Ecological Economics 99 (2014) 121129

Contents lists available at ScienceDirect

Ecological Economics
journal homepage: www.elsevier.com/locate/ecolecon

Analysis

Measuring the impact of nuclear accidents on energy policy

Zsuzsanna Csereklyei
Department of Economics, Vienna University of Economics and Business (WU), Welthandelsplatz 1, 1020 Vienna, Austria

a r t i c l e

i n f o

Article history:
Received 29 October 2013
Received in revised form 5 January 2014
Accepted 10 January 2014
Available online 14 February 2014
JEL Classication:
C33
C52
Q43
Q48
Keywords:
Nuclear energy
Nuclear accidents
Energy policy

a b s t r a c t
This paper investigates the effects of nuclear accidents on energy policy with the help of a panel dataset of 31
countries from 1965 to 2009, using annual data on the capacity of reactor construction starts, as well as the timing
of three nuclear accidents scaled ve or higher on the International Nuclear and Radiological Event Scale. After
determining the extent of the accident impact in the different countries, I nd that neither the Three Mile Island
(TMI) nor the Lucens accidents had a worldwide negative effect on construction starts, while Chernobyl did.
Three Mile Island had a lasting impact in the United States, however. I show that the effect of Chernobyl wore
off in certain geographical clusters, after ten to thirty years. An accident is likely to have a negative and long lasting impact in the country where it happened, and possibly in countries affected by the direct consequences. I nd
that nuclear capacity enlargement shows a signicant lock-in effect, but it was also driven by primary energy
consumption and energy security considerations in the past ve decades.
2014 Elsevier B.V. All rights reserved.

1. Introduction
In the rst decade of the 21st century, until the accident at the
Fukushima Dai-ichi nuclear power plant occurred, the notion of a nuclear renaissance emerged (Ahearne, 2011; Goodfellow et al., 2011) and
many, mostly Asian countries started large-scale civilian nuclear programs. The Fukushima accident in 2011 however, was followed by immediate policy reactions against nuclear energy in a number of
countries, though many other countries decided to continue with their
nuclear programs. For example, while Germany chose to phase out nuclear power (Wittneben, 2012), the Chinese government decided to
proceed following a short moratorium.
Already in a 2009 study, Joskow and Parsons (2009) claimed that
another signicant accident at an existing nuclear power plant anywhere
in the world could have very negative consequences for any hope of a
nuclear renaissance. However, the true impact of this accident on
reactor constructions will be visible only in a few years. The goal of this
paper is to quantify the effect of past nuclear accidents, in order both to
understand how long and how severely nuclear accidents reduce the
construction of new reactors and to gain a clearer picture about the possible extent of the aftermath of the Fukushima accident. While the effects
of nuclear accidents on energy policy have been discussed in a number of
studies, among others in Ebinger (2011), Nohrstedt (2005, 2008),
Joskow and Parsons (2012), Thomas (2012) and in Goodfellow et al.

E-mail address: zsuzsanna.csereklyei@wu.ac.at.

0921-8009/$ see front matter 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.ecolecon.2014.01.010

(2011), most of these articles dealt with the psychology or the politics
of post-accident policy making, rather than quantifying the effects on reactor construction.
There are only a handful of econometric studies, such as Fuhrmann
(2012) and Gourley and Stulberg (2013) examining the driving forces
of reactor construction. Furthermore, the impact of accidents in this
context is rarely measured. Fuhrmann (2012) investigates the motives
of nuclear power plant construction using a logit model to test the
signicance of nuclear accidents, economic development, nuclear
proliferation, energy security, the supply of nuclear technology and
of norms as the determinants of nuclear power plant construction.
He comes to the conclusion that economic development and energy
insecurity are positively and signicantly inuencing reactor construction. Fuhrmann (2012) does not nd the nuclear weapons or
nuclear non-proliferation treaty variables signicant, implying that
enrolling in a civilian nuclear program does not mean sinister intentions from the beginning, although there is no evidence that these
intentions might not change. Unlike the Chernobyl and TMI accident
dummies, the supply side variable is insignicant. Fuhrmann (2012)
concludes that nuclear power plant construction is likely to continue
in countries which had invested in nuclear technology and infrastructure before the Fukushima accident, but the probability that
new countries will enrol in civilian nuclear programs is drastically
lowered.
Gourley and Stulberg (2013) investigate the correlates of nuclear energy using a binary logistic regression on the occurrence of construction
starts. They investigate the characteristics systematically shared by
existing nuclear power states that distinguish them from non-nuclear

122

Z. Csereklyei / Ecological Economics 99 (2014) 121129

states. They nd that energy insecurity and real income are positively
correlated with nuclear energy, however they do not nd evidence for
common traits in economic growth, political governance, regime duration, or strategic considerations such as enduring rivalries (hostilities).
They note that aspirant nuclear states tend to be either large emerging
market economies or smaller, but fast growing nations. Nelson (2010)
conducted a stepwise regression on nuclear reliance (dened as the
fraction of national electricity generated from nuclear energy). He
found that coal reserves and the state of the fuel cycle are negatively
correlated, while international commerce and polity and energy insecurity are positively correlated with nuclear reliance. Gross domestic
product, gas reserves or electricity generation were not found signicant. The methodology applied was questioned however by Gourley
and Stulberg (2013).
Guidolin and Guseo (2012) examine the existence of a nuclear renaissance using innovation diffusion models, and investigate among
others shocks to reactor grid-connections. The authors identify a negative shock in 1987 corresponding to the Chernobyl accident, followed
by a consistent decline. Especially in the wake of Fukushima, they foresee a declining future pattern for reactor startups. Guidolin and Guseo
(2012) also note that nuclear expansion presently is a phenomenon in
countries with highly centralised government structures.
This study contributes to closing the gap in the empirical literature
by examining the statistical impact of the main accidents on reactor
construction starts worldwide, while controlling for energy consumption, and for a number of economic and strategic factors. The method
is novel, as the length of impact of each accident is allowed to vary country by country, thus accounting for their diminishing impact, where appropriate. Previous studies used accident variables running uniformly
from the date of the accident to the end of the examined period, or for
a uniform number of years. Also, no other study has utilised the capacities of construction starts before, which allows for a more accurate investigation of the subject and a different econometric approach.
Earlier papers used a binary variable approach to account for construction starts.
Of the three accidents examined in this paper, only Chernobyl had a
signicant negative worldwide impact on nuclear power plant construction, while the effect of TMI was signicant primarily in the
United States. I nd that the effect of nuclear accidents can, but need
not wear-off in a time span of ten to thirty years. Thus, an accident is
likely to have a long lasting negative impact in the country where it happened, and possibly in regions most affected by the nuclear fallout. I also
nd that next to energy consumption and energy security, the lock-in
effect is a very strong driving factor in the nuclear industry. This paper
is organised as follows: Section 2 examines the economic and policy environment of nuclear energy during the past ve decades. Section 3 discusses the data and the methodology used, Section 4 explains the
empirical results, and Section 5 presents the conclusions.

