Performative Mechanisms

Abstract

Performativity is grounded in a particular kind of social mechanisms that involve semiotic causation. I combine three different literatures: on performativity, on constitutive explanations in the social sciences, and on Peircian semiotics into a conceptual and methodological framework that can deal with the ubiquitous phenomena of contextualization and framing in experimental and behavioural economics. In the semiotic model, incentivesal ways operate via two channels of causation, the efficient-causal one and the semiotic one, with the latter operating via the cognitive system. For example, in social preferences, incentives efficient-causally impact on neuronal mechanisms that generate actions and on the cognitive processing of in-group/out-group boundaries. I present the example of corporate governance and managerial incentive systems as a case in point, arguing that these are performative mechanisms, as reflected in observations of historical and cultural embeddedness. This results in the impossibility of formulating universal regularities along the lines of the ‘legal origin’ literature.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

3.1 Towards Constitutive Explanations in Economics

This paper advances the idea that the current literature on performativity can be put on a stronger methodological footing if it is combined with the literature on mechanisms in the social sciences. I think that what authors such as Donald MacKenzie or Michel Callon actually did in their seminal contributions is presenting thick descriptions of performative mechanisms. Yet, what is missing is a general conceptual framework that allows to extend these thick descriptions into analytical approaches to causal explanations of the observed phenomena. This framework is provided by the methodology of constitutive explanations. In this paper, I merely sketch a few bare bones of this. The core task is to relate the notion of performativity to established bodies of research in economics. I think that the pivotal notion is that of incentives working on a given set of preferences in order to generate a certain behaviour, which underlies the different kinds of mechanisms about which economists propose generalizations that aspire to assuming the form of universal laws. Against this idea, I present the view that incentives and preferences are embedded into performative mechanisms that generate behaviour, which implies that universal laws cannot be formulated, as performative mechanisms are local and contextualized in essence. In particular, performative mechanisms endogenize the causal loops between incentives and preferences, thus suspending the analytical independence between the two, which is the core condition for the possibility of generalizations about the causal link to behaviour.

Most economists maintain the implicit methodological stance of the ‘covering law’ benchmark for scientific explanations. This mostly means that given certain ceteris paribus conditions, economics can identify regularities in economic phenomena that are universal. As a consequence, economists also believe that there are certain causal mechanisms with most general scope that can be implemented in policy interventions to achieve a certain goal. For example, if economists can show that private property rights have certain universal efficiency features, they would recommend privatization as a standard policy under most circumstances. In practical applications, this might require to make the ceteris paribus conditions explicit, which would result in a much more detailed analysis of conditions of application, but without affecting the theoretical and methodological core.¹

Consider as one example that I will further detail in Sect. 3.6 of this chapter, the ‘legal origin’ theories of corporate governance which claim that common law institutions are more efficient than civil law institutions in arranging for external finance. This claim is based on a conception of causality that underlies the design of pertinent econometric testing, suggesting recommendations for interventions that were rapidly picked up by organizations such as the World Bank. By this reasoning, a universal regularity was stated, and its application would follow the universal claim while judging certain conditions of application (such as the state of the court system). However, when this research was later scrutinized in more detail, it turned out that the causal parameters could not be defined in a de-contextualized way. For example, considering certain elements of corporate governance schemes, functional equivalences between apparently different elements in social practices were not identified, or the impact of extra-legal embedding determinants was overlooked. Once researchers try to catch these aspects in a more exact way, a principle of indeterminacy seems to hold: The causal relations in the econometrics vanish, and no universal regularity can be established anymore, resulting in an apparent trade-off between exactness and universality. The fundamental methodological issue that comes to the fore here is that of contextuality: Against the background of the standard econometric approaches, the different parameters and variables are mutually and externally contextualized, so that in the end, one can certainly maintain the idea that there is a causal interdependence between them, but has to acknowledge that this causality cannot be covered by generalizations that hold for a larger number of cases.

There are different possible reactions to problems like this one, which easily crop up in many specific areas of economic research. One is to maintain the covering law stance (and mostly refer to protective arguments such as the ceteris paribus clause), another is to search for alternative methodological frameworks. In this paper, I argue that this framework is the notion of constitutive explanations, or the analysis of ‘mechanisms’. The notion of mechanisms has made a rapid career in the social sciences recently, but has rarely been received in economics so far.² For sure, there is the notion of ‘mechanism design’, but this differs fundamentally from the approach of constitutive explanations. ‘Mechanisms’ in mechanism design are mathematical structures that identify certain rules of games that relate individual strategic choices with a social value function defined by the designer, aiming at achieving the social optimum while incentivizing all agents to reveal truthful information. These are equilibrium solutions that do not identify causal mechanisms in the real world. If these mathematical models are applied to design real-world institutions, they actually refer to what is a causal mechanism, yet without identifying this directly. In other words, the ‘mechanism’ is a mathematical structure that is projected on real-world mechanisms, but without firstly trying to identify those causal mechanisms by means of theory-driven empirical research.³

