Energy Research & Social Science 42 (2018) 184–192

Contents lists available at ScienceDirect

Energy Research & Social Science

journal homepage: www.elsevier.com/locate/erss

Original research article

Inequality and energy: Revisiting the relationship between disparity of T

income distribution and energy use from a complex systems perspective
⁎
Fouad Khana,b, , Paul Heineckerc
a
Central European University, Nador utca 9, Budapest 1056, Hungary
b
WWF Luc Hoffmann Institute, Avenue du Mont-Blanc 27, 1196 Gland, Switzerland
c
University of Applied Science, Weesperzijde 190, Amstelcampus, 1097 DZ Amsterdam, Netherlands

A R T I C LE I N FO A B S T R A C T

Keywords: To consider the impacts of economic inequality on energy consumption efficiency we need indicators that take
Disparity into account the complexity of the economic and energy systems. We also need decision support tools that help
Inequality incorporate such indicators into policy analysis. Drawing inspiration from urban studies and ecology, in this
Energy consumption efficiency paper we develop a scaling indicator for income disparity in national economies that is a measure of system
Complex systems
complexity and does not presuppose any distribution as ideal. The scaling indicator is calculated for 2010 in-
National economics
come distribution data for countries. We show that rising disparity – measured using this indicator calculated; a)
Urban development
for distributions of incomes across consecutive twentieth percentiles of population in national economies and;
b), for distributions of population density in census blocks in metropolitan statistical areas affects energy con-
sumption efficiency in a diametrically different manner in cities and nation states leading to a higher urban
carbon footprint while increasing energy efficiency nationally. The different nature of these two systems explains
the results. We then modify tools for visualizing complexity from urban studies and ecology to explore the
correlation between income disparity and energy efficiency in national economies.

The adverse impacts of inequality in economic systems especially on
human well-being have been extensively documented in literature
going back more than three decades [1]. Recent social, neurological [2]
and even evolutionary [3] evidence points toward the necessity for
addressing the rising inequality in economic systems. Literature that
presents evidence for rising inequality in economic systems especially
in developed nations post World War II [4,5], has reinvigorated the
debate on income inequality [6–9]. However, such studies are often
criticized for use of ‘mean’ or ‘average’ measures that do not capture
non-linearity in systems [10]. One way to study disparity or inequality
in complex systems is to look at scaling within systems. Living organ-
isms and many other dynamical systems have been shown to obey a
power law in scaling of the sizes of their various elements [11–13]. The
impact of scaling on sociopolitical conflict has been explored in detail in
literature [14]. In US the relationship between income distribution and
energy and environmental indicators has also been studied [15]. In
growing economies such as China and India too income inequality has
been shown to affect energy consumption and CO2 emissions [16]
though in the long run in some instances the effect was not found to be
statistically significant [17]. Part of the problem with exploring the
correlation between inequality or disparity in systems and its effects on
energy consumption is the dearth of indicator systems that can take into
account the complexities that underlie such an interaction [18]. In-
dicator systems in economics, environment or energy consumption by
and large continue to be ‘linear’ in that they measure change in parti-
cular characteristics or parameters as not critically affecting all other
elements of the system in a systemic way and being affected by the
system, but respond to a limited, often no more than one or two stimuli.
While complex integrated assessment models do take into account the
density of networked interactions that underlie an economic system to a
certain extent, the measurements or indicators themselves, almost by
their very nature do not consider ‘non-linearity’ in dynamics of the
system [19]. This fact in itself makes it difficult to study the correlation
between ‘inequality’ and other system wide indicators because the
impacts of inequality or ‘disparity’ in the system largely often emerge in
the form of non-linear responses. Between certain values any change in
inequality may not have an effect on system wide parameters but then
minor changes may result in sudden large perturbations [20,21]. While
mechanisms have been proposed in literature that show why response
to income inequality is nonlinear, there is a dearth of indicators and
quantitative assessments on the subject that bring empirical evidence to
bear on the theoretical understanding of economy wide non-linearity

⁎
Corresponding author at: Central European University, Nador utca 9, Budapest 1056, Hungary.
E-mail addresses: fouadmkhan@gmail.com, fouad.khan@us.nature.com (F. Khan).

