3. Performative Mechanisms

On the central role of ceteris paribus assumptions in the covering law approach to economic hypotheses, see Hausman (1992: 131ff, 2013: 14ff). Hausman argues that these clauses enable economics to maintain ‘inexact laws’. Of course, the precondition is that there is a precise and reliable method to distinguish between acceptable and unacceptable c.p. clauses (on the complex questions here, see Reutlinger et al. (2015)).

For a comprehensive survey of this literature, see Hedström and Ylikoski (2010) or the volume edited by Demeulenaere (2011). This is mostly pursued under the heading of ‘analytical sociology’, but contributing strands of thought are broader in scope, including seminal works such as Elster (1989). Philosophically, an important pacesetter was Bhaskar (1989), although this example also shows how the reception in economics was ending in a heterodox cul-de-sac that left no impact on mainstream economics (so-called ‘realism’ à la Tony Lawson). For a rare reception of the mechanism methodology in economics, see Vromen’s (2011) analysis of routines as multi-level mechanisms.

For example, an auction is a ‘mechanism’ in mechanism design theory with certain optimality features. In the real world, auctions often do not work as designers imagine. For example (see The Economist, August 29, 2015: 60), on the eBay website, the share of auctions has been declining continuously, partly because of the ‘hassle costs’ of auctions. One cause of these costs is ‘sniping’, when users wait until the last minute in order to submit a winning bid, thus disappointing other bidders. The experience of sniping drives many users away from auctions. Systematic research on empirical issues of mechanism design models include, for example, the experimental testing of the solutions of the hold-up theorem provided in the theory of incomplete contracts (Maskin and Tirole 1999; Maskin 2002), which were widely regarded to resolve this problem; Hart (2009) therefore declared it as obsolete for theories about contracting and the firm. Yet, this strong conclusion was only tested recently by experimentalists (e.g. Fehr et al. 2014; Erlei and Roß 2014). Erlei and Roß, for example, show that the sheer complexity of the theoretical mechanisms may give a role to bounded rationality in determining the experimental subjects’ choices, which systematically and strongly diverge from the theoretical predictions.

Neuroscience research has been the most important object of studying mechanistic explanations in philosophy of science, with path-breaking contributions such as Craver (2007).

Searle (1995) distinguishes between observer-independent and observer-relative facts, thus assigning the ontological status of existence to both. For example, a tree is an observer-independent fact, a holy tree is an observer-relative fact. The property of ‘holiness’ can cause a change of state on part of the observer (such as fear), and thus exists in terms of having causal powers.

Craver (2007: 107ff, 233ff) distinguishes between the reductionist and the systems tradition in neuroscience, showing that in spite of the fact that many neuroscientists pursue a reductionist agenda, the field advances in developing multi-level integrative theories about complex mechanisms that produce a certain phenomenon in question.

Typically, reference to interpretation appears to entail hermeneutic approaches. However, even purely naturalistic theories of communication such as Aunger’s (2002: 255ff) argue that communication in populations of agents communicating via signals cannot be viewed in the sender–receiver paradigm, but as a population-level phenomenon in which the effects of communication events on the receivers determine the meaning, and not the intentions of the senders. In fact, this amounts to the naturalization of Wittgenstein’s approach to meaning. In my definition of a ‘social’ mechanism, I actually stay in line with Max Weber’s definition of a ‘social action’ as being a type of action which intrinsically relates to actions of others.

This understanding of ‘naturalism’ follows Bhaskar (1989) and should not be misunderstood as ‘physicalism’, although the general assumption of physical closure of the world would hold (Papineau 2009). In Bhaskar’s view, assigning causal powers to entities is constitutive for defining the ontology that underlies the design and testing of theories. So, in the social sciences constructing mechanisms is tantamount to creating a social ontology.

The current literature on performativity includes a range of different uses of the term, which my most general definition covers. One important strand is to investigate into the performativity of economic theories, following seminal contributions such as MacKenzie (2006). I treat a ‘theory’ simply as an instance of a cognitive state which is mediated via the artefacts that embody the theory (such as books and experimental devices). Callon (2007) extends this approach by including all ideas, practices, and devices that enable economic action, such as accounting practices. I have further expanded this approach to include materially mediated cognitive states in general (Herrmann-Pillath 2010, 2012a). This notion is also more general in not only referring to strong ‘Barnesian’ performativity in the sense of MacKenzie (2007) but also including all phenomena that relate to the emergence of collective intentionality via human interaction, in the sense of Searle (2010) or Tuomela (2007). This use is grounded in the original meaning of performativity in speech act theory.

In fact, this analysis is standard lore in the philosophy of language, referring to the overcoming of referential theories of meaning to rule-based theories which relate meaning to conventions and practices in communities of language users, for a survey, see e.g. Lycan (1999). Searle’s theory of institutions transfers this fundamental shift of perspective to the analysis of institutions.

This diagram is a modification of standard graphic representations of Peirce’s semiotics in biosemiotics, see, for example, El-Hani et al. (2006) or Salthe (2009). For a more detailed exposition, see Herrmann-Pillath (2012b). Peirce laid the ground for the analytical distinction between two modes of causality that are involved in social interactions, efficient and final; for a comprehensive discussion of Peirce’s views on causality, see Short (2007).