2. Five Decades of Commercial Nuclear Energy

This section looks closely at the historical and economic factors that
shaped the nuclear industry as well as the surrounding energy policy
during the last fty years. In the wake of the Second World War, civilian
reactor technology was initially developed from the military applications of nuclear power. The International Atomic Energy Agency
(IAEA) was established in 1957 to promote the safe, secure and peaceful
uses of nuclear technologies. The rst commercial reactors appeared in
the rst half of the 1950s in the US, in the USSR, and in West Europe.
Arising from the novelty of the technology, many different reactor designs were experimented with, in this decade (Lester and Rosner,
2009), a number of which either did not prove economically viable, or
other technically better designs were favoured instead. This was the
era of the Cold War, when nuclear armament was a strategic priority
for both the West and the Eastern Bloc.

A rapid worldwide commercial expansion of nuclear reactors

followed at the end of the 1960s, irrespective of the ownership structure
of the plants (state vs. privately owned plants). However, the 1970s
brought a signicant change to the international energy landscape.
The oil crisis of 197374 sent the ination and interest rates skyrocketing in the Western World. As a result of high energy prices, the
demand for end user energy services broke in considerably, while energy efciencies began to kick-in. This phenomenon at the time was not
observable in the countries of the Eastern Bloc, as the USSR not only
had large own oil and gas reserves, but also regulated resource and energy prices. It is argued that in an environment of high interest rates and
high ination, large base load power plants with signicant initial investment costs became especially uncompetitive. According to Ebinger
(2011), the US demand for large base load plants was growing annually
67% in the 1960s, but started falling 12% after the rst oil crisis. Apart
from this, a very complicated authorisation process evolved, with the
federal and state regulators all doing their independent reviews on construction and operational licenses that also increased the time needed to
build a nuclear plant (Ebinger, 2011).
By 1979 the nuclear industry was already ghting with overcapacity
in the United States, high investment costs and ination, growing public, environmental, and proliferation concerns. It was in this year that
the Three Mile Island accident happened. The release of radioactivity
was well under the health safety limits, and even after thirty years, no
deaths or illnesses are attributable to TMI (Van Roey, 2009). The accident however received tremendous media attention, and a local public
panic ensued. Contributing factors could have been the release of the
lm China Syndrome two weeks before the event, or an evacuation
order given by the nuclear regulatory body based on false information
regarding the measurement of released radioactivity that was later
withdrawn. According to Ebinger (2011), after TMI nearly 100 orders
for nuclear reactors were cancelled in the United States, and only one
reactor was nished. Many investments were scrapped, among others
nearly-ready plants (Joskow and Parsons, 2009). It is argued that TMI,
although it had a large echo, has just nished off what the stagation
and high interest rates of the decade have started (Cohen, 1990). The
worldwide number of nuclear power plant construction starts had fallen already starting with 1977. TMI also likely triggered the Swedish
referendum on nuclear energy (Nohrstedt, 2005), which resulted in a
long-term nuclear phase out decision.
The beginning of the 1980s was marked once again by a severe oil
crisis, resulting in signicant reductions in end-energy use, as well as
in primary energy consumption. The effects were again not visible in
the Eastern Bloc, for the reasons outlined above. Oil prices broke in however considerably in 1986. It was in these circumstances that the most
severe nuclear accident of all times happened in Ukraine. The disaster
at Chernobyl was due not only to a bad reactor design, but also to enormous human errors, and to the complete disregard of safety procedures
(Villa, 2008). The accident contaminated signicant areas of Europe,
and caused a large public outcry against nuclear energy. Apart from a
few countries that kept their nuclear programs, in most parts of the
world, new construction starts fell dramatically.
The last twenty years were characterized worldwide by the wide
spreading of small and medium capacity gas and coal plants
(Joskow and Parsons, 2009), and by the remarkable increase in
commodity prices starting in the early 2000s until the great nancial depression of 2008. Liberalisation and quasi-liberalisation of
electric markets sped up not only in Western but also in Eastern
Europe and in the former USSR. Especially in the Eastern Bloc, the
adjustment to market prices caused drastic increases in energy
prices, along with a very strong downward pressure on energy consumption. The primary energy consumption of the former Eastern
Bloc decreased by 2010 to the level of the 1970s, after a peak around
1990. In the US, deregulation marked the 1990s, but these efforts
suffered considerably as a result of the California Electricity Crisis
in 20002001.

Z. Csereklyei / Ecological Economics 99 (2014) 121129

During the past decade, due to the high and often uctuating oil and
gas prices, as well as to the increasing concerns related to climate
change and to the looming problem of energy security, many states
again realistically considered utilising nuclear power (Joskow and
Parsons, 2009; Lester and Rosner, 2009). Among others, the US initiated
a stimulus package to boost the building of new nuclear power plants.
China, India, and a number of Asian states also turned to nuclear energy
to satisfy their strongly growing energy needs. China passed a Medium
and Long-term Nuclear Power Development Plan Zhou et al. (2011),
running until 2020, to build new nuclear capacity of around 40 GW(e).
Later these plans were reassessed and are currently around 70
80 GW(e), almost 78 fold of the existing nuclear capacity in 2010
(Zhou et al., 2011). The conditions faced by the nuclear industry in
China compared to those in the Western economies are distinctly different. While in China the industry is characterized by competing state
owned corporations, in the United States and in parts of Europe the privately owned nuclear industry is competing with natural gas and shale
gas plants, with much lower initial investment, faster construction
times, and less regulatory hazard. How much this nuclear renaissance
was a true renaissance in the West, or rather a hype is questioned by
many. Among others Glaser (2011), Joskow and Parsons (2012) and
Thomas (2012) argue that the renaissance failed due to technoeconomic issues already before the Fukushima nuclear accident in 2011.
The discussion whether nuclear energy will play a signicant role in
climate change mitigation has been extensive over the past years. The
question remains, whether enough nuclear capacity is expected to
come online in the future to contribute signicantly to reducing greenhouse gas emissions. While nuclear energy has considerably lower
emissions as fossil energy sources, and causes no air pollution
(Adamantiades and Kessides, 2011), the Fukushima accident is likely to
affect both the willingness to build new reactors, and the costs of new
plants due to the increased safety measures and longer construction
times (Hayashi and Hughes, 2013a). Apergis et al. (2010) nd that nuclear energy contributes to emission reductions both on the short and on
the long-run. At the same time, while nuclear energy offers an option
to mitigate climate change, climate change itself impacts on the costs
of nuclear power, (Kopytko and Perkins, 2011) as nuclear plants must
adapt to changing environmental conditions, such as inland oods or increased sea levels. The lingering challenges of nuclear waste disposal
(Adamantiades and Kessides, 2011), safety and security (Mez, 2012)
must also be addressed in a coordinated manner, if nuclear energy is to
play a signicant role in mitigating climate change. Newer reactor designs promise to offer a better solution to the latter two. The next sections investigate the impact past accidents had on reactor construction
in different parts of the world, as well as the prospects of future reactor
construction.
3. Data and Methods
3.1. Data
The main analysis is based on an unbalanced panel dataset for 31
countries covering the period from 1965 to 2009. All the examined
countries have presently, or had during the examined period, civilian
nuclear capacity. The dataset includes Argentina, Belgium, Brazil,
Bulgaria, Canada, China, Czech Republic, Finland, France, Germany,
Hungary, India, Iran, Italy, Japan, Kazakhstan, Republic of Korea,
Lithuania, Mexico, Netherlands, Pakistan, Romania, Russia, Slovak Republic, South Africa, Spain, Sweden, Switzerland, Ukraine, the United
Kingdom, and the United States.
The data related to the capacity of nuclear power plants and construction starts originates from the International Atomic Energy
Agency (2011). Power plants abandoned in-construction after Three
Mile Island are not included, as only limited data are available on
these plants. It is well-known that in the 1950s and 1960s many experimental design nuclear plants were built, a good portion of which were

