3.1. Constitutive Explanations: Basic Principles
Given the difficulties that I outlined in the previous section, we need a fresh methodological approach to the thermodynamics–economics relationship. I propose that this is the perspective of constitutive explanations, or the identification of causal mechanisms. This has been recently developed as an alternative to the regularity-based notion of causality, which is basically following the covering law model of scientific explanations, and introduces a mechanistic notion of causality that is independent from the notion of regularity. In the former view, causality is the regular co-occurrence of causes and effects that is governed by universal laws. The pertinent discussion in the Philosophy of Science has amply demonstrated that this methodological framework may only apply for certain parts of physics but fails to describe the methodology of most other sciences, beginning with chemistry [
45,
46]. This is particularly true for sciences dealing with complex systems. A foremost example is the neurosciences, for which so far the alternative approach has been elaborated in most detail, developing a mechanistic model of causality without referring to universal laws in an essential way [
47,
48]. I think that this explanatory model is most appropriate also for defining a methodological framework for the relationship between thermodynamics and economics.
Further, there is the important development that mechanistic explanations are also increasingly employed in the social sciences [
49,
50]. That means, we can unify the methodology of the sciences and the social sciences within one single framework, thus opening up the view on mechanisms in the domains of the former and the latter as closely interacting phenomena in specific integrative models of constitutive explanation. For example, in the neurosciences, the explanation of empathy rests upon the identification of specific neurophysiological mechanisms that generate emphatic behavior, but at the same time it is impossible to identify a general regularity without referring to the symbolic domain in human sociality in which the difference between the in-group and out-group is specified that drives the activation of emphatic mechanisms [
51]. In the following, I argue that the same applies for the thermodynamics–economics link: A constitutive explanation combines mechanisms of different levels into one overarching systems view. In the previous example of the comparison between China and England during the period of European industrialization, that would include larger ecological structures, physical and engineering aspects of energy utilization, the reflections of these in economic parameters such as prices, or the specific patterns of the social organization of the utilization of technology, including aspects such as the fundamental religious and philosophical views of nature.
This observation can be generalized. The systems in question are multi-level in the ontological sense: In the neurosciences, there are chemical mechanisms on the molecular level in the human brain, there are higher levels of neuronal connectivity, there are interactions between brain areas, and finally there is a direct causal involvement of extra-somatic phenomena such as signs mediated via sensory perceptions. Therefore, the neurosciences are actually a multi-disciplinary endeavor: the complexity of the phenomena is reflected in the complexity of interacting and diverse disciplinary explanations, as has been also transpiring in the previous remarks about comparing England and China. As a consequence, constitutive explanations are also non-reductionist: Contrary to common philosophical stances that relate the notion of causality to the goal of reductionism, thus also suggesting the ultimate reduction of social phenomena to natural ones, constitutive explanations adopt a naturalistic ontology of causal explanations, but recognize the independent ontological status of social phenomena, in particular [
52].
In simplest terms, a constitutive explanation describes the specific mechanisms that come together in generating a certain result, given certain inputs, and make the architecture explicit that governs the interactions of these mechanisms [
11,
53]. This is the “constitution” of the system in question. That means both causes and effects are seen as being aspects of the state of the system at a particular point of time, and there is no reference to a universal law that directly explains this co-occurrence (although for single and elementary causal processes as parts of the larger mechanistic set-up laws may hold). This deserves emphasis: If we aim at a constitutive explanation of the energy–growth link, we do not actually care about “prime movers”, but approach energy (
i.e., input) and growth (
i.e., output) as co-instantiations of the interplay of mechanisms in the productive system “economy”, which may manifest phenomena of non-linearities and interactions between bottom-up and top-down causal processes, to name just two examples of architectural complexity. For example, this is salient in the difficulties to unequivocally establish whether energy drives growth or growth drives energy in the highly aggregate black box models [
8]. This systems view on causality avoids the fallacy of interpreting the relationship between causes and effects in a reductionist manner, thus enforcing an “either-or” dilemma.
Thus, constitutive explanations open up the black box between inputs and outputs. Therefore, econometric analyses are not seen as having conclusive explanatory power. This is because in complex and evolutionary systems, correlations for high levels of aggregation do not explain why and how the regular co-occurrence of inputs and outputs is actually produced by the system. Especially, there are difficult problems in identifying and assessing the role of non-occurrences and omissions as causal forces, and there is no necessary relationship between high probability of inputs and high probability of outputs. Therefore, statistical correlations have heuristic value, but should be always undergirded by the detailed identification of underlying mechanisms. This defines an essential difference to common economic methodology.
