Chapter 5 Supply Chain Theories and Risk
5.2 Principal-agent Theory (P/A)
The game theoretic origins of agency theory can, according to Eisenhardt (1989a), be traced back to the 1960’s and early 1970’s when economists explored risk sharing among individuals or groups. The core of agency problems is summarized in Fama & Jensen (1983a):
“Agency problems arise because contracts are not costlessly written and enforced. Agency costs include the costs of structuring, monitoring, and bonding a set of contracts among agents with conflicting interests, plus the residual loss incurred because the cost of full enforcements of contracts exceed the benefits.” (p. 327).
Arguing that activities can be performed under a variety of organizational forms and claiming there is competition among organizational forms for survival in any activity, they continue:
“Absent fiat, the form of organization that survives in an activity is the one that delivers the product demanded by customers at the lowest price while still covering costs.” (p. 327).
In Fama & Jensen (1983b) they continue:
“An organization is the nexus of contracts, written and unwritten, among owners of factors of production and customers. … The central contracts in any organization specify (1) the nature of residual claims and (2) the allocation of the steps in the decision process among agents. These contracts distinguish organizations from each others and explain why specific organizational forms survive.” (p. 302).
An important construct, residual risk is described as:
“The residual risk – the risk of the difference between stochastic inflows of resources and promised payments to agents – is borne by those who contract for the rights to net cash flows. We call these agents the residual claimants or residual risk bearers.” (p. 302).
The continued existence of the company is dependent on making the right decisions:
“Organizational survival involves a balance of the costs of alternative decision systems and systems for allocating residual risk against the benefits.” (p. 307).
Fama & Jensen argue that the control of agency problems is important when decision makers initiating and implementing important decisions are not the major residual claimants. To effectively control the agency problems the four phases in decision making (initiation,
ratification, implementation, and monitoring) is proposed split into two terms: decision management (initiation & ratification) and decision control (implementation & monitoring) each performed by a separate group of individuals in the organization. The authors then investigate under which circumstances the three terms residual risk bearing, decision management, and decision control should be separated and when it should be performed by the same agent. These fundamental issues constituted a basis for the development of the theoretical framework.
Two Streams of Research and Theory
Agency theory is usually divided into two streams: “positive agency theory” and “principal-agent theory” (Eisenhardt, 1989a). The two streams share common unit of analysis (the contract) and assumptions, but where the former relies more on verbal expositions, the latter is more concerned with mathematical techniques. According to Eisenhardt (1989a):
“…positivist researchers have focused on identifying situations in which the principal and agent are likely to have conflicting goals and then describing the governance mechanisms that limit the agent’s self-serving behaviour.” (p. 59).
The principal-agent researchers are more concerned with a general theory of the agency relationship – and are interested in applying the formal theory to other areas after careful specification of assumptions. However, these two streams should not be perceived as conflicting, but as complementing each other:
“…positivist theory identifies various contract alternatives, and principal-agent theory indicates which contract is the most efficient under varying levels of output uncertainty, risk aversion, information asymmetry, and other variables…” (p. 60)
As the purpose of this subchapter is to create an understanding of the theory, the application oriented “positivist” stream is excluded from the following description and evaluation.
The Basic P/A Models
The basic model has two players, the principal who wants a task done, and an agent who can perform it. The theory is concerned with resolving two problems that occur in agency relationships: the agency problem and the risk sharing problem. The agency problem arises when there is goal conflict between principal and agent and it is difficult or expensive for the principal to verify the actions of the agent. The risk sharing problem arises when the principal and the agent have different attitudes toward risk. Due to this difference the principal and the agent may prefer different actions. The agency problem is further developed into the moral hazard problem, the adverse selection problem and the signaling problem37:
¾ The moral hazard problem describes the situation where the principal is unable to verify the agents level of effort, and therefore will have to design a contract to induce the agent
37 Other categorizations have been suggested, see e.g. Schuster (1990).
to select an appropriate (to the principal) level of effort (van Ackere, 1993). The contract type will depend on various parameters such as the risk attitudes of the principal and the agent.