In comparison, the notion of mechanism in the social sciences has been received from the sciences, in particular, the neurosciences. Here, mechanisms are conceived as multi-level complex causal structures, with different levels being approached by different disciplines or disciplinary subfields. The covering law criterion has been questioned by many philosophers of science over the recent decades, not only for principled reasons but also for the empirical observation that beyond physics, most sciences do not meet this benchmark, at least in current practice. In our context, this is certainly true not only for the social sciences in general but also for another field that focuses on the explanation of human behaviour, the neurosciences.⁴ Both areas have got into close touch recently via the emergence of a new field in economics, namely neuroeconomics. This is part of a broader movement to introduce science-based methods into economics, behavioural and experimental economics. The question arises whether in such cross-disciplinary interactions, unification of methodological principles is necessary in order to achieve conceptual and empirical integration. Against this backdrop, the recent convergence of views about mechanisms in both the neurosciences and the social sciences is highly significant.

In this paper, beyond elaborating on the mechanism methodology in economics, I advance the additional thesis that in a general taxonomy of mechanisms, the specificity of the social sciences is that certain core mechanisms are ‘performative’. I distinguish between two categories of mechanisms, causal mechanisms in general and performative mechanisms as a subset. The general notion of mechanism already includes the defining feature that mechanisms are productive: This means, the coming together of parts and levels in a composite structure generates effects that are novel in the sense of new combinations of causes and effects. Therefore, the notion of mechanism also plays an important role in evolutionary theories which aim at understanding the emergence of novelties: A neuronal structure is a mechanism that produces novel phenomena in the physical world. Mechanisms in the social sciences are a special case because the productive function has the additional property of performativity. A most important class of social mechanisms are institutions in the broadest sense, and, following Searle, we can approach these as being observer-relative facts.⁵ A mechanism in the general sense is an observer-independent fact (neurons fire independently from observers watching them), whereas a social mechanism is an observer-relative fact in the sense that causes and effects are necessarily mediated by the cognitive or, more general, neuronal systems of the individuals whose interactions are part of the mechanism.

Stated in this way, it is important avoiding the conceptual short-cut that a performative mechanism is simply a more complex chain of causes and effects that includes physical phenomena of neuronal systems. This would define a reductionist position; constitutive explanations, however, are non-reductionist.⁶ Neuronal systems enable symbolic behaviour, and so mediation means semiotic causation: An observer-relative fact is a fact that comes into existence because the productive mechanism incorporates signs that are produced by neuronal systems and that establish channels for information transmission between neuronal systems that are themselves physically mediated not by neuronal mechanisms but by signs (such as body movements, utterances, and artefacts). Semiotic causation enables performativity of social mechanisms. The crucial phenomenon in establishing semiotic causation is interpretation: The effect is constituted by the interpretation of the receiver of the sign, and hence in principle independent from the original intention of the sender. This theoretical framework, I argue, allows to provide the analytical foundation for the phenomenon of contextuality that I introduce in the example of ‘legal origin’ theories.⁷

Subsequently, I continue with outlining the basic conceptual framework for analysing performativity in terms of semiotic causality. I show that this is reflected in the cross-disciplinary literature on the relationship between incentives and preferences, thus violating the standard assumption in economics that preferences and incentives are perfectly separable. I demonstrate that an incentive structure is performative, implying that there are no universal regularities that allow transplanting certain general models of incentive structures into different contexts while keeping the chain between causes and effects stable and uniform. Finally, I draw conclusions for the analysis of corporate governance mechanisms and the related incentive structures.

3.2 A Basic Conceptual Frame for Modelling Performative Mechanisms

In this section, I add more detail to the basic framework for analysing performative mechanisms. I start out from further clarifying the two elementary terms: ‘mechanism’ and ‘performativity’.

• A mechanism-based explanation is a constitutive one in the sense that explanations are based on the analytical and empirical identification of causal processes that are specific to time and space, hence do not result in universal regularities (‘covering laws’). Mechanisms operate under the constraint of universal laws, but for the explanation of the observed regularities, the universal laws are not sufficient: Under conditions of scientific and disciplinary specialization, therefore, particular disciplines such as the neurosciences focus on the identification of mechanisms as the primary epistemic goal. Mechanisms are complex as they include different levels of aggregation, and mostly are part of larger structures in relation to which the mechanisms are separated via boundaries; across these, further causal interactions occur which involve the inputs and the outputs of the mechanisms. the most important aspect of disciplinary methodological standards is how they delineate the criteria of acceptable mechanistic explanations and for causal relevance. Generally, this defines a naturalistic ontology, structured according to the criterion which entities are seen as having causal powers.⁸