123

shut down shortly for different reasons. This turbulent, early period of
commercial nuclear history is however excluded from this paper, by setting the sample period to start with 1965.
This study considers accidents rated INES 5 or higher. The International Nuclear and Radiological Event Scale (INES) was adopted in
1990 (International Atomic Energy Agency, 1990). The scale encompasses seven levels, of which levels four (4) to seven (7) are classied
as accidents, and levels two (2) to three (3) as incidents. In practice,
the classication of an accident or incident is carried out by the national
authorities or by the operators of nuclear power plants. A potential bias
from using the INES scale arises from the fact that the INES database was
only developed in the early 1990s and does not contain information on
events prior to that (International Atomic Energy Agency, 1996), save
for a few spectacular cases. There were a number of accidents or incidents from the 1950s to the 1980s, where classication is currently
not available, although experts might rate them up to level ve (5). To
properly account for the potentially vanishing impact of accidents, I employ a series of individual regressions, estimating the time length of the
impact, which may differ from country to country. The accidents chosen
are Chernobyl (level 7) and TMI, as well as the Lucens accident in
Switzerland, which is the only public INES 5 accident next to TMI in
the examined time period. The Lucens accident in 1969 involved a
small 8.3 MW(e) pilot reactor meltdown, but resulted in no radioactive
contamination of the environment. There were two other nuclear accidents in the USSR in the sample period that would rate level 5 or 6
(Smythe, 2011), but were not made public and therefore not known
at the time (Chernobyl 1982, Lake Karachay 1967). All other INES 5 or
6 accidents happened before the examined period and therefore are excluded from the dataset.
Primary energy consumption data originates from the British Petrol
(2010) Statistical Review of World Energy database, from 1965 to 2009.
Economic variables, such as gross domestic product and economic
growth are sourced from the Penn World Table 7.0 (Heston et al.,
2011). It is often argued that high interest rates, ination and overcapacities have shuttered the nuclear industry years before Three Mile Island (Cohen, 1990; Ebinger, 2011). Therefore I test the effects of real
interest rates jointly with ination on the annual construction starts,
where possible. Ination and real interest rates between 1965 and
2009 are taken from the The World Bank (2012). Both datasets are unfortunately incomplete, often starting in the 1980s and in the mid 1990s
for the countries of the former Eastern Bloc.
Strategic security considerations, such as energy security of a country are also claimed to play a signicant role in driving nuclear power
plant construction (Nelson, 2010; Fuhrmann, 2012; Gourley and
Stulberg, 2013). The energy security variable for the period of 1965
2009 is constructed as in Fuhrmann (2012), by taking the World Bank's
measure of a country's net energy imports (ENSEC1) as a percentage of
its primary energy use, multiplying it with minus one, rescaling and logging it. Unfortunately the energy security variable for many Warsaw
Pact countries starts only in the 1990s.

ENSEC
log


ENSEC
1 :
100

The descriptive statistics relating to the main variables are found in

Table 1.

1
Net energy imports are estimated as energy use less production, both measured in oil
equivalents. A negative value indicates that the country is a net exporter. Energy use refers
to use of primary energy before transformation to other end-use fuels, which is equal to
indigenous production plus imports and stock changes, minus exports and fuels supplied
to ships and aircraft engaged in international transport (World Bank Database: World Development Indicators).

124

Z. Csereklyei / Ecological Economics 99 (2014) 121129

Table 1
Descriptive statistics of main variables.

Mean
Median
Maximum
Minimum
Std. dev.
Skewness
Kurtosis
Observations

Srescaled

317.99
1.00
23,246.00
1.00
1,252.89
8.78
113.94
1450

204.60
70.93
2,432.20
6.05
388.51
3.83
18.18
1346

998,193.80
361,448.40
13,191,482.00
22,846.07
1,713,977.00
4.05
22.62
1224

ENSEC*
0.54
0.46
2.80
2.38
0.75
0.18
3.69
1217

Notes: Descriptive statistics of the pooled variables, S = construction starts in a given year
(MWe), E = primary energy consumption in given year (MTOE), Y = real income in a
given year (million USD), ENSEC* = rescaled measure of the World Bank's measure of
net energy imports.

approach performs just as well or better than many alternatives already

at T = 30, using the RMSE criterion. Alternatives such as the generalised
method of moments (GMM) procedure might produce a signicantly
higher bias, when too many instruments are used for a given crosssection size (Doornik et al., 2001). Judson and Owen (1999) recommend
the usage of the LSDV (xed effects model) over the panel size of T =
30, for unbalanced macroeconomic panel data.
The dynamic xed effect specications are estimated as follows:
Construction starts (Si,t) measured in MW(e) capacities is the main
dependent variable, which is regressed on the lag of the dependent variable to control for autoregression, and on the lag of the primary energy
consumption (Ei,t 1), country (i) and period xed effects (t). Accident Model I is estimated thus as:
LogSi;t 0 1 logSi;t1 2 logEi;t1 i t i;t :

3.2. Methods
In the next Sections I examine the impact of these three accidents on
reactor construction starts over the past decades, while controlling for
the major macroeconomic variables. To determine the best modelling
approach, I investigate the underlying time-series properties of the chosen variables by using the CIPS panel unit root test from Pesaran (2007),
which allows for cross sectional dependence in the heterogeneous
panel, caused by unobserved common shocks. Instead of basing the
unit root tests on deviations from the estimated factors, Pesaran
(2007) augments the DickeyFuller regressions with the cross section
averages of the lagged levels and rst-differences of the individual series. The panel unit root statistic is thus the average of the cross sectionally augmented ADF statistics. The results of the CIPS test can be seen in
Table 2.
While the dependent variable, construction starts shows a strong
autoregressive feature in line with the lock-in theory, the variable is
clearly stationary, therefore a cointegration approach is not a viable option. As a result, the econometric approach used throughout the paper is
a dynamic panel model with xed effects, estimated with ordinary least
squares methodology, using White period standard errors, robust to arbitrary heteroskedasticity and within cross-section serial correlation. A
possible caveat of this approach is that the lagged dependent variable
will be correlated with the error term (Baltagi, 2005). The consistency
of the within estimator, according to Baltagi (2005) will thus depend
on the time dimension (T) being large. Judson and Owen (1999) conducted large scale Monte-Carlo experiments to determine the best
methodological choice for dynamic macroeconomic panel data. They
come to the conclusion that the least squares dummy variable (LSDV)
Table 2
CIPS unit root tests.
Variable

logE
logY
logS
log
logENSEC
logr

Order

0
1
0
1
0
1
0
1
0
1
0
1

No or trend

Intercept ()

and trend

Statistics

Prob

Statistics

Prob

Statistics

Prob

1.19
3.38
1.14
3.07
4.14
7.27
2.09
5.26
1.23
3.30

N0.10
b0.01
N0.10
b0.01
b0.01
b0.01
b0.01
b0.01
N0.10
b0.01

1.63
4.10
1.45
3.29
4.35
7.20
2.98
5.16
1.40
3.55
3.10
6.24

N0.10
b0.01
N0.10
b0.01
b0.01
b0.01
b0.01
b0.01
N0.10
b0.01
b0.01
b0.01

2.42
4.15
1.82
3.41
4.53
7.13
3.48
5.05
2.07
3.85
3.20
6.24

N0.10
b0.01
N0.10
b0.01
b0.01
b0.01
b0.01
b0.01
N0.10
b0.01
b0.01
b0.01

Notes: 0 Order of differencing: 0 = level, 1 = 1st difference.

