6 Agency Theory

Stijn Masschelein

The chapters on specific knowledge and human capital showed that top management needs to decentralise decisions to departments because the people in those departments have skills, knowledge, and capabilities that top management does not have. This creates a potential conflict of interest where the departments can use those decisions rights to take decisions that are in their best interest and no longer in the interest of the firm. Because the departments know better which decision is in the best interest of the firm, top management cannot judge whether the departments have taken the right decision directly. They have to rely on necessarily imperfect measures. The central trade-off of decentralisation is between better decision making by the departments and the contracting costs as the result of imperfect performance measures (Prendergast, 2002).

In contrast to market transactions, organisations do not transfer the ownership rights to their employees; they only grant them some decision rights. Sales staff in a car dealership do not have the right to do what they want with the car. There are a limited number of decisions they can take over what will happen with that car. The decisions they can and cannot take will be described in their (official) job description. Their level of autonomy will depend on whether or not they have to get permission from a supervisor to give price reductions, give away extras at reduced prices, or pay for the customer’s old car. To make sure that the sales people make decisions in the interest of the dealership, they will be paid a commission for each sale. The remainder of this chapter deals with the problem of designing these incentives so that the balance between decision making and contracting costs is right.

The Agency Theory of Incentive Contracts

The elements of incentive contracts

One of the tools to align the interests of the employees with the interests of the firm are incentive contracts. For our purpose contracts are broadly defined; they can be both written and implicit agreements. Incentives are similarly supposed to capture a broad range of possible rewards for employees: from monetary bonuses, to business trips and awards, or a promotion. The other crucial element in an incentive contract are the performance measures that determine whether the employees get a reward or that determine the size of the reward. The measure can be financial measures, non-financial measures, or even the subjective judgment by a superior.

In summary, the theory of incentive contracts can apply to a broad range of settings. In the remainder of this chapter, the focus will be on explicit bonuses for managers of a department but keep in mind that the theory is broader than this example [1]. I start by making goal of incentive contracts explicit in more detail which will help us to understand better why firms are using incentive contracts over just monitoring their employees. In the sections that follow, I will explain the characteristics of good performance measures and how the combination of multiple performance measures can improve incentive contracts. Lastly, I look at the problems that arise when the performance measures motivate employees only on a subset of all the actions that are the best interest of the firm.

The goal of incentive contracts

The main goal of incentive contracts is to encourage and direct the effort and attention of employees. The first goal that springs to mind is that bonus contracts make employees work harder because they will be rewarded for their additional effort. This is not the only function of incentive contracts and is the least interesting one for our perspective because it does not focus on the problem that departments with specific knowledge make decentralised decisions. Incentive contracts also direct the effort and attention of employees to the right task. For instance, fashion designers might spend most of their time on innovative and artful designs while their employer might be more interested in more mainstream designs. To make sure that the designer spends enough time on mainstream designs, the organisation can reward the designer for mainstream designs but not for avant-garde designs. With or without the incentive contract the designer will work equally hard but with the incentive contract, their effort will be more in the commercial interest of the organisation. Thus, the organisation can give the fashion designers the right to decide on which tasks to focus if they can design an appropriate incentive contract. Another related but slightly different role for incentives is to make sure employees invest in projects that are profitable for the firm and not in projects that only increase the prestige and status of the employee. CEOs are often accused of empire building which means that they are investing in big and popular projects to raise their own status but those projects are not necessarily in the best interest of shareholders. One way to solve this problem is to reward CEOs with stock rewards. Of course, coordinating the incentives of the organisation is the key task of top management according to transaction cost economics.

These incentive contracts are studied in sophisticated mathematical models in a field of study that is sometimes called agency theory or principal-agent theory (Holmstrom & Milgrom, 1994; Lambert 2001). To explain the main intuition, I rely on a simple story of top management (the principal) and one employee (the agent). The timeline of the labour relationship under an incentive contract is as follows. First, the principal and the agent agree on a basic salary and on a bonus for the employee depending on the outcome of a performance measure. Next, the employee works and takes decisions. The assumption is that the agent will try to maximise their total compensation taking into account the costs of maximising total compensation. Examples of these costs are extra hours of work, or doing tasks that do not help their future career, status or prestige. Another important cost stems from the risk that the measure might not capture the agent’s performance completely and that despite the agent’s best effort they do not get their reward. Finally, in the last step in an incentive contract, the outcome for the performance measure is established and the employee receives their salary and bonus.