• In the social sciences, some mechanisms have the special property that human cognitive systems are involved which enable intentional actions towards other individuals that are based on cognitive states through which interpretations of those actions are mediated. As a result, mechanisms include those cognitive states, and a full explanation requires the reconstruction of the specific ways how a mechanism emerges from cognitively mediated interactions between individuals. I call this ‘performativity’, and the causal process involved is semiotic or is ‘semiosis’. Hence, whereas in the sciences mechanisms are givens (such as the physical structure of neurons), in the social sciences mechanisms are part and parcel of a social ontology of observer-relative facts that is continuously being created and reproduced by the individuals involved in interactions. In a nutshell, chains between external causes as inputs and behaviour as outputs are always mediated via cognitive states, which are distributed across many individuals, and are thus also determinants of the external causes.⁹

As a consequence, a social mechanism always manifests what I call a ‘triadic’ pattern of causation, involving both physical interactions between individuals and semiotically mediated interactions which can be embodied in the same physical phenomenon, but need not be. For example, if an individual hands over a banknote to another individual, this is the cause of a behavioural effect, but the effect cannot be explained by the mere physical fact of moving the banknote in space by means of bodily movements. The banknote is a sign, and semiotic causation necessarily involves the physical movement; both causal modes are indispensable for producing the effect, namely, a particular action of the receiver of the banknote. At first glance, that would suggest that semiotic causation supervenes on the physical movement. But a brief reflection shows that this is not the case, unless one adopts a naive sender–receiver perspective on the relationship. In this case, one would assume that the sender has the intention to use the banknote as a sign, and that this meaning is transferred to the receiver together with the physical item. However, as economists well recognize for the case of money, this does not match with the way how banknotes actually adopt the role of a sign, as this is only constituted on the level of the collective of individuals who mutually recognize this sign. Hence, we cannot apply the simple framework of efficient causality here that underlies the explanation of the physical movement of the banknote, although that might appear to be reasonable as long as we look at the interaction in isolation.¹⁰

Social mechanisms have a triadic structure that can be conceptualized as in Fig. 3.1, which depicts a semiotically mediated performative mechanism_s in which a physical mechanism is a constituent part.¹¹ We consider a physical object O that is a cause in the mechanism that produces an effect. This is mediated via the physical structure of the mechanism_p M. In our example, this is the physical transfer of the banknote. I emphasize that in analysing social mechanisms, the explicit treatment of physical aspects is indispensable, such as arrangements of individuals in space, temporal sequences of actions, the shape and properties of artefacts, or technologies of interaction.¹² However, the ultimate effect of the mechanism_p is determined by the semiotic causation mediated via the sign S (mechanism_p is a necessary but not a sufficient cause of the effect). This effect is relative to the interpretant I which refers to the sign S. Thus, between O and I, a relationship is established that is conventionally called ‘meaning’. However, this is dependent on the embedding of the performative mechanism into a larger context, the social system, in which the effect in terms of the subsequent action has a function. Thus, semiotic causation establishes a conjunction of meaning and function. This is enabled by the role of the sign in categorizing the physical mechanism M. For example, the banknote has a particular value, and depending on the value, a physically similar movement will lead to different effects in terms of subsequent actions of the interpretant I. The assignment of value is context-dependent, more specifically established by the usages of the sign in a population of interacting individuals.

Fig. 3.1

Triadic causation in semiosis

I call the causal relationship between O and I ‘bimodal’, both physically and semiotically mediated. This turns O into an observer-relative fact, and is therefore, constitutive of the performativity of social mechanisms. It is important to emphasize that this does not depend on the individual intentionality of sign usages in social relationships: For example, the colour of the skin is a sign that operates independently from intentionally sending the sign. Further, the general category of ‘sign’ subdivides into different kinds, with different physical manifestations of the relationship between sign and object.¹³

I argue that performative mechanisms build on semiotic causation. Semiotic causation establishes performativity, which more specifically roots in the collective assignment of meaning to signs that trigger certain actions in a community of sign users. In this process, collectives of agents establish observer-relative facts, that is, social entities, and thereby, enrich the social ontology by creative acts. If we refer this view with the mechanism approach, we notice central elements, in particular, the identification of different levels (individual vs. collective) and objects (artefacts, embodied actions, individuals) and particular causal pathways that embed a mechanism into a larger unit (such as the embeddedness of the single action in a network of recurrent actions involving a collective). The triadic framework is just a most general conceptual structure guiding the more detailed analysis of performative mechanisms.¹⁴

3.3 Performing the Ultimatum Game: Which Way is the ‘Right’ One?

Analysing performative mechanisms as being based on semiosis has far-reaching methodological implications for understanding the relationship between external causes of actions and the actions that result from the causal impact. A covering law approach to incentives would assume that all human individuals are following the same principles of decision-making, so that a particular incentive structure would produce regular outcomes under ceteris paribus conditions. This is also the assumption that underlies the practical uses of ‘mechanism design’ theory: Then, results of game theory would be interpreted as stating universal laws across different actualizations of human behaviour. Interestingly, in practical applications of mechanism design theory, many additional activities are necessary that fine-tune the contexts and even the behavioural stances of actors in order to make sure that they ‘perform’ the mechanism in an appropriate way. This is what we have to expect against the background of the triadic model.¹⁵

