Introduction

In The Problem of Social Cost, Coase (1960) used examples to suggest that, in the presence of externalities, if transaction costs are nil and if property rights are clearly defined and allocated, agents bargain over rights and achieve an optimal output that is independent of the initial allocation of rights. This proposition was to be called the “Coase theorem” by Stigler (1966, p. 113).

The theorem has long been tested against the facts, whether already existing (empirical studies) or purposefully constructed (laboratory experiments). Strictly speaking, however, these studies do not in fact test the Coase theorem. The reasons for this are twofold: first, the conclusion of almost six decades of debate over the validity of the theorem is generally taken to be that all criticisms can be subsumed under the category of transaction costs and that this renders the “theorem” tautological (Medema and Zerbe 2000). Second, transaction costs never being nil, the theorem does not apply to the real world (most of Coase’s seminal article actually examined the consequences of these costs). Regarding experiments, they reduce transaction costs to a minimum. As for empirical studies, they can only test a “generalized Coase theorem” (Bertrand 2011): if the gain from a transaction concerning a right (the use of which provokes side effects) is greater than its cost, then the transaction takes place. I will insist on the difficulty of testing such a proposition without measuring transaction costs, and on another obstacle to negotiation over and above these costs: a moral or social prohibition on exchange. For a detailed review of the tests of the Coase theorem, see Medema and Zerbe (2000).

Empirical Studies of the Coase Theorem with Externalities

Some empirical tests of the theorem have concerned pretrial settlements (Galanter 1983) and transactions on nuisances after trials (Farnsworth 1999), but the majority test the prediction of the absence of any effects engendered by a change in the law: notably with regard to share tenancy arrangements (Cheung 1969), rules governing divorce (Peters 1986), the effect of a payment of bonuses to unemployed workers when they obtain a job or to employers when they hire unemployed people (Donohue 1989), and generally with a very specific attention to sports (Hylan et al. 1996; Cymrot et al. 2001; Szymanski 2007). I will focus on two tests which concern well-known examples of externalities: Coase’s cattle that destroy crops on neighboring land (Ellickson 1986), and Meade’s bees that pollinate orchards while foraging for nectar (Cheung 1973). The cattle case was also examined by Vogel (1987), who analyzed the effects of the changes in trespass law in the different counties of California during the second half of the nineteenth century: the assignment of rights to farmers tended to increase farm output, contrary to the prediction of the theorem (this movement was studied over a longer period than that of Ellickson, which explains their different results). Hanley and Sumner (1995) also studied the resolution of externalities due to red deer in Scotland, and their conclusions are close to Ellickson’s.

Ellickson (1986) intended to test the realism of the “Parable of the Farmer and the Rancher” (p. 624), the numerical example developed by Coase (1960) in which, on the basis of their formal entitlements, a cattle raiser and a crops farmer negotiate the size of the herd in exchange for a monetary payment, and arrive at the same size whatever the initial allocation of rights is. In 1982, therefore, Ellickson turned to Shasta County, in the north of California, where two legal regimes coexisted: most of the county was under the historical “open-range” regime, in which a cattleman is not liable for trespass damages even when negligent, but one district had switched to a “closed-range” regime in 1973, where the rancher is always liable for damage even in the absence of negligence.

Although the allocation of resources (e.g., spaces respectively devoted to crops and cattle) is not affected by changes in liability law, the explanation does not turn on the monetary negotiations invoked by Coase: neighbors rather refer to social norms to resolve their disputes; and since these norms are independent from formal law, the resolution of incidents is as well.

In particular, the parties to a dispute refer to an informal rule according to which the cattle owner is liable for the actions of his animals. This goes against the formal rule of the main part of the county, which implies that “norms, not legal rules, are the basic sources of entitlements” (Ellickson 1986, p. 672). Other norms detail the manner of resolving an incident: most of the time, the victim downplays the incident, which will be quickly solved by an exchange of civilities, and neighbors expect reciprocity. In any case, the resolution of the conflict must be informal, without recourse to law or courts, and monetary compensation is forbidden: inhabitants “regard a monetary settlement as an arms-length transaction that symbolizes an unneighborly relationship” (ibid.: 682).

This monograph therefore shows that neighbors do not solve their disputes with monetary transactions, but through social norms, which Ellickson explains by high transaction costs. But, first, he does not measure them: he simply infers them from the facts that exchanges do not take place and that the law would be costly to learn (a cost itself inferred from the fact that inhabitants do not know it). Second, it could be argued that transaction costs are low: small number of parties, easily identifiable, easy monetary assessment of the damages, sharing of the same social norms, etc. Third, other obstacles than transaction costs may explain the absence of exchange: social norms can impede a market from developing (Bertrand 2011). Take the norm that condemns monetary settlements: Ellickson explains this by reference to transaction costs, but we could just as well assert that it is this norm that prevents transactions in the first place, and hence that the norm lies at the origin of the impossibility of negotiations. Transaction costs would be irrelevant since agents are not willing to negotiate.

