Challenges of complexity economics
By Joachim H. Spangenberg and Lia Polotzek
In recent WEA Commentaries, the issue of complexity theory and its implications for economics have rightfully gained some prominence. However, while the authors picked up some relevant points, the issue deserves a more comprehensive treatment in new economics, beyond mobilising some arguments to bolster ongoing debates. It should be recognised instead that complexity requires a different way of thinking, and of asking questions in economics. Only then the specific tools used in complexity research, unconventional as they are from a standard economics point of view, come into play. Thus we will briefly describe what we see as core elements of complexity, the corresponding world view, and the tools used.
Complexity economics is a genuine theoretical approach based on applying complexity analysis to the economic system; it requires a world view different from the one of neoclassical economics, moving from reductionist linear thinking to non-linear approaches of conceptualising the economy. In system science parlance, a ‘system’ is any set of things within a common frame (the system boundary) that is ruled by a given set of interactions (the system rules). Applying the terminology to the whole of the outside world, three systems have to be distinguished (Sayer 2000; Spash 2012): The ‘real-world system’ or ‘the reality’ is the object we would like to know more about. However, this system is not accessible to direct human observation since our perception is limited by the senses and instruments we have and interpreted by our brain.
The result is a ‘mental model’, an imagination of reality, a simplified system which provides our ontology. The ontology shapes expectation and questions asked, is the basis of the interpretation of experiences and observations, and shapes the recommendations derived from them. However, it is usually neither reflected nor made explicit, often rather being an unconscious model of the world and its functioning the analyst holds. Ontologies, like all mental models, are best described in qualitative narratives or storylines. The third system, computer models, are the tools used to quantify a selected set of the expectations raised by the mental models. They are limited by the system margins and the necessarily relatively simple descriptions of a limited number of interactions within the system. Which elements are taken into account (i.e. realised in the computer model), and which interactions are considered and thus programmed, depend on both the ontology and the limitations imposed by the modelling technology chosen. Surprisingly, most public trust lies in these most simpliﬁed models.
While the system boundaries can be defined according to the research question analysed (choosing the subsystem of interest), this is not the case for the system rules which are defining the functioning of a system and its subsystems. The number of rules needed to describe the system functioning is a good measure of the respective system’s complexity – the more rules are given, the more system behaviour is constrained and less complex. On the basis of Allen (2001), we can deﬁne ﬁve distinct rules, which, if they all apply, signal maximum determination. The ﬁve system rules are, simply expressed:
- It is possible to distinguish between ‘the system’ and ‘its environment’. Deﬁning the border line is crucial as what the system can describe is only what is inside; it is a condition for the very existence of a system. When economists regret that their predictions did not correctly predict real-world developments, explaining that with unforeseen ‘external factors’, they essentially indicate that they have drawn the border line in the wrong place, excluding factors decisive for the functioning of the system.
- All system components can be recognised and distinguished, which means it is possible to describe and at best understand their interaction.
- The active system elements are all identical, or at least the range of their behaviour is normally distributed around the average. In an economic system, for instance, consumers and producers are key active system elements. Microeconomics tries to understand their interaction by analysing the interplay between ‘representative agents’– one consumer representing all consumers and one producer representing all producers. To be able to do so, one must assume that all consumers and all producers are identical regarding their behaviour in the situation analysed. In a biological system, the range of behaviour of individuals of the same population tends to be centred around an average (if two dominant patterns exist, they can be considered as the core of behaviourally different subpopulations, which are system elements).
- The individual behaviour of the system elements can be described by average interaction parameters which characterise the system behaviour. This implies that producers, consumers and others always follow the same set of behavioural rules and norms (with some stochastic variation); they are extremely stubborn, do not learn or change their behaviour towards others, at least not as groups (this is not a statement about individual behaviour and learning, and similarities with known professions are purely incidental). The rationality of the selfish human in standard economics embodies these characteristics. The result is a deterministic development, with at most a random variation around the predicted outcome.
- The system develops towards a stationary equilibrium, which permits deﬁning ﬁxed relations of system variables. If this is the case, the future is perfectly predictable as the development trajectory of the system is deﬁned and unchangeable. This is an abstraction, a mechanistic contract: machines behave like that but no natural, biological or social system does.