Constitutive explanations have a general structure which consists of four steps [
54]. The first step is the analysis of individual phenomena on a particular ontological level. For example, in the neurosciences this might be the chemical mechanisms that operate at synapses; in the social sciences, these might be individual decisions. The second step is to analyze the interaction between different individual phenomena in a certain context, such as the interplay of synapses in a certain structural segment of the brain, or the individual transactions on a market. The third step is to explain aggregate phenomena that emerge from these interactions, such as certain brain states which might relate to certain higher level phenomena such as certain perceptions, or certain states of markets such as trends of prices. The fourth step is to understand the feedback mechanisms that work from the aggregate level to the individual level, such as brain states impinging on the activity levels of individual synapses, or market states that impact on individual decisions. For all these steps, it is essential to empirically identify concrete and specific mechanisms that explain the observed phenomena, thus resulting in a specific architecture which is the conceptual frame for the constitutive explanation.
Finally, constitutive explanations lean towards a manipulability account of causality [
55]. That means, identifying causes relies on the criterion whether the relevant phenomenon can be manipulated in order to achieve a particular effect; so, the approach directly ties up with the central role of the experiment in the sciences. Manipulability also includes the identification of “natural experiments” in which we can clearly describe a constitutive set-up and trace crucial phenomena that operate as causes of subsequent developments. So, taking the example of China again, as we have seen, a broad and detailed discussion of the different possible causes for her failure to industrialize endogenously has increasingly focused on the central role of coal as a source of energy, and the differences between the induced relative prices of energy and labor in England and China. In this context, manipulability in the social sciences also goes along with the local nature of many processes, thus highlighting the importance of medium-range theorizing, as opposed to universal laws [
56]. In the example, the launch of the Industrial Revolution in Europe was initially driven by local price structures in the spatially circumscribed regions of coal mining, and the following diffusion dynamics was determined also by many other factors, as mentioned previously [
41].
3.2. Rebound Effects as a Case of Constitutive Explanations
What does the model of constitutive explanations imply for integrating thermodynamics and economics? Let me illustrate the logic of mechanisms with one important example, the so-called rebound effect in energy economics. Whereas in the aggregate approach to the energy–growth link the rebound effect would only play a marginal role (as in [
5]), in the constitutive explanations methodology rebound effects and their precise causal architecture would assume a core position. Indeed, many authors recognize this fact, but lacking a concise methodological beacon, would not reach the appropriate theoretical conclusions, thus approaching the phenomenon as a mere empirical issue (see e.g., [
17]: 271f).
The rebound effect relates to a most universal pattern of economic and ecological dynamics that already transpired in the previous brief discussion of HANPP: human appropriation of net primary production has grown incessantly, yet with continuously improving efficiency. The rebound effect means that enhancing the efficiency of energy usage might induce individual behaviors which do not result in absolute savings of energy but in growing absolute amounts of energy flows. This effect was first identified by the British economist Jevons in his famous book “The Coal Question” in which he argued that technological progress in mining and exploiting coal in England will even sharpen the scarcity of coal as demand will grow even stronger than supply [
57]. In the discussion of the rebound effect, we observe the same difficulties regarding the identification of causality as discussed in the previous section: Researchers investigate whether improvements in energetic efficiency are causes of the effect of absolutely growing energetic throughputs. In the constitutive explanations view, both causes and effects are aspects of the complex system that consists of many mechanisms that come together in generating this pattern.
There are different venues to explore the rebound effect. One venue is the economic approach in the narrow sense that would aim at presenting a theoretical model of growth in which the effect would occur as a universal regularity [
58,
59]. The basic idea is that enhancing energy efficiency would also increase total factor productivity, which feeds back on growth and triggers growing demand for energy. However, whether such a regularity can be theoretically established heavily depends on a number of additional assumptions, such as the mathematical form of the production function. Another reason for the inconclusiveness of these economic modelling efforts may be, as Sorrell states [
57]: “The ‘rebound effect’ is an umbrella term for a variety of mechanisms that reduce the potential energy savings from improved energy efficiency”. This observation indeed suggests the necessity of shifting the methodological perspective.
