Introduction to The Invisible Hand in Economics (Routledge, 2008)


This is the Introduction to The Invisible Hand in Economics: How Economists Explain Unintended Social Consequences (Routledge, 2008). If you prefer to read the PDF version, it is available on ResearchGate.

Everyone is familiar with the (aesthetically) unpleasant walking-paths on public green fields. Usually, around these fields there are constructed paths for the service of the beloved citizens. However, citizens seem to deviate from these designed paths and take shortcuts passing through the green fields, through the zones they should not pass. The outcome of this behaviour is the death of the plants and the emergence of paths on these fields. Is there a good and acceptable explanation of why these walking-paths emerge? What kind of an explanation would be acceptable?

One scenario about the emergence of these paths may be the following. The first person to take a shortcut through the aforementioned field believes that his behaviour will do no harm to the green field and proceeds to take the shortcut. Obviously, he will in

fact have a negligible impact on the field. Subsequently, a second person, without being aware of the fact that he is the second person, takes the same shortcut with similar thoughts in mind. However, although the second person’s impact is still negligible, in time the damage on the plants accumulates. After some time, if some other people take that very same shortcut the damage will soon become visible. As a consequence, even if some earlier citizens did not intend the outcome of their behaviour, the damage to the plants will become visible; and hence the emergence of the path.

The plants may recover if everybody ceases to use the path. However, people generally continue taking shortcuts given the fact that they see the damage previously done by others. What are they thinking? Do they intend to bring about and maintain a visible path with no plants on it? Probably (and hopefully) not! Their reasoning is perhaps the following: ‘Even if I do not take the shortcut, some others will, thus the path will stay there anyway. Then, why should not I take the shortcut?’ So in this scenario, although individuals do foresee the outcome of their behaviour, they think that the creation of the path would be inevitable, given that everyone takes the shortcut. Moreover, if one particular individual is the only one taking the shortcut, the consequence of her action would be negligible. Thus, she takes the shortcut without intending to create such a nasty path. In this scenario the damage done to the plants has reinforced other people

to use that same path for it already exists. Thus, unless someone intervenes, the path is there to stay.

Would this scenario explain the emergence of these walking-paths? It sounds reasonable, but definitely not everyone is likely to accept this explanation. In fact, this was just one possible explanation. By introspection some of you may think that there may be other motivations behind taking these shortcuts. You may argue that people do not think about the consequences of their action at all when taking these shortcuts or you may think that these paths are intentionally created. For example, you may argue that one of those paths emerged due to a poorly planned city park which resulted in people intentionally taking a particular shortcut to show their dissatisfaction with the planning of this particular public space.

The above is yet another possible way to explain the emergence of these paths. More importantly, this explanation contradicts the previous general scenario. Thus one is tempted to ask: ‘unless some evidence is provided why should we believe in such speculations?’ If we would ask different people to explain this fact, we would get many different explanations. Many of these explanations would contain a story that makes the emergence of the path plausible. Hence, we would be left with many different (but not necessarily mutually exclusive) speculations or conjectures instead of ‘proper’ explanations. The fact that there are walking paths on public green fields is ostensibly simple to explain; however, it seems we are left only with conjectures.

The Invisible Hand in Economics
The Invisible Hand in Economics (Routledge, 2008)

Broadly speaking, conjectures and their explanatory characteristics are the subject matters of this book. It examines one particular type of explaining practice in social sciences, namely explaining the emergence of institutions (e.g., conventions and norms) and macro-social structures as unintended consequences of human action, from a methodological and philosophical perspective.

Explanations of ‘unintended consequences’ show numerous similarities with the above example of walking-paths on public green fields. The basic similarity however is that they seem to lack empirical content and as such they can be criticised as being simple conjectures with no explanatory value. This book illustrates the merits and demerits of such explanations by examining some of these attempts to explain institutions and macro-social structures as unintended consequences.


This is a book about ‘unintended consequences of human action’ and the mechanisms that bring about these consequences. It investigates the explanatory role of the models that characterise institutions and macro-social structures as unintended consequences of human action.