In the triadic model, semiotic causation makes the role of the context explicit, thus rendering arguments obsolete that would refer such necessary adaptations of models to the real world as taking place in the ceteris paribus domain or during the necessary tinkering in turning theory practical. Some well-known examples are the simple experiments such as the ultimatum game: In approaching models in terms of the covering-law methodology, human individuals would be expected to manifest similar behaviour, independent from their actual contexts in everyday life, allowing for random variations. If this cannot be proven, explanations would have to consider the ceteris paribus conditions. These conditions include states of knowledge in the experiment, such as the beliefs of the individuals (hence cognitive states). However, these beliefs can differ from the conceptualization of the experiment by the researcher, resulting in different interpretations. A common phenomenon is that the individuals subsume the experimental situation under familiar types of interactions outside the experimental setting. This can explain systematic variations across different groups of individuals which share certain social contexts that result in these beliefs. Hence, the experiment cannot be fully controlled by the experimenter, in the sense that she would be the conductor who fully determines the way how the experiment is performed. The individuals perform the experiment autonomously, and consequently, we cannot identify universal regularities over different applications of a standard experimental setting. This performance is mediated via the semiotic causation that is driven by the interpretive acts of the test persons: The incentive structure causes their behaviour, but incentives are simultaneously signs.¹⁶

The game-theoretic structure underlying the experiment describes only one part of the real-world causal mechanism that links the incentives (pay-offs) with the results. The game-theoretic description of the performative mechanism is incomplete. Contrary to the expectations of the experimenter, the cognitive states are not fully described by the information that the participants obtain from the description of the game provided by the experimenter. The question is how far we can say that their state of knowledge is simply ‘false beliefs’: This may become evident if games are played recurrently, and individuals learn, so that they might finally converge to the ‘rational’ solution, hence perform the game according to the expectations of the experimenter. Yet, if this means de-contextualization in an artificial environment, we are not allowed to draw any conclusions about behaviour in a recontextualized real-life environment. Individuals simply perform the experiment differently. In the triadic framework, the experimenter achieves to impose another function on the interpretant I, which is to play the game properly; but this also changes the meaning of the sign, namely, the pay-offs.¹⁷

In the social sciences debate about mechanisms, this interdependence between cognitive states and outcomes of interactions is mostly referred to as ‘self-fulfilling prophecy’ in the sense of Merton’s.¹⁸ However, I argue that this is misleading, as the notion of self-fulfilling prophecy means that there is a belief that is initially false, but leads towards actions that changes the beliefs of others, resulting in further actions that ultimately confirm the original beliefs, thus rendering them truthful (like in the example of the bank-run triggered by wrong assessments of the financial status of the bank). If we call a mechanism ‘performative’, this is a much stronger proposition in the sense that the reference for the truth value of a belief is endogenous to the process even to the degree that it does not exist independently from the process under scrutiny. This reflects the creation of observer–relative facts via semiosis. The standard example for this is money, which does not exist before being performed collectively (to the opposite, the financial status of a bank in the bank-run does exist already).

Consider the case of the ultimatum game, again. Can we really say that playing the game ‘correctly’ turns originally false beliefs into true beliefs? This question touches upon a foundational issue in behavioural and experimental economics, namely the existence of social preferences in human individuals. Do subjects ‘unlearn’ false preferences in correct treatments of the experiment? Are these therefore ‘errors’? Are we justified in judging the learned preferences as the ‘true’ ones? I argue that these questions do not grasp the real meaning of the ultimatum experiment, which in all the realizations actually involves the working of performative mechanisms.

3.4 Incentives Performing Preferences

Interestingly, the issue of human sociality is raised in many different disciplinary contexts, under different labels. What is ‘social preferences’ in behavioural economics is ‘collectivism’ or ‘allocentrism’ in psychology, or ‘empathy’ in the neurosciences. In the neurosciences, interest is directed at the question whether there are universal human (neuro)biological features that condition certain kinds of social behaviour. If that could be proven, this certainly would have implications for the other disciplines. In psychology, the issue is partly seen as an aspect of individual psychological properties, but in the context of collectivism, mainly as a cultural feature. This differs from the neuroscience question about universal human properties in hypothesizing that there are cultural properties shared by individuals belonging to the same social group which are stable in the long run; some groups might manifest higher degrees of collectivism than others. At first sight, both the neurosciences and the psychological approach would suggest that there is a set of fixed properties that would also be reflected in certain ‘true beliefs’ about the degree of other-mindedness among a group of human individuals. However, this expectation has not materialized.