If we analyze the rebound effect in terms of the four step model of constitutive explanations, we start out from the individual level,
i.e., the individual decisions to produce and consume coal (this corresponds to “direct rebound effects”. Next, these are reflected in individual-level activities of substitution and complementary activities (the “indirect rebound effects”). We then look at the interactions between those individual decisions on the marketplace, resulting in specific market states, that is prices for coal, applied technologies and so forth. This results in adaptations of the economy through time (the “economy-wide rebound effects”). So, the rebound effect is a complex set of interacting mechanisms [
60]. There are mechanisms that are involved in individual decision making, such as connecting preferences, prices and decisions. These mechanisms can be complex in turn, for example when we introduce expectations about future prices. Next, there are mechanisms of the interactions between individuals: These are often describable in terms of “institutions” that shape the specific forms of markets and result in certain regularities of interactions, such as, for example, the contracts that govern the use of energy. There is a large number of such transactional mechanisms that interact on the marketplace: Hence, the market itself is a complex mechanism at a higher level of organization. One particular effect on the market level is the emergence of market prices. Market prices are essential determinants of individual decisions. Here, the loop is closed back to individual decisions.
We can also approach the rebound effect as a “type” of mechanism. A type of mechanism amounts to a theoretical hypothesis, but it is easy to see that this is not based on a universal theory, but is a medium-range theoretical concept: It applies in specific regions of time and space, and does not work universally. Whether and how the rebound effect works is an empirical question, and requires opening up its black box. Yet, at the same time, the rebound effect is essential for predicting future trends in the relationship between energy and economic growth. In the current context, what is most significant is the fact that the empirical rebound effects are mostly below unity when considering single mechanisms, such as the effects of energy-saving devices for the utilization of a single product consuming energy, but may turn out to be larger than unity once one considers full-scale technological systems (such as the steam engine and its ramifications). Therefore, the rebound effect also points towards an essential challenge in constitutive explanations: This is identifying the proper boundaries of the mechanisms in question, both in space and time. In the case of the rebound effect, one aspect is the embeddedness of single technological devices into larger technological systems, with the important example of so-called “general purpose technologies” which diffuse across a large number of specific domains of application, and the other is the spatial interaction between different economic systems via international economic linkages. For example, Information and Communication Technology (ICT) is a complex amalgam of technological devices that define a “general purpose technology”. Looking at single devices, many observers perceive strong potential for energy savings, but the picture is much less clear for the evolution of the entire pattern of ICT applications, including their worldwide diffusion [
61].
The ICT example is illuminating because it also shows that the analysis of mechanisms also needs to be based on universal theories, as far as their fundamental characteristics are concerned. For understanding the relationship between ICT and growth, it is absolutely necessary to properly identify the causes of energy consumption by ICT. At the heart of this issue lies a foundational ontological and physical question, namely what is the relationship between energy and information, and how does this play out under specific technological conditions [
62,
63]: The central, still disputed question is whether and how memory and the erasure and superscription of information go along with thermodynamic costs which obey to the Second Law; if so, one would end up with a most universal rationale as to why the expansion of ICT must go along with rebound effects that are larger than one, if we consider the entire ICT technological system on a global scale. It is not enough just to consider current energy savings and current energy consumption of particular ICT devices in order to ground predictions about future developments. So, mechanism analysis needs to be based on a precise understanding of the different constituent phenomena, which ultimately also includes reference to fundamental physical laws. However, that does not imply that these laws can directly offer causal explanations of the relationship between energy, information and growth. The causal explanation rests upon the analysis of the complete multi-level system of interacting mechanisms.
So, in the constitutive explanations framework, the rebound effect is not only an excellent case for illustrating the power of this approach in appreciating practiced approaches in empirical energy economics and examining them on a systematic methodological basis, but also reveals that it might suggest more general hypotheses about the types of mechanisms that determine the energy–growth link. So, the methodological appreciation has direct consequences also for theoretical and empirical work (compare the assessment by Sorrell [
57], also referring to Ayres and Warr [
5], with the marginal treatment of the effect in their work; similarly, Kümmel [
6] only mentions rebound effects in the context of a citation). In the next section, I show how this shift of methodological perspectives suggests fresh approaches to some of the issues raised in the second section of this paper.
3.3. The Mechanism of Growth: Urbanization as a Physical Phenomenon
One important example for rebound effects is the growth of electricity usage which has been recurrently triggered by declining costs of producing electricity. In this context, one particularly interesting phenomenon is the absolute growth of energy throughput used for lighting, which is driven by falling energetic and other costs. On the one hand, for single lighting devices, it is mostly straightforward establishing rebound effects that are clearly smaller than unity or even negative, thus indeed resulting not only in higher energetic efficiency, but even absolute savings (as a recent example, [
64]). However, if we look at the trends in the long run, and broaden the perspective to the global dimensions of the diffusion of lighting, the rebound effects appear to be at least unity or even larger than unity [
65,
66]: That would imply lighting alone is an important driver of expanding energetic throughputs in economic growth (consuming roughly 7 percent of total global energy consumption). In explaining this phenomenon, there are different cross-disciplinary aspects, such as the biological and cultural determinants of the human need for lighting, and the obvious fact that the saturation point has not yet been achieved. However, there is also the observation that improvements in lighting have effects on productivity and creativity, especially with regard to the extension of human activity into night times [
67]. This argument extends the mechanisms underlying the rebound effect considerably, as there is a feedback from growth to the demand for lighting which is not directly triggered by cost considerations.