Test specication: yi;t i i yi;t1 i yt1 i yt i;t (where i = (1 i),i =
N
(1 i) and yi,t = yi,t yi,t 1). CIPSN;T j N 1 i1 CAD F i; f H0: Existence of a Unit
Root in the Series (H0 : i = 0 for all i). Sample: 30 cross sections, excluding Kazakhstan
for computational reasons.
2
Real interest rate (r): tested with ordinary IPS (ImPesaranShin) unit root test.
3
User Specied Lags: 1.
1

The construction starts variable was rescaled by adding one MW(e)

capacity to each year, to allow the log-transformation. The results can be
seen in Table 3. The purpose of this specication was to extract and examine the time xed effects from the model.
To construct an accident dummy that takes into account the different
and potentially diminishing impact of nuclear accidents on my dataset,
rst I examine the individual regressions for all countries, allowing for
the impacts of the nuclear accidents to last for varying times.
LogSi;t 0 1 logSi;t1 2 logEi;t1 3 LUC 141 4 TMI131
5 CHER124 i i;t :

I let the Lucens dummy to take on the value of 1 in successive time

periods, allowing for differing impact extents. The variable thus takes on
a positive value rst in 1969, then in 1969 and 1970, and then progressively covers the entire sample period between 1969 and 2009. Similarly, I allow for the impact of the TMI accident to run from 1979 up to 2009
and for Chernobyl, to run from 1986 up to 2009. What the optimal
length of impact of an accident is, taking into account the other accidents, the constructions of the previous year and the energy consumption of the previous year, I ascertain by running the entire array of
models, in this case 41 31 24 = 30,504 regressions for each country.
The optimal length of impact has been determined by using the AIC
sample selection criterion, and the results are found in Table 4. Afterwards, I construct a pooled variable for each accident dummy, including
the optimal impact length for each country, and examine the impact of
these accidents on the construction starts, without (Accident Model II)
and with a time trend (Accident Model III).
LogSi;t 0 1 logSi;t1 2 logEi;t1 3 LUC i;t 4 TMIi;t
5 CHERi;t i i;t

LogSi;t 0 1 logSi;t1 2 logEi;t1 3 LUC i;t 4 TMIi;t

5 CHERi;t i 8 t i;t :

In the next section, I will analyse the output of these equations, and
test the robustness of the results by including additional variables,
claimed to have an impact on reactor constructions. These variables
are explained as they are introduced.
4. Results: The Impact of Nuclear Accidents
The regression outputs from Eqs. (3), (4) and (5) are found in
Table 3. Accident Model I corresponds to Eq. (3), while Models II and
III incorporate the pooled accident dummies in Eqs. (4) and (5).
Both the rst lag of primary energy consumption and the lag of the
construction starts are highly signicant, with an expected positive
sign. The coefcients indicate that 1% increase in the primary energy
consumption of the previous year results between 0.3% and 1.3%
increase in nuclear power plant construction starts measured in

Z. Csereklyei / Ecological Economics 99 (2014) 121129

125

Table 3
Accident impact on reactor construction.

Accident Model I
Accident Model II
Accident Model III

logSi,t 1

logEi,t 1

4.99
(0.00)
0.74
(0.59)
3.18
(0.02)

0.31
(0.00)
0.37
(0.00)
0.31
(0.00)

1.30
(0.00)
0.37
(0.25)
1.23
(0.00)

LUCi,t

0.34
(0.35)
0.04
(0.91)

TMIi,t

0.02
(0.96)
0.15
(0.67)

CHERi,t

0.82
(0.01)
0.54
(0.06)

Trend

0.05
(0.00)

Incl

Incl

Incl

Not incl

Incl

Not incl

Notes: St 1 = construction starts in year t 1, Et 1 = primary energy consumption in year t 1, LUCi,t = Lucens accident impact, TMIi,t = Three Mile Island Accident Impact,
CHERi,t = Chernobyl accident impact, i = country xed effects, t = period xed effects. Robust p-values are in parenthesis.

Table 4
Accident impact on reactor construction per country.

Argentina
Belgium
Brazil
Bulgaria
Canada
China
Czech Republic
Finland
France
Germany
Hungary
India
Iran
Italy
Japan
Republic of Korea
Mexico
Netherlands
Pakistan
Romania
Russia
Slovak Republic
South Africa
Spain
Sweden
Switzerland
Ukraine
United Kingdom
United States

logSt 1

logEt 1

L-IL

LUC

TMI-IL

TMI

C-IL

CHER

9.85
(0.12)
8.57
(0.27)
1.80
(0.57)
1.26
(0.40)
21.47
(0.02)
7.73
(0.02)
5.92
(0.24)
5.22
(0.11)
24.64
(0.15)
20.92
(0.43)
4.27
(0.33)
1.97
(0.65)
0.32
(0.73)
2.09
(0.24)
31.03
(0.00)
1.43
(0.46)
5.28
(0.21)
0.00
(1.00)
4.49
(0.06)
0.64
(0.70)
6.73
(0.00)
4.53
(0.23)
5.02
(0.32)
15.08
(0.00)
28.63
(0.07)
11.69
(0.53)
0.00
(0.06)
5.50
(0.68)
5.37
(0.17)

0.11
(0.28)
0.00
(0.99)
0.30
(0.13)
0.57
(0.00)
0.32
(0.02)
0.40
(0.02)
0.03
(0.42)
0.00
(0.99)
0.10
(0.27)
0.28
(0.06)
0.00
(0.84)
0.02
(0.91)
0.19
(0.30)
0.00
(0.75)
0.20
(0.25)
0.11
(0.47)
0.06
(0.90)
0.00
(0.06)
0.26
(0.10)
0.19
(0.54)
0.12
(0.25)
0.40
(0.02)
0.27
(0.26)
0.04
(0.85)
0.15
(0.14)
0.04
(0.88)
0.00
(0.24)
0.03
(0.49)
0.22
(0.04)

2.39
(0.12)
2.82
(0.21)
0.52
(0.57)
0.42
(0.40)
2.83
(0.11)
1.70
(0.00)
1.61
(0.23)
1.98
(0.12)
6.22
(0.06)
2.34
(0.61)
1.36
(0.32)
0.51
(0.49)
0.44
(0.34)
0.44
(0.24)
4.79
(0.00)
1.58
(0.00)
2.03
(0.17)
0.00
(0.84)
0.95
(0.08)
0.05
(0.92)
0.26
(0.26)
1.61
(0.22)
1.80
(0.29)
3.11
(0.00)
7.44
(0.07)
3.01
(0.63)
0.00
(0.06)
2.33
(0.35)
0.30
(0.55)

