficients by simply using the ranks of time. This method is
also called the Cox-proportional hazards model. Nevertheless, proportional hazard models
require that the impact of any individual covariate on the hazard rate is the same for all
values of t. Lastly, parametric models complement semi-parametric analysis by assuming
that the baseline hazard follows a certain distribution and attempting to model/determine
it. This paper focuses on non-parametric and semi-parametric methods.

HAZARD MODELS IN DIFFUSION STUDIES


Framing the theoretical issue of technology adoption in the concept of hazard models, leads
us to investigate what factors determine the conditional probability of technology adoption
in time t given that the technology has not already been adopted by that time. Within the
economics literature, hazard models have been used to analyzing labor economics issues,
such as unemployment spells, but they have to a more limited extent used to issues related
to technology adoption (Kiefer 1988).

AZARD ODELS IN ECONOMICS AND ENVIRONMENTAL ECONOMICS


Kiefer (1988) performed an investigation in the potential use of hazard models or duration
studies in the science of economics, although he was not the first to use them. Among other
potential usages, he recommends their utilization in the field of technology studies, and in
particular, when one examines the time to adoption of new technologies (Kiefer 1988).
Hazard models are not widely used in environmental economics, but there have been some
attempts to utilize them, mainly in order to investigate the impact of regulations on the
diffusion of some environmental innovations.
12

Snyder et al (2003) used hazard models in order to investigate the impact of regulation on
the use of chlorine. They used these econometric models because their interest was on the
timing of the introduction of the innovation rather than other factors such as causality,
which is better investigated under different econometric techniques. They examined
diffusion at the firm level, and in particular, the decision to adopt a new more friendly
technique for chlorine manufacturing. Their focus is the USA over a 30-year period, and their
findings suggest that regulation has no impact on the decision of firms to retrofit the
innovation, but rather influences solely the decision of firms to exit the industry.
Lovely and Popp (2008) firstly construct a general equilibrium model and form their
hypothesis on the impact of regulation in their fictional economy, and then they use hazard
models to test the diffusion of environmental regulation across countries. Kerr and Newell
(2003) use hazard models to investigate how lead reduction technologies were diffused in
US petroleum refineries in the period 1971-1995.

THEORETICAL MODEL
Cox models are useful for our analysis because the scope of this paper is not to determine
whether or not the probability of failure change over time, but which factors influence the
probability of failure. According to this model, the hazard function h(t) is equal to
or
where h(t) is the rate at which the country introduces electricity from wind at time t, given
that they have not by time t-1, h0 is the baseline hazard (which remains unspecified in this
model) when the values of all covariates are equal to zero, and X is a vector of all covariates
and their corresponding parameter vector that the model assumes to have an impact on
h(t).
The baseline hazard (h0) is assumed to be common across all agents of our analysis, and to
vary only with time and not with any other variable including the covariates. Moreover, it is
based on the assumption that all hazard functions across the different levels of variables in
the model are proportional to a baseline hazard function (Somers, & Birnbaum 1999).
For the Cox model, the baseline hazard remains unspecified. This presents the analysis with
some problems, such as some loss in the efficiency of the estimators, because some
information may be left out. However, this efficiency loss is generally small and can
disappear completely in asymptotic results (Moeller, & Molina 2003).
However, there have been various models which attempt to determine the shape of h0.
These models are known as fully parametric, with the simplest being the exponential which
assumes that h0 is constant over time and equal to . Other common functional
specifications of h0 are the Weibull, the Gompertz, the loglogistic, etc. Their main difference
from the exponential is that they assume the value of h0 to vary with time. The Weibull for
example assumes that the baseline hazard is a monotonic with respect to time, while the
lognormal assumes that the hazard increases and then decreases. Nevertheless, these
methods are not going to be used in this paper.

13

DATA
The two primary databases used in this research are the World Bank Development
Indicators, and the IEA database on Renewables. The IEA Renewables database has
information on renewable electricity production and for 134 countries for the period 19902010. The WDI have a similar amount of countries and a wider range of indicators, and a
larger time span, but we are limited by the dependent variable that comes from the IEA
database. Some of the data like the FIT indicator or the green party were collected from a
variety of web resources, as there is not any freely available database.
The following table illustrates the key summary statistics of our dataset:

14

RESULTS AND INTERPRE TATIONS


The assumption is technology improves over time, and by improvement a decrease in the
cost/increase in the benefits is implied. Thus, it could be assumed that the risk of failure (i.e.
the hazard) increases over time.