https://doi.org/10.1016/j.erss.2018.03.026
Received 15 September 2016; Received in revised form 14 January 2018; Accepted 25 March 2018
2214-6296/ © 2018 Elsevier Ltd. All rights reserved.
F. Khan, P. Heinecker
[18]. the social dynamics of energy systems still need further in-
vestigation from multiple perspectives [22] especially with incorpora-
tion of the complexity of the systems into analysis. This work takes
some steps towards addressing these shortcomings.
Recently it has been demonstrated that system resilience is a func-
tion of system heterogeneity among other factors [23]. This hetero-
geneity has been shown to arise from the hierarchical structure of the
system for ecological systems [24] and expresses itself structurally in
the form of very specific scaling in distribution of sizes of elements. The
classes of sizes of given parameter are distributed at various scales such
that the number of elements p, at each scale x are related according to
the equation pxm = constant [25] where m is the exponent of the power
law with values lying between 0.75 and 2.5, also sometimes called the
fractal dimension. In other words, typically these systems do not have
aberrantly sized elements and the number of component elements de-
creases as the scale to which the element belongs increases in size; the
bigger an element is, the lesser its population in the system [12,25–28].
In network terms the system exhibits ‘scale free’ structuring [29]. Such
structuring has been discovered in a number of anthropogenic dyna-
mical systems especially cities [25,30–34]. It was recently shown that
indicators based on the exponent of the power-law can be developed
based on this scaling for cities that help explain energy consumption
behavior in the urban system [35]. The exponent of power-law dis-
tribution can thus serve as a scaling indicator which provides a much
more comprehensive picture of disparity in systems.
In this paper we develop just such an indicator for measuring
comparative disparity in national economies. The indicator is for-
mulated and calculated for income distribution across consecutive 20th
percentiles. We then show that comparative disparity measured using
this scaling indicators affects energy consumption efficiency positively
in national economies in a manner dissimilar to the way it affects en-
ergy consumption efficiency in cities. We then explore the systemic
reasons for this difference in results. Further we modify two tools for
visualizing complexity from urban studies and ecology and apply them
to the study of correlation between energy efficiency and income in-
equality in national economies [35].
1. Methods and materials
Economies are complex adaptive systems and should also exhibit
similar scaling properties as other complex systems. In order to see how
scaling in economic systems affect environmental indicators we looked
at a fractal dimension based scaling indicator of distribution of income.
The environmental or direct sustainability indicator studied was per
capita energy usage. The data was obtained from the World Bank open
data platform [36]. Data from the year 2010 was used as that provided
us with the biggest set of countries for which income distribution data
was available. In this case the primary limitation was income dis-
tribution data which was available for only a small number of countries.
The data is shown in Appendix A.
Fractal dimension based scaling indicator of national income dis-
tribution was calculated by plotting cumulative income share against
the cumulative population percentage. Once plotted on log–log scales
the resulting slope of the line would be the fractal dimension based
scaling indicator of the distribution of income within the country.
To derive our exponent we first start with the formula for the box-
counting dimension, as expressed by Eq. (1) [25]. In box-counting
method a grid of ‘boxes’ is layered on a map of the city, that divides the
spatial spread of the city into different populated areas, each with a
different intensity of urban land use or in other words a different land
use coverage.
log Nx
D=
log ()
1
x (1)
Where,
185
Energy Research & Social Science 42 (2018) 184–192
D = box counting dimension
x = certain percentage (or range of percentages) of area of the box
covered by land use
Nx = Number of boxes falling within range x
Instead of a map we have an extensive data set of the distribution of
income by percentage of population. So instead of overlaying a grid of
‘boxes’ on a map, we will split the population into virtual boxes, each
covering 20% of population. So in our methodology the ‘box’ of the box-
counting method is any given 20% population block.
In box counting method for cities the next step is to count all the
boxes that fall within a certain range of land use coverage; say 3 out of
40 boxes have between thirty to forty percent of their area covered by
urban land use. This is designated by the term Nx in Eq. (1). For our
methodology the congruent count will be the percentage of income
points that fall within a certain population percentile block. Fig. 1 ex-
plains how our method compares to the box counting method.
So if, ai = percentage of income held by population block i and
p = percentage of population in each block, then cumulative income a for
jth, p percentile segment of population will be given by;
j
aj = ∑ ai
i=1
Since for our scaling exponent, the no. of boxes Nx falling within a
certain range x, is simply the percentage of income share falling within
the percentile j; aj, Nx can be written simply as;
j
Nx = aj = ∑ ai
i=1
Similarly x for any given percentile j will be given as the total po-
pulation falling within percentile j. Therefore x = 1/(j × p). The scaling
indicator for our calculation, say Ds can now be expressed as;
log Nx log aj
Ds = =
log ()
1
x
log ()
1
x (2)
Using Richardson-Mandelbrot slope [37] now the value of Ds will be
calculated by plotting aj and 1/x on log–log scale and estimating the
slope of the regression line as shown in Fig. 1. As aj and x are measured
in percentage, Ds is a dimensionless quantity.
The indicator is a measure of disparity within the system. The
greater the value of the indicator the lesser the disparity. Compare the
two countries in Fig. 2 for instance; United States has higher income
share held by a lesser percentage of population compared to Denmark.
Thus the starting point for income share for the US (income of 20%
richest people) is higher than Denmark and slope of the trend line is
lesser. This slope of the trend line thus captures the disparity in the
system. The higher the slope or scaling indicator, the lesser the dis-
parity.
The scaling indicator values were calculated by plotting on log–log
scale the cumulative income distribution and cumulative population
percentage and taking the slope of the regression line. The r-squared
value was greater than 0.95 for all the linear fits for all countries, in-
dicating strong power law distribution. The values are shown in the
Appendix A. It should be noted here however that of the variables
considered in this plotting the clustering of population or binning (into
20th percentiles) was constant and only the wealth distribution varied
leaving only one degree of freedom. The high R values are only in-
dicative of a power law distribution and not necessarily predictive ac-
curacy.
A similar indicator (though with opposing directionality) has been
developed for cities in literature and has been shown to have a negative
correlation with energy efficiency. For cities as the disparity measured
using this indicator goes up, the energy efficiency decreases [35].
F. Khan, P. Heinecker Energy Research & Social Science 42 (2018) 184–192

Fig. 1. The ‘box’ in box counting method is analogous to cumulative 20th percentile of income in proposed data based method and clustering is on the basis of income
distribution instead of % age of area of box covered by land use.

2. Results 3. Discussion

Energy use per capita data for countries was obtained from World Before we proceed with a discussion of the results the relationship
Bank open data platform [36]. In Fig. 3 while the curves point in similar between the indicator of disparity proposed here and the GINI as an
direction, they represent opposing tendencies. For national economies equity index needs to be clarified. The relationship of the proposed
the scaling indicator is inversely correlated with disparity while for indicator with Gini coefficient is shown in Fig. 4. The proposed in-
cities the scaling indicator represents disparity. In cities case thus the dicator is actually the convex expression of the concave Lorenz curve
gasoline consumption and emissions go up with increasing disparity that forms the basis of the Gini coefficient.
while for national economies the reverse is true. The best-fit curves
were obtained using least square regressions and are primarily intended
to show the trend as discussed further. For cities, Los Angeles was an
outlier because of its high population density and area and for national
economies Iceland was the only outlier (in the 5th percentile). Re-
moving these outliers from the estimation of best fit curves did not
decrease the objective function more than 10% in either case. The
curves shown in Fig. 3 include outliers in the calculation. Since there
appears to be some heteroscedasticity we ran the Breusch-Pagan test for
the national economic numbers and found that the p-value was greater
than 0.1 indicating that the heteroscedasticity may not be significant
enough to affect the trend analysis.

