In order to limit human impact on the global climate, it is necessary to decarbonize the energy supply of nations by adopting clean energy sources to replace fossil fuels. However, as I show here on the basis of an analysis of cross-national time-series data for the past five decades, reducing the carbon intensity of overall energy use is associated with higher energy use, and reducing the carbon intensity of electricity production is associated with higher electricity production. These findings suggest that adding noncarbon and low-carbon energy generation capacity may be connected with processes that spur energy demand. This has important environmental implications, since alternative energy sources have serious environmental impacts of their own. The policy challenge is to ensure that clean energy sources replace rather than add to carbon-based energy.

INTRODUCTION

The Fifth Assessment Report of the United Nations' Intergovernmental Panel on Climate Change (IPCC 2014) makes clear that average global temperatures are rising because of human activities and that the changes to climate stemming from this process could have disastrous effects on ecosystems and societies around the world by the end of this century. The emission of greenhouse gases (GHGs) is the most important factor contributing to climate change, and carbon dioxide from the combustion of fossil fuels is the greatest contributor among the GHGs. One of the key mitigation strategies suggested by the IPCC (2014) is to decarbonize the energy supply by producing more energy from noncarbon sources (e.g., hydro, nuclear, wind, solar) and by using lower-carbon fuels (e.g., natural gas) as part of the transition to clean energy, with the expectation that over time noncarbon energy sources will replace carbon-based sources.

Obviously, since all societies require energy, if societies are to stop using fossil fuels, other sources of energy must be developed. However, in the push to expand production of non- and low-carbon energy sources so as to reduce the carbon intensity of the energy supply, one important question is rarely asked: Is there a connection between the carbon intensity of the energy supply and the total amount of energy societies use? The answer to this question has important implications for how much noncarbon and low-carbon energy needs to be supplied, because if adding noncarbon and low-carbon energy sources is connected with rising energy demand, replacing high-carbon energy may require additional growth in energy production. Understanding the connections between the mix of sources of energy and the total amount of energy used is particularly important for developing nations, since many of them are building major parts of their energy systems, while struggling to address environmental problems and the energy demands of their populations.

Although institutional and cultural factors are important in theories of modernization (Hirschman 1958; Levy 1966; Portes 2015), technology and infrastructure, particularly electricity generation and distribution, for a long time have been recognized as central to the material aspects of economic development (Dietz 2015; Mazur 2013). Although research has established that the expansion of energy consumption is not clearly connected with rising quality of life in more developed countries, energy use in societies is generally closely tied with economic growth and is linked with improving standards of living in the poorest nations (Dietz 2015; Mazur 2013; Mazur and Rosa 1974). A key facet of moving toward sustainable development is to devise ways to improve the well-being of people without relying on high levels of energy consumption or carbon emissions (Dietz 2015). Therefore, progress toward understanding how to allow for rising standards of living in poor nations without undermining sustainability depends to some degree on further developing our understanding of the forces influencing energy use and the connections between energy use and carbon emissions.

A substantial and growing body of sociological research assesses how various structural characteristics of societies—including economic, demographic, and technological factors—influence the scale of energy use and carbon dioxide emissions (Jorgenson and Clark 2012; Rosa et al. 2015; York 2012). Much of this research is concerned with the degree to which the development of “green” technologies, such as renewable energy sources, helps to curtail a variety of environmental problems. Sociologists and other social scientists suggest that technological solutions may not be as effective at leading to resource conservation and pollution abatement as is often anticipated by policy makers and engineers because of a variety of potential feedbacks and interactions that occur in socioeconomic systems with technological change (York 2006). For example, Sellen and Harper (2002) show that a variety of social and economic processes and unforeseen interactions often prevented the rise of electronic storage mediums from reducing paper use and that the growth of computer use is often associated with more paper use in businesses and societies as a whole. As another example, McGee (2015) found that the growth of the production of certified organic food in the United States has been positively associated with national greenhouse gas emissions because of how agribusiness exploits the profitability of organics and works to reduce costs and expand markets. More directly related to the topic I address here, York (2012) has shown that in recent decades the expansion of non–fossil fuel energy sources in nations around the world has been only modestly effective at suppressing fossil fuel use.