Indeed, changes and trends of intensity of night lighting, as observable via satellites on a global scale, have been recently identified as one of the most reliable indicators of GDP growth and levels of GDP per capita [
68]. The underlying rationale is that the expansion of economic activity implies the extension of activities into nighttime, both work and leisure. This close relationship raises the interesting possibility to define alternative means of measuring economic activity that have a direct categorial affinity to thermodynamics, and avoid the difficulties that arise from the reconciliation between monetary valuations and thermodynamic measures. In other words, if economists can use lighting as an indirect indicator of GDP, we can also invert this relationship and treat lighting as one parameter by which the growth of economic activity can be measured in physical magnitudes, without referring to GDP at all. At the same time, we can straightforwardly establish that rebound effects would play a central role in dissecting the causal processes that link energy and growth.
Lighting is an important phenomenon also for the reason that this directly ties up with the size and the growth of population, so also overcomes the problematic partitioning between the two aspects of growth. In other words, we can envisage to measure economic growth by means of analyzing the expansion of human energetics, hence including technology not in the sense of production, but in the sense of extensions of the human phenotype, along the lines of recent theorizing in biology about the inclusion of external artefacts in the notion of biological organism [
35,
39,
69]. Extensions are, among others, artefacts such as clothing or technologies for food preparation, but most importantly the entire technology of human dwelling, of which lighting is one aspect. Remarkably, energy consumed by human dwellings,
i.e., buildings with the accompanying infrastructure, represents the bulk of human energy consumption, far surpassing energy consumption by industry or transport [
70]. From this follows that a possible measure of economic activity in thermodynamic terms is the growth of the stock of capital that makes up the human extended phenotype, in the aggregate.
This argument can be streamlined methodologically in the constitutive explanations framework. The central point is how we conceive, identify, circumscribe and characterize the system which displays rebound effects in general and also with special reference to aspects of human life on Earth, namely the artificial environments in which humans live today, of which lighting is one essential aspect (another essential one is heating and cooling). This methodological focus can be now grounded in general evolutionary theory, hence in the discipline of biology, thus mediating between physical (thermodynamic) and social and cultural phenomena. In this view, what matters are the means by which humans improve their biological functions by technological artefacts, thus evolving new structural and behavioral features of adaptation, a perspective that has also been explored in the Philosophy of Technology [
71].
We can now give a more precise description of the systems in question which can be approached in terms of mechanistic explanations. The specific process in question is “urbanization”, and the systems are human urban settlements. In economics, urbanization is mainly seen as a source of GDP growth, and in fact its driving force, as has been especially emphasized in the so-called “New Economic Geography” [
72]. In a thermodynamic perspective, however, urbanization is the primary expression of economic activity, apart from agriculture, resulting in vast extensions of the human phenotype into complex urban technological systems and infrastructure that intensifies the flow of resources and energy: To mention England
vs. China again, the extraordinary growth of London was a crucial factor in driving the demand for coal as a supplier of energy [
41]. Economic growth is manifest in the expansion of urbanization, where a large number of complex self-reinforcing mechanisms and economies of scale (or, in another term, superlinear interactions) work together in further propelling economic growth [
73,
74,
75]. This is driven by growing absolute amounts of energetic throughputs, though at higher levels of efficiency and productivity.
In order to understand these phenomena, it is again important approaching urban systems as a complex structure of operating mechanisms, in particular, firstly, as systems that organize the flow of resources and ultimately, energy, and secondly, as dynamic networks of interacting people [
76]. These systems can be analyzed by a large array of analytical approaches which so far have been only marginally employed in economics, such as network analysis, and which allow to make the mechanisms of urban growth explicit [
77]. Most importantly, this is also a research area where physical research and social science increasingly converge in similar methods and even theoretical hypotheses [
78]. So, we would no longer approach economic growth at the aggregate level as expressed in GDP data, but as a complex evolving system of mechanisms dubbed “urbanization” (for a congenial in historical sciences, see [
79]). Then, coming back to the issue of rebound effects, we would not investigate aggregate interdependencies between energy and growth, but aim at understanding how mechanisms of urbanization result in those effects, as discussed in the example of lighting. This also opens up the possibility to achieve substantial generalizations. Most significantly, as Bettencourt [
75] has recently shown, for both the energetics of urban flows and the productivity of social networks, superlinearity seems to apply, resulting in a stable relationship between inputs, throughputs and outputs (in other words, technological innovations of urbanization would not result into absolute reductions of energy consumption). Theoretical relationships like this would provide a systematic foundation for the more detailed mechanistic analysis of particular rebound effects. At the same time, the pivotal role of rebound effects in understanding the energy–growth link would be confirmed.