¾ The adverse selection problem can be perceived as a special case of the moral hazard problem (van Ackere, 1993). The model now contains more agents, each with a different level of ability. In this model, rather than attempting to induce the appropriate level of effort from a specific agent, the principal tries to select an agent with the appropriate level of effort. The design of the contract will thereby aim at getting the right agent to accept the contract.
¾ Finally, a signaling problem occurs when an agent has private information he wants to convey to the principal.
Since the unit of analysis is the contract between the two parties, the objective of the theory is to identify the most efficient contract, or relationship, between the two parties, given a set of assumptions. The assumptions in the theory are about people (e.g. self-interest & bounded rationality), organizations (e.g. goal conflict among members), and information (e.g.
information is a commodity). In its most basic form, the decision to be made is whether the contract is to be based on outcome (commissions, stock options, transfer of property rights, market governance) or behavior (salaries, hierarchical governance). The assumptions are shown in the table below.
Table 5-2: Comparison of assumptions (Agency vs. Organizational perspectives)38
Perspective
Assumption Political Contingency Organization
Control Transaction
Cost Agency
Self-interest X X X
Goal conflict X X X
Bounded rationality X X X X
Information asymmetry X X X
Preeminence of efficiency X X X X
Risk aversion X
Information as a commodity X
Besides outlining the assumptions Eisenhardt (1989a) develops propositions from the literature reviewed. This introduces new variables to the model, e.g. length of relationship, measurability of outcome and task programmability (see Table 5-3 below). Depending on the specific agency problem variables are in- or excluded in the model and used as contingency variables or assumptions/preconditions.
38 Source: Table 2 in Eisenhardt (1989a), p. 63.
Table 5-3: Variables in Agency Models39 Variable Value Comment
Outcome-based (commissions, stock options, transfer of property rights & market governance)
Contract type
(output from the models)
Behavior-based (salaries &
hierarchical governance)
Low Goal-conflict
High Assumption in the basic agency model Yes Assumption in the basic agency model Information asymmetry
No Short Relation length
Long
Risk averse Assumption in the basic agency model Risk neutral
Risk attitude (agent)
Risk taker Risk averse
Risk neutral Assumption in the basic agency model Risk attitude (principal)
Risk taker
Low Task-programmability
High Low Outcome-measurability
High Low Outcome uncertainty
High
From this portfolio of variables a number of applications have emerged. According to van Ackere (1993) agency theory has been applied to e.g. accounting, industrial organization, and marketing. In Melnyk, Stewart, & Swink (2004) it is argued that agency theory has been used within OM to study as diverse subjects as decentralized cross-functional decision-making, group technology, international manufacturing, scheduling, and inventory management.
In this context the application to (design of) supply chains is the primary interest.
5.2.1 P/A Theory and SCM
Interestingly, agency theory can be applied internally as well as across company boundaries.
In the sub-discipline managerial pay, for instance, the principals are the board of directors and the agent is the manager, but in e.g. adverse selection the agents may be internal as well as external entities. The theory may be used to decide whether to have certain activities performed within an organization or have them outsourced. Also the mode of cooperation can be analyzed or described as the contract type will determine (or assume) access to monitoring.
The issue of internal and external integration thereby does not constitute a problem, at least not at an overall level of analysis.
Relating agency theory to SCM does raise some issues, though:
39 Propositions from Eisenhardt (1989a). As described above the variables ‘Goal conflict’, ‘Information asymmetry’, ‘Risk attitude (agent)’, and ‘Risk attitude (principal)’ are assumptions in the basic agency problem model, with the values ‘High’, ‘Yes’, ‘Risk averse’, and ‘Risk neutral’, respectively. ‘Contract type’ is the output from the model
¾ First and foremost, it is not evident who the principals and the agents are. Intuitively, the first tier supplier is the agent as he is providing components to the final product.
But the principal might also have agents located downstream, e.g. in the form of a TPL provider. So, the position in the chain does not in itself dictate the distribution of roles.
Considering even more complex situations like mass customization one may imagine a principal-agent model containing multiple agents (own and external) and multiple principals (subsidiary and customer)40. From this setup a large number of distinct model configurations can be created by modifying the degree of customization, the contact between the end-customer and the external parties, the stock keeping policies, type of postponement etc.