There is no single explanation for the sometimes paradoxical effects of “green” technologies on resource use and pollution, but social scientists have identified a variety of processes that are likely to be at work. There are several reasons to expect that using more non- and low-carbon sources may drive up energy use. Zehner (2012) has argued that increasing clean energy production can create a “boomerang effect,” where the downward pressure on energy prices that can come with an increase in energy supply may spur energy consumption. Furthermore, since the public typically sees noncarbon energy sources as contributing to environmental protection, the expansion of clean energy production may face little political resistance, so expanding clean energy production may make growth in energy consumption more feasible. Since fossil fuel–based technologies remain in place even when other energy sources are developed, clean energy to some degree may add to, rather than entirely replace, fossil fuels. For these reasons, adding clean energy sources to the power supply, without focused efforts to directly suppress fossil fuel use, may serve to entrench high-energy lifestyles and undermine efforts at energy conservation.

Zehner's (2012) argument fits well with a more general political-economic understanding of how technology is deployed in modern economies. Schnaiberg (1980) and Foster, Clark, and York (2010) present Marxian analyses of how modern market economies are focused on endlessly expanding production and consumption to generate profits for corporations. As part of this process, corporations invest heavily in technological development because improved technologies can both replace workers and reduce the quantity of resources used per unit of production, outcomes that reduce costs and thereby increase profits. Although on the surface it may seem that reducing the resources used per unit of production will help conserve resources, it is not done for that purpose, and it often does not have that effect, since corporations typically invest their increased profits in further expanding production and consumption, thereby increasing overall resource use. In this process, technologies that have the potential to reduce resource consumption can have the opposite effect.

This political-economic view is grounded in the recognition that there is no necessary intrinsic desire in ordinary people to consume more and more and that this demand for increased consumption is instead created by corporations through marketing and the use of political power to structure policies to be conducive to higher levels of consumption (Dawson 2003). This perspective highlights that in many ways production sometimes spurs consumption or, put differently, supply sometimes can generate demand. With regard to energy, this suggests that adding noncarbon energy sources to the energy supply increases the capacity for production and may stimulate consumption.

The work of Zehner (2012), Schnaiberg (1980), and Foster et al. (2010), among others, suggests the hypothesis that decarbonizing the energy supply (i.e., reducing carbon emissions per unit of energy) may be associated with increased energy demand. I test this hypothesis here in two ways. First, I assess whether the overall energy use of nations is associated with changes in carbon dioxide emissions per unit of energy use. Examining overall energy use may not be ideal, since different forms of energy may not readily substitute for one another directly: for example, it is challenging to replace liquid fuels used in transportation with electricity. To address this concern, in a second analysis I focus on the electrical sector and assess the association between electricity production and the amount of carbon emitted per unit of electricity production.

METHODS

I estimate panel models using cross-national temporally differenced (i.e., change over time) data to assess the effects of the carbon intensity of energy use on energy use per capita, and the effects of the carbon intensity of electricity production on electricity production per capita, for the period 1960–2010, with five-year intervals of change (i.e., a total of 10 periods: 1960 to 1965, 1965 to 1970, etc.). I use five-year intervals to allow for appreciable change in the variables under consideration and for the likelihood that the dynamic connections between the energy mix and energy use or electricity production play out on longer than annual cycles. I use national-level data from the World Bank's (2014) World Development Indicators, including all nations and periods for which sufficient data are available. For some nations there are not sufficient data for all time periods, particularly earlier periods. The sample for the energy use model includes 122 nations and a total of 588 observations. The sample for the electricity production model includes 121 nations and a total of 591 observations.

I use the change per five years in the natural logarithm of all variables, producing an elasticity model, where each coefficient indicates the percentage change in the dependent variables for a 1 percent change in the independent variable. The dependent variable in the first model is energy use per capita in kilograms of oil equivalent, which represents “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 2014). The key independent variable in the first model is the carbon intensity of energy use, which is measured as all energy-based carbon dioxide emissions in metric tons per unit of energy use.1 The dependent variable in the second model is electricity production per capita in kilowatt hours (kWh). The key independent variable in the second model is the carbon intensity of electricity production, which is measured as metric tons of carbon dioxide emissions from the electrical sector per kWh of electricity production.2 