This point corresponds to the revival of ‘materiality’ in sociology, see the seminal volume edited by Pinch and Swedberg (2008), which also plays an important role in performativity theory, partly reflecting the intellectual impact of actor-network theory that emphasizes the emergence of agency in networks of human individuals and artefacts (Latour 2005). This goes back to the origins in science and technology studies, where the physical location and structure of the laboratory is a central concern. In economic sociology, this has centred interest on the role of ‘market devices’ in enabling economic interactions (see the contributions in Callon et al. 2007).

Peirce’s major contribution in creating the discipline of semiotics was to elaborate on a complex taxonomy of signs that starts out from studying the nature of the underlying mechanisms. For example, a sign can be embodied information, such as a facial expression signalling an emotion, or purely conventional, such as a linguistic expression. For a survey of the Peircian taxonomy, see Short (2007).

It is straightforward to relate the semiotic model to sketches of mechanistic explanations such as Schmid (2011). Schmid distinguishes four steps in analysing social mechanisms: The explanation of the individual action, the analysis of the interaction patterns, the aggregation process, and the feedbacks between the aggregate level and individual action.

See Muniesa and Callon (2007) who give many examples of how theoretical models of game theory need to be supported by transformational measures aiming at the particular group of actors that are intended to perform the models in a particular context. This can refer to design of locations, design of forms of interactions, and also the training of participants. For example, in real-world spectrum auctions even the economists themselves who designed the mechanisms would be hired by participating companies to perform the mechanism properly.

This exposition summarizes the famous experiments in testing the ultimatum game predictions across a number of ‘small scale’ societies (Henrich et al. 2005). Deviations from the predictions of the model are a standard result which is mostly interpreted in a twofold way. One is to argue that humans are more altruistic than assumed by standard theory. This would be an alternative covering law approach in trying to substitute one universalization by another. The other is to include a learning dimension, showing that after some period of learning, experimental subjects will not commit the ‘mistakes’ anymore and produce the predicted result. The importance of the Henrich et al. study lies in showing up a third solution: This is that the response pattern is systematically influenced by socially embedded interpretations of the subjects (e.g. in societies with cooperative hunting offers in the ultimatum game would also be higher). I think that this also introduces a third alternative to Guala’s (2007) methodological evaluation of experimental economics: He distinguishes between ‘testers’ and ‘builders’ and emphasizes that ‘builders’ aim at transforming the context of an experimental game in order to make subjects performing it. He thinks that this does not invalidate the predictions of the model, as one can see this procedure as an attempt to isolate behavioural determinants, which, after all, actually appear to work, given the setting of the experiment. The Henrich et al. experiments show that collectives of experimental subjects might systematically create autonomous ‘performances’ of the models. This is what is expected in the mechanism approach, such as argued by Little (1992) who champions the idea of medium-level theoretical conceptions in the social sciences, with limited reach in space and time.

Although the learning argument would be the most straightforward one in dealing with these issues from the viewpoint of standard economic theory, in fact even this only works under special ‘performative’ conditions, see Camerer (2003: 59f). I pointed to Guala’s (2007) assessment in the previous footnote. This problem is an aspect of the issue of external validity of experiments which is certainly taken very seriously by experimental economists. But this results in a very strong impact of basic convictions and intentions of experimenters on the actual empirical strategies and interpretation of results, which I would see as another instance of performativity, in this case with reference to the collective or community of researchers.

See Hedström and Ylikoski (2010: 61f). For the original contribution, see Merton (1948: 195): ‘The self-fulfilling prophecy is, in the beginning, a false definition of the situation evoking a new behavior which makes the original false conception come true. This specious validity of the self-fulfilling prophecy perpetuates a reign of error. For the prophet will cite the actual course of events as proof that he was right from the very beginning.’

Singer and Lamm (2009) provide a concise statement of this interaction between top-down and bottom-up processes in triggering empathy. It is particularly interesting because this is a clear case where we cannot say that a mental phenomenon supervenes on a neuronal structure, because the mental phenomenon involves extra-somatic mechanisms. On the consequences of this research for economics, see Kirman and Teschl (2010). Interestingly, they point to experimental evidence that emphatic behaviour even varies for the same individuals depending on specific interactions with others. On the biological foundations of in-group/out-group distinction, see Bowles et al. (2003).

For a survey of the ecological approach, see Yamagishi (2012). In a large number of experiments Yamagishi has shown that Japanese subjects only appear to act more collectivistic than Americans if they receive certain contextual clues. If the context is entirely anonymous and de-contextualized, they often even act less other-oriented than Americans. This contradicts a long tradition in social psychology (e.g. Triandis 1995) in assuming that culture imbues individuals with certain sets of internalized values that explain certain behavioural patterns in comparison to people from other cultures (for a survey in the context of economics, see Beugelsdijk and Maseland 2010).