Fig. 2. Scaling indicator as a measure of disparity of income distribution in

national economies (2010).

186
F. Khan, P. Heinecker
Fig. 3. Energy usage and carbon emissions increase in cities but decrease nationwide with increasing disparity in population density distribution and income
distribution respectively.
3.1. Comparison with existing inequality indicators
There have been extensive discussions in literature about the im-
pacts of income inequality on well-being, health and resilience of eco-
nomic systems [1–9] while concerns continue to be raised about the
indices we use to measure inequality [10]. However one of the con-
straints towards widespread adoption of indicators like Gini as an ex-
planatory variable is the lack of an identifiable mechanism that can
explain how Gini might influence well-being, health or environment
variables. Such a mechanism cannot be identified because Gini does not
derive out of fundamental system properties and is arbitrarily con-
structed through economy wide measurements. While correlations
maybe established no theoretical mechanisms can be identified that
may establish causation between income inequality measured through
Gini and development and well-being outcomes.
Fractal or scaling based indicators of disparity or heterogeneity
however have been theorized to express and measure resilience and
energy metabolism within the system [23,24]. As such they are re-
flective of fundamental system properties such as the scale at which
development happens in cities. This allows for theoretical mechanisms
to be identified to explain any observed correlations between inequality
indicators and other parameters of concern. In this paper we propose
Fig. 4. Unlike Gini the proposed indicator does not pre-suppose that inequality should be measured in reference to a linear income distribution.
187
Energy Research & Social Science 42 (2018) 184–192
such a mechanism to describe the correlation between our proposed
inequality indicator and energy consumption efficiency in cities and at
the national level.
Because of this conceptual advantage our proposed indicator does
not have to presupposes that 1:1 equality, i.e. 50% of income going to
half the population is the ideal as Gini does. Instead it accepts the
power-law distribution nature of wealth and incomes as a given and
through measuring the exponent of the power-law, provides a way to
assess how strong the middle is in the economic distribution. Since we
are not identifying or proposing any value of this indicator as ideal, the
indicator value can merely be used as a guide to orient policy without
presuming one distribution over other as better. Say for instance we
want country A to make its energy efficiency numbers look more like
country B. Then, based on the correlation presented here as well as the
mechanism identified for that correlation to emerge, country A, in ad-
dition to practical segmented measures imitating country B’s energy
policy, can also define economic policy aims that will make its income
distribution look similar to country B’s in an attempt to meet its energy
efficiency goal. Country A would not be aiming for some arbitrary ideal
income distribution of say 1 Gini but would actually be aiming to
imitate a distribution that has been empirically and theoretically shown
to support its energy efficiency goals.
F. Khan, P. Heinecker
In this paper we looked at income inequality as a special case of
disparity of distribution in anthropogenic complex systems and devel-
oped a fractal dimension based scaling indicator for measuring this
disparity. The indicator was derived and calculated for income dis-
tribution in national economies and population density distribution in
US cities [35]. Since energy consumption has relevance to a number of
significant geo-political concerns from energy security to climate
change adaptation and resilience, to study the relevance of the disparity
indicator, the indicator was correlated to energy consumption per ca-
pita and per unit area in nations and cities.
One of the problems with inequality indicators currently is that such
indicators do not capture the rapidly accumulating wealth at the top of the
distribution. While our proposed indicator does not improve on Gini in
capturing this increasing inequality it certainly does not do any worse. It
should be noted however that documenting this extreme accumulation of
wealth in a smaller population group at this would require development of
new indicators to study and monitor as well as data collection at a gran-
ularity that has not been standardized internationally. Consistent data
from around the world simply is not available to document income share
going to 0.1 or 0.01% of the population. Monitoring and measuring this
accumulation could be instructive towards study of energy consumption in
economies as our results demonstrate. If increasing wealth in the top 20th
percentile changes energy consumption efficiency it would be imperative
to study the impacts of wealth accumulation within the hands of say top
one percent or 0.1%.
3.2. Why disparity may affect energy efficiency
In order to understand the policy implications, the two correlations
that have been observed in this research which need to be analyzed for
identification of the mechanism that may be resulting in the observed
relationships. The relationship for national economies is easier to explain.
With the exception of a few countries as the GDP goes up the income
inequality or disparity seems to go down. This relationship has been ex-
plored in literature in detail. Rising GDP of course translates into grater
energy consumption and hence energy consumption efficiency seems to go
up with rising disparity. At a higher level analysis both figures represent
divergent findings for two anthropogenic complex systems. That primary
finding is that energy usage (or some measure of it) in complex systems
correlates with scaling within the complex system according to a power-
law. While power law relationships indicate that there may be a minima
beyond which the energy consumption might actually increase with flatter
scaling, the empirical data both for cities and economies does not show
scaling indicator values approaching or lower than the supposed minima.
It should also be noted that the power law correlations are weak with R
values of less than 0.5 indicating that the only assertion that can be made
with any certainty about this result concerns that trend of change in en-
ergy consumption with changing scaling indicator value. Conclusions be-
yond this may be indicative but not entirely supported by the data.
To understand why the trend exists, we need to reconsider what
‘scaling’ means. Greater the scaling indicator, greater the change in
distribution of incomes across different population groups. While for
cities, a higher fractal dimension would mean that greater urban area is
occupied by lower density housing compared to a city with lower
fractal dimension for a national economy, a lower scaling indicator
(higher disparity) would mean that a greater percentage of the income
has been concentrated in the richer percentiles of the population.
It may be that lower scaling indicator corresponds to higher energy
consumption because the richest 20% end up consuming a dis-
proportionally large percentage of energy per capita. Though based on
the results from cities we cannot generalize this mechanism to both
systems studied (cities and national economies). However there is some
direct [38] and other indirect evidence [15,38–45] in literature that
may be the case; that the rise in energy consumption with lower dis-
parity may be in part at least due to the inordinate increase in energy
consumption by the higher income segments of the economy.
188
Energy Research & Social Science 42 (2018) 184–192
From a system’s perspective we can try to imagine what happens
when the scaling of distribution of sizes within a system becomes more
disparate (e.g., high income inequality in economies). As disparity in-
creases, the segment of the system that is meant to act as a regulator,