I include a suite of control variables that have been established in the sociological literature to be connected with energy use (Jorgenson and Clark 2012; Rosa et al. 2015; York 2012). Both models include GDP per capita in constant 2005 US$ and, to allow for a non-log-linear relationship, its quadratic if it had a significant coefficient,3 the percentage of the population living in urban areas, the percentage of GDP from the manufacturing sector, and the age dependency ratio (people under 15 years of age and over 64 years of age to those 15 to 64 years of age). I also include period dummy variables to control for temporal trends common across nations, which include the international price of energy resources. Since I am using differenced data, I employ a generalized least squares model instead of a fixed-effects model.4 Like fixed-effects models, models using temporally differenced data control for any potential omitted control variables that differ across nations but that do not change over time, such as latitude and other major geographic features (e.g., whether a nation is landlocked, mountainous, etc.). The models use the Prais-Winsten correction for first-order autocorrelation.5 

RESULTS AND DISCUSSION

The results of the analyses are presented in table 1.6 In the first model, GDP per capita and the percentage of the economy from manufacturing are both significantly and positively associated with energy use per capita. The percentage of the population living in urban areas and the age dependency ratio do not have significant associations with energy use. Importantly for the issue addressed here, the carbon intensity of the energy supply has a significant negative association with energy use per capita, meaning that carbon intensity and energy use change in opposite directions to one another. This means that a decline in carbon intensity is associated with a rise in energy use, confirming the hypothesis tested here.

Table 1
Panel Regression Elasticity Model of Change (Five-year Intervals) in Energy Use per Capita (Tonnes of Oil Equivalent) and Electricity Production (kWh), 1960–2010
Independent VariablesEnergy Use p.c. Coefficient (S.E.)Electricity Prod. p.c. Coefficient (S.E.)
Carbon intensity     −.061** (.023)     −.117*** (.022) 
GDP per capita      .447*** (.040)     1.149*** (.300) 
(GDP per capita)2      −.044* (.019) 
Urban population (%)      .194 (.125)      .764*** (.203) 
Manufacturing (% GDP)      .076** (.023)      .143*** (.041) 
Age dependency ratio      .081 (.089)      .054 (.150) 
R2 (overall)      .269      .278 
N (total/nations)       588/122       591/121 
Independent VariablesEnergy Use p.c. Coefficient (S.E.)Electricity Prod. p.c. Coefficient (S.E.)
Carbon intensity     −.061** (.023)     −.117*** (.022) 
GDP per capita      .447*** (.040)     1.149*** (.300) 
(GDP per capita)2      −.044* (.019) 
Urban population (%)      .194 (.125)      .764*** (.203) 
Manufacturing (% GDP)      .076** (.023)      .143*** (.041) 
Age dependency ratio      .081 (.089)      .054 (.150) 
R2 (overall)      .269      .278 
N (total/nations)       588/122       591/121 
***

p < .001

**

p <. 01

*

p < .05 (two-tailed tests).

Note: All variables including the dependent variable were in natural logarithmic form before differencing. Models use the Prais-Winsten correction for first-order autocorrelation. The models include period dummy variables that are not shown.

In the second model, all factors other than the age dependency ratio have significant relationships with electricity production per capita. The negative coefficient for the quadratic of GDP per capita suggests a slight attenuation in the rate that electricity production increases as GDP per capita grows. Importantly, the carbon intensity of electricity production is negatively and significantly correlated with electricity production per capita. This indicates that as carbon intensity declines electricity production per capita grows, once again confirming the research hypothesis.

The results of both models suggest that over the past 50 years reducing the carbon intensity of the energy supply has been connected with higher energy use and electricity production. The rise in energy use and electricity production associated with declining carbon intensity is important, since, although curtailing carbon emissions is extremely important in light of their effects on the global climate, alternative energy sources have substantial environmental impacts of their own. For example, hydroelectric dams destroy river ecosystems, nuclear power produces radioactive waste and presents catastrophic risks in cases of meltdown, and wind turbines and solar panels require many natural resources, some of them toxic, to produce (Sovacool and Dworkin 2014; Zehner 2012). Also, “noncarbon” energy sources do in fact contribute indirectly to carbon emissions, in part because of the cement used in construction (particularly for nuclear plants and hydroelectric dams) and in part because of the energy (some of it carbon based) used to make the hardware and infrastructure (Sovacool and Dworkin 2014).