It is important to notice that economists normally treat all incentives as equivalent to monetary incentives; in the context of experimental economics, this is most explicitly done so, as monetary pay-offs are seen as indirect indicators of utility. Beyond economics, the notion of incentive is much broader and includes, for example, praise, awards, prizes, fame, and so on. As is well known from psychological research, these incentives can work very differently on motivation than material incentives that are directly targeted at producing a certain level of activity. I come back on this point below.

There are very complex constellations that need further scrutiny in terms of Peircian semiotics, as I have previously mentioned (footnote 13). Basically, every incentive is a physical object, namely a physically embodied stimulus. There are incentives in which object and sign are physically united such as in the case of extending bodily caress to a person as a positive gratification. In other cases, the sign is separate, such as bestowing a medal on a person, where the original incentive would be the psychological and social states that are expressed by that sign. Money raises very tricky issues here, as economists, but also some neuroscientists treat money as directly reflecting the underlying utility. Although this is regarded as a technical device that is limited to certain experimental settings, the use of money is normally extended far beyond them. That would imply that money plays the intricate role of a culturally conditioned ‘primary reinforcer’, coming close to an oxymoron (see Camerer et al. 2005: 35). In standard economic theory, to the contrary, money is treated as a sign (a ‘veil’) that represents other underlying incentives, which are the things money can buy (see the discussion in Harrison 2008: 306f).

A concise argument on this has been presented by Sliwka (2007), compare also Falk and Kosfeld (2006). Psychologists have shown that in such a setting, there are many degrees of freedom: For example, when the incentive system signals the dominance of cooperative types, this might induce more people to free-ride (see Chen et al. 2005).

There is ample psychological evidence of strong framing effects of money, as in priming experiments; for a survey, see Vohs et al. (2006). Framing effects can be various and differentiated (e.g. even framing with clean or dirty banknotes can make a difference in behaviour); as exemplary studies, see Yang et al. (2012). This literature also refutes the typical assumption in experimental economics that money directly reflects underlying utilities, see Amir et al. (2008).

Since the seminal survey of Deci et al. (1999), the general notion of a conflict between extrinsic reward and intrinsic motivation is well accepted, but rarely received in economics, with exceptions such as Frey (1997) or Falk and Kosfeld (2006). Most importantly, for our discussion of managerial incentive system in the next section, these effects are especially strong if extrinsic rewards are very large and fall into the same category as the goal pursued by the incentivized action (James 2005). Bonus systems for managers should have an especially strong negative effect on their intrinsic motivation.

As one example of the discussion, see the influential work of Bebchuk and Fried (2004); on the public debates, see Joutsenvirta (2013). This has already triggered many regulatory responses, such as imposing caps on bonuses and strengthening the role of shareholders in fixing remuneration schemes for CEOs. But all these measures so far do not question the fundamental principles of these schemes.

The seminal contributions were La Porta et al. (1997, 1998) and the generalization by Djankov et al. (2003). Interestingly, in the latter contribution, contextualization creeps back into the analysis because the performance of certain legal systems is seen as being dependent on the stock of civic capital.

For a devastating empirical critique of the legal origins theory, see Siems and Deakin (2010). Aoki (2010: 71ff) is a good survey of the discussion.

This refers to the phenomenon that the investors’ perspective tends to dominate the entire institutional design of the corporate sector, hence also changing the strategic orientation of business towards financial goals, see Krippner (2005). In Herrmann-Pillath (2013) I offer an interpretation of this in terms of performativity theory, including aspects such as the accounting and financial reporting institutions or the patent system.

Aoki (2010: 26ff). Aoki’s theory can be related to formal mathematical approaches to firm structure that point towards positive externalities between production factors, thus creating supermodular production functions, see e.g. Milgrom and Roberts (1990). Interestingly, in practical application of this thinking, ‘high commitment human resource management systems’ would not take the shape of high-powered extrinsic incentives, see Roberts (2004: 174f).

This is an important issue in social ontology: Economics just accepts the idea that cognitive states are confined to individuals, hence brains, whereas in recent developments of cognitive science these are seen as being partly externalized, see e.g. Clark (2011). Again, this can only properly appreciated in a mechanism view.

These arguments have been already deployed in Aoki’s (1988) classical comparison between the A-Firm and the J-Firm. In principle, the two governance schemes in the USA and Japan achieved similar performance levels at that time, because they were embedded into different institutional structures of the labour market, different distributions and contents of skills, and so on, hence were contextualized differently. However, it is also partly a question of technology which scheme works best: At that time, Aoki pointed towards automotive industry as an example for a good match with the Japanese model, whereas chemical industry might be better governed by the US model. This observation points towards the possibility of ‘counterperformativity’ (MacKenzie 2007), which is an important phenomenon in rendering performativity theory empirically meaningful.

Ghoshal (2005) accused business schools in educating students to become opportunistic agents, based on advanced theory. It is important to notice that in the light of performativity theory, this does not need to suppose that students actually ‘become’ more opportunistic, which is probably not the case empirically (Guala 2007). Ghoshal’s reasoning matches with the aforementioned theory by Sliwka (2007) in that the education transmits information about the composition of types in the economy, which triggers strategic responses of individuals. Such effects can be leveraged by endogenous sorting of different types across market and non-market domains of the economy, see Kranton (1996).