The Problem of Unilateralism in Agency Theory: Towards a Bilateral Formulation

ABSTRACT: Some business ethicists view agency theory as a cautionary tale—a proof that it is impossible to carry out successful economic interactions in the absence of ethical behaviour. The cautionary-tale view presents a nuanced normative characterisation of agency, but its unilateral focus betrays a limited understanding of the structure of social interaction. This article moves beyond unilateralism by presenting a descriptive and normative argument for a bilateral cautionary-tale view. Specifically, we discuss hat swaps and role dualism in asymmetric-information principal-agent relationships and argue that the norm of reciprocity can function as a moral solution to agency risks in adverse-selection and moral-hazard problems. Our bilateral cautionary-tale formulation extends the normative boundaries of agency theory, while leaving the fundamental economic assumptions of agency theory intact.


INTRODUCTION
A GENCY THEORY IS A DOMINANT FIXTURE in studies of economic interaction and organisation. In essence, it is concerned with investigating "situations in which one individual (the agent) acts on behalf of another (the principal) and is supposed to advance the principal's goals" (Milgrom & Roberts, 1992: 170). Agency theory purports to enable economists to move beyond abstract theory and to discuss micro-level issues of internal management and organisation. The analytic and rigorous nature of agency theory arguably makes it an attractive tool for decision making in complex, practical situations. As a result, the vocabulary of principals and agents, the problems of incentives and information, and the challenges of monitoring and evaluating performance have been widely applied and debated in many scholarly fields.
Social science is historically intertwined with ethics, and ethicists have taken an interest in evaluating the application of agency theory to social life. Within the field of business ethics, agency theory is a popular target for theorists concerned with the supposed ethical deficits of modern capitalist societies. For some business ethicists-let's call them agency critics-agency theory embodies the assumptions that guide the ethics-minimising models of strategic management, finance, and corporate governance. For example, agency theory has been singled out as the theoretical and ideological bedrock of corporate cultures that foster large-scale scandals (Ghoshal, 2005;Khurana, 2007;Kulik, 2005). The dominance of agency-theoretic economic frameworks in business school curricula has also been criticised (e.g. Ferraro, Pfeffer, & Sutton, 2005;Ghoshal & Moran, 1996;Khurana, Nohria, & Penrice, 2005). More fundamentally, ethicists are concerned about the "dangerous and demeaning" game-theoretic underpinning of the principal-agent model, which falsely advocates a "monstrous" characterisation of human behaviour and rationality (Solomon 1999: 11). The spheres of agency theory and morality have thus been characterised by some business ethicists as mutually exclusive, and agency theory has been cast as the villain.
In contrast, others have explored whether agency theory can be used in a more progressive manner to inform normative considerations in economic and organisational interactions. Business ethicists in this camp-let's call them agency optimists-do not categorically reject agency theory, but often criticise and propose revisions to it (e.g. Boatright, 1999;Bosse & Phillips, 2016;Bowie & Freeman, 1992;Buchanan, 1996;Dees, 1992;Donaldson, 2012;Heath, 2009;Mitnick, 1992;Orts, 1998). Some agency optimists view agency theory as a cautionary tale (Heath, 2009)-a proof that it is impossible to carry out successful economic interactions in the absence of moral norms and principles. Let's call this the cautionary-tale view. Agency theory is thus drawn upon not as a straightforward positive characterisation of the nature of reality, but as an analytic tool that outlines a cautionary tale or an "instructive parable" (Noreen, 1988: 360) of what might happen in the absence of ethical behaviour. On this view, moral obligations arising out of agency interactions are unilateral-they involve internal constraints adopted by agents against posing risks to principals (Buchanan, 1996;Heath, 2006). Tapping into a broadly contractarian take on the relation between morality and economic interactions, this view characterises self-interested individuals as motivated to follow rules of morality and justice because each is vulnerable to the opportunism of others and each stands to benefit from cooperative arrangements (Gauthier, 1986;Hobbes, 1651Hobbes, /1994.
But although the cautionary-tale view presents a nuanced normative conception of agency, it is vulnerable to two criticisms. First, its unilateral focus betrays a descriptively limited understanding of the structure of social interaction in principal-agent relationships. Second, cautionary-tale theorists formulate moral obligations unilaterally: as internal constraints on the part of one player, the agent, against posing risks to another, the principal. But by attending to the moral obligations of only one player in a two-player principal-agent relationship, the cautionary-tale view overlooks the bilateral moral considerations relevant to such a social interaction. This two-part criticism makes up what we call the problem of unilateralism in agency theory.
In this article, we address the problem of unilateralism by presenting a descriptive and normative argument for a bilateral cautionary-tale view. Descriptively, in order to underscore the two-player orientation of agency interactions, we engage with the asymmetric-information stream of agency theory. Two bilateral features are discussed: (1) hat swaps, which involve agents and principals switching roles back and forth; and (2) role dualism, which involves two players posing agency risks to one another in their simultaneous roles as principals and agents. Hat swaps have received some attention in the business ethics literature, but remain underdeveloped. Role dualism has yet to be analysed by business ethicists. We refer to these descriptive features as the bilateralism thesis. Normatively, we argue that the bilateral structure of moral obligations and the norm of reciprocity deserve critical scrutiny in principal-agent relationships. Specifically, this article demonstrates that reciprocity can function as a moral solution to agency risks in bilateral adverse-selection and moral-hazard problems, alongside other organisational responses (e.g. incentive design and monitoring). Applied to organisational-level analyses, this appeal to reciprocity has attractive managerial implications and potential since the design of principal-agent relationships can foster the norm of reciprocity.