The absence of monetary payments and of reference to formal rules thus raises questions about the legitimization of the legal structure of rights and of their alienability, which may deter agents from negotiating these rights, quite apart from transaction costs. In the Shasta County case, there is no norm indicating that you can transact over the right to destroy your neighbor’s crops. In fact, we here encounter the basic problem of externality: a market does not exist, and it may be difficult to create one out of nothing because of social norms. By contrast, the next example will stress that a market does seem to exist for what was long considered as externalities: pollination and nectar services.

Cheung’s “Fable of the Bees” (1973) tests the realism of Meade’s (1952) example of the externality between beekeepers and orchard owners. It was written at Coase’s request, who was not satisfied with Johnson’s study (1973) – the latter being more focused on the institutional setting of the pollination service market. Cheung’s investigation was conducted in the state of Washington in spring 1972; it covered a sample of 9 beekeepers and a total of approximately 10,000 spring colonies.

Cheung first observes that a marketplace exists for transactions on pollination and nectar services. On the one hand, an orchard owner can rent the pollination services of an apiarist who places her hives in the orchard; he pays her a monetary pollination fee that depends on the number, density, and strength of hives. On the other hand, an apiarist who wants to place her hives among crops for honey production pays an apiary rent to the orchard owner, mostly in the form of honey, and depending on the honey crop. Note, however, a difference with Coase’s parable of the rancher and the farmer: an orchard owner can choose a beekeeper from among others and vice versa; they are not compelled to negotiate or to find a solution with their neighbor.

Nevertheless, we here have evidence that contracts with monetary exchanges can deal with so-called externalities. It seems that the possibility of exchanges in pollination services (or social permission) developed in the period after WWI, alongside with the specialization of farms which required specific pollinators. The right of having his plants pollinated was said to be exchangeable and a price was suggested: the set of social norms necessary for the operation of a market was available. Still, regarding the habit of giving some honey to the orchard owner in exchange for the right to place hives in his orchard, the question remains whether this should be considered as a market transaction or a social norm.

In any case, the markets for pollination and nectar services are unusual in that the enforcement of contracts, whether oral or written, relies heavily on social norms. In addition, externalities remain at the margins of the market for pollination services, and they are dealt with by social norms. For example, my neighbor, who also cultivates apples, benefits from the hives I rent for my apple trees. These positive externalities are solved by a social norm of neighborliness called “the rule of the orchards,” by which we both rent the same number of hives per acre (Cheung 1973, p. 30). Cheung explains the absence of monetary exchange to solve this externality by the cost of such a negotiation (not measured, but itself inferred from the absence of such an exchange). As with trespass incidents, we could alternatively explain the absence of negotiation by the existence of an already satisfying norm or, prior to it, by other norms that would deter monetary negotiations between neighbors and make transaction costs irrelevant.

It would be excessive to infer from Ellickson’s and Cheung’s studies that the generalized Coase theorem is empirically confirmed, since both authors have a tendency to explain their observations precisely by this assumption. This kind of empirical study faces the problem of measuring the gains and costs of exchange. However, the control possible in the laboratory allows for a more precise determination of them.

Experiments of the Coase Theorem with Monetary Negotiations

Laboratory experiments concerning the theorem began in 1982 with the publication of studies by Prudencio (1982) and Hoffman and Spitzer (1982). It is the last protocol, closest to the Coasean parable of the rancher and the farmer, that has been used the most. Experiments of the theorem seek to test the robustness of results given variation of the implicit or explicit assumptions: high number of agents (Hoffman and Spitzer 1986); incomplete information (Prudencio 1982; McKelvey and Page 2000); introduction of transaction costs (time limit in Prudencio 1982 or cost of an offer in Rhoads and Shogren 1999); uncertainty about the payments (Shogren 1992); property rights earned and/or legitimized (Hoffman and Spitzer 1985a), not allocated (Harrison and McKee 1985) or uncertain (Cherry and Shogren 2005); non-convexity (Shogren et al. 2002); empty core (Aivazian et al. 2009); and physical discomfort (Coursey et al. 1987). Schwab’s (1988) experiment does not concern externalities but rather contract presumptions. Another series of experiments consists in a comparison of the efficiency of different solutions to externalities (e.g., Plott 1983; Harrison et al. 1987). The results of the experiments on the endowment effect were applied to refute some of the theorem’s predictions (see Kahneman et al. 1990) but are not specific to the theorem and do not exhibit theorem-like mechanisms.