¾ Secondly, agency theory is dealing with dyadic relationships, and does not (in its original/basic form) deal with more levels of participants. In the adverse selection model a choice of one agent from a larger number must be made, in the managerial pay scenario multiple principals act as one decision maker. Even if Agrell et al. (2004) introduce a middle-man in their contribution, agency theory does not handle multi-tier models well. In this respect P/A theory resembles TCE.
¾ Thirdly, as pointed out by van Ackere (1993), the basic principal/agent model is single-period. A number of multi-period models exist but de-composing the models and solving them for optimality quickly becomes very difficult.
¾ A fourth complaint might be the absence of attention paid to the process orientation so often referred to as the most fundamental prerequisite for effective SCM. The theory does not relate to the concept of processes, but to tasks (or activities) alone. When it comes to the point of the “interface” (or relationship) between principal and agent, the theory uses the contract dichotomy: behavior-based or output-based contract. After each contract renegotiation for another is performed, supposedly describing the
“interface” or relationship as shifting from absence to output-based contract to behavior-based contract (sequence of types not relevant).
¾ Lastly, relating to the dyadic nature of the models, the optimality criteria used is too simple for SCM. The SCM perspective focuses on win-win for the entire chain, whereas agency theory is focused on maximizing for the principal, and satisfying for the agent(s).
The theory might thereby not be able to design entire supply chains, but might give an indication on which type of relationship is optimal with each supplier and customer on a period-by-period basis. The idea of changing the contract from one period to the next does somehow conflict with the idea of integration and relationship-building often associated with SCM. The applicability of the model might improve dramatically in the context of SCM if the cost element is divided into a principal and an agent element, and a mechanism to adjust the distribution of cost is introduced as well.
40 And this may be further complicated by including corporate as the principal of the subsidiary, having corporate acting as both agent and principal in the same model. As the focus here is on the example of a (series of) single transaction(s) this issue is not further elaborated on.
5.2.2 Contributions Referencing P/A Theory Several contributions referencing P/A theory were identified.
As mentioned already the arguments put forward in Logan (2000) are somewhat superficial, albeit the model itself is quite interesting. In Agrell, Lindroth, & Norrmal (2004) the authors investigate practice of coordination between players in the telecommunications industry by creating a three-tier P/A model. The parameters for the model are clearly and precisely described, but the extension of the model from a two-person to a three-person model somehow conflicts with (the traditional) P/A models. It remains unclear if the model can be further extended to include whole networks, and if the model then still would have theoretical support from P/A theory.
Table 5-4: Prior Research on Supply Risk41
PRIOR RESEARCH ON SUPPLY RISK
Variable Definition Reference
Supply Risk Sources
Inability to handle volume demand changes
Demand fluctuations in quantity and type for a component or
service …
Failures to make delivery requirements Methods to distribute, handle, and transport inputs … Cannot provide competitive pricing The ability to lower the price for the same good or service … Technologically behind competitors The frequency of new ideas and emerging technology … Inability to meet quality requirements The ability of suppliers to conform to specifications … Behavior-based Management
Supplier certification Identify suppliers’ abilities to meet quality, cost, service, and
delivery requirements …
Implement quality management programs
Implementing programs to improve the abilities and activities of suppliers to satisfy the quality needs of the purchasing firm … Develop target costing with suppliers Setting a planned selling price and subtracting the desired
profit, marketing, and distributing costs, leaving the required manufacturing and procurement costs
… Supplier development Efforts of the purchasing organization to improve a supplier’s
performance and/or capabilities …
Buffer-oriented Management
Safety stock Additional stock or items for products, supporting activities,
and customer service held internally …
Using multiple supply sources Procurement of a good or service from more than one
independent source …
Requiring suppliers to hold inventory Additional stock or items for products, supporting activities, and customer service held at the suppliers’ firm …
Zsidisin & Ellram (2003) use agency theory to describe the management of supply risk, and propose a dichotomy: Behavior-based Management and Buffer-oriented Management, see Table 5-4 above. The list of risk sources is quite interesting, as most of the risk sources have both a structural and a process risk component. The ‘Inability to meet quality requirements’
risk source, for example, can be perceived as a process risk as the variation in quality will create the necessity to perform quality assurance on receipt, and will thereby create fluctuations in the flow of input. The structural component is represented in the possibility that the chosen supplier should have been dropped. Similarly, ‘Inability to handle volume