Prior cautionary-tale theorists have defined ethics as a response to agency risks (e.g. Buchanan, 1996;Heath, 2006), but have not engaged with the thesis of bilateralism or the problems of moral hazard and adverse selection. Other agency optimists have called for a bilateral moral evaluation (e.g. Dees, 1992;Donaldson, 2012) within broader critical debates on agency theory, but have not engaged with or developed a bilateral formulation. Some optimists have engaged with asymmetricinformation problems , but have done so unilaterally. Still others have reexamined the theory of the firm stream of agency theory in broadly bilateral terms-by highlighting the legal principles of authority, power, and hierarchy (Orts, 1998), and by revising the assumption of bounded self-interest in light of empirical research on reciprocity (Bosse & Phillips, 2016). In contrast, our novel bilateral formulation of the cautionary-tale view extends the normative boundaries of the asymmetric-information stream of agency theory, while leaving the fundamental economic assumptions of agency theory intact. Setting aside the interesting but contentious debates over the role of authority or power in economic debates of agency, we engage with the asymmetric-information stream of agency theory. By offering a socially informed understanding of principal-agent relationships, the bilateral cautionary-tale view underscores the structure of reciprocal, as opposed to merely unilateral, deontic constraints in social and economic agency relationships.
The article begins with an overview of the cautionary-tale view and its salient unilateral features. Next, we criticise descriptive unilateralism and introduce two bilateral principal-agent features: hat swaps and role dualism. Finally, we discuss the normative significance of the bilateral cautionary-tale view before considering some potential objections.

AGENCY THEORY AND UNILATERALISM
The literature on agency theory is vast and multidisciplinary. It is not possible to do justice to all agency approaches here or review the various controversies, misunderstandings, and abuses in the definitional literature. The following selective overview of some of agency theory's prominent ancestry (Mitnick, 1992) provides the relevant foundation for our purpose.
One long-standing interest in agency is closely connected with the theory of the firm. The theory of the firm literature involves distinct interpretations and approaches (e.g. transaction cost, contract, property, employment), but its essential concern is the proper role of firms in markets. Some agency theorists aim to draw on the theory of the firm as a positive approach for analysing economic activities within the firm (Alchian & Demsetz, 1972;Fama, 1980;Jensen & Meckling, 1976). A second stream of agency theory focuses on problems of asymmetric information and risk in market interactions, broadly construed (Arrow, 1963;Knight, 1921;Ross, 1973). In comparison to the largely empirical and non-mathematical agency literature on the theory of the firm, the asymmetric-information literature is often non-empirical and formal, and draws on game theory and probability. As prior business ethicists have acknowledged, the game-theoretic/asymmetric-information approach is distinct from and much broader than the theory of the firm stream, dealing with any interaction between two individuals, irrespective of organisational variables, whereby one tries to "influence" the actions of the other (Heath, 2009: 499;Mitnick, 1992: 79-81). These two streams of agency are in turn distinct from the legal approach to agency, which assigns agency a narrower meaning, informed by notions such as authority, delegation, and liability (Clarke, 1985;Orts, 1998).

The Cautionary-Tale View
Some business ethicists view agency theory as a dystopian "state of nature" thought experiment that enables us to formulate moral considerations for guiding social and economic interactions (Buchanan, 1996;Heath, 2006;Noreen, 1988). Agency theory is thus studied as a cautionary tale; this view does not draw on agency theory as a strictly positive theory, pace the theory of the firm stream of agency. This means that cautionary-tale theorists do not assert that individuals behave opportunistically, for example. Instead they ask: What if it were true that everyone behaved opportunistically and acted according to an instrumental model of rationality? The positive approach to agency theory is thus drawn upon as a conditional: If certain assumptions about opportunism, sequential rationality, and expected utility maximisation hold, then agency theory can act as a "critical-diagnostic" (Heath, 2009: 497) deductive tool that identifies the points at which economic interactions are vulnerable to breaking down in the absence of moral norms and principles. 1 To be clear, the cautionary-tale view is neither a positive theory nor a normative theory. Instead, this view draws upon agency theory in order to provide "a series of instructive parables that illustrate the adverse consequences on social and economic systems of unconstrained opportunistic behavior" (Noreen, 1988: 360). This instructive parable may (but need not) inform normative considerations in the principal-agent relationship. The instrumental (or "economic") model of rationality at the core of agency theory has long been used in political philosophy as a basis for the development of normative theory (e.g. Braybrooke, 1976;Gauthier, 1986;Heath, 1996). If we suppose that certain background assumptions of rational-choice theory about the world are true, then in the imaginary world of the thought experiment, the "nasty, brutish, and short" nature of social interactions is a proof of the need for normative principles and moral constraints (Hobbes, 1651(Hobbes, /1994. Agency theory's instrumental conception of rationality is thus used to construct a cautionary tale that allows us to "state with a great degree of precision what would happen in the absence of morality" (Heath, 2009: 518). For example, analysing prevalent agency risks can shed light on and inform the implicit and explicit moral codes and principles that guide bureaucratic organisations (Buchanan, 1996). According to the cautionary-tale view, agency theory is an analytic method that identifies "major stress lines" and potential future "cracks" within economic interactions; organisational ethical codes are then analysed "as the glue that (to a greater or lesser degree of success) holds things together" (Heath, 2009: 520;referring to Buchanan, 1996).