Using the five rules, we can distinguish the levels of complexity between different complex adaptive systems. Geo-physical systems like the climate system fulfilling rules 1 to 4 can evolve and adapt, which makes transitions towards different attractor basins possible when external conditions change, a phenomenon we also know as crossing tipping points. Biological systems only match rules 1 to 3. They have a higher degree of complexity due to the individual behaviour of agents which can deviate from the standard behaviour of a representative agent and its fuzzy borders, making transitions towards different attractor basins even easier. Again one dimension more complex are anthropogenic systems (societies, economies, etc.) restricted only by rule 1 and 2, as here the agents are capable of anticipation. Modification of behaviour not randomly but based on expectations can avoid structural changes, but – if as so often expectations are wrong – can also result in accelerated and intensified changes. Unlike for biological models, taking this trait into account is a necessary condition for suitable economic models (alternative mental models, imaginaries – computer models so far fail to deal with this level of complexity). Thus when Maria Alejandra Madi discusses complexity theory in Commentaries 8(4), she is right defining it not as tool driven, but a genuine theoretical approach, but when equating the complexity of natural and anthropogenic systems, she underestimates the systemic differences.
The complexity theory approach is part of a distinct world view. The philosophical literature on the concept of worldview dates back to Immanuel Kant, who coined the term “Weltanschauung” in 1790. In the literature, the elements most frequently discussed as constituents of a worldview are ontology, epistemology, axiology and anthropology (Hedlund-de Witt 2012). Ontology is a section of philosophy dealing with questions concerning the nature of being, and in particular questions regarding how and under what circumstances entities exist or may be said to exist and how such entities may be grouped, related within a hierarchy and subdivided according to similarities and differences. Epistemology is the branch of philosophy dealing with the theory of knowledge. It studies the nature of knowledge, justification and the rationality of belief, describing the kinds of knowledge we can have about an entity identified by the ontology (hence the distinction of three levels of models is already part of our epistemology). Axiology is another branch of philosophy, encompassing a range of approaches to understanding how, why, and to what degree humans should or do value objects (entities), whether the object is physical (a person, a thing) or abstract (an idea, an action), or anything else. According to Hedlund-de Witt (2012), it should include a societal vision. It also determines the ethics pursued and thus should be made explicit when developing proposals for action (the discount rate built into current economic models determines the value of future development, making it an implicit ethics causing a lack of transparency). Philosophical anthropology describes the conditio humana, the essentials of human existence, and the nature of human beings, the latter typically used in the context of ambiguous subjects such as moral concerns and human reflections on the meaning of life.
The neoclassical economics world view describes a world consisting only of monetary flows, with the economy the meta system. The ecological economics ontology which we endorse considers the environment as the meta system in which society is embedded, and the economy is a subsystem of society. Together they form a dynamic panarchy (Gunderson, Holling 2001). All three are complex evolving systems of different complexity, and tend to follow some variant of the Holling cycle of resilience (Holling 2001). In this cycle a phase of expansion, growth and development is followed by one of stabilisation, a metastable state with still dynamic changes, but fixed underlying structures and key relations. It is followed by a phase of disruption, usually a rapid process after passing a tipping point which can start as slow degradation accelerating or happen without previously observable indications. Then reorganisation happens, making use of resources left from the previous cycle but developing a new system. Phases of apparent stability should not be misinterpreted as equilibria: they are rather dissipative patterns far-from equilibrium, with their basic patterns maintained by the permanent throughput of matter and energy (Prigogine, Stengers 1984). The slow-to-no growth situation of most affluent economies can be understood as the metastable interlude between expansion and disruption.
The epistemology used in neoclassical economics is a positivist one, based on the assumption that the world can be fully understood and measured. As opposed to that, the one we use is rooted in the assumptions of critical rationalism. The world is a concrete reality, a complex system characterised by prevailing and unavoidable risk, uncertainty and ignorance. We can perceive reality only indirectly through senses and instruments, which influence our perception, often unconsciously (as critical realism postulates and environmental sociology shows). Our ontology influences the interpretation of observations with a tendency to realign them as long as possible. Models are recognised as delivering incomplete information which needs to be understood in the context of the mental models and ontologies behind them, and be critically reflected. While complexity economics is accepting diverse value systems, the axiology of neoclassical economics is dominated by “economic rationality”, considered an anthropogenic constant which – together with the methodological individualism considering each individual as independent from social influences – is also shaping its anthropology. As opposed to that, complexity economics accepts human beings in their ambivalence as social beings, their behaviour influenced by both egoistic instincts and genuine social practices, shaped by their respective social, institutional and infrastructure context (Spangenberg, Lorek 2019). While according to the insights of sociology, psychology and political science this is more realistic than the standard economic assumptions, it makes predictions almost impossible as there is not one binding logic all individuals must follow at all times.
Greg Daneke in Commentaries 9(2) rightfully describes complexity economics as using specific, unconventional tools such as “a variety of computational tools (nonlinear math, neural nets, cellular automata, adaptive algorithms, etc.) to simulate the co-evolutionary interaction of heterogeneous agents (exhibiting cooperative, reciprocal, and even altruistic behaviours) and their institutions”. To this list focussed on new models and algorithms qualitative methods, text and discourse analysis, empirical methods, polls and questionnaires should be added. Complexity economics is methodologically diverse; models do not play a dominant role as in standard economics but are rather support tools for more complexity bearing narratives.