NON-PARAMETRIC ANALYSIS

The above two graphs illustrate the Kaplan-Meier failure and survival estimates. The KaplaMeier failure plot illustrates the probability of a subject failing at time t given that it has
survived up to time t. In this case, the probability that a country starts using wind energy in
year 10 (i.e. year 2000) is somewhere between 15-25%3. Similarly, by looking at the survival
estimate, it can be inferred that the probability of not adopting wind by year 10 is
somewhere between 75-85%.
However, this analysis implicitly assumes that time is the sole determinant of adoption, an
assumption which clearly paints a distorted picture of reality. Time in itself does not
determine anything; rather, as time passes other variables change, which in turn have a
causal relationship to the probability of adoption of wind. To capture some of this factors,
semi-parametric analysis is used, the results of which are presented in the following section.

SEMI-PARAMETRIC ANALYSIS
The advantage of Cox models is that they do not restrict the baseline hazard into some
predetermined shape, thus allowing for a more flexible specification of the model. In
particular, the baseline hazard can be determined after specifying the model.
A total of 240 models were tested, from which roughly 50% succeeded in the various
specification tests; the indicators that were tested are tabulated below:

3 The shaded area represents the pointwise confidence bands of the Kaplan -meier function, which is the solid blue.

15

Starting with the level of emissions, the results suggest that it is positively related to the
hazard rate, i.e. the higher the level of CO2 per GDP the higher is the probability of adoption
of wind. The explanations for this result are multiple, but the most probable argues that the
higher the Co2 per GDP, the less clean the countrys electricity production is, and the higher
the pressure that is placed to the country by agents to clean its production.
The actual impact of these agents is captured in the model, and in particular by the
statistically significant hazard rate of the green party dummy. The existence of a green party
in the country increases the probability of adoption of wind by 15.68%. Nevertheless, EU
membership or the signing of the Kyoto protocol have proven statistically insignificant
factors in determining the probability of adoption.
An argument could be made that the change agent per se does not have an impact in the
countrys decision to adopt wind, unless the country is open and democratic. Therefore, the
polity2 indicator has been used. The indicator was used in the model in two ways. The first
was as a stand-alone indicator, and the other one was as an interaction with the dummy
variable for green party. In the first case, the aim was to capture the effect of democratic
openness on the probability of adoption; nevertheless, the indicator was statistically
insignificant, implying that the probability of first adoption of wind does not depend on the
level of the countrys democracy. In the second formulation, the aim was to moderate the
impact of green party; in other words, it was assumed that the impact of green party will
vary with the countrys freedom. The more open the country the stronger the impact of
green party on the society, and thus the higher the probability of adoption. Nonetheless,
this was not confirmed by the data, and the explanation might lie on the phenomenon
under observation. In other words, the aim of this model was to examine the determinants
of the speed of first adoption, rather than the determinants of full diffusion. Thus, even if
there is not an open democratic environment, just the existence of one green party might
reflect some kind of environmental awareness in the country, which could be enough to
incentivize the country to adopt at least one wind turbine. To allow for wind technology to
spread to the country however, the strength of the change agent (green party in this case)
must be significant, and thus the interaction variable could become statistically significant.
Another important determinant from literature is the size of the market; the larger the size
of the market the larger is the potential for profit for the supplier of the innovation, and thus
the larger his effort is in attempting to promote that technology in the market. In this case,
we assumed that the market is the electricity market, and we proxied that by the amount of