Something that is often counterintuitive is that in this baseline story, the behaviour of the agent does not change based on whether they get the bonus. The behaviour of the agent is already fixed at the time when the bonus is being paid. Or in other words, the behaviour of the agent determines the bonus and not other way around.

To better understand the role of incentive contracts, we can contrast incentive contracts with giving directions to employees and correct them when they do not adhere to the directions. One of the reasons why giving directions might not work is that employees know better how to do their job than top management. They have the experience of doing the job day in day out or they have the specific education. In short, employees have specific knowledge. In addition, communicating top management’s directions through the different hierarchical levels is costly or takes a long time. Lastly, the employer still needs a costly monitoring system to know whether employees have adhered to the directions they were given. The employer can not follow around all employees to check whether they are following up on the employer’s directions or not. To avoid all these problems and transaction costs, firms can rely on incentive contracts to align the interest of the employees. The next section will deal with costs that are associated with these incentive contracts in the simplified story.

Risk, sensitivity, and precision

The benefit of incentive contracts for top management is that contracts make employees spend their time at the right tasks and spend more effort for the firm. The benefit for some employees is that they enjoy the responsibility and decision rights instead of following directions. However, because top management can only imperfectly measure and reward the employees, there is a risk that despite spending more effort on an unattractive task, employees do not receive a reward. If the performance measure is affected by other factors such as the economy, competitors, or other employees than the employee might not receive the expected bonus despite their best effort. Most people are risk averse and prefer to avoid such risks. Employees will only accept a risky incentive contract if they are compensated for this additional risk of incomplete measurement (Holmstrom, 1979). The additional compensation to make-up for the risk [2] is a transaction cost or a contracting cost for the employer. The organisation will have to weigh this extra cost against the benefit of aligning the employee’s interest with those of the firm.

A good measure from both the organisation’s and the employee’s point of view is a measure that picks up only what the employee has done and nothing else. Remember that the main goal of a performance measure in an incentive contract is to direct the decisions and effort of the agent. Such a precise measure limits the risk of employees not getting their reward while they deserve it. The advantage for the firm is that they do not have to pay a higher compensation for the additional risk. A good measure will also be sensitive to what the employee has done. This means that when employees change their behaviour, the performance measure will change as well. If a measure is sensitive to their behaviour, it is easier for the employee to move the measure in the right direction and gain a reward. Because it is easy for the employee to influence the measure, firms do not need to hand out large rewards to motivate the employee and they can set more demanding targets without offering egregious large rewards (Banker & Datar, 1989; Holmstrom, 1979).

Precision and sensitivity of a measure are not the same characteristic. Measures can be affected by little outside influences and be precise but at the same time not respond to the behaviour of the employee and be insensitive. Similarly, a measure can be affected by the behaviour of the employee but also affected by a number of other factors. For such measures, the organisations need to trade-off whether the improvement in precision is worth the decrease in sensitivity.

In the simplest agency theory story, the risk premium is the main transaction cost of a bad performance measure. If it is difficult for the employee to improve on the performance measure and the measure can be affected by a number of outside factors, they will not want to work under an incentive contract but prefer a fixed salary. To persuade risk averse employees to sign an incentive contract with such a bad measure, organisations will have to pay high salaries and bonuses.

Multiple measures

One way to improve the incentive contract is to make the reward dependent on more than one measure. The idea is that at two measures that are very insensitive or very noisy can together be informative about the behaviour of an employee. A specific application of this idea is to include measures in a contract that are uncontrollable by employees but decrease the noise in the contract. For instance, organisations regularly use benchmarks that make rewards dependent on the performance of peers or competitors. If there is a lot of economic uncertainty and the peers or competitors are also subject to that uncertainty, including performance of peers as a benchmark in the contract makes the total compensation less risky.