In the neurosciences, research on empathy has resulted in a complex mechanism-based explanation that combines bottom-up and top-down processes, hence multi-level causal feedback loops. In a nutshell, there are certain species-specific neuronal mechanisms that enable other-oriented cognitive states and respective behaviours, but at the same time, these are only triggered under certain conditions which depend on higher-level states of knowledge mediated by symbolic systems, hence semiotically caused, in my parlance. In particular, this refers to cognitive categorizations of other individuals into in-group and out-group members, with no ‘natural’ delineation of the group in question. Although this behavioural tendency towards ‘groupishness’ is presumably a universal feature of humans, this fact alone cannot explain group boundaries under specific circumstances. Hence, one cannot identify universal regularities of human social behaviour rooted in shared biological properties but only complex mechanisms of empathy that involve different ontological levels and kinds. This result can be easily put into the triadic framework: The mechanism_p is the neurophysiological structure that is triggered by certain sensory inputs and generates certain bodily reactions, but these reactions can only be fully explained via the mechanism_s, that is, the semiotic categorizations. Emphatic behaviour has a function in the larger context of social systems, and there is no universal mechanism, as these functions are specific to particular social systems located in space and time. In other words, empathy is a performative mechanism that includes neuronal structures as mediating physical entities, but does not simply supervene on these structures, being embodied in mechanisms that reach far beyond the brain, including social systems and the signs used therein.¹⁹

In psychology, suffice to mention one line of thought dubbed the ‘ecological approach to culture’. This rejects the idea that people carry along certain inherent cultural characteristics such as ‘values’ that define their degree of ‘collectivism’, but posits that in social interactions, individuals take actions that create a certain environment unintendedly which incentivizes their behaviour in a way such that a stable pattern of interaction is achieved reflecting certain regularities which are context-specific (the ‘niche’). This results in a conjunction between those incentives and cognitive states, that is only broken when fundamental parameters of the interaction are changed, especially involving the symbolic representations through which the interactions are mediated. As a result, what appears to be an internal ‘value’ explaining behaviour in terms of efficient causality (the value causes an action pattern), turns out to be a context-dependent regularity that is triggered by certain semiotic mechanisms. Once the experimenter achieves de-contextualization, the ‘value’ disappears.²⁰

Both strands of research on human sociality therefore result in a general hypothesis about performative mechanisms that underlie social behaviour: Incentives, cognitive states, and symbolic media work together in generating performative actions that establish degrees of sociality specific to time and place. Therefore we cannot state a universal regularity about the degree of ‘other-mindedness’ of individuals. The ‘mechanism’ that is stated in the game-theoretic structure of the ultimatum game cannot be extended to a ‘covering law’ that is empirically meaningful. This also implies that we certainly can de-contextualize the behaviour of individuals by appropriate experimental settings, thus apparently producing evidence on the validity of ‘rationality’ as a universal human characteristic. However, this is only another instance of performativity, imposed by the experimenter.

Fortunately, a related view has been also articulated in economics recently, mostly under the heading of ‘framing’. There is a large and growing literature showing that individuals do not have a fixed endowment with preferences that manifest a certain degree of ‘sociality’, but that ‘social preferences’ are endogenous to the context of a certain situation of choice, and in particular how choice is framed. In their comprehensive review of this literature, Bowles and Polanía-Reyes (2012) employ a general structure which directly matches with my model of triadic semiotic causation (Fig. 3.2). They argue that the standard economic view posits the ‘separability’ thesis: Incentives work independently from the given level of social preferences in generating behavioural outcomes. To the contrary, there is ample evidence that incentives also change the level of social preferences, so that there can be both crowding in and crowding out in terms of the resulting behaviour. This results in a two-channel or, in my parlance, ‘bimodal’ model of causation that connects incentives and actions.

Fig. 3.2

Incentives and social preferences (following Bowles and Polanía-Reyes 2012)

Bowles and Polanía-Reyes categorize the precise effects into four groups, always considering the introduction or the strengthening of a material incentive to perform a certain action.²¹

• The incentive changes the framing of an action such that self-interested motivation becomes salient and is seen as legitimate. This even applies for behaviour that does not involve social preferences at all: For example, introducing a fine on a certain behaviour can increase its frequency because the fine is interpreted as a price that is paid for allowing deviance.

• The incentive is interpreted as an information about the relatively low share of individuals with social preferences in a population, and increases uncertainty about the motivation of observed cooperative behaviours, that is, whether this behaviour is signalling social preferences. Therefore, the incentive decreases individual motivation to act socially.

• The incentive signals information of the designer about her assessment of the distribution of types in a population, and also gives information about how she perceives the nature of the action to be performed.

• The incentive reduces the sense of individual autonomy and therefore triggers resistance.

Resulting from these four constellations, there is no universal regularity between certain incentive structures and observed behaviour. We notice that economic research concurs with the aforementioned research in other disciplines, and suggest the interpretation that there is no fixed pattern of sociality in human individuals, but that sociality is based on performative mechanisms that result from the interaction between the environment of choice and distributed cognitive states in the group of interacting individuals, mediated by semiotic causation. Thus, incentives are always being interpreted, and interpretation is always contextualized. Scientific analysis, however, has to move beyond this general statement and needs to make the performative mechanisms explicit that result in particular causal chains between incentives and actions.