Salient Unilateral Features
Descriptively, principal-agent relationships are often characterised unilaterally: agents carry out certain tasks on behalf of principals, not the other way around. The unilateral focus is commonplace in foundational texts on agency theory (e.g. Eisenhardt, 1989;Fama, 1980;Jensen & Meckling, 1976;Milgrom & Roberts, 1992;Walsh & Steward, 1990). Here, the structure of interactions is "top-down," and "sanctions flow from the principal, who occupies a higher rank in the organisational hierarchy, down towards the agent, who occupies a subordinate role" (Heath, 2009: 516). Ethical principles in this top-down organisational setting express "commitments on the part of agents to reduce risks" that they impose to principals (Buchanan, 1996: 422). Some common agency risks are: inefficient use of resources (e.g. wastefulness), misappropriation of resources (e.g. embezzlement), passive opposition (e.g. stalling), and shirking (Buchanan, 1996: 425-426).
Given this unilateral stance, moral obligations in the cautionary tale view are formulated unilaterally on the part of only one of the players (the agent) in any given interaction. So, if player 1 (the agent) poses a risk of opportunism to player 2 (the principal), then cautionary tale theorists study the obligation to avoid opportunism, owed by player 1 to player 2. The normative essence of agency, according to this line of argument, involves abating the agency risks that are posed by agents to principals. Proponents thus characterise the cautionary-tale view simply as "the agency risk minimisation" theory of business ethics (Buchanan, 1996: 436;Heath, 2006: 23). The "function" of ethics here is to "reduce" agency risks, and moral obligations are a response to agency risks (Buchanan, 1996: 419). For example, loyalty to the chain of command constitutes a "central obligation" (Heath, 2006: 23) within firms, and the principle of stewardship is an "ethical commitment" which works to "counter pressures of self-interest" that might otherwise lead to behaviours such as embezzlement (Buchanan, 1996: 427). The analysis of agency risks and the related moral obligations is a key to understanding and informing many real-world moral codes and norms that implicitly or explicitly structure organisational interactions.
In this manner, the structure of agency risks and moral obligations in the cautionarytale view is unilateral. In relation to player 1's unilateral obligation of loyalty, for example, there is no discussion of player 2's correlative obligations to provide, say, job security. Here, moral obligations are "deontic" (Heath, 2006: 23) constraints that guide what agents ought to do-what they are morally required, forbidden, or permitted to do-when engaging with principals. If agents remain loyal or avoid misappropriation of resources, even when they can renege and cheat without getting caught, then the deontic constraints are "internal" (Buchanan, 1996: 423). Internal constraints, or ethical commitments, contrast with external constraints such as pecuniary incentives in the form of bonuses or fines.
What is the content of moral obligations, according to the cautionary-tale approach? This article does not comment on the normative content or the "inner morality" (Fuller, 1964) of agency interactions, nor does it take a position regarding the justifiability of moral obligations in the cautionary-tale approach. In other words, we do not identify or catalogue specific obligations or provide an account of why they arise. That debate is important, but due to a number of established difficulties and hurdles (see Donaldson, 2012;andBoatright's 2007 critique of van Oosterhout, Heugens, &Kaptein, 2006), it merits separate attention. Our purpose is more minimalist and pertains to the structure within which norms and obligations arise and are formulated; specifically, we discuss in this article who owes obligations to whom.
To recap, according to the cautionary-tale view, agency relationships are unilateral and one-way, with risks posed from only one individual, the agent, to the other, the principal. At the same time, unilateralism is a normative feature of the cautionary-tale view. Moral obligations arising in these agency interactions involve unilateral commitments on the part of agents against posing risks to principals. This unilateral stance is vulnerable to a descriptive and normative criticism. In the next section, we present a descriptive argument for a bilateral cautionary-tale view. We then turn to discussing the normative features of the bilateral cautionary-tale view in the section that follows.

A DESCRIPTIVE ARGUMENT FOR BILATERALISM
Recall that the cautionary-tale view is formulated as a conditional-that is, as an instructive parable, as opposed to a factual statement about the nature of reality. This diagnostic tool thus outlines the states of affairs that would come about if certain positive propositions were to hold. The stream of agency theory we draw on in order to criticise the unilateral formulation of the cautionary-tale view is the game-theoretic, asymmetric-information approach. We believe the two-player orientation of the asymmetric-information approach enables it to nicely exhibit certain bilateral features. Methodologically, given its abstract, non-empirical features, the asymmetric-information approach is well suited for testing out revisions and theoretical boundary-setting exercises in agency theory (Jensen, 1983). 2 To set the stage for our argument and avoid definitional confusion, we provide a basic definitional discussion of the asymmetric-information stream of agency theory.