16

electricity that is consumed per capita. Although the absolute size of the market could have
been used, we assumed that the important factor for first diffusion would not be the actual
size of the market, but how energy rich was that country per each citizen. For example, even
though Russia consumes almost 25 times the amount of electricity consumed by Denmark,
we should expect that the supplier of the technology would see that the profit potential in
Russia is 25 times that of Denmark. However, if we compare the electricity consumption per
capita, then we see that the two countries are almost identical, a finding that yields a much
more realistic depiction of reality. Consequently, our model confirmed that the higher the
amount of electricity per capita, the larger is the probability of adopting wind.
On a similar note on profitability, apart from the profitability of the supplier of the
innovation another at least equally important factor suggested by the diffusion literature is
the profitability of the buyer of the innovation. In this case, the profitability was captured by
three factors: the price of the wind turbine, the level of Feed-in Tariffs, and the price of
crude oil. The price of the wind turbine and crude oil were not statistically significant, while
the level of FIT was. The first finding is somewhat surprising, but the explanation probably
lies in the quality of data available. The only available data was on Danish wind turbines
from 1989 to 2001. Then these figures were converted to local currencies, and we assumed
that these are the prices each country faced when deciding whether to purchase the turbine
or not. However, this is an unrealistic depiction of reality, since a significant amount of costs
for wind turbines are the transport costs, which vary significantly with distance.
As far as FITs are concerned, it is a fact that wind technologies are significantly more
expensive than conventional fossil fuel technologies. Therefore, it is a widely accepted fact
that for their adoption and diffusion there is need for government policy. Although there are
various ways of government intervening and promoting wind energy, one of the most
effective is based on the use of market instruments, and in particular the use of FIT. This
general finding is also confirmed in this paper, which argues that the implementation of FIT
in the country increases the probability of adoption of wind by 17%.
Continuing the investigation into the profitability aspects of adoption, the gap however
between wind and conventional fossil fuel technologies decreases with time, mainly because
technology improves, but also because the price of fossil fuel increases. To capture the
dynamics of this changing gap, the price of wind turbines was used as it was previously
mentioned, and the assumption made was that as the technology improves, the cost of the
turbines will go down. Apart from cost, the efficiency of turbines should also have been
captured, but no satisfactory data were found. Moreover, to capture the effect of changing
of fossil fuel prices, the price of crude oil was used, since gas and coal are a bit harder to
capture. Nonetheless, the price of oil also seemed insignificant, maybe because the world
price was used at local currencies rather than the local price converted into US dollars.
To capture the observability criterion, we tried three different indicators. One was distance
from Denmark, which is arguably the pioneer in wind energy; we argue that the closer a
country to the innovator, the higher is the probability that it will adopt wind. However,
physical distance is not enough; for example a country might be neighboring to another, but
they might have no trade relationships. Thus, the amount of trade with Denmark was also
tested. Neither of the two indicators was found to be statistically significant on their own, so
a third alternative was tested, which was the interaction of the two terms. Distance was
then moderated by trade intensity, and the results suggest that for two countries with the

17

same distance from Denmark, the one with the highest level of trade has the higher
probability of adoption, thus confirming Rogers observability criterion.
A prominent issue in technology diffusion studies is the issue of technological lock-in. The
stronger this lock-in, the harder is for the economy/sector to move to an alternative
technological configuration. In the context of wind adoption, two types of lock-in were
identified, a technological and a market lock-in. The technological refers to how much the
economys electricity sector is based on traditional fossil fuel technologies. This was
captured the indicator on energy use as a % of GDP. The more energy inefficient the country
is, the more we assume it is based on old technologies, and thus the higher the resistance of
incumbent energy players to change. This negative relationship between energy intensity
and probability of adoption was confirmed by our model.
Similarly, the other element of resistance to change was the market lock-in. An economy or
sector is locked-in a particular technology if it is making an accounting profit by continuing
to use that technology. In the case of the energy sector, an economy might be finding
profitable to use conventional fossil fuel technologies and these may account for a
significant amount of its GDP, thus preventing it from shifting to alternative renewable
technologies such as wind. Nevertheless, energy markets suffer from externalities, as many
of the costs of fossil fuel technologies are not captured by the market price.
This phenomenon of market lock-in was explicitly captured by the indicator of Natural
Resource rents as a % of GDP, which is a sum of oil rents, natural gas rents, coal rents (hard
and soft), mineral rents, and forest rents. Ideally, we should use an indicator which focuses
solely on oil, natural gas, and coal rents, but we expect a country with high reserves of oil,
gas and coal is likely to have mineral and forest reserves; thus, this indicator should capture
similar effects. The coefficient of this indicator was statistically significant, and its impact
was as expected, i.e the higher the countrys market lock-in, the lower the probability of
adoption of wind.
Another issue, which is particular to the energy sector is the issue of energy dependence.
The vast majority of countries have inadequate fossil fuel reserves to cover all their energy
needs, so their energy supply depends on imports from other countries. This poses
considerable risks, which have been widely investigated from a wide range of social
sciences: from geopolitics and international relations, to energy economics and political
economy. Moreover, the importance of this factor is also evident by the extent to which this
subject has dominated the agendas of almost all developed energy-importing countries,
particularly those of the EU.
A way that a country can decrease its energy dependency is either by decreasing its energy
needs and/or increasing its domestic energy production. Renewable energy, and wind in
particular contributes to the increasing of domestic production, assuming that the country
has adequate wind resources. This argument was supported by the model which found a
positive relationship between Imports of Energy as a % of total energy and the probability
of adoption.
Another determinant in diffusion and innovation studies is the economys technological
capability. There is a very wide and diverse literature on the ways to measure technological
capability, but two of the most conventional, and widely available indicators are the
economys GDP per capita, and the years of schooling. In this case however, neither of these