Many economists argue that certain institutions and/or social structures, such as money, language, rules of the road, fairness, segregated city patterns, and localisation are unintended consequences of human action. Similar ideas can be found in other disciplines, such as in philosophy of language (e.g., Lewis 1969), in political philosophy (e.g., Nozick 1974), in linguistics (e.g., Keller 1994) in philosophy of science (e.g., Hull 1988), in sociology (e.g., Merton 1936, Boudon 1982). The above list, which can definitely be extended, illustrates the importance of the problem of explaining unintended consequences in social sciences and in (political) philosophy. Popper (1962: 342) acknowledged the significance of unintended consequences for social sciences by arguing that the explanation of unintended consequences of human action is ‘the main task of theoretical social sciences.’ Yet unintended consequence is a vague concept and as such it may denote many different things. In this book we will be concerned with a subset of the set of possible unintended consequences; the one which is of paramount importance to economics in explaining institutions and macro-social structures. The rough description of this subset is as follows (see Mäki 1990b, and Chapter 2 & 5):

  • Individuals do not intend to bring about a social phenomenon (e.g., a social institution, or a macro-social structure).
  • The consequence of their action is a social phenomenon (i.e., an institution or a social structure)
  • One individual alone is not enough to bring about the ‘social’ consequence—that is, independent actions of similar (in the sense that they do not intend to bring about the consequence) agents are needed.

An explanation of unintentionally hurting your hand, or an explanation of the unintended consequences of a government tax plan does not fall under the above definition and hence are not examined in this book. The former is not examined because the consequence is not a social phenomenon (violates condition b—and c depending on the case); the latter is not examined because the intention is about a social institution (condition a is violated). Given our definition, ‘unintended consequence’ is a crucial component of the ‘theory of spontaneous order’, of Adam Smith’s ‘invisible hand’, of Carl Menger’s notion of ‘organic phenomena’ and of ‘invisible-hand explanations’. It lies at the core of many contemporary models of institutions and macro-social structures.

The theory of spontaneous order finds its origins in 18th century Scottish Thought and it is defined with its characterisation of the social order as ‘unintended consequence of countless individual actions’ (Hamowy 1987: 3). Adam Smith, the founder of modern economics, who was part of this tradition, presented a metaphorical statement of ‘spontaneous order’ with the ‘invisible hand’. Menger, on the other hand, presented a closely related account of spontaneously created social institutions where he considered them as being similar to ‘organic phenomena’. Many social scientists and philosophers followed Smith and Menger by trying to answer versions of their questions about institutions and macro-social structures. Their aim was to show how institutions and social structures could emerge (or persist) without any design. Generally it is believed that neoclassical economists followed Smith’s lead and tried to prove his insights. However, neoclassical economists’ approach is only one of the possible ways to interpret Smith’s insights. There are at least two different interpretations of the ‘invisible hand’: the one that stresses ‘processes’ and the other that emphasizes ‘end-states’ (see Chapter 5). Neoclassical economists’ approach is an end-state interpretation: The ‘invisible hand theorem’ in economics stresses the consequences (e.g., optimum allocation of resources), rather than the processes that bring about these consequences.

This book is mainly concerned with what may be called the process interpretation of the invisible hand. Under this interpretation the ‘invisible hand’ represents causal and structural relationships and processes that may bring about unintended social consequences. Explanations under this particular interpretation can be gathered under the notion of ‘invisible-hand explanations’. An invisible-hand explanation aims to show the process that brings about the unintended consequence. Rather than merely focusing on the properties of the end-state (e.g., equilibrium), it explicates the way in which the end-state may be reached (Nozick 1974, Ullmann-Margalit 1978). It is possible to find many examples of such process models in the contemporary literature. The most prominent examples are game-theoretic models of institutions that show how institutions may emerge (or persist) as an unintended consequence of human action (e.g., Bicchieri 1993, Sugden 1986, Ullmann-Margalit 1977, Young 1998 etc.). Another area where unintended consequences are important is the emerging field of agent based computational economics (e.g., Axelrod 1997, Axtell et al. 1999, Epstein and Axtell 1996, Marimon, McGrattan et al. 1990 etc.). The pioneers of the game-theoretic literature acknowledge David Hume, Adam Smith and Carl Menger as their forefathers (see Lewis 1969, Sugden 1986, Ullmann-Margalit 1977, Young 1998) and similarly agent based computational economists generally acknowledge their intellectual debts to Adam Smith and Thomas Schelling (e.g., see Epstein and Axtell 1996, Tesfatsion 2002).