The Asymmetric-Information Approach to Agency
According to the asymmetric-information approach to agency theory (Arrow, 1985;Ross, 1973), agency problems involve the microeconomic game-theoretic analysis of interactions between principals and agents. The core element of this definition is the presence of two or more individuals. 3 Here, the informed party is the agent, the party that lacks critically relevant information is the principal, and agents act on behalf of principals. The problem of agency becomes theoretically interesting in this context in the presence of uncertainty, information asymmetry, goal incongruity, and agent risk aversion. Agency theory highlights fundamental problems in economic interactions that are not explicitly addressed by other theoretical approaches: given the assumptions of asymmetric information, divergent objectives, and rampant opportunism, agency theory elegantly depicts a world in which perfect policing and governance are impossible, despite the presence of intricate monitoring measures and incentive design. The two parties agree on a fee schedule to be paid to the agent, and any choice made by the agent affects the welfare of both the agent and the principal. Contractual remedies can be negotiated before principals and agents join a given agreement (ex ante) and after they have joined a contract (ex post). Ex ante remedies address adverse selection problems, while ex post remedies respond to moral hazard problems.
For example, consider a tenancy interaction in the commercial leasing market and in particular the operation of an indoor shopping mall. Two parties within this sector are the mall managers that lease, administer, and maintain the space, and the retailers that lease the units. Consider the interaction from the viewpoint of mall management. Prior to signing the lease, managers are the uninformed party, and the prospective retailer has access to private information. Adverse-selection problems may arise: managers may struggle to evaluate the suitability of retailers for a unit, given the limited information regarding the retailer's level of experience, financial condition, and interest in remaining for the long term. If the lease contract is signed, moral-hazard problems may arise: Without regular inspections, managers may find it difficult to observe or evaluate how well the retailer adheres to the mall's many operating standards (e.g. regarding cleaning, music, signage, storefront, and in-store displays) which contribute to the mall's profitability goals. The mall managers depend on this division of labour and would not be able to operate every retail store and maintain comparable profitability thresholds on their own. Here, the mall managers can be seen as seeking to influence the behaviour of retailers, for example, through specific incentives or monitoring mechanisms. In both moral-hazard and adverseselection problems, the risks of agency are posed unilaterally: the agent (the retailer) poses agency risks to the principal (the mall managers).
It may be objected that this example is not an illustration of a principal-agent relationship. Rather, it appears to be a contractual transaction that has an asymmetricinformation element. Just because the two parties make promises to one another, the objection goes, it does not mean that the two are in an agency relationship. According to a legal understanding of agency, for example, this definitional stance conflicts with the essential firm-level, legal relationships of authority and hierarchy. Theory of the firm agency theorists may also question whether the mall example counts as an instance of proper delegation. The worry here may be that the boundaries of this contractual interaction have been overly extended so as to strip it of agency theoretic significance.
In response, we remind readers that the theoretical boundaries of agency relations are not one and the same across the law of agency, theory of the firm, and asymmetricinformation agency streams. Further, the proper theoretical scope and nature of agency relationships is an ongoing, contentious matter of debate in the multi-disciplinary literature on agency theory. In the asymmetric-information agency stream, the goals or interests of two parties are incongruent, and interactions occur under conditions of uncertainty and information asymmetry. The informed party is the agent, the uniformed party is the principal, and the principal seeks to influence the agent's actions. It is not necessary that one party employ the other, and no assumptions are made regarding the structure of authority or the presence of discretion and fiduciary obligations (Laffont and Martimort, 2002;Salanié, 1997). Although the asymmetricinformation approach does not focus on specific organisational structures and addresses decision-theoretic variables instead, it is an established, valuable theoretical tool for representing the state of technical competence in formalising agency problems. It is within this analytic lens that we suggest business ethicists observe the bilateral features of agency. Whether this suggestion is applicable or fruitful in other streams of agency is beyond the scope of this article.
Given this understanding of asymmetric-information principal-agent relationships, in the remainder of this section we discuss two bilateral features of agency relationships: hat swaps and role dualism. As we argue, the roles that players fulfill as principals and agents deserve critical scrutiny from business ethicists.

Hat Swaps
The asymmetric-information approach to agency defines the uninformed party as the principal and the informed party as the agent. However, while the threat of asymmetric information may be posed by the agent to the principal, the agency roles of individuals can shift within a single relationship.
Consider the commercial leasing interaction above, but from the point of view of the retailer as opposed to the mall managers. Prior to the start of a leasing arrangement, the retailer is the uninformed party, whereas the mall managers have access to private information about the units. Adverse-selection problems arise if the mall misrepresents the physical condition of the unit during the pre-contract viewing process. Here, the risk is posed by the mall's management (agent) to the retailer (principal), who does not have access to private information regarding the true state of the property. Moral-hazard problems may also arise if the managers neglect to carry out contracted obligations of care after the retailer moves in (e.g. tardy attendance to routine repairs), or impose unexpected assessments for major repairs or maintenance, such as repaving the parking lots. Here, the agent (the mall managers) may be seen as acting on behalf of the principal (the retailer), in so far as the retailer would not be able to operate at the same threshold of profitability on its own in the absence of this division of labour. In both moral-hazard and adverse-selection problems, the risks of agency are posed unilaterally-the agent (the mall's management) poses agency risks to the principal (the retailer).