For instance, a car dealership might want to reward the sales staff based on their individual sales. However, car sales depend on the general economic sentiment which is out of control of the sales staff. Furthermore, it is hard to actually measure the effect of economic sentiment on car sales. The sales staff will want to avoid that their bonus depends on weak economic demand. One possibility for the car retailer is to compare the sales from each staff member to the average and give them a bonus if they are above average. Because all sales staff are affected by economic demand, the average of the peers’ sales will capture the impact of changes in demand and make the bonus less noisy. Including an insensitive measure (i.e. average peer sales), increases the precision of the overall measurement (i.e employee sales minus average peer sales) because it filters out the common economic effects. This is the basic idea behind benchmarking explained within the agency theory framework [3]

Multiple actions and incongruence

So far we did not explicitly take into account that a measure can be affected by two different actions from the employee. Another way of looking at the issue is that managers and employees take a lot of different independent decisions that can all affect the performance measures. A lot of actual real-life performance measures are compiled out of information and reports gathered by the employees. This means that employees can influence the measure without improving task performance by changing the reports. In this example, the employee can take two actions that will influence the performance measure, i.e. they can manipulate the measure or they can improve performance. If a measure is more sensitive to measurement manipulation than to performance improvement (i.e. it is easier to fudge the number undetected than to actually do the work), incorporating the measure in the contract will motivate the employees to manipulate the measure (Holmstrom, 1979; Lambert, 2001). If employees can influence a measure in two different ways and one of them is not valuable for the firm, the measure is called incongruent with the firm’s interest.

The opposite is also possible. Sometimes a performance measure does not capture every aspect of the job. For instance, a lot of the criticism of accounting based measures (e.g. accounting profit) is that they only measure a decision’s effect on short-run profit of the firm but not necessarily the effect on long-term firm value. The firm can complement the accounting based measures with non-financial measures such as customer satisfaction, product quality or leadership to capture those long-term effects.

These two examples show that the principal needs to consider two aspects when thinking about the effect of the actions on the performance measures. (1) Are there any actions that can improve a measure but are not in the best interest of the organisation? (2) Are there measures that do not capture the results of the agent’s actions? A congruent measure is a measure that captures all the actions of an employee that are in the best interest of the firm and no other actions. Very few measures are at the same time precise, sensitive to desired actions and not sensitive to undesired actions. Thus, incongruent measures will have to be rectified by adding additional measures to the bonus contract (Holmstrom & Milgrom, 1991). This is the motivation behind the Balanced Scorecard which I will discuss in a future chapter.

However sometimes, the organisation will have to trade-off the different characteristics of the measures. For instance, the profit of a department is arguably more sensitive to the decisions by the department’s managers than firm profit. Departmental profit is also less noisy than firm profit. In contrast, firm profit is more congruent because it will capture the effect that the manager’s decisions have on the other departments. In the chapter on Specific Knowledge, we worked through a story of a production department and a sales department of a computer firm where both department were strongly interdependent. The trade-off might be difficult [4]. Let us assume that the head of the department can invest in two projects: both projects improve departmental profit but only one improves firm profit. One setting where that could occur is when the department receives the revenues of the project while another department bears the costs. This is a variation on the story in the chapter on specific knowledge. Rewarding the head of the department only based on departmental profit will increase the effort in the project that is in the best interest of the firm but also in the project that is not in the best interest of the firm. Depending on the circumstances it can be best for the firm to:

  1. use both firm and department profit to evaluate the department. In that case, the firm will combine the congruence of firm profit with the relatively stronger sensitivity and precision of the department profit.
  2. use only department profit. If the project that is in the best interest of the firm is also more profitable than the other one for the department, department profit is not too incongruent. It will still motivate the department to focus on the project that is in the best interest of the firm.
  3. use only firm profit. If firm profit is not too noisy, the cost of the risk premium will be smaller than the benefits of more congruence with the interest of the firm.

These issues are at the forefront in discussion on the reorganisation(s) at Microsoft. Whereas until around 2010, Microsoft had strong internal competition between divisions (such as the hardware, the Office software, or the Windows software division), they have reconsidered their structure. Because of the strong focus on division profit, each division refused to cooperate with other divisions and invested in projects that caused headaches for other divisions.

One of the main sources of competition between the divisions is for the limited budget to develop new projects. The divisions have to convince headquarters that they deserve investments in their projects. However, that competition can turn ugly when different divisions do not want to accommodate their projects to benefit others. For instance, the Office division was for a long time not willing to adapt their software for a touch and stylus interface.

Costs and Benefits of Decentralisation

The strategy of the firm will have a big impact on how a firm structures its different divisions. Some firms will split up the firm according to geographical locations, others choose to have divisions for different product lines, and others have different divisions for different business functions such as sales, production, and purchasing. Big international firms will often have an organisational structure over multiple dimensions in a so-called matrix structure. In this concluding section, I want to emphasise again how the decentralisation decision fits into the broader framework I have presented so far.