Carl Menger’s (1892a) story of the emergence of a medium of exchange, Thomas Schelling’s (1969, 1971, 1978) models of residential segregation, Peyton H. Young’s (1993a, 1996, 1998) model of emergence of the rules of the road and Joshua Epstein’s and Robert Axtell’s (1996) ‘Sugarscape’ (where they grow artificial societies from the bottom up) are well known examples of such models. These examples range from verbal models (or stories) to formal game theoretic and computational models. Despite the differences in their methods and research tools there is an important similarity between them. Each shows how institutions and macro-social structures may emerge (or persist) as an unintended consequence of the (inter)actions of individuals. In order to do this, the authors conjecture (in the latter case with the help of computers) about the conditions under which individual actions may lead to the social phenomena in question. The common feature of these examples is that the observed social phenomenon (i.e., institution or macro-social structure) is ‘produced’ within the model-world by conjecturing about the initial conditions (e.g., environmental conditions, characteristics of the agents etc.) that may bring about the social phenomenon in question as an unintended consequence of the interactions of the agents. Moreover, these models are typically ahistorical in the sense that historical facts about the social phenomenon in question do not seem to play any role in these models. They are general and thus they are supposed to be applicable to all instances of the social phenomenon in question in different times and places. These models illustrate the possible ways in which certain mechanisms may interact (or may have interacted) to produce the types of institutions or macro-social structures in question.

More strikingly, some of these models seem to challenge the common sense and the historical knowledge about these social phenomena. For example, while many believe that money is a matter of design and was issued by central authorities in the past, Menger argues that it was brought about by the (inter)actions of individuals who were pursuing their self-interests without the intention to bring about a commonly acceptable medium of exchange. Schelling, on the other hand, shows that if individuals cannot tolerate living as an extreme minority in their neighbourhood, then residential segregation cannot be avoided even if they are happy in a mixed neighbourhood. Of course this seems to go against our belief that strong discriminatory preferences (e.g., racism) and economic factors (e.g., wealth differences among ethnic groups) are the main causes of residential segregation. A more recent example is Young’s model of the rules of the road. He shows that the rules of the road may emerge with the accumulation of the precedent as an unintended consequence of the (inter)actions of the individuals. This model also goes against the belief that the rule that specifies on which side of the road one should drive was designed and imposed by central authorities like other traffic rules. Finally, Epstein and Axtell show how under certain conditions fundamental social structures and group behaviours (e.g., institutions, segregation, cooperation) could emerge from the micro level. In this example, social phenomena are quite literally grown by the authors.

All of the aforementioned examples pose difficult questions for social scientists and philosophers. How could these models explain anything if they are simply speculations about the initial conditions under which social phenomena may be brought about as unintended consequences? In other words, we understand that these authors are able to ‘produce’ a certain social phenomenon in their model world as an unintended consequence of the interactions of model agents. However, given that these models are so abstract, ahistorical and speculative, how could they be used to explain something about the real world?

The philosophical and methodological challenges posed by these models created many debates in related areas. These debates constitute a significant part of the controversies about the role of abstract modelling in social sciences. The related question here is whether we can learn anything about the real world by studying highly abstract models. This is one of the basic questions of philosophy of science. The usual defence of scientific models is the claim that they isolate the relevant parts of the real world and that such realistic representations of the real world give a close to true account of the phenomenon in question when other things are absent or constant. Some economists also use the argument from realistic representation in defence of their models. For example, Young (1998: 10) argues that the assumptions of his models ‘represent a fairly accurate picture of reality.’