Asymmetric information thus arises on both sides of a commercial leasing relationship: agency risks are posed by the retailer to the mall's management, and also by the mall's management to the retailer, each of whom may wear the principal's hat or the agent's hat in different domains-that is, in different aspects or facets-of the relationship. For example, parties to a commercial lease will find themselves interacting in various domains, such as finances, major maintenance, day-to-day operations, seasonal promotions, etc. Each party may be informed in a given realm, and pose risks to the uninformed other. The other, at the same time, might possess private information about some other domain, and so pose a different set of risks back to the first party. This involves recognising that an agency contract is a complex and multifaceted social interaction. A single contract can include differing sets of risks arising from different domains, posed from player 1 to player 2 and from player 2 to player 1. Since both players wear the principal's hat in different domains in this interaction, this relationship is not unilateral.
The phenomenon of hat swaps remains underdeveloped in the literature on business ethics. 4 One source of confusion may be definitional. The relation between agency and "employment" corresponds to a key distinction between the current asymmetric information and theory of the firm streams of agency theory. Some in the latter stream take it for granted that principals "employ" (e.g. Bosse & Phillips, 2016) agents in agency relations, whereas the former defines agency relations separately from individual organisational functions (Salanié, 1997). If the roles of principals and employers were necessarily one and the same, then the notion of hat swaps could not get off the ground. However, since we engage with the asymmetric-information stream of agency, the roles of principals and agents are interchangeable between two players. By moving beyond the theory of the firm stream, our socially informed reexamination of agency theory has the potential to be applied to different kinds of organisational/structural settings and to diverse employment arrangements in future research.
It may be objected that the relation of agency actually remains unilateral in the commercial leasing example. In both the moral-hazard and adverse-selection problems, prior to and after signing the lease agreement, the agent by definition poses the risk because she is the informed player in an asymmetric-information scenario. According to this view, agency in this example is not bilateral, but rather a series of interactions in different unilateral domains. This is a legitimate, technical stance, but it overlooks an essential social feature of the relationship. If player 1 poses a unilateral risk to player 2 in one domain, and player 2 poses a unilateral risk to player 1 in another domain within the same two-person relationship, arguably this interaction is no longer unilateral, even when the two-player game formally ends after a single iteration. In our example, the implication of the hat swap is that both the mall management and the retailers have a chance to play the role of agent and pose risks to the other.
The implications of hat swaps deserve attention, especially given the normative implications of role shifts among agents and principals. A number of agency optimists have been calling for an analysis of the moral obligations of principals (e.g. Dees, 1992;Donaldson, 2012) in principal-agent relationships. According to the cautionary-tale view, moral obligations are by definition formulated on the part of agents. However, although moral obligations would be assigned to agents in the cautionary-tale view, hat swaps provide a necessary step for considering bilateral moral considerations on the part of both players in a given principal-agent relations. We return to discussing these moral considerations later in the article.

Role Dualism
A second descriptive bilateral feature of principal-agent relationships is what we call role dualism. Here, each player is both principal and agent in the same agency relationship and in the same domain. In contrast, in hat swaps, each player is both principal and agent in relation to the same counterpart but in different domains. In the commercial leasing example, the retailer may act opportunistically regarding payment interactions (by stretching the payment terms), and the mall management may act opportunistically regarding maintenance interactions (by being slow to repair things). The retailer and mall manager swap hats here. The retailer wears the agent's hat during the payment interaction, but wears the principal's hat during the maintenance interaction. Now consider a variation of the commercial leasing example-a joint project-to demonstrate role dualism. A sports retailer in the mall wants to put in a parkour facility, which is financially beyond his or her reach. The retailer presents a formal proposal to the mall managers. The managers believe this facility could be a real draw, and so the two contemplate a joint agreement in which they will share the construction costs. Any number of scenarios could result, but consider just this one: Might the retailer intentionally underestimate costs, and when the project goes over budget try to foist the remaining costs onto the mall management? Alternatively, the mall managers might intentionally do the same, such that once the project is under way they will "plead poverty" and try to foist the remaining costs onto the retailer. Both parties have to consider the possibility of finding themselves with a half-completed project, substantial sunk costs, and an ugly stand off. Either might then have to pony up the funds that the other party is refusing to pay, or let the unfinished project sit idle while enduring a protracted legal battle with an uncertain outcome. In this adverse-selection scenario, both parties play the role of agents with regard to construction costs. They are both defined as agents in the information-asymmetry approach, because each is the informed party who carries out tasks on behalf of the other uninformed party.
Role dualism has received attention in the economics literature as the "double-sided" or "two-sided" agency problem. As in the parkour example, double-sided adverse selection can arise for both players prior to agreeing to a contract (Soberman, 1997). Moral hazard can also become double-sided when each player takes actions unobserved by the other, and these actions affect the pay-offs for both players (Rasmusen, 1994). In both agency problems, the output of the principal-agent interaction depends on the combined efforts of two input providers-the principal and the agent.
In sum, the descriptive account of the cautionary-tale view must be amended in light of two descriptive features of asymmetric-information principal-agent relationships: a) hat swaps, in which each player might play the role of agent and principal in different domains of the agency relationship; and b) role dualism, in which moral-hazard and adverse-selection problems are posed by each player to the other in the same domain. Together, these two descriptive features establish a socially informed understanding of the asymmetric-information theory of agency that is structurally bilateral. We refer to this argument as the bilateralism thesis.