1980

1.16
(0.32)
4.20
(0.00)
1.76
(0.15)
3.57
(0.02)
5.52
(0.00)
2.66
(0.00)
0.58
(0.19)
3.52
(0.01)
5.76
(0.00)
6.00
(0.00)
1.48
(0.29)
3.02
(0.02)
1.12
(0.31)
3.43
(0.19)
4.06
(0.00)
2.53
(0.01)
1.98
(0.16)
6.18
(0.00)
2.67
(0.04)
1.38
(0.14)
5.81
(0.00)
2.72
(0.02)
1.56
(0.27)
3.55
(0.00)
5.62
(0.00)
2.86
(0.02)
6.86
(0.00)
6.80
(0.00)
1.05
(0.02)

1981

1.78
(0.32)
2.34
(0.06)
0.56
(0.58)
4.17
(0.01)
3.41
(0.01)
2.01
(0.03)
7.52
(0.00)
1.54
(0.23)
1.92
(0.06)
5.72
(0.00)
6.66
(0.00)
3.41
(0.00)
1.33
(0.30)
0.08
(0.23)
1.80
(0.03)
3.69
(0.00)
3.56
(0.15)
0.00
(0.25)
0.80
(0.04)
2.31
(0.05)
2.55
(0.00)
2.83
(0.02)
2.46
(0.26)
5.00
(0.00)
7.71
(0.00)
2.12
(0.17)
1.10
(0.00)
3.65
(0.22)
7.72
(0.00)

1996

0.72
(0.10)
2.96
(0.08)
0.28
(0.57)
3.80
(0.09)
5.42
(0.00)
2.07
(0.09)
3.57
(0.19)
0.71
(0.35)
8.81
(0.00)
7.32
(0.00)
0.24
(0.30)
2.85
(0.02)
1.71
(0.29)
0.14
(0.23)
4.51
(0.00)
5.46
(0.00)
4.46
(0.15)
0.00
(0.00)
1.28
(0.04)
0.84
(0.28)
7.68
(0.00)
3.03
(0.02)
3.33
(0.26)
1.07
(0.01)
0.54
(0.08)
1.64
(0.45)
0.00
(0.42)
2.23
(0.29)
0.05
(0.37)

1971
1977
1974
1969
2004
1986
1975
1970
1969
1974
2004
1974
1970
1983
1988
1975
1969
2004
1981
2005
1988
1975
1972
1975
1972
1987
2009
1979

1985
2009
1983
1979
2000
1979
2004
1991
1985
1979
1983
1985
1985
2007
2000
1985
2001
2006
1983
2007
1982
1985
1980
1980
1985
1986
1980
2009

2009
2009
1988
2009
1995
1987
2005
2009
2009
1992
1999
2009
2009
1993
2005
2009
2009
2005
2009
1986
1986
2009
1997
2002
2009
1989
1988
1995

Notes: St = construction starts in year t, Et = primary energy consumption in year t, LUC = Lucens accident impact, TMI = Three Mile Island Accident Impact, CHER = Chernobyl
accident impact, IL = length of impact per country. Robust p-values are in parenthesis. Lithuania and Kazakhstan have been excluded due to missing data.

126

Z. Csereklyei / Ecological Economics 99 (2014) 121129

TMI

CHER

LUC

-1

Period Fixed Effects

capacities, while controlling for the autoregressive parameter and the

effect of the accidents. The signicant positive persistence parameter
of the series supports the lock-in theory that countries, which have already invested in nuclear infrastructure are more likely to build nuclear
power plants. This supports the ndings of (Fuhrmann, 2012), and
those of Miller and Sagan (2009) and Gourley and Stulberg (2013),
who claim the presence of historical inertia, and note that nuclear entrants in the past were likely to be Warsaw Pact or North Atlantic countries. Apart from the evident historical reasons, the results might be
explained by the fact, that operating nuclear plants economically requires large economies of scale, well-trained personnel, and adequate
grid infrastructure, all of which need heavy initial investment.
Of special interest are the period xed effects in Accident Model I,
which are shown in Fig. 1, presenting a possible decrease of the positive
xed effects already in 197677, and the permanent turn in the sign of
the period xed effects in 1986. This means that while controlling for
the effects of both energy consumption and the persistence of construction starts, there was a global downward trend in the construction of
new power plants since 1986.
How long a nuclear event impacts on a certain country can be seen in
Table 4. Clearly, some of the impacts, picked up by the AIC criterion, are
highly questionable. It is unlikely that Lucens had a lasting impact on
Chinese, or Russian nuclear energy expansion. These results are driven
by the economic and political situation, and by the fact that nuclear
power plants were built only much later in China. Another bias we are
facing is that Russian and Ukrainian primary energy consumption data
are only available from 1985 on, therefore earlier years have been ignored by the regression in these countries.
Next, these accidents with differing impact lengths are tested for
signicance, while controlling for the autoregressive component of
power plant construction, primary energy consumption, and the other
accidents. A signicant positive impact of an accident means that the
years following the accident have not in fact caused a negative impact
on construction, but brought an increase with them. In the case of
1969 this is likely attributable to the Cold War, and the beginning of
the nuclear boom, and in the case of 1979 to the second oil crisis. We
see however that the model attributes a very strong signicant negative
effect to TMI in the United States, which is consistent with the prevailing
literature.
In interpreting the results, it is reasonable to assume that the driving
factors of nuclear power capacity development could have been
different in economies where the nuclear industry was in private
hand from the beginning, such as the US, and in planned economies,
with large state ownership, due to central party decisions or to inherent
energy security considerations. The case of France deserves also an explanation, as by the middle of the 1980s approximately 6570% of
French electricity production was covered by nuclear energy, with nuclear power plants occasionally being temporarily ran down to avoid
overcapacity. The decrease in nuclear power plant construction in
France thus could be more due to capacity issues than to Chernobyl.
Lester and Rosner (2009) and Guidolin and Guseo (2012) also note
that by today the French nuclear capacity exceeds baseload demand.

1970

1980

1990

2000

Year
Fig. 1. Period xed effects (t) from Accident Model I.