However, the main criticism to these models is that they ignore the relevant facts, such as the history of the social phenomenon in question—and therefore they do not realistically represent the relevant parts of the real world. For example Menger’s ‘the origin of money’ does not take into account the way in which money was issued and introduced in history. Moreover, Schelling’s model of segregation seems to sidestep two of the most important facts about segregation—the presence of strong discriminatory preferences and the role of economic factors. Thus the argument concerning realistic representation either has to demonstrate why history is irrelevant, or to show the complementarity between these models and history.

As mentioned above, these models start with the problem (e.g., that there is residential segregation) and try to produce the conditions under which this problem may emerge as an unintended consequence of human (inter)action. This methodology invites criticism for the following two reasons. Firstly, it seems to be one sided, for it tries to construct a model that shows something that the author wishes to see (e.g., residential segregation as an unintended consequence). Secondly, it may be argued that if one devotes enough time and energy it should be possible to construct a reasonable model that is able to show whatever we wish.

Given the focus of this book the relevant place to start seeking solutions to these problems is the literature on ‘invisible-hand explanations.’ It is argued in this literature that invisible-hand explanations are valuable independently from their truth for they explicate the process that may have brought about the social phenomenon at hand (Ullmann-Margalit 1978). This suggests that these models are valuable even if they are false, or even if they do not get the facts right. It is argued that explication of a hypothetical process that is sufficient to bring about the social phenomenon in question is valuable for its own sake. However, it is not explained why this explication would be valuable, or in what sense it would help us understand the real world. Simply this argument does not help us much unless it is explicated! This is merely a statement of the author’s intuition about the value of these ‘explanations’. Economists use these models because they believe that they are valuable. For example, Robert Sugden (2000) argues that the reason we believe that these models are conveying a true message about the real world is that we find them ‘credible’—by way of examining Schelling’s segregation model. He argues that these models are credible like a good story or a novel. Sugden’s basic argument boils down to the statement that we think these models are valuable because we find them plausible. This argument cannot demonstrate the value of these models unless it explains why they are plausible and in what sense plausibility of a conjectural process sheds light on the real world.

Another type of justification comes from computer simulations. It is argued that artificial environments (models, computer models) are used to gain insights about the social phenomenon in question (e.g., see Gilbert and Doran 1994a, Liebrand et al. 1998) or that models and computer simulations are like experiments where we test our ideas (e.g., see Drogoul and Ferber 1994) or that they are similar to thought experiments (e.g., Liebrand 1998, Liebrand et al. 1998). Briefly it is argued that models and simulations help us in finding out the necessary conditions under which certain results (e.g., segregation) are brought about (within the computer model) and in easily exploring the properties of these model environments. In this account these models are not for explanation but for exploration. However, this does not answer our question about moving from the model world to the real world. Specifically, a satisfactory defence of these models would have to tell us how to translate the results of the model in order to interpret the real word.

The discussions about the interpretation of game theory are also relevant in this respect. As mentioned above, some economists like Young use the argument from realistic representation to justify game-theoretic models. Yet not every game theorist would agree with this. One of the most prominent scholars of this field, Robert Aumann (1985) (also see van Damme 1998) argues that realisticness of the models does not matter that much. According to him, the conclusions are much more important: If the model is applicable to many situations and is productive, then it is a good model. He also argues that game theory (and other sciences—in his opinion) ‘is not a quest for truth, but a quest for understanding.’ He says, ‘science makes sense of what we see, but it is not what is “really” there’ (van Damme 1998: 181, 182). Aumann basically argues that game-theoretical models help us in putting together what we observe in a coherent framework, that they help us in fitting things together. He also argues that they lead to prediction and control. However, if we accept Aumann’s interpretation of game theory (and science) we are still faced with the following questions: Firstly, if the models of institutions and macro-social structures do not represent the reality how could they lead to prediction and control, and most importantly to understanding? Secondly, Aumann emphasises the productiveness and applicability of the models. Yet the current state of the modelling of institutions and macro-social structures (as unintended consequences) cannot be considered to have many real world applications or satisfactory predictive power. Should we then conclude that these models are not valuable?