Recall that we set out to address the problem of unilateralism through a descriptive and normative argument. A descriptive argument for bilateralism has been presented in this section. We now turn to the second, normative argument.

A NORMATIVE ARGUMENT FOR BILATERALISM
As established earlier in the article, the standard cautionary-tale view is in effect an agency risk reduction theory of business ethics. Its essential features are loyalty and moral commitment to a hierarchical chain of command. To be sure, cautionarytale theorists do not claim to address more than a small part of a complete conception of morality through agency theory (Buchanan, 1996;Heath, 2006). We may wonder, however, whether the cautionary tale view can provide additional moral insight beyond a unilateral conception of morality.
Alongside a unilateral descriptive stance, moral obligations in the cautionary-tale view have so far been formulated unilaterally-as commitments on the part of one party, the agent, against posing agency risks to the other, the principal. According to the bilateralism thesis (discussed in the prior section), a socially informed understanding of the structure of asymmetric-information principal-agent relationships must acknowledge agency hat swaps and role dualism. Moral obligations arising in asymmetric-information principal-agent relationships can thus be formulated not just unilaterally, but also bilaterally, as responses on the part of both players to agency risks. If one is convinced about the theoretical attraction of the cautionarytale view and its parsimonious engagement with an instrumental model of rationality, then there is reason to welcome such a bilateral reexamination. As we argue next, within this bilateral structure, the norm of reciprocity, conceived as the practice of exchanging with others for mutual benefit, can operate as a solution to agency risks in moral hazard and adverse selection problems.

Moral Hazard and Reciprocity
Consider the mall commercial leasing example again. Maintaining cleanliness standards helps keep the space free from insects, mould, and common allergens such as dust. To save time and effort, some retailers may be tempted to cut corners in upholding these standards. This constitutes a moral-hazard risk posed from the retailer to the mall. According to the cautionary-tale approach, moral obligation on the part of managers would involve adopting internal, deontic constraints and adhering to cleanliness standards irrespective of the presence of incentives or monitoring mechanisms. At the same time, the mall management could pose an agency risk to the retailers if the mall management does not maintain cleanliness standards in common spaces such as public bathrooms. In response to this risk, ethical behaviour would involve deontic constraints on the part of the mall's managers to maintain the mall's common areas. This example involves role dualism. Now let's embed this interaction in a longer-term, repeat-game structure. Suppose mall managers and retailers sign a five-year contract. In this two-sided agency setting, risks flow bilaterally between the mall managers and retailers. Ethical behaviour on the part of each player would involve meeting cleanliness standards, though early on in the relationship effort levels and quality of cleaning may not be readily observable by the other. At the end of year 1, however, suppose both players' performance is more readily observable, since the physical space deteriorates with the passage of time. Mall managers may not keep their end of the agreement in year 2 if retailers have not kept theirs in year 1. In turn, mall managers may have reasons to work harder to keep their end of the agreement in year 2, if retailers keep their end of the bargain in year 1. The norm of reciprocity in this context involves the retailers reciprocating the cooperative gesture of mall managers when they provide a clean space, and vice versa, even though both parties could have cut corners in ways that would be difficult for the other to perceive or evaluate in the short term. Reciprocity in repeat games can thus function as a solution to agency risks, since each player knows that a cooperative gesture can elicit in-kind reciprocal reactions from other players in the future.
In this example, bilateral moral hazard problems arose with respect to the same domain-cleanliness standards. Now consider that the mall managers and retailers may be, at the same time, in other moral hazard problems in different domains. For example, retailers may have access to private information regarding their ability to fulfill their financial commitments, as well as their true efforts in promoting mallwide seasonal promotions. Thus, even without a multi-year relationship, the social implications of a multitude of risks and moral-hazard problems is that each player's deontic commitments in one domain might be reciprocated by the other player in a different domain.

Adverse Selection and Reciprocity
Agency risks and moral obligations in adverse-selection scenarios also flow bilaterally. Consider the joint project parkour example. The mall managers and retailers are each uninformed with respect to one domain-for example, the ability of the other to afford the construction costs-prior to entering the agency interaction. In this double-sided adverse-selection setting, in response to agency risks, each party might adopt internal constraints against underestimating or misrepresenting construction costs. Recognising the presence of this double-sided structure, mall managers and retailers may conceivably appeal to a quid-pro-quo structure-they may each adopt moral constraints. In this context, reciprocity can operate as a solution to agency risks in an adverse-selection problem. This can happen when in-kind reciprocal reactions are carried out within the same one-shot interaction in different domains, or when such interactions occur over repeat interactions. This is because alongside role dualism with respect to parkour construction costs, the mall managers and retailers take part in a series of alternative interactions in other domains (e.g. unexpected large-scale maintenance). In anticipation of a broadly cooperative social relationship, then, a player in one principal-agent domain may adopt reciprocal moral constraints against posing adverse-selection risks, and these constraints may be reciprocated by the other player in a different domain.