3.5 Semiotic Causation and Performativity of Incentive Systems

The economic problem of sociality raises an intricate question about the nature of incentives as signs. This is straightforward to grasp in the triadic model: The incentive as object O is at the same time the material embodiment of the sign S. Hence, there is a direct function of the incentive in creating the performative mechanism that produces the ‘performance’ of the incentivized individual. Bimodal causality works embodied in one single physical form. This is not a necessary feature of incentive systems but the most interesting case for our discussion.²²

In terms of the economic conceptualization of preferences and choices, bimodal causality means that the incentive exerts causal effects via two channels. One channel is the direct effect on choice mediated by the given structure of preferences. The other channel works via the circumstance that this structure of preferences is also shaped by the effects of the incentive on cognitive states. This implies that preferences are not independent from the incentives. The following action is caused bimodally, ending up with certain consequences (pay-offs). These consequences work as incentives, in turn, thus resulting in feedback mechanisms. We end up with a recursive structure as pictured in Fig. 3.3. Here, I add the role of the designer of the system, because the interpretation of incentives certainly is changed fundamentally when interpretation also involves reference to the supposed intentions of the designer to create the system. Yet, this remains a simplified picture because I do not include the obvious mediating role of the designer’s cognitive states in the perception of consequences and in her response.

Fig. 3.3

Bimodal causation in incentive systems

How do the incentives influence the structure of preferences? Building on the previous summary of Bowles and Polanía-Reyes, I add some detail and emphasize the following.

• Under conditions of incomplete and imperfect information, all individuals are aware of the fact that their action consequences will depend on the behaviour of others; therefore, they need to construct beliefs about the preferences of others and their cognitive states. Then, the incentive system provides information about the composition of types of individuals in a population. This can directly involve reference to the designer of the system, because individuals may reasonably conclude that the designer of the system has designed it according to her information about that distribution. For example, as an employer, she may have even selected individuals according to those criteria. If the incentive system is geared towards individualistic behaviours, individuals may adapt their preferences to this information.²³

• The qualitative nature of incentives influences the way how situations of choice are perceived, especially in terms of interactions with others. In particular, it is not warranted that monetary values are just approximations to underlying preferences. Monetary incentives change frames of valuations and even induce shifts across categories of valuations. For example, monetary expressions of value induce behaviour that is more individualistic, and they shift the reference frame towards an exchange context focusing on notions of reciprocity, in comparison with the direct benefits of action consequences.²⁴

• Most generally, any kind of incentive system shifts the balance between intrinsic and extrinsic motivation. This effect is the more pronounced if the incentive system is cast in quantitative terms and can be anticipated in terms of the action consequences. If we consider the previous point, the effects of monetary incentives are actually twofold: one is the effect of the quantification and targeting of the incentive, the other is its explicit casting into monetary terms.²⁵

The result of this analysis is unequivocal. Incentive systems are social mechanisms, and therefore, are performative. As performative mechanisms, they involve distinct elements, such as the roles of designer and actor, the physical entities that represent the incentive system (such as operating procedures), or the entities in which the incentives are embedded and embodied (such as numbers in bank accounts). These elements are causally connected in a bimodal way, mediated via semiosis, that is, interpretations on part of the actors. Finally, these mechanisms are embedded into the context of the social system, such as the organization that implements the system. There is no way to predict the effects of certain incentives based on a universal regularity of causation. Causation is always semiotically mediated, and the same incentives can lead to very different behavioural responses, depending on the way how they are interpreted.

Incentive systems are important elements in the real-world design of economic institutions and organizations. The upshot of my analysis is that one cannot directly transfer results of generic economic models of incentive systems into particular social contexts. A real-world incentive system is a performative mechanism that extends beyond the part of the social system that is subject to the direct intervention of the designer of the system. In practice, this means that the designer would have to contextualize the incentive system. Naturally, this raises the question how to identify the relevant contexts. A case in point is the aforementioned issue of corporate governance institutions.

3.6 Implications for the Analysis and the Design of Corporate Governance Mechanisms

The case of corporate governance is of interest here, because it involves different forms of performativity. In the recent decades, corporate governance schemes have diffused across the globe that build on modern economic theory, such as principal-agent theory, transaction cost economics, and incomplete contract modelling. The problem driving these developments is the division between ownership and management in the modern public corporation: How can managerial behaviour be aligned with shareholders’, that is, owners’ goals? The theoretically grounded corporate governance schemes are themselves complex mechanisms, involving institutional regulations of the interaction between different groups and bodies of the corporation, and including a range of high-powered incentive systems for top-level management. These schemes originated in the Anglo-Saxon world, and are being promoted as best practices grounded in theory until today, with prominent cases as the ongoing reforms of corporate governance in Japan. One conspicuous feature of this diffusion is that on the one hand, the corporate governance schemes are seen as particular mechanisms that can be copied in other societal, cultural, and political contexts, and that on the other hand, these contexts are often blamed for dysfunctional performances of corporate governance. At the same time, however, there is no general agreement about the idea that the corporate governance schemes being promoted are factually the most efficacious ones. One issue highlighted in the ‘post-Piketty’ world is the possible impact on rising inequality and a disconnection between managerial compensation and actual performance.²⁶