becomes less and less of an accurate model of the system, unless the
regulator size (as percentage of system) and energy cost of regulation is
increased proportionally [46,47]. The intuitive conjecture that can be
drawn here is that a more disparate system needs higher energy to be
effectively regulated.
In national economies ‘system regulation’ would mean provision of
energy to the citizens. Greater disparity in incomes would have two
effects. One, it would disproportionately increase the share of energy
consumed by the higher percentiles. Secondly, it would make the task
of regulation, i.e. efficiently providing energy in different forms to all
citizens that much more complicated. At a national level, the higher
energy cost of regulating a more disparate system may also show up in
the cost of political governance. It becomes more difficult to have a
regulatory body that is also a good ‘model’ of the system and has for
instance representation from various economic strata, thereby in-
creasing the energy cost of regulation. However outweighing some of
these factors is the fact that with rising incomes, overall income dis-
parity has been shown to go down.
It should also be considered that if reducing land use inequality at the
urban level increases energy efficiency, this transforms inequality reduc-
tion into a tool for efficiency improvements just as bound to the laws of
rebound effect or Jevon’s paradox as any other efficiency enhancing tool.
The practical implementation of this would see an increasing number of
more equal cities prop up which are more efficient individually but lead to
a higher per capita energy consumption at the national level. While
studying and documenting such a phenomenon would require further
empirical research this is an area that should be explored in the debate for
changing urban form to make cities more energy efficient.
3.3. Policy implications
The wide scale adoption of inequality and disparity indicators into policy
development and analysis continues to be an uphill task. To that end we are
proposing two tools here that will simplify consideration of any inequality or
disparity indicator into policy analysis. We also describe how the theoretical
implications of the work can help explain the connections between disparity
and well-being and growth indicators (which policy makers find more in-
teresting) in a more rigorous basis building on first principles.
The findings presented here have significant policy implications in
terms of designing systems that are efficient and resilient. The debate on
equity in complex human systems and the role or impact of energy on
equity as well as the vice versa have been a subject of debate and dis-
cussion for a while. Gini coefficients have been proposed in the past to
look at distribution of energy resources [48]. Certainly the links between
energy poverty and health and mortality have been well explored espe-
cially in the context of winters in Europe [49]. Inequality in energy con-
sumption across different countries has been explored in terms of its im-
pacts on GDP as well as its causes in differing GDP across the countries
[50]. By providing a numeric resilience and sustainability based justifi-
cation for strategic planning, the fractal dimension based scaling indicator
can evolve into an alternate paradigm for economic, infrastructural and
urban development investment that complements the otherwise purely
economic considerations that usually underpin such planning processes.
The use of fractal dimension as an indicator of system structures can
provide an objective higher-level matrix to assess the structural patterns
that may be indicative of the overall system sustainability. Such higher
level matrices can, through research, be linked to a number of direct
measures (such as per capita energy consumption) to establish how the
matrix; for instance fractal dimension, is an indicator of what’s hap-
pening at the level of individual variables within the system. Once such
relationships are established the higher-level matrices can be used to
provide guidance on policy decision-making. In this way the fractal
F. Khan, P. Heinecker
Fig. 5. ‘Fractal Spectra’ modified for national economies; a plotting of squares representing different income shares by consecutive twentieth percentile population
segments on a y-axis representing the income share percentage.
dimension of the distribution of sizes of various parameters within
complex systems can be a significant higher-level matrix guiding policy
for sustainable development. Scaling based indicators can contribute to
the science of indicator development for complex systems. Typically
index development usually involves weighted summation of different
indicators or parameters which are selected based on specific criteria
and represent key dimensions of a complex system or phenomenon. If it
can be shown that complexity based scaling indicators correlate with a
wide range of parameters, then the scaling indicator can be used as an
aggregate index (without the need for aggregation). The other im-
plication is the consideration for scaling of certain parameters in ana-
lysis besides just averages. Many planning processes are concerned
primarily with indicator averages which do not provide any assessment
of how the indicator changes in distribution over the population. This is
of course not a comment on systems where design process considers
some measure of peak parameter value for analysis (e.g. water treat-
ment plant designs). Such scaling indicators can be calculated for any
parameters which are averaged over a large population or sample and
should be used in conjunction with average values to temper or put in
context the findings from average values.
There is some potential for the use of scaling based indicators or
indices in ex-post anthropological and historical studies to identify
patterns of growth and development that lead to societal collapse.
However application in such scenarios continues limited by the accu-
racy and extent of data availability. Any such results would need to be
seen in the context of the accuracy and extent of data available.
In social dynamics the work is an attempt at pointing out and identi-
fying tipping points where small changes can affect large outcomes. These
outcomes can be both positive and negative and by considering non-line-
arity the planning process could become cognizant of the potential for
negative outcomes and be on-guard before such tipping points, while po-
sitioning the system to benefit from any positive tipping points. The un-
derlying theme of this work is to identify ways in which quality of life can
be improved without increasing the entropic footprint of our civilization.
Although cognizant of the complex nature of urban systems, fractal
dimension based indicator systems share most of the limitations of any
other indicator systems. An indicator by nature is a compromise be-
tween science-based understanding of complex phenomena and the
need for distilling complex information into bite-size summary form
that is necessary in most decision making processes. The important
thing to realize is that the structural development that is necessary to
conceive indicators is in essence a tiered compilation of assumptions,
choices and informed decisions. Starting from the fundamental para-
digm that dictates the creation of the indicator right down to an
eventual number, the structures that inform decision making are held
189
Energy Research & Social Science 42 (2018) 184–192
together by the scaffolding of conditionality and exceptions [51]. These
must never be forgotten while putting an indicator to use in the service
of decision making. While the data required for implementation of this
indicator is usually available in most regions of the world, in case ad-
ditional temporal resolution is needed than the usual census time step
of a decade, this can become an expensive indicator to calculate. One of
the necessary requirements for an effective indicator is that it should be
easy to understand so as to be able to create ‘buy-in’ amongst a wide
