Principal-Agent Theory at Work: When Employees Decide Differently

Why do employees decide differently from how the owner would?

Every owner knows the situation. A salesperson grants a customer a ten percent discount although the order would have come anyway. A technician bills an hour more than was needed. An employee in purchasing has ordered from the same supplier for years because he is friendly, not because he is cheap. None of them wants to harm the business. Nevertheless the decisions turn out differently from how the owner would have made them.

The obvious reaction is annoyance, followed by more control. Before reacting that way, it is worth asking whether the decision was not a reasonable response to the employee's situation. Put simply, the question is: why do employees' decisions deviate from the owner's, and what helps without damaging trust in the business?

To answer it, I first present principal-agent theory, apply it to typical cases, and describe the two remedies it recommends. Then a second theory has its say, one that disputes the view of human nature behind the first. At the end I weigh which view goes further in a small business.

What is principal-agent theory?

The theory describes a relationship in which one person, the principal, delegates a task to another person, the agent. Michael Jensen and William Meckling developed it in 1976 for the relationship between shareholders and managers. But it fits any relationship in which someone acts on behalf of another: owner and managing director, owner and employee, customer and tradesperson.

Two assumptions carry the theory. First, principal and agent do not pursue the same goals. The owner wants profit; the employee also wants pleasant work, recognition, security, or a bonus. Second, the agent knows more about his work than the principal. This information asymmetry, unequally distributed knowledge, makes it hard for the principal to judge whether the agent acted in his interest.

An everyday analogy makes this tangible. If you take a taxi in an unfamiliar city, you do not know whether the driver takes the shortest route. The driver knows. If he is paid by the mile, he has a reason for the detour. Most drivers do not take it anyway. But the passenger cannot check, and this gap is exactly what the theory describes.

Three everyday cases

With these terms, the opening examples can be read anew. The focus is not on whether someone is good or bad, but on which situation explains his behavior.

1. The discount

   The salesperson is paid on revenue. A discount raises his close rate and costs him almost nothing. He does not see the contribution margin, what remains after direct costs. His behavior is a reasonable response to the incentive.

2. The extra hour

   The technician knows how long the work really took. The owner does not. If generous times never stand out, while tight ones lead to trouble with customers, he logs generously.

3. The convenient supplier

   The buyer bears the risk if a new supplier delivers poorly. Nobody in the business sees the benefit of a lower price. For her the familiar supplier is the safe choice.

In all three cases the situation distributes benefit and risk unequally. The employee bears a disadvantage if he acts in the owner's interest and gains an advantage if he does not. If you want to change that, you have to address the situation, not the person.

If you are annoyed by your employees' decisions, first check what you pay them for and what you show them.

The two remedies: information or outcome-based pay

In 1989 Kathleen Eisenhardt prepared the theory for organizational research in a much-cited review. Her conclusion: the theory offers distinct insights into information systems, outcome uncertainty, incentives, and risk, and it is empirically well supported, especially when combined with complementary perspectives. Two basic routes can be derived from this for practice.

The choice depends on the task. Where good behavior can be described and observed, such as standardized installation, information is the cheaper route. Where the outcome is clearly measurable and depends on the employee, such as winning new customers, an outcome link can make sense. Where the outcome depends heavily on chance, an outcome link becomes expensive, because the employee bears a risk he cannot influence.

A worked example shows how much depends on what is measured. A salesperson brings in orders worth 200,000 euros in a quarter. With a commission on revenue, any discount that secures a deal pays off. With a commission on contribution margin, the same discount costs him something himself.

How to design pay so that it does not reward the wrong deal is described in the comparison Sales compensation: base, commission, and bonus. How to determine the contribution margin per order is shown in Contribution margin per order.

The counterposition: employees as stewards

Principal-agent theory has a particular view of human nature. Davis, Schoorman, and Donaldson (1997) described it like this: economic approaches such as agency theory portray subordinates as individualistic, opportunistic, and self-serving. Against this they set stewardship theory. A steward acts in the interest of the organization because he identifies with it and his own benefit lies in its doing well.

For the authors, neither view is always right. According to their model, whether someone behaves like an agent or like a steward depends on his characteristics and on the situation the organization creates. If you control a lot, you treat employees as agents and are more likely to get agents. If you offer involvement and trust, you make steward behavior possible, but you run the risk of being disappointed.

Empirical evidence of this interaction comes from Della Torre, Salimi, and Giangreco (2020). They analyzed data from more than 22,000 establishments in Europe. Where employees had a direct voice, establishments were more often organizationally innovative. This relationship was weaker in establishments with individual performance pay such as piece rates. Profit sharing for everyone, by contrast, did not weaken it. That fits the assumption that narrowly measured individual incentives can crowd out thinking along, while shared incentives do not.

Agent or steward: which view holds in a small business?

Both theories are right, but for different situations. Principal-agent theory accurately describes what happens when incentives are skewed and information is missing. Stewardship theory accurately describes that most people want to do good work and respond to distrust by withdrawing.

My assessment: in a small business the more important lever is almost always information, not control. That sounds like a contradiction but is not one. Control means the owner checks the behavior. Information means the employee himself sees what his decision causes. A salesperson who knows the contribution margin of his order does not need discount approval in every case. He has the same information as the owner and often decides the same way.

This closes the gap the theory describes from the other side. Instead of extracting the agent's knowledge, the principal hands over his own. That is cheaper and strengthens steward behavior instead of undermining it. How to put this into practice is described in the guide Fostering entrepreneurial thinking in the team. A place where the most important numbers are visible to everyone is the one-page KPI cockpit.

Control is still needed, but in a different place. Where a lot of money moves or interests clearly conflict, clear limits, a second signature, or a regular review are needed. That is not distrust of individuals but a rule that protects everyone. How much of this a small business needs is discussed in Corporate governance for small businesses.

The Interest Check: five questions before any rule

Before you respond to an unwanted decision with a new rule, five questions are worthwhile. I call them the Interest Check.

• What is the person actually paid or praised for, and does that reward the desired decision?
• What information does she have, what do I have, and could I give her mine?
• Who bears the risk if she decides in my interest, and who gets the benefit?
• Can good behavior be described, or only the outcome measured?
• Would a new control weaken the person's responsibility instead of strengthening it?

Often the first question already reveals the cause. If you praise employees only for revenue, you get revenue. If you praise them for responsiveness, you get responsiveness. The theory does not demand elaborate contracts, but an honest look at what the business actually rewards. How to hand over tasks so that result and framework are clear is described in Delegating tasks with a clear handover.

What you can put into practice right away

1. Write down the three decisions by your employees that annoyed you most recently.
2. Go through each with the five questions of the Interest Check.
3. Choose a number that so far only you know, such as the contribution margin per order, and show it to the team.
4. Check whether a commission or bonus rewards something other than what you actually want.

The third measure is probably the most effective and the one owners hesitate over most. Sharing numbers feels risky. The theory suggests that the risk of not sharing them is greater.

Conclusion: why do employees decide differently, and what helps?

The guiding question was why employees decide differently from the owner and what helps without damaging trust. Principal-agent theory answers: because goals and knowledge are unequally distributed. Its remedies are better information about behavior or pay based on outcomes. Stewardship theory adds that too much control can produce the very behavior it is meant to prevent.

For a small business this yields a judgment in favor of information. If you share your numbers and set incentives so that they measure the right thing, you close the knowledge gap from your own side. Control belongs where a lot is at stake. Deviating decisions are then less often a character problem than a sign of a gap the owner can close himself.