In sum, embedding transactions in implicit or explicit single-iteration or long-term structures of reciprocity can serve as a solution to agency risks in adverse-selection and moral-hazard problems. Not all principal-agent relationships are bilateral, but in those that are (or can be designed to be), moral solutions and the norm of reciprocity merits attention alongside other agency theoretic mechanisms. 5 Our discussion of reciprocity as a moral solution might seem overly optimistic regarding the players' trust in one another to reciprocate cooperative commitments. It might appear that we believe players will necessarily respond to cooperative gestures in kind, as opposed to taking advantage of others. But in fact we are making a different claim: if a player knows about the possibility of reciprocity, whether through multiple iterations of interaction or through multiple domains of cooperative interactions, then agents might reciprocate through unilateral deontic constraints within a quid-pro-quo cooperative structure. This is not to say that people can or will in fact act in this manner. Rather, players may choose to adopt deontic constraints that enable cooperation, even though they could choose to act opportunistically; each might recognise their vulnerability to the other's reneging and also the benefits of positive reciprocation (see Gauthier, 1986 for this line of argument outside the purview of agency theory). If actors have an altruistic, socially oriented affinity for trust and justice, then of course the reciprocal structure attains even further strength and robustness, but the point is that such enlightened assumptions are not necessary for the cautionary-tale argument to work.
We do not make the (mistaken) assumption that moral solutions to agency problems are always beneficial, or that they are necessarily more effective than other organisational solutions. We also do not claim that moral solutions are ever a complete or stand-alone solution to agency problems. After all, the reason agency theory is theoretically interesting is that it elegantly depicts a world where perfect policing and governance are impossible, despite the presence of intricate solutions. Norms do merit more attention than many business ethicists have thus far acknowledged, however, since they operate and interact with other organisational mechanisms like monitoring and incentive design. In adverse-selection problems, standard organisational solutions include signaling and screening mechanisms; in moral-hazard problems, monitoring or incentives are common organisational solutions (Rasmusen, 1994). Moral solutions may operate in conjunction with these existing mechanisms. In the mall example, reciprocity would work in conjunction with incentive arrangements that give a break on rent in exchange for maintaining high cleaning standards, for instance.
While reciprocity is an important, potential moral solution, it also has certain weaknesses. Crucially, the quid pro quo structure of bilateral obligations in asymmetric-information principal-agent relationships also leaves open the possibility for individuals to reciprocate negatively. Negative reciprocity involves neglectful behaviour in return for (or in revenge for) uncooperative behaviour. Players who perceive or experience cooperative treatment in a given principalagent interaction may reward this treatment through positively reciprocal behaviours; players who perceive or experience non-cooperative treatment in a given interaction often punish that behaviour through negatively reciprocal behaviours, even though this poses costs to themselves (Bosse, Phillips, & Harrison, 2009). If the norm of negative reciprocity kicks in, this can exacerbate agency risks and make the agency problem doubly complicated.

Normative Cash-Value
To recap, according to the cautionary-tale view, agency theory is an instructive parable that can inform ethical principles and norms as a response to agency risk. However, the cautionary-tale view is vulnerable to two criticisms. First, a unilateral focus betrays a descriptively limited understanding of the structure of social interaction in principal-agent relationships. Second, the unilateral moral formulation of obligations overlooks bilateralism and the norm of reciprocity in asymmetricinformation principal-agent relationships. As we have demonstrated, the norm of reciprocity can operate as a moral solution and a response to agency risks in moral-hazard and adverse-selection problems.
Conceiving of reciprocity as a bilateral response to agency risks is a novel feature of our formulation of the cautionary-tale view. Previously, the cautionary-tale view has defined ethics in agency theory as a response to agency risks, but has not engaged with the asymmetric-information features of moral hazard and adverse selection or with the bilateralism thesis. Outside the cautionary-tale view,  has formulated moral solutions to moral-hazard and adverse-selection problems, but he does not engage with the bilateralism thesis, and so discusses only those moral solutions that engage one player in a given two-player interaction. Descriptively, our bilateral formulation of the cautionary-tale view can capture the socially relevant features of hat swaps and role dualism, which are underdeveloped in the business ethics literature. Normatively, the shift to bilateralism means analysing not just one set of moral constraints on the part of one of the two players in a given interaction. The socially informed bilateral stance can capture deontic constraints on the part of both players in a given principal-agent relationship. This opens the possibility for the norm of reciprocity in single or repeat game interactions. Note that although this normative stance moves beyond loyalty and commitment to the chain of command, the implication of this move is minimalist. The claim is not that the bilateral cautionary-tale view necessarily serves purposes that go beyond promoting cooperative, mutually beneficial relationships. Rather, by offering a socially informed understanding of principal-agent relationships, this view can underscore the structure of reciprocal, in addition to merely unilateral, deontic constraints in agency relationships.
Agency optimists have been making gestures towards a bilateral formulation of agency theory for some time. Outside the cautionary-tale approach, Perrow (1986) has criticised agency theory for neglecting the risks posed by principals (e.g. lying about profit levels, failing to disclose threats of lost business). In response to such risks, Dees (1992) and Donaldson (2012) have argued that moral obligations should flow from principals and agents, though a bilateral argument for supporting this stance has been previously unavailable. Orts (1998) has insightfully analysed the problems of shirking and sharking by drawing on the legal constructs of authority, power, and hierarchy to inform the theory of the firm stream of agency. Our bilateral cautionary-tale view enables us to discuss descriptive and normative features of bilateralism, and to demonstrate the role of reciprocity as a possible solution to agency risks in the asymmetric-information stream of agency theory, without engaging with debates surrounding the role of authority in the theory of the firm or the legal stream of agency theory.