2010

The results using the pooled accident variable in Table 3 show a positive effect of the Lucens accident on reactor constructions, although the
result is not signicant. This implies that there was no negative worldwide effect of Lucens, but the years following the accident have brought
an increase in nuclear power plant constructions. The coefcient of TMI
is slightly negative but insignicant, implying that TMI had certainly impacted on the US nuclear industry, but not on all countries. Of most interest is the Chernobyl dummy, which is not only negative as expected,
but also highly signicant, showing massive reductions in new construction starts as the aftermath of this accident. In Accident Model III
we see that the trend is negative and signicant, as is the pooled Chernobyl dummy. Based on these results, I conclude that of all the examined accidents Chernobyl had a lasting and negative consequence on
worldwide nuclear power plant construction.
The more interesting question here is the length of the impact. The
signicant negative effect of Chernobyl stops for the countries China
(1995), India (1999), and South Korea (2005) after a time span of
nine to nineteen years. I cannot nd a negative effect of the Ukrainian
accident on Japan. Gourley and Stulberg (2013) also note that Japan
and Korea were building at a substantial rate when others were not.
This implies that issues of fossil resource scarcity and physical constraints on pipeline/transmission further increase the signicance of
the energy dependence question. The signicant negative effect of Chernobyl does not stop until the end of the sample period (2009), for
Belgium, Canada, Germany, whereas France, as mentioned earlier has
almost full nuclear capacity to supply its electricity needs, and has currently one reactor under construction. The dominance of European
countries in this group could be perhaps attributed to the proximity of
the accident. Other European countries such as Italy, Hungary,
Romania and Switzerland also have negative coefcients for the Chernobyl dummy, although these effects are not statistically signicant,
and the length of the impact is varying. An interesting case is Finland,
where not Chernobyl, but TMI inuences reactor construction until
2004. Finland was the rst country to decide upon the construction of
a European Pressurized Water Reactor in 2005, due to issues of energy
security and environmental concerns. Another noteworthy case is that
of the United States, where the construction of the rst plants since
TMI (Vogtle 3 & 4) has started, although this construction is not covered
by this sample.
Based on these ndings, the impact of an accident is likely to have a
long lasting negative effect in the country where the accident happened,
and possibly in countries which were affected by the direct consequences, such as the nuclear fallout. This is true for the United States
after TMI, but also for most of Europe after Chernobyl. No plant construction followed Chernobyl in the Ukraine up to now, and the rst
NPP in Russia was built only twenty years later, in 2006. Besides
China, Russia presently also holds ambitious nuclear plans. A correlation
analysis of the years of negative impact with the distance of capital cities
to Prypiat however has shown no discernible pattern.
4.1. A Note on Interest Rate and Ination
I test the robustness of the above results to the effect of ination and
real interest rates where possible. Among others Cohen (1990) suggested the importance of these variables due to the large nancing requirement of a nuclear power plant, especially under open market
conditions. A nuclear plant has to be authorized, built, and connected
to the grid, before it can start producing cash ows. This meant at the
beginning of the 1970s that utilities in the United States had to nance
approximately four years of construction time. Ination in the US has
been already high starting in 1973, with the rst oil crisis. Real interest
rates began to rise however in 1979 and stayed comparatively high
throughout the 1980s and even in the 1990s (Ebinger, 2011). TMI had
a regulatory aftermath in the US that foresaw stricter control and authorization process for nuclear power plants. Coupled with resistance from
the local population, this lengthened the construction time and

Z. Csereklyei / Ecological Economics 99 (2014) 121129

considerably increased the costs of nuclear plants. Additionally, banks

were suddenly unwilling to be exposed to the high risks of nancing nuclear investments (Thomas, 2012).
The real costs of an NPP in the USA and Europe have increased from
approximately 1700 $/KW to about 4500 $/KW (Harding, 2007) in the
last three decades. These high investment and nancing requirements
are often blamed to make nuclear energy uncompetitive against natural
gas or shale gas. On the other hand, the cost of nuclear power plants in
the Asia-Pacic region is considerably lower, around 23003000 $/KW.
This is partially due to different ownership and nancing structures, to
faster construction times, and also to lower labour and engineering expenses and standardisation. Joskow and Parsons (2009) claim that with
the accumulation of signicant construction experience everywhere in
the world, construction costs are possibly going to decline signicantly.
A good example is France which maintained comparatively low costs by
working with large economies of scale and a few suppliers (Morton,
2012).
Thus, I test where the data availability allows, the robustness of my
results to the inclusion of the joint impact of ination (i,t) and real interest rates (ri,t). A full time series is available for the United States and
for South Africa for the period between 1965 and 2009, and a partial
series for France (19652004), for Sweden (19702005), and for
Japan (19712009). Many other Western countries have statistics only
starting in the 1980s. Due to the political transitions, data availability
also starts for most countries of the ex-Warsaw Pact after the mid1990s, resulting in a loss of information and a bias. The results are
found in Table 5.
The individual regressions usually show both the real interest rate
and the ination coefcients to be negative, but insignicant. A panel regression also results in insignicant parameter estimates with values
around zero. My ndings relating to the impact of accidents and to the
lagged dependent variable representing the lock-in theory are robust
to the inclusion of interest rates and ination.
4.2. A Note on Real Income, Economic Growth, Energy Security
A number of macroeconomic variables, such as real income (Y), economic and energy consumption growth, or energy security were
claimed to be signicant drivers of nuclear plant construction in the
past. Therefore, I test in Table 6 the robustness of the earlier results to
the inclusion of real income, economic growth and energy consumption
growth to Accident Model II.
I nd the coefcient estimate of real income in the absence of primary energy consumption insignicant, and in the presence of energy consumption insignicant and negative. It would be dangerous to assume
however, that nuclear power plant construction is negatively driven
by real income. The reason for the negative coefcient could be that
real GDP and primary energy consumption were cointegrated over the
examined period (Csereklyei and Humer, 2012), and so due to a

127

misspecied multicollinear model, real GDP takes on a large negative, while primary energy consumption a large positive coefcient.
These results imply that a proper model specication would not include both real income and primary energy consumption. While
real GDP might be causing primary energy consumption, it is likely
that primary energy consumption contributes directly to nuclear
plant constructions, and not real GDP. I nd the earlier results on
the accident dummies robust to the inclusion of real income. The insignicant coefcient on real income in the absence of the energy
variable is in contrast with the ndings of Fuhrmann (2012), as
well as of Gourley and Stulberg (2013). A panel regression of economic growth on reactor construction brings also insignicant
coefcient estimates, with or without the presence of energy consumption. Substituting energy consumption growth (logEi,t 1 )
into the original model instead of the logarithm of energy consumption, brings positive but insignicant results, yet leaves the main
conclusions regarding the lock-in effect and the impact and signicance of the accidents unchanged.
Another crucial topic to be addressed is the question of energy security, as discussed by Fuhrmann (2012), Miller and Sagan (2009),
Gourley and Stulberg (2013), and Nelson (2010). Ensuring energy security in a country is relevant not only for the economy, but also for national security. In case of diplomatic or military hostilities, a country
that is heavily energy import dependent may be adversely disadvantaged without proper energy supply. Past decisions about nuclear strategy might have been driven by energy security considerations. For
example, Japan's long-term energy security concept was based on a
plan to substantially increase the use of nuclear-power (Hayashi and
Hughes, 2013b). Therefore I test the impact of energy security alongside
the nuclear accidents on power plant construction in Table 7, to see how
the results presented in Table 3 might change.
While the coefcient on the energy security variable is signicant
and negative at a 10% level, meaning that energy dependency (insecurity) contributes to nuclear power plant construction, the impact of accidents or the magnitude of their coefcients is almost unchanged. The
results thus support the view that civilian nuclear power has been
mostly driven next to the increasing energy demand, by historical
circumstances, as well as energy security considerations, rather than
economic factors. This is also in line with the ndings of Burke (2010),
who claims that own resources, that enhance energy security signicantly reduce the likelihood of nuclear power generation or investments in renewable energies.
A simultaneous regression of the energy security and real income
variables has not brought a change to either the signicance or the direction & magnitude of the coefcients. The Chernobyl accident resulted
in an enormous worldwide decrease in new nuclear plant construction
for decades, with the exception in countries where physical barriers to
energy imports made nuclear power to be a crucial energy source. Oil
shocks have been claimed to be a major motivation to engage in nuclear

Table 5
Impact of ination and real interest rate on reactor construction.