Thus complexity economics indeed uses different tools than standard economics, and for good reasons. Analysing the available tools from a complexity perspective makes it crystal clear that the tools of economics are undercomplex and will not be able to deliver results adequately describing economic developments (see also Ciarli, Savona 2019). Equilibrium models follow rules 1 to 5 and system dynamic models rules 1 to 4; both are deterministic and have problems dealing with uncertainty and ignorance (stochastic variation as in fuzzy models is not uncertainty). As relative equilibria are considered to be just an interim phase of the Holling cycle, equilibrium models are only justifiable – if at all – for analyses of short term developments. However, in standard economics and in the Integrated Assessment Models (IAM) used in climate science, they are used for the opposite. Agent-based modelling uses identical agents (but usually deﬁnes more than two groups of agents) to analyse the interaction mechanisms of societies and adheres to rules 1 to 3. No model matches the complexity of reality (and most mental models); the best available option appears to be a combination of agent based models for social and economic processes, embedded in a system dynamics environment model.
Generally speaking, in order for computer models to be adequate (scientifically rigorous and socially robust), though, the mental model –already a simpliﬁcation of reality – must ﬁrst capture the major behavioural traits of ‘reality’, reflect and integrate them as the basis for deriving strategies effective when applied in a real-world context. Only then can the required technical tools be chosen or developed, attempting to enable them to express the main characteristics of the mental model, many of them qualitative in kind, and quantitative functions derived complementing and illustrating the qualitative mental models. As the mental model, expressed in scenario narratives or story lines, can accommodate qualitative aspects in a way no computer model can, the mental model narrative is the matrix in which diverse and complementary computer models can be embedded, illustrating and quantifying speciﬁc aspects of the scenario (Alcamo 2001). As both mental models and even more so computer models are simpler than the reality they describe, we should be aware how the simpliﬁcations that are inherent to the model (and that indeed is, to a certain degree, its purpose) impact the recommendations derived. In particular, when ‘the reality’ makes itself felt, confronting our expectations with unexpected experiences in a way that cannot be overlooked, the prevailing construction of the two derived systems must be considered falsiﬁed and due to change. Unfortunately, this basic principle is not always adhered to in standard economics.
Alcamo, J. (2001). Scenarios as tools for international environmental assessments. EEA Expert Corner Report Prospects and Scenarios No. 5. Expert Corner Reports. EEA European Environment Agency. Luxembourg, Office for the Official Publications of the European Communities: 31.
Allen, P. M. (2001). The Dynamics of Knowledge and Ignorance: Learning the New Systems Science. Integrative Systems Approaches to Natural and Social Dynamics. H. M. W. Matthies, J. Kriz,. Berlin, Heidelberg, New York, Springer: 3-30.
Ciarli, T., Savona, M. 2019. Modelling the Evolution of Economic Structure and Climate Change: A Review. Ecological Economics 158: 51-64.
Daneke, G. 2019. Why Economics is Still Not a Science of Adaptive Systems. WEA Commentaries 9(2).
Gunderson, L.H., Holling, C.S. (eds.) 2001. Panarchy: understanding transformations in systems of humans and nature. Washington D.C., USA, Island Press.
Hedlund-de Witt, A. (2012). “Exploring worldviews and their relationships to sustainable lifestyles: Towards a new conceptual and methodological approach.” Ecological Economics 84: 74-83.
Holling, C.S. 2001. Understanding the Complexity of Economic, Ecological, and Social systems. Ecosystems 2001(4): 390-405.
Madi, M. A. 2018. Complexity in Economics. WEA Commentaries 8(4).
Prigogine, I., Stengers, I. 1984. Order out of Chaos: Man’s New Dialogue with Nature. Toronto-New York-London-Sydney, Bantam Books.
Sayer, A. (2000). Realism and Social Science. London, UK, Sage Publications Ltd.
Spangenberg, J. H. 2015. Sustainability and the Challenge of Complex Systems. J. C. Enders, M. Remig (eds.), Theories of Sustainable Development. Routledge, Abingdon, UK: 89-111.
Spangenberg, J.H. 2016. The world we see shapes the world we create. How the underlying worldviews lead to different recommendations from environmental and ecological economics – the Green Economy example. Int. J. Sustainable Development.
Spangenberg, J.H., Lorek, S. 2019. Sufficiency and consumer behaviour: From theory to policy. Energy Policy 129: 1070-1079.
Spangenberg, J.H., Polotzek, L. 2019. Like blending chalk and cheese – the impact of standard economics in IPCC scenarios. Real-World Economics Review 87: 196-211.
Spash, C.L. 2012. New foundations for ecological economics. Ecological Economics 77: 36-47.
From: pp.8-11 of WEA Commentaries 10(1), February 2020