Therefore, even if decarbonizing the energy supply did not increase energy use and electricity production, it would still involve substantial trade-offs. In light of the fact that decarbonizing is associated with greater demand for energy, these trade-offs become less favorable. This is not to suggest that these trade-offs are not worthwhile, since global climate change is a very serious threat to societies and ecosystems. However, the apparent connection between decarbonization and energy demand does point to the importance of working not only to change the energy mix but also to constrain growth in energy consumption, especially in affluent nations.

It is important to consider the possibility that causality flows in the direction opposite of that hypothesized here, at least in some nations at some times. That is to say, it is plausible to interpret these findings as suggesting that growth in demand to some degree has driven expansion of low- and noncarbon energy sources. If this is the case, the findings nonetheless have some similar implications to those I explain above, since they mean that although expanding low- and non-carbon energy sources does not cause demand, clean energy may be part of what allows energy consumption to grow as demand rises. Since the models take into account GDP per capita and other structural factors, high energy use or electricity production indicates excess use or production, in the sense that it is use or production above what would be expected on the basis of economic and structural characteristics. Therefore, if high demand leads to more clean energy development, this suggests that clean energy is more likely to come into play to meet excess demand, implying that it is to some degree added on top of fossil fuels as a luxury, whereas, when demand is low, traditional fossil fuels are more likely to be relied on. Without ironclad theoretical logic, the direction of causality cannot be sorted out with certainty using nonexperimental data, so it is difficult to resolve this issue. Nonetheless, regardless of the direction of causality (it seems reasonable to expect that causality goes in both directions at least some of the time in some places), the findings from my analysis clearly indicate that energy use and electricity production are generally high when carbon intensity is low.

CONCLUSION

The analyses presented here show that across nations around the world over the past five decades there is a general pattern where a decline in the carbon intensity of the energy supply is associated with a rise in energy demand. This pattern exists both for overall energy use and for electricity production. This means that efforts to decarbonize the energy supply may be connected with excess energy demand, which is not inconsequential, since all forms of energy contribute to environmental problems. The association found here, of course, does not represent an inevitable and unalterable relationship; rather, it is merely a general pattern that has emerged from the economic, political, social, and cultural contexts that have been prevalent over the past 50 or so years. However, these findings are consistent with the assertion that green technologies may have unintended consequences that to some degree counterbalance their environmental benefits (Zehner 2012).

These findings have clear implications for policy. If the relationship that has prevailed in the past continues to hold in the future, policy makers need to recognize that it may require supplying more clean energy than the amount of carbon-based energy that needs to be replaced. This represents a major challenge, since the potential to meet the current and projected global demand for energy from clean sources is limited (Trainer 2010). This suggests that in addition to developing clean energy sources, it will be necessary to limit growth in total energy use. Since energy use is not clearly connected with well-being in affluent nations, restraining total energy use at the global level without harming quality of life could be achieved by reductions in energy consumption in the most affluent nations, while allowing for some growth in energy use in the poorest nations (Dietz 2015). Well-designed policies could potentially alter the pattern identified here so that the addition of clean energy sources would not be connected with rising demand. Thus policy makers should focus not only on implementing policies to encourage the development of clean energy but also on designing policies in such a way that clean energy sources replace fossil fuels rather than simply being added to them.