Other agency optimists have successfully argued for a role for reciprocity in agency theory (Bosse & Phillips, 2016), but this is accomplished by revising a key assumption underlying agency theory-the assumption of bounded selfinterest. While plenty of empirical evidence indicates the presence of the norm of reciprocity as an underlying motivational assumption, 6 and while we applaud these developments, we do not need to draw on these findings. This is because the cautionary-tale view already acknowledges that bounded self-interest is not a necessary assumption in agency theory. According to this view, standard assumptions of agency theory are a hypothetical possibility in an instructive parable, as opposed to a positive statement about the nature of reality. Nevertheless, by drawing attention to the bilateralism thesis, by arguing for enhanced social investigation of the descriptive features of agency theory, and by extending the normative boundaries of principal-agent relationships, this article carries out a broadly progressive, agency-optimist goal.
Three outstanding objections now require attention, and we turn to these in the final subsection of the article.

Objections
The first objection pertains to the relevance of reciprocity to principal-agent relationships. According to this objection, each player in a reciprocal two-person interaction cooperates with the expectation that doing so might induce the other player to cooperate again in the next round of interactions. Here, the person to whom we supply a benefit is the same person from whom we expect to receive benefits. 7 Moreover, according to this view, benefits from others are received within the same domain in which benefits are supplied. For example, in an intra-firm setting, internal constraints of loyalty on the part of player 1 (the employee) may be met by correlative obligations of loyalty on the part of player 2 (the owner) regarding job security. These two obligations occur in the same domain and they involve the same two players. Understood in this narrow manner, reciprocity seems not to apply to many instances of hat swaps and role dualism, especially in single-iteration games.
In response, the notion of reciprocity we have in mind is broader and recognises the cooperative effect of explicit and implicit structures and social features surrounding principal-agent relationships. For example, internal constraints on the part of player 1 (the employee) against embezzling funds from a local franchise branch may be met by obligations of player 2 (the head office HR department) regarding the effective provision of the company's investments in a pension plan. In this case, obligations take place in different domains. Furthermore, the employee poses a risk most directly to the local franchise owner, while a different entity (the head office HR department) is the reciprocating party. The pension case is an acceptable example of reciprocity in our account, because a broader notion of reciprocity permits player 1's moral obligation to player 2 in domain X to be reciprocated by player 3's obligation to player 1 in domain Y so long as players 1, 2, and 3 are implanted in a broadly cooperative setting (e.g. due to organisational design involving culture/identity, or temporal design involving repeat games). This holds even in single-iteration interactions, because a principal and agent are likely to be related and involved in other social interactions prior to, after, or at the same time as their one-shot interaction with one another.
A second, and potentially devastating, challenge to this article's argument is as follows: some may argue categorically against drawing on principal-agent theory to explain ethics and moral obligations in institutional or corporate life. According to this criticism, agency theory is incompatible with an adequate theory of moral responsibility. Layers of institutional, interpersonal, social, and economic dynamics are left out of a characterisation of ethics that is informed by agency theory. Thus, given the dominance of agency theory in applied fields such as corporate governance, relying on the theory to learn about ethics in organisations is arguably dangerous.
In response, cautionary-tale theorists can point out that the principal-agent model illuminates only certain, limited aspects of social interactions. The agency approach must then be supplemented with an analysis of the special and general obligations that individuals have beyond their agency obligations, as well as their explicit role-related, personal, and institutional moral obligations, and the legitimate expectations arising from their various commitments. Thus, although moral observations regarding principal-agent relations fall far short of constituting a complete account of ethics, they are nonetheless an important part of the multifaceted set of ethically relevant considerations in social interactions (Buchanan, 1996). This article has argued that this part can be extended to include a bilateral as opposed to a unilateral lens. Crucially, we have steered clear of debates about the content of moral obligations, and have focused instead on the contours or structure of ethics in agency theory. This minimalist structural extension from unilateralism to bilateralism is morally significant, since it involves encompassing not just loyalty but also reciprocity as an essential deontic constraint that can be informed by the cautionary-tale view of agency theory.
A final challenge remains. In recent years, a slew of social-scientific research outside the field of business ethics has revealed empirical evidence indicating the presence of the norm of reciprocity as an underlying motivational assumption (see note 6). In response to this research, economists have developed more advanced theories that are superior to agency theory in incorporating social norms and modelling deontic significance. For example, mechanism theory treats institutions as games designed to evoke particular kinds of play, e.g. honest reporting, audit efforts, cooperation in teams, etc. In the face of the state of the art of empirical and game-theoretic literature, why bother reexamining the cautionary-tale view?
While this is an important challenge, it is not a reason not to engage with agency theory. This article's constructive treatment and evaluation of agency theory is important and pertinent, not because agency is the newest game-theoretic tool on offer by economists, or because business ethicists have not criticised agency theory enough already; rather, we work with the cautionary-tale view because agency theory continues to be among the most dominant and widely used theories in the study of organisation and management. Unlike the business ethics critics who opt to reject agency theory outright, this article accepts and progressively engages with this reality.