The ontology of complexity and the neo-Schumpeterian evolutionary theory of economic change

Verónica Robert and Gabriel Yoguel[1]

Introduction

The corpus of evolutionary neo-Schumpeterian is neither unique nor completely integrated. Instead, it is made up of several heterogeneous contributions that have appeared over the last 30 years. Recently, different sets of contributions to neo-Schumpeterian evolutionism have come closer to the complexity approach (Silverberg, Dosi, and Orsenigo, 1988; Dosi and Kaniovski, 1994; Dosi and Nelson, 1994; Witt, 1997; Arthur, Durlauf, and Lane 1997; Arthur, 1989; 1990, Metcalfe, Foster, and Ramlogan, 2005; Saviotti and Pyka, 2004; Foster, 2005; Durlauf, 2005; Frenken, 2005; Antonelli, 2007; Arthur, 2009; Antonelli, 2011). This approach allows the micro, meso, and macro levels of analysis to be integrated. Also, it can account for micro-heterogeneity, self-organization, path dependence, non-linear interactions, and divergent paths. In particular, the complexity approach represents a break with both reductionist and holistic explanations. Emergent properties are the result of micro-meso-macro interactions. This group of characteristics defines a set of ontological suppositions about complexity that are related to the ontological suppositions about current neo-Schumpeterian evolutionism.

Several elements of the ontology of complexity can also be found in different authors from the history of economic thought that are frequently quoted in evolutionary economics, like Smith, Marshall, Schumpeter, Hayek, Kaldor, Young, and Hirschman. In this regard, the precedents for complexity theory in economic thought may follow two possible paths: the first is mainly concerned with order and the transformation problem associated with it. This path, as pointed out by Metcalfe (2010), starts with Smith and ends with Hayek. The second path is mainly concerned with development, cumulative causations, and divergence. This path starts also with Smith but ends with structuralism and the development school (Robert and Yoguel, 2011).

The two paths are also precedents for the evolutionary contributions that form the basis for very divergent interpretations as two forms of intervention in economy, a situation which is reflected by the diversity of policy recommendations deriving from said interpretations. These include, on the one hand, bottom-up policies such as capacity building and the development of institutions that promote innovation and economic development as emergent properties. On the other hand, the recommendations also include top-down policies such as the selection of specialization sectors and fostering structural change. These two forms of conceiving public intervention have often been put forward at the same time by evolutionist authors, despite their different conceptual roots.

In this paper, we argue that the complexity approach could be a theoretical and conceptual starting point that would allow the integration of different contributions from neo-Schumpeterian evolutionism and the economic policy recommendations deriving from it.

This article has three interconnected aims. First, to compare different evolutionist streams of thought taken into account the various dimensions of the ontology of complexity that they support, be it explicitly or implicitly. Second, to evaluate how close each group is to the two historical traditions of economic thought mentioned above. Third, to discuss the political actions that derive from each of these two historical traditions and the way in which evolutionist contributions articulate them to different degrees. This articulation will depend on how far they adhere to the different dimensions of complexity: the more points they have in common, the easier it will be to derive policy recommendations that articulate both bottom-up and top-down processes.

This paper is organized into three main sets of building blocks. In Section One, on the basis of different transdisciplinary definitions of complexity, we present the ontological assumptions of complexity. As such, the first set of building blocks is made up of the five dimensions of complexity that we identify. In Section Two, we identify two traditions in the history of economic thought into which a set of complexity-related ideas can be read. These ideas constitute the second set of building blocks. In Section Three, we propose a taxonomy of the contributions that make up neo-Schumpeterian evolutionary economics. The taxonomy, based on the main concerns addressed by each group, is the third set of building blocks. In Section Four, we mix the three sets of blocks up, showing the different starting points for each of the groups’ contributions, how they are related to the abovementioned traditions in economic history, and how the various dimensions of complexity are emphasized differently. Finally, in Section Five, we present our conclusions and some policy implications.

1.Towards anontology of complexity

It is difficultto say that complexityis a theory; but it is, in any case, an approach that includes a set of conceptual fundamentals and methodological tools. As an approach, it is also associated with theories like dissipative systems and networks, with specific application areas. Prigogine and Stengers (1984) suggest that the complexity approach is essentially a new relationship between science and nature that comes in response to a new view of the world, characterized not only by its unpredictability but by the impossibility of addressing its structure and dynamics throughgeneral and immutable laws. Since then, complexity has been associated with self-organization, out-of-equilibrium dynamics, irreversibility, and indeterminacy, and the notions of irreversibility and uncertainty have come to predominate over stability and equilibrium in systems dynamics.

However, Prigogine’s position has been strongly criticized by Bricmont (1996) arguing that complexity is not in contradiction with Newtonian thinking. This illustrates how far complexity has been a subject of debate even within the hard sciences.

The development of complexity as a concept has led tothe descriptionof a set of deep features relating to the functioning of complex systems. These features can account for many different situations. In this sense, a definition of complexity should aim to cover this diversity but also should describe the main features of the system itself (components, rules, etc.). Its application to social phenomena has given rise to a variety of contributions showing how broad the reach of complexity can be.