REFERENCES

REFERENCES
Dawson, Michael.
2003
.
The Consumer Trap: Big Business Marketing in American Life.
Champaign
:
University of Illinois Press
.
Dietz, Thomas.
2015
.
“Prolegomenon to a Structural Human Ecology of Human Well-being.”
Sociology of Development
1
(
1
):
123
48
.
Foster, John Bellamy, Brett Clark, and Richard York.
2010
.
The Ecological Rift: Capitalism's War on the Earth
.
New York
:
Monthly Review Press
.
Hirschman, Albert O.
1958
.
The Strategy of Economic Development
.
New Haven, CT
:
Yale University Press
.
Intergovernmental Panel on Climate Change (IPCC)
.
2014
.
Climate Change 2014: Synthesis Report
.
New York
:
United Nations
.
Jorgenson, Andrew K., and Brett Clark.
2012
.
“Are the Economy and the Environment Decoupling? A Comparative International Study, 1960–2005.”
American Journal of Sociology
18
(
1
):
1
44
.
Levy, Marion J.
1966
.
Modernization and the Structure of Societies
.
Princeton, NJ
:
Princeton University Press
.
Mazur, Allan.
2013
.
Energy and Electricity in Industrial Nations: The Sociology and Technology of Energy
.
Oxford
:
Earthscan
.
Mazur, Allan, and Eugene A. Rosa.
1974
.
“Energy and Life-style: Massive Energy Consumption May Not Be Necessary to Maintain Current Living Standards in America.”
Science
186
:
607
10
.
McGee, Julius Alexander.
2015
.
“Does Certified Organic Farming Reduce Greenhouse Gas Emissions from Agricultural Production?”
Agriculture and Human Values
32
(
2
):
255
63
.
Portes, Alejandro.
2015
.
“The Sociology of Development: From Modernization to the ‘Institutional Turn.’”
Sociology of Development
1
(
1
):
20
42
.
Rosa, Eugene A., Thomas K. Rudel, Richard York, Andrew K. Jorgenson, and Thomas Dietz.
2015
. “The Human (Anthropogenic) Driving Forces of Global Climate Change.” Pp.
47
91
in
Climate Change and Society: Sociological Perspectives
, edited by R. E. Dunlap and R. J. Brulle.
New York
:
Oxford University Press
.
Schnaiberg, Allan.
1980
.
The Environment: From Surplus to Scarcity
.
Oxford
:
Oxford University Press
.
Sellen, Abigail J., and Richard H. R. Harper.
2002
.
The Myth of the Paperless Office
.
Cambridge, MA
:
MIT Press
.
Sovacool, Benjamin K., and Michael H. Dworkin.
2014
.
Global Energy Justice: Problems, Principles, and Practices
.
Cambridge
:
Cambridge University Press
.
Trainer, Ted.
2010
.
“Can Renewables Etc. Solve the Greenhouse Problem? The Negative Case.”
Energy Policy
38
:
4107
14
.
World Bank.
2014
.
World Development Indicators
. Retrieved
December
16
(http://data.worldbank.org/data-catalog/world-development-indicators).
York, Richard.
2006
.
“Ecological Paradoxes: William Stanley Jevons and the Paperless Office.”
Human Ecology Review
13
(
2
):
143
47
.
York, Richard.
2012
.
“Do Alternative Energy Sources Displace Fossil Fuels?”
Nature Climate Change
2
(
6
):
441
43
.
Zehner, Ozzie.
2012
.
Green Illusions: The Dirty Secrets of Clean Energy and the Future of Environmentalism
.
Lincoln
:
University of Nebraska Press
.

NOTES

NOTES
1.
The measure “energy-based carbon dioxide emissions” is the sum of emissions from solid, liquid, and gaseous fuels. It differs from the commonly used measure of industrial carbon dioxide emissions in that it excludes emissions from cement manufacturing.
2.
“Metric tons of carbon dioxide emissions from the electrical sector” includes emissions from heat production by utilities.
3.
I initially estimated the energy use model including a quadratic for GDP per capita, but the quadratic term was not statistically significant, so it was dropped from the model.
4.
Using differenced data and a random-effects model is an alternative to, and largely the equivalent of, using a fixed-effects model with undifferenced data. Note that if I use a fixed-effects model with the differenced data, I obtain substantively the same results as I report below with respect to the association between carbon intensity and energy use/electricity production per capita. One advantage of first differencing is that it can fix problems with nonstationary data. However, since the data set I use here is not strongly balanced and contains gaps, the options for tests of whether data are stationary are limited. I used the Augmented Dickey Fuller unit root test on the first-differenced data, where the null hypothesis assumes that all series are nonstationary. For each variable in the model, the test rejected the null hypothesis.
5.
I have also estimated models using robust standard errors to correct for residuals clustering by nation. These models produce substantively the same results as the models reported here.
6.
An examination of the distribution of the residuals from both models shows that they are approximately normally distributed with the exception of a few outliers. I assessed the robustness of the results of both models against the influence of outliers by reestimating the models excluding observations that had residuals that were two or more standard deviations from the mean. The results of these models are substantively the same as the models presented here, indicating that my findings are not driven by outliers in residuals. As another test for influential cases, I reestimated the models excluding observations that had values of the Cook's distance statistic in the top 5 percent and, in other analyses, in the top 1 percent, and the findings were substantively unchanged (i.e., the carbon intensity variable was significant and in the same direction in each of these alternative models). Since Cook's distance statistic measures the overall influence of each observation, this approach assesses if there is undue influence not only from outliers in residuals but also from leverage cases, where there are outliers in the space of independent variables.