United States
Japan
France
South-Africa
Sweden
Panel sample

logSi,t 1

logEi,t 1

LUCt

TMIt

CHERt

ri,t

i,t

4.35
(0.42)
15.08
(0.68)
11.28
(0.51)
6.63
(0.31)
60.03
(0.22)
2.08
(0.43)

0.22
(0.05)
0.25
(0.20)
0.15
(0.25)
0.33
(0.25)
0.10
(0.33)
0.48
(0.00)

0.44
(0.55)
2.10
(0.73)
3.28
(0.32)
2.14
(0.29)
15.60
(0.22)
0.19
(0.72)

1.14
(0.00)
6.06
(0.02)
5.87
(0.00)
1.69
(0.26)
4.37
(0.04)
0.01
(0.99)

7.83
(0.00)
3.13
(0.01)
1.91
(0.07)
3.16
(0.26)
7.83
(0.00)
0.37
(0.51)

0.07
(0.42)
5.88
(0.00)
5.74
(0.00)
3.86
(0.26)
1.28
(0.24)
0.87
(0.02)

0.00
(0.98)
0.55
(0.08)
0.18
(0.27)
0.01
(0.70)
0.03
(0.42)
0.01
(0.07)

0.02
(0.59)
0.24
(0.15)
0.26
(0.07)
0.07
(0.26)
0.03
(0.35)
0.00
(0.10)

Incl

Notes: St 1 = construction starts in year t 1, Et 1 = primary energy consumption in year t 1, LUCt = Lucens accident impact, TMIt = Three Mile Island Accident Impact,
CHERt = Chernobyl accident impact, ri,t = real interest rate, i,t = ination rate, i = country xed effects. Robust p-values are in parenthesis.

128

Z. Csereklyei / Ecological Economics 99 (2014) 121129

Table 6
Impact of real GDP and growth on reactor construction.

GDP model I
GDP model II
Growth model I
Growth model II

Energy growth I
Energy growth II

logSi.t 1

logEi.t 1

LUCi.t

TMIi.t

CHERi.t

logYi.t 1

5.92
(0.49)
0.38
(0.95)
1.29
(0.34)
0.82
(0.00)

0.36
(0.00)
0.38
(0.00)
0.38
(0.00)
0.39
(0.00)

1.07
(0.09)

0.14
(0.72)
0.12
(0.77)
0.31
(0.52)
0.13
(0.79)

0.09
(0.80)
0.02
(0.96)
0.02
(0.96)
0.05
(0.88)

0.89
(0.01)
0.80
(0.02)
0.87
(0.01)
0.69
(0.01)

0.74
(0.38)
0.10
(0.81)

logSi.t 1

logEi.t 1

LUCi.t

TMIi.t

CHERi.t

logYi.t 1

0.37
(0.00)
0.39
(0.00)

1.63
(0.15)
1.13
(0.33)

0.18
(0.65)
0.11
(0.82)

0.83
(0.00)
0.81
(0.00)

0.48
(0.12)

0.06
(0.87)
0.06
(0.87)

logYi.t 1

Incl
2.24
(0.21)
2.22
(0.21)

0.69
(0.01)
0.69
(0.01)

i
Incl

logYi.t 1

Incl
Incl

i
Incl

1.61
(0.35)

Incl

Notes: St 1 = construction starts in year t 1, Et 1 = primary energy consumption in year t 1, LUCi,t = Lucens accident impact, TMIi,t = Three Mile Island Accident Impact, CHERi,t =
Chernobyl accident impact, logYi,t 1 = real income in year t 1, logYi,t 1 = economic growth in year t 1, i = country xed effects. Robust p-values are in parenthesis.

power generation in the absence of own oil reserves. I tested this claim
by including the growth rate of international crude oil prices in Accident
Model II. While the coefcient is slightly positive on the oil price variable, it is insignicant. From the three countries with the largest nuclear
power generation, oil prices are only signicant in France at a 10% level.
The reason for this statistical insignicance may have to do with the different oil price movements coinciding with both TMI and Chernobyl.
We can state however, that at least in France oil prices tended to inuence reactor constructions.
The question what impact the possibility of nuclear proliferation has
on reactor construction is beyond the scope of the study. The problem of
possible nuclear proliferation is identied among others by Miller and
Sagan (2009), Socolow and Glaser (2009), and Fuhrmann (2012) as a
potential challenge and constraint for the future of nuclear energy, especially if non-democratic countries were to acquire nuclear
technology.
5. Conclusion and Implications for the Future of Nuclear Energy
This paper investigates the effects of nuclear accidents on energy
policy with the help of a panel dataset of 31 countries from 1965 to
2009, using annual data on the capacity of reactors under construction,
primary energy consumption, as well as the incidence of three nuclear
accidents scaled INES ve or higher on the International Atomic Energy
Agency's scale. After determining the extent of the accident impact in
the different countries, I nd that neither the Three Mile Island nor
the Lucens accidents had a worldwide negative effect on construction
starts, while Chernobyl did. Three Mile Island had a lasting impact in
the United States however. I show that the effect of Chernobyl wore
off in certain geographical clusters, after ten to thirty years. Thus, an accident is likely to have a negative and long lasting impact in the country
where it happened, and possibly in countries that were affected by the
direct consequences, such as nuclear fallout. I also nd that nuclear capacity enlargement shows a signicant lock-in effect, but it was also signicantly driven by primary energy consumption in the past ve
decades. The results are robust to the inclusion of real interest rates

with ination, gross domestic product and economic growth into the
model, none of which are found signicant.
Recently, The Economist called nuclear energy a dream that failed
(Morton, 2012). To the question of why nuclear energy has not continued worldwide to gain market share in electricity generation after the
1970s, the answers are complex, and often interrelated. The golden
age of nuclear power was certainly not only due to the inevitably increasing energy demand, but also to political, military, security, and
strategic considerations. As the political and sociological landscape
changed, so did the environment surrounding the nuclear industry.
Where and how far the negative policy impact of Fukushima might
spread, will likely depend on the efciency and success of the
Japanese management of the damaged nuclear plant, whether or
not there is existing nuclear infrastructure and know-how in a country, on the availability of energy supplies to satisfy future energy demand, and on the existing political environments. We have seen that
the length of impact of an accident can, but need not wear off anywhere from ten to thirty years.
How much is future nuclear power plant construction expected to be
impacted by the Fukushima accident? Hayashi and Hughes (2013b)
argue that due to the cost of electricity from fossil fuels and the need
to reduce greenhouse gas emissions, a radical shift in Japanese nuclear
policy seems unrealistic despite the severity of the accident. Based on
the results of this study, it seems likely, however, that new construction
in Japan will be adversely affected as a result of Fukushima, while
existing plants may continue to operate, due to the otherwise very difcult to manage gap between electricity production and demand.
Thomas (2012) outlined four possible scenarios in the aftermath of
Fukushima. He concludes that the accident may either change everything, the impact may vary according to location, it is too early to determine, or change very little, because at least in the West, the nuclear
renaissance was failing long before Fukushima. I argue that where the
major driving forces encouraging nuclear energy expansion, such as
drastically increasing energy consumption and energy security concerns are coupled with government programs and plant ownership, nuclear energy expansion is likely to continue. New builds might, however

Table 7
Impact of energy security on reactor construction.