stakeholder base. While the fundamentals of this indicator system are
not necessarily complicated, any wide scale adoption of the indicator
would have to be coupled with informational media campaigns to
communicate some of the ideas behind the indicator system within the
public sphere. This is largely due to the fact that many of the concepts
behind this indicator system would be new for the general public even
in some of the cities with highest education rates in the world. Adoption
of new indicators is often as much a politics and policy issue as it is an
issue of scientific debate and if the primary stakeholders don’t have a
general understanding and acceptance of the ideas behind important
decision influencing indicators, that can render their adoption un-
certain and vulnerable to political criticisms and dismantling.
Even while evolving our indicator systems to reflect some of the
complexity of systems, one must not forget that the primary lesson to be
derived from complexity science is not one of additional trust in numbers,
but less trust, no matter how rigorous the derivations of those numbers
are. As we begin to understand the nature of complex adaptive systems,
the calls for trying to avoid over-simplification in the study of such sys-
tems, continue to pour in. In industrial risk assessments consideration of
holistic, systemic risks has gained significance since Three Mile Island
[52]. The post-normal science literature has long been calling for new
analysis techniques that take into account risks emerging from system
complexity [53], and post 2008, even in the financial world, the impact of
Black Swans is studied and considered with interest [54].
It must be realized that even focusing more on indicators as com-
pared to comprehensive models with assured predictabilities is a new
way of scientific decision making, one in which administrators and
policy makers must see themselves as permaculture gardeners, tending
to a complex system they must let evolve organically with as little in-
terruption as possible. Even when interruption is necessary, it should be
realized that the only way to handle decision making in the face of
daunting complexity is to attempt to understand complex systems with
humility and with due respect for all forms of knowledge, scientific,
meta-scientific, numeric, fuzzy, narrative and expert.
We now modify two visualization tools from urban studies and
ecology for income inequality in national economies to explore the
correlation between energy consumption efficiency and inequality.
F. Khan, P. Heinecker
Fig. 6. Planning plane with colour change showing how energy use per capita
(kg of oil equivalent) changes with change in GDP per capita and scaling in-
dicator of income distribution (44 countries from World Bank 2005 data).
3.3.1. Fractal spectra for income inequality
A visualization of how this scaling indicator captures income in-
equality is shown in the ‘fractal spectra’ developed for five sample coun-
tries in Fig. 5. A fractal spectrum is defined as a plotting of squares re-
presenting different income shares by consecutive twentieth percentile
population segments on a y-axis representing the income share percentage
[35]. The square size and the location of the square represents the income
share for any given fifth of the population represented by shading or
colour. Rapid change in size across a thin income band is indicative of
greater disparity and higher concentration of wealth.
3.3.2. Planning plane
The tool we adapt here from urban planning [35] to facilitate the
use of disparity indicators of the kind we have proposed here, in policy
making can be useful in any policy analysis scenarios where more than
one variable needs to be considered. For instance an important focus of
research in sustainable development goals (SDG) is the analysis of co-
benefits between SDGs. This requires consideration of more than one
variable in planning. The planning plane is essentially a contoured
Appendix A
See Table A1.
190
Energy Research & Social Science 42 (2018) 184–192
plane that takes for input as x and y variables the two independent
variables, say GDP growth rates and fractal dimension based scaling
indicator proposed here. The colour or the third axes represents the
dependent variable, in this case energy consumption efficiency. If we
want to optimize for energy consumption efficiency we need to visua-
lize in policy discussion scenarios not just how GDP growth affects
energy consumption but how income inequality affects energy con-
sumption. The contours or colours in this example are calculated using
ordinary kriging.
The planning plane for how GDP and inequality scaling indicator
affects energy consumption is shown in Fig. 6. Since this is just ordinary
kriging the roughness captures non-linearity of the response of system
to changes in inequality and GDP, as composed to conventional re-
gression based modelling. To deploy the planning plane a policy maker
would identify either the desired GDP and income inequality and see
how that would affect energy consumption on the plane, or would
identify the energy consumption goals and based on that define what
GDP or income inequality targets should be.
4. Conclusions
In summary it is theorized here that the increase in energy consumption
with increasing disparity is owed to the increasing energy cost of ‘good
regulation’ in systems [46,47] as the system becomes more disparate and
unequal. In cities this difficulty in good regulation manifests itself as the
increasing cost of building a transport ‘model’ of a system which provides
good transportation services as a regulatory function in an increasingly
disparate city. In national economies the energy cost of regulation may
have to do with both the shrinking economy of scale in providing energy
intensive services to an increasingly disparate populace, as well the dis-
proportionately high energy consumption by population segments in
higher income brackets. We believe there is sufficient evidence here to
suggest that disparity does indeed affect energy consumption efficiency
adversely in urban and national economic systems and consideration of
scaling indicators of disparity in policy analysis as well as further research
in the area, is merited. From a policy perspective the results presented here
implore us to include sophisticated disparity measures in analysis for all
complex anthropogenic systems, especially for energy efficiency optimi-
zation. There are multiple ways in which the work presented here can be
used to include consideration of system disparity in the study of human
well-being. First of all the methods suggested such as planning plane could
be used to study the correlations between various well-being indicators
such as say life expectancy and disparity or inequality in systems along
with the typical economic indicators such as the GDP. The idea that GDP
alone does not capture the non-linear relationship between wealth and
well-being can be studied quantitatively through making sure that every
time GDP is used as an independent variable, the disparity or inequality in
GDP is also used as an independent variable and the effects of change are
considered both from the perspective of growth in volume and disparity.
Secondly the indicator presented here allows for development of credible
mechanisms to theoretically explain some of the empirical correlations
between inequality and well-being, economic and energy indicators. In
energy social sciences and transition studies the discussion and results
presented here will allow us to focus and explore in greater detail the non-
linear relationship between energy consumption and societal change. The
results also support the idea that inequality –as a special case of disparity in
systems- creates inherent inefficiencies in systems which may lead to
system collapse if not addressed adequately. The visual tools adopted from
urban studies and ecology to economic analysis can help plan for energy
efficiency improvements while exploring new economic policy.
F. Khan, P. Heinecker Energy Research & Social Science 42 (2018) 184–192