Based on these considerations, I conclude with sketching an alternative approach to measuring economic activity and growth. This approach would be firmly embedded in evolutionary theory in focusing on the human organism, hence including population growth as an essential parameter. However, economic growth would be also seen as relating to the qualitative growth of the population, which boils down to the evolution of the human extended phenotype. This extended phenotype is defined via all artefacts that further enhance or enable human organismic functions, hence, in most general terms, enhance adaptive functions of the human organism (for a pertinent extension of measures of adaptation, see [
80]). In arguing this way, we also need to consider the fact that human adaptation is “social”, if not “ultrasocial”, in the sense that even without considering technological artefacts, the human appropriation of natural resources is based on a social division of labor [
20]. However, we can also assert that this division of labor as a collective phenomenon is only enabled by the technological infrastructures, as is most salient in the phenomenon of urbanization. Therefore, we can envisage a measure of economic growth that directly builds on the measurement of urban infrastructure, including all supporting structures which are necessary to maintain it. As I have shown, this is, for example, the capital stock of urban buildings and the flows that maintain living in buildings, such as lighting (this compares with related approaches in the literature which refer to accumulated GDP as a measure of thermodynamically relevant output [
24,
25]).
Measuring this capital stock as an aggregate quantity, however, would eventually also involve the use of monetary prices. However, the difficulties differ from using GDP data, as the capital stock would include both private and public stocks, and as there are many approaches and databases that allow assigning economic value to this stock. However, we can also, as a first approximation, refer back to population numbers: After all, the entire urban infrastructure serves to maintain a living for people. Therefore, I propose to regard the absolute and relative growth of the urban population as a measure for economic growth in an integrated framework of thermodynamics and economics. This is a most parsimonious way to include the two modes of growth in one measure that is directly compatible with a thermodynamic perspective: We treat the expansion of the extended phenotype by means of artefacts as the expansion of human energetics from considering the human metabolism only towards the inclusion of the urban technological metabolism that maintains the growing share of urban population as a part of the expanding total population.
Interestingly, this perspective also ties up with measuring the ecological impact of human activity by means of HANPP as this also includes the use of land for infrastructure as a form of land use [
22]. We can infer also normative and design conclusions from this observation, thus establishing connections to the discourse of Ecological Economics as well. One simple ecological imperative is that land utilization by urban growth should be kept as small as possible, thus favoring expansion of urban infrastructure into the “third dimension”,
i.e., high-rise buildings with high degrees of energetic efficiency in cooling and heating; this speaks against the global trend of suburbanization [
81].
Coming back to my example of China, again, the two aforementioned modes of growth are directly reflected in the pattern of urbanization that was dominant until the mid-20th century. China had achieved a comparatively high level of urbanization in the Middle Ages, reflecting the mediaeval revolutions in agriculture and market organization. However, later, the growth of large cities was stalled, in favor of the spread of small-scale semi-urban settlements which remained deeply integrated with the rural areas [
82]. This agrarian regime was sustainable over centuries, achieving high levels of efficiency without investing ever growing energetic resources into the expansion of urban structures. In Western countries, the accessibility and activation of huge amounts of fossil resources allowed for investing into the build-up of urban physical structures (for a related theoretical argument on the relationship between resource abundance and fixed costs of urbanization
vs. agrarian structures, see [
83]). China’s recent explosive economic growth is accompanied by the rapid expansion of urban and metropolitan areas, so that we can directly approach this expansion, reflecting the extremely high rate of investment into urban infrastructures, as a measure of economic growth. Interestingly, China is also unique in achieving growth with extremely high rates of savings and investment: A considerable share of these investments is devoted to real estate, meaning urbanization. So, the distinct population dynamics of China past and present, corresponding to distinct patterns of growth in terms of GDP, can be directly interpreted as a measure of growth, with the exploding share of the urban population in total population, with now low rates of total population growth, indicating the switch between the two modes of growth. We can also directly apply the aforementioned design considerations, as one conspicuous feature of current Chinese urbanization are urban sprawl and the inefficient use of land [
84].