There are multiple reasons why headquarters decide to decentralise decisions to divisions. Geographically decentralised divisions often have better specific knowledge about the preferences of customers, the capabilities of suppliers, and employees, or the local political environment. As a result, a decentralised division will be more responsive to its environment than headquarters. Because a decentralised division does not have to wait for directions from headquarters it will not only detect changes in the local environment quicker but also react quicker to changes in customer preferences or supplier capabilities.

In addition to improving decision making, decentralisation improves the motivation of managers in the division. Because managers feel more empowered when they can make their decisions, managers with more responsibility are more likely to take initiative and work harder in the interest of the firm. For instance, lower level managers with more autonomy are more likely to develop management skills than managers who can take fewer decisions. This does not mean that the incentives we talked about in the previous section do not matter. When the managers can earn a bonus or a promotion, they will be more likely to put those improved skills in action.

In the section above I already highlighted one potential risk of decentralisation. It can lead to competition between departments and the organisation no longer has the benefit that it can coordinate multiple investment decisions. Another disadvantage of decentralisation is that some activities and functions of the firm might be duplicated. That is why firms often centralise some supporting functions such as HR administration, accounting, or purchasing. Centralisation of these functions is an example of how firms can economise on the transaction costs of running a business.

The main insight of agency theory for this textbook is that decentralisation of decisions depends on how well the organisation can measure the actions of the decentralised employees. If the measures as a whole are sufficiently sensitive, precise, and congruent, the organisation can design a performance evaluation system that motivates the employees to act in the interest of the organisation. In agency theory, the only transaction cost of such a performance evaluation system is the risk premium the organisation has to pay as the result of insensitive and imprecise measures and the cost of misaligned incentives when measures are incongruent. In practice, another considerable transaction cost will be the cost of collecting, refining, and storing the measures that the organisation is using to evaluate the employees.


Banker, R. D., & Datar, S. (1989). Sensitivity, Precision, and Linear Aggregation of Signals for Performance Evaluation. Journal of Accounting Research, 27(1), 21–39. https://doi.org/10.2307/2491205

Holmstrom, B. (1979). Moral Hazard and Observability. The Bell Journal of Economics, 10(1), 74–91. https://doi.org/10.2307/3003320

Holmstrom, B., & Milgrom, P. (1991). Multitask principal-agent analyses: Incentive contracts, asset ownership, and job design. Journal of Law, Economics and Organization, 7, 24–52. https://doi.org/10.1093/jleo/7.special_issue.24

Holmstrom, B., & Milgrom, P. (1994). The Firm as an Incentive System. The American Economic Review, 84(4), 972–991. https://www.jstor.org/stable/2118041

Lambert, R. A. (2001). Contracting theory and accounting. Journal of Accounting and Economics, 32(January), 3–87. https://doi.org/10.1016/S0165-4101(01)00037-4

Prendergast, C. (2002). The tenuous tradeoff between risk and incentives. Journal of Political Economy, 110(5), 1071–1102. https://doi.org/10.1086/341874

  1. For instance, the chapter on internal labour markets will explicitly explain how subjective, non-verifiable judgements and promotions work as incentive contracts.
  2. You can think of it as a risk premium in financial transactions.
  3. This is a good place to illustrate why I take the approach of building up the theory of why organisations need performance measures over different story. At this point, you can trace back when benchmarking will be useful. It helps when an organisation (1) decentralises decision making to knowledgeable employees, (2) needs to incentivise the decentralised decision makers, (3) those decision makers are risk averse, and (4) benchmark helps to remove unwanted noise from the decision makers performance measure. I find it useful to be able to identify those assumptions because it helps me to decide whether benchmarking should be a priority in improving an organisation's performance measurement system.
  4. The story that follows is not something I necessarily expect you to come up with on your own. However, it again illustrates that combining the different stories and mechanisms that I have presented so far can lead to advanced insights in real cases. The problem is that the actual outcomes will depend on the actual costs and benefits which will differ from case to case. It becomes impossible to give general explanations that will always be true.


Icon for the Creative Commons Attribution-NonCommercial 4.0 International License

Principles of Strategic Management Accounting Copyright © 2024 by Stijn Masschelein is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, except where otherwise noted.

Share This Book