It is Hoffman and Spitzer’s (1982) first set of experiments, with two persons and complete information, whose design is the closest to the Coasean parable. The respective monetary payoffs (net profit) of the two subjects, who were randomly assigned the letters A (rancher) and B (farmer), depend on the value of a discrete number (the size of the herd): a couple of payoffs are associated to each number (0, 1, etc.) and only one number maximizes the sum of these payoffs. A’s payoff increases when the number increases and conversely for B, which renders the “externality” negative. The allocation of the property right is translated by the designation of one of the subjects as the “controller”: she has the right to unilaterally choose the number (and hence the payoffs). The possibility of negotiation comes from the fact that the other subject may influence her choice by proposing to transfer a part of his own payoff. The controller is randomly designated through a heads or tails game.

As for the results, exchanges take place and are efficient (23 out of 24 decisions choose the number that maximizes the sum of payoffs). The authors infer that this result, confirmed by their other sets (1982) and other experiments (1985a; 1986), “creates a strong presumption in favor of the Coase Theorem” (1985b, p. 1011). Nevertheless, most of the exchanges are not mutually advantageous. The controller sacrifices a part of her gain for fairness. Typically, whereas the controller, B, could obtain $12 (and A 0) by unilaterally choosing the number 0, A and B sign an agreement by which the controller chooses the number 1 (B earns $10, and A 4) – which is the maximum total payoff – and share this total gain equally ($7 for each), which means the controller sacrifices $5 (Hoffman and Spitzer 1982, p. 86 and 92). As Harrison and McKee (1985, p. 655) conclude, “the Coase Theorem is [therefore] behaviorally ‘right for the wrong reasons.’” In these experiments, subjects agree on the optimal issue, but only because one of them sacrifices a part of her gains. Why does this subject accept the exchange?

Admittedly, fairness is a classic result of bargaining experiments, the robustness of which has been confirmed in large measure by Hoffman and Spitzer’s (1982) own experiments. Fairness is here favored by public face-to-face negotiation; complete information (McKelvey and Page 2000 obtain less good results with incomplete information); and a random allocation of the right, without legitimacy and without insistence on its meaning (the controller obtains her individual maximum more often when the initial allocation of the right follows a preliminary game and when her authority is legitimated by the monitor (Hoffman and Spitzer 1985a), or when she can learn the meaning of her unilateral right (Harrison and McKee 1985)).

But fairness is here obtained by sacrifice. A likely explanation of this non-mutually advantageous exchange is that of a moral or social incentive raised by the experimental design and the instructions (Bertrand 2014). First, contracts are given to the subjects, from which they infer that the right is exchangeable and how they can exchange it. This amounts to morally authorizing the exchange regarding this externality, and even to encouraging it by implying that it is expected from participants that they use this possibility. Then, the list of payoffs in function of the number gives the subjects the monetary value of the externality, which impedes the perception of the moral problem of giving value to something that should not have one (Kelman 1985, pp. 1038–1039). Finally, instructions are not indifferent regarding the question of exchange, and take care, under the shelter of neutrality, to avoid any vocabulary concerning externalities, constraint, or causality.

Confronted with some of these criticisms, Coursey et al. (1987) built an experiment precisely to test the existence of a moral ban on bargaining in the presence of externalities. They replicate the 1982 experiment, but with the possibility of physical discomfort for the subject who suffers from the externality (keeping in his mouth a very bitter-tasting liquid for 20s). This allegedly allows testing the moral ban against negotiating over something that insults dignity, what they call the “dignity hypothesis.” Pairs of subjects reached the optimal result, and authors hence reject this hypothesis. But here again, the sources of bargaining breakdown are in the most part avoided (Bertrand 2014, pp. 454–456). In particular, and paradoxically in comparison to what the authors wanted to test, the moral obstacles to the exchange are at least partially removed: the contract is provided, and the instructions and consent insist on the possibility of exchange and the possibility of A taking the liquid. Individuals placed in such a situation know what they have to do to please the monitor, or may simply internalize her authority (Milgram 1974).

Conclusion

Since the Coase theorem is circular, what is tested is precisely whether mutually advantageous bargains are indeed realized. Such realization may encounter three obstacles: (1) transaction costs, (2) problems of agreement over the distribution of the surplus, and (3) moral or social prohibition of exchange. This entry has shown that the authors of these tests commonly appeal to transaction costs to legitimize every result as efficient. They thereby rationalize their observations by appeal to the assumption of efficiency rather than test its correspondence to the real world. And they underestimate the other obstacles.

What are the lessons for the theorem? As stressed by Medema, these studies enlighten us about “situational behavioral norms” (1997, p. 129) and the limits to the assumption of individual maximization, hence calling into question the behavioral assumptions that ground the theorem, and indeed the law and economics movement more generally. Cheung’s bees and Hoffman and Spitzer’s experiment bring to light that markets for what was seen as an “externality” can exist: the right over an “externality” is sold and bought. Two elements have nevertheless been overlooked by Coase (1960): first, the entitlements to which agents refer are not only determined by the law; second, impediments to bargaining other than transaction costs exist, these being moral or social obstacles to such an exchange.

Cross-References