It is the argument of this book that all of the interpretations expressed above convey justifiable intuitions about these models. That is, ‘realisticness’, ‘explication’, ‘credibility’, ‘exploration’, and ‘fitting things together’ are all parts of a framework that would help us in making sense of these models. However, there is no existing framework where these things are presented coherently and satisfactorily. It is the main task of this book to develop such a framework and to use it to gain new insights into the contemporary literature that characterises institutions and macro-social structures as unintended consequences of human action.


To be able to develop such a framework one has to understand what these models really accomplish. The most obvious way to do this is to carefully examine these models and their methodology. But before doing this, a clarification of the very idea of ‘unintended consequences’ is needed. The Second Chapter analyses and explicates the concept of unintended consequences to prevent misunderstandings that may be caused by its vagueness. In particular, the subset of the set of possible unintended consequences, which is relevant for understanding the models of institutions and macro-social structures as unintended consequences, is specified.

In Chapters Three and Four, the most prominent examples of such models, Menger’s ‘origin of money’ and Schelling’s ‘checkerboard model of segregation’ are examined. It appears they are natural candidates for several reasons. First of all, these models are paradigmatic examples of ‘explaining unintended consequences of human action’ and of invisible-hand explanations. Contemporary authors consider these models as conveying the key insights about their subject matter and about the way in which related issues should be handled. Their models are the predecessors of contemporary research in modelling institutions and macro-social structures as unintended consequences of human (inter)action. Menger is considered to be one of the founding fathers of the theoretical approach to institutions as opposed to the historical approach (e.g., see Rutherford 1994, Schotter 1981). Schelling’s model is one of the main predecessors of agent-based computer models (e.g., see Epstein and Axtell 1996, Blume and Durlauf 2001, Pancs and Vriend 2003, Rosser 2000, and Casti 1989) and it is considered to be the paradigmatic example of explaining with mechanisms in social sciences (e.g., see Hedström and Sedberg 1998). Briefly, since their models play an important role in the history of ‘explaining unintended consequences of human action’ understanding Menger’s and Schelling’s models should shed light on the related areas of contemporary research.

Another good reason to start our examination with these models is the fact that both Menger and Schelling are explicit about their methodology. In their work, they explain why they prefer the type of research they are engaged in. Moreover, in the literature there is a considerable amount of philosophical discussion about their methodology. As previously mentioned, their works are predominantly considered to be paradigmatic examples of invisible-hand type of ‘individualistic’ explanations (e.g., see Nozick 1974, Pettit 1996, Rosenberg 1995, Rutherford 1994, Ullmann-Margalit 1977). It is also common to examine the recent game-theoretical models of institutions alongside invisible-hand explanations (e.g., see Langlois 1986bc, Mäki 1993, Rosenberg 1994, Rutherford 1995, Vanberg 1994). It has also been stated that the authors of these models (including the authors of the computational models) consider themselves as following the invisible-hand tradition, or providing the mechanisms behind the invisible hand. For these reasons there is a considerable amount of resources that may help us in our quest. Thus we are more likely to find hints about the nature of similar models by starting our examination from Menger and Schelling’s models and their relation to the invisible hand. In addition, this choice makes it easier to see the common misunderstandings about ‘explaining unintended consequences of human action’ for the literature on invisible hand type of ‘individualistic’ models is abound with controversies. The Fifth Chapter undertakes the task of examining ‘invisible-hand explanations’ in light of the chapters on unintended consequences, and Menger and Schelling’s models. This examination sheds light on the nature of invisible-hand explanations. Particularly, an important misunderstanding about the relation between ‘unintended consequences’ and the ‘invisible hand’ is removed. By way of removing this misunderstanding the chapter prepares the ground for examining contemporary examples of invisible-hand explanations.