41 Source: Table 1 in Zsidisin & Ellram (2003), p. 17.
demand changes’ and ‘Failures to make delivery requirements’ have both risk components, whereas ‘Technologically behind competitors’ and ‘Cannot provide competitive pricing’ are structural risk sources. The dichotomy for the management of these risks demonstrates the difficulty of merging the probability/impact construct with the behavior/buffer construct, and at the same time being true to the contract metaphor (outcome/behavior). Specifically, the third strategy in the Behavior-based Management class, the ‘Develop target costing with suppliers’ seems more like an outcome type strategy, but does obviously not fall in the
‘Buffer-based Management’ class. The strategies in the second class, on the other hand, do not fit well with outcome type contract, but rather as mitigation strategies for failing outcome based contracts.
In a later article Zsidisin et al. (2004) use P/A theory to classify supply risk assessment techniques. In the article the authors decompose the basic P/A models put forward in Eisenhardt (1989a) and use the individual variables to classify the techniques in play. Theory is not developed and suggestions are kept at the most overall level.
5.2.3 P/A Theory and Supply Chain Risks
It is imminent risk has a place within agency theory as risk attitudes of principal(s) and agent(s) are variables in the models. The risk component is quite limited though, as it focuses on the contract type proposed by the principal and the risk appetite of principal and agent(s) only. External risk sources are not available in the model albeit their existence is acknowledged (the result of an output based contract is not solely ascribed opportunistic behavior of the agent). Or in other words: uncertainty relating to the output of the activities is a basic assumption in the model, but external factors are not specified.
Management of the Supply Chain Risks
Accepting the dichotomy of output versus behavior based contracts, the ability to manage the supply chain risks can be described as follows:
¾ Process Risk: According to P/A theory a principal chooses to ‘sell’ the risk of deviations to the agent(s). The added cost could be perceived as an insurance premium relieving the principal of uncertainty for the contracted output over the specified period of time. The principal, based on the cost of obtaining access to the behavior of the agent, can decide if it is worth to monitor the agent. Managing the intentional disruptions of the contract is thereby directly addressed. Unfortunately the non-intentional is less so. From the agent’s point of view the burden of uncertainty must be calculated in when accepting an output-based contract, for the principal the only difference between the two contract types is the timing of recognition of an exception.
In case of a behavior-based contract the principal will monitor the agent’s progress and will (immediately) observe the exception, whereas in the output-based contract the discrepancy between contracted and delivered output is not recognized until contract end. Obviously, the contract will contain clauses stipulating compensation for non-performance, but in many cases the claims can not cover the damages. Therefore, the
difference between the two contract types is the time period mentioned above, and the resulting safety stock needed to ensure stability downstream. The contract type will thereby describe the attitude towards the poorly performing supplier: the behavior-based contract signals inclusion and intent to solve the problem at the source, whereas the output-based contract signals exclusion.
¾ Structure Risk: The structure risks, on the other hand, are not really addressed as contracts are supposed to cover a certain period. Dropping out of a contract midway is not really a concern of the framework, albeit the concept ‘Moral Hazard’ is concerned with non-performance. The intentional exit from a contract might be handled by insisting on a behavior-based contract, but this would basically invalidate the theory, as all contracts with critical supply chain partners thereby would be behavior-based. For the same reasons as above the enforcement of contracts might have limited effects, besides increasing the cost of making a supplier switch.
In conclusion, the usability of P/A theory on the management of the supply chain risks is somewhat limited by a number of issues, both in relation to SCM in general and in terms of risk management. Especially the focusing on periodic renegotiation and the implied shifts between contract types from period to period conflicts with SCM.
Of the two risk categories, the process risks are best supported by P/A as moral hazard is addressed directly, and an option exists in relation to poor performers. The structural risks, on the other hand, do not really fit with the theory, especially not the non-intentional exit of a critical supplier.
The third theory in the analysis, Resource Based Theory, is somewhat different from TCE and P/A.