18

two were found to be significant, a finding that maybe has to do with the process that we
examine, which is simply the first adoption rather than full diffusion.
Lastly, the amount of wind resources in the country has exerts a small but positive influence
on the probability of adoption. This indicator classifies a countrys land area into 10 different
groups according to the full load wind hours. In other words, it measures how many hours a
wind turbine could work at full capacity. Clearly, the larger the amount of land at the higher
full load wind hour group, the higher the amount of wind. The results suggest that the
higher the amount of land area the higher is the probability of adoption. This is a widely
expected result, since the more of the resource the country has, the more attractive the
technology is, assuming that it will be more profitable.

19

CONCLUSION
The scope of this paper was the identification of the factors that determine the first
adoption of wind power by a country. Despite the fact that a great deal of research has been
written on the determinants of innovation and diffusion of renewable technologies, much
less emphasis has been placed on the determinants of first adoption.
This paper aims to cover this gap by building on the theoretical framework provided by two
major schools of thoughts: that of neoclassical economics and that of sociology. Griliches
and Rogers were chosen as the best representatives of these two schools, and then their
theoretical approaches were combined with the particularities of the energy sector to build
a theoretical model that can explain the determinants of first adoption of wind energy in
countries around the world.
Hazard models were then used to determine which factors can explain first adoption, and
the data sources were taken from a multiplicity of international sources. The results mainly
confirmed the neoclassical economic approach, suggesting the various elements of the
market as the most important determinants of profitability these included the profitability
of adopters, of suppliers, the availability of wind, as well as the carbon and market lock in.
Moreover, the existence of a change agent in the form of a green party and the countrys
trade with the pioneers of wind energy were also significant determinants, while all other
institutional determinants were found insignificant.

20

REFERENCES
Boyd, G.A. & Karlson, S.H., 1993, The impact of energy prices on technology choice in the
United States steel industry, The Energy Journal, 14(2), pp. 47-56.
Burt, R.S., 1973, The differential impact of social integration on participation in the diffusion
of innovations* 1, Social Science Research, 2(2), pp. 125-44.
Dechezleprtre, A., Glachant, M., Hascic, I., Johnstone, N. & Mnire, Y., 2010, Invention and
transfer of climate change mitigation technologies on a global scale: a study drawing on
patent data, Post-Print.
Nancy L. Rose & Paul L. Joskow, 1988. "The Diffusion of New Technologies: Evidence From
the Electric Utility Industry," NBER Working Papers 2676, National Bureau of Economic
Research, Inc.
Fisher-Vanden, K., Jefferson, G.H., Jingkui, M. & Jianyi, X., 2006, Technology development
and energy productivity in China, Energy Economics, 28(5-6), pp. 690-705.
Geroski, P.A., 2000, Models of technology diffusion, Research Policy, 29(4-5), pp. 603-25.
Gray, W.B. & Shadbegian, R.J., 1998, Environmental Regulation, Investment Timing, and
Technology Choice, The Journal of Industrial Economics, 46(2), pp. 235-56.
Griliches, Z., 1957, Hybrid Corn: An Exploration in the Economics of Technological Change,
Econometrica, 25(4), pp. pp. 501-22.
Grubb, M.J., 1990, The cinderella options:: A study of modernized renewable energy
technologies Part 2-Political and policy analysis, Energy policy, 18(8), pp. 711-25.
Jacobsson, S. & Johnson, A., 2000, The diffusion of renewable energy technology: an
analytical framework and key issues for research, Energy Policy, 28(9), pp. 625 - 640.
Jaffe, A.B. & Stavins, R.N., 1994, The energy paradox and the diffusion of conservation
technology, Resource and Energy Economics, 16(2), pp. 91 - 122.
Karshenas, M. & Stoneman, P.L., 1993, Rank, stock, order, and epidemic effects in the
diffusion of new process technologies: An empirical model, The Rand Journal of Economics,
pp. 503-28.
Kerr, S. & Newell, R.G., 2003, Policy-Induced Technology Adoption: Evidence from the US
Lead Phasedown, The Journal of Industrial Economics, 51(3), pp. 317-43.
Kiefer, N.M., 1988, Economic duration data and hazard functions, Journal of economic
literature, 26(2), pp. 646-79.
Lanjouw, J.O. & Mody, A., 1996, Innovation and the international diffusion of
environmentally responsive technology, Research Policy, 25(4), pp. 549 - 571.
Levin, S.G., Levin, S.L. & Meisel, J.B., 1987, A dynamic analysis of the adoption of a new
technology: the case of optical scanners, The Review of Economics and Statistics, pp. 12-7.