Table A1
Economic and Energy Use Data (circa 2010).
Country Name Energy use per % of Income % of Income share % of Income share % of Income share % of Income share Fractal dimension based
capita (kg oil share of highest of the 2nd highest of the middle 20% of the 2nd lowest of the lowest 20% scaling indicator for
equivalent) 20% earners 20% earners earners 20% earners earners income distribution

Argentina 1909.73 49.38 22.42 14.54 9.3 4.36 0.44777

Armenia 837.93 40.31 21.38 16.58 12.79 8.94 0.57129
Australia 5648.7 42.16 22.79 16.14 11.83 7.09 0.54477
Austria 4079.86 38.4 22.89 17.47 13.21 8.04 0.60435
Bangladesh 200.647 41.35 21.24 16.01 12.5 8.89 0.55427
Belarus 2900.22 36.85 22.79 17.67 13.65 9.04 0.62864
Belgium 5598.31 37.23 22.56 17.81 13.95 8.45 0.6242
Bulgaria 2416.36 42.22 22.7 16.73 12.1 6.26 0.54696
Cambodia 368.936 42.72 21 15.59 12.25 8.44 0.53419
Canada 7391.73 40.97 22.75 16.79 12.39 7.1 0.56414
China 1845.74 47.09 23.2 15.31 9.74 4.67 0.47785
Colombia 679.53 60.03 18.72 11.16 6.9 3.19 0.32366
Costa Rica 1022.18 53.77 20.32 12.9 8.57 4.45 0.39263
Croatia 1938.54 36.9 22.42 17.54 13.77 9.37 0.62725
Cyprus 2213.09 40.45 21.76 16.42 12.77 8.6 0.56925
Czech Republic 4237.9 36.44 21.97 17.73 14.49 9.37 0.63639
Denmark 3510.8 37.2 22.66 17.89 14.27 7.98 0.62601
Dominican 739.887 52.8 20.8 13.19 8.55 4.67 0.40369
Republic
Ecuador 919.422 54.35 20.15 12.97 8.42 4.1 0.38668
El Salvador 698.287 50 21.6 14.22 9.44 4.74 0.43925
Estonia 4222.74 39.76 22.84 17.06 13.25 7.08 0.58388
Ethiopia 469.658 41.67 21.34 16.34 12.61 8.03 0.55193
Finland 6819.56 37 22.59 17.42 13.81 9.17 0.62559
France 4024.25 41.82 21.79 16.33 12.4 7.64 0.54995
Georgia 795.286 47.58 22.36 15.07 9.94 5.04 0.47076
Germany 3997.05 39.29 22.62 17.13 12.95 8.01 0.58945
Greece 2482.63 41.15 22.97 17.15 12.25 6.49 0.56315
Honduras 608.227 57.59 20.18 11.91 7.13 3.19 0.35008
Hungary 2568.38 37.73 22.89 17.71 13.4 8.26 0.61533
Iceland 17023.2 36.08 22.16 17.99 14.42 9.37 0.64299
Indonesia 866.827 43.65 21.83 15.56 11.33 7.63 0.52119
Ireland 3161.54 40.23 22.55 16.92 12.72 7.59 0.57516
Israel 3042.43 47.4 22.89 15.48 9.62 4.61 0.47423
Italy 2867.74 41.06 22.94 17.09 12.52 6.38 0.56473
Jordan 1174.87 42.41 21.45 15.84 12.11 8.19 0.53929
Kazakhstan 4234.93 38.25 22.05 16.89 13.36 9.46 0.60345
Kosovo 1405.04 41.5 22.37 16.21 12.14 7.78 0.55372
Kyrgyz Republic 505.402 39.07 21.96 16.89 13.27 8.81 0.59156
Latvia 2148.65 41.8 22.77 17.14 12.26 6.02 0.55438
Lithuania 2275.76 40.35 23.24 17.57 12.45 6.4 0.57628
Luxembourg 8329.43 38.9 22.93 17.36 12.92 7.89 0.59609
Mexico 1486.03 53.89 19.79 12.92 8.67 4.72 0.39093
Moldova 984.625 40.33 22.42 16.62 12.51 8.12 0.57181
Mongolia 1452.97 41.07 22.52 16.48 12.14 7.78 0.56054
Montenegro 1898.29 37.05 23.23 17.76 13.24 8.73 0.62548
Nepal 379.922 41.42 21.87 16.23 12.14 8.34 0.55413
Netherlands 5021 37.23 22.79 17.77 13.87 8.34 0.6241
Norway 6939.49 35.3 22.81 18.05 14.48 9.37 0.65616
Pakistan 496.342 39.52 21.25 16.47 13.18 9.59 0.58249
Panama 1024.6 56.38 20.07 12.52 7.75 3.3 0.36461
Paraguay 774.221 56.27 19.52 12.69 7.95 3.57 0.36583
Peru 654.019 51.04 21.59 14.15 8.97 4.24 0.427
Poland 2640.24 41.34 22.25 16.37 12.18 7.86 0.55637
Portugal 2222.62 43.37 21.88 15.93 11.79 7.03 0.52744
Romania 1730.09 36.88 23.17 17.76 13.56 8.65 0.62888
Russian 4827.81 47.44 21.85 14.68 10.19 5.84 0.47078
Federation
Serbia 2141.06 38.2 22.81 17.36 13.26 8.37 0.60674
Slovak Republic 3306.81 36.07 22.72 18.24 14.46 8.51 0.6449
Slovenia 3539.25 34.76 22.67 18.23 14.65 9.69 0.66564
Spain 2742.88 41.91 23.29 16.84 11.99 5.97 0.5518
Sweden 5427.93 36.04 22.78 17.83 14.37 8.98 0.64351
Switzerland 3346.62 40.43 22.66 16.8 12.46 7.65 0.57138
Thailand 1766.93 46.5 21.66 14.79 10.45 6.62 0.48219
Tunisia 974.84 42.92 22.62 16.14 11.6 6.73 0.53436
Turkey 1455.83 45.18 22.34 15.59 10.99 5.9 0.50287
Ukraine 2886.99 35.16 22.39 17.93 14.3 10.21 0.65671
United Kingdom 3224.86 41.98 22.54 16.49 11.91 7.08 0.54827
United States 7161.51 45.75 22.96 15.7 10.49 5.09 0.49615
Uruguay 1210.58 50.92 21.44 13.74 8.99 4.93 0.4266
Vietnam 677.674 49.39 20.71 14.01 9.97 5.92 0.44492
Zambia 590.88 61.05 17.89 10.49 6.76 3.81 0.31111