Menger and Schelling’s models and insights were reconsidered and remodelled by contemporary authors. For this reason there is an explicit link between these models and the contemporary literature we wish to understand. This gives us a chance to evaluate the progress of this ‘research program’. For example, some of the papers directly related to Menger’s account of the medium of exchange can be listed as follows: Duffy and Ochs (1999), Gintis (1997), Kiyotaki and Wright (1989), Marimon, McGrattan et al. (1990), Schotter (1981), Selgin and Klein (2000), Townsend (1980), Young (1998). Some of the follow-ups to Schelling’s segregation model are the following: Clark (1991), Epstein and Axtell (1996), Sander et. al. (2001), Young (2001), and Zhang (2000, 2004ab). By examining these reconsiderations we may indeed see whether there is any progress or whether contemporary tools (e.g., game theory and computer modelling) improve the way in which we understand and explain the origin of money and segregation. Accordingly, the Sixth Chapter examines the more recent models of the emergence of money in detail, while recent reconsiderations ‘residential segregation’ are used as examples in Chapter Seven.

Particularly, in Chapter Six it is argued that we should not evaluate models that characterise macro-social phenomena as unintended consequences in isolation from other related models of the same phenomenon. In order to substantiate this proposition, recent reconsiderations of Menger’s explanation of the origin of money are examined. The chapter shows how Menger’s intuitions are further explored in the modern literature in various ways. It is argued that these recent models test the logical soundness of Menger’s arguments but do not bring us any closer to the real world. Recent models of the origin of money do not introduce new mechanisms but test the plausibility of the mechanisms that were suggested by Menger. While these models increase the plausibility of the idea that media of exchange may be brought about unintendedly, it is argued here that the idea that fiat money may be considered as an unintended consequence of human action does not appear to have a firm basis. Moreover, the chapter examines and demonstrates the relation among these models. This examination supports the thesis that different models have different functions and different models of the same phenomenon may be considered as forming a loose framework for explaining particular instantiations of it.

Chapter Seven explores the philosophical literature on models and explanation to provide a firmer basis for the arguments of the previous chapters. In particular, firstly, the concept of partial potential explanations is explicated. Secondly, it is argued that models help us explain by way of providing a proper way to conceptualise the phenomenon under question. Yet this further implies that the relationship between the model world and the real world is rather complex. Thirdly, it examines this complex relationship by way of discussing the related philosophical literature in light of the previous chapters. Fourthly, it is argued that similarity between models that are examined in this book and the real world amounts to the existence of certain (known) tendencies (individual mechanisms) in the model world. For this reason these models may be interpreted as revealing the possible ways in which these tendencies may interact, even if some of the assumptions of these models do not hold. Fifthly, the chapter emphasises the importance of exploration. Particularly, it shows how one may gain confidence about the implications of an existing model by way of further exploring its premises and results. To do this the chapter discusses reconsiderations of the checkerboard model. Finally, the chapter fortifies the idea that no model of this sort should be evaluated in isolation from other related models.

Chapter Eight examines the modern game theoretical models of conventions in light of the ideas developed in previous chapters. These models may be considered as attempts to provide a general theory of the emergence of conventions. The chapter reviews some of the existing game-theoretic literature on conventions and shows that existent conventions and norms, particular institutions and history are crucial for explaining the emergence of conventions. Six arguments are put forth in the chapter: (i) Static models of coordination (and convention) are concerned with examining the conditions under which certain outcomes are plausible, rather than explaining why and how such outcomes are brought about. Hence such models are in line with the end-state interpretation of the invisible hand. (ii) Dynamic models of coordination provide partial potential (theoretical) explanations of the emergence of coordination and conventions. Hence such models are in line with the process interpretation of the invisible hand. (iii) None of these models rule out the possibility that coordination and conventions may be brought about intentionally. Rather they examine whether successful coordination and conventions may emerge as unintended consequences of human action. The interpretation of these models as providing partial potential explanations is well in line with this remark. (iv) Explaining particular cases (e.g., explaining the emergence of a particular convention) necessitates empirical research. Nevertheless, general models of coordination and conventions need not be empirical or historical. (v) The collection of different models of coordination and conventions may be considered as providing a general framework for empirical research and providing singular explanations. (vi) Game-theoretic models in general may be interpreted as providing a framework for analysis, rather than providing ultimate explanations concerning social phenomena and individual behaviour. The Final Chapter concludes the book with questions for further research.

To learn more about The Invisible Hand in Economics (Routledge, 2008), visit this page.  Or buy the book.