Compared to the real-world use of mechanisms derived from generic reasoning in economics, according to the results of the previous sections, a mechanism in the domain of the economy would be conceptualized as a causal structure that involves a complex social ontology, namely constituent units and levels of the underlying performative mechanisms. For example, one would consider individuals, groups, or institutions as parts of the mechanism of corporate governance, as well as different specific mechanisms that establish their interactions, in particular focusing on semiotic causation. We would no longer approach a corporate governance mechanism based on de-contextualized economic models. Instead, we would reconstruct empirically the social ontologies that embed these mechanisms and show how they result in a certain pattern of ‘performing performances’, in the sense of performatively producing certain outcomes of actions.

In the first section, I have already referred to the ‘legal origin’ theories of corporate governance which are of special interest here as they start out from empirical research about determinants of external finance, and ground the interpretation of the results on economic theory.²⁷ This research posits a causal linkage between a set of variables that affect performance in terms of making external finance available at lowest costs and largest scope. At the same time, the argument operates on a higher level of aggregation in claiming that entire legal systems historically determined the emergence of these differences in performance, such as civil law versus common law. This is a multi-level analysis which was then tested statistically without making the underlying mechanisms explicit. When this research was put under closer scrutiny, serious empirical flaws became apparent. These flaws can be interpreted in terms of inaccurate and biased conceptualization, identification, and measurement of social mechanisms, in particular²⁸:

• The list of potentially relevant elements of corporate governance mechanisms was incomplete;

• Functional equivalences of included elements were not properly identified; vice versa, different functions of similar elements were not recognized;

• Measurement criteria were implicitly referring to US conditions as benchmark;

• The boundaries of the mechanisms were not properly identified and the role of extra-legal factors in determining performance was not recognized.

If researchers improve the empirical approach, they reach the conclusion that, on the one hand, it is possible to explain improvements of corporate governance and their effects for single countries through historical times, but that at the same time, statistical regularities across countries are difficult to establish. This clearly indicates that the research factually resulted in the identification of mechanisms that are specific to time and place, but cannot achieve generalizations over time and place.

I take this result as an indication that corporate governance schemes are performative: The ‘legal origins’ theory failed to take account of the contextualization of elements of corporate governance. This compares with the development of mainstream theory of corporate governance that converged on a shareholder-value paradigm in the recent decades. Against the background of the aforementioned problem structure resulting from the separation of ownership and management in the public corporation, the proposed mechanism mainly focuses on directly activating the investor’s valuations of companies on the capital markets in incentivizing the executive managers’ behaviour via specific compensation schemes. This transition was part and parcel of a broader trend of ‘financialization’ of capitalist economies.²⁹

This process has met a lot of criticism, both scholarly and public. As in the case of the ‘legal origins’ theories, which are closely connected to it, the idea is that one single mechanism is the most efficient one, and that the entire institutional set-up has to be geared towards this goal. However, if corporate governance mechanisms are performative, this assumption is not warranted. This point can be made in two different, though closely related ways. The first is to consider the interaction between corporate governance institutions and the economic system in general, the other is to scrutinize specific incentive systems which are parts of the corporate governance arrangements.

Without referring to performativity explicitly, Aoki’s model of the cognitive division of labour is actually presenting corporate governance as a performative social mechanism.³⁰ Aoki argues that there are different ways how the knowledge is distributed in a company that determines its competitive success. In my framework, competitive success is a function, and the governance mechanism is designed to mediate between certain incentives, behaviours, and this function. Aoki uses the term ‘cognitive assets’ and adopts the perspective of distributed cognition theorists: Cognitive assets are not strictly individualized but are distributed across individuals and technological artefacts.³¹ This has implications for the distribution of decision rights. For example, if cognitive assets are mostly individualized on part of the managers, workers’ participation would introduce inefficiencies in incentivization, and vice versa, if workers’ cognitive assets are highly complementary to managers’ cognitive assets, a participatory corporate governance mechanism would be more efficient.

However, these interdependencies cannot be simply cast into a universal regularity in turn, because they are performative. Performativity emerges on different levels, thus revealing a complex structure of performative mechanisms. First, within the company, the distribution of cognitive assets is endogenous to the corporate governance scheme. For example, if workers are excluded from decision-making procedures in the company, they are lacking incentives for adopting skills and knowledge that are highly complementary to managers’ cognitive assets, and vice versa. Second, the distribution of cognitive assets is partly determined by other institutions in the economy, such as education and training and the labour market structures in general. The more portable skills are across companies, the lower are the incentives for forming company-specific or complementary skills. As a result of these and similar effects, similar levels of corporate performance can be achieved by different mechanisms of corporate governance. However, this is not simply a self-fulfilling mechanism, because performativity depends on certain determinants that remain givens, in particular technology. Even though technology can also be performative, the question is how far the temporal and spatial contiguities play together in enabling performativity. For example, the Silicon Valley Hi-Tech model typically even enables the outsourcing of the entire production process, because there are very low complementarities of cognitive assets between workers and high-skilled managers and engineers.³²

Evidently, it is not possible to define a ‘one size fits all’ corporate governance scheme only taking theoretical analyses of the generic problems of principal–agent relations, asymmetric information, and so on into consideration. On the systemic level, corporate governance schemes become performative via the endogenous adaptation of the ‘cognitive division of labour’ in the economic system, thus changing the perception of the incentives that emanate from a certain corporate governance structure.