191
F. Khan, P. Heinecker
References
[1] R.G. Wilkinson, K. Pickett, The Spirit Level, 2009.
[2] E. Tricomi, A. Rangel, C.F. Camerer, J.P. O'Doherty, Neural evidence for inequality-
averse social preferences, Nature 463 (7284) (2010) 1089–1091.
[3] F. Jaumotte, S. Lall, C. Papageorgiou, Rising income inequality: technology, or
trade and financial globalization, IMF Econ. Rev. 61 (2) (2013) 271–309.
[4] T. Piketty, Capital in the 21st Century, Harvard University Press, Cambridge, MA,
2014.
[5] T. Piketty, G. Zucman, Capital is back: wealth-income ratios in rich countries,
1700–2010, Q. J. Econ. 129 (3) (2014) 1255–1310.
[6] A. Roberts, Why the occupy movement failed, Public Adm. Rev. 72 (5) (2012)
754–762.
[7] K. Khatib, M. Killjoy, M. McGuire, D. Graeber, We Are Many: Reflections on
Movement Strategy from Occupation to Liberation, AK Press, Edinburgh, 2012.
[8] M. Buchanan, More equal than others, Nat. Phys. 10 (7) (2014) 471–471.
[9] M. Shermer, The myth of income inequality, Sci. Am. 311 (1) (2014) 86–86.
[10] N.N. Taleb, R. Douady, On the Super-Additivity and Estimation Biases of Quantile
Contributions, (2014) https://docs.google.com/file/d/
0B8nhAlfIk3QIbzRrRkhhc1RNY0U/edit.
[11] B.B. Mandelbrot, The Fractal Geometry of Nature, W. H. Freeman, Berlin, Germany,
1983.
[12] G.B. West, J.H. Brown, Life’s universal scaling laws, Phys. Today 57 (9) (2004)
36–42.
[13] G.B. West, J.H. Brown, B.J. Enquist, A general model for the origin of allometric
scaling laws in biology, Science 276 (5309) (1997) 122–126.
[14] A. Alesina, R. Perotti, Income distribution, political instability, and investment, Eur.
Econ. Rev. 40 (6) (1996) 1203–1228.
[15] U. Soytas, R. Sari, B.T. Ewing, Energy consumption, income, and carbon emissions
in the United States, Ecol. Econ. 62 (3–4) (2007) 482–489.
[16] C. Zhang, W. Zhao, Panel estimation for income inequality and CO2 emissions: a
regional analysis in China, Appl. Energy 136 (2014) 382–392.
[17] Y. Wolde-Rufael, S. Idowu, Income distribution and CO2 emission: a comparative
analysis for China and India, Renew. Sustain. Energy Rev. 74 (2017) 1336–1345.
[18] A. Berthe, L. Elie, Mechanisms explaining the impact of economic inequality on
environmental deterioration, Ecol. Econ. 116 (2015) 191–200.
[19] M. Camacho, G. Perez-Quiros, P. Poncela, Extracting nonlinear signals from several
economic indicators, J. Appl. Econ. 30 (7) (2015) 1073–1089.
[20] L. Kalliovirta, T. Malinen, Nonlinearity and Cross-Country Dynamics of Income
Inequality, (2017).
[21] D. Park, K. Shin, Economic growth, financial development, and income inequality,
Emerg. Mark. Finance Trade 53 (12) (2017) 2794–2825.
[22] B.K. Sovacool, What are we doing here? Analyzing fifteen years of energy scho-
larship and proposing a social science research agenda, Energy Res. Soc. Sci. 1
(2014) 1–29.
[23] J. Gao, B. Barzel, A.-L. Barabási, Universal resilience patterns in complex networks,
Nature 530 (7590) (2016) 307–312.
[24] S.E. Jørgensen, S.N. Nielsen, The properties of the ecological hierarchy and their
application as ecological indicators, Ecol. Indic. 28 (0) (2013) 48–53.
[25] N.A. Salingaros, B.J. West, A universal rule for the distribution of sizes, Environ.
Plann. B: Plann. Des. 26 (1999) 909–923.
[26] G.B. West, J.H. Brown, A general model for the origin of allometric scaling laws in
biology, Science 276 (5309) (1997) 122.
[27] G.B. West, J.H. Brown, B.J. Enquist, The fourth dimension of life: fractal geometry
and allometric scaling of organisms, Science 284 (5420) (1999) 1677–1679.
[28] L. Parrott, Measuring ecological complexity, Ecol. Indic. 10 (6) (2010) 1069–1076.
[29] A.-L. Barabási, R. Albert, Emergence of scaling in random networks, Science 286
(5439) (1999) 509–512.
[30] M. Batty, P. Longley, Fractal Cities: A Geometry of Form and Function, Academic
Press Professional, Inc., San Diego, CA, 1994.
192
Energy Research & Social Science 42 (2018) 184–192