Energy security I
Energy security II

logSi,t 1

logEi,t 1

LUCi,t

TMIi,t

CHERi,t

ENSECi,t 1

0.53
(0.77)
2.05
(0.23)

0.34
(0.00)
0.29
(0.00)

0.01
(0.98)
0.95
(0.02)

0.13
(0.79)
0.29
(0.55)

0.09
(0.81)
0.23
(0.53)

0.79
(0.01)
0.51
(0.08)

0.90
(0.06)
0.53
(0.23)

@t

i
incl

0.06
(0.00)

incl

Notes: St 1 = construction starts in year t 1, Et 1 = primary energy consumption in year t 1, LUCi,t = Lucens accident impact, TMIi,t = Three Mile Island Accident Impact, CHERi,t =
Chernobyl accident impact, ENSECi,t 1 = energy security measure in year t 1, @t = time trend, i = country xed effects. Robust p-values are in parenthesis.

Z. Csereklyei / Ecological Economics 99 (2014) 121129

be impacted much more negatively in countries without these driving

factors, or where nuclear energy is facing free market conditions.
Acknowledgements
The author is grateful to Jess Crespo Cuaresma, to Helmuth Bck,
for advice and suggestions, to Stefan Humer, Jeff Kessler, Peter Gunther
and two anonymous reviewers for very useful comments. I thankfully
acknowledge reactor capacity data from Hans-Holger Rogner. All remaining errors are mine.
References
Adamantiades, A., Kessides, I., 2011. Nuclear power for sustainable development: current
status and future prospects. Energy Policy 33, 51495166.
Ahearne, J.F., 2011. Prospects for nuclear energy. Energy Econ. 33, 572580.
Apergis, N., Payne, J.E., Menyah, K., Wolde-Rufael, Y., 2010. On the causal dynamics between emissions, nuclear energy, renewable energy, and economic growth. Ecol.
Econ. 69, 22552260.
Baltagi, B.H., 2005. Econometric Analysis of Panel Data. John Wiley and Sons.
British Petrol, 2010. Statistical Review of World Energy.
Burke, P.J., 2010. Income, resources and the electricity mix. Energy Econ. 32, 616626.
Cohen, B.L., 1990. The Nuclear Energy Option. Plenum Press978-0306435676.
Csereklyei, Z., Humer, S., 2012. Modelling primary energy consumption under model
uncertainty. WU Working Papers in Economics, Nr. 147.
Doornik, J., Arrelano, M., Bond, S., 2001. Panel Data Estimation Using DPD for Ox.
Ebinger, C.K., 2011. Three Strikes and Out? Nuclear Energy in the United States in the
Wake of Three Mile Island, Chernobyl and Fukushima. Brookings Institution.
Fuhrmann, M., 2012. Splitting atoms: why do countries build nuclear power plants? Int.
Interact. 38 (1), 128.
Glaser, A., 2011. After Fukushima: preparing for a more uncertain future of nuclear power.
Electr. J. 24 (6), 2735.
Goodfellow, M.J., Williams, H.R., Azapagic, A., 2011. Nuclear renaissance, public perception and design criteria: an exploratory review. Energy Policy 39, 61996210.
Gourley, B., Stulberg, A.N., 2013. Correlates of Nuclear Energy. In: Stulberg, A., Fuhrmann, M.
(Eds.), The Nuclear Renaissance and International Security. Stanford University Press.
Guidolin, M., Guseo, R., 2012. A nuclear power renaissance? Technol. Forecast. Soc. Chang.
79, 17461760.
Harding, J., December 2007. Economics of Nuclear Power and Proliferation Risks in a
Carbon-Constrained World. The Electricity 20 (10), 6576.
Hayashi, M., Hughes, L., 2013a. The Fukushima nuclear accident and its effect on global
energy security. Energy Policy 59, 102111.

129

Hayashi, M., Hughes, L., 2013b. The policy responses to the Fukushima nuclear accident
and their effect on Japanese energy security. Energy Policy 59, 86101.
Heston, A., Summers, R., Aten, B., 2011. Penn World Table Version 7.0.
International Atomic Energy Agency, 1990. International Nuclear and Radiological Event
Scale.
International Atomic Energy Agency, 1996. Signicant Incidents in Nuclear Fuel Cycle Facilities, IAEA-TECDOC-867. 1011-4289.
International Atomic Energy Agency, 2011. Nuclear Power Reactors in the World, Reference Data Series 2, 2011 edition. 978-92-0-117810-7.
Joskow, P.L., Parsons, J.E., 2009. The economic future of nuclear power. Daedalus 138 (4),
4559.
Joskow, P.L., Parsons, J.E., 2012. The Future of Nuclear Power after Fukushima. MIT CEEPR
Working Paper Series, (WP 2012001).
Judson, R.A., Owen, A.L., 1999. Estimating dynamic panel data models: a guide for
macroeconomists. Econ. Lett. 65, 915.
Kopytko, N., Perkins, J., 2011. Climate change, nuclear power, and the adaptationmitigation dilemma. Energy Policy 39, 318333.
Lester, R.K., Rosner, R., 2009. The growth of nuclear power: drivers and constraints. Daedalus 138 (4), 1930.
Mez, L., 2012. Nuclear energyany solution for sustainability and climate protection? Energy Policy 48, 5663.
Miller, S.E., Sagan, S.D., 2009. Nuclear power without nuclear proliferation. Daedalus 138
(4), 718.
Morton, O., 2012. Special report on nuclear energy. The Economist (March 10, 2012),
316.
Nelson, P., 2010. An Empirical Assessment of Elements of the Future of Civil Nuclear Energy. Nuclear Security Science and Policy Institute, NSSPI-10-001.
Nohrstedt, D., 2005. External shocks and policy change: Three Mile Island and Swedish
nuclear energy policy. J. Eur. Public Policy 12 (6), 10411059.
Nohrstedt, D., 2008. The politics of crisis policymaking: Chernobyl and Swedish nuclear
energy policy. Policy Stud. J. 36 (2), 257278.
Pesaran, M.H., 2007. A simple panel unit root test in the presence of cross section dependence. J. Appl. Econ. 22, 265312.
Smythe, D., 2011. An objective nuclear accident magnitude scale for quantication of severe and catastrophic events. Phys. Today (12 December, Online Daily Edition).
Socolow, R.H., Glaser, A., 2009. Balancing risks: nuclear energy & climate change. Daedalus
138 (4), 3144.
The World Bank, 2012. World Bank Database: World Development Indicators.
Thomas, S., 2012. What will the Fukushima disaster change? Energy Policy 45,
1217.
Van Roey, E., 2009. Three Mile Island, an SCK-CEN Compilation.
Villa, M. 2008. Reaktorphysik, AIAU 98402. Technische UniversittWien, Atominstitut.
Wittneben, B.B.F., 2012. The impact of the Fukushima nuclear accident on European energy policy. Environ. Sci. Pol. 15, 13.
Zhou, Y., Rengo, C., Chen, P., Hinze, J., 2011. Is China ready for its nuclear expansion?
Energy Policy 39, 771781.