According to Rosser, a broad definition of complexity may be attained by discarding what is undoubtedly not complex, i.e.“those systems that do not generate endogenously or deterministically well behaved dynamics.” Nevertheless, the vagueness of the definition goes against its practicality. Rosser also quotes Day’sdefinition (1994), which indicates that a system is complex if it tends, endogenously and asymptotically, to something different than a fixed point, a limited cycle, or an explosion. In this case, the definition ismore precise but it is circumscribed to anevaluation of the system in terms ofits results, not ofits composition and characteristics. Besides, these types of outcome are not always complex dynamic results. Kwapieńa and Drozdz (2012) provide a definitionthat aims to describethe system through its components and not its behaviour or outcome. According tothese authors,a complex system is comprised of alarge number of components which interact in a nonlinear way, exhibiting collective behavior and beingeasily able to modify their internal structure and/or activity patterns from data or energy exchanges with the environment. While these approaches are not stringent enough to enable one to decide whether a system is complex, they are useful for identifying when a system is clearly non-complex.

Beyond these few-word-definitions of complexity, and the epistemological explanations of Prigogine and Stengers, efforts to characterize complexity have tended to list the set of characteristics that a system must have in order to be called complex. For example, the definition proposed by Nekola and Brown identifies several characteristics that a complex system should present: (i) micro-heterogeneity, (ii) interactions between system components and with environment in many different ways and on multiple spatial and temporal scales, (iii) complex structures and nonlinear dynamics, (iv) dynamics neither completely stochastic nor entirely deterministic, but instead a combination of randomness and order, (v) positive and negative feedback mechanisms, causing either amplification or damping of temporal and spatial variation. According to these authors, these are open systems that exchange matter and energywith the environment to reachan organized state far from equilibrium. They are historically contingent, so that their configurations reflect the influence of initial conditions and subsequent perturbations. Often, these systems are nested within others, giving rise to hierarchical organizations.This kind of definition is predominant in economics and particularly in evolutionary economics. Over the following paragraphswe present definitions fromArthur, Durlauf, and Lane (1997), Metcalfe and Foster (2004), Kirman (2010), Martin and Sunley (2007), and Dopfer and Potts (2004).

First, Arthur, Durlauf, and Lane (1997b) cite six key features of complexity applied to economic systems: i) heterogeneous agents interact with each other within a specific local environment in a given space; ii) the absence of a global controller that can exploit all the opportunities or interactions of the economy, although there may be weak global interactions, iii) a hierarchical organization with many intersectoral interactions, iv) continuous adaptation through learning and evolutionary agents; v) continuous innovation, new markets, technologies, behaviors, and institutions that create new niches within the system, and vi) non-equilibrium dynamics with either no equilibrium states or multiple ones, which are unlikely to reach a global optimum.

Second, Metcalfe and Foster (2004) and Dopfer and Potts (2004), from an evolutionary perspective, consider that: (i) a complex systemis a network structure made up of components and linkages, (ii) its linkages allow information and knowledge to circulate, (iii) its structure is modular, (iv) it is open to new components and new interactions between them, and (v) it has a hierarchy since each component is also a complex system. From the evolutionary geography perspective,Martin and Sunley (2007) stress the following features: i) the distributed nature of the systems,iii) the several scales of analysis, iii) the systems are open to novelty that comes from environment, iv) there are non-linear dynamics and feedbacks between system parts, v) a complex system cannot be decomposable or its decomposability is limited, vi) it may exhibit emergence and self-organization, vii) the components are adaptive and they interact with the environment, and viii) the system is not deterministic or traceable.

Third, Kirman (2010) and Helbing and Kirman (2013) give a definition of complexity that emphasizes the relevance of interactions between system components. According to these authors: (i) The connections between heterogeneous components are incomplete and chosen by the components according todifferent criteria (cost, benefits, capabilities, history, etc.). Each component has a limited number of connections to others in the neighborhood. (ii) The system’s behavior cannot be understood from the properties of its components, but rather from the interactions between them. In this sense, the interaction processes and coordination of the network structure are more important to explaining the aggregate results than individual behavior. (iii) The system may exhibit emergent properties like network structure, the heterogeneity of components, and the rules that guide the components’ behavior. Therefore, system behavior is often counter-intuitive. (iv) Feedback and unexpected side effects are common. The system may feature cascade effects and extreme events. The probability of extreme events is higher than expected according to a normal (Gaussian) distribution, and their impact may take on almost any size (in particular, it may be global in scale). Since the interactions are constrained by the neighborhood, extreme events do not affect all components simultaneously, but the scope and velocity of these events depend on the system’s network structure. (v) System behavior is hard to control in a centralized or top-down fashion. The components will often fail to behave the way they wishor as they should, because they cannot act independently. Therefore, the system cannot be strictly optimized in realtime, even with the biggest supercomputers. (vi) The system may spend long periods of time far from equilibrium, even when an equilibrium existsin principle. The system may have multiple equilibria, but these equilibria may be unstable.