[31] L. Benguigui, D. Czamanski, Simulation analysis of the fractality of cities, Geogr.
Anal. 36 (1) (2004) 69–84.
[32] L.M.A. Bettencourt, J. Lobo, D. Strumsky, G.B. West, Urban scaling and Its devia-
tions: revealing the structure of wealth, innovation and crime across cities, PLoS
One 5 (11) (2010) 1–9.
[33] L.A. Coward, N.A. Salingaros, The information architecture of cities, J. Inf. Sci. 30
(2) (2004) 107–118.
[34] G. Shen, Fractal dimension and fractal growth of urbanized areas, Int. J. Geogr. Inf.
Sci. 16 (5) (2002) 419–437.
[35] F. Khan, L. Pinter, Scaling indicator and planning plane: an indicator and a visual
tool for exploring the relationship between urban form, energy efficiency and
carbon emissions, Ecol. Indic. 67 (2016) 183–192.
[36] World Bank, World Bank Open Data, (2011) http://data.worldbank.org/ .
(Accessed 26 February 2014).
[37] B. Mandelbrot, How long is the coast of Britain? Statistical self-similarity and
fractional dimension, Science 156 (3775) (1967) 636–638.
[38] B. Fischer, 2013. http://blog.opower.com/2013/03/americas-energy-distribution-
the-top-1-of-homes-consume-4-times-more-electricity-than-average-and-why-it-
matters/. http://blog.opower.com/2013/03/americas-energy-distribution-the-top-
1-of-homes-consume-4-times-more-electricity-than-average-and-why-it-matters/.
(Accessed 14 May 2015).
[39] M. Mehrara, Energy consumption and economic growth: the case of oil exporting
countries, Energy Policy 35 (5) (2007) 2939–2945.
[40] A.M.M. Masih, R. Masih, Energy consumption, real income and temporal causality:
results from a multi-country study based on cointegration and error-correction
modelling techniques, Energy Econ. 18 (3) (1996) 165–183.
[41] J.B. Ang, CO2 emissions, energy consumption, and output in France, Energy Policy
35 (10) (2007) 4772–4778.
[42] W. Xiaohua, F. Zhenming, Survey of rural household energy consumption in China,
Energy 21 (7–8) (1996) 703–705.
[43] I. Ozturk, A. Aslan, H. Kalyoncu, Energy consumption and economic growth re-
lationship: evidence from panel data for low and middle income countries, Energy
Policy 38 (8) (2010) 4422–4428.
[44] A.M.M. Masih, R. Masih, On the temporal causal relationship between energy
consumption, real income, and prices: some new evidence from Asian-energy de-
pendent NICs Based on a multivariate cointegration/vector error-correction ap-
proach, J. Policy Model. 19 (4) (1997) 417–440.
[45] S.M. Hsiang, M. Burke, E. Miguel, Quantifying the influence of climate on human
conflict, Science 341 (6151) (2013).
[46] R.C. Conant, W. Ross Ashby, Every good regulator of a system must be a model of
that system, Int. J. Syst. Sci. 1 (2) (1970) 89–97.
[47] F. Khan, An adaptive learning mechanism for selection of increasingly more com-
plex systems, Int. J. Adv. Comput. Sci. Appl. 6 (6) (2013).
[48] A. Jacobson, A.D. Milman, D.M. Kammen, Letting the (energy) Gini out of the
bottle: Lorenz curves of cumulative electricity consumption and Gini coefficients as
metrics of energy distribution and equity, Energy Policy 33 (14) (2005) 1825–1832.
[49] J.D. Healy, Excess winter mortality in Europe: a cross country analysis identifying
key risk factors, J. Epidemiol. Community Health 57 (10) (2003) 784–789.
[50] J.A. Duro, V. Alcántara, E. Padilla, International inequality in energy intensity le-
vels and the role of production composition and energy efficiency: an analysis of
OECD countries, Ecol. Econ. 69 (12) (2010) 2468–2474.
[51] D. Meadows, Indicators and Information Systems for Sustainable Development: A
Report to the Balaton Group, The Sustainability Institute, Hartland, Vermont, 1998
p. 78.
[52] C. Perrow, Normal Accidents: Living with High-Risk Technologies, Princeton
University Press, Cambridge, 1984.
[53] S.O. Funtowicz, J.R. Ravetz, The worth of a songbird: ecological economics as a
post-normal science, Ecol. Econ. 10 (3) (1994) 197–207.
[54] N. Taleb, The Black Swan: The Impact of the Highly Improbable, Penguin, New
York, NY, 2008.