The other aspect to be discussed here is the incentive systems in the narrow sense, such as the stock option schemes. Here, we can directly apply the results of the previous section. Interestingly, it has been argued that the transition to those systems was also bolstered by management education, thus also establishing a possible case for the performativity of economic theory in conjunction with incentive structures. This argument runs in the following way. If education of managers is based on the mainstream theories about corporate governance and incentive schemes, as explicitly done in textbooks of managerial economics, students actually learn about the distribution of types at least in the environment of companies. So they will adopt the respective patterns of individualistic preferences. This confirms the expectations of the theorists and the designers of incentive systems, and therefore even empirically vindicates the underlying theories. This mechanism can be also supported by separating equilibria in sorting individuals with different levels of social preferences into different environments.³³

Further, we can apply the entire range of particular hypotheses about the performativity of incentive systems here, such as regarding the trade-off between intrinsic and extrinsic motivation when extrinsic incentives are quantified and announced in advance, or on the priming effects of money on behaviour. In the light of these hypotheses, one can argue that the introduction of high-powered incentive systems that directly aim at individual behaviour will also change the regularities that are assumed to hold for this behaviour. Indeed, one would expect that the behavioural patterns become more similar to the assumptions of opportunism and individualist rationality.

If we take together these two perspectives on corporate governance schemes, the institutional and the incentive structure, we realize that such schemes are in fact complex performative mechanisms with a high degree of contextualization. One important consequence is that similar incentive systems can operate in a different way at different times and places. This applies also on different levels of analysis: A corporate governance scheme can operate under the contextualization of single companies, working for one case and failing in another, or can be contextualized on higher levels, such as referring to national-level institutions and culture. For example, the Japanese system was working well after World War II into the 1980s. Since then, strong pressures emerged reforming the system, theoretically conceived as convergence to the Anglo-Saxon model. However, the process is slow and protracted, thus reflecting the complexity of the causal determinants. New mechanisms will also show idiosyncratic features, though of a different kind.

3.7 Conclusion

In this paper, I propose that the literature on performativity should be combined with the literature on social mechanisms in order to create a powerful approach to understanding and explaining the variance of human behaviour across different institutional contexts. This requires a fundamental shift in the methodological conceptualization of the economist’s work, namely from a ‘covering law’ view to a ‘constitutive explanations’ view.

In more detailed work, constitutive explanations require the precise identification of levels of social ontology, kinds of particular mechanisms and types of social entities that are involved in a constitutive explanation. I have provided a few hints with the example of corporate governance schemes. One important consequence of this is that for the analysis of mechanisms in the economy, economics as it stands is not sufficient. There is a huge explanatory gap between economic theories and the derived models and the mechanisms that work in the real world. As in the reference case for constitutive explanations, the neurosciences, the analysis of economic mechanisms is multi-disciplinary, involving the entire range of the social sciences, and also disciplines such as psychology or biology. Therefore, the mechanism approach is also providing a framework of disciplinary integration. However, it is important to get the direction of the underlying theoretical effort right here: The theoretical achievement in terms of providing explanations of real-world phenomena is the identification and substantiation of a mechanism. It is the mechanism that defines the patterns of cross-disciplinary integration, and hence these patterns can differ across different mechanisms. We cannot achieve cross-disciplinary integration in directly linking theoretical premises and results of the different disciplines.

This is especially important when we consider the phenomenon of performativity. As has been amply demonstrated by experimental economics, human behaviour can be shaped by the proper establishment of mechanisms that trigger certain performances. As such, the experiments do not test given theoretical hypotheses, but are actually implicit instructions how to design real-world institutions in order to generate similar behavioural results. If we create incentive systems based on the assumption of rational opportunism, we also create the agents that behave according to these predictions. By implication, the patterns of cross-disciplinary linkages are also endogenous to the application of the theories in the real world, again making any vision of a particular pattern of cross-disciplinary integration obsolete.

In other words, theorizing about human behaviour and social systems means to work on inventories of performative mechanisms, possibly resulting in taxonomies, historical and evolutionary trajectories. This is the true sense in which the economist becomes a naturalist. The first task of the economist-as-naturalist is to grasp the complexity and diversity of human behaviour in institutionalized contexts, and only then to work out regularities that apply across them. These regularities may in turn be rooted in universal laws, which apply for certain aspects and elements of the mechanisms that define the social ontology of the economy. In spite of being universal, however, they can only offer partial explanations of the causal processes that are mediated by complex performative mechanisms.