It is interesting to note that when the definitions of systems are based on a list of features, it is not clear in general whether a system is complex if it satisfies one, several, or all of the features listed, therefore such definitions entail a strong underlying ambiguity. Moreover, many of these features are associated with oneother or are mutually implied. At the same time, although there is some overlapping among the listed characteristics, the coincidence is not perfect.

Combining all the above definitions, we propose the following five dimensionstosynthesize the fifteen elements which are present in the different definitions of complexity: i) micro-meso-macro heterogeneity, ii) disequilibrium and divergence,iii) interactions and partial information, iv) network architecture, and v) emergent properties (see Table 1)

Heterogeneity is related to the ability of the system’s components to generate variety, adapt, and evolve. This feature, in turn, is combined on the one hand with the possibility of generating novelty endogenously—creativity—and on the other by selecting the relevant attributes on the basis of interaction with the environment, the learning process, and capability building. These features make complex systemsadaptive. Heterogeneity manifests itself at different levels of analysis: firms, local or sectoral systems,and national economies.

The second dimensionis that of disequilibrium. There is order to complex systems, but this does not imply that they are in a state of equilibrium. Feedback processes exist between the components of the system and between them and the environment. This explains why such systems show far-from-equilibrium dynamics. In this context, the system is indeterminate. Thesystem dynamics are associated withits initial conditionsand itsown history(path dependency), whichcan result in the dynamics leading to divergent paths and lock-in situations. Therefore, because of the absence of a global controller, there is no guarantee of reaching a global optimum.

Table 1. Five dimensions of the ontology of complexity
Ontological assumptions of complexity
I. Heterogeneity / 1. Evolutionary heterogeneous agents, with creative capacity
2. Learning and adaptation
3. Heterogeneity of systems (meso-macro)
II. Interactions, non-linear dynamics, disequilibrium and divergence. / 4. Positive feedbacks
5. Out-of-equilibrium dynamics
6. Indeterminacy and uncertainty
7. Non-ergodic path dependence
III. Connections and information / 8. Linkages more relevant than components
9. No global controller. Partial and local information
10. Lack of global optima
IV. Network Architecture / 11. Hierarchical organization
12. Decomposable modular structure
V. Emergent properties / 13. Multiscale analysis
14. Novelty
15. Micro variability consistent with macro stability

Source: Authors’ elaboration based on the cited works.

The interactions between components of a complex social system are intentional and are located in a multidimensional space. This means that the components can change their location and their specific links from moving along different dimensions of space. This assumption is related to various issues associated with the characteristics of the information. On the one hand, the information is local and therefore partial, however, the overall system can process information based on its distributed operation. While global interactions are possible (each component simultaneously exchanges information with the rest of the system’s components), they will tend to be weaker than local interactions (each component exchanges information with neighboring components in the multidimensional space with which it is linked). In this regard, the prevailing partial information prevents the existence of a global controller. Interactions are crucial, and their characteristics are more relevant to the global system dynamics than the characteristics of the components by themselves.

The fourth dimension is associated with the type of architecture of the network of interactions that complex systems present. In this regard, there are two key attributes. On the one hand, the presence of hierarchy in the sense put forward by Simon (1969), according to which a complex system consists of other subsystems that are also complex. On the otherhand, the presence of modular structures explains why interactions within subsystems are denser than interactions with each other. The modular system is resilient: it is able to absorb exogenous shocks and remain functional.

Finally, emergent properties are the result of multiple interactions on different scales of analysis. The fact that complex systems present various scales of space and time means that the results of each scale cannot be derived linearly from lower scales, each of whichshow specific attributes in each case. The macroscopic regularities which support small-scale variability is itself an emergent property of the system.

2.Two paths of complexity in economic history

Different conceptual elements of the complexity approach that have currently been adopted by several economists can also be read in the work of different authors throughout the history of economic thought. Actually, the adoption of the complexity approach by evolutionary economics is grounded in the fact that the contributions of its predecessors are consistent with many of the ontological assumptions of complexity discussed in Section One. In this section we will show two possible paths for the history of economic thought in whichdifferent aspects of the five dimensions of the ontology of complexity can be recognized. The first path, identified by Metcalfe (2010), is focused on coordination problems and thelinks between this and economic change. Metcalfe thereforerefers to this path as one of self-organization and self-transformation problems. However, there is a second, alternative path that can be identified behind the concepts of feedback and divergence, and which is much more related to accumulation and transformation problems.

According to Metcalfe (2010), there are numerous predecessors for the ideas of complexity in economics. In this direction, he traces a path from Smith to Hayek, including Marshall, Schumpeter, and Knight, in whose work a connection between interdependenceand ordercan be found. Metcalfe proposes that these authors’ ideas are of great importance, particularly those related to our understanding of the division of labor and the role of innovation to stimulate the processes of coordination and self-transformation. Therefore, the economic system is in disequilibrium,which is generated by innovation—economic growth reflects the growth of human knowledge—and therefore order became a better concept than equilibrium for coping with the problem of coordination (Metcalfe, 2010:46).