21

Lovely, M. & Popp, D., 2008, Trade, technology and the environment: Why do poorer
countries regulate sooner? NBER Working Paper, 14286.
Masini, A. & Menichetti, E., 2010, The impact of behavioural factors in the renewable energy
investment decision making process: Conceptual framework and empirical findings, Energy
Policy, In Press, Corrected Proof, pp. -.
Moeller, T. & Molina, C.A., 2003, Survival and Default of Original Issue High-Yield Bonds,
Financial Management, 32(1), pp. 83-107.
Nakicenovic, N., 2002, Technological change and diffusion as a learning process,
Technological change and the environment, pp. 160-81.
Neij, L., 1997, Use of experience curves to analyse the prospects for diffusion and adoption
of renewable energy technology, Energy Policy, 25(13), pp. 1099 - 1107.
Newell, R.G., Jaffe, A.B. & Stavins, R.N. 2004, The economics of energy efficiency, in C
Cleveland (ed), Encyclopedia of Energy, Elsevier, Amsterdam, pp. 79-90.
Oltra, V., 2008, Environmental innovation and industrial dynamics: the contributions of
evolutionary economics, Cahiers du GREThA, p. 27.
Oster, S., 1982, The diffusion of innovation among steel firms: the basic oxygen furnace, The
Bell Journal of Economics, pp. 45-56.
Popp, D., 2004, International innovation and diffusion of air pollution control technologies:
the effects of NOX and SO2 regulation in the US, Japan, and Germany, NBER Working Paper,
10643
Popp, D., 2006, R&D Subsidies and Climate Policy: Is There a Free Lunch? Climatic Change,
77(3), pp. 311-41.
Popp, D., Hascic, I. & Medhi, N., 2010, Technology and the diffusion of renewable energy,
Energy Economics, In Press, Accepted Manuscript, pp. -.
Popp, D., Newell, R.G. & Jaffe, A.B., 2009, Energy, the environment, and technological
change, NBER Working Paper, 14832
Rehfeld, K.M., Rennings, K. & Ziegler, A., 2007, Integrated product policy and environmental
product innovations: an empirical analysis, Ecological Economics, 61(1), pp. 91-100.
Rogers, E.M., 1988, Diffusion of innovations, 3rd Edition ed. The Free Press, New York.
Saloner, G. & Shepard, A., 1992, Adoption of technologies with network effects: an
empirical examination of the adoption of automated teller machines, The Rand Journal of
Economics, pp. 479-501
Shama, A., 1983, Energy conservation in US buildings* 1:: Solving the high potential/low
adoption paradox from a behavioural perspective, Energy Policy, 11(2), pp. 148-67.
Snyder, L.D., Miller, N.H. & Stavins, R.N., 2003, The Effects of Environmental Regulation on
Technology Diffusion: The Case of Chlorine Manufacturing, The American Economic Review,
93(2), pp. 431-5.

22

Somers, M.J. & Birnbaum, D., 1999, Survival versus traditional methodologies for studying
employee turnover: Differences, divergences and directions for future research, Journal of
Organizational Behavior, 20(2), pp. 273-84.
Unruh, G.C., 2000, Understanding carbon lock-in, Energy policy, 28(12), pp. 817-30.
Wstenhagen, R., Wolsink, M. & Brer, M.J., 2007, Social acceptance of renewable energy
innovation: An introduction to the concept, Energy Policy, 35(5), pp. 2683 - 2691.

23

S-ar putea să vă placă și